Title: Bayesian adaptive stepped wedge cluster randomized trials based on posterior predictive probability
Authors: Song Zhang - University of Texas Southwestern Medical Center (United States) [presenting]
Abstract: Bayesian group sequential design has been widely used in clinical studies, especially in phase II and III studies. It is flexible and efficient in allowing early termination based on the accumulated data through Bayesian framework. However, so far there has been no discussion on its application to stepped wedge cluster randomized trials, which has become more popular in pragmatic trials in clinical and health care delivery studies. We propose a Bayesian strategy to design a cross-sectional stepped wedge cluster randomized trial based on posterior predictive probability. It provides additional flexibility for trialists to continuously evaluate interim observations, and make adaptive decisions accordingly to ensure trial success (for example, stopping the trial early for efficacy or futility, enrolling additional clusters, etc). Simulation algorithms and application examples are presented.