By Cory E. Goldstein, Stephanie N. Dixon, Monica Taljaard, Charles Weijer.
As bioethicists and statisticians, we are occasionally invited to consult with a research team planning a pragmatic clinical trial after the research question and a basic outline of objectives and methods have been developed. When the proposed design is a cluster randomized design (in which groups, such as clinics or hospitals, are the units of randomization), we are commonly asked whether consent is necessary. For instance, if a medical director provides permission to enrol their clinic, do we need to get consent from patients? Is consent required from patients whose hospitals are randomized to provide usual care? Is it permissible to give people the option to opt-out of a trial rather than soliciting prospective written consent in order to enhance pragmatism (i.e., include all or most patients with the clinical condition of interest)?
We respond to these queries by shifting the conversation away from consent and, instead, we direct researchers to important upstream questions such as “why has a cluster randomized design been chosen?”
This was the impetus for writing our manuscript, “Navigating the consent river: questions to consider before waiving consent requirements in pragmatic cluster randomized trials.” While cluster randomized designs can be useful, a general recommendation is to only use this design when necessary. As we explain in our paper, cluster randomized trials may have some advantages, but they often require more participants, more resources, and are less likely to produce scientifically valid result compared to individually randomized designs.
For these reasons, the use of a cluster randomized design requires a compelling justification. However, sometimes the choice to use cluster randomization for evaluating an individual-level intervention (i.e., an intervention delivered to individuals within a cluster such as drugs or vaccines) is not adequately justified. If the design choice cannot be justified, an individually randomized design is preferrable and, consequently, questions about whether consent is required drift away since informed consent is almost always required in individually randomized trials.
But, in some cases, the use of a cluster randomized design is defensible and once again the consent waters are muddied. Research teams may still feel like they are up the creek without a paddle. Hence, we distilled our paper into a flow chart for research teams and research ethics committees to use when designing and reviewing trials. The starting point asks researchers to clarify the nature of the intervention under evaluation: is the intervention being delivered to individuals within clusters, or to clusters as an intact unit? The justification for using cluster randomization, as well as how to mitigate biases and the suitability of a waiver of consent, depends on the answer to this question (and some other considerations). It is our hope that this flow chart will be useful not only for our future consults, but for anyone who may need to navigate the consent river in pragmatic cluster randomized trials.
Authors: Cory E. Goldstein;1,2 Monica Taljaard;1,2 Stephanie N. Dixon;3,4 Charles Weijer4,5,6
Affiliations:
1 Methodological and Implementation Research Institute, Ottawa Hospital Research Institute, Ottawa, Ontario, Canada.
2 School of Epidemiology and Public Health, University of Ottawa, Ottawa, Ontario, Canada.
3 London Health Sciences Centre Research Institute, London, Ontario, Canada.
4 Department of Epidemiology and Biostatistics, Western University, London, Ontario, Canada.
5 Departments of Medicine, Western University, London, Ontario, Canada.
5 Department of Philosophy, Western University, London, Ontario, Canada.
Competing interests: Charles Weijer receives consulting income from Eli Lilly & Company. Other authors declare no competing interests.
Social media accounts of post author(s): @coryegoldstein; @charlesweijer