Journal of Clinical Epidemiology 137 (2021) 23–30

ORIGINAL ARTICLE
Typology of drug discontinuation trials - Methodological
recommendations
Annika Viniol a,∗, Jörg Haasenritter a, Nina Grede a, Karl Wegscheider b, Annette Becker a,
Helmut Sitter c, Ildikó Gágyor d, Andreas Sönnichsen e, Achim Mortsiefer f,
Ulrike Junius-Walker g, Norbert Donner-Banzhoff a
a Department of General Practice, University of Marburg, Karl-von-Frisch Str. 4, 35043, Marburg, Germany
b Institute of Medical Biometry and Epidemiology, University Medical Center Hamburg, Hamburg, Germany
c Dr. Reinfried Pohl-Centre for Medical Education, University of Marburg, Marburg, Germany
d Department for General Practice, Julius-Maximilians-Universitat Würzburg, Würzburg, Germany
e Department of General Practice and Family Medicine, Center for Public Health, Medical University of Vienna, Vienna, Austria
f Institute of General Practice and Family Medicine, Faculty of Health, Witten/Herdecke University, Witten, Germany
g Institute for General Practice, Hannover Medical School, Hannover, Germany

Accepted 18 March 2021; Available online 26 March 2021

Abstract
Objective: Due to the increasing concerns about polypharmacy, there is a growing need for clinical recommendations for drug
discontinuation. This requires studies investigating the process on several levels. This paper addresses the methodological problems
of drug discontinuation trials (DDTs). To that end, we offer a new typology of research aims and corresponding methodological
recommendations for trials evaluating drug discontinuation.
Study Design and Setting: Multi-stage development process, including literature search and expert panels.
Results: Clinical trials are only required in cases of scientific uncertainty. We identified three situations of uncertainty associated
with drug discontinuation from which we derived three study types: 1) Uncertainty regarding the effectiveness and/or safety of a drug;
2) Uncertainty regarding the procedure of discontinuing a previously taken drug; 3) Uncertainty regarding the effectiveness of complex
strategies used to discontinue one or more drugs. We developed specific methodological recommendations for each study type.
Conclusion: We offer a comprehensive definition of research aims, study designs, and methodological recommendations regarding
DDTs. The typology we propose can help investigators clarify their research aims and study design. The type-specific methodological
recommendation should improve the quality of future drug discontinuation trials. © 2021 The Authors. Published by Elsevier Inc.
This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/ licenses/ by- nc- nd/ 4.0/ )

Keywords: Desprescribing; Drug discontinuation

1. Introduction falls and hospitalization, as well as adverse drug reactions

and drug-drug interactions [3,4]. Furthermore, polyphar-
Polypharmacy is a major health risk. Medication use has
macy is a financial burden to individuals and health care
steadily increased over the last two decades, particularly
systems [3,4].
among older people. [1,2]. While this is the result of more Stopping medication, however, can be difficult. In recent
and better treatment options, especially for chronic dis-
decades, few research groups (among them, the Bruyère
eases, polypharmacy is also associated with several adverse
Deprescribing Guidelines Research Team (Canada); Pri-
outcomes. The regular intake of five or more medications mary Health Tasmania (Australia), etc.) have pub-
is associated with an increased risk for higher mortality,
lished several medication-specific recommendations (e.g.,
withdrawal of antihypertensives, antipsychotics, benzodi-
azepines, opioids, proton pump inhibitors, etc.) [5,6]. In
Conflict of Interest: The authors declare that they have no known addition, several generic discontinuation strategies were
competing financial interests or personal relationships that could have
appeared to influence the work reported in this paper.
developed based on targeted questions to elicit appropri-
∗ Corresponding author: Annika Viniol, Karl-von-Frisch Str. 4, 35043, ateness of drugs, like the Medication Appropriateness In-
Marburg, Germany, Tel.: +49 6421 28-65130. dex (MAI), [7]. However, the evidence for deprescribing

https://doi.org/10.1016/j.jclinepi.2021.03.017
0895-4356/© 2021 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
24 A. Viniol et al. / Journal of Clinical Epidemiology 137 (2021) 23–30

