Abstract
Targeted cancer therapies aim to effectively treat patients with specific biomarker profiles. Nevertheless, these therapies may not always precisely hit their intended targets, leading to uncertainty about the specific subset of patients who will benefit. To address this uncertainty, the identification of sensitive patient subsets in clinical trials becomes crucial. Our proposed phase IIB/III clinical trial design seeks to pinpoint a biomarker signature with precision, ensuring the accurate identification of patients who will respond to a specific treatment. This approach allows for the selective enrollment of sensitive patients to maximize benefits for trial participants. We incorporate Bayesian methodology to facilitate response-adaptive randomization, enhancing the likelihood that each participant receives his/her optimal treatment. Furthermore, our design uses inverse-probability-of-treatment-weighted analysis to avoid selection bias and control for the type I error rate. The evaluation of this trial design is based on four criteria: the statistical power, response rate of all patients participating in the current trial, their individual loss, and probabilities of receiving their optimal treatment for both current trial participants and future patients. Simulations demonstrate the proposed design's potential for maximizing trial participants' benefits with little sacrifice on statistical power. Its key advantages include an improved overall response rate within the trial and a higher percentage of patients receiving the optimal treatment.