Article Text

Download PDFPDF

Original research
Omega-3 supplementation and cardiovascular disease: formulation-based systematic review and meta-analysis with trial sequential analysis
  1. Evangelos C Rizos1,2,
  2. Georgios Markozannes3,
  3. Apostolos Tsapas4,5,
  4. Christos S Mantzoros6,7,
  5. Evangelia E Ntzani3,8,9
  1. 1 Department of Internal Medicine, University Hospital of Ioannina, Ioannina, Greece
  2. 2 School of Medicine, European University of Cyprus, Nicosia, Cyprus
  3. 3 Evidence-Based Medicine Unit, Department of Hygiene and Epidemiology, University of Ioannina School of Medicine, Ioannina, Greece
  4. 4 Clinical Research and Evidence-Based Medicine Unit, Aristotle University of Thessaloniki, Hippokration Hospital, Thessaloniki, Greece
  5. 5 Harris Manchester College, University of Oxford, Oxford, UK
  6. 6 Department of Medicine, Beth Isreal Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts, USA
  7. 7 Section of Endocrinology, Boston VA Healthcare System, Harvard Medical School, Boston, Massachusetts, USA
  8. 8 Center for Evidence Synthesis in Health, Department of Health Services, Policy and Practice, School of Public Health, Brown University, Providence, Rhode Island, USA
  9. 9 Institute of Biosciences, University Research Center of loannina, University of Ioannina, Ioannina, Greece
  1. Correspondence to Dr Evangelia E Ntzani, Department of Hygiene and Epidemiology, University of Ioannina School of Medicine (UISM), Ioannina 45110, Greece; entzani{at}


Background Omega-3 supplements are popular for cardiovascular disease (CVD) prevention. We aimed to assess the association between dose-specific omega-3 supplementation and CVD outcomes.

Design We included double-blind randomised clinical trials with duration ≥1 year assessing omega-3 supplementation and estimated the relative risk (RR) for all-cause mortality, cardiac death, sudden death, myocardial infarction and stroke. Primary analysis was a stratified random-effects meta-analysis by omega-3 dose in 4 a priori defined categories (<1, 1, 2, ≥3 of 1 g capsules/day). Complementary approaches were trial sequential analysis and sensitivity analyses for triglycerides, prevention setting, intention-to-treat analysis, eicosapentaenoic acid, sample size, statin use, study duration.

Results Seventeen studies (n=83 617) were included. Omega-3 supplementation as ≤1 capsule/day was not associated with any outcome under study; futility boundaries were crossed for all-cause mortality and cardiac death. For two capsules/day, we observed a statistically significant reduction of cardiac death (n=3, RR 0.55, 95% CI 0.33 to 0.90, I2=0%); for ≥3 capsules/day we observed a statistically significant reduction of cardiac death (n=3, RR 0.82, 95% CI 0.68 to 0.99, I2=0%), sudden death (n=1, RR 0.70, 95% CI 0.51 to 0.97) and stroke (n=2, RR 0.74, 95% CI 0.57 to 0.95, I2=0%).

Conclusion Omega-3 supplementation at <2 1 g capsules/day showed no association with CVD outcomes; this seems unlikely to change from future research. Compared with the robust scientific evidence available for low doses, the evidence for higher doses (2–4 1 g capsules/day) is weak. The emerging postulated benefit from high-dose supplementation needs replication and further evaluation as to the precise formulation and indication.

  • meta-analysis
  • coronary artery disease
  • stroke
  • cardiac risk factors and prevention

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Omega-3 supplements are among the most popular over-the-counter products in the USA with almost 1 in 10 adults using them in 20121 for a substantial variety of diseases ranging from cardiovascular disease (CVD) protection to dementia and cancer. With regard to their medicinal use, regulatory agencies have now reach convergence and, despite their initial different approaches on the role of omega-3 supplementation for CVD protection, have currently approved them as triglyceride (TG)-lowering agents for overt hypertriglyceridaemia.2 3 Recent meta-analyses of randomised clinical trials (RCT) in the field have failed to show an indisputable benefit across patient-important clinical outcomes including mortality.4–6 Conversely, observational studies suggest that higher marine oil intake (including from dietary fish) is associated with lower CVD risk.7 Meanwhile, new RCTs are conducted using a range of omega-3 doses augmenting this conflicting evidence base.8 9

