Phase I (or phase II) dose-ranging clinical trials: proposal of a two-stage Bayesian design.

We propose a new design for phase I (or phase II) dose-ranging clinical trials aiming at determining a dose of an experimental treatment to satisfy safety (respectively efficacy) requirements, at treating a sufficiently large number of patients to estimate the toxicity (respectively failure) probability of the dose level with a given reliability, and at stopping the trial early if it is likely that no dose is safe (respectively efficacious). A two-stage design was derived from the Continual Reassessment Method (CRM), with implementation of Bayesian criteria to generate stopping rules. A simulation study was conducted to compare the operating characteristics of the proposed two-stage design to those reached by the traditional CRM. Finally, two applications to real data sets are provided.