|
Abstract:
|
In recent years, there has been a growing interest in individualized medicine utilizing biomarkers associated with treatment heterogeneity. A related but different question is to identify the subgroup of individuals for which measuring of a new biomarker (or biomarker panel) is useful for treatment recommendation, given that the biomarker's value can differ across individuals. Identifying subpopulations where the biomarker is useful can help guide the design of biomarker validation studies among the targeted subpopulation and spare biomarker measurement in individuals for whom the biomarker's treatment-selection impact is minimum. To achieve this goal, we propose to use the covariate-speci?c expected bene?t to quantify the treatment-selection performance of the biomarker conditional on covariates. We developed novel estimators of subgroups among which the covariate-specific expected benefits exceed the desired level.
|