Given the existing divergence of the accumulated evidence, we approached the clinical question under consideration acknowledging the administration of omega-3 as discrete dose-based formulations and the reality of the accumulated heterogeneous evidence. We thus embarked into an evidence synopsis for omega-3 supplements and patient-important CVD outcomes implementing three complementary approaches: a) a systematic review and meta-analysis for ‘hard’ clinical outcomes, b) a trial sequential analysis (TSA) and c) an assessment of the robustness of the evidence.


A detailed description of the methods is provided in the online supplementary file 1.

Supplemental material

Data sources and searches

We searched PubMed, EMBASE and the Cochrane Central Register of Controlled Trials up to September 2019 and we also screened references of pertinent systematic reviews.

Study selection

Trials were eligible if they were randomised, double-blinded and controlled using another supplement or placebo. Driven by the priority to minimise bias and to increase the generalisation of the meta-analysis findings, we excluded trials assessing non-supplement omega-3 (eg, food fortified with omega-3), trials with patients on haemodialysis or with an implantable cardioverter defibrillator, trials with treatment duration <1 year (not allowing enough time for the intervention to show any CVD effect) and studies where CVD outcomes were assessed as safety outcomes in a composite, non-specific fashion. Eligible outcomes included all-cause mortality, cardiac death, sudden death, non-fatal myocardial infarction (MI) and stroke (any type).

Data extraction and quality assessment, data synthesis, main and sensitivity analyses

We analysed each major outcome separately (all-cause mortality, cardiac death, sudden death, non-fatal MI and all types of stroke). Since there is clinical variability in the population characteristics, the type and the dosing of supplementation, we chose to conduct our primary analysis stratified by omega-3 dose formulation. The prescribed omega-3 supplements are either capsules of 1 g containing around 0.85 g of eicosapentaenoic acid (EPA)+docosahexaenoic acid (DHA) or capsules of 1 g containing only EPA. We thus created a priori four distinct dose categories according to the total gram of EPA+DHA: <1 capsule/day (<0.84 g); 1 capsule/day (0.84–1.68 g); 2 capsules/day (1.68–2.52 g) and ≥3 capsules/day (>2.52 g). We performed random-effects meta-analyses using the relative risk (RR) and the absolute risk difference (RD) as our summary measures. The extent of heterogeneity was assessed by the I2.s1

Trial sequential analysis

TSA is a tool for quantifying the statistical reliability of data in the cumulative meta-analysis adjusting significance levels for sparse data and repetitive testing on accumulating data.s2 We estimated the required information size (IS) to detect a 10% RR reduction in the intervention arm with 80% power (β=0.2) and type I error α=0.05 (two-sided test). We considered two cases for the RR reduction based on different approaches of estimating the control arm event rate (CER): a) by meta-analysing the event proportion of the control arm, using a random-effects meta-analysis with the Freeman-Tukey double arcsine transformation in order to stabilise the variances and b) by meta-analysing the annualised CER (CER divided by study duration). TSA was performed for each omega-3 dose, as well as for lower (<2) versus higher (≥2 capsules/day) dosing schemes.

Assessment for the robustness of the evidence

We followed a grading approach similar to the one used in umbrella reviews that incorporates quantitative criteria to assess the robustness of the evidence.s3,s4 Briefly, statistically significant associations that included ≥3 studies were graded as strong, highly suggestive, suggestive and weak based on predefined criteria (online supplementary table S1). We performed sensitivity analyses for associations showing at least weak evidence (p<0.05) using a 10% credibility ceiling in order to evaluate potential spurious precision of the combined effect estimates.

