Abstract
In this article, the problem of dose finding in Phase I clinical trials for a heterogeneous patient population is considered. A new interval-based nonparametric design is proposed that stratifies the patients into multiple groups of varying sizes according to their susceptibility to toxicity and finds the maximum tolerable dose (MTD) in each group. The proposed design allows for sharing of patient information between groups, thus increasing the efficiency in identifying the group MTDs correctly, especially in patient groups with low prevalence. The convergence results for the proposed algorithm are presented. Various simulation studies are performed to illustrate the new methodology.
Keywords:
Acknowledgments
The authors are indebted to two anonymous referees and Professor P. Vellaisamy for their helpful comments.