NTP National Toxicology Program

U.S. Department of Health and Human Services

NTP Monograph
on the State of the
Science Concerning
Fluoride Exposure
and Neurodevelopment
and Cognition:
A Systematic Review

August 2024
 NTP Monograph on the
State of the Science Concerning Fluoride
Exposure and Neurodevelopment and Cognition:
A Systematic Review
NTP Monograph 08

August 2024

National Toxicology Program

Public Health Service
U.S. Department of Health and Human Services
ISSN: 2378-5144

Research Triangle Park, North Carolina, USA

 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

About This Monograph

What This Monograph Does

This Monograph and Addendum evaluate the evidence of an association between exposures to fluoride and
human neurodevelopment and cognition.
The Monograph is a systematic review of human, experimental animal, and mechanistic studies that focuses
on human data because it is the most informative from available studies. Findings from human epidemiology
studies published through May 1, 2020, are evaluated and summarized in the main text. This includes studies
of various cognitive neurodevelopmental endpoints in children, a limited number of studies in adults, and
studies of potential mechanisms. The majority of the studies examine intelligence quotients (IQs) of children
with higher estimates of fluoride exposure compared to children with lower estimates of fluoride exposure.
The human studies reviewed in this Monograph and Addendum involve fluoride exposures from many sources
including drinking water, prepared beverages, foods, and dental products.
The children’s IQ studies discussed in the Monograph and Addendum were performed in 10 countries
including Mexico and Canada. No studies evaluating IQ were conducted in the United States.
What Evidence Was Evaluated
Published research from human, experimental animal, and mechanistic studies on fluoride exposure and
neurodevelopment and cognition was identified and evaluated. Studies from each evidence stream were fully
evaluated through April 1, 2019. In early drafts, it was apparent that the animal data were of poor quality and
that the human data were most informative and would be the basis of the confidence conclusions. Therefore,
findings from the animal evidence stream were determined to be inadequate and were removed from further
drafts. Mechanistic evidence from animal studies was retained in Appendix F. The epidemiology data were
updated and fully assessed through May 1, 2020, for the current Monograph, and the animal and mechanistic
literature identified between April 1, 2019, and May 1, 2020, were scanned for potentially major advances that
might impact the confidence conclusions. This effort did not find significant evidence of increased
understanding of how fluoride may affect children’s cognitive neurodevelopment and did not strengthen the
confidence assessment based on the human evidence available in 2020.
Considerable time elapsed during peer reviews of drafts of this Monograph necessitating an updated search of
the children’s IQ and associated mechanistic literature through October 2023, carried out according to the
same systematic review procedures. To accommodate the review, this update is summarized in an Addendum
following the conclusion of the main text of this Monograph. The systematic review of the updated literature
did not result in a change to the earlier moderate confidence conclusion. The experimental animal and
mechanistic literature were also searched and added to updated, interactive reference flow diagrams in the
Addendum for transparency.
What This Monograph Does Not Do
This Monograph and Addendum do not address whether the sole exposure to fluoride added to
drinking water in some countries (i.e., fluoridation, at 0.7 mg/L in the United States and Canada) is
associated with a measurable effect on IQ.
This Monograph and Addendum do not provide a quantitative estimate of the number of IQ points lost for a
given increase in fluoride exposure measures; however, references are provided to prior and concurrent meta-
analyses that do provide such estimates (DTT Meta-analysis, Taylor et al. 2024, in press).
This Monograph and Addendum do not assess benefits of the use of fluorides in oral health or provide a
risk/benefit analysis.

ii
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Foreword
The National Toxicology Program (NTP), established in 1978, is an interagency collaboration
within the Public Health Service of the U.S. Department of Health and Human Services. Its
activities are executed through a partnership of the National Institute for Occupational Safety and
Health (part of the Centers for Disease Control and Prevention), the Food and Drug
Administration (primarily at the National Center for Toxicological Research), and the National
Institute of Environmental Health Sciences (part of the National Institutes of Health), where this
virtual program is administratively located. NTP’s work focuses on the testing, research, and
analysis of agents of concern to identify toxic and biological effects, provide information that
strengthens the science base, and inform decisions by health regulatory and research agencies to
safeguard public health. NTP also works to develop and apply new and improved methods and
approaches that advance toxicology and better assess health effects from environmental
exposures.
Literature-based evaluations are one means by which NTP assesses whether exposure to
environmental substances (e.g., chemicals, physical agents, and mixtures) may be associated
with adverse health effects. These evaluations result in hazard conclusions or characterize the
extent of the evidence and are published in the NTP Monograph series, which began in 2011.
NTP monographs serve as an environmental health resource to provide information that can be
used to make informed decisions about whether exposure to a substance may be of concern for
human health.
These health effects evaluations follow prespecified protocols that apply the general methods
outlined in the “Handbook for Conducting a Literature-Based Health Assessment Using the
OHAT Approach for Systematic Review and Evidence Integration.Ӡ The protocol describes
project-specific procedures tailored to each systematic review in a process that facilitates
evaluation and integration of scientific evidence from published human, experimental animal,
and mechanistic studies.
Systematic review procedures are not algorithms, and the methods require scientific judgments.
The key feature of the systematic review approach is the application of a transparent framework
to document the evaluation methods and the basis for scientific judgments. This process includes
steps to comprehensively search for studies, select relevant evidence, assess individual study
quality, rate confidence in bodies of evidence across studies, and then integrate evidence to
develop conclusions for the specific research question. Draft monographs undergo external peer
review prior to being finalized and published.
NTP monographs are available free of charge on the NTP website and cataloged in PubMed, a
free resource developed and maintained by the National Library of Medicine (part of the
National Institutes of Health). Data for these evaluations are included in the Health Assessment
and Workspace Collaborative.
For questions about the monographs, please email NTP or call 984-287-3211.
†
OHAT is the abbreviation for Office of Health Assessment and Translation, which has become the Health
Assessment and Translation group in the Integrative Health Assessments Branch of the Division of Translational
Toxicology at the National Institute of Environmental Health Sciences.

iii
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Table of Contents
About This Monograph ................................................................................................................... ii
Foreword ........................................................................................................................................ iii
Tables ...............................................................................................................................................v
Figures............................................................................................................................................ vi
About This Review ....................................................................................................................... vii
Peer Review .................................................................................................................................. xii
NTP Board of Scientific Counselors Review .............................................................................. xiii
Publication Details ...................................................................................................................... xvii
Acknowledgements ..................................................................................................................... xvii
Conflict of Interest ...................................................................................................................... xvii
Abstract ...................................................................................................................................... xviii
Preface............................................................................................................................................ xi
Introduction ......................................................................................................................................1
Objective and Specific Aims.......................................................................................................3
Objective ..............................................................................................................................3
Specific Aims .......................................................................................................................3
Methods............................................................................................................................................5
Problem Formulation and Protocol Development ......................................................................5
PECO Statements ........................................................................................................................5
Literature Search .........................................................................................................................7
Main Literature Search ........................................................................................................7
Supplemental Chinese Database Literature Search .............................................................8
Databases Searched..............................................................................................................9
Searching Other Resources ................................................................................................10
Unpublished Data...............................................................................................................10
Study Selection .........................................................................................................................10
Evidence Selection Criteria................................................................................................10
Screening Process ..............................................................................................................10
Evaluation of SWIFT-Active Screener Results .................................................................11
Screening of the May 2020 Literature Search Update .......................................................12
Supplemental Chinese Database Searches and Human Epidemiological Studies .............12
Data Extraction .........................................................................................................................13
Extraction Process..............................................................................................................13
Data Availability ................................................................................................................13
Quality Assessment of Individual Studies ................................................................................14
Key Risk-of-bias Questions ...............................................................................................14
Risk-of-bias Considerations for Human Studies................................................................14
Organizing and Rating Confidence in Bodies of Evidence ......................................................20
Health Outcome Categories for Neurodevelopmental and Cognitive Effects ...................20

iv
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Considerations for Pursuing a Narrative or Quantitative Evidence Synthesis ..........................20

Confidence Rating: Assessment of Body of Evidence .............................................................21
Factors to Consider for Potential Downgrading ................................................................21
Factors to Consider for Potential Upgrading .....................................................................23
Results ............................................................................................................................................25
Literature Search Results ..........................................................................................................25
Human Neurodevelopmental and Cognitive Data ....................................................................26
IQ in Children ....................................................................................................................27
Other Neurodevelopmental or Cognitive Effects in Children ...........................................58
Cognitive Effects in Adults................................................................................................69
Mechanistic Data in Humans .............................................................................................73
Animal Learning and Memory Data .........................................................................................76
Mechanistic Data in Animals .............................................................................................76
In Vitro Data on Neurodevelopmental or Cognitive Effects ....................................................77
Discussion ......................................................................................................................................78
Limitations of the Evidence Base .............................................................................................80
Limitations of the Systematic Review ......................................................................................82
Summary ........................................................................................................................................84
Addendum: Updated Literature Search (May 2020 through October 2023) .................................85
Introduction ...............................................................................................................................85
Results .......................................................................................................................................86
Low Risk-of-bias Studies...................................................................................................87
High Risk-of-bias Studies ..................................................................................................96
Human Mechanistic Studies of Fluoride and Cognitive Neurodevelopment ....................96
Meta-analyses on Fluoride Exposure and Children’s IQ ...................................................98
Conclusion ..............................................................................................................................102
References ....................................................................................................................................103
Appendix A. Data Figures: Neurodevelopmental or Cognitive Effects and Outcomes ............. A-1
Appendix B. Literature Search and Document Review Details ..................................................B-1
Appendix C. Detailed Literature Search Results and List of Included Studies ...........................C-1
Appendix D. Risk-of-bias Figures .............................................................................................. D-1
Appendix E. Details for Low Risk-of-bias Studies ..................................................................... E-1
Appendix F. Mechanistic Data from Animal Studies .................................................................. F-1
Appendix G. Protocol History and Revisions ............................................................................. G-1
Appendix H. Supplemental Files ................................................................................................ H-1

Erratum.......................................................................................................................................... 87

Tables
Table 1. Human PECO (Population, Exposure, Comparator and Outcome) Statement ..................6

v
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Table 2. Animal PECO Statement ...................................................................................................7

Table 3. In Vitro/Mechanistic PECO Statement ..............................................................................7
Table 4. OHAT Risk-of-bias Questions and Applicability by Study Design ................................19
Table 5. The Four Risk-of-bias Rating Options.............................................................................20
Table 6. Studies on IQ in Children .................................................................................................31
Table 7. Studies on Other Neurodevelopmental and Cognitive Function in Children ....................59
Table 8. Studies on Cognitive Function in Adults ..........................................................................70
Addendum Table 1. Studies on IQ in Children since 2020 ...........................................................89
Addendum Table 2. Previous Meta-analyses on Exposures to Fluoride and Children’s IQ .......102

Figures
Figure 1. Assessing Confidence in the Body of Evidence .............................................................24
Figure 2. Study Selection Diagram ................................................................................................26
Figure 3. Number of Epidemiological Studies by Outcome and Age Categories .........................27
Figure 4. Number of High- and Low-quality Studies of Fluoride Exposure and IQ in
Children by Year of Publication.....................................................................................28
Figure 5. Number of Studies of Fluoride Exposure and IQ in Children by Country and
Year of Publication.........................................................................................................29
Figure 6. Important Covariates Considered in Low Risk-of-bias IQ Studies Conducted in
Children ..........................................................................................................................52
Figure 7. Number of Low Risk-of-bias Studies that Evaluated Thyroid Hormones in
Children and Adults by Endpoint and Direction of Association ....................................75
Figure 8. Number of High Risk-of-bias Studies that Evaluated Thyroid Hormones in
Children by Endpoint and Direction of Association ......................................................75
Addendum Figure 1. Reference Flow Diagram for Updated Literature Search ............................86
Addendum Figure 2. Risk-of-bias Heatmap for Children’s IQ Studies Identified During
Updated Literature Search ...........................................................................86

vi
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

About This Review

National Toxicology Program1
1
Division of Translational Toxicology, National Institute of Environmental Health Sciences,
Research Triangle Park, North Carolina, USA

Collaborators
Kyla W. Taylor, John R. Bucher, Robyn B. Blain, Christopher A. Sibrizzi, Pamela A. Hartman,
Kristen Magnuson, Sorina E. Eftim, and Andrew A. Rooney

Division of Translational Toxicology, National Institute of Environmental Health Sciences,

Research Triangle Park, North Carolina, USA
Developed protocol and conducted systematic review through oversight and review of literature
search, screening, data extraction, and risk-of-bias assessment; wrote and edited monograph
Kyla W. Taylor, Ph.D., Project Lead
John R. Bucher, Ph.D.
Andrew A. Rooney, Ph.D.

ICF, Reston, Virginia, USA

Conducted literature screening, data extraction, and risk-of-bias assessment; reviewed data,
results, and analyses; wrote, edited, and formatted monograph; provided database and HAWC
support
Robyn B. Blain, Ph.D.
Sorina E. Eftim, Ph.D.
Kristen Magnuson, M.E.S.M., Deputy Work Assignment Manager
Christopher A. Sibrizzi, M.P.H., Lead Work Assignment Manager

Conducted literature screening, data extraction, and risk-of-bias assessment; reviewed data,
results, and analyses; wrote, edited, and formatted monograph; provided database and HAWC
support; supported data visualizations
Pamela A. Hartman, M.E.M.

Contributors
Division of Translational Toxicology, National Institute of Environmental Health Sciences,
Research Triangle Park, North Carolina, USA
Provided oversight for external peer review
Mary S. Wolfe, Ph.D.

Provided literature search, screening, Distiller and HAWC support

Vickie R. Walker, B.S.

Conducted a technical review of the draft monograph

Freya Kamel, Ph.D. (Retired)
Mamta Behl, Ph.D. (currently Neurocrine Biosciences, Inc., San Diego, California, USA)

vii
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Jesse D. Cushman, Ph.D.

Kelly Ferguson, Ph.D., M.P.H.
Kimberly A. Gray, Ph.D.
Ruth M. Lunn, Dr.P.H.
Suril Mehta, Dr.P.H.

Conducted literature screening

Kris Thayer, Ph.D. (currently Chemical and Pollution Assessment Division, United States
Environmental Protection Agency, Washington, D.C., USA)

Division of Extramural Research and Training, National Institute of Environmental Health

Sciences, Research Triangle Park, North Carolina, USA
Conducted literature screening
Heather King, Ph.D.

Kelly Government Services, Research Triangle Park, North Carolina, USA

Supported external peer review
Elizabeth A. Maull, Ph.D. (retired from NIEHS, Research Triangle Park, North Carolina, USA)

ICF, Reston, Virginia, USA

Provided contract oversight
David F. Burch, M.E.M., Principal Investigator
Cara Henning, Ph.D.
Jessica A. Wignall, M.S.P.H.

Reviewed data, results, and analyses

William Mendez, Ph.D.

Reviewed data, results, and analyses, conducted literature screening and risk-of-bias
assessment, and provided database and HAWC support
Anna Engstrom, Ph.D.

Provided database and HAWC support

Steven Black, M.P.H.
Natalie Blanton, M.P.H.
Kaitlin Geary, B.S.
Courtney Lemeris, B.A.
Rachel McGill, B.S.
Daniel Peters, B.A.
Alessandria Schumacher, B.A.
Anna Stamatogiannakis, B.S.
River Williams, B.S.

Conducted literature screening

Carlye A. Austin, Ph.D.
Ryan Cronk, Ph.D.
Lauren Fitzharris, M.P.H.

viii
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Samantha Goodman, M.S.

Angela Hensel, B.A.
Melinda Hoang, M.P.H.
Allison Killius, B.S.
Alex Kliminsky, M.S.
Shannon McGinnis, Ph.D.
Devon Morgan, M.S.
Joei Robertson, M.Sc.
Courtney Rosenthal, M.S.
Amanda Ross, M.Ed.
Karen E. Setty, Ph.D.
Kelly Shipkowski, Ph.D.
Raquel A. Silva, Ph.D.
Parnian Soleymani, M.S.

Conducted literature screening and data extraction

Katie Duke, Ph.D.
Susan Goldhaber, M.P.H.
Revathi Muralidharan, B.S.
Jennifer Seed, Ph.D.
Codi Sharp, M.P.H.

Conducted literature screening, data extraction, and risk-of-bias assessment

Johanna Rochester, Ph.D.

Conducted literature screening and risk-of-bias assessment

Pam Ross, M.S.P.H.

Conducted data extraction

Robert Shin, M.H.S.

Conducted data extraction and risk-of-bias assessment

Tao Hong, Ph.D.
Alex Lindahl, M.P.H.
Samantha J. Snow, Ph.D., DABT

Conducted risk-of-bias assessment

Louise Assem, Ph.D.
Kristin Bornstein, Ph.D.
Alexandra Goldstone, M.P.H.
Jennifer Hsieh, M.S.P.H.
Yi Lu, Ph.D.

Coordinated peer review

Canden N. Byrd, B.S.

ix
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Designed and executed literature searches, and managed references

Michelle Cawley, M.A., M.L.S.
Jeremy S. Frye, M.S.L.S.
Nicole Vetter, M.S.L.S.

Supported data visualizations

Grace Cooney, B.A.

Edited and formatted monograph

Penelope Kellar, M.S.
Whitney Mitchell, B.S.
Kevin O’Donovan, B.A.

Integrated Laboratory Systems, LLC, Research Triangle Park, North Carolina, USA
Conducted a technical review of the draft monograph
Cynthia J. Willson, D.V.M., Ph.D., DACVP

Boston University, Boston, Massachusetts, USA

Reviewed the systematic review protocol
Thomas Webster, Ph.D.

Brown University, Providence, Rhode Island, USA

Reviewed the systematic review protocol
Joseph Braun, Ph.D.

Columbia University, New York, New York, USA

Reviewed the systematic review protocol
Gail Wasserman, Ph.D.

Health Research Board, Dublin, Ireland

Reviewed the systematic review protocol
Marie Sutton, Ph.D.

University of Massachusetts, Amherst, Massachusetts, USA

Reviewed the systematic review protocol
Thomas Zoeller, Ph.D.

x
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Preface
The National Toxicology Program (NTP) conducted a systematic review of the published
scientific literature because of public concern regarding the potential association between
fluoride exposure and adverse neurodevelopmental and cognitive health effects.
NTP initially published a systematic review of the experimental animal literature in 2016 that
was subsequently expanded to include human epidemiological studies, mechanistic studies, and
newer experimental animal literature (see Appendix B, Table B-1 for document and review
timeline). Because of the high public interest in fluoride’s benefits and potential risks, NTP
asked the National Academies of Sciences, Engineering, and Medicine (NASEM) to conduct an
independent evaluation of the draft NTP Monograph on Fluoride Exposure and
Neurodevelopmental and Cognitive Health Effects (2019 draft monograph dated September 6,
2019) and the revised draft (2020 draft monograph dated September 16, 2020), which addressed
the NASEM committee’s recommendations for improvement. The NASEM committee
determined that, “Overall the revised monograph seems to include a wealth of evidence and a
number of evaluations that support its main conclusion, but the monograph falls short of
providing a clear and convincing argument that supports its assessments….” Thus, NTP has
removed the hazard assessment step and retitled this systematic review of fluoride exposure and
neurodevelopmental and cognitive health effects as a “state-of-the-science” document to indicate
the change. This state-of-the-science document does not include the meta-analysis of
epidemiological studies or hazard conclusions found in previous draft monographs; however, it
provides a comprehensive and current assessment of the scientific literature on fluoride as an
important resource to inform safe and appropriate use. The meta-analysis is a separate peer-
reviewed journal publication (DTT Meta-analysis, Taylor et al. 2024, in press).

xi
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Peer Review
The National Toxicology Program (NTP) conducted a peer review of the monograph entitled
Draft NTP Monograph on the State of the Science Concerning Fluoride Exposure and
Neurodevelopmental and Cognitive Health Effects: A Systematic Review by letter in December
2021. Reviewer selection and document review followed established NTP practices. The
reviewers were charged to:
(1) Comment on the technical accuracy and whether the monograph entitled Draft NTP
Monograph on the State of the Science Concerning Fluoride Exposure and
Neurodevelopmental and Cognitive Health Effects: A Systematic Review is clearly
stated and objectively presented.
(2) Determine whether the scientific evidence supports the NTP’s confidence ratings for
the bodies of evidence regarding neurodevelopmental and cognitive health effects
associated with exposure to fluoride.
NTP carefully considered reviewers’ comments in finalizing this monograph.

Peer Reviewers
Ethan Balk, M.D., M.P.H.
Associate Professor (Research)
Brown University School of Public Health
Providence, Rhode Island, USA
Pam Factor-Litvak, Ph.D.
Professor of Epidemiology
Columbia University Medical Center
New York, New York, USA
Erin Haynes, Dr.P.H.
Kurt W. Deuschle Professor in Preventive Medicine and Environmental Health
University of Kentucky College of Public Health
Lexington, Kentucky, USA
Julie Obbagy, Ph.D., R.D.
Nutritionist, Center for Nutrition Policy and Promotion
Office of Nutrition Guidance and Analysis
U.S. Department of Agriculture
Medfield, Massachusetts, USA
Heather Volk, Ph.D.
Associate Professor
Johns Hopkins Bloomberg School of Public Health
Baltimore, Maryland, USA

xii
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

NTP Board of Scientific Counselors Review

The NTP carefully considered the external peer reviewer comments in preparing the NTP
Monograph on the State of the Science Concerning Fluoride Exposure and Neurodevelopmental
and Cognitive Health Effects: A Systematic Review with a target date for release of May 18, 2022
(hereafter called the May 2022 NTP Monograph). The NTP Director, Rick Woychik, Ph.D.,
decided to delay publication of the May 2022 NTP Monograph so that it could undergo
additional review prior to publication. In May 2022, the NTP Director asked the chair of the NTP
Board of Scientific Counselors (BSC) to review NTP’s responses to comments received from
external peer reviewers and federal agencies. The BSC Chair and the NTP Director then jointly
made the decision to convene an independent working group of subject-matter experts (Table 1.1
of the BSC Working Group [WG] report) to conduct this review and report its findings to the
BSC. An additional round of interagency review of the May 2022 NTP Monograph was
conducted. The NTP prepared responses to all comments received and carefully considered the
comments from federal agencies in preparing the September 2022 NTP Monograph which were
submitted to the BSC for review.
The BSC Working Group (BSC WG) was charged to evaluate the adequacy of the NTP authors’
responses to external peer review and/or federal agency comments received during development
of the State of the Science Monograph and the Meta-Analysis Manuscript. They were not
charged to provide independent peer review of the September 2022 NTP Monograph or Meta-
Analysis Manuscript. However, the BSC WG was asked to provide perspectives and suggest
revisions that might improve the quality of either document based on reviewers’ comments and
on NTP authors’ responses to those comments.
In a public meeting on May 4, 2023, the BSC WG presented its report on the September 2022
NTP Monograph to the BSC. Overall, the BSC WG agreed with most (87%) of the NTP authors’
responses to reviewers’ comments (283/325). For the 42 responses that the BSC WG considered
inadequate, they recommended revisions. In addition, the BSC WG identified a subset of these as
9 recurring issues for which they issued 9 global recommendations. The BSC WG rated the NTP
authors’ responses to the reviewer comments as follows:
• 232 reviewer comments where NTP authors’ responses rated adequate; no edits
suggested.
• 51 reviewer comments where NTP authors’ responses rated adequate; edits suggested.
• 42 reviewer comments where NTP authors’ responses rated inadequate; revisions
recommended.
After deliberations and discussion, the BSC voted to accept the BSC WG report in full, with the
condition of one correction, and encouraged the NTP to not delay publication of this important
document any further. The BSC accepted the BSC WG report at a meeting on May 16, 2023.
The NTP authors responded to the BSC WG comments and carefully considered suggested edits
and recommendations in finalizing this monograph. As part of the edits, the NTP authors added
an Addendum in response to a request from the BSC WG to update the literature assessing
fluoride exposures and children’s IQ to match the timeframe for literature included in a

xiii
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

companion systematic review and meta-analysis, which is a separate peer-reviewed journal

publication (DTT Meta-analysis, Taylor et al. 2024, in press). 1

1
The NTP authors of this monograph conducted a companion systematic review and meta-analysis of fluoride
exposure and children’s IQ. Reference to this meta-analysis is cited in this monograph as “(DTT Meta-analysis,
Taylor et al. 2024, in press).”

xiv
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

BSC Working Group Members

David L. Eaton, Ph.D., DABT, FATS (Chair)*
Emeritus Professor, University of Washington
Adjunct Professor of Pharmacology and Toxicology, University of Arizona
*BSC Chair until December 31, 2022; remained as BSC Working Group Chair
Antonia M. Calafat, Ph.D.
Chief, Organic Analytical Toxicology Branch
Division of Laboratory Sciences
National Center for Environmental Health
Centers for Disease Control and Prevention (CDC)
Pamela Den Besten, D.D.S., M.S.
Professor, Department of Orofacial Sciences
School of Dentistry
University of California San Francisco
Stephanie M. Engel, Ph.D.
Professor, Department of Epidemiology
Director, Center for Early Life Exposures and Neurotoxicity
Deputy Director, Center for Environmental Health and Susceptibility
Gillings School of Global Public Health
University of North Carolina Chapel Hill
Michael K. Georgieff, M.D.
Executive Vice Chair and Martin Lenz Harrison Land Grant
Professor, Department of Pediatrics
Director, Center for Neurobehavioral Development
Co-Director, Masonic Institute for the Developing Brain
University of Minnesota Medical School
Matthew J. Maenner, Ph.D.
Chief, Child Development and Disability Branch
Division of Human Development and Disability
National Center on Birth Defects and Developmental Disabilities
Centers for Disease Control and Prevention (CDC)
David Michaels, Ph.D., M.P.H.*
Professor
Department of Environmental and Occupational Health
The Milken School of Public Health
George Washington University
*BSC member until December 31, 2022; remained on BSC Working Group
Sally C. Morton, Ph.D., M.Sc.
Executive Vice President
Knowledge Enterprise
Professor, College of Health Solutions and School of Mathematical and Statistical Sciences
Arizona State University

xv
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Sharon K. Sagiv, Ph.D., M.P.H.

Associate Adjunct Professor, Division of Epidemiology and Biostatistics
Investigator, Center for Environmental Research and Children’s Health
School of Public Health
University of California, Berkeley
Ian J. Saldanha, M.B.B.S., Ph.D., M.P.H.
Associate Professor, Center for Clinical Trials and Evidence Synthesis, Department of
Epidemiology
Johns Hopkins Bloomberg School of Public Health
Adjunct Associate Professor, Department of Health Services Policy and Practice
Brown University School of Public Health

xvi
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Publication Details
Publisher: National Toxicology Program
Publishing Location: Research Triangle Park, NC
ISSN: 2378-5144
DOI: https://doi.org/10.22427/NTP-MGRAPH-8
Report Series: NTP Monograph Series
Report Series Number: 08
Official citation: National Toxicology Program (NTP). 2024. NTP monograph on the state of the
science concerning fluoride exposure and neurodevelopment and cognition: a systematic review.
Research Triangle Park, NC: National Toxicology Program. NTP Monograph 08.

Acknowledgements

This work was supported by the Intramural Research Program (ES103316, ES103317) at the
National Institute of Environmental Health Sciences, National Institutes of Health and performed
for the National Toxicology Program, Public Health Service, U.S. Department of Health and
Human Services under contract GS00Q14OADU417 (Order No. HHSN273201600015U).

Conflict of Interest

Individuals who reviewed the systematic review protocol or meta-analysis protocol, conducted a
technical review of the draft monograph, or served on the peer review panel have certified that
they have no known real or apparent conflict of interest related to fluoride exposure or
neurodevelopmental and cognitive health effects.

xvii
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Abstract
Background: Fluoride is a common exposure in our environment that comes from a variety of
sources and is widely promoted for its dental and overall oral health benefits. Contributions to an
individual’s total exposure come primarily from fluoride in drinking water, food, beverages and
dental products. A 2006 evaluation by the National Research Council (NRC) found support for
an association between consumption of high levels of naturally occurring fluoride in drinking
water and adverse neurological effects in humans and recommended further investigation. The
evidence reviewed at that time was from dental and skeletal fluorosis-endemic regions of China.
Since the NRC evaluation, the number and location of studies examining cognitive and
neurobehavioral effects of fluoride in humans have grown considerably, including several recent
North American prospective cohort studies evaluating prenatal fluoride exposure.
In 2016, the National Toxicology Program (NTP) published a systematic review of the evidence
from experimental animal studies on the effects of fluoride on learning and memory. That
systematic review found a low-to-moderate level of evidence that deficits in learning and
memory occur in non-human mammals exposed to fluoride.
Objective: To conduct a systematic review of the human, experimental animal, and mechanistic
literature to evaluate the extent and quality of the evidence linking fluoride exposure to
neurodevelopmental and cognitive effects in humans.
Method: A systematic review protocol was developed and utilized following the standardized
OHAT systematic review approach for conducting literature-based health assessments. This
monograph presents the current state of evidence associating fluoride exposure with cognitive or
neurodevelopmental health effects and incorporated predefined assessments of study quality and
confidence levels. Benefits of fluoride with respect to oral health are not addressed in this
monograph.
Results: The bodies of experimental animal studies and human mechanistic evidence do not
provide clarity on the association between fluoride exposure and cognitive or
neurodevelopmental human health effects. Human mechanistic studies were too heterogenous
and limited in number to make any determination on biological plausibility.
This systematic review identified studies that assessed the association between estimated fluoride
exposure and cognitive or neurodevelopmental effects in both adults and children, which were
evaluated separately. The most common exposure assessment measures were drinking water
concentrations and estimates of total fluoride exposure, as reflected in biomarkers such as
urinary fluoride. In adults, only two high-quality cross-sectional studies examining cognitive
effects were available. The literature in children was more extensive and was separated into
studies assessing intelligence quotient (IQ) and studies assessing other cognitive or
neurodevelopmental outcomes. Eight of nine high-quality studies examining other cognitive or
neurodevelopmental outcomes reported associations with estimated fluoride exposure. Seventy-
two studies assessed the association between fluoride exposure and IQ in children. Nineteen of
those studies were considered to be high quality; of these, 18 reported an inverse association
between estimated fluoride exposure and IQ in children. The 18 studies, which include 3
prospective cohort studies and 15 cross-sectional studies, were conducted in 5 different
countries. Forty-six of the 53 low-quality studies in children also found evidence of an inverse
association between estimated fluoride exposure and IQ in children.

xviii
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Discussion: Existing animal studies provide little insight into the question of whether fluoride
exposure affects IQ. In addition, studies that evaluated fluoride exposure and mechanistic data in
humans were too heterogenous and limited in number to make any determination on biological
plausibility. The body of evidence from studies in adults is also limited and provides low
confidence that fluoride exposure is associated with adverse effects on adult cognition. There is,
however, a large body of evidence on associations between fluoride exposure and IQ in children.
There is also some evidence that fluoride exposure is associated with other neurodevelopmental
and cognitive effects in children; although, because of the heterogeneity of the outcomes, there is
low confidence in the literature for these other effects. This review finds, with moderate
confidence, that higher estimated fluoride exposures (e.g., as in approximations of exposure such
as drinking water fluoride concentrations that exceed the World Health Organization Guidelines
for Drinking-water Quality of 1.5 mg/L of fluoride) are consistently associated with lower IQ in
children. More studies are needed to fully understand the potential for lower fluoride exposure to
affect children’s IQ.

xix
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Introduction
Fluoride is a common exposure in our environment from a variety of sources and is widely
promoted for its dental and overall oral health benefits. Approximately 67% of the U.S.
population receives fluoridated water through a community water system (CDC 2013). In other
countries, fluoride supplementation has been achieved by fluoridating food products such as salt
or milk. Fluoride supplementation has been recommended to prevent bone fractures (Jones et al.
2005). Fluoride also can occur naturally in drinking water. Other sources of human exposure
include other foods and beverages, industrial emissions, pharmaceuticals, and pesticides (e.g.,
cryolite, sulfuryl fluoride). Soil ingestion is another source of fluoride exposure in young
children (USEPA 2010).
The U.S. Public Health Service (PHS) first recommended that communities add fluoride to
drinking water in 1962. PHS guidance is advisory, not regulatory, which means that while PHS
recommends community water fluoridation as a public health intervention, the decision to
fluoridate water systems is made by state and local governments. For many years, most
fluoridated community water systems used fluoride concentrations ranging from 0.8 to 1.2
milligrams/liter (mg/L) (USDHHS 2015). For community water systems that add fluoride, PHS
now recommends a fluoride concentration of 0.7 mg/L (equal to 0.7 parts per million [ppm]).
Under the Safe Drinking Water Act, the U.S. Environmental Protection Agency (EPA) sets
maximum exposure level standards for drinking water quality. The current enforceable drinking
water standard for fluoride, or the maximum contaminant level (MCL), is 4.0 mg/L. This level is
the maximum amount of fluoride contamination (naturally occurring, not from water
fluoridation) that is allowed in water from public water systems and is set to protect against
increased risk of skeletal fluorosis, a condition characterized by pain and tenderness of the major
joints. EPA also has a non-enforceable secondary drinking water standard of 2.0 mg/L of
fluoride, which is recommended to protect children against the tooth discoloration and/or pitting
that can be caused by severe dental fluorosis during the formative period prior to eruption of
teeth. Although the secondary standard is not enforceable, EPA requires that public water
systems notify the public if and when average fluoride levels exceed 2.0 mg/L (NRC 2006). The
World Health Organization (WHO) set a safe water guideline of 1.5 mg/L of fluoride in drinking
water (first established in 1984 and reaffirmed in 1993 and 2011), which is recommended to
protect against increasing risk of dental and skeletal fluorosis (WHO 2017). We have chosen to
refer to the WHO Drinking water Guideline of 1.5 mg/L fluoride when describing “higher”
fluoride exposure. This example was chosen because, based on an overall assessment of the
epidemiology literature, it represents a useful total fluoride exposure equivalent metric and no
alternative safety guidelines for total fluoride exposure exist. Note that while drinking water
provides the majority of fluoride exposure in many of the studies, total exposure can vary widely
even in optimally fluoridated areas based on personal habits in the use of dental products and
consumption of beverages such as black tea that can contain fluoride.
As of April 2020, 1.08% of persons living in the United States (~3.5 million people) were served
by community water systems (CWS) containing ≥1.1 mg/L naturally occurring fluoride. CWS
supplying water with ≥1.5 mg/L naturally occurring fluoride served 0.59% of the U.S.
population (~1.9 million people), and systems supplying water with ≥2 mg/L naturally occurring
fluoride served 0.31% of the U.S. population (~1 million people) (CDC Division of Oral Health
2020).

1
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Elevated naturally occurring fluoride levels in groundwater (>1.5 mg/L) are prevalent globally in
areas including central Australia, eastern Brazil, sub-Saharan Africa, the southern Arabian
Peninsula, south and east Asia, and western North America (Podgorski and Berg 2022). Regions
of the United States where CWS and private wells contain natural fluoride concentrations of
more than 1.5 mg/L serve over 2.9 million U.S. residents (Hefferon et al. 2024). The U.S.
Geological Survey estimates that 172,000 U.S. residents are served by domestic wells that
exceed EPA’s enforceable standard of 4.0 mg/L fluoride in drinking water, and 522,000 are
served by domestic wells that exceed EPA’s non-enforceable standard of 2.0 mg/L fluoride in
drinking water (USGS 2020).
Health concerns cited in relation to fluoride are skeletal fluorosis, lower intelligence quotient
(IQ) and other neurological effects, cancer, and endocrine disruption. Effects on neurological
function, endocrine function (e.g., thyroid, 2 parathyroid, pineal), metabolic function (e.g.,
glucose metabolism), and carcinogenicity were assessed in the 2006 NRC report, Fluoride in
Drinking Water: A Scientific Review of EPA’s Standards (NRC 2006). The NRC review
considered adverse effects of water fluoride, focusing on a range of concentrations (2–4 mg/L)
above the current 0.7-mg/L recommendation for community water fluoridation. The NRC report
concluded that the Maximum Contaminant Level Goal (MCLG), 4 mg/L, should be lowered to
protect against severe enamel fluorosis and reduce the risk of bone fractures associated with
skeletal fluorosis (NRC 2006). Other than severe fluorosis, NRC did not find sufficient evidence
of negative health effects at fluoride levels below 4 mg/L; however, it concluded that the
consistency of the results of IQ deficits in children exposed to fluoride at 2.5 to 4 mg/L in
drinking water from a few epidemiological studies of Chinese populations appeared significant
enough to warrant additional research on the effects of fluoride on intelligence. The NRC report
noted several challenges to evaluating the literature, including deficiencies in reporting quality,
lack of consideration of all sources of fluoride exposure, incomplete consideration of potential
confounding, selection of inappropriate control subject populations in epidemiological studies,
absence of demonstrated clinical significance of reported endocrine effects, and incomplete
understanding of the biological relationship between histological, biochemical, and molecular
alterations with behavioral effects.
In 2016, the National Toxicology Program (NTP) published a systematic review of the evidence
from experimental animal studies on the potential effects of fluoride exposure on learning and
memory (NTP 2016). That systematic review found a low-to-moderate level of evidence that
deficits in learning and memory occur in experimental animals exposed to fluoride. Given these
findings, NTP decided to conduct additional animal studies before carrying out this full
systematic review and integrate human, animal, and potentially relevant mechanistic evidence in
order to reach human health hazard identification conclusions for fluoride and learning and
memory effects. The NTP (2016) report on the experimental animal evidence focused on
learning and memory and developed confidence ratings for bodies of evidence by life stage of
exposure (i.e., exposure during development or adulthood). This monograph also evaluates two
different age groups in humans (i.e., children and adults) with a focus on cognitive

2
The current review has evaluated the fluoride literature with an eye toward potential thyroid effects because a large
literature base has accumulated examining the interaction of fluoride with iodine uptake by the thyroid gland and
consequential effects on synthesis of thyroid hormones, which are recognized to play significant roles in
neurodevelopment in utero and during early childhood. This literature, along with a detailed proposed mechanism of
action, was recently reviewed by Waugh (2019).

2
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

neurodevelopmental effects in children and cognitive effects in adults. The evaluation of

experimental animal studies in this monograph has been conducted separately from the 2016
experimental animal assessment; however, like the 2016 assessment, it assessed mainly learning
and memory effects in experimental animal studies to determine whether the findings inform the
assessment of cognitive neurodevelopmental effects in children and cognitive effects in adults.
A committee convened by the National Academies of Sciences, Engineering, and Medicine
(NASEM) reviewed earlier drafts of this monograph (September 6, 2019, and September 16,
2020) (NASEM 2020; 2021). The current document incorporates changes stemming from those
reviews, and responses to the 2020 review are available as
Sup01_Monograph_NASEM_Feb_2021.pdf. See Appendix B, Table B-1 for a timeline of key
activities contributing to this 2024 NTP monograph, including document review activities that
have occurred since 2016.

Objective and Specific Aims

Objective
The overall objective of this evaluation was to undertake a systematic review to develop NTP
human health hazard identification conclusions on the association between exposure to fluoride
and neurodevelopmental and cognitive effects based on assessing levels of evidence from human
and non-human animal studies with consideration of the degree of support from mechanistic
data. However, the NASEM Committee’s reviews (NASEM 2020; 2021) of the 2019 and 2020
drafts of the monograph indicated that, “Overall the revised monograph seems to include a
wealth of evidence and a number of evaluations that support its main conclusion, but the
monograph falls short of providing a clear and convincing argument that supports its
assessments….” For this reason, our methods were revised to remove the hazard assessment step
(i.e., the section “Integrate Evidence to Develop Hazard Identification Conclusions” and the
associated section “Translate Confidence Ratings into Level of Evidence for Health Effect”). In
addition, a meta-analysis of the epidemiological studies examining children’s IQ in relation to
fluoride exposure added to the 2020 draft in response to NASEM comments (NASEM 2020) was
removed for further refinement in preparation for a separate publication and is not part of this
document.
Therefore, the objective of this monograph is to undertake a systematic review of the literature
concerning the association between fluoride exposure and neurodevelopmental and cognitive
effects and to determine the level of confidence in that evidence. The assessment was based on
evidence from human and non-human animal studies with consideration of mechanistic
information.

Specific Aims
• Identify literature that assessed neurodevelopmental and cognitive health effects,
especially outcomes related to learning, memory, and intelligence, following
exposure to fluoride in human, animal, and relevant in vitro/mechanistic studies.
• Extract data on potential neurodevelopmental and cognitive health effects from
relevant studies.

3
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Assess the internal validity (risk of bias) of individual studies using pre-defined
criteria.
• Assess effects on thyroid function to help evaluate potential mechanisms of impaired
neurobehavioral 3 function.
• Summarize the extent and types of health effects evidence available.
•Describe limitations of the systematic review, strengths and limitations of the
evidence base, identify areas of uncertainty, as well as data gaps and research needs
for neurodevelopmental and cognitive health effects of fluoride.
Depending on the extent and nature of the available evidence:
• Synthesize the evidence using a narrative approach.
• Rate confidence in the body of evidence for human and animal studies separately
according to one of four statements: High, Moderate, Low, or Very Low/No Evidence
Available.

3
The specific aim in the protocol refers to “impaired neurological function”; however, it was changed to “impaired
neurobehavior function” in this document to use more precise terminology. The overall aim from the protocol
remained the same for this evaluation.

4
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Methods
Problem Formulation and Protocol Development
The research question and specific aims stated above were developed and refined through a
series of problem formulation steps, including:
(1) receipt of a nomination from the public in June 2015 to conduct analyses of fluoride
and developmental neurobehavioral toxicity;
(2) analysis of the extent of evidence available and the merit of pursuing systematic
reviews, given factors such as the extent of new research published since previous
evaluations and whether these new reports address or correct the deficiencies noted in
the literature (OEHHA 2011; NRC 2006; SCHER 2011);
(3) request for information in a Federal Register notice (dated October 7, 2015);
(4) consideration of comments providing a list of studies to review through Federal
Register notice and public comment period from October 7, 2015, to November 6,
2015;
(5) release of draft concept titled Proposed NTP Evaluation on Fluoride Exposure and
Potential for Developmental Neurobehavioral Effects in November 2015;
(6) presentation of draft concept at the NTP Board of Scientific Counselors (BSC)
meeting on December 1–2, 2015;
(7) consideration of comments on NTP’s draft concept from the NTP BSC meeting in
December 2015; and
(8) consideration of input on the draft protocol from review by technical advisors.
The protocol used to conduct this systematic review was posted in June 2017 with updates
posted in May 2019 and September 2020 (https://ntp.niehs.nih.gov/go/785076). 4 The protocol
served as the complete set of methods followed for the conduct of this systematic review. The
OHAT Handbook for Conducting a Literature-Based Health Assessment
(http://ntp.niehs.nih.gov/go/38673) is a source of general systematic review methods that were
selected and tailored in developing this protocol. Options in the OHAT handbook that were not
specifically referred to in the protocol were not part of the methods for the systematic review.
A brief summary of the methods is presented below. Although the methods were revised to
remove the hazard assessment step and meta-analysis from this document, the protocol was not
further revised.

PECO Statements
PECO (Population, Exposure, Comparators and Outcomes) statements were developed as an aid
to identify search terms and appropriate inclusion/exclusion criteria for addressing the overall
research question (effects on neurodevelopmental or cognitive function and thyroid associated

4
NTP conducts systematic reviews following prespecified protocols that describe the review procedures selected and
applied from the general methods outlined in the OHAT Handbook for Conducting a Literature-Based Health
Assessment (http://ntp.niehs.nih.gov/go/38673). The protocol describes project-specific procedures tailored to each
systematic review that supersede the methods in the OHAT Handbook.

5
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

with fluoride exposure) for the systematic review (Higgins and Green 2011). The PECO
statements are listed below for human, animal, and in vitro/mechanistic studies (see Table 1,
Table 2, and Table 3).
Using the PECO statements, the evaluation searched human studies, controlled exposure animal
studies, and mechanistic/in vitro studies for evidence of neurodevelopmental or cognitive
function and thyroid effects associated with fluoride exposure. Mechanistic data can come from a
wide variety of studies that are not intended to identify a disease phenotype. This source of
experimental data includes in vitro and in vivo laboratory studies directed at cellular,
biochemical, and molecular mechanisms and attempt to explain how a substance produces
particular adverse health effects. The mechanistic data were first organized by general categories
(e.g., biochemical effects in the brain and neurons, neurotransmitters, oxidative stress) to
evaluate the available information. To prioritize and consider available mechanistic data, the
categories focused on were those with more robust data at levels of fluoride more relevant to
human exposure. The intent was not to develop a mechanism for fluoride induction of effects on
learning and memory but to evaluate whether a plausible series of mechanistic events exists to
support effects observed in the low-dose region (below approximate drinking-water-equivalent
concentrations of 20 ppm for animal studies) that may strengthen a hazard conclusion if one is
derived.

Table 1. Human PECO (Population, Exposure, Comparator and Outcome) Statement

PECO Element Evidence
Population Humans without restriction as to age or sex, geographic location, or life stage at exposure or
outcome assessment
Exposure Exposure to fluoride based on administered dose or concentration, biomonitoring data (e.g.,
urine, blood, other specimens), environmental measures (e.g., air, water levels), or job title or
residence. Relevant forms are those used as additives for water fluoridation:
• Fluorosilicic acid (also called hydrofluorosilicate; Chemical Abstracts Service
Registry Number [CASRN] 16961-83-4)
• Sodium hexafluorosilicate (also called disodium hexafluorosilicate or sodium
fluorosilicate; CASRN 16893-85-9)
• Sodium fluoride (CASRN 7681-49-4)
• Other forms of fluoride that readily dissociate into free fluoride ions (e.g., potassium
fluoride, calcium fluoride, ammonium fluoride)

Comparators Comparable populations not exposed to fluoride (e.g., exposure below detection levels) or
exposed to lower levels of fluoride 5
Outcomes Neurodevelopmental outcomes, including learning, memory, intelligence, other forms of
cognitive behavior, other neurological/neurobehavioral 6 outcomes (e.g., anxiety, aggression,
motor activity), and biochemical changes in the brain or nervous system tissue; measures of
thyroid function, biochemical changes, or thyroid tissue pathology

5
Note: The human PECO statement in this monograph has been revised since the publication of the protocol to
clarify that “populations not exposed to fluoride” may be due to exposure biomarker concentrations being below the
level of detection.
6
The human PECO statement in the protocol refers to “neurological outcomes”; however, it was changed to
“neurological/neurobehavioral outcomes” in this document to use more precise terminology for the outcomes
included.

6
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Table 2. Animal PECO Statement

PECO Element Evidence
Population Non-human mammalian animal species (whole organism)
Exposure Exposure to fluoride based on administered dose or concentration and biomonitoring data
(e.g., urine, blood, other specimens). Relevant forms are those used as additives for water
fluoridation:
• Fluorosilicic acid (also called hydrofluorosilicate; CASRN 16961-83-4)
• Sodium hexafluorosilicate (also called disodium hexafluorosilicate or sodium
fluorosilicate; CASRN 16893-85-9)
• Sodium fluoride (CASRN 7681-49-4)
• Other forms of fluoride that readily dissociate into free fluoride ions (e.g., potassium
fluoride, calcium fluoride, ammonium fluoride)

Comparators Comparable animals that were untreated or exposed to vehicle-only treatment

Outcomes Neurodevelopmental outcomes, including learning, memory, intelligence, other forms of
cognitive behavior, other neurological/neurobehavioral 7 outcomes (e.g., anxiety, aggression,
motor activity), and biochemical changes in the brain or nervous system tissue; measures of
thyroid function, biochemical changes, or thyroid tissue pathology

Table 3. In Vitro/Mechanistic PECO Statement

PECO Element Evidence
Population Human or animal cells, tissues, or biochemical reactions (e.g., ligand binding assays)
Exposure Exposure to fluoride based on administered dose or concentration. Relevant forms are those
used as additives for water fluoridation:
• Fluorosilicic acid (also called hydrofluorosilicate; CASRN 16961-83-4)
• Sodium hexafluorosilicate (also called disodium hexafluorosilicate or sodium
fluorosilicate; CASRN 16893-85-9)
• Sodium fluoride (CASRN 7681-49-4)
• Other forms of fluoride that readily dissociate into free fluoride ions (e.g., potassium
fluoride, calcium fluoride, ammonium fluoride)

Comparators Comparable cells or tissues that were untreated or exposed to vehicle-only treatment
Outcomes Endpoints related to neurological and thyroid function, including neuronal electrophysiology;
mRNA, gene, or protein expression; cell proliferation or death in brain or thyroid tissue/cells;
neuronal signaling; synaptogenesis, etc.

Literature Search
Main Literature Search
Search terms were developed to identify all relevant published evidence on developmental
neurobehavioral toxicity or thyroid-related health effects potentially associated with exposure to

7
The animal PECO statement in the protocol refers to “neurological outcomes”; however, it was changed to
“neurological/neurobehavioral outcomes” in this document to use more precise terminology for the outcomes
included.

7
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

fluoride by reviewing Medical Subject Headings for relevant and appropriate neurobehavioral
and thyroid-related terms and by extracting key neurobehavioral and thyroid-related health
effects and developmental neurobehavioral terminology from reviews and a sample of relevant
studies. 8 Combinations of relevant subject headings and keywords were subsequently identified.
A test set of relevant studies was used to ensure the search terms retrieved 100% of the test set.
Six electronic databases were searched (see Main Literature Database Search) using a search
strategy tailored for each database (specific search terms used for the PubMed search are
presented in Appendix B; the search strategy for other databases are available in the protocol
https://ntp.niehs.nih.gov/go/785076). A search of PubChem indicated that sodium fluoride was
not found in either the Tox21 or ToxCast databases; therefore, these databases were not included
in the search. No language restrictions or publication-year limits were imposed. These six
databases were searched in December 2016, and the search was regularly updated during the
review process through April 1, 2019.
An additional search was conducted on May 1, 2020, where human epidemiological studies with
primary neurodevelopmental or cognitive outcomes (learning, memory, and intelligence) were
prioritized during screening. The review of the 2020 search results focused only on the human
studies because they formed the basis of the confidence ratings (see Figure 1 for framework to
assess confidence) and conclusions in the September 6, 2019, draft. A supplemental literature
search of Chinese-language databases (described below) was also conducted. See Appendix B,
Table B-1 for a timeline of key activities contributing to this 2024 NTP monograph, including
information relevant to the timing of multiple literature searches.
Publications identified in these searches are categorized as “references identified through
database searches” in Figure 2. Studies identified from other sources or manual review that
satisfy the PECO criteria for inclusion are considered under “references identified through other
sources” in Figure 2. Literature searches for this systematic review were conducted
independently from the literature search conducted for NTP (2016). The current literature search
strategy was based on the search terms used for NTP (2016) and refined for the current
evaluation, including the addition of search terms to identify human studies. Although the review
process identified experimental animal studies prior to 2015, the current assessment did not
evaluate these studies and relied on the NTP (2016) assessment. The focus of the literature
searches for this systematic review was to identify and evaluate for relevance animal studies that
were published since completion of the literature searches for the NTP (2016) assessment in
addition to the human and mechanistic data that were not previously evaluated.

Supplemental Chinese Database Literature Search

In order to identify non-English-language studies that might not appear in databases for the main
literature search, additional searches were developed for non-English-language databases. No
definitive guidance was found on the most comprehensive, highest quality, or otherwise most
appropriate non-English-language databases for health studies of fluoride. Therefore, databases
were chosen that identified non-English-language studies that were not captured in searches of
databases from the main literature search—those previously identified from other resources (see
the Searching Other Resources section below). Multiple non-English-language databases were

8
The terms “study” and “publication” are used interchangeably in this document to refer to a published work drawn
from an original body of research conducted on a defined population.

8
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

explored before two were identified, China National Knowledge Infrastructure (CNKI) and
Wanfang, that covered studies previously identified from other sources. These two Chinese
electronic databases were searched in May 2020 with no language restrictions or publication year
limits. Search terms from the main literature search were refined to focus on human
epidemiological studies. The CNKI and Wanfang databases have character limits in the search
strings; therefore, key terms were prioritized using text analytics to identify the most prevalent
terms from neurodevelopmental or cognitive human epidemiological studies previously
identified as relevant. Search strings were designed to capture known relevant studies that were
previously identified from searching other resources without identifying large numbers of non-
relevant studies [the search strategy for both databases is available in the protocol
(https://ntp.niehs.nih.gov/go/785076)]. Publications retrieved were compared with publications
retrieved from the main literature search, and duplicates were removed. The remaining relevant
publications are categorized as “references identified through database searches” in Figure 2.
New animal and mechanistic references retrieved were scanned for evidence that might extend
the information in the September 6, 2019, draft. Although additional studies were identified, data
that would materially advance the animal and mechanistic findings were not identified; therefore,
these studies were not extracted nor were they added to the draft. A primary goal of the screening
of the newly retrieved human references in the supplemental search of Chinese databases was to
identify studies that evaluated primary neurodevelopmental or cognitive outcomes (i.e., learning,
memory, and intelligence) that may have been missed in previous searches that did not include
the Chinese databases. A secondary goal was to examine whether the non-English-language
studies on the Fluoride Action Network website (http://fluoridealert.org/)—a site used as another
resource to identify potentially relevant studies because it is known to index fluoride
publications—had been selectively presented to list only studies reporting associations with
fluoride 9. Newly retrieved human references were reviewed to identify studies that may have
been missed using previous approaches. Studies identified that evaluated primary
neurodevelopmental or cognitive outcomes were included and either translated or reviewed by an
epidemiologist fluent in Chinese.

Databases Searched
Main Literature Database Search
• BIOSIS (Thomson Reuters)
• EMBASE
• PsycINFO (APA PsycNet)
• PubMed (NLM)
• Scopus (Elsevier)
• Web of Science (Thomson Reuters, Web of Science indexes the journal Fluoride)

Supplemental Chinese Database Literature Search

• CNKI

9
Note: As a result of this examination, NTP found no indication that studies were selectively presented on the
Fluoride Action Network website.

9
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Wanfang

Searching Other Resources

The reference lists of all included studies; relevant reviews, editorials, and commentaries; and
the Fluoride Action Network website were manually searched for additional relevant
publications.

Unpublished Data
Although no unpublished data were included in the review, unpublished data were eligible for
inclusion, provided the owner of the data was willing to have the data made public and peer
reviewed [see protocol (https://ntp.niehs.nih.gov/go/785076) for more details].

Study Selection
Evidence Selection Criteria
In order to be eligible for inclusion, studies had to satisfy eligibility criteria that reflect the PECO
statements in Table 1, Table 2, and Table 3.
The following additional exclusion criteria were applied [see protocol
(https://ntp.niehs.nih.gov/go/785076) for additional details]:
(1) Case studies or case reports. Although there are various definitions of ‘case study’
and ‘case report,’ the terms are used here to refer to publications designed to share
health-related events on a single subject or patient with a disease, diagnosis, or
specific outcome in the presence of a specific exposure (see Table 4 for study design
definitions).
(2) Articles without original data (e.g., reviews, editorials, or commentaries). Reference
lists from these materials, however, were reviewed to identify potentially relevant
studies not identified from the database searches. New studies identified were
assessed for eligibility for inclusion.
(3) Conference abstracts, theses, dissertations, and other non-peer-reviewed reports.

Screening Process
References retrieved from the literature search were independently screened by two trained
screeners at the title and abstract level to determine whether a reference met the evidence
selection criteria. Screening procedures following the evidence-selection criteria in the protocol
were pilot tested with experienced contract staff overseen by NTP. For citations with no abstract
or non-English abstracts, articles were screened based on title relevance (the title would need to
indicate clear relevance); number of pages (articles ≤2 pages were assumed to be conference
reports, editorials, or letters unlikely to contain original data); and/or PubMed Medical Subject
Headings (MeSH). Using this approach, literature was manually screened for relevance and
eligibility against the evidence selection criteria using a structured form in SWIFT-Active
Screener (Sciome) (Howard et al. 2020). While the human screeners review studies, SWIFT-
Active Screener aids in this process by employing a machine-learning software program to
priority-rank studies for screening (Howard et al. 2020). SWIFT-Active Screener also refines a

10
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

statistical model that continually ranks the remaining studies according to their likelihood for
inclusion. In addition, SWIFT-Active Screener employs active learning to continually
incorporate user feedback during title and abstract screening to predict the total number of
included studies, thus providing a statistical basis for a decision about when to stop screening
(Miller et al. 2016). Title and abstract screening was stopped once the statistical algorithm in
SWIFT-Active Screener estimated that 98% of the predicted number of relevant studies were
identified.
Studies that were not excluded during the title and abstract screening were further screened for
inclusion with a full-text review by two independent reviewers using DistillerSR® (Evidence
Partners), a web-based, systematic-review software program with structured forms and
procedures to ensure standardization of the process. Screening conflicts were resolved through
discussion and consultation with technical advisor(s), if necessary. During full-text review,
studies that were considered relevant were tagged to the appropriate evidence streams (i.e.,
human, animal, and/or in vitro). Studies tagged to human or animal evidence streams were also
categorized by outcome as primary neurodevelopmental or cognitive outcomes (learning,
memory, and intelligence); secondary neurobehavioral outcomes (anxiety, aggression, motor
activity, or biochemical); or related to thyroid effects. In vitro data were tagged as being related
to neurological effects or thyroid effects. Translation assistance was sought to assess the
relevance of non-English studies. Following full-text review, the remaining studies were
“included” and used for the evaluation.

Evaluation of SWIFT-Active Screener Results

During the initial title and abstract screening of 20,883 references using SWIFT-Active Screener,
approximately 38% 10 of the studies were manually screened in duplicate to identify an estimated
98.6% of the predicted number of relevant studies using the software’s statistical algorithm
(13,023 references were not screened). SWIFT-Active Screener predicted that there were 739
relevant studies during the initial title and abstract screening, of which 729 were identified and
moved to full-text review. The SWIFT-Active Screener statistical algorithm predicted that 10
relevant studies at the title and abstract level (10 represents 1.4% × 739 predicted relevant
studies; or 739 predicted relevant studies minus 729 identified relevant studies during screening)
were not identified by not screening the remaining 13,023 studies.
To further consider the impact of using SWIFT-Active Screener for this systematic review, the
evaluation team assessed the SWIFT-Active screening results to gain a better understanding of
the relevance of the last group of studies that was screened before 98% predicted recall (i.e., 98%
of the predicted number of relevant studies were identified). The goal was to determine the
likelihood of having missed important studies by not screening all of the literature. To do this,
the evaluation team examined subsets of studies screened in SWIFT-Active Screener for trends
and followed those studies through to full-text review for a final determination of relevance and
potential impact (i.e., whether the studies had data on primary outcomes). Based on this

10
Howard et al. (2020) evaluated the performance of the SWIFT-Active Screener methods for estimating total
number of relevant studies using 26 diverse systematic review datasets that were previously screened manually by
reviewers. The authors found that on average, 95% of the relevant articles were identified after screening 40% of the
total reference list when using SWIFT-Active Screener. In the document sets with 5,000 or more references, 95% of
the relevant articles were identified after screening 34% of the available references, on average, using SWIFT-
Active Screener.

11
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

evaluation, it was estimated that the use of SWIFT-Active Screener may have resulted in missing
one to two relevant human studies and one to two relevant animal studies with primary
neurodevelopmental or cognitive outcomes. Therefore, the use of SWIFT-Active Screener saved
considerable time and resources and is expected to miss very few potentially relevant
publications.

Screening of the May 2020 Literature Search Update

For the May 1, 2020, literature search, only primary human epidemiological studies were
identified for data extraction. The study screening and selection process was focused on the
human studies with primary outcomes for the evaluation because they form the basis of the
confidence ratings and conclusions. Animal in vivo, human secondary outcome-only, and human
and animal mechanistic references were identified as part of the screening process. These studies
were then scanned for evidence that might extend the information in the September 6, 2019,
draft. All included studies from the May 2020 literature search update appear in Appendix C;
however, other than the primary human epidemiological studies, data from the new studies were
not extracted unless they would materially advance the findings.
Note that NTP is aware of a conference abstract by Santa-Marina et al. on a Spanish cohort study
that looked at fluoride exposure and neuropsychological development in children (Santa-Marina
et al. 2019). The evaluation team conducted a targeted literature search in April 2021 to see
whether the data from this study had been published. When no publication was found, the
evaluation team contacted the study authors to inquire about the publication of their data. The
response from the study authors indicated that the study report was being finalized but had not
yet been sent to a journal for review; therefore, it was not considered here. 11

Supplemental Chinese Database Searches and Human Epidemiological

Studies
Supplemental searches were conducted in non-English-language databases (CNKI and
Wanfang). Of the 910 references that were identified in the supplemental Chinese database
searches, 13 relevant studies published in Chinese with primary neurobehavioral or cognitive
outcomes were identified during title and abstract screening (which were not identified through
the main literature searches). Full texts were not found for four studies after an extensive search.
The remaining nine studies for which full texts were retrieved were included and were either
professionally translated or evaluated by an epidemiologist fluent in Chinese for the data
extraction and quality assessment steps described below. If necessary, author inquiries were
conducted in Chinese to obtain missing information relevant to the assessment of the key risk-of-
bias questions described below.

11
NTP is aware that this study was published after April 2021 (Ibarluzea et al. 2021) and, therefore, is not included
in this monograph because it is beyond the dates of the literature search. Even if it had been published earlier, the
study would not have contributed to the body of evidence on children’s IQ because the authors assessed other
neurodevelopmental or cognitive effects, specifically the association between fluoride exposure and
neuropsychological development in children aged 1 year using the Mental Development Index (MDI) of the Bayley
Scales of Infant Development and in children aged 4 years using the General Cognitive Index (GCI) of the
McCarthy Scales of Children’s Abilities (MSCA). The study is included in sensitivity analyses in the DTT meta-
analysis (DTT Meta-analysis, Taylor et al. 2024, in press).

12
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Data Extraction
Extraction Process
Data were collected (i.e., extracted) from included studies by one member of the evaluation team
and checked by a second member for completeness and accuracy. Any discrepancies in data
extraction were resolved by discussion or consultation with a third member of the evaluation
team.

Data Availability
Data extraction was completed using the Health Assessment Workspace Collaborative (HAWC),
an open-source and freely available web-based application. 12 Data extraction elements are listed
separately for human, animal, and in vitro studies in the protocol
(https://ntp.niehs.nih.gov/go/785076). Data for primary and secondary outcomes, as well as
thyroid hormone level data, were extracted from human studies. Studies evaluating only goiters
or thyroid size were not extracted because they do not provide specific information on thyroid
hormone levels that would inform whether a thyroid-mediated mechanism was involved in
fluoride-associated changes in neurodevelopment. All primary outcomes and functional
neurological secondary outcomes (e.g., motor activity) were extracted from animal studies
identified since the NTP (2016) report. For animal mechanistic data, studies were tiered based on
exposure dose (with preference given to fluoride drinking-water-equivalent exposures, which
were calculated using the method described in the NTP (2016) report, of 20 ppm or less as
deemed most relevant to exposures in humans), exposure duration or relevant time window (i.e.,
developmental), exposure route (with preference given to oral exposures over injection
exposures), and commonality of mechanism (e.g., inflammation, oxidative stress, changes in
neurotransmitters, and histopathological changes) were considered pockets of mechanistic data.
Thyroid data were not extracted for animal studies due to inconsistency in the available data in
humans. In vitro studies were evaluated, although data were not extracted from these studies as
none of the findings were considered informative with respect to biological plausibility. The data
extraction results for included studies are publicly available and can be downloaded in Excel
format through HAWC (https://hawcproject.org/assessment/405/) (NTP 2019). Methods for
transforming and standardizing dose levels and results from behavioral tests in experimental
animals are detailed in the protocol (https://ntp.niehs.nih.gov/go/785076).
In 2016, NTP published a systematic review of the evidence from experimental animal studies
on the potential effects of fluoride exposure on learning and memory (NTP 2016). The literature
searches for the current assessment identified and evaluated relevant animal studies published
since the 2016 assessment and also included human and mechanistic data that were not
previously evaluated. Although literature search activities for the current assessment identified
experimental animal studies prior to 2015, the current assessment did not re-evaluate animal
studies published prior to 2015 because these were reviewed in the NTP (2016) assessment.

HAWC (Health Assessment Workspace Collaborative): A Modular Web-based Interface to Facilitate

12

Development of Human Health Assessments of Chemicals (https://hawcproject.org/portal/).

13
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Quality Assessment of Individual Studies

Risk of bias was assessed for individual studies using the OHAT risk-of-bias tool
(https://ntp.niehs.nih.gov/go/riskbias) that outlines a parallel approach to evaluating risk of bias
from human, animal, and mechanistic studies to facilitate consideration of risk of bias across
evidence streams with common terms and categories. The risk-of-bias tool is comprised of a
common set of 11 questions that are answered based on the specific details of individual studies
to develop risk-of-bias ratings for each question. Study design determines the subset of questions
used to assess risk of bias for an individual study (see Table 4). When evaluating the risk of bias
for an individual study, the direction and magnitude of association for any specific bias is
considered.
Assessors were trained with an initial pilot phase undertaken to improve clarity of rating criteria
and to improve consistency among assessors. Studies were independently evaluated by two
trained assessors who answered all applicable risk-of-bias questions with one of four options in
Table 5 following prespecified criteria detailed in the protocol
(https://ntp.niehs.nih.gov/go/785076). The criteria describe aspects of study design, conduct, and
reporting required to reach risk-of-bias ratings for each question and specify factors that can
distinguish among ratings (e.g., what separates “definitely low” from “probably low” risk of
bias).

Key Risk-of-bias Questions

In the OHAT approach, some risk-of-bias questions or elements are considered potentially more
important when assessing studies because these issues are generally considered to have a greater
impact on estimates of the effect size or on the credibility of study results in environmental
health studies. There are three Key Questions for observational human studies: confounding,
exposure characterization, and outcome assessment. Based on the complexity of the possible
responses to these questions in epidemiological studies, considerations made and methods used
for evaluating the Key Questions are provided below. There are also three Key Questions for
experimental animal studies: randomization, exposure characterization, and outcome assessment.
In addition, for animal developmental studies, failure to consider the litter as the unit of analysis
was also a key risk-of-bias concern. When there was not enough information to assess the
potential bias for a risk-of-bias question and authors did not respond to an inquiry for further
information, a conservative approach was followed, and the studies were rated probably high risk
of bias for that question.

Risk-of-bias Considerations for Human Studies

The risk of bias of individual studies in the body of evidence was considered in developing
confidence ratings. The key risk-of-bias questions (i.e., confounding, exposure characterization,
and outcome assessment for human studies) are discussed in the consideration of the body of
evidence. For this assessment, the key risk-of-bias questions, if not addressed appropriately, are
considered to have the greatest potential impact on the results. The other risk-of-bias questions,
including selection of study participants, were also considered and were used to identify any
other risk-of-bias concerns that may indicate serious issues with a study that could cause it to be
considered high risk of bias. No study was excluded based on concerns for risk of bias; however,
the low risk-of-bias studies generally drive the ratings on confidence in the results across the

14
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

body of evidence. Human evidence was evaluated with and without high risk-of-bias studies to
assess the impact of these studies on confidence in the association.
High risk-of-bias studies: Studies rated probably high risk of bias for at least two key risk-of-
bias questions or definitely high for any single question are considered studies with higher
potential for bias (i.e., high risk-of-bias studies) and to be of low quality. Studies could also be
considered high risk of bias if rated probably high risk of bias for one key risk-of-bias question
along with other concerns, including potential for selection bias and concerns with statistical
methods.
Low risk-of-bias studies: The remaining studies (i.e., other than the high risk-of-bias studies)
were considered to have lower potential for bias (i.e., low risk of bias) and to be of high quality.
Appendix E describes strengths and limitations of the low risk-of-bias/high-quality studies
identified during the assessment and clarifies why they are considered to pose low risk of bias.
Details on the statistical analyses are provided in the “Other potential threats” domain in order to
evaluate the adequacy of the statistical approach for individual studies.
Given the number of non-English-language studies in this assessment, the potential for the
translation to introduce bias was examined as described below, and it was determined that
translation of non-English-language studies did not impact evaluation of risk of bias. Thirty-two
of 100 studies included in the entire human body of evidence on neurodevelopmental and
cognitive effects were initially published in a foreign language (Chinese) and were either
translated and published in volume 41 of the journal Fluoride (n = 19) or were translated by the
Fluoride Action Network (n = 13)
(http://fluoridealert.org/researchers/translations/complete_archive/). Most of these studies were
considered to have high potential for bias due to lack of information across the key risk-of-bias
questions. Therefore, in order to assess whether the lack of information relevant to key risk-of-
bias concerns was the result of a loss in translation, the original Chinese publications and the
translated versions of the five studies that had the most potential for being included in the low
risk-of-bias group of studies were reviewed by a team member with Chinese as first language to
determine whether the translations were accurate and whether any of the risk-of-bias concerns
could be addressed (An et al. 1992; Chen et al. 1991 [translated in Chen et al. 2008]; Du et al.
1992 [translated in Du et al. 2008]; Guo et al. 1991 [translated in Guo et al. 2008a]; Li et al.
2009). For all five studies, the translations were determined to be accurate, and there was no
impact of the translations on the key risk-of-bias concerns.

Confounding
Covariates were determined a priori based on factors that are associated with neurodevelopment
or cognition and could be related to fluoride exposure. Covariates that were considered key for
all studies, populations, and outcomes included age, sex, and socioeconomic status (e.g.,
maternal education, household income, marital status, crowding). Additional covariates
considered important for this evaluation, depending on the study population and outcome,
included race/ethnicity; maternal demographics (e.g., maternal age, body mass index [BMI]);
parental behavioral and mental health disorders (e.g., attention deficit hyperactivity disorder
[ADHD], depression); smoking (e.g., maternal smoking status, secondhand tobacco smoke
exposure); reproductive factors (e.g., parity); nutrition (e.g., BMI, growth, anemia); iodine
deficiency/excess; minerals and other chemicals in water associated with neurotoxicity (e.g.,
arsenic, lead); maternal and paternal IQ; and quantity and quality of caregiving environment

15
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

(e.g., Home Observation Measurement of the Environment [HOME] score). To be assigned a

rating of probably low risk of bias for the key risk-of-bias question regarding the confounding
domain, studies were not required to address every important covariate listed; however, studies
were required to address the three key covariates for all studies, the potential for co-exposures, if
applicable (e.g., arsenic and lead, both of which could affect cognitive function), and any other
potential covariates considered important for the specific study population and outcome. For
example, studies of populations in China, India, and Mexico, where there is concern about co-
exposures to high fluoride and high arsenic, were required to address arsenic. If the authors did
not directly specify that arsenic exposures were evaluated, groundwater quality maps were
evaluated (https://www.gapmaps.org/Home/Public) in order to identify areas of China, India, and
Mexico where arsenic is a concern (Podgorski and Berg 2020). If no arsenic measurements were
available for the area, the arsenic groundwater quality predictions from the global arsenic 2020
map were used (Podgorski and Berg 2020). If an area had less than 50% probability of having
arsenic levels greater than 10 µg/L (the WHO guideline concentration), the area was considered
not to have an issue with arsenic that needed to be addressed by the study authors.

Exposure
Fluoride ion is rapidly absorbed from the gastrointestinal tract and is rapidly cleared from serum
by distribution into calcified tissues and urinary excretion (IPCS 2002). There is general
consensus that the best measures of long-term fluoride exposure are bone and/or tooth
measurements, and other than measures of dental fluorosis, these were not performed in any of
the studies reviewed in this document. Prolonged residence in an area with a given fluoride
content in drinking water has been considered in many studies as a proxy for long-term exposure.
Exposure was assessed using a variety of methods in the human body of evidence. Studies
provided varying levels of details on the methods used and employed different exposure
characterization methods to group study subjects into exposed and reference groups. Exposure
metrics included spot urine (from children or mothers during at least one trimester of gestation),
serum, individual drinking water, intake from infant formula, estimated total exposure dose,
municipal drinking water (with residence information), evidence of dental or skeletal fluorosis,
area of residence (endemic versus a non-endemic fluorosis area, with or without individual
validation of exposure), burning coal (with or without fluoride), and occupation type.
Urinary fluoride levels measured during pregnancy and in children include all ingested fluoride
and are considered a valid measure to estimate total fluoride exposure (Villa et al. 2010;
Watanabe et al. 1995); however, the type and timing of urinary sample collection are important
to consider. Urinary fluoride is thought to reflect recent exposure but can be influenced by the
timing of exposure (e.g., when water was last consumed, when teeth were last brushed). When
compared with 24-hour urine samples, spot urine samples are more prone to the influence of
timing of exposure and can also be affected by differences in dilution; however, many studies
attempted to account for dilution either by using urinary creatinine or specific gravity. Good
correlations between 24-hour samples and urinary fluoride concentrations from spot samples
adjusted for urinary dilution have been described (Zohouri et al. 2006). Despite potential issues
with spot urine samples, if authors made appropriate efforts to reduce the concern for bias (e.g.,
accounting for dilution), studies that used this metric were generally considered to have probably
low risk of bias for exposure.

16
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Analytical methods to measure fluoride in biological or water samples also varied, some of
which included atomic absorption, ion-selective electrode methods, colorimetric methods, or the
hexamethyldisiloxane microdiffusion method. “Well-established” methods denote accepted
methods for measuring fluoride levels. As noted in the protocol
(https://ntp.niehs.nih.gov/go/785076), the preferred analytical method is the ion selective
electrode method. However, use of other standard methods such as NIOSH Method 8308 or
other governmental standard methods were considered well established. Any study noting that
they used these methods was rated probably low risk of bias for exposure. In order to be rated
definitely low risk of bias for exposure, a study also had to provide a detailed description of QC
procedures (i.e., direct evidence) that were followed, including such things as use of recovery
rates, blanks, or reference standards.
Individual-level measures of exposure were generally considered more accurate than group-level
measures; however, using group-level measures (e.g., endemic versus non-endemic area) in an
analysis was less of a concern if the study provided water or urinary fluoride levels from some
individuals to verify that there were differences in the fluoride exposure between groups. Studies
that provided results by area and also reported individual urinary or serum fluoride
concentrations or other biochemical measures, including dental fluorosis in the children or
urinary levels in mothers during pregnancy, were considered to have probably low risk of bias.
Ideally, these studies would still need to consider and adjust for area-level clustering; however,
these concerns are captured in evaluations of other potential threats to internal validity.

Outcome
Studies included in this evaluation used a wide variety of methods to measure IQ and other
cognitive effects. Measures of IQ were generally standardized tests of IQ; however, for these
standardized methods to be considered low potential for bias, they needed to be conducted in the
appropriate population or modified for the study population. Because results of many of the tests
to measure neurodevelopment and cognitive function can be subjective, it was important that the
outcome assessors were blind to the fluoride exposure when evaluating the results of the tests. If
the study reported that the assessor was blind to the exposure, this was assumed to mean that the
outcome assessor did not have any knowledge of the exposure, including whether the study
subjects were from high-fluoride communities. If cross-sectional studies collected biomarker
measurements at the time of an IQ assessment, this was considered indirect evidence that the
outcome assessor would not have knowledge of the fluoride exposure unless there was also
potential for the outcome assessor to have knowledge of varying levels of fluoride by study area.
In cases wherein the study did not specify that the outcome assessors were blind, the study
authors were contacted and asked whether the outcome assessors were, in fact, blind to exposure.
When authors responded and indicated that outcome assessors were blind to exposure or that it
was not likely that they would have had knowledge of exposure, this was considered direct or
indirect evidence, respectively, that blinding was not a concern for those studies.
Any discrepancies in ratings between assessors were resolved by a senior technical specialist and
through discussion when necessary to reach the final recorded risk-of-bias rating for each
question along with a statement of the basis for that rating. Members of the evaluation team were
consulted for assistance if additional expertise was necessary to reach final risk-of-bias ratings
based on specific aspects of study design or performance reported for individual studies. Study
procedures that were not reported were assumed not to have been conducted, resulting in an

17
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

assessment of “probably high” risk of bias. Authors were queried by email to obtain missing
information, and responses received were used to update risk-of-bias ratings.

18
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Table 4. OHAT Risk-of-bias Questions and Applicability by Study Design

Human Controlled

Case Report/Case
Cross-sectionale
Experimental

Case-controld
Animala

Cohortc
Trialsb

Seriesf
Risk-of-bias Questions

1. Was administered dose or exposure level adequately randomized? X X

2. Was allocation to study groups adequately concealed? X X
3. Did selection of study participants result in the appropriate comparison groups? X X X
4. Did study design or analysis account for important confounding and modifying variables? X X X X
5. Were experimental conditions identical across study groups? X
6. Were research personnel blinded to the study group during the study? X X
7. Were outcome data complete without attrition or exclusion from analysis? X X X X X
8. Can we be confident in the exposure characterization? X X X X X X
9. Can we be confident in the outcome assessment (including blinding of outcome assessors)? X X X X X X
10. Were all measured outcomes reported? X X X X X X
11. Were there no other potential threats to internal validity? X X X X X X
aExperimental animal studies are controlled exposure studies. Non-human animal observational studies can be evaluated using the design features of observational human studies
such as cross-sectional study design.
bHuman controlled trials are studies in humans with controlled exposure (e.g., randomized controlled trials, non-randomized experimental studies).
cCohort studies are observational studies in humans that examine a cohort prospectively or retrospectively in which all subjects are classified according to their exposure and

followed over time to ascertain disease incidence. Longitudinal cohort studies permit repeated observation of subject characteristics and outcome over time (Caruana et al. 2015).
Although cohort studies may include longitudinal analyses, it is not a prerequisite of the cohort study design.
dCase-control studies are observational studies in humans that compare exposures of individuals who have a specific health effect or disease with exposures of controls who do not

have the health effect or disease. Controls generally come from the same population from which the cases were derived.
eCross-sectional studies are observational studies in humans that examine the relationship between exposures and outcomes or health effects assessed contemporaneously. Cross-

sectional studies include population surveys with individual data (e.g., NHANES) and surveys with aggregate data (i.e., ecological studies).
fA case report (or case study) is a descriptive study of a single individual or small group in which the study of an association between an observed effect and a specific

environmental exposure is based on clinical evaluations and histories of the individual(s). A case series study in environmental epidemiology is designed to share health-related
events on a collection of case reports on subjects with the same or similar health outcome(s) and environmental exposure(s).

19
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Answers to the risk-of-bias questions result in one of the following four risk-of-bias ratings:

Table 5. The Four Risk-of-bias Rating Options

Symbol Description
Definitely Low risk of bias:
++ There is direct evidence of low risk-of-bias practices.
Probably Low risk of bias:
+ There is indirect evidence of low risk-of-bias practices, OR it is deemed that deviations from low
risk-of-bias practices for these criteria during the study would not appreciably bias results, including
consideration of direction and magnitude of bias.
Probably High risk of bias:
− NR There is indirect evidence of high risk-of-bias practices (indicated with “−”), OR there is insufficient
information provided about relevant risk-of-bias practices (indicated with “NR” for not reported).
Both symbols indicate probably high risk of bias.

Definitely High risk of bias:

−− There is direct evidence of high risk-of-bias practices.

Organizing and Rating Confidence in Bodies of Evidence

Health Outcome Categories for Neurodevelopmental and Cognitive Effects
After data were extracted from all studies, the health effects results within the category of
neurodevelopmental or cognitive effects were grouped across studies to develop bodies of
evidence or collections of studies with data on the same or related outcomes. The grouping of
health effect results was not planned a priori. The vast majority of the human studies evaluated
IQ in children as the single outcome; therefore, the discussion of cognitive neurodevelopmental
effects in children focuses on IQ studies with supporting information from data on other
endpoints. Cognitive function in adults was evaluated separately. Consistent with the NTP
(2016) assessment, the primary focus within the animal study body of evidence was on animal
studies with endpoints related to learning and memory.

Considerations for Pursuing a Narrative or Quantitative Evidence

Synthesis
This evaluation provides only a narrative review of the data; however, heterogeneity within the
available evidence was evaluated to determine whether a quantitative synthesis (i.e., meta-
analysis) would be appropriate. Choi et al. (2012) and Duan et al. (2018) conducted meta-
analyses and found that high fluoride exposure was associated with lower IQ scores. Choi et al.
(2012) was able to determine a risk ratio for living in an endemic fluorosis area but was unable to
develop a dose-response relationship. Duan et al. (2018) reported a significant non-linear dose-
response relationship between fluoride dose and intelligence with the relationship stated as most
evident with exposures from drinking water above 4 mg/L (or 4 ppm) fluoride. Duan et al.
(2018) found similar results as Choi et al. (2012) for the standardized mean difference; however,
the majority of the available studies in both analyses compare populations with high fluoride
exposure to those with lower fluoride exposure (with the lower exposure levels frequently in the
range of drinking water fluoridation in the United States). The meta-analysis conducted in

20
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

association with this systematic review further informs this issue (DTT Meta-analysis, Taylor et
al. 2024, in press). 13

Confidence Rating: Assessment of Body of Evidence

The quality of evidence for neurodevelopmental and cognitive function outcomes was evaluated
using the GRADE system for rating the confidence in the body of evidence (Guyatt et al. 2011;
Rooney et al. 2014). More detailed guidance on reaching confidence ratings in the body of
evidence as “high,” “moderate,” “low,” or “very low” is provided in the protocol
(https://ntp.niehs.nih.gov/go/785076). In brief, available human and animal studies on a
particular health outcome were initially grouped by key study design features (controlled
exposure, exposure prior to outcome, individual outcome data, and comparison group), and each
grouping of studies was given an initial confidence rating by those features. Studies with all four
features receive an initial rating of high confidence (e.g., randomized controlled trials,
experimental non-human mammal studies). A body of evidence with high confidence would
have the highest certainty in the conclusions. Studies with three features receive an initial rating
of moderate confidence and typically include epidemiological studies with a prospective cohort
study design or with a cross-sectional study design when there is evidence that exposure
preceded the outcome assessment. A body of evidence based on adequately conducted
observational epidemiological studies that cannot completely rule out residual confounding by
design (e.g., typical cohort studies) begins with an initial rating of moderate confidence. Starting
at this initial rating (see column 1 of Figure 1), potential downgrading of the confidence rating
was considered for factors that decrease confidence in the results (see column 2 of Figure 1).
Potential upgrading of the confidence rating was considered for factors that increase confidence
in the results (see column 3 of Figure 1). Short descriptions of the factors that can decrease or
increase confidence in the body of evidence for human studies are provided below [see protocol
(https://ntp.niehs.nih.gov/go/785076) for additional details related to the human body of
evidence, as well as considerations for experimental animal studies].

Factors to Consider for Potential Downgrading

• Risk of bias: Addresses whether the body of evidence did not account for critical
factors in study quality or design, including confounding bias, selection bias,
exposure assessment, and outcome assessment. Consideration for downgrading the
confidence rating is based on the entire body of evidence, and the evidence is
downgraded when there is substantial bias across most studies that could lead to
decreased confidence in the results and when the studies without substantial bias
could not support the confidence rating. Individual studies are evaluated for risk of
bias based on a set of criteria (as discussed above); magnitude and direction of the
bias are also considered.
• Unexplained inconsistency: Addresses inconsistencies in results across studies of
similar populations and design that can be determined by assessing similarity of point
estimates and extent of overlap between confidence intervals or more formally
through statistical tests of heterogeneity. Sensitivity analysis can be used to assess the

The NTP authors of this monograph conducted a companion systematic review and meta-analysis of fluoride
13

exposure and children’s IQ. Reference to this meta-analysis is cited in this monograph as “(DTT Meta-analysis,
Taylor et al. 2024, in press).”

21
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

impact of specific variables on the outcome. Inconsistencies that can be plausibly

explained by characteristics of the studies (e.g., sex-associated differences) are
typically not used to support a downgrade. A downgrade would only be applied when
there is an inconsistency that cannot be explained and results in reduced confidence in
the body of evidence.
• Indirectness: Addresses generalizability and relevance to the objective of the
assessment. As outlined in the Objective and consistent with the population specified
in the PECO statement, this systematic review evaluated the extent and quality of the
evidence linking fluoride exposure to neurodevelopmental and cognitive effects in
humans without restriction as to age, sex, geographic location, or life stage at
exposure or outcome assessment. Furthermore, the review did not exclude subjects
exposed in occupational settings. All exposure levels and scenarios encountered in
human studies are considered direct (i.e., applicable, generalizable, and relevant to
address the objective of the assessment); therefore, a downgrade for indirectness
would not be applied to bodies of evidence from human studies.
• Imprecision: Addresses confidence associated with variability in quantitative
measures such as effect sizes. The GRADE-based assessment was used to determine
whether there was no serious imprecision or very serious imprecision. Typically, 95%
confidence intervals are used as the primary method to assess imprecision, but
considerations can also be made on whether studies were adequately powered. Meta-
analyses can also be used to determine whether the data are imprecise. When meta-
analyses are available, OHAT evaluates serious imprecision based on the confidence
intervals of the pooled effect estimates. When meta-analyses are not available or
feasible, OHAT considers standard deviations or confidence intervals across studies.
There are two principal considerations in reaching a judgement of no serious
imprecision (see Table 12 of the OHAT Handbook). First, there are no or minimal
indications of large standard deviations (i.e., SD > mean). In addition, for ratio
measures, the ratio of the upper to lower 95% CIs for most studies (or meta-estimate)
is <10 and, for absolute measures (e.g., percent control response), the absolute
difference between the upper and lower 95% CIs for most studies (or meta-estimate)
is <100. A downgrade would occur if the body of evidence was considered to be
imprecise based on a meta-analysis, or if serious or very serious imprecision was
consistently present in the body of evidence. A downgrade is especially likely if
imprecision raised questions as to whether an overall effect was significant.
• Publication bias: Addresses evidence of biased publication practices. Downgrade if
one strongly detects publication bias. Publication bias is difficult to detect but may be
evident if major sections of the research community are not publishing (e.g., absence
of industry, academic, or government studies) on a topic or if there are multiple
instances wherein data from conference abstracts are never published in peer-
reviewed journals. In addition, there are methods included in conducting a meta-
analysis to detect whether there is potential for publication bias, including the use of
fit-and-trim models, which help identify how publication bias may affect the results
of the meta-analysis. Although a meta-analysis is not included in this systematic
review, there are two published meta-analyses (Choi et al. 2012; Duan et al. 2018) in

22
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

addition to the one associated with this systematic review (DTT Meta-analysis,
Taylor et al. 2024, in press) that can be used to address publication bias.

Factors to Consider for Potential Upgrading

• Large magnitude of effect: Factors to consider include the outcome being measured
and the dose or exposure range assessed. The confidence can be upgraded if the body
of evidence is suggestive of a large magnitude of effect. GRADE provides guidance
on what can be considered a large magnitude of effect based on relative risk (i.e.,
suggests one upgrade in confidence if relative risk is greater than 2 and two upgrades
in confidence if greater than 5). However, not all studies provide data as a risk
estimate, and smaller changes, such as increases in blood pressure, may have greater
impact on health at the population level. Consideration for an upgrade is not based on
a single study, and what constitutes a large magnitude of effect will depend on the
outcome and the potential public health impact.
• Dose response: Patterns of dose response are evaluated within and across studies.
Confidence in the body of evidence can be increased when there is sufficient
evidence of a dose-response pattern across multiple studies.
• Consistency: Does not apply in this evaluation. The consideration of a potential
upgrade for consistency is primarily for non-human animal evidence in which it
would be applied to address increased confidence based on an observation of
consistent effects across multiple non-human animal species. For human evidence,
this factor would generally not be applied. Human studies are instead evaluated for
issues of consistency that could result in downgrading confidence for unexplained
inconsistency (see “Factors to Consider for Potential Downgrading” above).
• Consideration of residual confounding: Applies to observational studies and refers to
consideration of unmeasured determinants that are likely to be distributed unevenly
across groups. Residual confounding can push results in either direction, but
confidence in the results is increased when the body of evidence is biased by factors
that counter the observed effect and would cause an underestimation of the effect.
Confounding that would cause an overestimation of the effect is considered under the
risk-of-bias considerations for decreasing confidence.

