Biostatistics Seminar guest speaker, Ying Huang, Fred Hutchinson Cancer Research Center, will present, "Inferential Procedures for the Use of Biomarkers in Treatment Selection".
Biomarkers associated with treatment-effect heterogeneity have the potential to make treatment recommendations more personalized to achieve the best clinical outcomes. While interest in the development of statistical methodology for estimating optimal treatment-selection biomarker panels has grown in recent years, the corresponding developments in inference have been inadequate. In this talk we will present two recent developments on inferential procedures for the use of biomarkers in treatment selection, motivated by data arising from HIV-1 vaccine trials. The first problem focuses on the personalized use of a treatment-selection biomarker panel developed using either a risk modeling approach or a policy-based approach. We propose inferential procedures on a new biomarker-panel’s covariate-specific incremental value, for the purpose of identifying subpopulations that will benefit most from the marker panel. The second problem focuses on the use of threshold regression models to identify subgroups with heterogeneous treatment effect. We propose procedures for testing the existence of biomarker combinations that naturally separate a population into such subgroups.