Biomarker Thresholding to Identify Subgroup of Patients with Treatment Benefit

• Activity Number: 163
• Type: Topic Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yi Liu*+ and Jason Hsu
• Companies: Millennium : The Takeda Oncology Company and Ohio State University
• Keywords: biomarker ; threshold ; subgroup

Abstract

In routine medical practice, continuous biomarker values are often dichotomized (e.g. IHC) to represent different patient populations. However for biomarkers developed with more recent technologies, cut off values may not have been well established. This presentation is to show how to choose the threshold for a biomarker of interest to find the best subgroup of patients that benefit from the treatment. Assuming higher biomarker value translates to more efficacy, the subgroup is represented by patients with biomarker value higher than the threshold. Treatment effect is tested in multiple subgroups by varying the threshold. The choice of potential threshold values takes into consideration: 1) market potential based on the size of the subgroup; 2) treatment effect size of the subgroup; 3) multiplicity adjustment for testing multiple subgroups. We show power can be maximized using multiple testing procedures based on the exact correlation structure between the test statistics among different subgroups.