Repeated inclusion cluster randomized trials: a new class of designs for assessing group-level interventions

Individually randomized trials allow participants to be included in a trial multiple times, with independent randomization at each inclusion. Often referred to as re-randomization designs, these trials have been shown to increase trial recruitment rates. Treatment effect estimators remain unbiased but are more precise than for designs with more participants but no repeat inclusions, but do rely on additional assumptions. Here, we introduce a new class of cluster randomized trial designs: repeated inclusion cluster randomized trials, where some clusters are randomized serially in the same trial. The trial in which clusters are initially included may have a standard cluster randomized design or a longitudinal variant, such as a cluster randomized crossover design. Allowing clusters to participate multiple times in the same trial could reduce the need to recruit new clusters; useful when cluster recruitment is difficult. Assuming a constant treatment effect across repeated inclusions, we show that when equal numbers of clusters and participants are included in each treatment group in each study period, power will be the same or higher as for a similar trial where clusters are not re-randomized but which has the same total number of measurements. Whether power is maintained or increased depends on the study design and the within-cluster correlation structure; increasing the number of within-cluster comparisons increases study power.