Plasma brain natriuretic peptide concentration: impact of age and gender

Population sampling, subject recruitment, and enrollment

A random sample of the population >44 years was invited to participate. A sampling fraction of 7% was applied within each of the gender and age (five-year) specific strata. Of the 4,203 subjects invited, 2,042 (47%) participated. Analysis of the medical records of 500 nonparticipants revealed no clinically significant differences between participants and nonparticipants. Subjects gave written consent and underwent echocardiography and phlebotomy.

Of the 2,042 participants, 1,020 had no history of cardiovascular, renal, or pulmonary disease and were on no cardiovascular medications. Of these, 75 were excluded because of an abnormal two-dimensional (2-D) echocardiogram, 64 were excluded by an inability to characterize diastolic function, and 114 had abnormal diastolic function. Thus, 767 subjects had normal EF (>50%), no wall motion abnormalities, normal diastolic function, no valve disease, and normal sinus rhythm; these make up the normal subgroup. The criteria used to establish the normal subgroup resulted in small numbers in the oldest (age 75 to 83) normal subgroup, where there were 18 women and two men. In this age group, the majority of excluded subjects were excluded because of the presence of known cardiovascular disease (72% of women excluded, 84% of men excluded) and a minority were excluded purely on the basis of diastolic dysfunction (10% of women excluded, 7% of men excluded).

Medical record review

All Olmsted County care providers have maintained a unified medical record, which is indexed by the Rochester Epidemiology Project. Each subject’s medical record was reviewed by trained nurse chart abstractors using established criteria for hypertension (12), myocardial infarction (13), and congestive heart failure (10). Clinical diagnoses of coronary artery disease, obstructive lung disease, renal failure, and diabetes were recorded. Medication questionnaires were completed by each subject.

Doppler echocardiography

All echocardiograms were performed by one of three registered diagnostic cardiac sonographers with the same echocardiographic instrument (HP-2500) according to protocol and interpreted by a single echocardiologist (M. M. R.). Two-dimensional and color Doppler imaging were performed to screen for valvular stenosis or regurgitation.

In each subject, measurement of EF by M-mode (modified Quinnones formula), quantitative 2-D (BiPlane Simpsons) and semiquantitative 2-D (visual estimate) methods was attempted as previously described (14–17). For analysis in the normal subgroup, subjects had normal EF (>40%) by all three methods. For analysis in the entire population, the visual estimation of EF was used.

Pulsed-wave Doppler examination of mitral (before and with Valsalva maneuver) and pulmonary venous inflow as well as Doppler tissue imaging of the mitral annulus was performed in each subject. Diastolic function was categorized as normal, impaired relaxation without evidence of increased filling pressures (“impaired relaxation”), impaired relaxation associated with moderate elevation of filling pressures (“pseudonormal filling”), and advanced reduction in compliance (“restrictive filling”), as previously described (18–21). Decreases in the ratio of the mitral early to late filling velocities (E/A ratio) indicate a reduction in early diastolic filling related to impairment of LV relaxation. We defined impaired relaxation as an E/A ratio ≤0.75, reflecting the average lower confidence interval for the E/A ratio in subjects aged >45 years in a previous normal cohort (22). Doppler indices used to discriminate pseudonormal or restrictive from normal are independent of age.

Left ventricular dimension and mass (M-mode) and left atrial volume (2-D) were indexed to body surface area.

BNP analysis

Blood for BNP was collected in the fasting state in ethylenediaminetetraacetic acid-treated tubes and placed on ice. After centrifugation at 2,500 rpm and 3°C, the plasma was stored at −80°C.

Plasma BNP was determined by immunoradiometric assay (nonextracted) using antibody to human BNP (Shionogi Co. Ltd., Tokyo, Japan), as previously described (23). The intra-assay (within-day) and interassay (total) coefficients of variation were both 8% and were consistent across concentrations of BNP.

One-milliliter aliquots of frozen plasma were shipped to Biosite Diagnostics (San Diego, California), and the nonextracted samples were batch-analyzed by a fluorescence immunoassay (5,24). The within-day coefficient of variation (9.4% to 15.2%) and total coefficient of variation (10.1% to 16.2%) increased from low to high BNP.

Statistical methods

Because the variability of BNP increased with its mean level, the natural log transformation was used in the regression analyses to satisfy modeling assumptions. Nomograms based on age and gender were established using the least-squares regression fit of log-transformed BNP with age and gender as predictor variables. From the fitted model, the 5th, 25th, 50th, 75th, and 95th percentiles were estimated and back-transformed to the natural scale. An interaction term with age and gender was also evaluated to determine if the association of age with BNP differed between the genders.

