Abstract
Escalation with overdose control (EWOC) is a Bayesian adaptive design for selecting dose levels in cancer Phase I clinical trials while controlling the posterior probability of exceeding the maximum tolerated dose (MTD). EWOC has been used by clinicians to design many cancer Phase I clinical trials, see e.g. [1-4]. However, this design treats the toxicity response as a binary indicator of dose limiting toxicity (DLT) and does not account for the number and specific grades of toxicities experienced by patients during the trial. Chen et al. (2010) proposed a novel toxicity score system to fully utilize all toxicity information using a normalized equivalent toxicity score (NETS). In this paper, we propose to incorporate NETS into EWOC using a quasi-Bernoulli likelihood approach to design cancer Phase I clinical trials. We call the design escalation with overdose control using normalized equivalent toxicity score (EWOC-NETS). Simulation results show that this design has good operating characteristics and improves the accuracy of MTD, trial efficiency, therapeutic effect, and overdose control relative to EWOC which is used as a representative of designs treating toxicity response as a binary indicator of DLT. We illustrate the performance of this design using real trial data in identifying the Phase II dose.