Jodi B. Segal, MD, MPH

Despite pharmacologic and electrical interventions, sinus rhythm cannot be restored and maintained in many patients with atrial fibrillation. For these patients, control of the ventricular rate is a primary goal of therapy, since a rapid rate may lead to worsening congestive heart failure, myocardial ischemia, or distressing breathlessness and palpitations.

A number of review articles have described strategies for rate control, principally involving the use of digoxin, calcium-channel blockers, and b-blockers.^1-6 A recent analysis of the trends in the use of drugs for ventricular rate control found that the use of digoxin and b-blockers decreased between 1980 and 1981 and 1994 and 1996, and the use of the nondihydropyridine calcium-channel blockers diltiazem and verapamil increased.⁷ These investigators, however, indicated that “current practices are dictated more by clinical tradition than by clinical science.”⁷ There has not been a systematic review of the trials evaluating the efficacy of both the familiar and the newer medications for ventricular rate control in atrial fibrillation. It is increasingly clear the drugs that are used most often for heart rate control at rest may not be the most efficacious during exercise, and exercise tolerance is compromised by some drugs.⁶

The purpose of our review was to characterize the strength of the evidence regarding the efficacy of drugs used for ventricular rate control in atrial fibrillation.

Methods

Study Design

We performed a systematic literature review and synthesis of randomized controlled trials on ventricular rate control in atrial fibrillation. To be eligible for inclusion in our review, trials needed to meet the following criteria: address management of nonpostoperative atrial fibrillation or atrial flutter; include human data; include adult subjects; and present original data. Studies that included patients with postoperative atrial fibrillation were not excluded as long as those patients were only a minority of the included patients.

Literature Identification and Search Strategies

The primary source of literature for our review was the CENTRAL database of The Cochrane Collaboration, a comprehensive collection of controlled clinical trials from 1948 to the present. As a secondary source, we searched MEDLINE from 1966 to May 1998 to ensure completeness. Additionally, we used the related articles feature of PubMed, as well as recent search results submitted to the Baltimore Cochrane Center, the contents pages of recent relevant journals, and programs from recent cardiology meetings. Our search strategy included using the MeSH terms “atrial fibrillation” and “atrial flutter” as subject headings and text words, as well as “random allocation,” “double-blind method,” and “single-blind method.” The publication types were “randomized controlled trials” and “controlled clinical trials.”

Abstracts of the citations of randomized controlled clinical trials were reviewed independently by 2 members of the study team to identify articles that met the inclusion criteria. Only English-language articles were reviewed. However, we reviewed all English-language abstracts of non-English-language publications to assess qualitative consistency with our results.

Data Abstraction

A form was developed to extract information from the eligible articles regarding study quality, characteristics, and findings. The section on study quality was created after our review of forms used in meta-analytic studies^8,9 and a literature review^10,11 and with the assistance of The Cochrane Collaboration. The resultant form incorporated⁶ key questions used by The Cochrane Collaboration in its reviews and¹⁴ key questions identified by Detsky and colleagues.¹⁰ Our form was pilot-tested for clarity and reproducibility and revised as needed. The final version contained 22 questions assessing quality in the following 5 areas: representativeness (how well the study population was described); potential for bias and confounding; description of therapy (eg, how similarly the groups were treated); outcomes and follow-up; and statistical reporting and interpretation [Table 1]. Each question was worth a maximum of 2 points, and the score in each category was the percentage of points received out of the total available. The overall quality score was calculated as the average of the scores for the 5 categories.

The portion of the form for quantitative data abstraction included sections for subject inclusion and exclusion criteria, baseline subject characteristics, therapeutic protocols, and outcomes. We recorded the mean heart rate at rest, the mean maximum heart rate during exercise or immediately after exercise, the proportion of subjects who reached the goal heart rate reduction in each treatment arm, and any measure of exercise tolerance.

The review of study quality was done independently by 2 reviewers, and differences were resolved by consensus. Quantitative data were abstracted by one primary reviewer and then checked for accuracy by a secondary reviewer. The reviewers were not blinded to the author, institution, or journal, since it seemed unlikely that this information would make a significant difference in the results.¹²

Presentation of the Data

We constructed 3 evidence tables with the trials grouped according to the regimens being compared. [Table 2a][Table 2b] displays the quality scores and key design elements of each trial. [Table 3] contains a listing of the trials of the most frequent comparisons and the absolute differences in heart rates of patients using those therapies. [Table 4] shows the results from the trials for which there were few, if any, given comparisons.

Data was synthesized by creating scatter plots of mean heart rates at rest and with exercise for each of the main drug comparisons (Figures 1-5). The data were not amenable to formal mathematical pooling (ie, meta-analysis) because of significant qualitative heterogeneity among the studies.*

Results

Literature Yield

We retrieved 74 abstracts. Of these, 8 were abstracts of articles from non–English-language literature. Forty-five trials were eligible for inclusion in our review; the authors of those studies evaluated 17 different drugs and several combinations of drugs. Some of the included trials were not designed as rate-control studies; they were studies of pharmacologic conversion that included heart rate data.

Qualitative Synthesis

The following comparisons were made in the trials: calcium-channel blockers,¹⁴-²¹ b-blockers,^19-28 digoxin,^{14,22,28,30-32,38} or other drugs and combinations compared with placebo;^{14,28,29,33-38} calcium-channel blockers,^14,39-42 b-blockers,^23,29,43,44 or other drugs and combinations compared with digoxin;^{14,27,40,44-49} and other drug comparison trials.^{18-20,27,42,50-58} Many of the trials involved more than 2 treatment arms. Nearly all of the trials of calcium-channel blockers, b-blockers, and digoxin were designed to evaluate rate control. The studies of the other drugs were principally trials evaluating atrial fibrillation conversion that also reported heart rate data. Two of the digoxin trials were also aimed at evaluating conversion to sinus rhythm.

Study Design

As shown in [Table 2a][Table 2b], there were important similarities and differences among the trials evaluating the same therapies. The duration of the trials and route of administration of the drug are presented to aid in interpretation of the results.

Notably, all of the trials of any given comparison were done within 10 years of each other. This should reduce the likelihood of secular trends affecting the outcomes of the trials. The range in study sizes is extreme, although most of the trials enrolled fewer than 50 patients. Several trials had fewer than 10 participants, and it was anticipated that these small trials would have little power to detect differences between treatments.

The regimens differed among the trials of the same medication. The intravenous diltiazem and verapamil doses were fairly uniform, although the oral dosages and frequency of administration differed. Some of the b-blocker trials involved titration of the medication to effectiveness, and digoxin was often dosed to a target blood level. All these differences may have had an impact on the outcomes. The followup times ranged from minutes to 6 weeks. Short trials may be appropriate for intravenous agents; however, several of the trials assessed the outcome a very short time after a single oral dose of medication, before a therapeutic blood level could be expected.

Another notable feature of these trials was the permissibility of other agents during the trial, as detailed in the evidence table. Permitting the use of digoxin in trials testing other medications, without reporting the number of participants in each arm receiving digoxin, can potentially confound the results.

Although not shown in the evidence table, most trials had explicit inclusion criteria. Some required atrial fibrillation lasting longer than 1 month or longer than 6 months, and several specified a ventricular rate required for entry, such as more than 120 beats per minute or “rapid rate.” Exclusion criteria varied from none to stringent; it was particularly stringent in those studies that involved exercise, from which subjects with angina or significant congestive heart failure were often excluded. None of the trials used echocardiographic data as inclusion or exclusion criteria.

Quality Scores

Many studies were weakest in their description of the participants in the study arms, so it was not always possible to tell if the groups were similar. This can be seen in the exceptionally low scores in the “representativeness” category. The potential for bias and confounding varied markedly across the trials, as did the description of the therapies. It was often unclear which other therapies the patients may have been receiving. Generally, the investigators described the outcomes completely and objectively measured them with Holter monitoring or telemetry. The completeness of statistical reporting was variable, with many studies only reporting a P value without reporting a measure of variability in the outcomes.

The studies published more recently had slightly higher total quality scores. Total quality scores were strongly associated with the size of the study, with the larger studies receiving higher scores (P < .001).

Outcomes

As shown in [Table 3], all of the trials reported either the heart rate reduction [on] for the active drug compared with the comparative drug or the proportion of patients who reached the target heart rate. Many of the trials also evaluated the efficacy of the drugs during exercise. The exercise test itself varied among the trials and included measurement of distance walked on a treadmill, measurement of oxygen consumption, and workload tolerated on a stationary bicycle.

