Bias-Corrected Estimation of Treatment Effects in Biomarker-Based Randomized Clinical Trials

• Activity Number: 435
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Kiichiro Toyoizumi* and Shigeyuki Matsui
• Companies: Shionogi & Co., Ltd. and Nagoya University Graduate School of Medicine
• Keywords: Phase III clinical trials ; Predictive biomarkers ; Adaptive analysis plan ; Estimation bias

Abstract

In the development of molecular targeted agents that might only benefit a subpopulation of patients, we also need to develop and validate biomarkers to predict treatment responsiveness for individual patients. In the definitive phase III clinical trials, such a predictive biomarker would be incorporated to evaluate treatment efficacy only for a biomarker-positive subpopulation (i.e., enrichment designs) or for both the overall population and biomarker-based subpopulations, depending on the degree of ones credentials on the biomarker. In the latter, all-comers trials with adaptive analysis plans based on the biomarker, a bias can arise in estimating treatment effects using standard methods without regard to the correlation between the planed multiple tests. In this paper, we evaluate the bias function and provide bias-correction methods when a test of interaction between treatment and biomarker is planned. An application to a real clinical trial will be provided.