Subgroup Identification for Tailored Therapies: Methods and Consistent Evaluation

Abstract

In contrast to the “one-size-fits-all” approach of traditional drug development, the paradigm of tailored therapeutics seeks to identify subjects with an enhanced treatment effect. In this chapter, we describe a statistical approach (TSDT) to subgroup identification that utilizes ensemble trees and resampling. For each potential subgroup identified, TSDT produces a multiplicity-adjusted strength of the subgroup finding as well as a bias-adjusted estimate of the treatment effect in the identified subgroup, both of which are important for decision-making in the development of tailored therapeutics. We describe a careful examination of simulation studies in a number of related publications, in order to determine the ideal framework to compare subgroup identification methods. A simulation study is performed to evaluate the performance of TSDT. The method has been implemented in a publicly available R package.