Prognostic Value of B-Type Natriuretic Peptides in Patients With Stable Coronary Artery Disease

The PEACE Trial

Study subjects

The design, entry criteria, and main results of the PEACE trial have been described previously (17). In brief, from November 1996 to June 2000, 8,290 patients were randomized to the ACE inhibitor trandolapril or placebo and followed up for a median of 4.8 years. Entry criteria included: age 50 years or more, documented coronary artery disease, and left ventricular ejection fraction >40% by contrast or radionuclide ventriculography or echocardiogram. Exclusion criteria included: serum creatinine >2.0 mg/dl, valvular heart disease requiring surgical intervention, or plans for coronary revascularization. All patients who had baseline ethylenediaminetetraacetic acid (EDTA) plasma samples for measurement of BNP and NT-proBNP were included in this substudy.

Blood sampling procedures and biochemical assays

Samples of EDTA-anticoagulated venous blood were obtained before randomization. The test tube was centrifuged at room temperature, and plasma was aspirated and frozen at −20°C at individual centers. Within 3 months of collection, plasma samples were shipped on dry ice to the central laboratory for storage at −70°C or colder pending analysis. For natriuretic peptide analyses, samples were shipped on dry ice to the University of Oslo, where samples were thawed and analyzed for BNP and NT-proBNP on the same day. Plasma samples were kept on ice throughout the day, avoiding refreezing and rethawing. The stability of samples during long-term storage was assessed by plotting the duration of storage versus peptide concentration, and no indication of degradation during long-term storage of either BNP or NT-proBNP was observed (data not shown).

The BNP concentrations in plasma were determined using a 2-step microparticle enzyme immunoassay on an Abbott Ax Sym analyzer (Abbott Laboratories, Abbott Park, Illinois). The interassay coefficient of variation was 12.7% at a concentration of 32 pg/ml (n = 142), 8.7% at a concentration of 146 pg/ml (n = 128), and 8.6% at a concentration of 480 pg/ml (n = 87). The measuring range for the assay was 15 to 4,000 pg/ml. The Food and Drug Administration (FDA)-approved BNP decision threshold for heart failure is 100 pg/ml.

The NT-proBNP concentrations in plasma were determined with an electrochemiluminescence immunoassay on a Modular platform (Roche Diagnostics, Basel, Switzerland). The interassay coefficient of variation was 4.5% at a concentration of 21 pg/ml (n = 100) and 2.2% at a concentration of 9,120 pg/ml (n = 100). The measuring range for the assay was 5 to 34,000 pg/ml. The FDA-approved NT-proBNP decision threshold for heart failure is 125 pg/ml for those age <75 years and 450 pg/ml for those age ≥75 years.

For the measurement of C-reactive protein, aliquots were shipped frozen on dry ice to the TIMI (Thrombolysis in Myocardial Infarction) Biomarker Laboratory (Boston, Massachusetts), where they were thawed and analyzed in batch. High-sensitivity testing for C-reactive protein was performed using the CRP-Latex (II) immunoturbidimetric assay (Denka Seiken, Tokyo, Japan) on a Hitachi 911 immunoanalyzer (Roche Diagnostics, Indianapolis, Indiana). This assay has a minimal detectable concentration of 0.03 mg/l and a total imprecision of 5.1% and 2.5% at concentrations of 0.2 and 1.9 mg/l, respectively (18). Creatinine and total cholesterol concentrations in serum were determined by routine laboratory methods. Glomerular filtration rate was estimated using the 4-variable Modification of Diet in Renal Disease equation (19). All biochemical testing was performed by personnel blinded to clinical outcomes and treatment allocation.

Statistical analysis

We used analysis of variance and chi-square tests to test for differences in continuous and categoric baseline characteristics between quartiles of BNP and NT-proBNP. The overall F-test is reported for the analysis of variance models.

