Skip to main content
The Cochrane Database of Systematic Reviews logoLink to The Cochrane Database of Systematic Reviews
. 2020 Feb 29;2020(3):CD003177. doi: 10.1002/14651858.CD003177.pub5

Omega‐3 fatty acids for the primary and secondary prevention of cardiovascular disease

Asmaa S Abdelhamid 1, Tracey J Brown 1, Julii S Brainard 1, Priti Biswas 2, Gabrielle C Thorpe 3, Helen J Moore 4, Katherine HO Deane 3, Carolyn D Summerbell 5, Helen V Worthington 6, Fujian Song 1, Lee Hooper 1,
Editor: Cochrane Heart Group
PMCID: PMC7049091  PMID: 32114706

Abstract

Background

Omega‐3 polyunsaturated fatty acids from oily fish (long‐chain omega‐3 (LCn3)), including eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA)), as well as from plants (alpha‐linolenic acid (ALA)) may benefit cardiovascular health. Guidelines recommend increasing omega‐3‐rich foods, and sometimes supplementation, but recent trials have not confirmed this.

Objectives

To assess the effects of increased intake of fish‐ and plant‐based omega‐3 fats for all‐cause mortality, cardiovascular events, adiposity and lipids.

Search methods

We searched CENTRAL, MEDLINE and Embase to February 2019, plus ClinicalTrials.gov and World Health Organization International Clinical Trials Registry to August 2019, with no language restrictions. We handsearched systematic review references and bibliographies and contacted trial authors.

Selection criteria

We included randomised controlled trials (RCTs) that lasted at least 12 months and compared supplementation or advice to increase LCn3 or ALA intake, or both, versus usual or lower intake.

Data collection and analysis

Two review authors independently assessed trials for inclusion, extracted data and assessed validity. We performed separate random‐effects meta‐analysis for ALA and LCn3 interventions, and assessed dose‐response relationships through meta‐regression.

Main results

We included 86 RCTs (162,796 participants) in this review update and found that 28 were at low summary risk of bias. Trials were of 12 to 88 months' duration and included adults at varying cardiovascular risk, mainly in high‐income countries. Most trials assessed LCn3 supplementation with capsules, but some used LCn3‐ or ALA‐rich or enriched foods or dietary advice compared to placebo or usual diet. LCn3 doses ranged from 0.5 g a day to more than 5 g a day (19 RCTs gave at least 3 g LCn3 daily).

Meta‐analysis and sensitivity analyses suggested little or no effect of increasing LCn3 on all‐cause mortality (risk ratio (RR) 0.97, 95% confidence interval (CI) 0.93 to 1.01; 143,693 participants; 11,297 deaths in 45 RCTs; high‐certainty evidence), cardiovascular mortality (RR 0.92, 95% CI 0.86 to 0.99; 117,837 participants; 5658 deaths in 29 RCTs; moderate‐certainty evidence), cardiovascular events (RR 0.96, 95% CI 0.92 to 1.01; 140,482 participants; 17,619 people experienced events in 43 RCTs; high‐certainty evidence), stroke (RR 1.02, 95% CI 0.94 to 1.12; 138,888 participants; 2850 strokes in 31 RCTs; moderate‐certainty evidence) or arrhythmia (RR 0.99, 95% CI 0.92 to 1.06; 77,990 participants; 4586 people experienced arrhythmia in 30 RCTs; low‐certainty evidence). Increasing LCn3 may slightly reduce coronary heart disease mortality (number needed to treat for an additional beneficial outcome (NNTB) 334, RR 0.90, 95% CI 0.81 to 1.00; 127,378 participants; 3598 coronary heart disease deaths in 24 RCTs, low‐certainty evidence) and coronary heart disease events (NNTB 167, RR 0.91, 95% CI 0.85 to 0.97; 134,116 participants; 8791 people experienced coronary heart disease events in 32 RCTs, low‐certainty evidence). Overall, effects did not differ by trial duration or LCn3 dose in pre‐planned subgrouping or meta‐regression. There is little evidence of effects of eating fish.

Increasing ALA intake probably makes little or no difference to all‐cause mortality (RR 1.01, 95% CI 0.84 to 1.20; 19,327 participants; 459 deaths in 5 RCTs, moderate‐certainty evidence),cardiovascular mortality (RR 0.96, 95% CI 0.74 to 1.25; 18,619 participants; 219 cardiovascular deaths in 4 RCTs; moderate‐certainty evidence), coronary heart disease mortality (RR 0.95, 95% CI 0.72 to 1.26; 18,353 participants; 193 coronary heart disease deaths in 3 RCTs; moderate‐certainty evidence) and coronary heart disease events (RR 1.00, 95% CI 0.82 to 1.22; 19,061 participants; 397 coronary heart disease events in 4 RCTs; low‐certainty evidence). However, increased ALA may slightly reduce risk of cardiovascular disease events (NNTB 500, RR 0.95, 95% CI 0.83 to 1.07; but RR 0.91, 95% CI 0.79 to 1.04 in RCTs at low summary risk of bias; 19,327 participants; 884 cardiovascular disease events in 5 RCTs; low‐certainty evidence), and probably slightly reduces risk of arrhythmia (NNTB 91, RR 0.73, 95% CI 0.55 to 0.97; 4912 participants; 173 events in 2 RCTs; moderate‐certainty evidence). Effects on stroke are unclear.

Increasing LCn3 and ALA had little or no effect on serious adverse events, adiposity, lipids and blood pressure, except increasing LCn3 reduced triglycerides by ˜15% in a dose‐dependent way (high‐certainty evidence).

Authors' conclusions

This is the most extensive systematic assessment of effects of omega‐3 fats on cardiovascular health to date. Moderate‐ and low‐certainty evidence suggests that increasing LCn3 slightly reduces risk of coronary heart disease mortality and events, and reduces serum triglycerides (evidence mainly from supplement trials). Increasing ALA slightly reduces risk of cardiovascular events and arrhythmia.

Plain language summary

Omega‐3 intake for cardiovascular disease

Review question

We reviewed randomised trials (where participants have an equal chance of being assigned to either treatment) examining effects of increasing fish‐ and plant‐based omega‐3 fats on heart and circulatory disease (called cardiovascular diseases, which include heart attacks and stroke), fatness and blood fats (lipids, including cholesterol, triglycerides, high‐density lipoprotein (HDL – 'good' cholesterol) and low‐density lipoprotein (LDL – 'bad' cholesterol)).

Background

The main types of omega‐3 fats are eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), both found in fish, and alpha‐linolenic acid (ALA) found in plant foods. Many people believe that taking omega‐3 supplements reduces risk of heart disease, stroke and death.

Trial characteristics

The evidence is current to February 2019. The review included 86 trials involving 162,796 people. These trials assessed effects of greater omega‐3 intake versus lower omega‐3 intake for at least a year on heart and circulatory disease. Twenty‐eight trials were very trustworthy (well‐designed so as not to give biased results). Participants were adults, some with existing illness and some healthy, living in North America, Europe, Australia and Asia. Most EPA and DHA trials provided capsules, few gave oily fish.

