Quantifying the added value of BNP in suspected heart failure in general practice: an individual patient data meta-analysis

• Published in: Heart (British Cardiac Society) Vol. 97; no. 12; pp. 959 - 963
• Main Authors: Kelder, Johannes C, Cowie, Martin R, McDonagh, Theresa A, Hardman, Suzanna M C, Grobbee, Diederick E, Cost, Bernard, Hoes, Arno W
• Format: Journal Article
• Language: English
• Published: London BMJ Publishing Group Ltd and British Cardiovascular Society 01.06.2011; BMJ Publishing Group; BMJ Publishing Group LTD
• Subjects: Accuracy; Aged; Beta blockers; Biological and medical sciences; Biomarkers - blood; Brain; Cardiology. Vascular system; Datasets; diagnosis; Electrocardiography; Family medical history; Family Practice; Female; general practice; Heart; Heart attacks; Heart failure; Heart Failure - diagnosis; Heart failure, cardiogenic pulmonary edema, cardiac enlargement; Humans; Male; Medical prognosis; Medical sciences; natriuretic Peptide; Natriuretic Peptide, Brain - blood; Patients; Point-of-Care Systems; Sensitivity and Specificity; Studies; Human; Heart failure; Data analysis; Heart disease; Individual; Cardiovascular disease; Patient; General practice; Circulatory system; Cardiology; Quantitative analysis
• ISSN: 1355-6037; 1468-201X
• DOI: 10.1136/hrt.2010.220426

BackgroundDiagnosing early stages of heart failure with mild symptoms is difficult. B-type natriuretic peptide (BNP) has promising biochemical test characteristics, but its diagnostic yield on top of readily available diagnostic knowledge has not been sufficiently quantified in early stages of heart failure.ObjectivesTo quantify the added diagnostic value of BNP for the diagnosis of heart failure in a population relevant to GPs and validate the findings in an independent primary care patient population.DesignIndividual patient data meta-analysis followed by external validation. The additional diagnostic yield of BNP above standard clinical information was compared with ECG and chest x-ray results.Patients and methodsDerivation was performed on two existing datasets from Hillingdon (n=127) and Rotterdam (n=149) while the UK Natriuretic Peptide Study (n=306) served as validation dataset. Included were patients with suspected heart failure referred to a rapid-access diagnostic outpatient clinic. Case definition was according to the ESC guideline. Logistic regression was used to assess discrimination (with the c-statistic) and calibration.ResultsOf the 276 patients in the derivation set, 30.8% had heart failure. The clinical model (encompassing age, gender, known coronary artery disease, diabetes, orthopnoea, elevated jugular venous pressure, crackles, pitting oedema and S3 gallop) had a c-statistic of 0.79. Adding, respectively, chest x-ray results, ECG results or BNP to the clinical model increased the c-statistic to 0.84, 0.85 and 0.92. Neither ECG nor chest x-ray added significantly to the ‘clinical plus BNP’ model. All models had adequate calibration. The ‘clinical plus BNP’ diagnostic model performed well in an independent cohort with comparable inclusion criteria (c-statistic=0.91 and adequate calibration). Using separate cut-off values for ‘ruling in’ (typically implying referral for echocardiography) and for ‘ruling out’ heart failure—creating a grey zone—resulted in insufficient proportions of patients with a correct diagnosis.ConclusionBNP has considerable diagnostic value in addition to signs and symptoms in patients suspected of heart failure in primary care. However, using BNP alone with the currently recommended cut-off levels is not sufficient to make a reliable diagnosis of heart failure.