Investigating dose finding methods and different priors in Bayesian continual reassessment method

Abstract

A well-designed and properly analyzed clinical trial is a powerful tool for the development of new drugs. Clinical trials are studies that explore whether a treatment, drug or device is safe and effective for humans. These studies can show which drug development method works best for certain diseases. The first step in drug discovery (phase I) is very important to determine maximum tolerated dose (MTD). In the first part of the study, the classical 3+3 design, Continual Reassessment Method (CRM), and Bayesian Continual Reassessment Method (B-CRM) are compared in terms of selection probability of MTD and the number of treated patients. Among these designs, the B-CRM produced better results than the 3+3 and the CRM. In the second part of the study, we considered different model structures and priors in the B-CRM design. We considered three model structures; power model, hyperbolic tangent model, logit model and three different prior distributions; gamma, uniform and lognormal prior respectively. It was found that if power or hyperbolic tangent model structure and uniform prior were selected, the MTD selection rates were the highest in B-CRM.