β-Adrenergic Blocking Agents in Heart Failure: Benefits of Vasodilating and Nonvasodilating Agents According to Patients' Characteristics: A Meta-analysis of Clinical Trials

Background  In patients with heart failure, β-adrenergic blocking agents reduce overall and cardiovascular mortality. This meta-analysis aimed at clarifying their effect on sudden death, the magnitude of their benefit according to the cause of heart failure, and whether there is any difference between vasodilating and nonvasodilating agents.

Methods  Randomized, clinical trials were included if they evaluated a β-adrenergic blocking agent without intrinsic sympathomimetic activity, included a control group receiving placebo or standard treatment, evaluated mortality on an intention-to-treat basis, and lasted at least 8 weeks.

Results  Twenty-one trials with 5849 patients (3130 receiving β-blockers) were included. Median length of treatment was 6 months. Most patients had mild or moderate heart failure and were treated with angiotensin-converting enzyme inhibitors, diuretics, and digitalis. The β-blockers significantly reduced overall mortality, cardiovascular mortality, and mortality due to pump failure and sudden death by 34% to 39%. The decrease in overall mortality in patients with ischemic heart disease (IHD) (30%) was no different from that among patients with non-IHD (26%) (P=.08). The reduction in overall mortality was greater with vasodilating than with nonvasodilating agents (45% vs 27%; P=.007), particularly in patients without IHD (62%), compared with those with IHD (22%; P=.03).

Conclusions  In patients with heart failure, β-blockers reduce total and cardiovascular mortality at the expense of a decrease in mortality due to pump failure and sudden death. The magnitude of the benefit is similar in patients with IHD and in those with non-IHD. Vasodilating β-blockers have a greater effect on overall mortality than nonvasodilating agents, particularly in patients with non-IHD.

DESPITE MAJOR advances in knowledge about heart failure, it remains a serious public health problem, with a growing prevalence of 0.2% to 4% in the general population and about 10% in patients older than 80 years.¹ Case-fatality rate is 10% to 20% within 1 year of diagnosis and 50% or more within 5 years.^1,2 Although death rates due to other cardiovascular diseases are falling, advanced heart failure progresses to rapid deterioration and is a cause of rising mortality rates.^3-6 This has stimulated research into a more intensive treatment of less advanced stages of the disease and treatments based on new pathophysiological targets.

In patients with heart failure, the renin-angiotensin and sympathetic nervous systems are activated and contribute to progressive deterioration of the patient's condition.^7,8 Long-term sympathetic activation has detrimental effects on the myocardium, through direct myocardial toxic effects and arrhythmogenesis caused by excess of catecholamines, and through indirect stimulation of the renin-angiotensin system.^9-11 The results of several clinical trials suggest that β-adrenergic blocking agents improve symptoms, left ventricular function, and the rate of hospital admission.^12-21 For carvedilol, a beneficial effect on mortality has also been reported.²² Three meta-analyses^23-25 have indicated that β-blockers reduce overall and cardiovascular mortality. However, it is not known whether they reduce sudden death, whether their efficacy is influenced by the cause of heart failure, or whether there is any difference in the effect of vasodilating and nonvasodilating agents. Our study addressed these questions, including additional information from the recently published Cardiac Insufficiency Bisoprolol Study II (CIBIS-II) trial.²⁶

Clinical trials were identified through a computerized bibliographic search of the MEDLINE (from January 1, 1966, to January 31, 1998) and EMBASE (from January 1, 1974, to May 30, 1997) databases and by browsing the bibliography from review articles on heart failure. A letter, asking about any unpublished or unknown trial on β-blockers in heart failure, was also sent to the first author of identified clinical trials, to experts in heart failure, and to the pharmaceutical companies manufacturing β-blocking agents. Furthermore, since January 1998, we have been following the progress and possible termination of a number of ongoing clinical trials.

