Abstract

As the first in human experiment, a phase I clinical trial is complex in ethics and methodology. To address individual ethics in a phase I dose finding clinical trial, we need to minimize the overall risk of toxicity and treat efficiently as many patients in the trial as possible. The collective ethics also requires us to reliably identify the maximum tolerated dose for subsequent trials. A good statistical design should be flexible and transparent, simple to apply in practice, and optimal in its performance. We introduce and discuss the use of some Bayesian designs that strike a good balance between the collective ethics and individual ethics.