Key results

Increasing EPA and DHA has little or no effect on deaths and cardiovascular events (high‐certainty evidence) and probably makes little or no difference to cardiovascular death, stroke, or heart irregularities (moderate‐certainty evidence). However, increasing EPA and DHA may slightly reduce risk of coronary death and coronary events (low‐certainty evidence, coronary events are illnesses of arteries supplying the heart). To prevent one person having a coronary event, 167 people would need to increase their EPA and DHA, and 334 people would need to increase their EPA and DHA to prevent one person dying from coronary disease. EPA and DHA reduce triglycerides by about 15% but do not affect fatness or other lipids (high‐certainty evidence).

Eating more ALA (for example, by increasing walnuts or enriched margarine) probably makes little or no difference to all‐cause, cardiovascular or coronary deaths or coronary events but probably slightly reduces cardiovascular events and heart irregularities (moderate‐ or low‐certainty evidence). To prevent one person having a coronary event, 500 people would need to increase their ALA, 91 people to prevent one person having arrhythmia.

There is little evidence of effects of eating fish. EPA and DHA reduce triglycerides. EPA, DHA and ALA may be slightly protective of some heart and circulatory diseases.

Summary of findings

Background

Description of the condition

Cardiovascular diseases are disorders of the heart and blood vessels. They comprise cerebrovascular disease (including stroke and transient ischaemic attack), coronary heart disease (including heart attack or myocardial infarction and angina), peripheral arterial disease (diseases of the blood vessels to the arms and legs), deep vein thrombosis and pulmonary embolism (blood clots formed in the legs which can move to the heart and lungs), as well as rheumatic and congenital heart disease (WHO 2017); these last two are not discussed in this review. Globally, 31% of all deaths are due to cardiovascular disease, more than from any other cause (WHO 2017). Of the estimated 17.7 million people who died from cardiovascular diseases in 2015, around 7.4 million were due to coronary heart disease and 6.7 million due to stroke. Of 17 million premature deaths in 2015 caused by non‐communicable diseases, 82% were in low‐ and middle‐income countries, and 37% were caused by cardiovascular diseases (WHO 2017).

Description of the intervention

Omega‐3 fats (also called Ω3 or n‐3 fats) from fish sources include eicosapentaenoic acid (EPA, or 20:5), docosahexaenoic acid (DHA, 22:6) and docosapentaenoic acid (DPA, 22:5); these are long‐chain omega‐3 fats (LCn3). Alpha‐linolenic acid (ALA or α‐linolenic, 18:3) is the short‐chain omega‐3 fat found in plants and grass‐fed meat, which is partially converted to LCn3 fatty acids within our bodies. There is some debate about the effectiveness of this conversion, which may differ depending on whether it is assessed over the short or long term as well as on other dietary factors (Li 1999; Pawlosky 2001). For this reason the effectiveness of ALA may differ from that of the LCn3 fats.

Since Bang and colleagues first suggested that the abundance of omega‐3 fatty acids in the diet of the Greenland Inuit people was responsible for their low mortality from ischaemic heart disease (Bang 1972; Bang 1976), there has been considerable interest in the protective role and possible mechanism of action of marine unsaturated fats. This interest has spread to encompass plant seeds and oils rich in ALA, including chia seed, flax (linseed) and rapeseed (canola) oils (Nettleton 1991), their derivatives (e.g. margarines), purslane leaves (Simopoulos 1992), and nuts (especially walnuts).

How the intervention might work

Proposed mechanisms for the protective role of omega‐3 fats against cardiovascular diseases include: lowering the blood pressure; altering the lipid profile, especially reduced serum triglyceride concentration; modulating arterial lipoprotein lipase levels; reducing thrombotic tendency; producing anti‐inflammatory effects and anti‐arrhythmic effects; improving vascular endothelial function and insulin sensitivity; and increasing plaque stability and paraoxonase levels (Bhatnagar 2003; BNF 1999; Calabresi 2004; Chang 2013; Geelen 2004).

Given that most omega‐3 fats are ingested in the form of oily fish or fish oil (often fish liver) capsules, reports of high levels of various toxic compounds such as mercury, dioxins and polychlorinated biphenyls (PCBs) in oily fish and fish oils are concerning (Bourdon 2010; FSA 2000; Levine 2005; Liem 1997; MAFF 1998A; SACN COT 2004; USFDA 1995). These are all fat soluble and accumulate over time in the body, so harm may be exhibited only after long‐term fish consumption or supplementation with fish oils. Animal intervention trials and human cohorts who have suffered accidental exposure to dioxins and PCBs suggest that pre‐natal exposure may cause sub‐fertility problems, and adult exposures may lead to an excess of total cancers (JECFA 2001). Human cohorts exposed to high levels of mercury exhibit neurological problems (USFDA 1995). As many people eat oily fish once or twice a week or take fish oil supplements, it is important to explore the potentially harmful effects of fish‐associated omega‐3 intake. It is also possible that omega‐3 fats themselves may exhibit harm, for example through extension of bleeding times or suppression of normal immune responses (USFDA 2000).

Cardiovascular effects of eating more oily fish may differ from those of taking a fish oil supplement because fish (not fish oil) is a rich source of nutrients including selenium, iodine, zinc, calcium and protein. Fish in the diet may also displace a variety of other foods including sources of saturated or trans fats, so it could alter cardiovascular disease risk in other ways.

Why it is important to do this review

There is a great deal of public belief in the cardiovascular benefits of omega‐3 fats. Analysis of US National Health and Nutrition Examination Survey data from 2003 to 2008 suggests that in the USA, adults' mean LCn3 intakes were greater from dietary supplements (0.72 g/d EPA and DHA) than from foods (0.41 g/d, Papanikolaou 2014). But public health advice differs across countries. For example, the National Institute for Health and Clinical Excellence in the UK now encourages fish intake but discourages supplementation: "people with or at high risk of CVD [cardiovascular disease] should be advised to consume at least 2 portions of fish per week, including a portion of oily fish". However, it advises that omega‐3 fatty acid compounds "should not be offered for primary or secondary prevention of CVD [cardiovascular disease]" (NICE 2016). The American Heart Association (AHA) also "recommends eating fish (particularly fatty fish) at least two times (two servings) a week". Although the AHA suggests that omega‐3 intake via foods is preferable, it is more positive about omega‐3 supplements: "those with coronary artery disease may not get enough omega‐3 by diet alone. These people may want to talk to their doctor about supplements" (AHA 2016). These recommendations are balanced with a warning about potential excessive bleeding in those taking doses of more than 3 g/d omega‐3 fatty acids (presumably LCn3 fats). The AHA have issued updated guidelines on use of omega‐3 fats to treat raised triglycerides, suggesting that "prescription n‐3 FAs [fatty acids, meaning LCn3] (EPA+DHA or EPA‐only) at a dose of 4 g/d (>3 g/d total EPA+DHA) are an effective and safe option for reducing triglycerides as monotherapy or as an adjunct to other lipid‐lowering agents" (Skulas‐Ray 2019). Such recommendations, and resulting increased fish consumption, have potentially negative long‐term consequences for our marine ecosystems (Brunner 2009).

