Author + information
- Received November 15, 2000
- Revision received April 13, 2001
- Accepted April 26, 2001
- Published online August 1, 2001.
- Brigitte Stanek, MDa,* (, )
- Bernhard Frey, MDa,
- Martin Hülsmann, MDa,
- Rudolf Berger, MDa,
- Barbara Sturm, MDa,
- Jeanette Strametz-Juranek, MDa,
- Jutta Bergler-Klein, MDa,
- Petra Mosera,
- Anja Bojic, MDa,
- Engelber Hartter, MD, PhDa and
- Richard Pacher, MDa
- ↵*Reprint requests and correspondence: Dr. Brigitte Stanek, Department of Cardiology, University of Vienna, A-1090 Vienna, Austria
The study assessed the relative predictive potency of neurohumoral factors in patients with advanced left ventricular (LV) dysfunction during neurohumoral blocking therapy.
The course of heart failure is characterized by progressive LV deterioration associated with an increase in cardiac (natriuretic peptides) and predominantly extracardiac (norepinephrine, big endothelin [big ET]) hormone plasma levels.
Plasma hormones were measured at baseline and months 3, 6, 12 and 24 in 91 patients with heart failure (left ventricular ejection fraction [LVEF] <25%) receiving 40 mg enalapril/day and double-blind atenolol (50 to 100 mg/day) or placebo. After the double-blind study phase, patients were followed up to four years. Stepwise multivariate regression analyses were performed with 10 variables (age, etiology, LVEF, symptom class, atenolol/placebo, norepinephrine, big ET, log aminoterminal atrial natriuretic peptide, log aminoterminal B-type natriuretic peptide [N-BNP] and log B-type natriuretic peptide [BNP]). During the study, the last values prior to patient death were used, and in survivors the last hormone level, New York Heart Association class and LVEF at month 24 were used.
Thirty-one patients died from a cardiovascular cause during follow-up. At baseline, log BNP plasma level (x2= 13.9, p = 0.0002), treatment allocation (x2= 9.5, p = 0.002) and LVEF (x2= 5.6, p = 0.017) were independently related to mortality. During the study, log BNP plasma level (x2= 21.3, p = 0.0001) remained the strongest predictive marker, with LVEF (x2= 11.2, p = 0.0008) log N-BNP plasma level (x2= 8.9, p = 0.0027) and treatment allocation (x2= 6.4, p = 0.0109) providing additional independent information.
In patients with advanced LV dysfunction receiving high-dose angiotensin-converting enzyme inhibitors and beta-blocker therapy BNP and N-BNP plasma levels are both independently related to mortality. This observation highlights the importance of these hormones and implies that they will likely emerge as a very useful blood test for detection of the progression of heart failure, even in the face of neurohumoral blocking therapy.
In the peripheral circulation of patients with heart failure, natriuretic peptides—systemic hormones produced by the heart—are abnormally elevated in conjunction with their aminoterminal (N-terminal) portions (1). As with other neurohumoral factors, these cardiac peptides increase in accordance with disease progression. Accordingly, their plasma levels are supposed to reflect the degree of cardiac jeopardy and injury (2–7). As B-type natriuretic peptide (BNP) is released from the myocardium of the ventricles, the BNP plasma level is a particularly sensitive marker of impending ventricular damage. As a hypothesis, the release of BNP into the circulation can be regarded as a backup provided by the ventricular myocytes as soon as atrial-natriuretic peptide (ANP) release fails to provide adequate compensation. Prognostic assessment in patients with heart failure recently identified plasma BNP as a potent mortality predictor (8).
