Carvedilol alone or in combination with digoxin for the management of atrial fibrillation in patients with heart failure?

Patients

We recruited patients with persistent AF (>1 month) and HF who were receiving digoxin and diuretics. Heart failure was defined as appropriate symptoms of HF for more than two months and echocardiographic evidence of cardiac dysfunction (left ventricular ejection fraction [LVEF] <40% or preserved LV systolic function, together with LV hypertrophy, suggesting diastolic dysfunction in the absence of an alternative potential cause of symptoms). The following were exclusion criteria: heart rate (HR) at rest <60 beats/min, systolic blood pressure (BP) <90 mm Hg, sick sinus syndrome or complete heart block, current treatment with a beta-blocker or HR-lowering calcium channel antagonist or >200 mg amiodarone, recent major cardiovascular event or procedure, asthma or reversible obstructive airways disease, serum creatinine >250 μmol/l or significant hepatic disease, uncorrected significant valvular heart disease, or any life-threatening noncardiac disease. All patients gave written, informed consent, and the protocol was approved by the local Ethics Committee.

Study design

In a randomized, controlled, double-blinded, parallel-arm study, the following treatments were assessed: digoxin alone versus the combination of digoxin and carvedilol (phase 1, duration four months) and digoxin alone versus carvedilol alone (phase 2, duration six months) (Fig. 1). Withdrawal of digoxin at the same time as initiating and uptitrating beta-blockers could increase the risk of worsening HF. Therefore, a complex study design was adopted to allow double-blinded initiation of carvedilol first, followed by double-blinded withdrawal of digoxin, once maintenance doses of carvedilol had been achieved.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Study design. *Randomization at this stage determined pairing of therapies in phase 2. Carv. = carvedilol; DB = double-blind; Dig. = digoxin; OL = open-label.

Investigations were performed at baseline and after each phase. Before randomization, patients had their digoxin dose increased, if necessary, to achieve optimal resting ventricular rate control (defined as a ventricular rate of at least <90 beats/min on a 12-lead electrocardiogram [ECG]). Plasma concentrations of digoxin were monitored to avoid toxicity, except in the second phase, which would have compromised the double-blinded nature of the study. The reference levels for toxicity were 1.2 to 2.6 nmol/l. Forty-seven patients were randomized into two groups: 24 into the carvedilol treated group and 23 into the placebo group. In phase 1, both groups continued open-label digoxin; in addition, one group received double-blind carvedilol while the other received double-blind placebo. The starting dose of carvedilol was 3.125 mg b.i.d., and this was increased at two-week intervals to the target dose of 25 mg b.i.d. (uptitration period of two months) or, for patients weighing more than 85 kg, 50 mg b.i.d. In phase 2, open-label digoxin was replaced by double-blind placebo in the group already receiving double-blind carvedilol. Open-label digoxin was replaced with double-blind digoxin in the group receiving double-blind placebo in phase 1. Patients were reviewed two weeks after open-label digoxin was withdrawn or substituted to monitor for worsening HF. Repeat investigations were carried out eight weeks after open-label digoxin withdrawal or substitution (or six months after initial randomization). The randomization process involved a paired coding system during the two phases, so that no patient was left without some ventricular rate control therapy. This design allowed a double-blinded comparison between digoxin alone and carvedilol alone at the end of six-month follow-up.

Investigations

The primary prespecified outcomes for the study were: 1) LVEF by ECG-gated radionuclide ventriculography; 2) ventricular rate control by 24-h ambulatory ECG; and 3) symptoms at the end of phase 2 of the study. We also investigated exercise tolerance by 6-min corridor walk distance in all patients. In a subset of 23 patients, ventricular rate control during steady-state bicycle exercise test was assessed. Plasma concentrations of brain natriuretic peptide (BNP) (C-terminal, Peninsula Labs, Belmont, California) were measured.

Primary end point measurements

Assessment of symptoms

We used a self-administered, quantitative questionnaire designed to measure the patient's perception of the frequency and severity of symptoms and their functional capacity. A 4-point scale (0 = absent; 3 = severe symptoms) was used for the following symptoms; chest pain/discomfort, fatigue, shortness of breath, and palpitations. For the first three symptoms, patients were asked to grade severity at rest, during walking at normal pace, and while climbing stairs. We also asked about the patient's “global” health or their sense of well-being. The responses were summed to produce a symptom score ranging from 0 (no symptoms) to 33 (worst symptoms).

Exercise tests

All patients underwent a standard 6-min corridor walk test with verbal encouragement given to the patient every 2 min. A subgroup of 11 patients in the carvedilol-treated arm and 12 patients in the placebo arm exercised on a bicycle against a resistance of 50 W at 40 rotations/min for 5 min. The remainder was either unable to exercise due to co-morbidities (n = 14), such as arthritis or hemiparesis due to stroke, or declined (n = 10). Blood pressure was recorded manually by a sphygmomanometer at rest (pre-exercise) and 3 and 5 min into exercise. The last 10 s of recorded ECG of each minute was used to determine the ventricular rate for that minute. All patients exercised in the morning between 9 amand 12 pm.