Epidemiological trials have supported the relationship between high omega‐3 intake and lower cardiovascular disease rates (Ballard‐Barbash 1987; Burr 1993; Kris‐Etherton 2002). However, these associations could be due to other characteristics of people who choose to eat fish. In many societies eating fish is associated with better social status and a health‐conscientious life view (Cade 2007), so eating fish is highly confounded by dietary quality, socioeconomic status and other markers of healthy lifestyles. As an example, the global attributable burden of eating a diet low in seafood omega‐3 fats was estimated as 1.1% of global disability‐adjusted life‐years (DALYs; 95% confidence interval (CI) 0.8 to 1.5), "with 22% of ischaemic heart disease DALYs attributable to low seafood intake" (Engell 2013). The data sources are not described, but when the estimate was derived from randomised controlled trials (RCTs) alone, rather than cohort trials and RCTs combined, the estimated global attributable burden was much smaller, 0.5% (95% CI −0.5 to 1.4). Information concerning cause and effect is more reliably supplied by intervention trials in which participants are randomly allocated to receive fish oil or advice to eat more fish.

Systematic reviews of RCTs have had various findings. A recent version of this review (Abdelhamid 2018a), included 79 long‐term RCTs and more than 112,000 participants, finding no effects of omega‐3 fats on all‐cause mortality or cardiovascular outcomes. Other systematic reviews have suggested a lack of effect for omega‐3 fats on all‐cause mortality or a variety of cardiovascular diseases (Campbell 2013; Chowdhury 2012; Khoueiry 2013; Kotwal 2012; Kwak 2012; Mariani 2013; Rizos 2012; Zheng 2014). However, some reviews have highlighted particular outcomes or circumstances in which cardiovascular disease prevention was evident: after heart surgery (He 2013), for preventing sudden cardiac death (Zhao 2009), for reducing cardiovascular disease mortality and sudden cardiac death, although with no effect on all‐cause mortality (Trikalinos 2012), for cardiovascular disease mortality (Sethi 2016), and for reducing the risk of stroke in women, albeit with no effect on stroke overall (Larsson 2012). Kwak 2012 reported marginal effects on cardiovascular death, though these were lost when a poor‐quality trial was removed, and a few others have reported only positive effects in their abstracts (reductions in cardiovascular events, cardiac death and coronary events) (Delgado‐Lista 2012). These disparate findings have fuelled both debate and confusion. A recent extensive Agency for Healthcare Research and Quality review meta‐analysed risk factors extensively but suggested there was only limited RCT data to assess the effects of omega‐3 fats on clinical cardiovascular disease outcomes (Balk 2016). The publication recently of a suite of large‐scale and long‐term trials of LCn3 (ASCEND 2018; REDUCE‐IT 2019; VITAL 2019), has prompted the need to update Abdelhamid 2018a.

This systematic review and meta‐analysis aimed to assess the evidence on the effects of omega‐3 fats (LCn3 and ALA separately) on all‐cause mortality and cardiovascular diseases. It also aimed to assess potentially harmful effects of omega‐3 fats or compounds associated with consuming LCn3 fats, such as excessive bleeding. A related review has formally systematically reviewed potential harms such as excessive cancers, rather than simply examining trials included in this review for cancer outcomes (Hanson 2019). We assessed mechanisms of action such as lipid and body weight changes and antiarrhythmic effects as primary or secondary outcomes in this review, and we have systematically reviewed these outcomes in a formal way by including trials that assessed adiposity, lipids and arrhythmic events, even where no cardiovascular disease events occurred or were reported. The World Health Organization (WHO) is currently updating its guidance on polyunsaturated fatty acid (PUFA) intake in adults and children. This is one of a set of systematic reviews commissioned by WHO in order to inform and contribute to the development of updated WHO recommendations. Sister systematic reviews have assessed effects of omega‐3, omega‐6 and total polyunsaturated fats on inflammation and inflammatory bowel disease (Thorpe 2017), diabetes and glucose metabolism (Brown 2019), depression and anxiety (Deane 2019), cognition and dementia (Brainard 2019), cancers (Hanson 2019) and functional status (Abdelhamid 2019). Separate reviews assess effects of omega‐6 fats and total polyunsaturated fat on mortality and cardiovascular outcome (Abdelhamid 2018b; Hooper 2018), and provide a detailed database of the relevant trials for use by others (Hooper 2019).

The results of this review including GRADE assessments were discussed and reviewed by the WHO Nutrition Guidance Expert Advisory Group (NUGAG) Subgroup on Diet and Health as part of WHO’s guideline development process.

Objectives

To assess the effects of increased intake of fish‐ and plant‐based omega‐3 for all‐cause mortality, cardiovascular events, adiposity and lipids.

The primary review question was, 'Do long‐chain omega‐3 fats (LCn3, fish‐based omega‐3 fats) or ALA (plant‐based omega‐3 fats) alter risk of all‐cause mortality, cardiovascular deaths, cardiovascular events, coronary heart disease deaths, coronary heart disease events, stroke, arrhythmia, adiposity and lipids?'

Secondary questions include the following.

  • If omega‐3 fatty acids confer protection:

    • does protection occur equally in those at low and at high risk of cardiovascular disease?

    • does protection depend on the dose of omega‐3 fats taken per day?

    • do effects differ between dietary and supplemental omega‐3 sources?

    • does protection depend on trial summary risk of bias?

  • Is protection or harm stronger with longer trial duration?

  • Are effects of omega‐3 fatty acids dependent on baseline triglyceride levels or diabetic status?

The latter was suggested by WHO NUGAG and added post‐hoc specifically for this update.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled clinical trials that included diet advice or dietary supplementation to promote omega‐3 fatty acid intake versus placebo, no supplementation, usual diet or lower‐dose omega‐3. One of our outcomes had to be measured and available (through publications or contact with trial authors), and trials had to follow participants for at least 12 months (52 weeks or 360 days). For advice trials, follow‐up must have been at least 12 months following advice, and for trials where participants received food or supplementation, provision must have continued for at least 12 months. We accepted randomisation of individuals or of clusters as long as there were at least six clusters randomised.

