ABC | Volume 110, Nº3, March 2018

Original Article Almeida Junior et al BNP, echo and clinical evaluation in heart failure Arq Bras Cardiol. 2018; 110(3):270-277 Figure 2 – Receiver operator characteristics curves for estimating left atrial pressure ≥ 15 mm Hg. Estimates were based on dichotomized variables. CS: clinical score; CR: chest radiography; BNP: B-type natriuretic peptide. 100 100 80 80 60 60 40 40 20 20 0 0 Sensitivity 100-Specificity CS (AUC: 0.53) CS+CR (AUC: 0.60) CS+CR+BNP400 (AUC: 0.62) CS+CR+BNP1000 (AUC: 0.66) Table 4 – Diagnostic characteristics of the B-type natriuretic peptide (BNP) and chest radiography to predict left atrial pressure ≥ 15 mm Hg Sensitivity Specificity PPV NPV Accuracy BNP > 400 73 44 83 30 67 BNP >1000 44 88 93 29 53 Chest radiography 79 44 84 36 72 PPV: positive predictive value; NPV: negative predictive value. Table 5 – Diagnostic characteristics of clinical score, chest radiograph (CR), B-type natriuretic peptide (BNP) and all combined to predict left atrial pressure ≥ 15 mm Hg Sensitivity Specificity PPV NPV Accuracy CS+ 64 33 78 20 58 CS+ plus CR 82 33 82 33 72 CS+ plus CR plus BNP > 400 91 22 81 40 76 CS+: positive clinical score; PPV: positive predictive value; NPV: negative predictive value. Moreover, in that study, a low diagnostic accuracy (58%) for the diagnosis of ADHF when utilizing all signs together was observed. In our study, S3 was present in less than half of patients and, when present, showed a positive predictive value of 79% for LAP > 15 mm Hg. When absent, an elevation of filling pressures could not be ruled out. In addition, S3 did not add any information regarding hemodynamic status. This is in accordance with other studies. 3,26 Of note, in our study, all physical examinations were done by a heart failure specialist. In the setting of a less experienced professional, the accuracy of physical exam (particularly S3) can be low, since studies suggest poor agreement between medical interns or residents and phonocardiographic findings. 27 On the other hand, reasonable agreement in S3 detection has been found among professionals of heart failure clinics. 28 Levels of BNP The strongest evidence for clinical use of BNP is to discriminate the cause of dyspnea in patients admitted in the emergency department 29 and to assess prognosis. 30,31 For other BNP purposes, data are less clear. In the Escape Trial, 32 the ROC curve for the performance of BNP in estimating an elevation in PCWP > 22 mmHg showed a poor performance (AUC = 0.55). Another study with 40 critically ill patients utilizing invasive hemodynamic monitoring has shown a weak correlation between BNP and PCWP (r = 0.58). 33 Our data were consistent with these studies, showing a weak correlation between BNP and LAP (r = 0.29). BNP was also tested for guiding treatment, because, theoretically, lowering BNP is a consequence of lowering filling pressures, 34 but this strategy failed to show clinical benefit. 35 In contrast, 274