Receiver operating characteristic (ROC) curves and the definition of threshold levels to diagnose coronary artery disease on electrocardiographic stress testing. Part I: The use of ROC curves in diagnostic medicine and electrocardiographic markers of ischaemia

Abstract

A common problem in diagnostic medicine, when performing a diagnostic test, is to obtain an accurate discrimination between 'normal' cases and cases with disease, owing to the overlapping distributions of these populations. In clinical practice, it is exceedingly rare that a chosen cut point will achieve perfect discrimination between normal cases and those with disease, and one has to select the best compromise between sensitivity and specificity by comparing the diagnostic performance of different tests or diagnostic criteria available. Receiver operating characteristic (or receiver operator characteristic, ROC) curves allow systematic and intuitively appealing descriptions of the diagnostic performance of a test and a comparison of the performance of different tests or diagnostic criteria. This review will analyse the basic principles underlying ROC curves and their specific application to the choice of optimal parameters on exercise electrocardiographic (ECG) stress testing. Part I will focus on theoretical description and analysis along with reviewing the common problems related to the diagnosis of myocardial ischaemia by means of exercise ECG stress testing. Part II will be devoted to applying ROC curves to available diagnostic criteria through the analysis of ECG stress test parameters.