On the basis of our qualitative assessment of the trials, we felt that any mathematical pooling of the results would result in invalid estimates of treatment effects. This was because of the markedly different treatment regimens within each drug class and the differing goals of treatment (acute or chronic management). The mixed quality of the trials also argued against pooling.

Study Results

Calcium-Channel Blockers Versus Placebo for Rate Control. All comparisons of calcium-channel blockers with placebo demonstrated that the calcium-channel blocker was more efficacious than placebo at reducing heart rate both at rest and during exercise. Five of the trials used diltiazem, 4 used verapamil, and 1 evaluated both drugs. An improvement in exercise tolerance was almost always seen, although different measures of tolerance were used. All but 2 of the trials allowed the participants to use digoxin but did not report what percentage of subjects in each treatment group received the drug. Despite different rates in the placebo arms, there was uniformity in treatment effects across the trials [Figure 1].

b-Blockers Versus Placebo for Rate Control. Seven different b-blockers were tested. Only 7 of the 12 comparisons demonstrated efficacy of the b-blocker at rest, although all were efficacious during exercise. The efficacy appears to be medication dependent. Atenolol (at 50 mg daily²⁸ or twice daily²⁰ or 100 mg daily²⁸) performed significantly better than placebo. Timolol (1 mg intravenously) allowed more subjects to reach the target heart rate compared with placebo.²⁴ Pindolol²⁸ (5 mg or 15 mg twice daily) and nadolol²⁵ (titrated dose) significantly reduced mean resting heart rate. The data regarding xamoterol were mixed.^20-23 Celiprolol²⁶ and labetalol were no more efficacious than placebo at rest.

All of the tested b-blockers demonstrated a significant reduction in heart rate with exercise compared with placebo; this included atenolol, labetalol, nadolol, celiprolol, and xamoterol.^{19,20,22,24,25} [Figure 2] shows that the effect of b-blockers on heart rate during exercise was more uniform than their impact on heart rate control at rest. However, these trials suggest that exercise tolerance in patients with atrial fibrillation may be reduced with b-blockers.

As in the calcium-channel blocker studies, most of the trials allowed subjects to continue on digoxin.Digoxin Versus Placebo. The outcomes with digoxin were mixed, as shown in [Table 3] and [Figure 3]. Two of the trials of digoxin versus placebo did not demonstrate a reduction in mean resting heart rate;^28,30 in 5 trials, however, there was a reduction.^{14,22,31,32,38} Two of these studies included patients on verapamil in both arms, so we could not attribute all of the rate reduction to digoxin alone.^31,32

Two studies evaluating digoxin during exercise did not find a significant heart rate reduction.^14,22 In one trial that suggested a difference, no measure of statistical significance was provided.²⁸ Four of the studies of digoxin and placebo evaluated exercise tolerance.^14,22,28,29 In one14 the cardiac output was higher for patients taking digoxin, and in another²² the time on the treadmill was longer with digoxin although the maximal attainable heart rate blood pressure product was higher with placebo.

Calcium-Channel Blockers Versus Digoxin for Rate Control. Three trials compared diltiazem with digoxin,^14,19,40 and 3 compared verapamil with digoxin^14,41,42 with the outcomes reported in [Table 3] and in [Figure 4]. The scatter plot is most useful for noting the trend toward improved control with calcium-channel blockers both at rest and with exercise. Notably, the cardiac output on digoxin during exercise was greater than in the 2 diltiazem groups (12.6 L/min vs 10.9 L/min and 9.1 L/min for 60 mg and 120 mg, respectively).¹⁴ Conversely, the group receiving verapamil was able to exercise longer on the treadmill than the digoxin group.⁴³ This latter study, however, had methodologic flaws, including little description of the participants.

b-Blockers Versus Digoxin for Rate Control. Four trials compared b-blockers with digoxin for rate control in atrial fibrillation, and the outcomes are reported in [Table 3] and in Figure 5.^22,28,42,43 Similar to the results of the trials of b-blockers compared with placebo, the efficacy of b-blockers was most convincing in the trials that evaluated their use during exercise. There appeared to be little difference between the efficacy of b-blockers and digoxin at rest.

Other Drugs and Combinations Versus Placebo and Digoxin. [Table 4] summarizes the outcomes for the few trials of other agents. Not surprisingly, of the 8 trials that compared digoxin with a combination of digoxin with a calcium-channel blocker,^{14,19,22,38,40,41,45,46} only one study did not find a significant decrease in mean resting heart rate with the addition of the calcium-channel blocker.⁴⁰ In 5 of the 6 studies with an exercise evaluation,^{14,19,38,41,45} the combination of a calcium-channel blocker and digoxin controlled the heart rate better than digoxin alone, while the sixth trial did not report the statistical significance of this outcome.¹⁴ Of the trials of a b}-blocker combined with digoxin, all were more effective than placebo, and all were more effective than digoxin alone except for the combination of digoxin and labetolol.²⁸ During exercise, however, this combination was more effective than either comparison arm.

Other Drugs Evaluated for Rate Control. There were 9 other randomized controlled trials of drugs for rate control in atrial fibrillation.^50-58 Two studies compared intravenous magnesium sulfate with intravenous verapamil for acute control.^50,51 In both studies, a higher percentage of subjects reached a heart rate of less than 100 beats per minute with verapamil than with magnesium sulfate.

Two studies evaluated rate control with propafenone or flecainide, both at 2 mg per kg intravenously for 1 hour; both significantly reduced the heart rate from baseline.^52,53 In both studies, subjects were allowed to continue on digoxin, calcium-channel blockers, and b-blockers. The side effects of flecainide were of more concern than those of propafenone, with conduction abnormalities in the flecainide group. Another study compared propafenone with quinidine for rate control.⁵⁴ Propafenone significantly slowed the heart rate at rest compared with quinidine. Either drug effectively slowed the heart rate compared with baseline.

Disopyramide did not reduce the mean resting heart rate from baseline.⁵⁵ The combination of diltiazem and digoxin reduced the mean resting heart rate to a greater degree than the combination of propranolol and digoxin, but all 3 drugs together were even more effective.⁵⁶ That study also demonstrated that with exercise the combination of propranolol and digoxin was more efficacious for heart rate control than diltiazem and digoxin and that the 3-drug combination was not better than just propranolol and digoxin. The combination of pindolol and digoxin reduced the maximum area under the heart rate curve significantly more than verapamil and digoxin.⁵⁷ Finally, the combination of amiodarone and digoxin slowed the resting heart rate when compared with baseline, while the combination of quinidine, verapamil, and digoxin did not; this was a small trial, however, and the baseline resting heart rates were not rapid.⁵⁸

Discussion

The randomized controlled trials of diltiazem and verapamil used by patients with atrial fibrillation provide strong evidence for their efficacy in reducing heart rate both at rest and with exercise when compared with placebo. In all of the studies that evaluated calcium-channel blockers compared with placebo during exercise, the calcium-channel blockers produced either an increase in cardiac output, oxygen consumption, or distance walked. There was also moderate evidence that diltiazem or verapamil was more effective at heart rate control both at rest and during exercise in the direct comparisons with digoxin, with a more rapid onset of action. Although digoxin appeared to increase cardiac output, verapamil prolonged time on the treadmill and increased oxygen consumption. Thus, the evidence strongly supports the use of diltiazem or verapamil for ventricular rate control in atrial fibrillation. Although they have a negative inotropic effect, reflex responses to vasodilatation usually result in a small increase in cardiac output. Therefore, except in moderate to severe heart failure, the negative inotropic effect is often not clinically apparent.⁶⁰

All of the tested b-blockers successfully reduced heart rate with exercise when compared with placebo, and most of them reduced resting heart rate. The effect on exercise tolerance was variable and may be due to the different receptor specificities of the tested drugs and the varying treatment times before testing.

When administered acutely, b-blocking agents depress myocardial function secondary to the withdrawal of adrenergically mediated inotropic and chronotropic support. However, in patients with congestive heart failure treated for longer than 1 month, b-blockers may improve myocardial function by improving intrinsic systolic function.^59,61-63 Extrapolating to patients with atrial fibrillation, b-blocker therapy may increase left ventricular ejection fraction compared with placebo if administered for longer than 1 month. None of the trials in this review lasted for longer than 4 weeks, so it is conceivable that the worsening exercise tolerance was a transient effect of the drug.