23
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure 1. Assessing Confidence in the Body of Evidence

Confidence ratings were assessed by the evaluation team for accuracy and consistency, and
discrepancies were resolved by consensus and consultation with technical advisors as needed.
Confidence ratings for the primary outcomes are summarized in evidence profile tables for each
outcome.

24
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Results
Literature Search Results
The electronic database searches retrieved 25,450 unique references with 11 additional
references 14 identified by technical advisors or obtained by manually searching the Fluoride
Action Network website or reviewing reference lists of published reviews and other included
studies. During title and abstract screening, 1,036 references were moved to full-text review and
24,425 were excluded (11,402 by manual screening for not satisfying the PECO criteria and
13,023 based on the SWIFT-Active Screener algorithm). Among the 1,036 references that
underwent full-text review, 547 studies were considered PECO-relevant (see Appendix C for list
of included studies). A few studies assessed data for more than one evidence stream (human,
non-human mammal, and/or in vitro), and several studies assessed more than one type of
outcome (e.g., primary and secondary outcomes). Included studies break down as follows:
• 167 human studies (84 primary only; 13 secondary only; 5 primary and secondary; 8
primary and thyroid; 2 secondary and thyroid; and 55 thyroid only);
• 339 non-human mammal studies (7 primary only; 186 secondary only; 67 primary
and secondary; 6 primary, secondary, and thyroid; 4 secondary and thyroid; and 69
thyroid only); and,
• 60 in vitro/mechanistic studies (48 neurological and 12 thyroid).
Additional details on the screening results are provided in Appendix C. These screening results
are outlined in a study selection diagram that reports numbers of studies excluded at each stage
and documents the reason for exclusion at the full-text review stage (see Figure 2) [using
reporting practices outlined in Page et al. (2021)].

14
These 11 studies (9 human and 2 animal studies) were not identified through the electronic database searches, as
they were not indexed in any of the electronic databases searched. Note that the supplemental search of non-English-
language databases was designed in part to identify non-English-language studies that are not indexed in traditional
bibliographic databases such as PubMed. It was successful in this goal, as multiple studies that were initially only
identified through “other sources” were subsequently captured in the supplemental Chinese database search, leaving
only 11 as identified through other sources. Note that omission of these 11 studies would not impact any confidence
conclusions for this systematic review.

25
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure 2. Study Selection Diagrama

aAn Interactive REFerence Flow (I-REFF) diagram of this information is available here:

https://hawcproject.org/summary/visual/assessment/405/Figure-2/. The I-REFF approach was used to enhance the transparency

of the literature search and screening process using an interactive format that allows readers to track screening decisions at each
level, including both inclusions and exclusions, or use direct searching for references of interest. See Walker et al. (2024) for
general I-REFF methodology.
*Studies from all literature searches conducted during the review excluded at the full-text level for pre-established criteria; see
the Methods section for extraction and search update information. Studies may have been excluded for more than one reason; the
first reason identified was recorded.
**Studies excluded from the 2020 literature searches for reasons other than pre-established criteria; see the Methods section for
extraction and search update information.
***Publications may contain more than one evidence stream, so the numbers will not total the 547 included studies.

Human Neurodevelopmental and Cognitive Data

The body of literature that evaluates the association between estimated fluoride exposure and
neurodevelopmental and cognitive effects in humans is relatively robust with a large number of
studies (n = 100) that cover a wide array of endpoints (see Figure 3). Seventy-two human studies
investigated IQ in children. Additional studies evaluated learning and memory (n = 9 studies) or
other cognitive developmental effects (e.g., total neurobehavioral scores and total mental
capacity index in children, cognitive impairment in adults; n = 15 studies). 15 For this review, the
evidence in children and adults was evaluated separately to address potential differences in the
health impact of fluoride exposure during development versus adulthood.

15
Some studies are included in more than one endpoint category (e.g., IQ and other cognitive developmental effects);
therefore, these counts are not mutually exclusive.

26
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure 3. Number of Epidemiological Studies by Outcome and Age Categoriesa

aInteractive figure and additional study details are available at
https://public.tableau.com/app/profile/ntp.visuals/viz/FluorideTableauDashboards/ReadMe.
Choi et al. (2015) used subtests of the omnibus IQ test reported by the authors as Wechsler Intelligence Scale for Children-
Revised (WISC-IV) to evaluate visuospatial abilities (using block design) and executive function (using digit span). These
endpoints are included in the intelligence (IQ) outcome category as they are subsets of the IQ tests.
Three additional publications based on subsamples (i.e., 50–60 children) of the larger Yu et al. (2018) cohort were identified
(Zhao et al. 2019; Zhao et al. 2020; Zhou et al. 2019) and are not included in the counts of this figure.

Because the majority of studies evaluated intelligence, the following section focuses on IQ in
children followed by separate discussions on other measures of cognitive function and
neurobehavioral effects in children and cognitive effects in adults. There were three studies
identified on the association of fluoride and cognitive neurodevelopment of children or adults in
exposed populations in the United States, but none of them assessed IQ (Malin and Till 2015;
Morgan et al. 1998; Rotton et al. 1982). Studies that evaluated mechanistic data in humans,
including effects on the thyroid, are discussed in the Mechanistic Data in Humans section. Note
that a few studies were identified on congenital neurological malformations and neurological
complications of fluorosis; however, they are not considered further due to the limited number of
studies and the heterogeneity of outcomes evaluated in those studies.

IQ in Children
Seventy-two epidemiological studies were identified that evaluated the association between
estimated fluoride exposure and children’s IQ. Nineteen of the 72 IQ studies were determined to
have low potential for bias (i.e., were of high quality). Looking across the literature, there has
been a progression over the years in the quality of studies conducted to assess the association
between fluoride exposure and IQ in children, with more recent studies including better study

27
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

designs, more precise estimates of fluoride exposure, larger sample sizes, and more sophisticated
statistical analysis. Older studies often had limitations related to study design or methods, and
most of the high risk-of-bias studies (i.e., studies of low quality) were published prior to the 2006
NRC evaluation of fluoride in drinking water. In contrast, 18 of the low risk-of-bias studies were
published after the 2006 NRC evaluation of fluoride in drinking water, and over half of those
were published between 2015 and 2020 (Figure 4).

Figure 4. Number of High- and Low-quality Studies of Fluoride Exposure and IQ in Children by
Year of Publication

Several characteristics of recent studies contribute to higher study quality in the overall body of
literature on children’s IQ and fluoride, including:
• Demonstration that exposure occurred prior to outcome assessment (an important
factor when considering confidence in study results; see Figure 1) either by study
design (e.g., for prospective cohort studies) or analysis (e.g., prevalence of dental
fluorosis in children, limiting study populations to children who lived in the same
area for long periods of time).
• Improved reporting of key study details that are necessary to evaluate study quality
and allow for a more precise analysis of risk of bias.
• Increased consideration of key covariates (e.g., socioeconomic status) including
potential co-exposures (e.g., arsenic or lead intake).
• Increased use of individual-level exposure assessment measures (urine or water) as
well as prenatal fluoride exposure to assess either individual-level fluoride exposure
or—if still using group-level data—to confirm that regions being compared had
differences in fluoride exposure.
• Utilization of more sophisticated sampling techniques for the study populations (e.g.,
stratified multistage random sampling).
• Application of more sophisticated regression approaches (e.g., piecewise linear
regression models, multi-level regression with random effects, or generalized additive
models for longitudinal measurements of fluoride).

28
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• For studies using individual-level exposure assessment measures, application of more

sophisticated regression techniques to account for clustering at the cohort level by
using cohort as a fixed or random effect and by accounting for numerous covariates
that capture the cohort effect.
In addition, newer studies represent more diverse study populations across several countries
(Figure 5), whereas all identified peer-reviewed studies that were published prior to 2006 took
place in a single country (China). The majority of high-quality, low risk-of-bias studies exhibit
these important study design and analysis characteristics, as discussed further in subsequent
sections.

Figure 5. Number of Studies of Fluoride Exposure and IQ in Children by Country and Year of
Publication

All available studies were considered in this evaluation; however, review of the body of evidence
focused on the high-quality, low risk-of-bias studies for two main reasons. First, there are fewer
limitations and greater confidence in the results of the high-quality studies. Second, there are a
relatively large number of high-quality studies (n = 19), such that the body of evidence from
these studies could be used to evaluate confidence in the association between fluoride exposure
and changes in children’s IQ. Therefore, the remainder of the discussion on IQ in children
focuses on the 19 studies with low risk of bias. The high risk-of-bias studies are discussed briefly
relative to their overall support of findings from the low risk-of-bias studies.

Low Risk-of-bias IQ Studies

Overview of Studies
Nineteen studies (3 prospective cohort and 16 cross-sectional studies) with low potential for bias
evaluated the association between estimated fluoride exposure and IQ in children (see Quality
Assessment of Individual Studies section for methods on determining which studies pose low
risk of bias). These IQ studies were conducted in 15 study populations across 5 countries and

29
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

included more than 7,000 children. Specifically, of the 19 low risk-of-bias studies of IQ in
children:
• ten were conducted in four areas of China on seven study populations, 16
• three were conducted in three areas of Mexico on three study populations,
• two were conducted in Canada using the same study population,
• three were conducted in three areas of India on three study populations, and
• one was conducted in Iran.
Most studies measured fluoride in drinking water (n = 15) and/or urine (child or maternal)
(n = 15). Two studies measured fluoride in serum. The IQ studies used a variety of tests to
measure IQ. Because IQ tests should be culturally relevant, the tests used often differed between
studies, reflecting adjustments for the range in populations studied (e.g., western vs. Asian
populations). In some cases, different IQ tests were used to study similar populations. Overall,
these studies used IQ tests that were population- and age-appropriate.
Table 6 provides a summary of study characteristics and key IQ and fluoride findings for the 19
low risk-of-bias studies (organized by country and then by year 17). Several of these studies
conducted multiple analyses and reported results on multiple endpoints. The purpose of the table
is to summarize key findings (independent of whether an association is indicated) from each
study and is not meant to be a comprehensive summary of all results from each study. For each
study, results are summarized for each exposure measure assessed, but results from multiple
analyses using the same exposure assessment measure may not be presented for all studies unless
multiple analyses yielded conflicting results. See Appendix E for additional information on each
study in Table 6, including strengths and limitations, clarifications for why studies are
considered to pose low risk of bias, and information regarding statistical analyses, important
covariates, exposure assessment, and outcome assessment.

16
In this document, “study population” refers to a defined population on which an original body of research was
conducted. The published work drawn from that original body of research is often referred to as a “study.” IQ
studies that report on the same study populations are identified in Table 6.
17
Note: Several ways to organize Table 6 were considered, including grouping studies using fluoride concentrations
in “low” and “high” areas together to illustrate the change in IQ scores, as well as grouping studies by IQ test. While
an association is consistently observed when comparing low to high fluoride areas, comparing changes in IQ scores
across these studies is challenging due to the variability in the exposure levels that are considered “low” and “high.”
There are no consistent definitions of “low” and “high” that apply across all cases, and therefore this organizational
structure for Table 6 was not considered an effective presentation of the data. Regarding grouping studies by IQ test,
as the Raven’s tests were almost exclusively conducted in China, India, and Iran, the current organization by
country, to a large extent, also organizes the studies by IQ test. Therefore, we find the current structure most
accommodating for focusing on results by IQ test and most clear and appropriate for providing a quick summary of
study characteristics and key findings per study.

30
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Table 6. Studies on IQ in Childrena

Study Design
Exposure Assessment Measures Assessment Outcome and
Study (Location/Subjects) Summary of IQ Resultsb,c
and Summary Statistics Timing Methods
[n]
China
Xiang et al. Cross-sectional Drinking water Children IQ: Combined Significant dose-related association of
(2003a)d Wamiao and Xinhuai Mean (SD): 0.36 (0.15) (control), (ages 8–13 Raven's Test for fluoride on IQ score based on drinking water
villages (Sihong 2.47 (0.79) (high fluoride) mg/L years) Rural China quintile levels with significantly lower IQ
County)/school children Children’s urine scores observed at water fluoride levels of
[512] 1.53 mg/L or higher; % of subjects with IQ
Mean (SD): 1.11 (0.39) (control), <80 was significantly increased at water
3.47 (1.95) (high fluoride) mg/L levels 2.46 mg/L or higher; significant
Village of residence (non-endemic inverse correlation between IQ and urinary
vs. endemic fluorosis) fluoride (Pearson correlation coefficient of
−0.164); mean IQ scores for children in non-
endemic region (100.41 ± 13.21) significantly
higher than endemic region (92.02 ± 13.00)
No statistical adjustment for covariates

Ding et al. Cross-sectional Children’s urine Children IQ: Combined Significant inverse association between
(2011) Inner Mongolia Range: 0.1–3.55 mg/L (ages 7–14 Raven’s Test for urinary fluoride and IQ score (each 1-mg/L
(Hulunbuir years) Rural China increase was associated with a decrease in IQ
Drinking water (reported but not score of 0.59 points; 95% CI: −1.09, −0.08)
City)/elementary school used in analyses)
children Adjusted for age
[331] Mean (SD): 1.31 (1.05) mg/L

Xiang et al. Cross-sectional Children’s serum Children IQ: Combined Significant linear trend across quartiles of
(2011)d Wamiao and Xinhuai Mean (SD): 0.041 (0.009) (ages 8–13 Raven’s Test for serum fluoride and children’s IQ score <80
villages (Sihong (control), 0.081 (0.019) (high years) Rural China (adjusted ORs for Q1 and Q2; Q1 and Q3;
County)/school children fluoride) mg/L and Q1 and Q4, respectively: 1; 2.22 [95%
[512] CI: 1.42, 3.47]; and 2.48 [95% CI: 1.85,
3.32]); significant associations at ≥0.05 mg/L
serum fluoride
Adjusted for age and sex

31
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design
Exposure Assessment Measures Assessment Outcome and
Study (Location/Subjects) Summary of IQ Resultsb,c
and Summary Statistics Timing Methods
[n]
Wang et al. Cross-sectional Children’s total fluoride intake Children IQ: Combined Significantly lower mean IQ in the endemic
(2012)d Wamiao and Xinhuai Mean (SD): 0.78 (0.13) (control), (ages 8–13 Raven’s Test for versus non-endemic regions, as reported in
villages (Sihong 3.05 (0.99) (high fluoride) mg/day years) Rural China Xiang et al. (2003a); when high-exposure
County)/school children group was broken into four exposure groups
Village of residence (non-endemic based on fluoride intake, a dose-dependent
[526] vs. endemic fluorosis) decrease in IQ and increase in % with low IQ
Drinking water (reported for observed; significant correlation between
villages but not used in analyses) total fluoride intake and IQ (r = −0.332); for
Mean (SD): 0.36 (0.11) (control), IQ <80, adjusted OR of total fluoride intake
2.45 (0.80) (high fluoride) mg/L per 1-mg/(person/day) was 1.106 (95% CI:
1.052, 1.163)
Adjusted for age and sex

Choi et al. Cross-sectional Drinking water Children IQ: WISC-IV Compared to normal/questionable fluorosis,
(2015) Mianning County/1st GM: 2.20 mg/L (ages 6–8 (block design presence of moderate/severe fluorosis
grade children years) and digit span) significantly associated with lower total
Children’s urine (adjusted β = −4.28; 95% CI: −8.22, −0.33)
[51]
GM: 1.64 mg/L and backward (adjusted β = −2.13; 95% CI:
Severity of fluorosis (Dean Index) −4.24, −0.02) digit span scores; linear
associations between total digit span and log-
transformed urinary fluoride (adjusted
β = −1.67; 95% CI: −5.46, 2.12) and log-
transformed drinking water fluoride (adjusted
β = −1.39; 95% CI: −6.76, 3.98) observed but
not significant; forward digit span had similar
results as backward and total but was not
statistically significant; block design (square
root transformed) not significantly associated
with any measure of fluoride exposure
Adjusted for age and sex, parity, illness
before 3 years old, household income last
year, and caretaker’s age and education

32
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design
Exposure Assessment Measures Assessment Outcome and
Study (Location/Subjects) Summary of IQ Resultsb,c
and Summary Statistics Timing Methods
[n]
Zhang et al. Cross-sectional Drinking water Children IQ: Combined Significant correlation between IQ score and
(2015b) Tianjin City (Jinnan Mean: 0.63 (control), 1.40 (ages 10–12 Raven’s Test for children’s serum fluoride (r = −0.47) and
District)/school children (endemic fluorosis) mg/L (SD not years) Rural China urinary fluoride (r = −0.45); significant
[180] reported) difference in mean IQ score for high-fluoride
area (defined as >1 mg/L in drinking water;
Children’s urine 102.33 ± 13.46) compared with control area
Mean (SD): 1.1 (0.67) (control), (109.42 ± 13.30); % of subjects with IQ <90
2.4 (1.01) (endemic fluorosis) significantly increased in high-fluoride area
mg/L (28.7%) vs. low-fluoride area (8.33%); not
Children’s serum significantly correlated with water fluoride
Mean (SD): 0.06 (0.03) (control), Adjusted for age and sex, if applicable
0.18 (0.11) (endemic fluorosis)
mg/L

Cui et al. Cross-sectional Children’s urine Children IQ: Combined Significant inverse association between IQ
(2018) Tianjin City (districts Median (Q1–Q3): 1.3 (0.9–1.7) (ages 7–12 Raven’s Test for score and log-transformed urinary fluoride
Jinghai and mg/L (boys), 1.2 (0.9–1.6) mg/L years) Rural China (adjusted β = −2.47; 95% CI: −4.93, −0.01)
Dagang)/school (girls) Adjusted for age, mother’s education, family
children member smoking, stress, and anger
[323]

33
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design
Exposure Assessment Measures Assessment Outcome and
Study (Location/Subjects) Summary of IQ Resultsb,c
and Summary Statistics Timing Methods
[n]
Yu et al. Cross-sectional Drinking water Children IQ: Combined Significant difference in mean IQ scores in
(2018)e,f Tianjin City (7 Mean (SD): 0.50 (0.27) (normal), (ages 7–13 Raven’s Test for high water fluoride areas (>1.0 mg/L;
towns)/children 2.00 (0.75) (high) mg/L years) Rural China 106.4 ± 12.3 IQ) compared to the normal
[2,886] water fluoride areas (≤1.0 mg/L;
Children’s urine 107.4 ± 13.0); distribution of the IQ scores
Mean (SD): 0.41 (0.49) (normal), also significantly different (p = 0.003); every
1.37 (1.08) (high) mg/L 0.5-mg/L increase in water fluoride was
associated with a decrease of 4.29 in IQ score
(95% CI: −8.09, −0.48) when exposure was
between 3.40 and 3.90 mg/L; no significant
association between 0.2 and 3.40 mg/L; every
0.5-mg/L increase in urinary fluoride was
associated with a decrease of 2.67 in IQ score
(95% CI: −4.67, −0.68) between 1.60 and
2.50 mg/L but not at levels of 0.01–
1.60 mg/L or 2.50–5.54 mg/L.
Adjusted for age and sex, maternal education,
paternal education, and low birth weight

Cui et al. Cross-sectional Children’s urine Children IQ: Combined Decreasing mean (± SD) IQ score with
(2020) Tianjin City (all <1.6–≥2.5 mg/L (ages 7–12 Raven’s Test increasing urinary fluoride levels (statistical
districts)/school years) significance not reached based on a one-way
children (potentially ANOVA)
some overlap with Cui <1.6 mg/L: 112.16 ± 11.50
et al. (2018)) 1.6–2.5 mg/L: 112.05 ± 12.01
[498]
≥2.5 mg/L: 110 ± 14.92
No statistical adjustment for covariates

34
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design
Exposure Assessment Measures Assessment Outcome and
Study (Location/Subjects) Summary of IQ Resultsb,c
and Summary Statistics Timing Methods
[n]
Wang et al. Cross-sectional Drinking water Children IQ: Combined Significant inverse associations between IQ
(2020b)e Tianjin City (villages Mean (SD): 1.39 (1.01) mg/L (ages 7–13 Raven’s Test for and water and urinary fluoride concentrations
not specified)/school years) Rural China in boys and girls combined based on both
Children’s urine quartiles and continuous measures (water:
children
[571] Mean (SD): 1.28 (1.30) mg/L 1.587 decrease in IQ score per 1-mg/L
increase; urine: 1.214 decrease in IQ score
per 1-mg/L increase); no significant effect
modification of sex
Adjusted for age and sex, BMI, maternal
education, paternal education, household
income, and low birth weight

Mexico
Rocha- Cross-sectional Drinking water Children IQ: WISC- Significant inverse associations between log-
Amador et al. Moctezuma and Salitral Mean (SD): 0.8 (1.4), 5.3 (0.9), 9.4 (ages 6–10 Revised Mexican transformed fluoride and IQ scores (full-scale
(2007) in San Luis Potosi State (0.9) mg/L (3 rural areas) years) Version IQ adjusted βs of −10.2 [water] and −16.9
and 5 de Febrero of [urine]; CIs not reported); arsenic also
Children’s urine present, but the association with arsenic was
Durango State
/elementary school Mean (SD): 1.8 (1.5), 6.0 (1.6), 5.5 smaller (full-scale IQ adjusted βs of −6.15
children (3.3) mg/L (3 rural areas) [water] and −5.72 [urine]; CIs not reported)
[132] Adjusted for blood lead, mother’s education,
SES, height-for-age z-scores, and transferrin
saturation

35
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design
Exposure Assessment Measures Assessment Outcome and
Study (Location/Subjects) Summary of IQ Resultsb,c
and Summary Statistics Timing Methods
[n]
Bashash et al. Cohort (prospective) Maternal urine during pregnancy Children IQ: WASI- Significantly lower child IQ score per 0.5-
(2017) Mexico City/Early Life Mean (SD): 0.90 (0.35) mg/L (ages 6–12 Spanish Version mg/L increase in maternal urinary fluoride
Exposures in Mexico to Children’s urine years) (adjusted β = −2.50; 95% CI: −4.12, −0.59);
Environmental no significant association with children’s
Toxicants (ELEMENT) Mean (SD): 0.82 (0.38) mg/L urine
participants [299] Adjusted for sex, gestational age; weight at
IQ analysis [211] birth; parity (being the first child); age at
outcome measurement; and maternal
characteristics, including smoking history
(ever smoked during the pregnancy vs.
nonsmoker), marital status (married vs. not
married), age at delivery, education, IQ, and
cohort

Soto-Barreras Cross-sectional Children’s urine Children IQ: Raven’s No significant difference in urinary fluoride,
et al. (2019) Chihuahua/school Range: 0.11–2.10 mg/L (ages 9–10 Colored drinking water fluoride, fluoride exposure
children years) Progressive dose, or fluorosis index in subjects across
Drinking water Matrices different IQ grades
[161]
Range: 0.05–2.93 mg/L No statistical adjustment for covariates
Fluoride exposure dose (summary
statistics not reported)
Fluorosis index (summary statistics
not reported)

36
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design
Exposure Assessment Measures Assessment Outcome and
Study (Location/Subjects) Summary of IQ Resultsb,c
and Summary Statistics Timing Methods
[n]
Canada
Green et al. Cohort (prospective) Maternal urine during pregnancy Children IQ: full-scale, Significantly lower full-scale IQ (adjusted
(2019)g 10 cities/Maternal- Mean (SD): 0.51 (0.36) mg/L (0.40 (ages 3–4 performance, and β = −4.49; 95% CI: −8.38, −0.60) and
Infant Research on [0.27] mg/L in non-fluoridated years) verbal using performance IQ (adjusted β = −4.63; 95% CI:
Environmental areas and 0.69 [0.42] mg/L in Wechsler −9.01, −0.25) per 1-mg/L increase in
Chemicals (MIREC) fluoridated areas) Preschool and maternal urinary fluoride in boys but not girls
[512] Primary Scale of (adjusted β = 2.40; 95% CI: −2.53, 7.33 and
Maternal fluoride intake during Intelligence, adjusted β = 4.51; 95% CI: −1.02, 10.05,
Non-fluoridated [238] pregnancy Third Edition respectively) or boys and girls combined
Fluoridated [162] Mean (SD): 0.54 (0.44) mg/day (WPPSI-III) (adjusted β = −1.95; 95% CI: −5.19, 1.28 and
Boys [248] (0.30 [0.26] and 0.93 adjusted β = −1.24; 95% CI: −4.88, 2.40,
[0.43] mg/day, respectively) respectively); significantly lower full-scale
Girls [264] IQ (adjusted β = −3.66; 95% CI: −7.16,
Drinking water
−0.15) per 1-mg increase in maternal fluoride
Mean (SD): 0.31 (0.23) mg/L (0.13 intake (no sex interaction); significantly
[0.06] and 0.59 [0.08] mg/L, lower full-scale IQ (adjusted β = −5.29; 95%
respectively) CI: −10.39, −0.19) per 1-mg/L increase in
water fluoride concentration (no sex
interaction); no significant associations
observed between measures of fluoride and
verbal IQ
Adjusted for sex, city, HOME score, maternal
education, race, and prenatal secondhand
smoke exposure

37
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design
Exposure Assessment Measures Assessment Outcome and
Study (Location/Subjects) Summary of IQ Resultsb,c
and Summary Statistics Timing Methods
[n]
Till et al. Cohort (prospective) Drinking water Children IQ: full-scale, Drinking water
(2020)g 10 cities/ MIREC [398] Mean (SD) (ages 3–4 performance, and Breastfed infants: Lower (not significant)
years) verbal using full-scale IQ (adjusted β = −1.34, 95% CI:
Non-fluoridated [247] For breastfed infants: 0.13 Wechsler
(0.06) mg/L in non-fluoridated −5.04, 2.38) per 0.5-mg/L increase in water
Fluoridated [151] Preschool and fluoride concentration; significantly lower
areas and 0.58 (0.08) mg/L in Primary Scale of
Breastfed as infants fluoridated areas performance IQ (adjusted β = −6.19, 95% CI:
[200] Intelligence, −10.45, −1.94)
For formula-fed infants: 0.13 Third Edition
Formula-fed as infants (0.05) mg/day in non-fluoridated (WPPSI-III) Formula-fed infants: Significantly lower full-
[198] areas and 0.59 (0.07) mg/L in scale IQ (adjusted β = −4.40, 95% CI: −8.34,
fluoridated areas −0.46) per 0.5-mg/L increase in water
fluoride concentration; significantly lower
Infant fluoride intake performance IQ (adjusted β = −9.26, 95% CI:
Mean (SD) −13.77, −4.76)
For breastfed infants: 0.02 Infant fluoride intake
(0.02) mg/day in non-fluoridated Breastfed: No results reported
areas and 0.12 (0.07) mg/day in
fluoridated areas Formula-fed: Lower (not significant) full-
scale IQ (adjusted β = −2.69, 95% CI: −709,
For formula-fed infants: 0.08 3.21) per 0.5-mg/L increase in fluoride intake
(0.04) mg/day in non-fluoridated from formula; significantly lower
areas and 0.34 (0.12) mg/day in performance IQ (adjusted β = −8.76, 95% CI:
fluoridated areas −14.18, −3.34)
Maternal urine during pregnancy Maternal urine during pregnancy+

38
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design
Exposure Assessment Measures Assessment Outcome and
Study (Location/Subjects) Summary of IQ Resultsb,c
and Summary Statistics Timing Methods
[n]
Mean (SD) Lower (not significant) full-scale IQ
Breastfed: 0.42 (0.28) mg/L in (adjusted β = −1.08, 95% CI: −1.54, 0.47) per
non-fluoridated areas and 0.70 0.5-mg/L increase in maternal urinary
(0.39) mg/L in fluoridated areas fluoride++; lower (not significant)
performance IQ (adjusted β = −1.31, 95% CI:
Formula-fed: 0.38 (0.27) mg/L in −3.63, 1.03)++
non-fluoridated areas and 0.64
(0.37) mg/L in fluoridated areas Lower (not significant) performance IQ
(adjusted β = −1.50, 95% CI: −3.41, 0.43) per
0.5-mg/L increase in maternal urinary
fluoride+++; significantly lower full-scale IQ
(adjusted β = −2.38, 95% CI: −4.62,
−0.27)+++
No association between verbal IQ scores and
any measure of fluoride exposure
+Maternal urinary fluoride analyzed as
covariate in the drinking water and infant
fluoride intake from formula models and not
in an individual model
++After additional adjustment for drinking
water and breastfeeding status
+++After additional adjustment for infant
fluoride intake from formula
All models adjusted for maternal education,
maternal race, age at IQ testing, sex, HOME
total score, and secondhand smoke status in
the child’s home (separate analysis also
adjusted for mother’s urinary fluoride)

39
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design
Exposure Assessment Measures Assessment Outcome and
Study (Location/Subjects) Summary of IQ Resultsb,c
and Summary Statistics Timing Methods
[n]
India
Sudhir et al. Cross-sectional Drinking water Children IQ: Raven’s Significant increase in mean and distributions
(2009) Nalgonda District Level 1: <0.7 mg/L (ages 13–15 Standard of IQ grades (i.e., increase in proportion of
(Andhra years) Progressive children with intellectual impairment) with
Level 2: 0.7–1.2 mg/L Matrices increasing drinking water fluoride levels
Pradesh)/school
children Level 3: 1.3–4.0 mg/L No statistical adjustment for covariates
[1,000] Level 4: >4.0 mg/L

Saxena et al. Cross-sectional Drinking water Children (age IQ: Raven’s Significant correlations between IQ grade and
(2012) Madhya Pradesh/school ≥1.5 mg/L (high fluoride group) 12 years) Standard water (r = 0.534) and urinary (r = 0.542)
children Progressive fluoride levels; in adjusted analyses,
Children’s urine Matrices significant increase in mean IQ grade (i.e.,
[170]
Range: 1.7–8.4 mg/L increase in proportion of children with
intellectual impairment) with increasing
urinary fluoride; no significant differences in
the levels of urinary lead or arsenic in
children with the different water fluoride
exposure levels
Covariates included in the analysis were not
reported

Trivedi et al. Cross-sectional Mean (SE) Children IQ: questionnaire Significantly lower mean IQ score in high
(2012) Kachchh, Low-fluoride villages: drinking (ages 12–13 prepared by fluoride villages (92.53 ± 3.13) compared to
Gujarat/school children water: 0.84 (0.38) mg/L years) Professor JH the low-fluoride villages (97.17 ± 2.54);
(6th and 7th grades) Shah (97% differences significant for boys and girls
Children’s urine: 0.42 (0.23) mg/L reliability rating) combined, as well as separately
[84]
High fluoride villages: drinking No statistical adjustment for covariates
water: 2.3 (0.87) mg/L
Children’s urine: 2.69 (0.92) mg/L

40
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design
Exposure Assessment Measures Assessment Outcome and
Study (Location/Subjects) Summary of IQ Resultsb,c
and Summary Statistics Timing Methods
[n]
Iran
Seraj et al. Cross-sectional Drinking water Children IQ: Raven’s Significant inverse association between water
(2012) Makoo/school children Mean (SD): 0.8 (0.3) (normal), 3.1 (ages 6–11 Colored fluoride and IQ score (adjusted β = −3.865
[293] (0.9) (medium), 5.2 (1.1) years) Progressive per 1-mg/L increase in water fluoride); CIs
(high) mg/L Matrices not reported); significantly higher mean IQ
score in normal area (97.77 ± 18.91)
compared with medium (89.03 ± 12.99) and
high (88.58 ± 16.01) areas
Adjusted for age, sex, child’s education level,
mother’s education level, father’s education
level, and fluorosis intensity
ANOVA = analysis of variance; GM = geometric mean; HOME = Home Observation Measurement of the Environment; IQ = intelligence quotient; Q1, Q3 = first and third
quartiles; SD = standard deviations; WASI = Wechsler Abbreviated Scale of Intelligence (Spanish version); WISC-IV = Wechsler Intelligence Scale for Children-Revised (as
reported by Choi et al. 2015).
aIncludes low risk-of-bias studies.
bAssociations between IQ and fluoride levels were reported quantitatively, when possible. For studies with multiple analyses and results, the table summarizes key findings and is

not a comprehensive summary of all findings. Results also indicate when a study reported no association between IQ and fluoride, provided as a qualitative statement of no
association.
cSee Figure A-1 through Figure A-8 for additional study results.
dXiang et al. (2003a), Xiang et al. (2011), and Wang et al. (2012) are based on the same study population.
eYu et al. (2018) and Wang et al. (2020b) are based on the same study population.
fThree additional publications based on a subsample (i.e., 50–60 children) of the larger Yu et al. (2018) cohort were identified (Zhao et al. 2019; Zhao et al. 2020; Zhou et al.

2019); however, these publications focused on mechanistic considerations and are not included in the study totals for IQ because the main study by Yu et al. (2018) is considered a
better representation of the IQ results.
gGreen et al. (2019) and Till et al. (2020) are based on the same study population.

41
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Summary of Results

Overall Findings
The results from 18 of the 19 high-quality (low risk-of-bias) studies (3 prospective cohort studies
from 2 different study populations and 15 cross-sectional studies from 13 different study
populations) that evaluated IQ in children provide consistent evidence of an inverse association
between estimated fluoride exposure and IQ scores (see “Summary of IQ Results” in Table 6)
(Bashash et al. 2017; Choi et al. 2015; Cui et al. 2018; Ding et al. 2011; Green et al. 2019;
Rocha-Amador et al. 2007; Saxena et al. 2012; Seraj et al. 2012; Sudhir et al. 2009; Till et al.
2020; Trivedi et al. 2012; Wang et al. 2012; Wang et al. 2020b; Xiang et al. 2011; Xiang et al.
2003a; Yu et al. 2018; Zhang et al. 2015b). Only one study (Soto-Barreras et al. 2019) did not
observe an association between estimated fluoride exposure and IQ; however, results were not
provided in a manner that allowed for a direct comparison with other low risk-of-bias studies
(see Appendix E for details). A strength of the findings across 18 of 19 low risk-of-bias studies
was the consistent inverse association between estimated fluoride exposure and IQ scores across
studies of varying study designs, exposure assessment measures, and study populations. In
studies that analyzed the sexes separately (n = 5 studies with 2 studies reporting on the same
study population), consistent findings of an inverse association between estimated fluoride
exposures and children’s IQ were generally reported for both sexes. There is some indication of
differential susceptibility between sexes, but ultimately, due to too few high-quality studies that
analyzed exposure and outcome by sex separately and a lack of consistent findings that one sex
is more susceptible, it is unclear whether one sex is more susceptible to the effects of fluoride
exposure than the other. The body of evidence from the 19 low risk-of-bias studies is described
in further detail below. Prospective cohort studies are discussed first, as this study design can
establish a temporal relationship between exposure and outcome, which would contribute to
demonstrating causality and, therefore, providing the strongest evidence for an association
between fluoride exposure during development and IQ in children.

Results by Study Design – Prospective Cohort Studies

As noted above, three prospective cohort studies, conducted in Mexico and Canada, were
identified and considered to reflect a low risk for bias. All three prospective cohort studies found
an inverse association between maternal or child estimated fluoride exposure and IQ in children
(Bashash et al. 2017; Green et al. 2019; Till et al. 2020). Two of the studies (Green et al. 2019;
Till et al. 2020) were based on the same Canadian study population, but one evaluated prenatal
fluoride exposure and the other evaluated postnatal fluoride exposure. The two Canadian studies
were performed in non-fluoridated areas (mean fluoride levels <= 0.13 mg/L) and fluoridated
areas with a mean water fluoride level around the PHS’s recommended fluoride concentration of
0.7 mg/L for community water systems (i.e., approximately half of the WHO Guidelines for
Drinking-water Quality of 1.5 mg/L of fluoride). Green et al. (2019) included maternal urinary
fluoride, maternal fluoride intake, and water fluoride concentrations, while Till et al. (2020) used
fluoride intake from formula or water concentrations in formula-fed versus breastfed infants.
Multiple analyses were conducted in each prospective study, and results by analysis for the three
prospective studies are discussed below. In summary, although not every analysis found a
statistically significant association, together the three studies provided consistent evidence of an
inverse association between maternal fluoride levels and IQ scores in the children.

42
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

In the Early Life Exposures in Mexico to Environmental Toxicants cohort, Bashash et al. (2017)
observed a statistically significant inverse association (p-value = 0.01) between IQ scores in
children and prenatal fluoride exposure measured by maternal urinary fluoride (measured during
all three trimesters and included if at least one measurement was available). An increase of
0.5 mg/L of maternal urinary fluoride was associated with a 2.5-point decrease in IQ score [95%
CI: −4.12, −0.59] in boys and girls combined (see Figure A-8). This study also reported an
inverse association between IQ level and children’s urinary fluoride levels (single spot urine
sample); however, this specific result did not achieve statistical significance (a 0.5-mg/L increase
of child urinary fluoride was associated with a 0.89-point decrease in IQ score [95% CI: −2.63,
0.85]) (Bashash et al. 2017).
In the Maternal-Infant Research on Environmental Chemicals cohort, consisting of 10 cities in
Canada, Green et al. (2019) also reported inverse associations between IQ scores in children and
multiple assessment measures of prenatal fluoride exposure, including maternal urinary fluoride,
maternal fluoride intake, and water fluoride concentrations. Green et al. (2019) observed a
statistically significantly lower IQ for boys associated with maternal urinary fluoride averaged
across trimesters (4.49-point decrease in IQ score [95% CI: −8.38, −0.60; p-value = 0.02] per 1-
mg/L increase in maternal urinary fluoride); however, results were not significant in boys and
girls combined (1.95-point decrease in IQ [95% CI: −5.19, 1.28]) and were positive but not
significant in girls (2.40-point increase in IQ [95% CI: −2.53, 7.33]). Other measures of prenatal
exposure (maternal fluoride intake or water fluoride concentrations) were associated with lower
IQ scores in boys and girls combined; the authors found no significant effect measure
modification between child sex and estimated fluoride exposure in these analyses so they did not
report boys and girls separately (Green et al. 2019). Specifically, when evaluating the association
between estimated maternal fluoride intake based on maternal water and beverage consumption
during pregnancy and IQ in children, a 1-mg increase in daily maternal consumption of fluoride
during pregnancy was associated with a significant decrease in IQ score of 3.66 points in boys
and girls combined (95% CI: −7.16, −0.15; p-value = 0.04). Similarly, based on drinking water
concentrations, a 1-mg/L increase of fluoride in drinking water was associated with a significant
5.29-point decrease in IQ score in both boys and girls combined (95% CI: −10.39, −0.19; p-value
<0.05) (Green et al. 2019).
In a study of the same study population as Green et al. (2019) that used fluoride intake from
formula or water concentrations in formula-fed versus breastfed infants, Till et al. (2020)
observed significant inverse associations between performance IQ scores and estimated fluoride
exposures regardless of the comparison used (p-values ≤0.004). They did not observe any
association with verbal IQ, and full-scale IQ was only significantly lower in formula-fed infants
using water fluoride concentrations as the exposure assessment measure (p-value = 0.03).
Breastfed infants and fluoride intake from formula also showed inverse associations but were not
significant.
Taken together, the three prospective cohort studies (based on two North American study
populations) indicate consistency in results across different types of analysis and across two
study populations of an inverse association between estimated fluoride exposure during
development and IQ scores.

43
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Results by Study Design – Cross-sectional Studies

As with the prospective cohort studies, the cross-sectional studies reported a consistent inverse
association between estimated fluoride exposure and IQ scores in children. Fifteen of the 16 low
risk-of-bias cross-sectional studies [i.e., all with the exception of Soto-Barreras et al. (2019)]
consistently demonstrate that exposure to fluoride is associated with lower IQ scores. Fourteen of
these 15 studies [with the exception of Cui et al. (2020)] reported significant associations.
Cross-sectional studies can have limitations, as the study design often cannot ensure that
exposure preceded outcome. This uncertainty reduces confidence in study findings compared
with prospective cohort studies—which, by design, establish that exposure occurred prior to
outcome—and is captured in the outcome assessment. In some cases, cross-sectional studies do
provide indicators of prior exposure (e.g., prevalence of dental fluorosis, limiting study
populations to subjects who lived in the same area for long periods of time). Evidence that
exposure occurred prior to the outcome of interest increases the confidence in results and any
potential association reported in these studies. Of the 16 low risk-of-bias cross-sectional studies,
12 established that exposure preceded the outcome assessment (Choi et al. 2015; Ding et al.
2011; Rocha-Amador et al. 2007; Saxena et al. 2012; Seraj et al. 2012; Soto-Barreras et al. 2019;
Sudhir et al. 2009; Wang et al. 2012; Wang et al. 2020b; Xiang et al. 2011; Xiang et al. 2003a;
Yu et al. 2018). Five studies from different study populations indicated that a large portion of the
exposed children had dental fluorosis (ranging from 43% to 100%) at the time of assessment
(Choi et al. 2015; Ding et al. 2011; Seraj et al. 2012; Sudhir et al. 2009; Yu et al. 2018). Because
dental fluorosis occurs when fluoride is consumed during enamel formation (usually during the
first 6–8 years of life), the presence of dental fluorosis suggests that exposures to fluoride
occurred prior to the outcome assessment. Nine studies from six study populations (including Yu
et al. (2018) and Sudhir et al. (2009) listed above) excluded subjects who had not lived in the
study area for a specified period of time, sometimes since birth (Rocha-Amador et al. 2007;
Saxena et al. 2012; Soto-Barreras et al. 2019; Sudhir et al. 2009; Wang et al. 2012; Wang et al.
2020b; Xiang et al. 2011; Xiang et al. 2003a; Yu et al. 2018). Because these areas were generally
known to be fluoride-endemic for long periods of time, it can generally be assumed that in these
nine studies, exposure occurred prior to the outcome. Taken together, 12 cross-sectional studies
from 9 study populations provide indicators of prior exposure.

Results by Study Design – Cross-sectional Study Variations

Overall, the cross-sectional studies provide consistent evidence that estimated fluoride exposure
is inversely associated with IQ scores in children. Several cross-sectional studies conducted
multiple analyses (e.g., reported results for multiple exposure metrics, endpoints,
subpopulations). Although some of these variations are heterogeneous and are not comparable
across studies, the consistency of the results across multiple metrics contributes to the confidence
in the data. Table 6 summarizes key results for each of the low risk-of-bias cross-sectional
studies, and a few examples of the within-study variations in results are provided below.
Nine cross-sectional studies (from six study populations) assessed the association between IQ
and multiple exposure assessment measures (Choi et al. 2015; Rocha-Amador et al. 2007;
Saxena et al. 2012; Wang et al. 2012; Wang et al. 2020b; Xiang et al. 2011; Xiang et al. 2003a;
Yu et al. 2018; Zhang et al. 2015b). Lower IQ was consistently observed across different
exposure assessment measures in these studies. Choi et al. (2015), a small pilot study (n = 51),
reported a consistent inverse association between estimated fluoride exposure across all exposure

44
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

assessment measures (drinking water, children’s urine, and severity of fluorosis) and digit span
measures (subtest of the WISC-IV omnibus IQ test); however, results were only statistically
significant when fluoride exposure was based on moderate or severe dental fluorosis in children
(see Figure A-7). Choi et al. (2015) also observed some variation in results by outcome assessed
(i.e., square root transformed block design and digit span [forward, backward, and total]). It was
the only cross-sectional study that did not provide a full IQ score but instead provided results by
specific subtests. The study authors consistently observed an inverse association between
estimated fluoride exposure and results from the digit span subtest (which specifically assesses
executive function); however, results from the block design (square root transformed), a subtest
of the WISC-IV omnibus IQ test that specifically assesses visuospatial function, was not
associated with estimated fluoride exposure. Note that Rocha-Amador et al. (2009) also assessed
visuospatial function, and the authors reported a significant inverse association (p-value <0.001)
between estimated fluoride exposure and visuospatial constructional ability using the Rey-
Osterrieth Complex Figure (ROCF) Test. Ultimately, too few studies were identified that
reported results by subtest of omnibus IQ tests or assessed domains other than IQ (e.g.,
visuospatial function) to examine or explain the variation by outcome observed in Choi et al.
(2015). The only other studies that provided a breakdown of the full IQ score were the
prospective cohort studies by Green et al. (2019) and Till et al. (2020), which provided results
for full-scale IQ as well as results for performance and verbal IQ. In both of these studies, lower
verbal IQ was not associated with estimated fluoride exposure, but lower performance and full-
scale IQ were associated with estimated fluoride exposure. There are too few studies to evaluate
whether there is a specific aspect of IQ testing that is affected by exposure to fluoride, but the
studies nonetheless consistently provide evidence that estimated fluoride exposure is inversely
associated with IQ.
Yu et al. (2018) reported an overall inverse association between IQ and estimated fluoride
exposure across multiple analyses but observed some variation in IQ results by urinary exposure
level. The authors reported inverse associations between IQ and children’s medium- and high-
range urinary fluoride levels (1.60–2.50 mg/L and 2.50–5.54 mg/L, respectively), although
change in IQ score was greater in the medium-range group (2.67 points decrease [95% CI: −4.67,
−0.68]) for every 0.5-mg/L increase of urinary fluoride than in the high-range group (0.84 points
decrease [95% CI: −2.18, 0.50]) (see Figure A-7). No association was reported at low-range
urinary fluoride levels (0.01–1.60 mg/L). An inverse association was also observed between IQ
and drinking water fluoride levels at 3.40–3.90 mg/L (4.29-point decrease in IQ score [95% CI:
−8.09, −0.48]) for every 0.5-mg/L increase in water fluoride). No association was reported
between IQ and drinking water fluoride levels at 0.20–3.40 mg/L (0.04-point decrease in IQ
score [95% CI: −0.33, 0.24] for every 0.5-mg/L increase in water fluoride). The variation by
exposure level in urine could not be verified in the analysis of drinking water exposures because
there were only two water exposure groups (low and high). In a second study (Wang et al.
2020b), authors conducted a categorical analysis using urinary fluoride quartiles with reported
betas per quartile. As observed in Yu et al. (2018), there were decreasing trends in IQ within
each quartile; however, unlike Yu et al. (2018), Wang et al. (2020b) observed a larger decrease
in IQ with each increasing urinary quartile and observed similar results using water fluoride
quartiles (Wang et al. 2020b). Note that Wang et al. (2020b) cannot be compared directly to Yu
et al. (2018) for evaluation at the higher exposure levels because the two studies do not use the
same categorical exposure ranges. Although additional studies may have looked at different
exposure levels, none of these studies provided results in the same manner as Yu et al. (2018)

45
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

and Wang et al. (2020b) (i.e., betas by exposure category). Instead, these other studies provided
an overall beta or mean IQ scores by exposure level. Despite the noted variations among these
studies, the overall results still consistently support an inverse association between estimated
fluoride exposure and IQ.
Two studies (Cui et al. 2018; Zhang et al. 2015b) observed inverse associations between IQ in
children and estimated exposure to fluoride, with variations in results in subpopulations of
children with different genetic variants (see Figure A-7). These were the only two studies that
considered genetic variants as a sub-analysis. Cui et al. (2018) observed a significant inverse
association between log-transformed children’s single spot urinary fluoride and IQ scores (2.47-
point decrease in IQ scores [95% CI: −4.93, −0.01; p-value = 0.049] per ln-mg/L increase in
urinary fluoride), and the association was strongest in subjects with a TT variant (compared with
children with a CC or CT variant) in the dopamine receptor D2 (DRD2) gene (12.31-point
decrease in IQ score [95% CI: −18.69, −5.94; p-value <0.001] per ln-mg/L increase in urinary
fluoride), which, according to the authors, probably resulted in a reduced D2 receptor density
(Cui et al. 2018). Similarly, Zhang et al. (2015b) observed a significant inverse association
between IQ scores and children’s single spot urinary fluoride (2.42-point decrease in IQ scores
[95% CI: −4.59, −0.24; p-value = 0.030] per 1-mg/L increase in urinary fluoride), and the
association was strongest in subjects with a val/val variant (compared with children who carried
the heterozygous or homozygous variant genotypes [met/val or met/met]) in the catechol-O-
methyltransferase (COMT) gene (9.67-point decrease in IQ score [95% CI: −16.80, −2.55; p-
value = 0.003] per 1-mg/L increase in urinary fluoride).
Overall, the cross-sectional studies consistently support a pattern of findings that estimated
fluoride exposure is inversely associated with IQ scores in children. Slight within-study
variations occur that may be associated with study variables such as IQ domains or subsets of IQ
tests in a few studies that conducted multiple analyses, but these variations are heterogenous and
cannot be further explored with the available studies. Despite these few variations, the overall
evidence of an inverse association with IQ is apparent.

Exposure Assessment Measure and Study Population Factors

Low risk-of-bias studies provide consistent evidence that estimated fluoride exposure is
inversely associated with IQ scores across studies using different exposure assessment measures.
In addition to water fluoride levels, studies measured fluoride exposure using single serum
samples in children (Xiang et al. 2011; Zhang et al. 2015b), single spot urine samples in children
(Cui et al. 2018; Ding et al. 2011; Rocha-Amador et al. 2007; Saxena et al. 2012; Wang et al.
2020b; Xiang et al. 2003a; Yu et al. 2018; Zhang et al. 2015b), and prenatal maternal urinary
measures (Bashash et al. 2017; Green et al. 2019), all of which were demonstrated to be
consistently inversely associated with IQ scores (see Figure A-6, Figure A-7, and Figure A-8).
Urine levels encompass all sources of fluoride exposure and provide a better measure of the
totality of exposure. As noted previously, even though some studies measured single spot
samples, which may not be representative of peak exposure, these studies generally provided
evidence that fluoride exposure had been occurring for some time. The consistency in the results
across studies that used different measures of fluoride exposure and different life stages at which
fluoride was measured strengthens confidence in the body of evidence.
The low risk-of-bias studies consistently provide evidence that estimated fluoride exposure is
inversely associated with IQ scores across studies of different study populations. These 19 high-

46
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

quality studies represent diverse populations (n = 15 study populations) across 5 countries.

Eighteen of the 19 studies conducted in Canada (n = 2), China (n = 10), India (n = 3), Iran
(n = 1), and Mexico (n = 2) provide evidence that estimated exposure to fluoride is inversely
associated with IQ scores; 1 study conducted in Mexico did not observe an association but
reported results in a manner that did not allow for a direct comparison with the other studies (see
Appendix E for details). The overall consistency in the study results across study populations
adds strength to the body of evidence.

Exposure Levels
As described in this section, there is evidence supporting the association between higher fluoride
exposure [e.g., as in approximations of exposure such as drinking water that approximates or
exceeds the WHO Guidelines for Drinking-water Quality of 1.5 mg/L of fluoride (WHO 2017)]
and lower IQ in children. However, there is less certainty in the evidence of an association in
populations with lower estimates of fluoride exposures. In the September 6, 2019, draft of this
monograph, NTP conducted a qualitative analysis of children’s IQ studies that 1) evaluated
lower fluoride exposures (<1.5 mg/L) in drinking water and/or urine and 2) provided information
to evaluate dose response (i.e., provided three or more fluoride exposure groups or a dose-
response curve in their publication) in the lower fluoride exposure range. Nine low risk-of-bias
studies met these criteria, which includes the three prospective cohort studies discussed in this
section. Based on the qualitative review of these studies, the evidence of an association between
fluoride exposure below 1.5 mg/L and lower IQ in children appeared less consistent than results
of studies at higher exposure levels.
A draft quantitative dose-response meta-analysis was prepared and included in the September 16,
2020, draft monograph (NTP 2020). This meta-analysis is now an updated, separate peer-
reviewed journal publication and may further inform a discussion on the association between
fluoride exposure levels and IQ in children (DTT Meta-analysis, Taylor et al. 2024, in press).

Sex Considerations
Recent literature suggests that adverse neurodevelopmental effects of early-life exposure to
fluoride may differ depending on timing of exposure and sex of the exposed subject. In a review
of the human and animal literature, Green et al. (2020) concluded that, compared with females,
male offspring appear to be more sensitive to prenatal but not postnatal exposure to fluoride,
with several potential sex-specific mechanisms.
Sex differences were examined in five of the low risk-of-bias studies (in four study populations)
(Green et al. 2019; Trivedi et al. 2012; Wang et al. 2012; Wang et al. 2020b; Xiang et al. 2003a).
In general, sex differences were difficult to assess for trends within different study populations
because few studies in the body of evidence analyzed exposure and stratified results by sex.
Although these five studies reported IQ scores separately for boys and girls, only two of these
studies analyzed fluoride exposure for boys and girls separately (Green et al. 2019; Wang et al.
2020b), which is essential for evaluating whether a differential change in IQ by sex may be
related to higher susceptibility in one sex or higher exposure in that sex. The remaining three
studies stratified results by sex (Trivedi et al. 2012; Wang et al. 2012; Xiang et al. 2003a), but
the analyses were based on area-level exposure data (e.g., low-fluoride village compared with
high fluoride village) and not drinking water or urinary fluoride concentrations. In the five
studies that reported results by sex separately, findings of IQ inversely associated with estimated

47
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

fluoride exposure were generally reported for both sexes. There was some variation in the results
between sexes across study populations and exposure assessment measures, but there is
insufficient evidence to determine whether one sex is more susceptible to the effects of fluoride
exposure than the other.
Green et al. (2019) observed a significant inverse association between maternal urinary fluoride
levels and IQ scores in boys (p-values ≤0.04) but not girls in a Canadian population. Green et al.
(2019) did not find any sex differences in the association between IQ and water fluoride
concentrations. Wang et al. (2020b) evaluated Chinese boys and girls separately and combined
and observed statistically significant decreasing trends in IQ in all groups by urinary fluoride
quartiles (p-values for trend ≤0.035) (see Figure A-7). Similarly, when evaluated as a continuous
variable, spot urinary fluoride levels (per 1-mg/L increase) were significantly associated with
lower IQ scores in girls (−1.379 [95% CI: −2.628, −0.129; p-value = 0.031]), boys (−1.037 [95%
CI: −2.040, −0.035; p-value = 0.043]), and in the sexes combined (−1.214 [95% CI: −1.987,
−0.442; p-value = 0.002]). According to water fluoride quartiles, Wang et al. (2020b) found that
there was a significant trend in the sexes combined, although the decreasing trend in boys and
girls separately did not achieve statistical significance (p-values = 0.077 and 0.055, respectively).
When water fluoride levels were evaluated as a continuous variable (per 1-mg/L increase), there
were significant associations with lower IQ scores in girls (−1.649 [95% CI: −3.201, −0.097]; p-
value = 0.037), boys (−1.422 [95% CI: −2.792, −0.053; p-value = 0.042]), and the sexes
combined (−1.587 [95% CI: −2.607, −0.568]; p-value = 0.002).
The remaining three studies that reported results by sex-based comparisons of areas of high and
low urinary or water fluoride did not report exposure levels separately for boys and girls, which
decreases the utility of the data to evaluate differential susceptibility by sex. Trivedi et al. (2012)
observed significantly lower IQ in children in high fluoride Indian villages compared with low-
fluoride villages with decreases observed in boys and girls separately or combined (p-values
≤0.05) (see Figure A-2). Xiang et al. (2003a) and Wang et al. (2012) provide data on the same
study population in China. There was a significantly lower IQ in the high fluoride area compared
with the low-fluoride area in boys and girls separately and in the sexes combined (p-values
<0.01), although the difference was greater in girls. Because fluoride exposure was not analyzed
for boys and girls separately, it is unclear whether the greater change in IQ scores in girls could
be attributed to higher susceptibility to fluoride exposure or differences in fluoride exposure by
sex.
In summary, it is unclear whether there is a stronger inverse association between fluoride
exposure and IQ in children in one sex over the other due to the limited number of studies that
analyzed exposure and outcome by sex and the lack of a consistent pattern of findings that one
sex is more susceptible. Green et al. (2019) did not observe an association between maternal
urinary fluoride levels and IQ scores in girls but did observe a significant inverse association in
boys. Although this is an indication of higher sensitivity in boys in this analysis, the authors did
not detect this sex difference using other measures of prenatal exposure (maternal fluoride intake
or water fluoride concentrations). Wang et al. (2020b) and Trivedi et al. (2012) reported
statistically significant associations in both boys and girls without indication that one sex may be
more susceptible. Although Xiang et al. (2003a) and Wang et al. (2012) reported a greater
change in IQ in girls than boys, the studies used area-level exposure data, and the authors did not
determine whether fluoride exposure differed in boys versus girls. Therefore, it is unclear
whether this differential result by sex is an indication of higher susceptibility in girls or whether

48
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

it could be explained by a difference in exposure by sex. Overall, there are too few studies that
analyzed exposure and outcome by sex separately to properly evaluate whether there is
differential susceptibility to fluoride exposure by sex, and results from the five low risk-of-bias
studies that do evaluate sex differences indicate that there is no consistent difference by sex
across the different study populations.

Summary of Key Findings for Low Risk-of-bias Children’s IQ Studies

In summary, the high-quality studies (i.e., studies with low potential for bias) consistently
demonstrate an inverse association between IQ scores and estimated fluoride exposure. The
consistency in the direction of the association is observed among studies of varying study
designs, exposure assessment measures, outcome assessment methods, and study populations.
Although some studies that conducted multiple analyses observed within-study variations in
results (e.g., differences between subsets of IQ tests), these variations were unique to individual
studies and did not detract from the overall consistency in the findings that estimated fluoride
exposure is inversely associated with IQ scores.

High Risk-of-bias IQ Studies

The results from 53 studies with higher potential for bias provide evidence that supports the
results from the 19 studies with lower potential for bias. Forty-six of the 53 studies reported an
inverse association between estimated fluoride exposure and IQ scores in children.

Risk of Bias for IQ Studies in Children

The confidence in the human body of evidence was based on studies with the lowest potential for
bias. A total of 19 studies on IQ in children had little or no risk-of-bias concerns, representing a
relatively large body of evidence for low risk-of-bias studies (i.e., 15 study populations across 5
countries evaluating more than 7,000 children). These 19 studies are considered low risk of bias
because they were rated probably low or definitely low risk of bias for at least two of the three
key risk-of-bias questions and did not have any other risk-of-bias concerns that would indicate
serious issues with the studies. Thirteen of the 19 studies were rated definitely low or probably
low risk of bias for all risk-of-bias questions, and the remaining 6 studies were rated probably
high risk of bias for a single question that was judged to have minimal impact on overall
potential for bias. None of the 19 studies had a rating of definitely high risk of bias for any
question. Risk-of-bias ratings for individual studies for all questions are available in Figure D-1
through Figure D-4, with risk-of-bias ratings for IQ studies in children available in Figure D-5
through Figure D-8 and Appendix E. Although the low risk-of-bias studies had minimal or no
concerns, the studies with high overall potential for bias had a number of risk-of-bias concerns,
including potential confounding, poor exposure characterization, poor outcome assessment, and,
in many cases, potential concern with participant selection. The key risk-of-bias questions are
discussed below.

Confounding for IQ Studies in Children

Low Risk-of-bias Studies

As discussed above, there are 19 studies considered to have low risk of bias when assessed
across all risk-of-bias domains. Sixteen of the 19 low risk-of-bias studies [i.e., all with the
exception of Cui et al. (2020), Ding et al. (2011), and Soto-Barreras et al. (2019)] were
considered to have low potential for bias due to confounding because the authors addressed the

49
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

three key covariates for all studies (i.e., age, sex, and socioeconomic status) through study design
or analysis. Other important covariates, including health factors, smoking, and parental
characteristics, were also addressed in many of the low risk-of-bias studies (see Figure 6).
Co-exposures to arsenic and lead were not considered a concern in 18 of 19 low risk-of-bias
studies [i.e., all except for Soto-Barreras et al. (2019)] because the studies addressed the potential
co-exposures, the co-exposures were not considered an issue in the study population, or the
impact of the potential bias on the results was not a concern. Fifteen of 19 low risk-of-bias
studies either addressed potential bias related to co-exposure to arsenic through study design or
analysis or co-exposure to arsenic was unlikely in the study area. All 15 studies observed an
inverse association between IQ and estimated fluoride exposure. Co-exposure to arsenic was not
accounted for in the remaining four low risk-of-bias studies and was the main potential concern
in these studies; however, three of these studies (Wang et al. 2012; Xiang et al. 2011; Xiang et al.
2003a) were still considered low risk of bias for confounding because although arsenic was
observed in the water in the low-fluoride (and not the high-fluoride) comparison areas, which
would bias the association toward the null, an association was still observed. In this case, the
lack of adjustment for arsenic strengthens the evidence for an association and does not represent
a potential concern. The other study did not address arsenic co-exposure and, as noted above,
was conducted in an area that had potential for arsenic exposure to occur (Soto-Barreras et al.
2019); it is also the only low risk-of-bias study that did not observe an association between IQ
and estimated fluoride exposure (see Appendix E for further discussion of the risk-of-bias
concern regarding arsenic for this study). Although Soto-Barreras et al. (2019) did not discuss
arsenic, there is no direct evidence that arsenic was present in the study area. Fourteen studies
accounted for co-exposure to lead through study design or analysis, and all observed an inverse
association between IQ and estimated fluoride exposure. Five studies did not consider co-
exposure to lead; however, for all of these studies, co-exposure to lead was considered unlikely
to have an impact in these study populations as there was no evidence that lead was prevalent or
occurring in relation to fluoride (Cui et al. 2020; Cui et al. 2018; Soto-Barreras et al. 2019; Till et
al. 2020; Trivedi et al. 2012).
There is considerable variation in the specific covariates considered across the 19 low risk-of-
bias studies. The consistency of results across these studies suggests that confounding is not a
concern in this body of evidence. Each of the 18 low risk-of-bias studies that observed an
association between fluoride and IQ (see Summary of Results section above) considered a
unique combination of covariates. The findings of these studies consistently provide evidence of
an inverse association between IQ in children and estimated exposure to fluoride regardless of
the inclusion or absence of consideration of any one or combination of covariates of interest. For
example, maternal or family member smoking was addressed in 7 of the 19 low risk-of-bias
studies, and this did not appear to affect the conclusions. All 7 studies that accounted for
smoking found evidence of an inverse association between estimated fluoride exposure and IQ
scores as did 11 of the 12 studies that did not account for smoking. Similarly, all 16 studies that
addressed the three key covariates (age, sex, SES) (16 of 16 studies) and two of the three studies
that did not fully account for them also found evidence of an inverse association between
estimated fluoride exposure and IQ scores. In summary, when considering the impact of each
covariate (or combinations of covariates) on the consistency of results, no trends are discernable
that would suggest that bias due to confounding has impacted or would explain the consistency

50
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

in findings across the body of evidence that estimated fluoride exposure is associated with IQ in
children.
Five of the low risk-of-bias studies confirmed the robustness of the results by conducting
sensitivity analyses (Bashash et al. 2017; Green et al. 2019; Till et al. 2020; Wang et al. 2020b;
Yu et al. 2018), and none of the sensitivity analyses adjusting for additional covariates found
meaningful shifts in the association between fluoride exposure and IQ or other measures of
cognitive function. Bashash et al. (2017) found that adjusting for HOME score increased the
association between maternal urinary fluoride and children’s IQ. Green et al. (2019) reported that
adjusting for lead, mercury, manganese, perfluorooctanoic acid, and arsenic concentrations did
not substantially alter the associations with IQ. Sensitivity analyses by Yu et al. (2018) that
adjusted for covariates (including age, sex, and socioeconomic status) did not find differences in
the results compared with the primary analyses. Wang et al. (2020b) found the results of the
sensitivity analysis to be the same as the results from the primary analysis. Till et al. (2020)
observed that adjusting for maternal urinary fluoride levels, as a way to consider postnatal
exposure, had little impact on the results.
Among the 19 low risk-of-bias studies, three were identified that have potential for bias due to
confounding (Cui et al. 2020; Ding et al. 2011; Soto-Barreras et al. 2019). This was mainly due
to a lack of details on covariates considered key for all studies (i.e., age, sex, and SES). See
Appendix E for further discussion of the risk-of-bias concerns regarding confounding for
individual studies. Although these three studies have some potential for bias due to confounding,
they are considered to be low risk of bias overall because they have low potential for bias for the
other two key risk-of-bias questions (exposure characterization and outcome assessment), and no
other major concerns for bias were identified. Consistent with the 16 studies that adequately
addressed confounding, two of these three studies also provide evidence of an inverse association
between estimated fluoride exposure and IQ scores in children.
Taken together and considering the consistency in the results despite the variability across
studies in which covariates were accounted for, bias due to confounding is not considered to be a
concern in the body of evidence. The potential for the consistency in results to be attributable to
bias due to confounding in the 19 low risk-of-bias studies is considered low.

51
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure 6. Important Covariates Considered in Low Risk-of-bias IQ Studies Conducted in Children

a
Includes all low risk-of-bias IQ studies in children. Studies are organized as those with an overall risk-of-bias rating for confounding as probably
low (green) followed by those with an overall risk-of-bias rating for confounding as probably high (yellow).
b
Covariates represented here are those considered important for this evaluation. Depending on the specific study population, individual covariates
may be considered a potential confounder, effect measure modifier, and/or co-exposure. See study details provided in HAWC (NTP 2019) for
information on additional covariates.
Factors outlined in blue are key covariates for all studies (subject age, subject sex, SES) and arsenic (which is of particular importance to some
study populations).

52
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

A √ indicates that a covariate was considered. Examples of what it means for a covariate to be “considered”: it was adjusted for in the final
model, it was considered in the model but not included in the final model because it did not change the effect estimate, it was reported to have the
same distribution in both the exposed and unexposed groups, it was reported to not be associated with the exposure or outcome in that specific
study population. For arsenic, a √ might also be used when arsenic was not expected to be an issue because there is no evidence to indicate that
the co-exposure was prevalent or occurring in relation to fluoride. See risk-of-bias explanations in Appendix E [or HAWC (NTP 2019) for
details. A hyphen (-) indicates that the factor was not considered.
c
See the “Notes” column for additional details.
d
Covariates considered measures of SES include SES scaled scores, household/family income, child education, caretaker/parental education, and
occupation/employment.
e
Extent of reported associations varies by study. “Yes” indicates that study authors provided evidence of an association between lower IQ scores
and fluoride exposure.
f
Study reported lower IQ scores with increasing fluoride exposure, but the results did not achieve statistical significance.

High Risk-of-bias Studies

Most high risk-of-bias studies (n = 53) considered important covariates to some degree through
study design or analysis; however, when considering the full scale of potential concerns of bias
due to confounding, all but three of these studies were rated probably or definitely high risk of
bias. The majority of high risk-of-bias studies accounted for one or two of the three covariates
considered key for all studies (age, sex, SES) but did not address all three and did not address
other covariates considered important for the specific study population and outcome. Potential
confounding related to important co-exposures (e.g., arsenic) was often not addressed in high
risk-of-bias studies. In studies in which there was high exposure to fluoride via drinking water
with high naturally occurring fluoride or from the use of coal-containing fluoride, most
researchers did not account for potential exposures to arsenic, which is commonly found in coal
and drinking water in fluoride-endemic areas of China and Mexico.
Despite the lack of adequate consideration of key covariates in the vast majority of high risk-of-
bias studies, the results across most of these studies (46 of 53) consistently provide evidence of
an inverse association between estimated fluoride exposure and IQ, supporting the results
observed in the low risk-of-bias studies. This finding suggests that confounding is likely less of a
concern for the body of evidence as a whole than for any individual study. Although the high
risk-of-bias studies may have more potential for bias due to confounding compared with the low
risk-of-bias studies, the consistent IQ findings across high and low risk-of-bias studies indicate
that the results cannot be explained solely by potential bias due to confounding.

Exposure Characterization in IQ Studies

Low Risk-of-bias Studies

In general, there were few, if any, risk-of-bias concerns regarding exposure characterization in
the low risk-of-bias studies. These studies mainly had individual exposure data based on urine or
water measures with appropriate analyses. Although there are concerns related to using urine
samples (see the Risk-of-bias Considerations for Human Studies section for details), the
evidence suggests that urinary fluoride is a reasonable measure of exposure (Villa et al. 2010;
Watanabe et al. 1995). Using three methods to account for urine dilution, Till et al. (2018)
reported that adjusted risk estimates did not differ from unadjusted estimates. Analyzing the
same study population as Till et al. (2018), Green et al. (2019) found that adjusting for time of
urine collection or time of collection since last void during pregnancy did not substantially affect
associations with IQ results in either boys or girls. In addition, adjusting maternal urinary
fluoride for creatinine did not substantially alter the observed association (Green et al. 2019). To
provide a more accurate and sensitive measurement of maternal urinary fluoride than a single
measurement provides, Green et al. (2019) included only participants with valid fluoride

53
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

measurements at all trimesters in their analysis. Other studies also measured urinary fluoride
multiple times throughout pregnancy (Bashash et al. 2017). Some studies demonstrated
correlations between urinary fluoride and fluoride in drinking water, fluorosis, or estimated dose
based on drinking water concentrations and consumption (Choi et al. 2015; Ding et al. 2011;
Green et al. 2019; Saxena et al. 2012; Yu et al. 2018; Zhang et al. 2015b). Till et al. (2018)
demonstrated that there was a linear association between urinary fluoride concentrations in
pregnant women and drinking water fluoride concentrations regardless of method used to correct
for urine dilution or whether adjustments were made for dilution. Bashash et al. (2017) excluded
exposure outliers and found that doing so did not substantively change the results. Taken
together, these studies suggest that urinary fluoride is a reasonable measure of exposure despite
some potential issues.
All but one low risk-of-bias study was rated probably or definitely low risk of bias for exposure
assessment. Seraj et al. (2012) had potential exposure misclassification and was rated probably
high risk of bias for exposure assessment. Villages were categorized as normal (0.5–1 ppm),
medium (3.1 ± 0.9 ppm), or high (5.2 ± 1.1 ppm) based on average fluoride content in drinking
water in varying seasons over a 12-year period. Mild fluorosis observed in children in the normal
fluoride level group indicates that there may have been higher exposure in this group at some
point in the past; however, this would bias the results toward the null, and the children in the
normal fluoride group had a significantly higher IQ score compared with the medium and high
fluoride groups (p-value = 0.001). There were also significant inverse associations between IQ
scores and fluorosis intensity (p-value = 0.014) and water fluoride concentration when evaluated
as a continuous variable (p-values <0.001). Although there is potential for exposure bias, the
apparent exposure misclassification and inclusion of children with higher fluoride exposure in
the normal group indicate that the association may be greater than what was observed in this
study.

High Risk-of-bias Studies

A frequent, critical limitation among the high risk-of-bias studies was lack of information
regarding exposure or poor exposure characterization. Many of the high risk-of-bias studies
compared only subjects living in two regions with differing levels of fluoride exposure, and
although most of them did provide some differentiation in levels of fluoride between the areas,
limited or no individual exposure information was reported. Among studies that provided
drinking water levels of fluoride in two areas being compared, sufficient information to
determine whether the individual study subjects were exposed to these levels was often not
reported. Some studies also lacked information on fluoride analysis methods and timing of the
exposure assessment measurements. In some cases (n = 3), study areas that were considered
endemic for dental and/or skeletal fluorosis were compared with non-endemic areas, or high-
fluoride areas were compared with low-fluoride areas, with no other information provided on
fluoride levels in the areas (Li et al. 2003 [translated in Li et al. 2008c]; Ren et al. 1989
[translated in Ren et al. 2008]; Sun et al. 1991). Although living in an area endemic for fluorosis
could be an indicator of exposure, these studies did not specify whether the study subjects
themselves had fluorosis. Another study used only dental fluorosis as a measure of fluoride
exposure in subjects who were all from an endemic area with similar drinking water fluoride
levels (Li et al. 2010). In one case, multiple sources of fluoride exposure were assessed
separately without properly controlling for the other sources of exposure, which could bias the
results (Broadbent et al. 2015). Broadbent et al. (2015) assessed fluoride exposure in three ways:

54
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

use of community water in a fluoridated area versus a non-fluoridated area, use of fluoride
toothpaste (never, sometimes, always), or use of fluoride tablets prior to age 5 (ever, never). The
same children were used for each analysis without accounting for fluoride exposure through
other sources. For example, there were 99 children included in the non-fluoridated area for the
community water evaluation, but there is no indication that these 99 children were not some of
the 139 children that had ever used supplemental fluoride tablets or the 634 children that had
always used fluoride toothpaste. Therefore, comparing fluoridated areas to non-fluoridated areas
without accounting for other sources of exposure that might occur in these non-fluoridated areas
would bias the results toward the null.

Outcome Assessment for IQ Studies

Low Risk-of-bias Studies

The low risk-of-bias studies have few concerns regarding outcome assessment. All 19 low risk-
of-bias studies used appropriate methods for measuring IQ in the study population being
assessed, and blinding of outcome assessors was not a concern in 18 of the 19 studies [i.e., all
low risk-of-bias studies except Sudhir et al. (2009)]. Fourteen of these 18 studies reported
blinding of the outcome assessors, or correspondence with the study authors confirmed that it
was not likely an issue. For the remaining 4 of the 18 studies, it was assumed that the outcome
assessors were most likely blind because exposure was assessed via urine or drinking water
obtained at the same time as the outcome assessment in the general population studies. One IQ
study (Sudhir et al. 2009) had concerns for potential bias in the outcome assessment due to lack
of information to determine whether blinding at the time of the outcome assessment was a
concern (see Appendix E for details).

High Risk-of-bias Studies

Among the studies with high risk of bias, the main limitation in the outcome assessment was the
lack of reporting on blinding of the outcome assessor (i.e., whether the outcome was assessed
without knowledge of exposure). Although there is little concern that the children’s knowledge
of their own exposure would bias the way they took the IQ tests, there is potential for bias if the
tests were administered by an interviewer, or if the scoring of results could be subjective (e.g.,
drawing tests), and the interviewer or scorer had knowledge of the children’s exposure. Most of
the studies did not provide sufficient information on the person scoring or administering the tests
or other information on the assessment methods to alleviate concerns for potential interviewer or
reviewer bias.
High risk-of-bias studies were mainly carried out in two separate populations without
information provided that the tests were conducted in a central location. In many cases, the
methods indicated that the tests were conducted at the schools in the study area (indicating that
there was likely knowledge of exposure). In some cases, the outcomes were not considered
sensitive measures (e.g., Seguin Form Board Test to test for IQ), or the test was not considered
appropriate for the study population (e.g., a test validated in a western population was used on a
rural Chinese population).

Confidence Assessment of Findings on IQ in Children

We conclude that there is moderate confidence in the body of evidence that estimated fluoride
exposure is inversely associated with IQ in children. This confidence rating was reached by

55
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

starting with an initial confidence rating based on key study design features of the body of
evidence and then considering factors that may increase or decrease the confidence in that body
of evidence. The initial moderate confidence rating is based on 15 of the 19 low risk-of-bias
studies that have 3 of the 4 key study design features shown in Figure 1 (i.e., exposure occurred
prior to outcome, individual-based outcomes were evaluated, and a comparison group was used).
Three of these studies were prospective cohort studies, and 12 were cross-sectional studies that
provided evidence of long-term, chronic fluoride exposure prior to outcome measurement.
There are nine factors to consider for increasing or decreasing the confidence in the body of
evidence (provided in Figure 1). Discussion of each of these factors in the body of evidence on
fluoride exposure and IQ in children is presented below.
• Risk of bias: Only studies that were considered to have low risk of bias were
included in the moderate confidence rating; therefore, there was no downgrade for
risk-of-bias concerns.
• Unexplained inconsistencies: There was no evidence of unexplained inconsistencies
among the studies of children’s IQ; therefore, there was no downgrade for this factor.
Eighteen of the 19 low risk-of-bias studies reported inverse associations between
estimated fluoride exposure and IQ scores in children. Seventeen of these were
statistically significant inverse associations. The inverse association remained despite
the following: differences in study location and populations (studies were conducted
in 5 different countries on more than 7,000 children from 15 different study
populations), having different study design (prospective and cross-sectional), and
using different fluoride exposure assessment measures (including urinary and
drinking water fluoride). The one study that found an inverse association and did not
reach statistical significance compared three different urinary fluoride categories and
did observe an inverse exposure-response relationship (Cui et al. 2020). The one
study that did not observe an association did not provide results in a comparable
manner and therefore this body of evidence is not considered to have unexplained
inconsistencies (Soto-Barreras et al. 2019).
• Indirectness: IQ in humans is a direct measure of the association of interest;
therefore, no adjustment in confidence is warranted.
• Imprecision: There is no evidence of serious imprecision that would warrant a
downgrade. The eighteen low risk-of-bias studies reported statistically significant
inverse associations between IQ and estimated fluoride exposure, and narrow
confidence intervals or standard deviations lower than the mean effect estimates
(Table 6, Figure A-2 through Figure A-5). Sample sizes were large and there was no
evidence that the studies were inadequately powered. In addition, two previously
published meta-analyses (Choi et al. 2012; Duan et al. 2018) estimated statistically
significant effect estimates with narrow confidence intervals, and included data from
27 and 26 studies, respectively, showing the precision of the association between
fluoride exposure and IQ.
• Publication bias: There is no strong evidence of publication bias; therefore, no
downgrade was applied for publication bias. Two published meta-analyses (Choi et
al. 2012; Duan et al. 2018) did not indicate strong evidence of publication bias. The
draft meta-analysis conducted by the NTP authors in the September 16, 2020, draft

56
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

monograph found no publication bias among the low risk-of-bias studies (NTP 2020).
Among high risk-of-bias studies, adjusting for publication bias using the trim-and-fill
analysis estimated that, in the absence of publication bias, the inverse direction of
association and statistical significance remained, thus indicating that there was no
need to downgrade for publication bias.
• Large magnitude of effect size: Although some individual studies indicated a large
magnitude of effect size, the magnitude of effect was not the same across all studies.
Therefore, the overall data would not support an upgrade due to a large magnitude of
effect size.
• Dose response: Evidence of an exposure-response relationship that could justify an
upgrade to the confidence in the body of evidence is not presented in this monograph.
While the overall findings qualitatively appear less clear in the lower exposure range,
many of the studies that provide data to evaluate exposure response were judged to be
high risk of bias. The meta-analysis conducted in association with this systematic
review further informs this issue and is a separate peer-reviewed journal publication
(DTT Meta-analysis, Taylor et al. 2024, in press).
• Residual confounding: Xiang et al. (2003a), Xiang et al. (2011), and Wang et al.
(2012) studied the same population where arsenic occurred in the area with low
fluoride but did not occur in the area with high fluoride. This would have biased the
results toward the null, but there were significantly lower IQ scores in the area with
high fluoride. The remaining studies do not provide enough information to consider
whether residual confounding occurred for the body of evidence. Note that parental
IQ has the potential to be an important factor when considering residual confounding
based on likely correlations between parental IQ and children’s IQ; however, there is
not sufficient evidence that parental IQ is associated with water fluoride content.
Taken together, the overall data would not support an upgrade due to residual
confounding.
• Consistency: The consideration of a potential upgrade for consistency in the methods
is primarily for non-human animal evidence, where it would be applied to address
increased confidence for consistent effects across multiple non-human animal species.
For human evidence, it is generally not applied, and the data would only be
considered in deciding whether to downgrade for unexplained inconsistency.
Therefore, no upgrade is applied for consistency.
As described above, there are no changes in confidence rating based on any of the possible
upgrade or downgrade factors. The magnitude of effect size and the overall strength and quality
of the human literature base provide moderate confidence in the body of evidence that estimated
fluoride exposure is inversely associated with IQ in children (see the Discussion section for
strengths and limitations of the evidence base). Note that additional, well-designed prospective
cohort studies with individual-level exposure data and outcome measures could provide
increased confidence in the association between fluoride exposure and lower IQ in children.

57
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Other Neurodevelopmental or Cognitive Effects in Children

Low Risk-of-bias Studies

Overview of Studies
Nine low risk-of-bias studies (three prospective cohort and six cross-sectional studies) evaluated
the association between fluoride exposure and cognitive neurodevelopmental effects other than
IQ in children. These nine studies were conducted in multiple study populations in three
countries, specifically:
• three were conducted in three areas of China on three study populations,
• four were conducted in two areas of Mexico on three study populations, and
• two were conducted in Canada using the same study population.
There is considerable heterogeneity across studies, particularly in the different health outcomes
evaluated and ages assessed. Most studies measured fluoride in the drinking water or urine (child
or maternal) with one study using severity of dental fluorosis as an exposure assessment measure
in addition to drinking water and children’s urine. Two of the studies were conducted on infants,
with one evaluating effects within 72 hours of birth (Li et al. 2004 [translated in Li et al. 2008a])
and the other evaluating effects at 3 to 15 months of age (Valdez Jiménez et al. 2017). The
remaining studies were conducted in children of varying ages, ranging from 4 to 17 years. Other
cognitive neurodevelopmental outcomes assessed include neurobehavioral effects in infants,
learning and memory impairment, and learning disabilities such as attention deficit hyperactivity
disorder (ADHD). Few studies measured the same health outcomes, used the same outcome
assessment methods, or evaluated the same age groups.
Table 7 provides a summary of study characteristics and key findings related to other cognitive
neurodevelopmental outcomes and fluoride exposure for the nine low risk-of-bias studies. The
different tests conducted and the populations on which the tests were conducted are also
indicated in Table 7. Several of these studies conducted multiple analyses and reported results on
multiple endpoints. The purpose of the table is to summarize key findings (independent of
whether an association was found) from each study and is not meant to be a comprehensive
summary of all results. For each study, results are summarized for each exposure assessment
measure assessed. Results from multiple analyses using the same exposure assessment measure
may not all be presented unless conflicting results were reported. See Appendix E for additional
information on studies in Table 7, including strengths and limitations, clarifications for why they
are considered to pose low risk of bias, and information regarding statistical analyses, covariates,
exposure assessment, and outcome assessment.

58
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Table 7. Studies on Other Neurodevelopmental and Cognitive Function in Childrena

Exposure Assessment
Study Design
Study Measures and Assessment Timing Outcome and Methods Neurobehavioral Outcome Summaryb
(Location/Subjects) [n]
Summary Statistics
China
Li et al. (2004) Cross-sectional Drinking water Neonates (24– Neurodevelopmental: Significant differences in neurobehavioral
[translated in Li et al. Zhaozhou County, Range: 0.5–1.0 mg/L 72 hours after Neonatal behavioral assessment total scores between high-
2008a] Heilongjiang Province/neonates (control); 1.7–6.0 mg/L delivery) neurological assessment fluoride (36.48 ± 1.09) and control groups
[91] (high) (NBNA) (38.28 ± 1.10) (subjects divided into high
fluoride group and control group based on
Maternal urine during drinking water fluoride levels in place of
pregnancy residence); significant differences in total
Mean (SD): 1.74 score of behavioral capability that includes
(0.96) mg/L (control); measures of non-biological visual
3.58 (1.47) mg/L (high) orientation reaction and biological visual
and auditory orientation reaction between
the two groups (11.34 ± 0.56 in controls
compared to 10.05 ± 0.94 in high-fluoride
group)
No statistical adjustment for covariates
Choi et al. (2015) Cross-sectional Drinking water Children (ages 6– Learning and memory: Outcomes unrelated to the IQ test not
Mianning County/1st grade GM: 2.20 mg/L 8 years) Neuropsychological tests significantly associated with any fluoride
children including WRAML exposure assessment measure
Children’s urine
[51] Visual motor ability: Adjusted for age, sex, parity, illness before
GM: 1.64 mg/L WRAVMA 3 years old, household income last year, and
Severity of fluorosis Motor ability: Finger tapping caretaker’s age and education
(Dean Index) task
Manual dexterity: Grooved
pegboard test
Wang et al. (2020a) Cross-sectional Children’s urine Children (ages 7–13 ADHD and behavior Significant association between
Tongxu County/school children Mean (SD): 1.54 years) measures: Conners’ Parent psychosomatic problems and urinary
[325] (0.89) mg/L Rating Scale-Revised fluoride level (per 1-mg/L increase;
(Chinese version) (CPRS-48) β = 4.01; 95% CI: 2.74, 5.28; OR for T-
score >70 = 1.97; 95% CI: 1.19, 3.27); no
associations between urinary fluoride level
and ADHD index or other behavioral
measures
Adjusted for age, sex, child’s BMI, urinary
creatinine, mother migrated, and father
migrated

59
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Exposure Assessment
Study Design
Study Measures and Assessment Timing Outcome and Methods Neurobehavioral Outcome Summaryb
(Location/Subjects) [n]
Summary Statistics
Mexico
Rocha-Amador et al. Cross-sectional Children’s urine Children (ages 6–11 Visuospatial organization Significant correlation between urinary
(2009) Durango/elementary school GM (SD): 5.6 years) and visual memory: Rey- fluoride and visuospatial organization
children (1.7) mg/L Osterrieth Complex Figure (r = −0.29) and visual memory scores
[80] Test, children’s version (r = −0.27); no significant correlation with
arsenic
Adjusted for age
Valdez Jiménez et al. Cohort (Prospective) Maternal urine Infants (ages 3–15 Mental development index Significant association between log10-mg/L
(2017) Durango City and Lagos de Range: 0.16–8.2 mg/L months) (MDI): Bayley Scales of maternal urinary fluoride and MDI score
Moreno/infants (all trimesters) Infant Development II during first trimester (adjusted β = −19.05;
[65] (BSDI-II) SE = 8.9) and second trimester (adjusted
Drinking water
Psychomotor developmental β = −19.34; SE = 7.46); no significant
Range: 0.5–12.5 mg/L index (PDI): Bayley Scales associations between maternal urinary
(all trimesters) of Infant Development II fluoride and PDI score; analyses of
(BSDI-II) outcomes using drinking water fluoride not
performed
Adjusted for age, gestational age,
marginality index, and type of drinking
water
Bashash et al. (2017)c Cohort (prospective) Maternal urine during Children (age 4 years) General cognitive index Significant association between maternal
Mexico City/Early Life pregnancy (GCI): McCarthy Scales of urinary fluoride and offspring GCI score
Exposures in Mexico to Mean (SD): 0.90 Children’s Abilities (MSCA) (per 0.5-mg/L increase adjusted β = −3.15;
Environmental Toxicants (0.35) mg/L 95% CI: −5.42, −0.87); associations with
(ELEMENT) participants [299] Children’s urine children’s urine not significant

GCI analysis [287] Adjusted for gestational age; weight at

Mean (SD): 0.82 birth; sex; parity (being the first child); age
(0.38) mg/L at outcome measurement; and maternal
characteristics, including smoking history
(ever smoked during the pregnancy vs.
nonsmoker), marital status (married vs. not
married), age at delivery, IQ, education, and
cohort

60
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Exposure Assessment
Study Design
Study Measures and Assessment Timing Outcome and Methods Neurobehavioral Outcome Summaryb
(Location/Subjects) [n]
Summary Statistics
Bashash et al. (2018)c Cohort (prospective) Maternal urine during Children (ages 6–12 ADHD: Conners’ Rating Significant associations between maternal
Mexico City/Early Life pregnancy years) Scales-Revised (CRS-R) urinary fluoride (per 0.5-mg/L increase) and
Exposures in Mexico to Mean 0.85 (95% CI: CRS-R scores, including Cognitive
Environmental Toxicants 0.81, 0.90) mg/L Problems + Inattention Index (adjusted
(ELEMENT) participants β = 2.54; 95% CI: 0.44, 4.63), DSM-IV
[210] Inattention Index (adjusted β = 2.84; 95%
CI: 0.84, 4.84), DSM-IV ADHD Total
Index (adjusted β = 2.38; 95% CI: 0.42,
4.34), and ADHD Index (adjusted β = 2.47;
95% CI: 0.43, 4.50)
Adjusted for gestational age; birth weight;
sex; parity; age at outcome measurement;
and maternal characteristics, including
smoking history (ever smoked vs.
nonsmoker), marital status (married vs. not
married), education, socioeconomic status,
and cohort
Canada
Barberio et al. Cross-sectional Children’s urine Children (ages 3–12 Learning disability, ADHD Significant increase in adjusted OR for
(2017b)d General population/Canadian Mean Cycle 2: 32.06 years) (Cycle 2 only): Parent or learning disability (adjusted OR = 1.02;
Health Measures Survey (95% CI: 29.65, child self-report 95% CI: 1.00, 1.03) per 1-µmol/L increase
(Cycles 2 and 3) 34.46) µmol/L in unadjusted urinary fluoride when Cycle 2
[2,221] Mean Cycle 3: 26.17 and 3 were combined; no significant
(95% CI: 22.57, associations found between urinary fluoride
29.76) µmol/L and ADHD (only evaluated in Cycle 2); no
significant associations found when using
creatinine- or specific gravity-adjusted
urinary fluoride
Adjusted for age and sex, household income
adequacy, and highest attained education in
the household

61
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Exposure Assessment
Study Design
Study Measures and Assessment Timing Outcome and Methods Neurobehavioral Outcome Summaryb
(Location/Subjects) [n]
Summary Statistics
Riddell et al. (2019)d Cross-sectional Drinking water Children (ages 6–17 Hyperactivity/inattention: Significantly increased risk of ADHD with
General population/Canadian Mean (SD): 0.23 years) Strengths and Difficulties fluoride in tap water (adjusted OR = 6.10
Health Measures Survey (0.24) mg/L [non- Questionnaire (SDQ) per 1-mg/L increase; 95% CI: 1.60, 22.8) or
(Cycles 2 and 3) fluoridated water: 0.04 ADHD: parent or self- community water fluoridation status (1.21;
[3,745] (0.06) mg/L; fluoridated reported physician diagnosis 95% CI: 1.03, 1.42) but not with urinary
water: 0.49 (0.22)] fluoride; similar results observed with
attention symptoms based on the SDQ
Community water scores
fluoridation status (yes
or no) Adjusted for age and sex, child’s BMI,
ethnicity, parental education, household
Children’s urine income, blood lead, and smoking in the
Mean (SD): 0.61 home
(0.39) mg/L [non-
fluoridated water: 0.46
(0.32) mg/L; fluoridated
water: 0.82 (0.54)]
ADHD = attention-deficit/hyperactivity disorder; BMI = body mass index; GCI = General Cognitive Index; GM = geometric mean; HOME = Home Observation Measurement of
the Environment; IQ = intelligence quotient; MSCA = McCarthy Scales of Children’s Abilities; SD = standard deviation; WASI = Wechsler Abbreviated Scale of Intelligence
(Spanish version); WISC-IV = Wechsler Intelligence Scale for Children-Revised (as reported by Choi et al. 2015); WRAML = Wide Range Assessment of Memory and Learning;
WRAVMA = Wide Range Assessment of Visual Motor Ability.
aIncludes low risk-of-bias studies.
bAssociations between other cognitive neurodevelopmental outcomes in children and fluoride levels were reported quantitatively, when possible. For studies with multiple analyses

and results, the table summarizes key findings and is not a comprehensive summary of all findings. Results also indicated when a study reported no association, provided as a
qualitative statement of no association.
cBashash et al. (2017) and Bashash et al. (2018) are based on the same study population.
dBarberio et al. (2017b) and Riddell et al. (2019) are based on the same study population.

