Analyses of drug combinations using missing data shortens trial periods in phase I/II oncology trials

• Main Authors: Shinjo Yada, Chikuma Hamada
• Format: Article
• Language: English
• Published: Elsevier 2017-09-01
• Series: Contemporary Clinical Trials Communications
• Online Access: http://www.sciencedirect.com/science/article/pii/S2451865416301314

In previous phase I/II oncology trials for drug combinations, a number of methods have been studied to determine the dose combination for the next cohort. However, there is a risk that trial durations will be unfeasibly long if methods for evaluating safety and efficacy are based on the best overall response and toxicity during trial design. In this study, we propose an approach to shorten the duration of drug trials in oncology. In this method, the dose combination to be allocated to the next cohort is decided before all data for patients in the current cohort is known and best overall response is determined. The efficacy of drug combinations in patients for whom the best overall response has not been determined is treated as missing data. The missing data mechanism is modeled by nonparametric prior processes. The probabilities of efficacy and toxicity are estimated after applying data augmentation to missing data, and the dose combination to be allocated to the next cohort is decided using these probabilities. Simulation studies from the present study show that this proposed approach would shorten trial durations without the low-performing of the trial design in comparison to existing approaches. Shortening trial durations would enable patients with the targeted disease to receive effective therapy at an earlier stage. This also enables clinical trial sponsors to use fewer patients in drug trials, which would lead to a reduction in the costs associated with clinical development. Keywords: Missing data, Data augmentation algorithm, Gamma process, Dose combinations, Seamless phase I/II trials