In separate models, we used Cox proportional hazards models to examine the association between BNP and NT-proBNP and the following prospectively defined outcome variables: cardiovascular mortality, fatal or nonfatal myocardial infarction, fatal or nonfatal stroke, and fatal or nonfatal new congestive heart failure requiring hospitalization. All events were confirmed by review of medical records, with blinded review by an end points committee for cardiovascular death, fatal and nonfatal myocardial infarction, and fatal and nonfatal stroke.

Patients were censored at their last visit. A test of the interaction between BNP and treatment and between NT-proBNP and treatment assignment was included. Subsequent models were adjusted for treatment assignment and baseline variables known to be important predictors of cardiovascular events: age, gender, body mass index, ejection fraction <50%, estimated glomerular filtration rate, current smoking, history of hypertension or measured hypertension, history of myocardial infarction, history of diabetes, history of stroke, history of percutaneous coronary intervention, history of coronary artery bypass grafting, total cholesterol, C-reactive protein, use of a beta-blocker, use of a lipid-lowering drug, use of aspirin or an antiplatelet medication, and use of a diuretic. To examine the potential influence of interim heart failure events on cardiovascular death, additional models were adjusted for interim heart failure as a time-dependent variable and censored patients at the time of their heart failure event. Both BNP and NT-proBNP were examined as continuous variables (per 1 standard deviation of the logarithmic transformation). To determine if there was a gradient effect, BNP and NT-proBNP were also examined using gender-specific quartiles. Kaplan-Meier curves were generated to show the cumulative incidence for each of the end points. Residual analysis was used to assess model fit. The collinearity index was used to check for linear combinations among covariates (20). To assess the discriminative ability of BNP and NT-proBNP, we computed the c-index for each outcome and compared them between BNP and NT-proBNP (21,22).

Pearson correlation was used to describe the association between log BNP, log NT-proBNP, and risk factors. The SAS analysis system version 9.1 was used for all analyses (SAS Institute, Cary, North Carolina).

Characteristics at baseline

Characteristics at baseline in patients included in this biomarker substudy (n = 3,761) did not differ markedly from those of patients not included in this analysis (n = 4,529; data not shown).

Characteristics of the patients according to their BNP and NT-proBNP concentrations at baseline are shown in Tables 1 and 2,⇓respectively. The median value of BNP was 52.6 pg/ml (interquartile range 24.9 to 101.4 pg/ml; normal value <100 pg/ml), and the median value of NT-proBNP was 139.3 pg/ml (interquartile range 71.3 to 272.1 pg/ml; normal value: age <75 years: <125 pg/ml; age ≥75 years: <450 pg/ml). B-type natriuretic peptide and NT-proBNP correlated significantly (r = 0.62; p < 0.001). Modest correlations were observed between natriuretic peptides and ejection fraction (BNP: r = −0.09; p < 0.001; NT-proBNP: r = −0.18; p < 0.001), estimated glomerular filtration rate (BNP: r = −0.14; p < 0.001; NT-proBNP: r = −0.22; p < 0.001), body mass index (BNP: r = −0.09; p < 0.001; NT-proBNP: r = −0.12; p < 0.001), and systolic blood pressure (BNP: r = 0.11; p < 0.001; NT-proBNP: r = 0.15; p < 0.001). Both BNP (r = 0.27; p < 0.001) and NT-proBNP (r = 0.39; p < 0.001) correlated more closely with patient age.

Relation to fatal and nonfatal cardiovascular events

During a median of 4.8 years of follow-up there were 127 cardiovascular deaths, 104 fatal or first nonfatal congestive heart failure events, 241 fatal or first nonfatal acute myocardial infarction events and 86 fatal or first nonfatal stroke events. By univariable analyses, both BNP and NT-proBNP were significant predictors of cardiovascular mortality, fatal or nonfatal stroke, and fatal or nonfatal congestive heart failure but not fatal or nonfatal acute myocardial infarction (Table 3).

