Primary care practitioners and cardiologists must choose from a number of available pharmacological antiarrhythmic agents for the management of non-postoperative AF without any consensus regarding the best agent. Today, multiple antiarrhythmic agents are used both for acute conversion of AF and for maintenance of sinus rhythm after cardioversion. In this systematic synthesis of the literature, we describe all of the randomized clinical trials of the various antiarrhythmic agents that address either or both outcomes of acute conversion and maintenance of sinus rhythm for non-postoperative AF. In this format, the rates of conversion, maintenance of sinus rhythm, and adverse effects can be interpreted in light of the characteristics of the subjects enrolled and other specifics of the treatment regimens. Our primary aim in systematically presenting these studies is to improve the understanding of the evidence regarding the relative efficacy and relative morbidity of the various antiarrhythmic agents for both acute conversion of AF and maintenance of sinus rhythm after cardioversion.
Although clinicians frequently choose one antiarrhythmic agent for both initial conversion and subsequent maintenance of sinus rhythm, we felt it was important to separate the evidence. This type of analysis will provide a clear picture of the efficacy of each agent for these two distinct aspects in the management of AF. To further this goal, we also separate the data, when possible, in terms of AF versus atrial flutter because a large percentage of clinical trials involved a mixture of subjects with AF and atrial flutter.
Designs of the Studies
This review of the literature identified 46 randomized clinical trials on acute pharmacological conversion of AF that included the following antiarrhythmic agents, according to the Vaughan-Williams (1970; 1984; Singh, 1998) classification system:
| Class Ia: | Quinidine (Quinidex, Cardioquin), procainamide (Pronestyl, Procan SR, Procanbid), disopyramide (Norpace) |
| Class Ic: | Flecainide (Tambocor), propafenone (Rhythmol) |
| Class II: | Beta-blockers (timolol, practolol) |
| Class III: | Amiodarone (Cordarone, Pacerone), sotalol (Betapace), ibutilide (Corvert), dofetilide |
| Class IV: | Verapamil, diltiazem |
| Class V: | Digoxin |
| Miscellaneous: | Cibenzoline, pirmenol, pilsicainide, magnesium |
For purposes of discussion of the evidence, we refer to the following as the major antiarrhythmic agents for acute conversion of AF: quinidine, procainamide, disopy¡ramide, flecainide, propafenone, amiodarone, sotalol, and ibutilide/dofetilide. The older agents of beta-blockers, calcium-channel-blockers (verapamil, diltiazem), and digoxin are currently viewed as relatively ineffective therapy for conversion. However, they are frequently adjunctive therapy via their rate control abilities (Cobbe, 1997). Our search identified them for this review based on their evaluation in relatively older clinical trials. For ease of presentation in the evidence table, we have grouped these older agents, some of the newest agents, and agents not commonly used into the Miscellaneous category at the end of the tables. These antiarrhythmic agents typically have only one identified clinical trial addressing their use.
The list of antiarrhythmic agents identified in 29 randomized clinical trials regarding maintenance of sinus rhythm in AF is as follows:
| Class Ia: | Quinidine, disopyramide |
| Class Ic: | Flecainide, propafenone |
| Class III: | Amiodarone, sotalol, N-acetylprocainamide |
| Class IV: | Verapamil |
| Miscellaneous: | Cibenzoline, bidisomide |
Comparable to the note above on the antiarrhythmic agents for conversion, we refer to the following as the major antiarrhythmic agents for maintenance of sinus rhythm in AF: quinidine, disopyramide, flecainide, propafenone, amiodarone, and sotalol. The calcium-channel-blocker verapamil is currently viewed as ineffective therapy for maintenance of sinus rhythm. It was identified in this review based on relatively older clinical trials. We group the evidence for cibenzoline and bidisomide at the end of the table in a manner similar to that used in the table for conversion. In addition, we identified one trial using N-acetylprocainamide. We report these results also at the end of the table because use of this agent is not common and because all subjects in that trial had atrial flutter.
The most established of the major antiarrhythmic agents for both conversion and subsequent maintenance of sinus rhythm in AF is quinidine, with the first published randomized clinical trial in 1971 (Hillestad, Bjerkelund, Dale et al., 1971). Because of limitations in quinidine's efficacy and questions of its safety, however, better agents for pharmacological management of AF have been sought. This has led to the current list of multiple agents with varying mechanisms of action and to a spurt of published randomized clinical trials beginning in the late 1980s. A 1998 review by Waldo and Prystowsky in the American Journal of Cardiology (Waldo and Prystowsky, 1998) listed eight agents as widely available in the United States: quinidine, procainamide, disopyramide, flecainide, propafenone, moricizine (a Ic-like drug), amiodarone, and sotalol. (Our literature review did not identify any randomized clinical trials using moricizine for the treatment of AF.) At present, only three of these drugs-quinidine, flecainide, and propafenone-have approval from the U.S. Food and Drug Admini¡stration for the treatment of AF. Quinidine has broad approval for treatment of AF, whereas flecainide and propafenone are approved for paroxysmal AF in the absence of structural heart disease (Waldo and Prystowsky, 1998).
In addition to trials evaluating these eight agents and trials that tested negative chronotropic agents (beta-blockers, verapamil, diltiazem, and digoxin), our review found trials with several newer agents. These include cibenzoline (Class I with features of Classes III and IV), bidisomide (Class I), pirmenol (Class Ia), pilsicainide (Class Ic), and ibutilide and dofetilide (Class III). Cibenzoline currently is not available in the United States (Waldo and Prystowsky, 1998). We also identified one trial on conversion using magnesium, which has electrophysiologic properties similar to Class IV, the calcium-channel-blockers.
Our review also identified a subsample of studies (n = 9) evaluating the use of direct current cardioversion in conjunction with a particular antiarrhythmic agent (e.g., quinidine, propafenone, sotalol, and verapamil).
Evidence Tables 1 and 2 present the pertinent study design details for the studies that assessed acute conversion and maintenance of sinus rhythm. A total of 46 clinical trials were reviewed regarding acute conversion of AF, and 29 trials were reviewed with respect to maintenance of sinus rhythm after cardioversion of AF. In addition, one trial on acute conversion and one trial on maintenance of sinus rhythm included only patients with atrial flutter. Although included in the tables, neither trial is incorporated in any data synthesis. Eight trials presented data on both outcomes and are included in both Evidence Tables 1 and 2. The order of studies in the tables is by the Vaughan-Williams classes; any trials using Class Ia agents are listed first, followed by Class Ic, and so forth. Miscellaneous drugs, those with only one published trial, and trials comparing only verapamil, diltiazem, and digoxin are reported at the end of the tables. Several issues need to be considered when reviewing these evidence tables, and these points will be highlighted in the following discussion.
Comparison Groups
The comparison groups used in the studies differ considerably. Some studies compared two pharmacological agents without a placebo group, whereas other studies incorporated a placebo group for comparison. Furthermore, some studies used as the comparison group verapamil, diltiazem, or digoxin-agents widely regarded to have no significant impact on conversion or maintenance of sinus rhythm rates (Cobbe, 1997).
A crossover study design was common. With particular respect to the outcome of acute conversion of AF, these evidence tables reflect only the first treatment group to go through each arm of the study. This orchestration of the data was required because the subjects who were successfully converted in the first phase were not able to participate in the second phase. As a result, the groups compared in the second phase of the crossover studies were less likely to be comparable than the groups in the first phase. The resulting potential bias was felt to outweigh the value of any additional data received from the crossover arms. For trials on maintenance of sinus rhythm, we were able to use all phases of crossover studies. The trials that used a crossover design can be identified in the tables by the fact that the overall subject number (N) is considerably less than the sum of subjects in each treatment arm.
Number of Studies
The number of studies for each particular antiarrhythmic agent ranges from 15 for propafenone, in acute conversion, to 1 for agents such as cibenzoline, pilsicainide, disopyramide, magnesium, pirmenol, timolol, and bidisomide. The strength of our summary data depends on the amount of data that is aggregated. Any conclusions made about agents with only one or two studies need to be viewed cautiously, especially because many of these studies were small.
Publication Year
The year of study publication ranged from 1970 to 1997. Of all the studies listed, however, only six were published in the 1970s. The vast majority of the randomized clinical trials occurred in the late 1980s and the 1990s.
Number of Subjects
With one exception, the number of enrolled subjects varied from 11 to 300. The exception was the Atrial Fibrillation Investigation with Bidisomide (AFIB) Investigators (1994) evaluating maintenance of sinus rhythm, which had a sample size of 1,189. Unfortunately, this trial was the only one evaluating bidisomide, and it was prematurely terminated due to lack of efficacy and the suggestion of increased mortality associated with bidisomide use. Thus, all of the evidence for the major antiarrhythmic agents is based on trials with small to modest numbers of subjects. The ability to detect small differences in efficacy is hampered by these sample sizes. This problem is minimized, however, by the availability of multiple studies for many of the antiarrhythmic agents.
Treatment Regimens
The treatment regimens for each individual antiarrhythmic agent are heterogeneous. This includes the actual dose administered; the route of administration; use of loading doses; and allowance of nonassigned additional pharmacological agents such as digoxin, beta-blockers, and calcium-channel-blockers.
Also, as previously mentioned, nine clinical trials on conversion incorporated direct current cardioversion in conjunction with pharmacological therapy. These regimens involved a pharmacological loading period followed in usually 1 to 7 days by direct current cardioversion. The results of these studies are presented in two tables. First, acute conversion rates due to the pharmacological intervention alone are presented with the other studies on pharmacological conversion (
Evidence Table 3). Second, we created an additional table on cardioversion results that shows the data from these nine trials only with respect to cardioversion via pharmacological and electrical means combined (
Evidence Table 4). One of these direct current cardioversion studies (
Hillestad, Bjerkelund, Dale et al., 1971) is shown only in
Evidence Table 4 because no data were provided on cardioversion rates for pharmacological therapy alone.
Inclusion and Exclusion Criteria
All studies had explicit inclusion and exclusion criteria. Of particular note relative to inclusion criteria, there was some variability in the inclusion of subjects with paroxysmal versus persistent AF, in the inclusion of subjects with atrial flutter, and in the specified duration of AF required for enrollment. Because the majority of studies did not provide their working definition of paroxysmal AF, we were unable to arrive at a consensus definition for paroxysmal AF for these evidence tables. Similarly, for the few trials that described using subjects with paroxysmal atrial flutter, a consensus definition was not obtainable. However, because the clinical outcomes of conversion and maintenance of sinus rhythm differ between paroxysmal and persistent AF and atrial flutter, we noted in the results tables whenever a study stated that it enrolled paroxysmal AF or atrial flutter subjects. Related to this issue, several studies reported including "recent onset" or "acute" AF. Because this type of wording may suggest paroxysmal AF in some cases and thus may affect reported outcomes, this is also noted in the comments sections of the results tables.
Several studies included subjects with atrial flutter. Overall, 11 studies on conversion and 8 on maintenance of sinus rhythm described inclusion of such subjects. In particular, the trials by Guo, Ellenbogen, Wood et al. (1996) for conversion and Feld (1989) for maintenance of sinus rhythm included only subjects with atrial flutter. Because antiarrhythmic agents may have different efficacy for conversion and maintenance of sinus rhythm in subjects with atrial flutter as opposed to AF, this inclusion is important to note. Whenever studies reported data separately for subjects with AF and subjects with atrial flutter, our main outcome data for both conversion and maintenance of sinus rhythm reflect subjects with AF only. Atrial flutter data in these trials is shown in the comments section of the results table. The subsequent mathematical pooling of data comparably reflects AF data only, as far as is possible based on the published data. When separation of AF and atrial flutter data was not possible, our later data pooling explored this issue by creating subgroups based on inclusion of subjects with atrial flutter.
The specified duration of AF for inclusion as a study subject varied considerably. Because of the potential association between length of time in AF and difficulty of both conversion and maintenance of sinus rhythm, our subsequent meta-analysis evaluated the outcomes stratifying the studies based on duration of AF greater than 2 weeks versus less than or equal to 2 weeks. This factor, in addition to the inclusion of subjects with atrial flutter, was felt a priori to potentially have significant impact on study outcomes.
In terms of exclusion criteria, we report on the following factors that could be associated with differences in outcomes based on their known association with the development of AF: recent myocardial infarction or history of angina pectoris, prior use of other antiarrhythmic agents, presence of congestive heart failure, and presence of other system disease (hepatic, renal, pulmonary, and/or endocrine). Based on the Manitoba Followup Study (Krahn, Manfreda, Tate et al., 1995), the age-adjusted relative risks (with 95 percent confidence intervals) for the development of AF were 3.5 (2.7 to 4.5) for ischemic heart disease and 9.9 (7.0 to 14.1) for congestive heart failure. A prior use of other antiarrhythmic agents may signify subjects whose AF is particularly resistant to pharmacological management. The presence of other system disease includes conditions known to be etiologic for AF (e.g., hyperthyroidism) and conditions that may affect choice of antiarrhythmic agent (e.g., hepatic failure). Overall, all of these four areas of exclusion involve factors with a known impact on the development of AF and, therefore, potential impact on the response of AF to pharmacological therapy. As a whole, most studies excluded subjects with a recent myocardial infarction or history of angina pectoris or use of other antiarrhythmic agents. Subjects with congestive heart failure or other system disease involving the hepatic, renal, pulmonary, and/or endocrine systems were systematically excluded by roughly half of the studies. When it appears that these factors may affect the observed trial results, comments to that effect are made in the ensuing discussion. Although not shown explicitly in Evidence Tables 1 and 2, within each of these exclusion categories we discovered a range of definitions. For example, criteria ranged from relatively objective New York Heart Association criteria to an array of subjective symptoms for the definition of congestive heart failure. Likewise, the presence of other system disease entailed numerous systems in some studies but only one pertinent system in others.
Overall Study Quality
As described in the methods section of this report, we evaluated study quality in five domains: representativeness of population, bias and confounding, therapy description, outcomes assessment, and statistical analysis. By reporting scores in all domains as well as an overall score, we hope to encourage readers to form their own impression of the strengths and weaknesses of each study. In addition, this review may point out issues to be addressed with future trials. It is important to keep in mind that a limitation of using published reports to assess study design quality is that the evaluation is a mixture of design assessment and reporting assessment.
The studies had the largest variability in terms of representativeness of the popu¡lation, with a range of scores from 0 to 100 percent of possible points on the quality assessment form developed for this project. Although a portion of this variability likely represents variability in reporting rather than true representativeness of the study population, it nevertheless makes it difficult for clinicians to know whether the results of the studies are likely to apply to the types of patients they see in their own practices. All of the other four domains of quality also had considerable variation in scores, with the following ranges: bias and confounding, 17 to 100 percent; therapy description, 25 to 100 percent; outcomes assessment, 25 to 90 percent; and statistical analysis, 0 to 100 percent. Again, percentages are out of all possible points for each domain on our quality assessment form.
