Introduction and objectives. Recent studies have shown that brain natriuretic peptide (BNP) and N-terminal proBNP (NT-proBNP) are useful in the diagnosis of heart failure in patients presenting with dyspnea. However, the cutoff values used with these markers vary according to patient characteristics and dyspnea severity. The aim of this study was to investigate the diagnostic accuracy of using the plasma NT-proBNP level for identifying heart failure in a heterogeneous population of patients with dyspnea. Methods. A multicentre study involving 247 conse cutive patients with recent-onset dyspnea was carded out at 12 Spanish hospitals. Patients previously diagnosed with heart failure or any other condition known to cause dyspnea were excluded. Results. Of the 247 patients, 161 (65%) had heart failure. The remaining 86 (35%) presented with dyspnea of non-cardiac origin. Plasma NT-proBNP levels were higher in patients with heart failure (5600 [7988] pg/mL vs 1182 [4406] pg/mL; P=.0001), and increased as functional status deteriorated (P=.036). The area under the receiver operating characteristic curve was 0.87 (0.02) (95% Cl, 0.81-0.91) for the optimum cutoff value of 1335 pg/mL. The sensitivity of this cutoff value for diagnosing heart failure was 77% (95% Cl, 70%-83%), the specificity was 92% (95% Cl, 84%-97%), the positive predictive value was 94%, and the negative predictive value was 68%. Conclusions. The plasma NT-proBNP concentration provides an accurate means of diagnosing heart failure. However, the negative predictive value found in this study was somewhat lower than the values found in previous studies involving more homogeneous patient populations.