It is well known that, in heart failure, hormones of extracardiac origin are also increased in the circulation, particularly norepinephrine (NE) and endothelin. In theory, the extracardiac hormones might reflect the overall peripheral response to cardiac impairment. Conventionally, the NE plasma level is used as an index of adrenergic activity, although spillover of NE from the nerve endings to the bloodstream is very small. Similarly, plasma endothelin, derived from the vasculature, is rapidly cleared and, in instances of paracrine action, may never reach the circulation. In contrast, big endothelin (ET), its precursor without biologic activity, circulates in higher concentration and integrates the secretory activity of endothelial linings and other endothelin-producing cells. The prognostic value of big ET in patients with advanced left ventricular (LV) dysfunction is well established (9–11). Thus, both high levels of cardiac natriuretic peptides and high levels of endothelin (or its prohormone) are generally capable of identifying a very high mortality group.
Reduction of deleterious neurohumoral factors is believed to play a key role in mediating the efficacy of beta-blocker therapy for heart failure. Carvedilol reduced plasma endothelin levels in a small sample of patients, and this decrement was significantly related to functional and hemodynamic treatment success (12). Carvedilol also interfered with the predictive value of prestudy BNP levels (13). Moreover, pilot data have shown that vasodilator treatment can be titrated to reduce BNP concentrations toward a normal range (14). A recent study showed also that circulating aminoterminal-BNP (N-BNP) concentrations can be reduced by intensification of drug therapy in heart failure (15). Employing drug treatment guided by plasma N-BNP concentrations, the total number of cardiovascular events compared with clinically guided treatment by the same range of therapies was reduced. As N-BNP concentrations reflect severity of LV hemodynamic dysfunction the investigators assumed that treatment that reduces N-BNP concentrations should unload the LV and reduce both LV wall stress and myocardial oxygen requirements, thereby possibly translating into a slowing in the decline of myocardial function. A potential weakness of the above study (15)was the low rate of beta-blocker use. Hence, the usefulness of N-BNP plasma concentrations as a guide to beta-blocker administration remained unclear.
So far, the relative prognostic importance of the two (antagonistic) neurohumoral factor types has not been fully elucidated. More importantly, it has been speculated that these markers might modify their role under the influence of long-term beta-blocker treatment. Therefore, we performed a neurohumoral substudy including repetitive measurements of NE, big ET, BNP, aminoterminal-ANP (N-ANP) and N-BNP plasma levels during prospective randomized double-blind treatment with the beta-1 selective beta-blocker atenolol or a placebo.
The aim of this study was to investigate the relative prognostic potency of cardiac versus extracardiac markers to identify patients at highest mortality risk both before and during beta-blocker treatment.
This is a neurohumoral substudy of a randomized trial of 100 heart failure patients with left ventricular ejection fraction (LVEF) <25% (by radionuclide ventriculography) who received double-blind treatment with atenolol (50 to 100 mg/day, mean dosage 89 mg/day) or placebo. Background therapy included digitalis and 40 mg enalapril/day in all patients who were outpatients at our institution and were clinically stable for more than four weeks. Prior to randomization, a challenge test with 12.5 mg atenolol was performed. Only patients whose heart rate remained >60 beats/min and whose systolic blood pressure remained >90 mm Hg 5 h after this dose were included. Informed consent was obtained from all patients for participation in the study according to a protocol approved by the local ethics committee. The original trial was stopped early after 395 ± 266 days at interim analysis due to a significant treatment benefit of atenolol (worsening heart failure or death) over placebo (16).
The end point of the neurohumoral substudy was cardiovascular death at four-year follow-up. During this period 28 patients who were randomized to atenolol and who completed the double-blind study without reaching an end point were continued on open atenolol added onto their previous background therapy, including 40 mg enalapril. Of 20 study-completing patients randomized to placebo, 10 completed 24 months of the double-blind study without reaching an end point. These patients received open atenolol during subsequent clinical visits, if tolerated. The remaining 10 patients who completed the double-blind phase of the study earlier were denied open atenolol therapy until month 24. Of the remaining 29 patients in the placebo group who dropped from double-blind treatment because of serious clinical events, worsening heart failure or death, 4 patients were dropped because of ventricular tachycardia. They received an automatic implanted cardioverter defibrillator and additional beta-blocker therapy, as clinically appropriate. Five patients developed worsening heart failure. After recompensation these patients were also treated with beta-blockers because they needed additional oral therapy.
