Randomized Phase II Cancer Clinical Trials to Validate Predictive Biomarkers

Abstract

Objectives: The design of cancer clinical trials incorporating biomarkers depends on various factors, including the trial phase, the type of biomarker and whether its role has been validated. This article aims to present a method for designing and analyzing phase II cancer clinical trials that validate predictive biomarkers. Methods: We propose a randomized trial design where patients are allocated between a targeted therapy and a non-targeted therapy stratified by biomarker status. Tumor response is used as the primary endpoint to validate the biomarker through interaction testing between treatment and biomarker positivity. Additionally we propose a sample size calculation method for this design, considering two types of interaction: one based on logit-transformed response rates and the other on raw response rates. Results: The proposed sample size method is applied to the design of a real randomized phase II trial. Extensive simulations are conducted to evaluate the performance of the test statistic and the sample size method under different scenarios. Conclusions: Our method provides a practical approach to validating predictive biomarkers in phase II cancer trials. The simulations demonstrate robust performance for both interaction models, offering guidance for the sample size selection and analysis strategy in biomarker-stratified trials.

Keywords

• interaction
• logistic regression
• sample size calculation
• stratified randomization trial