The current design of oncology phase I clinical trials: progressing from algorithms to statistical models.

In this article we examine how phase I trial designs for anti-cancer agents have developed over the past few decades. We review the use of algorithmic, relatively non- statistical designs, and then describe nonparametric and parametric statistically-driven designs that have better operating characteristics than algorithmic designs. We then follow with a description of how the original parametric design, known as the continual reassessment method (CRM), has been generalized and expanded to create designs for many complex settings that occur in early-phase oncology trials. We conclude with a discussion of outstanding issues and controversies that continue to exist with phase I trial designs.