medication is sparse in contrast to evidence provided by sign [8]. The large number of methodologically inadequate
classical clinical trials investigating efficacy of treatment. drug discontinuation studies motivated us to reflect upon
A closer look at trials evaluating drug discontinuation and discuss possible improvements to the evidence base of
reveals several limitations. The number of published stud- this important clinical topic.
ies is small, and many suffer from methodological flaws. We therefore employed the following multi-stage devel-
A systematic review by Grede and Thio et al. found inap- opment process:
propriate study designs, such as lack of control group, too
small sample size, and insufficient follow up periods [8,9].
2.1. Phase I – Type identification
A more thorough methodological approach to drug dis-
continuation studies is needed. Only a few research groups The working group consisted of the authors AV, NG,
are contributing to the evidence: Thompson et al. pub- JH, AB, and NDB. We collected clinical scenarios in which
lished a collection of articles addressing issues relevant the question of deprescribing drugs typically arises. These
to future investigations of deprescribing [10]. The authors scenarios were derived from relevant literature and clini-
formulated six priority areas: “1) Conducting high-quality cian members’ own experience as primary care physicians.
and long-term clinical trials that measure patient-important Based on typical clinical drug discontinuation scenarios,
outcomes, 2) Focusing on patient involvement and per- the team suggested the first draft of a typology for DDTs.
spectives, 3) Investigating the pharmacoeconomics of de-
prescribing interventions, 4) Understanding deprescribing 2.2. Phase II – Expert panel
interventions in different populations, 5) Generating evi-
dence on clinical management during deprescribing (e.g., The members of the expert panel were the authors IG,
managing adverse drug withdrawal effects, and subsequent AS, UJW, AM, HS, as well as our aforementioned working
re-prescribing), and 6) Implementing interventions in clin- group members. The participants are experts in the field of
ical practice.” [10]. Linsky et al. published a "deprescrib- multimorbidity, polypharmacy, drug trials, and drug dis-
ing conceptual framework", which separates the decision continuation. All expert panel members received the first
to deprescribe from the process of implementation [11]. version of the typology before a one day conference. They
They thereby emphasize the complex interaction between were asked to provide critical comments. These comments
patients, prescribers, and system influences [11]. In ad- were made available to members during the meeting and
dition, Blom et al. addressed unique features relevant to discussed in the group. Afterwards, disagreement in the
the reporting of deprescribing trials [12]. The results were comments was reviewed by the working group.
published as an elaboration of the CONSORT statement
[12].
2.3. Phase III - Development of the type-specific
The work listed above is an important prelude to the sci- methodological recommendation
entific debate on research designs for evaluating drug dis-
continuation, which we would like to expand upon through The types and the ideas of methodological recommen-
this project. We believe a first step should be to clarify the dations further developed in the panel were presented at the
objectives of a particular drug discontinuation trial (DDT). Institute of Medical Biometry and Epidemiology (Univer-
As a next step, details of study design and implementation sity of Hamburg, Germany). The methodological recom-
should be outlined. Here we propose a new typology of mendations were revised in collaboration with the working
research aims and offer methodological recommendations group of Prof. Wegscheider (expert for clinical studies).
for trial designs evaluating drug discontinuation.
2.4. Phase IV - Second voting expert panel + last
2. Methods revision
We established the working group “Methodology of After further revisions of the typology and type-specific
DDTs” at the Department for General Practice at the methodological recommendations, we again requested the
Philipps-Universität Marburg, Germany. As a first step, we expert panel’s opinions and comments. All members of the
conducted a systematic review of published DDTs (funded expert panel were asked to submit written comments. Our
by the German Federal Ministry of Education and Re- working group then developed the final version, presented
search) [8]. As expected, we found a wide range of clinical in this article. It was reviewed and unanimously endorsed
topics and study designs. The results also showed a sub- by all authors.
stantial heterogeneity of motivations for conducting discon-
tinuation studies, such as doubts concerning effectiveness 3. Results
and/or safety of drugs, proof of efficacy via discontinua-
tion, or changes in therapy regimes. In a large proportion The multi-stage development process mentioned above
of cases, study designs were incongruent with study ob- led to the following typology of research aims and result-
jectives; e.g., therapy was evaluated with observational de- ing methodological recommendations for DDTs.
 A. Viniol et al. / Journal of Clinical Epidemiology 137 (2021) 23–30 25

At inial prescripon: Changes during medicaon:

a b 2c d
2
Difference Difference Change of
between between individual paents
clinical populaon intervenon in pracce in For paents in a clinical
in pracce and clinical populaon context, the
and study intervenon
study populaon medicaon exceeds the
intervenon duraon
For example, regarding … For example, regarding … For example, regarding … of a trial.
- age - dosage - change in diseases severity
- gender - pharmaceucal form - increasing age or frailty
- degree of frailty - co-medicaon - changed co-medicaon
- comorbidies
- severity of diseases

If the scenarios raise doubts regarding the effecveness or safety of the long-term medicaon,
a drug disconnuaon trial may be necessary

Fig. 1. Clinical scenarios which lead to doubts on effectiveness and/or safety of long-term medication (Type 1)

