A Predictive Biomarker Model for Subgroup Selection

• Activity Number: 36
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Yu-Chuan Chen* and James J. Chen
• Companies: FDA/NCTR and FDA/NCTR
• Keywords: Subgroup selection ; Composite score ; Personalized medicine ; Biomarkers ; Changepoint ; Survival analysis

Abstract

Subgroup selection in personalized medicine is to divide patients into several different subgroups where each subgroup corresponds to an optimal treatment. For two subgroups, traditionally multivariate Cox proportional hazards model is fitted and used to calculate the risk score when outcome is survival time. Median is commonly chosen as the cutoff value to separate patients. However, this may lead to a bias if the size between two subgroups is not balanced. Moreover, multivariate Cox proportional hazards model tends to perform poor when too many significant biomarkers are fitted. To overcome these problems, we propose a new subgroup selection algorithm by first developing a composite score using the selected significant biomarkers from the interaction test. Next, instead of using median as the cutoff, we adopt changepoint algorithm to find the cutoff. A simulation study is conducted to compare the performance between our proposed method and the multivariate Cox model. The simulation result shows that our proposed method performs better than the multivariate Cox model. The applications of our proposed method to three public cancer data sets are also conducted for illustration.