Statistical requirements of phase I studies.

Methodological aspects of planning and evaluating phase I studies in oncology--the link between preclinical research and approval of clinical efficacy--include the human starting dose, maximum tolerable dose and dose escalation schemes. Statistical requirements of phase I studies are presented here, emphasizing the urgency of their application in current practice. For the maximum tolerable dose a distinction will be made between an individual- and a population-based approach, which is crucial for a correct definition and statistical parameter estimation. Weaknesses of the dose escalation scheme according to the modified Fibonacci scheme are shown and contrasted with the recently proposed pharmacokinetically guided dose escalation. Comprehensive phase I/II information processing is recommended for validating current practice.