Prognostic Impact of Plasma N-Terminal Pro–Brain Natriuretic Peptide in Severe Chronic Congestive Heart Failure
A Substudy of the Carvedilol Prospective Randomized Cumulative Survival (COPERNICUS) Trial
Background— The utility of N-terminal proBNP (NT-proBNP) to predict the occurrence of death and hospitalization was prospectively evaluated in the COPERNICUS study, which enrolled patients with an ejection fraction <25% and symptoms of chronic congestive heart failure at rest or on minimal exertion.
Methods and Results— Baseline plasma concentrations of NT-proBNP were measured in a subgroup of 814 men and 197 women with symptoms at rest or on minimal exertion who were enrolled in the COPERNICUS study and were randomized to placebo (n=506) or carvedilol (n=505). Values of NT-proBNP were markedly increased despite the requirement that patients be euvolemic before the start of treatment (mean±SD, 3235±4392 pg/mL; median, 1767 pg/mL). By univariate Cox regression analysis, NT-proBNP was found to be a powerful predictor of subsequent all-cause mortality (relative risk [RR], 2.7; 95% CI, 1.7 to 4.3; P=0.0001 for above versus below median) and all-cause mortality or hospitalization for heart failure (RR, 2.4; 95% CI, 1.8 to 3.4; P=0.0001 for above versus below median). The predictive value of NT-proBNP was similar when both placebo and carvedilol patients were analyzed separately. No significant interaction was found between NT-proBNP and treatment group (P=0.93 for above- versus below-median NT-proBNP).
Conclusions— NT-proBNP was consistently associated with increased risk for all-cause mortality and for all-cause mortality or hospitalization for heart failure in patients with severe congestive heart failure, even in those who were clinically euvolemic. This marker therefore may be a useful tool in risk stratification of patients with severe congestive heart failure.
Received June 26, 2002; de novo received October 25, 2003; revision received May 12, 2004; accepted May 21, 2004.
The association of neurohormonal activation with cardiac events in chronic congestive heart failure (CHF) is well established.1–4 Recently, natriuretic peptides, particularly brain natriuretic peptide (BNP), were introduced as biochemical indicators of impaired left ventricular function.5,6 Several assays measuring different peptides were developed and found to be suitable for risk stratification. Immunoreactive amino-terminal (NT) proBNP is present in human plasma with concentrations similar to that of BNP in subjects with normal left ventricular function. Compared with BNP, a higher absolute and relative increase in NT-proBNP has been observed in patients with heart failure,7–9 indicating that NT-proBNP might be a more sensitive cardiac marker than BNP.
NT-proBNP has been shown to be an independent indicator of survival in patients after myocardial infarction10,11 and in patients with established ischemic heart failure (NYHA class II to III).12 So far, however, the predictive value of NT-proBNP plasma measurements has not been prospectively evaluated in a large cohort of patients with symptoms of severe chronic heart failure.
The Carvedilol Prospective Randomized Cumulative Survival (COPERNICUS) trial included the most severe heart failure patients of all β-blocker trials conducted so far. This trial provided the unique opportunity to prospectively test the predictive power of biochemical indexes of heart failure when clinical findings indicated severe disease. We hypothesized that plasma levels of NT-proBNP might identify patients at particularly high risk for an adverse outcome in individuals with symptoms of severe chronic CHF. We further speculated that benefit from carvedilol treatment might be more pronounced in the patients with high NT-proBNP levels.
Methods and results of the COPERNICUS study have been reported elsewhere.13 The NT-proBNP substudy was approved by the COPERNICUS Steering Committee before the start of the main study and was coordinated by the University Hospital of Luebeck in cooperation with Roche Diagnostics GmbH. The institutional ethics committees of the participating study sites approved the substudy protocol. Written informed consent was obtained from all patients.
In COPERNICUS, 2289 patients with severe chronic CHF, left ventricular ejection fraction <25%, and symptoms at rest or on minimal exertion despite appropriate conventional therapy for heart failure were randomized to receive placebo or carvedilol. Patients had to be clinically euvolemic (defined as the absence of rales and ascites and no more than minimal peripheral edema) and could not be enrolled if they needed intensive care or treatment with intravenous inotropic agents or intravenous vasodilators within 4 days of screening. Serum creatinine had to be ≤2.8 mg/dL or 250 μmol/L at randomization and must not have increased >0.5 mg/dL or 44.2 μmol/L during the screening phase. Carvedilol/placebo doses were uptitrated at 2-week intervals to a target dose of 25 mg twice daily or, if the patient was intolerant, to the highest tolerated dose. Patients then continued on the highest tolerated dose until the end of the trial. Patients were followed up for up to 29 months. The Data and Safety Monitoring Board stopped the trial program early because of a marked effect of carvedilol on survival.