Clinical trials were included if they (1) were randomized and tested a β-adrenergic blocking agent devoid of intrinsic sympathomimetic activity in patients with heart failure; (2) had a control group treated with placebo or standard treatment; (3) had a parallel or crossover design (from the latter, only the results of the first part of the study before the crossover were included); (4) evaluated mortality on an intention-to-treat basis; and (5) had a minimum duration of 8 weeks. Trials with and without run-in phase were included.

Three investigators (A.A., J.M.A., X.V.) independently considered the inclusion of each trial without knowledge of the title, authors, and publishing journal. Differences were resolved by consensus. The information on each trial included in the meta-analysis usually was obtained from the corresponding publication. For each trial, the following information was independently collected by two of us (S.B., A.A.), using a standard form: sex, age, New York Heart Association (NYHA) functional classification, ejection fraction, standard treatments, duration of the study, β-adrenergic agent tested, number of withdrawals and deaths during the run-in phase, and mortality during the double-blind phase of each trial. When some of these items were missing in the article, the first investigator of the study was asked for them. When information on a particular variable ultimately was not obtained, the trial was excluded from the analysis of that particular variable.

Heterogeneity of treatment effects between studies was tested by using a χ² test for heterogeneity.²⁷ Relative risks (RR) and their 95% confidence intervals (CIs) were calculated using the DerSimonian and Laird random-effects methods²⁸ with the Rev-man program developed by the Cochrane Collaboration.²⁸ The random effects model was chosen because, in absence of heterogeneity, the results are the same as those obtained with the fixed effects model, and they can be more confidently generalized, as they are also more conservative. Comparisons between relative risk estimates were performed using the method described by Schlesselman.²⁹

Thirty-nine studies^{12-22,26,30-56} were evaluated, but only 21^{12-21,26,30-39} met the preestablished criteria and were included in the study (Table 1). Four studies were excluded because they lasted less than 8 weeks^40-43; 6 studies because they did not include a control group receiving placebo or standard treatment^44-49; 1 study because it was a subanalysis of patients with heart failure in the β-Blocker Heart Attack Trial⁵⁰; 1 study because labetalol hydrochloride, the agent used, has some intrinsic sympathomimetic activity⁵¹; 1 study because the assignment procedure was suboptimal⁵²; 1 study because it only described the design of a study⁵³; 1 study because it did not report on mortality⁵⁴; and 2 studies because of discrepancies in the numerical data.^55,56 Finally, 1 was the summary report²² of 4 randomized trials with carvedilol,^30,31,33,38 and its results were used only when data were not available from the individual trials.

A total of 5849 patients were studied (75% male; 3130 randomized to β-blocking agents and 2719 to the control group). The median duration of treatment was 6 months (range, 3-32 months). In 8 trials, the drug under study was carvedilol^{17,18,21,30-33,38}; in 7 studies, metoprolol (metoprolol tartrate in three studies^15,16,19; nonspecified salt in the others)^{12,14-16,19,34,36}; in 3 studies, bucindolol hydrochloride^13,35,37; in 2 studies, bisoprolol fumarate^20,26; and in 1 study, propranolol hydrochloride³⁹(Table 1). Eleven trials (1839 patients, of whom 1658 were included in trials with carvedilol) tested a β-blocker with vasodilating properties, and 10 (4010 patients) tested a nonvasodilating agent (3288 patients in trials with bisoprolol).

Heart failure due to ischemic heart disease (IHD) resulted in 3 trials,^16,21,39 to non-IHD (idiopathic) in 8,^{12,14,15,18,19,34,36,37} and to mixed in 9^{13,17,20,26,30,32,33,35,38}; cause was not stated in 1 study.³¹ In total, heart failure was judged to be ischemic in 2841 patients (48.6%) and nonischemic in 2903 (49.6%). Patients with NYHA class II or III disease constituted 88.7%. In trials that included information on concomitant treatment, 70% to 100% of patients received an angiotensin-converting enzyme (ACE) inhibitor (15 trials); 75% to 100%, a diuretic (14 trials); and 34% to 100%, digitalis (16 trials).

