Area under ROC curve, sensitivity, specificity of N-terminal probrain natriuretic peptide in predicting mortality in various subsets of patients with ischemic heart disease

Abstract

Sirs: Biomarker-based risk prediction has been widely used in the cardiovascular field to predict future events or stratify patients in various categories of risk. Receiver operating characteristic (ROC) curve is an effective method of evaluating the performance of diagnostic tests and the area under the ROC curve is widely accepted as a measure of the discriminatory power of a diagnostic test [1]. It has been sufficiently demonstrated that N-terminal probrain natriuretic peptide (NT-proBNP) levels increase during myocardial ischemia [2] and they predict mortality in patients with coronary artery disease (CAD) including stable CAD [3], non-ST-segment elevation acute coronary syndromes [4] and ST-segment elevation acute myocardial infarction [5]. In a previous work from our group, we assessed the accuracy of NT-proBNP to predict mortality using the same consecutive series of patients with CAD as in this research letter. An area under ROC curve of 0.76 (95% confidence interval 0.72–0.80) was found [6]. However, discriminatory power of NT-proBNP in various subsets of patients with CAD has not been reported. Therefore, we performed a ROC curve-based analysis to assess the ability of NT-proBNP to predict mortality in various subsets of patients with angiographically-proven CAD. Discriminatory power of NT-proBNP to predict mortality was performed in 1552 patients with stable or unstable CAD. Definition of clinical characteristics, measurement of NT-proBNP (Roche Diagnostics), C-reactive protein and creatinine were performed as previously described [7, 8]. Patients were divided into various subsets by dichotomizing clinical characteristics (Table 1). The area under ROC curve and 95% confidence limits were used to assess predictive power for mortality in each of the subsets of patients. The NT-proBNP cutoff value with the minimal difference between the sensitivity and specificity was defined. During a 3- to 5-year follow-up (median 3.6 years), 171 patients (11%) died. The area under ROC curve with 95% confidence interval, the best NT-prBNP cutoff value and its sensitivity and specificity regarding mortality are shown in Table 1. The area under ROC curve varied from 0.67 in patients younger than 60 years to 0.81 in patients with impaired renal function. The lowest best NT-proBNP cutoff value was in patients younger than 60 years (257.3 pg/ml) and the highest was in patients with atrial fibrillation (2666.0 pg/ml). Sensitivity and specificity with minimal difference between them ranged from 62.8 to 76.1% and from 61.9% to 76.1%, respectively.