The overall quality scores ranged from a low of 36 percent to a high of 94 percent; however, only nine studies had an overall quality score less than 50 percent. Two of these studies were in the small subsample that evaluated antiarrhythmic agents in conjunction with direct current cardioversion (Hillestad, Bjerkelund, Dale et al., 1971; Rasmussen, Wang, and Fausa, 1981). The third study had one of the smallest subject numbers, 15, and therefore will have minimal impact on summary data (Barranco, Sanchez, Rodriguez et al., 1994). The next five studies evaluated flecainide versus cibenzoline (Kuhlkamp, Schmid, Risler et al., 1991), short-acting quinidine versus long-acting quinidine (Normand, Legendre, Kahn et al., 1976), quinidine plus practolol versus quinidine plus placebo (Levi, Proto, and Rovetta, 1973), intravenous flecainide versus oral flecainide (Crijns, van Wijk, van Gilst et al., 1988), and disopyramide versus amiodarone for maintenance of sinus rhythm (Martin, Benbow, Leach et al., 1986). These are the only trials to use these pharmacological comparison groups, and therefore they cannot be combined with other trials for data synthesis. The last trial assessed flecainide versus propafenone for maintenance of sinus rhythm, with a sample size of 97 and an overall quality score of 48 (Aliot and Denjoy, 1996).
With respect to each outcome, the top 25 percent of trials on conversion had overall quality scores between 78 percent and 94 percent, and the top 25 percent of trials on maintenance of sinus rhythm had overall quality scores between 70 percent and 80 percent. Because we realized that any quality assessment reflected not only the methodological quality of the trial but also the reporting, we carefully looked at all the identified trials. Overall, we felt that all the randomized controlled trials had acceptable quality for inclusion in this evidence synthesis.
Despite the limitations in the quality and size of some of the studies, we concluded that there was enough evidence on most of the drug comparisons to warrant a systematic synthesis of the evidence.
Results of the Studies
Evidence Tables 3, 4, and 5 present data on subject characteristics, conversion success rates (
Evidence Table 3) or maintenance of sinus rhythm success rates (
Evidence Table 5), and adverse effect rates for the included studies. As mentioned previously,
Evidence Table 4 reports on only nine studies with the outcome being conversion by pharmacological therapy in conjunction with direct current cardioversion. Figures to present plots of absolute rates and meta-analysis results. These figures will be referenced as appropriate in the text. Supplements to Figures - and - are available upon request (Supplemental Figures, 1999).
Comparison across studies should be done with attention to the different subject characteristics as well as differences in regimens and inclusion and exclusion criteria from Evidence Tables 1 and 2. The clinical characteristics reported in these tables were selected because of their potential impact on choice of therapeutic options. Furthermore, these tables should allow a clinician to judge whether his or her patient resembles the enrolled subjects enough to justify use of the results of a particular clinical trial.
Subject Characteristics
The information regarding subject characteristics is displayed in Evidence Tables 3, 4, and 5. The mean age of subjects ranged from 50 to 69 years old for the majority of trials. This is considerably younger than would be expected from a population sample of subjects with AF. In fact, of the studies that reported mean ages, only 23 percent (10/44) of the conversion articles and only 7 percent (2/27) of the maintenance articles had mean ages greater than 65 years. Three trials did not report a mean age. Because the incidence of AF increases with increasing age (Krahn, Manfreda, Tate et al., 1995) and because response to pharmacological therapy may change with increasing age, it is important to evaluate this effect. Unfortunately, because the trials did not uniformly present age-stratified results, we were not able to explore this association further. Our evidence as a whole is derived from the cumulative population with a mean age range of 50 to 69 years. The age range of subjects was sporadically reported and is not included in these evidence tables.
The majority of trials had roughly even sex distributions, but seven trials were skewed with a distribution of at least 80/20 between the sexes (Suttorp, Kingma, Lie-A-Huen et al., 1989; Guo, Ellenbogen, Wood et al., 1996; Stambler, Wood, Ellenbogen et al., 1996; Ellenbogen, Stambler, Wood et al., 1996; Feld, Nademanee, Noll et al., 1989; Hou, Chang, Chen et al., 1995; Juul-Moller, Edvardsson, and Rehnqvist-Ahlberg, 1990). The first is a trial of flecainide versus verapamil, with 80 percent male subjects. The next three were all trials of the newer agents ibutilide/dofetilide, making one cautious in applying results of these newer antiarrhythmic agents to women. The next study (Feld, Nademanee, Noll et al., 1989) used all male subjects and evaluated an uncommon comparison of quinidine versus N-acetylprocainamide with the entire subject population having atrial flutter. This study, therefore, will not be combined with other studies for further analysis. The study by Hou (1995) looked at amiodarone versus digoxin with 86 percent male subjects, and the study by Juul-Moller (1990) with 81 percent male subjects evaluated quinidine versus sotalol. Based on results of the Framingham Study (Prystowsky, Benson, Fuster et al., 1996), there is evidence for differing incidence rates between the sexes. Because of the potential of differing responses to pharmacological therapy based on sex, it is important to evaluate this effect. Our evidence does not currently allow such an exploration. The outcomes reported here, in general, reflect a fairly evenly mixed population of men and women.
The mean left-atrial diameter in all trials in which it was reported ranged from 3.3 to 5.0 centimeters. Because of the known association between left-atrial enlargement and AF (Vaziri, Larson, Benjamin et al., 1993), this factor may be a potential confounder if it is also associated with response to pharmacological therapy. Our later mathematical pooling of the data includes a subgroup analysis of response based on mean left-atrial diameter less than or equal to 4.0 centimeters versus greater than 4.0 centimeters.
The percentage of subjects with coronary artery disease usually ranged from 0 to 35 percent. As discussed in the section on inclusion and exclusion criteria relative to study design, the presence of ischemic heart disease has an age-adjusted relative risk for AF of 3.5 (Krahn, Manfreda, Tate et al., 1995). Although this may be a potential confounder of the evidence, we did not perform subgroup analysis based on presence of coronary artery disease because the definition of coronary artery disease varied considerably among the trials or was not defined at all. Given concerns of ventricular arrhythmia in patients using Class Ia agents, presence of coronary artery disease would clearly influence choice of antiarrhythmic agent.
Hypertension was present in 5 to 52 percent of subjects in studies where it was reported. As previously mentioned, the Manitoba Followup Study (Krahn, Manfreda, Tate et al., 1995) estimated a relative risk of AF in the presence of hypertension to be 2.3. Comparable to the situation regarding presence of coronary artery disease, we did not perform subgroup analysis to explore this potential confounder because of the inconsistent definition of hypertension in the trials.
The presence of valvular heart disease had more variability, with a range of 0 to 71 percent, although in the majority of studies less than 40 percent of subjects had valvular heart disease. The Manitoba Followup Study (Krahn, Manfreda, Tate et al., 1995) estimated the relative risk of AF with valvular disease to be 7.0. Because of inconsistent definitions of valvular disease by the reviewed trials, however, a subgroup analysis of this potential confounder was not possible based on the evidence.
The inclusion of subjects with atrial flutter was reported sporadically but typically fell in the range of 0 to 40 percent, with two exceptions. Both Guo, Ellenbogen, Wood et al. (1996) and Feld (1989) reported that 100 percent of their subjects had atrial flutter. Because our target population for this literature review was subjects with AF, whenever possible we extracted the data separately for AF versus atrial flutter. Our tables and subsequent mathematical pooling of the data reflect as pure a population of AF as possible.
For the outcome of maintenance of sinus rhythm, we noted the percentage of subjects with paroxysmal AF due to the expected differences in successful maintenance of sinus rhythm between paroxysmal and more persistent forms of AF. Unfortunately, the percentage of subjects with paroxysmal AF also was reported sporadically. Several studies made blanket statements that the subjects had paroxysmal AF, and this is noted in the comments section of the results tables. As mentioned before, we also noted when studies stated that the enrolled subjects had "recent onset" or "acute" AF and paroxysmal atrial flutter. In examining the evidence, our discussion highlights instances where outcome differences may be influenced by the use of subjects with paroxysmal AF versus persistent AF. We were not able to do subgroup analysis of these two categories of AF because of the inconsistent definitions and reporting provided by the trials.
Acute Conversion of AF
Evidence tables for acute conversion of AF
The reported pharmacological conversion rates for each study are shown in
Evidence Table 3.
Evidence Table 4 reports on the combined outcome of pharmacological conversion in conjunction with direct current cardioversion and will be discussed at the end of this section. A crude summary of the rates for successful conversion of AF by each of the major antiarrhythmic agents and the number of trials for each agent is as follows:
| Quinidine (Quinaglute, Quinidex) | 11--86% | (n = 6) |
| Procainamide (Procan) | 63% | (n = 1) |
| Disopyramide (Norpace) | 23% | (n = 1) |
| Flecainide (Tambocor) | 52-95% | (n = 8) |
| Propafenone (Rythmol) | 6--91% | (n = 15) |
| Amiodarone Cordarone) | 25--92% | (n = 6) |
| Sotalol (Betapace) | 8--52% | (n = 3) |
| Ibutilide (Corvert)/Dofetilide | 10--49% | (n = 3) |
| Control treatment | 0--76% | (n = 25) |
These numbers must be interpreted in light of all of the study design characteristics mentioned previously. In particular, the broad range of conversion incidence rates for the control treatment groups likely reflects the inclusion of subjects with paroxysmal AF by some of the trials. Several agents, including beta-blockers, had scarce evidence and for ease of presentation are discussed at the end of this session as "miscellaneous antiarrhythmic agents."
For data synthesis purposes, we chose to combine the following as control treatment groups: placebo, verapamil, diltiazem, and digoxin. This was done to enhance combinability of the data given the known relative ineffectiveness of these agents for acute conversion of AF (Cobbe, 1997). As will be discussed later, our review of the literature supports this combination.
One trial (Guo, Ellenbogen, Wood et al., 1996) is presented in the table for completeness but is not included in later synthesis because 100 percent of subjects had atrial flutter.
For ease of examining this evidence, we group the results. First, we present the evidence comparing each major antiarrhythmic agent to control treatment. It should be noted that the bulk of evidence for any individual agent exists in this comparison. Second, we present the evidence for each individual major antiarrhythmic agent separately. Third, we present the evidence for miscellaneous comparisons and for use of direct current cardioversion in conjunction with pharmacological conversion.
The evidence on clinical predictors of success in pharmacological conversion was too sparse and disparate to be meaningfully summarized or presented in evidence tables.
A. Antiarrhythmic agents versus control treatment
In brief overview, the conversion success rates for the following major antiarrhythmic agents all showed a relative benefit of the various agents compared with control treatment: quinidine, flecainide, propafenone, amiodarone, and ibutilide/dofetilide. Notably, no trials were identified that evaluated procainamide versus control treatment. The data are equivocal on any benefit of disopyramide over control treatment. The one antiarrhythmic agent that is not supported for use over control treatment is sotalol. This information is displayed graphically in , which shows the absolute rates for each trial of the major antiarrhythmic agents. The magnitude of the difference between each agent and control treatment varied and, at least in some studies, is likely a result of the use of subjects with high spontaneous conversion rates due to underlying paroxysmal AF. Almost all studies that had control group conversion rates greater than 50 percent described their subjects as one of the following: "recent onset" AF, paroxysmal AF, or with a duration of AF less than 10 days. Such definitions may reflect a disease different from persistent AF. The one exception in this group is a relatively unusual study by Cowan, Gardiner, Reid et al. (1986). The subjects in this trial were all patients in the intensive care unit for proven or suspected myocardial infarction. Because short, self-terminating periods of AF can be seen with myocardial infarction and because 76 percent of the subjects were eventually proven to have suffered a myocardial infarction, it is difficult to determine the similarity of these subjects to ones in other trials. As mentioned above, one small trial evaluated sotalol (
Singh, Saini, DiMarco et al., 1991) and showed a 12 percent lower conversion success rate with sotalol versus placebo. These findings are in line with a recent review by Cobbe in the
European Heart Journal (1997), although that review did not look at the newer agents of ibutilide/dofetilide.
B. Individual antiarrhythmic agents versus all comparison groups
Figure 2. Proportion of subjects with successful pharmacological conversion
Notes:
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
**Vertical lines represent 95% confidence intervals for the proportion of subjects with successful pharmacological conversion
+ n equals the number of trials evaluating each comparison
Figure 9. Proportion of subjects with successful pharmacological conversion
Notes:
* Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
** Vertical lines represent 95% confidence intervals for the proportion of subjects with successful pharmacological conversion
+ n equals the number of trials evaluating each comparison
To provide clinicians with a sense of the relative efficacy for each of the major antiarrhythmic agents and to display the absolute rates, we have chosen to graphically summarize the evidence tables for each major agent in Figures to . These figures evaluate quinidine, procainamide, disopyramide, flecainide, propafenone, amiodarone, sotalol, and ibutilide and dofetilide. The majority of trials involving a comparison between two agents with high conversion rates for both treatment arms describe their subjects in terms suggestive of paroxysmal AF. This use of subjects with paroxysmal AF versus persistent AF may explain some of the observed differences in absolute rates shown in the figures.
As these tables and figures illustrate, not all possible treatment comparisons had published randomized clinical trials. We note these in the following discussion. In particular, the data were very sparse with respect to procainamide and disopyramide. The newer agents of ibutilide and dofetilide have had randomized clinical trials that compare the agents only to placebo administration.
Quinidine
For quinidine, summarizes all the identified trials on conversion of AF and specifies the comparison groups used. In general, the data are equivocal with respect to the comparison of quinidine with propafenone and do not support any benefit of using quinidine over amiodarone. The evidence does suggest that quinidine is more efficacious than sotalol. In addition, conversion rates in the quinidine groups were higher than the rates in the control treatment groups. Notably absent is any comparison of quinidine versus flecainide because no such randomized clinical trials were identified.
Procainamide and Disopyramide
Figure 3. Proportion of subjects with successful pharmacological conversion
Notes:
**Vertical lines represent 95% confidence intervals for the proportion of subjects with successful pharmacological conversion
+ n equals the number of trials evaluating each comparison
Figure 4. Proportion of subjects with successful pharmacological conversion
Notes:
* Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
** vertical lines represent 95% confidence intervals for the proportion of subjects with successful pharmacological conversion
+ n equals the number of trials evaluating each comparison
Figures and display the absolute rates for studies involving procainamide and disopyramide, respectively. In each case, only one trial was identified for each in terms of conversion of AF. These data do not support the benefit of procainamide over flecainide, and they are equivocal on any benefit of disopyramide over control treatment for conversion of AF.
Flecainide
Figure 5. Proportion of subjects with successful pharmacological conversion
Notes:
* Control treatment includes groups receiving placebo, Verapamil, diltiazem, or digoxin
** Vertical lines represent 95% confidence intervals for the proportion of subjects with successful pharmacological conversion
+ n equals the number of trials evaluating each comparison
The data for flecainide are shown in . Overall, these trials indicate that flecainide is more efficacious than procainamide, propafenone, and the control treatments. The data are equivocal for the comparison of flecainide versus amiodarone. As mentioned before, we did not identify any randomized trials comparing flecainide and quinidine. In addition, we did not identify any trials comparing flecainide and sotalol.
Propafenone
Figure 6. Proportion of subjects with successful pharmacological conversion
Notes:
* Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
** Vertical lines represent 95% confidence intervals for the proportion of subjects with successful pharmacological conversion
+ n equals the number of trials evaluating each comparison
shows the absolute rates for all trials involving propafenone. The data are equivocal in terms of propafenone versus quinidine, and they do not suggest a benefit of using propafenone compared with flecainide. However, the data do indicate an advantage in efficacy of propafenone as opposed to amiodarone and control treatments for acute conversion of AF. No trials were identified that addressed propafenone versus sotalol for conversion of AF.