By multivariable analyses, after adjustment for risk factors, NT-proBNP remained a significant independent predictor of cardiovascular mortality (Fig. 1),fatal or nonfatal congestive heart failure (Fig. 2),and fatal or nonfatal stroke (Fig. 3,Table 4).Consistent results were obtained regardless of whether the peptides were treated as logarithmically transformed continuous variable (Table 4) or as categoric variables (Table 5).Adjustment or censoring for interim congestive heart failure had only minor impact on the association between NT-proBNP and risk of cardiovascular death. For example, adjustment for interim congestive heart failure reduced the hazard ratio for the fourth versus first quartile of NT-proBNP marginally from 4.9 (95% confidence interval 2.2 to 10.7) to 4.5 (95% confidence interval 2.0 to 9.8).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

NT-ProBNP and Cardiovascular Death

Kaplan-Meier curves showing the cumulative incidence of death due to a cardiovascular cause in patients according to quartiles of plasma N-terminal pro-B-type natriuretic peptide (NT-proBNP) concentrations.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

NT-ProBNP and Heart Failure

Kaplan-Meier curves showing the cumulative incidence of fatal or nonfatal congestive heart failure (CHF) in patients according to quartiles of plasma N-terminal pro-B-type natriuretic peptide (NT-proBNP) concentrations.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

NT-ProBNP and Stroke

Kaplan-Meier curves showing the cumulative incidence of fatal or nonfatal stroke in patients according to quartiles of plasma N-terminal pro-B-type natriuretic peptide (NT-proBNP) concentrations.

Calculation of the c-index as a measure of overall prognostic accuracy demonstrated that in univariable analysis NT-proBNP performed significantly better than BNP as a predictor of cardiovascular mortality, fatal or nonfatal stroke, and fatal or nonfatal congestive heart failure; neither predicted fatal or nonfatal myocardial infarction (Table 6).For the end point of fatal and nonfatal congestive heart failure, addition of either BNP or NT-proBNP to the multivariable model resulted in statistically significant improved prognostic accuracy, as evaluated by the c-index, compared with the conventional risk factor model. For cardiovascular mortality, NT-proBNP, but not BNP, resulted in statistically significant improved prognostic accuracy (Table 7).

In the main PEACE trial the incidence of the primary end point—death from cardiovascular causes, myocardial infarction, or coronary revascularization—was not significantly different between the ACE inhibitor or placebo group. However, significantly fewer patients in the ACE inhibitor group than in the placebo group were hospitalized with or died of congestive heart failure. No interaction was observed between BNP or NT-proBNP and randomized treatment group. That is, the association between treatment group (ACE inhibition or placebo) and each outcome did not vary according to BNP or NT-proBNP concentration (data not shown).

BNPs and specific cardiovascular events

In contrast to earlier studies in patients with stable coronary artery disease (9–13), the present study provides data on a range of specific cardiovascular end points, which also permits analysis of the association between these end points and cardiovascular death. The observation that neither BNP nor NT-proBNP predicted fatal and nonfatal myocardial infarction suggests that the association between NT-proBNP and cardiovascular death is unlikely to be explained by prediction of myocardial infarction. These results are in line with those of 3 large acute coronary syndrome studies that failed to demonstrate an independent association between BNPs and the incidence of coronary events (4,6,8), Not unexpectedly, both BNP and NT-proBNP were powerful and independent predictors of heart failure. However, in the present low-risk population, prediction of heart failure appeared to be of modest importance for prediction of cardiovascular mortality, because BNP failed to independently predict cardiovascular mortality despite a strong association with the incidence of heart failure. Interestingly, a similar lack of association between BNP and cardiovascular mortality, after adjustment for echocardiographic indices of ventricular function, was observed in the Framingham Heart Study population that also excluded subjects with heart failure at baseline (23). Moreover, adjustment for or censoring of incident heart failure in the present study did not impact on the merit of NT-proBNP as a predictor of cardiovascular mortality. The association between NT-proBNP and cardiovascular mortality may therefore be ascribed to prediction of neither heart failure nor myocardial infarction.