Statistics

We used LVEF to determine the sample size. Because there are no data directly comparing digoxin to carvedilol for a change in LVEF in patients with HF, we powered the study on the expected differences in ejection fraction when comparing digoxin with the combination of carvedilol and digoxin. The mean increase in LVEF in controlled studies of carvedilol in HF patients is ∼8% (10); a large percentage of the population in these trials was on maintenance digoxin therapy. Given a postulated standard deviation of 9.5% for ejection fraction, the sample size was estimated to be 44 patients (22 in each limb; beta = 0.80, alpha = 0.05).

Absolute values between groups were not compared because of expected heterogeneity in individual electrophysiologic characteristics and consequently differing responses to pharmacologic intervention. Rather, for each patient, the change from baseline for the results of investigations was determined at the end of phase 1 (four months) and phase 2 (six months). The change from baseline (i.e., before the first randomization) was then compared between groups. A paired ttest for intragroup comparisons and a two-sided ttest for intergroup comparisons were used for normally distributed data. The Mann-Whitney Utest was used to compare differences for nonparametric data. For paired sample non-normal data, the Wilcoxon signed rank test was used. The Ztest was used for the comparison of proportions, as appropriate. A value of p < 0.05 indicated statistical significance. To estimate the time of peak ventricular response (acrophase) and diurnal variation in the hourly mean ventricular rate for treatments, the mean hourly ventricular rate of each group was modeled by cosinor analysis (11). Group data are presented as the mean ± SD for parametric data and the median value with interquartile ranges for nonparametric data. Analysis was undertaken by SPSS statistical software, version 9.

Our statistical analysis plan did not include analysis of variance (ANOVA), as the principle analysis of interest was the comparison between monotherapies, with combination therapy being a necessary intermediate step of novel mechanistic but not novel clinical significance (as far larger subsets of patients with AF were included in landmark trials of beta-blockers in HF). However, we also explored the results of repeated measures ANOVA on the three normally distributed variables (change from baseline): 24-h ambulatory mean HR, systolic BP, and 6-min walk distance. The results for this analysis did not differ from the predefined analysis plan.

Phase 1: digoxin alone versus carvedilol and digoxin in combination

Twenty of 24 patients randomized to carvedilol and 21 of 23 patients randomized to placebo completed phase 1. The mean daily dose of carvedilol achieved in the active arm was 47.6 mg. The plasma digoxin concentration did not differ significantly between groups either at baseline (Table 1) or at the end of phase 1 (1.2 ± 0.5 vs. 1.4 ± 0.5 nmol/l for digoxin alone vs. combination, respectively, p = 0.2). However, two patients on carvedilol developed possible symptoms of digoxin toxicity and had their digoxin dose reduced, although serum concentrations remained in the therapeutic range.

No significant change in any study variable occurred when adding placebo to digoxin. Compared with placebo, the addition of carvedilol to digoxin reduced the mean 24-h ventricular rate (p < 0.0001) and improved symptoms scores (p < 0.05) and LVEF (p < 0.05). New York Heart Association (NYHA) functional class also tended to improve (p = 0.08) with combination treatment. Changes in BNP and 6-min walk distances were not significant (Table 2).

Phase 2: digoxin alone versus carvedilol alone

In phase 2 of the study, double-blinded withdrawal of digoxin from patients on carvedilol led, by intragroup comparison, to a significant increase in the 24-h mean HR and a decline in LVEF. When the intergroup changes from baseline for these variables were compared, no significant difference between digoxin alone and carvedilol alone was noted. However, BNP values increased significantly more for carvedilol alone than for digoxin alone. There was conflicting evidence for symptomatic deterioration on withdrawal of digoxin. Symptom scores of patients who remained on therapy favored carvedilol (p = 0.007). However, three patients in the carvedilol-treated group developed worsening symptoms of HF, associated with a rise in HR, leading to an exit from the study on days 11, 18, and 24 days after withdrawal of digoxin. Preceding NYHA class, ventricular function, or plasma BNP did not predict deterioration. One patient developed worsening symptoms of HF when switching from open-label to double-blind digoxin. After assigning maximum symptom scores to withdrawals, a significant difference in symptoms between carvedilol alone and digoxin alone was no longer noted (Table 2).

Ventricular rate control (Figs. 2 and 3)

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

(A to C)Mean hourly heart rate (HR) as influenced by treatment regimen(s) during ambulatory 24-h electrocardiogram. Points with an asterisk represent significant differences in HR between groups. At baseline, by chance, the ventricular rate tended to be higher in the carvedilol-treated group throughout 24 h, but this difference was only statistically significant for a few hours. Analyses are intergroup changes from baseline values.

