Bayesian-Augmented Control Methods for Efficiently Incorporating Historical Information in Clinical Trials

• Activity Number: 500
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Carl Dicasoli*+ and Michael Kunz and Daniel Haverstock
• Companies: Bayer Healthcare Pharmaceuticals and Bayer Healthcare Pharmaceuticals and Bayer Healthcare Pharmaceuticals
• Keywords: Bayesian ; hierarchical ; noninferiority ; Type I error ; power ; covariate adjustment

Abstract

When planning a clinical trial, there is often historical clinical data available. Recently Viele et al.(2013) have presented approaches that incorporate this historical clinical data into an analysis procedure. We focus on the idea of dynamic borrowing in the framework of various hierarchical modeling strategies and derive the Type I error, power, and DIC. These strategies may include estimating parameters for each trial separately versus pooling, weighting the prior distribution corresponding to each historical study based on the sample size, and incorporating historical borrowing on the control arm by a separate random effect parameter. As a further refinement we present how in the setting of a non-inferiority trial a covariate adjustment approach can be implemented to recalibrate the non-inferiority margin based on the difference between active control and placebo.