In a prospective protocol, 1011 of the 1387 European patients were included in the NT-proBNP substudy. Plasma samples for baseline neurohormonal levels were obtained on the day of randomization (start of study medication; n=810) or, if not available, at the screening visit (3 to 14 days before randomization; n=201), with 10-mL EDTA syringes used for sampling. Plasma was separated and stored at −80°C until assays were performed. Plasma levels of NT-proBNP were determined with Elecsys proBNP (Roche Diagnostics GmbH), a quantitative electrochemiluminescence immunoassay.14 The analytical detection limit of the assay was 5 pg/mL. The intra-assay coefficient of variation was 2.4% at 355 pg/mL and 1.8% at 4962 pg/mL; the interassay CVs were 2.9% at 355 pg/mL and 2.3% at 4962 pg/mL. Upper limits of normal of 100 pg/mL in men and 150 pg/mL in women are proposed by the manufacturer. Measurements were performed with the Elecsys 2010 immunoassay analyzer (Roche Diagnostics GmbH).
Myocardial marker levels were presented descriptively. Survival distributions were compared by use of a 2-sided log-rank test. The effect of baseline NT-proBNP was characterized by hazard ratios (and corresponding 95% CIs) on a Cox proportional-hazard model. Kaplan-Meier curves were generated, and 1-year mortality and annual hazard rates were calculated by stratifying according to baseline NT-proBNP levels above and below the median. Analyses were conducted separately for groups receiving carvedilol or placebo and for the total population.
Multivariate Cox proportional-hazards regression model was used to investigate the effect of NT-proBNP levels, taking into account the influence of other potentially prognostic factors (in particular, treatment group, left ventricular ejection fraction, age, sex, cause of heart failure, creatinine, systolic blood pressure). To test for interaction between treatment and NT-proBNP, an interaction term was added to the multivariate model (interaction term is 1 for carvedilol treatment and NT-proBNP above the median and 0 otherwise). Statistical significance was defined as P<0.05 (2 tailed).
Treatment effects with carvedilol in the COPERNICUS study have been reported recently.13 In total, 2289 patients (1133 placebo, 1156 carvedilol) were analyzed. The Kaplan-Meier estimate of 1-year cumulative mortality rate was 14.9% (18.5% in the placebo group compared with 11.4% in the carvedilol group). The risk of death was reduced by 35% (relative risk [RR], 0.65, 95% CI, 0.52 to 0.81) for patients treated with carvedilol. The corresponding annual hazard rates were 19.7% for placebo and 12.8% for carvedilol. In the NT-proBNP substudy, 1011 European patients (506 placebo, 505 carvedilol) were analyzed. Patient characteristics indicate that the substudy population was comparable to the total COPERNICUS study population regarding age, sex, cause of heart failure, left ventricular ejection fraction, heart rate, and other clinical findings (Table 1). In addition, overall survival distributions were similar in both the substudy and the main COPERNICUS trial. In the NT-proBNP subpopulation, the Kaplan-Meier estimate of 1-year cumulative mortality rate was 13.4% (18.1% in the placebo group compared with 8.5% in the carvedilol group). The risk of death was reduced by 38% (RR, 0.625; 95% CI, 0.40 to 0.99) for patients treated with carvedilol. The corresponding annual hazard rates were 18.1% for placebo and 11.3% for carvedilol.
The baseline NT-proBNP concentrations ranged from the detection limit to 35830 pg/mL, with a median of 1767 (25th to 75th percentiles, 748 to 3927 pg/mL) and a mean±SD of 3235±4392 pg/mL. In the substudy, a total of 78 patients died (48 placebo versus 30 carvedilol) during follow-up (median, 159 days; range, 1 to 488 days) (primary end point). We found that 180 patients died or were hospitalized for heart failure (101 placebo versus 79 carvedilol), 205 patients died or were hospitalized for cardiovascular reasons (115 placebo versus 90 carvedilol), and 293 patients died or were hospitalized for any reason (160 placebo versus 133 carvedilol) (secondary end points).
NT-proBNP and Clinical Events
Table 2 indicates event rates of the primary and all secondary end points of the COPERNICUS trial according to baseline NT-proBNP levels above or below the median of 1767 pg/mL, regardless of treatment. Baseline NT-proBNP levels above median were associated with an increased risk for all-cause mortality and all combined end points.