62
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Summary of Results

Overall Findings
Although discussed together in this section, various health outcomes were assessed in the nine
low risk-of-bias studies of other neurodevelopmental outcomes, including neurobehavioral
scores in infants (two studies), cognitive tests in children other than IQ (three studies), and
ADHD or learning disabilities (four studies) in children. The results from eight of nine low risk-
of-bias studies (three prospective cohort studies and five cross-sectional studies from seven
different study populations) provide evidence of significant associations between higher fluoride
exposure and cognitive neurodevelopmental outcomes in children other than decrements in IQ
(see Figure A-9 through Figure A-11) (Barberio et al. 2017b; Bashash et al. 2018; Bashash et al.
2017; Li et al. 2004 [translated in Li et al. 2008a]; Riddell et al. 2019; Rocha-Amador et al.
2009; Valdez Jiménez et al. 2017; Wang et al. 2020a). Only one cross-sectional study did not
find a significant association between estimated fluoride exposure and a measure of cognitive
neurodevelopment (Choi et al. 2015).
Although there is heterogeneity in the outcomes assessed and a limited number of directly
comparable studies, the data provide additional evidence (beyond the consistent evidence of an
inverse association between estimated fluoride exposure and IQ) of an association between
higher fluoride exposure and cognitive or neurodevelopmental effects. The body of evidence
from the nine low risk-of-bias studies is described in further detail below, including the direction
of effect for statistically significant associations, and is grouped into outcome categories of
studies that are most comparable.

Results in Infants
Two studies evaluated neurobehavioral effects in infants either shortly after birth or at 3 to
15 months of age (Li et al. 2004 [translated in Li et al. 2008a]; Valdez Jiménez et al. 2017). Both
studies observed a statistically significant inverse association between estimated fluoride
exposure and neurobehavioral scores. In neonates (1–3 days old), the high fluoride group
(3.58 ± 1.47 mg/L fluoride based on spot maternal urine collected just prior to birth) had
significantly lower total neurobehavioral assessment scores (36.48 ± 1.09 versus 38.28 ± 1.10 in
controls; p-value <0.05) and total behavioral capacity scores (10.05 ± 0.94 versus 11.34 ± 0.56 in
controls; p-value <0.05) compared to the control group (1.74 ± 0.96 mg/L fluoride) as measured
by a standard neonatal behavioral neurological assessment (NBNA) method (Li et al. 2004
[translated in Li et al. 2008a]). In infants 3 to 15 months of age, the Mental Development Index
(MDI)—which measures functions including hand-eye coordination, manipulation,
understanding of object relations, imitation, and early language development—was significantly
inversely associated with maternal urinary fluoride in both the first and second trimesters
(adjusted βs per log10-mg/L increase = −19.05 with standard error of 8.9 for first trimester [p-
value = 0.04] and −19.34 with standard error of 7.46 for second trimester [p-value = 0.013])
(Valdez Jiménez et al. 2017). Note that this study did not find an association between maternal
fluoride during any trimester and the Psychomotor Developmental Index (PDI), which measures
gross motor development (adjusted βs = 6.28 and 5.33 for first and second trimesters,
respectively; no standard errors provided) (Valdez Jiménez et al. 2017).

63
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Results for Cognitive Tests Other Than IQ in Children

Three studies conducted tests on cognitive function in children that were not part of an IQ test
(Bashash et al. 2017; Choi et al. 2015; Rocha-Amador et al. 2009). None of the studies
conducted the same tests, but two of the three studies (Bashash et al. 2017; Rocha-Amador et al.
2009) observed associations (one statistically significant and one not statistically significant)
between fluoride exposure and lower test scores. The General Cognitive Index (GCI) of the
McCarthy Scales of Children’s Abilities (MSCA) in 4-year-old children was significantly
inversely associated with maternal creatinine-adjusted urinary fluoride levels during pregnancy
(collected during each trimester) (adjusted β per 0.5-mg/L increase = −3.15 [95% CI: −5.42,
−0.87; p-value = 0.01] in a model adjusting for main covariates including gestational age, weight
at birth, sex, maternal smoking, and indicators of socioeconomic status). The association
remained even after adjusting for maternal bone lead (adjusted β per 0.5-mg/L increase = −5.63
[95% CI: −8.53, −2.72; p-value <0.01]) (Bashash et al. 2017) (see Figure A-11). Choi et al.
(2015), however, evaluated cognitive function endpoints in addition to IQ and found no
significant associations between concurrent log-transformed water or urinary fluoride levels and
Wide Range Assessment of Visual Motor Ability (WRAVMA) scores, finger tapping test scores,
and grooved pegboard test scores, although there were some significant associations based on
degree of fluorosis (see Figure A-11). Another study using visuoconstructional and memory
scores from the Rey-Osterrieth Complex Figure Test in children 6–11 years old observed
significantly lower scores with increasing concurrent child single spot urinary fluoride even after
adjusting for age (partial correlation coefficients, per log-mg/L increase = −0.29 and −0.27 for
copy [p-value <0.001] and immediate recall [p-value <0.001], respectively [CIs not reported])
(Rocha-Amador et al. 2009). Although these children were also exposed to arsenic, the presence
of arsenic could not explain the changes because, in the area with natural contamination by
fluoride and arsenic (F–As), the test scores were not significantly associated with urinary arsenic
levels (partial correlation coefficients, per log-mg/L increase = −0.05 and 0.02 for copy and
immediate recall, respectively [CIs not reported]). The test scores were only marginally
increased from fluoride alone when both fluoride and arsenic were included simultaneously in
the model (partial correlation coefficients, per log-mg/L increase = −0.32 and −0.34 for copy and
immediate recall, respectively [CIs not reported]) (Rocha-Amador et al. 2009) (see Figure A-10).

Attention-related Disorders Including ADHD and Learning Disabilities in Children

Four studies evaluated attention-related disorders or learning disabilities (Barberio et al. 2017b;
Bashash et al. 2018; Riddell et al. 2019; Wang et al. 2020a). All four studies found a statistically
significant positive association between fluoride exposure and measures of ADHD or learning
disability; however, studies varied in the exposure metrics and outcomes measure. Bashash et al.
(2018) evaluated behaviors associated with ADHD in children ages 6–12 years using the
Conners Rating Scales-Revised (CRS-R) and observed significant associations between maternal
urinary fluoride (measured during each trimester) and ADHD-like symptoms, particularly those
related to inattention (an increase in 0.5 mg/L of maternal urinary fluoride was significantly
associated with a 2.84-point increase [95% CI: 0.84, 4.84; p-value = 0.0054] in the DSM-IV
Inattention Index and a 2.54-point increase [95% CI: 0.44, 4.63; p-value = 0.0178] in the
Cognitive Problems and Inattention Index). These two scales contributed to the global ADHD
Index and the DSM-IV ADHD Total Index, which were also significantly associated with higher
levels of prenatal fluoride exposure (an increase of 0.5 mg/L in maternal urinary fluoride was
associated with a 2.38-point increase [95% CI: 0.42, 4.34; p-value = 0.0176] in the DSM-IV

64
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

ADHD Total Index and a 2.47-point increase [95% CI: 0.43, 4.50; p-value = 0.0175] in the
ADHD Index) (see Figure A-11). Significant associations were not observed between maternal
urinary fluoride concentrations during pregnancy and child performance on measures of
hyperactivity, nor were there any significant results in children using Conners’ Continuous
Performance Test (CPT-II, 2nd Edition), a computerized test of sustained attention and inhibitory
control (Bashash et al. 2018). Wang et al. (2020a) also used Conners’ Parent Rating Scale
(Chinese version) to assess behavioral outcomes in children ages 7–13 years but found only a
significant association between spot urinary fluoride concentrations in children (model adjusted
for creatinine) and psychosomatic problems (adjusted OR for T-score >70 per 1-mg/L
increase = 1.97 [95% CI: 1.19, 3.27; p-value = 0.009] and adjusted β per 1-mg/L increase = 4.01
[95% CI: 2.74, 5.28; p-value <0.001]). No associations were found between spot urinary fluoride
and the ADHD index or other behavioral measures.
Barberio et al. (2017b) evaluated learning disabilities in children 3–12 years of age, including
ADHD, attention deficit disorder (ADD), and dyslexia, as part of the Canadian Health Measures
Survey and found that when compared to lower spot urinary fluoride levels, higher spot urinary
fluoride levels in children were associated with a small but significantly increased risk in self-
reported (children 12 years of age) or parent- or guardian-reported (children 3–11 years of age)
learning disabilities (adjusted OR per 1-µmol/L increase = 1.02; 95% CI: 1.00, 1.03; p-value
<0.05) (see Figure A-12). However, significant associations were not observed in analyses using
creatinine- or specific gravity-adjusted urinary fluoride (Barberio et al. 2017b). Barberio et al.
(2017b) also reported no associations between single spot urinary fluoride and ADHD in
children ages 3 to 12 years. Riddell et al. (2019) used the same Canadian Health Measured
Survey but evaluated children 6–17 years old. Riddell et al. (2019) found a significantly
increased risk for ADHD diagnosis with both tap water fluoride (adjusted OR per 1-mg/L
increase = 6.10; 95% CI: 1.60, 22.8; p-value <0.05) and community water fluoridation status
(adjusted OR per 1-mg/L increase = 1.21; 95% CI: 1.03, 1.42; p-value <0.05). A similar increase
in the hyperactivity-inattention symptoms score based on the Strengths and Difficulties
Questionnaire was observed with both tap water fluoride (adjusted β per 1-mg/L increase = 0.31;
95% CI: 0.04, 0.58; p-value <0.05) and community fluoridation status (adjusted β per 1-mg/L
increase = 0.11; 95% CI: 0.02, 0.20; p-value <0.05). As was observed with Barberio et al.
(2017b), Riddell et al. (2019) did not observe associations between specific gravity-adjusted spot
urinary fluoride concentrations and either ADHD diagnosis (adjusted OR per 1-mg/L
increase = 0.96; 95% CI: 0.63, 1.46) or hyperactivity-inattention symptoms (adjusted β per 1-
mg/L increase = 0.31; 95% CI: −0.04, 0.66).

Summary of Key Findings for Low Risk-of-bias Studies of Other Neurodevelopmental and
Cognitive Effects in Children
In summary, the high-quality studies (i.e., studies with low potential for bias), when compared to
lower estimated fluoride exposure, provide evidence of an association between higher estimated
fluoride exposure and neurodevelopmental and cognitive effects in children other than IQ;
however, the body of evidence is limited by heterogeneity in the outcomes evaluated and few
directly comparable studies. Across these outcomes, eight of nine studies reported a significant
association between estimated fluoride exposure and a measure of neurodevelopment or
cognition other than IQ, which provides support for the consistency in evidence based on
children’s IQ studies of an association between fluoride exposure and adverse effects on
cognitive neurodevelopment.

65
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

High Risk-of-bias Studies

High risk-of-bias studies (n = 6) also provide some evidence of associations between estimated
fluoride exposure and neurodevelopmental or cognitive effects in children other than effects on
IQ, but the results are inconsistent and address different outcomes (Jin et al. 2016; Li et al. 1994
[translated in Li et al. 2008b]; Malin and Till 2015; Morgan et al. 1998; Mustafa et al. 2018;
Shannon et al. 1986).

Risk of Bias for Neurodevelopmental or Cognitive Effect Studies in Children

The confidence in the human body of evidence was based on studies with the lowest potential for
bias (i.e., studies that rated probably low or definitely low risk of bias for at least two of the three
key risk-of-bias questions and did not have any other risk-of-bias concerns that would indicate
serious issues with the studies). Each of the nine low risk-of-bias studies on other
neurodevelopmental effects in children had little or no risk-of-bias concerns. Four of the nine
studies were rated definitely low or probably low risk of bias for all risk-of-bias questions, and
the remaining five studies were rated probably high risk of bias for a single question that was
judged to have minimal impact on overall potential bias. None of the nine studies had a rating of
definitely high risk of bias for any question. Although the nine low risk-of-bias studies had
minimal or no concerns, the six studies with high overall potential for bias had several risk-of-
bias concerns related to one or more of the three key risk-of-bias questions (confounding,
exposure characterization, and outcome assessment). The key risk-of-bias questions are
discussed below. Risk-of-bias ratings for other neurodevelopmental effect studies in children are
available in Figure D-9 through Figure D-12 and Appendix E for the low and high risk-of-bias
studies.

Confounding for Other Neurodevelopmental Studies in Children

Low Risk-of-bias Studies

As discussed above, there are nine studies considered to have low risk of bias when assessed
across all risk-of-bias domains. Seven of nine low risk-of-bias studies were considered to have
low potential for bias due to confounding because the authors addressed the three key covariates
for all studies (age, sex, and socioeconomic status) and also addressed arsenic as a potential co-
exposure of concern through study design or analysis. Other important covariates, including
health factors, smoking, and parental characteristics, were also addressed in many of the low
risk-of-bias studies. One of the studies (Bashash et al. 2018) examined several covariates in
sensitivity analyses involving subsets of participants, including HOME scores, child
contemporaneous fluoride exposure measured by child urinary fluoride adjusted for specific
gravity, and maternal lead and mercury exposures. The authors reported that none of the
sensitivity analyses indicated appreciable changes in the fluoride-related association with
behaviors related to ADHD, nor was there evidence of effect modification between maternal
urinary fluoride and sex.
Among the nine low risk-of-bias studies, two studies were identified that have potential for bias
due to confounding (Rocha-Amador et al. 2009; Valdez Jiménez et al. 2017). Although both of
these studies adjusted for several covariates through analysis or study design, Valdez Jiménez et
al. (2017) did not address a potential concern for co-exposure to arsenic, and Rocha-Amador et
al. (2009) does not appear to adjust for SES or address why it would not be a concern in the
study population (see Appendix E for further details). Although these two studies have some

66
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

potential for bias due to confounding, they are considered to have low potential for bias overall
because they have low potential for bias for the other two key risk-of-bias questions (exposure
characterization and outcome assessment), and no other major concerns for bias were identified.
Consistent with the IQ studies, bias due to confounding is not likely a concern for the low risk-
of-bias studies.

High Risk-of-bias Studies

The six high risk-of-bias studies in the human body of evidence did not adequately address
important covariates through study design or analysis. The same concerns due to potential
confounding noted previously for the high risk-of-bias children’s IQ studies were also present in
the other neurodevelopmental high risk-of-bias studies, including not addressing the three key
covariates for all studies (age, sex, SES) and/or not addressing potential co-exposures (e.g.,
arsenic) in areas of potential concern.

Exposure Characterization in Other Neurodevelopmental Studies in Children

Low Risk-of-bias Studies

There were no risk-of-bias concerns regarding exposure assessment in the low risk-of-bias
studies. All of the low risk-of-bias studies had individual exposure data based on urine or water
measures with appropriate analyses, and most of the urinary fluoride studies accounted for
urinary dilution when appropriate. Although there are concerns related to the timing of urine
samples (see the Risk-of-bias Considerations for Human Studies section for details), the studies
that used maternal urine measured urinary fluoride multiple times throughout pregnancy
(Bashash et al. 2018; Bashash et al. 2017; Valdez Jiménez et al. 2017). Another study
demonstrated correlations between urinary fluoride and fluoride in the drinking water, fluorosis,
or estimated dose based on water (Choi et al. 2015). Bashash et al. (2017) excluded exposure
measurement outliers but found that doing so did not change the results in a meaningful way.

High Risk-of-bias Studies

A frequent critical limitation among the high risk-of-bias studies was lack of information
regarding exposure or poor exposure characterization. In the high risk-of-bias studies that
assessed the association between fluoride exposure and other neurodevelopmental and cognitive
effects in children, fluoride exposure assessment was based on dental fluorosis, municipality-
level water fluoridation prevalence data, number of years living in an area with fluoridated water,
or group-level water samples. See the Exposure Characterization in IQ Studies section for further
discussion on the limitations of exposure assessments in high risk-of-bias studies.

Outcome Assessment in Other Neurodevelopmental Studies in Children

Low Risk-of-bias Studies

The low risk-of-bias studies have few concerns regarding outcome assessment. Seven of the nine
studies [i.e., all low risk-of-bias studies except Barberio et al. (2017b) and Riddell et al. (2019)]
used appropriate methods for measuring other neurodevelopmental effects in the study
population, and blinding of outcome assessors was either reported or not a concern in eight of the
nine studies [i.e., all with the exception of Wang et al. (2020a)].
Among the nine low risk-of-bias studies, three were identified that have a potential for bias due
to outcome assessment. One of the studies (Wang et al. 2020a) had potential concern for bias due

67
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

to lack of information regarding the blinding of outcome assessors. Two of the studies (Barberio
et al. 2017b; Riddell et al. 2019) were based on the same study population in Canada, where
different questions were asked in Cycles 2 (2009–2011) and 3 (2012–2013) of the Canadian
Health Measures Survey (CHMS) to ascertain learning disabilities including ADHD. In Cycle 2,
subjects were asked whether they had a learning disability diagnosed by a health professional
and, if yes, were asked what kind. In Cycle 3, CHMS did not ask what kind of learning disability
was diagnosed nor was a reason for the question omission provided. Because no reason was
provided for the removal of the question, and because a question on learning disability without
the specific diagnosis may be more prone to bias, this change in questioning from Cycles 2 to 3
is a potential concern. Blinding was not considered an issue in these two studies, but the methods
for obtaining the information are considered to be less than ideal for measuring learning
disabilities including ADHD. Although the questionnaire asked about a doctor’s diagnosis of a
learning disability, there was no confirmation with medical records. Moreover, these
questionnaires were not validated like Conners’ Rating Scales, which would have been a better
method for assessing ADHD. Although the outcome assessment methods are less than ideal,
there was no direct evidence that they were conducted incorrectly or that the methods would
have biased the results in any specific direction. Because this was the only concern in these
studies, they were considered to have low risk of bias overall.

High Risk-of-bias Studies

Among the studies on other neurodevelopmental effects with high potential for bias, there were
several reasons for studies to be considered probably or definitely high risk of bias for outcome
assessment. One study (Shannon et al. 1986) was considered to have probably high risk of bias
based on lack of information regarding blinding of outcome assessors. One study was considered
definitely high risk of bias because outcome was assessed based on a parent-completed
questionnaire, and the study authors noted that the parents were informed of the study’s intent
and were requested to provide information on fluoride history. Other studies used outcome
assessment methods that were not validated or utilized group-level measurements (i.e., school
performance, working memory scores).

Confidence Assessment of Findings on Other Neurodevelopmental Effects in Children

The high-quality studies (i.e., studies with low potential for bias) provide some evidence of an
association between estimated fluoride exposure and other cognitive neurodevelopmental effects,
including lower neurobehavioral scores in infants, cognitive effects other than IQ in children,
and increased attention-related disorders including ADHD in children. However, due to
limitations in the data set, including the heterogeneity in the outcomes assessed, a limited
number of directly comparable studies, and differences in outcome assessment methods even
when studies evaluated similar outcomes, there is low confidence based on this body of evidence
that estimated fluoride exposure is associated with other cognitive neurodevelopmental effects in
children. Due to these limitations, the confidence assessment is not described in the same manner
as the IQ in Children section or as outlined in Figure 1. Although there are limitations in the
body of evidence, the low risk-of-bias studies demonstrate, that when compared to lower
estimated fluoride exposure, there is a relationship between higher estimated fluoride exposure
and neurodevelopmental effects, even in very young children. This supports the consistency in
evidence shown in children’s IQ studies of an association between estimated fluoride exposure
and adverse effects on cognitive neurodevelopment.

68
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Cognitive Effects in Adults

Low Risk-of-bias Studies

Overview of Studies
Two low risk-of-bias cross-sectional studies evaluated the association between estimated fluoride
exposure and cognitive effect in adults (Jacqmin et al. 1994; Li et al. 2016). These two studies
used the same test for cognitive function (i.e., Mini-Mental State or MMS Examination) and
used drinking water fluoride levels to assess fluoride exposure. Li et al. (2016) also measured
urinary fluoride. Both studies were cross-sectional in design. One was conducted in France
(Jacqmin et al. 1994) and the other in China (Li et al. 2016). Both studies were conducted in
older populations (i.e., over 60 or 65 years of age).
Table 8 provides a summary of study characteristics and key findings related to estimated
fluoride exposure and cognitive effects in adults for the two low risk-of-bias studies. The
purpose of the table is to summarize key findings (independent of whether an association was
found) from each study and is not meant to be a comprehensive summary of all results. For each
study, results are summarized for each exposure measure assessed. Results from multiple
analyses using the same exposure assessment measure may not all be presented unless
conflicting results were reported.

69
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Table 8. Studies on Cognitive Function in Adultsa

Exposure Assessment
Study Design Assessment Neurobehavioral Outcome
Study Measures and Outcome and Methods
(Location/Subjects) [n] Timing Summaryb
Summary Statistics
Jacqmin et al. Cross-sectional Drinking water Adults (ages ≥65 Cognitive function: MMS No significant increase in the
(1994) France (Gironde and Range: 0.03–2.03 mg years) Examination prevalence of cognitive
Dordogne)/elderly adults impairment with increasing
[3,490] fluoride quartiles
No statistical adjustment for
covariates for prevalence rates

Li et al. (2016) Cross-sectional Drinking water daily Adults (ages ≥60 Cognitive function: MMS Subjects with cognitive
China (Inner fluoride intake years) Examination impairment had a significantly
Mongolia)/adults Mean (SD): 2.23 higher skeletal fluorosis score
[511] (2.23) (normal group), and urinary fluoride
3.62 (6.71) (cognitive concentrations; odds of
impairment group) mg increasing severity of cognitive
impairment increased with
Urine urinary fluoride concentrations
Mean (SD): 1.46 but were not statistically
(1.04) (normal group), significant; no significant
2.47 (2.88) (cognitive association with total daily water
impairment group) fluoride intake
mg/L Adjusted for sex, age, education,
Fluorosis score marital status (married vs. not
Mean (SD): 0.74 married), alcohol consumption
(0.98) (normal group), (non-drinkers, light drinkers,
1.29 (1.01) (cognitive moderate to heavy drinkers),
impairment group) smoking history (never smoker,
ex-smoker, light smoker, heavy
smoker), and serum
homocysteine levels
GM = geometric mean; MMS = Mini-Mental State.
aIncludes low risk-of-bias studies.
bAssociations between cognitive effects in adults and fluoride levels were reported quantitatively, when possible. For studies with multiple analyses and results, the table

summarizes key findings and is not a comprehensive summary of all findings. Results also indicate when a study reported no association, provided as a qualitative statement of no
association.

70
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Summary of Results
Results from two low risk-of-bias studies in adults did not provide enough evidence to evaluate
consistency when assessing evidence for a potential association between fluoride exposure and
cognitive impairment (based on the MMS Examination) (Jacqmin et al. 1994; Li et al. 2016).
Jacqmin et al. (1994) did not find an association between drinking water fluoride and cognitive
impairment in populations in France (n = 3,490) and found prevalence rates of cognitive
impairment to be the same regardless of fluoride exposure (see Figure A-13). In contrast, Li et al.
(2016) did find significantly higher urinary fluoride levels and skeletal fluorosis scores in the
cognitively impaired group compared with the control group in an analysis of 38 cognitively
impaired cases and 38 controls matched for several covariates, including age, sex, education,
alcohol consumption, and smoking (p-value <0.05). However, the authors found no significant
association between cognitive impairment and total daily water fluoride intake (adjusted ORs per
1-mg/day increase = 0.94 [95% CI: 0.85, 1.04] and 0.86 [95% CI: 0.69, 1.06] in the moderate
and severe cognitive impairment groups, respectively) or urinary fluoride levels (adjusted ORs
per 1-mgL increase = 1.12 [95% CI: 0.89, 1.42] and 1.25 [95% CI: 0.87, 1.81] in the moderate
and severe cognitive impairment groups, respectively) in subjects from fluorosis-endemic areas
of China (n = 511).

High Risk-of-bias Studies

The results from five out of eight high risk-of-bias studies provide evidence of cognitive
impairment in adults associated with higher levels of exposure to fluoride; however, there was
heterogeneity in the outcomes assessed, a limited number of directly comparable studies, and
some variability in results (e.g., variation in IQ results across studies). Due to the limited number
of low risk-of-bias studies identified that assess cognitive impairment in adults, the results from
the high risk-of-bias studies are summarized in greater detail below than had been done in this
document for bodies of evidence for IQ in children and other neurodevelopmental and cognitive
effects in children.
In aluminum factory workers (exposed to gaseous and particular fluoride emissions during the
production of aluminum metal), significant decreases in IQ (Duan et al. 1995), diminished
performance on several neurobehavioral core battery tests (NCTBs) (Guo et al. 2001 [translated
in Guo et al. 2008b]), and impaired psychomotor performance and memory were observed
(Yazdi et al. 2011). One study conducted on adult subjects with fluorosis (dental and skeletal)
from a fluorosis-endemic area compared with healthy subjects from a non-endemic area
observed significant differences for some cognitive function tests (i.e., tests of speech fluency,
recognition, and working memory) but not others and generally did not observe a significant
change in IQ except in the operation scores (Shao 2003). One prospective cohort study evaluated
exposure to fluoride in children at 5 years of age, based on whether the children resided in areas
with community water fluoridation or used fluoride toothpaste or fluoride tablets, and found no
clear differences in IQ scores of the subjects at 38 years of age (Broadbent et al. 2015). One
additional study suggested that populations living in areas with higher drinking water fluoride
had lower levels of dementia (Still and Kelley 1980); however, the study was not focused on
associations with fluoride but on whether fluoride was able to reduce the risk associated with
aluminum by competing with aluminum and reducing its bioavailability. Therefore, the study
was considered inadequate to evaluate the association between fluoride and dementia (Still and
Kelley 1980). A more recent study in Scotland evaluated dementia rates associated with
aluminum and fluoride drinking water concentrations and observed a significant increased risk of

71
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

dementia per standard deviation increase in fluoride (p-value <0.001) with the risk of dementia
more than double in the highest quartile of fluoride exposure (56.3 µg/L) compared to the lowest
quartile (<44.4 µg/L). The authors also found a significantly increased risk of dementia
associated with increased aluminum levels at all quartiles compared with the reference group (p-
values <0.05) but found no statistical interaction between aluminum and fluoride levels in
relation to dementia (Russ et al. 2019). Conversely, a study in China did not find a significant
association between fluoride concentrations in the drinking water and risk for dementia (Liang et
al. 2003). In addition to studies that reported on cognitive impairment and exposure to fluoride,
two high risk-of-bias studies were identified that reported impaired motor and sensory function
(Rotton et al. 1982) and a higher prevalence of self-reported headaches, insomnia, and lethargy
(Sharma et al. 2009) associated with fluoride exposure.

Risk of Bias for Cognitive Effect Studies in Adults

Due to the small number of studies with a low potential for bias (see Figure D-13 and
Figure D-14), the key risk-of-bias domains (confounding, exposure characterization, outcome
assessment) are not discussed separately in respective subsections, as was done for the IQ in
Children and Other Neurodevelopmental and Cognitive Effects in Children bodies of evidence.
The high risk-of-bias studies had concerns across several domains (see Figure D-15 and
Figure D-16), but there were still relatively few studies. Therefore, the discussion for high risk-
of-bias studies is also not separated into subsections by key domain.

Low Risk-of-bias Studies

Both low risk-of-bias studies on cognitive effects in adults had little or no risk-of-bias concerns.
One study was rated definitely low or probably low risk of bias for all risk-of-bias questions (Li
et al. 2016), and the other study was rated probably high risk of bias for a single question that
was judged to have minimal impact on overall potential bias (Jacqmin et al. 1994). Jacqmin et al.
(1994) had potential concern for bias due to confounding because smoking was not addressed,
which has the potential to impact risk for Alzheimer’s disease and rates could vary by parish (the
target population consisted of men and women from 75 civil parishes in southwestern France).

High Risk-of-bias Studies

There were several issues in the eight studies in adults considered to have high potential for bias.
Four of the eight studies had potential concern for bias due to lack of information on the
comparison groups, or the comparison groups were considered inappropriate. All eight studies
had potential concern for bias regarding covariates not being addressed, including possible co-
exposures in occupational studies (e.g., aluminum) and smoking. Five of the eight studies had
potential concern for bias due to lack of information regarding exposure characterization or poor
exposure characterization with the most utilized exposure assessment measure in these studies
being a comparison between exposed and unexposed areas. In one case (Broadbent et al. 2015),
multiple sources of fluoride exposure were assessed separately without properly controlling for
the other sources of exposure, which could bias the results (see Exposure Characterization in IQ
Studies for further details). Five studies also had potential for bias based on limitations in the
outcome assessment, which was mainly due to lack of blinding of outcome assessors, lack of
validation of the methods, or lack of sufficient details on how the outcomes were assessed.

72
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Confidence Assessment of Findings on Cognitive Effects in Adults

The body of evidence available to examine the association between exposure to fluoride and
cognitive effects in adults is limited to two low risk-of-bias cross-sectional studies. Due to the
limited number of studies and a lack of evidence of an effect, there is low confidence based on
this body of evidence that fluoride exposure is associated with cognitive effects in adults.

Mechanistic Data in Humans

Eight low risk-of-bias studies that evaluated fluoride exposure and mechanistic data in humans
were considered potentially relevant to neurological effects. Effects on the thyroid were
specifically evaluated because the NRC 2006 report identified this as a possible effect of fluoride
(NRC 2006), and changes in thyroid hormones have been identified as a mechanism for
neurodevelopmental effects (Haschek and Rousseaux 1991). These included effects on thyroid
hormones in children (Kheradpisheh et al. 2018a; Kheradpisheh et al. 2018b; Malin et al. 2018),
adults (Kheradpisheh et al. 2018a; Kheradpisheh et al. 2018b; Malin et al. 2018), or children and
adults combined (Barberio et al. 2017a). In addition, some studies evaluated self-reported thyroid
conditions in children and adults combined (Barberio et al. 2017a) and thyroid diseases in adults
(Kheradpisheh et al. 2018b; Peckham et al. 2015) (see Figure D-17 and Figure D-18). Although
the low risk-of-bias studies provide some evidence of mechanistic effects (primarily changes in
thyroid stimulating hormone [TSH] levels in children), the studies were too heterogeneous or
limited in number to make any determination on mechanism (see Figure 7).
Among the seven low risk-of-bias studies that reported on changes in thyroid hormones, three
studies were conducted in children (Kumar et al. 2018; Singh et al. 2014; Zhang et al. 2015b)
and reported increases in TSH levels. Zhang et al. (2015b) reported significant increases in TSH
in children from a fluorosis-endemic area (median fluoride drinking water
concentration = 1.40 mg/L; interquartile range = 1.23–1.57 mg/L) compared with a non-
fluorosis-endemic area (median fluoride drinking water concentration = 0.63 mg/L; interquartile
range = 0.58–0.68 mg/L), whereas 3,5,3’-triiodothyronine (T3) or thyroxine (T4) were not
significantly different between the two groups. Similarly, Singh et al. (2014) observed
significantly higher TSH levels in children without dental fluorosis who lived in a fluorosis-
endemic area (fluoride drinking water concentrations of 1.6–5.5 mg/L) compared with children
without dental fluorosis who lived in a non-fluorosis-endemic area (fluoride drinking water
concentrations of 0.98–1.00 mg/L). When all children (with and without dental fluorosis) in the
endemic area were compared with children from the non-endemic area, the TSH levels were
higher in children from the fluorosis-endemic area, although results did not reach statistical
significance (p = 0.057). Significant differences in T4 or T3 were not observed between groups
(Singh et al. 2014). Kumar et al. (2018) also observed a significant increase in TSH levels in
children from a fluorosis-endemic area (1.5–5.8 mg/L fluoride) compared with a control area
(0.94–1.08 mg/L fluoride). There were also decreases in T3 and T4, but results were not
statistically significant.
Barberio et al. (2017a) evaluated associations between fluoride and TSH levels in children and
adults combined and found no relationship between fluoride exposure (measures in urine and tap
water) and TSH levels. In the one study that evaluated thyroid hormone levels in adults but not
children, Kheradpisheh et al. (2018b) found a significant increase in TSH associated with higher
fluoride concentrations in drinking water in both adults with and without thyroid diseases such as
hypothyroidism, hyperthyroidism, thyroid nodules, or thyroid cancer. Significant increases in T3

73
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

were associated with higher fluoride in drinking water in adults without thyroid diseases but
increases in T3 were not significant in adults with thyroid diseases. A significant association
between T4 and higher fluoride in drinking water was not observed in adults with or without
thyroid diseases (Kheradpisheh et al. 2018b).
Other than changes in hormone levels, there is limited evidence of fluoride-related mechanistic
effects in the three low risk-of-bias studies that evaluated thyroid-related effects. Barberio et al.
(2017a) found no relationship between fluoride exposure and self-reported thyroid conditions in
children and adults (children were older than 12). Kheradpisheh et al. (2018b) also found no
association between fluoride exposure and hypothyroidism in an adult population in Iran. One
study found a significantly higher prevalence of hypothyroidism in areas with higher fluoride
concentrations in drinking water (>0.7 mg/L) compared with areas with lower fluoride drinking
water concentrations (≤0.7 mg/L) (Peckham et al. 2015).
Sixteen high risk-of-bias studies were available that evaluated mechanistic data in humans
associated with fluoride exposure, including effects on thyroid hormones in children (n = 9
studies), thyroid hormones in adults (Michael et al. 1996; Yasmin et al. 2013), catecholamines in
adults (Michael et al. 1996) or in subjects of unknown ages (Chinoy and Narayana 1992),
acetylcholinesterase (AChE) or serotonin levels in children (Lu et al. 2019; Pratap et al. 2013),
brain histopathology or biochemistry in aborted fetuses (Du et al. 1992 [translated in Du et al.
2008]; Yu et al. 1996 [translated in Yu et al. 2008]), and mitochondrial fission/fusion molecules
in children (Zhao et al. 2019). Similar to the low risk-of-bias studies, the high risk-of-bias studies
provide some evidence of mechanistic effects (primarily changes in TSH levels in children);
however, the data are insufficient to identify a clear mechanism by which fluoride causes
neurodevelopmental or cognitive effects in humans.
Among high risk-of-bias studies (see Figure D-19 and Figure D-20), varying results were
reported in 11 studies that evaluated associations between fluoride exposure and thyroid
hormones, and a few of these studies (Lin et al. 1991; Wang et al. 2001; Yang et al. 1994
[translated in Yang et al. 2008]) were complicated by high or low iodine in the high fluoride
area. When considering fluoride effects on each of the hormones individually, similar to results
from low risk-of-bias studies, the most consistent evidence of fluoride-associated effects on a
thyroid hormone was reported as changes in TSH levels in children, although there was some
variation in the direction of association. Six of the nine high risk-of-bias studies that evaluated
changes in TSH levels in children reported increases in TSH levels with higher fluoride (Lin et
al. 1991; Susheela et al. 2005; Wang et al. 2001; Yang et al. 1994 [translated in Yang et al.
2008]; Yao et al. 1996; Yasmin et al. 2013). Two of the nine high risk-of-bias studies reported
decreases in TSH levels in children with higher fluoride (Khandare et al. 2017; Khandare et al.
2018). One of the nine studies found no significant alterations in TSH levels in children from
fluorosis-endemic areas (Hosur et al. 2012) (see Figure 8).
When considering associations between fluoride exposure and TSH, T3, and T4 levels together,
studies that evaluated changes in all three thyroid hormones reported varying combinations of
increases, decreases, or no changes in levels across the three hormones, although among the
eight low and high risk-of-bias studies that evaluated associations between fluoride exposure and
TSH, T3, and T4 levels and reported increases in TSH levels in children, seven of the eight
studies found no alterations in T3 levels (one study found an increase in T3), and six of the eight
studies found no alterations in T4 levels (two studies found an increase in T4). Studies also

74
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

displayed variation by age in the associations between fluoride and TSH, T3, and T4. Due to the
dynamic relationship between the thyroid gland, the pituitary gland, and the production and
clearance of TSH, T3, and T4, the variations in results are not unexpected and do not eliminate
the possibility of a mechanistic link between thyroid effects and neurodevelopmental or
cognitive effects; however, the data do not support a clear indication that thyroid effects are a
mechanism by which fluoride causes these effects in humans.

Figure 7. Number of Low Risk-of-bias Studies that Evaluated Thyroid Hormones in Children and
Adults by Endpoint and Direction of Association

Interactive figure and additional study details are available at

https://public.tableau.com/app/profile/ntp.visuals/viz/FluorideTableauDashboards/ReadMe. This figure displays study counts for
low risk-of-bias studies in both children and adults, as these counts are most relevant to the summary of fluoride-related
mechanistic effects in low risk-of-bias studies. Counts for high risk-of bias studies and studies by age (i.e., children, adults, or
children/adults combined) can also be accessed in the interactive figure. Study counts are tabulated by significance (unless study
footnotes in the interactive figure indicate that statistical significance was not tested)—statistically significant increase (↑),
statistically significant decrease (↓), or not significant (NS). For example, the “↑” column displays numbers of unique studies
with significantly increased results.

Figure 8. Number of High Risk-of-bias Studies that Evaluated Thyroid Hormones in Children by
Endpoint and Direction of Association

Interactive figure and additional study details are available at

https://public.tableau.com/app/profile/ntp.visuals/viz/FluorideTableauDashboards/ReadMe. This figure displays study counts for
high risk-of-bias studies in children, as these counts are most relevant to the summary of associations between fluoride and
thyroid hormones in high risk-of-bias studies. Counts for low risk-of bias studies, studies in adults, or all studies combined can
also be accessed in the interactive figure. Study counts are tabulated by significance (unless study footnotes in the interactive
figure indicate that statistical significance was not tested)—statistically significant increase (↑), statistically significant decrease
(↓), or not significant (NS). For example, the “↑” column displays numbers of unique studies with significantly increased results.

In addition to evaluating thyroid hormone levels, a few high risk-of-bias studies evaluated other
mechanistic data associated with fluoride exposure; however, the data are insufficient to identify
a clear mechanism by which fluoride might cause neurodevelopmental or cognitive effects in
humans. Serum epinephrine and norepinephrine were significantly increased in a fluoride-

75
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

endemic region (it was not reported whether subjects were children or adults) compared with a
non-endemic region (Chinoy and Narayana 1992). A separate study reported that serum
epinephrine and norepinephrine (referred to as adrenaline and noradrenaline in the study) were
significantly increased in adults in a fluoride-endemic area (fluoride in the drinking water ranged
from 1.0–6.53 ppm) compared with a control area (fluoride in the drinking water ranged from
0.56–0.72 ppm) (Michael et al. 1996). Serum AChE was significantly reduced in children from a
high fluoride region compared with a lower fluoride region (Pratap et al. 2013). Serum serotonin
was significantly increased in children from Turkey who were drinking water containing
2.5 mg/L of fluoride compared with children drinking bottled water or water containing
<0.5 mg/L of fluoride (Lu et al. 2019). Aborted fetuses from high fluoride areas in China were
found to have histological changes in the brain and significant changes in neurotransmitter levels
compared with a control area (Du et al. 1992 [translated in Du et al. 2008]; Yu et al. 1996
[translated in Yu et al. 2008]).
There are also two more recent low risk-of-bias studies that evaluated genetic variants in
dopamine-related genes; however, a determination on mechanism cannot be made at this time
due to the limited number of studies. For children (10–12 years old) with a Val158Met variant in
the COMT gene (i.e., catechol-O-methyltransferase), which results in slower degradation and
greater availability of dopamine within the brain, a stronger association between increasing
urinary fluoride levels and decreasing IQ was reported (Zhang et al. 2015b). For children (7–
12 years old) with a dopamine receptor-2 (DRD2) Taq 1A variant (which is involved in reduced
D2 receptor density and availability) and the TT (variant) genotype, a significant inverse
association between log urinary fluoride and IQ was observed; however, this significant
relationship was not observed in children with the CC (wild-type) or CT (hybrid) genotypes (Cui
et al. 2018).

Animal Learning and Memory Data

NTP provided a review of the experimental animal evidence in the earlier draft monographs
(NTP 2020) and agrees with the NASEM committee’s comments (NASEM 2020; 2021) that the
experimental animal database is of poor quality, with many studies suffering from major
reporting deficiencies (Sup01_Monograph_NASEM_Feb_2021.pdf). NTP acknowledges that
further efforts to disentangle the potential for motor activity deficits to influence tests of learning
and memory in the fluoride literature are warranted. Overall, these general issues and
deficiencies with the experimental animal database led to NTP’s conclusion that the animal
studies are currently inadequate to inform the question of an association between fluoride
exposure and neurodevelopmental and cognitive effects in humans. Therefore, this systematic
review does not include an experimental animal section, although Appendix D contains a risk-of-
bias assessment of the newer animal studies reviewed up to 2019.

Mechanistic Data in Animals

There are a wide variety of studies in animals that evaluate mechanistic effects potentially related
to neurological changes following oral fluoride exposure (see Appendix F); however, the
mechanisms underlying fluoride-associated cognitive neurodevelopmental effects are not well
characterized, and review of the data did not identify a mode of action for fluoride effects on IQ
in children. Categories of mechanistic endpoints with the largest amount of available data
include changes in biochemical components of the brain or neurons, neurotransmitters, oxidative

76
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

stress, histopathology, and thyroid function. Limiting the data to studies with at least one
exposure at or below 20 ppm fluoride drinking water equivalents (gavage and dietary exposures
were back calculated into equivalent drinking water concentrations for comparison) still
provided a sufficient number of studies for evaluation of these mechanistic endpoints. This
evaluation is provided in Appendix F. Neurotransmitter and biochemical changes in the brain
and neurons were considered the mechanistic areas with the greatest potential to demonstrate
effects of fluoride on the brain of animals in the lower dose range and provide evidence of
changes in the brain that may relate to lower IQ in children (see Appendix F). Histological data
can be useful in determining whether effects are occurring in the brain at lower fluoride
concentrations; however, author descriptions of these effects may be limited, thereby making it
difficult to directly link histological changes in the brain to learning and memory effects.
Oxidative stress is considered a general mechanistic endpoint that cannot be specifically linked
to neurodevelopmental or cognitive effects in humans; however, like histopathology, it may help
in identifying changes in the brain occurring at lower concentrations of fluoride.

In Vitro Data on Neurodevelopmental or Cognitive Effects

Although in vitro studies were identified as part of the systematic review process, NTP
determined that the information on neurological effects from these studies is too general, and
results cannot necessarily be attributed to effects on learning and memory or other cognitive
functions at this time. The in vitro data may help support specific mechanisms identified from in
vivo mechanistic data; however, as described above, no specific mechanism has been determined
for fluoride effects on learning and memory or other neurodevelopmental or cognitive outcomes.

77
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Discussion
This systematic review evaluated the available animal and human literature concerning the
association between fluoride exposure and cognitive neurodevelopment. The available data on
potential mechanisms to evaluate biological plausibility were also assessed. The health benefits
of fluoride with respect to oral health are acknowledged but are not the focus of this review.
This review extended NTP’s previous evaluation of the experimental animal data (NTP 2016).
Although the animal data provide some evidence of effects of fluoride on neurodevelopment,
they give little insight into the question of whether fluoride influences IQ. This is due to
deficiencies identified in the animal body of evidence. Mechanistic studies in humans provide
some evidence of adverse neurological effects of fluoride. However, these studies were too
heterogenous and limited in number to make any determination on biological plausibility.
Mechanistic studies in animals also provide some evidence of consistency in mechanistic effects
(Appendix F); however, the mechanisms underlying fluoride-associated cognitive
neurodevelopmental effects are not well characterized, and review of the data did not identify a
mode of action for fluoride effects on IQ in children.
The literature on adults is also limited; therefore, it was determined that there is low confidence
in the body of evidence from studies that evaluate fluoride exposure and adult cognition.
Compared to the literature in adults, there is a much more extensive literature in children.
The literature in children was separated into studies assessing IQ and studies assessing other
cognitive or neurodevelopmental outcomes. There is low confidence in the body of evidence
from studies that evaluate fluoride exposure and other cognitive or neurodevelopmental
outcomes in children. The confidence in this body of evidence is low because the number of
studies is limited, and there is too much heterogeneity in the outcomes measured, ages assessed,
and methods used to directly compare studies of any one outcome. This body of evidence is
made up of nine high-quality studies (three prospective cohort and six cross-sectional studies
from seven different study populations) and six low-quality studies. Eight of the nine high-
quality studies observed significant associations between fluoride and other cognitive or
neurodevelopmental outcomes in children including ADHD, visuospatial organization and
memory, NBNA, MDI, GCI, and MSCA. The data also suggest that neurodevelopmental effects
occur in very young children. Additional studies on outcomes such as ADHD and other
attention-related disorders, where there is some evidence of an association with estimated
fluoride exposure (i.e., all four studies evaluating attention-related disorders or learning
disabilities found a statistically significant positive association between fluoride exposure
assessment measure and measures of ADHD or learning disability), would be necessary to
critically assess the data.
Most of the epidemiological studies (n = 72) assessed the association between estimated fluoride
exposure and IQ in children. Although all studies, both high- and low-quality, were considered,
this evaluation focuses on the high-quality, low risk-of-bias studies in children for two reasons.
First, there are fewer limitations and greater confidence in the results of the high-quality studies.
Second, there is a relatively large number of high-quality studies (n = 19), such that the body of
evidence from these studies could be used to evaluate confidence in the association between
fluoride exposure and changes in children’s IQ.

78
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

This review finds, with moderate confidence, that higher estimated fluoride exposures (e.g., as in
approximations of exposure such as drinking water fluoride concentrations that exceed the World
Health Organization Guidelines for Drinking-water Quality of 1.5 mg/L of fluoride) are
consistently associated with lower IQ in children. The inverse association between estimated
fluoride exposure and IQ in children was consistent across different study populations, study
locations, study quality/risk-of-bias determinations, study designs, exposure assessment
measures, and types of exposure data (group-level and individual-level). There were 19 low risk-
of-bias studies that were conducted in 15 study populations, across 5 countries, and evaluating
more than 7,000 children. Of these 19 studies, 18 reported an inverse association between
estimated fluoride exposure [e.g., as in approximations of fluoride exposure such as drinking
water fluoride concentrations that exceeded the WHO Guidelines for Drinking-water Quality of
1.5mg/L of fluoride] and lower IQ. These include 3 prospective cohort studies and 15 cross-
sectional studies (12 of which indicated that exposure likely preceded the outcome). Forty-six of
53 low-quality studies in children also reported an association between higher fluoride exposure
assessment measures and lower IQ.
The moderate confidence rating was reached by starting with an initial confidence rating based
on key study design features of the body of evidence and then considering factors that may
increase or decrease the confidence in that body of evidence. The initial moderate confidence
rating is based on 15 of the 19 low risk-of-bias studies that have 3 of the 4 key study design
features shown in Figure 1 (i.e., exposure occurred prior to outcome, individual-based outcomes
were evaluated, and a comparison group was used). Three of these studies were prospective
cohort studies, and 12 were cross-sectional studies that provided evidence of long-term, chronic
fluoride exposure prior to outcome measurement.
Many studies in this assessment relied on drinking-water fluoride levels (both group-level
measures and individual-level measures), rather than measures of total fluoride exposure, to
establish exposed versus “unexposed” or reference groups. Although fluoride in water is a major
source of exposure [comprising 40% to 70% of total exposure (USEPA 2010)], other sources of
fluoride provide variable amounts that depend on personal preferences and habits. The use of
dental products containing fluoride and consuming foods and beverages prepared with
fluoridated water can also result in measurable exposures (USEPA 2010). Green et al. (2019)
suggested that significant exposures occur from black tea consumption. Thus, drinking water
fluoride levels may, but usually do not, reflect total fluoride exposure. This could be a potential
limitation in studies that rely on water fluoride data to assess fluoride exposure (in particular,
earlier studies). Because water is only part of a person’s total exposure to fluoride, this limitation
would likely result in an underestimate of exposure to fluoride. In other words, in studies where
the exposure metric was drinking water fluoride concentrations that were lower than the WHO
Drinking Water Quality Guideline of 1.5 mg/L for example, actual exposures may be
underestimated due to exposures from other sources. This limitation is less of a concern in areas
where fluoride in the drinking water is high because drinking water likely contributes a larger
proportion of the total fluoride intake in those areas as compared with areas where fluoride in the
drinking water is lower.
This review found that the quality of exposure assessment measures has improved over the years.
More recent studies by Valdez Jiménez et al. (2017), Bashash et al. (2017), and Green et al.
(2019) that support the inverse associations between estimated total fluoride exposure and
children’s IQ and other cognitive neurodevelopmental effects, used individual measures of

79
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

urinary fluoride, either maternal urine collected prenatally or children’s urine. Studies using
different types of exposure assessment measures reported similar findings of an association
(NRC 2006), which strengthens confidence in earlier studies that reported IQ deficits with high
group-level fluoride exposure. However, there is less certainty in the quantitative estimates of the
magnitude of IQ deficits from earlier studies that used group-level exposure assessment
measures than the estimates from more recent studies that used individual-level exposure
assessment measures.
It is worth noting that there are circumstances wherein typical children’s water consumption
considered with water fluoride levels may substantially underestimate total fluoride exposure.
One example is bottle-fed infants wherein nutrition is provided by powdered formula that is
rehydrated with fluoridated water (Till et al. 2020). To decrease an exclusively formula-fed
infant’s exposure to fluoride, for the purpose of reducing risk of dental fluorosis, the Centers for
Disease Control and Prevention recommends using low-fluoride bottled water to mix with infant
formula (CDC 2015). A few studies also support the hypothesis that individuals with certain
genetic variants in dopamine receptor D2 or catechol-O-methyltransferase may be at heightened
sensitivities to the potential detrimental cognitive effects of fluoride exposure (Cui et al. 2018;
Zhang et al. 2015b), potentially impacting dopamine catabolism and receptor sensitivity. Given
the growing body of evidence suggesting an inverse association between estimated total fluoride
intake and certain neurodevelopmental effects in children, differential exposures to total fluoride
intake and genetic susceptibilities of children to fluoride may represent special situations that
would appear to warrant further research.
The following section briefly recaps the strength of the epidemiological evidence for an
association between fluoride exposure and cognitive neurodevelopmental deficits. This is
followed by a more detailed listing of limitations of the evidence base and limitations of the
systematic review, with some suggestions of areas where further research may be most
beneficial.

Limitations of the Evidence Base

Limitations in the epidemiological studies with low risk of bias include:
• Few studies are available that assessed the association between fluoride exposure and
cognitive function (particularly IQ) in adults and attention-related disorders including
ADHD in children and adults.
• Heterogeneity in outcomes was assessed for other neurobehavioral outcomes, limiting
the assessment of other possible effects in children.
• Studies rarely separated the results by sex or provided information to indicate that sex
was not a modifying factor.
• Associations between lower estimated total fluoride exposure [e.g., as in
approximations of exposure such as drinking water fluoride concentrations that were
lower than the WHO Guidelines for Drinking-water Quality of 1.5 mg/L of fluoride
(WHO 2017)] and children’s IQ remain unclear. More studies estimating lower
exposure levels are needed to fully understand potential associations in ranges
typically found in the United States (i.e., <1.5 mg/L in water). However, it should be
noted that, as of April 2020, CWS supplying water with ≥1.5 mg/L naturally

80
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

occurring fluoride served 0.59% of the U.S. population (~1.9 million people) (CDC
Division of Oral Health 2020). This indicates that the moderate confidence in the
association between higher fluoride exposure and lower IQ is relevant, at a minimum
to children living in these areas of the United States where fluoride in drinking water
is known to be at or above 1.5 mg/L. This is only compounded by additional
exposures to fluoride from other sources.
• No high-quality studies investigating the association between fluoride exposure and
neurodevelopmental or cognitive effects in adults or children have been conducted in
the United States.
• No studies are available to evaluate lifelong exposure in adults or fluoride exposure
over a child’s lifetime and neurodevelopmental or cognitive changes over time.
• Biomarkers that reflect total fluoride exposure are exposure estimates and not a
measure of actual exposure. Although we have noted above why spot urine samples
are considered low potential for bias, they do represent recent exposure and can vary.
Although human serum levels tend to reflect fluoride levels in water (IPCS 2002),
they vary widely during the day, and only rarely were they measured or reported in
the literature that was evaluated.
• The database does not allow for comparison of ages and possible changes at different
developmental stages in children to assess if there is a delay in development or if
associations persist.
• The database does not allow for establishing clear correlations between prenatal and
postnatal exposures.
Limitations in the epidemiological studies with high risk of bias include:
• Many of the original publications were in a non-English language and provided
limited details on methodology.
• Most studies lacked information regarding exposure and/or had serious limitations in
the exposure assessment measures. Exposure assessment measure concerns include
limited individual exposure information, a lack of information on fluoride sampling
methods and timing of the exposure assessment measurements, a lack of quantitation
of levels of fluoride in drinking water in a few studies, and a lack of individual-level
information on fluorosis in areas reported to be endemic for fluorosis.
• The comparison groups in studies conducted in areas endemic for fluorosis still may
have been exposed to high levels of fluoride or levels similar to those used in water
fluoridation in the United States. This factor may have limited the ability to detect
true effects.
• Many studies did not provide sufficient direct information (e.g., participation rates or
methods for selection) to evaluate selection bias.
• Failure to address important covariates was an issue for most of the studies. Some
studies conducted simple statistical analyses without accounting for any covariates in
the analysis, although many noted similarities between the study populations. In cases
where adjustments in analyses were made, often these studies did not account for

81
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

covariates considered critical for that study population and outcome including co-
exposures.
• Studies conducted in areas with high, naturally occurring fluoride levels in drinking
water often did not account for potential exposures to arsenic or iodine deficiencies in
study subjects in areas where these substances were likely to occur.
• Many studies lacked information on whether the outcome assessors were blind to the
exposure group, including studies that examined children in their schools and subjects
from high-fluoride communities.
Limitations in the animal and mechanistic evidence base include:
• The overall quality of the experimental animal studies is poor, and there are relatively
few well-designed and well-performed studies at lower fluoride exposure levels (i.e.,
<20 ppm, which is roughly equivalent to human exposure of <4 ppm).
• The understanding of the specific molecular events responsible for fluoride’s adverse
effects on neurobehavioral function is poor.
A key data gap in the human and animal bodies of evidence includes the need for mechanistic
insight into fluoride-related neurodevelopmental or cognitive changes.

Limitations of the Systematic Review

This systematic review has relatively few limitations. The human body of evidence included a
large database of observational studies. Most of the observational studies were cross-sectional;
however, 12 of these were considered to provide sufficient evidence that exposure occurred prior
to the outcome. In addition, the systematic review covered a wide range of study designs,
populations, and measures of fluoride exposure. The systematic review was designed to cover
reports on all potential mechanistic data including effects on the thyroid. After review of the
studies evaluating thyroid effects, studies that only evaluated goiters and other effects on thyroid
size were not considered in this review. This is not considered a limitation because these studies
did not include specific information on thyroid hormones that could indicate a mechanism for
thyroid involvement in neurodevelopment. In addition, review of the mechanistic data was
limited to in vivo studies with at least one concentration below 20 ppm. This is not considered a
limitation for the systematic review because the mechanistic body of evidence was used to
evaluate biological plausibility for the effects observed in humans; therefore, data were limited to
concentrations that would be more reflective of human exposures. The decision to not more
closely evaluate the in vitro data is not considered a limitation because there were sufficient in
vivo data, and no key events were identified where in vitro data would provide additional insight.
The supplemental literature search for non-English-language studies not indexed in traditional
databases supports the comprehensive nature of the literature search strategy for this systematic
review. In the absence of guidance on the most complete non-English-language databases that
may contain health studies of fluoride, a standard systematic review approach for database
selection was followed whereby a set of exemplar documents, called ‘seed studies’, were used.
Databases were selected that identified non-English-language studies of fluoride that we were
aware of and were not captured in searches of databases from the main literature search. This
informed approach influenced the selection process; however, this is not considered a limitation
because it provided an objective measure by which to compare databases. Following the

82
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

recommendation of the NASEM committee in its review of the September 16, 2020, draft
monograph, the experimental animal section has been removed and is not included in this
monograph. Although the deficiencies identified in the animal body of evidence support this
removal (see Animal Learning and Memory Data for further explanation), NTP acknowledges
that the absence of the experimental animal data is a limitation of this systematic review. For the
purpose of this review, NTP considers the experimental animal data to be inadequate to inform
whether fluoride exposure is associated with cognitive effects (including cognitive
neurodevelopmental effects) in humans.

83
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Summary
This systematic review evaluated the available animal and human literature concerning the
association between fluoride exposure and cognitive neurodevelopment. The available data on
potential mechanisms to evaluate biological plausibility were also assessed. Existing animal
studies provide little insight into the question of whether fluoride exposure affects IQ. Human
mechanistic studies were too heterogenous and limited in number to make any determination on
biological plausibility. The body of evidence from studies on adults is also limited and provides
low confidence that fluoride exposure is associated with adverse effects on adult cognition.
There is, however, a large body of evidence on inverse associations between total fluoride
exposure and IQ in children. There is also some evidence that higher fluoride exposure is
associated with other neurodevelopmental and cognitive effects; although, because of the
heterogeneity of the outcomes, there is low confidence in the literature for these other effects.
This review finds, with moderate confidence, that higher estimated fluoride exposures (e.g., as in
approximations of exposure such as drinking water fluoride concentrations that exceed the World
Health Organization Guidelines for Drinking-water Quality of 1.5 mg/L of fluoride) are
consistently associated with lower IQ in children. The moderate confidence in the inverse
association between fluoride exposure and children’s IQ is based primarily on studies with
estimated fluoride exposures higher than what is generally associated with consumption of
optimally fluoridated water in the United States. Associations between lower total fluoride
exposure [e.g., as in approximations of exposure such as drinking water fluoride concentrations
that were lower than the WHO Guidelines for Drinking-water Quality of 1.5 mg/L of fluoride
(WHO 2017)] and children’s IQ remain unclear. However, because people receive fluoride from
multiple sources (not just drinking water), individuals living in areas with optimally fluoridated
water can have total fluoride exposures higher than the concentration of their drinking water. In
addition, there are people living in the United States who live in areas with naturally occurring
fluoride in drinking water that is higher than 1.5 mg/L. As of April 2020, community water
systems supplying water with ≥1.5 mg/L naturally occurring fluoride served 0.59% of the U.S.
population (~1.9 million people) (CDC Division of Oral Health 2020). This indicates that the
moderate confidence in the inverse association between fluoride exposure and children’s IQ is
relevant to some children living in the United States, including at a minimum those living in
areas where fluoride in drinking water is known to be at or above 1.5 mg/L.
Additional exposures to fluoride from other sources would increase total fluoride exposure. The
moderate confidence conclusions may also be relevant to people living in optimally fluoridated
areas of the United States depending on the extent of their additional exposures to fluoride from
sources other than drinking water. Because no studies of fluoride exposures and IQ have been
performed in children in the United States and no nationally representative urinary fluoride
levels are available, targeted research that prospectively examines the association between
fluoride exposure and children’s IQ in optimally fluoridated areas of the United States is needed
to add clarity to the existing data.

84
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Addendum: Updated Literature Search (May 2020 through

October 2023)
Introduction
The multiple scientific peer reviews of this monograph are outlined in Appendix B, Table B-1.
During the final stage of review by the NTP Board of Scientific Counselors (BSC), a
recommendation was made by the BSC Working Group, which was adopted by the full BSC, to
update the literature for this monograph that examined the association between fluoride exposure
and children’s IQ. To comply with this recommendation, an update to the May 2020 literature
search was completed in October 2023. The updated literature search was conducted using the
same methods as those described in the Main Literature Search section of this document and in
the protocol (https://ntp.niehs.nih.gov/go/785076).
The results of this updated search are outlined below and are shown in the reference flow
diagram (Addendum Figure 1). Six low risk-of-bias studies reported on new cohorts of children
(i.e., study populations that had not previously been considered in the confidence assessment of
the draft monograph reviewed by the BSC Working Group). Therefore, the results of this
updated literature search have been included as this Addendum to the monograph rather than
incorporated into the monograph. As such, this Addendum was prepared to describe relevant
information from the updated literature search on the association between fluoride exposure and
children’s IQ and to discuss any potential impact of this information on the confidence
conclusions of the monograph. This Addendum also includes a discussion of meta-analyses of
studies examining the relationship between children’s estimated fluoride exposure and IQ to
provide further context of the findings from the updated literature search.
In addition to children’s IQ studies, several studies (including Cantoral et al. 2021; and Ibarluzea
et al. 2021) were found during the updated literature search that fall under the “Other
Neurodevelopmental or Cognitive Effects in Children” or “Cognitive Effects in Adults” outcome
categories of the monograph. Therefore, these studies are not discussed in this Addendum;
however, they are included in sensitivity analyses in the DTT meta-analysis, which is a separate
peer-reviewed journal publication (DTT Meta-analysis, Taylor et al. 2024, in press).

85
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Addendum Figure 1. Reference Flow Diagram for Updated Literature Searcha

aAn interactive version of this figure is available here.
*Studies may have been excluded for more than one reason; the first reason identified was recorded.

Results
Twenty-eight studies were published on the association between fluoride exposure and children’s
IQ between May 2020 (the end date of the literature search for the monograph) and October
2023 (the date of this updated literature search). Twelve studies (seven cross-sectional and five
prospective cohort studies from nine different study populations) had lower potential for bias.
Sixteen cross-sectional studies had higher potential for bias (see Addendum Figure 2).

Addendum Figure 2. Risk-of-bias Heatmap for Children’s IQ Studies Identified During Updated
Literature Search

An interactive version of this figure is available here.

86
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Low Risk-of-bias Studies

The updated literature search identified 12 low risk-of-bias IQ studies in children. None of the
new low risk-of-bias studies evaluated populations in the United States.
Six of the 12 low risk-of-bias studies were conducted on populations already discussed in the
main portion of the monograph (see Addendum Table 1). Results from these studies (Farmus et
al. 2021; Goodman et al. 2022a; Goodman et al. 2022b; Wang et al. 2021c; Yu et al. 2021; Zhou
et al. 2021) were similar to the results of the original studies, but included additional analyses
and, in some cases, additional mechanistic data.
Three of the studies that evaluated new populations (i.e., did not evaluate populations that
overlapped with studies included in the main portion of the monograph) reported statistically
significant inverse associations (at p-value < 0.05) between urinary fluoride exposures and
children’s IQ (Feng et al. 2022; Zhao et al. 2021). All three were cross-sectional studies
conducted in China. Zhao et al. (2021) conducted a cross-sectional study in Tianjin City in 2018
with a stratified multistage random sampling of 6–11-year-olds from five randomly selected
primary schools from five randomly selected towns (four with historically high fluoride and one
without high fluoride). The study reported a statistically significant inverse association between
children’s urinary fluoride (median [Q1–Q3] = 1.03 [0.72–1.47] mg/L) and IQ scores (adjusted
β = −5.957, 95% CI: −9.712, −2.202 per 1-log-mg/L increase). Feng et al. (2022) conducted a
cross-sectional study in Tongxu County in 2017 that evaluated children in four randomly
selected primary schools. Using the median level of creatinine-adjusted urinary fluoride (i.e.,
1.33 mg/L), the children were separated into high and control fluoride groups (mean [SD] = 2.15
[0.91] mg/L and 0.83 [0.30] mg/L, respectively). There was a statistically significant inverse
association between creatinine-adjusted urinary fluoride and IQ scores in the high fluoride group
(adjusted β = −2.502, 95% CI: −4.411, −0.59 per 1-mg/L increase). There were no statistically
significant associations among the control group or when the two groups were combined or
between the mean IQ scores when comparing the two groups. Xia et al. (2023) conducted a
cross-sectional study in Jiangsu Province by selecting three communities with high fluoride
(described as >1.0 mg/L) and three control communities with fluoride below 1.0 mg/L based on
monitoring results from 2018 to 2019. Mean IQ of primary school children was significantly
lower in the high fluoride communities compared to the control communities (p < 0.001). There
was a statistically significant inverse association between urinary fluoride and IQ scores in the
high fluoride communities (β = −4.08, 95% CI: −3.04, −1.32 per 1-mg/L increase). The risk of
below-normal IQ scores increased with excessive urinary fluoride exposure, presence of dental
fluorosis, and increases of dental fluorosis grading.
Three of the six studies that evaluated new populations reported non-statistically significant
inverse associations between fluoride and children’s IQ. Lin et al. (2023) conducted a cross-
sectional study in Taichung, Taiwan from 2019 to 2020 by randomly selecting five elementary
schools. Two children classes from each grade were invited to participate. 18 The study reported
an inverse association between urinary fluoride and IQ scores (β = −0.97, 95% CI: −2.90, 0.97
per ln-mg/L increase) and creatinine-adjusted urinary fluoride and IQ scores (β = −2.21, 95% CI:

18
ERRATUM: An error was identified in the NTP Monograph on the State of the Science Concerning Fluoride
Exposure and Neurodevelopment and Cognition: A Systematic Review (NTP Monograph 08). This sentence
originally read, “Two children from each grade...,” which contained an error. It has been corrected to: “Two classes
from each grade...” (and the correction is italicized). [September 4, 2024]

87
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

−4.99, 0.57 per ln-mg/g increase). Grandjean et al. (2023) conducted a study using the Odense
Child Cohort (OCC). Creatinine-adjusted maternal urinary fluoride was inversely associated with
full-scale IQ scores of 7-year-old children when using 24-hour maternal urine samples (β =
−0.72, 95% CI: −3.24, 1.80 per doubling of maternal urinary fluoride). When using spot urine
samples, there was an inverse association with IQ in girls (β = −0.78, 95% CI: −3.64, 2.08) and a
positive association with IQ in boys (β = 2.14, 95% CI: −0.92, 5.20). Dewey et al. (2023) was a
unique prospective ecological cohort study that evaluated exposures before and after fluoridation
ceased in Calgary, Canada on May 19, 2011. Exposure was based on when the pregnancy
occurred in relation to when water fluoridation had stopped. Participants were considered fully
exposed to 0.7 mg/L fluoride if the women were pregnant the whole time before water
fluoridation stopped, partially exposed when part of the pregnancy occurred before water
fluoridation stopped, or not exposed if pregnancy began after fluoridation stopped. Using the
nonexposed participants as the reference group, this study reported inverse associations between
fluoride exposure and verbal comprehension index and visual spatial index in 3–5-year-old
children. There were inverse associations between fluoride exposure and full-scale IQ in fully
exposed boys (β = −0.71, 95% CI: −5.46, 4.04) compared to nonexposed boys and in partially
exposed girls (β = −0.62, 95% CI: −4.67, 3.43) compared to nonexposed girls. In this study, IQ
was assessed using the Wechsler Preschool and Primary Scale of Intelligence Fourth Edition:
Canadian (WPPSI-IVCND) which does not include a performance IQ score. As noted in the
main section of this document, fluoride exposure has been found to be more strongly associated
with performance scores compared to other IQ subtests (e.g., verbal IQ) (Green et al. 2019; Till
et al. 2020). In addition, the exposure was based solely on fluoridation status during pregnancy
and did not consider the quantity of water intake by the mothers or exposure to fluoride from
sources other than drinking water.
Nine low risk-of-bias studies examined specific aspects of the inverse association between
fluoride exposure and children’s IQ. Five of the studies evaluated mechanistic aspects associated
with fluoride exposure assessment measures including genetic variants, gene-environment
interactions, or effects on the cholinergic system. Other studies assessed the specific timing of
exposure (prenatal, infancy, and childhood) (Farmus et al. 2021), evaluated multiple IQ domains
(Dewey et al. 2023; Goodman et al. 2022a), or assessed the role of iodine on the association
between fluoride exposure and IQ (Goodman et al. 2022b).

88
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Addendum Table 1. Studies on IQ in Children since 2020a

Study Design Exposure Measures and Summary Assessment Outcome and
Study Summary of IQ Resultsb
(Location/Subjects) [n] Statistics Timing Methods
China
Xia et al. (2023) Cross-sectional Drinking water Children (ages IQ: Combined Significant inverse association between
Jiangsu Province (school Overall mean (SD): 1.18 (0.11) mg/L 8–12 years) Raven’s Test for children’s urinary fluoride and IQ scores in the
children) Rural China high fluoride group (adjusted β per 1.0 mg/L =
Control group: 0.73 (0.08) mg/L −4.08; 95% CI: −3.04, −1.32).
[711]
High fluoride group: 1.89 (0.17) High fluoride group had significantly lower
mg/L mean IQ scores than the control group
Children’s urine (p < 0.001).
Control group, median (min, max): Adjusted for age, sex, BMI, parental education
0.98 (0.33, 3.59) mg/L
High fluoride group: 2.63 (0.57, 8.84)
mg/L
Feng et al. Cross-sectional Children’s urine Children (ages IQ: Combined Significant inverse association between child’s
(2022) Tongxu County, Henan (creatinine adjusted), mean (SD): 8–12 years) Raven's Test for urinary fluoride and IQ scores in the high
Province (school 1.49 (0.95) mg/L Rural China fluoride group (adjusted β per 1.0 mg/L =
children) Control group: 0.83 (0.30) mg/L −2.502; 95% CI: −4.411, −0.593); probability of
[683] having “excellent” IQ were significantly
High fluoride group: 2.15 (0.91) decreased in the high fluoride group (adjusted
mg/L OR per 1 mg/L = 0.537; 95% CI: 0.290, 0.994);
no significant difference in mean IQ scores
between the control and high fluoride group.
Significant interactions observed between
fluoride exposure and methylenetetrahydrofolate
dehydrogenase, cyclohydrolase, and
formyltetrahydrofolate synthetase 1 (MTHFD1)
variants.
Adjusted for age, sex, BMI, gestational weeks,
maternal and paternal education, birth weight,
birth modes, and maternal age at pregnancy

89
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design Exposure Measures and Summary Assessment Outcome and

Study Summary of IQ Resultsb
(Location/Subjects) [n] Statistics Timing Methods
Wang et al. Cross-sectional Drinking water, mean (SD): 1.20 Children (ages IQ: Combined Significant inverse association between fluoride
(2021c) Tianjin City, 7 (0.95) mg/L 7–13 years) Raven’s Test for exposures (both water and urine) and IQ scores
Same towns/children Children’s urine: Rural China in both boys and girls (p-values for trend <0.05)
population as [709] 0.66 (0.67) mg/L Significant inverse association between water
Yu et al. (2018) fluoride levels >1.0 mg/L and IQ score for all
children (adjusted β for Q3 [1.00 – 1.60 mg/L]
vs. Q1 [≤ 0.30 mg/L]: −2.77; 95% CI: [−5.44,
−0.10; for Q4 [>1.60 mg/L] vs. Q1
[≤ 0.30 mg/L] −4.10; 95% CI: −6.71, −1.48).
Significant inverse association between urinary
fluoride levels and IQ score for all children
(adjusted β for Q3 [0.48–0.90 mg/L] vs. Q1 [≤
0.20 mg/L]: −3.02; 95% CI: –5.71, –0.33; for
Q4 [>0.90 mg/L] vs. Q1 [≤ 0.20 mg/L]: −4.49;
95% CI: −7.21, −1.77).
Significantly reduced probabilities of superior
and above IQ (≥120) for both water fluoride
(adjusted OR per 1 mg/L: 0.69; 95% CI: 0.54,
0.90) and urinary fluoride (OR per 1 mg/L: 0.67;
95% CI: 0.46, 0.97).
Water fluoride was significantly positively
associated with serum AChE and significantly
negatively associated with serum ChAT and
ACh; trends were similar for urinary fluoride,
except ACh did not have a statistically
significant trend.
Adjusted for age, sex, BMI, low birth weight,
maternal and paternal education, family income,
and urine creatinine (for urinary fluoride)

90
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design Exposure Measures and Summary Assessment Outcome and

Study Summary of IQ Resultsb
(Location/Subjects) [n] Statistics Timing Methods
Yu et al. (2021) Cross-sectional Drinking water, median (Q1–Q3) Children (ages IQ: Combined Higher fluoride associated with lower
Same Tianjin City, 7 High IQ group: 0.70 (0.40–1.00) 7–13 years) Raven’s Test for probability of high IQ for all 4 exposure
population as towns/children mg/L Rural China matrices (p-values for trend <0.001).
Yu et al. (2018) [952] Non-high IQ group: 1.00 (0.50–1.90) Water: adjusted OR for tertile 2 (0.61–1.40
High IQ (IQ ≥120) [173] mg/L mg/L) vs. tertile 1 (≤60 mg/L): 0.94; 95% CI:
0.64, 1.37; for tertile 2 (>1.40 mg/L) vs. tertile 1
Non-high IQ Children’s urine (≤60 mg/L): 0.39; 95% CI: 0.25, 0.61.
(70≤IQ<120) [779] High IQ group: 0.33 (0.13–0.81) Urine: adjusted OR for tertile 2 (0.23–1.80
mg/L mg/L) vs. tertile 1 (≤0.22 mg/L): 1.26; 95% CI:
Non-high IQ group: 0.60 (0.16–2.22) 0.87, 1.84; for tertile 3 (>1.80 mg/L) vs. tertile 1
mg/L (≤0.22 mg/L): 0.41; 95% CI: 0.26, 0.66.
Children’s hair Significant inverse association between fluoride
High IQ group: 8.26 (5.72–10.48) exposures and IQ scores for all 4 matrices (water
µg/g and urine data below).
Non-high IQ group: 14.39 (10.25– Water, adjusted β per 0.5-mg/L increase in water
20.56) µg/g fluoride at levels between 0.20–3.40 mg/L:
−1.16; 95% CI: −1.41, −0.91; at levels between
Children’s nail 3.40–3.90 mg/L: −4.21; 95% CI: −7.54, −0.87.
High IQ group: 11.71 (8.53–14.64) Urine, adjusted β per 0.5-mg/L increase in
µg/g urinary fluoride at levels between 0.01–1.60
Non-high IQ group: 19.76 (14.16– mg/L: 1.01; 95% CI: 0.34, 1.68; at levels
27.32) µg/g between 1.60–2.50 mg/L: −5.23; 95% CI: −7.07,
−3.39); at levels between 2.50–5.54 mg/L:
−0.34; 95% CI: −0.98, 0.30.
Significant interactions between water, urinary
and hair fluoride exposures and a SNV set of 5
genes related to neurodevelopment or
mitochondrial function.
Adjusted for age, sex, maternal and paternal
education

91
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design Exposure Measures and Summary Assessment Outcome and

Study Summary of IQ Resultsb
(Location/Subjects) [n] Statistics Timing Methods
Zhao et al. Cross-sectional Children’s urine, median (Q1–Q3): Children (ages IQ: Combined Significant inverse association between urinary
(2021) Tianjin City, 5 towns 1.03 (0.72–1.47) mg/L 6–11 years) Raven's Test for fluoride and IQ scores (adjusted β per 1-log-
(school children) Log-transformed urine, Mean (SD): Rural China mg/L = −5.957; 95% CI: −9.712, −2.202).
[567] 0.015 (0.252) Interactions with dopamine-related genes
(COMT, ANKK1, DAT1, and MAOA) were also
observed.
Adjusted for age, sex, BMI, maternal and
paternal education, household income, abnormal
birth, and maternal age at delivery
Zhou et al. Cross-sectional Drinking water, median (Q1–Q3) Children (ages IQ: Combined Significant inverse association between fluoride
(2021) Tianjin City, 30 Moderate fluoride group: 2.20 (1.50– 7–13 years) Raven’s Test for exposure and IQ scores:
Same villages/children 2.80) mg/L Rural China Water: adjusted β per 1-mg/L increase = −1.29;
population as [605] Normal fluoride group: 0.70 (0.40– 95% CI: −2.05, −0.52)
Yu et al. (2018) 0.70) mg/L Urine: adjusted β per 1-mg/L increase = −0.75;
Children’s urine 95% CI: −1.35, −0.15)
Moderate fluoride group: 2.29 (1.94– Significant inverse association between water
2.88) mg/L and urine fluoride exposures and circulating
mitochondrial DNA (mtDNA). Significantly
Normal fluoride group: 0.15 (0.08– increased odds of excellent intelligence in
0.27) mg/L children in the highest tertile of mtDNA
compared to children in the lowest tertile.
Significant association between mtDNA and
higher odds of having excellent intelligence in
girls only. No significant association between
mtDNA and children’s IQ scores.
Adjusted for age, sex, BMI, low birth weight,
household income, maternal and paternal
education

92
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design Exposure Measures and Summary Assessment Outcome and

Study Summary of IQ Resultsb
(Location/Subjects) [n] Statistics Timing Methods
Taiwan
Lin et al. (2023) Cross-sectional Children’s urine, mean (SD) Children (ages IQ: Raven’s No significant associations between children’s
Taichung (school Reference: 0.38 (0.25) mg/L 6–12 years) Colored urinary fluoride exposure or dental fluorosis and
children) Progressive children’s IQ; nonsignificant inverse association
Dental fluorosis: 0.46 (0.32) mg/L Matrices-Parallel, between urinary fluoride and IQ scores (adjusted
[562]
Taiwan edition β per ln-mg/L increase = −0.97, 95% CI: −2.90,
0.97) and creatinine-adjusted urinary fluoride
and IQ scores (β per ln-mg/g increase = −2.21,
95% CI: −4.99, 0.57).
Adjusted for age, area of school, BMI, sex,
history of allergic rhinitis, parental education
Mexico
Goodman et al. Cohort (prospective) Maternal urine during pregnancy Children (ages Full-scale, Significant inverse associations between
(2022a) Mexico City/Early Life (creatinine adjusted), mean (SD): 6–12 years) performance, and maternal urinary fluoride and child’s full scale
Same Exposures in Mexico to 0.90 (0.39) mg/L verbal IQ: WASI, IQ score (adjusted β per 0.5-mg/L = −2.01; 95%
population as Environmental Toxicants Spanish Version CI: −3.66, −0.46); performance IQ (adjusted β
Bashash et al. (ELEMENT) participants per 0.5-mg/L = −1.80; 95% CI: −3.39, −0.21);
(2017) [348] and verbal IQ (adjusted β per 0.5-mg/L = −1.93;
95% CI: −3.67, −0.18).
IQ analysis [278]
Adjusted for gestational age, birth weight, sex,
parity (being the first child), age at outcome
measurement, time of testing and maternal
characteristics including smoking history,
marital status, age at delivery, education,
maternal education, and cohort
Denmark
Grandjean et al. Cohort (prospective) Maternal urine (creatinine adjusted), Children (age 7 Full-scale IQ: No significant associations between maternal
(2023) Odense Child Cohort, mean (SD): 0.58 (0.32) mg/L years) Wechsler urinary fluoride exposure and full-scale IQ;
Odense Intelligence Scales however, there was an inverse association for
[837] for Children, 24-hr urine samples (adjusted β per doubling of
Danish version maternal urinary fluoride = −0.72, 95% CI:
IQ analysis [460] −3.24, 1.80). Spot urine samples had a
Boys [239], girls [221] nonsignificant inverse association with IQ in
girls (β per doubling of maternal urinary fluoride
= −0.78, 95% CI: −3.64, 2.08) and a

93
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design Exposure Measures and Summary Assessment Outcome and

Study Summary of IQ Resultsb
(Location/Subjects) [n] Statistics Timing Methods
nonsignificant positive association with IQ in
boys (β per doubling of maternal urinary
fluoride = 2.14, 95% CI: −0.92, 5.20).
Adjusted for parental education, preterm birth,
age at the time of testing, examiner,
breastfeeding duration, school grade, school
type, smoking status, alcohol consumption
Canada
Dewey et al. Cohort (prospective Not exposed/partially exposed/fully Children Full-scale IQ: No significant associations between fluoride
(2023) ecological) exposed to fluoridated drinking water (ages 3−5 years) Wechsler exposure and full-scale IQ.
APrON Cohort, Calgary (0.7 mg/L) during pregnancy Preschool and
[616] Exposure groups based on when the Primary Scale of
pregnancy occurred in relation to Intelligence,
when water fluoridation had stopped. Fourth Edition
(WPPSI-IV CND)
Farmus et al. Cohort (prospective) Maternal urine during pregnancy. Children Full-scale, Significant inverse associations between
(2021) 6 of 10 cities/Maternal- mean (SD) (ages 3−4 years) performance, and maternal urinary fluoride and full-scale IQ in
Same Infant Research on 0.53 (0.37) mg/L verbal IQ: boys (adjusted β per 0.5-mg/L = −2.48; 95% CI:
population as Environmental Chemicals First trimester: 0.44 (0.46) mg/L Wechsler −4.30, −0.66) and for performance IQ in boys
Green et al. (MIREC) [596] Preschool and and all children (adjusted β per 0.5-mg/L
(2019) and Till Second trimester: 0.51 (0.48) mg/L Primary Scale of = −4.02; 95% CI: −6.15, −1.89 and adjusted β
Maternal urine [593] Intelligence, Third per 0.5-mg/L = −3.15; 95% CI: −4.85, −1.44,
et al. (2020) Third trimester: 0.65 (0.53) mg/L
Infant fluoride intake Edition (WPPSI- respectively). No significant associations
[442] Infant fluoride intake: 0.14 (0.13) III) between maternal urinary fluoride and full-scale
mg/day IQ and performance IQ in girls (adjusted β per
Children’s urine [434]
Children’s urine: 0.51 (0.39) mg/L 0.5-mg/L = −0.31; 95% CI: −2.76, 2.14 and
Boys [291], girls [305] adjusted β per 0.5-mg/L = −1.58; 95% CI:
−4.43, 1.28, respectively).
Significant inverse associations between infant
fluoride intake and performance IQ in girls
(adjusted β per 0.1-mg/day = −2.03; 95% CI:
−3.43, −0.63) and all children (adjusted β per
0.1-mg/day = −1.58; 95% CI: −2.59, −0.57).
No significant associations between fluoride
exposure measures (maternal urine, child urine

94
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Study Design Exposure Measures and Summary Assessment Outcome and

Study Summary of IQ Resultsb
(Location/Subjects) [n] Statistics Timing Methods
or infant intake) and verbal IQ. No significant
associations between child’s urinary fluoride
levels and full-scale, performance or verbal IQ
scores.
Adjusted for HOME score, maternal education,
race, age at urine sampling, and prenatal
secondhand smoke exposure
Goodman et al. Cohort (prospective) Maternal urine during pregnancy Children Full-scale IQ: Significant inverse association between maternal
(2022b) 6 of 10 cities/Maternal- (corrected for creatinine), median (ages 3−4 years) Wechsler urinary fluoride and full-scale IQ (adjusted β
Same Infant Research on (IQR): 0.61 (0.49) mg/g Preschool and [SE] per 0.5-mg/g creatinine = −5.89 [1.85]).
population as Environmental Chemicals Boys: 0.63 (0.52) mg/g Primary Scale of Significant 3-way interaction between maternal
Green et al. (MIREC) [601] Intelligence, Third urinary fluoride, maternal urinary iodine, and
Girls: 0.61 (0.48) mg/g Edition (WPPSI-
(2019) and Till Maternal urine [366] child’s sex for full-scale IQ; significant
et al. (2020) III) interaction between maternal urinary fluoride
Boys [186] and maternal urinary iodine in boys, but not
Girls [180] girls; in boys, low maternal urinary iodine was
associated with a larger decrease in full-scale IQ
compared with boys whose mothers had
adequate iodine (adjusted β per 0.5 mg/g
creatinine: −4.65 and −2.95, respectively).
Adjusted for maternal education, maternal race,
study site, and HOME score
HOME = Home Observation Measurement of the Environment; IQ = intelligence quotient; IQR = interquartile range; OR = odds ratio; Q1, Q3 = first and third quartiles; SD = standard
deviation; SNV = single nucleotide variant; WASI = Wechsler Abbreviated Scale of Intelligence (Spanish version); WPPSI = Wechsler Preschool and Primary Scale of Intelligence.
aIncludes low risk-of-bias studies only.
bAssociations between fluoride assessment measures and IQ were reported quantitatively, when possible. For studies with multiple analyses and results, the table summarizes key findings

and is not a comprehensive summary of all findings reported in the study. Results also indicate when a study reported no significant association between fluoride assessment measures and
IQ.

95
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

High Risk-of-bias Studies

There were 16 high risk-of-bias IQ studies in children published since the last literature search
update. Thirteen of the 16 studies reported an inverse association between fluoride exposure and
IQ in areas of China (6 studies), India (5 studies), Indonesia (1 study), and Pakistan (1 study).
Among the three studies that did not report an inverse association between fluoride exposure and
IQ, one study conducted in the Dominican Republic reported that, “Almost 99% of the evaluated
students [living in an area of endemic fluorosis] got intellectually lower results than the average
score according to Raven’s test” and that there was no statistically significant difference in IQ
percentiles across dental fluorosis scores. One study in Pakistan reported significantly higher IQ
scores in urban areas with higher fluoride compared with IQ scores in rural areas with lower
fluoride. One study in India reported that there was no significant association between fluoride
levels above and below 0.5 ppm and risk of low intelligence level in adolescents ages 10 to
14 years.