Careful work by Browning suggests that supplements of EPA and DHA equivalent to one weekly portion of oily fish results in 95% of maximal incorporation by 5 days for EPA in plasma phosphatidylcholine (95% CI 0 to 18 days) to 273 days for DHA into blood mononuclear cells (95% CI 0 to 670 days; FISH 2012). While this suggests individual variability, on average all compartments except blood mononuclear cells had equilibrated by 117 days (both EPA and DHA into plasma phosphatidylcholine, plasma cholesteryl esters, plasma nonesterified fatty acids, plasma triglycerides, erythrocytes and platelets). The authors stated, "EPA and DHA reached a maximum in platelets in 3–4 weeks and 1–2 months, respectively, and in blood mononuclear cells in 6–9 months". For this reason we chose 12 months as the minimum duration of intervention, as it allows equilibration of most body compartments with EPA and DHA as well as time for this change in body composition to have some effect on cardiovascular risk or mortality.

In previous reviews of dietary effects on cardiovascular outcomes, we limited trials to at least two years' duration (Hooper 2015), as the proposed mechanism of effects was via LDL cholesterol, atherosclerosis and its sequelae, and this takes time to develop. The 4S trial showed separation of the survival curves at around two years (Scandinavian Simvastatin Survival Study Group 1994). Potential mechanisms for effects of PUFAs are broader, including what could be rapid effects on arrhythmias or inflammation, so we decided to include trials of at least 12 months to ensure we did not miss these effects.

Types of participants

Trials in adults (18 years or older, men and women) at any risk of cardiovascular disease (with or without existing cardiovascular disease) were eligible, including those in participants with increased risk of cancer, those undergoing or who have undergone coronary artery bypass grafting or angioplasty, and those with current or previous cardiovascular disease, nephritis in systemic lupus erythematosus, breast cysts, diabetes mellitus, rheumatoid arthritis, multiple sclerosis, psoriasis, hay fever, asthma or ulcerative colitis. Including these populations allows us to understand both development and progression of cardiovascular disease (primary and secondary prevention). We excluded participants who were pregnant or acutely ill (with acute‐stage cancer, undergoing heart or renal transplantation, with HIV or AIDS, on haemodialysis, with IgA glomerulonephritis, or any other renal problem except in diabetes).

Types of interventions

The intervention must have been dietary supplementation, a provided diet or advice on diet. The foodstuffs or supplements must have been: oily fish (including mackerel, dogfish, salmon, herring, trout, tuna, sturgeon, stablefish, anchovy, sprat, coho, capelin, sardines, swordfish, sild, pilchard, brisling, menhaden, bloater, whitebait, crab and conger eel); fish oils (made from any of the above or a mixture of fish, or cod liver oil); linseed (flax), canola (rapeseed), perilla, purslane, mustard seed, candlenut, stillingia or walnut as a food, capsule, oil, made into a spreading fat or supplementing another food (such as bread or eggs). For ALA sources the product consumed had to have an omega‐3 fat content of at least 10% of the total fat content. Refined eicosapentaenoic acid (EPA), docosahexaenoic acid (DHA) or ALAs, or concentrated fish or algal oils, were also accepted. Supplementation may have been in oil or capsule form or as foodstuffs provided to be consumed by mouth (excluding enteral and parenteral feeds and enemas).

We excluded trials using multiple risk factor interventions on lifestyle factors (such as weight reduction, smoking or physical activity goals), or differential dietary interventions not involving dietary fats, except where that other intervention was a direct replacement for polyunsaturated fats or the effect of diet or supplementation could be separated out from the other interventions.

Trials were eligible if they compared the effect of dietary advice or supplementation to increase omega‐3 fats with the usual diet, no advice, no supplementation, placebo or lower‐dose omega‐3.

Types of outcome measures

Primary outcomes

Primary outcomes included numbers of participants experiencing:

  • all‐cause mortality (deaths);

  • cardiovascular mortality (cardiovascular deaths);

  • cardiovascular events (cardiovascular events);

  • coronary heart disease mortality (coronary heart disease deaths);

  • coronary heart disease events;

  • stroke; and

  • arrhythmia (atrial fibrillation).

We analysed coronary heart disease using the first of the following to be reported: number of participants experiencing coronary heart disease or coronary events, total myocardial infarction, acute coronary syndrome or angina (stable and unstable). This meant that if trial authors reported coronary heart disease events, we used these in analysis; where trials did not report coronary heart disease events but did report total myocardial infarction, we used that (and so on). Combined cardiovascular events included fatal and non‐fatal myocardial infarction, angina, stroke, heart failure, peripheral arterial disease, sudden death and non‐scheduled cardiovascular interventions – coronary artery bypass surgery or angioplasty. We included all available outcomes where we could be sure that the same participant was not being counted twice.

At the request of WHO NUGAG Subgroup on Diet and Health, we added coronary heart disease mortality post hoc as a primary outcome. Data used were the first of the following list reported: coronary death, ischaemic heart disease death, fatal myocardial infarction, cardiac death. We only used cardiac death when no other outcomes in this category were available, and we ran a sensitivity analysis omitting cardiac death. The reason for excluding cardiac death in sensitivity analysis was that it goes slightly outside our area of interest, including other causes of death in addition to coronary heart disease, such as cardiomyopathies and congenital and valvular heart diseases. We wanted to include cardiac death in the main analysis as we felt that otherwise we would be missing some important cases of coronary heart mortality, but we decided to exclude it in sensitivity analysis as we were potentially including a few outcomes that coronary heart disease mortality did not encompass.

Secondary outcomes

Secondary outcomes included:

  • major adverse cerebrovascular or cardiovascular events (MACCEs) or individual cardiovascular events (total, fatal or non‐fatal myocardial infarction, sudden cardiac death, angina, heart failure, revascularisation, peripheral arterial disease or acute coronary syndrome);

  • body weight and other measures of adiposity; and

  • lipids (total, LDL or HDL cholesterol and triglycerides).

We defined MACCEs as participants experiencing myocardial infarction, unstable angina, stroke or death. We did not consider trials that did not provide data on all these health events for this outcome.

The review included trials if any of their participants experienced or were assessed for any primary or secondary outcome. These could have been reported in publications (as outcomes or reasons for dropout or adverse events), supplied by trial authors, or which clearly happened even if exact numbers were not available. However, as almost all trials note if a death or cardiovascular event occurs in a trial participant (so all trials assessed for our primary outcomes) we only included trials where at least one event occurred, or where a continuous outcome was measured.

Tertiary outcomes

We extracted the following outcomes where available within included trials.

  • Blood pressure

  • Serious adverse events (any other reported illnesses)

  • Side effects

  • Dropouts

  • Quality‐of‐life measures

  • Economic costs

We originally intended to assess type 2 diabetes diagnoses, measures of glucose metabolism, cancers, breast cancer, neurocognitive outcomes such as dementia, depression and anxiety within included trials. However, as part of the larger set of reviews we formally systematically reviewed effects of omega‐3 fats on type 2 diabetes diagnoses and measures of glucose metabolism (Brown 2019), cancers including breast cancer (Hanson 2019), neurocognitive outcomes such as dementia (Brainard 2019), irritable bowel disease (IBD) and inflammatory factors (Thorpe 2017), depression and anxiety (Deane 2019), and functional outcomes (Abdelhamid 2019), so a partial assessment within this review would be unhelpful and potentially misleading. For this reason we exclude these specific outcomes from our reporting of serious adverse events.

