We propose a new dose-assignment rule for phase I clinical trials in conjunction with a model of toxicity probabilities that uses a binomial likelihood and conjugate beta priors. The proposed dose-assignment rule is based on a partition of the toxicity probability parameter space into high, acceptable, and low toxicity probability intervals. The rule associates each interval with a de-escalate, stay, or escalate dose-assignment action, and takes the action corresponding to the interval with the largest posterior probability. All dose-assignment actions for allocating patients can be pre-determined and tabulated. Therefore, in practice, investigators can easily monitor trials by using a table that contains these dose-assignment actions. We conduct simulation studies to demonstrate that the proposed method performs well compared with several methods in the literature.
Ji, Yuan; Li, Yisheng; and Bekele, Benjamin, "Practical Dose-finding in Phase I Clinical Trials Based on Toxicity Probability Intervals" (March 2006). UT MD Anderson Cancer Center Department of Biostatistics Working Paper Series. Working Paper 24.