Reducing the patients-at-risk (PaR) in a response-adaptive trial: A numerical study

Abstract

This article investigates the dichotomy between higher statistical power and higher allocation to better treatment in an ethical-optimal response-adaptive design. Although many response-adaptive designs in the literature promise higher allocation to the superior treatment, this is not always guaranteed due to the variability of the designs. A new criterion for evaluating response-adaptive designs, motivated by the value-at-risk measure, is proposed to address this problem. We also provide an illustration of applying this criterion in a real clinical trial.

Keywords

• Clinical trial
• Ethics
• Optimality
• Response adaptive design
• Bandit algorithm