Bayesian Hierarchical Modeling and Dose Finding in Alzheimer's Disease

• Activity Number: 484
• Type: Topic Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Guosheng Yin*+
• Companies: University of Hong Kong
• Keywords: Dose finding ; Efficacy ; Longitudinal data ; Posterior predictive ; Toxicity

Abstract

We consider designing phase II clinical trials for Alzheimer's disease, which is typically the proof-of-concept stage for drug development. In such a dose ranging study, multiple doses of the new agent are investigated to identify the most appropriate dose or multiple doses that may have disease modification or symptomatic effects, which would be moved forward to phase III clinical trials for confirmative testing. For this purpose, we develop Bayesian hierarchical modeling for longitudinal measurements, and incorporate the posterior predictive outcome of disease deterioration into decision making. At the interim analysis, any dose arm that shows futility would be dropped earlier, so that the rest of patients in the trial would be allocated to the remaining active arms. We conduct extensive simulation studies to show the satisfactory performance of the proposed Bayesian hierarchical modeling and dose finding approach. Our method can pinpoint the right doses in terms of both tolerability and therapeutic effects.