Scribd
Upload
52 views55 pages

Research Methods 2

This document discusses strengths and weaknesses of different study designs for identifying disease risk factors. It covers concepts like external and internal validity, confounding, bias, and how different study types address these issues. Specifically, it discusses how randomized trials, cohort studies, case-control studies, and cross-sectional studies handle confounding differently. It also provides examples of types of bias like selection bias, information bias, and how they can be addressed. The goal is to understand how different study designs incorporate these concepts to make inferences about disease causation.

Uploaded by

Dr Ahmed Nabil
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
52 views55 pages

Research Methods 2

This document discusses strengths and weaknesses of different study designs for identifying disease risk factors. It covers concepts like external and internal validity, confounding, bias, and how different study types address these issues. Specifically, it discusses how randomized trials, cohort studies, case-control studies, and cross-sectional studies handle confounding differently. It also provides examples of types of bias like selection bias, information bias, and how they can be addressed. The goal is to understand how different study designs incorporate these concepts to make inferences about disease causation.

Uploaded by

Dr Ahmed Nabil
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 55

Strengths and Weaknesses

of Different Study Designs

Jo Freudenheim, PhD
Department of Social and Preventive Medicine
Disclosure Statement
 Jo Freudenheim, PhD
 Honoraria: none
 Consultancies: NCI
 Common stock: none
 Research: NIH, DOD, Avon
Goals and objectives
 Overall goal:
Identification of strengths and
weaknesses of study designs
 External and internal validity
 Confounding/Interaction
 Bias
 Incorporating different study designs
 External and internal validity
 Confounding/Interaction
 Bias
 Incorporating different study designs
Study Validity

 Our ability to make inferences based


on the study findings
Study Validity: external validity

 The generalizability of the study


findings
 To whom do the study findings apply
Study Validity: external validity

 Example: case control study of breast


cancer among women aged 35-70 in
Western New York
 How generalizable are those findings?
Study Validity: external validity

 Example: cohort study among


postmenopausal women in the
Women’s Health Initiative
 How generalizable are those findings?
Study Validity: internal validity

 Are the methods of the study


appropriate, no major concerns with
methodology
Study Validity: internal validity
 Are the methods of the study
appropriate, no major concerns with
methodology
 Confounding/Interaction
 Bias
 External and internal validity
 Confounding/Interaction
 Bias
 Incorporating different study designs
Causal Association: Direct

 The factor under study causes the


disease
 The observed association is causal

 Ex: HPV causes cervical cancer


Causal Relations
Causal relations
 Factor is in the causal pathway
 Not a confounder
 Is another step on the path from
exposures of interest to outcome
Types of Causal Relations
Types of Causal Relations
Types of Causal Relations
Types of Causal Relations
Non-Causal Association: Confounding

 The factor under study does not


cause the disease even though it is
observed to be associated with the
disease.
 There is another factor that is
correlated with both the factor under
study and with the disease which is
the true causal factor
Non-Causal Association: Confounding

 People with yellow staining on their


fingers are more likely to get lung
cancer
 Does yellow stain cause lung cancer?
Non-Causal Association: Confounding

 It is reported that people who drink


more coffee are more likely to get
pancreatic cancer
 Does coffee cause pancreatic cancer?
Non-Causal Association: Confounding

 It is reported that women with a


history of herpes infection are more
likely to get cervical cancer
 Does herpes cause cervical cancer?
Confounding
 Smokers have “yellow” finger
 Coffee drinkers are more likely to be
smokers
 Those who are more sexually active,
more likely to be exposed both to
herpes and to HPV
 In both cases, causal agent is a
different, associated factor
Non-causal association
(confounding)

 Important to understanding public


health significance of a finding
 If the prevalence of a risk factor in
the population changes, will the
incidence of disease change?
Confounding
Confounding

 Approaches to handling confounding


 In designing and carrying out the study
 Individual matching
 Group matching
 In the analysis of data
 Stratification
 Adjustment
Example of Confounding
Sucrose intake and risk of endometrial cancer: Swedish
Mammography Cohort
Total Cases Person- Age- Multivariate
sucrose g/d years adjusted RR adj* RR
< 15 112 203213 1.00 (ref) 1.00 (ref)
16-25 248 342793 1.35 1.50
26-35 158 239495 1.25 1.41
> 36 211 338431 1.16 1.36
*Adjusted for age, BMI, coffee, energy, diabetes

Friberg et al, CEBP 2011;20:1831-7


One More Concept . . .

