N‐3 Polyunsaturated Fatty Acids to Prevent Atrial Fibrillation: Updated Systematic Review and Meta‐Analysis of Randomized Controlled Trials

Background Previous studies have suggested that n‐3 polyunsaturated fatty acids (n‐3 PUFAs) have antiarrhythmic effects on atrial fibrillation (AF). We aimed to assess the effects of therapy with n‐3 PUFAs on the incidence of recurrent AF and on postoperative AF. Methods and Results Electronic searches were conducted in Web of Science, Medline, Biological Abstracts, Journal Citation Reports, and the Cochrane Central Register of Controlled Trials databases. In addition, data from the recently completed FORωARD and OPERA trials were included. We included randomized controlled trials comparing treatment with n‐3 PUFAs versus control to (1) prevent recurrent AF in patients who underwent reversion of AF or (2) prevent incident postoperative AF after cardiac surgery. Of identified studies, 12.9% (16 of 124) were included, providing data on 4677 patients. Eight studies (1990 patients) evaluated n‐3 PUFA effects on AF recurrence among patients with reverted AF and 8 trials (2687 patients) on postoperative AF. Pooled risk ratios through random‐effects models showed no significant effects on AF recurrence (RR, 0.95; 95% CI, 0.79 to 1.13; I2, 72%) or on postoperative AF (0.86; 95% CI, 0.71 to 1.04; I2, 53.1%). A funnel plot suggested publication bias among postoperative trials but not among persistent AF trials. Meta‐regression analysis did not find any relationship between doses and effects (P=0.887 and 0.833 for recurrent and postoperative AF, respectively). Conclusions Published clinical trials do not support n‐3 PUFAs as agents aimed at preventing either postoperative or recurrent AF. Clinical Trial Registration URL: http://www.crd.york.ac.uk/PROSPERO. Unique Identifier: CRD42012002199.

A trial fibrillation (AF) is the most common arrhythmia in adults, and its incidence is increasing worldwide. 1,2 Classic antiarrhythmic drugs used to preserve normal sinus rhythm in patients with previous AF have shown limited efficacy as well as frequent and serious harmful effects. [3][4][5] For this reason, actively searching for antiarrhythmic agents without the common adverse events of classic antiarrhythmic drugs has become increasingly important. [6][7][8][9] Although statins and angiotensin II receptor blockers may favorably affect the atrial remodeling associated with AF, 6-8 the results of clinical trials have been neutral. [9][10][11] In addition, new antiarrhythmic drugs proved to be neither more effective nor safer than classic therapies. 12 N-3 polyunsaturated fatty acids (PUFAs) from animal sources have shown antiarrhythmic properties on ventricular arrhythmia in patients with previous myocardial infarction, 13,14 although recent findings failed to replicate these results. 15,16 Encouraged by previous basic, 17,18 epidemiological, 19,20 and clinical data 13,14 on ventricular arrhythmias, a body of experimental data has suggested a potential role of these compounds in treating atrial arrhythmias. 21,22 Clinical trials have focused their attention on 2 different populations: patients with persistent/paroxysmal AF for whom the principal objective is to preserve normal sinus rhythm after reversion and patients who have undergone cardiac surgery to prevent the onset of new AF. Overall, the results of clinical trials led to conflicting results. Previous systematic reviews failed to provide a definitive answer because the numbers of patients and events were relatively small, [23][24][25] but since their publication, the number of patients available for assessment has more than doubled. Therefore, we have conducted a systematic review and meta-analysis to evaluate the effects of n-3 PUFAs on sinus rhythm maintenance after AF reversion and on AF incidence after cardiac surgery.

Methods
The protocol for our study is registered in the international prospective register of systematic reviews (PROSPERO). Registration number CRD42012002199 (available from http://www.crd.york.ac.uk/PROSPERO/display_record.asp? ID=CRD42012002199). This systemic review is reported following recommendations of the PRISMA statement. 26

Eligibility Criteria
To be included in the meta-analysis, studies had to be randomized controlled trials evaluating any dose and formulation of n-3 PUFAs, administered as pharmacological prep-arations and conducted in either of the following settings: sinus rhythm maintenance after spontaneous electrical or pharmacological cardioversion or AF prevention in patients undergoing cardiac surgery.
Studies could be double-blind, placebo-controlled, or unexposed controlled trials. In sinus rhythm maintenance trials, patients could be randomized with AF or in sinus rhythm (ie, before or after reversion).
In cardiac surgery trials, all patients had to be in sinus rhythm at randomization. No restriction criterion on type of surgery was adopted.
We excluded nonrandomized studies, those that did not reported data on atrial fibrillation occurrence during follow up, those with no follow-up (ie, evaluating the electrophysiological effects of 1 or few doses of n-3 PUFAs), and those that were reported in languages other than English.
Trials reported as proceeding abstracts were included if other inclusion criteria were met.

