Incorporating biomarkers into adaptive clinical trials.

Abstract

e13009 Background: Multi-arm randomized controlled trials with Bayesian interim analyses improve efficiencies over conventional two arm trials. We expand on this to include biomarkers and a set of heterogeneous drugs that are linked to distinct treatment-biomarker interaction hypotheses. While biomarker hypotheses can be incorporated into clinical trials in a myriad of ways, we developed a novel methodology to incorporate prior information to gain the efficiency associated with biomarker-only trials while preserving the valuable biomarker information generated by all patient trials. Experimental arms are linked to biomarker-defined subgroups within the enrolled population, with treatment/biomarker links estimated with various degrees of uncertainty by previous studies including preclinical experiments. Methods: We propose statistical designs and analyses of multi arm trials using a priori information on treatment-biomarker interactions formalized through Bayesian modeling. A priori information is represen...