Selection of studies for inclusion.
Funnel plot based on results for mortality. OR indicates odds ratio; SE, standard error.
Overall mortality. Some studies compared more than 2 drugs, so the total numbers of studies and patients in the figure are greater than the absolute numbers of studies and patients included. CI indicates confidence interval; NA, not applicable; and OR, odds ratio. Boldface values are significant.
Withdrawals due to adverse effects and proarrhythmia. Some studies compared more than 2 drugs, so the total numbers of studies and patients in the table are greater than the absolute numbers of studies and patients included. When the numbers of studies pooled were different for the 2 outcomes, the number combined to evaluate withdrawals is given first, followed by the number combined to evaluate proarrhythmia. CI indicates confidence interval; OR, odds ratio; PAFAC, Prevention of Atrial Fibrillation After Cardioversion; SOPAT, Suppression of Paroxysmal Atrial Tachyarrhythmias; asterisk, PAFAC and SOPAT trials in both cases, which showed heterogeneity compared with other studies on quinidine or on sotalol; and dagger, the OR was calculated by the random-effects model, because the test for heterogeneity between pooled studies was significant.
Withdrawals due to adverse effects. Some studies compared more than 2 drugs, so the total numbers of studies and patients shown are greater than the absolute numbers of studies and patients included. CI indicates confidence interval; OR, odds ratio; PAFAC, Prevention of Atrial Fibrillation After Cardioversion study; and SOPAT, Suppression of Paroxysmal Atrial Tachyarrhythmias. Comparisons marked with an asterisk were from the PAFAC and SOPAT trials in both cases, which showed heterogeneity compared with other studies of quinidine or sotalol. Dagger indicates that OR was calculated by random effects model because the test for heterogeneity between pooled studies was significant for those comparisons. Boldface P values are significant.
Proarrhythmia (either bradycardia or tachycardia). Some studies compared more than 2 drugs, so the total numbers of studies and patients shown are greater than the absolute numbers of studies and patients included. CI indicates confidence interval; NA, not applicable; OR, odds ratio. Boldface P values are significant.
Atrial fibrillation recurrence. Some studies compared more than 2 drugs, so the total numbers of studies and patients in the figure are greater than the absolute numbers of studies and patients included. CI indicates confidence interval; OR, odds ratio.
Lafuente-Lafuente C, Mouly S, Longás-Tejero MA, Mahé I, Bergmann J. Antiarrhythmic Drugs for Maintaining Sinus Rhythm After Cardioversion of Atrial FibrillationA Systematic Review of Randomized Controlled Trials. Arch Intern Med. 2006;166(7):719-728. doi:10.1001/archinte.166.7.719
Copyright 2006 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2006
A variety of antiarrhythmic drugs have been used to prevent recurrence of atrial fibrillation after conversion to sinus rhythm. We performed a systematic review to determine the effect of long-term treatment with those drugs on death, embolisms, adverse effects, and atrial fibrillation recurrence.
We searched MEDLINE, EMBASE, the Cochrane Library (all up to May 2005), and the reference lists of retrieved articles. We included randomized controlled trials that compared any antiarrhythmic against control (placebo or no treatment) or another antiarrhythmic, for more than 6 months. Postoperative atrial fibrillation was excluded. Two evaluators independently reviewed the retrieved studies and extracted all data. Disagreements were resolved by discussion. All results were calculated at 1 year of follow-up.
Forty-four trials were included, with a total of 11 322 patients. Several class IA (disopyramide phosphate, quinidine sulfate), class IC (flecainide acetate, propafenone hydrochloride), and class III (amiodarone, dofetilide, sotalol hydrochloride) drugs significantly reduced recurrence of atrial fibrillation (number needed to treat, 2-9), but all increased withdrawals due to adverse effects (number needed to harm [NNH], 9-27) and all but amiodarone and propafenone increased proarrhythmia (NNH, 17-119). Class IA drugs, pooled, were associated with increased mortality compared with controls (Peto odds ratio, 2.39; 95% confidence interval, 1.03-5.59; P = .04; NNH, 109). No other antiarrhythmic showed a significant effect on mortality compared with controls. We could not analyze other outcomes because data were lacking.
Class IA, IC, and III drugs are effective in maintaining sinus rhythm but increase adverse effects, and class IA drugs may increase mortality.
