Heart failure is a major public health problem. The clinical diagnosis is imprecise and a simple test of left ventricular function would greatly improve diagnostic accuracy. Measurement of brain natriuretic peptide (BNP) has been shown in some studies to be a potential marker of left ventricular dysfunction.1 Heart failure is more common in the elderly and is often treated in the community where a simple test would be of considerable benefit. We have therefore measured plasma BNP values and compared them with echocardiographic measures of left ventricular function in a large, community based population.

Eight general practices of a total list size of 65 820 were surveyed. All patients currently prescribed loop diuretics were invited to take part in the study as it was thought the majority of patients with the clinical syndrome of heart failure would be treated with them. Of the 1425 patients asked to participate, informed consent, echocardiograms, and measurement of plasma BNP were obtained in 653. The median age of the group was 76 years with an interquartile range (IQR) of 70–82 years. The group was composed of 254 males and 399 females. Patients took all usual medication and were not subjected to any dietary restrictions. The study was approved by the local committee on medical research ethics.

Left ventricular volumes were measured by two dimensional (cross sectional) echocardiography using a phased array sector scanner (Vingmed CFM 700). The images were analysed using a computer assisted, video overlay, echocardiographic analysis system (Thoraxcenter, Erasmus University, Rotterdam, The Netherlands). An apical four chamber view was used for imaging and a modified Simpson's single plane disc method for analysis. The method has been described previously.2All echocardiograms were performed by three experienced operators and all analyses by two technicians.

Venous blood for measurement of BNP was collected into chilled tubes containing 1 mg/ml EDTA and 1000 KIU/ml aprotinin. Plasma was separated within one hour and stored at −70°C until analysis. All samples were extracted within four weeks and assayed by radioimmunoassay.

Plasma samples were acidified and extracted using preactivated Sep Pak C18 cartridge (Waters Corporation, Milford, Massachusetts, USA). The eluates were dried under vacuum using a centrifugal evaporator and stored frozen until assayed. The precipitates were resuspended in assay buffer and assayed by radioimmunoassay (Peninsula Laboratories, St Helens, UK).

Extraction of plasma samples with known amounts of standard hBNP-32 added (n = 12) gave a mean (SD) recovery of 83 (12.5)%. The intra- and inter assay coefficients of variation for the assay were 5.6% (n = 10) and 9.6% (n = 20), respectively. All the samples for the study were extracted and assayed by either of two experienced operators, who were blinded to the echocardiograph data. Blood samples taken from 50 healthy subjects aged from 20 to 81 years with no evidence of cardiac disease gave a median plasma BNP of 33 pg/ml (22–40 pg/ml IQR).

There was a significant negative correlation between plasma BNP and left ventricular ejection fraction (fig 1A;r = −0.1695, p < 0.001 (Spearman correlation coefficient), n = 653). There was also a significant correlation between plasma BNP and both left ventricular end diastolic volume index (r = 0.2640, p < 0.001, n = 652) and left ventricular end systolic volume index (r = 0.2710, p < 0.001, n = 652).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

(A) Brain natriuretic peptide (BNP) concentration against left ventricular ejection fraction (LVEF). (B) Graph showing distribution of plasma BNP concentrations when the cohort is divided using an ejection fraction of 40% as the cut off point. Solid bars indicate median values, dashed bars represent the lower and upper quartiles.

In patients with a left ventricular ejection fraction above 40% the median (IQR) plasma BNP concentration was 47 pg/ml (32–79 pg/ml) compared to 60 pg/ml (38–97 pg/ml) in those with an LVEF of < 40% (Wilcoxon rank sum test; p < 0.001). The plasma BNP concentrations for both groups are presented as a scatterplot in fig 1B.

A receiver operator characteristic (ROC) curve was constructed (SAS statistical software) to assess the sensitivity and specificity of BNP to detect left ventricular systolic dysfunction. The area under the curve, which provides a measure of the overall diagnostic accuracy of the test, was 0.587, indicating poor sensitivity and specificity.

This study has demonstrated a significant correlation between plasma BNP and echocardiographic measures of left ventricular systolic dysfunction. However, the correlation coefficient indicates the relation is not close and the area under the ROC curve was much less than reported in other studies.3 ,4 This suggests that the accuracy of BNP concentrations to provide a measure of left ventricular function and help diagnose heart failure is not as good as previously reported. The reason for the differences between our results and those of others is not clear. We believe our BNP assay is robust and the echocardiograms were performed by a small number of experienced operators and analysed in a standardised way. It is possible that our patients were different to those included in other studies. We deliberately chose patients who general practitioners thought might have heart failure and had accordingly treated them with diuretics. The diagnostic difficulties are clearly more important in the community. Furthermore the age of the patients included in this study was much higher than in others, but also the prevalence of heart failure is much higher in the elderly. We do not believe the previous prescription of diuretics would have adversely affected our results, as there is no evidence diuretics change BNP concentrations or have significant effects on echocardiographic measures of ejection fraction. The results of this study suggest that BNP concentrations may not be as accurate an aid to the diagnosis of heart failure as previously stated.