BNP and NT-proBNP as prognostic indicators

Existing data on the relative prognostic value of BNP and NT-proBNP are still scarce. One recent publication of patients with coronary artery disease suggested that the two are comparable as predictors of all-cause mortality and heart failure hospitalizations, but the study included a large proportion of patients with chronic heart failure and impaired systolic function and the results may not be generalizable to low-risk patients with stable coronary artery disease (13). Moreover, that study restricted their analysis to the end points of all-cause mortality or all-cause mortality and heart failure. The study also included a substantially lower number of study subjects than the present investigation and may as a consequence have had less statistical power to detect differences between BNP and NT-proBNP. The use of in-house biochemical assays for analysis of BNP and NT-proBNP also make direct comparisons with the present data and other studies using commercially available assays somewhat difficult. Moreover, the duration of time between blood sampling and determination of circulating peptide concentrations was substantially longer for NT-proBNP than for BNP, which may have introduced bias in favor of BNP in that study.

The relative prognostic value of BNP and NT-proBNP has also been evaluated in 2 other recent studies. The association between BNP and NT-proBNP and all-cause mortality was assessed in a community-based cohort of subjects free of heart failure at baseline (24). In a model containing both peptides, only NT-proBNP was predictive of outcome. However, because no overall index of prognostic value was provided, the clinical significance of the statistical difference is somewhat difficult to assess. Moreover, end points other than mortality were not examined. The relative value of BNP and NT-proBNP has also been examined in a large population of patients with symptomatic chronic heart failure (25). The predictive ability of BNP and NT-proBNP did not differ concerning the end point of all-cause mortality, but NT-proBNP provided superior prognostic information concerning the combined mortality and morbidity end point and concerning heart failure hospitalizations.

The reasons for the observed differences in prognostic value between BNP and NT-proBNP remain somewhat unclear. Early reports suggested that NT-proBNP is a more sensitive marker of asymptomatic left ventricular dysfunction than BNP (26,27), and this observation has been confirmed by a recent population-based study (28). A closer relation to renal function may not be fully accounted for by adjustment for estimated glomerular filtration rate and blood pressure and may represent a potential link to adverse outcome. A third possibility is that a longer half-life of NT-proBNP makes this peptide less prone to intraday variability than BNP, making a single measurement of NT-proBNP a more accurate and robust index of cardiovascular function and prognosis than a single measurement of BNP.

Finally, we cannot rule out the possibility that differences in in vitro stability and analytical precision between BNP and NT-proBNP assays may have contributed to the observed differences in prognostic merit. In a substudy of the SAVE (Survival and Ventricular Enlargement) trial, NT-pro-A-type natriuretic peptide (ANP) was found to be a stronger prognostic indicator than ANP (29), and this difference in prognostic value may have been based on the marked in vitro stability difference between ANP and NT-proANP. Data concerning the in vitro stability of BNP have been somewhat conflicting. Early studies suggested that BNP is stable in whole blood and EDTA plasma for at least 24 h (30,31). More recent data suggest that this may depend on the assay used. According to the manufacturer of the BNP assay used in the present study, BNP is stable for 4 h in whole blood and in EDTA plasma, and the recovery of BNP may potentially be diminished if collected in glass tubes. Moreover, long-term storage of frozen samples may also be associated with degradation of BNP. For instance, in one recent study using the Abbott AxSym assay, BNP immunoreactivity in EDTA plasma samples stored at −20°C for 2 to 4 months were reported to be <50% compared with levels measured within 4 h of blood collection (32). On the other hand, no association between the duration of storage and natriuretic peptide concentrations were observed in the present study. The observation of a strong and independent relationship to incident heart failure also argues strongly against significant degradation of BNP in the present study. However, the differences in prognostic merit between BNP and NT-proBNP for other end points may potentially be attributed to differences in in vitro stability, and these findings may not be relevant for samples measured immediately after blood collection in clinical day-to-day testing.

Differences in analytic performance of the assays may also have affected the results. In accordance with a study comparing the assays used in the present analysis (33), the low-range sensitivity and analytic precision of the NT-proBNP assay appeared to be superior to those of the BNP assay. Combined, these factors may have contributed to a lower signal-to-noise ratio for BNP than for NT-proBNP, translating into reduced prognostic power for BNP for non-heart failure end points.

