The coincidence of increased risk of atrial fibrillation in randomized control trials of omega-3 fatty acids: a meta-analysis

Abstract

Watanabe and Tatsuno [1] in their state-of-the-art review on omega-3 fatty acids for the prevention of cardiovascular diseases commented that the effects of omega-3 fatty acids on the risk of atrial fibrillation (AF) have not been established. However, the Long-Term Outcomes Study to Assess Statin Residual Risk with Epanova in High Cardiovascular Risk Patients with Hypertriglyceridemia (STRENGTH) double-blind, randomized trial [2] though demonstrated no significant reduction in the primary endpoint consisting of a composite of cardiovascular death, nonfatal myocardial infarction, nonfatal stroke, coronary revascularization, or unstable angina requiring hospitalization, among patients treated with omega–3 carboxylic acids (CA) versus patients treated with corn oil (hazard ratio [HR]: 0.99 95% confidence interval [CI] 0.90–1.09), but there was an increased risk of new-onset atrial fibrillation in the omega-3 CA group compared with the corn oil group (HR: 1.69; 95% CI 1.29–2.21). Indeed, this was one of the reasons for the early termination of the STRENGTH trial. The detection of an increased risk of AF was not unique to the STRENGTH trial. The previous Reduction of Cardiovascular Events with Icosapent Ethyl–Intervention (REDUCE-IT) doubleblind, randomized trial [3] reported that the endpoint of hospitalization for AF or flutter occurred more frequently in patients randomized to icosapent ethyl (highly purified eicosapentaenoic acid ethyl ester) than placebo (3.1% vs. 2.1%, P = 0.004). While Watanabe and Tatsuno [1] commented that a definitive conclusion cannot be drawn at this stage due to the variation across the available trials, but we believe there are adequate large-scale randomized controlled trials for us to perform a meta-analysis to summarize the existing evidence on the overall effect of omega-3 fatty acids on the development of AF. Two investigators (CSK and SSH) performed a systematic literature search in electronic databases including PubMed, Embase, and Cochrane Central Register of Controlled Trials for randomized controlled trials that reported new-onset incident/recurrent AF following exposure to omega-3 fatty acids. We used the following keywords for our literature search: ‘atrial fibrillation’ and ‘omega-3’. The search was limited to studies published in the English language and performed in human participants. Study abstracts were screened against inclusion and exclusion criteria. The inclusion criteria included randomized controlled trials comparing the use of omega-3 fatty acids supplementation and containing eicosapentaenoic acid (EPA) alone or EPA plus docosahexaenoic acid (DHA) versus control, a sample size of 500 participants or more, and reported new-onset incident/recurrent AF following exposure to omega-3 fatty acids. We excluded non-randomized trials, observational studies, and trials with less than 1 year of follow-up duration. Two investigators independently extracted the following data from the included trials: number of participants, demographics of participants,, the dose of omega-3 fatty acids, the incidence of new-onset/recurrent AF, and duration of follow-up. The meta-analysis was performed by computing the incident rate ratio for the incidence of new-onset/recurrent AF using both the random-effects model and inverse variance heterogeneity (IVhet) model along with 95% confidence intervals. Heterogeneity across included trials was assessed with Cochran’s Q method with I statistics to evaluate the magnitude of the heterogeneity across studies. We also conducted subgroup analysis stratified by the dose of omega-3 fatty acids supplementation (highdose [>1 g/day EPA or EPA + DHA] and low-dose [≤1 g/day EPA or EPA + DHA]). Two investigators (CSK and SSH) assessed the risk of bias of the trials included with Version 2 of the Cochrane risk-of-bias tool for randomized trials (RoB 2) [4], which is a standardized method for assessing potential bias in reports of randomized interventions. RoB 2 is structured into a fixed set of domains of bias, focusing on different aspects of trial design, conduct, and reporting. A proposed judgment about the risk of bias arising from each domain is generated by an algorithm, where judgment can be ‘Low’ or ‘High’ risk of bias or can express ‘Some concerns’. All analyses were performed using Meta XL, version 5.3 (EpiGear International, Queensland, Australia). Our initial search retrieved 510 titles; upon removal of duplicated records, a total of 315 full-text articles were identified for detailed assessment; 6 randomized controlled trials were ultimately included in our systematic review and meta– analysis. [2,3,5–8] Our data synthesis included 75,120 patients