To determine the diagnostic accuracy of the measurement of plasma B-type natriuretic peptide (BNP) and N-terminal pro-BNP (NTproBNP) in patients referred by their general practitioners (GPs) with symptoms suggestive of heart failure. Additionally, to compare the diagnostic accuracy of the resting 12-lead electrocardiogram (ECG) with that of the peptides. A diagnostic accuracy study. Rapid-access heart failure clinics in five hospitals. 306 patients referred by their GPs with suspected heart failure. Sensitivity, specificity, positive and negative predictive values (PPV and NPV) and positive and negative likelihood ratios for BNP, NTproBNP and the ECG for the diagnosis of heart failure. Area under the receiver operating characteristics (ROC) curves for the two natriuretic peptides. The diagnosis of heart failure was confirmed in 104 (34%) patients. The area under the ROC curve was 0.84 [95% CI 0.79-0.89] for BNP and 0.85 [0.81-0.90] for NTproBNP. At the manufacturers' recommended decision cut-points, NTproBNP provided a higher NPV (0.97) than BNP (0.87), but at lower PPV (0.44 versus 0.59). An abnormal ECG did not add any further predictive value to that of NTproBNP. We have confirmed the value of the measurement of plasma BNP or NTproBNP as a 'rule-out' test for heart failure in patients currently referred by GPs to rapid access diagnostic clinics. A simple classification of the 12-lead ECG into 'normal' or 'abnormal' adds little value to ruling out heart failure in these circumstances. Further work is necessary to establish the best decision cut-points for use in clinical practice.