Atrial fibrillation (AF) is the most common sustained arrhythmia and is associated with important morbidity and mortality related to stroke, other embolic complications, and heart failure.1,2 In developed countries, AF has grown progressively as a contributing cause of hospitalization and death in recent decades.3 Many patients, as many as 70% in some studies,4 recover sinus rhythm spontaneously after an episode of recent-onset AF. If not, electrical and pharmacologic cardioversion are very effective in restoring sinus rhythm. However, the problem lies in the fact that the recurrence rate of AF is high: without treatment, only 20% to 30% of patients who converted remain in sinus rhythm at 1 year.5,6
Therefore, a variety of antiarrhythmic drugs (AAs) have been widely used to prevent recurrence of AF. However, their effects on outcomes other than merely maintaining sinus rhythm are not well known. Serious adverse events are possible, as some of these drugs, such as quinidine sulfate7,8 or flecainide acetate,9 have the potential to induce life-threatening arrhythmias. Overall, a rhythm-control strategy, using AAs to maintain sinus rhythm, has not shown clear differences when compared with a rate-control strategy in outcomes such as mortality or stroke.10 In addition, the relative effectiveness and safety of the different AAs used for this indication are not well defined.
Attempts to summarize the available multitude of studies on AAs in this setting have been incomplete. Existing meta-analysis and reviews have focused on individual specific drugs,7,11,12 have pooled studies using AAs for acute cardioversion together with long-term treatment,13 or did not evaluate outcomes other than sinus rhythm maintenance.14 Consequently, we aimed to conduct a comprehensive systematic review of randomized controlled trials studying long-term use of AAs, in patients converted to sinus rhythm after having AF, with the objective of determining the effect of the different AAs not only on the recurrence of AF but also on other important clinical outcomes: death, stroke and other embolisms, drug adverse effects, and proarrhythmia.
We searched the Cochrane Central Register of Controlled Trials, MEDLINE (PubMed), and EMBASE (Ovid), all up to May 2005, by using the following terms: (Atrial Fibrillation OR [(atrial OR atrium OR auricular) AND fibrillat*]) AND (Anti-Arrhythmia Agents OR antiarrhythmi* OR anti-arrhythm* OR procainamide OR disopyramide OR quinidine OR mexiletine OR flecainide OR propafenone OR bisoprolol OR esmolol OR amiodarone OR dofetilide OR sotalol OR ibutilide OR azimilide OR dronedarone OR moricizine OR cibenzoline).
Search terms were combined with the strategy to identify randomized controlled trials developed by the Cochrane Collaboration.15 In addition, we checked the reference lists of retrieved studies, recent guidelines, meta-analyses, and general reviews on AF. Any article that seemed to possibly meet the criteria listed in the next section was retrieved. No limitation by language was applied. During the process, publications in English, German, Italian, French, Spanish, and Swedish were retrieved, translated, and evaluated.
We included only randomized controlled trials that included patients older than 16 years who had AF of any type and duration and in whom sinus rhythm had been restored (spontaneously or by any therapeutic means), and compared long-term treatment (at least 6 months) with any available AA against a control (placebo, no treatment, or drugs for rate control) or against other AAs. In each study, all treatment groups had to be similar with regard to (1) cardiac disease (frequency, type, and severity); (2) type and duration of AF; and (3) management of anticoagulation, heart failure, and hypertension. Finally, studies had to evaluate at least 1 of the following outcomes: all-cause mortality, embolic complications (stroke, peripheral embolisms), adverse events leading to withdrawal of treatment, proarrhythmia, recurrence of AF, and anticoagulation use at the end of follow-up. We considered the following as proarrhythmia: sudden death, any new symptomatic arrhythmia (including symptomatic bradycardia), worsened preexisting arrhythmias (ie, rapid AF), and newly appeared QRS or QT widening when they forced treatment to stop.16 Crossover studies and studies on AF after cardiac surgery were excluded.
Two of us (C.L.-L. and S.M., M.A.L.-T., or J.-F.B.) independently read the full text of the studies retrieved and selected the trials that met the inclusion criteria, then assessed methodologic quality and extracted data on an intention-to-treat basis. Any difference between reviewers was decided by discussion and consensus. Records of the study selection process were kept and a Quality of Reporting of Meta-analyses statement was prepared.17 Quality was rated according to the adequacy of allocation concealment (concealing assignment until treatment had been allocated), ranked as A (explained and adequate) or B (unclear or not well explained). Studies where allocation was not concealed were not considered truly randomized and were not included. When necessary, the authors of primary studies were contacted for additional information.