 Interaction
 When the incidence of disease in the
presence of two or more risk factors
differs from the incidence expected to
result from their individual effects (cohort
study)
 2 or more factors modify the effects of
each other with respect to disease
 Effect modification
Interaction

 Is there an association?
 Is the association due to confounding?
 Is the association equally strong in
strata formed on the basis of the 3rd
variable?
 If no, then interaction is present
 If yes, then no interaction is present
Interaction

 Is the association equally strong in strata


formed on the basis of the 3rd variable?
 If no, then interaction is present
 If yes, then no interaction is present
 Interaction is a biological difference
 The association is different in the different
strata
Confounding and study design
 Cross sectional studies
 Case control studies
 Cohort studies
 Randomized trials
Interaction and study design

 Cross sectional studies


 Case control studies
 Cohort studies
 Randomized trials
 External and internal validity
 Confounding/Interaction
 Bias
 Incorporating different study designs
Bias

 Bias =
 Any systematic error in the design,
conduct or analysis of a study that
results in a mistaken estimate of an
exposure’s effect on the risk of disease
(Schlesselman and Stolley, 1982)
Sources and Types of Bias

 Selection bias
Systematic differences in those included in
a study and those who are not
Sources and Types of Bias

 Selection bias:
 If selection of subjects does not reflect
population:
 Lack of generalizability or external validity
 Not selection bias
 If selection of subjects is different between
groups within a study:
 Lack of internal validity
 Selection bias
Sources and Types of Bias

 Information bias:
 Problems with the information obtained
about subjects
 Information about disease status
 Information about exposure
Sources and Types of Bias

 Information bias can lead to:


 Misclassification bias:
 Non-differential
 If the misclassification is not related to the
exposure status or disease status
 Similar misclassification of exposed and non-
exposed or those with disease and without
disease
 RR or OR is biased towards 1.0
 Less likely to detect a difference
Sources and Types of Bias

 Information bias can lead to:


 Misclassification bias:
 Differential:
 If the misclassification is related to the exposure
status or disease status
 Differential misclassification in exposed and
non-exposed, or in those with disease and
without disease
 RR or OR is biased away or towards 1.0
 Less or more likely to detect a difference
Sources and Types of Bias

 Information bias:
 Surveillance bias:
 Diseases are more likely to be diagnosed
in persons under medical surveillance
 Example: Thrombophlebitis and oral contraceptive
use. MDs may examine women using OCs more
often and more thoroughly
 Possible solutions:
 Stratify cases and controls according to some
index of medical care utilization
 In a prospective study, systematically assess
outcome in both exposed and non-exposed
Sources and Types of Bias

 Information bias:
 Observer bias:
 When exposure status is known to the
observer, assignment to outcome (yes/no)
may be biased
 Example: Knowledge of exposure to alcohol
may increase likelihood of alcoholic cirrhosis dx
 Possible solutions:
 Mask exposure status
 Standardize procedures
Sources and Types of Bias

 Information bias:
 Interviewer bias:
 Knowledge of case-control status may affect
way in which questions on past exposure are
asked
 Possible solutions:
 Mask disease status
 Standardize procedures
 Monitor quality during study (taping)
Sources and Types of Bias

 Information bias:
 Recall bias:
 Bias from a difference in the ability to recall past
exposure between cases and controls
 Example: OR of maternal rubella with congenital
malformations in offspring were greater in the
case-control studies than RR in prospective studies
 Possible solutions:
 Cohort studies
 Validate responses
 Nested case-control studies
Sources and Types of Bias

 Information bias:
 In abstracting records
 Surrogate interviews
 Reporting bias
 Perceptions, beliefs, attitudes
 Over or underreporting
Bias and Confounding

 Bias:
 Any systematic error in the design,
conduct or analysis of a study
 Confounding:
 Not an error in a study
 External and internal validity
 Confounding/Interaction
 Bias
 Incorporating different study
designs
Studying Etiology of Disease

 Conclusions about what causes a disease


is based on the totality of evidence
 Animal models
 In-vitro systems
 Human population studies
 Our understanding of what causes a
disease and also how best to prevent it
continues to change and to be revised
Studying Etiology of Disease

 Animal models
 Advantages
 disadvantages
 In-vitro systems:
 Advantages
 disadvantages
 Human population studies
 Advantages
 Disadvantages
Studying Etiology of Disease

 Animal models
 Controlled exposures
 Detail regarding impact on tissues and specific
organs
 Between species differences
 In-vitro systems:
 Frequently use human cells
 Cellular level
 Difficult to extrapolate to functioning body
 Issues of dose
 Human population studies
Studying Etiology of Disease

 Animal models
 In-vitro systems
 Human population studies
 Randomization not always ethical
 Concerns of confounding and bias
 Issues of measurement of exposure
 Report of exposure
 Changes in exposure
 Other correlated exposures
1940’s,
United Kingdom

You might also like