Kaplan-Meier analyses for all-cause mortality revealed 1-year mortality rates of 7.0% and 21.6% in patients with NT-proBNP levels ≤1767 and >1767 pg/mL, respectively. Thus, the risk of death was 2.7-fold higher for the patients with baseline NT-proBNP values above the median (95% CI, 1.7 to 4.3; P=0.0001, log-rank test; Figure 1).
The combined 1-year risk for all-cause mortality or hospitalization for heart failure was 18.9% compared with 38.0% in patients with below- or above-median NT-proBNP levels, respectively. Thus, the combined risk of death or hospitalization for heart failure was 2.5-fold higher for patients with baseline NT-proBNP values above the median (95% CI, 1.8 to 3.4; P=0.0001, log-rank test) (Figure 2).
The impact of various baseline variables on the end points of the study was investigated by univariate Cox regression analysis. Clinically relevant classes were defined for continuous variables to further characterize their impact on outcome. Besides NT-proBNP (regardless of whether continuous or categorized), systolic blood pressure (continuous and categorized), age (continuous and categorized), creatinine (continuous), treatment group, hospitalization for heart failure within the year before enrollment (recent hospitalization), and any high-risk feature (any patient with ascites, rales, or edema at randomization, who was hospitalized at screening or randomization, who was hospitalized >2 times within the past year or received intravenous positive inotropic agents or other intravenous drugs for heart failure within 14 days of randomization, or whose baseline left ventricular ejection was ≤15%) were shown to have a statistically significant effect on mortality (P=0.05 level). Left ventricular ejection fraction (continuous and categorized), cause of heart failure, and creatinine (categorized) were not significant on the P=0.05 level. However, left ventricular ejection fraction predicted the combined end point of all-cause mortality and hospitalizations for heart failure with P=0.010 and P=0.003 (continuous and categorized) and all other combined end points. Similarly, ischemic etiology predicted the combined end point of all-cause mortality and hospitalizations for heart failure with P=0.008. No impact of treatment with amiodarone or angiotensin type II receptor antagonists on the end points was observed. Predictive values of all other risk indicators were similar for the primary and all secondary end points. Risk ratios and probability values are indicated in Table 3.
Investigation of the prognostic baseline parameters with a stepwise-backward selection method in a multivariate Cox proportional-hazards regression using the same categories as for the univariate model revealed prognostic significance for all-cause mortality in a model containing NT-proBNP, treatment group, age, systolic blood pressure, recent hospitalization, and high-risk combination. NT-proBNP was a highly significant risk indicator in the model with continuous variables and in the model with categorized variables. Further relevant (P<0.05) prognostic variables for death in the multivariate model were treatment group, systolic blood pressure <100 mm Hg, age ≥65 years, recent hospitalization, and high-risk combination. Significant independent prognostic variables for combined death and heart failure hospitalizations were NT-proBNP classes, treatment group, systolic blood pressure <100 mm Hg, recent hospitalization, and high-risk combination (Table 4). Similar findings were observed for the other combined end points.
NT-proBNP and Treatment Effect
Subgroups were formed according to treatment with carvedilol or placebo and were again stratified according to median NT-proBNP. In patients with plasma NT-proBNP levels above the median, those receiving placebo incurred a 1-year mortality of 30.9%, whereas mortality was 13.8% in those receiving carvedilol (P=0.068, log-rank test). In patients entering the study with an NT-proBNP level below the median, the 1-year mortality rates were 10.2% in those receiving placebo and 3.4% in those receiving carvedilol (P=0.068, log-rank test). Therefore, administration of carvedilol in high NT-proBNP patients resulted in 17.1 fewer deaths per 100 patient-years compared with 6.8 fewer deaths per 100 patient-years in those receiving carvedilol who had lower NT-proBNP levels (Figure 3).
Interaction between treatment group and NT-proBNP level was further investigated by including an interaction term in the multivariate model. This analysis indicated that in the COPERNICUS NT-proBNP substudy patients, the difference in risk reduction of carvedilol treatment in patients with above- versus below-median NT-proBNP levels was not significant (P=0.93; RR, 0.96).
The COPERNICUS NT-proBNP substudy is the first study to investigate the prognostic value of the newly developed cardiac marker NT-proBNP in a large number of patients with severe CHF resulting from ischemic and nonischemic cardiomyopathy. In these high-risk CHF patients who were symptomatic at rest or on minimal exertion and had a left ventricular ejection fraction <25%, high NT-proBNP plasma levels on admission were strongly associated with an increased risk of all-cause mortality and heart failure during follow-up. The association of NT-proBNP levels with cardiac events was independent of treatment with placebo or carvedilol, cause of heart failure, systolic blood pressure, left ventricular ejection fraction, age, and recent hospitalization. The beneficial effect of carvedilol was independent of NT-proBNP levels. The combination of below-median pretreatment NT-proBNP levels and treatment with carvedilol was associated with an extremely low mortality risk of 3.4% in the first year. This rate contrasted to the high mortality rate of 30.9% in patients with above-median NT-proBNP and placebo treatment. The substudy results indicate that risk stratification with NT-proBNP adds important information to clinical risk predictors in patients with very severe heart failure, even those who were clinically euvolemic.