Data from AAs were pooled and analyzed individually (each specific drug) and grouped by pharmacologic class. Peto odds ratios (ORs) with 95% confidence intervals (CIs) were calculated for all outcomes by means of a fixed-effects model. Homogeneity between included studies was tested by the Mantel-Haenszel χ2 test. If significant heterogeneity between studies was observed, differences in clinical characteristics of the studies were searched and either a random-effects model was used or studies were not combined if clinically dissimilar. We used the Cochrane Collaboration software RevMan (version 4.2.2; available at: http://www.cc-ims.net/RevMan) for all statistics. Significant results were also expressed as the number needed to treat or the number needed to harm, to prevent or produce, respectively, one adverse outcome. A funnel plot was constructed, based on the data for mortality.
Sensitivity analyses to test the robustness of the results were performed by (1) calculating both extremes of intention-to-treat analysis possibilities, ie, the “best case” counting all missing patients as being free of events, and the “worst case,” counting all missing patients as having events; and (2) selectively pooling best-quality studies and studies with more than 250 patients.
Subgroup analyses were planned as follows: (1) recent-onset or persistent AF; (2) structurally normal heart or heart failure; and (3) studies where warfarin sodium treatment was mandatory.
From a total of 2576 REFERENCES found, we assessed 151 articles in more detail. Forty-four studies fulfilled inclusion criteria and had usable data.18- 61 They represented a total of 11 322 patients. Figure 1 illustrates the selection process. Agreement between reviewers was excellent. The funnel plot was asymmetrical, indicating that publication bias is possible (Figure 2).
The Table details the characteristics of included studies. All were prospective, randomized, parallel-group, controlled trials. Twenty-one trials (5935 patients) compared an AA with a control, 9 trials (3265 patients) compared 2 AAs with a control, and 14 trials (2122 patients) compared 2 or more AAs with each other. The control was placebo in 25 trials, β-blockers in 1, digoxin in 1, and no treatment in 3. Most trials comparing AAs vs control were single or double blind; in contrast, most trials comparing 2 different AAs were open label.
The type of AF most frequently studied was persistent AF (60% of patients in the pooled population). Only 6 studies included exclusively paroxysmal or recent-onset AF. The proportion of patients having underlying heart disease, as defined by each study, varied widely, from 33% to 100%; only 1 study selectively included patients without any structural heart abnormality.30 Nonetheless, the mean left ventricular ejection fraction was greater than 50% in all except 3 trials.28,35,45 The most frequent abnormalities were, unsurprisingly, coronary disease (5% to 50% of patients), hypertension, and valvular heart disease, the last more frequent in older studies.
The percentage of patients lost during follow-up was reported in 30 trials and was small: 5% to 10%. Imputing those missing patients as events—the worst-case intention-to-treat scenario—seldom modified results, so the best-case intention-to-treat analysis—missing patients counted as being free of events—is given by default, and when differences existed, they are stated.
All results are calculated at 1 year of follow-up.
Figure 3 shows results for overall mortality. Mortality rate was low: 0% to 4.4%. No death at all was reported in trials of flecainide. The only exception to the general low mortality was the Danish Investigations of Arrhythmia and Mortality on Dofetilide study,28 which specifically recruited patients with advanced heart failure and had a mortality of 31% at 1 year.
Compared with controls, a nonsignificant trend to increased mortality appeared with quinidine (OR, 2.26; 95% CI, 0.93-5.45; P = .07). This increase in mortality was significant (OR, 2.39; 95% CI, 1.03-5.59; P = .04) when drugs of class IA—quinidine and disopyramide phosphate—were combined. The corresponding number needed to harm for combined class IA drugs was 109 patients treated for 1 year to have 1 excess death; the 95% CI was very large (34-4895 patients).
In sensitivity analyses, counting missing patients as deaths confirmed these results, showing a significant increase for quinidine alone. However, selectively pooling trials with adequate allocation concealment or those including more than 250 patients left only 2 trials—Prevention of Atrial Fibrillation After Cardioversion (PAFAC)44 and Suppression of Paroxysmal Atrial Tachyarrhythmias (SOPAT)56—in which no effect on mortality was apparent. These 2 trials used a lower dosage of quinidine sulfate (320-480 mg/d) than other studies (800 to 1800 mg/d), and combined it with verapamil hydrochloride.