3.1. Definition ‘Drug discontinuation trial’ Several scenarios are conceivable in which doubts arise
as to whether extrapolation from studies to clinical popu-
We define DDTs as studies which aim to evaluate the
lations can still be justified (Fig. 1).
discontinuation of one or more drugs that were prescribed (a) A difference between the clinical population in prac-
to patients in a clinical context.
tice and the study population that already exists at
the time of the first prescription. Possible reasons
are, for example, age, gender, degree of frailty, co-
3.2. Typology of research aims for evaluating drug morbidities, or severity of diseases.
discontinuation (b) A difference between the intervention in practice and
the study intervention that already exists at the time
Clinical trials are only justified if uncertainty exists. We
of the first prescription. These differences can refer
identified three types of uncertainty that are associated with
to, for example, dosage, pharmaceutical form, or co-
drug discontinuation: Type 1) Uncertainty regarding the
medication.
effectiveness and/or safety of a drug; Type 2) Uncertainty
(c) Patients in the clinical population undergoing
regarding the process of discontinuing a previously taken
changes during medication. These differences can re-
drug; Type 3) Uncertainty regarding the effectiveness of
late to, for example, disease severity, diagnostic tests
complex strategies used to discontinue one or more drugs.
becoming more sensitive, increasing age or frailty, or
changes in co-medication.
3.2.1. Type 1 – Doubts on effectiveness and/or safety (d) Patients are prescribed medication for much longer
Research question: Can drug X be discontinued with- periods than follow-up in clinical trials.
out harm compared to the continuation of medication? The These doubts refer to the effectiveness and/or safety of
motivation for this research question stems from doubts long-term medication. Several scenarios may occur simul-
about the efficacy and/or safety of a continuing drug ther- taneously. In these cases, a Type 1 DDT can be indicated.
apy. Table 1 illustrates examples of clinical scenarios from al-
Clinical trials never fully reflect the situations clinicians ready published discontinuation trials.
are confronted with in their daily practice. Participants in
clinical trials can differ from patients in a clinical con-
text in numerous ways, for example, with regard to the 3.2.2. Type 2 – Discontinuation procedure
severity of the target disease, demographic characteristics, Research question: How can drug X be discontinued
comorbidities, comedications, or general health state. In or- without rebound and/or withdrawal reactions? The clinical
der to let their patients benefit from evidence-based treat- starting point for this question is uncertainty regarding the
ments, clinicians, and also guideline authors, extrapolate discontinuation procedure. We define the discontinuation
from the results of published studies to their individual procedure as the practical steps required to terminate a
patients, even in the face of considerable discrepancies. particular long-term medication. In this setting, rebound or
26 A. Viniol et al. / Journal of Clinical Epidemiology 137 (2021) 23–30

Table 1. Examples for type-specific research questions

Type Clinical scenario Research Question Intervention References
Doubts on “… patients with “To evaluate whether a “Disease activity guided dose van Herwaarden N., et al.
effectiveness rheumatoid arthritis … disease activity guided reduction (advice to stepwise Disease activity guided
and/or safety using adalimumab or strategy of dose increase the injection interval dose reduction and
etanercept … with reduction of two tumour every three months, until flare withdrawal of adalimumab
stable low disease necrosis factor (TNF) of disease activity or or etanercept compared
activity …” inhibitors, adalimumab discontinuation) or usual care with usual care in
“…Treatment with TNF or etanercept, is (no dose reduction advice).” rheumatoid arthritis: open
inhibitors is not without non-inferior in label, randomised
its drawbacks: they are maintaining disease controlled, non-inferiority
associated with adverse control in patients with trial. BMJ. 2015 Apr
effects … [and] rheumatoid arthritis 9;350:h1389. doi:
treatment is costly…” compared with usual 10.1136/bmj.h1389. [1]
care.”
Discontinuation “Antiepileptic drug “We aim to establish if “Patients will be randomized Gasparini S., et al. Rapid
procedure withdrawal may be an a slow or a rapid to a slow (160 days) or a rapid versus slow withdrawal of
option for patients who withdrawal schedule of (60 days) schedule.” antiepileptic monotherapy
have been seizure free antiepileptic in 2-year seizure-free adult
for some years. The monotherapy influences patients with epilepsy
best withdrawal rate is relapse rate in adult (RASLOW) study: a
questionable…” patients with focal or pragmatic multicentre,
generalized epilepsy prospective, randomized,
who have been seizure controlled study. Neurol
free for at least 2 Sci. 2016 Apr;
years.” 37(4):579-83. doi:
10.1007/s10072-016-
2483-3. Epub 2016 Jan
25. [2]
Complex “Residents with a life “To examine successful “Multidisciplinary Multistep Wouters H., et al.
discontinuation expectancy greater than discontinuation of Medication Review (3MR) Discontinuing
strategy 4 weeks who consented inappropriate consisting of an assessment of Inappropriate Medication
to treatment with medication use and to the patient perspective, Use in Nursing Home
medication.” improve prescribing in medical history, critical Residents: A Cluster
nursing home appraisal of medications, a Randomized Controlled
residents.” meeting between the treating Trial. Ann Intern Med.
elder care physician and the 2017 Nov
pharmacist, and 7;167(9):609-617. doi:
implementation of medication 10.7326/M16-2729. Epub
changes.” 2017 Oct 10. [3]
References
[1] van Herwaarden N, van der Maas A, Minten MJM, van den Hoogen FHJ, Kievit W, van Vollenhoven RF et al. Disease activity guided dose
reduction and withdrawal of adalimumab or etanercept compared with usual care in rheumatoid arthritis: Open label, randomised controlled,
non-inferiority trial. BMJ 2015;350:h1389. https://doi.org/10.1136/bmj.h1389.
[2] Gasparini S, Ferlazzo E, Giussani G, Italiano D, Cianci V, Sueri C et al. Rapid versus slow withdrawal of antiepileptic monotherapy in
2-year seizure-free adult patients with epilepsy (RASLOW) study: A pragmatic multicentre, prospective, randomized, controlled study. Neurol
Sci 2016;37(4):579–83. https://doi.org/10.1007/s10072- 016- 2483- 3.
[3] Wouters H, Scheper J, Koning H, Brouwer C, Twisk JW, van der Meer H et al. Discontinuing Inappropriate Medication Use in Nursing Home
Residents: A Cluster Randomized Controlled Trial. Ann Intern Med 2017;167(9):609–17. https://doi.org/10.7326/M16-2729

