Beta-blockers are a cornerstone treatment of patients with heart failure (HF) (1). Randomized trials with carvedilol (U.S. Carvedilol Study (2) and COPERNICUS [Carvedilol Prospective Randomized Cumulative Survival] [(3),]), metoprolol (MERIT-HF [Metoprolol CR/XL Randomised Intervention Trial in Congestive Heart Failure] 5), bisoprolol (CIBIS-II [Cardiac Insufficiency Bisoprolol Study II] 6), and nebivolol (SENIORS [Study of Effects of Nebivolol Intervention on Outcomes and Rehospitalization in Seniors With Heart Failure] 7) showed that beta-blockers reduce morbidity and mortality in HF patients. As a result, these agents have received a Class IA recommendation in current HF guidelines (1). Atrial fibrillation (AF) is common in HF and occurs in 30% to 40% of all patients (8). The HF trials that led to the recommendations also included a proportion of patients with AF. In current guidelines for HF therapy (1), the recommendation for beta-blockers is not restricted to patients with sinus rhythm and includes all HF patients (i.e., also for those with AF) but it is unknown whether beta-blockers are as effective in those patients as they are in patients with sinus rhythm.

In patients with sinus rhythm with and without HF, lower heart rate is associated with a better outcome ((9),(10),11) and reduction of heart rate (by beta-blockers) probably plays an important role in the beneficial effect of these drugs. In patients with AF, with or without HF, lower heart rate, however, is not associated with a better outcome as was shown recently (12).

Although patients with AF were included in the large HF trials, the absolute number of patients with AF in each individual study was limited ((13),(14),(15),16). The aim of the present meta-analysis therefore was to assess the effect of beta-blockade on outcome (i.e., mortality and hospitalization for HF) in patients with both HF and AF.

We searched MEDLINE using search tools provided by PubMed and OVID. These search tools have been validated by Haynes et al. (17) to optimize retrieval. We used the keywords atrial fibrillation, heart failure, beta-blocker therapy, beta-blockade, and medical therapy and a combination of these terms and included papers published in English. Furthermore, we reviewed reference lists from eligible studies, used the “see related articles” feature for key publications in PubMed, consulted the Cochrane Library, and searched the ISI Web of Knowledge for publications that cited key publications.

Studies were included that investigated the effect of placebo-controlled, randomized beta-blocker therapy in patients with AF documented by electrocardiography (ECG) at baseline and HF with reduced systolic left ventricular ejection fraction (LVEF) <40%. We restricted our final search to beta-blockers that are registered for HF treatment (i.e., metoprolol, carvedilol, bisoprolol, nebivolol). For this reason, one large outcome trial which examined bucindolol (BEST [Beta-Blocker Evaluation of Survival Trial]) was not included (18). One study (SENIORS [Study of Effects of Nebivolol Intervention on Outcomes and Rehospitalization in Seniors With Heart Failure]) included both patients with reduced and preserved LVEF. For the present analysis, we included only patients with LVEF <35%, because this was the cutoff value used in that study, both in the methodology in the main study (7) and in the separate publications of the 2 groups (19). After study selection, we extracted data for the sinus rhythm group for comparison with those for AF patients.

Articles were excluded if: 1) no data were available for clinical outcomes; 2) data were published only in abstract form; 3) no definition for HF was given, either by combination of symptoms and signs (using New York Heart Association [NYHA] functional class or physical examination), imaging (impaired LVEF), or a combination of both; and 4) no distinction was made between AF and sinus rhythm. The primary outcome measure was defined as all-cause mortality. The secondary outcome variable was hospital admission for worsening HF. For each study, we evaluated the effect of beta-blocker treatment on both of these outcomes, separately, in patients with AF and in those in sinus rhythm included.

The quality of the individual studies was assessed by 11 factors: 1) sufficiently specified inclusion and exclusion criteria; 2) sufficient explanation of sample selection; 3) specification of clinical and demographic variables; 4) representation of the study sample for the mentioned patient population; 5) specification of outcome measures; 6) definition of AF; 7) assessment of the dose-response relationship between beta-blocker therapy and outcome; 8) adjustment for possible confounders in the analysis; 9) reporting of rates of patients lost to follow-up; 10) study design; and 11) duration of follow-up. Grading was as follows: good quality included 8 to 11 criteria, fair quality included 5 to 7 criteria, and poor quality included <5 criteria (20).

Meta-analysis was performed using a fixed-effects model to determine risk associated with beta-blocker therapy and all-cause mortality as measured by combined crude mortality rates. In secondary analysis, hospital admission for HF was studied in a similar manner. For comparison with patients in sinus rhythm, subgroup analysis was carried out by testing of heterogeneity across subgroups and by testing the null-hypothesis that the proportion of total variation in subgroup estimates was due to genuine variation across subgroups, rather than sampling error. Second, we carried out interaction analysis between subgroups of patients with AF and sinus rhythm, based on methods described by Altman and Bland (21). Among studies, heterogeneity of risk estimates was examined using a standard chi-square test and I² statistic for heterogeneity. Reasons for diversity in study results were explored using metaregression analysis. Variables explored included age, sex, baseline rhythm, hypertension, diabetes, ischemic heart disease, NYHA functional class, LVEF, heart rate, heart rate reduction, blood pressure, and medical treatments, including use of diuresis and digitalis and renin angiotensin system inhibitor. Results are presented as odds ratios (ORs) with 95% confidence intervals (CIs) and p values. Evidence of publication bias was assessed by visual inspection of the funnel plot. A p value of <0.05 was considered statistically significant. Statistical analyses were performed using Stata version 11.0 software (Stata, College Station, Texas) and RevMan version 5.1 software (22).

