Bayesian Methods for Combining Medical Device Studies under FDA Review

• Activity Number: 182
• Type: Topic Contributed
• Date/Time: Tuesday, August 10, 2004 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Gene A. Pennello*+
• Companies: U.S. Food and Drug Administration
• Keywords: hierarchical model ; effective sample size ; Dirichlet process prior ; synthetic control ; historical control ; meta-analysis

Abstract

To support pre-market approval of a medical device under review by FDA, Bayesian methods are sometimes used that combine studies on the same or very similar devices. For example, Bayesian methods have been used to allow a study on the rate of target vessel failure for a new generation coronary stent to borrow strength from studies on previously approved, earlier generation stents. As another example, Bayesian methods have been used to summarize multiple studies of historical controls that are being compared with the investigational device. The typical approach has been to use a Bayesian hierarchical model to account for variation between studies. I will (1) review advantages, potential pitfalls, and useful summaries when using hierarchical models to combine medical device studies, (2) discuss various approaches to using hierarchical models to combine historical control studies, including the derivation of synthetic controls, and (3) compare hierarchical models with other Bayesian methods for combining studies (e.g,. using a Dirichlet process prior).