DODII: Bayesian Dose Optimization Design for Randomized Phase II Trials

The traditional MTD-based dose selection paradigm commonly used for cytotoxic chemotherapies might not be optimal for targeted therapies because a higher dose does not necessarily result in improved anti-cancer activity. With the initiation of "Project Optimus" at the FDA, a randomized dose optimization study at the early stage of drug development has become necessary for oncology drug development. We propose a Bayesian Dose Optimization Design for Randomized Phase II trials (DODII) that integrates Bayesian continuous monitoring and Bayesian pick-the-winner approach in a randomized design, where efficacy and toxicity endpoints are jointly used to inform the dose selection. The adaptive feature of the DODII design will terminate a suboptimal dose at an interim analysis under two circumstances: (a) when efficacy or safety data is unpromising, (b) when the other dose option is clearly better. The critical values of the decision rule for the DODII design can be enumerated prior to the initiation of the clinical study, which makes it practically easy to implement. Through simulation studies, we showed that the DODII design has favorable operating characteristics with controlled selection error, higher power to select the optimal dose option, and reduced total sample size as compared to the other commonly used methods.