Amiodarone
Figure 7. Proportion of subjects with successful pharmacological conversion
Notes:
* Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
** Vertical lines represent 95% confidence intervals for the proportion of subjects with successful pharmacological conversion
+ n equals the number of trials evaluating each comparison
With respect to amiodarone, as shown in , the absolute rates suggest a benefit of amiodarone as opposed to quinidine and control treatments. The data are equivocal in regard to amiodarone versus flecainide, and they do not indicate an advantage of amiodarone as compared with propafenone for acute conversion. No trials were identified addressing amiodarone versus sotalol.
Sotalol
Figure 8. Proportion of subjects with successful pharmacological conversion
Notes:
* Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
** Vertical lines represent 95% confidence intervals for the proportion of subjects with successful pharmacological conversion
+ n equals the number of trials evaluating each comparison
displays the data for trials using sotalol. This antiarrhythmic agent had three identified trials, and all used the comparison group of either quinidine or control treatment. Despite the limited number of trials, the data do not suggest an advantage of sotalol over either quinidine or control treatment for conversion of AF.
Ibutilide/dofetilide
shows the absolute rates for studies evaluating ibutilide/dofetilide. As mentioned, the only randomized trials identified all used control treatment comparison arms. These data uniformly indicate a benefit of ibutilide/dofetilide over control treatment for conversion of AF. Of note for this presentation of the evidence, one of the two studies evaluating ibutilide was completed for dose-finding purposes. Based on reviewing other published literature on ibutilide, the lowest dose arm in this trial of 0.005 mg/kg likely represents a sub-therapeutic dose. We therefore did not include the data from this arm of the trial when generating because this would have biased the study against finding ibutilide efficacious.
C. Miscellaneous antiarrhythmic agents
Figure 10. Proportion of subjects with successful pharmacological conversion
Notes:
* Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
** Vertical lines represent 95% confidence intervals for the proportion of subjects with successful pharmacological conversion
+ n equals the number of trials evaluating each comparison
In addition to these major antiarrhythmic agents, we identified several randomized trials using unusual agent combinations, uncommon agents, or older agents now known to be relatively ineffective for conversion of AF. displays the absolute rates for all of the studies that we term "miscellaneous" trials. Several key points are shown by these data. First, adding verapamil to quinidine therapy appears to improve conversion success relative to digoxin and quinidine. Unfortunately, no trials simply compared verapamil to digoxin to help answer the question of whether verapamil is efficacious alone or only when given in conjunction with quinidine. Second, use of the beta-blocking agent practolol-in conjunction with quinidine-or timolol did not show any superiority over placebo. Third, the relatively new agents of pilsicainide and pirmenol seem to show enhanced efficacy relative to control treatment, while the data on cibenzoline are equivocal relative to flecainide. Fourth, the data on using magnesium versus control treatment are equivocal. Fifth, the one trial evaluating intravenous flecainide versus oral flecainide for acute conversion of AF had equivocal results. Sixth, all comparisons of digoxin versus placebo show no apparent benefit of digoxin for acute conversion of AF. Comparably, a head-to-head match of verapamil and diltiazem showed very low conversion rates for both arms. These last comparisons of digoxin, verapamil, and diltiazem all support our use of these agents, in addition to placebo arms, as control treatment arms for data synthesis purposes.
D. Individual antiarrhythmic agents in conjunction with direct current cardioversion
With respect to antiarrhythmic agent use for conversion in conjunction with direct current cardioversion, the reported rates for successful conversion of AF and number of trials were as follows:
| Quinidine | 73--100% | (n = 5) |
| Propafenone | 75--84% | (n = 3) |
| Sotalol | 50--91% | (n = 3) |
| Control treatment | 60--96% | (n = 7) |
These incidence rates reflect a strategy of the sequential use of pharmacological therapy alone, then in conjunction with direct current cardioversion. In terms of a relative conversion rate as compared with a control treatment group that received direct current cardioversion without use of an antiarrhythmic agent, these studies do not suggest a benefit for quinidine, propafenone, or sotalol in conjunction with direct current cardioversion. Two trials evaluating direct current cardioversion in conjunction with quinidine versus sotalol do indicate a benefit of quinidine over sotalol. This benefit may reflect a benefit at the level of pharmacological conversion alone given the previously described data that support use of quinidine over sotalol for pharmacological conversion of AF. displays the data for all these trials involving pharmacological and electrical cardioversion of AF.
Meta-analysis for acute conversion of AF
In the interest of presenting our meta-analysis results for conversion of AF consistent with the evidence table discussion, we have grouped the data into tables similar to the discussed figures displaying absolute rates. Overall, our analysis deals primarily with the calculated odds ratio for each comparison. In order to present evidence usable to the reader, our final Summary and Conclusions (Efficacy of Antiarrhythmic Agents) of the evidence also reports the important comparisons in terms of the number of patients needed to be treated by one agent in order to see a benefit compared with the alternative treatment. Comparable to the preceding section on the evidence tables and figures of absolute rates, we present the results of our mathematical pooling of the data in the following order: antiarrhythmic agents compared with control treatment, each individual agent versus all comparison groups, miscellaneous comparisons, and use of direct current cardioversion in conjunction with pharmacological conversion.
Realizing it is difficult to categorize the strength of evidence, we felt it was important to facilitate interpretation of the odds ratio (OR). The concept of "strength of evidence" depends on the estimated magnitude of effect, the precision of that estimate, and our confidence that the true effect is different from zero. Quantitatively, all of these dimensions are captured by the point estimate and confidence interval, but it may still be difficult for all readers to interpret results presented only in that fashion. Therefore, we wanted to provide verbal descriptors of the strength of evidence. For the evidence on conversion, an OR greater than 1.0 represents a higher odds of conversion compared with the comparison group. We chose the following categorization of strength of evidence by noting the placement of the point estimate and the width of the confidence interval (CI) surrounding it:
-
Strong evidence of efficacy: OR > 1.0, 99 percent CI does not include 1.0 (p < 0.01).
-
Moderate evidence of efficacy: OR > 1.0, 95 percent CI does not include 1.0, but 99 percent CI includes 1.0 (0.01 < p <0.05).
-
Suggestive evidence of efficacy: 95 percent CI includes 1.0 in the lower tail (0.05< p <0.2 to 0.3) and the OR is in a clinically meaningful range.
-
Inconclusive evidence of efficacy: 95 percent CI widely distributed around 1.0.
-
Strong evidence of lack of efficacy: OR near 1.0, 95 percent CI is narrow.
When the point estimate was less than 1.0, we termed this negative efficacy and used the same categorization of strong, moderate, and suggestive evidence based on the CI.
In some cases, the previous categorizations did not completely capture some element of the result that was important for interpretation, particularly in cases with large point estimates and/or very wide confidence intervals. In those instances, we added additional descriptive text. The use of the 99 percent confidence interval (i.e., p = 0.01) as a boundary point for the descriptor "strong" was based on published statistical research on the relationship of p-values to Bayesian evidential summaries (Bayes factors), as well as the recognition that all meta-analyses have some degree of qualitative heterogeneity that is not reflected in the quantitative summaries.
A. Antiarrhythmic agents versus control treatment
With respect to each major antiarrhythmic agent relative to control treatment for acute conversion of AF, the meta-analysis results are displayed in . The bulk of evidence behind any individual antiarrhythmic agent occurred in comparison to control treatment. Overall, there is strong evidence of efficacy for the following agents compared with control treatment: flecainide, propafenone, and ibutilide/dofetilide. There is moderate evidence of efficacy for quinidine in regard to conversion of AF. The data suggest efficacy for both disopyramide and amiodarone compared with control treatment and suggest negative efficacy for sotalol compared with control treatment.
In more detail, the strongest evidence of efficacy relative to control treatment exists for flecainide and ibutilide/dofetilide in acute conversion of AF. The summary odds ratio for flecainide is 24.7 (CI 9.0 to 68.3) based on four trials; and for ibutilide/dofetilide the ratio is 29.1 (CI 9.8 to 86.1) based on three trials. Both point estimates are consistent with a large treatment effect size. One non-English-language article (Kondili, Kastrati, and Popa, 1990) with an English abstract supports the finding of flecainide having significantly better efficacy than control treatment.
Although they are both relatively new Class III agents, we thought it would be helpful to look at least briefly at the efficacy of ibutilide and dofetilide separately. Combining the two ibutilide studies, the data (OR 31.8, CI 9.9 to 102.8) remain relatively unchanged from the combined drug treatment estimate. Looking at the one trial on dofetilide, the treatment effect overall (OR 17.0, CI 1.0 to 296.1) is suggestive of efficacy of dofetilide compared with control treatment. Moreover, looking only at the high dofetilide dose arm (8 ´g/kg), because a clear dosage recommendation does not exist yet for dofetilide, the treatment effect (OR 28.3, CI 1.6 to 513.0) is comparable to that of ibutilide alone and to the combined ibutilide/dofetilide treatment effect estimate. The precision of this estimate, however, is substantially less than it is with the combined analysis because of smaller sample size.
Although the magnitude of treatment benefit compared with control treatment is less for propafenone (OR 4.6, CI 2.6 to 8.2), our meta-analysis yielded strong evidence of efficacy for conversion of AF for this agent. Though smaller than the effect size for flecainide and ibutilide/dofetilide, the evidence for propafenone is consistent with a fairly large effect size. It should be noted that using a fixed-effects model, this meta-analysis had significant quantitative heterogeneity between the studies that were pooled, making the validity of the summary data questionable. We, therefore, performed this mathematical pooling using a random-effects model. Both modeling results were consistent with strong evidence in favor of propafenone for conversion of AF. In particular, our analysis reveals that the heterogeneity of the results exists primarily among the four trials using intravenous propafenone. Although the point estimates of three of these trials suggest a moderate benefit with odds ratios of 2.0 to 6.0, the fourth trial suggests a substantial benefit with an odds ratio of 20.9. Overall, all four studies are consistent with a benefit of intravenous propafenone versus control treatment, and when pooled with the studies evaluating oral propafenone, strong evidence is shown of efficacy for using propafenone versus control treatment for acute conversion of AF. A non¡English-language article (Kondili, Kastrati, and Popa, 1990) found a higher, but not statistically significant, conversion rate for propafenone compared with control treatment.
The data on quinidine reflects a combination of three trials with a pooled odds ratio of 2.9 (CI 1.2 to 7.0), consistent with moderate evidence of efficacy for quinidine as opposed to control treatment. A non-English-language article by Luciardi (Luciardi, Berman, Santana et al., 1996) supports this by finding quinidine statistically more efficacious than control treatment. We should note that the categorization of evidence for quinidine as moderate is based on the fact that the 99 percent confidence interval for the point estimate includes unity (OR 2.9, 99 percent CI 0.9 to 9.2). The overall estimated effect size of quinidine treatment is modest.
The summary data for amiodarone (OR 5.7, CI 1.0 to 33.4) are consistent with suggestive evidence of efficacy for conversion compared with control treatment. The estimated effect size is fairly large. Similar to the situation with propafenone, this data pooling had significant quantitative heterogeneity between the included studies when using a fixed-effects model. We therefore used a random-effects model for this mathematical data pooling. Specifically, in the three trials evaluating the use of amiodarone versus control treatment, all of the point estimates are consistent with a benefit of amiodarone. Two trials suggest a moderate advantage with odds ratios of 1.4 and 4.9. The third trial, however, is compatible with a substantial benefit, having an odds ratio of 69.0. Overall, the mathematical pooling of all trials evaluating amiodarone provides suggestive evidence of an advantage of amiodarone versus control treatment for acute conversion of AF. It should be noted that unlike the propafenone comparison, the results of our fixed-effects and random-effects modeling differed for amiodarone. The fixed-effects model yielded strong evidence in favor of amiodarone relative to control treatment. The random-effects model yielded only suggestive evidence in favor of amiodarone. Of all comparisons for acute conversion of AF, amiodarone compared with control treatment is the most problematic from a qualitative standpoint. This is likely a result of two issues related to the specific trials being combined. First, all three trials were relatively small (Noc, Stajer, and Horvat, 1990; Cowan, Gardiner, Reid et al., 1986; Hou, Chang, Chen et al., 1995). Second, one trial (Cowan, Gardiner, Reid et al., 1986) involved all subjects who were in an intensive care unit for proven or suspected myocardial infarction. Because short, self-terminating periods of AF can be seen with myocardial infarction and because 76 percent of the subjects were eventually proven to have suffered a myocardial infarction, it is difficult to determine the similarity of these subjects to those in other trials. Therefore, the comparison of amiodarone to control treatment for acute conversion has the greatest qualitative, as well as quantitative, heterogeneity of all comparisons for acute conversion of AF.
The summary data for disopyramide and sotalol reflect the results of only one trial each, and the confidence intervals around the estimates include unity. These data, therefore, are suggestive of efficacy for disopyramide versus control treatment (OR 7.0, CI 0.3 to 153.0), although the precision of this estimate is poor. The evidence is suggestive of negative efficacy of sotalol compared with control treatment (OR 0.4, CI 0.0 to 3.0), although the precision of this estimate is also poor.
Because some antiarrhythmic agents are commonly considered together based on the Vaughan-Williams classification (1984), we grouped the data for Class Ia and Class Ic agents, in comparison to control treatment, together for one meta-analysis. The combined Ia agents are quinidine and disopyramide. The combined Ic agents are flecainide and propafenone. We did not think that any combination of the Class III agents (amiodarone, sotalol, ibutilide/dofetilide) was clinically relevant. Not surprisingly, these meta-analysis results show strong evidence of efficacy of both Class Ia (OR 3.2, CI 1.4 to 7.3) and Class Ic (OR 6.2, CI 3.5 to 10.8) agents relative to control treatment for acute conversion of AF. These results are also shown in . Given the known quantitative heterogeneity of the propafenone versus control treatment studies, we performed the mathematical pooling of data for the Class Ic agents using a random-effects model.
B. Individual antiarrhythmic agents versus all comparison groups
Quinidine
Figure 13. Meta-analysis on pharmacological conversion
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
* Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
Individual study data are available on request (Supplemental Figures, 1999)
The results of the meta-analysis of evidence regarding the use of quinidine for acute conversion of AF are shown in . The data showing moderate evidence of efficacy relative to control treatment have been illustrated already in . The one identified non-English article on quinidine versus control treatment (
Luciardi, Berman, Santana et al., 1996) also supports this finding. Evidence is sparse for the comparison of quinidine to other antiarrhythmic agents. The data for quinidine versus propafenone and quinidine versus amiodarone each reflect only one trial. As discussed in the section on the evidence tables, these data are inconclusive in terms of quinidine compared with propafenone (OR 0.4, CI 0.1 to 2.0). There is moderate evidence of negative efficacy for quinidine as opposed to amiodarone (OR 0.2, CI 0.1 to 0.9), with potentially a fairly large effect size based on the point estimate. The evidence behind the comparison of quinidine and sotalol reflects a pooling of two trials. This summary is consistent with strong evidence of efficacy and a fairly large effect size for quinidine compared with sotalol (OR 5.8, CI 2.4 to 14.2) for acute conversion of AF. In reviewing non-English-language articles, we found two that evaluated quinidine compared with other antiarrhythmic agents. The first one (
César, Serrano, Pamplona et al., 1994) found quinidine to be significantly more efficacious than amiodarone or procainamide. The second article (
Satullo, Arrigo, Cavallaro et al., 1996) found no difference in conversion rates between quinidine and propafenone.