Figures 2A to 2C, illustrates the 24-h ventricular rate profile for each phase of the study. Compared with digoxin alone, the addition of carvedilol lowered the mean 24-h ventricular rate during both day and night (Fig. 2B). Carvedilol alone and digoxin alone were equally effective in controlling the daytime ventricular rate, but digoxin lowered the nocturnal HR to a greater extent (Fig. 2C). Carvedilol significantly reduced day-night differences in the mean hourly HR, in combination with digoxin or alone, principally by attenuating the increase in HR during waking hours (Figs. 2B and 2C). The peak ventricular rate occurred at ∼15:00 h for all treatment groups. The median maximum ventricular rate was 176 beats/min on digoxin alone versus 134 beats/min on combination treatment (phase 1; p < 0.05). Carvedilol alone and digoxin alone appeared equally effective in reducing the median maximal ventricular rate.

Compared with digoxin, combination therapy reduced the ventricular rate at rest and throughout steady-state exercise (Fig. 3B) (peak ventricular rate 106 vs. 123 beats/min on digoxin; p < 0.05). Carvedilol alone and digoxin alone appeared equally effective in controlling exercise HR (Fig. 3C). The rate of rise in ventricular rate during exercise was similar for carvedilol or digoxin alone and combination treatment.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Ventricular rate during submaximal exercise. (A)Pre-carvedilol (open triangles)and placebo (solid triangles)groups at baseline. (B)Phase 1 (four months): combination therapy (solid squares)versus digoxin alone (open squares). (C)Phase 2 (six months): digoxin alone (open circles)versus carvedilol alone (solid circles).

The peak rate-pressure product (RPP) was reduced on combination therapy compared with digoxin alone (15,003 ± 6,684 vs. 19,176 ± 5,463; p < 0.007). However, in phase 2, comparing carvedilol and digoxin monotherapies, peak RPP was not significantly different (13,583 ± 3,213 vs. 17,017 ± 3,557, p = 0.23). The peak ventricular rate was similar (114 ± 21 vs. 116 ± 18 beats/min, p = 0.8), but systolic BP (127 ± 12 vs. 157 ± 19 mm Hg, p = 0.002) was lower for carvedilol than digoxin.

Pauses

Compared with digoxin, combination therapy resulted in a greater mean maximum pause (2.9 ± 0.8 vs. 2.4 ± 0.6 s, p < 0.05). The mean maximum pause was greater for digoxin alone compared to carvedilol alone in phase 2 (2.3 ± 0.7 s vs. 1.8 ± 0.73 s for digoxin and carvedilol, respectively; p < 0.05). However, there was no significant difference in the numbers of pauses >3 s, either during the day (8 amto 8 pm) or night (8 pmto 8 am) for any treatment group.

Adverse events

In phase 1, three patients withdrew because of adverse effects subsequent to the initiation of carvedilol (one each with gastrointestinal disturbance, tiredness, and bronchospasm), and one withdrew in the placebo group (self-withdrawal). Four patients, as noted earlier, withdrew because of worsening HF during phase 2. There were three deaths during the study: two in the group randomized to carvedilol (myocardial infarction in phase 1 and stroke in phase 2) and one in the group randomized to placebo (postoperative sepsis after transurethral resection of the prostate during phase 1).

Ventricular rate control and LV function

A reduction in the ventricular rate during exercise and 24-h ambulatory ECG monitoring was less than that anticipated with carvedilol alone compared with digoxin alone (12). This suggests that both increased vagal tone and reduced sympathetic activation are important in controlling ventricular rate in patients with HF and AF (Figs. 2B and 3B). The sympatholytic effects of digoxin may also be complementary to those of beta-blockade (13). Digoxin was more effective at reducing the nocturnal HR, reflecting the importance of parasympathetic activation when sympathetic tone is low.

Beta-blockers consistently reduce end-diastolic volume and improve LVEF both in patients with chronic HF in sinus rhythm (10)and in those with AF (5), but the mechanisms underlying this effect are uncertain. An increase in LVEF with the addition of carvedilol could reflect a nonspecific response to a reduction in the ventricular rate. There is some evidence of HR-dependent cardiomyopathy in patients with AF and suboptimal ventricular rate control (12).