Seven hundred eighty-three patients died, 333 (10.6%) of 3130 treated with β-blockers, and 450 (16.6%) of 2719 in the control group (RR, 0.71; 95% CI, 0.63-0.80) (Table 2). Thus, 1 death can be prevented by treating 17 patients for 6 months. In 14 trials, a trend toward reduced mortality was seen, but only in 3 was this statistically significant (Figure 1). Including deaths during the run-in phase of trials (n=15) in the analysis did not materially change the estimate of mortality reduction (OR, 0.71; 95% CI, 0.63-0.80).

Information on cardiovascular mortality was obtained from 18 studies.^{12-21,26,30,32,34-36,38,39} There were 262 (9.0%) of 2913 cardiovascular deaths among patients treated with β-blockers, compared with 347 (13.7%) of 2530 in the control group (RR, 0.71; 95% CI, 0.59-0.86). This corresponds to 1 cardiac death prevented for every 21 patients treated. Reduced cardiovascular mortality was the result of a decrease in the number of deaths due to heart failure (RR, 0.66 [95% CI, 0.47-0.92]) and sudden death (RR, 0.70 [95% CI, 0.54-0.89]). There were few deaths due to myocardial infarction (12 vs 14; RR, 0.81 [95% CI, 0.38-1.72]) and other cardiovascular causes (41 vs 45; RR, 0.87 [95% CI, 0.48-1.57]).

The inclusion of the CIBIS-II trial did not change the estimates of overall, cardiovascular, or heart failure mortality (Table 2), but it did increase the estimate of the benefit on sudden death.

Both groups of drugs reduced overall mortality, and this effect was greater with vasodilating agents (RR, 0.55 [95% CI, 0.38-0.78] vs RR, 0.73 [95% CI, 0.64-0.83]; P=.007) (Table 3 and Table 4). The point estimate of the reduction in cardiovascular mortality was also greater with vasodilating agents (RR, 0.50 [95% CI, 0.22-1.17]) than with nonvasodilating agents (RR, 0.74 [95% CI, 0.64-0.86]), but the CIs were wide, and the difference was not statistically significant (P=.14). A similar nonsignificant trend was seen when the effect of vasodilating (RR, 0.33 [95% CI, 0.08-1.35]) and nonvasodilating agents (RR, 0.72 [95% CI, 0.51-1.01]) (P=.24) on mortality due to pump failure was considered. Both groups of drugs similarly reduced mortality due to sudden death (RR, 0.63 [95% CI, 0.35-1.12] vs RR, 0.76 [95% CI, 0.53-1.10]).

Before the publication of the CIBIS-II trial, a trend toward a greater protective effect of vasodilating agents on sudden death was seen (Table 5), but this disappeared when the CIBIS-II was included.

Information on the causes of heart failure was available from 15 trials.^{12,14-16,18-21,26,32,34-37,39} Reduction in overall mortality with β-blockers was slightly greater in patients with IHD (RR, 0.70 [95% CI, 0.60-0.81]) (Table 6) than in those with non-IHD (RR, 0.74 [95% CI, 0.60-0.91]) (Table 7), but this difference was not statistically significant (P=.08). In patients with IHD, overall mortality was similarly reduced by vasodilating (RR, 0.61 [95% CI, 0.35-1.06]) and nonvasodilating agents (RR, 0.71 [95% CI, 0.60-0.83]) (Table 5). In patients with non-IHD, the decrease in overall mortality was greater with vasodilating (RR, 0.38 [95% CI, 0.18-0.83]) than with nonvasodilating β-blockers (RR, 0.78 [95% CI, 0.63-0.96]; P=.03) (Table 5).

Four trials gave information on all-cause hospital admissions,^21,26,34,36 and 9 gave information on hospitalization for heart failure.^{15,18,20,21,26,30,34,36,38} The β-blockers reduced the rate of all-cause hospital admissions by 15% (RR, 0.85 [95% CI, 0.77-0.92]), and the rate of hospital admissions for heart failure by 33% (RR, 0.67 [95% CI, 0.58-0.77]). Vasodilating (RR, 0.64 [95% CI, 0.43-0.94]) and nonvasodilating agents (RR, 0.67 [95% CI, 0.57-0.78]) similarly reduced hospitalization for heart failure (Table 5).

