Objective: We aimed to assess the predictive value of high NT-proBNP for all-cause long-term mortality in heart failure (HF) patients for different subsets of renal function defined by the estimated glomerular filtration rate (eGFR).
Materials and methods: HF patients admitted consecutively to our cardiology department from 2011 to 2014 were retrospectively included in the cohort. Patients without contemporary guideline-directed medical therapy, with in-hospital mortality, or with incomplete data were excluded.
All-cause mortality was assessed in June 2020 after a median follow-up of 96 months. eGFR was calculated using the CKD-EPI formula.
Results: The study sample included 1,262 patients with confirmed HF and with a mean age of 72.21 ± 10.47 years. 52.06% were women. The mean EF was 42.79% ± 12.14%. All-cause long term-mortality was 44.36%.
In multivariable analysis eGFR (p=0.001) and NT-proBNP (p<0.001) were independent predictors of mortality. The overall predictive value of NT-proBNP for mortality associated an AUC of 0.709, 95%CI 0.679–0.739, p<0.001.
The NT-proBNP predictive power decreased with decreasing eGFR: from AUC 0.749 (95% CI [0.672–0.819]; p<0.001) for eGFR >90ml/min/1.73m2, to AUC 0.685 (95% CI [0.636–0.733]; p<0.001) for eGFR between 60–90 ml/min/1.73m2, to 0.688 (95% CI [0.621–0.756]; p<0.001) for eGFR between 45–60 ml/min/1.73m2, to 0.660 (95% CI [0.568–0.751]; p=0.001) for eGFR between 30–45ml/min/1.73m2.
In patients with eGFR <30ml/min/1.73m2, NT-proBNP lost statistical significance for mortality prediction (AUC 0.626, 95% CI [0.455–0.798]; p=0.137).
Conclusion: In HF patients NT-proBNP and eGFR were independent predictors of mortality. However, at low eGFR (30ml/min/1.73m2), the predictive value of NT-proBNP is lost.