Procainamide and disopyramide
Figure 14. Meta-analysis on pharmacological conversion
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
Individual study data available on request
Figure 15. Meta-analysis on pharmacological conversion
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
Individual study data available on request
The evidence shown in Figures and , for procainamide and disopyramide respectively, is limited to one trial each. As reviewed earlier with , these data on acute conversion of AF give strong evidence of negative efficacy of procainamide compared with flecainide (OR 0.1, CI 0.0 to 0.5) and suggestive evidence of efficacy of disopyramide over control treatment (OR 7.0, CI 0.3 to 153.0), although the precision for the estimate is poor. A non-English article (
César, Serrano, Pamplona et al., 1994) did not find procainamide more efficacious than quinidine. We were not able to determine the results of comparing procainamide and amiodarone in this trial (
César, Serrano, Pamplona et al., 1994).
Flecainide
Figure 16. Meta-analysis on pharmacological conversion
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
Individual study data available on request
summarizes the data for use of flecainide in acute conversion of AF. The summary odds ratio for flecainide is consistent with strong evidence of efficacy (OR 24.7, CI 9.0 to 68.3) relative to control treatment. The non-English article by Kondili (1990) supports this finding. In addition, the summary evidence is consistent with a large effect size. As discussed above, the evidence consists of only one trial showing strong evidence and a fairly large effect size of flecainide over procainamide (OR 7.4, CI 1.9 to 28.3). There is also strong evidence of an advantage of efficacy and a fairly large effect size for flecainide compared with propafenone (OR 5.1, CI 2.3 to 11.0), based on three trials. The comparison of flecainide and amiodarone has only one trial with inconclusive evidence for efficacy of flecainide relative to amiodarone (OR 2.5, CI 0.2 to 29.6) for conversion of AF.
Propafenone
Figure 17. Meta-analysis on pharmacological conversion
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
**Confidence interal using random-effects model
Individual study data available on request
The bulk of evidence behind propafenone is relative to control treatment and has been extensively discussed in the previous section on antiarrhythmic agents versus control treatment. Overall, the pooled data are consistent with strong evidence of efficacy of propafenone (OR 4.6, CI 2.6 to 8.2) relative to control treatment, with a fairly large effect size for acute conversion of AF. This is shown in . We should note that one non-English article (
Kondili, Kastrati, and Popa, 1990) found a higher, but not statistically significant, conversion rate for propafenone compared with control treatment. The evidence for propafenone versus quinidine consists of only one inconclusive trial (OR 2.3, CI 0.5 to 10.1). This is supported by a non-English-language article (
Satullo, Arrigo, Cavallaro et al., 1996). The comparison of propafenone and flecainide has three trials with strong evidence of negative efficacy of propafenone as opposed to flecainide (OR 0.2, CI 0.1 to 0.4). This is supported by the non-English article by Kondili (1990). The comparison of propafenone and amiodarone involves only one trial with strong evidence of efficacy and a potentially large effect size for propafenone compared with amiodarone (OR 13.0, CI 2.1 to 79.6) for acute conversion of AF. Unfortunately, the precision of this estimate is poor. One non-English-language article (
Treglia, Alfano, and Rossini, 1994) found no difference in conversion rates between propafenone and amiodarone.
Amiodarone
Figure 18. Meta-analysis on pharmacological conversion
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
**Confidence interval using random-effects model
Individual study data available on request
The evidence for amiodarone versus control treatment for acute conversion of AF has been discussed and is suggestive of efficacy (OR 5.7, CI 1.0 to 33.4). This is shown in . It should be again noted that this comparison is the most problematic in the area of acute conversion because of the substantial qualitative and quantitative heterogeneity between the combined studies. The comparisons of amiodarone with quinidine, flecainide, and propafenone all involve only one trial each. These trials give moderate evidence of efficacy of amiodarone instead of quinidine (OR 4.5, CI 1.2 to 17.4), are inconclusive for amiodarone relative to flecainide (OR 0.4, CI 0.0 to 4.9), and give strong evidence of negative efficacy of amiodarone compared with propafenone (OR 0.1, CI 0.0 to 0.5), although the precision of this potentially large effect size is poor. Two non-English articles evaluated amiodarone. The first (
César, Serrano, Pamplona et al., 1994) did not find amiodarone more efficacious than quinidine. The second (
Treglia, Alfano, and Rossini, 1994) found no difference in conversion rates between amiodarone and propafenone.
Sotalol
Figure 19. Meta-analysis on pharmacological conversion
Notes:
Vertical bars and point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
Individual study data available on request
The antiarrhythmic agent sotalol has relatively few trials evaluating efficacy. The summary data, shown in , for sotalol versus control treatment, based on only one trial, are suggestive of negative efficacy (OR 0.4, CI 0.0 to 3.0). The evidence for sotalol compared with quinidine consists of two trials. This result gives strong evidence of negative efficacy of sotalol as opposed to quinidine for acute conversion (OR 0.2, CI 0.1 to 0.4).
Ibutilide/dofetilide
Figure 20. Meta-analysis on pharmacological conversion
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
Individual study data available on request
As for the major antiarrhythmic agents for acute conversion of AF, the evidence for ibutilide/dofetilide consists only of trials with comparison groups of control treatment and is shown in . The summary evidence shows strong evidence of efficacy for ibutilide/dofetilide versus control treatment (OR 29.1, CI 9.8 to 86.1). Furthermore, the evidence is consistent with a large treatment effect. As discussed earlier, one trial on ibutilide (
Ellenbogen, Stambler, Wood et al., 1996) was primarily a dose-finding study. Based on our review of the literature, we excluded the lowest dose arm of ibutilide (0.005 mg/kg) in this study because it likely represents an underdosage of the agent. A subgroup analysis of the therapeutically dosed ibutilide and dofetilide arms versus this sub-therapeutic arm shows a significant benefit of the therapeutic dose (OR 5.1, CI 1.9 to 13.9). When only the two ibutilide studies are pooled, the data remain consistent with strong evidence of ibutilide benefit relative to control treatment (OR 31.8, CI 9.9 to 102.8). For dofetilide, there is strong evidence in support of the high dose arm (0.008 mg/kg) relative to control treatment (OR 28.3, CI 1.6 to 513.0) and suggestive evidence for the low dose arm (0.004 mg/kg) relative to control treatment (OR 9.6, CI 0.5 to 187.0). The wide confidence intervals reflect the paucity of data on dofetilide.
C. Miscellaneous antiarrhythmic agents
Figure 21. Meta-analysis on pharmacological conversion
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
Individual study data available on request
shows the meta-analysis results for all the miscellaneous antiarrhythmic trials. Nine of these ten comparisons involve only one trial. The other comparison, of digoxin versus placebo, incorporates three trials. In brief overview, the only trials with strong evidence of efficacy are as follows: pilsicainide is beneficial relative to control treatment (OR 8.7, CI 2.3 to 33.2) with a fairly large effect size, and pirmenol is beneficial relative to control treatment (OR 8.5, CI 1.9 to 38.8) with a fairly large effect size. There is moderate evidence of efficacy for quinidine with verapamil compared with quinidine with digoxin (OR 6.4, CI 1.4 to 28.4). The pooling of three studies evaluating digoxin versus placebo (OR 1.3, CI 0.8 to 2.0) gives strong evidence of lack of efficacy of digoxin and supports our use of digoxin as a control treatment for other comparison purposes.
D. Individual antiarrhythmic agents in conjunction with direct current cardioversion
represents the meta-analysis of trials evaluating pharmacological conversion of AF in conjunction with direct current cardioversion. The summary comparison of quinidine versus control treatment for use in conjunction with direct current cardioversion consists of three trials and is strong evidence of lack of efficacy (OR 0.9, CI 0.5 to 1.8). Comparably, the summary comparison of propafenone versus control treatment for use in conjunction with direct current cardioversion consists of three trials and is strong evidence of lack of efficacy (OR 1.2, CI 0.7 to 2.2). The comparison of sotalol versus placebo treatment for use in conjunction with direct current cardioversion involves only one trial and is inconclusive evidence of efficacy (OR 0.7, CI 0.2 to 3.0). With respect to the use of direct current cardioversion, two trials evaluated quinidine versus sotalol. The pooled results of these trials give moderate evidence of efficacy of quinidine compared with sotalol (OR 4.0, CI 1.1 to 15.0), with potentially a fairly large effect size when used in conjunction with direct current cardioversion for AF. As mentioned previously, this result may reflect the apparent benefit of quinidine over sotalol for pharmacological conversion alone, as supported by other evidence in this review.
Subgroup analysis for acute conversion of AF (Appendix P)
Because several important factors were felt a priori to be potential confounders, we examined the following factors in subgroup analysis: type of control treatment utilized (placebo versus verapamil/diltiazem versus digoxin), route of antiarrhythmic administration (intravenous versus oral), followup time for conversion outcome of less than 24 hours versus greater than or equal to 24 hours, duration of AF less than or equal to 2 weeks versus greater than 2 weeks, mean left-atrial size less than or equal to 4.0 centimeters versus greater than 4.0 centimeters, and inclusion of subjects with atrial flutter (no subjects versus some subjects with atrial flutter). It should be noted that the ability of a systematic literature review to detect subgroup differences is limited because the data are not at the level of each individual study subject but rather at the level of the overall study. For example, if one trial had a mean left-atrial size for all subjects of 3.7 centimeters, that trial as a whole was taken as evidence for conversion with left-atrial size less than or equal to 4.0 centimeters. We did not have enough data from the publications to separate subjects within any given trial for analysis of subgroups. In addition, incomplete information and inconsistent definitions of subgroups hampered our subgroup analyses.
Starting with type of control treatment used (placebo versus verapamil/diltiazem versus digoxin), we found no significant differences between results for the various control treatments. In particular, this analysis was relevant to the following antiarrhythmic agents: quinidine, flecainide, propafenone, amiodarone, and ibutilide/dofetilide. In each of these cases, there were no important differences in the magnitude of treatment effect based on which comparison group was used.
Comparably, our subgroup analysis evaluating route of antiarrhythmic agent administration revealed no important differences. This subgroup analysis applied only to the agents of flecainide and propafenone. Furthermore, two trials specifically addressed this issue. For flecainide, the study by Crijns, van Wijk, van Gilst et al. (1988) involved two treatment arms: intravenous flecainide versus oral flecainide. The result was equivocal for any differences between the two regimens, as is shown in . For propafenone, the trial by Boriani, Capucci, Lenzi et al. (1995) involved three treatment arms: intravenous propafenone, oral propafenone, and placebo. Analysis of the conversion outcome numbers for the first two arms yields no evidence for any signifi¡cant difference between the two administration routes for propafenone, as is shown in
Evidence Table 3.
Next we examined the subgrouping of followup time for conversion outcome of less than 24 hours versus greater than or equal to 24 hours. We felt that this analysis was important to detect any antiarrhythmic agent that had greater efficacy with use for greater than or equal to 24 hours as opposed to relatively rapid efficacy seen in less than 24 hours. In general, the majority of comparisons did not have enough data for performing this analysis. Furthermore, when the analysis was possible, all comparisons did not show any important difference in treatment effect based on goal followup time. The summary of this meta-analysis and supplemental figures are available upon request (Supplemental Figures, 1999).
Our subgroup analysis on duration of AF less than or equal to 2 weeks versus greater than 2 weeks was somewhat problematic. Although we felt this factor was an important potential confounder with respect to acute conversion of AF, our chosen cut point of 2 weeks was not the chosen cut point for the majority of trials. Therefore, we broadened our definition to read "any study that could have had subjects with a duration greater than 2 weeks." This rewording of our initial definition allowed us to subgroup trials with alternative cut points for duration of AF (e.g., greater than 3 days). As with the subgroup analysis regarding goal followup time, the majority of trials did not have enough data for performing this analysis. When the analysis was possible, only two comparisons had any significant difference in treatment effect based on duration of AF: quinidine versus control treatment and flecainide versus control treatment. In the case of quinidine, this analysis revealed that while the overall evidence behind quinidine compared with control treatment gives moderate evidence in favor of quinidine (OR 2.9, CI 1.2 to 7.0), the estimated treatment effect is inconclusive if the duration of AF is less than or equal to 2 weeks (OR 1.2, CI 0.4 to 4.2). The estimated treatment effect of quinidine compared with control treatment for subjects with a duration of AF greater than 2 weeks persists as moderate evidence of efficacy of quinidine (OR 5.4, CI 1.0 to 28.8). This subgroup analysis, therefore, suggests that the majority of quinidine efficacy compared with control treatment for acute conversion of AF may be in subjects with persistent AF. It should be noted, however, that this subgroup analysis had only one trial in each grouping, with 50 to 60 subjects in each trial. In regard to flecainide compared with control treatment, our overall analysis showed strong evidence of efficacy of flecainide compared with control treatment for acute conversion of AF (OR 24.7, CI 9.0 to 68.3). The subgroup analysis based on duration of AF separates this support of flecainide use. While there clearly continues to be strong evidence of efficacy of flecainide for subjects with a duration less than or equal to 2 weeks (OR 10.8, CI 2.1 to 56.5), the evidence and estimated treatment effect are greater for subjects with duration of AF greater than 2 weeks (OR 66.5, CI 13.8 to 321.6). This conclusion is based on a total of three trials with 141 subjects in all trials. This subgroup analysis with respect to amiodarone is difficult to evaluate because of the small number of subjects in each arm and the especially wide confidence intervals. The summary of this meta-analysis and supplemental figures are available upon request (Supplemental Figures, 1999).
Our subgroup analyses regarding mean left-atrial size less than or equal to 4.0 centimeters versus greater than 4.0 centimeters and regarding inclusion of subjects with atrial flutter were limited based on the information provided in the identified clinical trials. Of the few trials with sufficient information for performing these analyses, no significant treatment effect differences were noted based on mean left-atrial size or inclusion of subjects with atrial flutter. Please see the section on Echocardiography at the end of this chapter for a discussion of the individual trials that give results based on left-atrial size. Individual study data are available upon request (Supplemental Figures, 1999).
