Objective: To evaluate the value of NT-proBNP in predicting in-hospital mortality in patients with decompensated systolic heart failure.
Methods: Plasma NT-proBNP levels within 24 hours of admission were obtained in 366 patients with decompensated systolic heart failure. The levels were compared between dying patients in hospital and survival patients at discharge. ROC analyses were performed to evaluate if NT-proBNP was a predictor for in-hospital mortality and identify the optimal NT-proBNP cut-off point for predicting in-hospital mortality. A binary logistic regression analysis was used to evaluate if NT-proBNP was an independent predictor for in-hospital mortality.
Results: 19 cases of the 366 patients died in hospital. NT-proBNP levels of the dying cases were much higher than those of the survivals 3970 (3452, 6934) pmol/L vs 2340 (1132, 4002) pmol/L respectively, P < 0.01). ROC analysis of NT-proBNP to predict in-hospital mortality had an area under the curve (AUC) of 0.762 (95% CI: 0.657-0.857, P < 0.01), the optimal NT-proBNP cut-off point for predicting in-hospital mortality was 3500 pmol/L with a sensitivity of 73.7%, a specificity of 66.9%, an accuracy of 67.6% and a negative predictive value of 97.9%. Patients whose NT-proBNP levels were equal or more than 3500 pmol/L had an in-hospital mortality of 10.9%, compare with 2.1% in those NT-proBNP levels less than 3500 pmol/L (P < 0.01). Binary logistic regression analysis demonstrated that NT-proBNP was an independent predictor for in-hospital mortality in patients with decompensated systolic heart failure (P < 0.01).
Conclusion: Admission plasma NT-proBNP level is an independent predictor for in-hospital mortality in patients with decompensated systolic heart failure. The optimal NT-proBNP cut-off point for predicting in-hospital mortality is 3500 pmol/L.