Key outcomes

When the World Health Organization (WHO) NUGAG Subgroup on Diet and Health requested this review update they named the following as key outcomes to inform their planned dietary guidance.

  • All‐cause mortality

  • Cardiovascular disease mortality

  • Cardiovascular disease events

  • Coronary heart disease mortality

  • Coronary heart disease events

  • Stroke

  • Arrhythmia (atrial fibrillation)

  • Serum lipids including total cholesterol, fasting triglycerides, high‐density lipoprotein (HDL) and low‐density lipoprotein (LDL)

  • Measures of adiposity (body weight and body mass index (BMI)

We were not able to make all of these outcomes into primary outcomes (as the number of primary outcomes are restricted for Cochrane Reviews). However, because WHO NUGAG Subgroup on Diet and Health will use these outcomes to underpin guidance, we carried out sensitivity analyses, subgroup analyses and GRADE assessment of certainty of evidence for them, even when they were not primary outcomes.

Search methods for identification of studies

Electronic searches

We searched the following electronic databases on 13 February 2019 to identify reports of relevant randomised clinical trials:

  • Cochrane Central Register of Controlled Trials (CENTRAL; 2019, Issue 2) in the Cochrane Library;

  • Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, MEDLINE Daily and MEDLINE (Ovid, 1946 to 12 February 2019);

  • Embase Classic and Embase (Ovid, 1947 to 2019 week 6).

We applied date limits to the terms from the original strategies so that the search included only new records (Appendix 1). The RCT filter for MEDLINE was the Cochrane sensitivity and precision‐maximising RCT filter, and for Embase, we applied the terms as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Lefebvre 2011).

Appendix 2 shows the MEDLINE search strategy for the original version of this review (Hooper 2004), Appendix 3 and Appendix 4 show searches used to update the previous version of this review (Abdelhamid 2018a).

For the previous (27 April 2017) update we also ran searches for a new systematic review of the effects of polyunsaturated fats on cardiovascular disease (Abdelhamid 2018b), as well as updating and extending a Cochrane Review of the effects of omega‐6 polyunsaturated fats on health outcomes (Hooper 2018). We ran searches for these reviews using the same RCT filters (Appendix 4). The results of these searches were de‐duplicated against the omega‐3 searches, and all the titles and abstracts assessed as a single set for all three reviews. We created a data set of RCTs that lasted at least six months and compared higher versus lower omega‐6, omega‐3 or total PUFA in adults. We used this data set as the wider study pool from which we selected included trials for all reviews (Abdelhamid 2018a; Abdelhamid 2018b; Abdelhamid 2019; Brainard 2019; Brown 2019; Deane 2019; Hanson 2019; Hooper 2018; Hooper 2019; Thorpe 2017). We did not repeat these additional searches in 2019.

We also searched two trials registers, ClinicalTrials.gov (clinicaltrials.gov), and the WHO International Clinical Trials Registry Platform (ICTRP, www.who.int/ictrp/en), to 31 July 2019 for registry entries for relevant completed and ongoing trials.

Searching other resources

We assessed titles and abstracts retrieved during these electronic searches for relevant RCTs and also relevant systematic reviews. We handsearched the included trials in all relevant systematic reviews up to April 2017 for new trials and additional publications of included trials. We contacted authors of all large included trials (at least 100 participants) included up to April 2017 and some smaller trials for further trial data, methodological details and references to trials not yet identified, including published, unpublished or ongoing trials.

For all included trials we carefully searched and data‐extracted trials registry entries, protocols, supplementary materials, letters, conference abstracts and additional publications to help us locate complete data sets.

Data collection and analysis

Selection of studies

At least two review authors independently assessed titles and abstracts resulting from the electronic and bibliographic searches. We rejected titles and abstracts on initial screen only if the review author could determine from the title and abstract that the article was not a report of a RCT; did not address omega‐3 intake (or total polyunsaturated fat or omega‐6 fat for the other two reviews); were exclusively in children or young adults (less than 18 years old), pregnant women or the critically ill; were of less than 12 months' duration; or if the intervention was multi‐factorial and we could not separate out the effect of dietary fat.

We rejected trials only when it was certain that no primary or secondary outcome events occurred, and none of the secondary outcome risk factors were measured. When we could not reject a title or abstract with certainty, we obtained the full text of the article for further evaluation. We made attempts to obtain full‐text translations or evaluations, or both, of all potentially relevant non‐English articles.

We used an in/out form to assess full‐text papers and trials for inclusion (or otherwise) into the review. We contacted the authors of all potentially included RCTs for further information on trial methodology and outcomes. Two review authors independently decided on inclusion of full‐text RCTs, resolving any differences by discussion and, when necessary, in consultation with the review author team.

Data extraction and management

We designed a data extraction form for this review, which each of the review authors tested on a common 'training' trial (SCIMO 1999), and we adapted the form as needed. We extracted data concerning participants, interventions, and outcomes, as described above in the selection criteria section. We extracted dichotomous data from dietary advice trials at the latest point available in the trial (regardless of the amount of reinforcement of the original dietary message), while for supplement trials, we extracted dichotomous data to the point that supplementation or the trial ended, whichever was earlier. We extracted continuous data at the nearest time point to 12 months and also the latest point available in fixed‐term trials, but in trials where participants were followed up for varying durations (aside from dropouts), we extracted the participants' data from the first time point following the mean trial duration. We never used data from periods following the end of a trial in meta‐analysis.

We also extracted data on risk of bias, assessed using the Cochrane 'Risk of bias' tool, along with data on potential effect modifiers, including existing cardiovascular disease (primary or secondary prevention), trial duration, intensity of intervention (dietary advice, diet provided, supplemental foods, supplements (capsules) and any combination), LCn3 fats or ALA and dose, replacement, medications used (including statins, antihypertensive, antiarrhythmic or antithrombotic medication), fatty acid data (from plasma, platelets or adipose tissue) and smoking status.

For primary and secondary dichotomous outcomes, we extracted numbers of participants experiencing an outcome and total numbers of participants randomised (or in whom the outcome was assessed where known) for each trial arm. For continuous outcomes, we extracted the number of participants assessed, means and standard deviations of the final readings in each treatment arm; we calculated standard deviations from other variance data where appropriate. Where data were available on both change and final readings, we used data on change.

Two review authors independently extracted original reports of trial results. We resolved differences between review authors' results by discussion and, when necessary, in consultation with a third review author or the review author team.

Assessment of risk of bias in included studies

Two review authors independently assessed risk of bias for each included trial, using Cochrane criteria (Higgins 2017), including in the domains of sequence generation; allocation concealment; blinding of participants and personnel, blinding of outcome assessors; incomplete outcome data; and selective outcome reporting. Additional review‐specific criteria included similarity of type and intensity of intervention in both arms (attention) and evidence of appropriate moderate to high compliance (to establish that the intervention group were receiving a different intake of omega‐3 fats than the control group). Table 4 presents specific details of how we interpreted these criteria for this review.

