Optimal two-stage designs for single arm phase II cancer trials.

In this article, we provide a minimax design for the two-stage single binomial trial with given exact upper bounds for type I and type II errors (operating characteristics). Amongst all two-stage designs exceeding the specified operating characteristics, the proposed design is minimax in the sense that it has the smallest globally maximized expected sample size. Tables are provided for interested users, covering planning differences in response rates of 15 and 20%.