Clinical value of NT-proBNP assay in the emergency department for the diagnosis of heart failure (HF) in very elderly people

Objective Scanty data are available on the accuracy of NT-proBNP in the diagnosis of HF and effects of comorbidities in very elderly patients. Methods Symptoms, signs, NT-proBNP, eGFR, Ht, CRP and the presence of cardiomegaly and pleuric effusion were assessed in 895 consecutive patients aged 86 ± 4.3 years admitted to Emergency Department and used to define the diagnosis of HF according to Framingham criteria. Receiver operating characteristic curves (ROC) were used to calculate diagnostic performance and cutoff of NT-proBNP. Sensitivity (Sn), specificity (Sp), positive predictive value (PPV) and negative predictive value (NPV) were computed for all NT-proBNP cutoffs. Results Satisfactory diagnostic performance was obtained with a lower threshold of 980 pg/mL (Sn 0.95; NPV 0.90) and a higher threshold of 5340 (Sp 0.85; PPV 0.76) but with 42.4% of patients in the uncertainty area. We determined a second couple of cutoffs (1470–4200) that reduced the gray-area to 27.4%, maintaining an acceptable diagnostic performance compared to commonly used cutoffs (300–1800). Ht, CRP and eGFR all correlated with NT-proBNP in groups with and without HF but none affected diagnostic performance. Conclusion NT-proBNP performs satisfactorily for the diagnosis of HF in very elderly patients. Proposed threshold couple, compared with the most used cutoffs, showed a gain in Sp and PPV with a slightly lower performance in Sn and NPV and with a decrease in the gray-area with the second one. Our data do not support the use of different NT-proBNP cutoffs depending on eGFR, Ht and CRP.