Methods of Biomarker and Subgroup Identification for Personalized Medicine

• Activity Number: 223 - Clinical Trials: Recent Statistical Advances for Enabling Personalized Medicine
• Type: Topic Contributed
• Date/Time: Monday, July 31, 2017 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Ilya Lipkovich* and Bohdana Ratitch and Alex Dmitrienko
• Companies: QuintilesIMS and QuintilesIMS and Mediana Inc
• Keywords: Personalized medicine ; Subgroup identification ; Recursive partitioning ; Variable importance

Abstract

This talk provides a relatively high-level description of a broad class of statistical methods dealing with exploratory subgroup analysis, i.e., subgroup search/biomarker discovery methods that can be applied both in early and late-phase clinical trials. Discussion of exploratory subgroup analysis methods begins with a review of common approaches to subgroup identification in the context of personalized medicine and then focuses on the SIDES method (Lipkovich et al. 2011) and its extensions SIDEScreen and Stochastic SIDEScreen (Lipkovich and Dmitrienko 2014, Lipkovich et al. 2017).Stochastic SIDEScreen is an extension of the earlier SIDES methods that introduces randomness in the subgroup generation process borrowing ideas from bagging (bootstrap aggregation) methods to produce a broader collection of subgroups used as the basis for computing Variable Importance indices for biomarker selection. The new approach leads to a more reliable biomarker selection which is especially important for smaller early phase studies when biomarker selection is typically carried out. The methodology will be illustrated with a case study.