NT-proBNP as a Predictor of Prognosis in Severe CHF
Risk prediction in heart failure patients is a prerequisite for selecting treatment strategies, including cardiac transplantation. However, results from large trials have consistently revealed the limitations of common indicators used either singly or in combination.15 This relates also to left ventricular ejection fraction, complex arrhythmias, and heart failure symptoms. We anticipated that biochemical indexes and circulating neurohormone concentrations may add precision to the generally less objective clinical indicators. Several groups have demonstrated the usefulness but also the limitations of neurohormonal markers such as catecholamines and particularly natriuretic peptides, including atrial natriuretic peptide and BNP.1–4,16,17 BNP is more predictive than atrial natriuretic peptide for cardiac mortality in patients with chronic CHF and has provided prognostic information independently of other variables previously associated with poor outcome.17
In the Australia/New Zealand (ANZ) Heart Failure Study, the recently detected cardiac marker NT-proBNP was evaluated for prediction of all-cause mortality and heart failure in 297 patients with established ischemic left ventricular dysfunction (NYHA class II to III; left ventricular ejection fraction <45%) treated with carvedilol or placebo. Above-median NT-proBNP levels conferred >4.5-fold-increased risk of mortality. NT-proBNP was superior to left ventricular ejection fraction in predicting mortality and heart failure.12
The COPERNICUS NT-proBNP substudy is the first study to extend this important finding to patients with severe CHF and left ventricular ejection fraction <25% resulting from ischemic or nonischemic cardiomyopathy. The COPERNICUS patients represent a high-risk population with an overall 1-year placebo mortality rate of 18%. In the NT-proBNP substudy, the risk of dying was 2.7-fold higher in placebo patients with above-median baseline NT-proBNP concentration compared with patients with below-median NT-proBNP values. Thus, even in these severely compromised patients who were symptomatic at rest or on minimal exertion, NT-proBNP plasma concentrations could separate the population into subgroups with more or less favorable outcomes. The prognostic significance of NT-proBNP was independent of the thresholds used and was superior or similar to established risk indicators such as treatment modalities, cause of heart failure, left ventricular ejection fraction, age, systolic blood pressure, and recent hospitalization.
NT-proBNP as a Predictor of Response to Treatment
In an analysis of the Survival and Ventricular Enlargement (SAVE) data, neurohumoral activation at the time of hospital discharge failed to identify those patients who would derive benefit from treatment with β-blockers.18 In a previous report of the ANZ Carvedilol Study Group, above-median BNP concentrations combined with below-median norepinephrine levels predicted benefit from carvedilol treatment in ischemic left ventricular dysfunction.19 Recently, similar findings were reported for NT-proBNP and adrenomedullin in the same patient population. Carvedilol significantly halved the risk of heart failure hospitalization in patients with above-median NT-proBNP levels, whereas treatment effect in the patients with below-median NT-proBNP did not reach significance. In addition, the investigators reported statistically significant interactions of treatment with NT-proBNP for the prediction of all analyzed end points, indicating that the predictive power of the marker is influenced by treatment with carvedilol.12
In the present substudy, administration of carvedilol in high-NT-proBNP patients resulted in 17.1 fewer deaths per 100 patient-years compared with 6.8 fewer deaths per 100 patient-years in those with lower NT-proBNP levels receiving carvedilol. Both differences did not reach statistical significance. However, the difference in impact of 10.3 lives per 100 patient-years is at least noteworthy, even though further statistical analyses failed to show significant interactions between treatment group and NT-proBNP concentrations. However, our study was not primarily powered to detect a statistical significant difference of prognostic values of the marker in predefined subgroups. Thus, this finding is to be interpreted with caution.
The reasons for this possible difference from the ANZ study findings are unknown but may be due, at least in part, to differences in patient population. Heart failure in the ANZ study was less severe, and there was no overall survival benefit with carvedilol.
Thus, in COPERNICUS, NT-proBNP did not convincingly allow stratification into groups most likely to benefit from carvedilol therapy. Because it currently is not clear which patients will benefit most from such therapy, this medication should, according to the COPERNICUS data, be administered liberally in advanced CHF patients.
