A Prospective Bayesian Adaptive Trial with Hierarchical Borrowing from a Prior Single Arm Study
View Presentation View Presentation
Kristine Broglio, Berry Consultants  *Jason T Connor, Berry Consultants 

Keywords: Bayesian, adaptive, historical data

We describe an adaptive randomized controlled Bayesian device trial with historical hierarchical borrowing for a cardiac device with both prospective efficacy and safety outcomes. This prospective 2-arm trial incorporates an adaptive sample size plus adaptive borrowing from a 1-armed trial of a similar device in Europe. Using a Bayesian hierarchical model, the design is structured so if US data is similar to EU data, then more borrowing occurs, and less US data is needed. If the US and EU data are different, less borrowing will occur and the US trial will run to a larger sample size, providing independent data to substantiate the device's safety & effectiveness in an American patient population. We illustrate the trial, its operating characteristics and show how the actual trial progressed, stopping early for success, and leading to FDA approval.