Nominal statistical significance was defined at a p value <0.05. Analyses were performed using Stata V.13 and TSA (V. beta). The study is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses checklist.s5


Evidence base overview

Out of 6481 originally retrieved citations, 17 studies8–24 were deemed eligible including 83 617 randomised participants and reporting 7227 deaths, 3830 cardiac deaths, 960 sudden deaths, 2483 MIs and 1704 strokes (figure 1). Trials were published as early as 1995 and 88% of the included trials had been conducted during the period where statins were routinely recommended for CVD risk modification (1998 onwards). Sample size ranged from 72 to 25 871 participants and the mean omega-3 dose was 1.64 g/day (0.96 g/day EPA, 0.59 g/day DHA); nine, seven and one trial assessed secondary, mixed and primary CVD prevention settings, respectively (table 1). Most studies were of high methodological quality: although all studies were double-blind, the methods used to ensure adequate allocation concealment were not always clearly reported; 14 studies (82%) used intention-to-treat (ITT) approach (online supplementary table S2).

Figure 1

Meta-analysis flow chart.

Table 1

Characteristics of the eligible randomised clinical trials

Table 2

Meta-analysis and trial sequential analysis by outcome with observed and required (diversity-adjusted) information size

Assessed outcomes and evidence synthesis

The forest plots by omega-3 dose category for all outcomes are presented in detail in online supplementary figure S1.

All-cause mortality

Sixteen studies were included, reporting 7227 deaths among 83 286 participants8–23 (table 1). Omega-3 supplements were not statistically significantly associated with reduced all-cause mortality for any of the assessed formulations (<1 capsule/day: four studies, (RR, 95% CI) 1.00, 0.85 to 1.19; I2=0%; (RD, 95% CI) –0.001, –0.016 to 0.013, I2=25%; 1 capsule/day: seven studies, (RR, 95% CI) 0.99, 0.94 to 1.04; I2=9%; (RD, 95% CI) 0.001, –0.003 to 0.005; I2=8%; 2 capsules/day: two studies, (RR, 95% CI) 0.58, 0.31 to 1.07; I2=0%; (RD, 95% CI) –0.022, –0.05 to 0.006, I2=0%; ≥3 capsules/day: three studies, (RR, 95% CI) 0.89, 0.76 to 1.03; I2=0%; (RD, 95% CI) –0.009, –0.02 to 0.002, I2=0%) (figure 2). The results remained unchanged in all sensitivity and meta-regression analyses (online supplementary tables S3,4).

Figure 2

Results from the main analysis on omega-3 dosage formulation. CI, confidence interval; MI, myocardial infarction; RR, relative risk. Bold denotes a statistically significant p value. Embedded Image: <1 capsule/day (<0.84 g);

Embedded Image
: 1 capsule/day (0.84–1.68 g);
Embedded Image
Embedded Image: 2 capsules/day (1.68–2.52 g);
Embedded Image
Embedded Image Embedded Image+: ≥3 capsules/day (>2.52 g).

Cardiac death

From the 12 included studies there were 3830 cardiac deaths among 72 360 participants8 9 11 13 16 17 19–24 (table 1). For formulations containing <1 capsule/day (two studies, (RR, 95% CI) 0.99, 0.73 to 1.33; I2=0%; (RD, 95% CI) 0.000, –0.010 to 0.010, I2=0%) or 1 capsule/day (four studies, (RR, 95% CI) 0.96, 0.90 to 1.02; I2=0%; (RD, 95% CI) 0.000, –0.003 to 0.002; I2=0%) omega-3 supplements were not statistically significantly associated with reduced cardiac death. In contrast, for formulations corresponding to 2 capsules/day (three studies, (RR, 95% CI) 0.55, 0.33 to 0.90; I2=0%; (RD, 95% CI) –0.022, –0.053 to 0.009; I2=44%) or to ≥3 capsules/day (three studies, (RR, 95% CI) 0.82, 0.68 to 0.99; I2=0%; (RD, 95% CI) –0.009, –0.018 to 0.000; I2=0%), we observed a statistically significantly association with cardiac death (figure 2). The results remained significant when we analysed studies with: baseline median TG >150 mg/dL (three studies, (RR, 95% CI) 0.75, 0.58 to 0.98; I2=10%; (RD, 95% CI) –0.016, –0.041 to 0.009; I2=50%); ITT only (10 studies, (RR, 95% CI) 0.94, 0.88 to 0.99; I2=0%; (RD, 95% CI) –0.002, –0.006 to 0.001; I2=25%) and EPA only (2 studies, (RR, 95% CI) 0.82, 0.67 to 0.99; I2=0%; (RD, 95% CI) –0.009, –0.018 to 0.000; I2=0%). However, the meta-regression on EPA dose did not show a statistically significant association (online supplementary tables S3,4).