Clinical Implications and Limitations of the COPERNICUS NT-proBNP Substudy
The short duration of the study may have limited the statistical power of the prognostic findings. However, our study shows for the first time a powerful association of NT-proBNP with death and heart failure hospitalization in patients with severe heart failure who suffer from shortness of breath at rest or on minimal exertion. The combination of a low NT-proBNP level and treatment with carvedilol was associated with a high probability of event-free survival. Therefore, such patients appear to be appropriately treated with medical therapy. Patients with high NT-proBNP levels who are not treated with carvedilol aggregate on the opposite extreme of the risk scale. Their 1-year mortality rate of 31% could be halved by carvedilol treatment. Whether these patients with high levels of NT-proBNP may further benefit from alternative treatment options such as cardiac transplantation warrants further investigation.
So far, the general application of NT-proBNP in heart failure has been restricted by the fact that NT-proBNP blood levels and the discriminator values related only to the specific assays used for the different trials. In the present study, NT-proBNP was measured by a newly developed, highly sensitive and standardized electrochemiluminescence immunoassay. The availability of such a standardized and easy-to-use assay is prerequisite for general application of NT-proBNP as a biochemical marker in CHF patients. Under these conditions, it is anticipated that measurement of cardiac hormones will become part of the routine workup of heart failure patients and will be a useful tool for risk stratification of such patients.
The study and substudy were supported by a grant from Roche Pharmaceuticals, and NT-proBNP plasma level measurements were performed in a core laboratory at Roche Diagnostics GmbH. We gratefully appreciate the valuable and irreplaceable impact of Christoph Staiger, Juergen Spinke, and Matthias Baumann during planning, execution, and evaluation of the study.
Drs Packer, Coats, Fowler, Katus, Krum, Mohacsi, Rouleau, Tendera, and Castaigne served as consultants to Roche Pharmaceuticals. Drs Amann-Zalan and Hoersch are employees of or on committees at Roche Diagnostics GmbH.
Francis GS, Benedict C, Johnstone DE, et al. Comparison of neuroendocrine activation in patients with left ventricular dysfunction with and without congestive heart failure: a substudy of the Studies of Left Ventricular Dysfunction (SOLVD). Circulation. 1990; 82: 1724–1729.
Omland T, Aakvaag A, Bonarjee VV, et al. Plasma brain natriuretic peptide as an indicator of left ventricular systolic function and long-term survival after acute myocardial infarction: comparison with plasma atrial natriuretic peptide and N-terminal proatrial natriuretic peptide. Circulation. 1996; 93: 1963–1969.
Richards AM, Nicholls MG, Yandle TG, et al. Plasma N-terminal pro-brain natriuretic peptide and adrenomedullin: new neurohormonal predictors of left ventricular function and prognosis after myocardial infarction. Circulation. 1998; 97: 1921–1929.
Talwar S, Squire IB, Downie PF, et al. Profile of plasma N-terminal proBNP following acute myocardial infarction: correlation with left ventricular systolic dysfunction. Eur Heart J. 2000; 21: 1514–1521.
Richards AM, Doughty R, Nicholls MG, et al. Plasma N-terminal pro-brain natriuretic peptide and adrenomedullin: prognostic utility and prediction of benefit from carvedilol in chronic ischemic left ventricular dysfunction: Australia/New Zealand Heart Failure Group. J Am Coll Cardiol. 2001; 37: 1781–1787.
Karl J, Borgya A, Gallusser A, et al. Development of a novel, N-terminal-proBNP (NT-proBNP) assay with a low detection limit. Scand J Clin Lab Invest. 1999; 230 (suppl): 177–181.
Cohn JN, Johnson GR, Shabetai R, et al. Ejection fraction, peak exercise oxygen consumption, cardiothoracic ratio, ventricular arrhythmias, and plasma norepinephrine as determinants of prognosis in heart failure: the V-HeFT VA Cooperative Studies Group. Circulation. 1993; 87 (suppl VI): VI-5–VI-16.
Tsutamoto T, Wada A, Maeda K, et al. Attenuation of compensation of endogenous cardiac natriuretic peptide system in chronic heart failure: prognostic role of plasma brain natriuretic peptide concentration in patients with chronic symptomatic left ventricular dysfunction. Circulation. 1997; 96: 509–516.
Richards AM, Doughty R, Nicholls MG, et al. Neurohumoral prediction of benefit from carvedilol in ischemic left ventricular dysfunction. Australia-New Zealand Heart Failure Group. Circulation. 1999; 99: 786–792.