Amino-terminal pro-B-type natriuretic peptide (NT-proBNP) has been associated with prognosis in heart failure. The aim of this study was to assess whether NT-proBNP enhances risk prediction in acute decompensated heart failure (ADHF). The authors enrolled 453 patients hospitalized for ADHF. The primary outcome was 12-month mortality. Two separate multivariable predictive models were built by using standard variables–only and NT-proBNP concentrations. The models were internally validated using the bootstrapping method. Twelve-month mortality was 28.3%. There was no statistical evidence of overfit. Compared with the standard variables–only model, the NT-proBNP model had a better predictive performance as judged by the Nagelkerke R2 (0.410 vs 0.374) and the Brier score (0.136 vs 0.141), which are measures of overall performance; the Akaike Information Criterion (399.2 vs 415.0), which is a tool for model selection; and the C index (0.844 vs 0.831), which is a measure of the discriminative ability. Both models were well calibrated, as judged by the Hosmer-Lemeshow chi-square test. Both models predicted 12-month mortality significantly better than the Enhanced Feedback for Effective Cardiology Treatment risk score. In conclusion, the NT-proBNP model improved risk prediction compared with the standard variables–only model.
氨基末端前b型利钠肽(NT-proBNP)与心力衰竭的预后相关。本研究的目的是评估NT-proBNP是否能增强急性失代偿性心力衰竭(ADHF)的风险预测。作者招募了453名ADHF住院患者。主要终点为12个月死亡率。分别使用标准变量和NT-proBNP浓度建立了两个独立的多变量预测模型。采用自举法对模型进行内部验证。12个月死亡率为28.3%。没有统计学上的过拟合证据。通过Nagelkerke R2 (0.410 vs 0.374)和Brier评分(0.136 vs 0.141)来判断,NT-proBNP模型与标准变量模型相比具有更好的预测性能;赤池信息标准(399.2 vs 415.0),这是模型选择的工具;C指数(0.844 vs 0.831),是衡量辨别能力的指标。根据Hosmer-Lemeshow卡方检验,两个模型都经过了很好的校准。两种模型对12个月死亡率的预测都明显优于有效心脏病治疗的增强反馈风险评分。综上所述,NT-proBNP模型比标准变量模型提高了风险预测。