Search Strategy
We conducted an electronic search in the Web of Science database, simultaneously searching in the Web of Science  Search terms were "(n-3 PUFA OR n 3 polyunsaturated OR fatty acids OR fish oil OR docosahexaenoic OR eicosapentaenoic OR polyunsaturated fatty acids) AND (atrial fibrillation OR atrial flutter) AND (random* OR randomised OR randomized)," searched in titles or as topics.
Additional searches included reference lists of relevant articles and reference lists of previous systematic reviews on this topic. Data from the recently released FORxARD (Fish Oil Research with x-3 for Atrial fibrillation Recurrence Delaying, Trial Registration Identifier: NCT00402363) trial were also included in the analyses.

Data Collection
Two investigators (J.M. and A.M.) independently collected information from studies retrieved by the initial search in an unblinded fashion. Titles and abstracts were scrutinized to check eligibility, and when inclusion and exclusion criteria were unclear, the full text report was evaluated.
Data abstracted from the included studies were authors, date of publication, design, comparator, dosage and formulation of n-3 PUFA, loading dose, recurrent/incident AF definition, number of participants, patient characteristics, target population (ie, persistent AF or postoperative AF), type of surgery, and outcomes of interest. Abstracted data were collected in paper form and then entered in a database designed for the study. All discrepancies were solved by consensus with a third investigator (D.F.). No attempt was made to standardize definitions of end points. Quality of the studies was assessed using the score suggested by Jadad et al. 27

Outcomes
The primary outcome was the occurrence of AF. For persistent AF studies, this was recurrent AF, and for postoperative studies, incident AF. Secondary outcomes were all-cause mortality and length of ICU stay (only for postoperative studies).

Statistics
All analyses were conducted separately for trials of persistent AF and postoperative AF. For every study, we computed risk ratios and corresponding 95% confidence intervals (CIs) for outcomes in the n-3 PUFA group compared with control/ placebo group. Risk ratios from each individual trial were Table 1. pooled using the random-effects model approach as described by DerSimonian and Laird. 28 For postoperative AF trials, we also computed mean length of ICU stay and pooled all means using weighted mean differences (also with a random-effects model).

Continued
Heterogeneity was assessed through the Cochran Q test, with a P<0.1 indicating statistically significant heterogeneity. To further measure heterogeneity, inconsistency (ie, the I 2 statistic) was computed, considering >50% as moderate inconsistency. 29 Additional prespecified analyses to explore possible sources of heterogeneity between studies include a repeated pooled analysis excluding studies with less than the median quality score. Other sensitivity analyses by b-blocker therapy, amiodarone therapy, age (≤ or >median), and sex were conducted. To further explore potential sources of heterogeneity in the estimated effect sizes between studies, we conducted meta-regression analyses, in which the dependent variable (the [log] risk ratios) was weighted-regressed against covariates at the study level (n-3 PUFA dose, AF rate in the control group, quality score, mean age, proportion who were male, rate of b-blocker and amiodarone use at baseline, left ventricular ejection fraction). Meta-regression was conducted separately for recurrent AF studies and postoperative AF studies. 30 Publication bias was evaluated using visual inspection of the funnel plot and Egger test, with P<0.1 indicating evidence of statistically significant asymmetry in the funnel plot. 31
Among trials evaluating n-3 PUFAs to prevent recurrence of AF after reversion, follow-up ranged from 6 to 12 months. Assessment of outcomes included transtelephonic monitoring in 2 studies, 24-hour Holter monitoring in 2 studies, and ECG in all studies that reported this information.
Studies evaluating n-3 PUFAs to prevent AF after cardiac surgery had follow-up limited to hospitalization: 1 study followed patients up to 14 days through telephone contact and the other up to 30 days. 44,47 Methods to assess incident AF consisted of continuous monitoring for 2 to 5 days postsurgery and ECG thereafter in most studies, with only 1 study continuously monitoring patients for the complete hospital stay. 42 Median sample size was 186 patients (189 among persistent AF studies and 181 among postoperative AF) ( Table 2). Median Jadad's score was 4 points, with 3 reports having <3 points. 37,38,45 Studies evaluating n-3 PUFAs to prevent recurrent AF had higher scores than postoperative studies (median, 5 versus 4 points, respectively). Table 2 shows demographic, clinical, and echocardiographic characteristics of study participants. Most patients were male, and the mean age ranged between 55.5 and 69.5 years. There was high prevalence of hypertension, diabetes, and previous myocardial infarction or coronary artery disease. Concomitant therapy with b-blockers ranged from 25.5% to 84.5%. Use of amiodarone was an exclusion criterion in 8 studies and mandatory in 2 and varied from 3.8% to 63% in the remaining studies (1 study, reported as proceeding abstract, did not report this information). Mean left atrial dimension (or area) was mildly dilated in most studies, and mean left ventricular ejection fraction was in the normal range.

Outcomes
Maintenance of sinus rhythm after AF reversion studies Figure 2 shows the cumulative results of n-3 PUFAs on AF. Treatment had no effect on AF recurrence (RR, 0.95; 95% CI, 0.79 to 1.13), with moderate inconsistency across trials  (I 2 , 72.0%). Overall death rates were low among the 5 studies that reported this information, and there was no effect of n-3 PUFAs on mortality (RR, 0.85; 95% CI, 0.26 to 2.77; Figure 3).

