Adaptive Enrichment Designs with Population Selection Based on an Ordinal Biomarker

• Activity Number: 406
• Type: Topic Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Michael Rosenblum and Yu Du* and Aaron Fisher and Tianchen Qian
• Companies: Johns Hopkins Bloomberg School of Public Health and The Johns Hopkins University and The Johns Hopkins University and
• Keywords: adaptive design ; personalized medicine ; treatment heterogeneity

Abstract

We propose a general method for subpopulation selection in an adaptive enrichment design, where the subpopulation is defined by an ordinal valued biomarker measured at baseline. We optimize the design and analysis procedure in a decision theoretic framework, while also guaranteeing strong control of the familywise Type I error rate. At each interim analysis, we use semiparametric, locally efficient estimators. This increases the information at each interim analysis, leading to better decisions and more powerful testing procedures. We illustrate our designs in the context of a trial of a surgical intervention for stroke.