Sudden death

Seven studies provided data for 36 469 participants and 960 sudden deaths9–11 15 17 19 24 (table 1). No study evaluated sudden death using a dosage of <1 capsule/day. For formulations containing 1 capsule/day (four studies, (RR, 95% CI) 0.96, 0.8 4 to 1.10; I2=0%; (RD, 95% CI) 0.000, –0.002 to 0.002; I2=0%) or 2 capsules/day (two studies, (RR, 95% CI) 0.39, 0.14 to 1.11; I2=0%; (RD, 95% CI) –0.023, –0.064 to 0.019, I2=61%) omega-3 supplements were not statistically significantly associated with reduced sudden death. There was only one study9 evaluating ≥3 capsules/day showing a reduction in sudden death on the omega-3 arm (RR, 95% CI 0.70, 0.51 to 0.97) (figure 2). This trial was also the only one in the EPA-only sensitivity analysis. The results remained non-significant for the rest of the analyses (online supplementary tables S3,4).

Myocardial infarction

Fourteen studies (13 non-fatal,8–11 14 16 17 19–24 1 fatal/non-fatal MI13) were included with 79 064 participants and 2483 events (table 1). For formulations containing <1 capsule/day (three studies, (RR, 95% CI) 1.01, 0.75 to 1.34; I2=0%; (RD, 95% CI) 0.000, –0.007 to 0.007, I2=0%); 1 capsule/day (five studies, (RR, 95% CI) 0.90, 0.78 to 1.05; I2=51%; (RD, 95% CI) –0.003, –0.005 to −0.001; I2=0%); 2 capsules/day (three studies, (RR, 95% CI) 0.76, 0.28 to 2.05; I2=54%; (RD, 95% CI) –0.025, –0.081 to 0.030, I2=76%); ≥3 capsules/day (three studies, (RR, 95% CI) 0.90; 0.54 to 1.48; I2=49%; (RD, 95% CI) –0.012, –0.042 to 0.019, I2=21%) omega-3 were not statistically significantly associated with reduced MI (except for RD in the 1 capsule/day group) (figure 2). In the sensitivity analyses, we observed a statistically significant reduction of MI for: baseline TG >150 mg/dL (three studies, (RR, 95% CI) 0.72, 0.61 to 0.84; I2=0%; (RD, 95% CI) –0.030, –0.065 to 0.004, I2=53%) and EPA only (two studies, (RR, 95% CI) 0.74, 0.63 to 0.88; I2=0%%; (RD, 95% CI) –0.019, –0.029 to 0.008, I2=0%) (online supplementary table S3), but not in the meta-regression on EPA dose (online supplementary table S4).


Ten studies with 1704 events among 73 124 participants were included8–11 13 14 19 21–23 (table 1). For formulations containing <1 capsule/day (two studies, (RR, 95% CI) 0.98, 0.60 to 1.58; I2=3%; (RD, 95% CI) 0.001, –0.02 to 0.023, I2=13%) or 1 capsule/day (five studies, (RR, 95% CI) 1.08, 0.94 to 1.24; I2=34%; (RD, 95% CI) 0.002, –0.001 to 0.004; I2=13%) omega-3 were not statistically significantly associated with reduced stroke. There was only one non-significant study evaluating a 2 capsules/day dosage. Finally, for ≥3 capsules/day (two studies, (RR, 95% CI) 0.74, 0.57 to 0.95; I2=0%; (RD, 95% CI) –0.009, –0.016 to −0.001, I2=0%) omega-3 were statistically significantly associated with reduced stroke (figure 2). In the sensitivity analyses, we observed a statistically significant reduction of stroke only for baseline statin use >50% and only for RD (three studies, (RR, 95% CI) 0.87, 0.74 to 1.01; I2=23%; (RD, 95% CI) –0.006, –0.011 to −0.001, I2=0%), while the meta-regression on EPA dose, although not statistically significant, also indicated a protective association with stroke (nine studies, (RR, 95% CI) 0.90, 0.80 to 1.01) (online supplementary tables S3,4).