The search retrieved 248 citations, 4 of which fulfilled all criteria as they investigated the randomized allocation of beta-blocker therapy in patients with HF and AF (Figure 02_gr1). All these reports were specific AF substudies of the large HF outcome trials (U.S.-Carvedilol [(13)], CIBIS II [(14)], MERIT-HF [(15)], and SENIORS [16]) that compared the effect of beta-blockers with those of placebo. We were not able to retrieve data from one other large HF beta-blocker study (COPERNICUS), because the presence of AF documented at baseline was not reported (although new onset AF was documented in 1 article) (4). Study quality was scored as “good” for all but one, the U.S.-Carvedilol study, which was scored as “fair.” All 4 studies reported the effect on all-cause mortality, and 3 of the 4 studies also reported HF hospitalizations.

The main characteristics of the studies included in the analysis are reported in (Table 1). Overall, 8,680 patients were included. Patient characteristics are shown in (Table 2). In total, 1,677 patients (19%) were in AF at baseline (mean age, 68 years of age; 30% women), of whom 842 received a beta-blocker and 835 received placebo. The other 7,003 patients (mean age 63 years; 24% women) had documented sinus rhythm at baseline (3,640 in the beta-blocker group and 3,363 in the placebo group). Baseline heart rate, change in heart rate, and achieved heart rate in each patient group are shown in (Table 3) and (Figure 02_gr2) (no data were available for the U.S.-Carvedilol study). Heart rate reduction in patients in AF was similar to that in patients in sinus rhythm, although the baseline heart rate was higher in patients with AF. Doses of beta-blockers were similar in CIBIS-II, MERIT-HF, and SENIORS (no data were available for the U.S.-Carvedilol study).

Follow-up varied between a maximum of 13 months in the U.S.-Carvedilol study to a mean of 21 months in the SENIORS trial. The crude mortality rates for AF patients receiving beta-blocker therapy and those who were not were 13.5% and 15.7%, respectively, and 8.3% and 13.1%, respectively, for sinus rhythm patients receiving beta-blocker therapy and those who were not. This resulted in a combined mortality risk OR of 0.86 for AF patients (95% CI: 0.66 to 1.13; p = 0.28) receiving beta-blocker therapy versus a combined mortality risk OR of 0.63 for sinus rhythm patients (95% CI: 0.54 to 0.73; p < 0.00001) receiving beta-blocker therapy (Figure 02_gr3). Interaction analysis showed there was significant interaction on the effect of beta-blocker therapy in AF versus sinus rhythm (p = 0.048). There was no heterogeneity observed among the studies with AF included (I² = 0%; p = 0.46). In metaregression analysis, we found no confounding by any of the variables explored. (Figure 02_gr4) shows the funnel plot for the main outcome analysis, which shows no evidence of publication bias.

Three of the 4 studies (no data were available for the U.S.-Carvedilol trial) reported the effect of beta-blocker therapy on hospital admission for HF, including 7,586 HF patients (1,541 [20%] AF patients). Beta-blocker therapy in AF patients was not associated with a reduction of HF hospitalizations (16.2% vs. 14.8% events), resulting in an OR of 1.11 (95% CI: 0.85 to 1.47; p = 0.44) (Figure 02_gr5). For patients in sinus rhythm (8.5% vs. 14.3% events), beta-blocker therapy was associated with a reduction of HF hospitalizations (OR: 0.58 [0.49 to 0.68]; p < 0.0001). Interaction analysis showed a significant interaction on the effect of beta-blocker therapy in AF versus sinus rhythm for HF hospitalizations (p < 0.001).

The main finding of the present meta-analysis indicates that the effect of beta-blockers in patients with HF and AF is significantly different from the effect of these drugs in patients with HF and sinus rhythm. Indeed, beta-blockers were not found to have a favorable effect on HF hospitalizations or mortality in 1,677 AF patients who had been enrolled in placebo-controlled, randomized studies.

This finding is important as most patients with HF and AF receive beta-blocker treatment. Beta-blockade is recommended in the current guidelines for HF and AF treatment, albeit for different indications ((1),). In HF treatment guidelines, beta-blockers are recommended for all patients in order to reduce morbidity and mortality, without differentiation regarding rhythm. As such, these drugs are part of the standard medical therapy for all patients with HF and reduced LVEF. In addition, beta-blocker therapy has been shown to prevent new onset or recurrent AF in patients with HF ((15),) after myocardial infarction (25), and also in a relatively low-risk (most hypertension) population (26).

