A Bayesian adaptive dose-finding algorithm for balancing individual- and population-level ethics in Phase I clinical trials. Issue 4 (1st October 2016)

• Record Type: Journal Article
• Authors: Kim, Steven B.; Gillen, Daniel L.
• Is Part Of: Sequential analysis. Volume 35:Issue 4(2016)
• Page Start: 423
• Page End: 439
• Publication Date: 2016-10-01
• Subjects: Adaptive design -- Individual- and population-level ethics -- Phase I clinical trials; 62C10 -- 62F15 -- 62K05 -- 62L05; Sequential analysis -- Periodicals; 519.54
• DOI: 10.1080/07474946.2016.1238250
• Languages: English
• ISSNs: 0747-4946

Abstract

ABSTRACT: Multiple Bayesian adaptive designs have been proposed for Phase I clinical trials since the continual reassessment method (CRM) was proposed by O'Quigley et al. (1990). Focused on dose-finding in cancer studies, the CRM seeks to allocate new patients to an estimated maximum tolerable dose (MTD). Later, Whitehead and Brunier (1995 ) applied Bayesian decision theory to maximize statistical information for the MTD when allocating new patients. The two allocation rules reflect conflicting perspectives. The CRM emphasizes individual-level ethics, whereas the method of Whitehead and Brunier (1995 ) emphasizes population-level ethics. In the design of a Phase I clinical trial to investigate hyperthermic intraperitoneal chemotherapy (HIPEC) we sought to compromise the two perspectives. To this end, we propose a novel dose allocation design referred to as the balanced information gain method. We first decompose the loss function used by Whitehead and Brunier and then modify it with a tuning parameter that allows a trialist to differentially weigh individual- and population-level ethics based on their particular clinical setting. Simulation studies show that the proposed design provides a reasonable compromise between the distribution of the estimated MTD and the distribution of the number of observed adverse events per trial when compared to the two existing methods.