Abstract:
|
It is often desirable and sometimes crucial to identify, based on data collected from randomized clinical trials, subgroup of patients defined by biomarkers who are expected to have an enhanced response to an investigational treatment. This effort has been made more feasible by recent advances in statistical methods for effective identification of such patient subgroups. In this talk, I will review some of these methods; furthermore, I will present approaches to evaluate subgroup identification methods and optimize the analysis in any given application, thereby improving the various key aspects of subgroup identification, namely maximized power to identify relevant subgroups and improved quality of the subgroups identified, while still adequately controlling the type I error rate. Insights generated from applying these approaches to real settings will be shared.
|