Bayesian Adaptive Randomization Design Using Posterior Predictive Probability

• Activity Number: 350
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Xuemin Gu*+ and Brenda Gaydos
• Companies: Eli Lilly and Company and Eli Lilly and Company
• Keywords: Adaptive Design ; Adaptive Randomization ; Predictive Probability

Abstract

Outcome-based adaptive randomization has been increasingly used as an adaptation element in Bayesian adaptive design for clinical trials, especially in the exploratory phase of drug development, where the most common randomization schemes involve using patient allocation ratios proportional to the probability of one treatment being better than all the others or posterior point estimate of treatment efficacy after proper scaling. However during the planning stage of a clinical trial, these methods, intuitively appealing, all show great variations of allocation ratios in the early stage of the trial. Various remedies for the early variation are ad hoc. In the present study, we show that the use of predictive probability approach can solve the problem satisfactorily. Additionally, other benefits of predictive probability approach are demonstrated. In particular, for example one set of decision rule will be used for both early stopping and final decision, which makes the decision process more consistent.