Maintenance of Sinus Rhythm in AF
Evidence tables for maintenance of sinus rhythm in AF
Evidence Table 5 shows reported maintenance of sinus rhythm rates for AF and goal followup time for all of the identified randomized clinical trials. For purposes of this review, successful maintenance of sinus rhythm is defined as no recurrences of AF. It is important to note that for a significant proportion of trials, the overall study size (N) is considerably larger than the sum of the treatment arms. Two factors account for this. First, because the outcome of maintenance of sinus rhythm is time dependent, these studies were subject to losses to followup. Unfortunately, relatively few trials attempted to account for this by some form of life-table analysis. Therefore, the resultant rates for maintenance of sinus rhythm often are based on smaller denominators than the original sample size. For the majority of trials, we were unable to account for all subjects in order to correct this, and therefore we chose to present the rates as reported in the article. The reported rates for adverse events, however, varied from trial to trial, from including all subjects enrolled to including only subjects felt to be "valuable" in the denominator. Second, about one-third of the trials actually looked at both outcomes of acute conversion and maintenance of sinus rhythm for AF. These trials, therefore, had fewer subjects followed for the maintenance of sinus rhythm phase because only those with successful conversion were eligible. A summary of the crude rates for successful maintenance of sinus rhythm in AF for the major antiarrhythmic agents follows:
| Quinidine | 11--91% | (n = 11) |
| Disopyramide | 54--72% | (n = 4) |
| Flecainide | 24--72% | (n = 7) |
| Propafenone | 27--67% | (n = 8) |
| Amiodarone | 79--92% | (n = 2) |
| Sotalol | 35--76% | (n = 5) |
| Control treatment | 0--90% | (n = 16) |
The majority of trials had mean followup times of 3 to 12 months. Studies with an overall size (N) considerably smaller than the sum of treatment arms represent studies employing a crossover design. As mentioned above, interpretation of these numbers must be done in light of study design characteristics, in particular the length of followup for maintenance of sinus rhythm in AF.
One trial (Feld, Nademanee, Noll et al., 1989) is presented in the table for completeness but is not included in later data synthesis because 100 percent of the subjects had atrial flutter.
For ease of examining this evidence, we group the results. First, we present the evidence comparing each major antiarrhythmic agent to control treatment. It should be noted that the bulk of evidence for any individual agent exists in this comparison. Second, we present the evidence for each individual major antiarrhythmic agent separately. Third, we present the evidence for miscellaneous comparisons.
A. Antiarrhythmic agents versus control treatment
Evaluation of the relative efficacy for maintenance of sinus rhythm in AF compared with control treatment suggests efficacy for all of the following antiarrhythmic agents: quinidine, disopyramide, flecainide, propafenone, and sotalol. This is graphically displayed in . No data were identified regarding amiodarone compared with control treatment for maintenance of sinus rhythm. In comparison with the Cobbe review from 1997 (
Cobbe, 1997), we have two areas of difference. First, we identified two randomized clinical trials of disopyramide versus placebo. Both of these trials indicate a relative benefit of diso¡pyramide for maintenance of sinus rhythm. Second, our review identified no trials evaluating amiodarone compared with control treatment, whereas Cobbe (1997) reported positive efficacy of amiodarone for maintenance of sinus rhythm. This last difference may be because we used only randomized clinical trials for these evidence tables.
B. Individual antiarrhythmic agents versus all comparison groups
As with the rates for successful conversion, we have chosen to summarize the evidence tables on maintenance of sinus rhythm in AF for each major antiarrhythmic agent graphically in order to show more easily the relative efficacy for each of the major antiarrhythmic agents and in order to display the absolute rates.
Quinidine
Figure 24. Proportion of subjects with successful maintenance of sinus rhythm
Notes:
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
**Vertical lines represent 95% confidence intervals for the proportion of subjects with successful maintenance of sinus rhythm
+ n equals the number of trials evaluating each comparison
The evidence regarding quinidine is shown in . In general, these data suggest lack of efficacy of quinidine compared with propafenone for maintenance of sinus rhythm in AF. The data are equivocal for the comparisons of quinidine with flecainide, amiodarone, and sotalol. The only apparent relative benefit of quinidine is in comparison to control treatment.
Disopyramide
Figure 25. Proportion of subjects with successful maintenance of sinus rhythm
Notes:
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
**Vertical lines represent 95% confidence intervals for the proportion of subjects with successful maintenance of sinus rhythm
+ n equals the number of trials evaluating each comparison
As with the conversion data, only a handful of trials evaluated disopyramide. These are summarized in . The evidence is equivocal for disopyramide compared with propafenone and shows a lack of efficacy of disopyramide compared with amiodarone. This last comparison is based on interim results (
Martin, Benbow, Leach et al., 1986). Our literature review did not identify the final results of this trial. Comparable to quinidine, the only relative benefit of disopyramide is in comparison with control treatment.
Flecainide
Figure 26. Proportion of subjects with successful mainenance of sinus rhythm
Notes:
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
**Vertical lines represent 95% confidence intervals for the proportion of subjects with successful maintenance of sinus rhythm
+ n equals the number of trials evaluating each comparison
shows the data for maintenance of sinus rhythm with flecainide. In general, the data regarding use of flecainide are equivocal with respect to the comparisons with both quinidine and propafenone. Notably, no identified trials evaluated flecainide relative to amiodarone or sotalol for maintenance of sinus rhythm in AF. The only relative benefit supported by the literature for flecainide is in comparison with control treatment for maintenance of sinus rhythm in AF.
Propafenone
Figure 27. Proportion of subjects with successful maintenance of sinus rhythm
Notes:
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
**Vertical lines represent 95% confidence intervals for the proportion of subjects with successful maintenance of sinus rhythm
+ n equals the number of trials evaluating each comparison
The evidence on propafenone is displayed in . These data show efficacy of propafenone over both quinidine and control treatment for maintenance of sinus rhythm in AF. The evidence is equivocal for propafenone compared with disopyramide and flecainide, and for propafenone compared with sotalol. No trials were identified that compared propafenone and amiodarone for maintenance of sinus rhythm in AF.
Amiodarone
Figure 28. Proportion of subjects with successful maintenance of sinus rhythm
Notes:
** Vertical lines represent 95% confidence intervals for the proportion of subjects with successful maintenance of sinus rhythm
+ n equals the number of trials evaluating each comparison
Few trials examined the use of amiodarone for maintenance of sinus rhythm in AF. These are shown in . The data are equivocal for amiodarone compared with quinidine. Efficacy of amiodarone compared with disopyramide is supported by the data, although these are reportedly interim results by the trial (
Martin, Benbow, Leach et al., 1986). As previously mentioned, our literature review did not identify the final results of this trial. No trials were identified that evaluated amiodarone relative to flecainide, propafenone, sotalol, or control treatment for maintenance of sinus rhythm.
Sotalol
Figure 29. Proportion of subjects with successful maintenance of sinus rhythm
Notes:
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
**Vertical lines represent 95% confidence intervals for the proportion of subjects with successful maintenance of sinus rhythm
+ n equals the number of trials evaluating each comparison
The trials using sotalol for maintenance of sinus rhythm are summarized in . These data are equivocal for sotalol compared with quinidine and for sotalol compared with propafenone. The only suggested efficacy of sotalol is in comparison with control treatment. Notably absent are any trials comparing sotalol with flecainide or amiodarone.
C. Miscellaneous antiarrhythmic agents
With respect to maintenance of sinus rhythm in AF, we identified three trials using unusual comparison groups or newer agents. These are summarized in . Two trials examining flecainide in comparison with cibenzoline both reported equivocal data. A relatively old trial from 1976 (
Normand, Legendre, Kahn et al., 1976) looked at two different preparations of quinidine and found a "long-acting" form superior to a "short-acting" form.
Meta-analysis for maintenance of sinus rhythm in AF
In the interest of presenting our meta-analysis results for maintenance of sinus rhythm in AF consistent with the evidence table discussion, we have grouped the data into tables similar to the figures displaying absolute rates. Overall, our analysis deals primarily with the calculated odds ratio for each comparison. In order to present evidence usable to the reader, our final summary section (Efficacy of Antiarrhythmic Agents) of the evidence also reports the important comparisons in terms of the number of patients needed to be treated by one agent in order to see a benefit compared with the alternative treatment. Comparable to the preceding section on the evidence tables and figures of absolute rates, we present the results of our mathematical pooling of the data in the following order: antiarrhythmic agents compared with control treatment, each individual agent versus all comparison groups, and miscellaneous comparisons.
Of particular note for the outcome of maintenance of sinus rhythm in AF, a small percentage of trials used life-table analysis. The resultant data reported were cumulative percentages of successful maintenance of sinus rhythm over the course of followup. These cumulative percentages are displayed on the figures of absolute rates. However, the publication of such analyses frequently did not provide easily extractable data for incorporation into a meta-analysis. Thus, for our meta-analysis purposes, we applied the cumulative percentages to the initial overall subject number in each trial arm (n) to derive a proportion for inclusion in the meta-analysis. We note in the comments section of
Evidence Table 5 when a life-table analysis was done. This note, therefore, also signifies when the above discussion regarding calculation of study proportions for inclusion in the meta-analysis applies.
For the evidence on maintenance of sinus rhythm, an OR greater than 1.0 represents a higher odds of maintenance of sinus rhythm than in the comparison group. As indicated earlier, we chose the following categorization of strength of evidence by noting the placement of the point estimate and the width of the confidence interval (CI) surrounding it:
-
Strong evidence of efficacy: OR > 1.0, 99 percent CI does not include 1.0 (p < 0.01).
-
Moderate evidence of efficacy: OR > 1.0, 95 percent CI does not include 1.0, but 99 percent CI includes 1.0 (0.01
<
p
<
0.05).
-
Suggestive evidence of efficacy: 95 percent CI includes 1.0 in the lower tail (0.05 < p < 0.2 to 0.3) and the OR is in a clinically meaningful range.
-
Inconclusive evidence of efficacy: 95 percent CI widely distributed around 1.0.
-
Strong evidence of lack of efficacy: OR near 1.0, 95 percent CI is narrow.
When the point estimate was less than 1.0, we termed this negative efficacy and used the same categorization of strong, moderate, and suggestive evidence based on the CI.
In some situations, the above categorizations did not completely capture some element of the result that was important for interpretation, particularly in cases with large point estimates and/or very wide confidence intervals. In those instances, we added additional descriptive text. The use of the 99 percent confidence interval (i.e., p = 0.01) as a boundary point for the descriptor "strong" was based on published statistical research on the relationship of p-values to Bayesian evidential summaries (Bayes factors), as well as the recognition that all meta-analyses have some degree of qualitative heterogeneity not reflected in the quantitative summaries.
A. Antiarrhythmic agents versus control treatment
The meta-analysis results for maintenance of sinus rhythm in AF for each of the major antiarrhythmic agents relative to control treatment are shown in . As with the conversion evidence, the bulk of evidence behind any individual antiarrhythmic agent occurred in comparison to control treatment. Overall, in terms of maintenance of sinus rhythm, the following antiarrhythmic agents have strong evidence of efficacy relative to control treatment: quinidine, disopyramide, flecainide, propafenone, and sotalol. There is no evidence for the comparison of amiodarone to control treatment.
In more detail, the evidence behind use of quinidine consisted of four trials with pooled data finding strong evidence of efficacy of using quinidine over control treatment (OR 4.1, CI 2.5 to 6.7) for maintenance of sinus rhythm. The estimated treatment effect size is fairly large. It should be noted that a fifth trial (Rasmussen, Wang, and Fausa, 1981) evaluated quinidine compared with control treatment. This trial did not report data in such a way that it could be incorporated into our meta-analysis. However, the overall finding of the trial was in support of quinidine, consistent with the other evidence.
The data for disopyramide compared with control treatment involved two trials and also had strong evidence of efficacy and a fairly large effect size for disopyramide (OR 3.4, CI 1.6 to 7.1). The comparison of flecainide versus control treatment comprised three trials with strong evidence of efficacy and a fairly large effect size of flecainide compared with control treatment (OR 3.1, CI 1.5 to 6.2). This is supported by a non-English-language article (Carunchio, Fera, Mazza et al., 1995). Comparably, the propafenone relative to control treatment shows strong evidence of efficacy and a fairly large effect size for propafenone (OR 3.7, CI 2.4 to 5.7) based on four trials.
There is strong evidence of efficacy of sotalol versus control treatment (OR 7.1, CI 3.8 to 13.4) for maintenance of sinus rhythm in AF. This is supported by a non-English-language article (Carunchio, Fera, Mazza et al., 1995).
Because some antiarrhythmic agents are commonly considered together based on the Vaughan-Williams classification (1984), we have grouped the data for Class Ia and Class Ic agents, in comparison with control treatment, in this meta-analysis. The combined Ia agents are quinidine and disopyramide. The combined Ic agents are flecainide and propafenone. We did not think that any combination of the Class III agents (amiodarone, sotalol) was clinically relevant. Not surprisingly, the results of these meta-analyses show strong evidence of efficacy of both Class Ia (OR 4.2, CI 2.9 to 6.1) and Class Ic (OR 3.5, CI 2.4 to 5.1) agents relative to control treatment for maintenance of sinus rhythm in AF. These results are shown in .
B. Individual antiarrhythmic agents versus all comparison groups
Quinidine
Figure 32. Meta-analysis on maintenance of sinus rhythm
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
Individual study data available on request
As for quinidine for maintenance of sinus rhythm, as shown in , the only comparison with strong evidence of efficacy of quinidine is relative to control treatment (OR 4.1, CI 2.5 to 6.7), as discussed previously. The estimated effect size for this comparison is fairly large. The data regarding quinidine compared with flecainide consist of three trials and are inconclusive (OR 0.7, CI 0.4 to 1.2). The comparisons of quinidine versus propafenone and of quinidine versus amiodarone comprise only one clinical trial each. The data give strong evidence of negative efficacy of quinidine compared with propafenone (OR 0.3, CI 0.1 to 0.7), with a fairly large effect size, and are inconclusive for quinidine versus amiodarone (OR 0.9, CI 0.1 to 16.5). Two trials compared quinidine and sotalol for maintenance of sinus rhythm. These resulted in strong evidence of lack of efficacy of quinidine compared with sotalol (OR 0.9, CI 0.5 to 1.5). We should note that one non-English language article (
Richiardi, Gaita, Greco et al., 1991) found no difference in rates of maintenance of sinus rhythm between quinidine and propafenone, with greater than 6 months of followup time.
Disopyramide
Figure 33. Meta-analysis on maintenance of sinus rhythm
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
Individual study data available on request
depicts the evidence for use of disopyramide for maintenance of sinus rhythm in AF. A total of four trials deal with disopyramide. Two of the trials give strong evidence of efficacy of disopyramide relative to control treatment (OR 3.4, CI 1.6 to 7.1), with a fairly large effect size. The third trial is suggestive of efficacy of disopyramide versus propafenone (OR 1.8, CI 0.6 to 5.1), with only a modest effect size. The fourth trial gives moderate evidence of negative efficacy of disopyramide compared with amiodarone (OR 0.3, CI 0.1 to 1.0), with a modest effect size, although this represents only interim results of this trial (
Martin, Benbow, Leach et al., 1986). A non-English language article (
Villani, Zoletti, Veniani et al., 1992) found significant evidence of lack of efficacy of disopyramide compared with amiodarone, with approximately 12 months of followup time.
Flecainide
Figure 34. Meta analysis on maintenance of sinus rhythm
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
Individual study data available on request
The evidence for flecainide in terms of maintenance of sinus rhythm, shown in , only gives strong evidence of efficacy relative to use of control treatment (OR 3.1, CI 1.5 to 6.2), based on three trials. Furthermore, the estimated treatment effect size is fairly large. This is supported by a non-English language article (
Carunchio, Fera, Mazza et al., 1995). The comparison of flecainide to quinidine (OR 1.4, CI 0.8 to 2.3) is inconclusive, and the comparison of flecainide to propafenone (OR 0.9, CI 0.4 to 2.2) is inconclusive. The non-English language article by Carunchio, Fera, Mazza et al. (1995) found no difference in maintenance of sinus rhythm between flecainide and sotalol, with 12 months of followup time.
