Omega 3 Fatty Acids and Cardiovascular Outcomes: Systematic Review and Meta-Analysis

Some decades ago, the Inuit people of Greenland were noted to have a low rate of cardiovascular disease,¹ and this was attributed to their high intake of oily fish. Early trials testing this hypothesis found a beneficial effect of fish oil on mortality,^2,3 with subsequent recommendations developed for the use of omega 3 fatty acids (ω-3 FA) for the primary and secondary prevention of cardiovascular disease. More recent trials have, however, failed to replicate these initial positive results, with several large studies reporting null effects. Systematic reviews of the accumulating data^4–8 done during the last decade have also delivered variable findings. In part, this is because more recent overviews have included new data from large neutral trials,^9,10 and in part it is because different overviews have sought to address particular questions for specific patient groups.^4,7,8 This is the first overview to include all recent trials, systematically address the effects on all-important outcomes, and fully explore the potentially different effects achieved with particular interventions in major patient subgroups and in primary and secondary prevention. With several large trials completed in the past 18 months, we sought to more precisely and reliably define the effects of ω-3 FA on a broad range of clinical outcomes, overall and in major patient subsets.

The findings of this large systematic review, that includes data from 20 trials, >60000 individuals and >6000 major cardiovascular events raises important questions about the use of fish oil for the prevention of cardiovascular disease. Recommendations for the use of fish oil supplements are included in a number of guidelines,^34,35 but the neutral outcomes of recent large trials^9,10,16 have served to weaken rather than strengthen the evidence base. Although it remains possible that fish oil supplements will produce health benefits through the prevention of vascular complications, the size of these gains are probably smaller than previously believed, and both physician and patient expectations may need to be reset.

A key strength of this overview is the attempt to extract data on all commonly reported vascular outcomes from all trials and to systematically report the summary estimates of effect in each case. The impact of this approach has been to move the focus of attention from the positive or negative findings for particular outcomes identified in individual studies or overviews, to the overall estimates of effect across the entire body of evidence. This attempt to apply greater objectivity to the analysis of the data has not, however, been without its challenges. In part because the reporting of outcomes across studies is inconsistent and in part because there is significant heterogeneity between the trials’ results for several of the outcomes studied. This heterogeneity may also contribute to the absence of positive findings in this meta-analysis.

The heterogeneity in results between trials was large for several analyses and is unlikely to simply reflect the play of chance. Although not easy to explain in every case, the analyses of trial subgroups suggest a number of explanations based on plausible biological phenomena and correlations in the data. Greater average effects on the primary outcome were observed in the studies that used higher doses of ω-3 FA^{3,18,19,22,23,30,31} and trials done in patients with higher baseline levels of triglycerides.^3,19,31 These findings are similar to those reported for fibrates, which is notable given that both classes of agents have similar effects on lipid profiles, and especially on triglyceride levels.³⁶ We also observed greater average effects in those with younger average ages,^3,23,31 and in trials with lower proportions of hypertensives^2,3,18,19,30 and lower proportions of diabetes mellitus.^3,18,19,30 These differences between subgroups are difficult to explain on the basis of physiology, although there are also animal data that suggest a mechanism of action for lesser effects in patients with diabetes mellitus.³⁷ The subgroup findings also seem to be heavily influenced by the characteristics and results of a few large trials. The GISSI-Prevenzione trial,³ for example, studied mostly younger individuals who were nonhypertensive, and was reported before the year 2000. As a major contributor of events to all 3 of these subgroup analyses, it is easy to see how the heterogeneity of effects by these trial characteristics arises even if it is not entirely explained. The neutral findings of the more recent trials might also be associated with underlying levels of marine oil intake, but with background consumption data reported by only 7 trials^{3,9,10,17,24,29,30} that used very different definitions this was difficult to robustly investigate. Variation in the composition of the placebo compound may also contribute to the neutrality of the findings, although a subgroup analysis based on the use of an inactive control compared with corn oil or olive oil identified significant heterogeneity.

Vascular death was the only outcome for which a significant benefit was observed. This result does not seem to have been driven by an effect on sudden death or arrhythmia, both of which had more moderate and nonsignificant estimates of effect. The JELIS trial¹⁹ as well as the GISSI-Prevenzione trial³ however, found a reduction in major cardiovascular events. The absence of benefit in the more recent trials raises the possibility that the effects of ω-3 FA are determined by some as yet unquantified external factor.

The overview identified no effect of fish oil on the overall risk of stroke, although there was a trend toward harm for intracerebral hemorrhage. This is an effect that might be anticipated on the basis of the known effects of ω-3 FA on bleeding time and platelet aggregation.^38–40

The conduct of the analyses on tabular data extracted from the original trial reports was a limitation of our study design. Access to the original trial datasets would likely uncover additional outcome events from trials that did not publish data on each of the outcomes we studied. This would increase the power of the analyses by raising the number of events available as well as permitting analyses based on more directly comparable definitions than has been possible here. Analyses done on individual participant data would also allow for much more sophisticated exploration of the effects in different patient subgroups, and provide a better understanding of the sources of heterogeneity in the trial findings. A notable deficit in the current data are the systematic reporting of side effects. Total adverse events, although mostly fairly benign, were clearly higher with fish oil, and if the benefits are smaller than previously believed the risk-benefit trade-off will require more careful evaluation. Therapies that ameliorate cardiovascular risk, such as antihypertensives, statins, and antiplatelet agents have become prevalent and potent over time and, therefore, trials, such as the ORIGIN trial that include patients with moderate cardiovascular risk at baseline, are likely to comprise patients that are on these therapies. This might further dilute the detectable effect of ω-3 FA. Finally, the generalizability of the overview findings may be questioned, given that aside from a handful of trials,^18,19 the data derive entirely from the secondary prevention setting and white populations.

In conclusion, these results raise further uncertainty about the net effects of ω-3 fish oil therapy and reinforce the importance of the forthcoming ASCEND⁴¹ and R&P⁴² trials. Although it is probably reasonable for patients with existing vascular disease who are currently using fish oil to continue to do so, better evidence is required to support the more widespread promulgation of this strategy, particularly among lower risk patients. Individuals with high triglyceride levels or IgA nephropathy may be especially worthy of investigation, and higher rather than lower doses seem more likely to produce benefit. Further research in the primary prevention setting would be welcome given the potential implications of evidence about fish oil intake to dietary advice about fish consumption in the general population.

S. Kotwal, M. Jun, D. Sullivan, V. Perkovic, and B. Neal were responsible for the design of the study. S. Kotwal and M. Jun were responsible for data collection and analysis. All authors were responsible for interpretation and manuscript preparation. All authors contributed to data interpretation and critical revision of the publication. S. Kotwal had full access to all data in the study and takes responsibility for the integrity of the data and accuracy of the analysis.