When compared with digoxin, the trials favored the use of b-blockers, as digoxin was less efficacious than metoprolol in resting heart rate reduction and less efficacious than labetalol at rate reduction during exercise. Furthermore, time on the treadmill was longer with both labetalol and metoprolol than with digoxin.

It will be interesting to see the effect of the third-generation b-blockers, such as carvedilol, on heart rate control in atrial fibrillation. We anticipate that they will be effective at ventricular rate control, with an improvement in exercise tolerance. Celiprolol and xameterol are not available in the United States and are no longer being evaluated for approval by the Food and Drug Administration.

Although several of the studies comparing digoxin with placebo were limited by subtherapeutic serum levels of digoxin at the time of evaluation, the others did show a resting heart rate reduction. There is little evidence to support the efficacy of digoxin for heart rate control with exercise. However, in the 2 trials that evaluated exercise tolerance on digoxin compared with placebo, cardiac output and time on the treadmill were greater with digoxin.

The trials evaluating other drugs, including propafenone, clonidine, and amiodarone yielded insufficient evidence to support their use for rate control at this time. These drugs appear promising and may prove efficacious when more evidence becomes available.

The optimal degree of heart rate control for patients with atrial fibrillation is unclear, particularly during exercise. Certainly an excessively rapid rate impairs ventricular filling and decreases cardiac output. However, severely limiting the heart rate acceleration that is needed to maintain cardiac output can also limit exercise tolerance.⁶³ There are few empirical trials of this. One recent study found that ventricular rate control in atrial fibrillation had no impact on cardiovascular performance as measured by endurance on a treadmill.⁶³ The studies in our review that report only on heart rate control during exercise without mention of exercise tolerance may be of less value to clinicians.

It is important to recognize that both heart rate control and exercise tolerance are surrogate outcomes for what is truly important to clinicians and their patients: their well-being, ability to conduct their daily activities, and mortality. Although several of the studies did inquire about symptoms such as palpitations or breathlessness, the use of validated quality-of-life questionnaires was rare. A study comparing pharmacologic treatment of atrial fibrillation with atrioventricular junction ablation and pacing demonstrated the use of several different quality-of-life measures: a Quality-of-Life Questionnaire, Specific Symptoms Scale, New York Heart Association Classification, and a Specific Activity Scale.⁶⁴ The authors of that study commented that they were uncertain whether the sensitivities of these scales were high enough to use as they did.

Our study is the first comprehensive evidence-based review that focused on this aspect of the management of atrial fibrillation. Several recent reviews described ventricular rate control trials in the context of other pharmacologic therapies for managing atrial fibrillation,^65,2-4 but those studies did not comprehensively review all the rate control trials. Furthermore, they contained recommendations for drugs that have never been evaluated in controlled trials of patients with atrial fibrillation exclusively (such as intravenous esmolol and intravenous propranolol).²

Limitations

Our systematic review was subject to the same limitations that are common to most reviews. The important differences among the trials preclude mathematical pooling and have to be taken into account when drawing conclusions using these studies. As in most assessments of study quality, our assessment tool was tailored to fit the topic, so the scores cannot be directly compared with quality assessments of studies on other topics. Our intent was to be able to compare the trials.

Few of the studies evaluated adverse effects in any systematic way. This inconsistent reporting of adverse effects limited our ability to comment on them. Many of the trials incompletely described the enrolled participants, so applying these results to all patients with atrial fibrillation should be done cautiously. Also, since results were seldom stratified by the clinical features of the enrolled patients, we could not report on the evidence supporting the use of these drugs in different patient populations.

Similarly, the results were not reported stratified by whether the patient had atrial fibrillation or atrial flutter, thus we cannot report the evidence separately for those 2 conditions. We feel this is appropriate, however, because those 2 arrhythmias frequently coexist.^66,67 None of the trials had echocardiographic data as inclusion or exclusion criteria, but that information is more relevant to decisions regarding anticoagulation or cardioversion. We know of no study that associates echocardiographic data with ventricular rate control.

We cannot exclude the presence of publication bias, although we are confident that our search strategy did capture the published literature. In our review of the 8 non–English-language abstracts, we found them in agreement with the articles published in the English literature.

Future research should address the outcomes most relevant to patients — well-being and functionality—particularly since the relationship between heart rate control and exercise tolerance is unclear. We encourage the use of validated instruments for assessment, although which instruments are most appropriate is unknown at this time. Similarly, systematic recording of adverse events should be a regular component of all future trials of these drugs.

Recommendations for clinical practice

For adults with nonpostoperative atrial fibrillation, the evidence supports the following statements. The nondihydropyridine calcium-channel blockers, diltiazem, and verapamil are efficacious for heart rate control at rest and with exercise without decrement in exercise tolerance. Selected b-blockers, such as the noncardioselective b-antagonist nadolol or the second-generation b1-antagonists atenolol and metoprolol, are efficacious at rest and with exercise. There is some evidence, however, that b-blockers cause a transient decrease in exercise tolerance. For patients unlikely to exercise, such as those markedly incapacitated by other illness, digoxin should provide acceptable control.

· Acknowledgments ·

This study was conducted by the Johns Hopkins Evidence-based Practice Center through contract no. 290-97-006 from the Agency for Health Care Policy and Research, Rockville, Maryland. The authors are responsible for its content, including any clinical recommendations. No statement of this article should be construed as an official position of the Agency for Health Care Policy and Research or the US Department of Health and Human Services. We would like to thank Dr Francis Chesley of the Agency for Health Care Policy and Research and Drs Hanan Bell and Michael LeFevre of the American Academy of Family Physicians for their helpful suggestions regarding this project, Dr David Haines of the American College of Cardiology and Drs Ronald Berger and Gary Gerstenblith for their expert advice, Paul Abboud for assistance with data abstraction, and Donna Lea for extensive help with the manuscript.

Despite pharmacologic and electrical interventions, sinus rhythm cannot be restored and maintained in many patients with atrial fibrillation. For these patients, control of the ventricular rate is a primary goal of therapy, since a rapid rate may lead to worsening congestive heart failure, myocardial ischemia, or distressing breathlessness and palpitations.

A number of review articles have described strategies for rate control, principally involving the use of digoxin, calcium-channel blockers, and b-blockers.^1-6 A recent analysis of the trends in the use of drugs for ventricular rate control found that the use of digoxin and b-blockers decreased between 1980 and 1981 and 1994 and 1996, and the use of the nondihydropyridine calcium-channel blockers diltiazem and verapamil increased.⁷ These investigators, however, indicated that “current practices are dictated more by clinical tradition than by clinical science.”⁷ There has not been a systematic review of the trials evaluating the efficacy of both the familiar and the newer medications for ventricular rate control in atrial fibrillation. It is increasingly clear the drugs that are used most often for heart rate control at rest may not be the most efficacious during exercise, and exercise tolerance is compromised by some drugs.⁶

The purpose of our review was to characterize the strength of the evidence regarding the efficacy of drugs used for ventricular rate control in atrial fibrillation.

Methods

Study Design

We performed a systematic literature review and synthesis of randomized controlled trials on ventricular rate control in atrial fibrillation. To be eligible for inclusion in our review, trials needed to meet the following criteria: address management of nonpostoperative atrial fibrillation or atrial flutter; include human data; include adult subjects; and present original data. Studies that included patients with postoperative atrial fibrillation were not excluded as long as those patients were only a minority of the included patients.

Literature Identification and Search Strategies

The primary source of literature for our review was the CENTRAL database of The Cochrane Collaboration, a comprehensive collection of controlled clinical trials from 1948 to the present. As a secondary source, we searched MEDLINE from 1966 to May 1998 to ensure completeness. Additionally, we used the related articles feature of PubMed, as well as recent search results submitted to the Baltimore Cochrane Center, the contents pages of recent relevant journals, and programs from recent cardiology meetings. Our search strategy included using the MeSH terms “atrial fibrillation” and “atrial flutter” as subject headings and text words, as well as “random allocation,” “double-blind method,” and “single-blind method.” The publication types were “randomized controlled trials” and “controlled clinical trials.”

Abstracts of the citations of randomized controlled clinical trials were reviewed independently by 2 members of the study team to identify articles that met the inclusion criteria. Only English-language articles were reviewed. However, we reviewed all English-language abstracts of non-English-language publications to assess qualitative consistency with our results.

