Bayesian Adaptive Design in a Dose-Finding Study

• Activity Number: 178
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Zijiang Yang*+
• Companies: Merck
• Keywords: Bayesian Design ; Dose Finding ; Adaptive Design

Abstract

When the primary endpoint is percentage reduction of a PD parameter and multiple candidate doses are to be tested, traditional designs are costly and need potential multiplicity adjustments. Further, patients randomized to high (low) doses are subject to toxicity (futility) concerns. A simple Bayesian adaptive design is proposed to find optimal doses that satisfy stringent and complicated criteria and aim to minimize sample size without sacrificing the integrity of the trial. This design comes with minimally informative priors and no pre-set dose response model to reduce potential bias. Simulation results show the design seeks to find all qualifying doses with minimal number of subjects.