A model-based method for selecting an acceptable dose in phase IIa chemoprevention trials

In a cancer chemoprevention trial, it is vitally important to use a dose that will limit intolerable side-effects. Unlike the situation with cancer drug treatment, the side-effects are relatively moderate, only rather low levels will be permitted, and the time to development of intolerability may be long and variable. These characteristics all argue in favor of a fixed-sample size, and a fixed-dose design. Here we give a very general procedure for selecting an acceptable dose for subsequent testing based on such a design. We also show how a simple program can display the characteristics of various designs graphically, permitting a common-sense selection.