Data Abstraction

A form was developed to extract information from the eligible articles regarding study quality, characteristics, and findings. The section on study quality was created after our review of forms used in meta-analytic studies^8,9 and a literature review^10,11 and with the assistance of The Cochrane Collaboration. The resultant form incorporated⁶ key questions used by The Cochrane Collaboration in its reviews and¹⁴ key questions identified by Detsky and colleagues.¹⁰ Our form was pilot-tested for clarity and reproducibility and revised as needed. The final version contained 22 questions assessing quality in the following 5 areas: representativeness (how well the study population was described); potential for bias and confounding; description of therapy (eg, how similarly the groups were treated); outcomes and follow-up; and statistical reporting and interpretation [Table 1]. Each question was worth a maximum of 2 points, and the score in each category was the percentage of points received out of the total available. The overall quality score was calculated as the average of the scores for the 5 categories.

The portion of the form for quantitative data abstraction included sections for subject inclusion and exclusion criteria, baseline subject characteristics, therapeutic protocols, and outcomes. We recorded the mean heart rate at rest, the mean maximum heart rate during exercise or immediately after exercise, the proportion of subjects who reached the goal heart rate reduction in each treatment arm, and any measure of exercise tolerance.

The review of study quality was done independently by 2 reviewers, and differences were resolved by consensus. Quantitative data were abstracted by one primary reviewer and then checked for accuracy by a secondary reviewer. The reviewers were not blinded to the author, institution, or journal, since it seemed unlikely that this information would make a significant difference in the results.¹²

Presentation of the Data

We constructed 3 evidence tables with the trials grouped according to the regimens being compared. [Table 2a][Table 2b] displays the quality scores and key design elements of each trial. [Table 3] contains a listing of the trials of the most frequent comparisons and the absolute differences in heart rates of patients using those therapies. [Table 4] shows the results from the trials for which there were few, if any, given comparisons.

Data was synthesized by creating scatter plots of mean heart rates at rest and with exercise for each of the main drug comparisons (Figures 1-5). The data were not amenable to formal mathematical pooling (ie, meta-analysis) because of significant qualitative heterogeneity among the studies.*

Results

Literature Yield

We retrieved 74 abstracts. Of these, 8 were abstracts of articles from non–English-language literature. Forty-five trials were eligible for inclusion in our review; the authors of those studies evaluated 17 different drugs and several combinations of drugs. Some of the included trials were not designed as rate-control studies; they were studies of pharmacologic conversion that included heart rate data.

Qualitative Synthesis

The following comparisons were made in the trials: calcium-channel blockers,¹⁴-²¹ b-blockers,^19-28 digoxin,^{14,22,28,30-32,38} or other drugs and combinations compared with placebo;^{14,28,29,33-38} calcium-channel blockers,^14,39-42 b-blockers,^23,29,43,44 or other drugs and combinations compared with digoxin;^{14,27,40,44-49} and other drug comparison trials.^{18-20,27,42,50-58} Many of the trials involved more than 2 treatment arms. Nearly all of the trials of calcium-channel blockers, b-blockers, and digoxin were designed to evaluate rate control. The studies of the other drugs were principally trials evaluating atrial fibrillation conversion that also reported heart rate data. Two of the digoxin trials were also aimed at evaluating conversion to sinus rhythm.

Study Design

As shown in [Table 2a][Table 2b], there were important similarities and differences among the trials evaluating the same therapies. The duration of the trials and route of administration of the drug are presented to aid in interpretation of the results.

Notably, all of the trials of any given comparison were done within 10 years of each other. This should reduce the likelihood of secular trends affecting the outcomes of the trials. The range in study sizes is extreme, although most of the trials enrolled fewer than 50 patients. Several trials had fewer than 10 participants, and it was anticipated that these small trials would have little power to detect differences between treatments.

The regimens differed among the trials of the same medication. The intravenous diltiazem and verapamil doses were fairly uniform, although the oral dosages and frequency of administration differed. Some of the b-blocker trials involved titration of the medication to effectiveness, and digoxin was often dosed to a target blood level. All these differences may have had an impact on the outcomes. The followup times ranged from minutes to 6 weeks. Short trials may be appropriate for intravenous agents; however, several of the trials assessed the outcome a very short time after a single oral dose of medication, before a therapeutic blood level could be expected.

Another notable feature of these trials was the permissibility of other agents during the trial, as detailed in the evidence table. Permitting the use of digoxin in trials testing other medications, without reporting the number of participants in each arm receiving digoxin, can potentially confound the results.

Although not shown in the evidence table, most trials had explicit inclusion criteria. Some required atrial fibrillation lasting longer than 1 month or longer than 6 months, and several specified a ventricular rate required for entry, such as more than 120 beats per minute or “rapid rate.” Exclusion criteria varied from none to stringent; it was particularly stringent in those studies that involved exercise, from which subjects with angina or significant congestive heart failure were often excluded. None of the trials used echocardiographic data as inclusion or exclusion criteria.

Quality Scores

Many studies were weakest in their description of the participants in the study arms, so it was not always possible to tell if the groups were similar. This can be seen in the exceptionally low scores in the “representativeness” category. The potential for bias and confounding varied markedly across the trials, as did the description of the therapies. It was often unclear which other therapies the patients may have been receiving. Generally, the investigators described the outcomes completely and objectively measured them with Holter monitoring or telemetry. The completeness of statistical reporting was variable, with many studies only reporting a P value without reporting a measure of variability in the outcomes.

The studies published more recently had slightly higher total quality scores. Total quality scores were strongly associated with the size of the study, with the larger studies receiving higher scores (P < .001).

Outcomes

As shown in [Table 3], all of the trials reported either the heart rate reduction [on] for the active drug compared with the comparative drug or the proportion of patients who reached the target heart rate. Many of the trials also evaluated the efficacy of the drugs during exercise. The exercise test itself varied among the trials and included measurement of distance walked on a treadmill, measurement of oxygen consumption, and workload tolerated on a stationary bicycle.

On the basis of our qualitative assessment of the trials, we felt that any mathematical pooling of the results would result in invalid estimates of treatment effects. This was because of the markedly different treatment regimens within each drug class and the differing goals of treatment (acute or chronic management). The mixed quality of the trials also argued against pooling.

Study Results

Calcium-Channel Blockers Versus Placebo for Rate Control. All comparisons of calcium-channel blockers with placebo demonstrated that the calcium-channel blocker was more efficacious than placebo at reducing heart rate both at rest and during exercise. Five of the trials used diltiazem, 4 used verapamil, and 1 evaluated both drugs. An improvement in exercise tolerance was almost always seen, although different measures of tolerance were used. All but 2 of the trials allowed the participants to use digoxin but did not report what percentage of subjects in each treatment group received the drug. Despite different rates in the placebo arms, there was uniformity in treatment effects across the trials [Figure 1].

b-Blockers Versus Placebo for Rate Control. Seven different b-blockers were tested. Only 7 of the 12 comparisons demonstrated efficacy of the b-blocker at rest, although all were efficacious during exercise. The efficacy appears to be medication dependent. Atenolol (at 50 mg daily²⁸ or twice daily²⁰ or 100 mg daily²⁸) performed significantly better than placebo. Timolol (1 mg intravenously) allowed more subjects to reach the target heart rate compared with placebo.²⁴ Pindolol²⁸ (5 mg or 15 mg twice daily) and nadolol²⁵ (titrated dose) significantly reduced mean resting heart rate. The data regarding xamoterol were mixed.^20-23 Celiprolol²⁶ and labetalol were no more efficacious than placebo at rest.

All of the tested b-blockers demonstrated a significant reduction in heart rate with exercise compared with placebo; this included atenolol, labetalol, nadolol, celiprolol, and xamoterol.^{19,20,22,24,25} [Figure 2] shows that the effect of b-blockers on heart rate during exercise was more uniform than their impact on heart rate control at rest. However, these trials suggest that exercise tolerance in patients with atrial fibrillation may be reduced with b-blockers.

As in the calcium-channel blocker studies, most of the trials allowed subjects to continue on digoxin.Digoxin Versus Placebo. The outcomes with digoxin were mixed, as shown in [Table 3] and [Figure 3]. Two of the trials of digoxin versus placebo did not demonstrate a reduction in mean resting heart rate;^28,30 in 5 trials, however, there was a reduction.^{14,22,31,32,38} Two of these studies included patients on verapamil in both arms, so we could not attribute all of the rate reduction to digoxin alone.^31,32

Two studies evaluating digoxin during exercise did not find a significant heart rate reduction.^14,22 In one trial that suggested a difference, no measure of statistical significance was provided.²⁸ Four of the studies of digoxin and placebo evaluated exercise tolerance.^14,22,28,29 In one14 the cardiac output was higher for patients taking digoxin, and in another²² the time on the treadmill was longer with digoxin although the maximal attainable heart rate blood pressure product was higher with placebo.

