Clinical trial design for rare cancer

Abstract

How to manage clinical trials for rare cancer including pediatric cancer is a burning issue globally involving Japanese regulatory authorities. The conditions that can loosen limitations in study designs are as follows: (1) allowance for a one-arm trial design, (2) setting a surrogate endpoint (e.g., response rate), and (3) increasing type-I or type-II error probabilities. Also, several efficient study designs are suggested, such as study designs based on genomic information (e.g., umbrella design and basket design), N-of-1 design, adaptive designs, and Bayesian designs. The advantages of the Bayesian approach are as follows: (1) the interpretation for the probability is intuitive, (2) the flexibility that enables the probability update during the study is suitable for adaptive design, and (3) the variety of prior beliefs, which sometimes have clinically important meanings, can be explicitly integrated as a prior distribution. A trade-off exists between scientific validity and efficiency in trial designs. Therefore, the pros and cons of each design should be thoroughly discussed at the planning stage. Effective collaborations between clinicians and statisticians are necessary to improve the quality of trial design.