ABSTRACT
This chapter begins with a discussion of algorithm-based designs and their advantages and disadvantages. It focuses on model-based designs, specifically the continual reassessment method and dose escalation with overdose control. The chapter deals with a discussion of the hybrid designs such as the modified toxicity probability interval. Assuming that the toxic response increases as the dose level increases, the continual reassessment method (CRM) uses a parametric model to describe the dose–toxicity relationship. Essentially a Bayesian approach, the CRM updates this relationship based on the posterior distribution of the model parameters after each patient response is obtained. The key feature of the CRM is that each patient is entered in the trial at the dose closest to the maximum tolerated dose. Phase I dose-finding clinical trials are intended to find the maximum tolerated dose of a new drug for a specific mode of administration. Phase I dose-finding trials may be carried out with healthy volunteers if the drugs are relatively non-toxic.
