On Bayesian Design and Analysis for Trials with Multiple Primary Endpoints

• Activity Number: 298
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 8:30 AM to 10:30 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Kimberly Crimin*+ and Robb J Muirhead
• Companies: Novartis and
• Keywords: Bayesian ; multiple endpoints ; sample size

Abstract

Many diseases are characterized by multiple co-primary endpoints and regulatory agencies require that efficacy be demonstrated on all of the endpoints for the drug to be efficacious. We present a Bayesian approach which incorporates information from a study to help size a new study which will use the same multiple co-primary endpoints. The approach is based on the joint posterior distribution of all the co-primary endpoints. The Bayesian approach incorporates correlations among the co-primary endpoints through the sample covariance matrix. We focus on conjugate prior distributions that are appropriate for multivariate normal models. We discuss considerations that should be taken into account in sample sizing problems and highlight the differences between Bayesian and frequentist approaches. We present results from a simulation study and illustrate the approach with an example.