Modified continual reassessment method in phase I clinical trials

Phase I clinical trials represent the first application of a new drug or treatment to human subjects. The primary goal of Phase I clinical trial is to establish the maximum tolerated dose (MTD) for further phases of trials. O’Quigley, Pepe and Fisher [1990] proposed a model-based clinical trial design,calledContinualReassessmentMethod(CRM),which later became the main stream in theresearchfieldandapplications. Its various modified versions have been extensively studied and shown to improve the performance in finding MTD adaptively in the case of dichotomous toxicity responses in phase I clinical trials.

In this research study, we extend the concept of the overall MTD by Yang [2011] which makes the CRM study of both discrete and continuous response models under a unified framework. Connections between the relation of dose level and the toxicity measure and the relation of the toxicity measure and the true toxicity level are made through one regression model and the target toxicity probability function. The target toxicity probability function may be established by practitioners as well as the drug developers. Various possible toxicity probability curve functions are investigated in this research. Latent variables are used in the analysis for a toxicity response measure that is either dichotomous or polychotomous. The extensive simulation for the unified framework has been done for various response variables, target toxicity probability curves, modeling scenarios and dose range selections. Performance of the proposed framework are investigated.

Furthermore, to avoid more allocations of the experimental dosage that might cause severer toxicity reaction to the MTD level for the experimental subjects in the phase I study, a modification in the methodology using the idea similar to the Escalation With Overdose Control (EWOC) method is proposed in this research and comparisons between EWOC and non-EWOC are made by using simulation.