1. Risk of bias assessment methods in greater detail.
Risk of bias element Criteria for low risk of bias Criteria for unclear Criteria for high risk of bias
Selection bias: random sequence generation The trial authors needed to have described the method used to generate the allocation sequence in sufficient detail to allow an assessment of whether it should produce comparable groups. For example "the randomisation sequence was computer generated". We allowed that a good method of randomisation was strongly implied if the authors discussed stratification and/or blocking. Therefore, if the authors were not explicit about their randomisation method but did describe stratification or blocking we assessed this as corresponding to low risk. The trial authors have not described their method in sufficient detail for the assessment of whether it would produce comparable groups. For example, the authors state "the trial was randomised" and provide no further information. The randomisation method was assessed as not truly random, and may not produce comparable groups.
Selection bias: allocation concealment The trial authors needed to have described the method used to conceal allocation sequence in sufficient detail to determine whether the allocations could have been foreseen in advance of, or during, enrolment. Good methods included putting allocation codes in opaque sealed envelopes (ideally prepared by someone outside the treatment or assessment teams and sequentially numbered), using a telephone allocation system after the participants had consented to participate or providing a random number that links to a specific set of capsules prepared and distributed centrally or by an arms‐length pharmacist. The trial authors gave insufficient detail as to method. The allocation was known in advance of participants consenting to take part in the trial.
Performance bias: blinding of participants and personnel The trial authors needed to have described all measures used, if any, to blind trial participants and personnel from knowledge of which intervention a participant received. Ideally, they should also have provided information relating to whether the intended blinding was effective. For example, the authors could say "both the intervention and placebo capsules looked and tasted the same." However, if the trial authors did not provide information on whether the blinding was effective, but sufficient detail was given on a good method of blinding, then we assumed that the blinding was effective and the risk of bias was low. Insufficient methodological details were provided e.g. "the trial was blinded." The trial was unblinded or where blinding was broken, e.g. "the capsules were visually identical but the participants reported a strong fishy flavour in the intervention group only."
Detection bias: blinding of outcome assessment Trial authors needed to have described measures used, if any, to blind outcome assessors from knowledge of which intervention a participant received. Ideally, they should also have provided information relating to whether the intended blinding was effective. For example, the authors could say "the outcome assessors had no knowledge of the group allocation, and both the intervention and placebo capsules looked and tasted the same so the self‐assessment scales were also blinded." However if the trial authors did not provide information on whether the blinding was effective, but sufficient detail was given on a good method of blinding of the assessors, then we assumed that the blinding was effective and the risk of bias is low. All biochemical assessment (lipids, glucose, CRP, insulin, PSA, etc.) were considered at low risk of detection bias if outcome assessor blinding or double‐blinding was stated. Insufficient methodological details were provided e.g. "the trial was blinded." The trial was unblinded or blinding was broken, e.g. for a self‐assessment measure "the capsules were visually identical but the participants reported a strong fishy flavour in the intervention group only."
Because the level of blinding could vary by outcome, assessment of risk of bias was based on blinding of the review's primary outcome(s). Where primary outcomes had different assessments we opted for the higher risk of bias but noted that risk of bias was lower for other outcomes.
Attrition bias: incomplete outcome data The trial authors needed to describe the completeness of outcome data for each main outcome, including attrition and exclusions from the analysis. They needed to report the number of attrition/exclusions, the numbers in each group at each time point, reasons for attrition/exclusion and any re‐inclusions in analyses. Ideally, they would report how they imputed any missing data e.g. last observation carried forward. There needed to be a reasonable balance of attrition/exclusions between trial arms and ≤ 20% of the sample should be lost over a year. The authors didn't state reasons for attrition/exclusion, or were unclear about the numbers lost to attrition/exclusion in each trial arm. The authors demonstrated a substantial difference in the rates of attrition/exclusions between the trial arms and/or > 20% of the baseline sample was lost over a year (> 10% over 6 months).
Reporting bias: selective outcome reporting The trial authors needed to have published their trial protocol or trials registry entry before the end of the trial's recruitment period, i.e. prospectively. They needed to have reported on all of the primary and secondary outcomes listed in the protocol/registry entry. We deemed reporting additional secondary outcomes in the results paper(s), although not ideal, to still be low risk. No trial protocol or trials registry entry was found, it was registered retrospectively, or the dates of registration and participant recruitment were unclear. The trial authors did not report at least one primary or secondary outcome listed in the protocol/registry entry or the results paper(s) reported a primary outcome that was not listed at all in the protocol or not listed as a primary outcome in the protocol.
Other sources of bias: attention bias The trial authors needed to have reported that participants in all trial arms received the same amount of attention and time from researchers and clinical teams. For example, "All participants attended the clinic for a baseline assessment which took 2 hours. They were then followed with monthly telephone calls, and finally attended for a 6 month assessment at the clinic which took 1 hour." If the trial only differed by the content of the capsules, and the assessment schedule was not stated to differ between the two arms, we assumed it to be at low risk. The authors did not state the attention each arm received. Participants in different arms received different amounts of attention. For example "the intervention group only attended for additional assessments at months 2, 4, and 6" or "the rates of relapse differed substantially between the groups which led to differing amounts of treatment time and attention," or "the intervention group received a 40 minute dietary education session."
Other sources of bias: limited compliance The trial authors needed to have reported on the level of compliance in all arms in sufficient detail to determine whether the trial results were robust. We followed a flow chart to make this determination. A statistically significant difference between the intervention and control groups in a body measure of at least 50% of the test fatty acids. Where no body measures were reported then estimated compliance needed to be greater than 64% (proportion complying multiplied by compliance threshold). Compliance not reported or not in a way that could be interpreted. Measures of compliance were reported but fell below the appropriate thresholds.
Other sources of bias: other In the absence of any additional issues this item was coded "low risk of bias" If fraud concerns had been raised and the paper had been withdrawn, or the author had been found guilty of fraud by a legal or medical entity the paper was excluded from the review. However if fraud concerns were raised, but the journal had not withdrawn the paper, and the author had not been formally sanctioned; then the trial was included in the review, but concerns were raised here, and the risk of bias for this item was high.
CRP: C‐reactive protein; PSA: prostate specific antigen

In brief, we considered a trial to be at low risk of attention bias when participants were given the same amount of time and attention from trial staff and health professionals whether they were in the intervention or control arms, and at low risk of compliance bias when adherence was assessed, results of that assessment were clearly reported for both intervention and control arms, and where most participants appeared to have taken at least 75% of the intended PUFA dose.