Human Mechanistic Studies of Fluoride and Cognitive Neurodevelopment

Although multiorgan toxicity of fluoride has long been recognized (IPCS 2002), there remains
no clear established mechanism for fluoride’s adverse effect on cognitive neurodevelopment (Li
et al. 2023). This is not due to a lack of potential mechanisms, but more likely due to the myriad
of ways that fluoride and fluoride complexes affect fundamental cellular biochemistry. These
include, but are not limited to, binding to and inhibiting enzymes involved in intermediary
metabolism and cell signaling, disrupting organelle structure and function (particularly of
mitochondria), and altering electrolyte balance and cellular pH. Nonetheless, the contribution of
these mechanisms to neurotoxicity at current levels of fluoride exposure to humans has been
questioned (reviewed in Johnston and Strobel 2020).
Five of the low risk-of-bias studies identified in this updated literature search explored potential
mechanisms of fluoride effects on children’s IQ (Feng et al. 2022; Wang et al. 2021c; Yu et al.
2021; Zhao et al. 2021; Zhou et al. 2021). Many of the mechanistic studies that are reviewed in
the main body of this monograph examined deficits in thyroid hormones associated with higher
fluoride exposures. Although not directly related to the question of fluoride’s effect on thyroid
function, one new study (Goodman et al. 2022b) noted a larger deficit in IQ for a 0.5-mg/g
increase in maternal urinary fluoride exposure for boys born to mothers whose urinary iodine
levels were low compared with boys born to mothers whose urinary iodine levels reflected an
adequate iodine intake.
An earlier high risk-of-bias study by Pratap et al. (2013) [reviewed in the monograph main text]
reported lower IQ scores and reduced serum acetylcholinesterase (AChE) activities for children
in areas with water fluoride levels above 2.0 ppm compared with control children living in areas
with water fluoride less than 1.5 ppm. In contrast, a recent study by Wang et al. (2021c) reported
that serum AChE was significantly increased in children with higher measures of fluoride in their
drinking water or urine and was associated with lower IQs. The Wang et al. (2021c) study also
looked at serum choline acetyl transferase (ChAT) and acetylcholine (ACh) levels and reported
significant inverse associations between serum ChAT and ACh levels and children’s drinking
water and urine fluoride measures, but associations with IQ were not significant. Assuming that
ChAT and its degradation enzyme (AChE) levels reflect those in the brain, the findings from
Wang et al. (2021c) are more consistent with current concepts of the central role of adequate

96
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

brain acetylcholine-mediated neurotransmission in learning and memory (Huang et al. 2022) and
are inconsistent with the findings reported in Pratap et al. (2013).
Three studies have examined possible relationships between fluoride exposure and dopamine
metabolism. A study by Zhang et al. (2015b) [reviewed in the monograph main text] examined
the interaction between fluoride exposure and genetic variants of the catechol-O-
methyltransferase (COMT) gene, which is partially responsible for dopamine degradation. The
study reported a stronger inverse association between measures of fluoride exposure and IQ in
children with the reference COMT gene variant (val/val) that exhibited lower activity, and
presumably higher brain dopamine levels, than among children who carried the heterozygous or
homozygous variant genotypes (met/val or met/met). Cui et al. (2018) [also reviewed in the
monograph] reported that an inverse association between fluoride exposure and children’s IQ
was found only in children homozygous for the TT variant genotype of the dopamine receptor-2
Taq 1A variant and not in children with the hybrid-type CT or wild-type CC genotypes. The TT
genotype is associated with the greatest D2 receptor availability (McDonell et al. 2018).
In a recent study of school-based children from endemic and non-endemic fluorosis areas in
Tianjin, China—not related to the Yu et al. (2018) study population in the same area—Zhao et
al. (2021) examined genetic variants of several dopamine-related genes as effect modifiers of the
inverse association between fluoride exposure and IQ. The study evaluated high and low activity
variants of genes thought to be involved in dopamine synthesis (ANKK1 Taq1A), reuptake
(DAT1 40 bp VNTR), and catabolism (COMT Val158Met and MAOA uVNTR). The study
observed no interactions between urinary fluoride and variants of the single genes but found a
high-dimensional interaction on IQ among urinary fluoride, ANKK1, COMT, and MAOA, and
concluded that dopamine-related pathways may play a critical role in the neurotoxicity of
fluoride. Overall, the three studies discussed above provide evidence that the inverse association
between fluoride exposure and children’s IQ may involve or interact with genetic differences in
dopamine-related pathways.
Feng et al. (2022) examined possible interactions between fluoride exposure measures and
children’s genetic variants in four specific loci in the multifunctional enzyme complex MTHFD1
(methylenetetrahydrofolate dehydrogenase, cyclohydrolase, formyltetrahydrofolate synthetase
1), which functions in folate metabolism, on the intelligence of children living in areas of
endemic fluorosis. This potential interaction was of interest to the authors because folate
deficiency is believed to be associated with neuropathological lesions. The study did not find
evidence of an interaction with any single locus but observed evidence of interactions between
multiple loci in several different statistical models, suggesting that “MTHFD1 polymorphisms
may be involved in the effects of fluoride exposure on intelligence in school-age children.”
Yu et al. (2021) examined potential interactions of genetic variants associated with nervous
system development and mitochondrial processes with fluoride-induced deficits in IQ in a subset
of children from Tianjin, China (also part of Yu et al. 2018 study). There were 53 functional
single nucleotide variants (SNVs) from 17 candidate genes selected for analysis. Of the 53
SNVs, 5 were on genes [catechol-O-methyl transferase (COMT), nitric oxide synthase1 (NOS1),
translocase of outer mitochondrial membrane (TOMM40), mitochondrial translational elongation
factor (TUFM), SLC25A12] that gave an SNV-set score associated with lower IQ. No single
gene was found associated with high intelligence, but clusterin protein (CLU) and translocase of
outer mitochondrial membrane 40 (TOMM40) interacted significantly with IQ and fluoride

97
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

measurements in hair of females. Pathway analyses did not show significant interactions with
fluoride exposures and IQ.
Zhou et al. (2021) evaluated mitochondrial DNA (mtDNA) copy number in relation to single-
copy nuclear DNA in lymphocytes and its potential role in the inverse association between
fluoride exposure measures and IQ in a subset of children from Tianjin, China (also part of Yu et
al. 2018 study). The study observed a significantly downward trend between mtDNA copy
number across tertiles of fluoride drinking water and urine measures. There were significantly
increased odds of having excellent intelligence in children in the highest tertile of mtDNA
compared with children in the lowest tertile, and a significant association between mtDNA and
higher odds of having excellent intelligence was found in girls, but not in boys. However, there
was no significant association between circulating mtDNA levels and children’s IQ scores. The
authors noted that further research is needed to characterize the relationship between mtDNA
content in peripheral blood and mtDNA content in neuronal tissues.
The studies outlined above provide valuable additional mechanistic information, potential
insights, and further suggestive evidence that there may be genetic variants that may affect the
susceptibility of some individuals to greater cognitive impairment than others associated with
exposures to fluoride during development. However, a clear mechanism for a fluoride effect on
cognitive neurodevelopment remains to be established.

Meta-analyses on Fluoride Exposure and Children’s IQ

Seven meta-analyses found statistically significant inverse associations between fluoride
assessment measures and children’s IQ (Addendum Table 2). Many of the studies included in
these meta-analyses lacked the information necessary to evaluate study quality, and most used
group-level estimates of fluoride exposure. The meta-analyses used a range of methods and
varied in the transparency, objectivity (e.g., predefined protocol), and rigorousness of the
analytic approaches applied, including procedures such as: study selection, quantitative
evaluation of overall associations, and selection of effect measures. For the effect measure, four
meta-analyses used a standardized mean difference (SMD) (Choi et al. 2012; DTT Meta-
analysis, Taylor et al. 2024, in press; Duan et al. 2018; Kumar et al. 2023), two used weighted
mean difference (WMD) (Tang et al. 2008; Veneri et al. 2023), and one used an odds ratio
(Miranda et al. 2021).
Tang et al. (2008) conducted a meta-analysis of 16 case-control studies with a pooled WMD
calculated using the Mantel-Haenszel method. The study reported significant inverse
associations that indicate “children who live in a fluorosis area have five times higher odds of
developing low IQ than those who live in a non-fluorosis area or a slight fluorosis area.” The
authors’ interpretation of the WMD as “odds of developing low IQ” is lacking because the
WMD compares mean differences in IQ scores, not probabilities. Other limitations of this
analysis included incorrect statements on study design assignment (all included studies are cross-
sectional, not case-control as labeled in the analysis); lack of investigation of sources of
heterogeneity (I2 not reported or incorrectly listed as “not applicable”); unclear choice of
sensitivity analyses; and no reporting of a predefined protocol or software used for the analyses.
Choi et al. (2012) assessed 27 studies comparing mean IQ scores in high fluoride exposure
groups to those in low fluoride exposure groups. Most of these study populations were from

98
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

China and were exposed to fluoride through drinking water. Choi et al. (2012) found an inverse
association between fluoride exposure and IQ in children (Addendum Table 2). In meta-
regression analyses, the study found that year of publication (but not mean age of the study
children) was a significant source of heterogeneity. When the analyses were restricted to the 16
studies that used the Combined Raven’s Test–The Rural edition in China (CRT-RC), the mean
age of the study children (but not year of publication) was a significant predictor of the estimated
SMD. While the study estimated a risk ratio for living in an endemic fluorosis area, authors
excluded studies with individual-level measures of exposure and were not able to perform a
formal dose-response analysis. Although software used for the meta-analysis was reported, the
study lacked a predefined protocol.
A more recent meta-analysis (Duan et al. 2018) assessed 26 studies that evaluated intelligence
levels in children exposed to high or low drinking water fluoride, including 15 that were also
included in the Choi et al. (2012) meta-analysis and 7 that were published after the 2011
inclusion period from Choi et al. (2012). Thirteen studies included in the Choi et al. (2012) meta-
analysis were not considered in the Duan et al. (2018) evaluation for unclear reasons. The Duan
et al. (2018) study included four studies from Iran and four studies from India, with the
remaining studies conducted in China. Duan et al. (2018) found an inverse association between
fluoride exposure and IQ in children (Addendum Table 2). In meta-regression analyses, the mean
age of study children significantly affected the relationship between high water fluoride levels
and children’s intelligence levels. Subgroup analyses included country, age (<10 or ≥10 years),
water fluoride level, type of intelligence assessment, and sex. Duan et al. (2018) also performed
a dose-response meta-analysis that suggested a significant association between increased water
fluoride exposure and lower intelligence levels; however, it is unclear which studies were
included in this analysis. The study reported both linear and nonlinear inverse relationships
between fluoride exposure and children’s intelligence levels. Software used in the meta-analysis
was reported; however, the study lacked a predefined protocol.
Miranda et al. (2021) performed a meta-analysis of 10 studies and found an association between
high fluoride exposure and decreased IQ as reflected by odds ratios (Addendum Table 2). The
meta-analysis has many serious methodological and reporting limitations. Regardless of how the
original studies reported the data, Miranda et al. (2021) chose to “classify the studies according
to the WHO guidelines that consider optimal levels between 0.5–1.0 mg/L (low levels) and > 2
mg/L, as higher levels for water fluoridation” thus limiting the body of evidence to the few
studies reporting the “optimal” low levels, and also excluding studies that reported exposure
levels qualitatively (e.g., just described as high and low). In addition, only “low risk of bias”
studies were included, and study selection raises serious concerns since many relevant studies
were excluded. Other limitations included inappropriate risk-of-bias assessment methodology
(i.e., best practices were not followed, and no rationale presented to support ratings that were
reported); improper conduct of the quantitative analysis; and use of an effect measure that is
limited in its interpretation and usefulness for assessing fluoride-IQ associations.
Veneri et al. (2023) performed a meta-analysis of 30 studies using the WMD as an effect
measure. The study reported significant inverse associations between fluoride exposure and IQ in
children (Addendum Table 2). There are several concerns with and limitations of this meta-
analysis. When the outcome assessment approaches are not identical—such as when IQ is
measured with various tests with different scales—a WMD is not an appropriate effect measure
(as compared to the SMD which accounts for test heterogeneity). Most importantly, the overall

99
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

pooled WMD is not valid. Several studies are counted multiple times for each exposure measure
of fluoride provided in the study (that result in the same WMD) with no evidence that the
correlation between study-specific effects was considered. This occurs for 6 of the 30 included
studies (Ahmad et al. 2022; Broadbent et al. 2015; Trivedi et al. 2007; Wang et al. 2021c; Yu et
al. 2021; Zhang et al. 2015b). For example, identical estimates from the same study (Zhang et al.
2015b) were included in the overall WMD calculation for three different exposure metrics
(water, serum and hair). Another limitation is lack of adequate justification for not considering or
including several studies that would appear to meet their inclusion criteria (Farmus et al. 2021;
Goodman et al. 2022a; Rocha-Amador et al. 2007). In addition, although the study research
question explicitly called for a separate assessment of prenatal exposure, and several studies had
maternal urinary fluoride measures, there is no analysis of prenatal fluoride exposure. There is
also no discussion of heterogeneity or evidence to indicate that sources of heterogeneity were
investigated using stratified analyses or meta-regression.
The Veneri et al. (2023) study also conducted a dose-response meta-analysis which suggested a
significant association between increased water or urinary fluoride exposure and lower
intelligence levels. However, the lack of detailed discussion of the shape of the dose-response
association at levels <1.5 mg/L (WHO Guidelines for Drinking-water Quality) or <0.7 mg/L is a
major limitation, especially given the current discourse about the association between fluoride
exposure and children’s IQ at lower levels of exposure.
Lastly, for assessing risk of bias, Veneri et al. (2023) employ the ROBINS-E tool (Higgins et al.
2024). This tool uses an algorithm to determine an individual study’s overall risk of bias, which
is based on the domain with the greatest risk of bias. For example, if one domain receives a
judgement of “high” risk of bias, the study is determined to be “high” risk of bias overall. This
approach assumes, without supporting evidence, that each type of bias has an equal influence
which may distort true study quality measures. Approaches that give different types of biases
equal weight or influence are discouraged by the National Academies of Science
(http://bit.ly/2CbAd1A) and others (Jüni et al. 1999; Singla et al. 2019; Stang 2010) as they
hinder a more thoughtful and informative critical examination of the evidence and do not
acknowledge that some potential biases are more important and influential than others (Arroyave
et al. 2021; Savitz et al. 2019; Steenland et al. 2020). In contrast, the OHAT approach is
consistent with Cochrane guidance (Sterne et al. 2023) in which studies may be considered
“lower” risk of bias (as compared to “higher” risk of bias) based on concern for bias on key
domains (e.g., exposure assessment, outcome assessment, confounding) which are determined on
a project-specific basis. Veneri et al. (2023) cautioned that the observational design of the
reviewed studies may have resulted in unmeasured or residual confounding, reducing confidence
in the associations. Although we agree that unmeasured or residual confounding may be a
concern, when assessing the impact of each covariate in the DTT Meta-analysis, Taylor et al.
(2024, in press), we carefully considered this issue and found that no trends were discernable,
suggesting that bias due to confounding could explain the consistency of the inverse association
across the body of evidence.
Kumar et al. (2023) performed a meta-analysis of 28 studies using SMD as an effect measure.
The study reported significant inverse associations between fluoride exposure and children’s
intelligence measures (IQ and cognition scores) (Addendum Table 2). The analysis included
double counting of three studies that contributed data to both endemic and nonendemic fluoride
areas (i.e., per Kumar et al. (2023) “> and < ~1.5 mg/L fluoride levels” respectively) (Sebastian

100
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

and Sunitha 2015; Xiang et al. 2003a; Xu et al. 1994). When restricting analyses to eight studies
from the nonendemic fluorosis area, the study found no significant associations. There was also
no significant association in nonlinear modeling with restricted cubic splines. Meta-analyses of
regression coefficients from studies with children’s and maternal urinary fluoride found
nonsignificant inverse associations. There are several serious concerns and limitations of these
analyses which failed to report many details critical to the conduct of a systematic literature
review and meta-analysis. These missing details include, but are not limited to: lacking a
predefined protocol; providing rationale or criteria for the risk-of-bias assessments; providing
details about how the SMDs were calculated; reporting the data from the individual studies that
were used to calculate the SMDs; describing the selection process for which data to use if a study
reported results of multiple models; describing assessment of heterogeneity (other than reporting
I2 values); describing the method for assessment of publication bias; and describing rationale for
choice of sensitivity analyses.
The DTT Meta-analysis, Taylor et al. (2024, in press)19 includes a meta-analysis and dose-
response meta-analysis using group-level exposure data, and a regression slopes meta-analysis
using individual-level exposure data. The group-level meta-analysis of 59 studies (n = 20,932
children) used SMD as the effect measure and reported statistically significant inverse
associations between fluoride exposure measures and children’s IQ. There was also a significant
dose-response relationship between group-level fluoride exposure and IQ. In stratified dose-
response meta-analyses of the low risk-of-bias studies, the direction of association remained
consistent when group-level exposure was restricted to <4 mg/L, <2 mg/L, and <1.5 mg/L
fluoride in drinking water and <4 mg/L, <2 mg/L, and <1.5 mg/L fluoride in urine. The
regression slopes meta-analysis of 13 studies (n = 4,475 children) with individual-level measures
of fluoride found a significant decrease in IQ of 1.63 points (95% CI: −2.33, −0.93; p-value
<0.001) per 1-mg/L increase in urinary fluoride. In subgroup analyses of both group-level and
individual-level data, the direction of the association remained inverse when stratified by study
quality (high versus low risk of bias), sex, age group, outcome assessment, study location,
exposure timing, and exposure metric.
Although the use of various effect measures and methods makes comparison of the magnitude of
the associations difficult across meta-analyses, there is a consistent reporting of inverse
associations between fluoride exposure assessment measures and children’s IQ.

19
The NTP authors of this monograph conducted a companion systematic review and meta-analysis of fluoride
exposure and children’s IQ. Reference to this meta-analysis is cited in this monograph as “(DTT Meta-analysis,
Taylor et al. 2024, in press).”

101
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Addendum Table 2. Previous Meta-analyses on Exposures to Fluoride and Children’s IQ

Number of Pooled Effect Type, Estimate Heterogeneity
Analysis
Studies (95% CI) p-value I2
Tang et al. (2008) 16 WMD, −5.03 (−6.51, −3.55) NR NR
Choi et al. (2012) 27 SMD, −0.45 (−0.56, −0.34) <0.001 80%
Duan et al. (2018) 26 SMD, −0.52 (−0.62, −0.42) <0.001 69.1%
Miranda et al. (2021) 10 OR, 3.88 (2.41, 6.23) <0.0001 77%
Veneri et al. (2023) 30 (38 results) WMD, −4.68 (−6.45, −2.92) NR 98.75%
Kumar et al. (2023) 28 (31 results) SMD, −0.33 (−0.44, −0.22) <0.001 83%
DTT Meta-analysis, 59 SMD, −0.45 (−0.57, −0.33) <0.001 94%
Taylor et al. (2024, in
press)a
CI = confidence interval; NR = not reported; SMD = standardized weighted mean difference; OR = odds ratio of low IQ in the
high fluoride versus low fluoride groups; WMD = weighted mean difference.
aThe NTP authors of this monograph conducted a companion systematic review and meta-analysis of fluoride exposure and

children’s IQ (DTT Meta-analysis, Taylor et al. 2024, in press).

Conclusion
The current monograph concludes with moderate confidence that higher estimated fluoride
exposures (e.g., as in approximations of exposure such as drinking water fluoride concentrations
that exceed the WHO Guidelines for Drinking-water Quality of 1.5 mg/L of fluoride) are
consistently associated with lower IQ in children. The moderate confidence in the inverse
association between fluoride exposure and children’s IQ is based primarily on studies with
estimated fluoride exposures higher than what is generally associated with consumption of
optimally fluoridated water in the United States.
Compared to the body of literature reviewed in the current monograph that supports the existing
confidence statement, the studies identified in the updated literature search had similar study
designs and patterns of findings. Recent meta-analyses of the inverse association between
children’s IQ and fluoride exposures provide additional evidence of a dose-response relationship.
However, uncertainty remains in findings at the lower fluoride exposure range. As this body of
evidence matures, consideration for upgrading the moderate confidence conclusion to high
confidence based on additional evidence of dose-response relationships at lower fluoride levels
may be warranted.

102
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

References
Adedara IA, Abolaji AO, Idris UF, Olabiyi BF, Onibiyo EM, Ojuade TD, Farombi EO. 2017a.
Neuroprotective influence of taurine on fluoride-induced biochemical and behavioral deficits in
rats. Chem Biol Interact. 261:1-10. https://doi.org/10.1016/j.cbi.2016.11.011
Adedara IA, Olabiyi BF, Ojuade TD, Idris UF, Onibiyo EM, Farombi EO. 2017b. Taurine
reverses sodium fluoride-mediated increase in inflammation, caspase-3 activity, and oxidative
damage along the brain-pituitary-gonadal axis in male rats. Can J Phys Pharmacol. 95(9):1019-
1029. https://doi.org/10.1139/cjpp-2016-0641
Ahmad MS, Sarker MNI, Ahmad MN, Ali M, Sadiq, Abbas M. 2022. Does high fluoride intake
cause low IQ? A case of Islamic religious schools (madrassas) in rural and urban areas of Sindh,
Pakistan. Fluoride. 55(1):49-62.
Akinrinade ID, Memudu AE, Ogundele OM. 2015a. Fluoride and aluminium disturb neuronal
morphology, transport functions, cholinesterase, lysosomal and cell cycle activities.
Pathophysiology. 22(2):105-115. https://doi.org/10.1016/j.pathophys.2015.03.001
Akinrinade ID, Memudu AE, Ogundele OM, Ajetunmobi OI. 2015b. Interplay of glia activation
and oxidative stress formation in fluoride and aluminium exposure. Pathophysiology. 22(1):39-
48. https://doi.org/10.1016/j.pathophys.2014.12.001
An JA, Mei SZ, Liu AP, Fu Y, Wang CF. 1992. [Effect of high level of fluoride on children’s
intelligence]. Chin J Control Endem Dis. 7(2):93-94.
Arbuckle TE, Fraser WD, Fisher M, Davis K, Liang CL, Lupien N, Bastien S, Velez MP, von
Dadelszen P, Hemmings DG, et al. 2013. Cohort profile: The maternal-infant research on
environmental chemicals research platform. Paediatr Perinat Epidemiol. 27(4):415-425.
https://doi.org/10.1111/ppe.12061
Arroyave WD, Mehta SS, Guha N, Schwingl P, Taylor KW, Glenn B, Radke EG, Vilahur N,
Carreón T, Nachman RM, et al. 2021. Challenges and recommendations on the conduct of
systematic reviews of observational epidemiologic studies in environmental and occupational
health. J Expo Sci Environ Epidemiol. 31(1):21-30. https://doi.org/10.1038/s41370-020-0228-0
Baba NA, Raina R, Verma PK, Sultana M. 2014. Alterations in plasma and tissue
acetylcholinesterase activity following repeated oral exposure of chlorpyrifos alone and in
conjunction with fluoride in Wistar rats. Proc Natl Acad Sci India Sect B Biol Sci. 84:969-972.
https://doi.org/10.1007/s40011-013-0286-3
Banala RR, Nagapuri KK, Mohd KP, Reddy MM, Karnati PR. 2018. Carica Papaya leaf extract
as a neuroprotective agent against behavioral and neurotransmitter changes in brain of the rat
treated with sodium fluoride in pre- and post-natal periods. Pharmacogn Mag. 14(55 Suppl
1):S123-S131. https://doi.org/10.4103/pm.pm_378_17
Barberio AM, Hosein FS, Quiñonez C, McLaren L. 2017a. Fluoride exposure and indicators of
thyroid functioning in the Canadian population: Implications for community water fluoridation. J
Epidemiol Community Health. 71(10):1019-1025. https://doi.org/10.1136/jech-2017-209129

103
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Barberio AM, Quiñonez C, Hosein FS, McLaren L. 2017b. Fluoride exposure and reported
learning disability diagnosis among Canadian children: Implications for community water
fluoridation. Can J Public Health. 108(3):e229-e239. https://doi.org/10.17269/cjph.108.5951
Bartos M, Gumilar F, Gallegos CE, Bras C, Dominguez S, Mónaco N, Esandi MDC, Bouzat C,
Cancela LM, Minetti A. 2018. Alterations in the memory of rat offspring exposed to low levels
of fluoride during gestation and lactation: Involvement of the alpha7 nicotinic receptor and
oxidative stress. Reprod Toxicol. 81:108-114. https://doi.org/10.1016/j.reprotox.2018.07.078
Bashash M, Marchand M, Hu H, Till C, Martinez-Mier EA, Sanchez BN, Basu N, Peterson KE,
Green R, Schnaas L, et al. 2018. Prenatal fluoride exposure and attention deficit hyperactivity
disorder (ADHD) symptoms in children at 6–12 years of age in Mexico City. Environ Int. 121(Pt
1):658-666. https://doi.org/10.1016/j.envint.2018.09.017
Bashash M, Thomas D, Hu H, Martinez-Mier EA, Sanchez BN, Basu N, Peterson KE, Ettinger
AS, Wright R, Zhang Z, et al. 2017. Prenatal fluoride exposure and cognitive outcomes in
children at 4 and 6-12 years of age in Mexico. Environ Health Perspect. 125(9):097017.
https://doi.org/10.1289/ehp655
Bhatnagar M, Rao P, Sushma J, Bhatnagar R. 2002. Neurotoxicity of fluoride:
Neurodegeneration in hippocampus of female mice. Indian J Exp Biol. 40(5):546-554.
Bhatnagar M, Sukhwal P, Suhalka P, Jain A, Joshi C, Sharma D. 2011. Effects of fluoride in
drinking water on NADPH-diaphorase neurons in the forebrain of mice: A possible mechanism
of fluoride neurotoxicity. Fluoride. 44(4):195-209.
Broadbent JM, Thomson WM, Moffitt TE, Poulton R. 2015. Community water fluoridation and
intelligence: Broadbent et al. respond. Am J Public Health. 105(4):e3-e4.
https://doi.org/10.2105/AJPH.2015.302647
California Office of Environmental Health Hazard Assessment (OEHHA). 2011. Meeting
synopsis and slide presentations: Carcinogen Identification Committee meeting held on October
12, 2011. Sacramento, CA: California Office of Environmental Health Hazard Assessment.
[Accessed: August 19, 2019]. http://oehha.ca.gov/prop65/public_meetings/cic101211synop.html
Cantoral A, Téllez-Rojo MM, Malin AJ, Schnaas L, Osorio-Valencia E, Mercado A, Martínez-
Mier EA, Wright RO, Till C. 2021. Dietary fluoride intake during pregnancy and
neurodevelopment in toddlers: A prospective study in the progress cohort. Neurotoxicology.
87:86-93. https://doi.org/10.1016/j.neuro.2021.08.015
Caruana EJ, Roman M, Hernández-Sánchez J, Solli P. 2015. Longitudinal studies. J Thorac Dis.
7(11):E537-E540. https://doi.org/10.3978/j.issn.2072-1439.2015.10.63
CDC Division of Oral Health. 2020. Personal communication. September 3, 2020.
Centers for Disease Control and Prevention (CDC). 2013. Community water fluoridation:
Fluoridation statistics: 2012. Atlanta, GA: U.S. Department of Health and Human Services,
Public Health Service, Centers for Disease Control and Prevention. [Accessed: August 19, 2019].
https://web.archive.org/web/20190705104746/https://www.cdc.gov/fluoridation/statistics/2012st
ats.htm

104
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Centers for Disease Control and Prevention (CDC). 2015. Community water fluoridation FAQs:
Infant formula. Atlanta, GA: U.S. Department of Health and Human Services, Public Health
Service, Centers for Disease Control and Prevention. [Accessed: September 22, 2021].
https://web.archive.org/web/20210808212314/https:/www.cdc.gov/fluoridation/faqs/infant-
formula.html
Chen Y. 2012. Organophosphate-induced brain damage: Mechanisms, neuropsychiatric and
neurological consequences, and potential therapeutic strategies. Neurotoxicology. 33(3):391-400.
https://doi.org/10.1016/j.neuro.2012.03.011
Chen YX, Han FL, Zhou ZL, Zhang HQ, Jiao XS, Zhang SC, Huang MC, Chang TQ, Dong YF.
1991. [Research on the intellectual development of children in high fluoride areas]. Chin J
Control Endem Dis. 6 Suppl:99-100.
Chen YX, Han FL, Zhou ZL, Zhang HQ, Jiao XS, Zhang SC, Huang MC, Chang TQ, Dong YF.
2008. Research on the intellectual development of children in high fluoride areas. Fluoride.
41(2):120-124.
Chinoy NJ, Narayana MV. 1992. Studies on fluorosis in Mehsana District of North Gujarat. Proc
Zool Soc Calcutta. 45(2):157-161.
Choi AL, Sun G, Zhang Y, Grandjean P. 2012. Developmental fluoride neurotoxicity: A
systematic review and meta-analysis. Environ Health Perspect. 120(10):1362-1368.
https://doi.org/10.1289/ehp.1104912
Choi AL, Zhang Y, Sun G, Bellinger DC, Wang K, Yang XJ, Li JS, Zheng Q, Fu Y, Grandjean
P. 2015. Association of lifetime exposure to fluoride and cognitive functions in Chinese children:
A pilot study. Neurotoxicol Teratol. 47:96-101. https://doi.org/10.1016/j.ntt.2014.11.001
Chouhan S, Flora SJS. 2008. Effects of fluoride on the tissue oxidative stress and apoptosis in
rats: Biochemical assays supported by IR spectroscopy data. Toxicology. 254(1-2):61-67.
https://doi.org/10.1016/j.tox.2008.09.008
Chouhan S, Lomash V, Flora SJS. 2010. Fluoride-induced changes in haem biosynthesis
pathway, neurological variables and tissue histopathology of rats. J Appl Toxicol. 30(1):63-73.
https://doi.org/10.1002/jat.1474
Cui Y, Yu J, Zhang B, Guo B, Gao T, Liu H. 2020. The relationships between thyroid-
stimulating hormone and/or dopamine levels in peripheral blood and IQ in children with
different urinary iodine concentrations. Neurosci Lett. 729:134981.
https://doi.org/10.1016/j.neulet.2020.134981
Cui Y, Zhang B, Ma J, Wang Y, Zhao L, Hou C, Yu J, Zhao Y, Zhang Z, Nie J, et al. 2018.
Dopamine receptor D2 gene polymorphism, urine fluoride, and intelligence impairment of
children in China: A school-based cross-sectional study. Ecotoxicol Environ Saf. 165:270-277.
https://doi.org/10.1016/j.ecoenv.2018.09.018
Dewey D, England-Mason G, Ntanda H, Deane AJ, Jain M, Barnieh N, Giesbrecht GF,
Letourneau N, APrON Study Team. 2023. Fluoride exposure during pregnancy from a
community water supply is associated with executive function in preschool children: A

105
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

prospective ecological cohort study. Sci Total Environ. 891:164322.

https://doi.org/10.1016/j.scitotenv.2023.164322
Ding Y, Gao Y, Sun H, Han H, Wang W, Ji X, Liu X, Sun D. 2011. The relationships between
low levels of urine fluoride on children's intelligence, dental fluorosis in endemic fluorosis areas
in Hulunbuir, Inner Mongolia, China. J Hazard Mater. 186(2-3):1942-1946.
https://doi.org/10.1016/j.jhazmat.2010.12.097
DTT Meta-analysis, Taylor KW, Eftim SE, Sibrizzi CA, Blain RB, Magnuson K, Hartman PA,
Rooney AA, Bucher JR. 2024. Fluoride exposure and children's intelligence: A systematic
review and meta-analysis. In Press.
Du L, Wan C, Cao X, Liu J. 1992. [The effect of fluorine on the developing human brain]. Chin
J Pathol. 21(4):218-220.
Du L, Wan C, Cao X, Liu J. 2008. The effect of fluorine on the developing human brain.
Fluoride. 41(4):327-330.
Duan J, Zhao M, Wang L, Fang D, Wang Y, Wang W. 1995. A comparative analysis of the
results of multiple tests in patients with chronic industrial fluorosis. Guizhou Med J. 18(3):179-
180.
Duan Q, Jiao J, Chen X, Wang X. 2018. Association between water fluoride and the level of
children's intelligence: A dose-response meta-analysis. Public Health. 154:87-97.
https://doi.org/10.1016/j.puhe.2017.08.013
Farmus L, Till C, Green R, Hornung R, Martinez Mier EA, Ayotte P, Muckle G, Lanphear BP,
Flora DB. 2021. Critical windows of fluoride neurotoxicity in Canadian children. Environ Res.
200:111315. https://doi.org/10.1016/j.envres.2021.111315
Feng Z, An N, Yu F, Ma J, Li N, Du Y, Guo M, Xu K, Hou X, Li Z, et al. 2022. Do
methylenetetrahydrofolate dehydrogenase, cyclohydrolase, and formyltetrahydrofolate
synthetase 1 polymorphisms modify changes in intelligence of school-age children in areas of
endemic fluorosis? Chin Med J. 135(15):1846-1854.
https://doi.org/10.1097/CM9.0000000000002062
Gais S, Schönauer M. 2017. Untangling a cholinergic pathway from wakefulness to memory.
Neuron. 94(4):696-698. https://doi.org/10.1016/j.neuron.2017.05.010
Gao Q, Liu YJ, Guan ZZ. 2009. Decreased learning and memory ability in rats with fluorosis:
Increased oxidative stress and reduced cholinesterase activity in the brain. Fluoride. 42(4):277-
285.
Gao Q, Liu YJ, Wu CX, Long YG, Guan ZZ. 2008a. [Effects of fluoride on learning and
memory and cholinesterase activity in rat brains]. Chin J Endemiol. 27(2):128-130.
https://doi.org/10.3760/cma.j.issn.1000-4955.2008.02.004
Gao Q, Liu YJ, Wu CX, Long YG, Guan ZZ. 2008b. [Level of oxidative stress in rat brains and
learning and memory function of rats with chronic fluorosis]. Chin J Endemiol. 27(4):371-373.
https://doi.org/10.3760/cma.j.issn.1000-4955.2008.04.006

106
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Garman RH, Li AA, Kaufmann W, Auer RN, Bolon B. 2016. Recommended methods for brain
processing and quantitative analysis in rodent developmental neurotoxicity studies. Toxicol
Pathol. 44(1):14-42. https://doi.org/10.1177/0192623315596858
Goodman CV, Bashash M, Green R, Song P, Peterson KE, Schnaas L, Mercado-García A,
Martínez-Medina S, Hernández-Avila M, Martinez-Mier A, et al. 2022a. Domain-specific effects
of prenatal fluoride exposure on child IQ at 4, 5, and 6–12 years in the ELEMENT cohort.
Environ Res. 211:112993. https://doi.org/10.1016/j.envres.2022.112993
Goodman CV, Hall M, Green R, Chevrier J, Ayotte P, Martinez-Mier EA, McGuckin T,
Krzeczkowski J, Flora D, Hornung R, et al. 2022b. Iodine status modifies the association
between fluoride exposure in pregnancy and preschool boys' intelligence. Nutrients.
14(14):2920. https://doi.org/10.3390/nu14142920
Grandjean P, Meddis A, Nielsen F, Beck IH, Bilenberg N, Goodman CV, Hu H, Till C, Budtz-
Jørgensen E. 2023. Dose dependence of prenatal fluoride exposure associations with cognitive
performance at school age in three prospective studies. Eur J Public Health. 34(1):143-149.
https://doi.org/10.1093/eurpub/ckad170
Green R, Lanphear B, Hornung R, Flora D, Martinez-Mier EA, Neufeld R, Ayotte P, Muckle G,
Till C. 2019. Association between maternal fluoride exposure during pregnancy and IQ scores in
offspring in Canada. JAMA Pediatr. 173(10):940-948.
https://doi.org/10.1001/jamapediatrics.2019.1729
Green R, Rubenstein J, Popoli R, Capulong R, Till C. 2020. Sex-specific neurotoxic effects of
early-life exposure to fluoride: A review of the epidemiologic and animal literature. Curr
Epidemiol Rep. 7(4):263-273. https://doi.org/10.1007/s40471-020-00246-1
Güner S, Uyar-Bozkurt S, Haznedaroğlu E, Menteş A. 2016. Dental fluorosis and catalase
immunoreactivity of the brain tissues in rats exposed to high fluoride pre- and postnatally. Biol
Trace Elem Res. 174(1):150-157. https://doi.org/10.1007/s12011-016-0695-2
Guo XC, Wang RY, Cheng CF, Wei WS, Tang LM, Wang QS, Tang DX, Liu GW, He GD, Li
SL. 1991. [A preliminary investigation of the IQs of 7-13 year-old children from an area with
coal burning-related fluoride poisoning]. Chin J Endemiol. 10(2):98-100.
Guo XC, Wang RY, Cheng CF, Wei WS, Tang LM, Wang QS, Tang DX, Liu GW, He GD, Li
SL. 2008a. A preliminary investigation of the IQs of 7-13 year-old children from an area with
coal burning-related fluoride poisoning. Fluoride. 41(2):125-128.
Guo ZY, He YH, Zhu QX. 2001. [Research on the neurobehavioral function of workers
occupationally exposed to fluoride]. Ind Health Occup Dis. 27(6):346-348.
Guo ZY, He YH, Zhu QX. 2008b. Research on the neurobehavioral function of workers
occupationally exposed to fluoride. Fluoride. 41(2):152-155.
Guyatt G, Oxman AD, Akl EA, Kunz R, Vist G, Brozek J, Norris S, Falck-Ytter Y, Glasziou P,
deBeer H, et al. 2011. GRADE guidelines: 1. Introduction-GRADE evidence profiles and
summary of findings tables. J Clin Epidemiol. 64(4):383-394.
https://doi.org/10.1016/j.jclinepi.2010.04.026

107
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Haschek WM, Rousseaux CG. 1991. Handbook of toxicologic pathology. San Diego, CA:
Academic Press.
Health Canada. 2015. Third report on human biomonitoring of environmental chemicals in
Canada: Results of the Canadian Health Measures Survey Cycle 3 (2012–2013). Ottawa,
Ontario: Health Canada. https://www.canada.ca/content/dam/hc-sc/migration/hc-sc/ewh-
semt/alt_formats/pdf/pubs/contaminants/chms-ecms-cycle3/chms-ecms-cycle3-eng.pdf
Hefferon R, Goin DE, Sarnat JA, Nigra AE. 2024. Regional and racial/ethnic inequalities in
public drinking water fluoride concentrations across the US. J Expo Sci Environ Epidemiol.
34(1):68-76. https://doi.org/10.1038/s41370-023-00570-w
Higgins JPT, Green S. 2011. Cochrane handbook for systematic reviews of interventions.
Hoboken, NJ: John Wiley & Sons.
Higgins JPT, Morgan RL, Rooney AA, Taylor KW, Thayer KA, Silva RA, Lemeris C, Akl EA,
Bateson TF, Berkman ND, et al. 2024. A tool to assess risk of bias in non-randomized follow-up
studies of exposure effects (ROBINS-E). Environ Int. 186:108602.
https://doi.org/10.1016/j.envint.2024.108602
Hosur MB, Puranik RS, Vanaki S, Puranik SR. 2012. Study of thyroid hormones free
triiodothyronine (FT3), free thyroxine (FT4) and thyroid stimulating hormone (TSH) in subjects
with dental fluorosis. Eur J Dent. 6(2):184-190. https://doi.org/10.1055/s-0039-1698949
Howard BE, Phillips J, Tandon A, Maharana A, Elmore R, Mav D, Sedykh A, Thayer K,
Merrick BA, Walker V, et al. 2020. SWIFT-Active Screener: Accelerated document screening
through active learning and integrated recall estimation. Environ Int. 138:105623.
https://doi.org/10.1016/j.envint.2020.105623
Huang Q, Liao C, Ge F, Ao J, Liu T. 2022. Acetylcholine bidirectionally regulates learning and
memory. J Neurorestoratology. 10(2):100002. https://doi.org/10.1016/j.jnrt.2022.100002
Ibarluzea J, Gallastegi M, Santa-Marina L, Jiménez Zabala A, Arranz E, Molinuevo A, Lopez-
Espinosa MJ, Ballester F, Villanueva CM, Riano I, et al. 2021. Prenatal exposure to fluoride and
neuropsychological development in early childhood: 1-to 4 years old children. Environ Res.
207:112181. https://doi.org/10.1016/j.envres.2021.112181
International Programme on Chemical Safety (IPCS). 2002. Fluorides. Geneva, Switzerland:
World Health Organization, International Programme on Chemical Safety. Environmental Health
Criteria 227. https://inchem.org/documents/ehc/ehc/ehc227.htm
Jacqmin H, Commenges D, Letenneur L, Barberger-Gateau P, Dartigues JF. 1994. Components
of drinking water and risk of cognitive impairment in the elderly. Am J Epidemiol. 139(1):48-57.
https://doi.org/10.1093/oxfordjournals.aje.a116934
Jia B, Zong L, Lee JY, Lei J, Zhu Y, Xie H, Clemens JL, Feller MC, Na Q, Dong J, et al. 2019.
Maternal supplementation of low dose fluoride alleviates adverse perinatal outcomes following
exposure to intrauterine inflammation. Sci Rep. 9(1):2575. https://doi.org/10.1038/s41598-018-
38241-8

108
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Jiang C, Zhang S, Liu H, Guan Z, Zeng Q, Zhang C, Lei R, Xia T, Wang Z, Yang L, et al. 2014.
Low glucose utilization and neurodegenerative changes caused by sodium fluoride exposure in
rat's developmental brain. Neuromolecular Med. 16(1):94-105. https://doi.org/10.1007/s12017-
013-8260-z
Jin T, Han T, Wei Y, Wu Y, Wang Z, Zhang H. 2016. Investigation on working memory level of
children aged 8-12 years in coal-burning fluorosis area. Journal of Environment and Health.
33(5):409-411.
Johnston NR, Strobel SA. 2020. Principles of fluoride toxicity and the cellular response: A
review. Arch Toxicol. 94(4):1051-1069. https://doi.org/10.1007/s00204-020-02687-5
Jones S, Burt BA, Petersen PE, Lennon MA. 2005. The effective use of fluorides in public
health. Bull World Health Organ. 83(9):670-676.
Jüni P, Witschi A, Bloch R, Egger M. 1999. The hazards of scoring the quality of clinical trials
for meta-analysis. JAMA. 282(11):1054-1060. https://doi.org/10.1001/jama.282.11.1054
Khan AM, Raina R, Dubey N, Verma PK. 2017. Effect of deltamethrin and fluoride co-exposure
on the brain antioxidant status and cholinesterase activity in Wistar rats. Drug Chem Toxicol.
41(2):123-127. https://doi.org/10.1080/01480545.2017.1321009
Khandare AL, Gourineni SR, Validandi V. 2017. Dental fluorosis, nutritional status, kidney
damage, and thyroid function along with bone metabolic indicators in school-going children
living in fluoride-affected hilly areas of Doda District, Jammu and Kashmir, India. Environ
Monit Assess. 189(11):579. https://doi.org/10.1007/s10661-017-6288-5
Khandare AL, Validandi V, Gourineni SR, Gopalan V, Nagalla B. 2018. Dose-dependent effect
of fluoride on clinical and subclinical indices of fluorosis in school going children and its
mitigation by supply of safe drinking water for 5 years: An Indian study. Environ Monit Assess.
190(3):110. https://doi.org/10.1007/s10661-018-6501-1
Kheradpisheh Z, Mahvi AH, Mirzaei M, Mokhtari M, Azizi R, Fallahzadeh H, Ehrampoush MH.
2018a. Correlation between drinking water fluoride and TSH hormone by ANNs and ANFIS. J
Environ Health Sci Eng. 16(1):11-18. https://doi.org/10.1007/s40201-018-0290-x
Kheradpisheh Z, Mirzaei M, Mahvi AH, Mokhtari M, Azizi R, Fallahzadeh H, Ehrampoush MH.
2018b. Impact of drinking water fluoride on human thyroid hormones: A case-control study. Sci
Rep. 8(1):2674. https://doi.org/10.1038/s41598-018-20696-4
Kumar JV, Moss ME, Liu H, Fisher-Owens S. 2023. Association between low fluoride exposure
and children's intelligence: A meta-analysis relevant to community water fluoridation. Public
Health. 219:73-84. https://doi.org/10.1016/j.puhe.2023.03.011
Kumar V, Chahar P, Kajjari S, Rahman F, Bansal DK, Kapadia JM. 2018. Fluoride, thyroid
hormone derangements and its correlation with tooth eruption pattern among the pediatric
population from endemic and non-endemic fluorosis areas. J Contemp Dent Pract. 19(12):1512-
1516.
Li D, Zhao Q, Xie L, Wang C, Tian Z, Tang H, Xia T, Wang A. 2023. Fluoride impairs
mitochondrial translation by targeting miR-221-3p/c-Fos/RMND1 axis contributing to

109
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

neurodevelopment defects. Sci Total Environ. 869:161738.

https://doi.org/10.1016/j.scitotenv.2023.161738
Li F, Chen X, Huang R, Xie Y. 2009. The impact of endemic fluorosis caused by the burning of
coal on the development of intelligence in children. Journal of Environment and Health.
26(4):838-840.
Li J, Yao L, Shao QL, Wu CY. 2004. [Effects of high fluoride level on neonatal neurobehavioral
development]. Chin J Endemiol. 23(5):463-465. https://doi.org/10.3760/cma.j.issn.1000-
4955.2004.05.023
Li J, Yao L, Shao QL, Wu CY. 2008a. Effects of high fluoride level on neonatal neurobehavioral
development. Fluoride. 41(2):165-170.
Li M, Gao Y, Cui J, Li Y, Li B, Liu Y, Sun J, Liu X, Liu H, Zhao L, et al. 2016. Cognitive
impairment and risk factors in elderly people living in fluorosis areas in China. Biol Trace Elem
Res. 172(1):53-60. https://doi.org/10.1007/s12011-015-0568-0
Li X, Hou G, Yu B, Yuan C, Liu Y, Zhang L. 2010. [Investigation and analysis of children’s IQ
and dental fluorosis in high fluoride area]. Chin J Pest Control. 26(3):230-231.
Li Y, Jing X, Chen D, Lin L, Wang Z. 2003. [Effects of endemic fluoride poisoning on the
intellectual development of children in Baotou]. Chin J Public Health Manag. 19(4):337-338.
Li Y, Jing X, Chen D, Lin L, Wang Z. 2008c. Effects of endemic fluoride poisoning on the
intellectual development of children in Baotou. Fluoride. 41(2):161-164.
Li Y, Li X, Wei S. 1994. [Effects of high fluoride intake on child mental work capacity:
Preliminary investigation into the mechanisms involved]. J West Chin University Med Sci.
25(2):188-191.
Li Y, Li X, Wei S. 2008b. Effects of high fluoride intake on child mental work capacity:
Preliminary investigation into the mechanisms involved. Fluoride. 41(4):331-335.
Liang C, Ji R, Cao J, Jiang Y, Yu B, Ma F, Wu Y, Ying B, Zhang Y, Sun S, et al. 2003. [Study
on the relationship between drinking water trace elements and cognitive ability of the elderly]. J
Hyg Res. 32(5):436-440.
Lin FF, Ai HT, Zhao HX, Lin J, Jhiang JY, et al. 1991. [High fluoride and low iodine
environment and subclinical cretinism in Xinjiang]. Endem Dis Bull. 6(2):62-67.
Lin YY, Hsu WY, Yen CE, Hu SW. 2023. Association of dental fluorosis and urinary fluoride
with intelligence among schoolchildren. Children. 10(6):987.
https://doi.org/10.3390/children10060987
Liu YJ, Gao Q, Wu CX, Guan ZZ. 2010. Alterations of nAChRs and ERK1/2 in the brains of
rats with chronic fluorosis and their connections with the decreased capacity of learning and
memory. Toxicol Lett. 192(3):324-329. https://doi.org/10.1016/j.toxlet.2009.11.002

110
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Lou DD, Guan ZZ, Liu YJ, Liu YF, Zhang KL, Pan JG, Pei JJ. 2013. The influence of chronic
fluorosis on mitochondrial dynamics morphology and distribution in cortical neurons of the rat
brain. Arch Toxicol. 87(3):449-457. https://doi.org/10.1007/s00204-012-0942-z
Lu F, Zhang Y, Trivedi A, Jiang X, Chandra D, Zheng J, Nakano Y, Uyghurturk DA, Jalai R,
Onur SG, et al. 2019. Fluoride related changes in behavioral outcomes may relate to increased
serotonin. Physiol Behav. 206:76-83. https://doi.org/10.1016/j.physbeh.2019.02.017
Ma Q, Huang H, Sun L, Zhou T, Zhu J, Cheng X, Duan L, Li Z, Cui L, Ba Y. 2017. Gene-
environment interaction: Does fluoride influence the reproductive hormones in male farmers
modified by ERα gene polymorphisms? Chemosphere. 188:525-531.
https://doi.org/10.1016/j.chemosphere.2017.08.166
Malin AJ, Riddell J, McCague H, Till C. 2018. Fluoride exposure and thyroid function among
adults living in Canada: Effect modification by iodine status. Environ Int. 121(Pt 1):667-674.
https://doi.org/10.1016/j.envint.2018.09.026
Malin AJ, Till C. 2015. Exposure to fluoridated water and attention deficit hyperactivity disorder
prevalence among children and adolescents in the United States: An ecological association.
Environ Health. 14:17. https://doi.org/10.1186/s12940-015-0003-1
McDonell KE, van Wouwe NC, Harrison MB, Wylie SA, Claassen DO. 2018. Taq1A
polymorphism and medication effects on inhibitory action control in Parkinson disease. Brain
Behav. 8(7):e01008. https://doi.org/10.1002/brb3.1008
McPherson CA, Zhang G, Gilliam R, Brar SS, Wilson R, Brix A, Picut C, Harry GJ. 2018. An
evaluation of neurotoxicity following fluoride exposure from gestational through adult ages in
Long-Evans hooded rats. Neurotox Res. 34(4):781-798. https://doi.org/10.1007/s12640-018-
9870-x
Mesram N, Nagapuri K, Banala RR, Nalagoni CR, Karnati PR. 2016. Quercetin treatment
against NaF induced oxidative stress related neuronal and learning changes in developing rats. J
King Saud Univ Sci. 29(2):221-229. https://doi.org/10.1016/j.jksus.2016.04.002
Michael M, Barot VV, Chinoy NJ. 1996. Investigations of soft tissue functions in fluorotic
individuals of North Gujarat. Fluoride. 29(2):63-71.
Miller K, Howard BE, Phillips J, Shah M, Mav D, Thayer K, Shah R. 2016. SWIFT-Active
Screener: Reducing literature screening effort through machine learning for systematic reviews.
24th Cochrane Colloquium, Seoul, South Korea, October 23-27, 2016.
Miranda GHN, Alvarenga MOP, Ferreira MKM, Puty B, Bittencourt LO, Fagundes NCF, Pessan
JP, Buzalaf MAR, Lima RR. 2021. A systematic review and meta-analysis of the association
between fluoride exposure and neurological disorders. Sci Rep. 11(1):22659.
https://doi.org/10.1038/s41598-021-99688-w
Morgan L, Allred E, Tavares M, Bellinger D, Needleman H. 1998. Investigation of the possible
associations between fluorosis, fluoride exposure, and childhood behavior problems. Pediatr
Dent. 20(4):244-252.

111
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Mustafa DE, Younis UM, Elhag SAA. 2018. The relationship between the fluoride levels in
drinking water and the schooling performance of children in rural areas of Khartoum State,
Sudan. Fluoride. 51(2):102-113.
Nageshwar M, Sudhakar K, Reddy KP. 2018. Quercetin ameliorates oxidative stress, neural
damage of brain and behavioral impairment of rat with fluoride exposure. Int J Pharm Sci Res.
9(8):3247-3256. https://doi.org/10.13040/IJPSR.0975-8232.9(8).3247-56
National Academies of Sciences, Engineering, and Medicine (NASEM). 2020. Review of the
draft NTP monograph: Systematic review of fluoride exposure and neurodevelopmental and
cognitive health effects. Washington, DC: The National Academies Press.
https://doi.org/10.17226/25715
National Academies of Sciences, Engineering, and Medicine (NASEM). 2021. Review of the
revised NTP monograph on the systematic review of fluoride exposure and neurodevelopmental
and cognitive health effects: A letter report. Washington, DC: The National Academies Press.
https://doi.org/10.17226/26030
National Institute for Occupational Safety and Health (NIOSH). 1984. Fluoride in urine: Method
8308. In: NIOSH Manual of Analytical Methods, Volume 11. 3rd ed. Cincinnati, OH: U.S.
Department of Health and Human Services, Public Health Service, Centers for Disease Control
and Prevention, National Institute for Occupational Safety and Health.
National Research Council (NRC). 2006. Fluoride in drinking water: A scientific review of
EPA’s standards. Washington, DC: The National Academies Press.
https://doi.org/10.17226/11571
National Toxicology Program (NTP). 2016. NTP research report on systematic literature review
on the effects of fluoride on learning and memory in animal studies. Research Triangle Park,
NC: U.S. Department of Health and Human Services, Public Health Service, National
Toxicology Program. NTP Research Report 1. https://doi.org/10.22427/NTP-RR-1
National Toxicology Program (NTP). 2019. Health Assessment Workspace Collaborative
(HAWC): Fluoride (2019). Research Triangle Park, NC: U.S. Department of Health and Human
Services, National Institutes of Health, National Toxicology Program. [Accessed: 2019].
https://hawcproject.org/assessment/405/
National Toxicology Program (NTP). 2020. Revised draft NTP monograph on the systematic
review of fluoride exposure and neurodevelopmental and cognitive health effects. Research
Triangle Park, NC: U.S. Department of Health and Human Services, National Institutes of
Health, National Institute of Environmental Health Sciences, Division of the National
Toxicology Program, Office of Health Assessment and Translation. https://fluoridealert.org/wp-
content/uploads/ntp.revised-monograph.9-16-2020.pdf
National Toxicology Program (NTP). 2024. Data and protocol for systematic review of fluoride
exposure and neurodevelopmental and cognitive health effects. Research Triangle Park, NC:
U.S. Department of Health and Human Services, Public Health Service, National Toxicology
Program. https://doi.org/10.22427/NTP-DATA-MGRAPH-8

112
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Niu Q, Chen J, Xia T, Li P, Zhou G, Xu C, Zhao Q, Dong L, Zhang S, Wang A. 2018. Excessive
ER stress and the resulting autophagic flux dysfunction contribute to fluoride-induced
neurotoxicity. Environ Pollut. 233:889-899. https://doi.org/10.1016/j.envpol.2017.09.015
Nkpaa KW, Onyeso GI. 2018. Rutin attenuates neurobehavioral deficits, oxidative stress, neuro-
inflammation and apoptosis in fluoride treated rats. Neurosci Lett. 682:92-99.
https://doi.org/10.1016/j.neulet.2018.06.023
Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, Shamseer L,
Tetzlaff JM, Akl EA, Brennan SE, et al. 2021. The PRISMA 2020 statement: An updated
guideline for reporting systematic reviews. BMJ. 372:n71. https://doi.org/10.1136/bmj.n71
Peckham S, Lowery D, Spencer S. 2015. Are fluoride levels in drinking water associated with
hypothyroidism prevalence in England? A large observational study of GP practice data and
fluoride levels in drinking water. J Epidemiol Community Health. 69(7):619-624.
https://doi.org/10.1136/jech-2014-204971
Podgorski J, Berg M. 2020. Global threat of arsenic in groundwater. Science. 368(6493):845-
850. https://doi.org/10.1126/science.aba1510
Podgorski J, Berg M. 2022. Global analysis and prediction of fluoride in groundwater. Nat
Commun. 13(1):4232. https://doi.org/10.1038/s41467-022-31940-x
Pratap SV, Singh CD, Sandeep T, Sandeep K, Vikas G, Mukesh T, Anurag T. 2013. A
correlation between serum vitamin, acetylcholinesterase activity and IQ in children with
excessive endemic fluoride exposure in Rajasthan, India. Int Res J Medical Sci. 1(3):12-16.
Pulungan ZSA, Sofro ZM, Partadiredja G. 2016. Sodium fluoride does not affect the working
memory and number of pyramidal cells in rat medial prefrontal cortex. Anat Sci Int. 93(1):128-
138. https://doi.org/10.1007/s12565-016-0384-4
Reddy YP, Tiwari SK, Shaik AP, Alsaeed A, Sultana A, Reddy PK. 2014. Effect of sodium
fluoride on neuroimmunological parameters, oxidative stress and antioxidative defenses. Toxicol
Mech Methods. 24(1):31-36. https://doi.org/10.3109/15376516.2013.843224
Ren D, Li K, Liu D. 1989. [A study of the intellectual ability of 8-14 year-old children in high
fluoride, low iodine areas]. Chin J Control Endem Dis. 4(4):251.
Ren D, Li K, Liu D. 2008. A study of the intellectual ability of 8-14 year-old children in high
fluoride, low iodine areas. Fluoride. 41(4):319-320.
Riddell JK, Malin AJ, Flora D, McCague H, Till C. 2019. Association of water fluoride and
urinary fluoride concentrations with attention deficit hyperactivity disorder in Canadian youth.
Environ Int. 133(Pt B):105190. https://doi.org/10.1016/j.envint.2019.105190
Rocha-Amador D, Navarro M, Trejo-Acevedo A, Carrizales L, Pérez-Maldonado I, Díaz-Barriga
F, Calderón J. 2009. Use of the Rey-Osterrieth Complex Figure Test for neurotoxicity evaluation
of mixtures in children. Neurotoxicology. 30(6):1149-1154.
https://doi.org/10.1016/j.neuro.2009.09.003

113
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Rocha-Amador D, Navarro ME, Carrizales L, Morales R, Calderón J. 2007. Decreased

intelligence in children and exposure to fluoride and arsenic in drinking water. Cad Saude
Publica. 23 Suppl 4:S579-S587. https://doi.org/10.1590/s0102-311x2007001600018
Rooney AA, Boyles AL, Wolfe MS, Bucher JR, Thayer KA. 2014. Systematic review and
evidence integration for literature-based environmental health science assessments. Environ
Health Perspect. 122(7):711-718. https://doi.org/10.1289/ehp.1307972
Rotton J, Tikofsky RS, Feldman HT. 1982. Behavioral effects of chemicals in drinking water. J
Appl Psychol. 67(2):230-238. https://doi.org/10.1037/0021-9010.67.2.230
Russ TC, Killin LOJ, Hannah J, Batty GD, Deary IJ, Starr JM. 2019. Aluminium and fluoride in
drinking water in relation to later dementia risk. Br J Psychiatry. 216(1):29-34.
https://doi.org/10.1192/bjp.2018.287
Santa-Marina L, Jimenez-Zabala A, Molinuevo A, Lopez-Espinosa M, Villanueva C, Riano I,
Ballester F, Sunyer J, Tardon A, Ibarluzea J. 2019. Fluorinated water consumption in pregnancy
and neuropsychological development of children at 14 months and 4 years of age. Environ
Epidemiol. 3:386-387. https://doi.org/10.1097/01.EE9.0000610304.33479.18
Savitz DA, Wellenius GA, Trikalinos TA. 2019. The problem with mechanistic risk of bias
assessments in evidence synthesis of observational studies and a practical alternative: Assessing
the impact of specific sources of potential bias. Am J Epidemiol. 188(9):1581-1585.
https://doi.org/10.1093/aje/kwz131
Saxena S, Sahay A, Goel P. 2012. Effect of fluoride exposure on the intelligence of school
children in Madhya Pradesh, India. J Neurosci Rural Pract. 3(2):144-149.
https://doi.org/10.4103/0976-3147.98213
Scientific Committee on Health and Environmental Risks (SCHER). 2011. Critical review of any
new evidence on the hazard profile, health effects, and human exposure to fluoride and the
fluoridating agents of drinking water. Brussels, Belgium: European Commission, Directorate-
General for Health and Consumers.
http://ec.europa.eu/health/scientific_committees/environmental_risks/docs/scher_o_139.pdf
Sebastian ST, Sunitha S. 2015. A cross-sectional study to assess the intelligence quotient (IQ) of
school going children aged 10-12 years in villages of Mysore District, India with different
fluoride levels. J Indian Soc Pedod Prev Dent. 33(4):307-311. https://doi.org/10.4103/0970-
4388.165682
Seraj B, Shahrabi M, Shadfar M, Ahmadi R, Fallahzadeh M, Eslamlu HF, Kharazifard MJ. 2012.
Effect of high water fluoride concentration on the intellectual development of children in
Makoo/Iran. J Dent (Tehran). 9(3):221-229.
Shan KR, Qi XL, Long YG, Nordberg A, Guan ZZ. 2004. Decreased nicotinic receptors in PC12
cells and rat brains influenced by fluoride toxicity—a mechanism relating to a damage at the
level in post-transcription of the receptor genes. Toxicology. 200(2-3):169-177.
https://doi.org/10.1016/j.tox.2004.03.013

114
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Shannon FT, Fergusson DM, Horwood LJ. 1986. Exposure to fluoridated public water supplies
and child health and behaviour. N Z Med J. 99(803):416-418.
Shao Q. 2003. [Study of cognitive function impairment caused by chronic fluorosis]. Chin J
Endemiol. 22(4):336-338. https://doi.org/10.3760/cma.j.issn.1000-4955.2003.04.016
Sharma JD, Sohu D, Jain P. 2009. Prevalence of neurological manifestations in a human
population exposed to fluoride in drinking water. Fluoride. 42(2):127-132.
Singh N, Verma KG, Verma P, Sidhu GK, Sachdeva S. 2014. A comparative study of fluoride
ingestion levels, serum thyroid hormone & TSH level derangements, dental fluorosis status
among school children from endemic and non-endemic fluorosis areas. Springerplus. 3:7.
https://doi.org/10.1186/2193-1801-3-7
Singla VI, Sutton PM, Woodruff TJ. 2019. The Environmental Protection Agency Toxic
Substances Control Act systematic review method may curtail science used to inform policies,
with profound implications for public health. Am J Public Health. 109(7):982-984.
https://doi.org/10.2105/ajph.2019.305068
Soto-Barreras U, Escalante-Villalobos KY, Holguín-Loya B, Perez-Aguirre B, Nevárez-Rascón
A, Martínez-Martínez RE, Loyola-Rodriguez JP. 2019. Effect of fluoride in drinking water on
dental caries and IQ in children. Fluoride. 52(3 Pt 3):474-482.
Stang A. 2010. Critical evaluation of the Newcastle-Ottawa scale for the assessment of the
quality of nonrandomized studies in meta-analyses. Eur J Epidemiol. 25(9):603-605.
https://doi.org/10.1007/s10654-010-9491-z
Steenland K, Schubauer-Berigan MK, Vermeulen R, Lunn RM, Straif K, Zahm S, Stewart P,
Arroyave WD, Mehta SS, Pearce N. 2020. Risk of bias assessments and evidence syntheses for
observational epidemiologic studies of environmental and occupational exposures: Strengths and
limitations. Environ Health Perspect. 128(9):095002. https://doi.org/10.1289/EHP6980
Sterne JAC, Hernán MA, McAleenan A, Reeves BC, Higgins JPT. 2023. Chapter 25: Assessing
risk of bias in a non-randomized study. In: Higgins JPT, Thomas J, Chandler J, Cumpston M, Li
T, Page MJ, Welch VA, editors. Cochrane Handbook for Systematic Reviews of Interventions
version 6.4 (updated August 2023). London, UK: The Cochrane Collaboration.
https://training.cochrane.org/handbook
Still CN, Kelley P. 1980. On the incidence of primary degenerative dementia vs. water fluoride
content in South Carolina. Neurotoxicology. 1(4):125-131.
Sudhakar K, Reddy KP. 2018. Protective effects of Abelmoschus moschatus seed extract on
neurotransmitter system of developing brain of Wistar rats with gestational and post-natal
exposure of sodium fluoride. Int J Green Pharm. 12(1 Suppl):S131-S139.
https://doi.org/10.22377/ijgp.v12i01.1609
Sudhir KM, Chandu GN, Prashant GM, Subba Reddy VV. 2009. Effect of fluoride exposure on
intelligence quotient (IQ) among 13-15 year old school children of known endemic area of
fluorosis, Nalgonda District, Andhra Pradesh. J Indian Assoc Public Health Dent. 2009(13):88-
94.

115
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Sun MM, Li SK, Wang YF, Li FS. 1991. [Measurement of intelligence by drawing test among
the children in the endemic area of Al-F combined toxicosis]. J Guiyang Med Coll. 16(3):204-
206.
Sun ZR, Liu FZ, Wu LN, Lu Y, Yu DK. 2000. [Effects of high fluoride drinking water on the
cerebral function of mice]. Chin J Endemiol. 19(4):262-263.
https://doi.org/10.3760/cma.j.issn.1000-4955.2000.04.009
Sun ZR, Liu FZ, Wu LN, Lu Y, Yu DK. 2008. Effects of high fluoride drinking water on the
cerebral function of mice. Fluoride. 41(2):148-151.
Susheela AK, Bhatnagar M, Vig K, Mondal NK. 2005. Excess fluoride ingestion and thyroid
hormone derangements in children living in Delhi, India. Fluoride. 38(2):98-108.
Tang QQ, Du J, Ma HH, Jiang SJ, Zhou XJ. 2008. Fluoride and children's intelligence: A meta-
analysis. Biol Trace Elem Res. 126(1-3):115-120. https://doi.org/10.1007/s12011-008-8204-x
Till C, Green R, Flora D, Hornung R, Martinez-Mier EA, Blazer M, Farmus L, Ayotte P, Muckle
G, Lanphear B. 2020. Fluoride exposure from infant formula and child IQ in a Canadian birth
cohort. Environ Int. 134:105315. https://doi.org/10.1016/j.envint.2019.105315
Till C, Green R, Grundy JG, Hornung R, Neufeld R, Martinez-Mier EA, Ayotte P, Muckle G,
Lanphear B. 2018. Community water fluoridation and urinary fluoride concentrations in a
national sample of pregnant women in Canada. Environ Health Perspect. 126(10):107001.
https://doi.org/10.1289/ehp3546
Trivedi MH, Sangai NP, Patel RS, Payak M, Vyas SJ. 2012. Assessment of groundwater quality
with special reference to fluoride and its impact on IQ of schoolchildren in six villages of the
Mundra Region, Kachchh, Gujurat, India. Fluoride. 45(4):377-383.
Trivedi MH, Verma RJ, Chinoy NJ, Patel RS, Sathawara NG. 2007. Effect of high fluoride water
on intelligence of school children in India. Fluoride. 40(3):178-183.
Tsunoda M, Aizawa Y, Nakano K, Liu Y, Horiuchi T, Itai K, Tsunoda H. 2005. Changes in
fluoride levels in the liver, kidney, and brain and in neurotransmitters of mice after subacute
administration of fluoride. Fluoride. 38(4):284-292.
U.S. Department of Health and Human Services Federal Panel on Community Water
Fluoridation (USDHHS). 2015. U.S. Public Health Service recommendation for fluoride
concentration in drinking water for the prevention of dental caries. Public Health Rep.
130(4):318-331. https://doi.org/10.1177/003335491513000408
U.S. Environmental Protection Agency (USEPA). 2010. Fluoride: Exposure and relative source
contribution analysis. Washington, DC: U.S. Environmental Protection Agency. EPA Report No.
820-R-10-015. http://www.epa.gov/dwstandardsregulations/fluoride-risk-assessment-and-
relative-source-contribution
U.S. Geological Survey (USGS). 2020. Fluoride in groundwater: Too much or too little of a
good thing? Lakewood, CO: U.S. Geological Survey, Water Resources Mission Area, Colorado
Water Science Center. https://www.usgs.gov/news/comprehensive-assessment-fluoride-
groundwater

116
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Valdez Jiménez L, López Guzmán OD, Cervantes Flores M, Costilla-Salazar R, Calderón

Hernández J, Alcaraz Contreras Y, Rocha-Amador DO. 2017. In utero exposure to fluoride and
cognitive development delay in infants. Neurotoxicology. 59:65-70.
https://doi.org/10.1016/j.neuro.2016.12.011
Veneri F, Vinceti M, Generali L, Giannone ME, Mazzoleni E, Birnbaum LS, Consolo U,
Filippini T. 2023. Fluoride exposure and cognitive neurodevelopment: Systematic review and
dose-response meta-analysis. Environ Res. 221:115239.
https://doi.org/10.1016/j.envres.2023.115239
Villa A, Anabalon M, Zohouri V, Maguire A, Franco AM, Rugg-Gunn A. 2010. Relationships
between fluoride intake, urinary fluoride excretion and fluoride retention in children and adults:
An analysis of available data. Caries Res. 44(1):60-68. https://doi.org/10.1159/000279325
Walker VR, Lemeris CR, Magnuson K, Sibrizzi CA, Shipkowski KA, Taylor KW, Rooney AA.
2024. I-REFF diagrams: Enhancing transparency in systematic review through interactive
reference flow diagrams. Syst Rev. 13(1):33. https://doi.org/10.1186/s13643-023-02420-0
Wang A, Duan L, Huang H, Ma J, Zhang Y, Ma Q, Guo Y, Li Z, Cheng X, Zhu J, et al. 2020a.
Association between fluoride exposure and behavioural outcomes of school-age children: A pilot
study in China. Int J Environ Health Res. 32(1):232-241.
https://doi.org/10.1080/09603123.2020.1747601
Wang G, Gao M, Zhang M, Yang M, Xiang Q. 2012. Correlation between total fluoride intake
and children's IQ. J Southeast Univ Med Ed. 31(6):743-746.
Wang M, Liu L, Li H, Li Y, Liu H, Hou C, Zeng Q, Li P, Zhao Q, Dong L, et al. 2020b. Thyroid
function, intelligence, and low-moderate fluoride exposure among Chinese school-age children.
Environ Int. 134:105229. https://doi.org/10.1016/j.envint.2019.105229
Wang S, Zhao Q, Li G, Wang M, Liu H, Yu X, Chen J, Li P, Dong L, Zhou G, et al. 2021c. The
cholinergic system, intelligence, and dental fluorosis in school-aged children with low-to-
moderate fluoride exposure. Ecotoxicol Environ Saf. 228:112959.
https://doi.org/10.1016/j.ecoenv.2021.112959
Wang X, Wang L, Hu P, Guo X, Luo X. 2001. [Effects of high iodine and high fluorine on
children’s intelligence and thyroid function]. Chin J Endemiol. 20(4):288-290.
https://doi.org/10.3760/cma.j.issn.1000-4955.2001.04.017
Watanabe M, Kono K, Orita Y, Dote T, Usuda K, Takahashi Y, Yoshida Y. 1995. Influence of
dietary fluoride intake on urinary fluoride concentration and evaluation of corrected levels in
spot urine. Fluoride. 28(2):61-70.
Waugh DT. 2019. Fluoride exposure induces inhibition of sodium/iodide symporter (NIS)
contributing to impaired iodine absorption and iodine deficiency: Molecular mechanisms of
inhibition and implications for public health. Int J Environ Res Public Health. 16(6):1086.
https://doi.org/10.3390/ijerph16061086
World Health Organization (WHO). 2008. Guidelines for drinking-water quality: Incorporating
the first and second addenda, volume 1: Recommendations. 3rd ed. Geneva, Switzerland: World

117
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Health Organization.
https://apps.who.int/iris/bitstream/handle/10665/204411/9789241547611_eng.pdf?sequence=1&i
sAllowed=y
World Health Organization (WHO). 2017. Guidelines for drinking-water quality: Fourth edition
incorporating the first addendum. Geneva, Switzerland: World Health Organization.
https://apps.who.int/iris/handle/10665/254637
Xia Y, Xu Y, Shi M, Liu S, Liu S, Wang H, Dai C, Ye Y, Liu M, Shang L, et al. 2023. Effects of
high-water fluoride exposure on IQ levels in school-age children: A cross-sectional study in
Jiangsu, China. Expo Health. 16(3):885-895. https://doi.org/10.1007/s12403-023-00597-2
Xiang Q, Liang Y, Chen B, Chen L. 2011. Analysis of children's serum fluoride levels in relation
to intelligence scores in a high and low fluoride water village in China. Fluoride. 44(4):191-194.
Xiang Q, Liang Y, Chen L, Wang C, Chen B, Chen X, Zhou M. 2003a. Effect of fluoride in
drinking water on children's intelligence. Fluoride. 36(2):84-94.
Xiang Q, Wang Y, Yang M, Zhang M, Xu Y. 2013. Level of fluoride and arsenic in household
shallow well water in Wamiao and Xinhuai villages in Jiangsu Province, China. Fluoride.
46(4):192-197.
Xiang QY, Liang YX, Zhou MS, Zang HB. 2003b. Blood lead of children in Wamiao-Xinhuai
intelligence study. Fluoride. 36(3):198-199.
Xu Y, Lu C, Zhang X. 1994. [The effect of fluorine on the level of intelligence in children].
Endem Dis Bull. 9(2):83-84.
Yang Y, Wang X, Guo X, Hu P. 2008. The effects of high levels of fluoride and iodine on child
intellectual ability and the metabolism of fluoride and iodine. Fluoride. 41(4):336-339.
Yang Y, Wang X, X. G, Hu P. 1994. [The effects of high levels of fluoride and iodine on child
intellectual ability and the metabolism of fluoride and iodine]. Chin J Epidemiol. 15(4):296-298.
Yao L, Zhou J, Wang X, Cui Q, Lin F. 1996. Analysis on the correlation between TSH and
intelligence in children with dental fluorosis from endemic fluorosis regions. Lit Inf Prev Med.
2(1):26-27.
Yasmin S, Ranjan S, D'Souza D. 2013. Effect of excess fluoride ingestion on human thyroid
function in Gaya Region, Bihar, India. Toxicol Environ Chem. 95(7):1235-1243.
https://doi.org/10.1080/02772248.2013.847619
Yazdi SM, Sharifian A, Dehghani-Beshne M, Momeni VR, Aminian O. 2011. Effects of fluoride
on psychomotor performance and memory of aluminum potroom workers. Fluoride. 44(3):158-
162.
Yu X, Chen J, Li Y, Liu H, Hou C, Zeng Q, Cui Y, Zhao L, Li P, Zhou Z, et al. 2018. Threshold
effects of moderately excessive fluoride exposure on children's health: A potential association
between dental fluorosis and loss of excellent intelligence. Environ Int. 118:116-124.
https://doi.org/10.1016/j.envint.2018.05.042

118
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Yu X, Xia L, Zhang S, Zhou G, Li Y, Liu H, Hou C, Zhao Q, Dong L, Cui Y, et al. 2021.
Fluoride exposure and children’s intelligence: Gene-environment interaction based on SNP-set,
gene and pathway analysis, using a case-control design based on a cross-sectional study. Environ
Int. 155:106681. https://doi.org/10.1016/j.envint.2021.106681
Yu YN, Yang WX, Dong Z, Wan CW, Zhang JT, Liu JL, Xiao KQ, Huang YS, Lu BF. 1996.
[Neurotransmitter and receptor changes in the brains of fetuses from areas of endemic fluorosis].
Chin J Endemiol. 15(5):257-259.
Yu YN, Yang WX, Dong Z, Wan CW, Zhang JT, Liu JL, Xiao KQ, Huang YS, Lu BF. 2008.
Neurotransmitter and receptor changes in the brains of fetuses from areas of endemic fluorosis.
Fluoride. 41(2):134-138.
Zhang KL, Lou DD, Guan ZZ. 2015. Activation of the AGE/RAGE system in the brains of rats
and in SH-SY5Y cells exposed to high level of fluoride might connect to oxidative stress.
Neurotoxicol Teratol. 48:49-55. https://doi.org/10.1016/j.ntt.2015.01.007
Zhang S, Zhang X, Liu H, Qu W, Guan Z, Zeng Q, Jiang C, Gao H, Zhang C, Lei R, et al.
2015b. Modifying effect of COMT gene polymorphism and a predictive role for proteomics
analysis in children's intelligence in endemic fluorosis area in Tianjin, China. Toxicol Sci.
144(2):238-245. https://doi.org/10.1093/toxsci/kfu311
Zhao L, Yu C, Lv J, Cui Y, Wang Y, Hou C, Yu J, Guo B, Liu H, Li L. 2021. Fluoride exposure,
dopamine relative gene polymorphism and intelligence: A cross-sectional study in China.
Ecotoxicol Environ Saf. 209:111826. https://doi.org/10.1016/j.ecoenv.2020.111826
Zhao Q, Niu Q, Chen J, Xia T, Zhou G, Li P, Dong L, Xu C, Tian Z, Luo C, et al. 2019. Roles of
mitochondrial fission inhibition in developmental fluoride neurotoxicity: Mechanisms of action
in vitro and associations with cognition in rats and children. Arch Toxicol. 93(3):709-726.
https://doi.org/10.1007/s00204-019-02390-0
Zhao Q, Tian Z, Zhou G, Niu Q, Chen J, Li P, Dong L, Xia T, Zhang S, Wang A. 2020. SIRT1-
dependent mitochondrial biogenesis supports therapeutic effects of resveratrol against
neurodevelopment damage by fluoride. Theranostics. 10(11):4822-4838.
https://doi.org/10.7150/thno.42387
Zhou G, Tang S, Yang L, Niu Q, Chen J, Xia T, Wang S, Wang M, Zhao Q, Liu L, et al. 2019.
Effects of long-term fluoride exposure on cognitive ability and the underlying mechanisms: Role
of autophagy and its association with apoptosis. Toxicol Appl Pharmacol. 378:114608.
https://doi.org/10.1016/j.taap.2019.114608
Zhou G, Zhao Q, Luo C, Liu H, Li P, Cui Y, Yu X, Chen J, Liu L, Zhang S, et al. 2021. Low-
moderate fluoride exposure and intelligence among Chinese school-aged children: Role of
circulating mtDNA content. Sci Total Environ. 786:147330.
https://doi.org/10.1016/j.scitotenv.2021.147330
Zhou T, Duan L, Ding Z, Yang R, Li S, Xi Y, Cheng X, Hou J, Wen S, Chen J, et al. 2012.
Environmental fluoride exposure and reproductive hormones in male living in endemic fluorosis
villages in China. Life Sci J. 9(4):1-7.

119
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Zohouri FV, Swinbank CM, Maguire A, Moynihan PJ. 2006. Is the fluoride/creatinine ratio of a
spot urine sample indicative of 24-h urinary fluoride? Community Dent Oral Epidemiol.
34(2):130-138. https://doi.org/10.1111/j.1600-0528.2006.00269.x

120
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Appendix A. Data Figures: Neurodevelopmental or Cognitive

Effects and Outcomes

Figures
Figure A-1. Distribution of IQ in Children by Fluoride Exposure (Low Risk-of-bias Studies;
Presented as % in Area or % of Total Group) ........................................................ A-2
Figure A-2. Mean IQ in Children by Fluoride Exposure (Low Risk-of-bias Studies) ............... A-3
Figure A-3. Intelligence Grade in Children by Fluoride Exposure (Low Risk-of-bias Studies;
Presented as Mean) ................................................................................................. A-4
Figure A-4. Mean Change in IQ in Children by Fluoride Exposure (Low Risk-of-bias
Studies) ................................................................................................................... A-4
Figure A-5. Associations between Fluoride Exposure and IQ Scores in Children (Low Risk-of-
bias Studies; Presented as Adjusted OR) ................................................................ A-5
Figure A-6. Correlations between Fluoride Exposure and IQ Score in Children (Low Risk-of-
bias Studies; Presented as Correlation Coefficient) ................................................ A-6
Figure A-7. Associations between Fluoride Exposure and IQ Score in Children (Low Risk-of-
bias Studies; Presented as Adjusted Beta)—China ................................................. A-7
Figure A-8. Associations between Fluoride Exposure and IQ Score in Children (Low Risk-of-
bias Studies; Presented as Adjusted Beta)—Areas Other Than China ................... A-8
Figure A-9. Mean Motor/Sensory Scores in Children by Fluoride Exposure (Low Risk-of-bias
Studies) ................................................................................................................... A-9
Figure A-10. Correlations between Fluoride Exposure and Other Cognitive Effects in Children
(Low Risk-of-bias Studies; Presented as Correlation Coefficient) ....................... A-9
Figure A-11. Associations between Fluoride Exposure and Other Neurodevelopmental Effects in
Children (Low Risk-of-bias Studies; Presented as Adjusted Beta)..................... A-10
Figure A-12. Associations between Fluoride Exposure and Other Neurodevelopmental Effects in
Children (Low Risk-of-bias Studies; Presented as Adjusted OR) ...................... A-11
Figure A-13. Cognitive Impairment in Adults by Fluoride Exposure (Low Risk-of-bias Studies;
Presented as % of Total Group) .......................................................................... A-11

A-1
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure A-1. Distribution of IQ in Children by Fluoride Exposure (Low Risk-of-bias Studies;

Presented as % in Area or % of Total Group)

Reference group indicated by blue bars; other bars represent response estimates with red indicating statistical significance
compared with the reference group.
An interactive version of Figure A-1 and additional study details in HAWC here (NTP 2019). “F” represents fluoride. For IQ
distribution results by drinking water fluoride level provided in Xiang et al. (2003a), Trivedi et al. (2012), Sudhir et al. (2009),
and Seraj et al. (2012), and rate of low IQ scores by fluoride intake provided in Wang et al. (2012), statistical significance was
not evaluated.

A-2
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure A-2. Mean IQ in Children by Fluoride Exposure (Low Risk-of-bias Studies)

Reference group indicated by blue triangles; circles represent response estimates with red indicating statistical significance.
An interactive version of Figure A-2 and additional study details in HAWC here (NTP 2019) “F” represents fluoride. Three
additional publications based on subsample of the larger Yu et al. (2018) cohort were identified (Zhao et al. 2019; Zhao et al.
2020; Zhou et al. 2019); however, results from these studies are not presented here. The main study by Yu et al. (2018) is
considered a better representation of the IQ results. For all studies, SDs are available and can be viewed in HAWC (NTP 2019)
by clicking the data points within the plot area; however, 95% CIs could not be calculated for Seraj et al. (2012) because Ns are
not available for exposure groups.

A-3
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure A-3. Intelligence Grade in Children by Fluoride Exposure (Low Risk-of-bias Studies;
Presented as Mean)

Reference group indicated by blue triangles; circles represent response estimates with red indicating statistical significance.
An interactive version of Figure A-3 and additional study details in HAWC here (NTP 2019). For Saxena et al. (2012), children’s
intelligence was measured using Raven’s Standard Progressive Matrices. Children’s scores were converted to percentile, and
specific grades were allotted based on the percentiles. Grades ranged from intellectually superior (Grade I) to intellectually
impaired (Grade V). Results for Soto-Barreras et al. (2019) are not presented here. Outcomes in the study were presented as
levels of fluoride exposure associated with each intelligence grade. Results reported were not significant.

Figure A-4. Mean Change in IQ in Children by Fluoride Exposure (Low Risk-of-bias Studies)

Reference group indicated by blue triangles; circles represent response estimates with red indicating statistical significance.
An interactive version of Figure A-4 and additional study details in HAWC here (NTP 2019). For Ding et al. (2011), SDs are
available and can be viewed in HAWC (NTP 2019) by clicking the data points within the plot area; however, 95% CIs could not
be calculated because Ns for each exposure group are not available.

A-4
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure A-5. Associations between Fluoride Exposure and IQ Scores in Children (Low Risk-of-bias
Studies; Presented as Adjusted OR)

Reference group indicated by blue triangles; circles represent response estimates with red indicating statistical significance.
Cutoffs for the dichotomous outcome are listed in the Endpoint column.
An interactive version of Figure A-5 and additional study details in HAWC here (NTP 2019). For Xiang et al. (2011), there was a
significant linear trend across different levels of serum fluoride for IQ score <80 (p < 0.001). For Yu et al. (2018), significance
levels by IQ score were not reported.

A-5
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure A-6. Correlations between Fluoride Exposure and IQ Score in Children (Low Risk-of-bias
Studies; Presented as Correlation Coefficient)

Circles represent response estimates with red indicating statistical significance.

An interactive version of Figure A-6 and additional study details in HAWC here (NTP 2019). “F” represents fluoride. For Saxena
et al. (2012), a significant relationship between water fluoride level and intelligence grade was observed. Increasing intelligence
grades reflected increasing levels of impairment (reduced intelligence) in children.

A-6
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure A-7. Associations between Fluoride Exposure and IQ Score in Children (Low Risk-of-bias
Studies; Presented as Adjusted Beta)—China

Circles represent response estimates with red indicating statistical significance.

An interactive version of Figure A-7 and additional study details in HAWC here (NTP 2019). “F” represents fluoride. For Yu et
al. (2018), authors note an obvious decrease in the IQ score at water fluoride exposure levels between 3.40 mg/L and 3.90 mg/L
and a similar adverse effect on IQ scores at urinary fluoride exposure levels from 1.60 mg/L to 2.50 mg/L, and so the changes in
IQ score are indicated as significant; however, significance levels for change in IQ score were not reported.

A-7
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure A-8. Associations between Fluoride Exposure and IQ Score in Children (Low Risk-of-bias
Studies; Presented as Adjusted Beta)—Areas Other Than China

Circles represent response estimates with red indicating statistical significance.

An interactive version of Figure A-8 and additional study details in HAWC here (NTP 2019). “F” represents fluoride.

A-8
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure A-9. Mean Motor/Sensory Scores in Children by Fluoride Exposure (Low Risk-of-bias
Studies)

Reference group indicated by blue triangles; circles represent response estimates with red indicating statistical significance.
An interactive version of Figure A-9 and additional study details in HAWC here (NTP 2019). “F” represents fluoride. 95% CIs
are small and are within figure symbols and may be difficult to see. Values for SDs and 95% CIs can be viewed in HAWC (NTP
2019) by clicking the data points within the plot area. Total neonatal behavioral neurological assessment (NBNA) score was also
significantly reduced in the endemic F region versus reference region (not shown).

Figure A-10. Correlations between Fluoride Exposure and Other Cognitive Effects in Children
(Low Risk-of-bias Studies; Presented as Correlation Coefficient)

Circles represent response estimates with red indicating statistical significance.

An interactive version of Figure A-10 and additional study details in HAWC here (NTP 2019). “F” represents fluoride.

A-9
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure A-11. Associations between Fluoride Exposure and Other Neurodevelopmental Effects in
Children (Low Risk-of-bias Studies; Presented as Adjusted Beta)

Reference group indicated by blue triangles; circles represent response estimates with red indicating statistical significance.
An interactive version of Figure A-11 and additional study details in HAWC here (NTP 2019). “F” represents fluoride. Bashash
et al. (2018) observed significant associations between maternal urinary fluoride and ADHD-like symptoms related to inattention
(an increase in 0.5 mg/L of maternal urinary fluoride was associated with a 2.84-point increase in the DSM-IV Inattention Index
and a 2.54-point increase in Cognitive Problems and Inattention Index). These two scales contributed to the global ADHD Index
and the DSM-IV ADHD Total Index shown here.

A-10
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure A-12. Associations between Fluoride Exposure and Other Neurodevelopmental Effects in
Children (Low Risk-of-bias Studies; Presented as Adjusted OR)

Circles represent response estimates with red indicating statistical significance.

An interactive version of Figure A-12 and additional study details in HAWC here (NTP 2019). “F” represents fluoride. Drinking
water results for Barberio et al. (2017b) have a large confidence interval and are not completely visible in the figure. 95% CIs are
0.068–11.33 and can be viewed in HAWC (NTP 2019) by clicking the OR within the plot area.

Figure A-13. Cognitive Impairment in Adults by Fluoride Exposure (Low Risk-of-bias Studies;
Presented as % of Total Group)

Reference group indicated by blue triangles; circles represent response estimates with red indicating statistical significance.
An interactive version of Figure A-13 and additional study details in HAWC here (NTP 2019). Results from Li et al. (2016)
suggested that fluoride exposure may be a risk factor for cognitive impairment in elderly subjects; however, results from the
study were not conducive to presentation in this visualization.

A-11
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Appendix B. Literature Search and Document Review Details

Table of Contents
B.1. Introduction ..........................................................................................................................B-2

Tables
Table B-1. Literature Search and Document Review Timeline ...................................................B-2
Table B-2. PubMed Search Terms ...............................................................................................B-4

B-1
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

B.1. Introduction

NTP initially published a systematic review of the experimental animal literature in 2016 that
was subsequently expanded to include human epidemiological studies, mechanistic studies, and
newer experimental animal literature. Table B-1 provides a timeline of key activities contributing
to the 2024 NTP monograph including the multiple literature searches, draft monographs, and
document review activities that have occurred since 2016.
Table B-2 is a summary of the specific search terms used for the PubMed database. In order to
ensure inclusion of relevant papers, the strategy for this search was broad for the consideration of
neurodevelopmental or cognitive endpoints and comprehensive for fluoride as an exposure or
treatment. The specific search strategies for other databases are available in the protocol
(https://ntp.niehs.nih.gov/go/785076).

