Copula-Based Model for Incorporating Single-Agent Historical Data into Dual-Agent Phase I Cancer Trials

Phase I clinical trials for testing the combination of anticancer drugs (combination trials) generally start after the tolerability of each drug is evaluated in phase I clinical trials for monotherapy (single-agent trials). In either trial, one of the crucial objectives is to identify the maximum tolerated dose (or dose combination). Several methods to incorporate data from single-agent trials (SA historical data) into subsequent combination trials have been proposed. However, the effect of using SA historical data under the presence of heterogeneity in data between single-agent and combination trials has not been discussed sufficiently. In this study, we examined the relationships among the presence and absence of heterogeneity, the amount of SA historical data incorporated, and the performance of dose-finding designs (e.g., the percentages of correct selection of maximum tolerated dose combination) by conducting an extensive simulation study with 100,000 random toxicity scenarios. In this evaluation, we used a dose-finding design based on a copula-based model that could easily leverage SA historical data. Although our simulations showed that incorporating SA historical data under homogeneity between SA historical and the current data is beneficial, we found advantages in incorporating heterogeneous data, provided heterogeneity is not extremely large.