2014 年 34 巻 2 号 p. 81-97
We discuss toxicity-based dose-finding methods for two-agent combinations in phase I oncology trials. We focus on the four dose-finding methods recently developed, the methods based on 1) a copula-type model, 2) a hierarchical Bayesian model, 3) continual reassessment method with partial ordering (POCRM), and 4) a shrinkage logistic model. We summarize the characteristic of each method and compare the performance among them through simulation studies. In the simulation studies, we examined recommendation rates (RDs) for both true maximum tolerated dose combinations (MTDCs) and unacceptable toxicity dose combinations (UTDCs) under 12 scenarios with 3 × 3, 4 × 4, 2 × 4, and 3 × 5 dose combination matrices. Simulation studies demonstrated that the POCRM and method using the shrinkage logistic model outperformed the other two methods in terms of recommending true MTDCs. The RDs for the UTDCs in the method using the shrinkage logistic model were lower than the other three methods.