Robust Optimal Interval Design for High-Dimensional Dose Finding in Multi-agent Combination Trials

Abstract

In the era of precision medicine, combination therapy is playing a more and more important role in drug development. However, drug combinations often lead to a high-dimensional dose searching space compared to conventional single-agent dose finding, especially when three or more drugs are combined for treatment. To overcome the burden of calibration of multiple design parameters, which often intertwine with each other, we propose a robust optimal interval (ROI) design to locate the maximum tolerated dose (MTD) in phase I clinical trials. The optimal interval is determined by minimizing the probability of incorrect decisions under the Bayesian paradigm. Our method only requires specification of the target toxicity rate, which is the minimal design parameter. Neither does ROI impose any parametric assumption on the underlying distribution of the toxicity curve, nor it needs to calibrate any other design parameters. To tackle high-dimensional drug combinations, we develop a random-walk ROI design to identify the MTD combination in the multi-agent dose space. Both the single- and multi-agent ROI designs enjoy convergence properties with a large sample size. We conduct simulation studies to demonstrate the finite-sample performance of the proposed methods under various scenarios. The proposed ROI designs are simple and easy to implement, while their performances are competitive and robust.