withdrawal reactions are frequent. The procedure tries to palpitation and rise in blood pressure after cessation of
minimize the physiological reactions of drug elimination. beta blockers. Management of withdrawal syndromes and
‘Withdrawal syndromes’ are medication-specific acute rebound typically includes schemes of down tapering and
symptoms which typically appear immediately after dis- the use of rescue medication.
continuation and may resemble symptoms of the under- Investigators conducting Type 2 studies should start
lying disease; they are usually transient [13]. The term with the assumption that the discontinuation of a certain
rebound describes an intensified recurrence of the same drug is desirable. Research should focus on finding the
symptoms which had originally motivated treatment. This optimal way to achieve this for the individual patient.
effect is based on counter-regulatory mechanisms activated
by treatment and excessive counter-regulation after stop- 3.2.3. Type 3 – Complex discontinuation strategy
ping the medication [13]. Typical examples are worsening Research question: Does a complex discontinuation
reflux symptoms after stopping proton pump inhibitors, or strategy effectively reduce the intake of one or more drugs?
 A. Viniol et al. / Journal of Clinical Epidemiology 137 (2021) 23–30 27

We define a discontinuation strategy as a complex interven- 3.4.1. Research aim

tion to help clinicians and patients discontinue a long-term Independent of the type-specific focus (effectiveness
medication that is no longer considered beneficial. It typi- and/or safety vs. discontinuation procedure vs. complex
cally consists of information, advice, motivation, and emo- discontinuation strategy), the research questions should ad-
tional support. During this process, health care providers dress efficacy and safety (discontinuation without harm).
have to take the following steps: (1) provide the patient Further information about formulation of hypothesis (su-
with information on the benefits and/or risks of long-term periority vs. non-inferiority) is described in the sec-
medications; (2) explore the patient’s motivation for medi- tion 3.4.7 Outcomes.
cation and possible discontinuation; (3) address associated
fears, clarify preferences and goals that the patient consider 3.4.2. Study design
relevant; (4) if there are several candidate drugs, discuss In order to address confounders, the research questions
priorities; (5) propose practical steps including dose reduc- from all study types should be investigated within a ran-
tion, timing, use of rescue medication, follow-up visits, and domized controlled design. Type 3 studies typically require
help in case of emergencies [14–16]. Once a decision has a cluster randomization because health care providers con-
been made, it must be documented. A written plan should ducting the complex study intervention cannot be expected
be handed to the patient covering the issues mentioned to switch between intervention and control procedure. Ran-
above. Like in other attempts to change health-related be- domization at the health care unit level, such as practice,
haviors, such as smoking cessation or change in diet, the hospital ward, etc., is thus indicated.
challenge is approached with a realistic attitude [17,18].
Providers must be aware that such interventions are only 3.4.3. Study population
successful in a minority of cases [19]. DDTs include patients who have been prescribed a cer-
A well-known example for a complex discontinuation tain medication as part of their routine treatment. It is often
strategy is the Geriatric-Palliative Approach for Improving assumed that the long-term medication under consideration
Drug Therapy by Garfinkel et al. [20]. To evaluate this was prescribed without a planned withdrawal date. This
strategy, the authors compared the algorithm in the inter- applies, for example, to antihypertensive or lipid-lowering
vention group with a usual care control group. Outcome drugs. Oral anticoagulation after deep vein thrombosis pro-
variables were death rate, referrals to hospital, and costs in vides a different case, as this treatment is usually only
conjunction with change in medications [20]. For further prescribed for a limited time span.
examples, see Table 1. Another example for these types
of studies are trials that evaluated strategies for helping 3.4.4. Intervention-/ Control group
patients stop taking benzodiazepines. In Type 1 studies, medication is discontinued in the in-