No previous study has investigated the utility of digoxin in patients with HF and persistent AF treated with beta-blockers. This is important given concerns about polypharmacy in patients with HF. For those who tolerated digoxin withdrawal, the ejection fraction declined by a median of 9%, whereas the mean 24-h ventricular rate rose by 22 beats/min to a mean of 88 beats/min. The latter was unexpected, as digoxin withdrawal has not been shown to cause such a marked increase in HR or reduction in ventricular function in patients in sinus rhythm (14). It suggests that the atrioventricular node may be less susceptible than the sinoatrial node to the effects of beta-blockers in patients with HF. Alternatively, the withdrawal of digoxin may reduce baroreflex receptor sensitivity, removing the tonic inhibition of central sympathetic drive, thereby limiting the effects of beta-blockade (13). Our findings, therefore, do not support the notion that the introduction of beta-blockers as standard treatment of HF obviates the need for digoxin in the presence of AF.

A potential criticism of this study is the accuracy of the assessment of LV function in patients with AF, due to the large variation in the RR interval. However, Bacharach et al. (9)showed that the average of single-beat values plotted against preceding RR intervals were very similar to one ECG-gated time-activity curve (as used in this study). Inagaki et al. (15)also showed that ECG-gated radionuclide ventriculography is reliable, reproducible, and sensitive to change in patients with AF.

Symptoms

There were significant improvements in symptom scores with the double-blinded addition of carvedilol to digoxin (Table 2), as demonstrated in other studies of HF, presumably reflecting the effects of therapy on underlying ventricular function (3). Most patients did not deteriorate symptomatically upon withdrawal of digoxin, despite a decline in ventricular function. This could reflect a protective effect of increased BNP levels or some other action of carvedilol, such as peripheral vasodilation. No patient with severe HF deteriorated clinically after digoxin withdrawal. All three patients who deteriorated symptomatically, necessitating withdrawal from study, were in NYHA class I or II at the end of phase 1. Longer follow-up on combination therapy might have prevented deterioration on digoxin withdrawal. However, longer follow-up on carvedilol monotherapy might have allowed declining cardiac function to be translated into a worse clinical outcome more often.

Natriuretic peptides

Plasma concentrations of BNP tended to rise when carvedilol was added to digoxin, rose further when digoxin was withdrawn, and did not predict which patients would deteriorate upon withdrawal of digoxin. Increases in natriuretic peptides have also been noted in studies of beta-blockers in patients in sinus rhythm (16)and may reflect increases in ventricular filling pressure, conversely this could be part of their therapeutic action of beta-blockers (17). These data suggest that natriuretic peptides may not be a useful aid to the management of HF with beta-blockers.

Previous studies

Synergistic effects of digoxin and beta-blockade in controlling the ventricular rate in AF have been demonstrated in previous studies, although these small studies had few patients with HF (1). It is uncertain whether all beta-blockers would exert the same benefits in patients with AF and HF, most of whom are currently treated with digoxin. A post hoc analysis of Cardiac Insufficiency BIsoprolol Study (CIBIS II) suggested that, in patients with AF, bisoprolol had no impact on survival or hospitalization for HF (18). However, retrospective analyses of the U.S. carvedilol HF trials program suggested that the benefit of carvedilol was similar in the presence or absence of AF (5), whereas a greater benefit of carvedilol on death and hospitalization was reported among patients taking digoxin, although only a minority had AF (19). Recently, the Carvedilol Or Metoprolol European Trial (COMET) has suggested that carvedilol exerts a greater effect on mortality than metoprolol, but data on patients with AF are not yet available (20).

Potential pharmacologic alternatives to beta-blockers for the control of ventricular rate in patients with AF and HF include diltiazem and amiodarone. Concerns exist about the safety of diltiazem in patients with HF (21). Deedwania et al. (22), in a post hoc subgroup analysis of the Congestive Heart Failure-Survival Trial of Anti-arrhythmic Therapy (CHF-STAT) trial, demonstrated that the ventricular rate was significantly reduced with amiodarone. Furthermore, spontaneous cardioversion occurred in 31% of those on amiodarone but only 8% of those on placebo at one year. The risk of long-term side effects with amiodarone and the known mortality benefits of beta-blockers render the latter group of agents preferable for the control of ventricular rate in patients with persistent AF and LV systolic dysfunction.

An alternative treatment strategy in this patient population is repeated cardioversion and prophylactic anti-arrhythmic therapy to maintain sinus rhythm. However, in the presence of HF and long-established AF, the failure of cardioversion, toxicity of anti-arrhythmic drugs, and early relapse are major drawbacks (12). A number of trials have investigated the utility of cardioversion versus a strategy of rate control for atrial fibrillation (23–25). None has suggested that cardioversion is superior in terms of improving symptoms, reducing stroke, or improving survival. Ventricular rate control with anticoagulation therefore seems an appropriate strategy in this population until either a subgroup is identified that would benefit from a policy of repeated cardioversion or more effective and safer methods of maintaining sinus rhythm are developed.

Conclusions

This study suggests that the combination of digoxin and carvedilol reduces symptoms, improves ventricular function, and leads to better ventricular rate control than either agent alone and therefore should be considered the standard treatment for HF patients with persistent AF.