Cost-effective design strategies for development of breakthrough personalized medicines
View Presentation View Presentation
*Cong Chen, Merck & Co. 

Keywords: bayesian, decision analysis, false discovery rate, pathway

In this presentation, we will discuss two issues associated with the development of breakthrough personalized medicines in the oncology therapeutic area. First, how to design Phase II POC trials efficiently given that the number of possible POC hypotheses often far exceeds available public or private resources? We explicitly maximize the return on socioeconomic investment via risk adjusted benefit-cost ratio, and arrive at a cost-effective design that has a ~5% type I error rate (one-sided) and ~80% power for detecting an effect size ~50% greater than under conventional design. Second, how to get a personalized medicine approved in as many indications and as fast as possible? We propose a basket design for Phase III that enrolls patients by a common biomarker across multiple tumor types into a single study after POC is demonstrated. Each tumor has adequate power for detecting a clinically meaningful treatment effect in an accepted surrogate endpoint (e.g., PFS) at 2.5% (one-sided). Unlike a conventional design, OS for an individual tumor in the proposed design is under powered which easily cuts the sample size by 60-80%. The drug may be approved in all the tumors investigated if PFS is positive in each tumor, and OS is positive in the pooled population and generally consistent across the tumors. The drug may have a narrower label or be rejected otherwise.