Calcium-Channel Blockers Versus Digoxin for Rate Control. Three trials compared diltiazem with digoxin,^14,19,40 and 3 compared verapamil with digoxin^14,41,42 with the outcomes reported in [Table 3] and in [Figure 4]. The scatter plot is most useful for noting the trend toward improved control with calcium-channel blockers both at rest and with exercise. Notably, the cardiac output on digoxin during exercise was greater than in the 2 diltiazem groups (12.6 L/min vs 10.9 L/min and 9.1 L/min for 60 mg and 120 mg, respectively).¹⁴ Conversely, the group receiving verapamil was able to exercise longer on the treadmill than the digoxin group.⁴³ This latter study, however, had methodologic flaws, including little description of the participants.

b-Blockers Versus Digoxin for Rate Control. Four trials compared b-blockers with digoxin for rate control in atrial fibrillation, and the outcomes are reported in [Table 3] and in Figure 5.^22,28,42,43 Similar to the results of the trials of b-blockers compared with placebo, the efficacy of b-blockers was most convincing in the trials that evaluated their use during exercise. There appeared to be little difference between the efficacy of b-blockers and digoxin at rest.

Other Drugs and Combinations Versus Placebo and Digoxin. [Table 4] summarizes the outcomes for the few trials of other agents. Not surprisingly, of the 8 trials that compared digoxin with a combination of digoxin with a calcium-channel blocker,^{14,19,22,38,40,41,45,46} only one study did not find a significant decrease in mean resting heart rate with the addition of the calcium-channel blocker.⁴⁰ In 5 of the 6 studies with an exercise evaluation,^{14,19,38,41,45} the combination of a calcium-channel blocker and digoxin controlled the heart rate better than digoxin alone, while the sixth trial did not report the statistical significance of this outcome.¹⁴ Of the trials of a b}-blocker combined with digoxin, all were more effective than placebo, and all were more effective than digoxin alone except for the combination of digoxin and labetolol.²⁸ During exercise, however, this combination was more effective than either comparison arm.

Other Drugs Evaluated for Rate Control. There were 9 other randomized controlled trials of drugs for rate control in atrial fibrillation.^50-58 Two studies compared intravenous magnesium sulfate with intravenous verapamil for acute control.^50,51 In both studies, a higher percentage of subjects reached a heart rate of less than 100 beats per minute with verapamil than with magnesium sulfate.

Two studies evaluated rate control with propafenone or flecainide, both at 2 mg per kg intravenously for 1 hour; both significantly reduced the heart rate from baseline.^52,53 In both studies, subjects were allowed to continue on digoxin, calcium-channel blockers, and b-blockers. The side effects of flecainide were of more concern than those of propafenone, with conduction abnormalities in the flecainide group. Another study compared propafenone with quinidine for rate control.⁵⁴ Propafenone significantly slowed the heart rate at rest compared with quinidine. Either drug effectively slowed the heart rate compared with baseline.

Disopyramide did not reduce the mean resting heart rate from baseline.⁵⁵ The combination of diltiazem and digoxin reduced the mean resting heart rate to a greater degree than the combination of propranolol and digoxin, but all 3 drugs together were even more effective.⁵⁶ That study also demonstrated that with exercise the combination of propranolol and digoxin was more efficacious for heart rate control than diltiazem and digoxin and that the 3-drug combination was not better than just propranolol and digoxin. The combination of pindolol and digoxin reduced the maximum area under the heart rate curve significantly more than verapamil and digoxin.⁵⁷ Finally, the combination of amiodarone and digoxin slowed the resting heart rate when compared with baseline, while the combination of quinidine, verapamil, and digoxin did not; this was a small trial, however, and the baseline resting heart rates were not rapid.⁵⁸

Discussion

The randomized controlled trials of diltiazem and verapamil used by patients with atrial fibrillation provide strong evidence for their efficacy in reducing heart rate both at rest and with exercise when compared with placebo. In all of the studies that evaluated calcium-channel blockers compared with placebo during exercise, the calcium-channel blockers produced either an increase in cardiac output, oxygen consumption, or distance walked. There was also moderate evidence that diltiazem or verapamil was more effective at heart rate control both at rest and during exercise in the direct comparisons with digoxin, with a more rapid onset of action. Although digoxin appeared to increase cardiac output, verapamil prolonged time on the treadmill and increased oxygen consumption. Thus, the evidence strongly supports the use of diltiazem or verapamil for ventricular rate control in atrial fibrillation. Although they have a negative inotropic effect, reflex responses to vasodilatation usually result in a small increase in cardiac output. Therefore, except in moderate to severe heart failure, the negative inotropic effect is often not clinically apparent.⁶⁰

All of the tested b-blockers successfully reduced heart rate with exercise when compared with placebo, and most of them reduced resting heart rate. The effect on exercise tolerance was variable and may be due to the different receptor specificities of the tested drugs and the varying treatment times before testing.

When administered acutely, b-blocking agents depress myocardial function secondary to the withdrawal of adrenergically mediated inotropic and chronotropic support. However, in patients with congestive heart failure treated for longer than 1 month, b-blockers may improve myocardial function by improving intrinsic systolic function.^59,61-63 Extrapolating to patients with atrial fibrillation, b-blocker therapy may increase left ventricular ejection fraction compared with placebo if administered for longer than 1 month. None of the trials in this review lasted for longer than 4 weeks, so it is conceivable that the worsening exercise tolerance was a transient effect of the drug.

When compared with digoxin, the trials favored the use of b-blockers, as digoxin was less efficacious than metoprolol in resting heart rate reduction and less efficacious than labetalol at rate reduction during exercise. Furthermore, time on the treadmill was longer with both labetalol and metoprolol than with digoxin.

It will be interesting to see the effect of the third-generation b-blockers, such as carvedilol, on heart rate control in atrial fibrillation. We anticipate that they will be effective at ventricular rate control, with an improvement in exercise tolerance. Celiprolol and xameterol are not available in the United States and are no longer being evaluated for approval by the Food and Drug Administration.

Although several of the studies comparing digoxin with placebo were limited by subtherapeutic serum levels of digoxin at the time of evaluation, the others did show a resting heart rate reduction. There is little evidence to support the efficacy of digoxin for heart rate control with exercise. However, in the 2 trials that evaluated exercise tolerance on digoxin compared with placebo, cardiac output and time on the treadmill were greater with digoxin.

The trials evaluating other drugs, including propafenone, clonidine, and amiodarone yielded insufficient evidence to support their use for rate control at this time. These drugs appear promising and may prove efficacious when more evidence becomes available.

The optimal degree of heart rate control for patients with atrial fibrillation is unclear, particularly during exercise. Certainly an excessively rapid rate impairs ventricular filling and decreases cardiac output. However, severely limiting the heart rate acceleration that is needed to maintain cardiac output can also limit exercise tolerance.⁶³ There are few empirical trials of this. One recent study found that ventricular rate control in atrial fibrillation had no impact on cardiovascular performance as measured by endurance on a treadmill.⁶³ The studies in our review that report only on heart rate control during exercise without mention of exercise tolerance may be of less value to clinicians.

It is important to recognize that both heart rate control and exercise tolerance are surrogate outcomes for what is truly important to clinicians and their patients: their well-being, ability to conduct their daily activities, and mortality. Although several of the studies did inquire about symptoms such as palpitations or breathlessness, the use of validated quality-of-life questionnaires was rare. A study comparing pharmacologic treatment of atrial fibrillation with atrioventricular junction ablation and pacing demonstrated the use of several different quality-of-life measures: a Quality-of-Life Questionnaire, Specific Symptoms Scale, New York Heart Association Classification, and a Specific Activity Scale.⁶⁴ The authors of that study commented that they were uncertain whether the sensitivities of these scales were high enough to use as they did.

Our study is the first comprehensive evidence-based review that focused on this aspect of the management of atrial fibrillation. Several recent reviews described ventricular rate control trials in the context of other pharmacologic therapies for managing atrial fibrillation,^65,2-4 but those studies did not comprehensively review all the rate control trials. Furthermore, they contained recommendations for drugs that have never been evaluated in controlled trials of patients with atrial fibrillation exclusively (such as intravenous esmolol and intravenous propranolol).²

Limitations

Our systematic review was subject to the same limitations that are common to most reviews. The important differences among the trials preclude mathematical pooling and have to be taken into account when drawing conclusions using these studies. As in most assessments of study quality, our assessment tool was tailored to fit the topic, so the scores cannot be directly compared with quality assessments of studies on other topics. Our intent was to be able to compare the trials.

