A Bayesian Adaptive Dropping-Arm Design Using Multiple Interim Evaluations for Decisionmaking

• Activity Number: 130
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Yili L. Pritchett*+ and Shihua Wen
• Companies: Abbott Laboratories and Abbott Laboratories
• Keywords: Adaptive dropping-arm design ; Bayesian predictive probability ; Futility criteria

Abstract

Adaptive dropping-arm design allows studies to initially evaluate multiple doses with equal sample size, and then drop "losers" from further enrollment by applying pre-specified criteria to accumulated interim data in the middle course of the study. In most dropping-arm designs, the decision is based on the results of a single interim evaluation, which requires the right choice of interim timing and consistence of trial data to make the right decision. While easy to implement, the decision made at interim in such design is prone to mistake due to possible errors included in interim data or heterogeneity of data between those collected at early and late stage of the study. In this presentation, we introduce a dropping-arm design where futility criteria are made based on the pattern of consecutively calculated Bayesian predictive probabilities at multiple interim evaluations. The adaptive algorithm and decision rules will be described. Simulation details and the design operating characteristics will also be presented.