Alternative designs of phase II trials considering response and toxicity

Phase II clinical trials in oncology are used to initially evaluate the therapeutic efficacy of a new treatment regimen. Simon's two-stage design is commonly used for such trials. However, he only focused on the "response rate", the proportion of patients experiencing tumor regression. In clinical practice, it is preferred of a sequential design to monitor antitumor activity as well as toxicity. Conaway and Petroni proposed a method for designing phase II trials on the basis of both treatment efficacy and safety, which imply an equal weighing of response and toxicity. In this paper, we developed an alternative test to cope with the trade-off between safety and efficacy. The main idea is to control for the marginal type I errors of response rate and toxicity rate separately. We provide guides on searching the stopping and rejecting regions and determination of sample size. The proposed method has advantage over other designs, including those of Conaway and Petroni's and Bryant and Day's, that it can definitely control one type I error of the interests such as treatment antitumor activity or safety and is robust against the real association parameter. Furthermore, it is conceptive intuitive, very simple to implement, and also feasible for the requirement of small sample size in a phase II trial.