Practical implementation of the continual reassessment method for dose finding cancer trials.

摘要

The objective of dose finding trials is the determination of the maximum tolerated dose (MTD). Algorithm and model based methods have been proposed throughout the years. Although, algorithm based methods are easy to implement, model based methods such as the continual reassessment method (CRM) offer clear advantages in terms of specification of target probabilities of toxicity and the possibility of incorporating data regarding multiple grades and types of toxicities in the determination of the MTD. Concerns have been raised since the dose limiting toxicity (DLT) categorization used in traditional approaches does not differentiate between severity of toxicities and does not account for multiple non-DLTs that may have an aggregate effect. We are proposing ways to make the CRM easier to implement and more accessible to investigators as well as extending the CRM to incorporate the information on multiple toxicities and toxicity grades in the estimation of the MTD. An algorithm is proposed to provide a systematic approach for selecting dose levels used for the CRM in a much less time consuming manner. In addition, a method that incorporates information on toxicity types and grades, and allows for the specification of objectives related to probabilities of toxicity is proposed. Our findings indicate that by including the information on toxicity grades, our method decreases the frequency with which doses above the MTD are recommended.