Online Program Home
My Program

Abstract Details

Activity Number: 644 - Statistical Methods for the Co-Development of Drug and Companion Diagnostic in Oncology
Type: Topic Contributed
Date/Time: Thursday, August 1, 2019 : 10:30 AM to 12:20 PM
Sponsor: Section on Medical Devices and Diagnostics
Abstract #304707
Title: Evaluation of Biomarker Threshold Designs in Cancer Therapy
Author(s): Kui Shen* and Xiaowen Tian and Jonathan Siegel
Companies: Bayer U.S. LLC and University of Washington and Bayer HealthCare Pharmaceuticals Inc.
Keywords: Biomarker; adaptive threshold design; bootstrap; permutation test; robust

Biomarkers are an important component in developing effective anti-cancer drugs. Efficient clinical trial designs that can combine biomarker subgroup assessment and efficacy evaluations with controlled FWER would be valuable tools to speeding development. While discrete biomarkers (e.g. genetic) have received the most research, continuous biomarkers are also important to drug development. FWER control is a more difficult problem in a continuous context. This simulation study evaluated the operating characteristics and usability of two proposed adoptive threshold designs for continuous biomarkers, the adaptive threshold design model of Jiang, Freidlin, and Simon (J Natl Cancer Inst 99:1036-43, 2007) (JF&S), and the residual bootstrap method of Gavanji, Chen, and Jiang (Stat Biosci 10:202-216, 2018) (GC&J). Limitations of the JF&S method, noted by GC&J, are incompete evaluation of its small-sample operating characteristics, and dependence of its permutation exchangeability on an assumption of no biomarker effect below the threshold. We assessed the robustness of both methods to relaxation of this assumption by parametrizing biomarker level/efficacy functions as logistic curves.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program