Diagnostic yield of a heart failure referral pathway using N-terminal pro-brain natriuretic peptide

Objective To determine the diagnostic yield of a 'high' N-terminal pro-brain natriuretic peptide (NT-proBNP) in patients with suspected heart failure (HF) referred from primary to secondary care. Methods In this retrospective study, cardiac diagnoses were quantified in consecutive patients with an NT-proBNP>400 ng/L referred from primary care centres to a specialist HF service. Results Among 654 consecutive patients (age: 78.5±9.72 years; 45.9% men; left ventricular ejection fraction (LVEF): 55.4±12.5% (mean±SD)), the primary diagnoses were: valvular disease (39.4%), HF (29.2%; 13.3% with LVEF<40%) and atrial fibrillation (AF; 17.3%). In terms of primary or secondary diagnoses, 68% of patients had valve disease, 46.9% had AF and 29.2% had HF. A cardiac diagnosis was made in 85.9%. In multivariable analyses, NT-proBNP predicted HF with LVEF<40% (OR: 10.2, 95% CI: 5.63 to 18.3) and HF with any LVEF (OR: 6.13, 95% CI: 3.79 to 9.93). In canonical linear discriminant analyses, NT-proBNP correctly identified 54.5% of patients with HF. The remainder were misclassified as valvular disease, AF or no cardiac diagnosis. Conclusion Among patients with an NT-proBNP>400 ng/L referred through a primary care HF pathway, most patients had valve disease or AF rather than HF. NT-proBNP cannot discriminate among HF, valve disease and AF. On this basis, NT-proBNP may be best employed in detecting cardiac disease in general rather than HF per se.