Propafenone
Figure 35. Meta-analysis on maintenance of sinus rhythm
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
Individual study data available on request
In regard to propafenone, as shown in , the meta-analysis shows strong evidence of efficacy and a fairly large effect size of propafenone compared with control treatment (OR 3.7, CI 2.4 to 5.7), based on four trials. Propafenone compared with quinidine (OR 3.6, CI 1.5 to 9.0), based on one trial, has strong evidence of efficacy for propafenone. The evidence is suggestive of negative efficacy of propafenone compared with disopyramide (OR 0.6, CI 0.2 to 1.6), with a moderate effect size, and is also suggestive of negative efficacy of propafenone compared with sotalol (OR 0.7, CI 0.4 to 1.1), with a modest effect size, for maintenance of sinus rhythm in AF. The evidence is inconclusive for the comparison of propafenone and flecainide (OR 1.1, CI 0.5 to 2.6). We should note that one non-English language article (
Richiardi, Gaita, Greco et al., 1991) found no difference in rates of maintenance of sinus rhythm between propafenone and quinidine, with greater than 6 months of followup time.
Amiodarone
Figure 36. Meta-analysis on maintenance of sinus rhythm
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
Individual study data available on request
deals with the meta-analysis regarding amiodarone for maintenance of sinus rhythm. The evidence for amiodarone versus quinidine (OR 1.1, CI 0.1 to 20.0) is inconclusive, based on one trial. The evidence for amiodarone compared with disopyramide, based on interim results of one trial (
Martin, Benbow, Leach et al., 1986), is consistent with moderate evidence of efficacy of amiodarone (OR 3.2, CI 1.0 to 9.6). As mentioned, our review of the literature did not identify the final results of this trial. A non-English language article (
Villani, Zoletti, Veniani et al., 1992) found significant evidence of efficacy of amiodarone compared with disopyramide for maintenance of sinus rhythm, with approximately 12 months of followup time.
Sotalol
Figure 37. Meta-analysis on maintenance of sinus rhythm
Notes:
Vertical bars are point estimates. Horizontal bars and numbers in brackets are 95% confidence intervals using fixed-effects model unless otherwise indicated.
*Control treatment includes groups receiving placebo, verapamil, diltiazem, or digoxin
Individual study data available on request
The evidence for sotalol in maintenance of sinus rhythm is shown in . The evidence regarding sotalol versus control treatment consists of two trials and is consistent with strong evidence of efficacy and a large effect size for sotalol (OR 7.1, CI 3.8 to 13.4). This is supported by a non-English language article (
Carunchio, Fera, Mazza et al., 1995). The comparison of sotalol versus quinidine (OR 1.1, CI 0.7 to 2.0) has strong evidence of lack of efficacy. There is, however, suggestive evidence of efficacy of sotalol compared with propafenone (OR 1.6, CI 0.9 to 2.6), with only modest effect size, based on two trials. The non-English language article by Carunchio, Fera, Mazza et al. (1995) found no difference in maintenance of sinus rhythm between flecainide and sotalol, with 12 months of followup time.
C. Miscellaneous antiarrhythmic agents
shows the results for all miscellaneous antiarrhythmic trials with respect to maintenance of sinus rhythm. Two trials compared flecainide to cibenzoline, with inconclusive evidence (OR 1.4, CI 0.5 to 4.0), based on two trials. Four non-English language articles evaluated cibenzoline. In general, these found cibenzoline either more efficacious than quinidine, propafenone, and flecainide or equally as efficacious. The first article (
Frances, Luccioni, Delaage et al., 1985) found cibenzoline superior to quinidine for maintenance of sinus rhythm, at 6 months of followup time. The second article (
Touboul, Brembilla-Perrot, Scheck et al., 1995) found cibenzoline equally as efficacious as quinidine, with 12 months of followup time. The third study (
Lardoux, Maison Blanche, Marchand et al., 1996), with 6 months of followup time, found cibenzoline as efficacious as propafenone. A study by Maison-Blanche, Brembilla-Perrot, Fauchier et al. (1997) found cibenzoline equal in efficacy to flecainide for maintaining sinus rhythm, with 6 months of followup time. The evidence comparing "long-acting" versus "short-acting" quinidine is suggestive of efficacy of the "long-acting" preparation (OR 3.5, CI 0.9 to 13.0).
Subgroup analysis for maintenance of sinus rhythm in AF (Appendix Q)
Comparable to the analysis of conversion, several important factors were felt a priori to be potential confounders regarding maintenance of sinus rhythm in AF. We therefore examined the following factors in subgroup analysis: type of control treatment utilized (placebo versus verapamil/diltiazem versus digoxin), followup time for maintenance of sinus rhythm of less than 6 months versus greater than or equal to 6 months, mean left-atrial size less than or equal to 4.0 centimeters versus greater than 4.0 centimeters, and inclusion of subjects with atrial flutter (no subjects versus some subjects with atrial flutter). It should again be noted that the ability of a systematic literature review to detect subgroup differences is limited because the data are not at the level of each individual study subject but rather at the overall study level. For example, if one trial had a mean left-atrial size for all subjects of 3.7 centimeters, that trial as a whole was taken as evidence for maintenance of sinus rhythm with left-atrial size less than or equal to 4.0 centimeters. We did not have enough data from the publications to separate subjects within any given trial for subgroup analysis. In addition, incomplete information and inconsistent definitions of subgroups hampered our subgroup analyses.
The analysis of type of control treatment used (placebo versus verapamil/diltiazem versus digoxin) was not applicable to the outcome of maintenance of sinus rhythm because all of the identified trials of any individual antiarrhythmic agent used placebo as the control treatment. This enhances the comparability of the various studies regarding antiarrhythmic agents versus control treatment. The evidence table includes one study (Rasmussen, Wang, and Fausa, 1981) that used verapamil as the control treatment. This trial is not included in our meta-analysis because the results were not reported in a way that was usable for that purpose. Consistent with the meta-analysis, however, this trial did report a significant benefit of quinidine relative to control treatment for maintenance of sinus rhythm in AF.
The next subgroup analysis of goal followup time for maintenance of sinus rhythm of less than 6 months versus greater than or equal to 6 months revealed only one important finding in all comparisons where it was possible to do the analysis. We felt this was an important subgroup analysis because followup for greater than 6 months is a better test of an antiarrhythmic agent for the outcome of maintenance of sinus rhythm. The one important finding was in terms of sotalol versus propafenone. The overall meta-analysis resulted in evidence suggestive of efficacy of sotalol compared with propafenone (OR 1.6, CI 0.9 to 2.6). Looking only at trials with greater than or equal to 6 months of followup, however, there is moderate evidence of efficacy of sotalol (OR 1.9, CI 1.1 to 3.5) compared with propafenone.
Comparable to this finding, our results for the subgrouping of mean left-atrial size less than or equal to 4.0 centimeters versus greater than 4.0 centimeters revealed only one important finding, and this was with respect to sotalol versus propafenone. The overall evidence was suggestive of efficacy of sotalol versus propafenone (OR 1.6, CI 0.9 to 2.6). Looking only at the subgroup of subjects with mean left-atrial diameter greater than 4.0 centimeters, there is moderate evidence of efficacy of sotalol (OR 1.9, CI 1.1 to 3.5) compared with propafenone. Please see the Echocardiography section for a discussion of the individual trials that give results based on left-atrial size.
We explored whether the inclusion of subjects with atrial flutter may have influenced the likelihood of successful maintenance of sinus rhythm by separating the trials as to whether any subjects had atrial flutter. This subgroup analysis did not reveal any important treatment effect differences between trials with no subjects having atrial flutter and trials in which some subjects had atrial flutter. Individual study data are available upon request (Supplemental Figures, 1999).
Adverse effects
We report the incidence rates for adverse effects as described by each study in Evidence Tables 3
and
5. More specifically, we report the incidence rates of ventricular dysrhythmias, other nontransient dysrhythmias, and adverse events causing study drug cessation or dosage decrease. In this discussion, we combine the information from trials on conversion and trials on maintenance of sinus rhythm in AF.
If a particular study did not specifically mention a lack of serious ventricular dysrhythmias (ventricular fibrillation, polymorphic ventricular tachycardia, or torsade de pointes), we left that section of the result table blank for that study. We felt it was important to avoid making assumptions about such serious adverse effects in creating these systematic evidence tables. Overall, for all our categories of adverse effects, the data were sporadically reported in the identified clinical trials. It should be noted that randomized clinical trials are not the best method to assess uncommon adverse effects because of their controlled environments, relatively small samples, and relatively short followup times.
Only about one-half of the included studies specifically mentioned ventricular fibrillation, polymorphic ventricular tachycardia, and/or torsade de pointes as a potential adverse effect of treatment. In more detail, one trial evaluating quinidine (Hohnloser, van de Loo, and Baedeker 1995) reported the highest ventricular fibrillation rate, 12 percent, noted in our table. A trial evaluating ibutilide (Stambler, Wood, Ellenbogen et al., 1996) reported an incidence of 9 percent, and another (Ellenbogen, Stambler, Wood et al., 1996) reported an incidence of 5 percent. The one trial of dofetilide (Falk, Pollak, Singh et al., 1997) and one trial evaluating propafenone (UK Propafenone PSVT Study Group, 1995) found incidences of 3 percent. One trial investigating flecainide (Donovan, Dobb, Coombs et al., 1992) and one evaluating propafenone (Aliot and Denjoy, 1996) reported incidences of 2 percent. Finally, one trial investigating both quinidine and sotalol found incidences of 1 percent for each agent. One trial, evaluating quinidine and practolol (Levi, Proto, and Rovetta, 1973), reported an incidence of 4 percent. It is important to note that the majority of trials did not specifically mention these serious adverse effects. Given this fact, the small sample sizes, and limited followup times, our reported incidence rates need to be viewed cautiously.
The results on ventricular arrhythmias are particularly important in light of the suggestion of increased mortality in patients taking quinidine in observational studies (Coplen, Antman, Berlin et al., 1990) and the increased mortality of patients taking Class Ic agents after a myocardial infarction (The Cardiac Arrhythmia Suppression Trial [CAST] Investigators, 1989). Given incomplete reporting and inconsistent definitions of coronary artery disease, however, we were unable to analyze these relationships adequately.
As for other nontransient dysrhythmias, again only about one-half of the studies reported results. The types of dysrhythmias included significant and/or symptomatic bradycardia usually requiring intervention, functional rhythms, nonsustained ventricular tachycardias, and monomorphic ventricular tachycardias. The range of incidence rates of these types of dysrhythmia in general was 0 to 20 percent for the major antiarrhythmic agents. It was not always easy to identify which of these dysrhythmias were considered clinically significant based on the article review.
In terms of other adverse events requiring a change in drug dosage or cessation of the drug, again only about one-half of the studies reported results. The reported incidence of these adverse events was generally 0 to 50 percent for the major antiarrhythmic agents. These rates need to be cautiously interpreted because each study independently made decisions regarding dosage decreases or study withdrawals. In addition, studies involving a one-time drug administration by definition could not have adverse events requiring dosage decrease or, thereby, study withdrawal.
Summary and Conclusions, Efficacy of Antiarrythmic Agents
This systematic review of the evidence regarding the efficacy of antiarrhythmic agents in the management of non-postoperative AF yielded a substantial amount of information. In general, the bulk of evidence exists in terms of comparing each antiarrhythmic agent to control treatment as opposed to direct comparisons between different antiarrhythmic agents. This fact makes a definitive ranking of the antiarrhythmic agents in terms of efficacy difficult because not all relevant comparisons have been done. In addition, it should be noted that the evaluation of serious adverse events related to antiarrhythmic agent therapy is significantly limited when a pool of randomized clinical trials is used.
Pharmacological conversion of AF
Strong evidence of efficacy
The antiarrhythmic agents with the strongest evidence of efficacy relative to control treatment for conversion of AF are flecainide and ibutilide/dofetilide. The odds ratios for these agents relative to control treatment are 25 and 29, respectively. Furthermore, the evidence for both flecainide and ibutilide/dofetilide is consistent with a large treatment effect size. Assuming a control treatment conversion rate of 30 percent and using the upper and lower 95 percent confidence limits on these odds ratios, the estimated number of subjects needed to be treated (NNT) in order to see a benefit relative to control treatment ranged from 1.5 to 2.0 for both flecainide and ibutilide/dofetilide.
Strong evidence of efficacy relative to control treatment also exists for propafenone, with an odds ratio of 4.6. Again assuming a control treatment conversion rate of 30 percent and using the upper and lower 95 percent confidence limits on these odds ratios, the NNT range with propafenone in order to see a benefit relative to control treatment is 2.0 to 4.5.
We should note here that the NNT range is sensitive to the baseline assumption of the control treatment conversion rate. For example, if we vary the control treatment conversion rate from 5 to 75 percent for propafenone, the NNT ranges become 4.0 to 14.3 and 4.7 to 7.5, respectively. Both of these ranges are higher and broader than the range obtained by assuming a 30 percent control treatment conversion rate. This is because, at a given odds ratio, a very low or very high baseline conversion rate requires more subjects to see an effect of treatment.
Moderate evidence of efficacy
There was moderate evidence of efficacy for quinidine compared with control treatment (OR 2.9, CI 1.2 to 7.0), for conversion of AF. The estimated treatment effect size is modest. Assuming a control treatment conversion rate of 30 percent and using the 95 percent confidence limits, the range of number of subjects needed to be treated with quinidine in order to see a benefit relative to control treatment is 2.0 to 25.0.
Suggestive evidence of efficacy
Disopyramide and amiodarone have suggestive evidence of benefit relative to control treatment for acute conversion of AF relative to control treatment. The comparison of amiodarone to control treatment, however, is particularly problematic because of the substantial qualitative and quantitative heterogeneity between the combined studies. It should be noted that direct comparisons of amiodarone to other antiarrhythmic agents yielded moderate evidence of amiodarone efficacy relative to quinidine and inconclusive evidence for amiodarone compared with flecainide.
Suggestive evidence of negative efficacy
The evidence is suggestive of negative efficacy of sotalol for acute conversion of AF.
Minimal evidence
Minimal evidence existed for agents in Classes II, IV, and V and in the miscellaneous category.
Pharmacological maintenance of sinus rhythm in AF
Strong evidence of efficacy
All of the following antiarrhythmic agents have strong evidence of efficacy relative to control treatment and fairly large estimated effect size for maintenance of sinus rhythm in AF: quinidine, disopyramide, flecainide, propafenone, and sotalol. The estimated odds ratios for successful maintenance of sinus rhythm relative to control treatment range from 3.1 to 7.1. Our review does not support any definitive ranking of these five agents for efficacy. Based on the evidence for control treatment, we made the assumption that only 30 percent of control treated subjects are still in sinus rhythm at 6 months. Using this assumption and the upper and lower 95 percent confidence limits for each agent, the NNT range for each agent in order to see a benefit relative to control treatment is as follows: quinidine 2.3 to 4.6; disopyramide 2.2 to 9.4; flecainide 2.3 to 10.9; propafenone 2.4 to 4.8; and sotalol 1.8 to 3.1.
