Speakers
Monica Taljaard, PhD
Senior Scientist, Clinical Epidemiology Program
Ottawa Hospital Research Institute
Full Professor, Epidemiology and Community Medicine
University of Ottawa
David Magnus, PhD
Thomas A. Raffin Professor of Medicine and Biomedical Ethics and Professor of Pediatrics, Medicine, and by courtesy of Bioengineering
Director, Stanford Center for Biomedical Ethics
Associate Dean for Research
Keywords
Ethics, Stepped Wedge Cluster Randomized Trial, Study Design
Key Points
- There are two main types of clinical trials: patient randomized trial and cluster randomized trial (CRT). The patient randomized trial is always preferable; cluster randomized trials should only be used when there is no other choice because CRTs will always require a larger sample size, have higher risks of bias, be more vulnerable to chance imbalance between arms, and are more complicated to design and analyze.
- There are several good reasons to adopt a cluster randomization trial, including that the intervention is a cluster-level intervention or the research question of interest pertains to cluster-level effects.
- If a cluster randomized trial is justified there are different types of CRT designs, the parallel arm CRT, parallel arm before and after CRT, cluster cross-over, and stepped wedge. Investigators should provide a justification for choice of stepped wedge CRT.
- Five methodological arguments for stepped wedge CRTs are to improve rigor, to facilitate recruitment, to reduce the required sample size, to simplify logistics, and to reduce bias; however, most methodological arguments in favor of stepped wedge trials have a counter argument. Trialists should work with a biostatistician to come up with the most scientifically robust design given the practical constraints of the study. In most cases, a parallel arm or cluster cross would be a better choice.
- In a study of more than 100 stepped wedge cluster randomized trials, the most common reason for the design was to be able to make the intervention available to all clusters by the end of the trial based on ethical or equity grounds rather than anything methodological.
- During the Ebola crisis, the ethical argument for stepped wedge CRTs was that anything involving a placebo in the control arm was automatically considered unethical if the intervention arm holds a chance of benefit. This led to a shift to stepped wedge design trials, but in Ebola there was no evidence that benefits outweighed the harms of proposed therapeutics and vaccines and there was confusion that all participants get the intervention in a stepped wedge CRT, where each cluster may get the intervention but all participants may not get the intervention, depending on the design.
- Arguments in favor often confuse individual belief in the benefit of intervention with equipoise (which requires consensus in the field that the intervention is of benefit). If truly not in equipoise, delay in providing the intervention is no more justified than the placebo. True protection is clinical equipoise and it is neutral between stepped wedge CRT and parallel research designs.
Discussion Themes
– Should we be allowing stepped wedge trials in the Collaboratory? As a statistician, I agree but it’s important to do the extra work and look at alternative designs. If stepped wedge design is the best option, you have to ensure that the analysis is robust and sound and that there is not recruitment bias. The policy level point is well taken; saying that they are prohibited might be too strong, but you do need a justification or rationale for stepped wedge design, especially over parallel.
– Could stepped wedge be a good design for implementation trials? It’s still important to look at how effective it is and is the juice worth the squeeze in this context? You should use the best statistical methods to answer the question you are trying to answer.
Tags
#pctGR, @Collaboratory1