Table B-1. Literature Search and Document Review Timeline

Date Action
July 2016 Published 2016 NTP monograph of the systematic literature review on the
effects of fluoride on learning and memory in animals only
June 2017 Published protocol for a new NTP monograph on systematic review on effects
of fluoride on neurodevelopment and cognition from evidence in human,
experimental animal, and mechanistic data
April 2019 Completed final literature search for 2019 draft NTP monograph on human,
experimental animal, and mechanistic data (i.e., updated through April 2019)
May 2019 Published 2019 revised protocol for 2019 draft NTP monograph
September 2019 Sent 2019 draft NTP monograph for review by NASEM committee
February 2020 Received NASEM committee’s review report of 2019 draft NTP monograph;
began the following key changes in response to NASEM report:
• Expanded literature search to non-English-language databases
• Conducted meta-analysis on children’s IQ and fluoride exposure
• Revised protocol for monograph to include additional information.
May 2020 Completed final literature search for 2020 draft NTP monograph on human
experimental animal and mechanistic data (i.e., updated through May 2020
and expanded to include non-English-language databases)
September 2020 Published 2020 revised protocol for 2020 draft NTP monograph
September 2020 Sent 2020 draft NTP monograph for second review by NASEM committee
February 2021 Received NASEM committee’s review report of revised 2020 draft NTP
monograph; made the following key changes in response to NASEM report:
• Removed hazard step and hazard conclusions
• Removed meta-analysis to publish separately.
December 2021 Sent 2021 draft NTP monograph on the state of the science for external peer
review
May 2022 Sent May 2022 NTP Monograph for interagency review in preparation for
BSC Review performed by BSC Working Group (WG).

B-2
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Date Action
July 2022 Received final set of interagency review comments. Comments received from:
• Centers for Disease Control and Prevention (CDC)
• U.S. Food and Drug Administration (FDA)
• Eunice Kennedy Shriver National Institute of Child Health and
Human Development (NICHD)
• National Institute of Dental and Craniofacial Research (NIDCR)
• Office of the Director, National Institutes of Health (NIH OD)
• Office of the Assistant Secretary for Health (OASH)
September 2022 Sent the September 2022 NTP Monograph and responses to comments from
NASEM review, external peer review, and interagency review to BSC WG for
review of the adequacy of the NTP authors’ responses to comments.
May 2023 Received BSC WG report on the September 2022 NTP Monograph; made the
following key changes in response to the BSC WG report:
• Added an Addendum to update the literature assessing fluoride
exposures and children’s IQ to match the timeframe for literature
included in the meta-analysis for separate publication.

B-3
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Table B-2. PubMed Search Terms

Database Search Terms
PUBMED ((Fluorides[mh:noexp] OR fluorides, topical[mh] OR sodium fluoride[mh] OR Fluorosis, Dental[mh]
OR fluorosis[tiab] OR fluorid*[tiab] OR flurid*[tiab] OR fluorin*[tiab] OR florin*[tiab]) NOT
(18F[tiab] OR f-18[tiab] OR 19F[tiab] OR f-19[tiab] OR f-labeled[tiab] OR "fluorine-18"[tiab] OR
"fluorine-19"[tiab] OR pet-scan[tiab] OR radioligand*[tiab]))
AND ((Aryl Hydrocarbon Hydroxylases[mh] OR Aryl Hydrocarbon Receptor Nuclear
Translocator[mh] OR Behavior and Behavior Mechanisms[mh] OR Gene Expression Regulation[mh]
OR Glucuronosyltransferase[mh] OR Intelligence tests[mh] OR Malate Dehydrogenase[mh] OR
Mediator Complex Subunit 1[mh] OR Mental disorders[mh] OR Mental processes[mh] OR
Monocarboxylic Acid Transporters[mh] OR Myelin Basic Protein[mh] OR nervous system[mh] OR
nervous system diseases[mh] OR nervous system physiological phenomena[mh] OR Neurogranin[mh]
OR Oligodendroglia[mh] OR Peroxisome Proliferator-Activated Receptors[mh] OR Psychological
Phenomena and Processes[mh] OR Receptors, thyroid hormone[mh] OR Receptors, thyrotropin[mh]
OR Retinoid X Receptors[mh] OR thyroid diseases[mh] OR thyroid hormones[mh] OR Thyrotropin-
releasing hormone[mh] OR Thyroxine-Binding Proteins[mh] OR Pregnane X Receptor[supplementary
concept] OR thyroid-hormone-receptor interacting protein[supplementary concept] OR Constitutive
androstane receptor[supplementary concept] OR Academic performance[tiab] OR auditory[tiab] OR
cortical[tiab] OR delayed development[tiab] OR developmental impairment[tiab] OR developmental-
delay*[tiab] OR developmental-disorder*[tiab] OR euthyroid[tiab] OR gait[tiab] OR glia*[tiab] OR
gliogenesis[tiab] OR hyperactiv*[tiab] OR impulse-control[tiab] OR iodide peroxidase[tiab] OR
IQ[tiab] OR ischemi*[tiab] OR locomotor[tiab] OR mental deficiency[tiab] OR mental
development[tiab] OR mental illness[tiab] OR mental-deficit[tiab] OR mobility[tiab] OR mood[tiab]
OR morris-maze[tiab] OR morris-water[tiab] OR motor abilit*[tiab] OR Motor activities[tiab] OR
motor performance[tiab] OR nerve[tiab] OR neural[tiab] OR neurobehav*[tiab] OR Neurocognitive
impairment[tiab] OR neurodegenerat*[tiab] OR Neurodevelopment*[tiab] OR neurodisease*[tiab] OR
neurologic*[tiab] OR neuromuscular[tiab] OR neuron*[tiab] OR neuropath*[tiab] OR obsessive
compulsive[tiab] OR OCD[tiab] OR olfaction[tiab] OR olfactory[tiab] OR open-field-test[tiab] OR
passive avoidance[tiab] OR plasticity[tiab] OR senil*[tiab] OR sociab*[tiab] OR speech*[tiab] OR
spelling[tiab] OR stereotypic-movement*[tiab] OR synap*[tiab] OR tauopath*[tiab] OR
Thyroglobulin[tiab] OR Thyroid disease*[tiab] OR thyroid gland[tiab] OR thyroid hormone*[tiab] OR
thyronine*[tiab] OR visual motor[tiab] OR Visuospatial processing[tiab] OR water maze[tiab]) OR
((active-avoidance[tiab] OR ADHD[tiab] OR alzheimer*[tiab] OR amygdala[tiab] OR antisocial[tiab]
OR anxiety[tiab] OR anxious[tiab] OR asperger*[tiab] OR attention deficit[tiab] OR autism[tiab] OR
autistic[tiab] OR behavioral[tiab] OR behaviors[tiab] OR behavioural[tiab] OR behaviours[tiab] OR
bipolar[tiab] OR cerebellum[tiab] OR cognition[tiab] OR cognitive[tiab] OR communication-
disorder*[tiab] OR comprehension[tiab] OR cranial[tiab] OR dementia[tiab] OR dendrit*[tiab] OR
dentate-gyrus[tiab] OR depression[tiab] OR dextrothyroxine[tiab] OR diiodothyronine*[tiab] OR
diiodotyrosine[tiab] OR down syndrome[tiab] OR dyslexia[tiab] OR entorhinal cortex[tiab] OR
epilep*[tiab] OR gangli*[tiab] OR goiter[tiab] OR graves-disease[tiab] OR hearing[tiab] OR
hippocamp*[tiab] OR human development[tiab] OR hyperthyroid*[tiab] OR hypothalam*[tiab] OR
hypothyroid*[tiab] OR impulsiv*[tiab] OR Intellectual disability[tiab] OR intelligence[tiab] OR
language[tiab] OR learning[tiab] OR lewy bod*[tiab] OR long-term potentiation[tiab] OR long-term
synaptic depression[tiab] OR memory[tiab] OR mental disorder*[tiab] OR mental recall[tiab] OR
monoiodotyrosine[tiab] OR Motor activity[tiab] OR motor skill*[tiab] OR multiple sclerosis[tiab] OR
myxedema[tiab] OR Nervous system[tiab] OR nervous-system[tiab] OR neurit*[tiab] OR optic[tiab]
OR palsy[tiab] OR panic[tiab] OR parahippocamp*[tiab] OR paranoia[tiab] OR paranoid[tiab] OR
parkinson*[tiab] OR perception[tiab] OR perforant*[tiab] OR personality[tiab] OR phobia[tiab] OR
problem solving[tiab] OR proprioception[tiab] OR psychomotor[tiab] OR reflex[tiab] OR risk
taking[tiab] OR schizophrenia[tiab] OR seizure*[tiab] OR sensation*[tiab] OR sleep[tiab] OR
smell[tiab] OR spatial behavior[tiab] OR stroke[tiab] OR substantia-nigra[tiab] OR taste[tiab] OR
thyroiditis[tiab] OR thyrotoxicosis[tiab] OR Thyrotropin[tiab] OR thyroxine[tiab] OR
triiodothyronine[tiab] OR vision[tiab]) NOT medline[sb]))

B-4
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Appendix C. Detailed Literature Search Results and List of

Included Studies

Table of Contents
C.1. Detailed Literature Search Results .......................................................................................C-2
C.2. List of Included Studies........................................................................................................C-3

C-1
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

C.1. Detailed Literature Search Results

C.1.1. Literature Search Results Counts and Title and Abstract Screening
The electronic database searches retrieved 25,450 unique references in total (20,883 references
during the initial search conducted in December 2016, 3,657 references during the literature
search updates [including the final updated search conducted for the primary epidemiological
studies on May 1, 2020], and 910 references from the supplemental Chinese database searches);
11 additional references were identified by technical advisors or from reviewing reference lists in
published reviews and included studies. As a result of title and abstract screening, 1,036
references were moved to full-text review, and 24,425 references were excluded (11,402 by
manual screening for not satisfying the PECO criteria and 13,023 based on the SWIFT-Active
Screener algorithm).

C.1.2. Full-text Review

Among the 1,036 references that underwent full-text review, 489 were excluded at that stage
with reasons for exclusion documented; 333 references were excluded for not satisfying the
PECO criteria; and 156 references from the May 2020 searches (main literature search update
and supplemental Chinese database searches) were excluded for not including information that
would materially advance the human, animal in vivo, or mechanistic findings (see the Main
Literature Search section for a description of the methodology). These screening results are
outlined in a study selection diagram that reports numbers of studies excluded for each reason at
the full-text review stage (see Figure 2) [using reporting practices outlined in Page et al. (2021)].
After full-text review, 547 studies were considered relevant with primary neurodevelopmental or
cognitive outcomes, secondary neurobehavioral outcomes, and/or outcomes related to thyroid
function. A few studies assessed data for more than one evidence stream (human, non-human
mammal, and/or in vitro), and several human and animal studies assessed more than one type of
outcome (e.g., primary and secondary outcomes). The number of included studies is summarized
below:
• 167 human studies (84 primary only; 13 secondary only; 5 primary and secondary; 8
primary and thyroid; 2 secondary and thyroid; and 55 thyroid only);
• 339 non-human mammal studies (7 primary only; 186 secondary only; 67 primary
and secondary; 6 primary, secondary, and thyroid; 4 secondary and thyroid; and 69
thyroid only); and,
• 60 in vitro/mechanistic studies (48 neurological and 12 thyroid).
One publication contained human, experimental non-human mammal, and in vitro data. Three
publications contained both human and experimental non-human mammal data. Fourteen
publications contained data relevant to both experimental non-human mammal studies and in
vitro studies.

C-2
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

C.2. List of Included Studies

C.2.1. Studies in Humans

As described in Figure 2, 167 human studies were included; however, full data extraction was
conducted only on studies with neurological outcomes or thyroid hormone data. Data extraction
was completed using HAWC (NTP 2019). Data were extracted from a subset of included studies
in humans (n = 124) and are available in HAWC based on outcome. The following lists of
references are organized as studies that are available in HAWC followed by studies that are not
available in HAWC. Specifically, data for primary neurodevelopmental or cognitive outcomes
(learning, memory, and intelligence) and secondary neurobehavioral outcomes (anxiety,
aggression, motor activity, or biochemical changes), as well as thyroid hormone level data, were
extracted from included human studies and are available in HAWC. Data for included studies
identified through the 2020 literature search update were extracted only for primary
neurodevelopmental or cognitive outcomes; a subset of these studies (n = 7) also included
secondary neurobehavioral outcomes and/or thyroid hormone level data that were not extracted
because those data would not materially advance the human or mechanistic findings. Included
human studies that evaluated only other thyroid-related effects such as goiters or thyroid size
(n = 43) were not extracted and are not available in HAWC. The list below presents the 167
human studies that were included in the review. An overview of the screening results is outlined
in the study selection diagram (Figure 2) that reports numbers of included studies as well as
numbers of studies excluded for each reason at the full-text review stage.
C.2.1.1. Studies Available in HAWC
An J, Mei S, Liu A, Fu Y, Wang C. 1992. [Effect of high level of fluoride on children’s
intelligence]. Chin J Control Endem Dis 7(2): 93-94.
Aravind A, Dhanya RS, Narayan A, Sam G, Adarsh VJ, Kiran M. 2016. Effect of fluoridated
water on intelligence in 10-12-year-old school children. J Int Soc Prev Community Dent 6(Suppl
3): S237-S242.
Bai A, Li Y, Fan Z, Li X, Li P. 2014. [Intelligence and growth development of children in coal-
burning-borne arsenism and fluorosis areas: An investigation study]. Chin J Endemiol 33(2):
160-163.
Barberio AM, Hosein FS, Quinonez C, McLaren L. 2017. Fluoride exposure and indicators of
thyroid functioning in the Canadian population: Implications for community water fluoridation. J
Epidemiol Community Health 71: 1019-1025.
Barberio AM, Quinonez C, Hosein FS, McLaren L. 2017. Fluoride exposure and reported
learning disability diagnosis among Canadian children: Implications for community water
fluoridation. Can J Public Health 108: 229-239.
Bashash M, Thomas D, Hu H, Martinez-Mier EA, Sanchez BN, Basu N, Peterson KE, Ettinger
AS, Wright R, Zhang Z, Liu Y, Schnaas L, Mercado-Garcia A, Tellez-Rojo MM, Hernandez-
Avila M. 2017. Prenatal fluoride exposure and cognitive outcomes in children at 4 and 6-12
years of age in Mexico. Environ Health Perspect 125(9): 1-12.

C-3
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Bashash M, Marchand M, Hu H, Till C, Martinez-Mier EA, Sanchez BN, Basu N, Peterson KE,
Green R, Schnaas L, Mercado-Garcia A, Hernandez-Avila M, Tellez-Rojo MM. 2018. Prenatal
fluoride exposure and attention deficit hyperactivity disorder (ADHD) symptoms in children at
6-12 years of age in Mexico City. Environ Int 121(Pt 1): 658-666.
Broadbent JM, Thomson WM, Moffitt TE, Poulton R. 2015. Community water fluoridation and
intelligence response. Am J Public Health 105: 3-4.
Chen YX, Han FL, Zhoua ZL, Zhang HQ, Jiao XS, Zhang SC, Huang MC, Chang TQ, Dong YF.
1991. [Research on the intellectual development of children in high fluoride areas]. Chin J
Control Endem Dis 6(Suppl): 99-100.
Chen YX, Han FL, Zhoua ZL, Zhang HQ, Jiao XS, Zhang SC, Huang MC, Chang TQ, Dong YF.
2008. Research on the intellectual development of children in high fluoride areas. Fluoride 41:
120-124.
Chinoy NJ, Narayana MV. 1992. Studies on fluorosis in Mehsana District of North Gujarat. Proc
Zool Soc 45: 157-161.
Choi AL, Zhang Y, Sun G, Bellinger DC, Wang K, Yang XJ, Li JS, Zheng Q, Fu Y, Grandjean
P. 2015. Association of lifetime exposure to fluoride and cognitive functions in Chinese children:
A pilot study. Neurotoxicol Teratol 47: 96-101.
Cui Y, Zhang B, Ma J, Wang Y, Zhao L, Hou C, Yu J, Zhao Y, Zhang Z, Nie J, Gao T, Zhou G,
Liu H. 2018. Dopamine receptor D2 gene polymorphism, urine fluoride, and intelligence
impairment of children in China: A school-based cross-sectional study. Ecotoxicol Environ Saf
165: 270-277.
Cui Y, Yu J, Zhang B, Guo B, Gao T, Liu H. 2020. The relationships between thyroid-
stimulating hormone and/or dopamine levels in peripheral blood and IQ in children with
different urinary iodine concentrations. Neurosci Lett 729: 134981.
Das K, Mondal NK. 2016. Dental fluorosis and urinary fluoride concentration as a reflection of
fluoride exposure and its impact on IQ level and BMI of children of Laxmisagar, Simlapal Block
of Bankura District, W.B., India. Environ Monit Assess 188: 218.
Ding Y, Sun H, Han H, Wang W, Ji X, Liu X, Sun D. 2011. The relationships between low
levels of urine fluoride on children's intelligence, dental fluorosis in endemic fluorosis areas in
Hulunbuir, Inner Mongolia, China. J Hazard Mater 186: 1942-1946.
Du L, Wan C, Cao X, Liu J. 1992. [The effect of fluorine on the developing human brain]. Chin
J Pathol 21(4): 218-220.
Du L, Wan C, Cao X, Liu J. 2008. The effect of fluorine on the developing human brain.
Fluoride 41: 327-330.
Duan J, Zhao M, Wang L, Fang D, Wang Y, Wang W. 1995. A comparative analysis of the
results of multiple tests in patients with chronic industrial fluorosis. Guizhou Med J 18(3): 179-
180. Erickson JD, Hay S. 1976. Water fluoridation and congenital malformations: No
association. J Am Dent Assoc 93: 981-984.

C-4
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Erickson JD, Hay S. 1976. Water fluoridation and congenital malformations: No association. J
Am Dent Assoc 93: 981-984.
Erickson JD. 1980. Down syndrome, water fluoridation, and maternal age. Teratology 21: 177-
180.
Eswar P, Nagesh L, Devaraj CG. 2011. Intelligent quotients of 12-14 year old school children in
a high and low fluoride village in India. Fluoride 44: 168-172.
Fan Z, Dai H, Bai A, Li P, Li T, Li G. 2007. Effect of high fluoride exposure in children’s
intelligence. J Environ Health 24(10): 802-803.
Green R, Lanphear B, Hornung R, Flora D, Martinez-Mier EA, Neufeld R, Ayotte P, Muckle G,
Till C. 2019. Association between maternal fluoride exposure during pregnancy and IQ scores in
offspring in Canada. JAMA Pediatr: E1-E9.
Guo XC, Wang RY, Cheng CF, Wei WS, Tang LM, Wang QS, Tang DX, Liu GW, He GD, Li
SL. 1991. [A preliminary investigation of the IQs of 7-13 year-old children from an area with
coal burning-related fluoride poisoning]. Chin J Epidemiol 10(2): 98-100.
Guo XC, Wang RY, Cheng CF, Wei WS, Tang LM, Wang QS, Tang DX, Liu GW, He GD, Li
SL. 2008. A preliminary investigation of the IQs of 7-13 year-old children from an area with coal
burning-related fluoride poisoning. Fluoride 41: 125-128.
Guo ZY, He YH, Zhu QX. 2001. [Research on the neurobehavioral function of workers
occupationally exposed to fluoride]. Ind Hlth & Occup Dis 27(6): 346-348.
Guo ZY, He YH, Zhu QX. 2008. Research on the neurobehavioral function of workers
occupationally exposed to fluoride. Fluoride 41: 152-155.
He H, Cheng ZS, Liu WQ. 1989. [Effects of fluorine on the human fetus]. J Control Endem Dis
4(3): 136-138.
He H, Cheng ZS, Liu WQ. 2008. Effects of fluorine on the human fetus. Fluoride 41: 321-326.
He MX, Zhang CN. 2010. [Investigation of children's intelligence quotient and dental fluorosis
in drinking water-type of endemic fluorosis area in Pucheng County, Shaanxi Province before
and after drinking water change]. Chin J Endemiol 29: 547-548.
Hong F, Wang H, Yang D, Zhang Z. 2001. [Investigation on the intelligence and metabolism of
iodine and fluoride in children with high iodine and fluoride]. Chin J Control Endem Dis 12-14.
Hong FG, Cao YX, Yang D, Wang H. 2001. [Research on the effects of fluoride on child
intellectual development under different environmental conditions]. Chin Prim Health Care
15(3): 56-57.
Hong FG, Cao YX, Yang D, Wang H. 2008. Research on the effects of fluoride on child
intellectual development under different environmental conditions. Fluoride 41: 156-160.
Hosur MB, Puranik RS, Vanaki S, Puranik SR. 2012. Study of thyroid hormones free
triiodothyronine (FT3), free thyroxine (FT4) and thyroid stimulating hormone (TSH) in subjects
with dental fluorosis. Eur J Dent 6: 184-190.

C-5
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Jacqmin H, Commenges D, Letenneur L, Barberger-Gateau P, Dartigues JF. 1994. Components

of drinking water and risk of cognitive impairment in the elderly. Am J Epidemiol 139: 48-57.
Jin T, Han T, Wei Y, Wu Y, Wang Z, Zhang, H. 2016. [Investigation on working memory level
of children aged 8-12 years in coal-burning fluorosis area]. J Environ Health 33(5): 409-411.
Jin T, Wang Z, Wei Y, Wu Y, Han T, Zhang H. 2017. [Investigation on intelligence level of
children aged 8-12 years old in coal-burning fluorosis area]. J Environ Health 34(3): 229-231.
Kang J, Cheng Y, Wu K, Lin S, He G, Jin Y. 2011. Effect of exposure to fluoride and arsenic in
drinking water of Hangjinhouqi on children's intelligence. Chinese School Health: 679-681.
Karimzade S, Aghaei M, Mahvi AH. 2014. Investigation of intelligence quotient in 9-12 year-old
children exposed to high- and low-drinking water fluoride in West Azerbaijan Province, Iran.
Fluoride 47: 9-14.
Khan SA, Singh RK, Navit S, Chadha D, Johri N, Navit P, Sharma A, Bahuguna R. 2015.
Relationship between dental fluorosis and intelligence quotient of school going children in and
around Lucknow District: A cross-sectional study. J Clin Diagn Res 9(11): 10-15.
Khandare AL, Gourineni SR, Validandi V. 2017. Dental fluorosis, nutritional status, kidney
damage, and thyroid function along with bone metabolic indicators in school-going children
living in fluoride-affected hilly areas of Doda District, Jammu and Kashmir, India. Environ
Monit Assess 189: 579.
Khandare AL, Validandi V, Gourineni SR, Gopalan V, Nagalla B. 2018. Dose-dependent effect
of fluoride on clinical and subclinical indices of fluorosis in school going children and its
mitigation by supply of safe drinking water for 5 years: An Indian study. Environ Monit Assess
190: 110.
Kheradpisheh Z, Mahvi AH, Mirzaei M, Mokhtari M, Azizi R, Fallahzadeh H, Ehrampoush MH.
2018. Correlation between drinking water fluoride and TSH hormone by ANNs and ANFIS. J
Environ Health Sci Eng 16(1): 11-18.
Kheradpisheh Z, Mirzaei M, Mahvi AH, Mokhtari M, Azizi R, Fallahzadeh H, Ehrampoush MH.
2018. Impact of drinking water fluoride on human thyroid hormones: A case-control study. Sci
Rep 8: 2674.
Kumar V, Chahar P, Kajjari S, Rahman F, Bansal DK, Kapadia JM. 2018. Fluoride, thyroid
hormone derangements and its correlation with tooth eruption pattern among the pediatric
population from endemic and non-endemic fluorosis areas. J Contemp Dent Pract 19(12): 1512-
1516.
Kundu H, Basavaraj P, Singla A, Gupta R, Singh K, Jain S. 2015. Effect of fluoride in drinking
water on children's intelligence in high and low fluoride areas of Delhi. J Indian Assoc Public
Health Dent 13(2): 116-121.
Lamberg M, Hausen H, Vartiainen T. 1997. Symptoms experienced during periods of actual and
supposed water fluoridation. Community Dent Oral Epidemiol 25: 291-295.

C-6
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Li F, Chen X, Huang R, Xie Y. 2009. The impact of endemic fluorosis caused by the burning of
coal on the development of intelligence in children. J Environ Health 26(4): 838-840.
Li J, Yao L, Shao QL, Wu CY. 2004. [Effects of high fluoride level on neonatal neurobehavioral
development]. Chin J Endemiol 23(5): 463-465.
Li J, Yao L, Shao QL, Wu CY. 2008. Effects of high fluoride level on neonatal neurobehavioral
development. Fluoride 41: 165-170.
Li M, Gao Y, Cui J, Li Y, Li B, Liu Y, Sun J, Liu X, Liu H, Zhao L, Sun D. 2016. Cognitive
impairment and risk factors in elderly people living in fluorosis areas in China. Biol Trace Elem
Res 172: 53-60.
Li X, Hou G, Yu B, Yuan C, Liu Y, L Z. 2010. Investigation and analysis of children’s
intelligence and dental fluorosis in high fluoride area. J Med Pest Control 26(3): 230-231.
Li XS, Zhi JL, Gao RO. 1995. Effect of fluoride exposure on the intelligence of children.
Fluoride 28: 189-192.
Li Y, Li X, Wei S. 1994. [Effects of high fluoride intake on child mental work capacity:
Preliminary investigation into the mechanisms involved]. J West Chin University Med Sci 25(2):
188-191.
Li Y, Li X, Wei S. 2008. Effects of high fluoride intake on child mental work capacity:
Preliminary investigation into the mechanisms involved. Fluoride 41: 331-335.
Li YP, Jing XY, Chen D, Lin L, Wang ZJ. 2003. [Effects of endemic fluoride poisoning on the
intellectual development of children in Baotou]. Chin J Public Health Manag 19(4): 337-338.
Li YP, Jing XY, Chen D, Lin L, Wang ZJ. 2008. Effects of endemic fluoride poisoning on the
intellectual development of children in Baotou. Fluoride 41: 161-164.
Liang C, Ji R, Cao J, Jiang Y, Yu B, Ma F, Wu Y, Ying B, Zhang Y, Sun S, Li Y, Emsley CL,
Gao S, Hall KS, Hendrie HC. 2003. [Study on the relationship between drinking water trace
elements and cognitive ability of the elderly]. Health Res 436-440.
Lin F, Ai H, Zhao H, Lin J, Jhiang J, Maimaiti. 1991. High fluoride and low iodine environment
and subclinical cretinism in Xinjiang. Endem Dis Bull 6(2): 62-67.
Liu S, Lu Y, Sun Z, Wu L, Wang X, Yan S. 2000. [Report on the intellectual ability of children
living in high-fluoride water areas]. Chin J Control Endem Dis 15(4): 231-232.
Liu SL, Lu Y, Sun ZR, Wu L, Lu WL, Wang XW, Yan S. 2008. Report on the intellectual ability
of children living in high-fluoride water areas. Fluoride 41: 144-147.
Lu F, Zhang Y, Trivedi A, Jiang X, Chandra D, Zheng J, Nakano Y, Uyghurturk DA, Jalai R,
Guner S, Mentes A, DenBesten PK. 2019. Fluoride related changes in behavioral outcomes may
relate to increased serotonin. Physiol Behav 206: 76-83.
Lu Y, Sun ZR, Wu LN, Wang X, Lu W, Liu SS. 2000. Effect of high-fluoride water on
intelligence in children. Fluoride 33: 74-78.

C-7
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Luo Y, Ma R, Liu Z, Guan Z, Lou D, Zheng D. 2018. [Intelligence investigation and forensic
significance of children in coal-burning fluorosis area]. Chin J Forensic Medicine 33(6): 590-
593.
Malin AJ, Till C. 2015. Exposure to fluoridated water and attention deficit hyperactivity disorder
prevalence among children and adolescents in the United States: An ecological association.
Environ Health 14: 17.
Malin AJ, Riddell J, McCague H, Till C. 2018. Fluoride exposure and thyroid function among
adults living in Canada: Effect modification by iodine status. Environ Int 121(Pt 1): 667-674.
Michael M, Barot VV, Chinoy NJ. 1996. Investigations of soft tissue functions in fluorotic
individuals of North Gujarat. Fluoride 29: 63-71.
Mondal D, Dutta G, Gupta S. 2016. Inferring the fluoride hydrogeochemistry and effect of
consuming fluoride-contaminated drinking water on human health in some endemic areas of
Birbhum District, West Bengal. Environ Geochem Health 38: 557-576.
Morgan L, Allred E, Tavares M, Bellinger D, Needleman H. 1998. Investigation of the possible
associations between fluorosis, fluoride exposure, and childhood behavior problems. Pediatr
Dent 20: 244-252.
Mustafa DE, Younis UM, Elhag SAA. 2018. The relationship between the fluoride levels in
drinking water and the schooling performance of children in rural areas of Khartoum State,
Sudan. Fluoride 51(2): 102-113.
Nagarajappa R, Pujara P, Sharda AJ, Asawa K, Tak M, Aapaliya P, Bhanushali N. 2013.
Comparative assessment of intelligence quotient among children living in high and low fluoride
areas of Kutch, India: A pilot study. Iran J Public Health 42: 813-818.
Peckham S, Lowery D, Spencer S. 2015. Are fluoride levels in drinking water associated with
hypothyroidism prevalence in England? A large observational study of GP practice data and
fluoride levels in drinking water. J Epidemiol Community Health 69: 619-624.
Poureslami HR, Horri A, Garrusi B. 2011. A comparative study of the IQ of children age 7-9 in a
high and a low fluoride water city in Iran. Fluoride 44: 163-167.
Pratap SV, Singh CD, Sandeep T, Sandeep K, Vikas G, Mukesh T, Anurag T. 2013. A
correlation between serum vitamin, acetylcholinesterase activity and IQ in children with
excessive endemic fluoride exposure in Rajasthan, India. Int Res J Medical Sci 1(3): 12-16.
Qin LS, Huo SY, Chen RL, Chang YZ, Zhao MY. 1990. [Using the Raven's Standard
Progressive Matrices to determine the effects of the level of fluoride in drinking water on the
intellectual ability of school-age children]. J Control Endem Dis 5(4): 203-204.
Qin LS, Huo SY, Chen RL, Chang YZ, Zhao MY. 2008. Using the Raven's Standard Progressive
Matrices to determine the effects of the level of fluoride in drinking water on the intellectual
ability of school-age children. Fluoride 41: 115-119.
Ratan SK, Rattan KN, Pandey RM, Singhal S, Kharab S, Bala M, Singh V, Jhanwar A. 2008.
Evaluation of the levels of folate, vitamin B12, homocysteine and fluoride in the parents and the

C-8
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

affected neonates with neural tube defect and their matched controls. Pediatr Surg Int 24: 803-
808.
Razdan P, Patthi B, Kumar JK, Agnihotri N, Chaudhari P, Prasad M. 2017. Effect of fluoride
concentration in drinking water on intelligence quotient of 12-14-year-old children in Mathura
District: A cross-sectional study. J Int Soc Prev Community Dent 7: 252-258.
Ren D, Li K, Liu D. 1989. [A study of the intellectual ability of 8-14 year-old children in high
fluoride, low iodine areas]. Chin J Control Endem Dis 4(4): 251.
Ren D, Li K, Liu D. 2008. A study of the intellectual ability of 8-14 year-old children in high
fluoride, low iodine areas. Fluoride 41: 319-320.
Riddell JK, Malin AJ, Flora D, McCague H, Till C. 2019. Association of water fluoride and
urinary fluoride concentrations with attention deficit hyperactivity disorder in Canadian youth.
Environ Int 133: 105190.
Rocha-Amador D, Navarro ME, Carrizales L, Morales R, Calderon J. 2007. Decreased
intelligence in children and exposure to fluoride and arsenic in drinking water. Cad Saude
Publica 23 Suppl 4: S579-587.
Rocha-Amador D, Navarro M, Trejo-Acevedo A, Carrizales L, Perez-Maldonado I, Diaz-Barriga
F, Calderon J. 2009. Use of the Rey-Osterrieth Complex Figure Test for neurotoxicity evaluation
of mixtures in children. Neurotox 30: 1149-1154.
Rotton J, Tikofsky RS, Feldman HT. 1982. Behavioral effects of chemicals in drinking water. J
Appl Psychol 67: 230-238.
Russ TC, Killin LOJ, Hannah J, Batty GD, Deary IJ, Starr JM. 2019. Aluminium and fluoride in
drinking water in relation to later dementia risk. Brit J Psychol 14: 1-6.
Saxena S, Sahay A, Goel P. 2012. Effect of fluoride exposure on the intelligence of school
children in Madhya Pradesh, India. J Neurosci Rural Pract 3: 144-149.
Sebastian ST, Sunitha S. 2015. A cross-sectional study to assess the intelligence quotient (IQ) of
school going children aged 10-12 years in villages of Mysore District, India with different
fluoride levels. J Indian Soc Pedod Prev Dent 33: 307-311.
Seraj B, Shahrabi M, Falahzade M, Falahzade F, Akhondi N. 2006. Effect of high fluoride
concentration in drinking water on children’s intelligence. J Dent Med 19(2): 80-86.
Seraj B, Shahrabi M, Shadfar M, Ahmadi R, Fallahzadeh M, Eslamlu HF, Kharazifard MJ. 2012.
Effect of high water fluoride concentration on the intellectual development of children in Makoo,
Iran. J Dent 9: 221-229.
Shannon FT, Fergusson DM, Horwood LJ. 1986. Exposure to fluoridated public water supplies
and child health and behaviour. New Zeal Med J 99: 416-418.
Shao Q. 2003. Study of cognitive function impairment caused by chronic fluorosis. Chin J
Endemiol 22(4): 336-338.

C-9
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Sharma JD, Sohu D, Jain P. 2009. Prevalence of neurological manifestations in a human

population exposed to fluoride in drinking water. Fluoride 42: 127-132.
Shivaprakash PK, Ohri K, Noorani H. 2011. Relation between dental fluorosis and intelligence
quotient in school children of Bagalkot District. J Indian Soc Pedod Prev Dent 29: 117-120.
Singh A, Jolly SS, Devi P, Bansal BC, Singh SS. 1962. Endemic fluorosis: An epidemiological,
biochemical and clinical study in the Bhatinda District of Panjab. Indian J Med Res 50: 387-398.
Singh N, Verma KG, Verma P, Sidhu GK, Sachdeva S. 2014. A comparative study of fluoride
ingestion levels, serum thyroid hormone & TSH level derangements, dental fluorosis status
among school children from endemic and non-endemic fluorosis areas. Springerplus 3: 7.
Soto-Barreras U, Escalante-Villalobos KY, Holguin-Loya B, Perez-Aguirre B, Nevarez-Rascon
A, Martinez-Martinez RE, Loyola-Rodriguez JP. 2019. Effect of fluoride in drinking water on
dental caries and IQ in children. Fluoride 52: 474-482.
Still CN, Kelley P. 1980. The incidence of primary degenerative dementia vs. water fluoride
content in South Carolina, USA. Neurotox 1: 125-132.
Sudhir KM, Chandu GN, Prashant GM, Subba Reddy VV. 2009. Effect of fluoride exposure on
intelligence quotient (IQ) among 13-15 year old school children of known endemic area of
fluorosis, Nalgonda District, Andhra Pradesh. J Indian Assoc Public Health Dent 2009(13): 88-
94.
Sun M, Li S, Wang Y, Li F. 1991. Measurement of intelligence by drawing test among the
children in the endemic area of Al-F combined toxicosis. J Guiyang Med Coll 16(3): 204-206.
Susheela AK, Bhatnagar M, Vig K, Mondal NK. 2005. Excess fluoride ingestion and thyroid
hormone derangements in children living in Delhi, India. Fluoride 38: 98-108.
Tamboli BL, Mathur GM, Mathur AP, Lalla SK, Goyal OP. 1980. Prevalence of fluorosis in
Pratabpura and Surajpura villages, District Ajmer (Rajasthan). Indian J Med Res 71: 57-67.
Till C, Green R, Flora D, Hornung R, Martinez-Mier EA, Blazer M, Farmus L, Ayotte P, Muckle
G, Lanphear B. 2020. Fluoride exposure from infant formula and child IQ in a Canadian birth
cohort. Environ Int 134: 105315.
Tripathi P, Sultana N. 2007. Fluoride content of groundwater and prevalence of dental, skeletal
and neurological stage of fluorosis in Tehsil Purwa of Unnao. Indian J Environ Prot 27: 737-
739.
Trivedi MH, Verma RJ, Chinoy NJ, Patel RS, Sathawara NG. 2007. Effect of high fluoride water
on intelligence of school children in India. Fluoride 40: 178-183.
Trivedi M, Sangai N, Patel R, Payak M, Vyas S. 2012. Assessment of groundwater quality with
special reference to fluoride and its impact on IQ of schoolchildren in six villages of the Mundra
Region, Kachchh, Gujurat, India. Fluoride 45(4): 377-383.

C-10
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Valdez Jimenez L, Lopez Guzman OD, Cervantes Flores M, Costilla-Salazar R, Calderon

Hernandez J, Alcaraz Contreras Y, Rocha-Amador DO. 2017. In utero exposure to fluoride and
cognitive development delay in infants. Neurotox 59: 65-70.
Wang A, Duan L, Huang H, Ma J, Zhang Y, Ma Q, Guo Y, Li Z, Cheng X, Zhu J, Zhou G, Ba
Y. 2020. Association between fluoride exposure and behavioural outcomes of school-age
children: A pilot study in China. Int J Environ Health Res: 1-10.
Wang G, Yang D, Jia F, Wang H. 1996. [A study of the IQ levels of four- to seven-year-old
children in high fluoride areas]. Endem Dis Bull 11(1): 60-62.
Wang G, Yang D, Jia F, Wang H. 2008. A study of the IQ levels of four- to seven-year-old
children in high fluoride areas. Fluoride 41: 340-343.
Wang G, Gao M, Zhang M, Yang M, Xiang Q. 2012. Correlation between total fluoride intake
and children's IQ. J Southeast Univ Med Ed: 743-746.
Wang G, Zhang M, Wang Q, Han A, Gao M, Lin P, Xiang Q. 2017. [Investigation on the
relationship between serum fluoride content and IQ of children before and after reducing fluoride
to water]. Capital Public Health 11(6): 274-277.
Wang M, Liu L, Li H, Li Y, Liu H, Hou C, Zeng Q, Li P, Zhao Q, Dong L, Zhou G, Yu X, Liu
L, Guan Q, Zhang S, Wang A. 2020. Thyroid function, intelligence, and low-moderate fluoride
exposure among Chinese school-age children. Environ Int 134: 105229.
Wang S, Wang L, Hu P, Guo S, Law S. 2001. [Effects of high iodine and high fluorine on
children’s intelligence and thyroid function]. Chin J Endemiol 20(4): 288-290.
Wang SX, Wang ZH, Cheng XT, Li J, Sang ZP, Zhang XD, Han LL, Qiao XY, Wu ZM, Wang
ZQ. 2005. [Investigation and evaluation on intelligence and growth of children in endemic
fluorosis and arsenism areas]. Chin J Endemiol 24: 179-182.
Wang SX, Wang ZH, Cheng XT, Li J, Sang ZP, Zhang XD, Han LL, Qiao XY, Wu ZM, Wang
ZQ. 2007. Arsenic and fluoride exposure in drinking water: Children's IQ and growth in Shanyin
County, Shanxi Province, China. Environ Health Perspect 115: 643-647.
Wang SY, Zhang HX, Fan W, Fang SJ, Kang PP, Chen XG, Yu MJ. 2005. [The effects of
endemic fluoride poisoning caused by coal burning on the physical development and intelligence
of children]. J Appl Clin Ped 20(9): 897-899.
Wang SY, Zhang HX, Fan W, Fang SJ, Kang PP, Chen XG, Yu MJ. 2008. The effects of
endemic fluoride poisoning caused by coal burning on the physical development and intelligence
of children. Fluoride 41: 344-348.
Wang Z, Wang S, Zhang X, Li J, Zheng X, Hu C. 2006. Investigation of children’s growth and
development under long-term fluoride exposure. Chin J Control Endem Dis 21(4): 239-241.
Wei N, Li Y, Deng J, Xu S, Guan Z. 2014. [The effects of comprehensive control measures on
intelligence of school-age children in coal-burning-borne endemic fluorosis areas]. Chin J
Endemiol 33(2): 320-322.

C-11
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Xiang Q, Liang Y, Chen L, Wang C, Chen B, Chen X, Zhou M. 2003. Effect of fluoride in
drinking water on children's intelligence. Fluoride 36: 84-94.
Xiang Q, Liang Y, Chen B, Chen L. 2011. Analysis of children's serum fluoride levels in relation
to intelligence scores in a high and low fluoride water village in China. Fluoride 44: 191-194.
Xu Y, Lu C, Zhang X. 1994. Effect of fluoride on children’s intelligence. Endem Dis Bull 2: 83-
84.
Yang Y, Wang X, Guo X, Hu P. 1994. [The effects of high levels of fluoride and iodine on child
intellectual ability and the metabolism of fluoride and iodine]. Chin J Epidemiol 15(4): 296-298.
Yang Y, Wang X, Guo X, Hu P. 2008. The effects of high levels of fluoride and iodine on child
intellectual ability and the metabolism of fluoride and iodine. Fluoride 41: 336-339.
Yao L, Zhou J, Wang S, Cui K, Lin F. 1996. Analysis of TSH levels and intelligence of children
residing in high fluorosis areas. Lit Inf Prev Med 2(1): 26-27.
Yao Y. 1997. Comparative assessment of the physical and mental development of children in
endemic fluorosis area with water improvement and without water improvement. Lit Inf Prev
Med 3(1): 42-43.
Yasmin S, Ranjan S, D'Souza D. 2013. Effect of excess fluoride ingestion on human thyroid
function in Gaya Region, Bihar, India. Toxicol Environ Chem 95: 1235-1243.
Yazdi SM, Sharifian A, Dehghani-Beshne M, Momeni VR, Aminian O. 2011. Effects of fluoride
on psychomotor performance and memory of aluminum potroom workers. Fluoride 44: 158-162.
Yu YN, Yang WX, Dong Z, Wan CW, Zhang JT, Liu JL, Xiao KQ, Huang YS, Lu BF. 1996.
[Neurotransmitter and receptor changes in the brains of fetuses from areas of endemic fluorosis].
Chin J Endemiol 15(5): 257-259.
Yu YN, Yang WX, Dong Z, Wan CW, Zhang JT, Liu JL, Xiao KQ, Huang YS, Lu BF. 2008.
Neurotransmitter and receptor changes in the brains of fetuses from areas of endemic fluorosis.
Fluoride 41: 134-138.
Yu X, Chen J, Li Y, Liu H, Hou C, Zeng Q, Cui Y, Zhao L, Li P, Zhou Z, Pang S, Tang S, Tian
K, Zhao Q, Dong L, Xu C, Zhang X, Zhang S, Liu L, Wang A. 2018. Threshold effects of
moderately excessive fluoride exposure on children's health: A potential association between
dental fluorosis and loss of excellent intelligence. Environ Int 118: 116-124.
Zhang J, Yao H, Chen Y. 1998. [Effect of high level of fluoride and arsenic on children’s
intelligence]. Chin J Public Health 17(2): 57.
Zhang P, Cheng L. 2015. [Effect of coal-burning endemic fluorosis on children’s physical
development and intellectual level]. Chin J Control Endem Dis 30(6): 458-459.
Zhang S, Zhang X, Liu H, Qu W, Guan Z, Zeng Q, Jiang C, Gao H, Zhang C, Lei R, Xia T,
Wang Z, Yang L, Chen Y, Wu X, Cui Y, Yu L, Wang A. 2015. Modifying effect of COMT gene
polymorphism and a predictive role for proteomics analysis in children's intelligence in endemic
fluorosis area in Tianjin, China. Toxicol Sci 144: 238-245.

C-12
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Zhao LB, Liang GH, Zhang DN, Wu XR. 1996. Effect of a high fluoride water supply on
children's intelligence. Fluoride 29: 190-192.
Zhao Q, Niu Q, Chen J, Xia T, Zhou G, Li P, Dong L, Xu C, Tian Z, Luo C, Liu L, Zhang S,
Wang A. 2019. Roles of mitochondrial fission inhibition in developmental fluoride
neurotoxicity: Mechanisms of action in vitro and associations with cognition in rats and children.
Arch Toxicol 93(3): 709-726.
Zhao Q, Tian Z, Zhou G, Niu Q, Chen J, Li P, Dong L, Xia T, Zhang S, Wang A. 2020. SIRT1-
dependent mitochondrial biogenesis supports therapeutic effects of resveratrol against
neurodevelopment damage by fluoride. Theranostics 10: 4822-4838.
Zhao Y, Cui Y, Yu J, Zhang B, Nie J, Zhao L, Zhang Z, Liu H. 2018. [Study on the relationship
between water-borne high iodine and thyroid hormone and children's intelligence level]. J
Environ Health 35(1): 6-9.
Zhou G, Tang S, Yang L, Niu Q, Chen J, Xia T, Wang S, Wang M, Zhao Q, Liu L, Li P, Dong
L, Yang K, Zhang S, Wang A. 2019. Effects of long-term fluoride exposure on cognitive ability
and the underlying mechanisms: Role of autophagy and its association with apoptosis. Toxicol
Appl Pharmacol 378: 114608.
C.2.1.2. Studies Not Available in HAWC
Bachinskii PP, Gutsalenko OA, Naryzhniuk ND, Sidora VD, Shliakhta AI. 1985. [Action of the
body fluorine of healthy persons and thyroidopathy patients on the function of hypophyseal-
thyroid the system]. Probl Endokrinol 31: 25-29.
Balabolkin MI, Mikhailets ND, Lobovskaia RN, Chernousova NV. 1995. [The interrelationship
of the thyroid and immune statuses of workers with long-term fluorine exposure]. Ter Arkh 67:
41-42.
Baum K, Boerner W, Reiners C, Moll E. 1981. [Bone density and thyroid function in adolescents
in relation to fluoride content of drinking water]. Fortschr Med 99: 1470-1472.
Berry WTC, Whittles JH. 1963. Absence of effect of fluoride upon the incidence of thyroid
enlargements in Wiltshire schoolgirls. Mon Bull Minist Health Public Health Lab Serv 22: 50-
52.
Cherkinskii SN, Zaslavskaia RM. 1956. [Significance of fluorides in potable water in the
development of endemic goiter]. Probl Endokrinol Gormonoter 2: 70-75.
Choubisa SL. 2001. Endemic fluorosis in southern Rajasthan, India. Fluoride 34: 61-70.
Chuka A, Zhukovskil V, Mirku I, Postel'Niku D. 1964. Prezhdevremennoe starenie naseleniya v
zone rasprostraneniya endemicheskogo zoba. Vestnik Akad Med Nauk Sssr 19: 23-27.
Dai HX, Zeng P, Wang KY, Zhang XG, Ma ZJ, Zhou YG, Fan ZX, Guo SH. 2013. [Analysis of
a survey results of patients with suspected high iodine goiter in Liuji Town Fuping County of
Shaanxi Province]. Chin J Endemiol 32: 408-411.
Day T, Powell-Jackson P. 1972. Fluoride, water hardness, and endemic goitre. Lancet
299(7761): 1135-1138.

C-13
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Desai VK, Solanki DM, Bansal RK. 1993. Epidemiological study of goitre in endemic fluorosis
district of Gujarat. Fluoride 26: 187-190.
Díaz-Cadórniga FJ, Delgado E, Tartón T, Valdés MM, Méndez A, Fernández MT, Rojo C. 2003.
Endemic goiter associated with high iodine intake in primary school children in the Saharawi
Arab Democratic Republic. Endocrinol Nutr 50: 357-362.
Eichner R, Borner W, Henschler D, Kohler W, Moll E. 1981. [Osteoporosis therapy and thyroid
function. Influence of 6 months of sodium fluoride treatment on thyroid function and bone
density]. Fortschr Med 99: 342-348.
Fiorentini S, Galeazzi M, Visintin B. 1947. II fluoro in natura come agente morbigeno II. La
fluorosi die Campagnano di Roma. III. Un focolaio di fluorosi umana a Campagnano di Roma.
IV. Osservazioni radiologiche sui processi alveolari, sulle ossa mascellari, e sul paradenzio degli
abitanti die Campagnano. V. Zona fluorotica intorno a Campagnano di Roma. VI. Frequenza e
caratteri clinici della carie dentale in soggetti fluorotici. Rend Ist Superiore Sanita 10: 721-804.
Gabovich RD, Verzhikovskaia NV. 1958. [Effect of fluoride compounds on absorption of
radioactive iodine by the thyroid gland in humans and in experimental conditions]. Probl
Endokrinol Gormonoter 4: 49-54.
Galletti P, Held HR, Korrodi H, Wegmann T. 1956. [Investigations on the possible damaging
effect of fluorine on the thyroid gland]. Helv Med Acta 23: 601-605.
Galletti PM, Joyet G, Jallut O. 1957. [Effect of sodium fluoride on thyroid function in Basedow's
Disease]. Helv Med Acta 24: 209-215.
Galletti PM, Joyet G. 1958. Effect of fluorine on thyroidal iodine metabolism in
hyperthyroidism. J Clin Endocrinol Metab 18: 1102-1110.
Gas'kov AI, Savchenkov MF, Iushkov NN. 2005. [The specific features of the development of
iodine deficiencies in children living under environmental pollution with fluorine compounds].
Gig Sanit: 53-55.
Gedalia I, Brand N. 1963. The relationship of fluoride and iodine in drinking water in the
occurrence of goiter. Arch Int Pharmacodyn Ther 142: 312-315.
Grimm H. 1973. [The physical development of schoolchildren under the influence of drinking
water fluoridation in Karl Marx Stadt]. Dtsch Gesundheitsw 28: 2363-2369.
Hasling C, Nielsen HE, Melsen F, Mosekilde L. 1987. Safety of osteoporosis treatment with
sodium fluoride, calcium phosphate and vitamin D. Miner Electrolyte Metab 13: 96-103.
Hidehiko T. 1958. On the relation between the distribution of endemic goiter and the fluorine
content of natural water in Hidaka Province, Hokkaido. Folia Pharmacol Jpn 54: 225-229.
Hoffmann-Axthelm W. 1953. [Observations on the influence of fluorine on dental enamel and
thyroid gland]. Dtsch Zahnarztl Z 8: 757-765.
Jentzer A. 1956. [Effect of fluorine on the iodine content of the human thyroid gland]. Bull
Schweiz Akad Med Wiss 12: 539-543.

C-14
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Jooste PL, Weight MJ, Kriek JA, Louw AJ. 1999. Endemic goitre in the absence of iodine
deficiency in schoolchildren of the Northern Cape Province of South Africa. Eur J Clin Nutr 53:
8-12.
Kolomiitseva MG. 1961. [The content of fluorine in the external environment of the Upper Altai
autonomous region and its role in the etiology of endemic goiter]. Gig Sanit 26: 101-103.
Korrodi H, Wegmann T, Galleti P, Held HR. 1955. [Caries prophylaxis and the untoward effects
of fluor on the thyroid gland]. Schweiz Med Wochenschr 85: 1016-1019.
Kutlucan A, Kale Koroglu B, Numan Tamer M, Aydin Y, Baltaci D, Akdogan M, Ozturk M,
Vural H, Ermis F. 2013. The investigation of effects of fluorosis on thyroid volume in school-age
children. Med Glas 10: 93-98.
Latham MC, Grech P. 1967. The effects of excessive fluoride intake. Am J Public Health 57:
651-660.
Leone NC, Leatherwood EC, Petrie IM, Lieberman L. 1964. Effect of fluoride on thyroid gland:
Clinical study. J Am Dent Assoc 69: 179-180.
Levi JE, Silberstein HE. 1955. Lack of effect of fluorine ingestion on uptake of iodine 131 by the
thyroid gland. J Lab Clin Med 45: 348-351.
McGlashan N, Chelkowska E, Sasananan S. 2010. A survey of goiter morbidity in Ban Mae
Toen, northwest Thailand. Southeast Asian J Trop Med Public Health 41: 1200-1208.
Rathore S, Meena C, Gonmei Z, Dwivedi S, Toteja GS, Bala K. 2018. Study of excess fluoride
ingestion and thyroid hormone derangement in relation with different fluoride levels in drinking
water among children of Jodhpur District, Rajasthan, India. Asian J Microbiol Biotechnol
Environ Sci 20: 327-331.
Reisenauer R, Rezler D, Křemenová J, Preininger Q. 1961. [Fluorization of the waters in
Czechoslovakia. IV. Endocrinological control of results of two years' fluorization of drinking-
water in school children]. Cesk Stomatol 61: 91-97.
Romer TEZ, Kowalczyk B, Kacprzak M, Wiktorowski M. 1976. [Incidence of goiter in pubertal
girls of the Piotrkow Region and iodide content in drinking water]. Endokrynol Pol 27: 373-380.
Savchenkov MF, Efimova NV, Manueva RS, Nikolaeva LA, Shin NS. 2016. [Thyroid gland
pathology in children population exposed to the combination of iodine deficiency and fluoride
pollution of environment]. Gig Sanit 95: 1201-1205.
Shtifanova AK. 1962. [The fluorine content in water, soil and vegetal products of the Alma-
Atinsk District areas and its role in the etiology of dental caries and endemic goiter].
Zdravookhranenie Kazakhstana: 60-63.
Siddiqui AH. 1969. Incidence of simple goiter in areas of endemic fluorosis in Nalgonda
District, Andhra Pradesh, India. Fluoride 2: 200-205.

C-15
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Sidora VD, Shliakhta AI, Iugov VK, Kas'ianenko AS, Piatenko VG. 1983. [Indices of the
pituitary-thyroid system in residents of cities with various fluorine concentrations in drinking
water]. Probl Endokrinol 29: 32-35.
Sung FC, Chen KP, Chen CY, Tai PW, Yang CF. 1973. Studies of the effect of salt iodization on
endemic goiter in Taiwan. IV. A survey of drinking water in relation to endemic goiter. J
Fomosan Med Assoc 72: 96-103.
Tokar VI, Voroshnin VV, Sherbakov SV. 1989. [Chronic effects of fluorides on the pituitary-
thyroid system in industrial workers]. Gig Tr Prof Zabol: 19-22.
Wespi HJ. 1954. [Iodized-fluoridized salt for the prevention of goiter and caries]. Schweiz Med
Wochenschr 84: 885-890.
Yu YN. 1985. [Effects of chronic fluorosis on the thyroid gland]. Chin Med J 65: 747-7479.

C.2.2. Studies in Non-human Animals

As described in Figure 2, 339 non-human mammal studies were included; however, full data
extraction was conducted only on studies with primary neurological outcomes and/or secondary
functional neurological outcomes (e.g., motor activity). Data extraction was completed using
HAWC (NTP 2019). Data were extracted from a subset of included studies in animals (n = 123)
and are available in HAWC based on outcome. The following lists of references are organized as
studies that are available in HAWC followed by studies that are not available in HAWC.
Specifically, all primary outcomes and functional neurological secondary outcomes (e.g., motor
activity) were extracted from animal studies and are available in HAWC, including studies from
the NTP (2016) assessment. Studies are also available in HAWC that evaluated mechanistic
effects related to oral fluoride exposure at or below 20 ppm fluoride drinking water equivalents
for categories of mechanistic endpoints with the largest amount of available data (i.e.,
biochemistry of the brain or neurons, neurotransmission, oxidative stress, and histopathology
[n = 70]); however, these mechanistic data were generally not extracted. Several animal studies
assessed primary neurological outcomes and/or functional neurological secondary outcomes and
mechanistic effects in the four mechanistic categories listed above (n = 56). In total, 140 animal
studies are available in HAWC (70 with primary neurological outcomes and/or secondary
functional neurological outcomes without relevant mechanistic data; 15 with relevant
mechanistic data only; and 55 with primary and/or secondary functional neurological outcomes
with relevant mechanistic data). Studies that evaluated other mechanistic endpoints, as well as
studies that assessed only mechanistic effects at fluoride levels above 20 ppm fluoride drinking
water equivalents, are not available in HAWC (n = 199). The list below presents the 339 non-
human animal studies that were included in the review. An overview of the screening results is
outlined in the study selection diagram (Figure 2) that reports numbers of included studies as
well as numbers of studies excluded for each reason at the full-text review stage.
C.2.2.1. Studies Available in HAWC
Adedara IA, Abolaji AO, Idris UF, Olabiyi BF, Onibiyo EM, Ojuade TD, Farombi EO. 2017.
Neuroprotective influence of taurine on fluoride-induced biochemical and behavioral deficits in
rats. Chem Biol Interact 261: 1-10.

C-16
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Adedara IA, Olabiyi BF, Ojuade TD, Idris UF, Onibiyo EM, Farombi EO. 2017. Taurine
reverses sodium fluoride-mediated increase in inflammation, caspase-3 activity, and oxidative
damage along the brain-pituitary-gonadal axis in male rats. Can J Phys Pharmacol 95: 1019-
1029.
Agustina F, Sofro ZM, Partadiredja G. 2018. Subchronic administration of high-dose sodium
fluoride causes deficits in cerebellar purkinje cells but not motor coordination of rats. Biol Trace
Elem Res 188(2): 424-433.
Ahmad KR, Noor S, Jabeen S, Nauroze T, Kanwal MA, Raees K, Abbas T. 2017. Amelioration
by jambul fruit extract of fluoride-induced hepato-nephronal histopathologies and impaired
neuromotor capacity in mice. Fluoride 50: 2-14.
Akinrinade ID, Memudu AE, Ogundele OM. 2015. Fluoride and aluminium disturb neuronal
morphology, transport functions, cholinesterase, lysosomal and cell cycle activities.
Pathophysiology 22: 105-115.
Akinrinade ID, Memudu AE, Ogundele OM, Ajetunmobi OI. 2015. Interplay of glia activation
and oxidative stress formation in fluoride and aluminium exposure. Pathophysiology 22: 39-48.
Baba NA, Raina R, Verma PK, Sultana M. 2014. Alterations in plasma and tissue
acetylcholinesterase activity following repeated oral exposure of chlorpyrifos alone and in
conjunction with fluoride in Wistar rats. Proc Indian Natl Sci Acad B Biol Sci 84: 969-972.
Bagmut I, Kolisnyk I, Titkova A, Babiy L, Filipchenko S. 2018. The antioxidant system
enzymes’ activity in rats’ brain, intoxicated with sodium fluoride in subtoxic doses. Arch Balkan
Med Union 53(4): 506-511.
Balaji B, Kumar EP, Kumar A. 2015. Evaluation of standardized bacopa monniera extract in
sodium fluoride-induced behavioural, biochemical, and histopathological alterations in mice.
Toxicol Ind Health 31: 18-30.
Balayssac D, Richard D, Authier N, Nicolay A, Jourdan D, Eschalier A, Coudore F. 2002.
Absence of painful neuropathy after chronic oral fluoride intake in Sprague-Dawley and Lou/C
rats. Neurosci Lett 327: 169-172.
Banala RR, Karnati PR. 2015. Vitamin A deficiency: An oxidative stress marker in sodium
fluoride (NaF) induced oxidative damage in developing rat brain. Int J Dev Neurosci 47: 298-
303.
Banala R, Nagapuri K, Mohd K, Reddy M, Karnati P. 2018. Carica Papaya leaf extract as a
neuroprotective agent against behavioral and neurotransmitter changes in brain of the rat treated
with sodium fluoride in pre- and post-natal periods. Pharmacogn Mag 14(55): 123-131.
Banji D, Banji OJ, Pratusha NG, Annamalai AR. 2013. Investigation on the role of spirulina
platensis in ameliorating behavioural changes, thyroid dysfunction and oxidative stress in
offspring of pregnant rats exposed to fluoride. Food Chem 140: 321-331.
Baran-Poesina V, Negres S, Dobrescu D, Dimcevici-Poesina N, Dimcevici-Poesina A, Feghiu A,
Soare T, Militaru M. 2013. Experimental pharmacological researches regarding the influence of
sodium fluoride in allopathic and homeopathic doses on central nervous system's performances:

C-17
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

A correlation between behavioral response in classic maze test and morphological aspects of
cerebral cortex. Farmacia 61: 781-799.
Bartos M, Gumilar F, Bras C, Gallegos CE, Giannuzzi L, Cancela LM, Minetti A. 2015.
Neurobehavioural effects of exposure to fluoride in the earliest stages of rat development.
Physiol Behav 147: 205-212.
Bartos M, Gumilar F, Gallegos CE, Bras C, Dominguez S, Monaco N, Esandi MDC, Bouzat C,
Cancela LM, Minetti A. 2018. Alterations in the memory of rat offspring exposed to low levels
of fluoride during gestation and lactation: Involvement of the alpha7 nicotinic receptor and
oxidative stress. Reprod Toxicol 81: 108-114.
Basha PM, Rai P, Begum S. 2011. Fluoride toxicity and status of serum thyroid hormones, brain
histopathology, and learning memory in rats: A multigenerational assessment. Biol Trace Elem
Res 144: 1083-1094.
Basha PM, Sujitha NS. 2012. Combined impact of exercise and temperature in learning and
memory performance of fluoride toxicated rats. Biol Trace Elem Res 150: 306-313.
Bataineh HN, Nusier MK. 2006. Impact of 12-week ingestion of sodium fluoride on aggression,
sexual behavior, and fertility in adult male rats. Fluoride 39: 293-301.
Bera I, Sabatini R, Auteri P, Flace P, Sisto G, Montagnani M, Potenza MA, Marasciulo FL,
Carratu MR, Coluccia A, Borracci P, Tarullo A, Cagiano R. 2007. Neurofunctional effects of
developmental sodium fluoride exposure in rats. Eur Rev Med Pharmacol Sci 11: 211-224.
Bhatnagar M, Rao P, Sushma J, Bhatnagar R. 2002. Neurotoxicity of fluoride:
Neurodegeneration in hippocampus of female mice. Indian J Exp Biol 40: 546-554.
Bhatnagar M, Sukhwal P, Suhalka P, Jain A, Joshi C, Sharma D. 2011. Effects of fluoride in
drinking water on NADPH-diaphorase neurons in the forebrain of mice: A possible mechanism
of fluoride neurotoxicity. Fluoride 44: 195-209.
Chen H, Geng D. 2011. [The change of cognition induced by chronic fluoride in rats]. Acta
Academiae Medicinae Xuzhou 31(5): 319-322.
Chen J, Niu Q, Xia T, Zhou G, Li P, Zhao Q, Xu C, Dong L, Zhang S, Wang A. 2018. ERK1/2-
mediated disruption of BDNF-TrkB signaling causes synaptic impairment contributing to
fluoride-induced developmental neurotoxicity. Toxicology 410: 222-230.
Chinoy NJ, Shah SD. 2004. Biochemical effects of sodium fluoride and arsenic trioxide toxicity
and their reversal in the brain of mice. Fluoride 37: 80-87.
Chioca LR, Raupp IM, Da Cunha C, Losso EM, Andreatini R. 2008. Subchronic fluoride intake
induces impairment in habituation and active avoidance tasks in rats. Eur J Pharmacol 579: 196-
201.
Chouhan S, Flora SJS. 2008. Effects of fluoride on the tissue oxidative stress and apoptosis in
rats: Biochemical assays supported by IR spectroscopy data. Toxicology 254: 61-67.

C-18
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Chouhan S, Lomash V, Flora SJ. 2010. Fluoride-induced changes in haem biosynthesis pathway,
neurological variables and tissue histopathology of rats. J Appl Toxicol 30: 63-73.
Cui YS, Zhong Q, Li WF, Liu ZH, Wang Y, Hou CC. 2017. [Effects of fluoride exposure on
thyroid hormone level and intelligence in rats]. Chin J Ind Hyg Occup Dis 35: 888-892.
Dabrowska E. 1997. Effect of different fluorine doses on the supraoptic nucleus of the rat. Folia
Histochem Cytobiol 35: 115-116.
Dong Y, Wang Y, Wei N, Guan Z. 2015. [Expression levels of brain muscarinic acetylcholine
receptor in offspring rats of drinking-water borne fluorosis]. Chin J Endemiol 34: 326-330.
Dong YT, Wang Y, Wei N, Zhang QF, Guan ZZ. 2015. Deficit in learning and memory of rats
with chronic fluorosis correlates with the decreased expressions of M1 and M3 muscarinic
acetylcholine receptors. Arch Toxicol 89: 1981-1991.
Dong YT, Wei N, Qi XL, Liu XH, Chen D, Zeng XX, Guan ZZ. 2017. Attenuating effect of
vitamin E on the deficit of learning and memory of rats with chronic fluorosis: The mechanism
may involve muscarinic acetylcholine receptors. Fluoride 50: 354-364.
Dong YW, Y. Wei, N. Guan, Z. 2015. [Expression of muscarinic acetylcholine receptors in the
brain of rats with chronic fluorosis]. Chin J Endemiol 34(2): 84-88.
Ekambaram P, Paul V. 2001. Calcium preventing locomotor behavioral and dental toxicities of
fluoride by decreasing serum fluoride level in rats. Environ Toxicol Pharmacol 9: 141-146.
Ekambaram P, Paul V. 2002. Modulation of fluoride toxicity in rats by calcium carbonate and by
withdrawal of fluoride exposure. Pharmacol Toxicol 90: 53-58.
Ekambaram P, Paul V. 2003. Effect of vitamin D on chronic behavioral and dental toxicities of
sodium fluoride in rats. Fluoride 36: 189-197.
El-lethey HS, Kamel MM, Shaheed IB. 2010. Neurobehavioral toxicity produced by sodium
fluoride in drinking water of laboratory rats. J Am Sci 6(5): 54-63.
El-lethey HS, Kamel MM. 2011. Effects of black tea in mitigation of sodium fluoride potency to
suppress motor activity and coordination in laboratory rats. J Am Sci 7(4): 243-254.
El-lethey HS, Shaheed IB. 2011. Potential health impact of black tea against Na-F-induced
alterations in territorial aggression, sexual behaviour and fertility of male rats. Life Sci J 8: 828-
839.
Elliott L. 1967. Lack of effect of administration of fluoride on the central nervous system of rats.
Acta Pharmacol Toxicol (Copenh) 25: 323-328.
Flace P, Benagiano V, Vermesan D, Sabatini R, Inchingolo AM, Auteri P, Ambrosi G, Tarullo
A, Cagiano R. 2010. Effects of developmental fluoride exposure on rat ultrasonic vocalization,
acoustic startle reflex and pre-pulse inhibition. Eur Rev Med Pharmacol Sci 14: 507-512.
Gabovich RD. 1962. [On the problem of the effect of fluorine in drinking water on the functional
state of the central nervous system]. Gig Sanit 27: 10-12.

C-19
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Gao Q, Liu YJ, Wu CX, Long YG, Guan ZZ. 2008. [Level of oxidative stress in rat brains and
learning and memory function of rats with chronic fluorosis]. Chin J Endemiol 27: 371-373.
Gao Q, Liu YJ, Wu CX, Long YG, Guan ZZ. 2008. [Effects of fluoride on learning and memory
and cholinesterase activity in rat brains]. Chin J Endemiol 27: 128-130.
Gao Q, Liu YJ, Guan ZZ. 2009. Decreased learning and memory ability in rats with fluorosis:
Increased oxidative stress and reduced cholinesterase activity in the brain. Fluoride 42: 277-285.
Gao Y, Liu L, Young L, Huan L, Jin H. 2009. Effects of learning and memory of fluoride and
the antagonism of selenium in rats. Studies of Trace Elements and Health 26(2): 1-3.
Ge QD, Tan Y, Luo Y, Wang WJ, Zhang H, Xie C. 2018. MiR-132, miR-204 and BDNF-TrkB
signaling pathway may be involved in spatial learning and memory impairment of the offspring
rats caused by fluorine and aluminum exposure during the embryonic stage and into adulthood.
Environ Toxicol Pharmacol 63: 60-68.
Ge Y, Chen L, Yin Z, Song X, Ruan T, Hua L, Liu J, Wang J, Ning H. 2018. Fluoride-induced
alterations of synapse-related proteins in the cerebral cortex of ICR offspring mouse brain.
Chemosphere 201: 874-883.
Gopal K, Saxena R, Gupta GSD, Rana MD, Agrawal D. 2006. Fluoride induced alterations in
neurobehavioural and cardiovascular responses in rats. J Adv Zool 27: 1-7.
Gui CZ, Ran LY, Wu CX, Long YG, He J, Zhang H, Guan ZZ. 2009. [Changes in learning and
memory ability and brain cholinesterase activity in the rats with coal burning fluorosis]. Chin J
Endemiol 28: 497-500.
Gui CZ, Ran LY, Li JP, Guan ZZ. 2010. Changes of learning and memory ability and brain
nicotinic receptors of rat offspring with coal burning fluorosis. Neurotoxicol Teratol 32: 536-
541.
Gui CZ, Ran LY, Guan ZZ. 2011. [Expression levels of brain nicotinic acetylcholine receptor
mRNA and protein in coal-burning type of fluorosis rats]. Chin J Endemiol 30: 239-242.
Guner S, Uyar-Bozkurt S, Haznedaroglu E, Mentes A. 2016. Dental fluorosis and catalase
immunoreactivity of the brain tissues in rats exposed to high fluoride pre- and postnatally. Biol
Trace Elem Res 174: 150-157.
Han H, Du W, Zhou B, Zhang W, Xu G, Niu R, Sun Z. 2014. Effects of chronic fluoride
exposure on object recognition memory and mRNA expression of SNARE complex in
hippocampus of male mice. Biol Trace Elem Res 158: 58-64.
Hong JH, Ge YM, Ning HM, Wang JD. 2005. [Effects of High Fluoride and Low Iodine on
Learning-Memory and TchE of Brain in Offspring Rats]. Chin Prev Med 6: 489-491.
Inkielewicz I, Krechniak J. 2004. Fluoride effects on glutathione peroxidase and lipid
peroxidation in rats. Fluoride 37: 7-12.
Jain A, Mehta VK, Chittora RA, Mahdi A, Bhatnagar M. 2015. Melatonin ameliorates fluoride
induced neurotoxicity in young rats: An in vivo evidence. Asian J Pharm Clin Res 8: 164-167.

C-20
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Jetti R, Raghuveer CV, Mallikarjuna RC. 2016. Protective effect of ascorbic acid and Ginkgo
biloba against learning and memory deficits caused by fluoride. Toxicol Ind Health 32: 183-187.
Jia B, Zong L, Lee JY, Lei J, Zhu Y, Xie H, Clemens JL, Feller MC, Na Q, Dong J, McLane
MW, Jones-Beatty K, Burd I. 2019. Maternal supplementation of low dose fluoride alleviates
adverse perinatal outcomes following exposure to intrauterine inflammation. Sci Rep 9(1): 2575.
Jiang C, Zhang S, Liu H, Guan Z, Zeng Q, Zhang C, Lei R, Xia T, Wang Z, Yang L, Chen Y,
Wu X, Zhang X, Cui Y, Yu L, Wang A. 2014. Low glucose utilization and neurodegenerative
changes caused by sodium fluoride exposure in rat's developmental brain. Neuromolecular Med
16: 94-105.
Jiang S, Su J, Yao S, Zhang Y, Cao F, Wang F, Wang H, Li J, Xi S. 2014. Fluoride and arsenic
exposure impairs learning and memory and decreases mGluR5 expression in the hippocampus
and cortex in rats. PLoS One 9(4): e96041.
Khan AM, Raina R, Dubey N, Verma PK. 2017. Effect of deltamethrin and fluoride co-exposure
on the brain antioxidant status and cholinesterase activity in Wistar rats. Drug Chem Toxicol
41(2): 1-5.
Kinawy AA, Al-Eidan AA. 2018. Impact of prenatal and postnatal treatment of sodium fluoride
and aluminum chloride on some hormonal and sensorimotor aspects in rats. Biol Trace Elem
Res: 1-8.
Kivrak Y. 2012. Effects of fluoride on anxiety and depression in mice. Fluoride 45: 302-306.
Li M, Cui J, Gao YH, Zhang W, Sun LY, Liu XN, Liu Y, Sun DJ. 2015. Pathological changes
and effect on the learning and memory ability in rats exposed to fluoride and aluminum. Toxicol
Res 4: 1366-1373.
Li X, Zhang J, Niu R, Manthari RK, Yang K, Wang J. 2019. Effect of fluoride exposure on
anxiety- and depression-like behavior in mouse. Chemosphere 215: 454-460.
Liu F, Ma J, Zhang H, Liu P, Liu YP, Xing B, Dang YH. 2014. Fluoride exposure during
development affects both cognition and emotion in mice. Physiol Behav 124: 1-7.
Liu WX. 1989. [Experimental study of behavior and cerebral morphology of rat pups generated
by fluorotic female rat]. Chin J Pathol 18: 290-292.
Liu YJ, Gao Q, Wu CX, Long YG, Guan ZZ. 2009. [Modified expression of extracellular signal-
regulated protein kinase signal transduction in rat brains and changed capacity of learning and
memory of rats with chronic fluorosis]. Chin J Endemiol 28: 32-35.
Liu YJ, Gao Q, Wu CX, Guan ZZ. 2010. Alterations of nAChRs and ERK1/2 in the brains of
rats with chronic fluorosis and their connections with the decreased capacity of learning and
memory. Toxicol Lett 192: 324-329.
Liu YJ, Gao Q, Long YG, Yu YN, Guan ZZ. 2011. [Influence of chronic fluorosis on expression
of phospho-Elk-1 in rat brains]. Chin J Endemiol 30: 251-255.

C-21
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Lou DD, Guan ZZ, Liu YJ, Liu YF, Zhang KL, Pan JG, Pei JJ. 2013. The influence of chronic
fluorosis on mitochondrial dynamics morphology and distribution in cortical neurons of the rat
brain. Arch Toxicol 87: 449-457.
Lu F, Zhang Y, Trivedi A, Jiang X, Chandra D, Zheng J, Nakano Y, Uyghurturk DA, Jalai R,
Guner S, Mentes A, DenBesten PK. 2019. Fluoride related changes in behavioral outcomes may
relate to increased serotonin. Physiol Behav 206: 76-83.
Ma J, Liu F, Liu P, Dong YY, Chu Z, Hou TZ, Dang YH. 2015. Impact of early developmental
fluoride exposure on the peripheral pain sensitivity in mice. Int J Dev Neurosci 47: 165-171.
Manusha S, Sudhakar K, Reddy KP. 2019. Protective effects of allium sativum extract against
sodium fluoride induced neurotoxicity. Int J Pharm Sci Res 10(2): 625-633.
McPherson CA, Zhang G, Gilliam R, Brar SS, Wilson R, Brix A, Picut C, Harry GJ. 2018. An
evaluation of neurotoxicity following fluoride exposure from gestational through adult ages in
Long-Evans hooded rats. Neurotoxicol Res: 1-18.
Mesram N, Nagapuri K, Banala RR, Nalagoni CR, Karnati PR. 2016. Quercetin treatment
against NaF induced oxidative stress related neuronal and learning changes in developing rats. J
King Saud Univ Sci 29: 221-229.
Mullenix PJ, Denbesten PK, Schunior A, Kernan WJ. 1995. Neurotoxicity of sodium fluoride in
rats. Neurotoxicol Teratol 17: 169-177.
Nageshwar M, Sudhakar K, Reddy NCC, Reddy KP. 2017. Neuroprotective effects of curcumin
on sodium fluoride induced behavioural and enzymatic changes in brain and muscles of rat. J
Environ Biol 38: 675-681.
Nageshwar M, Sudhakar K, Reddy KP. 2018. Quercetin ameliorates oxidative stress, neural
damage of brain and behavioral impairment of rat with fluoride exposure. Int J Pharm Sci Res
9(8): 3247-3256.
Nian W, Wang X, Shao D, Yu Q, Ouyang W, Zhang Z, Ruan Q. 2018. Effects of subchronic
exposure to fluorine on hippocampal injury in mice and its molecular mechanism. Acta Sci
Circumst 38(11): 4512-4519.
Niu Q, Chen J, Xia T, Li P, Zhou G, Xu C, Zhao Q, Dong L, Zhang S, Wang A. 2018. Excessive
ER stress and the resulting autophagic flux dysfunction contribute to fluoride-induced
neurotoxicity. Environ Pollut 233: 889-899.
Niu R, Sun Z, Wang J, Cheng Z. 2008. Effects of fluoride and lead on locomotor behavior and
expression of nissl body in brain of adult rats. Fluoride 41: 276-282.
Niu R, Sun Z, Cheng Z, Li Z, Wang J. 2009. Decreased learning ability and low hippocampus
glutamate in offspring rats exposed to fluoride and lead. Environ Toxicol Pharmacol 28: 254-
258.
Niu R, Liu S, Wang J, Zhang J, Sun Z. 2014. Proteomic analysis of hippocampus in offspring
male mice exposed to fluoride and lead. Biol Trace Elem Res 162: 227-233.

C-22
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Niu R, Xue X, Zhao Y, Sun Z, Yan X, Li X, Feng C, Wang J. 2015. Effects of fluoride on
microtubule ultrastructure and expression of Tubalpha1a and Tubbeta2a in mouse hippocampus.
Chemosphere 139: 422-427.
Niu R, Chen H, Manthari RK, Sun Z, Wang J, Zhang J, Wang J. 2018. Effects of fluoride on
synapse morphology and myelin damage in mouse hippocampus. Chemosphere 194: 628-633.
Nkpaa KW, Onyeso GI. 2018. Rutin attenuates neurobehavioral deficits, oxidative stress, neuro-
inflammation and apoptosis in fluoride treated rats. Neurosci Lett 682: 92-99.
Paul V, Ekambaram P, Jayakumar AR. 1998. Effects of sodium fluoride on locomotor behavior
and a few biochemical parameters in rats. Environ Toxicol Pharmacol 6: 187-191.
Pereira M, Dombrowski PA, Losso EM, Chioca LR, Da Cunha C, Andreatini R. 2011. Memory
impairment induced by sodium fluoride is associated with changes in brain monoamine levels.
Neurotoxicol Res 19: 55-62.
Pulungan ZS, Sofro ZM, Partadiredja G. 2016. Sodium fluoride does not affect the working
memory and number of pyramidal cells in rat medial prefrontal cortex. Anat Sci Int 93(1): 128-
138.
Raghu J, Raghuveer VC, Rao MC, Somayaji NS, Babu PB. 2013. The ameliorative effect of
ascorbic acid and Ginkgo biloba on learning and memory deficits associated with fluoride
exposure. Interdiscip Toxicol 6: 217-221.
Raju S, Sivanesan SK, Gudemalla K. 2019. Cognitive enhancement effect of Ginkgo biloba
extract on memory and learning impairments induced by fluoride neurotoxicity. Int J Res Pharm
Sci 10(1): 129-134.
Reddy MM, Karnati PR. 2015. Protective effects of aqueous extract of fruit pulp of tamarindus
indica on motor activity and metabolism of the gastrocnemius muscle of rats treated with
fluoride. Int J Toxicol Pharmacol Res 7: 241-246.
Reddy YP, Tiwari SK, Shaik AP, Alsaeed A, Sultana A, Reddy PK. 2014. Effect of sodium
fluoride on neuroimmunological parameters, oxidative stress and antioxidative defenses. Toxicol
Mech Methods 24: 31-36.
Rumiantsev GI, Novikov SM, Mel'nikova NN, Levchenko NI, Kozeeva EE. 1988. [Experimental
study of the biological effect of salts of hydrofluosilicic acid]. Gig Sanit: 80-82.
Sarkozi K, Horvath E, Vezer T, Papp A, Paulik E. 2015. Behavioral and general effects of
subacute oral arsenic exposure in rats with and without fluoride. Int J Environ Health Res 25:
418-431.
Shah SD, Chinoy NJ. 2004. Adverse effects of fluoride and/or arsenic on the cerebral
hemisphere of mice and recovery by some antidotes. Fluoride 37: 162-171.
Shalini B, Sharma JD. 2015. Beneficial effects of emblica officinalis on fluoride-induced toxicity
on brain biochemical indexes and learning-memory in rats. Toxicol Int 22: 35-39.

C-23
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Shan KR, Qi XL, Long YG, Nordberg A, Guan ZZ. 2004. Decreased nicotinic receptors in PC12
cells and rat brains influenced by fluoride toxicity: A mechanism relating to a damage at the
level in post-transcription of the receptor genes. Toxicology 200: 169-177.
Sharma C, Suhalka P, Bhatnagar M. 2018. Curcumin and resveratrol rescue cortical-
hippocampal system from chronic fluoride-induced neurodegeneration and enhance memory
retrieval. Int J Neurosci: 1-15.
Shen X, Zhang Z, Xu X. 2004. [Effect of iodine and selenium on learning memory impairment
induced by fluorosis and blood biochemical criterion of rats]. Occupation and Health 20(1): 6-8.
Sudhakar K, Nageshwar M, Pratap Reddy K. 2017. Seed extract of Abelmoschus moschatus
medik reverses NAF-induced behavioral changes through neurodegeneration and oxidative stress
in brain of rat. Asian J Pharm Clin Res 10: 165-171.
Sudhakar K, Nageshwar M, Reddy KP. 2018. Protective effect of okra, Abelmoschus moschatus
seed extract on developing brain of rats during pre- and post-natal fluoride exposure. Int J Pharm
Sci Res 9: 1519-1528.
Sudhakar K, Nageshwar M, Reddy KP. 2018. Abelmoschus moschatus extract reverses altered
pain and neurohistology of a rat with developmental exposure of fluoride. J Appl Pharm Sci 8(6):
94-104.
Sudhakar K, Reddy KP. 2018. Protective effects of Abelmoschus moschatus seed extract on
neurotransmitter system of developing brain of Wistar rats with gestational and post-natal
exposure of sodium fluoride. Int J Green Pharm 12: S131-S139.
Sun ZR, Liu FZ, Wu LN, Lu Y, Yu DK. 2000. [Effects of high fluoride drinking water on the
cerebral function of mice]. Chin J Epidemiol 19: 262-263.
Sun ZR, Liu FZ, Wu LN, Lu Y, Yu DK. 2008. Effects of high fluoride drinking water on the
cerebral function of mice. Fluoride 41: 148-151.
Sun Z, Zhang Y, Xue X, Niu R, Wang J. 2018. Maternal fluoride exposure during gestation and
lactation decreased learning and memory ability, and glutamate receptor mRNA expressions of
mouse pups. Hum Exp Toxicol 37: 87-93.
Trivedi MH, Verma RJ, Chinoy NJ. 2007. Amelioration by black tea of sodium fluoride-induced
changes in protein content of cerebral hemisphere, cerebellum and medulla oblongata in brain
region of mice. Acta Poloniae Pharm 64: 221-225.
Trivedi MH, Verma RJ, Chinoy NJ. 2009. Mitigation of sodium fluoride induced toxicity in mice
brain by black tea infusion. Fluoride 42: 29-33.
Trivedi MH, Verma RJ, Sangai NP, Chinoy NJ. 2011. Black tea extract mitigation of NaF-
induced lipid peroxidation in different regions of mice brains. Fluoride 44: 243-254.
Trivedi MH, Verma RJ, Sangai NP, Chinoy NJ. 2012. Mitigation by black tea extract of sodium
fluoride induced histopathological changes in brain of mice. Fluoride 45: 13-26.

C-24
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Tsunoda M, Aizawa Y, Nakano K, Liu Y, Horiuchi T, Itai K, Tsunoda H. 2005. Changes in

fluoride levels in the liver, kidney, and brain and in neurotransmitters of mice after subacute
administration of fluoride. Fluoride 38: 284-292.
Varner JA, Jensen KF, Horvath W, Isaacson RL. 1998. Chronic administration of aluminum-
fluoride or sodium-fluoride to rats in drinking water: Alterations in neuronal and cerebrovascular
integrity. Brain Res 784(1-2): 284-298.
Verma RJ, Trivedi MH, Chinoy NJ. 2007. Black tea amelioration of sodium fluoride-induced
alterations of DNA, RNA, and protein contents in the cerebral hemisphere, cerebellum, and
medulla oblongata regions of mouse brain. Fluoride 40: 7-12.
Wang G, Li J, Zhu H, Zhu J. 2006. Effect of different doses of chronic exposure of fluoride on
rat learning and memory behavior. Studies of Trace Elements and Health 23(2): 1-2.
Wang JD, Ge YM, Ning HM, Wang SL. 2004. Effects of high fluoride and low iodine on
biochemical indexes of the brain and learning-memory of offspring rats. Fluoride 37: 201-208.
Wang J, Zhang Y, Guo Z, Li R, Xue X, Sun Z, Niu R. 2018. Effects of perinatal fluoride
exposure on the expressions of miR-124 and miR-132 in hippocampus of mouse pups.
Chemosphere 197: 117-122.
Wei N, Dong Y, Wang Y, Guan Z. 2014. [Effects of chronic fluorosis on neurobehavioral
development in offspring of rats and antagonistic effect of vitamin E]. Chin J Endemiol 33: 125-
128.
Whitford GM, Whitford JL, Hobbs SH. 2009. Appetitive-based learning in rats: Lack of effect of
chronic exposure to fluoride. Neurotoxicol Teratol 31: 210-215.
Wu CX, Gu XL, Ge YM, Zhang JH, Wang JD. 2006. Effects of high fluoride and arsenic on
brain biochemical indexes and learning-memory in rats. Fluoride 39: 274-279.
Wu NP, Zhao ZL, Gao WH, Li XQ. 1995. [Behavioral teratology in rats exposed to fluoride.]
Chin J Endemiol 12(5): 271-273.
Wu NP, Zhao ZL, Gao WH, Li XQ. 2008. Behavioral teratology in rats exposed to fluoride.
Fluoride 41: 129-133.
Xu X, Shen X, Zhang Z. 2001. Effect of fluorosis on mice learning and memory behaviors and
brain SOD activity and MDA content China Public Health 17(1): 8-10.
Yang L, Jin P, Wang X, Zhou Q, Lin X, Xi S. 2018. Fluoride activates microglia, secretes
inflammatory factors and influences synaptic neuron plasticity in the hippocampus of rats.
Neurotox 69: 108-120.
Yu Q, Shao D, Zhang R, Ouyang W, Zhang Z. 2019. Effects of drinking water fluorosis on L-
type calcium channel of hippocampal neurons in mice. Chemosphere 220: 169-175.
Yuan J, Li Q, Niu R, Wang J. 2019. Fluoride exposure decreased learning ability and the
expressions of the insulin receptor in male mouse hippocampus and olfactory bulb. Chemosphere
224: 71-76.