These nine patients of the placebo group were excluded from the prognostic evaluation in the neurohumoral substudy because they received beta-blocker therapy prior to 24 months (during the double-blind phase of the study) for safety reasons. This decision was based on the concern that these nine patients would have biased the result of the neurohumoral substudy, which was designed to evaluate the contribution of randomized treatment (two-year treatment with atenolol vs. placebo) in the prognostic process. Plasma levels of N-ANP, N-BNP, BNP, NE and big ET were determined at months 0, 3, 6, 12 and 24 in addition to LVEF (at months 0, 12, 24) and symptoms (New York Heart Association [NYHA] functional class) to define independent mortality risk factors.
Blood samples were drawn from an antecubital vein after 30 min of rest, transferred into chilled tubes, placed on ice and centrifuged at 4°C. Plasma was frozen at −20°C until assay. For determination of N-ANP (by enzyme-linked immunosorbent assay [ELISA]), N-BNP (by ELISA), BNP (by radioimmunoassy [RIA]), NE (by high-pressure liquid chromatography) and big ET levels (by RIA) commercially available assay kits were used, all purchased from Biomedica, Austria.
Continuous variables are expressed as mean ± SD. For group comparison of continuous variables, the two-tailed Student ttest was used; for categorical variables, the Fisher exact test was used. The Spearman rank correlation coefficient was computed to assess correlation between continuous variables. Repeated measures analysis of variance (ANOVA) was used to analyze the time course of hormone plasma levels during treatment in the study-completing cohorts, and ANOVA, with Bonferroni adjustments was used, if appropriate. A Cox proportional hazards regression analysis was performed to identify independent predictors of four-year mortality. The model was built stepwise; the p value for entering and staying in the model was set at 0.05. Because BNP, N-BNP and N-ANP were not normally distributed, log BNP, log N-BNP and log N-ANP plasma levels were used in the multivariate analysis. Ten variables were entered: age; ischemic/nonischemic heart failure; LVEF; NYHA functional class; NE plasma level; big ET plasma level; log BNP plasma level; log N-ANP plasma level; log N-BNP plasma level; allocation to atenolol/placebo in the original trial: a) before and b) during the study. For prognostic evaluation during the study period, the last available values of NYHA functional class, LVEF and hormones prior to death were used, and in survivors the values of NYHA functional class, LVEF and hormones at month 24 were used.
Kaplan-Meier survival estimates and the log-rank test were used to compare survival between patients with BNP plasma levels ≥50 fmol/ml and <50 fmol/ml at time of randomization. A p value <0.05 was considered statistically significant. For all statistical analyses, SAS (SAS Institute, Cary, North Carolina) version 7 statistical package software was used.
The original study population included 100 patients, randomized to atenolol (51 patients) and placebo (49 patients) of whom 78% were in NYHA functional class II, 13% in functional class III and 2% in functional class IV while treated with digoxin and 40 mg/day enalapril. All hormone levels were abnormally elevated at baseline and were similar in both treatment groups (Table 1).
Correlations between hormones and LVEF at baseline
All hormones correlated positively with each other, albeit the coefficients varied widely. As expected, the closest relationships were detected among BNP, N-BNP and N-ANP (Table 2). In addition, inverse correlations were detected between the values of LVEF and the plasma concentrations of BNP (r = −0.37, p < 0.001), N-BNP (r = −0.32, p < 0.01), N-ANP (r = −0.26, p < 0.01) and NE (r = −0.22, p < 0.05).