Previously published meta-analyses, which had each included from 3023 to 3141 patients, had already suggested a protective effect of β-blocking agents on mortality due to heart failure.^23-25 Our study, which includes 5849 patients, confirms previous estimates^23-25 of the benefit of β-adrenergic blocking agents on total, cardiovascular, and pump failure mortality in patients with heart failure. In addition, our results indicate that reduced cardiovascular mortality was the result of reduced mortality due to heart failure and sudden death, that the benefits in terms of mortality were of the same order in patients with IHD and non-IHD, and that vasodilating agents are more effective than nonvasodilating agents.

The results of several clinical trials had suggested that patients with heart failure resulting from IHD did not benefit from β-blockers. In contrast, in 1 meta-analysis,²⁴ overall mortality was similarly reduced in patients with IHD and non-IHD heart failure. Our results show that the magnitude of the benefit of β-blockers in patients with IHD does not differ from that in patients with non-IHD.

Although β-adrenergic blocking agents are a relatively heterogeneous group of drugs, pharmacological differences rarely translate into therapeutic differences. However, in patients with heart failure, the vasodilating effect of certain β-blockers would compensate for the initial negative inotropic effect,^57,58 and it has been suggested that this, rather than β-adrenergic blockade, would account for their beneficial effect in heart failure.^59,60 In our analysis, vasodilating and nonvasodilating β-blockers reduced overall mortality, but this effect was greater with vasodilating agents, and mainly in patients with non-IHD cardiomyopathy. Our results suggest that the effect of vasodilating β-blockers could be greater on cardiovascular mortality and on mortality due to progressive pump failure. On the other hand, although a relative advantage of vasodilating β-blockers on sudden death had been suggested,²⁴ adding the results of the CIBIS-II trial suggests that the latter do not significantly differ from nonvasodilating β-blockers in this respect.

A previously published meta-analysis²⁵ had suggested a greater effect of nonselective agents on overall mortality compared with the selective agents. These results are in accord with ours, because, of 1600 patients treated with nonselective β-blocking agents and included in that meta-analysis, 1564 had received carvedilol or bucindolol, which are nonselective and vasodilating. The results of a recently early stopped trial with metoprolol, where a 34% reduction in mortality was seen,⁶¹ are also in accord with our results regarding selective nonvasodilating β-blockers.

We suggest that in patients with non-IHD, vasodilating β-blockers are more effective, and that the benefits of vasodilating and nonvasodilating β-blockers may be the result of different mechanisms. Although the exact mechanisms by which these drugs benefit patients with heart failure are unknown, it seems that the vasodilator component would be more important in reducing mortality due to heart failure, whereas the β-blocker component (or the antiarrhythmic effect) would mainly contribute to reducing sudden death.

In conclusion, β-blockers reduce total and cardiovascular mortality, as a result of a decrease in pump failure mortality and sudden death. Our results also show that the effect of β-blockers in patients with IHD is at least equal to that in patients with non-IHD. Compared with nonvasodilating agents, vasodilating β-blockers have a greater effect on overall mortality, particularly in patients with non-IHD. This may be attributable to a greater effect of vasodilating β-blockers on cardiac and pump failure mortality, rather than on sudden death.

Accepted for publication July 1, 1999.

This study was supported by grant 042996 from the Catalan Agency for Health Technology Assessment, Barcelona, Spain.

We deeply acknowledge additional information on the trials provided by Wilbert S. Aronow, MD; Eric J. Eichhorn, MD; John G. F. Cleland, MD; Jay N. Cohn, MD; Henry J. Dargie, MD; Leif Erhardt, MD; Marc A. Pfeffer, MD; Willem J. Remme, MD; and Karl Swedberg, MD.

Reprints: Joan-Ramon Laporte, MD, Fundació Institut Català de Farmacologia, Hospitals Vall d'Hebron, 08035, Barcelona, Spain (e-mail: jrl@icf.uab.es).