A nonsignificant trend to increased mortality appeared also with sotalol hydrochloride (OR, 2.09; 95% CI, 0.97-4.49; P = .06). This trend became significant if missing patients were counted as deaths but was less pronounced if only high-quality trials or trials having more than 250 patients were analyzed. Amiodarone, when compared with combined class I drugs, showed a significant reduction in mortality (OR, 0.39; 95% CI, 0.19-0.79; P = .009; number needed to treat, 17; 95% CI, 13-52). This effect was mainly due to the weight of 1 study18 and persisted in sensitivity analysis. Amiodarone showed no difference when compared with placebo.
No other significant difference in mortality appeared in the remaining comparisons.
Figure 4 shows results for these outcomes (see also eFigure 1 and eFigure 2). Withdrawals due to adverse effects were significantly more frequent with all AAs compared with controls, with few exceptions: aprindine hydrochloride and dofetilide, both having results from only 1 study. Heterogeneity between studies was detected for quinidine (P<.001) and sotalol (P = .03). In both cases that finding was due to 2 trials, again PAFAC44 and SOPAT,56 in which neither quinidine nor sotalol showed more withdrawals than placebo, contrary to other studies.
Almost all AAs showed significantly increased proarrhythmic effects, the only exceptions being amiodarone and propafenone hydrochloride. Pooled event rates varied depending on the drug used, from 9% to 23% for withdrawals due to adverse effects and from 1% to 7% for proarrhythmia. Number needed to harm ranged, for withdrawals, from 9 with quinidine to 27 with amiodarone, propafenone, or sotalol; and for proarrhythmia, between 17 with flecainide and 119 with dofetilide.
When the different AAs were compared, quinidine caused more withdrawals than the other class I drugs (OR, 2.25; 95% CI, 1.45-3.51; P<.001) but not more proarrhythmia. Amiodarone was associated with significantly fewer withdrawals (OR, 0.52; 95% CI, 0.34-0.81; P = .004) and fewer proarrhythmic events (OR, 0.28; 95% CI, 0.13-0.59; P<.001) than combined class I drugs; results were not modified in the sensitivity analysis.
Results for atrial fibrillation recurrence are presented in Figure 5. All studied class IA and IC drugs and all class III drugs except dronedarone proved to significantly reduce recurrences of AF. Pooled recurrence rates at 1 year were high: 71% to 84% in controls and reduced to 44% to 67% in treated patients. Corresponding average numbers needed to treat were 3 with amiodarone, 4 with flecainide, 5 with dofetilide and propafenone, and 8 with quinidine and sotalol; the 95% CIs varied between 2 and 14.
In comparisons between AAs, amiodarone reduced recurrences of AF significantly more than combined class I drugs (OR, 0.31; 95% CI, 0.21-0.45; P<.001) and more than sotalol (OR, 0.43; 95% CI, 0.33-0.56; P<.001). No other differences between AAs were detected.
Long-term anticoagulation with warfarin was mandatory for every patient during all the follow-up in only 3 studies.26,32,59 In the rest, the decision was left to the judgment of the attending physician. No trial reported the frequency of anticoagulation during the follow-up. Only 7 of the studies comparing AAs with a control reported strokes occurring during the trial,23,32,36,41,50,55,56 but it is not certain that reporting was exhaustive. They reported a total of 20 strokes in 1755 patients treated with AAs and 6 strokes in 650 patients in the corresponding control groups. Five trials reported data on the incidence of heart failure,30,33,40,46,48 which was low and without obvious differences between groups.
Subgroup analysis of patients with persistent AF replicated these results. The rest of the planned subgroup analyses were not possible.
Our study confirms that several AAs belonging to different pharmacologic classes (all class IA and IC drugs, and all class III drugs except dronedarone) proved to be effective in maintaining sinus rhythm after conversion of AF. Depending on the drug, recurrence of AF was reduced by 30% to 50% with respect to controls. Amiodarone seemed to be the most effective. However, the success of AAs was limited: AF still recurred in 42% to 67% of treated patients. Most trials focused on effects on electrical rhythm as the main outcome, as restoring and keeping normal auricular function is believed to be the most advantageous to the patient. We wanted to know whether, in addition to maintaining sinus rhythm, long-term treatment with these drugs carried other measurable clinical benefits.