Plasma hormone levels as influenced by atenolol and placebo
Table 3summarizes the plasma levels of N-BNP, N-ANP, BNP, big ET and NE obtained during treatment with atenolol and placebo. Atenolol resulted in a decrease in N-ANP plasma levels after 6 months (p < 0.01) and in a decrease in N-BNP plasma levels after 6, 12 and 24 months (all p < 0.01), whereas placebo did not change any hormone level significantly.
Ninety-one patients were available for the prognostic analysis. Nine patients, all in the placebo group, were excluded from the prognostic evaluation because they did not fulfill the criteria set for the neurohumoral substudy (see Methods section). Thus, the placebo group of the substudy comprised 40 patients (36 men, 4 women) age 52 ± 10 years, of whom 24 patients had idiopathic, 10 had ischemic and 6 had other etiologies of heart failure. Thirty patients were in NYHA functional class II and 10 in NYHA functional class III, with a mean LVEF of 17 ± 6%. At closing, 60 patients were still alive, whereas 31 had died. Of these, 10 patients were randomized to atenolol in the original trial, and 21 patients were randomized to placebo. Causes of death are given in Table 4.
Effect of atenolol and placebo by outcome
In survivors, LVEF increased over time in both treatment groups, from an average of 18% to 30% on atenolol and from 20% to 25% on placebo. The N-BNP levels decreased in parallel, but in survivors of the placebo group, NE plasma levels increased slightly. No significant changes occurred in nonsurvivors of the atenolol group. In the placebo group, however, BNP plasma levels increased fourfold in nonsurvivors (Table 5).
Multivariate stepwise Cox regression analysis and risk stratification
At baseline, 3 of 10 factors (see Methods section) were independently related to four-year mortality: 1) log BNP plasma level (x2= 13.9, p = 0.0002); 2) treatment allocation (x2= 9.5, p = 0.002); and 3) LVEF (x2= 5.6, p = 0.017). Mortality was significantly higher (log rank p < 0.0004) in 30 patients with baseline BNP levels ≥50 fmol/ml (17 deaths) than in 61 patients with BNP levels below this cut-off point (14 deaths) (Fig. 1). During the study, 4 of 10 factors (see Methods section) were independently related to four-year mortality: 1) log BNP (last available plasma level, x2= 21.3, p = 0.0001); 2) LVEF (last available value, x2= 11.2, p = 0.0008); 3) log N-BNP (last available plasma level, x2= 8.9, p = 0.0027); and 4) treatment allocation (x2= 6.4, p = 0.011).
In the next step, patients were dichotomized according to a cut-off value of 300 fmol/ml N-BNP (last available plasma level). Fifty patients had lower N-BNP plasma levels. Of these, 44 were alive after four years, whereas six had died. In contrast, of 41 patients with N-BNP levels ≥300 fmol/ml, 16 were alive and 25 had died. In this group, 22 patients had plasma BNP (last available levels) of ≥50 fmol/ml, and 19 patients had lower plasma BNP levels. In the group defined by N-BNP and BNP plasma levels in excess of the respective cut-off points (n = 22), 19 patients died and 3 were alive, corresponding to a 86% correct prediction of mortality. In the group defined by N-BNP in excess of the cut-off point in conjunction with BNP plasma levels <50 fmol/ml), 6 patients had died and 13 patients were alive, corresponding to a 68% correct prediction of survival.
In the current study, various neurohumoral plasma levels were evaluated in regard to their relative prognostic impact for patients with chronic heart failure. All patients received randomized double-blind treatment with atenolol or placebo on top of high-dose enalapril. However, the original study stopped early because of a significant treatment benefit of atenolol after an average of 295 days (16). Before the study the plasma levels of cardiac (N-ANP, N-BNP, BNP) as well as extracardiac hormones (NE, big ET) were abnormally increased as expected. Thereby the diagnostic significance of the relation to the degree of LV dysfunction was obvious for the cardiac peptides (BNP, N-BNP, and N-ANP in descending order), whereas big ET plasma levels did not correlate with LVEF. The majority of the patients were in NYHA functional class II, however, despite low LVEF values. It was previously shown that big ET plasma levels (reflecting endothelin production) are only moderately elevated in such patients (9).
