BACKGROUND:N-terminal pro-B-type natriuretic peptide (NT-proBNP) concentrations predict cardiovascular outcome in many settings. There are very few data assessing the utility of NT-proBNP concentrations in the prediction of long-term outcome after cardiac surgery. We assessed the ability of NT-proBNP to predict 3 yr mortality compared with validated clinical risk prediction tools. METHODS:A secondary analysis of a prospectively recruited patient cohort of 1010 patients undergoing cardiac surgery. Baseline clinical details were obtained including EuroSCORE. Multi-variable modelling, area under the receiver operating characteristic curves (AUCs), and net reclassification improvement were utilized. RESULTS:NT-proBNP was a univariable predictor of 3 yr mortality but was no longer a significant predictor in a multivariable model (hazard ratio 1.00 per 250 ng litre(-1), 95% confidence interval 0.98-1.02, P=0.80). The relative and additive predictive values of the preoperative EuroSCORE (both additive and logistic versions) and NT-proBNP concentrations were compared. All were predictive of 3 yr mortality (P<0.001) with almost identical AUCs (0.71 for EuroSCORE, 0.70 for NT-proBNP). When either the EuroSCORE or NT-proBNP concentrations are known, the addition of the other does not improve the ability to predict 3 yr mortality. CONCLUSIONS:Preoperative NT-proBNP concentrations and the EuroSCORE have equivalent, and moderate, predictive accuracy for mortality 3 yr after cardiac surgery. EuroSCORE uses clinical data but is not routinely used for individual clinical risk prediction. NT-proBNP measurement would incur additional costs but can be measured quickly and objectively. With such similar predictive accuracy, factors such as the ease of calculation and cost will likely determine their use in clinical practice.