How Bayesian Hierarchical Models Handle Multiplicity Issues Automatically: a Case Study in a Medical Device Trial

• Activity Number: 187
• Type: Topic Contributed
• Date/Time: Monday, August 7, 2006 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Feng Tang*+ and Lou Sherfesee and Andrew Mugglin
• Companies: Medtronic, Inc. and Medtronic, Inc. and University of Minnesota
• Keywords: Bayesian hierarchical model ; multiplicity ; stratification ; type I error

Abstract

We present how Bayesian hierarchical models handle multiplicity issues automatically using an example in a medical device trial. Initially, a Bayesian design was used to study the efficacy of a therapy for patients with indication type A. However, due to the slow enrollment of this group of patients into the trial and the potential similar benefit of the therapy on another group of patients with indication type B, the decision was made to include type B patients in the current trial. As a result, the initial model and design were revised. Rather than making separate inferences for the two hazard ratios, which requires adjustment of existing decision criteria to control overall simulated type I error, we added a layer of hierarchy to the existing model to allow us to keep the same decision criteria, yet with desirable frequentist properties.