In the AF treatment guidelines, however, beta-blockers are recommended for rate control in order to reduce AF-related symptoms but not to improve prognosis (23). In line with the guidelines are recent data from the CHARM (Candesartan in Heart failure: Assessment of Reduction in Mortality and Morbidity) program, which showed no predictive value of higher heart rates in HF patients with AF, in contrast to the observations in sinus rhythm patients (11).

How can these different effects of beta-blockers between HF patients with AF and sinus rhythm be explained? First, heart rate reduction by beta-blocker therapy may be less effective in patients with AF than in those with sinus rhythm because the mode of action of beta-blockers is different during AF and sinus rhythm. During sinus rhythm, beta-blockers exert their heart rate lowering effect by targeting the sinus node, whereas during AF their main site of action is the atrioventricular node. In the present analysis, however, we found a similar mean reduction in heart rate for patients with both AF and sinus rhythm with comparable dosages of beta-blockers; however, achieved heart rate was not available and may have been different in AF and sinus rhythm. Second, heart rates were measured only at rest. Heart rate reduction during (moderate) exercise may have been different between AF and sinus rhythm patients. Indeed, there may be differences in the optimal heart rate at rest and during exercise and optimal heart rate reduction by beta-blockers between both groups of patients. In patients with sinus rhythm, it has been proven that a pronounced reduction in heart rate is associated with improved morbidity and mortality, independent of beta-blocker dose or by additive therapy with selective If-channel blockade ((9),). For patients with permanent AF, it was recently demonstrated that stricter rate control was not superior to a lenient rate control (12). Third, because of the loss of the atrial kick and irregularity in ventricular response during AF, patients with AF may need a higher heart rate to maintain a similar cardiac output, possibly even more so during HF (27). Fourth, a low heart rate in patients with AF may be an expression of an underlying conduction disorder, which may be associated with impaired outcome itself. Finally, AF in patients with HF may be a marker of a poorer clinical condition leading to a worse outcome, less modifiable by beta-blocker treatment (28).

In addition to these potential explanations, we also cannot exclude the fact that the present findings could apply to some but not all beta-blockers, as differences in pharmacological profiles of beta-blockers may have played a role. Metoprolol and bisoprolol are selective beta-1 receptor antagonists, and carvedilol and nebivolol are beta-blockers with additional vasodilating properties. A subanalysis of the COMET (Carvedilol or Metoprolol European Trial) (including 600 patients with AF) demonstrated that carvedilol had a better effect on outcome than metoprolol (29). The main COMET has been criticized because the dose of the 2 drugs might not have been comparable, as they lowered heart rate to a different extent. However, given the absence of a relationship between heart rate lowering and outcome in AF patients, this criticism may be less relevant in this subpopulation of AF patients. It must be noted that carvedilol had a relatively favorable effect in the present analysis in the AF patients in the U.S.-Carvedilol study (13), but these patients had milder disease than in the other studies, which also may have affected the results.

Beta-blockers are standard therapy for HF. Other drugs that are generally recommended for HF are angiotensin-converting enzyme (ACE) inhibitors, angiotensin receptor blockers, and aldosterone receptor blockers (or mineralocorticoid antagonists). It is remarkable that all these classes of drugs have been shown to be at least as effective in patients with AF as they are in patients with sinus rhythm in analyses similar to the present study ((28),).

We conducted a meta-analysis of non–pre-specified subgroups of large randomized trials, and this analysis has some limitations that merit consideration. Although the number of AF patients in the included randomized studies was 1,677 with 145 events, this is still low for survival analysis, and we cannot exclude the possibility that lack of power may have played a role. On the other hand, when ORs of beta-blocker therapy in AF patients were similar to those in sinus rhythm patients, there would have been sufficient power to detect an effect as large as found in sinus rhythm patients. However, if the effect of beta-blocker therapy is attenuated in AF, larger sample sizes are needed to draw definite conclusions. Nevertheless, there was a significant interaction with regard to this (beta-blocker) treatment effect between AF and sinus rhythm patients, which further supports our findings. Also, in the present analysis, we pooled the effects of different beta-blocker therapies and thereby assumed a class effect. However, specific differences in pharmacologic profiles may have added to the heterogeneity of our cohort and thereby the results. Inherent limitations of pooled analysis of studies include the limited availability of confounding variables, including history of AF, duration of AF, pattern of AF (paroxysmal vs persistent and/or permanent AF), new onset AF, dose response, and tolerance to the drugs. In metaregression analysis, we explored possible study characteristics that might have influenced the pooled estimates. However, given the small number of studies included, we had only limited power to find significant confounders. Finally, this analysis pooled study group estimates and did not assess individual patient data, which limits the possibility of adjustment for individual patient characteristics.

The present analysis suggests that the effect of beta-blockade in HF patients with AF with regard to outcome is less than in HF patients with sinus rhythm. Clearly, prospective randomized controlled trials in HF specifically aiming at AF patients are warranted to study the prognostic effects of beta-blockers in this population.