Summary risk of bias

Schulz 1995 found that poorly concealed allocation was associated with a 40% greater effect size, so randomisation and allocation concealment are core issues for all trials. Lack of blinding is associated with bias, though smaller levels of bias than lack of allocation concealment (Savovic 2012), especially in trials with objectively measured outcomes (Wood 2008), such as those we primarily used in our review. Although we originally planned to assess summary risk of bias for all included trials in the same way across the whole set of reviews (Abdelhamid 2018a; Abdelhamid 2018b; Abdelhamid 2019; Brainard 2019; Brown 2019; Deane 2019; Hanson 2019; Hooper 2018; Hooper 2019; Thorpe 2017), we adopted a different approach after discussing the different nature of supplement trials compared to dietary advice or food provision trials with the WHO NUGAG Subgroup on Diet and Health.

We considered supplement or capsule type trials to be at low summary risk of bias where we judged randomisation, allocation concealment, blinding of participants and personnel, and blinding of outcome assessors to be adequate. We considered all other trials to be at moderate or high risk of bias (a single category).

We considered dietary advice or all‐food‐provided type trials to be at low summary risk of bias where we judged randomisation, allocation concealment, and blinding of outcome assessors to be adequate. We considered all other trials to be at moderate or high risk of bias (a single category).

Measures of treatment effect

We pooled dichotomous data using risk ratios (RR) to describe effect sizes and continuous data using mean differences (MD). Where effects were described by different but comparable measures or scales in different trials, we planned to combine them using standardised mean difference (SMD), but this was not needed in this review.

Unit of analysis issues

We considered that we could reduce patient numbers in cluster‐randomised trials to an effective sample size, as described by Hauck 1991, however, we identified no such trials. For combined outcomes (e.g. combined cardiovascular events), we made attempts to add numbers of individuals experiencing specific outcomes within trials, but only where we could be certain that we were not counting individual participants more than once within any one of our review outcome categories.

For trials with intervention arms providing different omega‐3 doses, we combined data for the intervention groups for binary outcomes and used higher‐dose data versus control for continuous outcomes. We used arms with different doses separately when subgrouping by dose. Where factorial trials ran more than one intervention included in this review (AlphaOmega ‐ ALA 2010; AlphaOmega ‐ EPA+DHA 2010), we did not pool both comparisons in the same meta‐analysis.

Dealing with missing data

We sought trials registry entries and trial protocols to help us assess which trials measured each outcome. Where trials appeared to have collected – but did not report – data, we wrote to trial authors to ask for information. We wrote to authors of all trials that randomised at least 100 participants as well as to those of many smaller trials (although not to all due to limited resources), prioritising our efforts on larger trials that would tend to provide more information to the review. For larger trials where we found no trials registry entry or protocol, we wrote to trial authors to ask whether they had collected information on any outcomes of interest that we had not yet located. Where it was clear that data existed but could not be located to use within the review, we noted this and assessed the potential effect of this missing data on effect sizes narratively.

A recent meta‐analysis of effects of omega‐3 fats on cardiovascular outcomes, Aung 2018, worked with 10 large RCTs in depth to formulate their outcome data to match precisely between trials. It is difficult as a reviewer to access complex outcomes (such as 'cardiovascular events') where we need to count people experiencing events, rather than counting events ‐ events are additive, but people experiencing events are not (a participant experiencing a stroke, non‐fatal myocardial infarction and angina must be counted as only one person experiencing cardiovascular events, rather than three events). As the review had worked with trial authors to optimise their data, we used data from this review to both fill in missing data (some data had not been previously published) and also to update our data where the data in Aung 2018 differed from previously reported data for any outcome. The trials that this applies to, those included in Aung 2018, were AlphaOmega ‐ ALA 2010; AlphaOmega ‐ EPA+DHA 2010; AREDS2 2014; DO IT 2010; GISSI‐HF 2008; GISSI‐P 1999; JELIS 2007; OMEGA 2009; ORIGIN 2012; Risk & Prevention 2013; SU.FOL.OM3 2010.

Assessment of heterogeneity

We assessed heterogeneity using the I2 test and assumed it to be important when I2 test value was more than 60% (Higgins 2003).

Assessment of reporting biases

We used funnel plots to assess for evidence of bias for primary outcomes where at least 10 trials contributed to the meta‐analysis (Egger 1997). We also compared effects from random‐effects and fixed‐effect meta‐analyses to understand whether small study effects may be important (Page 2019), alongside reporting known missing data (above).

Data synthesis

Primary measures of interest were effects of dietary advice or supplementation of fish‐based (LCn3) fats, and ALAs, on primary outcomes. We separated out effects of LCn3 and ALA in all analyses and thus present two separate sets of results: one for LCn3 and one for ALA.

We combined treatment/control differences in the outcomes across trials using risk ratios (RR) or mean differences (MD) in random‐effects meta‐analysis. For combined outcomes (e.g. combined cardiovascular events), we made attempts to add numbers of individuals experiencing specific outcomes within trials, but only where we were certain that we were not counting individual participants more than once within any one of our review outcome categories. However, individuals may have been counted for more than one of the review outcomes (in separate forest plots).

Subgroup analysis and investigation of heterogeneity

We explored the effects of LCn3 and ALA separately on all primary review outcomes and also on key review outcomes where these were secondary outcomes in our review and included at least six trials by subgrouping. The planned subgroup analyses were as follows.

  • Type of intervention: dietary advice, supplemental foods (for example margarine fortified with rapeseed, tins of sardines or oils to use in cooking) provided by the trial, supplements (capsules or oils) provided to take as medicine or any combination

  • Replacement of saturated fatty acids, mono‐unsaturated fatty acids (MUFA), omega‐6 fats, fat mixture, carbohydrates or sugars, non‐fat or no placebo, or unclear, with LCn3 or ALA

  • Primary prevention versus secondary prevention of cardiovascular disease (where trials that recruited participants for cardiovascular disease at baseline were considered secondary prevention trials, and trials that did not recruit on the basis of cardiovascular disease, so may include some or no people with existing cardiovascular disease, were considered primary prevention trials)

  • LCn3 dose: at least 150 mg/d, 250 mg/d, 400 mg/d from all sources including supplements (above or below each threshold); low dose 0.4 g/d to 2.4 g/d, medium dose 2.5 g/d to 4.4 g/d, and high dose ≥ 4.5 g/d of combined LCn3 fats;

  • ALA dose: higher versus lower levels of intake (≥ 5 g/d versus < 5 g/d);

  • Trial duration: trials with medium follow‐up (12 to 23 months), medium follow‐up (24 to 47 months) and long follow‐up (≥ 48 months)

  • Statin use (< 50% of control group on statins, ≥ 50% of control group on statins, use of statins unclear)

  • Baseline LCn3 intake, and baseline ALA intake

There were insufficient data on baseline omega‐3 or ALA intake (or intake in control groups, which could have been used as a proxy) to subgroup by baseline omega‐3 or ALA intake.

Post‐hoc, WHO NUGAG asked us to assess whether effects of omega‐3 fats differed by baseline triglyceride or diabetes status. We carried out post‐hoc subgroups on primary outcomes whenever the other subgroups were created, including the following.