Few of the studies evaluated adverse effects in any systematic way. This inconsistent reporting of adverse effects limited our ability to comment on them. Many of the trials incompletely described the enrolled participants, so applying these results to all patients with atrial fibrillation should be done cautiously. Also, since results were seldom stratified by the clinical features of the enrolled patients, we could not report on the evidence supporting the use of these drugs in different patient populations.

Similarly, the results were not reported stratified by whether the patient had atrial fibrillation or atrial flutter, thus we cannot report the evidence separately for those 2 conditions. We feel this is appropriate, however, because those 2 arrhythmias frequently coexist.^66,67 None of the trials had echocardiographic data as inclusion or exclusion criteria, but that information is more relevant to decisions regarding anticoagulation or cardioversion. We know of no study that associates echocardiographic data with ventricular rate control.

We cannot exclude the presence of publication bias, although we are confident that our search strategy did capture the published literature. In our review of the 8 non–English-language abstracts, we found them in agreement with the articles published in the English literature.

Future research should address the outcomes most relevant to patients — well-being and functionality—particularly since the relationship between heart rate control and exercise tolerance is unclear. We encourage the use of validated instruments for assessment, although which instruments are most appropriate is unknown at this time. Similarly, systematic recording of adverse events should be a regular component of all future trials of these drugs.

Recommendations for clinical practice

For adults with nonpostoperative atrial fibrillation, the evidence supports the following statements. The nondihydropyridine calcium-channel blockers, diltiazem, and verapamil are efficacious for heart rate control at rest and with exercise without decrement in exercise tolerance. Selected b-blockers, such as the noncardioselective b-antagonist nadolol or the second-generation b1-antagonists atenolol and metoprolol, are efficacious at rest and with exercise. There is some evidence, however, that b-blockers cause a transient decrease in exercise tolerance. For patients unlikely to exercise, such as those markedly incapacitated by other illness, digoxin should provide acceptable control.

· Acknowledgments ·

This study was conducted by the Johns Hopkins Evidence-based Practice Center through contract no. 290-97-006 from the Agency for Health Care Policy and Research, Rockville, Maryland. The authors are responsible for its content, including any clinical recommendations. No statement of this article should be construed as an official position of the Agency for Health Care Policy and Research or the US Department of Health and Human Services. We would like to thank Dr Francis Chesley of the Agency for Health Care Policy and Research and Drs Hanan Bell and Michael LeFevre of the American Academy of Family Physicians for their helpful suggestions regarding this project, Dr David Haines of the American College of Cardiology and Drs Ronald Berger and Gary Gerstenblith for their expert advice, Paul Abboud for assistance with data abstraction, and Donna Lea for extensive help with the manuscript.

Despite pharmacologic and electrical interventions, sinus rhythm cannot be restored and maintained in many patients with atrial fibrillation. For these patients, control of the ventricular rate is a primary goal of therapy, since a rapid rate may lead to worsening congestive heart failure, myocardial ischemia, or distressing breathlessness and palpitations.

A number of review articles have described strategies for rate control, principally involving the use of digoxin, calcium-channel blockers, and b-blockers.^1-6 A recent analysis of the trends in the use of drugs for ventricular rate control found that the use of digoxin and b-blockers decreased between 1980 and 1981 and 1994 and 1996, and the use of the nondihydropyridine calcium-channel blockers diltiazem and verapamil increased.⁷ These investigators, however, indicated that “current practices are dictated more by clinical tradition than by clinical science.”⁷ There has not been a systematic review of the trials evaluating the efficacy of both the familiar and the newer medications for ventricular rate control in atrial fibrillation. It is increasingly clear the drugs that are used most often for heart rate control at rest may not be the most efficacious during exercise, and exercise tolerance is compromised by some drugs.⁶

The purpose of our review was to characterize the strength of the evidence regarding the efficacy of drugs used for ventricular rate control in atrial fibrillation.

Methods

Study Design

We performed a systematic literature review and synthesis of randomized controlled trials on ventricular rate control in atrial fibrillation. To be eligible for inclusion in our review, trials needed to meet the following criteria: address management of nonpostoperative atrial fibrillation or atrial flutter; include human data; include adult subjects; and present original data. Studies that included patients with postoperative atrial fibrillation were not excluded as long as those patients were only a minority of the included patients.

Literature Identification and Search Strategies

The primary source of literature for our review was the CENTRAL database of The Cochrane Collaboration, a comprehensive collection of controlled clinical trials from 1948 to the present. As a secondary source, we searched MEDLINE from 1966 to May 1998 to ensure completeness. Additionally, we used the related articles feature of PubMed, as well as recent search results submitted to the Baltimore Cochrane Center, the contents pages of recent relevant journals, and programs from recent cardiology meetings. Our search strategy included using the MeSH terms “atrial fibrillation” and “atrial flutter” as subject headings and text words, as well as “random allocation,” “double-blind method,” and “single-blind method.” The publication types were “randomized controlled trials” and “controlled clinical trials.”

Abstracts of the citations of randomized controlled clinical trials were reviewed independently by 2 members of the study team to identify articles that met the inclusion criteria. Only English-language articles were reviewed. However, we reviewed all English-language abstracts of non-English-language publications to assess qualitative consistency with our results.

Data Abstraction

A form was developed to extract information from the eligible articles regarding study quality, characteristics, and findings. The section on study quality was created after our review of forms used in meta-analytic studies^8,9 and a literature review^10,11 and with the assistance of The Cochrane Collaboration. The resultant form incorporated⁶ key questions used by The Cochrane Collaboration in its reviews and¹⁴ key questions identified by Detsky and colleagues.¹⁰ Our form was pilot-tested for clarity and reproducibility and revised as needed. The final version contained 22 questions assessing quality in the following 5 areas: representativeness (how well the study population was described); potential for bias and confounding; description of therapy (eg, how similarly the groups were treated); outcomes and follow-up; and statistical reporting and interpretation [Table 1]. Each question was worth a maximum of 2 points, and the score in each category was the percentage of points received out of the total available. The overall quality score was calculated as the average of the scores for the 5 categories.

The portion of the form for quantitative data abstraction included sections for subject inclusion and exclusion criteria, baseline subject characteristics, therapeutic protocols, and outcomes. We recorded the mean heart rate at rest, the mean maximum heart rate during exercise or immediately after exercise, the proportion of subjects who reached the goal heart rate reduction in each treatment arm, and any measure of exercise tolerance.

The review of study quality was done independently by 2 reviewers, and differences were resolved by consensus. Quantitative data were abstracted by one primary reviewer and then checked for accuracy by a secondary reviewer. The reviewers were not blinded to the author, institution, or journal, since it seemed unlikely that this information would make a significant difference in the results.¹²

Presentation of the Data

We constructed 3 evidence tables with the trials grouped according to the regimens being compared. [Table 2a][Table 2b] displays the quality scores and key design elements of each trial. [Table 3] contains a listing of the trials of the most frequent comparisons and the absolute differences in heart rates of patients using those therapies. [Table 4] shows the results from the trials for which there were few, if any, given comparisons.

Data was synthesized by creating scatter plots of mean heart rates at rest and with exercise for each of the main drug comparisons (Figures 1-5). The data were not amenable to formal mathematical pooling (ie, meta-analysis) because of significant qualitative heterogeneity among the studies.*

Results

Literature Yield

We retrieved 74 abstracts. Of these, 8 were abstracts of articles from non–English-language literature. Forty-five trials were eligible for inclusion in our review; the authors of those studies evaluated 17 different drugs and several combinations of drugs. Some of the included trials were not designed as rate-control studies; they were studies of pharmacologic conversion that included heart rate data.

Qualitative Synthesis

The following comparisons were made in the trials: calcium-channel blockers,¹⁴-²¹ b-blockers,^19-28 digoxin,^{14,22,28,30-32,38} or other drugs and combinations compared with placebo;^{14,28,29,33-38} calcium-channel blockers,^14,39-42 b-blockers,^23,29,43,44 or other drugs and combinations compared with digoxin;^{14,27,40,44-49} and other drug comparison trials.^{18-20,27,42,50-58} Many of the trials involved more than 2 treatment arms. Nearly all of the trials of calcium-channel blockers, b-blockers, and digoxin were designed to evaluate rate control. The studies of the other drugs were principally trials evaluating atrial fibrillation conversion that also reported heart rate data. Two of the digoxin trials were also aimed at evaluating conversion to sinus rhythm.