Trial sequential analysis

For the low-dose scheme, futility boundaries were crossed for all-cause mortality and cardiac death denoting that this null result is conclusive (table 2, online supplementary figures S2,S3). For the remaining outcomes the required IS had not been reached, indicating that the effect may change when new evidence accumulates (Table 2, online supplementary figures S4–S6). Conversely, the analysis of higher doses was inconclusive due to relatively low IS (online supplementary figures S7–S11).

Assessment for the robustness of the evidence

No association was assessed as strong (convincing evidence), suggestive or highly suggestive. Four associations (2 capsules/day with cardiac death and ≥3 capsules/day dosage with cardiac death, sudden death and stroke, respectively) presented statistically significant results; only two of those (2 and ≥3 capsules/day dosage with cardiac death) had at least three studies. Both associations were graded in the weak evidence category due to small number of events and high p value. Neither association survived the sensitivity analysis based on 10% credibility ceiling.


In the present synthesis effort, we summarised the association between omega-3 supplementation, formulated in four distinct dosing schemes, with ‘hard’ clinical outcomes using randomised evidence from 17 double-blind RCTs (83 617 participants). We found that omega-3 supplements at doses corresponding to ≤1 capsule/day were not associated with any of the assessed clinical outcomes, and this is further supported by a conclusive TSA analysis for all-cause and cardiac mortality. For doses corresponding to 2 capsules/day, we observed a statistically significant association with cardiac death only, while for the higher doses (≥3 capsules/day) we observed a statistically significant association with cardiac or sudden death and stroke; the supporting evidence derives from a small number of trials (one to three at best) with a corresponding insufficient information size for a conclusive association.

Our findings are in partial agreement with the results of the latest published meta-analyses, reporting no overall benefit, or minimal benefit on coronary heart disease-related events or mortality, regardless of the prevention setting, TG levels, statin use or omega-3 dose.5 6 These efforts did not focus in discrete dosing schemes or the selected dosing schemes -contrary to our approach- did not accurately reflect everyday clinical practice; in addition, we assessed the robustness of the evidence and the information size of the evidence base. With regard to mortality, we only found a benefit of higher omega-3 doses on cardiac death, and of the highest dose on sudden death. These two outcomes are frequently inter-related to each other: their definition varies (‘cardiac’ or ‘coronary’ or ‘death related to any vascular event including stroke’) and reporting in the literature is problematic. An approach to synthesise the two outcomes together was not deemed appropriate due to fundamental differences between them and the lack of individual patient data.

Suggested biological pathways for the high omega-3 dose beneficial effect include TG reduction, anti-arrhythmic properties, plaque stabilisation, anti-inflammatory effects and blood pressure reduction.25 The postulated association of high-dose omega-3 supplements with CV events cannot be explained only through the expected 20%–30% reduction of TG levels26 or through the differential effect of EPA alone or of a specific omega-3 ratio; should an emerging benefit exists, it seems to be restricted to high dosing schemes for secondary prevention in hypertriglyceridemic subjects. The American Heart Association Science Advisory Committee describes the potential CVD benefits25 for EPA+DHA or EPA-only at a dose of 4 g/day as an effective and safe option for TG reduction. Conversely, the European guidelines report that icosapent ethyl 2×2 g/day on top of statins can be considered for high-risk or very high-risk patients with TG levels 135–499 mg/dL.27

Despite the cloudy scientific evidence, omega-3 continue to consolidate their place in the market becoming the third fastest-growing type of supplement globally between 2007 and 2012 with an estimated 12% annual growth rate and estimated US$7.3 billion sales by 2020.1 Giant food companies have been added to the list of pharmaceutical companies and offer a significant portfolio of omega-3 fortified food signifying the ‘brave omega-3 new world’.28 The Trojan horse for this market penetration is the claim that they are essential for a healthy heart even in low doses. However, diet alone can provide limited amount of omega-3 with once-daily consumption of seafood serving equals <1 g EPA/DHA.29 Furthermore, although omega-3 supplementation is generally considered safe and well-tolerated,27 fish oils are at risk of contamination with chemicals and are associated with a tendency to bleeding and potential interaction with the concurrent use of warfarin.28 Moreover, the evidence on any adverse event related to high doses is limited.