What BNP and NT-proBNP concentrations define increased risk?

Identification of threshold concentrations of BNP and NT-proBNP that define increased risk in clinically stable patients would be helpful for the clinician. The approved cut-off for BNP for heart failure is derived from trials of acutely dyspneic patients and does not take into account the confounding effects of age and gender. Because it is not based on healthy individuals it should not be regarded as a conventional upper normal value and may be imperfect for defining risk in stable coronary artery disease patients. In fact, this was a main argument for not using those cut-offs in the present analyses. Instead, we analyzed BNP and NT-proBNP as continuous variables and by quartile. Although there is some degree of peptide-specific and end point–specific variations, the current data suggest that the risk associated with increasing BNP and NT-proBNP concentrations is largely monotonic and does not allow definition of a specific threshold value.

Incremental prognostic value of BNPs

Multiple biomarkers have been proposed as prognostic tools in patients with coronary artery disease, but few have been demonstrated to provide clinically relevant incremental prognostic value to conventional risk factors (14). A recent publication from the HOPE (Heart Outcomes Prevention Evaluation) study in patients with cardiovascular risk factors suggests that NT-proBNP provides incremental prognostic information concerning the combined end point of cardiovascular death, stroke, and myocardial infarction above that obtained by models of traditional risk factors (34). The present study extends these findings by showing that, in patients with established coronary artery disease, adding BNP and NT-proBNP to the best available model results in statistically significant improvement in predictive ability for the individual end point of heart failure, as evaluated by the c-index. Moreover, NT-proBNP improved the predictive ability of the model for the end point of cardiovascular death. In contrast to the HOPE substudy, the best available model in the present study included an objective measure of ventricular function.

Prediction of the effect of ACE inhibition

In contrast to the HOPE (16) and EUROPA (EURopean trial On Reduction of cardiac events with Perindopril in stable coronary Artery disease) (15) studies, in the PEACE trial, ACE inhibition did not result in any reduction in the primary outcome (i.e., cardiovascular mortality, myocardial infarction, or coronary revascularization) (17). One possible explanation for the discrepant result is the considerably lower risk of the patients included in the PEACE trial than in the HOPE and the EUROPA trials. Because BNP and NT-proBNP strongly relate to mortality in a wide spectrum of cardiovascular disease, including patients with stable coronary artery disease (9–13), it seemed reasonable to hypothesize that measurement of BNP and NT-proBNP might identify patients who benefit from ACE inhibition. The failure of BNP and NT-proBNP to be helpful in this respect may be related to the inability of BNPs to predict coronary ischemic events, which constituted the majority of events in the PEACE trial. It is also conceivable that the neutral overall results of the PEACE trial may have reduced the likelihood of detecting a significant interaction between natriuretic peptide levels and benefit of ACE inhibition.

Study limitations

Limitations of the present study include the selection of patients participating in clinical trials who may not be representative of all stable coronary artery disease patients, as well as the use of biobank specimens obtained at multiple sites and shipped for peptide analysis.

Study strengths

Strengths of the study include the prospective design of the PEACE biomarker substudy, the large and well characterized patient sample, the long duration of follow-up, accuracy of recording of cardiovascular end points, and adjustment for important cardiovascular risk factors.

Clinical implications

In low-risk patients with stable coronary artery disease and preserved left ventricular function, BNPs provide strong and incremental prognostic information to traditional risk factors. The clinical implications of elevated BNPs are less clear. Lack of prediction of myocardial infarction suggests that determination of BNPs may not be useful for selection of anti-ischemic therapies in low-risk patients. Contrary to our expectations, BNP and NT-proBNP failed to identify stable coronary artery disease patients who benefit from ACE inhibition. In conclusion, current data support measurement of BNPs in low-risk populations for prognostic assessment but do not provide a mandate for the use of these peptides for tailoring therapy in individual patients.