Optimal Design Theory in Early-Phase Dose-Finding Problems

Phase I clinical trials concerns the estimation of the MTD (maximum tolerated dose), the dose level corresponding to the target toxicity rate. A great deal of methods have been proposed to address the MTD estimation problem, among which the CRM (continual reassessment method) stands out due to its simplicity and outstanding performance. We extend the classic CRM by incorporating the idea of optimal design theory. We denote this new approach the OD-CRM, which indicates that this strategy is developed within the CRM framework, and coupled with the optimal design theory. Then we move on to a more practical problem encountered in the oncology clinical studies, the late-onset toxicities. We adopt the weighting mechanism discussed in Cheung and Chappell (2000) which essentially assigns each toxicity response to a weight that depends on the patient's enrollment time and the observed data. We also offer a general dose- finding algorithm based on the OWEA (optimal weight exchange algorithm, Yang, Biedermann, and Tang, 2013), to explore the performance of the OD-CRM under a broader clinical trial setup.