Potentially strong evidence of efficacy
Evidence was scarce on amiodarone for maintenance of sinus rhythm in AF. It should be noted that one trial (Martin, Benbow, Leach et al., 1986), presenting only interim results, had moderate evidence of amiodarone efficacy relative to disopyramide. In addition, the trial by Zehender, Hohnloser, Muller et al. (1992) yielded inconclusive evidence of any benefit of amiodarone compared with quinidine. These two comparisons could, therefore, be consistent with strong evidence of efficacy of amiodarone. Notably, however, we identified no trials of amiodarone compared with control treatment for maintenance of sinus rhythm.
Minimal evidence
Minimal evidence existed for N-acetylprocainide and agents in Class IV and the miscellaneous category.
Adverse events
The use of adverse event information to guide antiarrhythmic agent choice is significantly hampered in this review. This is the result of inconsistent and sporadic reporting of the events in the articles describing the clinical trials. In addition, randomized clinical trials have significant limitations in assessing important but uncommon adverse events.
Generalizability of the Evidence
Examining the inclusion/exclusion criteria and baseline subject characteristics of all the reviewed trials, we do not believe that subject-specific factors significantly influenced the accumulated evidence. However, two points should be noted. First, the age range of the subjects in these trials was somewhat younger than would be seen in a population- based sample of AF. Because it is possible that response to pharmacological therapy may differ with age, this needs to be kept in mind. Second, our target population consisted of non-postoperative AF. The accumulated evidence, therefore, may not be applicable to subjects with postoperative AF.
Efficacy of Drugs for Achieving Rate Control in Patients With AF
Control of the ventricular rate is important in the management of subjects in atrial fibrillation. A rapid ventricular response may lead to worsening of congestive heart failure, pulmonary edema, myocardial ischemia, and systemic under-perfusion, as well as the distressing symptoms of palpitations and breathlessness. A number of drugs have been investigated for their ability to expediently control the rate, and others have been evaluated for their long-term effectiveness. For long-term effectiveness, control during exertion-when activation of the sympathetic nervous system tends to accelerate the rate-is particularly relevant.
Some of the trials presented here were not designed as rate-control studies; some were studies of pharmacological conversion cardioversion that included heart rate data. This presentation of the evidence includes a summary of the trials conducted, the quality of the studies, the effectiveness of the drugs in controlling ventricular rate, and the significant adverse effects of the drugs. We also identify drug comparisons that are notably absent and may be appropriate for further study.
Designs of the Studies
Evidence Tables 6A
--
6G list the studies that were relevant for inclusion in this review. The comparisons made in the studies include the following:
-
Calcium-channel-blockers versus placebo.
-
Beta-blockers versus placebo.
-
Digoxin (Lanoxin), other drugs, and drug combinations versus placebo.
-
Calcium-channel-blockers versus digoxin.
-
Beta-blockers versus digoxin.
-
Other drugs and drug combinations versus digoxin.
-
Other drug comparison trials.
The following medications were evaluated: diltiazem (Cardizem, Tiazac, Dilacor), verapamil (Calan, Covera, Isoptin, Verelan), atenolol (Tenormin), xamoterol, timolol (Blocarden), nadolol (Corgard), celiprolol, pindolol (Visken), propranolol (Inderal), labetalol (Normodyne, Trandate), clonidine (Catapres), propafenone (Rythmol), sotalol (Betapace), digoxin (Lanoxin), betaxolol (Kerlone), magnesium sulfate, metoprolol (Lopressor), disopyramide (Norpace), flecainide (Tambocur), quinidine (Quinaglute, Quinidex), and amiodarone (Cordarone).
The designs of the studies differ importantly in a number of ways that could influence the results.
The followup times of the studies are markedly different, ranging from less than 30 minutes to 8 weeks.
The medication dosages in the trials differ for the same medication. The intravenous diltiazem and verapamil doses are fairly uniform across the trials. Some of the beta-blocker trials titrated the medication to effectiveness; these report the ranges of dosages used. Some of the digoxin trials adjusted the medication based on blood levels, while others used fixed dosing. The trials of digoxin in combination with other drugs often continued the patient's pretrial dose of digoxin and did not report a blood level or dose.
Some of the trials that were not evaluating digoxin for rate control permitted digoxin to be continued during the trial, if the patient entered the trial while receiving digoxin therapy. This typically occurred in all arms of the trials, but the number of subjects in each arm who remained on digoxin during the trial was seldom reported.
The inclusion criteria of the trials varied. Most specified adults over the age of 18 years; some required a duration of atrial fibrillation longer than 1 month or 6 months. Several specified a ventricular rate required for entry, such as greater than 120 beats per minute or "rapid rate." Exclusion criteria varied from none specified to stringent criteria, particularly in studies that involved exercise, from which subjects with angina or significant congestive heart failure were excluded because of safety concerns. Most of the trials excluded subjects with untreated hyperthyroidism, or renal or hepatic failure.
Most of the trials were small crossover studies with two to six treatment arms. One study (Lewis, Laing, Moreland et al., 1988) enrolled only 6 subjects, while the largest study (Salerno, Dias, Kleiger et al., 1989) enrolled 113 subjects. Most of the studies enrolled fewer than 25 subjects.
The dates of the trials range from 1982 to 1997. The earliest study is by Stern, Pitchon, King et al. (1982), who compared verapamil plus digoxin with digoxin alone. The most recent trials studied diltiazem and digoxin (Schreck, Rivera, Tricarico et al., 1997); sotalol compared with placebo (Lok and Lau, 1997); and digoxin compared with placebo (The Digitalis in Acute Atrial Fibrillation [DAAF] Trial Group, 1997; Jordaens, Trouerbach, Calle et al., 1997).
Quality of the Studies
Generally, the studies were of high quality in their description of the outcomes and for having objective measures of the outcomes. The studies were weaker in their description of the study groups; it was not always possible to tell if the groups were similar. This was of less concern in the crossover studies, except in those instances where not all of the participants crossed into all arms of the study.
The statistical analyses in the trials often lacked the presentation of confidence intervals or other measures of variability. The p-values for comparisons were provided, but with such small studies the confidence intervals would have been instructive.
Abstraction of the data from the crossover studies required careful attention to the numbers of subjects who withdrew before crossover and whether they were considered to be treatment failures or censored observations. This was not always clear. Some of the crossover studies reported data from a nonrandom treatment assignment occurring after the patient had failed therapy in the randomized portion of the trial. The data we abstracted were for only the randomized portions of the trial.
The overall quality scores for the studies were variable, ranging from 31 to 81 out of 100, but with most having overall scores in the '50s and '60s. The highest quality study, from Botto, Capucci, Bonini et al. (1997), was one of the largest (105 subjects) and most recent.
Results of the Studies
In reporting the results of these rate-control studies, we abstracted the data that were most consistently reported in the trials and that were most relevant to clinicians. Some studies noted that the drugs that are most useful for heart rate control at rest may not be the most efficacious during exercise (Hohnloser and Li, 1997). Similarly, drugs that adequately control the heart rate during exercise do not always improve exercise tolerance. The data abstracted from the articles included the mean heart rate at rest on medication, the mean maximum heart rate during exercise or immediately after exercise, and the proportion of subjects who reached the goal heart rate reduction. Many of the studies also reported the distance tolerated during the exercise study or the number of minutes tolerated on the treadmill. Some studies reported VO2, a measure of oxygen consumption, while others reported cardiac output during exercise. Most, although not all, reported the side effects requiring drug withdrawal and the pro-arrhythmic effects of the medications.
In Evidence Tables 7A
--
7G
, we present the results described above and the measure of statistical significance as provided in the trials, with the reference group specified. Because many of the trials included an assessment of heart rate during exercise, we provide a brief description of the exercise protocol used and the workload if known.
Calcium-channel-blockers versus placebo for rate control
As depicted graphically in , all of the trials of calcium-channel-blockers (diltiazem and verapamil) compared with placebo demonstrated their effectiveness in controlling the heart rate at rest and with exertion. Five trials evaluated diltiazem for rate control. Three used intravenous diltiazem and followed the subjects for less than 24 hours for acute control at rest (
Goldenberg, Lewis, Dias et al., 1994;
Ellenbogen, Dias, Plumb et al., 1991;
Salerno, Dias, Kleiger et al., 1989). All three trials reported that the number of subjects reaching the target heart rate, under 100 beats per minute, was significantly greater than with placebo. Two of the trials studied oral diltiazem. Of these, one evaluated a single oral dose of 60 mg or 120 mg and followed the subjects for several hours (
Lewis, Laing, Moreland et al., 1988). The other study evaluated oral diltiazem administered 3 times daily for 3 weeks (
Lundstrom and Ryden, 1990). The single oral dose significantly reduced the heart rate during a modified Bruce treadmill exercise test compared with placebo. The longer diltiazem study also found a significant reduction in mean resting heart rate on drug compared with placebo and a decrease in mean maximum heart rate with exercise. Additionally, the tolerated workload on the treadmill was greater (136 watts versus 127 watts); there was little change in VO
2
.
Five studies evaluated verapamil, all of which used oral dosing (Lewis, Laing, Moreland et al., 1988; Lundstrom and Ryden, 1990; Lundstrom, Moor, and Ryden, 1992; Lewis, McMurray, and McDevitt, 1989; Panidis, Morganroth, and Baessler, 1983). All of the studies demonstrated a significant reduction in mean resting heart rate and reduction in maximum heart rate with exercise compared with placebo. Two of the trials had both verapamil and diltiazem arms but did not report a measure of statistical significance comparing the two drugs. The response appears to be similar with the two drugs.
Importantly, all but two of the trials evaluating verapamil or diltiazem mentioned above allowed digoxin to be used by the participants. Because it is not reported what percentage of subjects in each treatment group received digoxin, this may have confounded the results. This may be less of an issue in the crossover trials because it is likely that the subjects would have continued on the same nonstudy medications throughout all phases of the trial. Nevertheless, it is important to note that the absolute degree of heart rate control achieved by use of the calcium-channel-blockers in some of the studies described above was attributable in part to concurrent use of digoxin.
A non-English-language abstract comparing diltiazem with verapamil found them to be equally effective at rate control (Tezcan, Okucu, Fak et al., 1996).
Beta-blockers versus placebo for rate control
Figure 40. Rate control trials of beta-blockers versus placebo for subjects with atrial fibrillation
The trials reported here include placebo versus atenolol (
Lewis, McMurray, and McDevitt, 1989;
Channer, James, MacConnell et al., 1994), xamoterol (
Lewis, McMurray, and McDevitt, 1989;
Lundstrom, Moor, and Ryden, 1992;
Ang, Chan, Cleland et al., 1990), nadolol (
DiBianco, Morganroth, Freitag et al., 1984), labetalol (
Wong, Lau, Leung et al., 1990), celiprolol (
Myers, Atwood, Sullivan et al., 1987;
Lin, Morganroth, Heng et al., 1986), pindolol (
Channer, James, MacConnell et al., 1994), and timolol (
Sweany, Moncloa, Vickers et al., 1985). Some of the studies included more than one of the above beta-blockers with followup ranging from hours to 4 weeks. The results are graphically displayed in . For lowering mean heart rate at rest, atenolol was significantly better than placebo in both studies; one used 50 mg daily and the other used 50 mg twice daily or 100 mg daily. Xamoterol was not better than placebo at 100 mg twice daily, or at 200 mg twice daily in two studies. One study (
Lundstrom, Moore, and Ryden, 1992) did demonstrate effectiveness for xamoterol at 200 mg twice daily. Timolol treatment resulted in more subjects reaching the target heart rate of under 100 bpm than placebo. Pindolol at 5 mg twice daily or 15 mg twice daily significantly reduced mean resting heart rate, as did nadolol at a titrated dose. Celiprolol and labetalol were no better than placebo at rest.
All of the beta-blockers that were evaluated against placebo demonstrated a significant reduction in heart rate with exercise. Atenolol, celiprolol, xamoterol, labetalol, and nadolol all reduced maximum heart rate with exercise compared with placebo. Timolol and pindolol were not evaluated during exercise. It is important to note, however, that exercise tolerance may be less on medication than on placebo. In one study (Lewis, McMurray, and McDevitt 1989), the distance walked on the treadmill was less on atenolol and on xamoterol than on placebo. Similarly in the nadolol and celiprolol trials, the time spent walking on the modified Bruce protocol was less than on placebo. However, the other xamoterol trial (Ang, Chan, Cleland et al., 1990) found that the beta-blocker recipients spent longer on the treadmill than the placebo-treated subjects (10.7 min versus 9.7 min). The clinical importance of these differences is not known.
As in the calcium-channel-blocker studies, most of the trials allowed subjects to continue use of digoxin if they were already taking it.
Digoxin, other drugs, and combinations compared with placebo
Three of the trials of digoxin versus placebo did not demonstrate a reduction in mean resting heart rate. One study (Lewis, Laing, Moreland et al., 1988) evaluated the subjects just hours after a single oral dose of 0.25 mg of digoxin. Similarly, Falk, Knowlton, Bernard et al. (1987) evaluated the subjects 15 minutes after an oral dose of 0.6 mg of digoxin and found no difference in heart rates. The other trial (Wong, Lau, Leung et al., 1990) used 0.25 mg of digoxin daily for 10 to 14 days, but no significant difference was demonstrated (91 beats per minute for digoxin and 102 beats per minute for placebo). Each arm had only 11 subjects, which limits the power to detect a significant difference. Four digoxin trials did find a significant difference in resting heart rate compared with placebo (Ang, Chan, Cleland et al., 1990; Jordaens, Trouerbach, Calle et al., 1997; Digitalis in Acute Atrial Fibrillation (DAAF) Trial Group, 1997; and Koh, Kwon, Park et al., 1995). The DAAF Trial Group (1997) and Jordaens (1997) both included patients on verapamil; all of the rate reduction cannot be attributed to digoxin alone. The Ang (1990) study involved adjusting the digoxin dose to a serum concentration, which suggests that careful early attention to dosing may be important.
The Ang (1990) study did not find a significant benefit of digoxin with exercise (heart rate of 150 beats per minute on digoxin versus 159 beats per minute on placebo). The other two studies (Lewis, Laing, Moreland et al., 1988; Wong, Lau, Leung et al., 1990) also found no significant heart rate reduction. Koh does not report the statistical significance of the difference in heart rate, but there is the suggestion of a difference (165 beats per minute with digoxin and 196 beats per minute with placebo with exercise) (Koh, Kwon, Park et al., 1995).
