Adaptive Cohort Size Determination Method for Bayesian Optimal Interval Phase I/II Design to Shorten Clinical Trial Duration.

Abstract

Recently, the strategy for dose optimization in oncology has shifted toward conducting phase II randomized controlled trials with multiple doses. Optimal biologic dose (OBD) selection from phase I trial data to determine candidate doses for phase II trials has been gaining attention. Trials to identify the OBD have a fixed cohort size, which increases the trial duration. We propose a method to increase the cohort size using trial data and shorten the trial duration while maintaining accuracy. We propose a novel adaptive cohort size determination method in which the increase of cohort size is determined using desirability probability on the basis of toxicity and efficacy data. The desirability probability is a measure of how desirable a dose is and thus how close it is to the OBD. However, during the trial, the desirability probability does not need to be calculated. Instead, the cohort size expansion can be determined by a simple table generated in advance from toxicity and efficacy data. An illustrated example is provided and the performance was evaluated in a simulation study with 16 scenarios. In the simulation study, the trial duration was reduced by an average of 20% compared with the conventional design. The percentages of correct OBD selection are almost the same as those with the conventional design. The proposed adaptive cohort size determination method described in this study reduces trial duration while maintaining accuracy.