Unadjusted and adjusted associations of BNP were evaluated using Spearman’s correlation coefficient and linear least-squares regression, respectively. For adjusted associations, a model with age and gender was fit and, in turn, variables were added to see if there was significant residual association that could be explained by these characteristics.

The ability of BNP to detect an EF ≤40% was evaluated using receiver operating characteristic (ROC) curves within age/gender-specific strata. The optimal discriminatory value was identified as the BNP value that had a combined sensitivity/specificity at the smallest distance to 100%/100%, respectively (23).

Discriminatory values of BNP for the detection of systolic dysfunction in the entire cohort (n = 2,042)

Figure 3 displays the ROC curves for the detection of an EF ≤40% within each age stratum by gender and assay. Forty subjects (eight women) had an EF <40%. On each curve, both the BNP representing the upper (95th percentile) reference limit, as derived from the regression fit, and the BNP representing the optimal point on the ROC curve are indicated. The values for the BNP at the upper reference limit and at the optimal point on the ROC curve, and the corresponding sensitivity and specificity for the detection of an EF ≤40%, are shown in Table 5. The areas under the ROC curves were similar within each age/gender stratum for each assay. However, in general the optimal discriminatory value for BNP increases with age and is higher in women. Only eight female subjects had an EF ≤40% and this significantly limited the power to demonstrate age/gender differences in the optimal discriminatory value for BNP. The available data in women are included to demonstrate the effect on specificity of age/gender-specific partition values derived from the reference population.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Receiver operating curves defining the sensitivity and specificity of brain natriuretic peptide (BNP) (by Shionogi and Biosite assays) for the detection of systolic dysfunction (ejection fraction ≤40%) in each age strata of men and women. The open circle indicates the BNP value at the upper reference limit as derived from the regression fit in each age/gender/assay stratum. The open triangle indicates the BNP value with optimal sensitivity and specificity (the “optimal point”) on the receiver operating characteristic (ROC) curve.

Implications for use of BNP as a diagnostic test

Previous studies have reported the sensitivity and specificity of BNP for the detection of CHF or LV dysfunction (2–5). These studies derived a single abnormal (partition) value for BNP by ROC analysis and have not examined the use of age- or gender-specific partition values. For the Biosite assay, values of 75 to 80 pg/ml have been reported as partition values for detection of ventricular dysfunction (with and without symptoms) or diagnosis of CHF (5,24). For the Shionogi assay, values of 17.9 to 79 pg/ml have been reported for detection of ventricular dysfunction or CHF in highly variable populations (2,8,23,27,28). Although we found that the area under the ROC curves was similar among the age/gender strata, the optimal partition value for the detection of systolic dysfunction differed by assay, age, and gender.

These data suggest that age-, gender-, and assay-specific partition values should be used to interpret BNP. The ultimate partition values used may also depend on the abnormality being screened for (CHF vs. asymptomatic LV dysfunction), its severity, the implications of false positives or negatives and, importantly, on the population being studied. The operating characteristics of diagnostic tests may vary according to the population to which they are applied (11). Even though patients with severe CHF have very high BNP, age/gender effects still need to be considered when choosing a partition value. A previous study that utilized the Biosite assay to measure BNP reported that a BNP of 80 pg/ml is diagnostic of CHF in symptomatic patients presenting for urgent care (5). Although the average BNP value in patients with CHF was very high, and the performance of the test was outstanding, the study population was young (mean age 63 years) and predominately men (94%). In the current study, a value of 80 pg/ml would be abnormal in most male age groups (Table 2, Fig. 1) but would be well within the 95th percentile for women. A higher partition value may be needed in women or older patients where use of age- and gender-specific partition values may enhance the predictive characteristics of the test, particularly avoiding specificity problems in the elderly.

Limitations

The population of Olmsted County, Minnesota, is primarily white, and thus we are unable to investigate potential influence of ethnicity on BNP. Our study was confined to subjects >44 years of age.

Conclusions

In this population-based cohort, we found that BNP increases with age and is higher in women among subjects without cardiovascular disease or cardiac dysfunction. The magnitude of the effects of age and gender on BNP in the normal subgroup suggested that both parameters need to be considered when interpreting BNP, and indeed, the optimal discriminatory value of BNP for the detection of systolic dysfunction in the population was higher in women and older persons. Although the ultimate partition value for BNP used in clinical practice will be influenced by a number of variables, these data suggest that age-, gender-, and assay-specific values will be needed.