Significant modulations of the pronatriuretic plasma levels were observed in response to atenolol, with N-ANP and N-BNP levels falling in tandem after six months. Subsequently, the lower N-BNP level was well sustained, while N-ANP returned to baseline levels. This reflects the concept that BNP sensitively responds to physiologic stimuli from the ventricles and acts as a backup hormone to ANP only if cardiac filling pressures remain high. In the current study atenolol did not affect plasma NE levels, in contrast to a previous report (17). However, our patients had considerably lower plasma NE levels, which might be a result of high-dose enalapril treatment.
This neutral finding with atenolol is of particular interest because other beta-blockers—for example, bucindolol—remarkably diminish plasma NE levels over time (18). An emerging concept is that selective beta1-blockers, which fail to block the beta2-presynaptic receptor, are more likely to have little or no effect on circulating NE levels because they continue to facilitate NE release. Conversely, potent beta1/beta2receptor blockade, such as produced by bucindolol may result in a dramatic decline in plasma NE levels. It was also suggested that carvedilol lowers cardiac adrenergic drive, whereas metoprolol does not (19–21). However, although carvedilol may be more cardioselective, beta-blockers such as bucindolol and propranolol may be too potent in this regard. It is possible that this effect was the reason that bucindolol failed to improve survival in patients with heart failure, not unlike moxonidine, which centrally inhibits sympathetic outflow and also suppresses circulating NE levels by this mechanism (22,23).
Hormone-hormone relations between plasma BNP and the aminopeptides N-ANP and N-BNP were close at entry, confirming their part as portions of the natriuretic peptides during biosynthesis. We sought to discover whether using the precursors that circulate in higher molar concentration due to a longer plasma half-life would offer prognostic options. Of the 91 patients, 31 died during four-year follow-up, which was less than previously reported in similar cohorts (24).
A key finding of the study was that BNP and N-BNP were substantially lower throughout in patients who survived with an improved ejection fraction, whether they were treated with atenolol or not. The multivariate Cox regression analysis of repetitive hormone levels obtained during the study confirmed the superiority of BNP and N-BNP levels as an index to identify patients at highest mortality risk. This observation highlights the importance of BNP or N-BNP and implies that it will likely emerge as a very useful blood test for detection of the progression of heart failure, even in the face of neurohumoral blocking therapy. These cardiac (ventricular) hormones appear to reflect the hidden extent of myocardial jeopardy and damage more accurately than do previously used extracardiac markers (e.g., endothelin or big ET levels). Nevertheless, a benefit of carvedilol on the endothelin system was reported by Krum et al. (12)to the effect that, in patients who improved clinically with carvedilol, endothelin plasma levels decreased. However, to interpret this finding, direct inhibition of endothelin synthesis by beta-blockade, as shown with selective and nonselective beta-blockers in vitro, has to be taken into account (25). This drug-induced effect could in theory invalidate the prognostic significance of endothelin (or big ET) plasma levels.
The findings concur well with similarly sized studies relating BNP plasma levels to cardiac function and prognosis (8,26). In the 1997 study by Tsutamoto et al. (8)plasma levels of BNP were clearly superior to ANP in predicting two-year mortality in 25 of 85 patients with LVEF <45%. Mean LVEF in nonsurvivors was 24%, however, indicating that the patients had considerably less LV damage. Selvais et al. (26)reported that both BNP and N-ANP plasma levels of 101 ambulatory patients were related to the two-year prognosis after the NYHA classification. However, few studies considered the impact of treatment. Richards et al. (13)reported the effect of carvedilol on the prognostic value of plasma ANP and BNP levels. Although patients with levels above the median generally had a poorer prognosis, this association was no longer significant when survival curves of patients receiving carvedilol and patients receiving placebo were compared. This suggested a positive impact of beta-blocker treatment in patients with (ischemic) heart failure, particularly in the presence of high ANP and BNP levels, probably by reducing the stimuli responsible for increased cardiac peptide production.
