A Case Study for Radiation Therapy Dose Finding Utilizing Bayesian Sequential Trial Design

Abstract

Dose escalation trials for identifying the maximum tolerable dose (MTD) is commonly considered in phase 1 cancer clinical research. For this purpose, an algorithm-based design such as a standard escalation design with traditional escalation rule (TER) and a model-based design such as the method of continued reassessment method (CRM) under a well-established dose toxicity model are commonly employed. In practice, relative merits and limitations of these two different types of designs are not fully understood. Besides, most dose escalation studies do not provide scientific justification for sample size and design selection. In this article, the validity and efficiency of these two different types of study designs are evaluated based on the criteria of the number of subjects expected, the number of DLT expected, the probability of correctly achieving the MTD, and the probability of overdosing. A case study regarding a radiation therapy for treatment of certain solid tumors is discussed to illustrate the criteria for design selection.

Keywords: Algorithm-based design; Model-based design; “3+3” TER design; CRM design; Bayesian sequential design