tervention group; in the control group, medication is taken
as before. Type 2 and 3 studies aim to test the new discon-
3.3. Obsolete treatments tinuation procedure or strategy in the intervention group.
In the control group, an established procedure or strategy
Obsolete treatments may present Type 2 and Type 3 un-
should be conducted. The definition of cut-offs for safety
certainties. Thyroid hormones prescribed for years or even
indications and rescue measures can be indicated in Type
decades provide an extreme example. Diagnosis of relevant
1, 2, and 3 studies. Type 3 studies require the development
thyroid disease often dates back many years, the original
and exploratory testing of a complex intervention, which
documentation is often lost, and doctors re-prescribe the
usually includes training health care providers, information
medication with the assumption that there has been some
materials for patients, specific documentation, reminders
justification in the past. Attempts to stop the drug often
for health care teams, procedures for feedback, and audit.
meet patients’ resistance [21]. In order to gain evidence
See the British Medical Research Council Guidance [22].
for this dilemma, a DDT trial investigating discontinua-
tion procedures and evaluating complex strategies would
3.4.5. Randomization
be helpful.
Type 1 and 2 studies are typically randomized at the
patient level. In Type 3 studies, on the other hand, ran-
domization is usually carried out at the level of the work-
3.4. Methodological recommendations for discontinuation
trials ing units of the health care providers who carry out the
intervention.
Current frameworks for the design and conduct of clin-
ical trials have been developed with the introduction and 3.4.6. Blinding
evaluation of new treatments in mind. Most requirements Blinding patients and study and clinical personnel to
to avoid bias also apply to discontinuation trials. The ex- treatment allocation reduces observation bias and ensures
pert panel, however, made the following additional recom- adherence. Whether blinding is appropriate in a drug dis-
mendations specific to drug discontinuation (Table 2). continuation study depends on the research objectives of
28 A. Viniol et al. / Journal of Clinical Epidemiology 137 (2021) 23–30

Table 2. Type-specific methodological considerations

Type 1 Doubts on effectiveness Type 2 Discontinuation Type 3 Complex discontinuation
and/or safety procedure I strategy II
Study Patients with a long-term Patients with a long-term Patients with a long-term
population medication. medication. medication; typically nested at
The decision for healthcare units.
discontinuation has already
been made.
Research Can drug X be discontinued How can drug X be Does a complex
question without harm compared to the discontinued without rebound discontinuation strategy
continuation of medication? and/or withdrawal reactions? effectively reduce the intake of
one or more drug(s)?
Study design: Intervention Medication will be Discontinuation procedure Discontinuation strategy
RCT group discontinued (which should be evaluated) (which should be evaluated)
Control group Medication will be taken as Discontinuation procedure Discontinuation strategy
before (established procedure, e.g. (established strategy, e.g.
usual care) usual care)
Yes (Randomization at patient Yes (Randomization at patient Yes (Cluster randomization at
Randomization level) level) the level of the working units
of the health care providers
who carry out the
intervention). Randomization
after patient recruitment.
Blinding Double-blinding of providers Double-blinding of providers In case of provider
and patients. and patients. behavior-based interventions,
Placebo in the intervention blinding is often not possible
group is indicated. or explicitly not required.
Follow-up Follow-up time depends on the outcomes. Regular clinical monitoring to ensure patient safety.
Outcome Primary Co-primary endpoint consisting of a clinical safety endpoint (non-inferiority hypothesis) and the
discontinuation rate as an efficacy parameter (superiority hypothesis).
Secondary Further clinical efficacy and safety endpoint, economic outcomes, etc.
Analysis Intention-to-treat analysis + Per-protocol analysis
I Discontinuation procedures include practical steps required to terminate a particular long-term medication at patient level (aim: avoidance
of withdrawal or rebound phenomena)
II Discontinuation strategy implies a complex intervention for helping doctors go through the process of discontinuing a long-term medication.

