Project Summary
Pragmatic cluster randomized trials are often designed for patient-centered outcome and comparative effectiveness research with real-world populations. While the overall treatment effect has been the cornerstone in these studies, there is a growing interest in understanding how patient- and provider-level characteristics moderate the impact of new care innovations. In this context, the concept of heterogeneity of treatment effect (HTE) refers to potentially different treatment effects across pre-defined patient or provider subgroups that can arise due to various reasons, such as diverse practices, varying responses to treatment, or differing vulnerability to certain diseases. While the understanding of HTE has been a recognized goal in individually randomized trials, guidance on planning cluster randomized trials with HTE analysis is scarce, and it is currently unclear how to rigorously design pragmatic cluster randomized trials to ensure sufficient power for confirmatory HTE analysis. The objective of this research is to develop rigorous statistical methods, guidance, and software for planning cluster randomized trials to ensure sufficient power for confirmatory HTE analysis, addressing a critical methodological gap in designing definitive pragmatic trials.
First, the team will develop new methods for planning parallel cluster randomized trials with confirmatory HTE analysis. Under the parallel design, half of the clusters are simultaneously randomized to the intervention condition at baseline. The team focuses on a common hierarchical scenario in healthcare research where patients seek care from providers who are included in each clinic. For accurate sample size determination in relation to the HTE analysis, the team will first characterize the appropriate correlation structures for the outcome as well as the effect modifiers acknowledging this hierarchical structure, and then develop analytical expressions or numerical algorithms taking into account the within-clinic correlation structures. The team will further generalize the design methodology to accommodate anticipated missing data, which frequently happens in pragmatic trials due to patient attrition.
Second, the team will develop new methods for planning stepped wedge cluster randomized trials with confirmatory HTE analysis. Under the stepped wedge design, each cluster starts out from the control condition and randomly crosses over to the intervention condition at pre-defined time periods, or steps. The team’s development will separately consider the cross-sectional design, where different patients are included in different time periods for each cluster, and the closed-cohort design, where outcomes are repeatedly measured for the same cohort in each cluster. The team will develop variance expressions of the treatment effect that appropriately account for the longitudinal correlation structures of the outcomes and the effect modifiers, and therefore formally quantify the resources required to achieve adequate power for the HTE test.
Finally, the team will develop open-source software for planning parallel and stepped wedge cluster randomized trials with confirmatory HTE analysis. The team will illustrate its methods using three real-world pragmatic cluster randomized trials. The team will work with clinician and statistician stakeholders to translate its technical developments to practical guidance for designing pragmatic trials with HTE analysis and therefore make its results accessible to the broader research community.
To summarize, the team will create new statistical methods, guidance, and user-friendly software to design pragmatic cluster randomized trials with HTE analysis. This proposal fills an important methodological gap by expanding the current cluster randomized design toolbox to accommodate confirmatory HTE analysis, and will enable researchers to operationalize the PCORI methodology standards at the intersection of HTE and Research Designs Using Clusters. The proposed methods can be applied to pragmatic cluster randomized trials with any disease population where HTE assessment is of scientific interest.