Designing a Confirmatory Trial with a Continuous Biomarker Endpoint and an Undetermined Threshold Tested at Both the Interim and Final

• Activity Number: 406
• Type: Topic Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Adarsh Joshi* and Jenny Zhang and Nusrat Rabbee and Liang Fang
• Companies: Gilead Sciences and Gilead Sciences and UC Berkeley and Gilead Sciences
• Keywords: Personalized medicine ; Biomarker cutoff selection ; Subgroup identification ; Clinical trial simulation ; Group sequential design

Abstract

With targeted therapies, it is often hypothesized that the drug effect is specific to the subpopulation in which the target pathway is activated. We consider the problem of designing a confirmatory trial when the target biology is well understood but limited data is available to pre-define a subpopulation based on continuous biomarker values. The study design is further complicated if interim evaluations of the biomarker based subpopulations are also being considered. Based on extensive simulations, we will provide a comparison of several strategies for identifying subpopulations including modifications of the "explore and confirm" strategy proposed by Friedlin et al. (2005) and the approach by Holmgren as described above. With multiple evaluations of the ITT and biomarker subpopulation(s), appropriate Type I error adjustment for multiplicity is an important consideration as methods that ignore the positive correlations between the various tests (e.g., Bonferroni adjustment) will be very conservative in this setting. Hence, we will also discuss alpha allocation rules similar to those proposed by Holmgren for all the strategies compared.