Study Design

As shown in [Table 2a][Table 2b], there were important similarities and differences among the trials evaluating the same therapies. The duration of the trials and route of administration of the drug are presented to aid in interpretation of the results.

Notably, all of the trials of any given comparison were done within 10 years of each other. This should reduce the likelihood of secular trends affecting the outcomes of the trials. The range in study sizes is extreme, although most of the trials enrolled fewer than 50 patients. Several trials had fewer than 10 participants, and it was anticipated that these small trials would have little power to detect differences between treatments.

The regimens differed among the trials of the same medication. The intravenous diltiazem and verapamil doses were fairly uniform, although the oral dosages and frequency of administration differed. Some of the b-blocker trials involved titration of the medication to effectiveness, and digoxin was often dosed to a target blood level. All these differences may have had an impact on the outcomes. The followup times ranged from minutes to 6 weeks. Short trials may be appropriate for intravenous agents; however, several of the trials assessed the outcome a very short time after a single oral dose of medication, before a therapeutic blood level could be expected.

Another notable feature of these trials was the permissibility of other agents during the trial, as detailed in the evidence table. Permitting the use of digoxin in trials testing other medications, without reporting the number of participants in each arm receiving digoxin, can potentially confound the results.

Although not shown in the evidence table, most trials had explicit inclusion criteria. Some required atrial fibrillation lasting longer than 1 month or longer than 6 months, and several specified a ventricular rate required for entry, such as more than 120 beats per minute or “rapid rate.” Exclusion criteria varied from none to stringent; it was particularly stringent in those studies that involved exercise, from which subjects with angina or significant congestive heart failure were often excluded. None of the trials used echocardiographic data as inclusion or exclusion criteria.

Quality Scores

Many studies were weakest in their description of the participants in the study arms, so it was not always possible to tell if the groups were similar. This can be seen in the exceptionally low scores in the “representativeness” category. The potential for bias and confounding varied markedly across the trials, as did the description of the therapies. It was often unclear which other therapies the patients may have been receiving. Generally, the investigators described the outcomes completely and objectively measured them with Holter monitoring or telemetry. The completeness of statistical reporting was variable, with many studies only reporting a P value without reporting a measure of variability in the outcomes.

The studies published more recently had slightly higher total quality scores. Total quality scores were strongly associated with the size of the study, with the larger studies receiving higher scores (P < .001).

Outcomes

As shown in [Table 3], all of the trials reported either the heart rate reduction [on] for the active drug compared with the comparative drug or the proportion of patients who reached the target heart rate. Many of the trials also evaluated the efficacy of the drugs during exercise. The exercise test itself varied among the trials and included measurement of distance walked on a treadmill, measurement of oxygen consumption, and workload tolerated on a stationary bicycle.

On the basis of our qualitative assessment of the trials, we felt that any mathematical pooling of the results would result in invalid estimates of treatment effects. This was because of the markedly different treatment regimens within each drug class and the differing goals of treatment (acute or chronic management). The mixed quality of the trials also argued against pooling.

Study Results

Calcium-Channel Blockers Versus Placebo for Rate Control. All comparisons of calcium-channel blockers with placebo demonstrated that the calcium-channel blocker was more efficacious than placebo at reducing heart rate both at rest and during exercise. Five of the trials used diltiazem, 4 used verapamil, and 1 evaluated both drugs. An improvement in exercise tolerance was almost always seen, although different measures of tolerance were used. All but 2 of the trials allowed the participants to use digoxin but did not report what percentage of subjects in each treatment group received the drug. Despite different rates in the placebo arms, there was uniformity in treatment effects across the trials [Figure 1].

b-Blockers Versus Placebo for Rate Control. Seven different b-blockers were tested. Only 7 of the 12 comparisons demonstrated efficacy of the b-blocker at rest, although all were efficacious during exercise. The efficacy appears to be medication dependent. Atenolol (at 50 mg daily²⁸ or twice daily²⁰ or 100 mg daily²⁸) performed significantly better than placebo. Timolol (1 mg intravenously) allowed more subjects to reach the target heart rate compared with placebo.²⁴ Pindolol²⁸ (5 mg or 15 mg twice daily) and nadolol²⁵ (titrated dose) significantly reduced mean resting heart rate. The data regarding xamoterol were mixed.^20-23 Celiprolol²⁶ and labetalol were no more efficacious than placebo at rest.

All of the tested b-blockers demonstrated a significant reduction in heart rate with exercise compared with placebo; this included atenolol, labetalol, nadolol, celiprolol, and xamoterol.^{19,20,22,24,25} [Figure 2] shows that the effect of b-blockers on heart rate during exercise was more uniform than their impact on heart rate control at rest. However, these trials suggest that exercise tolerance in patients with atrial fibrillation may be reduced with b-blockers.

As in the calcium-channel blocker studies, most of the trials allowed subjects to continue on digoxin.Digoxin Versus Placebo. The outcomes with digoxin were mixed, as shown in [Table 3] and [Figure 3]. Two of the trials of digoxin versus placebo did not demonstrate a reduction in mean resting heart rate;^28,30 in 5 trials, however, there was a reduction.^{14,22,31,32,38} Two of these studies included patients on verapamil in both arms, so we could not attribute all of the rate reduction to digoxin alone.^31,32

Two studies evaluating digoxin during exercise did not find a significant heart rate reduction.^14,22 In one trial that suggested a difference, no measure of statistical significance was provided.²⁸ Four of the studies of digoxin and placebo evaluated exercise tolerance.^14,22,28,29 In one14 the cardiac output was higher for patients taking digoxin, and in another²² the time on the treadmill was longer with digoxin although the maximal attainable heart rate blood pressure product was higher with placebo.

Calcium-Channel Blockers Versus Digoxin for Rate Control. Three trials compared diltiazem with digoxin,^14,19,40 and 3 compared verapamil with digoxin^14,41,42 with the outcomes reported in [Table 3] and in [Figure 4]. The scatter plot is most useful for noting the trend toward improved control with calcium-channel blockers both at rest and with exercise. Notably, the cardiac output on digoxin during exercise was greater than in the 2 diltiazem groups (12.6 L/min vs 10.9 L/min and 9.1 L/min for 60 mg and 120 mg, respectively).¹⁴ Conversely, the group receiving verapamil was able to exercise longer on the treadmill than the digoxin group.⁴³ This latter study, however, had methodologic flaws, including little description of the participants.

b-Blockers Versus Digoxin for Rate Control. Four trials compared b-blockers with digoxin for rate control in atrial fibrillation, and the outcomes are reported in [Table 3] and in Figure 5.^22,28,42,43 Similar to the results of the trials of b-blockers compared with placebo, the efficacy of b-blockers was most convincing in the trials that evaluated their use during exercise. There appeared to be little difference between the efficacy of b-blockers and digoxin at rest.

Other Drugs and Combinations Versus Placebo and Digoxin. [Table 4] summarizes the outcomes for the few trials of other agents. Not surprisingly, of the 8 trials that compared digoxin with a combination of digoxin with a calcium-channel blocker,^{14,19,22,38,40,41,45,46} only one study did not find a significant decrease in mean resting heart rate with the addition of the calcium-channel blocker.⁴⁰ In 5 of the 6 studies with an exercise evaluation,^{14,19,38,41,45} the combination of a calcium-channel blocker and digoxin controlled the heart rate better than digoxin alone, while the sixth trial did not report the statistical significance of this outcome.¹⁴ Of the trials of a b}-blocker combined with digoxin, all were more effective than placebo, and all were more effective than digoxin alone except for the combination of digoxin and labetolol.²⁸ During exercise, however, this combination was more effective than either comparison arm.

Other Drugs Evaluated for Rate Control. There were 9 other randomized controlled trials of drugs for rate control in atrial fibrillation.^50-58 Two studies compared intravenous magnesium sulfate with intravenous verapamil for acute control.^50,51 In both studies, a higher percentage of subjects reached a heart rate of less than 100 beats per minute with verapamil than with magnesium sulfate.

Two studies evaluated rate control with propafenone or flecainide, both at 2 mg per kg intravenously for 1 hour; both significantly reduced the heart rate from baseline.^52,53 In both studies, subjects were allowed to continue on digoxin, calcium-channel blockers, and b-blockers. The side effects of flecainide were of more concern than those of propafenone, with conduction abnormalities in the flecainide group. Another study compared propafenone with quinidine for rate control.⁵⁴ Propafenone significantly slowed the heart rate at rest compared with quinidine. Either drug effectively slowed the heart rate compared with baseline.

