Identifying Predictive Biomarkers in Oncology Trials: A Practical Example from a Randomized Phase II Study in Neo-Adjuvant Breast Cancer (BC)

• Activity Number: 112
• Type: Topic Contributed
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Pralay Mukhopadhyay*+ and Guan Xing and Li-An Xu and David Liu and Christine Horak
• Companies: Bristol-Myers Squibb and Bristol-Myers Squibb and Bristol-Myers Squibb and Bristol-Myers Squibb and Bristol-Myers Squibb
• Keywords: biomarkers ; breast cancer ; cutoff ; enriched population

Abstract

Identifying biomarkers that can differentially predict for response between experimental and comparator arms can be a key to success in oncology drug development. Such markers, if validated in Phase II settings could result in smaller, faster and more rationally deigned Phase III trials, conducted in selected patient populations. Here we discuss the design and analysis of a biomarker guided randomized phase II trial that was utilized to select potential biomarkers. The trial randomized 295 patients on a 1:1 basis to either paclitaxel or ixabepilone-based neoadjuvant therapy. Tissues samples were available for approximately 80% of the patients. One of the key objectives of the trial was to estimate an optimal cutoff for a pre-defined biomarker enriched population and estimate the response rates in that population. Logistic regression modeling was used to evaluate any interaction between treatment and biomarker. A five-fold cross validation method was used to estimate the optimal cutoff. The study failed to demonstrate presence of a marker that differentiates between treatment arms. Reasons why the trial failed to confirm the initial findings are discussed.