C-25
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Zhang C, Ren C, Chen H, Geng R, Fan H, Zhao H, Guo K, Geng D. 2013. The analog of Ginkgo
biloba extract 761 is a protective factor of cognitive impairment induced by chronic fluorosis.
Biol Trace Elem Res 153: 229-236.
Zhang KL, Lou DD, Guan ZZ. 2015. Activation of the AGE/RAGE system in the brains of rats
and in SH-SY5Y cells exposed to high level of fluoride might connect to oxidative stress.
Neurotoxicol Teratol 48: 49-55.
Zhang J, Zhu W, Zhang Z. 2009. [The effect of fluorine exposure of pregnant rats on the learning
and memory capabilities of baby rats]. Chinese Journal of Public Health 25(11): 1347-1348.
Zhang J, Zhu WJ, Xu XH, Zhang ZG. 2011. Effect of fluoride on calcium ion concentration and
expression of nuclear transcription factor kappa-B rho65 in rat hippocampus. Exp Toxicol Pathol
63: 407-411.
Zhang J, Zhang Z. 2013. Effects of chronic fluorosis on camkiiα, c-FOS, BAX, and BCL-2
channel signaling in the hippocampus of rats. Fluoride 46: 135-141.
Zhang Z, Shen X, Xu X. 2001. [Effects of selenium on the damage of learning-memory ability of
mice induced by fluoride]. J Hyg Res 30: 144-146.
Zhang Z, Xu X, Shen X, Xua XH. 1999. [Effect of fluoride exposure on synaptic structure of
brain areas related to learning-memory in mice]. J Hyg Res 28(4): 210-212.
Zhang Z, Xu X, Shen X, Xua XH. 2008. Effect of fluoride exposure on synaptic structure of
brain areas related to learning-memory in mice. Fluoride 41: 139-143.
Zhao Q, Niu Q, Chen J, Xia T, Zhou G, Li P, Dong L, Xu C, Tian Z, Luo C, Liu L, Zhang S,
Wang A. 2019. Roles of mitochondrial fission inhibition in developmental fluoride
neurotoxicity: Mechanisms of action in vitro and associations with cognition in rats and children.
Arch Toxicol 93(3): 709-726.
Zhao XL, Wu JH. 1998. Actions of sodium fluoride on acetylcholinesterase activities in rats.
Biomed Environ Sci 11: 1-6.
Zheng X, Sun Y, Ke L, Ouyang W, Zhang Z. 2016. Molecular mechanism of brain impairment
caused by drinking-acquired fluorosis and selenium intervention. Environ Toxicol Pharmacol 43:
134-139.
Zhu W, Zhang J, Zhang Z. 2011. Effects of fluoride on synaptic membrane fluidity and PSD-95
expression level in rat hippocampus. Biol Trace Elem Res 139: 197-203.
Zhu YL, Zheng YJ, LV XM, Ma Y, Zhang J. 2012. Effects of fluoride exposure on performance
in water labyrinth and monoamine neurotransmitters of rats. Journal of Xinjiang Medical
University 3: 014.
Zhu YP, Xi SH, Li MY, Ding TT, Liu N, Cao FY, Zeng Y, Liu XJ, Tong JW, Jiang SF. 2017.
Fluoride and arsenic exposure affects spatial memory and activates the ERK/CREB signaling
pathway in offspring rats. Neurotox 59: 56-64.

C-26
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

C.2.2.2. Studies Not Available in HAWC

Abdelaleem MM, El-Tahawy NFG, Abozaid SMM, Abdel-Hakim SA. 2018. Possible protective
effect of curcumin on the thyroid gland changes induced by sodium fluoride in albino rats: Light
and electron microscopic study. Endocr Regul 52: 59-68.
Abd-Elhakim YM, Mohammed AT, Ali HA. 2018. Impact of subchronic exposure to triclosan
and/or fluoride on estrogenic activity in immature female rats: The expression pattern of
calbindin-D9k and estrogen receptor alpha genes. J Biochem Mol Toxicol 32(2): 22027.
Abdumajidov OR. 2004. [Sex differences in lipid peroxidation and antioxidant defense of the
brain tissue in intoxication with low doses of inorganic compounds]. Uzbekiston Tibbiet
Zhurnali: 58-60.
Adebayo OL, Shallie PD, Salau BA, Ajani EO, Adenuga GA. 2013. Comparative study on the
influence of fluoride on lipid peroxidation and antioxidants levels in the different brain regions
of well-fed and protein undernourished rats. J Trace Elem Med Biol 27: 370-374.
Adedara IA, Ojuade TJD, Olabiyi BF, Idris UF, Onibiyo EM, Ajeigbe OF, Farombi EO. 2016.
Taurine ameliorates renal oxidative damage and thyroid dysfunction in rats chronically exposed
to fluoride. Biol Trace Elem Res: 1-8.
Ahmed SK, Kalleny NK, Attia AAEM, Elkateb LA. 2015. The possible protective role of
chromium chloride against sodium fluoride-induced changes in the structure of the cerebellar
cortex of the adult male albino rat. Egypt J Histol 38: 402-414.
Al Badawi MH, Mahmoud OM, Salem NA. 2016. Therapeutic potential of omega-3 against
sodium fluoride toxicity on the cerebellar cortex of adult male albino rats: Histological and
immunohistochemical study. Egypt J Histol 39: 170-178.
Alhayani A, Elshal EB, Aal IHA, Al-Shammeri E, Kabra H. 2013. Does vitamin E protect
against sodium fluoride toxicity on the cerebellar cortex of albino rats? Middle East J Sci Res 16:
1019-1026.
Ameeramja J, Raghunath A, Perumal E. 2018. Tamarind seed coat extract restores fluoride-
induced hematological and biochemical alterations in rats. Environ Sci Pollut Res Int 25(26):
26157-26166.
Antonyan OA. 1980. [Lipid per oxidation in fluorosis and the protective role of dietary factors].
Zh Eksp Klin Med 20: 381-388.
Ardelean I, Racoveanu N, Manescu S, Lupulescu A, Diaconescu M, Ghelerter L. 1964.
Experimental investigations concerning the effect of fluorine on the thyroid gland. Rum Med Rev
8: 17-20.
Atmaca N, Atmaca HT, Kanici A, Anteplioglu T. 2014. Protective effect of resveratrol on
sodium fluoride-induced oxidative stress, hepatotoxicity and neurotoxicity in rats. Food Chem
Toxicol 70: 191-197.
Auskaps AM, Shaw JH. 1955. Hemoglobin concentration, thyroid weight and growth rate in rats
during minimum fluoride ingestion. J Nutr 55: 611-621.

C-27
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Bagmut I, Kolisnyk I, Titkova A, Petrenko T, Filipchenko S. 2018. Content of catecholamines in

blood serum of rats under fluoride intoxication. Georgian Med News (280-281): 125-129.
Bakalyan PH, Antonyan OA. 1981. [Effect of fluorosis on glutathione peroxidase and
glutathione reductase activities and sulfhydryl groups]. Zh Eksp Klin Med 21: 10-14.
Basha PM, Madhusudhan N. 2010. Pre and post natal exposure of fluoride induced oxidative
macromolecular alterations in developing central nervous system of rat and amelioration by
antioxidants. Neurochem Res 35: 1017-1028.
Basha PM, Madhusudhan N. 2011. Effect of maternal exposure of fluoride on oxidative stress
markers and amelioration by selected antioxidants in developing central nervous system of rats.
Biologia 66: 187-193.
Basha PM, Rai P, Begum S. 2011. Evaluation of fluoride-induced oxidative stress in rat brain: A
multigeneration study. Biol Trace Elem Res 142: 623-637.
Basha PM, Sujitha NS. 2012. Combined influence of intermittent exercise and temperature stress
on the modulation of fluoride toxicity. Biol Trace Elem Res 148: 69-75.
Basha PM, Saumya SM. 2013. Suppression of mitochondrial oxidative phosphorylation and
TCA enzymes in discrete brain regions of mice exposed to high fluoride: Amelioration by panax
ginseng (ginseng) and lagerstroemia speciosa (banaba) extracts. Cell Mol Neurobiol 33: 453-464.
Basha MP, Begum S, Madhusudhan N. 2014. Antioxidants in the management of fluoride
induced neural oxidative stress in developing rats. Int J Pharm Sci Res 5: 201-206.
Benetato G, Giuran AM, Cirmaciu R, Cirje M, Petrescu A, Vacariu A. 1970. [Effect of fluorine
in drinking water on the metabolism of Ca and Mg and on neuromuscular excitability:
Experimental studies and clinical observations]. Rev Roum Physiol 7: 335-352.
Bharti VK, Srivastava RS. 2009. Fluoride-induced oxidative stress in rat's brain and its
amelioration by buffalo (Bubalus bubalis) pineal proteins and melatonin. Biol Trace Elem Res
130: 131-140.
Bhatnagar M, Rao P, Saxena A, Bhatnagar R, Meena P, Barbar S, Chouhan A, Vimal S. 2006.
Biochemical changes in brain and other tissues of young adult female mice from fluoride in their
drinking water. Fluoride 39: 280-284.
Bilgili A, Akdogan M, Yildiz M, Eraslan G, Cetin N. 2004. The effects of fluoride on thyroid
hormones in rabbits. Indian Vet J 81: 986-988.
Bobek S, Kahl S, Ewy Z. 1976. Effect of long-term fluoride administration on thyroid hormones
level blood in rats. Endocrinol Exp 10: 289-295.
Bouaziz H, Ammar E, Ghorbel H, Ketata S, Jamoussi K, Ayadi F, Guermazi F, Zeghal N. 2004.
Effect of fluoride ingested by lactating mice on the thyroid function and bone maturation of their
suckling pups. Fluoride 37: 133-142.
Bouaziz H, Soussia L, Guermazi F, Zeghal N. 2005. Fluoride-induced thyroid proliferative
changes and their reversal in female mice and their pups. Fluoride 38: 185-192.

C-28
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Bouaziz HB, Amara I, Essefi M, Croute F, Zeghal N. 2010. Fluoride-induced brain damages in
suckling mice. Pestic Biochem Physiol 96: 24-29.
Chauhan SS, Ojha S, Mahmood A. 2013. Effects of fluoride and ethanol administration on lipid
peroxidation systems in rat brain. Indian J Exp Biol 51: 249-255.
Chen J, Chen X, Yang K, Xia T, Xie H. 2002. [Studies on DNA damage and apoptosis in rat
brain induced by fluoride]. Chin J Prev Med 36: 222-224.
Chirumari K, Reddy PK. 2007. Dose-dependent effects of fluoride on neurochemical milieu in
the hippocampus and neocortex of rat brain. Fluoride 40: 101-110.
Chouhan S, Yadav A, Kushwah P, Kaul RK, Flora SJS. 2011. Silymarin and quercetin abrogates
fluoride induced oxidative stress and toxic effects in rats. Mol Cell Toxicol 7: 25-32.
Clay AB, Suttie JW. 1987. Effect of dietary fluoride on dairy cattle: Growth of young heifers. J
Dairy Sci 70: 1241-1251.
Czechowicz K, Osada A, Slesak B. 1974. Histochemical studies on the effect of sodium fluoride
on metabolism in Purkinje's cells. Folia Histochem Cytochem 12: 37-44.
Demole V, Lerch P. 1956. [Normality of fixation of radioactive iodine in the thyroid of rats
during experimental fluorosis]. Helv Physiol Pharmacol Acta 14(4): 62-63.
Dhurvey V, Patil V, Thakare M. 2017. Effect of sodium fluoride on the structure and function of
the thyroid and ovary in albino rats (rattus norvegicus). Fluoride 50: 235-246.
Domzalska E. 1966. [Influence of sodium fluoride on hypophysis, thyroid gland, parathyroid,
and adrenal gland in the white rat]. Czas Stomatol 19: 839-844.
El-lethey HS, Kamel MM, Shaheed IB. 2011. Perinatal exposure to sodium fluoride with
emphasis on territorial aggression, sexual behaviour and fertility in male rats. Life Sci J 8: 686-
694.
Flora SJS, Mittal M, Mishra D. 2009. Co-exposure to arsenic and fluoride on oxidative stress,
glutathione linked enzymes, biogenic amines and DNA damage in mouse brain. J Neurol Sci
285: 198-205.
Flora SJS, Mittal M, Pachauri V, Dwivedi N. 2012. A possible mechanism for combined arsenic
and fluoride induced cellular and DNA damage in mice. Metallomics 4: 78-90.
Gabovich RD, Verzhikovskaia NV. 1958. [Effect of fluoride compounds on absorption of
radioactive iodine by the thyroid gland in humans and in experimental conditions]. Probl
Endokrinol Gormonoter 4: 49-54.
Galamini-Ligori M, Di Blasi F. 1961. [Action of sodium fluoride on the thyroid of
hypophysectomized rats]. Boll Soc Ital Biol Sper 37: 1503-1506.
Galletti P, Held HR, Korrodi H, Wegmann T. 1956. [Investigations on the possible damaging
effect of fluorine on the thyroid gland]. Helv Med Acta 23: 601-605.

C-29
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Ge Y, Ning H, Feng C, Wang H, Yan X, Wang S, Wang J. 2006. Apoptosis in brain cells of
offspring rats exposed to high fluoride and low iodine. Fluoride 39: 173-178.
Ge Y, Niu R, Zhang J, Wang J. 2011. Proteomic analysis of brain proteins of rats exposed to
high fluoride and low iodine. Arch Toxicol 85: 27-33.
Ge YM, Ning HM, Wang SL, Wang JD. 2005. DNA damage in thyroid gland cells of rats
exposed to long-term intake of high fluoride and low iodine. Fluoride 38: 318-323.
Ge YM, Ning HM, Wang SL, Wang JD. 2005. Comet assay of DNA damage in brain cells of
adult rats exposed to high fluoride and low iodine. Fluoride 38: 209-214.
Ge YM, Ning HM, Wang SL, Wang JD. 2005. Effects of high fluoride and low iodine on brain
histopathology in offspring rats. Fluoride 38: 127-132.
Ge YM, Ning HM, Gu XL, Yin M, Yang XF, Qi YH, Wang JD. 2013. Effects of high fluoride
and low iodine on thyroid function in offspring rats. J Integr Agric 12: 502-508.
Guan ZZ. 1986. [Morphology of the brain of the offspring of rats with chronic fluorosis]. Chin J
Pathol 15: 297-299.
Guan Z, Wang Y, Xiao K. 1997. [Influence of experimental fluorosis on phospholipid content
and fatty acid composition in rat brain]. Chin Med J 77: 592-596.
Guan Z-Z, Wang Y-N, Xiao K-Q, Dai D-Y, Chen Y-H, Liu J-L, Sindelar P, Dallner G. 1998.
Influence of chronic fluorosis on membrane lipids in rat brain. Neurotoxicol Teratol 20: 537-542.
Guan ZZ, Shan KR, Wang YN, Dallner G. 2006. [Changes of lipids and nicotinic receptors in rat
brains and pheochromocytoma with fluorosis]. Chin J Endemiol 25: 121-124.
Gushchin SK. 1951. [Effect of sodium fluoride on iodine metabolism in rabbit tissue organs; on
the etiology of endemic goiter]. Gig Sanit 2: 45-48.
Hamza RZ, Al-Harbi MS. 2014. Sodium fluoride induced neurotoxicity and possible antioxidant
role of selenium and curcumin in male mice. Biosci Biotechnol Res Asia 11: 81-87.
Hamza RZ, El-Shenawy NS, Ismail HAA. 2015. Protective effects of blackberry and quercetin
on sodium fluoride-induced oxidative stress and histological changes in the hepatic, renal, testis
and brain tissue of male rat. J Basic Clin Physiol Pharmacol 26: 237-251.
Haojun Z, Yaoling W, Ke Z, Jin L, Junling W. 2012. Effects of NaF on the expression of
intracellular Ca2+ fluxes and apoptosis and the antagonism of taurine in murine neuron. Toxicol
Mech Methods 22: 305-308.
Hara K. 1980. Studies on fluorosis especially effects of fluoride on thyroid metabolism. J Dent
Health 30: 42-57.
Harris NO, Hayes RL. 1955. A tracer study of the effect of acute and chronic exposure to sodium
fluoride on the thyroid iodine metabolism of rats. J Dent Res 34: 470-477.
Hassan HA, Abdel-Aziz AF. 2010. Evaluation of free radical-scavenging and anti-oxidant
properties of black berry against fluoride toxicity in rats. Food Chem Toxicol 48: 1999-2004.

C-30
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Hoogstratten B, Leone NCLG, Shupe J, Greenwood DA, Lieberman J. 1965. Effect of fluorides
on hematopoietic system, liver, and thyroid gland in cattle. J Amer Med Assoc 192: 26-32.
Inkielewicz I, Rogowska M, Krechniak J. 2006. Lipid peroxidation and antioxidant enzyme
activity in rats exposed to fluoride and ethanol. Fluoride 39: 53-59.
Inkielewicz I, Czarnowski W. 2008. Oxidative stress parameters in rats exposed to fluoride and
aspirin. Fluoride 41: 76-82.
Inkielewicz-Stepniak I, Czarnowski W. 2010. Oxidative stress parameters in rats exposed to
fluoride and caffeine. Food Chem Toxicol 48: 1607-1611.
Jiang P, Li G, Zhou X, Wang C, Qiao Y, Liao D, Shi D. 2018. Chronic fluoride exposure induces
neuronal apoptosis and impairs neurogenesis and synaptic plasticity: Role of GSK-3beta/beta-
catenin pathway. Chemosphere 214: 430-435.
Jiang SF, Xi SH, Yao SQ, Tong JW, Zhang YS, Wang Q, Su J, Li MY. 2013. [Effects of
fluoride, arsenic and co-exposure on expression of Bcl-2 and Bax in hippocampus and cerebral
cortex of rats]. Chin J Endemiol 32: 365-369.
Jiang Y, Guo X, Sun Q, Shan Z, Teng W. 2016. Effects of excess fluoride and iodide on thyroid
function and morphology. Biol Trace Elem Res 170: 382-389.
Jin TX, Guan ZZ, Zhang H. 2011. [The effect of fluoride on α subunit of calcium/calmodulin-
dependent protein kinase-II mRNA and protein expression in central nervous system]. Chin J
Endemiol 30: 247-250.
Jonderko G, Kita K, Pietrzak J, Primus-Slowinska B, Ruranska B, Zylka-Wloszczyk M,
Straszecka J. 1983. [Effect of subchronic sodium fluoride poisoning on the thyroid gland of
rabbits with normal and increased supply of iodine]. Endokrynol Pol 34: 195-203.
Kahl S, Bobek S. 1975. [Effect of fluoride administration on radiothyroxine turnower in rats].
Endokrynol Pol 26: 391-396.
Kahl S, Ewy Z. 1975. Effect of single and long term sodium fluoride administration on
biosynthesis of the thyroid hormone in rats. Fluoride 8: 191-198.
Kapoor V, Prasad T, Paliwal VK. 2001. Blood biochemical constituents in calves following
subclinical levels of fluoride toxicosis. Fluoride 34: 126-131.
Karawya FS, Zahran NM, Azzam EZ. 2015. Is water fluoridation a hidden cause of obesity?
Histological study on thyroid follicular cells of albino rats. Egypt J Histol 38: 547-557.
Kaur T, Bijarnia RK, Nehru B. 2009. Effect of concurrent chronic exposure of fluoride and
aluminum on rat brain. Drug Chem Toxicol 32: 215-221.
Kelimu A, Liu KT, Lian J, Hu HH, Zheng YJ, Wang TM. 2008. [Effects of vitamin C and E on
the ultrastructure in liver, kidney and brain of fluorosis rats]. Chin J Endemiol 27: 378-381.
Kinawy AA. 2019. Synergistic oxidative impact of aluminum chloride and sodium fluoride
exposure during early stages of brain development in the rat. Environ Sci Pollut Res Int 26(11):
10951-10960.

C-31
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Knizhnikov VA. 1959. [Effect of potable water with high fluoride concentration on thyroid
function]. Gig Sanit 24: 20-25.
Knizhnikov VA, Tsypin AB, Shcherbova EN, Bugryshev PF. 1963. [The effect of drinking water
with an increased fluorine content on the bioelectrical activity of the brain and heart under
experimental conditions]. Gig Sanit 28: 16-19.
Kondo T, Yoshida M, Kasahara K. 1976. [Acute fluorosis in female rats: Time of inhibition and
recovery of cholinesterase in serum and salivary glands]. Jpn J Dent Health 26: 187-192.
Kowalewska M. 1974. [Biopotentials of the organ of hearing in chronic sodium fluoride
poisoning]. J Pol Otolaryngol 28: 417-424.
Krechniak J, Inkielewicz I. 2005. Correlations between fluoride concentrations and free radical
parameters in soft tissues of rats. Fluoride 38: 293-296.
Leonard BE. 1972. Effect of phentolamine on the increase in brain glycolysis following the
intraventricular administration of dibutyryl-3,5-cyclic adenosine monophosphate and sodium
fluoride to mice. Biochem Pharmacol 21: 115-117.
Liu G, Zhang W, Jiang P, Li X, Liu C, Chai C. 2012. Role of nitric oxide and vascular
endothelial growth factor in fluoride-induced goitrogenesis in rats. Environ Toxicol Pharmacol
34: 209-217.
Li H, Cai Q, Wang D. 2012. [Effect of fluoride on the expression of rat thyroid peroxidase
mRNA]. Chin J Endemiol 31: 515-517.
Li H, Cai Q, Wang D. 2012. [Effects of fluoride on rat thyroid morphology, thyroid peroxidase
activity and the expression of thyroid peroxidase protein]. Chin J Endemiol 31: 271-274.
Liu H, Hou C, Zeng Q, Zhao L, Cui Y, Yu L, Wang L, Zhao Y, Nie J, Zhang B, Wang A. 2016.
Role of endoplasmic reticulum stress-induced apoptosis in rat thyroid toxicity caused by excess
fluoride and/or iodide. Environ Toxicol Pharmacol 46: 277-285.
Liu YJ, Gao Q, Wu CX, Guan ZZ. 2010. [Changes of the c-Jun N-terminal kinase in the brains
of rats with chronic fluorosis]. Chin J Endemiol 29: 608-612.
Liu YJ, Guan ZZ, Gao Q, Pei JJ. 2011. Increased level of apoptosis in rat brains and SH-SY5Y
cells exposed to excessive fluoride-A mechanism connected with activating JNK
phosphorylation. Toxicol Lett 204: 183-189.
Lohakare J, Pattanaik AK. 2013. Effects of addition of fluorine in diets differing in protein
content on urinary fluoride excretion, clinical chemistry and thyroid hormones in calves.
Brazilian J Anim Sci 42: 751-758.
Long YG, Wang YN, Chen J, Jiang SF, Nordberg A, Guan ZZ. 2002. Chronic fluoride toxicity
decreases the number of nicotinic acetylcholine receptors in rat brain. Neurotoxicol Teratol 24:
751-757.

C-32
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Lou DD, Liu YF, Zhang KL, Yu YN, Guan ZZ. 2011. [Changes of reactive oxygen species level
and mitochondria fission-fusion in cortical neurons of rats with chronic fluorosis]. Chin J
Endemiol 30: 256-260.
Lou DD, Liu YF, Qin SL, Zhang KL, Yu YN, Guan ZZ. 2012. [Changed transcription level of
mitochondrial fission and fusion gene loci in cortical neurons of rats with chronic fluorosis].
Chin J Endemiol 31: 125-129.
Lou DD, Zhang KL, Qin SL, Liu YF, Yu YN, Guan ZZ. 2012. [Alteration of mitochondrial
distribution and gene expression of fission 1 protein in cortical neurons of rats with chronic
fluorosis]. Chin J Pathol 41: 243-247.
Lou DD, Pan JG, Zhang KL, Qin SL, Liu YF, Yu YN, Guan ZZ. 2013. [Changed expression of
mito-fusion 1 and mitochondrial fragmentation in the cortical neurons of rats with chronic
fluorosis]. Chin J Prev Med 47: 170-174.
Lou DD, Zhang KL, Pan JG, Qin SL, Liu YF, Yu YN, Guan ZZ. 2013. [Influence of chronic
fluorosis on the expression of mitochondrial fission protein dynamin-related 1 in the cortical
neurons of rats]. Chin J Prev Med 47: 561-564.
Lou DD, Zhang KL, Qin SL, Liu YF, Liu YJ, Guan ZZ. 2013. [Effects of chronic fluorosis on
4.8 kb mitochondrial DNA in liver, kidney and brain of rats]. Chin J Endemiol 32: 121-124.
Lou DD, Guan ZZ, Pei JJ. 2014. Alterations of apoptosis and expressions of Bax and Bcl-2 in
the cerebral cortices of rats with chronic fluorosis. Fluoride 47: 199-207.
Luo GY, Niu RY, Sun ZL, Zhang JH, Wang JM, Wang C, Wang JD. 2011. Reduction of
CaMKII expression in the hippocampus of rats from ingestion of fluoride and/or lead. Fluoride
44: 63-69.
Ma T, Liu D, Song K. 1999. Cytochemical study of neuron enzyme at anterior horn of spinal
cord in rats with experimental fluorosis. J Chin Med Univ 28: 81-82.
Ma TX, Yu HT, Song KQ. 2008. [Expression of c-fos and Caspase 8 in cerebral cortex of rats
with experimental fluorosis]. Chin J Endemiol 27: 131-133.
Mach Z, Zygulska-Machowa H. 1959. O wplywie fluoru na przemiane J131 [Russian and
English summ.]. Endokrynol Pol 10: 157-162.
Machida H. 1989. [A study on the rabbit thermoregulatory system effects of high dose sodium
fluoride]. Dent Sci Rep 89: 607-626.
Madan J, Puri JP, Singh JK. 2009. Growth, feed efficiency and blood profile of buffalo calves
consuming high levels of fluoride. Trop Anim Health Prod 41: 295-298.
Madhusudhan N, Basha PM, Begum S, Ahmed F. 2009. Fluoride-induced neuronal oxidative
stress and its amelioration by antioxidants in developing rats. Fluoride 42: 179-187.
Madhusudhan N, Basha PM, Rai P, Ahmed F, Prasad GR. 2010. Effect of maternal fluoride
exposure on developing CNS of rats: Protective role of Aloe vera, Curcuma longa and Ocimum
sanctum. Indian J Exp Biol 48: 830-836.

C-33
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Manocha SL, Warner H, Olkowski ZL. 1975. Cytochemical response of kidney, liver and
nervous system of fluoride ions in drinking water. Histochem J 7: 343-355.
Mansour HH, Tawfik SS. 2012. Efficacy of lycopene against fluoride toxicity in rats. Pharm Biol
50: 707-711.
Mietkiewski K, Walczak M, Trojanowicz R. 1966. [Effect of sodium fluoride on the
neurosecretory system in guinea pigs]. Endokrynol Pol 17: 121-131.
Mohamed NE. 2016. The role of calcium in ameliorating the oxidative stress of fluoride in rats.
Biol Trace Elem Res 170: 128-144.
Muhlemann HR, Schneider R. 1956. [Mitotic activity of rat thyroid epithelium after
administration of fluoridated drinking water]. Schweiz Med Wochenschr 86: 625-627.
Nabavi SF, Eslami S, Moghaddam AH, Nabavi SM. 2011. Protective effects of curcumin against
fluoride-induced oxidative stress in the rat brain. Neurophysiology 43: 287-291.
Nabavi SF, Moghaddam AH, Nabavi SM, Eslami S. 2011. Protective effect of curcumin and
quercetin on thyroid function in sodium fluoride intoxicated rats. Fluoride 44: 147-152.
Nabavi SF, Habtemariam S, Jafari M, Sureda A, Nabavi SM. 2012. Protective role of gallic acid
on sodium fluoride induced oxidative stress in rat brain. Bull Environ Contam Toxicol 89: 73-77.
Nabavi SF, Nabavi SM, Latifi AM, Mirzaei M, Habtemariam S, Moghaddam AH. 2012.
Mitigating role of quercetin against sodium fluoride-induced oxidative stress in the rat brain.
Pharm Biol 50: 1380-1383.
Nabavi SF, Nabavi SM, Habtemariam S, Moghaddam AH, Sureda A, Mirzaei M. 2013.
Neuroprotective effects of methyl-3-O-methyl gallate against sodium fluoride-induced oxidative
stress in the brain of rats. Cell Mol Neurobiol 33: 261-267.
Nabavi SM, Sureda A, Nabavi SF, Latifi AM, Moghaddam AH, Hellio C. 2012. Neuroprotective
effects of silymarin on sodium fluoride-induced oxidative stress. J Fluor Chem 142: 79-82.
Narayanaswamy M, Piler MB. 2010. Effect of maternal exposure of fluoride on biometals and
oxidative stress parameters in developing CNS of rat. Biol Trace Elem Res 133: 71-82.
Narbutt B, Romer TE, Grabski J, Szymik N. 1971. [Influence of natrium fluoride on the structure
of the rat thyroid]. Endokrynol Pol 22: 445-451.
Narbutt B, Romer TE, Grabski J, Szymik N. 1971. [The influence of sodium fluoride on the
morphology of the thyroid gland in rats]. Endokrynol Pol 22: 361-365.
Niu RY, Sun ZL, Cheng ZT, Liu HT, Chen HC, Wang JD. 2008. Effects of fluoride and lead on
N-methyl-D-aspartate receptor 1 expression in the hippocampus of offspring rat pups. Fluoride
41: 101-110.
Niu R, Wang J, Sun Z, Xue X, Yan X, Zhang J. 2015. Transcriptional regulatory dynamics of the
hypothalamic-pituitary-testicular axis in male mice exposed to fluoride. Environ Toxicol
Pharmacol 40: 557-562.

C-34
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Niu R, Zhang Y, Liu S, Liu F, Sun Z, Wang J. 2015. Proteome alterations in cortex of mice
exposed to fluoride and lead. Biol Trace Elem Res 164: 99-105.
Ogilvie AL. 1952. Histological findings in the kidney, liver, pancreas, adrenal and thyroid gland
of the rat following sodium fluoride administration. J Dent Res 31: 598-598.
Okayasu I, Tsuchida M, Yanagisawa F. 1985. Hyperplastic nodules of thyroid parafollicular cells
(C cells) in rats induced by prolonged low dose ingestion of NaF. Fluoride 18: 111-117.
Pal S, Sarkar C. 2014. Protective effect of resveratrol on fluoride induced alteration in protein
and nucleic acid metabolism, DNA damage and biogenic amines in rat brain. Environ Toxicol
Pharmacol 38: 684-699.
Pan Y, Lu P, Yin L, Chen K, He Y. 2015. Effect of fluoride on the proteomic profile of the
hippocampus in rats. Z Naturforsch C 70: 151-157.
Phillips PH, Lamb AR. 1934. Histology of certain organs and teeth in chronic toxicosis due to
fluorin. Arch Path 17: 169-176.
Portela ML. 1972. [Biochemical effects in the prolonged ingestion of fluorides in rats]. Arch
Latinoam Nutr 22: 291-308.
Prestes DS, Filappi A, Schossler DR, Duarte FA, Dressler VL, Flores EMM, Cecim M. 2009.
Functional and histological evaluations of thyroid of sheep submitted to sodium fluoride
administration. Arq Bras Med Vet Zootec 61: 293-298.
Puentes F, Cremer HD. 1966. Experiments on fluoride-iodine antagonism in the thyroid gland.
Adv Fluorine Res 4: 213-220.
Qian W, Miao K, Li T, Zhang Z. 2013. Effect of selenium on fluoride-induced changes in
synaptic plasticity in rat hippocampus. Biol Trace Elem Res 155: 253-260.
Qing-Feng S, Ying-Peng X, Tian-Tong X. 2019. Matrix metalloproteinase-9 and p53 involved in
chronic fluorosis induced blood-brain barrier damage and neurocyte changes. Arch Med Sci
15(2): 457-466.
Qiu YH, Kong DM, Yang Q, Zhao N. 2010. [Influence of high-fluoride on thyroid function and
brain damage in rats]. Chin J Endemiol 29: 146-149.
Raghavendra M, Ravindra RK, Raghuveer YP, Narasimha JK, Uma MRV, Navakishor P. 2016.
Alleviatory effects of hydroalcoholic extract of cauliflower (brassica oleracea var. botrytis) on
thyroid function in fluoride intoxicated rats. Fluoride 49: 84-90.
Rakhov GM. 1964. [Effect of calcium and fluorine in drinking water on the iodine metabolism
and status of the thyroid gland in iodine insufficiency in food]. Gig Sanit 29: 12-17.
Ranpariya VL, Parmar SK, Sheth NR, Chandrashekhar VM. 2011. Neuroprotective activity of
matricaria recutita against fluoride-induced stress in rats. Pharm Biol 49: 696-701.
Reddy KP, Sailaja G, Krishnaiah C. 2009. Protective effects of selenium on fluoride induced
alterations in certain enzymes in brain of mice. J Environ Biol 30: 859-864.

C-35
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Rogalska A, Kuter K, Zelazko A, Glogowska-Gruszka A, Swietochowska E, Nowak P. 2017.

Fluoride alteration of [3H]glucose uptake in Wistar rat brain and peripheral tissues. Neurotoxicol
Res 31: 436-443.
Saka O, Hallac P, Urgancioğlu I. 1965. The effect of fluoride on the thyroid of the rat. New
Istanbul Contrib Clin Sci 8: 87-90.
Samanta A, Chanda S, Bandyopadhyay B, Das N. 2016. Establishment of drug delivery system
nanocapsulated with an antioxidant (+)-catechin hydrate and sodium meta borate chelator against
sodium fluoride induced oxidative stress in rats. J Trace Elem Med Biol 33: 54-67.
Sarkar C, Das N, Pal S, Dinda B. 2014. Oxidative stress induced alteration of protein and nucleic
acid metabolism in fluoride-intoxicated rat brain: Protection by 3α-hydroxy olean-12-en-27-oic
acid isolated from neanotis wightiana. Int J Pharm Sci Res 5: 3047-3066.
Sarkar C, Pal S. 2014. Ameliorative effect of resveratrol against fluoride-induced alteration of
thyroid function in male Wistar rats. Biol Trace Elem Res 162: 278-287.
Sarkar C, Pal S, Das N, Dinda B. 2014. Ameliorative effects of oleanolic acid on fluoride
induced metabolic and oxidative dysfunctions in rat brain: Experimental and biochemical
studies. Food Chem Toxicol 66: 224-236.
Seffner W, Teubener W, Runde H, Wiedner H, Vogt J, Otto G, Zschau E, Geinitz D, Franke J.
1990. Boron as an antidote to fluorosis? II. Studies on various organs of pigs. Fluoride 23: 68-
79.
Selim AOA, El-Haleem MR, Ibrahim IH. 2012. Effect of sodium fluoride on the thyroid gland of
growing male albino rats: Histological and biochemical study. Egypt J Histol 35: 470-482.
Shao Q, Wang Yn, Guan Z. 2000. [Influence of free radical inducer on the level of oxidative
stress in brain of rats with fluorosis]. Chin J Prev Med 34: 330-332.
Sharma C, Suhalka P, Sukhwal P, Jaiswal N, Bhatnagar M. 2014. Curcumin attenuates
neurotoxicity induced by fluoride: An in vivo evidence. Pharmacogn Mag 10: 61-65.
Shashi A. 1992. Studies on alterations in brain lipid metabolism following experimental
fluorosis. Fluoride 25: 77-84.
Shashi A. 1993. Nucleic acid levels in thyroid gland in acute and chronic fluoride intoxication.
Fluoride 26: 191-196.
Shashi A, Singh JP, Thapar SP. 1994. Effect of long-term administration of fluoride on levels of
protein, free amino acids and RNA in rabbit brain. Fluoride 27: 155-159.
Shashi A. 2003. Histopathological investigation of fluoride-induced neurotoxicity in rabbits.
Fluoride 36: 95-105.
Shashi A, Neetika S, Bhardwaj M. 2009. Neuronal DNA damage and apoptosis in brain of rat
exposed to fluoride. Asian J Microbiol Biotechnol Environ Sci 11: 629-632.
Shayiq RM, Raza H, Kidwai AM. 1986. Fluoride and lipid peroxidation a comparative study in
different rat tissues. Bull Environ Contam Toxicol 37: 70-76.

C-36
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Shen QF, Li HN, Xu TT, Xia YP. 2012. [Damage of blood brain barrier of spinal cord in rats
with chronic fluorosis]. Chin Med J 92: 2357-2361.
Shen Q, Tian R, Li H, Xu T, Xia Y. 2014. [White matter injury of spinal cord in rats with
chronic fluorosis and recovery after defluoridation]. Chin Med J 94: 1189-1192.
Shen X, Zhang Z, Xu X. 2004. [Influence of combined iodine and fluoride on phospholipid and
fatty acid composition in brain cells of rats]. J Hyg Res 33: 158-161.
Shivarajashankara YM, Shivashankara AR, Bhat PG, Rao SH. 2001. Effect of fluoride
intoxication on lipid peroxidation and antioxidant systems in rats. Fluoride 34: 108-113.
Shivarajashankara YM, Shivashankara AR, Bhat PG, Rao SH. 2002. Brain lipid peroxidation
and antioxidant systems of young rats in chronic fluoride intoxication. Fluoride 35: 197-203.
Shivarajashankara YM, Shivashankara AR, Bhat PG, Rao SM, Rao SH. 2002. Histological
changes in the brain of young fluoride-intoxicated rats. Fluoride 35: 12-21.
Siebenhuner L, Miloni E, Burgi H. 1984. [Effects of fluoride on thyroid hormone biosynthesis:
Studies in a highly sensitive test system]. Klin Wochenschr 62: 859-861.
Singh R, Srivastava AK, Gangwar NK. 2017. Clinico-pathological studies on the co-exposure of
cypermethrin and fluoride in experimental rats with ameliorative action of Vitamin E. Vet Pract
18(2): 207-210.
Soni KK, Shrivastava VK. 2007. Sodium fluoride induced histopathological changes in thyroid
gland of male mus musculus. Biochem Cell Arch 7: 317-320.
Stee EW. 1968. Effect of sodium fluoride and AMOX (NF3O) on growth and thyroid function in
the rat. No. AMRL-TR-67-189. Wright-Patterson Air Force Base, OH: pp. 67.
Štolc V, Podoba J. 1960. Effect of fluoride on the biogenesis of thyroid hormones. Nature 188:
855-856.
Sugivama Y. 1967. [The effect of sodium fluoride administration on the parathyroid glands].
Hirosaki Med J 19: 520-529.
Sun Y, Ke L, Zheng X, Li T, Ouyang W, Zhang Z. 2016. Effects of different levels of calcium
intake on brain cell apoptosis in fluorosis rat offspring and its molecular mechanism. Biol Trace
Elem Res: 1-12.
Takata H. 1958. The effect of fluorine upon the uptake of I131 by the thyroid glands. Folia
Pharmacol Jpn 54: 230-236.
Teng Y, Zhang J, Zhang Z, Feng J. 2017. The effect of chronic fluorosis on calcium ions and
CaMKIIα, and c-fos expression in the rat hippocampus. Biol Trace Elem Res: 295-302.
Trabelsi M, Guermazi F, Zeghal N. 2001. Effect of fluoride on thyroid function and cerebellar
development in mice. Fluoride 34: 165-173.

C-37
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Tsuchida M, Okayasu I, Kohyama Y, Kurihara H, Tanaka H, Yanagisawa F, Date C, Hayashi M,

Mui K, Asada M. 1986. Effects of long term, low dose ingestion of fluoride on the thyroid gland
in rats. Stud Environ Sci 27: 307-312.
Vani ML, Reddy KP. 2000. Effects of fluoride accumulation on some enzymes of brain and
gastrocnemius muscle of mice. Fluoride 33: 17-26.
Wang C, Liang C, Ma J, Manthari RK, Niu R, Wang J, Wang J, Zhang J. 2018. Co-exposure to
fluoride and sulfur dioxide on histological alteration and DNA damage in rat brain. J Biochem
Mol Toxicol 32.
Wang H, Yang Z, Zhou B, Gao H, Yan X, Wang J. 2009. Fluoride-induced thyroid dysfunction
in rats: Roles of dietary protein and calcium level. Toxicol Ind Health 25: 49-57.
Wang J, Niu R, Sun Z, Lv L, Smith GW. 2008. Effects of protein and calcium supplementation
on bone metabolism and thyroid function in protein and calcium deficient rabbits exposed to
fluoride. Fluoride 41: 283-291.
Wang JD, Ge YM, Ning HM, Wang SL. 2004. Effects of high fluoride and low iodine on
oxidative stress and antioxidant defense of the brain in offspring rats. Fluoride 37: 264-270.
Wang JL. 2007. [Effect of fluoride on the intracellular Ca2+ in neurons of mice]. Chin J
Endemiol 26: 505-507.
Wang Y, Guan Z, Xiao K. 1997. [Changes of coenzyme Q content in brain tissues of rats with
fluorosis]. Chin J Prev Med 31: 330-333.
Wang Y, Dong Y, Wei N, Guan Z. 2015. [Influence of chronic fluorosis on expression of
quinone oxidoreductase-1 and heme oxygenase-1 in rat brains]. Chin J Endemiol 34: 250-253.
Wedzisz A, Cieciura J. 1988. Effect of small sodium fluoride feed supplements on the serum
thyroid hormone content of rats. Bromatol Chem Toksykol 21: 174-175.
Wei N, Dong YT, Deng J, Wang Y, Qi XL, Yu WF, Xiao Y, Zhou JJ, Guan ZZ. 2018. Changed
expressions of N-methyl-d-aspartate receptors in the brains of rats and primary neurons exposed
to high level of fluoride. J Trace Elem Med Biol 45: 31-40.
Yan N, Liu Y, Liu S, Cao S, Wang F, Wang Z, Xi S. 2016. Fluoride-induced neuron apoptosis
and expressions of inflammatory factors by activating microglia in rat brain. Mol Neurobiol 53:
4449-4460.
Yang H, Xing R, Liu S, Yu H, Li P. 2016. Gamma-Aminobutyric acid ameliorates fluoride-
induced hypothyroidism in male Kunming mice. Life Sci 146: 1-7.
Yang H, Xing R, Liu S, Yu H, Li P. 2019. Analysis of the protective effects of gamma-
aminobutyric acid during fluoride-induced hypothyroidism in male Kunming mice. Pharm Biol
57(1): 29-37.
Yang M, Ren Z, Zhou B, Guan Z, Yu W. 2017. [Expression of endonuclease G in the brain
tissue of rats with chronic fluorosis]. Chin J Endemiol 36: 327-332.

C-38
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Yuan SD, Xie QW, Lu FY. 1993. Changes of serotonin content and turnover rate in
hypothalamus of female rat during fluorosis. Fluoride 26: 57-60.
Zhai JX, Guo ZY, Hu CL, Wang QN, Zhu QX. 2003. [Studies on fluoride concentration and
cholinesterase activity in rat hippocampus]. Chin J Ind Hyg Occup Dis 21: 102-104.
Zhan CW, Huo DJ. 1988. Ultrastructural findings in liver, kidneys, thyroid-gland and cardiac-
muscle of rabbits following sodium-fluoride administration. Fluoride 21: 32-38.
Zhan XA, Xu ZR, Li JX, Wang M. 2005. Effects of fluorosis on lipid peroxidation and
antioxidant systems in young pigs. Fluoride 38: 157-161.
Zhan XA, Li JX, Wang M, Xu ZR. 2006. Effects of fluoride on growth and thyroid function in
young pigs. Fluoride 39: 95-100.
Zhang KL, Lou DD, Liu YF, Qin SL, Guan ZZ. 2012. [Changes of P-glycoprotein and nuclear
factor κB in the cerebral cortex of rat with chronic fluorosis]. Chin J Endemiol 31: 613-616.
Zhang KL, Lou DD, Guan ZZ. 2013. [Expression of receptor for advanced glycation
endproducts and nuclear factor κB in brain hippocampus of rat with chronic fluorosis]. Chin J
Endemiol 32: 625-628.
Zhang WD, Zhang Y, Liu GY, Jiang P, Chai CY. 2008. [Effects of fluoride on ultrastructure of
thyroids in rats]. Chin J Endemiol 27: 622-624.
Zhang ZG, Wang XY, Nian WW, Liao QX, Zhang R, Ouyang W. 2017. Effects of calcium on
drinking fluorosis-induced hippocampal syntaptic plasticity impairment in the offspring of rats.
Transl Neurosci 8: 191-200.
Zhao W, Zhu H, Yu Z, Aoki K, Misumi J, Zhang X. 1998. Long-term effects of various iodine
and fluorine doses on the thyroid and fluorosis in mice. Endocr Regul 32: 63-70.
Zhao WY. 1988. [A preliminary study of the interaction of iodine and fluoride in experimental
iodine goiter and fluorosis]. Chin J Prev Med 22: 146-148.
Zhao XL, Gao WH, Zhao ZL. 1994. [Effects of sodium fluoride on the activity of Ca2+Mg(2+)-
ATPase in synaptic membrane in rat brain]. Chin J Prev Med 28: 264-266.
Zhavoronkov AA, Polyakova GA. 1973. Morphological and functional state of the hypothalamo-
hypophyseal neurosecretory system in experimental fluorosis. Bull Exp Biol Med 75: 194-196.
Zhou B, Luo G, Wang C, Niu R, Wang J. 2014. Effects of fluoride on expression of cytokines in
the hippocampus of adult rats. Fluoride 47: 191-198.

C.2.3. In Vitro Experimental Studies

As described in Figure 2, 60 in vitro experimental studies were included; however, data
extraction was not conducted on in vitro studies. Therefore, in vitro experimental studies are not
available in HAWC (NTP 2019) with the exception of in vitro studies that also reported in vivo
non-human animal data that met the relevant criteria for being made available in HAWC. The
following lists of references are organized as studies that are available in HAWC (n = 6)
followed by studies that are not available in HAWC (n = 54).

C-39
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

C.2.3.1. Studies Available in HAWC

Chen J, Niu Q, Xia T, Zhou G, Li P, Zhao Q, Xu C, Dong L, Zhang S, Wang A. 2018. ERK1/2-
mediated disruption of BDNF-TrkB signaling causes synaptic impairment contributing to
fluoride-induced developmental neurotoxicity. Toxicology 410: 222-230.
Niu Q, Chen J, Xia T, Li P, Zhou G, Xu C, Zhao Q, Dong L, Zhang S, Wang A. 2018. Excessive
ER stress and the resulting autophagic flux dysfunction contribute to fluoride-induced
neurotoxicity. Environ Pollut 233: 889-899.
Shan KR, Qi XL, Long YG, Nordberg A, Guan ZZ. 2004. Decreased nicotinic receptors in PC12
cells and rat brains influenced by fluoride toxicity: A mechanism relating to a damage at the
level in post-transcription of the receptor genes. Toxicology 200: 169-177.
Zhang KL, Lou DD, Guan ZZ. 2015. Activation of the AGE/RAGE system in the brains of rats
and in SH-SY5Y cells exposed to high level of fluoride might connect to oxidative stress.
Neurotoxicol Teratol 48: 49-55.
Zhao Q, Niu Q, Chen J, Xia T, Zhou G, Li P, Dong L, Xu C, Tian Z, Luo C, Liu L, Zhang S,
Wang A. 2019. Roles of mitochondrial fission inhibition in developmental fluoride
neurotoxicity: Mechanisms of action in vitro and associations with cognition in rats and children.
Arch Toxicol 93(3): 709-726.
Zhao XL, Wu JH. 1998. Actions of sodium fluoride on acetylcholinesterase activities in rats.
Biomed Environ Sci 11: 1-6.
C.2.3.2. Studies Not Available in HAWC
Ardelean I, Racoveanu N, Manescu S, Lupulescu A, Diaconescu M, Ghelerter L. 1964.
Experimental investigations concerning the effect of fluorine on the thyroid gland. Rum Med Rev
8: 17-20.
Chen J, Chen X, Yang K. 2000. [Effects of selenium and zinc on the DNA damage caused by
fluoride in pallium neural cells of rats]. J Hyg Res 29: 216-217.
Chen L, Ning H, Yin Z, Song X, Feng Y, Qin H, Li Y, Wang J, Ge Y, Wang W. 2017. The
effects of fluoride on neuronal function occurs via cytoskeleton damage and decreased signal
transmission. Chemosphere 185: 589-594.
Chen R, Zhao LD, Liu H, Li HH, Ren C, Zhang P, Guo KT, Zhang HX, Geng DQ, Zhang CY.
2017. Fluoride induces neuroinflammation and alters Wnt signaling pathway in BV2 microglial
cells. Inflammation 40: 1123-1130.
Cheng TJ, Chen TM, Chen CH, Lai YK. 1998. Induction of stress response and differential
expression of 70 kDa stress proteins by sodium fluoride in HeLa and rat brain tumor 9L cells. J
Cell Biochem 69: 221-231.
Deng MF, Zhu D, Liu YP, He WW, Gui CZ, Guan ZZ. 2018. Attenuation by 7-nitroindazole of
fluoride-induced toxicity in SH-SY5Y cells exposed to high fluoride: Effects on nitric oxide,
nitric oxide synthetase activity, nNOS, and apoptosis. Fluoride 51(4): 328-339.

C-40
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Flores-Mendez M, Ramirez D, Alamillo N, Hernandez-Kelly LC, Del Razo LM, Ortega A. 2014.
Fluoride exposure regulates the elongation phase of protein synthesis in cultured Bergmann glia
cells. Toxicol Lett 229: 126-133.
Gao Q, Liu YH, Guan ZZ. 2008. Oxidative stress might be a mechanism connected with the
decreased alpha 7 nicotinic receptor influenced by high-concentration of fluoride in SH-SY5Y
neuroblastoma cells. Toxicol In Vitro 22: 837-843.
Goschorska M, Gutowska I, Baranowska-Bosiacka I, Piotrowska K, Metryka E, Safranow K,
Chlubek D. 2018. Influence of acetylcholinesterase inhibitors used in Alzheimer's Disease
treatment on the activity of antioxidant enzymes and the concentration of glutathione in THP-1
macrophages under fluoride-induced oxidative stress. Int J Environ Res Pub Health 16(1).
Guan ZZ, Shan KR, Xiu J, Long YG. 2005. [Fluorosis on expression of nicotinic acetylcholine
receptors in protein and gene levels in human SH-SY5Y neuroblastoma cells]. Chin J Prev Med
39: 26-29.
Guan ZZ, Shan KR, Wang YN, Dallner G. 2006. [Changes of lipids and nicotinic receptors in rat
brains and pheochromocytoma with fluorosis]. Chin J Endemiol 25: 121-124.
Haojun Z, Yaoling W, Ke Z, Jin L, Junling W. 2012. Effects of NaF on the expression of
intracellular Ca2+ fluxes and apoptosis and the antagonism of taurine in murine neuron. Toxicol
Mech Methods 22: 305-308.
Hong-Liang L, Qiang Z, Yu-Shan C, Lei Z, Gang F, Chang-Chun H, Liang Z, Aiguo W. 2014.
Fluoride-induced thyroid cell apoptosis. Fluoride 47: 161-169.
Inkielewicz-Stepniak I, Radomski MW, Wozniak M. 2012. Fisetin prevents fluoride- and
dexamethasone-induced oxidative damage in osteoblast and hippocampal cells. Food Chem
Toxicol 50: 583-589.
Jin TX, Guan ZZ, Zhang H. 2011. [The effect of fluoride on α subunit of calcium/calmodulin-
dependent protein kinase-II mRNA and protein expression in central nervous system]. Chin J
Endemiol 30: 247-250.
Kariya T, Kotani M, Field JB. 1974. Effects of sodium fluoride and other metabolic inhibitors on
basal and TSH stimulated cyclic AMP and thyroid metabolism. Metab Clin Exper 23: 967-973.
Ke L, Zheng X, Sun Y, Ouyang W, Zhang Z. 2016. Effects of sodium fluoride on lipid
peroxidation and PARP, XBP-1 expression in PC12 cell. Biol Trace Elem Res 173: 161-167.
Lee J, Han YE, Favorov O, Tommerdahl M, Whitsel B, Lee CJ. 2016. Fluoride induces a volume
reduction in CA1 hippocampal slices via MAP kinase pathway through volume regulated anion
channels. Exp Neurobiol 25: 72-78.
Levesque L, Mizzen CA, McLachlan DR, Fraser PE. 2000. Ligand specific effects on aluminum
incorporation and toxicity in neurons and astrocytes. Brain Res 877: 191-202.
Li H, Gao MT, Xu KY, Wang CY. 2007. Effect of sodium fluoride on the primary porcine
thyroid cells and thyroid peroxidase activity. J Clin Rehabil Tissue Eng Res 11: 7425-7428.

C-41
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Li H, Gao MT, Xu KY, Cui MY, Dai X. 2008. [Effect of fluoride on thyroid functioning in
primary porcine thyrocyte]. Chin J Endemiol 27: 38-40.
Li H, Huang H, Xu Y, Gao Y, Liu Z. 2010. [Toxic effects of fluoride on rat cerebral cortex
astrocytes in vitro]. J Hyg Res 39: 86-88.
Liu H, Zeng Q, Cui Y, Yu L, Zhao L, Hou C, Zhang S, Zhang L, Fu G, Liu Y, Jiang C, Chen X,
Wang A. 2014. The effects and underlying mechanism of excessive iodide on excessive fluoride-
induced thyroid cytotoxicity. Environ Toxicol Pharmacol 38: 332-340.
Liu HL, Zeng Q, Cui YS, Zhao L, Zhang L, Fu G, Hou CC, Zhang S, Yu LY, Jiang CY, Wang
ZL, Chen XM, Wang AG. 2014. The role of the IRE1 pathway in excessive iodide- and/or
fluoride-induced apoptosis in Nthy-ori 3-1 cells in vitro. Toxicol Lett 224: 341-348.
Liu YJ, Guan ZZ, Gao Q, Pei JJ. 2011. Increased level of apoptosis in rat brains and SH-SY5Y
cells exposed to excessive fluoride-A mechanism connected with activating JNK
phosphorylation. Toxicol Lett 204: 183-189.
Liu Y, Gao Q, Tang Z, Zhang X, Guan Z. 2015. [The expression and correlation between neural
nicotinic acetylcholine receptor subunit α3 and mitogen-activated protein kinase cell signaling
transduction pathway in human neuroblastoma cell line SH-SY5Y overexposed to fluoride].
Chin J Endemiol 34: 553-558.
Madaoui S, Rappaport L, Nunez J. 1974. Prostaglandins and in vitro TSH-dependent iodide
binding by rat thyroid glands. Biochimie 56: 109-113.
Nakagawa-Yagi Y, Saito Y, Kitoh N, Ogane N, Fujisawa E, Nakamura H. 1993. Fluoride causes
suppression of neurite outgrowth in human neuroblastoma via an influx of extracellular calcium.
Biochem Biophys Res Commun 191: 727-736.
Ong J, Kerr DIB. 1995. Interactions of N-ethylmaleimide and aluminium fluoride with
GABA(B) receptor function in rat neocortical slices. Eur J Pharmacol 287: 197-200.
Pastan I, Macchia V, Katzen R. 1968. Effect of fluoride on the metabolic activity of thyroid
slices. Endocrinology 83: 157-160.
Rubakhova VM. 1977. [Effect of serotonin and sodium fluoride on visceral nerve conductors].
Vyestsi Akademii Navuk BSSR Syeryya Biyalahichnykh Navuk 1: 117-119.
Shayiq RM, Raza H, Kidwai AM. 1986. Fluoride and lipid peroxidation a comparative study in
different rat tissues. Bull Environ Contam Toxicol 37: 70-76.
Shuhua X, Ziyou L, Ling Y, Fei W, Sun G. 2012. A role of fluoride on free radical generation
and oxidative stress in BV-2 microglia cells. Mediators Inflamm 2012: 1-8.
Singh P, Das TK. 2019. Ultrastructural localization of 4-hydroxynonenal adducts in fluoride-
exposed cells: Protective role of dietary antioxidants. Fluoride 52(1): 49-58.
Taylor P. 1972. Comparison of the effects of various agents on thyroidal adenyl cyclase activity
with their effects on thyroid hormone release. J Endocrinol 54: 137-145.

C-42
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Tu W, Zhang Q, Liu Y, Han LY, Wang Q, Chen PP, Zhang S, Wang AG, Zhou X. 2018.
Fluoride induces apoptosis via inhibiting SIRT1 activity to activate mitochondrial p53 pathway
in human neuroblastoma SH-SY5Y cells. Toxicol Appl Pharmacol 347: 60-69.
van der Voet GB, Schijns O, de Wolff FA. 1999. Fluoride enhances the effect of aluminium
chloride on interconnections between aggregates of hippocampal neurons. Arch Physiol Biochem
107: 15-21.
Wang JL. 2007. [Effect of fluoride on the intracellular Ca2+ in neurons of mice]. Chin J
Endemiol 26: 505-507.
Wang J, Gao Y, Cheng X, Yang J, Zhao Y, Xu H, Zhu Y, Yan Z, Manthari RK, Mehdi OM,
Wang J. 2019. GSTO1 acts as a mediator in sodium fluoride-induced alterations of learning and
memory related factors expressions in the hippocampus cell line. Chemosphere 226: 201-209.
Wei N, Dong YT, Deng J, Wang Y, Qi XL, Yu WF, Xiao Y, Zhou JJ, Guan ZZ. 2018. Changed
expressions of N-methyl-d-aspartate receptors in the brains of rats and primary neurons exposed
to high level of fluoride. J Trace Elem Med Biol 45: 31-40.
Willems CB-V, Sande J, Dumont JE. 1972. Inhibition of thyroid secretion by sodium fluoride in
vitro. Biochim Biophys Acta 264: 197-204.
Woodward JJ, Harms J. 1992. Potentiation of N-methyl-D-aspartate-stimulated dopamine release
from rat brain slices by aluminum fluoride and carbachol. J Neurochem 58: 1547-1554.
Wu J, Cheng M, Liu Q, Yang J, Wu S, Lu X, Jin C, Ma H, Cai Y. 2015. Protective role of tert-
butylhydroquinone against sodium fluoride-induced oxidative stress and apoptosis in PC12 cells.
Cell Mol Neurobiol 35: 1017-1025.
Xia T, Zhang M, He WH, He P, Wang AG. 2007. [Effects of fluoride on neural cell adhesion
molecules mRNA and protein expression levels in primary rat hippocampal neurons]. Chin J
Prev Med 41: 475-478.
Xu B, Xu Z, Xia T, He P, Gao P, He W, Zhang M, Guo L, Niu Q, Wang A. 2011. Effects of the
Fas/Fas-L pathway on fluoride-induced apoptosis in SH-SY5Y cells. Environ Toxicol 26: 86-92.
Xu Z, Xu B, Xia T, He W, Gao P, Guo L, Wang Z, Niu Q, Wang A. 2013. Relationship between
intracellular Ca2+ and ROS during fluoride-induced injury in SH-SY5Y cells. Environ Toxicol
28: 307-312.
Yamashita K, Field JB. 1972. Elevation of cyclic guanosine 3,5; monophosphate levels in dog
thyroid slices caused by acetylcholine and sodium fluoride. J Biol Chem 247: 7062-7066.
Yan L, Liu S, Wang C, Wang F, Song Y, Yan N, Xi S, Liu Z, Sun G. 2013. JNK and NADPH
oxidase involved in fluoride-induced oxidative stress in BV-2 microglia cells. Mediators
Inflamm 2013: 895-975.
Zhang CY, Chen R, Wang F, Ren C, Zhang P, Li Q, Li HH, Guo KT, Geng DQ, Liu CF. 2016.
EGb-761 attenuates the anti-proliferative activity of fluoride via DDK1 in PC-12 cells.
Neurochem Res 42(2): 606-614.

C-43
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Zhang M, Wang A, He W, He P, Xu B, Xia T, Chen X, Yang K. 2007. Effects of fluoride on the

expression of NCAM, oxidative stress, and apoptosis in primary cultured hippocampal neurons.
Toxicology 236: 208-216.
Zhang M, Wang A, Xia T, He P. 2008. Effects of fluoride on DNA damage, S-phase cell-cycle
arrest and the expression of NF-kappaB in primary cultured rat hippocampal neurons. Toxicol
Lett 179: 1-5.
Zhang S, Zheng X, Sun Y, Wang Y, Zhang Z. 2015. Alterations in oxidative stress and apoptosis
in cultured PC12 cells exposed to fluoride. Fluoride 48: 213-222.
Zhao L, Xiao Y, Deng CM, Tan LC, Guan ZZ. 2016. Protective effect of lovastatin on
neurotoxicity of excessive fluoride in primary hippocampal neurons. Fluoride 49: 36-46.
Zhao XL, Gao WH, Zhao ZL. 1994. [Effects of sodium fluoride on the activity of Ca2+Mg(2+)-
ATPase in synaptic membrane in rat brain]. Chin J Prev Med 28: 264-266.

C-44
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Appendix D. Risk-of-bias Figures

Figures
Figure D-1. Risk-of-bias Heatmap for Low Risk-of-bias Human Neurodevelopmental or
Cognitive Studies Following Fluoride Exposure .................................................... D-3
Figure D-2. Risk-of-bias Bar Chart for Low Risk-of-bias Human Neurodevelopmental or
Cognitive Studies Following Fluoride Exposure .................................................... D-3
Figure D-3. Risk-of-bias Heatmap for High Risk-of-bias Human Neurodevelopmental or
Cognitive Studies Following Fluoride Exposure .................................................... D-3
Figure D-4. Risk-of-bias Bar Chart for High Risk-of-bias Human Neurodevelopmental or
Cognitive Studies Following Fluoride Exposure .................................................... D-4
Figure D-5. Risk-of-bias Heatmap for Low Risk-of-bias Children’s IQ Studies Following
Fluoride Exposure ................................................................................................... D-4
Figure D-6. Risk-of-bias Bar Chart for Low Risk-of-bias Children’s IQ Studies Following
Fluoride Exposure ................................................................................................... D-4
Figure D-7. Risk-of-bias Heatmap for High Risk-of-bias Children’s IQ Studies Following
Fluoride Exposure ................................................................................................... D-5
Figure D-8. Risk-of-bias Bar Chart for High Risk-of-bias Children’s IQ Studies Following
Fluoride Exposure ................................................................................................... D-5
Figure D-9. Risk-of-bias Heatmap for Low Risk-of-bias Children’s Other Neurodevelopmental
Effect Studies Following Fluoride Exposure .......................................................... D-5
Figure D-10. Risk-of-bias Bar Chart for Low Risk-of-bias Children’s Other Neurodevelopmental
Effect Studies Following Fluoride Exposure ........................................................ D-6
Figure D-11. Risk-of-bias Heatmap for High Risk-of-bias Children’s Other Neurodevelopmental
Effect Studies Following Fluoride Exposure ........................................................ D-6
Figure D-12. Risk-of-bias Bar Chart for High Risk-of-bias Children’s Other
Neurodevelopmental Effect Studies Following Fluoride Exposure ...................... D-7
Figure D-13. Risk-of-bias Heatmap for Low Risk-of-bias Adult Cognitive Studies Following
Fluoride Exposure ................................................................................................. D-7
Figure D-14. Risk-of-bias Bar Chart for Low Risk-of-bias Adult Cognitive Studies Following
Fluoride Exposure ................................................................................................. D-8
Figure D-15. Risk-of-bias Heatmap for High Risk-of-bias Adult Cognitive Studies Following
Fluoride Exposure ................................................................................................. D-8
Figure D-16. Risk-of-bias Bar Chart for High Risk-of-bias Adult Cognitive Studies Following
Fluoride Exposure ................................................................................................. D-9
Figure D-17. Risk-of-bias Heatmap for Low Risk-of-bias Human Mechanistic Studies Following
Fluoride Exposure ................................................................................................. D-9
Figure D-18. Risk-of-bias Bar Chart for Low Risk-of-bias Human Mechanistic Studies
Following Fluoride Exposure .............................................................................. D-10
Figure D-19. Risk-of-bias Heatmap for High Risk-of-bias Human Mechanistic Studies Following
Fluoride Exposure ............................................................................................... D-10
Figure D-20. Risk-of-bias Bar Chart for High Risk-of-bias Human Mechanistic Studies
Following Fluoride Exposure .............................................................................. D-10
Figure D-21. Risk-of-bias Heatmap for New Developmental Animal Learning and Memory
Studies Following Fluoride Exposure ................................................................. D-11

D-1
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-22. Risk-of-bias Bar Chart for New Developmental Animal Learning and Memory
Studies Following Fluoride Exposure ................................................................. D-11
Figure D-23. Risk-of-bias Heatmap for New Adult Animal Learning and Memory Studies
Following Fluoride Exposure .............................................................................. D-12
Figure D-24. Risk-of-bias Bar Chart for New Adult Animal Learning and Memory Studies
Following Fluoride Exposure .............................................................................. D-12
Figure D-25. Risk-of-bias Heatmap for Low Risk-of-bias Animal Biochemical Studies
Following Fluoride Exposure .............................................................................. D-13
Figure D-26. Risk-of-bias Bar Chart for Low Risk-of-bias Animal Biochemical Studies
Following Fluoride Exposure .............................................................................. D-13
Figure D-27. Risk-of-bias Heatmap for High Risk-of-bias Animal Biochemical Studies
Following Fluoride Exposure .............................................................................. D-14
Figure D-28. Risk-of-bias Bar Chart for High Risk-of-bias Animal Biochemical Studies
Following Fluoride Exposure .............................................................................. D-14
Figure D-29. Risk-of-bias Heatmap for Low Risk-of-bias Animal Neurotransmission Studies
Following Fluoride Exposure .............................................................................. D-15
Figure D-30. Risk-of-bias Bar Chart for Low Risk-of-bias Animal Neurotransmission Studies
Following Fluoride Exposure .............................................................................. D-15
Figure D-31. Risk-of-bias Heatmap for High Risk-of-bias Animal Neurotransmission Studies
Following Fluoride Exposure .............................................................................. D-16
Figure D-32. Risk-of-bias Bar Chart for High Risk-of-bias Animal Neurotransmission Studies
Following Fluoride Exposure .............................................................................. D-16
Figure D-33. Risk-of-bias Heatmap for Low Risk-of-bias Animal Oxidative Stress Studies
Following Fluoride Exposure .............................................................................. D-17
Figure D-34. Risk-of-bias Bar Chart for Low Risk-of-bias Animal Oxidative Stress Studies
Following Fluoride Exposure .............................................................................. D-17
Figure D-35. Risk-of-bias Heatmap for High Risk-of-bias Animal Oxidative Stress Studies
Following Fluoride Exposure .............................................................................. D-18
Figure D-36. Risk-of-bias Bar Chart for High Risk-of-bias Animal Oxidative Stress Studies
Following Fluoride Exposure .............................................................................. D-18
Figure D-37. Risk-of-bias Heatmap for Low Risk-of-bias Animal Histopathology Studies
Following Fluoride Exposure .............................................................................. D-19
Figure D-38. Risk-of-bias Bar Chart for Low Risk-of-bias Animal Histopathology Studies
Following Fluoride Exposure .............................................................................. D-19
Figure D-39. Risk-of-bias Heatmap for High Risk-of-bias Animal Histopathology Studies
Following Fluoride Exposure .............................................................................. D-20
Figure D-40. Risk-of-bias Bar Chart for High Risk-of-bias Animal Histopathology Studies
Following Fluoride Exposure .............................................................................. D-20

D-2
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

D.1. Studies in Humans

Figure D-1. Risk-of-bias Heatmap for Low Risk-of-bias Human Neurodevelopmental or Cognitive
Studies Following Fluoride Exposure

An interactive version of Figure D-1 and additional study details in HAWC here (NTP 2019).

Figure D-2. Risk-of-bias Bar Chart for Low Risk-of-bias Human Neurodevelopmental or Cognitive
Studies Following Fluoride Exposure

An interactive version of Figure D-2 and additional study details in HAWC here (NTP 2019).

Figure D-3. Risk-of-bias Heatmap for High Risk-of-bias Human Neurodevelopmental or Cognitive
Studies Following Fluoride Exposure

An interactive version of Figure D-3 and additional study details in HAWC here (NTP 2019).

D-3
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-4. Risk-of-bias Bar Chart for High Risk-of-bias Human Neurodevelopmental or Cognitive
Studies Following Fluoride Exposure

An interactive version of Figure D-4 and additional study details in HAWC here (NTP 2019).

Figure D-5. Risk-of-bias Heatmap for Low Risk-of-bias Children’s IQ Studies Following Fluoride
Exposure

An interactive version of Figure D-5 and additional study details in HAWC here (NTP 2019).

Figure D-6. Risk-of-bias Bar Chart for Low Risk-of-bias Children’s IQ Studies Following Fluoride
Exposure

An interactive version of Figure D-6 and additional study details in HAWC here (NTP 2019).

D-4
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-7. Risk-of-bias Heatmap for High Risk-of-bias Children’s IQ Studies Following Fluoride
Exposure

An interactive version of Figure D-7 and additional study details in HAWC here (NTP 2019).

Figure D-8. Risk-of-bias Bar Chart for High Risk-of-bias Children’s IQ Studies Following Fluoride
Exposure

An interactive version of Figure D-8 and additional study details in HAWC here (NTP 2019).

Figure D-9. Risk-of-bias Heatmap for Low Risk-of-bias Children’s Other Neurodevelopmental
Effect Studies Following Fluoride Exposure

An interactive version of Figure D-9 and additional study details in HAWC here (NTP 2019).

D-5
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-10. Risk-of-bias Bar Chart for Low Risk-of-bias Children’s Other Neurodevelopmental
Effect Studies Following Fluoride Exposure

An interactive version of Figure D-10 and additional study details in HAWC here (NTP 2019).

Figure D-11. Risk-of-bias Heatmap for High Risk-of-bias Children’s Other Neurodevelopmental
Effect Studies Following Fluoride Exposure

An interactive version of Figure D-11 and additional study details in HAWC here (NTP 2019).

D-6
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-12. Risk-of-bias Bar Chart for High Risk-of-bias Children’s Other Neurodevelopmental
Effect Studies Following Fluoride Exposure

An interactive version of Figure D-12 and additional study details in HAWC here (NTP 2019).

Figure D-13. Risk-of-bias Heatmap for Low Risk-of-bias Adult Cognitive Studies Following
Fluoride Exposure

An interactive version of Figure D-13 and additional study details in HAWC here (NTP 2019).

D-7
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-14. Risk-of-bias Bar Chart for Low Risk-of-bias Adult Cognitive Studies Following
Fluoride Exposure

An interactive version of Figure D-14 and additional study details in HAWC here (NTP 2019).

Figure D-15. Risk-of-bias Heatmap for High Risk-of-bias Adult Cognitive Studies Following
Fluoride Exposure

An interactive version of Figure D-15 and additional study details in HAWC here (NTP 2019).

D-8
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-16. Risk-of-bias Bar Chart for High Risk-of-bias Adult Cognitive Studies Following
Fluoride Exposure

An interactive version of Figure D-16 and additional study details in HAWC here (NTP 2019).

Figure D-17. Risk-of-bias Heatmap for Low Risk-of-bias Human Mechanistic Studies Following
Fluoride Exposure

An interactive version of Figure D-17 and additional study details in HAWC here (NTP 2019).

D-9
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-18. Risk-of-bias Bar Chart for Low Risk-of-bias Human Mechanistic Studies Following
Fluoride Exposure

An interactive version of Figure D-18 and additional study details in HAWC here (NTP 2019).

Figure D-19. Risk-of-bias Heatmap for High Risk-of-bias Human Mechanistic Studies Following
Fluoride Exposure

An interactive version of Figure D-19 and additional study details in HAWC here (NTP 2019).

Figure D-20. Risk-of-bias Bar Chart for High Risk-of-bias Human Mechanistic Studies Following
Fluoride Exposure

An interactive version of Figure D-20 and additional study details in HAWC here (NTP 2019).

D-10
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

D.2. Studies in Non-human Animals

Figure D-21. Risk-of-bias Heatmap for New Developmental Animal Learning and Memory Studies
Following Fluoride Exposure

An interactive version of Figure D-21 and additional study details in HAWC here (NTP 2019).

Figure D-22. Risk-of-bias Bar Chart for New Developmental Animal Learning and Memory Studies
Following Fluoride Exposure

An interactive version of Figure D-22 and additional study details in HAWC here (NTP 2019).

D-11
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-23. Risk-of-bias Heatmap for New Adult Animal Learning and Memory Studies
Following Fluoride Exposure

An interactive version of Figure D-23 and additional study details in HAWC here (NTP 2019).

Figure D-24. Risk-of-bias Bar Chart for New Adult Animal Learning and Memory Studies
Following Fluoride Exposure

An interactive version of Figure D-24 and additional study details in HAWC here (NTP 2019).

D-12
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-25. Risk-of-bias Heatmap for Low Risk-of-bias Animal Biochemical Studies Following
Fluoride Exposure

An interactive version of Figure D-25 and additional study details in HAWC here (NTP 2019).

Figure D-26. Risk-of-bias Bar Chart for Low Risk-of-bias Animal Biochemical Studies Following
Fluoride Exposure

An interactive version of Figure D-26 and additional study details in HAWC here (NTP 2019).

D-13
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-27. Risk-of-bias Heatmap for High Risk-of-bias Animal Biochemical Studies Following
Fluoride Exposure

An interactive version of Figure D-27 and additional study details in HAWC here (NTP 2019).

Figure D-28. Risk-of-bias Bar Chart for High Risk-of-bias Animal Biochemical Studies Following
Fluoride Exposure

An interactive version of Figure D-28 and additional study details in HAWC here (NTP 2019).

D-14
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-29. Risk-of-bias Heatmap for Low Risk-of-bias Animal Neurotransmission Studies
Following Fluoride Exposure

An interactive version of Figure D-29 and additional study details in HAWC here (NTP 2019).

Figure D-30. Risk-of-bias Bar Chart for Low Risk-of-bias Animal Neurotransmission Studies
Following Fluoride Exposure

An interactive version of Figure D-30 and additional study details in HAWC here (NTP 2019).

D-15
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-31. Risk-of-bias Heatmap for High Risk-of-bias Animal Neurotransmission Studies
Following Fluoride Exposure

An interactive version of Figure D-31 and additional study details in HAWC here (NTP 2019).

Figure D-32. Risk-of-bias Bar Chart for High Risk-of-bias Animal Neurotransmission Studies
Following Fluoride Exposure

An interactive version of Figure D-32 and additional study details in HAWC here (NTP 2019).

D-16
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-33. Risk-of-bias Heatmap for Low Risk-of-bias Animal Oxidative Stress Studies
Following Fluoride Exposure

An interactive version of Figure D-33 and additional study details in HAWC here (NTP 2019).

Figure D-34. Risk-of-bias Bar Chart for Low Risk-of-bias Animal Oxidative Stress Studies
Following Fluoride Exposure

An interactive version of Figure D-34 and additional study details in HAWC here (NTP 2019).

D-17
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-35. Risk-of-bias Heatmap for High Risk-of-bias Animal Oxidative Stress Studies
Following Fluoride Exposure

An interactive version of Figure D-35 and additional study details in HAWC here (NTP 2019).

Figure D-36. Risk-of-bias Bar Chart for High Risk-of-bias Animal Oxidative Stress Studies
Following Fluoride Exposure

An interactive version of Figure D-36 and additional study details in HAWC here (NTP 2019).

D-18
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-37. Risk-of-bias Heatmap for Low Risk-of-bias Animal Histopathology Studies Following
Fluoride Exposure

An interactive version of Figure D-37 and additional study details in HAWC here (NTP 2019).

Figure D-38. Risk-of-bias Bar Chart for Low Risk-of-bias Animal Histopathology Studies
Following Fluoride Exposure

An interactive version of Figure D-38 and additional study details in HAWC here (NTP 2019).

D-19
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Figure D-39. Risk-of-bias Heatmap for High Risk-of-bias Animal Histopathology Studies Following
Fluoride Exposure

An interactive version of Figure D-39 and additional study details in HAWC here (NTP 2019).

Figure D-40. Risk-of-bias Bar Chart for High Risk-of-bias Animal Histopathology Studies
Following Fluoride Exposure

An interactive version of Figure D-40 and additional study details in HAWC here (NTP 2019).

D-20
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Appendix E. Details for Low Risk-of-bias Studies

Table of Contents
E.1. IQ Studies ............................................................................................................................. E-2
E.2. Other Neurodevelopmental Studies .................................................................................... E-65

E-1
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

E.1. IQ Studies

E.1.1. Bashash et al. (2017)

E.1.1.1. Study Details

• Study design: Prospective cohort

• Population: Early Life Exposures in Mexico to Environmental Toxicants
(ELEMENT) participants (pregnant mothers and their children aged 4 or 6–12 years).
• Study area: Mexico City, Mexico
• Sample size: 299 mother-child pairs, of whom 211 had data for the IQ analyses and
287 had data for the general cognitive index (GCI).
• Data relevant to the review: Adjusted and unadjusted associations between IQ scores
and maternal or child’s urinary fluoride concentrations.
• Reported association with fluoride exposure: Yes: Significant inverse association
between maternal urinary fluoride and IQ score (adjusted β = −2.50 per 0.5-mg/L
increase; 95% CI: −4.12, −0.59). Significant inverse association between maternal
urinary fluoride and GCI score (adjusted β per 0.5 mg/L increase = −3.15; 95% CI:
−5.42, −0.87). No significant associations with children’s urinary fluoride.
E.1.1.2. Risk of Bias

• Author contacts:
o Authors were contacted for additional information on whether clustering was
addressed. The authors provided results from additional models with cohort as a
random effect, which informed the rating decision for the following risk-of-bias
domains: Other.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Study participants were selected from two different cohorts from three
hospitals in Mexico City that serve low-to-moderate income populations. One
cohort was from an observational study of prenatal lead exposure and
neurodevelopment outcomes, and the other was from a randomized trial of the
effect of calcium on maternal blood lead levels. The authors state that participants
had no history of psychiatric disorders, high-risk pregnancies, gestational
diabetes, illegal drug use, or continuous prescription drugs, but no information on
smoking habits was considered. Study populations appear to be similar, but there
may be some differences because subjects were selected from two different
cohorts that were recruited from slightly different time periods.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposure groups were similar despite the subjects coming from different original
study populations wherein different methods were used for recruitment.
• Confounding:
o Rating: Probably low risk of bias (+)

E-2
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Summary: Data were collected via questionnaire on maternal age, education,

marital status at first prenatal visit, birth order, birth weight, gestational age at
delivery, maternal smoking, maternal IQ, and HOME scores. All models were
adjusted for gestational age at birth, sex, birth weight, birth order, age at testing,
maternal marital status, smoking history, age at delivery, maternal IQ, education,
and cohort, with additional testing for children’s urinary fluoride, mercury, lead,
and calcium. Sensitivity analyses additionally adjusted for HOME score.
Important covariates not considered included BMI, iodine deficiency, arsenic, and
maternal mental health and nutrition. Arsenic is assumed not to be a potential co-
exposure in this population because the study authors did not discuss it as an
issue, although other co-exposures were considered. Arsenic is included in the
water quality control program in Mexico City and is thus not considered a
concern in this population.
o Potentially important study-specific covariates: All key covariates were
addressed.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias based on direct evidence that key
covariates, including other potential co-exposures, were addressed and indirect
evidence that the methods used to collect the information were valid and reliable
and that arsenic is not likely to be an issue in this study population.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: Although there was a large amount of attrition, the study authors
clearly describe all reasons for attrition and also provide characteristics to
compare those participants included to those excluded. There were some slight
differences between those included and those excluded, but there is no evidence
to indicate that the attrition would potentially bias the results.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Definitely low risk of bias (++)
o Summary: Urinary fluoride concentrations were determined in spot urine samples
(second morning void) collected from mothers (during at least one trimester) and
children ages 6–12 years. Fluoride content was measured using ion-selective
electrode-based assays. Quality control methods were described, including
between-laboratory correlations. All samples were measured in duplicate.
Extreme outliers were excluded. Urinary dilution was addressed by using
creatinine-adjusted levels.
 Direction/magnitude of effect size: Not applicable.

E-3
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Definitely low risk of bias based on direct evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Definitely low risk of bias (++)
o Summary: Outcome was assessed using the McCarthy Scales of Children’s
Abilities (MSCA) in 4-year-old children (translated into Spanish) and the
Wechsler Abbreviated Scale of Intelligence (WASI) in 6–12-year-olds. The
WASI is a well-established test, and the validity of both tests is well documented
by the authors. Inter-examiner reliability was evaluated and reported with a
correlation of 0.99 (++ for methods). The study report stated that psychologists
were blind to the children’s fluoride exposure (++ for blinding). Overall rating for
methods and blinding = ++.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessor was blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Definitely low risk of bias (++)
o Summary:
 Statistical analyses: Statistical analyses used were appropriate for the study.
Statistical tests of bivariate associations (using Chi-square tests for categorical
variables and analysis of variance [ANOVA]) were used to compare the
means of the outcomes or exposure within groups based on the distribution of
each covariate. Generalized additive models (GAMs) were used to estimate
the adjusted association between fluoride exposure and measures of children’s
intelligence. Residual diagnostics were used to examine model assumptions
and identify any potentially influential observations. Results are reported as
adjusted regression slopes and 95% CIs. In sensitivity analyses, regression
models accounted for clustering at the cohort level by using cohort as a fixed
effect in the models. Although using cohort as a random effect would be more
appropriate, using individual-level exposure data and accounting for
numerous important covariates in the models likely captured the cohort effect.
Additional models with cohort as a random effect were also subsequently
made available via personal communication with the study authors and
showed similar results to the main model.
 Other potential concerns: None identified.

E-4
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Definitely low risk if bias is based on direct evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Definitely or probably low
risk-of-bias ratings in confounding, exposure, and outcome. Study strengths include
individual exposure assessment measurements, blinding of outcome assessor to
participants’ fluoride exposure, and the prospective cohort study design.

E.1.2. Choi et al. (2015)

E.1.2.1. Study Details

• Study design: Cross-sectional

• Population: First-grade children (ages 6–8 years)
• Study area: Mianning County in southern Sichuan, China
• Sample size: 51 first-grade children
• Data relevant to the review: Associations between urinary or drinking water fluoride
levels and IQ measures (digit span for auditory span and working memory and block
design for visual organization and reasoning components of WISC-IV only). Study
also had information based on dental fluorosis score.
• Reported association with fluoride exposure: Yes: Significant association between
dental fluorosis and digit span scores. Compared to the normal/questionable dental
fluorosis, the moderate/severe dental fluorosis group was associated with significantly
lower total (adjusted β = −4.28; 95% CI: −8.22, −0.33) and backward (adjusted
β = −2.13; 95% CI: −4.24, −0.02) digit span scores. No associations between urinary
fluoride and total digit span and log-transformed fluoride in urine (adjusted
β = −1.67; 95% CI: −5.46, 2.12) and between drinking water and total digit span
(adjusted β = −1.39; 95% CI: −6.76, 3.98). Other outcomes were not significantly
associated with fluoride exposure.
E.1.2.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Definitely low risk of bias (++)
o Summary: Subjects were selected during the same time frame using the same
methods. Fifty-one first-grade children residing in Mianning County in southern
Sichuan, China were included in this pilot study. It is not specified whether the 51
children represented all the first-grade children from this area or whether some
refused to participate. Children who did not speak Chinese, were not students at
the Primary School of Sunshui Village in Mianning County, or those with chronic
or acute disease that might affect neurobehavioral function tests were excluded.
Demographic characteristics are presented in Table 1 of the study, which indicates

E-5
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

that subjects were similar. Important covariates are adjusted for in the statistical
analyses.
o Basis for Rating: Definitely low risk of bias based on direct evidence that the
exposure groups were similar and were recruited within the same time frame
using the same methods with no evidence of differences in participation/response
rates.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: The parents or guardians completed a questionnaire on demographic
and personal characteristics of the children (sex, age at testing, parity, illnesses
before age 3, and past medical history) and caretakers (age, parity, education and
occupational histories, residential history, and household income). A 20-μL
capillary blood sample was collected at the school by a Mianning County Center
for Disease Control (CDC) health practitioner and tested for possible iron
deficiency, which could have been used as a covariate of neurodevelopmental
performance. Important covariates that were not assessed include maternal BMI,
parental mental health, maternal smoking status, maternal reproductive factors,
parental IQ, and HOME score. However, the study authors noted that
confounding bias appeared to be limited due to the minimal diversity in the social
characteristics of the subjects. The study authors indicated that CDC records
documented that levels of other contaminants, including arsenic and lead, were
very low in the area. Iodine differences were not specifically addressed, but there
is no indication from the information provided that this might have been a
concern.
o Potentially important study-specific covariates: All key covariates were
considered in this study.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias because there is direct evidence that
the key covariates were considered and indirect evidence that co-exposure to
arsenic was likely not an issue in this area and that methods used for collecting
the information were valid and reliable.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: The majority of results were reported for the 51 children stated to be
included in the pilot study. In Table 5 of the study, the N for each dental fluorosis
category totals only 43, but the text indicates 8 children did not have a Dean Index
because permanent teeth had not erupted.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:

E-6
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Rating: Probably low risk of bias (+)

o Summary: The study used three different measurements of fluoride exposure: well
water fluoride concentrations from the residence during pregnancy and onwards,
fluoride concentrations from children’s first morning urine samples, and degree of
children’s dental fluorosis. Fluoride concentrations in community well water were
measured and recorded by Mianning County CDC; specific analytic methods
were not reported, but it is likely that standard methods were used because the
analyses were conducted by the CDC and were likely the same as those used to
measure the fluoride in urine. Migration of subjects was noted to be limited. Well
water fluoride concentrations of the mother’s residence during pregnancy and
onward were used to characterize a child’s lifetime exposure. To provide a
measure of the accumulated body burden, each child was given a 330-mL (11.2-
oz) bottle of Robust© distilled water (free from fluoride and other contaminants)
to drink the night before the clinical examinations, after emptying the bladder and
before bedtime. The first urine sample the following morning was collected at
home, and the fluoride concentration was determined on a 5-mL sample using an
ion-specific electrode at the Mianning CDC. There is no indication that urinary
fluoride levels accounted for dilution, nor was it clear that the method of
administering water to the children and collection methods sufficiently controlled
for differences in dilution. One of the investigators, a dentist, performed a blinded
dental examination on each child’s permanent teeth to rate the degree of dental
fluorosis using the Dean Index. The Dean Index is a commonly used index in
epidemiological studies and remains the gold standard in the dentistry
armamentarium. The Index has the following classifications: normal,
questionable, very mild, mild, moderate, and severe. Quality control (QC)
procedures are not reported but were likely appropriate.
 Direction/magnitude of effect size: Current levels were used to assess lifetime
exposure. This is likely to be a non-differential exposure misclassification,
and direction of bias is unknown. Because subject migration appears to be
limited, it is likely that the current fluoride levels are adequate reflections of
past exposure. Dental fluorosis would be an indicator that exposure occurred
in the past, and there was a fair correlation between degree of dental fluorosis
and current urine and water fluoride levels, with both increasing with
increasing levels of dental fluorosis.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measure exposure.
• Outcome:
o Rating: Probably low risk of bias (+)
o Summary: The study authors adopted culture-independent tests considered
feasible for children aged 6 to 8 years. The Wide Range Assessment of Memory
and Learning (WRAML) was used for the assessment of memory and learning.
Three subtests were also used. The Finger Windows subtest assesses sequential
visual memory. The Design Memory subtest assesses the ability to reproduce
designs from memory following a brief exposure. The Visual Learning subtest

E-7
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

assesses the ability to learn the locations of pictured objects over repeated
exposures. The Wechsler Intelligence Scale for Children-Revised (WISC-IV)
includes digit span for auditory span and working memory and block design for
visual organization and reasoning. The grooved pegboard test assesses manual
dexterity. The tests used have been validated on a Western population. Although
there is no information provided to indicate that the tests were validated on the
study population, the study authors indicated that the tests were culture-
independent (+ for methods). Blinding of the outcome assessors to participants’
fluoride exposure, or steps to minimize potential bias were not reported. However,
it is unlikely that the assessors had knowledge of the individual exposure as
children all came from the same area, and water and urine levels were tested at the
CDC. (+ for blinding). Overall = +.
o Basis for rating: Probably low risk of bias based on indirect evidence that all
outcomes were assessed using instruments that were valid and reliable in the
study population, and that the outcome assessors were blind to participants’
fluoride exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Statistical analyses are appropriate. Multiple regression
models evaluate the associations between exposure indicators and test scores
after adjusting for covariates. Specific regression models are not described or
refenced, just stated to be “standard regression analysis with confounder
adjustment.” The distributions of fluoride concentrations in urine and water
are skewed and log10-transformed to approximate a Gaussian distribution
(test not specified). Results are reported as adjusted regression slopes and 95%
CIs. There is no evidence that residual diagnostics were used to examine
model assumptions; however, the impact on the effect estimates is expected to
be minimal.
 Other potential concerns: It should be noted that this study was a pilot study
and, therefore, had a relatively small sample size (i.e., 51 children).
o Basis for rating: Probably low risk if bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in the confounding, exposure, and outcome domains. Study strengths include
individual fluoride measurements with blinding at outcome assessment likely. All key

E-8
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

covariates and many other important covariates were considered in the study design
or analysis.

