|
Abstract:
|
Precision medicine has emerged from the awareness that many human diseases are intrinsically heterogeneous with respect to their pathogenesis and composition among patient populations. Its application depends on identification of predictive biomarkers which are used as targets to devise treatment strategies that exploit current understanding of the biological mechanisms of the disease. In this paper, we present a novel method to compare the relative effectiveness of biomarker-based strategy when evaluated with biomarker validation design. Our method establishes a Bayesian framework for integrating information based on (1) treatment response surface, (2) biomarker-based treatment allocation rule, and (3) prognostic features balance in study cohorts. We also illustrate how our method can be adjusted to correct cohort bias when randomization to treatment is infeasible. Through simulation study, both reductions in bias, in the absence of truly predictive markers, and MSE, in presence of predictive markers, where evident when compared to competing methods based on generalized linear models. The methodology is also demonstrated through a proteomic study of lower grade glioma.
|
