Improved diagnostic performance of high-sensitivity cardiac Troponin assays is an artifact of censored data

Abstract

BACKGROUND: High-sensitivity cardiac troponin assays with improved analytical performance at low concentrations are credited with increased diagnostic sensitivity in acute coronary syndrome patients. We investigated the relationship between analytical sensitivity (detection capability) and diagnostic accuracy and tested the effect of censoring data with a software model. METHOD: We generated 4 sets of results with decreasing detection capability and compared the ROC curves with and without censored data. RESULTS: There was no relationship between diagnostic performance and detection capability. When data were censored the diagnostic accuracy decreased progressively with an increase in the threshold concentration for censoring. The ROC curves constructed with censored data have a characteristic appearance with a straight line between the censoring point and the top right hand corner. CONCLUSIONS: There is not a direct relationship between the diagnostic accuracy and the detection capability of cardiac troponin assays. The artifactual decrease in diagnostic accuracy can be added to the list of reasons why data should not be censored and this practice should be disclosed in studies on diagnostic accuracy.