the trial. For a Type 1 study evaluating the effect of a ical condition, such as pain intensity reported by patients
drug on disease episodes, such as exacerbations of COPD undergoing discontinuation of analgesics. Monitoring TSH
or cardiovascular events, blinding with an identical placebo and fT4 when stopping thyroid hormone would be another
in the discontinuation arm is appropriate. Investigators can example. The clinical safety endpoint should be evaluated
thus obtain an estimate of the biological effect of the drug. by a non-inferiority hypothesis implying that drug discon-
Cognitive or emotional reactions to the (non-) treatment tinuation can be undertaken without the disease worsening.
can thus be eliminated. In Type 3 trials, however, these As an efficacy outcome, in most instances the proportion
reactions are the very focus of the study. Dealing with the of patients in whom discontinuation has been successful is
patient’s subjective experience to discontinuation and her appropriate. The endpoint ‘discontinuation rate’ should be
psychological and physiological reactions, is an important evaluated by a superiority hypothesis.
task for clinicians. In our view, in Type 3 studies blinding For safety reasons, drug discontinuation should be care-
generally does not match the research objectives. For Type fully monitored (e.g., by patient interviews, clinical exami-
2 studies the choice of comparator and possible blinding nations, biomarker test, and medical imaging). Not achiev-
depends on the focus of study (physiological versus psy- ing discontinuation can be associated with several mech-
chological). anisms: deterioration of the disease, unmanageable with-
drawal symptoms, a psychological reaction by the patient,
3.4.7. Outcomes or interference by other health care providers, e.g., after
For all types of discontinuation trials, we recommend hospital admission. For a clinical recommendation to be
two co-primary endpoints: clinical safety, and discontinua- justified, both the safety and the efficacy evaluations must
tion rate as marker for efficacy. The clinical safety endpoint be successfully completed.
should be clinically relevant and related to the initial med-
 A. Viniol et al. / Journal of Clinical Epidemiology 137 (2021) 23–30 29

3.4.8. Follow-up priate” [24]. These two groups seem to be similar to Type
The duration of the follow-up depends on the condition, 3 (complex discontinuation strategy) and Type 1 (doubts on
the study drug, and the outcomes of interest. Symptom- effectiveness and/or safety) of our typology, respectively.
based measures, such as pain intensity after discontinuing However, we think that our Type 2 (discontinuation proce-
an analgesic, require only a short follow-up period. For dure) is an important addition to the framework. Evidence
drugs prescribed to prevent future disease events, such as for the discontinuation procedures is currently lacking, es-
vascular disease or exacerbations of COPD, more than 12 pecially in cases of medication with a high risk of de-
months are often required. pendency, such as benzodiazepines, opiates, or those with
a high risk of rebound, such as antidepressants or proton
3.4.9. Analysis pump inhibitors.
With respect to the co-primary endpoints mentioned Clinicians planning a drug discontinuation study often
above, we recommend both per per-protocol (PP) and require information which should ideally be obtained from
intent-to-treat analysis (ITT). The implementation of both other types of DDTs. For example, when planning a Type
procedures is based on the use of the co-primary end- 3 study, knowledge regarding discontinuation procedures
point. For the efficacy endpoint, usually part of a supe- is useful. This knowledge may be derived from formal
riority hypothesis, an ITT analysis is preferable [23]. For studies, but sometimes only clinical experience is available.
the non-inferiority hypothesis referring to the co-primary Do DDTs represent a particular study design entity? The
safety outcome, an additional PP analysis is justified [23]. suggestions regarding research aims, study population, and
All other statistical requirements for planning a DDT, like outcomes of DDTs made above support this consideration.
sample size calculation or the selection of statistical tests, On the other hand, the philosophy underlying discontin-
are concordant with established standards. uation studies and most design features are identical to
established clinical trials. For this reason, we do not pos-
3.4.10. Ethical considerations tulate an own design entity for drug discontinuations tri-
In order to ensure study patients’ safety, the evidence als, but suggest that certain aspects be taken into account
available prior to the study must be carefully reviewed. when designing a discontinuation study. We hope to mo-
As a useful analogy, researchers should have research pro- tivate investigators to undertake DDT studies and improve
grams in mind needed for the evaluation of novel treat- the quality of future studies in this area.
ments. Extensive animal as well as phase 1 and 2 stud- In their narrative review, Gnjidic et al. discuss the
ies in humans are conducted before the main efficacy and methodology of previously published DDT studies [25].
safety hypotheses are evaluated in phase 3 trials. Although They found that clinical endpoints as well as discontin-
this kind of evidence will neither be available nor relevant uation rates are used as outcomes in DDT studies [25].
to the issue of deprescribing, the success of a discontinu- In their recommendations for DDT studies, the Bruyère
ation study must be sufficiently likely to justify the inclu- Deprescribing Guidelines Research Team also formulate
sion and randomization of patients. In studies investigating that both clinical endpoints and discontinuation rates can
drugs preventing future disease episodes, an interim safety be used as outcomes [5]. We conclude that both are rel-
analysis can be appropriate. evant and should be evaluated concurrently as co-primary
outcomes; the discontinuation rate represents the focus of
4. Discussion DDTs, and the clinical endpoint highlights safety aspects.
Blom et al. developed recommendations for reporting
We propose a typology of DDTs addressing research DDTs extending the CONSORT statement [12]. Our work
objectives and study designs. The three types are derived complements their proposal with a typology of research
from clinical discontinuation situations, which are typically questions and recommendations for type-specific methods.
associated with uncertainty: 1) Uncertainty regarding the
effectiveness and/or safety of a drug; 2) Uncertainty re-
garding the procedure of discontinuing a previously taken 4.1. Implication for the future
drug; 3) Uncertainty regarding the effectiveness of complex Compared to classical clinical trials evaluating novel
strategies used to discontinue one or more drugs. Addi- treatments, deprescribing studies are a recent addition
tionally, we developed specific methodological recommen- to the world of clinical research. Study designs, choice
dations for each study type. of comparisons, outcomes, methods of analysis, to name
The need for a more differentiated view on the research but a few, are still in a state of flux. Before making
questions of drug discontinuation has already been de- these choices, however, investigators must define their
scribed elsewhere [10,24]. Reeve et al. describe two groups research objectives. Different objectives imply different,
of deprescribing studies: 1) “Studies focusing on whether sometimes even opposing, methodological consequences.
an intervention (e.g., educational intervention, medication We hope that our typology will help researchers make this
review) is effective”; 2) “Study targets a specific medica- important first step as they prepare studies investigating
tion…where use of this medication is considered inappro- drug discontinuation.
30 A. Viniol et al. / Journal of Clinical Epidemiology 137 (2021) 23–30