E.1.3. Cui et al. (2018)

E.1.3.1. Study Details

• Study design: Cross-sectional

• Population: School children aged 7–12 years from four schools in two districts in
China with different fluoride levels
• Study area: Jinghai and Dagang in Tianjin City, China
• Sample size: 323 school children
• Data relevant to the review: IQ scores by urine fluoride levels.
• Reported association with fluoride exposure: Yes: Significant inverse association
between urinary fluoride and IQ score (adjusted β = −2.47 per 1-log-mg/L increase;
95% CI: −4.93, −0.01).
E.1.3.2. Risk of Bias

• Author contacts:
o Authors were contacted in June 2019 to obtain additional information for risk-of-
bias evaluation. Additional information provided by the authors informed the
rating decision for the following risk-of-bias domains: Detection (outcome
assessment).
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Four schools were selected from the same district in China. The
schools were selected based on levels of fluoride in the local drinking water and
the degree of school cooperation. No details were provided on the number of
schools in given areas or the difficulty in getting school cooperation. It was noted
that the residents in the four areas had similar living habits, economic situations,
and educational standards. Although authors do not provide the specific data to
support this, fluoride levels and IQ scores were provided by different subject
characteristics. The areas were classified as historically endemic fluorosis and
non-fluorosis. Cluster sampling was used to select the grades in each school
according to previously set child ages, and classroom was randomly selected with
all students within a selected classroom included. Reasons for exclusion do not
appear to be related to exposure or outcome.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposure groups were similar and recruited within the same time frame using the
same methods, with no evidence of differences in participation/response rates.
• Confounding:
o Rating: Probably low risk of bias (+)

E-9
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Summary: The measurements of all covariates were obtained by structured

questionnaires that were completed by children with the help of their parents.
Covariates that were assessed include: sex, age, child’s ethnicity, child’s BMI,
birth (normal vs. abnormal), mother’s age at delivery, mother’s education, income
per family member, mother’s smoking/alcohol during pregnancy, family member
smoking, environmental noise, iodine region (non-endemic vs. iodine-excess-
endemic area), factory within 30 m of residence, iodine salt, diet supplements,
seafood/pickled food/tea consumption, surface water consumption, physical
activity, stress, anger, anxiety/depression, trauma, having a cold 5 times a year,
thyroid disease in relatives, mental retardation in relatives, and cancer in relatives.
Covariates not considered include parity, maternal and paternal IQ, and quantity
and quality of caregiving environment (e.g., HOME score). The authors report
that there were no other environmentally toxic substances that might have
affected intelligence, such as high arsenic or iodine deficiency according to the
Tianjin Centers for Disease Prevention and Control.
o Potentially important study-specific covariates: All key covariates were
considered in this study.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias because there is indirect evidence that
the key covariates were considered, methods for collecting the information were
valid and reliable, and co-exposure to arsenic was likely not an issue in this area.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: Of the 400 children enrolled, 35 were excluded because they did not
have informed consent signed by a guardian or they moved out of the area. Forty-
two children were excluded because they did not have a DRD2 genotyping
measurement. It is unclear whether these children were from the same schools or
whether they were evenly distributed throughout the study area. It is also unclear
whether the excluded subjects were similar to those included in the study. In the
study, some analyses had fewer than the 323 subjects, but this seems reasonable
given the subgroups that were being evaluated.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Although children were selected based on area fluoride levels, fluoride
in the urine was used in the analysis. Urine was collected from each child during
the morning of enrollment and analyzed within a week. Fluoride levels were
measured using an ion-selective electrode according to the China standard. A
brief description of the method was provided, but no QC methods were reported.
The study authors did not account for urinary dilution in the spot samples.

E-10
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

 Direction/magnitude of effect size: Not accounting for dilution could cause

some exposure misclassification. The direction and magnitude would depend
on where the differences occurred.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using acceptable methods that provide
individual levels of exposure.
• Outcome:
o Rating: Probably low risk of bias (+)
o Summary: IQ was measured by professionals using the Combined Raven’s Test–
The Rural in China method, which is the appropriate test for the study population
(++ for methods). Blinding or other methods to reduce bias were not reported.
Although it was unlikely that the outcome assessor would have knowledge of the
child’s urine fluoride levels, there was potential that they would know whether the
child was from an endemic or non-endemic area if the IQ tests were conducted at
the child’s school, and there was no information provided on how the IQ tests
were administered. Correspondence with the study author noted the cross-
sectional nature of the study with outcome and exposure assessed at the same
time, making the outcome assessors blind to the exposure status of participants.
However, there was still potential for knowledge of the area (+ for blinding).
o Basis for rating: Probably low risk of bias based on indirect evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessors were blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes in the abstract, introduction, and methods are reported in
sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Statistical analyses were appropriate. Multiple linear
regression models were applied to evaluate the relationship between urine
fluoride levels and IQ scores, accounting for numerous important covariates.
The urinary fluoride levels were log-transformed due to a skewed distribution.
Residual diagnostics were used to examine model assumptions. Model
robustness was tested through bootstrap, sensitivity analysis after excluding
potential outliers, and cross-validation techniques. Results are reported as
adjusted regression slopes and 95% CIs. The analysis did not account for
clustering at the school level or at the grade level (students were from four
schools in grades selected via a clustered sampling method). There is no
evidence that the sampling strategy was otherwise accounted for via sampling

E-11
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

weights. The impact of these factors on the effect estimates is expected to be

minimal given the use of individual-level data and adjustment for several
important covariates.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate, and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in confounding, exposure, and outcome. Study strengths include individual
exposure assessment measurements but are limited by the cross-sectional study
design and lack of accounting for urine dilution. All key covariates were considered
in the study design or analysis.

E.1.4. Cui et al. (2020)

E.1.4.1. Study Details

• Study design: Cross-sectional

• Population: School children aged 7–12 years
• Study area: Tianjin City, China (one randomly selected school from each district
based on iodine levels in the water), presumably was an expansion of the Cui et al.
(2018) study
• Sample size: 498 school children
• Data relevant to the review: Mean IQ scores by urine fluoride levels.
• Reported association with fluoride exposure: Yes: A 2-point decrease in IQ was
observed in the highest urinary fluoride group compared to the lowest urinary fluoride
group (i.e., 110.00 in ≥2.5-mg/L group versus 112.16 in <1.6-mg/L group); however,
the results did not achieve statistical significance based on a one-way ANOVA
comparing the three different urinary fluoride categories only.
E.1.4.2. Risk of Bias

• Author contacts:
o Authors were not contacted for the 2020 publication. Authors were contacted in
June 2019 for additional information on the Cui et al. (2018) publication.
Additional information provided by the authors regarding Cui et al. (2018)
informed the rating decision for the following risk-of-bias domains: Detection
(outcome assessment).
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Subjects were recruited from 2014 to 2018. One school was selected
from each district where water concentrations of water iodine were <10, 10–100,
100–150, 150–300 and >300 µg/L. In each school, classes were randomly
sampled for the appropriate age group of 7–12 years old. A table of subject
characteristics was provided by IQ. A total of 620 children were recruited, and

E-12
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

122 children who did not have complete information or enough blood sample
were excluded. Reasons for exclusion do not appear to be related to exposure or
outcome. The characteristics of the 498 included children are presented.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposure groups were similar and were recruited within the same time frame
using the same methods, with no evidence of differences in participation/response
rates.
• Confounding:
o Rating: Probably high risk of bias (−)
o Summary: It was noted by the study authors that there were no other
environmental poisons except water fluoride. Other studies also conducted in this
area of China noted specifically that arsenic was not a concern. Iodine was
addressed as that was one of the main points of the study. Twenty-one factors
(provided in Table 1 of the study) were selected as covariates, and a homemade
questionnaire of unspecified validity was used for obtaining the information. It
was noted that child age, stress, and anger were significantly associated with IQ
although it is unclear whether these varied by fluoride level. However, Cui et al.
(2018) indicate that fluoride was not significantly associated with stress and
anger, and it was assumed that results would be similar for this study even though
more children were included.
o Potentially important study-specific covariates: Age (children 7–12 years old)
 Direction/magnitude of effect size: Age is a key covariate for IQ, even in the
narrow age range evaluated in this study. The direction of the association may
depend on the number of children in each age group within the different
urinary fluoride categories; however, these data were not provided. In general,
there were fewer subjects ≤9 years of age (i.e., 111) compared to >9 years of
age (i.e., 387) with a significantly higher IQ in the ≤9-year-old age group.
Therefore, if exposure were higher in the older subjects, this could likely bias
the association away from the null.
o Basis for rating: Probably high risk of bias because there is indirect evidence that
age was not addressed as a key covariate, and it may be related to both IQ and
exposure.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: Of the 620 children recruited, 122 (20%) were excluded due to
incomplete information or inadequate blood sample. No information was provided
to indicate whether there were similarities or differences in the children included
versus the children excluded, but exclusion is unlikely to be related to either
outcome or exposure.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.

E-13
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Children’s morning urine was collected with a clean polyethylene tube,
and fluoride was measured using a fluoride ion-selective electrode following
Chinese standard WS/T 89-2015. A brief description was provided, but no QC
methods were reported. The study authors do not account for urinary dilution in
the spot samples.
 Direction/magnitude of effect size: Not accounting for dilution could cause
some exposure misclassification. The direction and magnitude would depend
on where the differences occurred.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using acceptable methods that provide
individual levels of exposure.
• Outcome:
o Rating: Probably low risk of bias (+)
o Summary: IQ was measured using the Combined Raven’s Test, which is an
appropriate test for the study population (++ for methods). Blinding was not
mentioned; however, the outcome assessors would not likely have had knowledge
of the child’s urinary fluoride. Subjects appear to have been recruited based on
iodine levels; therefore, it is unlikely that there would have been any knowledge
of potential fluoride exposure. Correspondence with the study authors for the Cui
et al. (2018) study also indicated that the outcome assessors would have been
blind.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessors were blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes in the abstract, introduction, and methods are reported in
sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: One-way ANOVA was used to make comparisons
between mean IQ by urinary fluoride levels. Consideration of heterogeneity of
variances was not reported. There is no adjustment for covariates or for
clustering of children at the school level. There is no evidence that the
sampling strategy was otherwise accounted for (i.e., via sampling weights).
The impact of these factors on the effect estimates is expected to be minimal

E-14
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

given the use of individual-level data. The primary focus of the study was to
evaluate associations between IQ and thyroid hormone or dopamine levels
(not between IQ and fluoride levels). It should also be noted that more
advanced analyses used for thyroid hormone- and dopamine-IQ associations
still lacked adjustment for school and accounting for clustering of children
from the same school.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate, and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in exposure and outcome. Study strengths include individual exposure
assessment measurements, but the study is limited by the cross-sectional study
design, lack of accounting for urine dilution, and lack of addressing age as a key
covariate.

E.1.5. Ding et al. (2011)

E.1.5.1. Study Details

• Study design: Cross-sectional

• Population: Elementary school children aged 7–14 years old
• Study area: Hulunbuir City, Inner Mongolia, China
• Sample size: 331 school children
• Data relevant to the review: IQ mean difference based on 10 categories of urine
fluoride. Adjusted associations between urinary fluoride and IQ scores.
• Reported association with fluoride exposure: Yes: Significant inverse association
between urinary fluoride and IQ score (adjusted β = −0.59 per 1-mg/L increase; 95%
CI: −1.09, −0.08).
E.1.5.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: The study randomly selected 7–14-year-olds (n = 340) from four
nearby elementary schools in Hulunbuir. The four elementary schools appeared to
be very similar in teaching quality. The study authors noted that they followed the
principles of matching social and natural factors like economic situation,
educational standards, and geological environments as much as possible;
however, how this was done is unclear and no table of study subject
characteristics by group was provided.

E-15
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposure groups were similar and were recruited within the same time frame
using the same methods, with no evidence of differences in participation/response
rates.
• Confounding:
o Rating: Probably high risk of bias (−)
o Summary: It was noted that none of the four sites had other potential neurotoxins,
including arsenic, in their drinking water. Details were not provided, except for a
reference supporting the statement. In addition, iodine deficiency was noted as not
being an issue in any of the four areas. Age was the only key covariate adjusted
for in the regression model. Although dental fluorosis severity by % female was
reported, not enough data were provided to determine whether sex should have
been considered in the regression model. The study authors note that future
studies will include covariates such as parents’ educational attainment, mother’s
age at delivery, and household income.
o Potentially important study-specific covariates: Sex
 Direction/magnitude of effect size: There is not enough information to
determine whether there was an effect from sex. There were some differences
in dental fluorosis level by sex, but it is unclear how this might impact the
results or whether the distribution of sex differed by age.
o Basis for rating: Probably high risk of bias based on indirect evidence that there
were differences in sex that were not considered in the study design or analyses.
• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: Data were relatively complete (i.e., <5% loss). Of the 340 subjects
selected for inclusion, 5 were excluded because they lived in the area for less than
a year with an additional 4 not consenting to participate.
o Basis for rating: Definitely low risk of bias based on direct evidence that
exclusion of subjects from analysis was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Spot urine samples were collected and measured using China CDC
standards. All samples were analyzed twice using a fluoride ion-selective
electrode. Recovery rates were specified as 95%–105% with an LOD of
0.05 mg/L. Water samples were collected from small-scale central water supply
systems and tube wells with handy pumps and were processed using standard
methods, similar to the urine samples. Quality assurance validation was reported.
A blind professional examiner evaluated the children for dental fluorosis using
Dean’s Index. All urine and water samples were above the LOD. Urine levels
were the primary exposure assessment measures used in the analysis. The study

E-16
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

authors did not account for urinary dilution in the spot samples. The mean urine
fluoride concentration was correlated with the dental fluorosis levels.
 Direction/magnitude of effect size: Spot urine samples that did not account for
dilution could have exposure misclassification. The misclassification is likely
non-differential, and the potential direction of bias is unknown.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measure exposure.
• Outcome:
o Rating: Probably low risk of bias (+)
o Summary: IQ was determined using the Combined Raven’s Test–The Rural in
China (CRT-RC3) (++ for methods). Although blinding was not reported, it is
unlikely that the IQ assessors had knowledge of the children’s urine levels or even
of the water levels from the four sites, as these were sent to a separate lab for
testing (+ for blinding). Overall rating for methods and blinding = +.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessors were blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Statistical analyses were reasonable (ANOVA and
multiple linear regression), but consideration of homogeneity of variance was
not reported. The NASEM committee’s review (NASEM 2021) pointed out a
potential concern regarding the lack of accounting for clustering at the school
level because children were selected from four elementary schools. However,
as outlined in the Selection domain, the authors stated that they followed the
principles of matching social and natural factors like economic situation,
educational standards, and geological environments to the extent possible and
that the four elementary schools appeared to be very similar in teaching
quality. There is no evidence that the sampling strategy was otherwise
accounted for (i.e., via sampling weights). The impact of these factors on the
effect estimates is expected to be minimal given the use of individual-level
data and adjustment for age as a key covariate.
 Other potential concerns: None identified.

E-17
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and that there were no other potential threats
of risk of bias.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in exposure and outcome. Study strengths include individual exposure
assessment measurements, but the study is limited by the cross-sectional study
design, lack of accounting for urine dilution, and lack of consideration of sex as a key
covariate.

E.1.6. Green et al. (2019)

E.1.6.1. Study Details

• Study design: Prospective cohort

• Population: Maternal-Infant Research on Environmental Chemicals (MIREC)
participants (pregnant mothers and their children aged 3–4 years)
• Study area: 10 cities, Canada
• Sample size: 512 mother-child pairs (238 from non-fluoridated areas, 162 from
fluoridated areas; 264 females, 248 males)
• Data relevant to the review: Adjusted associations between maternal urinary fluoride
across all three trimesters, estimated maternal fluoride intake, or drinking water
fluoride and IQ.
• Reported association with fluoride exposure: Yes: Significant inverse association
between maternal urinary fluoride and IQ in boys (adjusted β = −4.49 per 1-mg/L
increase). No significant associations in girls or in both sexes combined. Significant
inverse associations between maternal intake and IQ in both sexes combined
(adjusted β = −3.66 per 1-mg/day increase; 95% CI: −7.16, −0.15). Significant
inverse associations between drinking water fluoride and IQ in both sexes combined
(adjusted β = −5.29 per 1-mg/L increase; 95% CI: −10.39, −0.19). No significant
effect modification by sex for maternal fluoride intake and drinking water fluoride
associations.
E.1.6.2. Risk of Bias

• Author contacts:
o Authors were contacted in June 2019 for additional information for the risk-of-
bias evaluation. Additional information provided by the authors informed the
rating decision for the following risk-of-bias domains: Other.
• Population selection:
o Rating: Definitely low risk of bias (++)
o Summary: Pregnant women were recruited from the same population during the
same time frame and using the same methods as the MIREC program. Methods
were reported in detail.

E-18
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Definitely low risk of bias based on direct evidence that the
exposed groups were similar and were recruited with the same methods during the
same time frame.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: The study considered several possible covariates, including maternal
age, pre-pregnancy BMI, marriage status, birth country, race, maternal education,
employment, income, HOME score, smoking during pregnancy, secondhand
smoke in the home, alcohol consumption during pregnancy, parity, sex, age at
testing, gestational age, birth weight, time of void, and time since last void. The
study also conducted secondary analyses to test for lead, mercury, arsenic, and
PFOA. There is no indication of any other potential co-exposures in this study
population. Iodine deficiency or excess could not be assessed but is not expected
to differentially occur. The study was not able to assess parental IQ or mental
health disorders. Methods used to obtain the information included questionnaires
and laboratory tests.
o Potentially important study-specific covariates: All key covariates were
addressed.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
methods used to collect the information were valid and reliable and direct
evidence that key covariates, including potential co-exposures, were addressed.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: Of the 610 recruited children, 601 (98.5%) completed testing. Of the
601 mother-child pairs, 512 (85.2%) had all three maternal urine samples and
complete covariate data, and 400 (66.6%) had data available to estimate fluoride
intake.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Spot urine samples from all three trimesters of pregnancy were
evaluated using appropriate methods, and results were adjusted for creatinine and
specific gravity. Fluoride intake was estimated based on fluoride water levels, and
information on consumption of tap water and other water-based beverages (e.g.,
tea, coffee) was obtained via questionnaire.
 Direction/magnitude of effect size: There is not any specific direction or
magnitude of bias expected. Urinary fluoride levels are reflective of a recent
exposure. Having measurements from all three trimesters of pregnancy

E-19
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

provides a better representation of actual exposure than a single measurement,

although the potential for missed high exposure is possible. However, the
possibility of the occurrence of missed high exposure would be similar in all
females and would be non-differential. For the fluoride intake, exposure was
based on the fluoride levels in the water at the residence. If women worked
outside the home and the majority of intake occurred from areas outside the
home (and were different from levels in the home), there is potential to bias
toward the null.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Probably low risk of bias (+)
o Summary: The Wechsler Preschool and Primary Scale of Intelligence was
normalized for ages 2.5–<4.0 and sex using the U.S population-based norms.
Blinding was not reported, but it is unlikely that the outcome assessors had
knowledge of the maternal fluoride level or were aware of whether the city had
fluoridated water.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessors were blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes were reported.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Definitely low risk of bias (++)
o Summary:
 Statistical analyses: Multivariate linear regression analyses were used to
evaluate the associations between maternal urinary fluoride and fluoride
intake and children’s IQ scores. Regression diagnostics were used to test
assumptions for linearity, normality, and homogeneity. There were no
potential influential observations (based on Cook’s distance). Sensitivity
analyses showed that the effects of maternal urinary fluoride (MUF), fluoride
intake, and water fluoride were robust to the exclusion of two very low IQ
scores in males (<70). City was accounted for as a covariate in the regression
models published. Additional models with city as a random effect were also
subsequently made publicly available and showed similar results to the main
model.
 Other potential concerns: None identified.

E-20
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Definitely low risk of bias based on direct evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in confounding, exposure, and outcome. Study strengths include individual
exposure assessment measurements, prospective cohort design, and the consideration
of key covariates.

E.1.7. Rocha-Amador et al. (2007)

E.1.7.1. Study Details

• Study design: Cross-sectional

• Population: Children aged 6–10 years
• Study area: Moctezuma (low fluoride, low arsenic) and Salitral (high fluoride, high
arsenic) of San Luis Potosí State and 5 de Febrero (high fluoride, high arsenic) of
Durango State, Mexico
• Sample size: 132 children
• Data relevant to the review: Adjusted associations between urinary or drinking water
fluoride and full-scale IQ, performance IQ, verbal IQ.
• Reported association with fluoride exposure: Yes: Significant inverse associations
between water fluoride and full-scale IQ scores (adjusted β = −10.2 per log-mg/L
increase) and between urinary fluoride and full-scale IQ scores (adjusted β = −16.9
per log-mg/g-creatinine increase; CIs not reported, all p-values < 0.001). Arsenic also
present, but the effect from arsenic in water or urine was smaller (adjusted β = −6.15
per log-µg/L and −5.72 per l-µg/g-creatinine increase, respectively; CIs not reported,
all p-values < 0.01).
E.1.7.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information to inform the risk-of-bias
evaluation because it was not necessary.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: All children in 1st through 3rd grades in three rural areas in Mexico
(n = 480) were screened for study eligibility, including age, time at residence, and
address. Authors report that the three selected communities were similar in
population and general demographic characteristics. Children who had lived in
the area since birth and were 6–10 years old were eligible to participate (n = 308).
Of the 308 children, 155 were randomly selected and the response rate was 85%,
but participation was not reported by area. It was noted, however, that no
significant differences in age, sex, or time of residence were observed between
participants and non-participants. Time frame for selection was not mentioned but
appears to be similar. Sociodemographic characteristics of subjects were provided

E-21
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

in Table 1 of the study. There was a significant difference in SES and transferrin
saturation, but these were considered in the analysis.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
populations were similar, and differences were noted and addressed in the
analysis.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: The study design or analysis accounted for age, sex, SES, transferrin
saturation, weight, height, blood lead levels, and mother's education. Arsenic
levels were highly correlated with fluoride levels; however, arsenic and fluoride
were evaluated alone, and arsenic was found to have less of an effect on IQ than
fluoride. This provides evidence that arsenic had been addressed as a co-exposure
and cannot explain the association between fluoride exposure and decreased IQ.
Smoking was not addressed and methods for measuring many of the covariates
were not reported.
o Potentially important study-specific covariates: Arsenic
 Direction/magnitude of effect size: The presence of arsenic in this study,
which also demonstrated an association, would likely bias the association
away from the null. Although arsenic may contribute to some of the
magnitude of the observed effect of fluoride (the exact impact of arsenic on
the magnitude cannot be assessed), the presence of arsenic does not fully
explain the observed association between fluoride exposure and IQ. The
presence of arsenic may affect the magnitude of the association between
fluoride and IQ, but it has no impact on the direction of the association.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
methods used to collect the information were valid and reliable and direct
evidence that key covariates were addressed.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: Of 155 children randomly selected for study participation, 85%
responded to enroll. According to the authors, there were no significant
differences in age, sex, or time of residence between responders and non-
responders. However, no data were provided to support this, and no breakdown of
responders/non-responders by region was provided. Data were provided for the
132 children agreeing to participate.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Definitely low risk of bias (++)

E-22
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Summary: Urine samples were collected on the same day as psychological

evaluations and were analyzed for fluoride according to NIOSH Method 8308
(Fluoride in Urine). For QC, a reference standard was also used (NIST SRM
2671a). Urine samples were also analyzed for arsenic by using the Atomic
Absorption Spectrophotometer with hydride system and a reference standard for
QC. Levels were adjusted for urinary creatinine levels to account for dilution in
the spot samples. Tap water samples were collected from each child’s home on
the day of biological monitoring. Fluoride was measured with a sensitive, specific
ion electrode. Detailed methods are provided including internal quality controls. It
was noted that in the high fluoride group, it was common to drink bottled water
low in fluoride and to use the tap water only for cooking; therefore, urine was
considered the most appropriate measure of exposure. Only children who had
lived at the same residence since birth were included.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Definitely low risk of bias based on direct evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Probably low risk of bias (+)
o Summary: Neuropsychological profiles were assessed through the WISC-RM
(revised for Mexico). This is a well-established test appropriately adjusted for the
study population. However, no additional validation was provided (+ for
methods). The study report stated that the test assessors were masked to both
arsenic and fluoride water levels (++ for blinding). Overall rating for methods and
blinding = +.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessors were blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Probably high risk of bias (−)
o Summary: It was reported that an interaction between fluoride and arsenic was
measured, but it was noted only in the discussion that the study design precluded
testing statistical interaction between fluoride and arsenic. This provides indirect
evidence of selective reporting.
o Basis for rating: Probably high risk of bias based on indirect evidence that there
was selective reporting.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Statistical analyses used were appropriate for the study.
Multivariate linear analyses were used to evaluate the associations between

E-23
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

fluoride in water and urine and children’s IQ scores. Exposures were natural
log-transformed, but the rationale was not provided. Regression diagnostics
were not used to test model assumptions for linearity, normality, and
homogeneity. The analyses did not account for clustering at the community
level. The three selected communities were similar in population and general
demographic characteristics. Although the analysis used individual-level
exposures rather than area‐level exposures, if the exposure levels within a
certain area were highly correlated (which might be expected), then the results
might still be biased. However, the overall impact on the effect estimates is
expected to be minimal given the use of individual-level data and adjustment
for multiple important covariates.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Definitely or probably low
risk-of-bias ratings in confounding, exposure, and outcome. Study strengths include
individual exposure assessment measurements and blinding of outcome assessors to
participants’ fluoride exposure, but it is limited by the cross-sectional study design
and the inability to completely rule out the influence of arsenic in the results.

E.1.8. Saxena et al. (2012)

E.1.8.1. Study Details

• Study design: Cross-sectional

• Population: Children aged 12 years
• Study area: Madhya Pradesh, India
• Sample size: 170 children
• Data relevant to the review: Mean IQ grade (not standard scores; higher IQ grades
are associated with lower intelligence) by water fluoride quartiles. Correlations
between urinary and water fluoride and IQ scores. Adjusted associations between
water and urinary fluoride and IQ.
• Reported association with fluoride exposure: Yes: Significant negative correlations
between water (r = 0.534) and urinary fluoride and IQ scores (r = 0.542). Significant
increase in mean IQ grade (i.e., increase in proportion of children with intellectual
impairment) with increasing urinary fluoride.
E.1.8.2. Risk of Bias

• Author contacts:
o Authors were contacted in August 2017 to obtain additional information for risk-
of-bias evaluation. Additional information provided by the authors informed the
rating decision for the following risk-of-bias domains: Detection (outcome
assessment).

E-24
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: There was indirect evidence that subjects were similar and were
recruited using the same methods during the same time frame. The study
participants were selected from a stratified cluster of geographic areas based on
fluoride concentration in groundwater. According to the authors, the selected
villages were similar in population and demographic characteristics. Data are
provided to show the breakdown in SES, parental education, height/age, and
weight/height, and no significant differences were noted. Participation was stated
to be voluntary, but participation rates were not provided. It is unclear whether the
170 subjects were selected with 100% participation or whether the 170 subjects
were all who were asked to participate, but it appears that all subjects participated.
Timing of the recruitment was not provided but is assumed to occur during the
same time frame.
o Basis for rating: Probably low risk of bias based on indirect evidence that
subjects were similar and recruited using the same methods during the same time
frame.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: There was indirect evidence that key covariates, including potential co-
exposures, were addressed using reasonable methods. A questionnaire, completed
with the assistance of parents, was used to collect information on child
characteristics (age, sex, height, weight), residential history, medical history
(including illness affecting the nervous system and head trauma), educational
level of the head of the family (in years), and SES of the family. The SES was
recorded according to the Pareek and Trivedi classification. The nutritional status
of the children was calculated using Waterlow’s classification, which defines two
groups for malnutrition using height-for-age ratio (chronic condition) and weight
for height ratio (acute condition). Within both groups, it categorizes the
malnutrition as normal, mildly impaired, moderately impaired, or severely
impaired. Urinary lead and arsenic were analyzed using the atomic absorption
spectrophotometer. Urinary iodine was measured using the Dunn method.
Authors do not report which covariates were included in the multivariate
regression models; however, there was no difference in reported demographic
characteristics. All subjects were the same age, and there was no difference in
iodine, lead, or arsenic between the groups. Mean urinary arsenic levels increased
with increasing fluoride even though there was no significant difference by group.
o Potentially important study-specific covariates: All key covariates were
considered in this study.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
methods used to collect the information were valid and reliable and that key
covariates, including potential co-exposures, were addressed.

E-25
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: Results were provided for all 170 children stated to be included in the
study.
o Basis for rating: Definitely low risk of bias based on direct evidence of no
attrition.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: A sample of 200 mL of drinking water was collected at each child’s
home. The fluoride levels were analyzed by a fluoride ion-selective electrode.
Each subject was also asked to collect a sample of his/her first morning urine. The
fluoride content in the urine was determined using a fluoride ion-selective
electrode. QA/QC and LOD were not reported, and urinary dilution was not
assessed. Although only current levels were measured, children who had changed
their water source since birth were excluded.
 Direction/magnitude of effect size: Spot urine samples that did not account for
dilution could have exposure misclassification. The misclassification is likely
non-differential and not likely to bias in any specific direction. Children who
had changed water source since birth were excluded, but it was not
specifically noted that the fluoride in the water source was stable over the
years.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Probably low risk of bias (+)
o Summary: Intelligence was assessed using Raven’s Standard Progressive Matrices
and categorized into five grade levels. Although it was not noted that the test was
validated to the study population, the test is visual and would be applicable to
most populations (+ for methods). There is no mention of blinding by test
administrators or evaluators, and the exposure groups come from different
geographic areas. It was also not reported who measured the levels of fluoride
from the home or urine samples. Correspondence with the study authors indicated
that the outcome assessors were blind to the children’s fluoride status (++ for
blinding). Overall rating for methods and blinding = +.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessors were blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)

E-26
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: One-way analysis of variance (ANOVA), simple linear
regression, and multiple linear regression were used to compare mean
intelligence grades by water fluoride levels and to assess the association
between grades and urinary fluoride. Consideration of heterogeneity of
variance (for ANOVA) was not reported. Regression diagnostics were not
used to test model assumptions for linearity, normality, and homogeneity.
Given the ordinal nature of the intelligence grade variable (score from 1 to 5),
ordinal logistic regression would have been a more appropriate method. There
was no adjustment for area-level clustering in multivariate analyses (although
subjects were selected via stratified cluster sampling from two areas).
Although the analysis used individual-level exposures rather than area‐level
exposures, if the exposure levels within a certain area were highly correlated
(which might be expected), then the results might still be biased. However, the
overall impact on the effect estimates is expected to be minimal given the use
of individual-level data and adjustment for important covariates.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate, and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in confounding, exposure, and outcome. Study strengths include individual
exposure assessment measurements and the consideration of key covariates, but it
was limited by the cross-sectional study design and lack of addressing dilution in the
urine samples.

E.1.9. Seraj et al. (2012)

E.1.9.1. Study Details

• Study design: Cross-sectional

• Population: Children aged 6–11 years
• Study area: five villages, Makoo, Iran
• Sample size: 293 children
• Data relevant to the review: IQ (mean and distribution) assessed by Raven’s Colored
Progressive Matrices and presented by fluoride area. Adjusted associations between
water fluoride and IQ scores.

E-27
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Reported association with fluoride exposure: Yes: Significant inverse association

between water fluoride and IQ score (adjusted β = −3.865 per 1-mg/L increase ; CIs
not reported); significantly higher IQ score in normal area (97.77 ± 18.91) compared
with medium (89.03 ± 12.99) and high (88.58 ± 16.01) fluoride areas.
E.1.9.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Subjects were selected from five villages in Makoo. The villages were
stated to all be rural with similar general demographic and geographic
characteristics and were comparable in terms of SES and parental occupations.
Children were 6–11 years old. Age, sex, and education were taken into account in
the analysis. No other characteristics were provided or discussed. Participation
rates were not reported. There is indirect evidence that the populations were
similar, and some possible differences were addressed.
o Basis for rating: Probably low risk of bias based on indirect evidence that
subjects were similar and recruited using the same methods during the same time
frame.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Age, sex, dental fluorosis intensity, and educational levels (child’s and
parents’) were evaluated as important covariates. Other covariates such as
smoking were not discussed. Information was obtained from a detailed
questionnaire. Lead was measured but found only in low levels in the drinking
water throughout the study regions. Iodine in the water was also stated to be
measured, and residents were receiving iodine-enriched salt. Arsenic was not
addressed, but there is no evidence that arsenic levels would vary across villages
in this area. Based on water quality maps, co-exposure to arsenic is likely not a
major concern in this area.
o Potentially important study-specific covariates: Arsenic.
 Direction/magnitude of effect size: Conceptually, if there were differential
amounts of arsenic in the different villages, co-exposure to arsenic could bias
the association, with the direction of the bias dependent on where the arsenic
was present; however, arsenic was not expected to be a major concern in this
study area based on water quality maps.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
methods used to collect the information were valid and that key covariates,
including potential co-exposures, were addressed or were not likely to be an issue
in the study area.
• Attrition:

E-28
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Rating: Definitely low risk of bias (++)

o Summary: Attrition was low if it occurred. It was noted that 293 out of 314
children living in the villages were recruited. It is not clear whether 21 children
were excluded based on exclusion criteria or whether they refused to participate;
however, this accounts for less than 10% of the population, and results were
available for all 293 subjects.
o Basis for rating: Definitely low risk of bias based on direct evidence that
exclusion of subjects from analyses was minimal, adequately addressed, and
reasons were documented when subjects were removed from the study or
excluded from analyses.
• Exposure:
o Rating: Probably high risk of bias (−)
o Summary: Exposure was primarily based on area of residence. Fluoride in the
groundwater was analyzed by the SPADNS (Sulfophenylazo
dihydroxynaphthalene-disulfonate) method, utilizing the 4000 UV-Vis
spectrophotometer in the environmental health engineering laboratory of the
Public Health School of the Tehran University of Medical Sciences. Specific
details were not provided on methods of collection or sample locations or whether
these locations represented the primary sources of drinking water for the subjects.
Villages were categorized into normal (0.5–1 ppm), moderate (3.1 ± 0.9 ppm),
and high (5.2 ± 1.1 ppm) fluoride based on the mean fluoride content of all
seasons presumably for the stated 12-year time period. Subjects were stated to be
long-life residents of the village. Dental fluorosis was also measured and
increased in severity with fluoride levels; however, all areas had some degree of
dental fluorosis. Although authors used an average fluoride level in varying
seasons over presumably 12 years, they used a less-established method without
reporting reliability or validity, and they did not provide data to indicate that the
mean was truly representative of the fluoride levels over time and throughout the
village. Although dental fluorosis severity increased with increasing fluoride
levels, the data could also indicate potential exposure misclassification.
 Direction/magnitude of effect size: The presence of dental fluorosis in all
groups indicates that there may have been different exposures in some
children at a younger age. Although there were only about 20 children in the
“normal” fluoride group with very mild to mild dental fluorosis, this could
bias the results toward the null because those children may have experienced a
higher level of fluoride at some point. The other two fluoride groups were
exposed to fluoride levels that likely exceeded those in the “normal” fluoride
group.
o Basis for rating: Probably high risk of bias based on indirect evidence that
exposure was assessed using insensitive methods.
• Outcome:
o Rating: Probably low risk of bias (+)

E-29
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Summary: Intelligence was evaluated using Raven’s Color Progressive Matrices.

This is a well-established method. Although the study authors did not provide
data to indicate that the methods were valid in this study population, the test is
designed to be culturally diverse (+ for methods). The study report stated that test
administrators were blinded to subjects’ exposure status (++ for blinding). Overall
rating for methods and blinding = +.
o Basis for rating: Probably low risk of bias based on indirect evidence that
outcomes were assessed using instruments that were valid and reliable in the
study population, and that the outcome assessors were blind to participants’
fluoride exposure.
• Selective Reporting:
o Rating: Probably low risk of bias (+)
o Summary: All outcomes outlined in the abstract, introduction, and methods were
reported. However, because the study author did not report the method for
obtaining the betas in Table 4 of the study, it is not clear whether these were
adjusted or unadjusted regression coefficients.
o Basis for rating: Probably low risk of bias based on direct evidence that all the
study’s measured outcomes were reported, but the results were not sufficiently
reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Statistical methods for comparisons of IQ level by
exposure groups were reasonable (ANOVA, post hoc test, and Kruskal-Wallis
test), but consideration of heterogeneity of variance was not reported.
Clustering at the village levels was not accounted for in multivariate analyses,
which used area‐level water fluoride levels. Because the exposure levels
within a certain area are highly correlated (which might be expected), the
results are likely to be biased. There was adjustment for some individual-level
important covariates, and the children were from five rural areas with similar
general demographic and geographic characteristics and were comparable in
terms of SES and parental occupations. These factors are expected to mitigate
some of the impact of lack of accounting for clustering, and the overall impact
on the effect estimates is expected to be minimal.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in confounding and outcome. Study strengths include addressing potential key
covariates, but it was limited by the cross-sectional study design and the group-level
exposure data.

E-30
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

E.1.10. Soto-Barreras et al. (2019)

E.1.10.1. Study Details

• Study design: Cross-sectional

• Population: Children aged 9–10 years
• Study area: Chihuahua, Mexico
• Sample size: 161 children
• Data relevant to the review: Water fluoride, urinary fluoride, exposure dose, and
dental fluorosis index by IQ grade.
• Reported association with fluoride exposure: No: Results were not presented to
evaluate an association between fluoride exposure and IQ but to compare fluoride
levels within IQ grades. For this reason, the results of this study are not comparable to
other studies that evaluated IQ scores by fluoride exposure levels. We also note that
no significant differences in measured fluoride levels across IQ grades were observed.
E.1.10.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information to inform the risk-of-bias
evaluation because it was not necessary.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Subjects were selected using a multistage cluster sampling. During the
first stage, 13 public elementary schools were randomly selected from a pool of
73 using a cluster sample design. Secondly, only fourth-grade students were
included. Authors stated that they wanted to keep the same grade level, but there
were no specific details as to why fourth graders were selected as opposed to any
other grade. Lastly, only children whose parents or guardians attended and
responded to the survey were included. There is no information provided on how
the 13 schools selected may have been similar to or different from the 60 schools
not selected. There is no information provided on the number of children in the
fourth grade to know participant rates. It was only noted that 245 children were
examined, but 161 were included after the exclusion rules were applied. Inclusion
and exclusion criteria are presented. Reasons for exclusion do not appear to be
related to exposure or outcome. Characteristics of participants and non-
participants are not compared; however, characteristics of the 161 included
children were provided, and any differences were taken into account in the
analysis.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposed groups were similar and were recruited using similar methods during the
same time frame.
• Confounding:
o Rating: Probably high risk of bias (−)

E-31
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Summary: No covariates were considered when evaluating associations between

fluoride exposure and intelligence; covariates were considered only when
evaluating associations between fluoride levels and dental caries. According to
Table 4 of the study, there was no significant association between IQ grade and
age, sex, parental education, or SES status. No other information was reported or
considered. There is no information on potential co-exposures. According to
water quality maps, the arsenic prediction indicates a greater than 50% probability
of exceeding the WHO guidelines for arsenic of 10 µg/L in areas of Chihuahua,
Mexico.
o Potentially important study-specific covariates: Arsenic.
 Direction/magnitude of effect size: The impact on the direction and magnitude
of effect size is unknown. There is potential for arsenic to occur in the study
area, but it is not known how it relates to fluoride exposure. If they occur
together in the water, it would likely bias the association away from the null;
however, if they occur in different areas, there is potential to bias the
association toward the null.
o Basis for rating: Probably high risk of bias based on indirect evidence that there is
potential for exposure to arsenic that was not sufficiently addressed.
• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: A total of 161 of 245 children were included in the study. Exclusion
criteria are presented and are unrelated to outcome or exposure. For the 161
children, there are no missing outcome data.
o Basis for rating: Definitely low risk of bias based on direct evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Probably low risk of bias (+); Probably high risk of bias (−)
o Summary: Urinary Fluoride (probably low risk of bias): First morning void
urine samples were collected based on NIOSH methods. Water samples were also
stated to be collected, but it does not appear that methods followed any particular
standard, and there is no indication that subjects were provided with collection
containers. Analysis was based on a calibration curve using fluoride ion-selective
electrode. QC methods were mentioned. Based on results, there were values
below detection limits, but LODs or % below LOD were not reported.
Daily fluoride exposure (probably high risk of bias): Daily fluoride exposure
was based on the water fluoride level, drinking water consumption (based on
parental report of how many glasses of water consumed), and body weight.
 Direction/magnitude of effect size: Spot urine samples that did not account for
dilution could have exposure misclassification. The misclassification is likely
non-differential and is not likely to bias in any specific direction. Daily

E-32
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

exposure was based partially on parental report of water consumption. The

direction and magnitude of effect is unknown.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure. The daily fluoride exposure is probably high risk of bias
because there is indirect evidence that the exposure was assessed using methods
of unknown validity.
• Outcome:
o Rating: Probably low risk of bias (+)
o Summary: Intellectual ability was evaluated using Raven’s Colored Progressive
Matrices by an independent examiner. Some details were provided, but it was not
stated that the tests were assessed blind; however, there is no indication that
subjects were from high fluoride areas, and the assessor would not have
knowledge of the urine or water fluoride levels. Results for children were
converted into a percentile according to age (details not provided), and overall
scores were assigned an intellectual grade of I to V as described in the report.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessor was blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods were
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: The Kolmogorov-Smirnov test was used to determine
variable distribution. The Kruskal-Wallis test was used to compare exposure
levels between IQ grades with Dunn’s post hoc test. Multivariate logistic
regression was used to estimate the association between presence of dental
caries and various risk factors. Fluoride levels in drinking water and urine and
fluoride exposure dose were compared across intellectual grades. Children
were from 13 schools selected via stratified cluster sample design. There was
no adjustment for clustering at the school level or for the sampling design.
Although the analysis used individual-level exposures rather than area‐level
exposures, if the exposure levels within a certain school were highly
correlated (which might be expected), then the results might still be biased.
The large number of clusters (13 schools) makes clustering less of a concern,
and the impact on the effect estimates is expected to be minimal.

E-33
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

 Other potential concerns: None identified.

o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in exposure and outcome. Study strengths include individual exposure
assessment measurements and blinding of outcome assessor to participants’ fluoride
exposure, but it is limited by the cross-sectional study design, lack of accounting for
urine dilution, and lack of consideration for potential exposures to arsenic in the study
area. Although the study is considered to have low potential for bias overall, the focus
of the study was to evaluate the relationship between fluoride exposure and lower
rates of dental caries. In terms of evaluating an association between fluoride exposure
and IQ scores, the study is limited by the way the data were reported.

E.1.11. Sudhir et al. (2009)

E.1.11.1. Study Details

• Study design: Cross-sectional

• Population: Children aged 13–15 years
• Study area: Nalgonda district (Andhra Pradesh), India
• Sample size: 1,000 children
• Data relevant to the review: Mean IQ grade (not standard scores) or IQ distribution
by water fluoride strata (<0.7, 0.7–1.2, 1.3–4.0, and >4.0 ppm).
• Reported association with fluoride exposure: Yes: Significant increase in mean and
distributions of IQ grades (i.e., increase in proportion of children with intellectual
impairment) with increasing drinking water fluoride levels.
E.1.11.2. Risk of Bias

• Author contacts:
o Authors were contacted in September of 2017 for additional information related
to risk-of-bias evaluation, but no response was received.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Children were selected from the same general population during the
same time frame and were then broken down into nearly equal exposure groups.
A cross-sectional study was conducted among 13–15-year-old school children of
Nalgonda district, Andhra Pradesh, between August and October 2006. Data were
collected from the school children who were lifelong residents of Nalgonda
district, Andhra Pradesh, and who consumed drinking water from the same source
during the first 10 years of life. A stratified random sampling technique was used.
The entire geographical area of Nalgonda district was divided into four strata
based on different levels of naturally occurring fluoride in the drinking water
supply. Children were randomly selected from schools in the different strata. It

E-34
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

was noted that the 1,000 selected children were equally divided among all four
strata; however, each group did not have 250 children (rather, each had 243–267).
Participation rates were not reported. Exclusion criteria included children who
had a history of brain disease and head injuries, children whose intelligence had
been affected by congenital or acquired disease, children who had migrated or
were not permanent residents, children with orthodontic brackets, and children
with severe extrinsic stains on their teeth. Age and sex data are presented in
Table 1 of the study, but this information is not presented by the different fluoride
groups.
o Basis for rating: Probably low risk of bias based on indirect evidence that
subjects were similar and were recruited using the same methods during the same
time frame.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Data were collected using a self-administered questionnaire and
clinical examination. The questionnaire requested information on demographic
data (appears to cover age and sex), permanent residential address, staple food
consumed, liquids routinely consumed, and aids used for oral hygiene
maintenance (fluoridated or non-fluoridated). SES was measured using the
Kakkar socioeconomic status scale (KSESS) with eight closed-ended questions
related to parental education, family income, father’s occupation, and other
factors. All children were asked to fill out the form, and the answers obtained
were scored using Kakkar socioeconomic status scoring keys. Based on this
scoring, children were divided into three groups: lower class, middle class, or
upper class. Age, sex, and SES were not found to be significantly associated with
IQ. Other covariates, including smoking, were not addressed. Co-exposures such
as arsenic and lead were not addressed; however, there is no indication that lead is
a co-exposure in this population, and arsenic is not likely a major concern in this
area based on water quality maps.
o Potentially important study-specific covariates: Key covariates age, sex, and
measures of SES were similar between exposure groups; however, arsenic was
not considered. Arsenic often occurs in the drinking water along with fluoride in
some Indian populations; however, based on water quality maps, this does not
appear to be an issue in the Nalgonda district of Andhra Pradesh. Iodine
deficiencies are not mentioned.
 Direction/magnitude of effect size: Conceptually, the presence of arsenic
would potentially bias the association away from the null if present with
fluoride. Deficiencies in iodine would likely bias the association away from
the null if present in areas of high fluoride but toward the null if present in
areas of non-high fluoride. Neither of these were considered issues in this
study for reasons noted above.
o Basis for rating: Probably low risk of bias based on indirect evidence that the key
covariates were considered, co-exposure to arsenic was likely not an issue in this
area, and methods used for collecting the information were valid and reliable.

E-35
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: Results were available for the 1,000 children selected to participate.
o Basis for rating: Definitely low risk of bias based on direct evidence of no
attrition.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Children were placed into one of four strata based on the level of
fluoride in drinking water. Collection of water samples was done in the districts.
The placement into strata was based on fluoride levels obtained from documented
records of the District Rural Water Works Department. Once the children were
assigned to strata, it was confirmed that the fluoride level of their drinking water
was within the strata assigned. This was done using the methodology followed in
the National Oral Health Survey and Fluoride Mapping 2002–2003. During the
initial visits to the schools, the children were interviewed regarding their history
of residence and source of drinking water from birth to 10 years. The first child
meeting the criteria was given a bottle for water collection, and the next child was
given a bottle for collection only if the water source was different from that of a
previous child. Children were asked to collect a water sample from the source that
was used in the initial 10 years of their life (and that sample was collected the
next day). It was not reported whether all bottles were returned. The water
samples collected were subjected to water fluoride analysis using an ion-specific
electrode, Orion 720A fluoride meter at District Water Works, Nalgonda to
confirm the fluoride levels in the water before commencement of clinical
examination. LOD and QA/QC details were not reported.
 Direction/magnitude of effect size: There is some potential for exposure
misclassification based on recall of the children on the source of water used in
their first 10 years of life. The misclassification is likely non-differential and
not likely to bias in any specific direction. Children who had changed water
since birth were excluded, but it was not specifically noted that the fluoride in
the water source was stable over the years.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Probably high risk of bias (NR)
o Summary: Raven’s standard progressive matrices (1992 edition) was used to
assess IQ. Raven’s test is a standard test; although there is no information
provided to indicate that the methods were reliable and valid in this study
population, the test was created to be culturally fair (+ for methods). Blinding or
other methods to reduce potential bias were not reported (NR for blinding). No
response was received to an email request for clarification in September 2017.
Overall rating for methods and blinding = NR.

E-36
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Probably high risk of bias based on indirect evidence that the
outcome assessors were not blind to participants’ fluoride exposure and could bias
the results.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Chi-square test and Spearman rank correlation were used
to assess the association between four different fluoride levels and IQ grades.
Area-level exposures were used. Clustering of children within the four areas
was not accounted for in the analysis; however, because multiple villages
were included in each fluoride exposure level, clustering was less of a concern
and the impact on the effect estimates was expected to be minimal.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in confounding and exposure. Study strengths include verification of exposure
assessment measurements and consideration of key covariates, but it was limited by
the cross-sectional study design and lack of information on blinding during outcome
assessment.

E.1.12. Till et al. (2020)

E.1.12.1. Study Details

• Study design: Prospective cohort

• Population: MIREC participants (pregnant mothers and their children aged 3–
4 years)
• Study area: 10 cities, Canada
• Sample size: 398 mother-child pairs (247 from non-fluoridated areas, 151 from
fluoridated areas; 200 breastfed as infants, 198 formula-fed as infants)
• Data relevant to the review: Adjusted associations between water fluoride
concentration (with or without adjusting for maternal urine) in formula-fed or
breastfed infants or fluoride intake from formula and IQ scores.

E-37
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Reported association with fluoride exposure: Yes: Significant inverse association

between water fluoride and IQ by breastfeeding status (adjusted β = −9.26 formula-
fed, −6.19 breastfed per 0.5-mg/L increase) and between fluoride intake from formula
and performance IQ (adjusted β = −8.76 per 0.5-mg/day increase); significant inverse
association between water fluoride and full-scale IQ in formula-fed children (adjusted
β = −4.40 per 0.5-mg/L increase); no significant changes in full-scale IQ for water
fluoride in breastfed children or fluoride intake from formula-fed children; no
significant changes in verbal IQ scores with fluoride exposure.
E.1.12.2. Risk of Bias

• Author contacts:
o Authors were not contacted for the 2020 publication. Authors were contacted in
June 2019 for additional information on the Green et al. (2019) publication.
Information obtained from that correspondence may have been used for additional
information in the 2020 publication.
• Population selection:
o Rating: Definitely low risk of bias (++)
o Summary: Pregnant women were recruited between 2008 and 2011 by the
MIREC program from 10 cities across Canada. Inclusion and exclusion criteria
were provided. Additional details were stated to be available in Arbuckle et al.
(2013). A total of 610 children were recruited to participate in the developmental
follow-up with 601 children completing all testing. The demographic
characteristics of women included in the current analyses (n = 398) were not
substantially different from the original MIREC cohort (n = 1,945) or the subset
without complete water fluoride and covariate data (n = 203). A table of
characteristics of the study population was provided. Approximately half of the
children lived in non-fluoridated cities and half lived in fluoridated cities.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
exposed groups were similar and were recruited with the same methods during the
same time frame.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Covariates were selected a priori that have been associated with
fluoride, breast feeding, and children’s intellectual ability. Final covariates
included sex and age at testing, maternal education, maternal race, secondhand
smoke in the home, and HOME score. City was considered but excluded from the
models. Covariates that were not assessed include parental mental health, iodine
deficiency/excess, parental IQ, and co-exposure to arsenic and lead. Co-exposure
to arsenic is less likely an issue in this Canadian population because it receives
water mainly from municipal water supplies that monitor for lead and arsenic, and
the lack of information is not considered to appreciably bias the results. In
addition, a previous study on this population (Green et al. 2019) conducted
sensitivity analyses on co-exposures to lead and arsenic. Results from these

E-38
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

sensitivity analyses support the conclusion that co-exposures to lead and arsenic
are not likely a major concern in this study population.
o Potentially important study-specific covariates: All key covariates were
considered in this study.
 Direction/magnitude of effect: Not applicable.
o Basis for rating: Probably low risk of bias based on direct evidence that key
covariates were considered and indirect evidence that the methods used to collect
the information were valid and reliable and co-exposures were not an issue.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: Of 610 children, 601 (98.5%) in the MIREC developmental study who
were ages 3–4 years completed the neurodevelopment testing. Of the 601 children
who completed the neurodevelopmental testing, 591 (99%) completed the infant
feeding questionnaire and 398 (67.3%) reported drinking tap water. It was noted
that the demographic characteristics were not substantially different from the
original MIREC cohort or the 203 subjects without complete water fluoride or
covariate data.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Information on breastfeeding was obtained via questionnaire at 30–
48 months. Fluoride concentration in the drinking water was assessed by daily or
monthly reports provided by water treatment plants. Water reports were first
linked with mothers’ postal codes, and the daily or weekly amounts were
averaged over the first 6 months of each child’s life. Additional details can be
found in Till et al. (2018). Maternal urinary exposure was used to assess fetal
fluoride exposure. Procedures can be found in Green et al. (2019).
 Direction/magnitude of effect size: There is not any specific direction or
magnitude of bias expected. Urinary fluoride levels are reflective of recent
exposure. The possibility of exposure misclassification would be similar in all
subjects and would be non-differential. For the fluoride intake from formula,
exposure was based on the fluoride levels in the water at the residence and the
proportion of time that the infant was not exclusively breastfed. This exposure
misclassification would also be non-differential.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Probably low risk of bias (+)

E-39
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Summary: Intelligence was tested using the Wechsler Preschool and Primary
Scale of Intelligence III, which is considered a gold standard test. It is appropriate
for both the study population and age group. It was not reported whether the
evaluators were blind to the child’s fluoride exposure status during the
assessment. Although it is unlikely that the assessors had knowledge of the
specific drinking water levels or maternal urine levels, there is potential that the
outcome assessors had knowledge of the city the child lived in and whether the
city was fluoridated or non-fluoridated. Correspondence with the study authors on
the outcome assessment for Green et al. (2019) indicated that it was unlikely that
the testers had knowledge of the city’s fluoridation. The same is assumed here.
Specific measurements included were identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessors were blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods were
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Regression diagnostics were used to test assumptions for
linearity, normality, and homogeneity. There were two potential influential
observations (based on Cook’s distance), and sensitivity analyses re-estimated
the models without these two variables. Effect modification by breastfeeding
status was evaluated. Interestingly, all regression coefficients were divided by
2 to represent change in IQ per 0.5-mg/L change in fluoride. One concern is
posed by the lack of accounting for city in the regression models, ideally as a
random effect. The authors explored including city as a covariate in the
models; however, city was not included either because it was strongly multi-
collinear with water fluoride concentration (VIF > 20) (model 1, with water
fluoride concentration) or because fluoride intake from formula is a function
of water fluoride concentration (assessed at the city level) and was therefore
deemed redundant (model 2). However, the models use city-level water
fluoride concentrations—and, in sensitivity analyses, adjust for maternal
urinary fluoride—which warrants exploration of city as a random effect rather
than a fixed effect (as would be the case by having it included as a covariate).
Even including individual-level maternal urinary fluoride might not fully
account for lack of a city effect, given that the subjects were from six different
cities, with half of them fully on fluoridated water. Hence, even individual-
level exposures are likely to be correlated at the city level. Based on a

E-40
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

previous analysis (Green et al. 2019), it is unlikely that exclusion of city from
models (as a fixed or random effect) would significantly impact the effect
estimates.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in confounding, exposure, and outcome. Study strengths include individual
exposure assessment measurements, prospective cohort design, and consideration of
key covariates.

E.1.13. Trivedi et al. (2012)

E.1.13.1. Study Details

• Study design: Cross-sectional

• Population: Children aged 12–13 years
• Study area: Kachchh, Gujarat, India
• Sample size: 84 children
• Data relevant to the review: Mean IQ scores and distribution by low and high
fluoride villages.
• Reported association with fluoride exposure: Yes: Significantly lower mean IQ
score in the high fluoride villages (92.53 ± 3.13) compared with the low-fluoride
villages (97.17 ± 2.54) in boys and girls combined (and by sex).
E.1.13.2. Risk of Bias

• Author contacts:
o Authors were contacted in September of 2017 to obtain additional information for
risk-of-bias evaluation. Additional information provided by the authors informed
the rating decision for the following risk-of-bias domains: Selection, Attrition,
Detection (exposure assessment), Detection (outcome assessment).
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: There is insufficient information provided on the sampling methods to
determine whether the populations were similar. Although it was noted that
samples were obtained for groundwater quality from March to May of 2011, there
is no indication that the children were selected at the same time or during a
similar time frame. Correspondence with the author indicates that children were
selected within a week of the water collection based on random selection of a
school in the village. Study participants were selected from six different villages
of the Mundra region of Gujarat, India. Subjects were grouped into high and low
villages based on the level of fluoride in the drinking water of those villages. The
number of subjects per village was not reported, but it was noted that there were

E-41
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

50 children in the low-fluoride group and 34 children in the high fluoride group. It
is not clear whether the differences in numbers were based on different
participation rates or whether there were fewer children in the high fluoride
villages. Recruitment methods, including any exclusion criteria and participation
rates, were not provided. SES was stated to be low and equal based on
questionnaire information, but the results were not provided. It should also be
noted that only regular students (having attendance more than 80%) of standard
6th and 7th grades were selected, but it was not noted whether attendance varied
by village. Correspondence with the study author indicated that there was an
average of 20 students per class with an average of 40 students per village. It
appears that keeping the requirement of 80% attendance was a limiting factor that
resulted in different numbers of children by area; however, this was applied
similarly to both groups.
o Basis for rating: Probably low risk of bias based on indirect evidence that
subjects were similar and recruited using the same methods during the same time
frame.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Children were stated to be students of the 6th and 7th standard grades.
Age was not addressed, but the children would all be of similar ages based on the
grades included. Results were reported for males and females separately as well
as combined. SES and iodine consumption were stated to be analyzed via a
questionnaire and were standardized on the basis of the 2011 census of India.
Although it was noted in the abstract that the SES was equal (no data provided),
the study report did not mention the iodine results. Although arsenic and lead
were not considered, the study authors provided physicochemical analyses for the
water samples from the six different villages. While the authors did not
specifically analyze lead or arsenic in the water samples, these physicochemical
analyses suggest that differential lead or arsenic exposure was unlikely.
Moreover, based on water quality maps, arsenic was not expected to be a major
concern in this study area. According to the information from the water quality
maps and the physiochemical analysis of the water provided, there is indirect
evidence that neither arsenic nor lead were a concern in this study population.
o Potentially important study-specific covariates: Key covariates age, sex, and
measures of SES were similar between exposure groups; however, arsenic was
not considered. Arsenic often occurs in the drinking water along with fluoride in
some Indian populations; however, based on water quality maps, arsenic does not
appear to be an issue in the study area.
 Direction/magnitude of effect size: Conceptually, the presence of arsenic
would potentially bias the association away from the null if present with
fluoride, or toward the null if present in the reference group; however, for
reasons noted above, arsenic is not considered a concern in this study
population.

E-42
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Probably low risk of bias based on indirect evidence that the
methods used to collect the information were valid and reliable, that potential co-
exposures were not an issue, and that key covariates were addressed.
• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: Results were provided for 84 children, but the methods do not indicate
how many children were initially selected to participate, nor were any exclusion
criteria provided. It was noted in the results that 84 children had their groundwater
and urine tested, but it was not noted whether analyses were restricted to these
children or whether exposures were assessed in all the children who had IQ
measurements. Correspondence with the study author indicated that the main
reason for exclusion was a <80% attendance rate, with fluoride and IQ measured
on all 84 children who met the criteria.
o Basis for rating: Definitely low risk of bias based on direct evidence of no
attrition.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Children in villages were grouped based on fluoride levels that were
assessed in groundwater (low fluoride villages versus high fluoride villages). The
average concentration of these levels was considered to be the levels in the
drinking water with confirmation using urinary fluoride levels. The groundwater
samples were selected to cover major parts of the taluka and represent overall
groundwater quality. Ten samples were obtained from each village. Fluoride was
measured in the groundwater using ion exchange chromatography. Although urine
levels were also significantly higher in the high fluoride village, no information
was provided on how or when the urinary samples were obtained or how they
were measured. However, correspondence with the study author indicated that the
groundwater and urine fluoride levels were available for all 84 children,
indicating that the urine measures were available for the children that had IQ
measures. The urine samples were stated to be collected at the same time the
second water sample was collected.
 Direction/magnitude of effect size: Fluoride levels were measured in both the
drinking water and urine. Although there is some variability in the
measurements, there is no overlap between the two groups, and the urine and
drinking water levels in the children support each other. Any potential
exposure misclassification would be non-differential, and the impact on the
direction and magnitude of the effect size is unknown.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:

E-43
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Rating: Probably low risk of bias (+)

o Summary: Outcome methods were only noted to be reported in Trivedi et al.
(2007), which was scored as follows: IQ was measured in the children of both
areas using a questionnaire prepared by Professor JH Shah, copyrighted by Akash
Manomapan Kendra, Ahmedabad, India, and standardized on the Gujarati
population with a 97% reliability rate in relation to the Stanford-Binet Intelligence
Scale (+ for methods). Blinding or other methods to reduce bias were not
reported, but correspondence with the study author indicated that the teachers
were blind to the status of fluoride. The teachers administered the tests in the
presence of a research fellow. It is not completely clear who scored the tests, but
it is assumed the teachers (+ for blinding). Overall rating for methods and
blinding = +.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
outcomes were assessed using instruments that were valid and reliable in the
study population, and that the outcome assessors were blind to participants’
fluoride exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably high risk of bias (−)
o Summary:
 Statistical analyses: Mean IQ scores in low and high fluoride villages were
compared using a t-test. Consideration of heterogeneity of variances was not
reported. Results are reported as means and standard errors of the means, with
p-values for significant differences. Area-level exposures were used. There
was no accounting for clustering of children within the villages, and
comparative analyses did not account for covariates. Urinary fluoride was not
considered in the comparative analyses. The lack of individual exposure levels
and the lack of accounting for clustering are likely to bias the standard error of
the difference in mean IQ levels between the high- and low-fluoride villages
and make the differences appear stronger than they actually are.
o Basis for rating: Probably high risk of bias based on indirect evidence that the
statistical analyses did not account for clustering, and this lack of accounting
could bias the association. There were no other potential threats of risk of bias
identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in confounding, exposure, and outcome. Study strengths include individual
exposure assessment measurements and the addressing of potential key covariates,
but the study was limited by the cross-sectional study design. Another limitation was

E-44
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

the lack of accounting for clustering, which may bias the standard error of the
differences, making the effect appear stronger than it actually is; however, this does
not change the nearly 5-point difference in IQ scores between the two villages.

E.1.14. Wang et al. (2012)

E.1.14.1. Study Details

• Study design: Cross-sectional

• Population: Children aged 8–13 years [possibly the same study population as Xiang
et al. (2003a)]
• Study area: Wamiao and Xinhuai villages located in Sihong County, Jiangsu
Province, China
• Sample size: 526 school children
• Data relevant to the review: Mean IQ and % low IQ (<80) by total fluoride intake.
• Reported association with fluoride exposure: Yes: Significantly lower mean IQ in
the endemic versus non-endemic regions, as reported in Xiang et al. (2003a); when
the high-exposure group was divided into four groups based on fluoride intake, a
dose-dependent decrease in IQ and increase in % with low IQ was observed;
significant inverse correlation between total fluoride intake and IQ (r = −0.332);
significant positive association between total fluoride intake and risk of IQ <80
(adjusted OR = 1.106 per 1-mg/day increase; 95% CI: 1.052, 1.163).
E.1.14.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: The study appears to have the same study population as Xiang et al.
(2003a) and Xiang et al. (2011); however, it does not cite these studies as
providing additional information, and the numbers of children differ; therefore, it
may be a separate analysis on the same villages. The years of testing were not
provided, so it cannot be determined whether study subjects were the same. Two
villages, Wamiao and Xinhuai, located 64 km apart in Sihong County, Jiangsu
Province, were selected for the study. Wamiao is a village in a region with severe
endemic fluorosis, and Xinhuai is a village in a non-endemic fluorosis region.
Neither village has fluoride pollution from coal or industrial sources. Villages
were stated to be similar in terms of annual per capita income, transportation,
education, medical conditions, natural environment, and lifestyle. All primary
students ages 8–13 years currently in school in either village were surveyed with
exclusions noted. Of 243 children from Wamiao, 236 (97.12%) were included,
and of 305 children from Xinhuai, 290 (95.08%) were included. No table of
subject characteristics was provided.

E-45
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposure groups were similar and were recruited using the same methods within
the same time frame, with direct evidence that there was no difference in
participation/response rates.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Logistic regression of low IQ rate and total fluoride intake adjusted for
age and sex. Both villages had hand-pumped well water for drinking water, but
the authors do not mention whether arsenic was also present in the drinking water.
However, a publication by Xiang et al. (2013) in the same study areas indicates
that Xinhuai (the low-fluoride area) had significantly higher arsenic levels
compared with Wamiao (the endemic fluorosis area), which would bias the
association toward the null. Areas were stated to be similar in annual per capita
income, transportation, education, medical conditions, natural environment, and
lifestyle; however, no details were provided. This study did not address other co-
exposures, but other studies on populations in these villages (Xiang et al. 2011;
Xiang et al. 2003a) indicate that iodine and lead are not concerns.
o Potentially important study-specific covariates: Arsenic often occurs in the
drinking water along with fluoride in some Chinese populations; however, based
on information provided in Xiang et al. (2013), arsenic concentrations were
higher in the low-fluoride area compared with the high fluoride area. Because
there were significant effects on IQ observed in the high fluoride areas, the impact
of co-exposure to arsenic is less of a concern. The presence of arsenic in the
control village may cause an underestimation of the effect of fluoride, but despite
this potential impact, a significant association between fluoride exposure and IQ
was reported.
 Direction/magnitude of effect size: Presence of arsenic in this study population
would potentially bias the association toward the null.
o Basis for rating: Probably low risk of bias because there is indirect evidence that
the key covariates were considered, methods used for collecting the information
were valid and reliable, and co-exposures to arsenic and lead and iodine
deficiency were not attributing to the association observed in this study. The
potential for bias toward the null combined with the reported significant
association increases confidence in the observed effect.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: Data are reported for all 526 children noted to be included in the study.
There is a slight discrepancy in the reported total number of children from the
high-fluoride village and the number of participants from the high-fluoride village
between this paper (236 participated of 243 total children) and the 2003 and 2011
publications on the same study population (222 of 238). This discrepancy is not
explained but is not expected to appreciably bias the results.

E-46
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Probably low risk of bias (+); Probably high risk of bias (−)
o Summary: Water fluoride (+ probably low risk of bias): Exposure was based
on drinking water levels and fluoride intake. Residents in the Wamiao village
were divided into five groups based on fluoride levels in the drinking water.
Clean, dry polyethylene bottles were used to collect 50 mL of drinking water from
each student’s household, and fluoride content was measured.
Total fluoride intake (− probably high risk of bias): Six families from each of
the five Wamiao groups were randomly selected as dietary survey households.
Intakes of various foods by each person at each meal and intakes of unboiled
water, boiled water, and tea were surveyed for four consecutive days. Methods for
food collection were described. Five representative households from each village
were selected based on geographic location, population distribution, housing
structure, and other conditions. Indoor air samples were collected once daily for
five consecutive days; outdoor air was sampled at two points once daily for five
days. Methods for determining fluoride content in samples were noted to follow
specific guidelines. Calculation of total fluoride intake was stated to follow
Appendix A of the People’s Republic of China Health Industry Standard with
some details provided. Although it is assumed the method is valid, it was not
detailed how each fluoride determination was made for each subject, and it
appears that total fluoride intake was determined based on data from select
subjects and not all subjects.
 Direction/magnitude of effect size: There is potential for exposure
misclassification based on calculating fluoride intake based on measurements
from a few select subjects rather than all subjects. The potential impact on the
direction and magnitude of effect size cannot be assessed based on the
information provided.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure. The total fluoride intake is probably high risk of bias because
there is indirect evidence that the exposure was assessed using methods of
unknown validity.
• Outcome:
o Rating: Definitely low risk of bias (++)
o Summary: The IQ of each child was measured with the Combined Raven’s Test
for Rural China (CRT-RC) (++ for methods). The test was stated to be
administered to the children independently in a school classroom under the
supervision of three exam proctors. Testing methods, testing language, and testing
conditions were all in strict accordance with the CRT-RC guidebook. Major

E-47
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

testing personnel received necessary training by the Psychology Department of

East China Normal University. The children undergoing IQ testing and the test
scorers were kept double-blind throughout the testing process (++ for blinding).
Overall rating = ++.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessors were blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Logistic regression analysis was used to determine the
odds of having low IQ with increasing fluoride intake. Analyses and methods
are not well described. There is no mention of what tests were used for the
mean IQ comparison by village; however, statistical software (SPSS) was
used, suggesting appropriate tests were applied. Simple linear regression
analyses were conducted to evaluate associations between total fluoride intake
and children’s IQ or low IQ rate. There is no evidence that regression
diagnostics were used to test model assumptions for linearity, normality, and
homogeneity. Clustering at the village level was not accounted for in the
analyses. The overall impact of these factors on effect estimates is expected to
be minimal given the use of individual-level data and adjustment for
important covariates.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Definitely or probably low
risk-of-bias ratings in confounding, exposure, and outcome. Study strengths include
individual exposure assessment measurements with blinding at outcome assessment
but are limited by the cross-sectional study design and lack of individual
measurements to calculate fluoride intake. All key covariates were accounted for in
the study design or analysis, but there is potential for the presence of arsenic to bias
the association toward the null.

E-48
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

E.1.15. Wang et al. (2020b)

E.1.15.1. Study Details

• Study design: Cross-sectional

• Population: School children aged 7–13 years
• Study area: Tianjin City, China [possibly a subset of the children from Yu et al.
(2018)]
• Sample size: 571 school children
• Data relevant to the review: Adjusted associations between urine and water fluoride
levels and IQ scores.
• Reported association with fluoride exposure: Yes: Significant inverse associations
between water fluoride and IQ scores (adjusted β = −1.587 per 1-mg/L increase; 95%
CI: −2.607, −0.568) and between urinary fluoride and IQ scores (adjusted β = −1.214
per 1-mg/L increase; 95% CI: −1.987, −0.442) in boys and girls combined. No
significant effect modification by sex.
E.1.15.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Definitely low risk of bias (++)
o Summary: Subjects were from a cross-sectional study conducted in 2015, but no
citation was provided on this cohort [presumably the Yu et al. (2018) cohort]. It
was noted that the subjects in that cohort were from districts with historically high
or normal fluoride levels. Subjects for this study were selected by using a
stratified and multistage random sampling approach. Brief description was
provided. The study area consisted of three historically high fluoride areas and
four non-endemic areas. A flow diagram was provided for inclusion and
exclusion, but this detail was given for all children and not by area. Therefore, it
cannot be determined whether the participation differed by area. However, there
was a 93% recruitment rate, and the 13 excluded due to missing data were not
likely excluded due to exposure. Detailed characteristics of the study population
are provided. Exclusion criteria included: “children who had congenital or
acquired diseases affecting intelligence, or a history of cerebral trauma and
neurological disorders, or those with a positive screening test history (like
hepatitis B virus infection, Treponema palladium infection and Down's syndrome)
and adverse exposures (smoking and drinking) during maternal pregnancy, prior
diagnosis of thyroid disease, and children who had had missing values of
significant factors (2.2%) were also excluded.”
o Basis for rating: Definitely low risk of bias based on direct evidence that the
exposed groups were recruited using similar methods during the same time frame

E-49
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

and that any differences between the exposed groups were accounted for in the
statistical analyses.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Study authors noted that the study areas were not exposed to other
neurotoxins such as lead, arsenic, or mercury nor were they iodine deficient. Final
models included age, sex, child’s BMI, maternal and paternal education,
household income, and low birth weight. The other covariates that were
considered are unclear as the authors only noted that the covariates were selected
based on current literature. Reasons for exclusion included history of disease
affecting intelligence, history of trauma or neurological disorders, positive
screening test history, or exposures such as smoking or drinking during
pregnancy. Information was obtained by questionnaire or measurements.
Covariates such as parental BMI, behavioral and mental health disorders, IQ, and
quantity and quality of the caregiving environment were not considered.
o Potentially important study-specific covariates: All key covariates were
considered in this study.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias because there is direct evidence that
the key covariates were considered and indirect evidence that the methods for
collecting the information were valid and reliable and that co-exposure to arsenic
was not an issue in this area.
• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: A detailed chart of the recruitment process is presented. The study had
a 93% recruitment rate, and only 2.2% of subjects with missing data for certain
covariates were excluded.
o Basis for rating: Definitely low risk of bias based on direct evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Children provided spot urine samples, presumably at the time of
examination. Water samples were randomly collected from public water supplies
in each village. Fluoride concentrations were analyzed using fluoride ion-selective
electrode according to the national standardized method in China. There is no
indication of whether the urine samples accounted for dilution.
 Direction/magnitude of effect size: Not accounting for dilution could cause
some exposure misclassification. The impact on the direction and magnitude
of effect size would depend on where the differences occurred.

E-50
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using acceptable methods that provide
individual levels of exposure.
• Outcome:
o Rating: Definitely low risk of bias (++)
o Summary: Assessments of IQ scores were conducted by graduate students at the
School of Public Health, Tongji Medical College, Huazhong University of
Science and Technology. Each team member was assigned a single task, meaning
that only one person would have conducted the IQ tests. A Combined Raven’s
Test for Rural China was used. Therefore, the test was appropriate for the study
population (++ for method). It was noted that the examiner was trained and blind
to the exposure (++ for blinding). Overall = ++
o Basis for rating: Definitely low risk of bias based on direct evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessor was blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes in the abstract, introduction, and methods are reported in
sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Logistic and multivariate regression models accounting
for covariates were used. Results are presented as betas or ORs and 95% CIs.
Regression diagnostics were conducted for all models, including examination
of multicollinearity, heteroscedasticity, and influential observations.
Mediation and interaction analyses were appropriate. There is no evidence
that the stratified and multistage random sampling approach for subject
selection was accounted for in the analyses by using sampling weights or
accounting for clustering using random effect models; however, selected
villages were similar in population and general demographic characteristics.
Given the use of individual-level data and adjustment for important
covariates, the impact on the regression coefficients is likely to be minimal.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and no other potential threats of risk of bias
were identified.
• Basis for classification as low risk-of-bias study overall: Definitely or probably low
risk-of-bias ratings in confounding, exposure, and outcome. Study strengths include

E-51
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

individual exposure assessment measurements but are limited by the cross-sectional

study design and lack of accounting for urine dilution. All key covariates were
considered in the study design or analysis.

E.1.16. Xiang et al. (2003a)

E.1.16.1. Study Details

• Study design: Cross-sectional

• Population: Children aged 8–13 years
• Study area: Wamiao and Xinhuai villages located in Sihong County, Jiangsu
Province, China
• Sample size: 512 school children
• Data relevant to the review: Comparison of IQ (mean and distribution) between
Wamiao County (a severe endemic fluorosis area) and Xinhuai County (a non-
endemic fluorosis area); additional breakdown of the Wamiao area into five water
fluoride exposure groups. Correlations between water fluoride and IQ scores.
• Reported association with fluoride exposure: Yes: Significantly lower IQ scores
observed with water fluoride levels of 1.53 mg/L or higher. The percentage of
subjects with IQ scores below 80 was significantly increased at water fluoride levels
of 2.46 mg/L or higher. Significant inverse correlation between urinary fluoride and
IQ (r = −0.164). Mean IQ scores for children in the non-endemic region
(100.41 ± 13.21) were significantly higher than the endemic region (92.02 ± 13.00).
E.1.16.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Two villages, Wamiao and Xinhuai, located 64 km apart in Sihong
County, Jiangsu Province, were selected for this study, which was conducted
between September and December 2002. Wamiao is located in a severe fluorosis
endemic area, and Xinhuai is located in a non-endemic fluorosis area. Neither
village has fluoride pollution from burning coal or other industrial sources. All
eligible children in each village were included; children who had been absent
from either village for 2 years or longer or who had a history of brain disease or
head injury were excluded. In Wamiao, 93% of the children (222 out of 238) were
included in the study; in Xinhuai, 95% were included (290 out of 305). The
children in Wamiao were divided into five subgroups according to the level of
fluoride in their drinking water: <1.0 mg/L (group A), 1.0–1.9 mg/L (group B),
2.0–2.9 mg/L (group C), 3.0–3.9 mg/L (group D), and >3.9 mg/L (group E).
Children in Xinhuai (0.18–0.76 mg/L in the drinking water) served as a control
group (group F). Demographic characteristics are not presented, and statistical

E-52
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

analyses are not adjusted, but mean IQ scores are stratified by age, sex, family
income, and parental education.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposure groups were similar and were recruited using the same methods within
the same time frame, with direct evidence that there was no difference in
participation/response rates.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Although information was stated to be collected on personal
characteristics, medical history, education levels of the children and parents,
family SES, and lifestyle, only sex, age, family income, and parental education
were considered. Potential co-exposures, such as arsenic, were not addressed. A
separate publication in 2003 [(Xiang et al. 2003b), letter to the editor] indicated
that blood lead levels were not significantly different between the two areas.
Although arsenic was not addressed specifically in this publication, Xiang et al.
(2013) measured both fluoride and arsenic in the Wamiao and Xinhuai areas.
Xinhuai (the low-fluoride area) had significantly higher arsenic levels compared
with Wamiao (the endemic fluorosis area). This is likely to bias the association
toward the null; however, the study observed a significantly lower IQ score in the
endemic fluorosis area. Iodine was tested in a subset of the children and found not
to be significantly different between the two groups.
o Potentially important study-specific covariates: Arsenic often occurs in the
drinking water along with fluoride in some Chinese populations; however, based
on information provided in Xiang et al. (2013), arsenic concentrations were
higher in the low-fluoride area compared with the high fluoride area. Because
there were significant effects on IQ observed in the high fluoride areas, the impact
of co-exposure to arsenic is less of a concern. The presence of arsenic in the
control village may cause an underestimation of the effect of fluoride, but despite
this potential impact, there was still a significant association between fluoride
exposure and IQ.
 Direction/magnitude of effect size: Presence of arsenic in this study population
would potentially bias the association toward the null.
o Basis for rating: Probably low risk of bias because there is indirect evidence that
the key covariates were taken into account, methods used for collecting the
information were valid and reliable, and co-exposures to arsenic and lead and
iodine deficiency were not attributing to the effect observed in this area. The
potential for bias toward the null, combined with the reported significant
association increases confidence in the observed effect.
• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: Data are complete. IQ results were reported for all 512 children
included in the study (222 in the endemic area and 290 in the nonendemic area).

E-53
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Definitely low risk of bias based on direct evidence that there
was no attrition.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Exposure was based on drinking water and urinary levels of fluoride.
The two study areas were selected to reflect a severe endemic area and a non-
endemic area. Drinking water was collected from wells, and early-morning spot
urine samples were collected from a randomly selected subsample of children.
Both water and urine samples were measured using fluoride ion-selective
electrode, but no quality control was discussed. Both absolute and creatinine-
adjusted urine results were reported.
 Direction/magnitude of effect size: There is potential for exposure
misclassification because only current levels were assessed. Migration of
subjects in or out of the area was not assessed, but the study authors noted that
if the children had been absent from the village for 2 or more years, they were
excluded. Misclassification would likely be non-differential, which could
likely bias the association in either direction.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Definitely low risk of bias (++)
o Summary: The IQ of each child was measured with the Combined Raven’s Test
for Rural China (CRT-RC) (++ for methods). The test was stated to be
administered to the children independently in a school classroom, in a double-
blind manner, under the supervision of an examiner and two assistants, and in
accordance with the directions of the CRT-RC manual regarding test
administration conditions, instructions to be given, and test environment (++ for
blinding). Overall rating = ++
o Basis for rating: Definitely low risk of bias based on direct evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessor was blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:

E-54
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

 Statistical analyses: There is no mention of the tests conducted, but data were
stated to be analyzed using SAS, suggesting appropriate tests were applied.
Results provided in the tables indicate that t-tests comparing IQ values
between the villages (overall and by sex) were conducted, but it was not
reported that heterogeneity of variance was assessed. In addition, correlations
between IQ and age, family income, and parents’ education level were tested
with Pearson’s correlation. There is no evidence that a test for trend was
conducted to evaluate the stated “significant inverse concentration-response
relationship between the fluoride level in drinking water and the IQ of
children.”
 A potential concern raised by the NASEM (2020) committee’s review was the
lack of accounting for relationships in exposure between persons from the
same village. Given only two villages were included and the analyses
consisted of village-level comparisons (no use of individual-level covariate
data), it is likely that the standard error of the difference in mean IQ between
fluoride in water exposure groups will be biased, making differences appear
stronger than they actually are. Without controlling for village effects and
given the large differences in fluoride concentrations and IQ levels between
villages, the apparent dose-response relationship could be due to a village
effect in addition to a fluoride effect. However, a dose-response relationship is
apparent within the “exposed” village, diminishing the concern for a village-
only effect and likely minimizing the impact on the effect estimates.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that
statistical analyses were appropriate and that there were no other threats of risk of
bias.
• Basis for classification as low risk-of-bias study overall: Definitely or probably low
risk-of-bias ratings in confounding, exposure, and outcome. Study strengths include
individual exposure assessment measurements and blinding of outcome assessor to
exposure but is limited by the cross-sectional study design and lack of accounting for
urine dilution. All key covariates were considered in the study design or analysis, but
there is potential for the presence of arsenic to bias the association toward the null.

E.1.17. Xiang et al. (2011)

E.1.17.1. Study Details

• Study design: Cross-sectional

• Population: Children aged 8–13 years [same study population as Xiang et al.
(2003a)]
• Study area: Wamiao and Xinhuai villages located in Sihong County, Jiangsu
Province, China
• Sample size: 512 school children
• Data relevant to the review: Mean IQ scores and odds ratio for having an IQ <80 by
serum fluoride quartiles.

E-55
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Reported association with fluoride exposure: Yes: Significant linear trend across
quartiles of serum fluoride and children’s IQ score <80 (adjusted ORs for Q1 and Q2;
Q1 and Q3; and Q1 and Q4, respectively: 1; 2.22 [95% CI: 1.42, 3.47]; and 2.48
[95% CI: 1.85, 3.32]); significant associations observed at levels ≥0.05 mg/L serum
fluoride.
E.1.17.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: The study population was the same as that used in the Xiang et al.
(2003a) study, but a few more measurements were available and different
analyses were conducted. The comparison population was considered the same
based on the study populations being recruited from similar populations, using
similar methods, during the same time frame. Demographic characteristics were
not provided.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposure groups were similar and were recruited using the same methods within
the same time frame, with direct evidence that there was no difference in
participation/response rates.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: As was noted in the 2003 publication (Xiang et al. 2003a), information
was collected on personal characteristics, medical history, education levels in the
children and parents, family SES, and lifestyle. In the logistic regression model
age and sex were adjusted for in the analysis. In the previous report, no significant
associations were observed between groups for family income and parents’
education (Xiang et al. 2003a). Urinary iodine and blood lead levels were also
stated to be measured and were noted not to be significantly different between the
groups. Although the iodine levels were reported in the previous publication, the
lead levels were not and neither were the methods. Lead information is reported in
a letter to the editor (Xiang et al. 2003b) and was not significantly different
between the areas. Although arsenic was not addressed specifically in this
publication, Xiang et al. (2013) measured both fluoride and arsenic in the Wamiao
and Xinhuai areas. Xinhuai (the low-fluoride area) had significantly higher
arsenic levels compared with Wamiao (the endemic fluorosis area). This is likely
to bias the association toward the null; however, the study observed a
significantly lower IQ score in the endemic fluorosis area and with increasing
serum fluoride.
o Potentially important study-specific covariates: Arsenic often occurs in the
drinking water along with fluoride in some Chinese populations; however, based
on information provided in Xiang et al. (2013), arsenic concentrations were

E-56
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

higher in the low-fluoride area compared to the high fluoride area. Because there
were significant effects on IQ observed in the high fluoride areas, the impact of
co-exposure to arsenic is less of a concern. The presence of arsenic in the control
village may cause an underestimation of the effect of fluoride, but despite this
potential impact, there was still a significant association between fluoride
exposure and IQ.
 Direction/magnitude of effect size: Presence of arsenic in this study population
would potentially bias the association toward the null.
o Basis for rating: Probably low of risk bias because there is indirect evidence that
the key covariates were considered, methods used for collecting the information
were valid and reliable, and co-exposures to arsenic and lead and iodine
deficiency were not attributing to the effects observed in this area. The potential
bias toward the null, combined with the reported significant association increases
confidence in the observed effect.
• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: Data are reported for all 512 children noted to be included in the study.
o Basis for rating: Definitely low risk of bias based on direct evidence that there
was no attrition.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Fluoride levels were measured in serum with a fluoride ion-selective
electrode. A fasting venous blood sample was used. No details are provided on
validation (including correlation with drinking water levels) or QA. Children who
did not reside in their village for at least 2 years were excluded. Results were
provided in quartiles, but the authors combined the lower two quartiles. After
combining the two lower quartiles, the exposure levels ranged from <0.05 mg/L
(Q1 + Q2) to >0.08 mg/L (Q4).
 Direction/magnitude of effect size: Serum fluoride may not be the best
estimate for exposure. There is potential for exposure misclassification
because only current levels were assessed. Migration of subjects in or out of
the area was not assessed, but the study authors noted that if the children had
been absent from the village for 2 or more years, they were excluded.
Misclassification would likely be non-differential, which could bias results in
either direction.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Definitely low risk of bias (++)
o Summary: IQ was assessed as part of the 2003 evaluation. IQ was measured with
the Combined Raven’s Test for Rural China, which is appropriate for this

E-57
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

population (++ for methods). Although this study does not provide details, the
original study article from 2003 provides specific details. The study authors
indicate in the 2003 publication that the tests were conducted in a double-blind
manner, and these are the same results and population (++ for methods). Overall
rating = ++
o Basis for rating: Definitely low risk of bias based on direct evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessor was blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Statistical analyses conducted were appropriate for the
study. Chi-square tests were used to compare categorical variables, and
multiple logistic regression was used to evaluate the association between
serum fluoride levels and risk of low IQ. A potential concern raised by the
NASEM (2020) peer review was the lack of accounting for relationships in
exposure between persons from the same village. Although only two villages
were included, in the analyses that consisted of village-level comparisons, it is
likely that the standard error of the difference in mean IQ between villages is
biased. This is less of a concern for the mean IQ comparisons across quartiles
of serum fluoride levels and for the logistic regression analyses of risk of low
IQ and individual-level serum fluoride levels. Without controlling for village
effects and given the large differences in fluoride concentrations and IQ
between villages, the apparent dose-response relationship could be due to a
village effect in addition to a fluoride effect. However, the dose-response
relationship is still present within the “exposed” village, diminishing the
concern for a village-only effect and likely minimizing the impact on the
effect estimates.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Definitely or probably low
risk-of-bias ratings in confounding, exposure, and outcome. Study strengths include
individual exposure assessment measurements with blinding at outcome assessment
but is limited by the cross-sectional study design and use of serum concentrations. All

E-58
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

key covariates were considered in the study design or analysis, but there is potential
for the presence of arsenic to bias the association toward the null.