Concerning mortality, which was low, our results show that only class IA drugs (quinidine and disopyramide together) demonstrated a significant effect on mortality, increasing it. These results were not reproduced when only the PAFAC and SOPAT studies were analyzed. These 2 trials are recent, high-quality, large (848 and 1033 patients, respectively) studies that compared quinidine, sotalol, and placebo and showed no increase in mortality. A probable explanation is that both used a lower dose of quinidine than other studies and that quinidine was combined with verapamil, which has been shown to reduce some of the proarrhythmic effects of quinidine, such as accelerated atrioventricular conduction. In addition, the proportion of patients having structural heart disease was lower in the PAFAC and SOPAT studies than in older trials.
When AAs were compared with each other, amiodarone produced less mortality than class I drugs altogether. However, it is important to emphasize that amiodarone proved no advantage in mortality compared with placebo. To date, 2 previous meta-analyses have assessed mortality in this setting: Coplen et al7 focused on quinidine and also found an increased mortality with this drug, while Nichol et al13 found no difference with any AA regarding mortality, but most of the trials they pooled had very short follow-up periods.
We also found evidence of increased adverse effects. Virtually all of the AAs caused more withdrawals due to adverse effects than controls did. For proarrhythmia, only amiodarone and propafenone showed no significant difference compared with controls. Amiodarone was associated with significantly fewer adverse effects and proarrhythmic events than class I drugs combined. However, these are results at 1 year of follow-up, and adverse effects of amiodarone are well known to increase in frequency over time.
Unfortunately, we could not evaluate other important clinical outcomes, such as stroke and heart failure, because they were rarely reported. Moreover, the frequency of use of long-term anticoagulation was not available, complicating this evaluation. While the frequency of stroke and heart failure, in the few studies that reported it, was very low and did not show apparent differences between treatment groups, the actual effect of long-term AAs on these important end points remains unknown.
To the best of our knowledge, this is the most exhaustive systematic review performed to date in this setting, assessing not only the effect of AAs on maintaining sinus rhythm but also other relevant outcomes. We have synthesized the accumulated experience of 44 good-quality randomized controlled trials and focused on diverse AAs, representing approximately 30 years of research, some recently published,37,44,50,56 and involving more than 11 000 patients in total. A variety of sensitivity analyses produced the same results.
The main limitation of this study is the lack of data on some outcomes. Other limitations are as follows: (1) In many studies patients were followed up until AF recurred and not thereafter; hence, events between that point and the complete 1 year of follow-up might have been missed. (2) Mean age of included patients was about 60 years, and most of them had a normal left ventricular ejection fraction. We do not know whether our results can be extrapolated to other patient populations, such as older people or patients with impaired left ventricular function.
In conclusion, after cardioversion of AF, various AAs appear to be moderately effective in maintaining sinus rhythm in the long term, but all show evidence of adverse effects, and data on various important clinical outcomes, such as stroke, embolisms, and heart failure, are sparse among available trials. Therefore, the final risk-benefit ratio of long-term treatment with those drugs remains unclear. With respect to the relative effectiveness and risks of each drug, class IA drugs should be used most carefully for this indication because of the risk of increasing mortality. On the other hand, on the basis of results at 1 year, amiodarone could be preferred because it seems to be the most effective in preventing recurrences of AF, produces fewer adverse events, and is associated with less mortality than class I drugs while not increasing mortality compared with controls. However, we do not know whether those advantages are maintained with longer treatment. A more precise estimation of the effects of long-term treatment with AAs on appropriate clinical outcomes, as compared with placebo or alternative strategies to manage AF—ie, rate control,10 drugs only for terminating AF recurrences,62 and radiofrequency ablation63—would be needed.
Correspondence: Carmelo Lafuente-Lafuente, MD, Service de Médecine Interne A, Hôpital Lariboisière, 2 rue Ambroise Paré, 75010 Paris, France (firstname.lastname@example.org).
Accepted for Publication: September 12, 2005.
Author Contributions: Dr Lafuente-Lafuente had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Financial Disclosure: None.
Funding/Support: This work was supported by the Unité de Recherche Thérapeutique, Hôpital Lariboisière, Paris.
Role of the Sponsor: The sponsor had no role in the conduct of the study or in the collection, analysis, and interpretation of the data. It had no role regarding the preparation, review, and approval of the manuscript, or the decision to publish.
Additional Resources: The online-only eFigure 1 and eFigure 2 are available.
Acknowledgment: The protocol for this review, with the detailed search strategy and methods, has been published in the Cochrane Library. We thank the Cochrane Heart Group for their review process and help in searching and obtaining articles; all of the authors who answered our request for additional data; Charles Caulin, MD, PhD, for his most valuable suggestions; and Barbara Stadler, MD, for translating articles from German.