CRediT authorship contribution statement [10] Thompson W, Reeve E, Moriarty F, Maclure M, Turner J, Stein-
man MA, et al. Deprescribing: future directions for research. Res
Annika Viniol: Conceptualization, Project administra- Social Adm Pharm 2018. doi:10.1016/j.sapharm.2018.08.013.
tion, Supervision, Writing - original draft, Writing - re- [11] Linsky A, Gellad WF, Linder JA, Friedberg MW. Advancing the sci-
view & editing. Jörg Haasenritter: Writing - review & ence of deprescribing: a novel comprehensive conceptual framework.
J Am Geriatr Soc 2019;67(10):2018–22. doi:10.1111/jgs.16136.
editing. Nina Grede: Conceptualization, Writing - review
[12] Blom JW, Muth C, Glasziou P, McCormack JP, Perera R,
& editing. Karl Wegscheider: Conceptualization, Writing Poortvliet RKE, et al. Describing deprescribing trials better: an elab-
- review & editing. Annette Becker: Conceptualization, oration of the CONSORT statement. J Clin Epidemiol 2020;127:87–
Writing - review & editing. Helmut Sitter: Conceptualiza- 95. doi:10.1016/j.jclinepi.2020.07.011.
tion, Writing - review & editing. Ildikó Gágyor: Concep- [13] Henssler J, Heinz A, Brandt L, Bschor T. Antidepressant withdrawal
and rebound phenomena - a systematic review. Dtsch Arztebl Int
tualization, Writing - review & editing. Andreas Sönnich- 2019;116:355–61.
sen: Conceptualization, Writing - review & editing. Achim [14] Duncan P, Duerden M, Payne RA. Deprescribing: a primary care
Mortsiefer: Conceptualization, Writing - review & edit- perspective. Eur J Hosp Pharm Sci Pract 2017;24(1):37–42. doi:10.
ing. Ulrike Junius-Walker: Conceptualization, Writing - 1136/ejhpharm- 2016- 000967.
review & editing. Norbert Donner-Banzhoff: Conceptu- [15] Scott I, Anderson K, Freeman C. Review of structured guides for
deprescribing. Eur J Hosp Pharm 2017;24(1):51–7.
alization, Supervision, Writing - original draft, Writing - [16] Scott IA, Hilmer SN, Reeve E, Potter K, Le Couteur D, Rigby D,
review & editing. et al. Reducing inappropriate polypharmacy: the process of de-
prescribing. JAMA Intern Med 2015;175(5):827–34. doi:10.1001/
jamainternmed.2015.0324.
Acknowledgements [17] Baird J, Jarman M, Lawrence W, Black C, Davies J, Tinati T, et al.
The effect of a behaviour change intervention on the diets and phys-
We would like to thank Juliette Rautenberg for provid-
ical activity levels of women attending sure start children’s cen-
ing English-language editing, improving the precision and tres: results from a complex public health intervention. BMJ Open
fluency of the manuscript. 2014;4(7):e005290. doi:10.1136/bmjopen- 2014- 005290.
[18] Tarquinio C, Kivits J, Minary L, Coste J, Alla F. Evaluating com-
plex interventions: perspectives and issues for health behaviour
References change interventions. Psychol Health 2015;30(1):35–51. doi:10.
1080/08870446.2014.953530.
[1] Guthrie B, Makubate B, Hernandez-Santiago V, Dreischulte T. The
[19] Johnson MJ, May CR. Promoting professional behaviour change in
rising tide of polypharmacy and drug-drug interactions: Population
healthcare: what interventions work, and why? a theory-led overview
