A Case Study of a Bayesian Adaptive Cardiology Device Trial Leading to Approval

• Activity Number: 165
• Type: Topic Contributed
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Jason Connor*+
• Companies: Berry Consultants
• Keywords: Bayesian ; adaptive ; FDA ; medical device

Abstract

I review a 'Goldilocks' Bayesian adaptive trial that stopped accrual early for predicted success and led to an approval. I illustrate a single-arm trial with both efficacy and safety endpoints compared to objective performance criteria. The trial uses predictive probabilities with predefined stopping bounds along with a longitudinal model during the accrual phase to decide when to stop accruing patients. All patients are followed for 6-months after the final patient is enrolled. I will discuss the decision structure, simulations shown to FDA during the design process, the role of the historical prior in the longitudinal model and how it preserves Type I error, the stopping decision and trial results. I will also describe lessons learned -- when trying to get the sample size 'just right' it can be too easy to get it 'just wrong'.