Martingale-Based Estimators Following Group Sequential/Adaptive Designs

• Activity Number: 516
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Venkata Goteti and Rongqi Chen and Navdeep Boparai
• Companies: Merck & Co., Inc. and University of California at Davis and Merck & Co., Inc.
• Keywords: Group Sequential Design ; Adaptive Design ; Martingale ; Clinical Trial

Abstract

In the context of multi-stage (group sequential or adaptive) designs, the naive estimators,which equal the difference of the sample means between the treatment groups up to the stopping stage, are biased. For group sequential designs, two common approaches were used to adjust the bias explicitly or get unbiased estimators from sufficient statistics.However, these methods either apply to a restricted set of group sequential designs or are inapplicable to adaptive designs. We now propose a brand new series of estimators which are solutions of equations based on Martingale processes; meanwhile, another series of estimators which combine the advantages of naive estimators and martingale-based estimators are also proposed. Our approaches can be easily extended to the adaptive designs and confidence intervals can also be obtained. These approaches show high accuracy and flexibility in both simulations and a case study from pharmaceutical industry.