To the Editor,

In their article, Spence et al. argue that cluster randomized crossover trials are “the preferred method to evaluate questions of effectiveness” in anesthesia, and that “individual RCTs are not the preferred design.”1 To illustrate the utility of this design, the authors discuss the B-Free pilot trial and “expect to begin the [full-scale] trial in the summer of 2018.”1

In our commentary, we argue, “individual (not cluster) randomization is the gold standard study design to evaluate questions of both efficacy and effectiveness.”2 Our arguments are not specific to the B-Free trial since, to date, no protocol has been published. Our conclusion is general: “the cluster crossover design… raises methodological and ethical concerns that must be carefully addressed.”2

Lee et al. agree that we make “correct statements about methodological concerns for cluster crossover trials in general.”3 They agree that these include increased risks of bias, limited external validity, imbalances in baseline characteristics, and carry-over and period effects, and they explain how each will be addressed in the full-scale B-Free trial.3 We note that Lee et al. make no effort to defend the original thesis that cluster crossover trials are generally the preferred design to evaluate questions of effectiveness.

We disagree with several points made by Spence et al.4 First, they conflate units of randomization and intervention. A cluster level intervention, such as community public health messages, is indivisible at the individual level. But the institutional policy of benzodiazepine administration is an individual-level intervention, as it is divisible individually. Patients in the intervention arm will receive benzodiazepines “unless there are contraindications”1—and refusal to consent should be a contraindication.

Second, they assert that we take the “extreme position that informed consent is always required for individual-level interventions.”4 This is false. Our position is that “[g]enerally . . . [for] individual-level interventions, informed consent is required.”2 The qualifier “generally” implies that exceptions may be justifiable. The presumption of consent is shared by other ethical analyses and guidelines specific to cluster randomized trials.5,6,7,8,9

Third, they claim that “established requirements to justify a waiver” and the existence of “many completed and ongoing cluster trials… where a waiver has been accepted” show the permissibility of conducting the B-Free trial without consent.4 Yet this ignores the novelty of the cluster crossover design. Canadian and U.S. guidelines were not written with cluster trials in mind.10,11 As a result, local research ethics committees that review cluster randomized trials may be unfamiliar with the specific guidelines of this unique design.8,9

Finally, with respect to the flu vaccine trial, contributor roles are outlined in the trial publication12: the trial was in progress when Dr. Taljaard joined as an independent collaborator to supervise the statistical analysis.