Postoperative studies
Among studies evaluating the effects of n-3 PUFAs to prevent postoperative AF, treatment resulted in nonsignificant reduction of AF during hospitalization (RR, 0.86; 95% CI, 0.71 to 1.04), with moderate inconsistency (I 2 , 53.1%) (Figure 2). There were no effects on in-hospital death across the 6 studies that reported this outcome (RR, 0.86; 95% CI, 0.41 to 1.84; Figure 3).
Three studies reported information on ICU length of stay. Treatment with n-3 PUFAs resulted in significantly shorter stay compared with the control (weighted mean difference, À0.69; 95% CI, À1.27 to À0.12), with low inconsistency across trials (I 2 , 0.0%) (Figure 4). Figure 5 shows a funnel plot. Formal testing of publication bias showed no evidence of bias among studies evaluating n-3 PUFAs to prevent recurrent AF (P=0.87). Visual inspection of the funnel plot and formal testing of its asymmetry showed evidence of publication bias among studies assessing effects of n-3 PUFAs to prevent postoperative AF (P=0.09).

Sensitivity Analyses
Repeated analyses excluding studies with a quality score lower than the median did not materially change the results. For AF recurrence studies, the risk ratio was 0.99 (95% CI, 0.83 to 1.20), and for postoperative AF, it was 0.89 (95% CI, 0.71 to 1.12). Other sensitivity analyses, including previous/ concomitant use of b-blockers, amiodarone, age, and sex did not provide different results (Table 3). Meta-regression analyses suggested that the dose of n-3 PUFAs did not influenced their effects (for recurrent AF, P=0.887; for postoperative AF, P=0.833). Additional meta-regression analyses did not find  associations between several study-level covariates and the effect-size estimates (Table 4 and Figures 6 and 7).

Discussion
The present systematic review was sufficiently powered (93% of power at the conventional type I error level of 0.05 to detect a 20% reduction in recurrent AF and 88% to detect a 15% reduction; these numbers for postoperative AF were 96% and 80%, to detect 20% and 15% reductions, respectively) to close the uncertainty that existed regarding potential effects of n-3 PUFAs for the prevention of AF. The data obtained with >4500 patients and 1753 events regarding the efficacy of n-3 PUFAs in preventing recurrent AF  showed that the effect of pharmacological supplementation with these compounds resulted in no benefit. That these results were obtained evaluating studies that included patients who were receiving amiodarone or b-blockers and others that excluded these treatments and across a wide range of doses of n-3 PUFA strengthen the main finding.
In the setting of secondary prevention of AF and beyond the heterogeneity observed in clinical trial designs, the number of patients and events collected in this meta-analysis allows confidence in the main conclusion of neutral effects of n-3 PUFAs for this clinical indication. Moreover-and for reasons that remain unclear-the 2 most recent trials conducted, which contributed a large number of patients and events, 36,39 showed an excess accumulation of AF among patients randomized to n-3 PUFAs compared with those assigned to placebo. When separately considering the results of individual clinical trials, it seemed appropriate to propose a large, "definitive" randomized trial, but this meta-analysis would call for caution. The economic and logistic effort to conduct such a trial would be substantial, and the cumulative results seem to be confirmatory of no benefit.
Similarly, results obtained for the prevention of postoperative AF have also been disproved according to this analysis. The incorporation of a large and well-conducted clinical trial 47 provided strength to this systematic review, a characteristic that was not present in previous meta-analyses. [23][24][25] Other systematic reviews had appraised the evidence on the effects of n-3 PUFAs to prevent AF [23][24][25]48 with no definitive results, but our analysis, which more than doubled the number of patients and events, precludes any beneficial effects of n-3 PUFAs for the prevention of secondary and postoperative AF.
Trials had different designs, and this resulted in an important degree of heterogeneity. Of note, the doses of n-3 PUFAs varied nearly 10-fold across studies. The present analysis also explored whether the effects of n-3 PUFAs on AF prevention would vary with the dose used in different trials. The result of the meta-regression does not support the hypothesis, as no relationship was observed between dose and effect. However, it also should be noted that the studies were too few to draw any definitive conclusion regarding the dose effect.
Besides AF prevention, cumulative results also failed to demonstrate any benefit of n-3 PUFA supplementation on other relevant end points, including mortality, although all trials were underpowered to detect differences on this end point.
It is important to note that these trials were restricted to a particularly high-risk population (ie, those under secondary prevention and those undergoing cardiovascular surgery), and these results should not be extrapolated to a potential beneficial effect of these compounds in the context of primary prevention of AF.
In addition, the trials had an inherent short duration. It could be the case that n-3 PUFA supplementation would require a more prolonged duration of exposure to see a potential benefit.
In conclusion, the present meta-analysis provides confident evidence of the lack of usefulness of oral supplementation of n-3 PUFAs for the secondary prevention of AF and for the incidence of new AF in patients undergoing cardiovascular surgery.