Recent attention has focused on the problem of how antagonistic neurohumoral systems (e.g., cardiac peptides and endothelin) are interlinked in the prognostic setting of heart failure. At first glance, enhanced endothelin production might simply reflect the extracardiac response (including the lungs) to cardiac dysfunction, whereas high BNP plasma levels originating from the myocardium might reflect adverse changes in the ventricular wall directly. The heart itself is a target organ for endothelin; endothelin favors LV remodeling on one side and stimulates BNP synthesis on the other (27). Synthesis of BNP and its augmented release into the circulation could represent an emergency aid provided by the damaged ventricles. However, the BNP receptor coupled to guanylate cyclase is downregulated in advanced LV dysfunction (8,26). Therefore, the actions of this cardiac peptide, although suggesting a favorable compensatory role in cardiac injury, are probably impaired (28). These possibilities might account for the finding that high plasma levels of BNP—despite beneficial intrinsic actions of this peptide—have negative prognostic importance.
Most prognostic studies designed to identify the predictive value of clinical variables have been derived from large multicenter heart failure trials. In contrast, the current study evaluated only 91 patients comprising a clinically heterogenous patient population. Regression analysis showed, however, that the power of BNP and N-BNP plasma levels to predict mortality of this cohort was independent of differences in clinical variables (e.g., etiology of heart failure or NYHA functional class) and provided additional information to LVEF. Nevertheless, given the small sample size, the hypothesis generated in this study will need to be tested in a larger number of patients.
The study suggests a role for BNP and N-BNP plasma levels not only for diagnosis but also for therapeutic monitoring in patients receiving combined neurohumoral therapy with angiotensin-converting enzyme inhibitors and beta-blockers.
We are grateful to Eva Moser for her expert technical assistance and management skills and to Sabine Kopetzky for the artwork.
☆ This study was supported by a grant of the Ö;-12usterreichische Nationalbank (Jubiläumsfonds).
- analysis of variance
- atrial natriuretic peptide
- B-type natriuretic peptide
- enzyme-linked immunosorbent assay
- left ventricle, ventricular
- left ventricular ejection fraction
- aminoterminal atrial natriuretic peptide
- aminoterminal B-type natriuretic peptide
- New York Heart Association
- Received November 15, 2000.
- Revision received April 13, 2001.
- Accepted April 26, 2001.
- American College of Cardiology
- Yasue H,
- Yoshimura M,
- Sumida H,
- et al.
- Davidson N.C,
- Naas A.A,
- Hanson J.K,
- Kennedy N.S.J,
- Coutie W.J,
- Struthers A.D
- Yamamoto K,
- Burnett J.C,
- Jougasaki M,
- et al.
- Daggubati S,
- Parks J.R,
- Overton R.M,
- Cintron G,
- Schocken D.D,
- Vesely D.L
- Tsutamoto T,
- Wada A,
- Maeda K,
- et al.
- Pacher R,
- Stanek B,
- Hülsmann M,
- et al.
- Hülsmann M,
- Stanek B,
- Frey B,
- et al.
- Australia-New Zealand Heart Failure Group,
- Richards A.M,
- Doughty R,
- Nicholls M.G,
- et al.
- Murdoch D.R,
- McDonagh T.A,
- Byme J,
- et al.
- Eichhorn E.J,
- Heesch C.M,
- Barnett J.H,
- et al.
- Gilbert E.M,
- Abraham W.T,
- Olsen S,
- et al.
- Newton G.E,
- Parker J.D
- Swedberg K,
- Bergh C.H,
- Dickstein K,
- McNay J,
- Steinberg M
- Bruneau B.G,
- Piazza L.A,
- DeBold A.J