database analysis 1995-2010. BMC Med 2015;13:74. doi:10.1186/
of systematic reviews. BMJ Open 2015;5(9):e008592. doi:10.1136/
s12916- 015- 0322- 7.
bmjopen- 2015- 008592.
[2] Gao L, Maidment I, Matthews FE, Robinson L, Brayne C. Medica-
[20] Garfinkel D, Zur-Gil S, Ben-Israel J. The war against polypharmacy:
tion usage change in older people (65+) in England over 20 years:
A new cost-effective geriatric-palliative approach for improving drug
Findings from CFAS I and CFAS II. Age Ageing 2018;47(2):220–5.
therapy in disabled elderly people. Isr Med Assoc J 2007;9(6):430–4.
doi:10.1093/ ageing/ afx158.
[21] Viniol A, Bösner S, Baum E, Donner-Banzhoff N. Forgotten drugs:
[3] Midão L, Giardini A, Menditto E, Kardas P, Costa E. Polypharmacy
long-term prescriptions of thyroid hormones - a cross-sectional
prevalence among older adults based on the survey of health, ageing
study. Int J Gen Med 2013;6:329–34. doi:10.2147/IJGM.S43187.
and retirement in Europe. Arch Gerontol Geriatr 2018;78:213–20.
[22] Craig P, Dieppe P, Macintyre S, Michie S, Nazareth I, Petticrew M.
doi:10.1016/j.archger.2018.06.018.
Developing and evaluating complex interventions: the new Medical
[4] van den Akker M, Vaes B, Goderis G, van Pottelbergh G, Burgh-
Research Council guidance. BMJ 2008;337:a1655. doi:10.1136/bmj.
graeve T de, Henrard S. Trends in multimorbidity and polypharmacy
a1655.
in the Flemish-Belgian population between 2000 and 2015. PLoS
[23] The European agency for the evaluation of medicinal products,
ONE 2019;14(2):e0212046. doi:10.1371/journal.pone.0212046.
committee for proprietary medicinal products. Points to consider
[5] Bruyère deprescribing guidelines research team; Available from:
on switching between superiority and non-inferiority; Available
https:// deprescribing.org/ .
from: https:// www.ema.europa.eu/ en/ documents/ scientific-guideline/
[6] Primary Health Tasmania; Available from: https://www.
points- consider- switching- between- superiority- non- inferiority_en.
primaryhealthtas.com.au/.
pdf.
[7] Hanlon JT, Schmader KE, Samsa GP, Weinberger M, Uttech KM,
[24] Reeve E, Thompson W, Farrell B. Deprescribing: a narrative re-
Lewis IK, et al. A method for assessing drug therapy appro-
view of the evidence and practical recommendations for recogniz-
priateness. J Clin Epidemiol 1992;45(10):1045–51. doi:10.1016/
ing opportunities and taking action. Eur J Intern Med 2017;38:3–11.
0895- 4356(92)90144- c.
doi:10.1016/j.ejim.2016.12.021.
[8] Grede N, Kuss K, Donner-Banzhoff N, Viniol A. Methods of drug
[25] Gnjidic D, Le Couteur DG, Kouladjian L, Hilmer SN. Deprescrib-
discontinuation trials - a systematic review. Congress Abstract (Ger-
ing trials: methods to reduce polypharmacy and the impact on pre-
man Network for Evidence based Medicine) 2017.
scribing and clinical outcomes. Clin Geriatr Med 2012;28(2):237–
[9] Thio SL, Nam J, van Driel ML, Dirven T, Blom JW. Effects of
53. doi:10.1016/j.cger.2012.01.006.
discontinuation of chronic medication in primary care: a systematic
review of deprescribing trials. Br J Gen Pract 2018;68(675):e663–
72. doi:10.3399/bjgp18X699041.