A0289
Title: Identifying optimal biomarker combinations for treatment selection using data from randomized controlled trials
Authors: Ying Huang - Fred Hutchinson Cancer Research Center (United States) [presenting]
Abstract: Biomarkers associated with treatment-effect heterogeneity can be used to make treatment recommendations that optimize individual clinical outcomes. Statistical methods are needed to generate marker-based treatment-selection rules that can effectively reduce the population burden due to disease and treatment. Compared to the standard approach of risk modeling to combine markers, a more robust approach is to directly minimize an unbiased estimate of total disease and treatment burden among a pre-specified class of rules. We frame this into a general problem of minimizing a weighted sum of 0-1 loss and propose a penalized minimization method based on the difference of convex function algorithm, using data from randomized trials. The corresponding estimator has a kernel property that allows flexible modeling of linear and nonlinear combinations of markers. We further expand the method with a L1-penalty to allow for feature selection and develop an algorithm based on the coordinate descent method. We compare the proposed methods with existing methods for optimizing treatment regimens such as the logistic regression, the weighted logistic regression, and the weighted support vector machine. Performances of different weight functions are also investigated. We illustrate the application of the proposed methods in host-genetics data from an HIV vaccine trial.
