Assessment of Biomarker Prediction Accuracy Under Marker-Dependent Sampling

• Activity Number: 410
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Xiaofei Wang and Junling Ma*+ and Stephen George
• Companies: Duke University Medical Center and Shanghai University of Finance and Economics and Duke University Medical Center
• Keywords: AUC ; Biased sampling ; Biomarker evaluation ; partial AUC ; ROC

Abstract

The receiver operating characteristic (ROC) curve and its summary measure, the area under the ROC curve (AUC), are useful statistical tool to evaluate the performance of a biomarker predicting disease status or clinical condition. Motivated by the need for novel study designs with better estimation efficiency and reduced study cost in biomarker validation, we consider statistical inference on ROC and AUC under marker-dependent sampling scheme (MDS). The MDS design is a biased sampling scheme that consists of a simple random component and a supplemental marker-dependent component. Using such design, investigators can over-sample or under-sample subjects falling into certain ranges of the biomarker score, yielding an improved precision for the estimation of ROC and AUC with a fixed size of subjects. Marker-dependent sampling will introduce bias in the predictive accuracy of the biomarker if standard ROC and AUC estimation methods are used. In this presentation, we will discuss semiparametric methods for analyzing data with a marker-dependent structure. The asymptotic properties of the proposed estimators will be discussed as well as their finite sample properties.