Evaluating Classification Performance of Biomarkers in Two-Phase Case-Control Studies

• Activity Number: 126 - Diagnostics, Classification, and Agreement
• Type: Contributed
• Date/Time: Monday, July 31, 2017 : 8:30 AM to 10:20 AM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Lu Wang* and Ying Huang
• Companies: Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center
• Keywords: Two-phase sampling ; ROC curve ; AUC ; Partial AUC ; Inverse probability weighting ; Calibration

Abstract

Biomarkers have played an increasingly important role in disease early detection and risk prediction. Two-phase case-control sampling design has been widely used in biomarker evaluation. Oftentimes sampling probabilities for cases and controls in phase-two can depend on other covariates (sampling strata). Such biased sampling can lead to invalid inference on a biomarker's classification accuracy if not properly accounted for. In this paper, we adopt the idea of inverse probability weighting(IPW) and propose IPW estimators of the biomarker's classification performance, including points on the receiver operating characteristics (ROC) curve, area under the ROC curve(AUC) and partial AUC. We consider sampling weights estimated conditional on sampling strata and further improve the efficiency of classification accuracy estimators by estimated weights or calibration via use of auxiliary variables. Asymptotic properties of the proposed estimators are developed. Our simulation studies illustrate that the proposed weighted estimators provide valid inference while the traditional empirical method can produce severely biased estimators. We apply the proposed method to a prostate cancer study.