Disopyramide did not reduce the mean resting heart rate from baseline.⁵⁵ The combination of diltiazem and digoxin reduced the mean resting heart rate to a greater degree than the combination of propranolol and digoxin, but all 3 drugs together were even more effective.⁵⁶ That study also demonstrated that with exercise the combination of propranolol and digoxin was more efficacious for heart rate control than diltiazem and digoxin and that the 3-drug combination was not better than just propranolol and digoxin. The combination of pindolol and digoxin reduced the maximum area under the heart rate curve significantly more than verapamil and digoxin.⁵⁷ Finally, the combination of amiodarone and digoxin slowed the resting heart rate when compared with baseline, while the combination of quinidine, verapamil, and digoxin did not; this was a small trial, however, and the baseline resting heart rates were not rapid.⁵⁸

Discussion

The randomized controlled trials of diltiazem and verapamil used by patients with atrial fibrillation provide strong evidence for their efficacy in reducing heart rate both at rest and with exercise when compared with placebo. In all of the studies that evaluated calcium-channel blockers compared with placebo during exercise, the calcium-channel blockers produced either an increase in cardiac output, oxygen consumption, or distance walked. There was also moderate evidence that diltiazem or verapamil was more effective at heart rate control both at rest and during exercise in the direct comparisons with digoxin, with a more rapid onset of action. Although digoxin appeared to increase cardiac output, verapamil prolonged time on the treadmill and increased oxygen consumption. Thus, the evidence strongly supports the use of diltiazem or verapamil for ventricular rate control in atrial fibrillation. Although they have a negative inotropic effect, reflex responses to vasodilatation usually result in a small increase in cardiac output. Therefore, except in moderate to severe heart failure, the negative inotropic effect is often not clinically apparent.⁶⁰

All of the tested b-blockers successfully reduced heart rate with exercise when compared with placebo, and most of them reduced resting heart rate. The effect on exercise tolerance was variable and may be due to the different receptor specificities of the tested drugs and the varying treatment times before testing.

When administered acutely, b-blocking agents depress myocardial function secondary to the withdrawal of adrenergically mediated inotropic and chronotropic support. However, in patients with congestive heart failure treated for longer than 1 month, b-blockers may improve myocardial function by improving intrinsic systolic function.^59,61-63 Extrapolating to patients with atrial fibrillation, b-blocker therapy may increase left ventricular ejection fraction compared with placebo if administered for longer than 1 month. None of the trials in this review lasted for longer than 4 weeks, so it is conceivable that the worsening exercise tolerance was a transient effect of the drug.

When compared with digoxin, the trials favored the use of b-blockers, as digoxin was less efficacious than metoprolol in resting heart rate reduction and less efficacious than labetalol at rate reduction during exercise. Furthermore, time on the treadmill was longer with both labetalol and metoprolol than with digoxin.

It will be interesting to see the effect of the third-generation b-blockers, such as carvedilol, on heart rate control in atrial fibrillation. We anticipate that they will be effective at ventricular rate control, with an improvement in exercise tolerance. Celiprolol and xameterol are not available in the United States and are no longer being evaluated for approval by the Food and Drug Administration.

Although several of the studies comparing digoxin with placebo were limited by subtherapeutic serum levels of digoxin at the time of evaluation, the others did show a resting heart rate reduction. There is little evidence to support the efficacy of digoxin for heart rate control with exercise. However, in the 2 trials that evaluated exercise tolerance on digoxin compared with placebo, cardiac output and time on the treadmill were greater with digoxin.

The trials evaluating other drugs, including propafenone, clonidine, and amiodarone yielded insufficient evidence to support their use for rate control at this time. These drugs appear promising and may prove efficacious when more evidence becomes available.

The optimal degree of heart rate control for patients with atrial fibrillation is unclear, particularly during exercise. Certainly an excessively rapid rate impairs ventricular filling and decreases cardiac output. However, severely limiting the heart rate acceleration that is needed to maintain cardiac output can also limit exercise tolerance.⁶³ There are few empirical trials of this. One recent study found that ventricular rate control in atrial fibrillation had no impact on cardiovascular performance as measured by endurance on a treadmill.⁶³ The studies in our review that report only on heart rate control during exercise without mention of exercise tolerance may be of less value to clinicians.

It is important to recognize that both heart rate control and exercise tolerance are surrogate outcomes for what is truly important to clinicians and their patients: their well-being, ability to conduct their daily activities, and mortality. Although several of the studies did inquire about symptoms such as palpitations or breathlessness, the use of validated quality-of-life questionnaires was rare. A study comparing pharmacologic treatment of atrial fibrillation with atrioventricular junction ablation and pacing demonstrated the use of several different quality-of-life measures: a Quality-of-Life Questionnaire, Specific Symptoms Scale, New York Heart Association Classification, and a Specific Activity Scale.⁶⁴ The authors of that study commented that they were uncertain whether the sensitivities of these scales were high enough to use as they did.

Our study is the first comprehensive evidence-based review that focused on this aspect of the management of atrial fibrillation. Several recent reviews described ventricular rate control trials in the context of other pharmacologic therapies for managing atrial fibrillation,^65,2-4 but those studies did not comprehensively review all the rate control trials. Furthermore, they contained recommendations for drugs that have never been evaluated in controlled trials of patients with atrial fibrillation exclusively (such as intravenous esmolol and intravenous propranolol).²

Limitations

Our systematic review was subject to the same limitations that are common to most reviews. The important differences among the trials preclude mathematical pooling and have to be taken into account when drawing conclusions using these studies. As in most assessments of study quality, our assessment tool was tailored to fit the topic, so the scores cannot be directly compared with quality assessments of studies on other topics. Our intent was to be able to compare the trials.

Few of the studies evaluated adverse effects in any systematic way. This inconsistent reporting of adverse effects limited our ability to comment on them. Many of the trials incompletely described the enrolled participants, so applying these results to all patients with atrial fibrillation should be done cautiously. Also, since results were seldom stratified by the clinical features of the enrolled patients, we could not report on the evidence supporting the use of these drugs in different patient populations.

Similarly, the results were not reported stratified by whether the patient had atrial fibrillation or atrial flutter, thus we cannot report the evidence separately for those 2 conditions. We feel this is appropriate, however, because those 2 arrhythmias frequently coexist.^66,67 None of the trials had echocardiographic data as inclusion or exclusion criteria, but that information is more relevant to decisions regarding anticoagulation or cardioversion. We know of no study that associates echocardiographic data with ventricular rate control.

We cannot exclude the presence of publication bias, although we are confident that our search strategy did capture the published literature. In our review of the 8 non–English-language abstracts, we found them in agreement with the articles published in the English literature.

Future research should address the outcomes most relevant to patients — well-being and functionality—particularly since the relationship between heart rate control and exercise tolerance is unclear. We encourage the use of validated instruments for assessment, although which instruments are most appropriate is unknown at this time. Similarly, systematic recording of adverse events should be a regular component of all future trials of these drugs.

Recommendations for clinical practice

For adults with nonpostoperative atrial fibrillation, the evidence supports the following statements. The nondihydropyridine calcium-channel blockers, diltiazem, and verapamil are efficacious for heart rate control at rest and with exercise without decrement in exercise tolerance. Selected b-blockers, such as the noncardioselective b-antagonist nadolol or the second-generation b1-antagonists atenolol and metoprolol, are efficacious at rest and with exercise. There is some evidence, however, that b-blockers cause a transient decrease in exercise tolerance. For patients unlikely to exercise, such as those markedly incapacitated by other illness, digoxin should provide acceptable control.

· Acknowledgments ·

This study was conducted by the Johns Hopkins Evidence-based Practice Center through contract no. 290-97-006 from the Agency for Health Care Policy and Research, Rockville, Maryland. The authors are responsible for its content, including any clinical recommendations. No statement of this article should be construed as an official position of the Agency for Health Care Policy and Research or the US Department of Health and Human Services. We would like to thank Dr Francis Chesley of the Agency for Health Care Policy and Research and Drs Hanan Bell and Michael LeFevre of the American Academy of Family Physicians for their helpful suggestions regarding this project, Dr David Haines of the American College of Cardiology and Drs Ronald Berger and Gary Gerstenblith for their expert advice, Paul Abboud for assistance with data abstraction, and Donna Lea for extensive help with the manuscript.