Figure 41. Rate control trials of digoxin versus placebo for subjects with atrial fibrillation
Studies evaluating digoxin are displayed in . Clonidine was evaluated against placebo in two trials. The study by Roth, Kaluski, Felner et al. (1992) evaluated its use in the emergency room with 4 hours of followup and found that 8 of 9 subjects treated with clonidine reached a heart rate of under 100 beats per minute, including 3 subjects who had reverted to sinus rhythm. The placebo treatment in that trial resulted in 2 of 9 subjects with a heart rate under 100 beats per minute, including one in sinus rhythm. This is suggestive of a significant treatment effect. In neither arm was the absolute heart rate reduction significantly different from baseline, but, as stated above, more subjects achieved the goal heart rate with clonidine. The other trial of clonidine used 0.075 mg orally twice daily for 3 days and found a significant reduction in mean resting heart rate compared with placebo (90 beats per minute versus 104 beats per minute). Neither trial evaluated the subjects with exercise. One small trial evaluated magnesium sulfate versus placebo, along with digoxin to all subjects, for rate control at rest (
Brodsky, Saini, Bellinger et al., 1994). More subjects achieved adequate rate control with magnesium sulfate than with placebo, but this included patients who had converted to sinus rhythm.
Another drug evaluated against placebo was propafenone. Botto, Capucci, Bonini et al. (1997) found a significant reduction in resting heart rate compared with placebo at a dose of 450 mg orally or 600 mg orally. Three combinations were evaluated against placebo, digoxin/diltiazem (Lewis, Irvine, and McDevitt, 1988; Koh, Kwon, Park et al., 1995; Koh, Song, Kwon et al., 1995), digoxin/labetalol (Wong, Lau, Leung et al., 1990), and digoxin/betaxolol (Koh, Kwon, Park et al., 1995; Koh, Song, Kwon et al., 1995). The digoxin/diltiazem combination was effective both at rest and with exercise with no change in cardiac output compared with placebo. In the 1995 Koh, Song, Kwon trial, exercise capacity increased compared with placebo, but no difference was reported in the Koh, Kwon, Park et al. paper (1995). The digoxin/labetalol combination was not effective at rest but was effective with exercise when compared with placebo (heart rate of 154 beats per minute versus 175 beats per minute). Additionally, the time spent on the treadmill was longer with the combination than with placebo (16.1 minutes versus 14.1 minutes using the modified Bruce protocol). The digoxin/betaxolol combination significantly reduced the heart rate compared with placebo both at rest and with exercise.
Calcium-channel-blockers compared with digoxin for rate control
Figure 42. Rate control trials of calcium-channel-blocker versus digoxin for subjects with atrial fibrillation
Three trials compared diltiazem with digoxin (
Schreck, Rivera, and Tricarico, 1997;
Lewis, Irvine, and McDevitt, 1988;
Lewis, Laing, Moreland et al., 1988), and three trials compared verapamil with digoxin (
Lewis, Irvine, and McDevitt, 1988;
Pomfret, Beasley, Challenor et al., 1988;
Ahuja, Sinha, Saran et al., 1989). These are shown in . The Schreck (1997) study compared intravenous diltiazem with intravenous digoxin with followup of only 3 hours. The mean resting heart rate was lower after 3 hours on digoxin compared with baseline. Diltiazem was more quickly effective, with a significant heart rate reduction from baseline in 5 minutes. This difference was significantly different from the rate in the digoxin group at 3 minutes. This is not unexpected given the pharmacokinetics of these drugs. One study (
Lewis, Laing, Moreland et al., 1988) evaluated a single oral dose of 0.25 mg of digoxin against a single oral dose of diltiazem (60 mg or 120 mg) after several hours of followup. The statistical significance of the difference is not reported. With exercise, the rates appeared to be different, although the statistical significance was not presented. The mean maximum heart rate on digoxin was 159 beats per minute, and on diltiazem it was 143 beats per minute for the 60 mg dose and 133 beats per minute for the 120 mg dose. Notably, the cardiac output on digoxin during exercise was greater than in the two diltiazem groups (12.6 liters/minute versus 10.9 liters/minute and 9.1 liters/minute for 60 mg and 120 mg, respectively). Another study (
Lewis, Irvine, and McDevitt, 1988) used digoxin adjusted by serum level and diltiazem dosed 3 times daily for 4 weeks. The statistical significance of the results at rest and with exercise was not presented, but there was the suggestion of a greater treatment effect with diltiazem.
Lewis, Laing, Moreland et al. (1988) investigated verapamil as a single oral 80 mg dose. It reduced the mean resting heart rate to 79 beats per minute, while it was 97 beats per minute after a dose of digoxin. In the Pomfret (1988) study, verapamil was evaluated at 40 mg 3 times daily and 80 mg 3 times daily versus daily digoxin for 2 weeks. The only statistically significant result was for the reduction in mean maximum heart rate with exercise on the higher verapamil dose compared with either digoxin or the lower verapamil dose. The resting heart rates did not differ significantly. The Ahuja (1989) trial of digoxin versus verapamil found that digoxin did not decrease the resting heart from baseline, while verapamil did. Both drugs decreased the maximum heart rate with exercise from baseline. The group receiving verapamil was able to exercise longer on the treadmill than was the group receiving digoxin. This latter study (Ahuja, 1989) received a low overall quality score, losing most points for not adequately describing the enrolled subjects and for not specifying any inclusion or exclusion criteria.
One non-English-language abstract reported that diltiazem was more efficacious than digoxin both at rest and during exercise (Vitale, Auricchio, De Stefano et al., 1989).
Beta-blockers compared with digoxin for rate control
Figure 43. Rate control trials of beta-blockers versus digoxin for subjects with atrial fibrillation
The comparison of beta-blockers with digoxin was made in four studies (
Lawson-Matthew, McLean, Dent et al., 1995;
Ahuja, Sinha, Saran et al., 1989;
Ang, Chan, Cleland et al., 1990;
Wong, Lau, Leung et al., 1990) and is shown in . Lawson-Matthew (1995) and Ang (1990) evaluated xamoterol 200 mg twice daily versus daily digoxin. In both studies, digoxin reduced the mean resting heart rate significantly more than did xamoterol. The rates in both arms of the former study were much higher than in the latter study. Only the Ang (1990) trial evaluated the drugs with exercise. With the modified Bruce treadmill protocol, xamoterol controlled heart rate more effectively than digoxin (136 beats per minute versus 150 beats per minute). The time on the treadmill was similar. Ahuja (1989) found that metoprolol reduced the mean resting heart rate compared with baseline, while digoxin did not. With exercise, both drugs reduced the heart rate from baseline. The time on the treadmill was longer with metoprolol than with digoxin; the significance of this is unknown. Wong (1990) evaluated labetalol versus digoxin. There was no difference between the groups' heart rates at rest with labetalol compared with digoxin. However, with exercise, labetalol controlled the rate significantly better (156 beats per minute versus 177 beats per minute). The one non-English-language abstract reviewed reported similar findings-heart rate reduction at rest with either drug but greater rate reduction with exercise with the beta-blocker (
Lanas, Salvatici, Castillo et al., 1995).
Other drugs and combinations compared with digoxin for rate control
Eight trials evaluated digoxin versus a combination of digoxin with a calcium-channel-blocker (
Channer, Papouchado, James et al., 1987;
Stern, Pitchon, King et al., 1982;
Lang, Klein, Di Segni et al., 1983;
Schreck, Rivera, and Tricarico 1997;
Lewis, Irvine, and McDevitt 1988;
Pomfret, Beasley, Challenor et al., 1988;
Lewis, Laing, Moreland et al., 1988;
Koh, Kwon, Park et al., 1995). Seven are displayed in . The eighth (
Channer, Papouchado, James et al. 1987) reported the outcome as the area under the curve, plotting time against mean maximum resting heart rate. Only the study by Schreck (1997) did
not
find a significant decrease in mean resting heart rate with the addition of a calcium-channel-blocker to digoxin. This study, however, may have been too small to find a clinically significant difference. In all of the other trials, the addition of diltiazem or verapamil to digoxin significantly lowered the mean resting heart rate compared with digoxin alone. In the Channer (1987) study, the combination significantly reduced the heart rate compared with digoxin alone but at the expense of a greater percentage of subjects having ventricular pauses longer than 2.5 seconds. Six of the studies included an exercise evaluation. In five of the six, the calcium-channel-blocker/digoxin combination controlled the rate better than digoxin alone. The sixth did not report the statistical significance of this outcome.
One trial compared digoxin with two different doses of sotalol versus digoxin alone (Brodsky, Orlov, Capparelli et al., 1994). Both combination arms significantly reduced the mean resting heart rate and the mean heart rate with exercise compared with digoxin alone.
One trial of digoxin plus nadolol (Zoble, Brewington, Olukotun et al., 1987), one trial of digoxin plus xamoterol (Lawson-Matthew, McLean, Dent et al., 1995), and one trial of digoxin plus betaxolol (Koh, Kwon, Park et al., 1995) demonstrated significant reductions in the resting heart rate compared with digoxin alone. The resting study of digoxin plus labetalol (Wong, Lau, Leung et al., 1990) did not find a difference from digoxin alone. The Zoble (1987) study demonstrated impressive heart rate control with exercise, with the combination of digoxin plus nadolol compared with digoxin alone (120 beats per minute versus 162 beats per minute) with equivalent time on the treadmill. Similarly, in the Wong (1990) study, the combination of digoxin plus labetalol was more effective with exercise than was digoxin alone (154 beats per minute versus 177 beats per minute), and in the Koh study (Koh, Kwon, Park et al., 1995), digoxin plus betaxolol was more effective with exercise than digoxin alone (133 beats per minute versus 165 beats per minute). The Lawson-Matthew (1995) study reported fewer pauses on the xamoterol/digoxin combination than on digoxin alone (427 versus 1,207 total/24 hours).
The single trial of magnesium sulfate plus digoxin (Hays, Gilman, and Rubal, 1994) found no significant difference in mean resting heart rate compared with digoxin alone (120 beats per minute versus 131 beats per minute), but this was a very small trial.
Channer (1987) investigated double-dose digoxin up to 0.50 mg per day versus the usual maintenance dose of digoxin and found that the area under the resting heart rate curve was significantly reduced with the higher-dose digoxin but at the expense of more subjects with ventricular pauses (75 percent versus 29 percent). One non-English-language abstract studied daily dosing of digoxin compared with discontinuous dosing and found that daily dosing controlled heart rate better (Soto Pedre, Castro Beiras, and Cuna Estevez, 1990).
Other drugs evaluated for rate control
Nine other randomized trials evaluated drugs for rate control in atrial fibrillation. Two studies evaluated intravenous magnesium sulfate versus intravenous verapamil for acute control (Joshi, Deshmukh, and Salkar, 1995; Gullestad, Birkeland, Molstad et al., 1993). The Joshi study used two bolus doses of each medication, while the Gullestad study used two boluses and then a maintenance intravenous drip of medication. In both studies, a higher percentage of subjects reached a heart rate under 100 beats per minute with verapamil than with magnesium sulfate. In the Gullestad study, however, 6 of the 19 subjects treated with verapamil had symptomatic hypotension or congestive heart failure during the 24 hours of treatment.
Two studies reported on rate control with propafenone or flecainide, both at 2 mg/kg intravenously for 1 hour (Suttorp, Kingma, Jessurun et al., 1990; Kingma and Suttorp, 1992). In both studies, subjects were allowed to continue on digoxin, calcium-channel-blockers, and beta-blockers. Both drugs significantly reduced the heart rate from baseline, but data are not presented to allow the comparison of one drug to the other. The side effects with flecainide were of greater concern than with propafenone. In one study 3 of 25 subjects, and in the other 4 of 45 subjects, in the flecainide arms developed conduction abnormalities (left bundle branch block, junctional escape rhythm, and sinus arrest). Another study evaluated propafenone compared with quinidine for rate control (Lee, Chen, Chiang et al., 1996). Propafenone significantly slowed the heart rate at rest compared with quinidine. Either drug effectively slowed the heart rate compared with baseline.
The remainder of the trials support the following observations. Disopyramide did not reduce the mean resting heart rate from baseline (Boudonas, Lefkos, Efthymiadis et al., 1995). 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 three drugs together were even more effective (Dahlstrom, Edvardsson, Nasheng et al., 1992). This same study demonstrated that, with exercise, the combination of propranolol and digoxin was more effective than diltiazem and digoxin, and that the three-drug combination was
not
better than just propranolol and digoxin. The combination of pindolol and digoxin reduced the heart rate significantly compared with verapamil and digoxin (James, Channer, Papouchado et al., 1989). The combination of amiodarone and digoxin slowed the resting heart rate compared with baseline, while the combination of quinidine, verapamil, and digoxin did not, but this was a small trial with relatively low resting heart rates (Zehender, Hohnloser, Muller et al., 1992).
Summary
Compared with placebo, the calcium-channel-blockers, diltiazem and verapamil, are efficacious in reducing the resting heart rate of patients with atrial fibrillation. They are also efficacious at reducing maximum heart rate with exercise. The evidence is strong supporting these observations but tempered by the fact that some of the subjects were also receiving digoxin. Little information is available to compare diltiazem with verapamil. The trials comparing the calcium-channel-blockers with digoxin support the conclusion that the calcium-channel-blockers work more quickly for rate control. The long-term comparison of diltiazem versus digoxin was a small study and did not adequately report the statistical significance of the results. The longer study of verapamil versus digoxin supports the conclusion that verapamil is superior at heart rate control with exercise but that there is little difference in resting heart rate between the two therapies.
The effect of beta-blockers on resting heart rate is mixed. All of the beta-blockers tested reduced the heart rate during exercise compared with placebo. Exercise tolerance, however, was decreased on beta-blockers in a number of the studies; it was not clear if this caused the lower heart rate or was a result of the lower heart rate and possibly lower cardiac output. As with the calcium-channel-blocker studies, most of these trials allowed the participants to continue on digoxin. The results of the four studies that compared beta-blockers with digoxin were mixed, with xamoterol appearing to be less effective than digoxin for reducing resting heart rate, while metoprolol was more effective than digoxin. With exercise, all of the beta-blockers were more effective than digoxin for rate control. The differences in performance of the beta-blockers could be a result of the differences in receptor-specificities of the drugs, notably xamoterol, which is a selective beta-1 adrenoreceptor partial agonist and then a beta-1 adrenoreceptor competitive antagonist when sympathetic nervous system activation increases.
The trials that combined a calcium-channel-blocker with digoxin provide strong evidence that the addition of this drug to digoxin provides better rate control at rest and with exercise than does digoxin alone. However, in only two studies was the combination significantly more efficacious than the calcium-channel-blocker alone, and in two other studies the combination was not more efficacious. Thus, the evidence about the combination remains equivocal.
Few studies, mostly small, involved other drugs and drug combinations. Any conclusions regarding use of these drugs is therefore based only on suggestive evidence. Digoxin alone may reduce resting heart rate, but it may require careful attention to serum levels for maximum effectiveness. It is unlikely to be useful by itself during exercise. Clonidine may reduce resting heart rate, as may propafenone. The combination of digoxin and labetalol is efficacious at rest and with exercise compared with placebo. The combination of sotalol with digoxin is more efficacious at rest and with exercise than digoxin alone, as are the combinations of digoxin plus nadolol and digoxin plus xamoterol. Magnesium sulfate does not appear to be as efficacious as verapamil in controlling heart rate at rest, although it is better tolerated. The triple combination of digoxin, propranolol, and diltiazem is more effective than digoxin and diltiazem together, but not more so than propranolol and digoxin together. Similarly, the combination of pindolol and digoxin reduces the resting heart rate more than does the combination of verapamil and digoxin.