E.1.18. Yu et al. (2018)

E.1.18.1. Study Details

• Study design: Cross-sectional

• Population: Children aged 7–13 years
• Study area: Tianjin City, China
• Sample size: 2,886 school children
• Data relevant to the review: IQ for normal (≤1 mg/L) versus high (>1 mg/L) water
fluoride; adjusted associations between water and urine fluoride and IQ scores.
• Reported association with fluoride exposure: Yes: Significant difference in mean IQ
scores in high water fluoride areas (>1.0 mg/L; 106.4 ± 12.3 IQ) compared to the
normal water fluoride areas (≤1.0 mg/L; 107.4 ± 13.0). Distribution of IQ scores was
also significantly different (p = 0.003). Significant inverse association between water
fluoride and IQ scores (adjusted β = −4.29 per 0.5-mg/L increase; 95% CI: −8.09,
−0.48).
E.1.18.2. Risk of Bias

• Author contacts:
o Authors were contacted in September 2018 to obtain additional information for
the risk-of-bias evaluation. Additional information provided by the authors
informed the rating decision for the following risk-of-bias domains: Detection
(outcome assessment).
• Population selection:
o Rating: Definitely low risk of bias (++)
o Summary: School children (2,886), aged 7–13 years, were recruited from the rural
areas of Tianjin City, China. After exclusion, 1,636 children were assigned to the
“normal-fluoride” exposure group, and 1,250 were assigned to the “high-fluoride”
exposure group based on a cut-off water fluoride level of 1.0 mg/L. A multistage
random sampling technique, stratified by area, was performed to select
representative samples among local children who were permanent residents since
birth. Detailed characteristics of the study population were provided. Exclusion
criteria included: 1) children who had congenital or acquired diseases affecting
intelligence, 2) children with a history of cerebral trauma and neurological
disorders, 3) children with a positive screening test history (like hepatitis B virus
infection, Treponema palladium infection and Down's syndrome), and 4) children
with adverse exposures (smoking and drinking) during maternal pregnancy. A
table of characteristics was provided by fluoride level with differences adjusted in
the analysis.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
exposed groups were recruited using similar methods during the same time frame

E-59
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

and that any differences between the exposed groups were considered in the
statistical analyses.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Demographic data were collected by trained investigators during a
face-to-face interview with the recruited children and their parents.
Questionnaires were not stated to be validated. The developmental status of the
children was further assessed by calculation of BMI, and all measurements were
conducted by nurses based on recommended standard methods. Variables that
presented differential distribution between the normal-fluoride and high-fluoride
exposure groups were adjusted in the linear regression analysis of IQ data and
included age, sex, paternal and maternal education levels, and low birth weight.
Children exposed to smoking in utero were excluded from the study. Sensitivity
analyses were conducted by modifying covariates adjusted in multivariable
models among demographics (age and sex); development (BMI); socioeconomics
(maternal education, paternal education, and household income); history of
maternal disease during pregnancy (gestational diabetes, malnutrition, and
anemia); and delivery conditions (hypoxia, dystocia, premature birth, post-term
birth, and low birth weight). None of the study sites selected were in areas
endemic for iodine deficiency disorders, nor were other potential neurotoxins like
lead, arsenic, and mercury present. Variables such as parental BMI and behavioral
and mental health disorders were not addressed.
o Potentially important study-specific covariates: All key covariates were
considered in this study.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias based on indirect evidence that
methods of obtaining the information were valid and reliable and direct evidence
that all key covariates and co-exposures were considered.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: There were 1,636 children assigned to the “normal-fluoride” exposure
group based on water fluoride and 1,250 children assigned to the “high-fluoride”
exposure group. Exclusion from the original group of 2,886 children was
adequately described. A total of 2,380 children provided urine samples. There is
no indication that the data presented excludes any additional children or urine
samples, but results do not indicate a sample size for all results.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Probably low risk of bias (+)

E-60
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Summary: According to the annual surveillance data from the CDC, the drinking
water sources and water fluoride concentrations in each village had remained at
stable levels over the past decade. During the investigation, water samples were
collected randomly from the public water supplies in each village. Spot (early-
morning) urine samples from every child and water samples from each village
were collected in pre-cleaned, labeled polythene tubes and transported to the lab
within 24 hours while frozen. Samples were stored at −80°C until analysis.
Concentrations of fluoride ions (mg/L) were analyzed using the national
standardized ion-selective electrode method in China; the detection limit was
0.01 mg/L. Samples were diluted with an equal volume of total ionic strength
adjusted buffer (TISAB) of pH 5–5.5 for optimal analysis. Double-distilled
deionized water was used throughout the experiment. There is no reporting of any
QC methods.
 Direction/magnitude of effect size: Spot urine samples may lead to non-
differential exposure misclassification. The large population size likely dilutes
any potential effects of occasional misclassification. Because the drinking
water sources of fluoride had been noted to be stable for the past decade and
the children were 13 years or younger, there would only be exposure
misclassification if there was a lot of migration between areas.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Definitely low risk of bias (++)
o Summary: IQ scores were measured using the second edition of the Combined
Raven’s Test–The Rural in China (CRT-RC2) for children aged 7–13 years (++
for methods). The test was completed by each participant within 40 minutes,
according to the instruction manual. For each test, 40 children were randomly
allocated to one classroom to take the test independently under the supervision of
four trained professionals. There is no mention of whether the evaluators were
blinded to the fluoride group of each child (normal vs. high fluoride) or whether
there were steps taken to ensure consistency in scoring across the evaluators. It is
also not clear whether the 40 children randomly assigned to the classroom were
specific to the village or whether a local center was used. Correspondence with
the study authors indicated that the four professionals worked together throughout
the examination without knowledge of the child’s fluoride exposure (++ for
blinding).
o Basis for rating: Definitely low risk of bias based on the direct evidence that the
outcome was assessed using instruments that were valid and reliable, and that the
outcome assessors were blind to participants’ fluoride exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.

E-61
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Statistical analyses used were appropriate for the study.
Univariate and multivariable piecewise linear regression models were used to
estimate the associations between water fluoride or urinary fluoride levels and
IQ scores. Multiple logistic regression analysis was used to evaluate the
association between water or urinary fluoride levels and IQ degree using the
normal intelligence group as the control. Sensitivity analyses were conducted.
There is no evidence that residual diagnostics were used to examine model
assumptions or that the complex sampling design (stratified multistage
random sampling) was accounted for in the analysis using sampling weights
and adjustment for clustering. The impact of these factors on the effect
estimates is expected to be minimal given the use of individual-level data and
adjustment for numerous important covariates.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Definitely or probably low
risk-of-bias ratings in confounding, exposure, and outcome. Study strengths include
individual exposure assessment measurements with blinding at outcome assessment
but is limited by the cross-sectional study design and lack of accounting for urine
dilution. All key covariates, including potential co-exposures, were considered in the
study design or analysis.

E.1.19. Zhang et al. (2015b)

E.1.19.1. Study Details

• Study design: Cross-sectional

• Population: Children aged 10–12 years
• Study area: Tianjin City, China
• Sample size: 180 children
• Data relevant to the review: Mean IQ scores by control and high fluoride groups;
correlations between water, serum, or urinary fluoride and IQ scores; adjusted
associations between urinary fluoride levels and IQ scores (by genotypes).
• Reported association with fluoride exposure: Yes: Significant inverse correlation
between children’s serum fluoride (r = −0.47) and urinary fluoride (r = −0.45) and IQ
scores; significant difference in mean IQ score for high-fluoride area (defined as
>1 mg/L in drinking water; 102.33 ± 13.46) compared with control area (<1 mg/L;
109.42 ± 13.30).

E-62
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

E.1.19.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Definitely low risk of bias (++)
o Summary: Subjects were similar and recruited during the same time frame using
the same methods. Authors recruited schoolchildren from a high fluoride area
(1.40 mg/L) and a control area (0.63 mg/L) in Tianjin City, China. In accordance
with the principles of matching social and natural factors such as educational
standard, economic situation, and geological environments as much as possible,
two areas with different fluoride concentrations in the groundwater were selected
by a stratified cluster random sampling of this region. A total of 180 5th grade
children aged 10 to 12 years from two primary schools located 18 km apart in the
Jinnan District were recruited—Gegu Second Primary School (from an endemic
fluorosis area) and Shuanggang Experimental Primary School (from a non-
endemic fluorosis area). The areas are not affected by other drinking water
contaminants, such as arsenic or iodine. All subjects were unrelated ethnic Han
Chinese and residents in Tianjin with similar physical and mental health status.
The authors excluded subjects with known neurological conditions, including
pervasive developmental disorders and epilepsy. Descriptive statistics of the study
population are presented by exposure group in Table 1 of the study. A number of
potential differences were considered in the statistical analyses.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
exposure groups were similar and recruited using similar methods during the
same time frame.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Covariates included in the statistical models were age, sex, educational
levels of parents, drinking water fluoride (mg/L), and levels of thyroid hormones
(T3, T4, and TSH). Authors report that the study areas were not affected by other
contaminants such as arsenic or iodine, and residents were of similar physical and
mental health status. Other important covariates (maternal demographics,
smoking, reproductive health) were not considered. Covariate data were obtained
from a study questionnaire.
o Potentially important study-specific covariates: All key covariates were
considered in this study.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
methods used to collect the information were valid and reliable and direct
evidence that key covariates, including potential co-exposures, were considered.
• Attrition:

E-63
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Rating: Definitely low risk of bias (++)

o Summary: Results are complete for the 180 children selected for the study.
o Basis for rating: Definitely low risk of bias based on direct evidence that there
was no attrition.
• Exposure:
o Rating: Definitely low risk of bias (++)
o Summary: Drinking water samples (10 mL) were collected from the tube wells of
each child’s household. Three fasting venous blood samples were also collected.
Urine samples were collected in the early morning before breakfast. Fluoride
content in drinking water (W-F), serum (S-F), and urine (U-F) was measured
using an ion analyzer EA940 with a fluoride ion-selective electrode (Shanghai
Constant Magnetic Electronic Technology Co, Ltd, China), according to the
China standard GB 7484-87. All reference solutions for the fluoride
determinations were double-deionized water. Parallel samples were set for
determination, and averages were taken. The quantitation limits of this method for
W-F, S-F, and U-F were 0.2, 0.012, and 0.5 mg/L, respectively. Recovery rates
for this method were in the range of 94.3%–106.4%. The intra- and inter-assay
coefficients of variation for fluoride were 2.7% and 6.7%, respectively. Dilution
of the urinary fluoride was not addressed.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Definitely low risk of bias (++)
o Summary: A Combined Raven’s Test for Rural China (CRT-RC) was taken to
evaluate the IQ of each child (++ for methods). The study report stated that all
tests were administered at school by a trained examiner who was masked to
participants’ drinking water fluoride levels (++ for blinding). Overall rating for
methods and blinding = ++.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessor was blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All results outlined in the abstract, introduction, and methods sections
were reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:

E-64
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Rating: Probably low risk of bias (+)

o Summary:
 Statistical analyses: Associations between serum and urinary fluoride levels
and IQ score were estimated using general linear models and multivariate
linear regression by COMT variant. Normality (Kolmogorov-Smirnov test)
was evaluated for all continuous variables. There is no evidence that residual
diagnostics were used to examine model assumptions or that the complex
sampling design (stratified multistage random sampling) was accounted for in
the analysis using sampling weights and adjustment for clustering. The impact
of these factors on the regression effect estimates is expected to be minimal
given the use of individual-level data and adjustment for numerous covariates.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on direct evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Definitely or probably low
risk-of-bias ratings in confounding, exposure, and outcome. Study strengths include
individual exposure assessment measurements, blinding of outcome assessor to
participants’ fluoride exposure, and consideration of key covariates including
potential co-exposures.

E.2. Other Neurodevelopmental Studies

E.2.1. Barberio et al. (2017b)

E.2.1.1. Study Details

• Study design: Cross-sectional

• Population: Canadian Health Measures Survey (cycles 2 and 3) participants (children
aged 3–12 years)
• Study area: general population of Canada
• Sample size: 2,221 children (1,120 from Cycle 2, 1,101 from Cycle 3)
• Data relevant to the review: Adjusted associations between urinary fluoride and
learning disability or ADHD (Cycle 2 only) assessed by parent or child self-report.
• Reported association with fluoride exposure: Yes: Significant positive association
between urinary fluoride (unadjusted for creatinine) and risk of learning disability
(adjusted OR = 1.02 per 1-µmol/L increase; 95% CI: 1.00, 1.03) when Cycles 2 and 3
were combined. No significant associations with creatinine-adjusted or specific
gravity-adjusted urinary fluoride. No significant association between urinary fluoride
and ADHD.
E.2.1.2. Risk of Bias

• Author contacts:

E-65
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Definitely low risk of bias (++)
o Summary: The comparison groups were selected from Cycles 2 and 3 of the
Canadian Health Measures Survey. This is a nationally representative sample of
residents living in 10 provinces, with clear exclusion criteria provided. Exclusion
represented only about 4% of the target population (all Canadian residents 3–
79 years old living in 10 provinces). A table of characteristics of the study
population is provided.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
subjects were recruited from the same population using the same methods during
the same time frame, and exposure groups were similar.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: The study adjusted for sex, age (3–12 years old), household education,
and household income adequacy. Variables to discern fluoride source, including
drinking water and dental products, were also considered. Cycle 2 data also
included adjustments for: 1) children for whom tap water (vs. bottled or other)
was the primary source of drinking water at home or away from home and
2) children who had lived in their current home for 3 or more years. Covariates
such as parental behavioral and mental health disorders, smoking, and nutrition
were not discussed. The study used data from the Canadian Health Measures
Survey, which consists of a nationally representative sample of Canadians. Most
Canadians (~89%) receive water from municipal water supplies, which monitor
for levels of lead and arsenic. Therefore, co-exposure to lead and arsenic are less
likely an issue in this population and the lack of information is not considered to
appreciably bias the results.
o Potentially important study-specific covariates: All key covariates were
considered in this study.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias based on direct evidence that key
covariates were addressed and indirect evidence that the methods used to collect
the information were valid and reliable and that co-exposures were not an issue.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: Covariate data were missing for less than 5% of all analyses, apart
from household income; household income was reported for only 71%–77% of
participants and was imputed for the remainder.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were

E-66
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Estimates of urinary fluoride (µmol/L) from spot urine were available
for a subsample of respondents. Analysis was performed under standardized
operating procedures at the Human Toxicology Laboratory of the Institut National
de Santé Publique du Québec (accredited under ISO 17025). Fluoride content of
urine samples was analyzed using an Orion pH meter with a fluoride ion-selective
electrode with limits of detection of 20 μg/L (Cycle 2) and 10 μg/L (Cycle 3).
Urinary dilution was addressed by using creatinine-adjusted levels as well as
specific gravity-adjusted levels. In Cycle 3 only, estimates of the fluoride
concentration of tap water samples collected from randomly selected households
were available. The subsample of households selected for tap water sample
collection corresponded to the person-level urine fluoride subsample. Analysis of
the fluoride concentration of tap water was performed using a basic anion
exchange chromatography procedure, with a limit of detection of 0.006 mg/L. QC
methods were not addressed.
 Direction/magnitude of effect size: There is not any specific impact on the
direction or magnitude of effect size expected. Urinary fluoride levels are
reflective of a recent exposure. Having a single concurrent measurement may
not be reflective of the exposure associated with the outcome, but if subjects
lived in the same area throughout life, the exposure may be an adequate
representation. Although there is possible exposure misclassification, it would
likely be non-differential.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Probably high risk of bias (−)
o Summary: The primary outcome variable, diagnosis of a learning disability by a
health professional, was based on a single item from a household survey asked to
all respondents: “Do you have a learning disability?” Answer options were: “yes,”
“no,” “don’t know,” or the participant refused to answer. For Cycle 2, those who
indicated having a learning disability were also asked what kind, with the answer
options of: “ADD,” “ADHD,” “dyslexia,” or “other.” This question was omitted
in Cycle 3, and the reason for omission was not described. Parents or guardians
answered all questions for children aged 3–11 years, while children 12 years and
older answered questions themselves. The self-reporting of a learning disability
did not appear to have been confirmed by medical records or a health professional
(− for methods based on self-report of diagnosis by a health care professional;
also, in Cycle 3, no specific disabilities were described). Blinding was not a
concern as spot urine samples were sent to a separate lab, and self-reports would

E-67
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

not have knowledge of their urine or tap water exposure level (+ for blinding).
Overall rating = −.
o Basis for rating: Probably high risk of bias based on indirect evidence that the
outcome was measured using an insensitive method in the study population.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods
sections were reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Definitely low risk of bias (++)
o Summary:
 Statistical analyses: Logistic regression analyses, adjusted and unadjusted for
covariates, examined the associations between fluoride exposure and
diagnosis of learning disability. Analyses were performed for Cycle 2 only
(urinary fluoride and type of learning disability diagnosis), Cycle 3 only
(urinary fluoride, water fluoride, and learning disability diagnosis), and Cycles
2 and 3 combined. Analyses used survey weights and bootstrapped weights to
ensure proper computation of variance estimates. Results are reported as
unadjusted and adjusted ORs with 95% CIs.
 Other potential concerns: None identified.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in confounding and exposure. Study strengths include individual exposure
assessment measurements and the consideration of key covariates but were limited by
the cross-sectional study design and insensitive outcome measures.

E.2.2. Bashash et al. (2017)

E.2.2.1. Study Details

• Study design: Prospective cohort

• Population: Early Life Exposures in Mexico to Environmental Toxicants
(ELEMENT) participants (pregnant mothers and their children aged 4 or 6–12 years).
• Study area: Mexico City, Mexico
• Sample size: 299 mother-child pairs, of whom 287 had data for the general cognitive
index (GCI).
• Data relevant to the review: Adjusted and unadjusted associations between maternal
or child’s urinary fluoride concentrations and GCI.

E-68
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Reported association with fluoride exposure: Yes: Significant inverse association

between maternal urinary fluoride and GCI score (adjusted β = −3.15 per 0.5-mg/L
increase; 95% CI: −5.42, −0.87). No significant associations with children’s urinary
fluoride.
E.2.2.2. Risk of Bias

• Author contacts:
o Authors were contacted for additional information on whether clustering was
addressed. The authors provided results from additional models with cohort as a
random effect, which informed the rating decision for the following risk-of-bias
domains: Other.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Study participants were selected from two different cohorts from three
hospitals in Mexico City that serve low-to-moderate income populations. One
cohort was from an observational study of prenatal lead exposure and
neurodevelopmental outcomes, and the other was from a randomized trial of the
effect of calcium on maternal blood lead levels. The authors state that participants
had no history of psychiatric disorders, high-risk pregnancies, gestational
diabetes, illegal drug use, or continuous prescription drugs, but information on
smoking habits was not included. Study populations appear to be similar, but
there may be some differences because subjects were selected from two different
cohorts that were recruited during slightly different time periods.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposure groups were similar despite the subjects coming from different original
study populations for whom different methods were used for recruitment.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Data were collected via questionnaire on maternal age, education,
marital status at first prenatal visit, birth order, birth weight, gestational age at
delivery, maternal smoking, maternal IQ, and HOME scores. All models were
adjusted for gestational age at birth, sex, birth weight, birth order, age at testing,
maternal marital status, smoking history, maternal age at delivery, maternal IQ,
education, and cohort, with additional testing for children’s urinary fluoride,
mercury, lead, and calcium. Sensitivity analyses were additionally adjusted for
HOME score. Covariates not considered included BMI, iodine deficiency,
arsenic, and maternal mental health and nutrition. Arsenic is assumed not to be a
potential co-exposure in this population as the study authors did not discuss it as
an issue, although other co-exposures were considered. Arsenic is included in the
water quality control program in Mexico City and thus is not considered a
concern in this population.
o Potentially important study-specific covariates: All key covariates were
addressed.

E-69
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

 Direction/magnitude of effect size: Not applicable.

o Basis for rating: Probably low risk of bias based on direct evidence that key
covariates, including other potential co-exposures, were considered, and indirect
evidence that the methods used to collect the information were valid and reliable
and that arsenic was not likely to be an issue in this study population.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: Although there was a large amount of attrition, the study authors
clearly describe all reasons for attrition and also provide characteristics to
compare those participants included to those excluded. There were some slight
differences between those included and those excluded, but there is no evidence
to indicate that the attrition would potentially bias the results.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Definitely low risk of bias (++)
o Summary: Urinary fluoride concentrations were determined in spot urine samples
(second morning void) collected from mothers (during at least one trimester) and
children ages 6–12 years. Fluoride content was measured using ion-selective
electrode-based assays. Quality control methods were described, including
between-laboratory correlations. All samples were measured in duplicate.
Extreme outliers were excluded. Urinary dilution was addressed by using
creatinine-adjusted levels.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Definitely low risk of bias based on direct evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Definitely low risk of bias (++)
o Summary: Outcome was assessed using the McCarthy Scales of Children’s
Abilities (MSCA) in 4-year-old children (translated into Spanish) and the
Wechsler Abbreviated Scale of Intelligence (WASI) in 6–12-year-olds. The
WASI is a well-established test, and the validity of both tests is well documented
by the authors. Inter-examiner reliability was evaluated and reported with a
correlation of 0.99 (++ for methods). The study report stated that psychologists
were blind to the children’s fluoride exposure (++ for blinding). Overall rating for
methods and blinding = ++.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
outcome was assessed using instruments that were valid and reliable in the study

E-70
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

population, and that the outcome assessor was blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Definitely low risk of bias (++)
o Summary:
 Statistical analyses: Statistical analyses used were appropriate for the study.
Statistical tests of bivariate associations (using Chi-square tests for categorical
variables and analysis of variance [ANOVA]) were used to compare the
means of the outcomes or exposures within groups based on the distribution of
each covariate. Generalized additive models (GAMs) were used to estimate
the adjusted association between fluoride exposure and measures of children’s
intelligence. Residual diagnostics were used to examine model assumptions
and identify any potentially influential observations. Results are reported as
adjusted regression slopes and 95% CIs. In sensitivity analyses, regression
models accounted for clustering at the cohort level by using cohort as a fixed
effect in the models. Although using cohort as a random effect would be more
appropriate, using individual-level exposure data and accounting for
numerous important covariates in the models likely captured the cohort effect.
Additional models with cohort as a random effect were also subsequently
made available via personal communication with the study authors and
showed similar results to the main model.
 Other potential concerns: None identified.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Definitely or probably low
risk-of-bias ratings in confounding, exposure, and outcome. Study strengths include
individual exposure assessment measurements, blinding of outcome assessor to
participants’ fluoride exposure, and the prospective cohort study design.

E.2.3. Bashash et al. (2018)

E.2.3.1. Study Details

• Study design: Prospective cohort

• Population: ELEMENT participants (pregnant mothers and their children aged 6–
12 years)
• Study area: Mexico City, Mexico

E-71
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Sample size: 210 mother-child pairs

• Data relevant to the review: Adjusted associations between maternal urinary fluoride
concentrations and ADHD and other attention/impulsivity scores.
• Reported association with fluoride exposure: Yes: Significant positive associations
between maternal urinary fluoride and Conners’ Rating Scales-Revised (CRS-R)
scores, including Cognitive Problems and Inattention Index (adjusted β = 2.54 per
0.5-mg/L increase; 95% CI: 0.44, 4.63), DSM-IV Inattention Index (adjusted β = 2.84
per 0.5-mg/L increase; 95% CI: 0.84, 4.84), DSM-IV ADHD Total Index (adjusted
β = 2.38 per 0.5-mg/L increase; 95% CI: 0.42, 4.34), and ADHD Index (adjusted
β = 2.47 per 0.5-mg/L increase; 95% CI: 0.43, 4.50).
E.2.3.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Participants were a subset of mother-child dyads enrolled in various
longitudinal birth cohort studies of the Early Life Exposure in Mexico to
Environmental Toxicants (ELEMENT) project. Subjects were included from two
of the four cohorts for which maternal urinary samples were available.
Participants in cohort 2A were recruited between 1997 and 1999, and participants
in cohort 3 were recruited from 2001 to 2003. Inclusion and exclusion criteria
were applied consistently across the two cohorts. A table of subject characteristics
was provided in the study, and any differences were considered in the analysis.
Study populations appear to be similar, but there may be some differences
because subjects were selected from two different cohorts: one from an
observational study on prenatal lead exposure and the other from a randomized
trial on the effects of calcium on blood lead levels. In addition, they were
recruited from slightly different time periods.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposed groups were similar, and any differences were considered in the analysis.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Questionnaires were used to collect information on maternal age,
maternal education, history of smoking, and marital status during the first
pregnancy visit. Child information at birth included birth weight, sex, birth order,
and gestational age as calculated by the nurse. Mothers also responded to an SES
questionnaire during the visit when the psychometric tests were administered. The
Home Observation for Measurement of the Environment (HOME) score was
evaluated in a subset of participants. Covariates were selected a priori. Models
were adjusted for maternal age at delivery, years of education, marital status,
smoking history, gestational age at birth, age at outcome assessment, sex, birth

E-72
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

order, SES, cohort, and calcium intervention. Arsenic is included in the water
quality control program in Mexico City and is not considered a concern in this
population.
o Potentially important study-specific covariates: None identified, although this
study did not specifically address arsenic or other co-exposures. Bashash et al.
(2017) addressed potential co-exposure to lead and mercury but did not address
arsenic. Arsenic was potentially addressed as part of the water quality program in
Mexico City.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias based on direct evidence that key
covariates were addressed, and indirect evidence that the methods used to collect
the information were valid and reliable and that arsenic and other potential co-
exposures were not likely to be an issue in this study population.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: Although there was a large amount of attrition from the original
cohorts, it was unlikely related to outcome or exposure, and there were very little
missing data from those included in the study. Of the 231 mothers with a
minimum of one maternal urine fluoride measurement and matching outcome
identified for the project, only 17 were excluded based on incomplete
demographic and outcome information.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Definitely low risk of bias (++)
o Summary: Mothers provided at least one spot urine sample during pregnancy. As
described in Bashash et al. (2017), urinary concentrations were determined on
second morning void. Fluoride content was measured using ion-selective
electrode-based assay. Bashash et al. (2017) describe QC methods. All samples
were measured in duplicate, and extreme outliers were excluded. Urinary dilution
was addressed by using creatinine-adjusted levels.
 Direction/magnitude of effect: N/A
o Basis for rating: Definitely low risk of bias based on direct evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Definitely low risk of bias (++)
o Summary: Behaviors associated with ADHD were assessed using the Spanish
version of Conners’ Rating Scales-Revised, which has been validated for the
evaluation of ADHD. Mothers completed the CRS-R at the same follow-up visit

E-73
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

in which the child completed the CPT-II tests. All tests were applied under the
supervision of an experienced psychologist (++ for methods). Use of only parent
reports and not teacher reports was noted by the authors as a study limitation
because there is considerable variation between the two sources in terms of
identifying ADHD-associated behaviors. Blinding was not reported, but it is
unlikely that the mothers were aware of their urinary fluoride levels. Although
mothers may have had knowledge that they were receiving fluoride through
fluoridated salt or naturally occurring fluoride in their water, they would not have
knowledge that this was relevant to the study purpose as the ADHD tests were
conducted for the original cohort (as was acknowledged by the study authors in
the discussion) (++ for blinding). Overall rating = ++.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessors were blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods were
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Definitely low risk of bias (++)
o Summary:
 Statistical analyses: Bivariate analyses included Chi-square tests for
categorical variables and ANOVA for continuous outcomes. Appropriate
univariate statistics and transformations were performed before bivariate
analyses. Residuals from fully adjusted linear regressions were checked and
suggested skewness. Gamma regression with an identity link was used to
examine the adjusted association between prenatal fluoride and each
neurobehavioral outcome (instead of using log transformation). Generalized
additive models were used to visually examine potential non-linearity.
Sensitivity analyses examined impact of other covariates. Diagnostics tests
were used to assess violations of the model assumptions and to identify
remaining influential observations. The Benjamini-Hochberg false discovery
rate (FDR) procedure was used to correct for multiple testing.
 Other potential concerns: None identified.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Definitely or probably low
risk-of-bias ratings in confounding, exposure, and outcome. Study strengths include

E-74
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

individual exposure assessment measurements, blinding of outcome assessor to

participants’ fluoride exposure, and the prospective cohort study design.

E.2.4. Choi et al. (2015)

E.2.4.1. Study Details

• Study design: Cross-sectional

• Population: First-grade children (ages 6–8 years)
• Study area: Mianning County in southern Sichuan, China
• Sample size: 51 first-grade children
• Data relevant to the review: Adjusted associations between urine or drinking water
fluoride levels and learning, memory, visual motor ability, motor ability, and manual
dexterity. Study also had information based on dental fluorosis score.
• Reported association with fluoride exposure: No: None of the outcomes were
significantly associated with fluoride exposure.
E.2.4.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Definitely low risk of bias (++)
o Summary: Subjects were selected during the same time frame using the same
methods. Fifty-one first-grade children residing in Mianning County in southern
Sichuan, China were included in this pilot study. It is not specified whether the 51
children represented all first-grade children from this area or whether some
refused to participate. Children who did not speak Chinese, were not students at
the Primary School of Sunshui Village in Mianning County, or those with chronic
or acute disease that might affect neurobehavioral function tests were excluded.
Demographic characteristics are presented in Table 1 of the study, which indicates
that subjects were similar. Covariates were adjusted for in the statistical analyses.
o Basis for Rating: Definitely low risk of bias based on direct evidence that the
exposure groups were similar and were recruited within the same time frame
using the same methods with no evidence of differences in participation/response
rates.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: The parents or guardians completed a questionnaire on demographic
and personal characteristics of the children (sex, age at testing, parity, illnesses
before age 3, and past medical history) and caretakers (age, parity, education and
occupational histories, residential history, and household income). A 20-μL
capillary blood sample was collected at the school by a Mianning County Center

E-75
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

for Disease Control (CDC) health practitioner and tested for possible iron
deficiency, which could be used as a covariate of neurodevelopmental
performance. Covariates that were not assessed include maternal BMI, parental
mental health, maternal smoking status, maternal reproductive factors, parental
IQ, and HOME score. However, the study authors noted that confounding bias
appeared to be limited due to the minimal diversity in the social characteristics of
the subjects. The study authors indicated that CDC records documented that levels
of other contaminants, including arsenic and lead, were very low in the area.
Iodine differences were not specifically addressed, but there is no indication from
the information provided that this might have been a concern.
o Potentially important study-specific covariates: All key covariates were
considered in this study.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias because there is direct evidence that
the key covariates were considered and indirect evidence that co-exposure to
arsenic was likely not an issue in this area and that methods used for collecting
the information were valid and reliable.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: The majority of results were reported for the 51 children stated to be
included in the pilot study. In Table 5 of the study, the N for each dental fluorosis
category totals only 43, but the text indicates 8 children did not have a Dean Index
because permanent teeth had not erupted.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: The study used three different measurements of fluoride exposure: well
water fluoride concentrations from the residence during pregnancy and onwards,
fluoride concentrations from children’s first morning urine samples, and degree of
children’s dental fluorosis. Fluoride concentrations in community well water were
measured and recorded by Mianning County CDC; specific methods were not
reported, but standard methods were likely used because analyses were conducted
by the CDC and were likely the same as those used to measure the fluoride in
urine. Migration of subjects was noted to be limited. Well water fluoride
concentrations of the mother’s residence during pregnancy and onward were used
to characterize a child’s lifetime exposure. To provide a measure of the
accumulated body burden, each child was given a 330-mL (11.2-oz) bottle of
Robust© distilled water (free from fluoride and other contaminants) to drink the
night before the clinical examinations, after emptying the bladder and before
bedtime. The first urine sample was collected at home the following morning, and

E-76
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

the fluoride concentration was determined on a 5-mL sample using an ion-specific

electrode at the Mianning CDC. There is no indication that urinary fluoride levels
accounted for dilution, nor was it clear that the method of administering water to
the children and collection methods sufficiently controlled for differences in
dilution. One of the investigators, a dentist, performed a blinded dental
examination on each child’s permanent teeth to rate the degree of dental fluorosis
using the Dean Index. The Dean Index is commonly used in epidemiological
studies and remains the gold standard in the dentistry armamentarium. The Index
has the following classifications: normal, questionable, very mild, mild, moderate,
and severe. Quality control (QC) procedures are not reported but were likely
appropriate.
 Direction/magnitude of effect size: Current levels were used to assess lifetime
exposure. This is likely to be a non-differential exposure misclassification,
and direction of bias is unknown. Because subject migration appears to be
limited, it is likely that the current fluoride levels are adequate reflections of
past exposure. Dental fluorosis would be an indicator that exposure occurred
in the past, and there was a fair correlation between degree of dental fluorosis
and current urine and water fluoride levels, with both increasing with
increasing levels of dental fluorosis.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measure exposure.
• Outcome:
o Rating: Probably low risk of bias (+)
o Summary: The study authors adopted culture-independent tests considered
feasible for children aged 6 to 8 years. The Wide Range Assessment of Memory
and Learning (WRAML) was used for the assessment of memory and learning.
Three subtests were also used. The Finger Windows subtest assesses sequential
visual memory. The Design Memory subtest assesses the ability to reproduce
designs from memory following a brief exposure. The Visual Learning subtest
assesses the ability to learn the locations of pictured objects over repeated
exposures. The Wechsler Intelligence Scale for Children-Revised (WISC-IV)
included digit span for auditory span and working memory and block design for
visual organization and reasoning. The grooved pegboard test assesses manual
dexterity. The tests used have been validated on a Western population. Although
there is no information provided to indicate that the tests were validated on the
study population, the study authors indicated that the tests were culture-
independent (+ for methods). Blinding of the outcome assessors or steps to
minimize potential bias was not reported. However, it is unlikely that the
assessors had knowledge of the individual exposure as children all came from the
same area, and water and urine levels were tested at the CDC (+ for blinding).
Overall = +.
o Basis for rating: Probably low risk of bias based on indirect evidence that all
outcomes were assessed using instruments that were valid and reliable in the

E-77
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

study population, and that the outcome assessors were blind to participants’
fluoride exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods are
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Statistical analyses were appropriate. Multiple regression
models evaluated the associations between exposure indicators and test scores
after adjusting for covariates. Specific regression models are not described or
referenced, just stated to be “standard regression analysis with confounder
adjustment.” The distributions of fluoride concentrations in urine and water
were skewed and were log10-transformed to approximate a Gaussian
distribution (test not specified). Results were reported as adjusted regression
slopes and 95% CIs. There was no evidence that residual diagnostics were
used to examine model assumptions; however, the impact on the effect
estimates is expected to be minimal.
 Other potential concerns: It should be noted that this study was a pilot study
and, therefore, had a relatively small sample size (i.e., 51 children).
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in the confounding, exposure, and outcome domains. Study strengths include
individual exposure assessment measurements with blinding at outcome assessment
likely. All key covariates and many other covariates were considered in the study
design or analysis.

E.2.5. Li et al. (2004) [translated in Li et al. 2008a]

E.2.5.1. Study Details

• Study design: Cross-sectional

• Population: Full-term, normal neonates 24–72 hours old from healthy mothers
• Study area: Zhaozhou County, Heilongiang Province, China
• Sample size: 91 neonates (46 males and 45 females)
• Data relevant to the review: Comparison of neurobehavioral capacity between
children in the high-fluoride area compared to the control area.

E-78
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Reported association with fluoride exposure: Yes: Significant differences in

neurobehavioral assessment total scores between high-fluoride (36.48 ± 1.09) and
control (38.28 ± 1.10) groups; significant differences in total neurobehavioral
capacity scores as measured by non-biological visual orientation reaction and
biological visual and auditory orientation reaction between the two groups
(11.34 ± 0.56 in controls compared to 10.05 ± 0.94 in high-fluoride group).
E.2.5.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: There is indirect evidence that the exposure groups were similar.
Participants were recruited during the same time frame using the same methods.
From 2002 to 2003, 273 neonates were born in a hospital in Zhaozhou County,
China. Ninety-one of 273 full-term neonates (46 males, 45 females) were
randomly selected. Mothers ranged in age from 20 to 31 years, met multiple
health criteria, and had not changed residence during pregnancy. Authors report
that the two study groups were located in the same area with similar climate,
living habits, economic and nutritional conditions, and cultural backgrounds, but
do not provide these data in the manuscript. There is no statistically significant
difference in the mode of delivery, birth weight, infant length, or sex. Subjects
were separated into exposure groups after random selection.
o Basis for Rating: Probably low risk of bias based on indirect evidence that the
exposure groups were similar and were recruited within the same time frame
using the same methods with no evidence of differences in participation/response
rates.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: No covariates were specifically considered in the analysis. The study
authors note similarities in characteristics in the two populations (i.e., living
habits, economic and nutritional conditions, and cultural backgrounds) but do not
provide these data nor do they indicate which specific characteristics were
considered. There were no significant differences in infant sex, birth method,
gestational age, or infant weight and length. All tests were conducted when
children were 1–3 days old. No potential co-exposures were discussed. Although
arsenic is considered a potential issue in China, water quality maps indicate that
there is a 25%–50% probability that the drinking water in that area exceeds the
WHO guideline for arsenic of 10 µg/L.
o Potentially important study-specific covariates: Key covariates, including age,
sex, and measures of socioeconomic status (SES), were similar between exposure
groups; however, arsenic was not considered. Arsenic often occurs in the drinking
water along with fluoride in some Chinese populations; however, based on water

E-79
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

quality maps, arsenic does not appear to be an issue in Zhaozhou County of the
Heilongjiang Province. Iodine deficiencies are not mentioned.
 Direction/magnitude of effect size: Conceptually, the presence of arsenic
would potentially bias the association away from the null if it were present
with fluoride. Deficiencies in iodine would potentially bias the association
away from the null if it were present in areas of higher fluoride but toward the
null if it were present in areas of lower fluoride. Neither of these are
considered a concern in this study for reasons detailed above.
o Basis for rating: Probably low risk of bias based on indirect evidence that the key
covariates were considered, co-exposure to arsenic was likely not an issue in this
area, and methods used for collecting the information were valid and reliable.
• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: Although authors did not discuss why only 91 of the 273 neonates
available were randomly selected, results were available for all 91 subjects.
o Basis for rating: Definitely low risk of bias based on results being available for all
subjects.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Subjects were split into control and high-fluoride groups based on
fluoride levels in their places of residence. Although the levels were provided
(1.7–6.0 mg/L for the high-fluoride group compared to 0.5–1.0 mg/L for the
control group), it was not reported how or when these levels were measured.
Urine was collected when women were hospitalized but before labor began. Urine
samples were sent to a specific lab for measurement using fluoride ion-selective
electrode. It was noted that this procedure strictly followed the internal controls of
the laboratory, indicating quality control. Level of detection (LOD) was not
provided. Urinary fluoride levels were significantly higher in the high-fluoride
mothers (3.58 ± 1.47 mg/L) compared to the control-group mothers
(1.74 ± 0.96 mg/L). There was indirect evidence that exposure was consistently
assessed using well-established methods that directly measure exposure. Although
results were mainly based on exposure area, they were supported by urine data,
making exposure misclassification less of a concern.
 Direction/magnitude of effect size: There is high variability in both water
fluoride and urine fluoride in the subjects from the high-exposure area.
Although there is no overlap in the water fluoride levels in the exposure areas,
there is some overlap in the urine concentrations in the mothers from the two
areas. This may reflect the single measurement and pose no specific bias, or it
could indicate that some mothers in the high-fluoride area have lower fluoride
exposure, which could bias the association toward the null.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measure exposure.

E-80
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Outcome:
o Rating: Probably low risk of bias (+)
o Summary: A standard neonatal behavioral neurological assessment method was
carried out by professionals in the pediatric department working in a neonatal
section trained specifically for these programs and passing the training exams (+
for methods). The examinations were carried out 1 to 3 days after delivery.
Because urine samples were collected on the day of delivery and sent to a separate
laboratory, it is likely that the outcome assessors were blind. Although the
subjects were separated by fluoride exposure area, it is not likely that the
professionals were aware of the exposure as the tests were conducted in the
hospital (+ for blinding).
o Basis for rating: Probably low risk of bias based on indirect evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessors were blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Probably low risk of bias (+)
o Summary: The study authors reported numerous outcomes in sufficient detail;
however, because a list of outcomes tested was not provided, there is no direct
evidence that all were reported.
o Basis for rating: Probably low risk of bias based on indirect evidence that all the
study’s measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Statistical analyses are described only as a t-test.
Consideration of heterogeneity of variance was not reported. Results are
reported as mean and standard deviations of neurological scores. Maternal
urinary fluoride levels were used only to compare exposures between exposed
and control groups. Infants in the control group were from four villages, and
those in the exposed group were from five villages within the same district.
Infants were randomly selected before they were assigned to exposed or
control groups. In the comparisons, there was no accounting for clustering at
the village level. It is likely that the standard error of the difference in mean
neurobehavioral assessment scores between the high fluoride group and
control group will be biased, making differences appear stronger than they
actually are. However, the use of multiple villages per exposure group is
likely to mitigate some of the impact of this lack of accounting for clustering,
and the overall impact on effect estimates is expected to be minimal.
 Other potential concerns: It should be noted that although the study states that
subjects were randomly selected, it is unclear why only 91 subjects were
included and whether they were randomly selected to obtain equal numbers in
the high-fluoride and control groups.

E-81
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Probably low risk of bias based on indirect evidence that
statistical analyses were appropriate and that there were no other potential threats
of risk of bias.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in the confounding, exposure, and outcome domains. Study strengths include
individual exposure assessment measurements to support the differences in the two
areas. Tests were noted to be conducted at the hospital, providing indirect evidence
that blinding was not a concern during the outcome evaluation. Although there was
some potential for bias due to the lack of accounting for arsenic or iodine
deficiencies, co-exposure to arsenic was likely not a major concern according to
groundwater quality maps.

E.2.6. Riddell et al. (2019)

E.2.6.1. Study Details

• Study design: Cross-sectional

• Population: Canadian Health Measures Survey (Cycles 2 and 3) participants
(children aged 6–17 years)
• Study area: General population, Canada
• Sample size: 3,745 children
• Data relevant to the review: Adjusted associations between water and urinary
fluoride and risk of ADHD and attention symptoms by water fluoride in the tap water
or community fluoridation status.
• Reported association with fluoride exposure: Yes: Significant positive association
between tap water fluoride and risk of ADHD diagnosis (adjusted OR = 6.10 per 1-
mg/L increase; 95% CI: 1.60, 22.8) and hyperactivity/inattentive symptoms (adjusted
β = 0.31 per 1-mg/L increase; 95% CI: 0.04, 0.58). Significant positive association
between community water fluoridation status and risk of ADHD diagnosis (adjusted
OR = 1.21 comparing fluoridated with non-fluoridated areas; 95% CI: 1.03, 1.42) and
hyperactivity/inattentive symptoms (adjusted β = 0.11; 95% CI: 0.02, 0.58). No
significant associations with urinary fluoride levels.
E.2.6.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Definitely low risk of bias (++)
o Summary: Subjects were part of Cycles 2 and 3 of the Canadian Health Measures
Survey. This is a nationally representative sample of residents living in 10
provinces. Specific inclusion criteria were provided. This study was restricted to
children 6–17 years of age with different fluoride measurements that consisted of
three participant samples. One of the samples was available only in Cycle 3.

E-82
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Definitely low risk of bias based on direct evidence that the
exposed groups were similar and were recruited with the same methods during the
same time frame.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: Covariates included in all models included age at testing, sex,
ethnicity, BMI, parents’ education, total household income, exposure to cigarette
smoke inside the home, and log-transformed concurrent blood lead levels.
Covariates such as parental behavioral and mental health disorders, quantity and
quality of caregiving environment, and co-exposure to arsenic were not discussed.
The study used data from the Canadian Health Measures Survey, which consists
of a nationally representative sample of Canadians. Most Canadians (~89%)
receive water from municipal water supplies, which monitor for levels of arsenic.
Therefore, co-exposure to arsenic is not likely an issue in this population.
Rationale for selection of covariates was based on relationship to ADHD
diagnosis and to fluoride metabolism based on literature review and consultation
with an ADHD expert. There is no information of the source of data for
covariates, but it is likely the questionnaires from the Canadian Health Measures
Survey, which are considered standardized and validated.
o Potentially important study-specific covariates: All key covariates were
considered in this study.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Probably low risk of bias because there is indirect evidence that
the key covariates were considered, co-exposure to arsenic was likely not an
issue, and methods used for collecting the information were valid and reliable.
• Attrition:
o Rating: Probably low risk of bias (+)
o Summary: There is no information indicating that there were any data excluded
due to missing covariates. All exclusions of children were described and
reasonable (i.e., drinking bottled water when considering city fluoridation as a
measure of fluoride exposure). Outliers were stated to be excluded, but methods
for determining this were provided, and it was noted that the outliers were 0.27%
of the values.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Urinary Fluoride: Spot urine samples were collected under normal
non-fasting conditions and analyzed using an Orion pH meter with a fluoride ion-
selective electrode after being diluted with an ionic adjustment buffer. Analysis
was performed at the Human Toxicology Laboratory of the Institut National de

E-83
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Santé Publique du Québec. The precision and accuracy of the fluoride analyses,
including quality control and quality assurance, were described by Health Canada
(2015). The limits of detection were 20 µg/L for Cycle 2 and 10 µg/L for Cycle 3
with no values below detection. Fluoride levels were adjusted for specific gravity.
Water Fluoride in Tap Water: Tap water was collected at the subjects’ homes
in Cycle 3 only. Samples were analyzed for fluoride concentrations using anion
exchange chromatography procedure with an LOD of 0.006 mg/L. Values below
the LOD were imputed with LOD/square root(2). Of the 980 samples, 150 (15%)
were below detection.
Chlorinated Water Fluoride Status: This was determined by viewing reports on
each city’s website or contacting the water treatment plant (provided in
supplemental material). Children were excluded if they drank bottled water, had a
well, had a home filtration system, lived in the current residence for 2 years or
less, or lived in an area with mixed city fluoridation.
 Direction/magnitude of effect size: There is not any specific impact on the
direction or magnitude of effect size expected. Urinary fluoride levels are
reflective of a recent exposure, but the study authors adjusted to account for
dilution. The possibility of exposure misclassification would be similar in all
subjects and would be non-differential. There is less potential for exposure
misclassification due to tap water or chlorinated water fluoride status, since
children who drank bottled water were excluded and children who had a home
filtration system were excluded from the chlorinated water status.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Probably high risk of bias (−)
o Summary:
Strengths and Difficulties Questionnaire (SDQ): The questionnaire was
administered to youths under 18 years. Children aged 6–11 years had SDQ ratings
provided by parents and guardians, but youths aged 12–17 years completed the
questionnaire themselves. Tests consist of 25 items with a 3-point scale. Items
were divided into five subscales: emotional problems, conduct problems,
hyperactivity-inattention, peer problems, and prosocial behavior. The current
study used only the hyperactivity-inattention subscale. Validation of this method
was not reported (− for methods).
ADHD: Ninety percent of youths with ADHD are diagnosed after age 6. For
children aged 6–11 years, ADHD diagnosis was provided by parents, but youths
aged 12–17 years completed the questionnaire themselves. Cycle 2 asked “Do you
have a learning disability?”; if the subject answered “yes,” he/she was asked to
specify the type (four options were available and described). In Cycle 3, parents
were asked directly whether they had ADHD, and children 12 years and older
were asked whether they had a physician diagnosis of ADHD and, if so, what

E-84
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

subtype (− for methods because different methods were used, and only the
children 12 years and older in Cycle 3 were asked specifically about a doctor’s
diagnosis). Both were measured in both cycles. Blinding is likely not an issue as
subjects would not have knowledge of the urine or tap water fluoride levels.
However, they would likely have knowledge of the city.
o Basis for rating: Probably high risk of bias based on indirect evidence that the
outcome was assessed using insensitive methods that varied based subject age.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods
sections were reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Robust logistic regression was used to examine the
association between fluoride exposure and ADHD diagnosis, adjusting for
covariates. Box-Tidewell tests were used to check the linearity of the
relationship with the continuous predictors. Linear regression was used for the
SDQ scores using Huber-White standard errors. Multicollinearity was
evaluated using variance inflation factor (VIF) statistics. Outliers with high
studentized residuals, high leverage, or large Cook’s distance values were
removed from all analyses with urinary fluoride. All regressions were tested
for interactions between fluoride exposure and age and between fluoride
exposure and sex. Sensitivity analyses were conducted to test the different
survey cycles. There is no mention of adjustment for the complex survey
design using survey weights or bootstrapped weights to ensure appropriate
calculation of the estimated variances; however, the overall impact on effect
estimates is expected to be minimal.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in confounding and exposure. Study strengths include individual exposure
assessment measurements and the addressing of key covariates but were limited by
the cross-sectional study design and insensitive outcome measures.

E.2.7. Rocha-Amador et al. (2009)

E.2.7.1. Study Details

• Study design: Cross-sectional

E-85
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Population: Children aged 6–11 years

• Study area: Durango, Mexico
• Sample size: 80 children
• Data relevant to the review: Associations between urinary fluoride levels and
visuospatial organization and visual memory (using the Rey-Osterrieth Complex
Figure Test, children’s version).
• Reported association with fluoride exposure: Yes: Significant inverse correlations
between urinary fluoride and visuospatial organization (r = −0.29) and visual memory
(r = −0.27) scores. No significant correlations with urinary arsenic.
E.2.7.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Subjects were from the same population and were recruited during the
same time frame using the same methods. Although this study compared three
sites with antecedents of environmental pollution to mixtures of either F-As, Pb-
As, or DDT-PCBs, authors evaluated each contaminant separately. The only area
of interest with F and As contamination is in Durango state (5 de Febrero) where
drinking water is polluted naturally with F and As at levels exceeding 6 and 19
times, respectively, the World Health Organization (WHO) limits (WHO 2008).
Children attending public schools were screened through personal interviews for
study eligibility. Inclusion criteria were children between 6 and 11 years old,
living in the study area since birth, whose parents signed the agreement to
participate. Children with a neurological disease diagnosed by a physician and
reported by the mother were excluded from the study. The final sample for the F-
As group was 80. Participation rates were not reported. Selected demographic
characteristics are presented in Table 1 of the study.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
populations were similar and recruited during the same time frame using the same
methods.
• Confounding:
o Rating: Probably high risk of bias (−)
o Summary: Covariates included blood lead (PbB), age, sex, and height-for-age z-
scores; only age had significant associations and was included in the final
analysis. Arsenic was also assessed and analyzed separately from fluoride.
Arsenic in urine was analyzed by atomic absorption spectrophotometer coupled to
a hydride system (Perkin-Elmer model AAnalyst 100). Although the model did
not adjust for arsenic, arsenic in the F-As group was not associated with either
outcome; therefore, arsenic co-exposure is not considered a major concern in this
study. PbB was analyzed with a Perkin-Elmer 3110 atomic absorption

E-86
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

spectrophotometer using a graphite furnace. Authors note that the mean blood
lead level in the F-As study area was 5.2 µg/dL, and 8% of the children had
values above the reference value of 10 µg/dL. PbB was stated not to affect results
and was not included in the final analysis. Other covariate data were obtained
during the study interview. Father’s education was provided and, in the F-As
group, was stated to range from 0–16 years, but this was not considered. Maternal
education, smoking, and SES were also not considered. The authors provide an
SES score of 5.9 ± 1.4 for the 5 de Febrero region (the fluoride region). It is not
clear whether this would vary by fluoride or arsenic levels.
o Potentially important study-specific covariates: SES.
 Direction/magnitude of effect size: There are insufficient data to determine the
impact on the magnitude or direction of effect size. The impact on the
direction of the association would likely depend on the association between
fluoride exposure and SES.
o Basis for rating: Probably high risk of bias based on indirect evidence that the
SES was not considered in the study design or analysis and may have varied by
fluoride levels.
• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: Data are complete. All 80 participants stated to be the final sample for
the site of interest (F-As) were included in all analyses.
o Basis for rating: Definitely low risk of bias based on direct evidence that there
was no attrition.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Fluoride in urine (FU) was analyzed according to method 8308
(“fluoride in urine”) from the National Institute for Occupational Safety and
Health (NIOSH 1984) with a sensitive specific ion electrode. As a quality control
check, reference standard “fluoride in freeze dried urine” (NIST SRM 2671a) was
analyzed. The accuracy was 97.0% ± 6.0%. Levels of FU and AsU were adjusted
for urinary creatinine, which was analyzed by a colorimetric method (Bayer
Diagnostic Kit, Sera-Pak1 Plus). However, details on the collection methods were
not reported.
 Direction/magnitude of effect size: Spot urine samples in a small sample size
(i.e., 80 children) may have some exposure misclassification. Adjusting for
dilution reduces the potential for misclassification based on differences in
dilution. Exposure misclassification would likely be non-differential.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Probably low risk of bias (+)

E-87
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Summary: IQ was assessed through the Rey-Osterrieth Complex Figure Test

(ROCF). This is a less well-established method, although the authors provide
citations suggesting it has been validated and standardized for the Mexican
population (+ for methods). According to the study report, the neuropsychologist
who administered the test was blinded to all exposure types and levels (++ for
blinding). Overall rating for methods and blinding = +.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessor was blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes outlined in the abstract, introduction, and methods were
reported in sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Statistical analyses used log-transformed exposure
variables (although rationale was not provided). Crude and partial correlations
were calculated to evaluate associations between serum fluoride levels and
TOCF scores. There is no other description of the regression model, and
regression diagnostics to evaluate model assumptions are not presented;
however, the overall impact on effect estimates is expected to be minimal.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in exposure and outcome. Study strengths include individual exposure
assessment measurements and blinding of outcome assessor to participants’ fluoride
exposure, but it is limited by the cross-sectional study design, lack of consideration of
SES in the study population, co-exposure with arsenic, and use of spot samples in a
small population.

E.2.8. Valdez Jiménez et al. (2017)

E.2.8.1. Study Details

• Study design: Prospective cohort

• Population: Infants aged 3–15 months
• Study area: Durango City and Lagos de Moreno, Jalisco, Mexico

E-88
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

• Sample size: 65 infants

• Data relevant to the review: The Bayley Scales of Infant Development II was used to
assess Mental Development Index scale and the Psychomotor Development Index
scale in children aged 3 to 15 months and evaluated for associations with first and
second trimester maternal urine fluoride.
• Reported association with fluoride exposure: Yes: Significant association between
maternal urinary fluoride and MDI score during first trimester (adjusted β = −19.05
per log10-mg/L increase; SE = 8.9) and second trimester (adjusted β = −19.34 per
log10-mg/L increase; SE = 7.46). No association between maternal fluoride during
any trimester and Psychomotor Developmental Index (PDI).
E.2.8.2. Risk of Bias

• Author contacts:
o Authors were not contacted for additional information because it was not
necessary.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Subjects were recruited from two endemic areas in Mexico. The study
authors do not provide information on the similarities or differences between the
two areas, nor do they indicate whether there were different participation rates.
However, recruitment methods were the same. Women receiving prenatal care in
health centers located in Durango City and Lagos de Moreno, Jalisco, Mexico
were recruited in 2013–2014. Participation rates are not likely to be an issue as
characteristics were similar between those who participated and those who did
not. Although the authors did not provide characteristics by area, the
characteristics provided do not indicate any differences that may be biased by the
selection. Considering the age range for the non-participants, the mean age for
non-participants appears to be incorrect (or the age range is incorrect); however,
there does not appear to be a difference that would potentially indicate selection
bias.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposure groups were similar and were recruited with the same methods in the
same time frame, with no evidence of differences or issues with
participation/response rates.
• Confounding:
o Rating: Probably high risk of bias (−)
o Summary: Questionnaires were used to obtain information about
sociodemographic factors, prenatal history, mother’s health status before
pregnancy (e.g., use of drugs, vaccines, diseases), and the type of water for
drinking and cooking. The marginalization index (MI) was obtained from the
National Population Council (CONAPO). Two additional surveys were conducted
during the second and third trimester of pregnancy to get information about the
mother’s health, pregnancy evolution, and sources of water consumption. A

E-89
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

survey was also conducted to get information about childbirth (type of birth, week
of birth, weight and length of the baby at birth, Apgar score and health conditions
of the baby during the first month of life). This information was corroborated with
the birth certificate. Linear regression models included gestational age, children’s
age, marginality index, and type of drinking water. Bivariate analyses were
conducted on the other factors, including sex, prior to conducting multivariable
regression models. Some important covariates were not considered, including
parental mental health, IQ, smoking, and potential co-exposures. Water quality
maps indicate a potential for arsenic to be present in the study area.
o Potentially important study-specific covariates: Arsenic is a potential co-exposure
in this area of Mexico.
 Direction/magnitude of effect size: If arsenic were present as a co-exposure, it
would likely bias the association away from the null.
o Basis for rating: Probably high risk of bias based on indirect evidence that there is
a potential for co-exposure with arsenic that was not addressed.
• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: Out of the 90 women selected for inclusion in the study, 65 approved
the participation of their infants. The authors provide a table of characteristics
between women who consented to their children’s cognitive evaluation and those
who participated only in biological monitoring. There were no significant
differences between the groups. There were fewer women who provided urine
during the second and third trimesters. All specified children are included in the
relevant analyses.
o Basis for rating: Definitely low risk of bias based on direct evidence that
exclusion of subjects from analyses was adequately addressed, and reasons were
documented when subjects were removed from the study or excluded from
analyses.
• Exposure:
o Rating: Definitely low risk of bias (++)
o Summary: Fluoride exposure was assessed through morning urine samples and
water fluoride levels collected from the children’s homes. Sampling methodology
was appropriately documented, and water levels were quantified through specific
ion-sensitive electrode assays. QC was described, and accuracy was >90%.
Urinary fluoride was corrected by specific gravity.
 Direction/magnitude of effect size: Not applicable.
o Basis for rating: Definitely low risk of bias based on direct evidence that
exposure was consistently assessed using well-established methods that directly
measured exposure.
• Outcome:
o Rating: Definitely low risk of bias (++)

E-90
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Summary: Neurodevelopment was assessed with the Bayley Scales of Infant

Development II (BSDI-II) that was noted to be reliable and valid for evaluating
children from 3 months to 5 years of age. The average age of children assessed
was 8 months, with a range of 3–15 months) (++ for methods). The study report
stated that a trained psychologist who was blinded to the mother’s fluoride
exposure evaluated the infants at home (++ for blinding). Overall rating for
methods and blinding = ++.
o Basis for rating: Definitely low risk of bias based on direct evidence that the
outcome was assessed using instruments that were valid and reliable in the study
population, and that the outcome assessor was blind to participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Probably low risk of bias (+)
o Summary: All outcomes outlined in the abstract, introduction, and methods were
reported. Table 4 of the study displays only data for trimesters 1 and 2. Although
third trimester data were collected, they were not reported, likely because they
were available for only 29 subjects. No discussion of this was provided.
o Basis for rating: Probably low risk of bias because, although it appears some data
were not reported, it is likely because there were insufficient data and not because
the authors were selectively reporting the results.
• Other potential threats:
o Rating: Probably low risk of bias (+)
o Summary:
 Statistical analyses: Statistical analyses used log10-transformed exposure
variables. Normality, homoscedasticity, and linearity assumptions were tested
and satisfied for MDI and PDI scores. Bivariate analyses included
correlations, t-tests, and ANOVA. Multiple linear regression models by the
first and second trimester of pregnancy were used to evaluate the association
between maternal fluoride exposure and MDI and PDI scores. The best-fit
model was selected using a “stepwise method,” and the best-fit line was
evaluated using “the curve fitting method.” It is not further specified or cited
what these methods entailed. Best-fit or goodness-of-fit statistics are not
reported. It is unclear how a best-fit model could be selected when the authors
state that all models adjusted for the same set of covariates regardless of
significance, and these covariates also appear in the final model—presumably
the best-fit model. It is unlikely that a stepwise method would retain all those
covariates unless they were forced in the model. Residual analysis was
conducted to assess model validity; however, there is no description of the
results of the residual analysis. Nonetheless, the impact on effect estimates is
expected to be minimal.
 Other potential concerns: No other potential concerns were identified. In the
peer-review report, NASEM (2020) cited the following as potential concerns:
“the large difference in numbers of males and females in the offspring (20

E-91
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

males, 45 females), and apparently incorrect probabilities were reported for

age differences between participants and nonparticipants, high rates of
cesarean deliveries and premature births among participants (degree of
overlap not reported), and incorrect comparisons of observed prematurity rates
with national expected rates.” However, these concerns were taken into
consideration in other domains (Selection, Confounding).
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and there were no other potential threats of
risk of bias identified.
• Basis for classification as low risk-of-bias study overall: Definitely low risk-of-bias
ratings in exposure and outcome. Study strengths include individual exposure
assessment measurements and blinding of outcome assessor to participants’ fluoride
exposure, but it is limited by the cross-sectional study design and lack of accounting
for potential co-exposures to arsenic.

E.2.9. Wang et al. (2020a)

E.2.9.1. Study Details

• Study design: Cross-sectional

• Population: School children aged 7–13 years
• Study area: Tongxu County, China
• Sample size: 325 school children
• Data relevant to the review: Adjusted associations between urine fluoride
concentrations and ADHD and other measures of learning disability.
• Reported association with fluoride exposure: Yes: Significant positive association
between urinary fluoride and psychosomatic problems (adjusted β = 4.01 per 1-mg/L
increase; 95% CI: 2.74, 5.28) and increased risk of a T-score >70 (adjusted OR = 1.97
per 1-mg/L increase; 95% CI: 1.19, 3.27). No significant associations with ADHD or
other measures of learning disability.
E.2.9.2. Risk of Bias

• Author contacts:
o Authors were contacted in July of 2020 to obtain additional information for risk-
of-bias evaluation. No response was received.
• Population selection:
o Rating: Probably low risk of bias (+)
o Summary: Subjects were recruited in 2017 from Tongxu County, China. Children
were selected from four randomly selected primary schools in the area. Selection
was based on specified inclusion rules. It was noted that the living habits and diets
of the participants from the four schools were well matched, but details were not
provided. The area did not have industrial pollution within 1 km of the living
environment of the children, and it was noted that the children were not exposed
to other neurodevelopmental toxicants (lead, cadmium, arsenic, or mercury). A

E-92
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

table of subject characteristics was provided in the study but not by school or
exposure. This was a pilot study, and it was not explicitly stated whether all
eligible subjects participated in the study. There is no information on participation
rates or whether they varied by school.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
exposed groups were recruited using similar methods during the same time frame
and that any differences between the exposed groups were considered in the
statistical analyses.
• Confounding:
o Rating: Probably low risk of bias (+)
o Summary: It was noted that subjects were well matched in terms of living habits
and diets, but there were no specifics provided. It was noted that there was no
industrial exposure or exposure to other neurotoxins such as lead, cadmium,
arsenic, or mercury. Covariates were collected using a standardized and structured
questionnaire completed by the children and their guardians under the direction of
investigators, but reliability or validity of the questionnaire was not reported.
Information collected included age, sex, weight, height, parental education level,
and parental migration (or work as migrant workers). IQ scores evaluated by the
Combined Raven’s Test–the Rural in China were used to represent basic
cognitive function. Models were adjusted for age, BMI, sex, mother and father
migration, and urinary creatinine. Adjustments were not made for parental
education, race/ethnicity, maternal demographics (e.g., maternal age, BMI),
parental behavioral and mental health disorders (e.g., ADHD, depression),
smoking (e.g., maternal smoking status, secondhand tobacco smoke exposure),
reproductive factors (e.g., parity), iodine deficiency/excess, maternal (and
paternal) IQ, quantity and quality of caregiving environment (e.g., HOME score),
or SES other than parental migration. There is no evidence to suggest that SES
would differ substantially among the four rural schools in the same area of China
that were randomly selected.
o Potentially important study-specific covariates: SES.
 Direction/magnitude of effect size: The impact on the direction and magnitude
of effect size are unknown. It was noted that the subjects were matched in
terms of living habits and diet, and this could be an indication that SES was
not different among the groups, but details were not provided.
o Basis for rating: Probably low risk of bias because there is indirect evidence that
the key covariates were considered, that the methods for collecting the
information were valid and reliable, and that co-exposure to arsenic was not an
issue in this area.
• Attrition:
o Rating: Definitely low risk of bias (++)
o Summary: Data are complete. It was noted that there were 325 subjects included,
and results were available on all subjects.

E-93
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Basis for rating: Definitely low risk of bias based on direct evidence that there
was no attrition.
• Exposure:
o Rating: Probably low risk of bias (+)
o Summary: Spot urine samples were collected from each child in the early morning
into cleaned polyethylene tubes. Fluoride concentrations were measured using
fluoride ion-selective electrode [with reference to Ma et al. (2017); however, that
reference cites Zhou et al. (2012)]. Therefore, no QC methods or LODs were
available. Fluoride concentrations were creatinine-adjusted.
 Direction/magnitude of effect size: Spot urine samples account for only recent
exposure. Although this could cause some exposure misclassification, the
number of subjects should help dilute any issues with the non-differential
misclassification.
o Basis for rating: Probably low risk of bias based on indirect evidence that
exposure was consistently assessed using acceptable methods that provide
individual levels of exposure.
• Outcome:
o Rating: Probably high risk of bias (NR)
o Summary: Children’s behavior was assessed by the Chinese version of Conners’
Parent Rating Scale-Revised (CPRS-48). The homogeneity reliability of
Cronbach α in the Chinese version of CPRS-48 was 0.932, the correlation of
Spearman-brown split-half was 0.900, and the retest reliability of total score was
0.594. Raw scores for each subscale were converted into sex- and age-adjusted T-
scores within a mean ± standard deviation (SD) of 50 ± 10. The guardians
independently completed the CPRS-48 according to the instruction manual under
the direction of trained investigators (++ for methods). Blinding is not reported.
Although it is unlikely that the outcome assessors were aware of the fluoride
levels in the urine, it is unclear whether subjects were selected based on areas
with endemic fluoride or whether parents were aware of fluoride concentrations in
the areas (NR for blinding). Overall rating for methods and blinding = NR.
o Basis for rating: Probably high risk of bias based on no information provided to
indicate that the outcome assessors were blind to the participants’ fluoride
exposure.
• Selective Reporting:
o Rating: Definitely low risk of bias (++)
o Summary: All outcomes in the abstract, introduction, and methods are reported in
sufficient detail.
o Basis for rating: Definitely low risk of bias based on direct evidence that all
measured outcomes were reported.
• Other potential threats:
o Rating: Probably low risk of bias (+)

E-94
Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

o Summary:
 Statistical analyses: Multiple linear regression models were used to assess the
association between urinary fluoride exposure and each behavioral outcome.
Logistic regression was used to assess the risk of behavioral problems (T-
scores >70) due to fluoride exposure. Sensitivity analyses were performed,
with models adjusting for combinations of age, BMI, sex, mother migrated,
father migrated, and urinary creatinine levels. Regression diagnostics to
evaluate model assumptions are not described; however, the overall impact on
effect estimates is expected to be minimal.
 Other potential concerns: None identified.
o Basis for rating: Probably low risk of bias based on indirect evidence that the
statistical analyses were appropriate and no other potential threats of risk of bias
were identified.
• Basis for classification as low risk-of-bias study overall: Probably low risk-of-bias
ratings in confounding and exposure. Study strengths include individual exposure
assessment measurements, but it is limited by the cross-sectional study design and
lack of details on blinding of the outcome assessment. All key covariates were
considered in the study design or analysis.

E-95
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Appendix F. Mechanistic Data from Animal Studies

Table of Contents
F.1. Neurotransmitters ................................................................................................................. F-2
F.2. Biochemistry (Brain/Neurons).............................................................................................. F-4
F.3. Histopathology...................................................................................................................... F-4
F.4. Oxidative Stress .................................................................................................................... F-5
F.5. Apoptosis/Cell Death ............................................................................................................ F-7
F.6. Inflammation......................................................................................................................... F-7
F.7. Thyroid ................................................................................................................................. F-7

Figures
Figure F-1. Number of Animal Mechanistic Studies for Fluoride by Mechanistic Category and
Exposure Level ......................................................................................................... F-2
Figure F-2. Number of Low Risk-of-bias Animal Studies That Evaluated Biochemical,
Neurotransmission, and Oxidative Stress Effects at or below 20 ppm by Mechanism
Subcategory and Direction of Effect......................................................................... F-7

F-1
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

A number of animal studies were available that presented mechanistic data in several effect
categories (see Figure F-1). Limiting the data to studies with at least one exposure at or below
20 ppm fluoride drinking water equivalents (gavage and dietary exposures were backcalculated
into equivalent drinking water concentrations for comparison) still provided a sufficient number
of studies for evaluation of several mechanistic endpoints while allowing for a more focused
look at exposure levels most relevant to human exposures. The following sections summarize the
mechanistic data by effect category. Although there is some evidence of consistency in
mechanistic effects, overall these data are insufficient to increase confidence in the assessment of
findings from human epidemiological studies.

Figure F-1. Number of Animal Mechanistic Studies for Fluoride by Mechanistic Category and
Exposure Level

An interactive version of Figure F-1 and additional study details are available at
https://public.tableau.com/app/profile/ntp.visuals/viz/FluorideTableauDashboards/ReadMe. The number of studies that evaluated
mechanistic effects associated with at least one exposure at or below 20 ppm fluoride is tabulated in the “≤20 ppm” column. The
total number of studies per mechanistic category is summarized in the “All” column.

F.1. Neurotransmitters

Neurotransmitter and biochemical changes in the brain and neurons were considered the
mechanistic areas with the greatest potential to demonstrate effects of fluoride on the brain of
animals in the lower dose range and provide evidence of changes in the brain that may relate to
lower IQ in children (see Figure F-2). Twenty of 23 neurotransmitter studies assessed changes in
brain cholinesterase activity associated with fluoride exposure at or below 20 ppm fluoride.
Acetylcholine is a major neurotransmitter involved in learning, memory, and intelligence (Chen
2012; Gais and Schönauer 2017). AChE is responsible for the breakdown of acetylcholine in the
synapses of nerve cells. Changes in cholinesterase, acetylcholine, or AChE could be related to
effects on memory. Evidence of an effect varied among the low risk-of-bias studies that assessed
changes in cholinesterase or acetylcholine (n = 11 drinking water studies) (Adedara et al. 2017a;
Akinrinade et al. 2015a; Baba et al. 2014; Chouhan et al. 2010; Gao et al. 2009; Gao et al. 2008a;
Khan et al. 2017; Liu et al. 2010; Mesram et al. 2016; Nkpaa and Onyeso 2018; Sun et al. 2000
[translated in Sun et al. 2008]), with the majority reporting evidence of an effect that is
considered inconsistent with the phenotypic outcome (see Quality Assessment of Individual

F-2
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Studies section for methods on determining which studies pose low risk of bias). Decreases in
cholinesterase will cause increases in acetylcholine, which can have a positive effect on learning
and memory; however, long-term decreases in cholinesterase can lead to secondary neuronal
damage occurring in the cholinergic region of the brain (Chen 2012).
Five of the 11 studies with low risk of bias (Adedara et al. 2017a; Baba et al. 2014; Gao et al.
2009; Khan et al. 2017; Nkpaa and Onyeso 2018) found statistically significant decreases in
cholinesterase or AChE in brain homogenates (with some brains dissected into specific regions
prior to homogenizing) with fluoride concentrations in drinking water at or below 20 ppm, and
four of the five studies found statistically significant decreases in cholinesterase or AChE below
10 ppm. The five studies were conducted in rats (Wistar or Sprague-Dawley) with exposure
ranging from 28 days to 6 months. An additional 2 out of 11 studies (Akinrinade et al. 2015a;
Gao et al. 2008a) reported decreases in brain homogenate AChE at concentrations at or below
20 ppm fluoride in drinking water, but statistical significance was not reached. These studies
were also conducted in rats with exposure for 30 days or 3 months. Gao et al. (2008a) reported a
dose-dependent decrease in brain homogenate AChE in the low (5 ppm fluoride) and high
(50 ppm fluoride) treatment groups compared with the control group, but the decrease was
statistically significant only in the high-dose group. Similarly, Akinrinade et al. (2015a) observed
a dose-dependent decrease in percent intensity of AChE immunohistochemistry in the prefrontal
cortex associated with 2.1 and 10 ppm sodium fluoride in drinking water, but neither result was
statistically significant. Gao et al. (2009) found lower brain homogenate AChE levels in the 5-
ppm animals compared with the 50-ppm animals; therefore, the results were not always dose-
dependent.
Relative to the above-mentioned studies, 2 of the 11 low risk-of-bias studies observed opposite
effects on brain cholinesterase levels. Sun et al. (2000) [translated in Sun et al. (2008)] observed
a significant increase in brain cholinesterase in Kunming mice associated with fluoride drinking
water concentrations from 10 to 100 mg/L but did not observe a dose response. Chouhan et al.
(2010) did observe a dose-related increase in AChE levels in brain homogenate of Wistar rats
with sodium fluoride concentrations of 1 to 100 ppm for 12 weeks and noted statistically
significant results at 1, 50, and 100 ppm but not at 10 ppm.
Mesram et al. (2016) did not assess changes in AChE but observed a significant decrease in
acetylcholine levels in cerebral cortex homogenate through 30 days of age in rats treated in utero
with 20 ppm sodium fluoride, which may suggest an increase in AChE levels. Likewise, Liu et
al. (2010) did not assess changes in AChE but measured nicotinic acetylcholine receptors
(nAChRs) in brain homogenate of rats following drinking water fluoride exposure, which the
authors stated could modulate physiological and pharmacological functions that are involved in
learning- and memory-related behaviors. Significant decreases in the protein expressions of
nAChR subunits at 2.26 ppm fluoride were observed; however, the corresponding receptor
subunit mRNAs did not exhibit any changes (Liu et al. 2010).
The studies that assessed other neurotransmitters of the brain and neurons were too
heterogeneous or limited in number to make any determination on mechanism, even before
limiting the review of the data to low risk-of-bias studies. There were only five studies that
evaluated dopamine and/or metabolites (Banala et al. 2018; Chouhan et al. 2010; Reddy et al.
2014; Sudhakar and Reddy 2018; Tsunoda et al. 2005). Four of the studies observed decreases in
dopamine levels in the brain with exposures of less than 20 ppm fluoride (Banala et al. 2018;

F-3
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Chouhan et al. 2010; Reddy et al. 2014; Sudhakar and Reddy 2018); however, the fifth study
(Tsunoda et al. 2005) observed increased dopamine and metabolites at fluoride exposures below
20 ppm (with statistical significance achieved only for the metabolite homovanillic acid in one
brain region). No differences from the control group were observed at levels above 20 ppm
fluoride. Other neurotransmitters were evaluated at or below 20 ppm fluoride exposure, but
generally only in a couple of studies.

F.2. Biochemistry (Brain/Neurons)

Similar to the above, the endpoints measured in brain biochemistry studies were too
heterogeneous or limited in number to make any determination on potential relevance of
mechanism, even before limiting the review of the data to low risk-of-bias studies (see
Figure F-2). Endpoints related to biochemical changes in the brain or neurons included
carbohydrate or lipid changes, RNA or DNA changes, changes in gene expression, or changes in
protein expression. For the most part, only a single study was available for any given endpoint.
The largest body of evidence on biochemistry was on protein level in various brain regions.
Eleven low risk-of-bias studies were identified that evaluated protein levels; however, few
studies evaluated the same proteins or areas of the brain. In the few cases in which the same
protein was evaluated, results were not always consistent. These data are insufficient to increase
confidence or support a change to hazard conclusions.

F.3. Histopathology

Histological data can be useful in determining whether effects are occurring in the brain at lower
fluoride concentrations; however, author descriptions of these effects may be limited, thereby
making it difficult to directly link histological changes in the brain to learning and memory
effects. Histopathology of the brain was evaluated in 31 studies with concentrations at or below
20 ppm fluoride, of which 15 were considered low risk-of-bias studies (Adedara et al. 2017b;
Akinrinade et al. 2015a; Bhatnagar et al. 2002; Bhatnagar et al. 2011; Chouhan et al. 2010;
Güner et al. 2016; Jia et al. 2019; Jiang et al. 2014; Lou et al. 2013; McPherson et al. 2018;
Mesram et al. 2016; Nageshwar et al. 2018; Niu et al. 2018; Pulungan et al. 2016; Zhao et al.
2019). In all but one low risk-of-bias study [Pulungan et al. (2016); gavage], animals were
exposed to fluoride via drinking water. All low risk-of-bias studies were conducted in rodents,
and all but three were conducted in rats (Wistar [seven studies], Sprague-Dawley [four studies],
Long-Evans hooded [one study]). Overall, the low risk-of-bias studies that evaluated
histopathology in the brain had low potential for bias for key questions regarding randomization
and exposure characterization; however, eight studies were rated as probably high risk of bias for
the key risk-of-bias question regarding outcome assessment based on lack of reporting of
blinding of outcome assessors and/or inadequate description of outcome measures or lesions.
Moreover, low image quality in some of the studies hampered the ability to verify the quality of
the data. Further technical review of the 15 low risk-of-bias studies was conducted by a board-
certified pathologist. Based on confidence in the results for each study, the technical reviewer
further categorized the low risk-of-bias studies as studies with higher or lower confidence in the
outcome assessment, which is reflected in the following summary of the brain histopathology
results. Main limitations of the histopathology data identified by the pathologist included lack of
information on methods of euthanasia and fixation. Perfusion fixation is generally considered the

F-4
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

best practice for lesions of the central nervous system in addition to complete fixation of the
brain prior to its removal from the skull (Garman et al. 2016). Four of the low risk-of-bias
studies reported that they used this method (Bhatnagar et al. 2002; Bhatnagar et al. 2011;
McPherson et al. 2018; Pulungan et al. 2016). Two of the low risk-of-bias studies handled the
brains before fixation was complete, which can produce artifacts that can resemble dead neurons
(Nageshwar et al. 2018; Zhao et al. 2019). Fixation and brain removal details were inadequately
described in the remaining low risk-of-bias studies.
Although there was heterogeneity in the endpoints reported (e.g., cell size, shape, and counts;
nuclei fragmentation; increased vacuolar spaces) and some variation in the consistency of the
evidence based on the area of the brain evaluated, the majority of the low risk-of-bias studies (11
of 14 drinking water studies) found some histological change in the brain of rats or mice treated
with fluoride at concentrations at or below 20 ppm, of which 8 studies reported histological
changes in the brain at or below 10 ppm. Histological changes in the hippocampus (one of the
areas of the brain most evaluated for histological changes) associated with fluoride exposure at
or below 20 ppm were reported in three of four low risk-of-bias studies with higher confidence in
the outcome assessment (Bhatnagar et al. 2002; Bhatnagar et al. 2011; Güner et al. 2016) and in
three of four low risk-of-bias studies with lower confidence in the outcome assessment (Jiang et
al. 2014; Nageshwar et al. 2018; Niu et al. 2018). McPherson et al. (2018) was the only drinking
water study (with higher confidence in the histopathology outcome assessment) that did not
observe any histological changes in hippocampus at 10 or 20 ppm fluoride in male Long-Evans
hooded rats exposed in utero through adulthood (>PND 80). Although there are too few studies
to definitively explain the inconsistency in results, McPherson et al. (2018) also did not observe
any associations between fluoride exposure and impairments to learning and memory, which is
inconsistent with the majority of developmental exposure studies that observed learning and
impairments associated with fluoride exposure for other strains of rats. Similarly, histological
changes in the cortex were reported in three of the four low risk-of-bias drinking water studies
with higher confidence in the outcome assessment (Akinrinade et al. 2015a; Bhatnagar et al.
2011; Chouhan et al. 2010) and in three of four low risk-of-bias studies with lower confidence in
the outcome assessment (Lou et al. 2013; Mesram et al. 2016; Nageshwar et al. 2018).
Histological changes were also consistently reported in other areas of the brain in studies with
higher confidence in the outcome assessment, including the amygdala, caudate putamen,
cerebellum, and hypothalamus, although each of these areas of the brain was evaluated in only
one low risk-of-bias study (Bhatnagar et al. 2011; Güner et al. 2016). Pulungan et al. (2016), one
of two low risk-of-bias studies with higher confidence in the outcome assessment that did not
report histological changes in the brain, observed a decreasing trend in the number of pyramidal
cells in the prefrontal cortex with increasing dose, but this was not changed at concentrations
below 20 ppm (the study administered sodium fluoride via gavage; the 5-mg/kg/day dose was
considered equivalent to 15.3 ppm fluoride in drinking water), nor were any of the results
statistically significant.

F.4. Oxidative Stress

Oxidative stress is considered a general mechanistic endpoint that cannot be specifically linked
to neurodevelopmental or cognitive effects in humans; however, like histopathology, it may help
in identifying changes in the brain occurring at lower concentrations of fluoride. Oxidative stress

F-5
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

in the brain was evaluated in 25 studies that examined concentrations at or below 20 ppm
fluoride, of which 15 studies had low potential for bias (Adedara et al. 2017a; Adedara et al.
2017b; Akinrinade et al. 2015b; Bartos et al. 2018; Chouhan and Flora 2008; Chouhan et al.
2010; Gao et al. 2009; Gao et al. 2008b; Güner et al. 2016; Khan et al. 2017; Mesram et al. 2016;
Nageshwar et al. 2018; Nkpaa and Onyeso 2018; Shan et al. 2004; Zhang et al. 2015). All of the
low risk-of-bias studies were conducted in rats (mainly Wistar or Sprague-Dawley) and
administered fluoride via drinking water with exposure durations ranging from 28 days to
7 months. Although there was heterogeneity in the endpoints reported (i.e., varying measures of
protein oxidation, antioxidant activity, lipid peroxidation, and reactive oxygen species [ROS])
and some variation in the consistency of the evidence based on the endpoint, the majority of the
studies (13 of 15) (Adedara et al. 2017a; Adedara et al. 2017b; Akinrinade et al. 2015b; Bartos et
al. 2018; Gao et al. 2009; Gao et al. 2008b; Güner et al. 2016; Khan et al. 2017; Mesram et al.
2016; Nageshwar et al. 2018; Nkpaa and Onyeso 2018; Shan et al. 2004; Zhang et al. 2015)
found evidence of oxidative stress in the brains of rats treated with fluoride at concentrations at
or below 20 ppm, of which 10 studies reported oxidative stress in the brain below 10 ppm
fluoride. The most consistent evidence of oxidative stress in the brain was reported through
changes in antioxidant activity. Eleven of the 12 low risk-of-bias studies that evaluated
antioxidant activity reported an effect at concentrations at or below 20 ppm (Adedara et al.
2017a; Adedara et al. 2017b; Akinrinade et al. 2015b; Bartos et al. 2018; Gao et al. 2009; Gao et
al. 2008b; Güner et al. 2016; Khan et al. 2017; Mesram et al. 2016; Nageshwar et al. 2018;
Nkpaa and Onyeso 2018). Decreases in antioxidant activity using measures of superoxide
dismutase (SOD) activity were reported in seven of eight low risk-of-bias studies (Adedara et al.
2017a; Adedara et al. 2017b; Akinrinade et al. 2015b; Khan et al. 2017; Mesram et al. 2016;
Nageshwar et al. 2018; Nkpaa and Onyeso 2018), and, among these seven studies, all that also
measured changes in catalase (CAT) activity (n = 6 studies) also reported decreased activity
(Adedara et al. 2017a; Adedara et al. 2017b; Khan et al. 2017; Mesram et al. 2016; Nageshwar et
al. 2018; Nkpaa and Onyeso 2018). A decrease in total antioxidant capacity (T-AOC) as a
measure of antioxidant activity was also consistently reported in two low risk-of-bias studies
(Gao et al. 2009; Gao et al. 2008b), and a decrease in glutathione peroxidase (GPx) activity was
reported in two of three low risk-of-bias studies (Adedara et al. 2017b; Nkpaa and Onyeso 2018).
Relative to the above-mentioned studies, 2 of the 15 low risk-of-bias studies (Chouhan and Flora
2008; Chouhan et al. 2010) did not observe statistically significant effects on oxidative stress in
the brain with concentrations at or below 20 ppm fluoride; however, the measure of oxidative
stress evaluated in Chouhan and Flora (2008) and Chouhan et al. (2010) (glutathione [GSH] to
oxidized glutathione [GSSG] ratio as an indication of antioxidant activity and ROS levels) were
not evaluated in any other low risk-of-bias study. Chouhan and Flora (2008) observed a dose-
dependent increase in ROS levels associated with 10, 50, and 100 mg/L sodium fluoride in
drinking water; however, results were not statistically significant at any dose. In Chouhan et al.
(2010), the levels of ROS were significantly higher at 50 ppm sodium fluoride in drinking water,
but statistical significance was not met at doses below 20 ppm fluoride (1 and 10 ppm sodium
fluoride) or at 100 ppm sodium fluoride; yet, hydrogen peroxide levels as a measure of ROS
were found to be significantly increased at 15 ppm sodium fluoride in drinking water in studies
conducted by another group of authors (Adedara et al. 2017a; Adedara et al. 2017b).

F-6
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

F.5. Apoptosis/Cell Death

Seven low risk-of-bias studies were identified that evaluated apoptosis with concentrations at or
below 20 ppm fluoride. Results from these studies were inconsistent and were insufficient for
evaluating fluoride-induced apoptosis. These data are insufficient to increase confidence or
support a change to hazard conclusions.

F.6. Inflammation

Five low risk-of-bias studies were identified that evaluated potential effects of fluoride on
inflammation with concentrations at or below 20 ppm. The inflammation markers were too
heterogeneous or limited in number to make any determination on potential relevance of
mechanism, even before limiting the review of the data to low risk-of-bias studies. These data
are insufficient to increase confidence or support a change to hazard conclusions.

F.7. Thyroid

Seventeen studies were identified that evaluated potential effects of fluoride on the thyroid with
concentrations at or below 20 ppm (see Figure F-1). These animal thyroid data are not further
described because this endpoint has been directly evaluated in a number of human studies that
have failed to identify consistent evidence to suggest that thyroid effects are a requisite
mechanism by which fluoride causes neurodevelopmental or cognitive effects in humans.

Figure F-2. Number of Low Risk-of-bias Animal Studies That Evaluated Biochemical,
Neurotransmission, and Oxidative Stress Effects at or below 20 ppm by Mechanism Subcategory
and Direction of Effect

An interactive version of Figure F-2 and additional study details are available at
https://public.tableau.com/app/profile/ntp.visuals/viz/FluorideTableauDashboards/ReadMe. This figure displays study counts for
low risk-of-bias studies, as these counts are most relevant to the text in this section. Counts for high risk-of bias studies or all
studies combined can be accessed in the interactive figure. Study counts are tabulated by significance—statistically significant
increase (↑), statistically significant decrease (↓), or not significant (NS). For example, the “↑” column displays numbers of
unique studies with at least one endpoint in the mechanistic subcategory with significantly increasing results at fluoride exposure
levels of ≤20 ppm. These columns are not mutually exclusive (i.e., a study may report on multiple endpoints with varying results
within a single mechanistic subcategory and therefore may be reflected in the counts for the “↑”, “↓”, and NS columns but would
be counted only once in the Grand Total column). Endpoints, species, strain, sex, and exposure duration are available for each
study in the interactive figure.

F-7
 Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review

Appendix G. Protocol History and Revisions

Table G-1. Protocol History and Revisions
Date Activity or Revision
December 14, 2016 Draft evaluation protocol reviewed: sent to technical advisors for peer review

April 10, 2017 Draft human risk-of-bias protocol reviewed: sent to technical advisors for peer
review
May 2, 2017 Draft animal risk-of-bias protocol reviewed: sent to technical advisors for peer
review
June 2017 Evaluation protocol finalized: Review protocol finalized for use and posting

May 29, 2019 Revised protocol: Revised review protocol posted

September 16, 2020 Revised protocol: Revised review protocol posted

G-1
Appendix H. Supplemental Files
The following supplemental files are available at https://doi.org/10.22427/NTP-DATA-
MGRAPH-8 (NTP 2024).

H.1. Protocol

NTP Protocol for Systematic Review of Human, Animal, and Mechanistic Evidence -
Second Revision (September 16, 2020)
ntpprotocol_revised20200916_508.pdf
NTP Protocol for Systematic Review of Human, Animal, and Mechanistic Evidence - First
Revision (May 29, 2019)
protocol_fluoridemay2019_508.pdf
NTP Protocol for Systematic Review of Human, Animal, and Mechanistic Evidence (June
2017)
protocol_fluoridejune2017_508.pdf

H.2. Datasets

Interactive Reference Flow Diagram (NTP Monograph Figure 2)

interactive_reference_flow_diagram_monograph8_figure_2.xlsx
Interactive Reference Flow Diagram for Updated Literature Search (NTP Monograph
Addendum Figure 1)
interactive_reference_flow_diagram_monograph8_addendum_figure_1.xlsx
NTP Monograph Interactive Figure 3
ntp_monograph8_interactive_figure_3.xlsx
NTP Monograph Interactive Figures 7 and 8
ntp_monograph8_interactive_figures_7_and_8.xlsx
NTP Monograph Interactive Figure F-1
ntp_monograph8_interactive_figure_f1.xlsx
NTP Monograph Interactive Figure F-2
ntp_monograph8_interactive_figure_f2.xlsx

H-1
National Toxicology Program
National Institute of Environmental Health Sciences
National Institutes of Health
P.O. Box 12233, MD K2-05
Durham, NC 27709
Tel: 984-287-3211
[email protected]

https://ntp.niehs.nih.gov ISSN 2378-5144