To minimise the expected limitations of a meta-analysis, we included double-blind RCTs protected from selection and performance bias. Regarding attrition bias, attrition was low, an ITT approach was endorsed and no effect variation was observed. We argue against the presence of substantial information bias for the low omega-3 dosing, as a fairly large number of studies is available, publication dates span between 1995 and 2018, and we found low heterogeneity with absent publication bias indicators. For the high omega-3 doses, studies were published decades apart, the number of studies is small and the study sample size is also generally small with the exception of one mega trial. Information bias assessment is not possible with such a small number of studies, although all available databases were interrogated including CENTRAL. We acknowledge that the observed efficacy for the high doses is largely driven by the beneficial results of one study including 8179 patients with high CV risk with elevated TG levels, who were randomised to 4 g/day EPA versus placebo (mineral oil).9 The use of mineral oil as placebo could have attributed to the observed increases in the levels of TG, LDL and non-HDL cholesterol, apolipoprotein B and C reactive protein in the placebo group and raises the concern of a lower underlying cardiovascular EPA effect.30

Besides the major categories of bias, we also conducted sensitivity analyses based on a number of clinical parameters. The majority of the trials were conducted during a period of wide implementation of statins and we also analysed separately those with baseline statin use >50%. Subgroup analysis limited to EPA and the meta-regression for EPA dosage did not yield any significant effect variation. Finally, two of the older studies in the high omega-3 group (published in 199724 and 200120) included patients with acute post-MI who have radically different prognosis.


For many years, omega-3 fatty acid supplementation was assessed as one exposure group regardless of the administered dose. We argue that omega-3 supplements should be considered as distinct interventions based on the administered doses and ratios, supported by robust criteria of dose response, directionality, consistency, precision and residual confounding. Omega-3 supplementation as <2 1 g capsules/day showed no association with CVD outcomes and seems unlikely to change from future research. An emerging postulated benefit for higher omega-3 doses needs replication and further evaluation as to the precise formulation and indication (dose, ratio, type, baseline lipid profile, lipid profile alteration, comorbidities). Ongoing trials are greatly anticipated; their synthesis will shed light, provided that methodological comprehensiveness, reporting transparency, outcome harmonisation and extensive data availability and sharing are available.

Key questions

What is already known on this subject?

  • Meta-analyses of randomised clinical trials (RCT) failed to show an indisputable benefit from omega-3 supplements across patient-important clinical outcomes including mortality.

  • Latest RCTs using a range of omega-3 type and dose augment this conflicting evidence base.

What might this study add?

  • Omega-3 supplementation at conventional doses of ≤1 capsule/day does not reduce all-cause mortality, cardiac death, sudden death, myocardial infarction and stroke.

  • This finding is unlikely to change from future research.

  • For higher doses, the robustness of the evidence is weak to support an emerging benefit.

  • Additional research investigating high omega-3 doses is needed to evaluate any benefit according to the precise formulation and indication (dose, ratio, type, baseline lipid profile, lipid profile alteration, comorbidities).

How might this impact on clinical practice?

  • Omega-3 supplements should be considered as distinct interventions based on the administered doses and ratios.

  • If the aim of supplementation is the prevention of cardiovascular end points, doses of ≤1 capsule/day are unlikely to provide any real benefit.



  • ECR and GM are joint first authors.

  • ECR and GM contributed equally.

  • Contributors EEN 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. Concept and design: all authors. Acquisition, analysis or interpretation of data: all authors. Drafting of the manuscript: ECR, GM, EEN. Critical revision of the manuscript for important intellectual content: all authors. Statistical analysis: GM, EEN. Supervision: EEN.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.

  • Patient consent for publication Not required.

  • Ethics approval Ethical approval was not required by our institution because this study retrieved and synthesised anonymous data only from already published studies (not raw data provided by the authors).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.

Linked Articles