  • Triglyceride status: raised triglycerides at baseline (inclusion criteria relate to triglycerides and mean triglycerides in control group at baseline were > 200 mg/dL or > 2.26 mmol/L) versus normal triglycerides (triglycerides did not relate to inclusion criteria, or mean triglycerides in control group at baseline was < 200 mg/dL or < 2.26 mmol/L)

  • Diabetes status: diabetic at baseline (at least half of participants had diabetes) versus diabetes risk factors at baseline (insulin resistance, non‐alcoholic fatty liver disease, non‐alcoholic steatohepatitis or obesity) versus inclusion criteria that did not relate to diabetic status.

Meta‐regression

We used meta‐regression to further explore effects of LCn3 dose, ALA dose, omega‐6 dose, total PUFA dose (looking for evidence of dose response for each), trial duration, primary or secondary prevention, food or capsule intervention (food included dietary advice and supplemental foods), and risk of bias (summary risk of bias low or moderate to high) on primary outcomes. We performed random‐effects meta‐regression using the STATA 16 command meta regress (Berkley 1995; Sharp 1998): log(e) risk ratio versus (dose or primary/secondary prevention or type of intervention or risk of bias or duration), weighted by the standard error of the log(e) risk ratio. Where there were no events in one arm, we added 0.1 to the numbers for both groups (so a trial with 10 people experiencing stroke in one arm but none in the other arm would be entered as 10.1 and 0.1). We analysed all included trials (of at least 12 months' duration) that reported each outcome from this review and its sister reviews (omega‐3 trials from this review, omega‐6 trials from Hooper 2018, and total PUFA trials from Abdelhamid 2018b). We carried out meta‐regression of each variable singly, then a multivariate meta‐regression of the three variables with lowest P values in single regression for each outcome. Given that we generally included data from fewer than 40 trials and there were some missing data for some trials, we did not run meta‐regressions with more than three variables at one time.

Sensitivity analysis

We carried out sensitivity analyses on all primary outcomes (regardless of the number of included trials) and on key outcomes that were secondary outcomes in this review.

We used sensitivity analyses to assess robustness of results to:

  • trial quality (removing trials at moderate or high summary risk of bias);

  • trial size (retaining only trials that randomised at least 100 participants across all trial arms);

  • fixed‐effect analysis; and

  • compliance (retaining only trials where we assessed compliance as conferring low risk of bias).

We tabulated the type and frequency of side effects and adverse effects (with the other extracted data on adverse effects) and compared between different trials and designs.

'Summary of findings' tables

We interpreted outcome data as follows.

  1. Is there an effect? (options were ‘increased risk’, ‘decreased risk’, or ‘little or no effect’). Our main outcome measure was RR so we decided on existence of an effect using RR. RR more than 8% (RR < 0.92 or > 1.08) for the highest‐certainty evidence suggested increased or decreased risk (otherwise little or no effect). The presence or not of an effect was decided on the RR for the main analysis and sensitivity analyses, the highest‐certainty evidence (the main analysis, the sensitivity analyses of trials at low summary risk of bias and at low risk of compliance problems).

  2. For continuous outcomes, we considered increasing ALA or LCn3 to have little or no effect unless effect sizes represented at least 5% change from baseline (or 2% in the case of cumulative outcomes such as adiposity).

  3. We assessed certainty of evidence using GRADE assessment (GRADE Working Group 2004), for key outcomes. We used the five GRADE considerations (risk of bias, consistency of effect, imprecision, indirectness and publication bias) to assess the certainty of the body of evidence as it related to the trials that contributed data to the meta‐analyses for the prespecified outcomes. We used methods and recommendations described in Section 8.5 (Higgins 2017), and Chapter 11 (Schünemann 2017), of the Cochrane Handbook for Systematic Reviews of Interventions, plus GRADEpro GDT software (GRADEpro GDT). We justified all decisions to downgrade the certainty of trials using footnotes and made comments to aid readers' understanding of the review.

  4. Where there was a suggested effect, we assessed the size of effect using the number needed to treat for an additional beneficial outcome (NNTB), number needed to treat for an additional harmful outcome (NNTH) or absolute risk reduction (ARR).

We included three 'Summary of findings' tables: for effects of LCn3 on primary outcomes (Table 1), effects of ALA on primary outcomes (Table 2), and for key outcomes that were not included in the review primary outcomes (measures of adiposity and serum lipids, Table 3).

Summary of findings for the main comparison. High versus low long‐chain omega‐3 fats for preventing cardiovascular disease and mortality (primary outcomes).
High versus low long‐chain omega‐3 fats for preventing cardiovascular disease and mortality (primary outcomes)
Patient or population: adults with or without existing CVD
 Setting: participants were living at home for most or all of the duration of their trials. Most trials were carried out in high‐income economies (World Bank 2018), but four were carried out in upper‐middle‐income countries (Argentina, Iran, Turkey and China). No trials took place wholly in low‐ or low‐middle income countries.
 Intervention: higher intake of LCn3 fats
 Comparison: lower intake of LCn3 fats
The intervention was dietary supplementation, a provided diet or advice on diet. Supplementation may have been in oil or capsule form or as foodstuffs provided, to be consumed by mouth (excluding enteral and parenteral feeds and enemas). The foodstuffs or supplements must have been oily fish or fish oils as a food, oil, made into a spreading fat, or supplementing another food (such as bread or eggs). Refined eicosapentaenoic acid (EPA), docosahexaenoic acid (DHA) or concentrated fish or algal oils, were also accepted.
Outcomes Anticipated absolute effects* (95% CI) Relative effect
 (95% CI) № of participants
 (trials) Certainty of the evidence
 (GRADE) Comments
Risk with lower LCn3 Risk with higher LCn3
All‐cause mortality – deaths
 Assessed with number of participants dying of any cause, whether reported as an outcome or a reason for dropout
Duration: range 12‐88 months
80 per 1000 78 per 1000
 (74 to 81) RR 0.97
 (0.93 to 1.01) 143,693
(45 RCTs)
⊕⊕⊕⊕
 Higha LCn3 fat intake makes little or no difference to risk of all‐cause mortality
Cardiovascular mortality – cardiovascular deaths
 Assessed with deaths from any cardiovascular cause. Where this was not available, we used cardiac death instead where known
Duration: range 12‐88 months
50 per 1000 46 per 1000
 (43 to 49) RR 0.92
 (0.86 to 0.99) 117,837
 (29 RCTs) ⊕⊕⊕⊝
 Moderateb LCn3 fat intake probably makes little or no difference to risk of cardiovascular death
Cardiovascular events
Assessed with number of participants experiencing any cardiovascular event
Duration: range 12‐88 months
128 per 1000 123 per 1000
 (118 to 129) RR 0.96
 (0.92 to 1.01) 140,482
 (43 RCTs) ⊕⊕⊕⊕
 Highc LCn3 fat intake makes little or no difference to risk of cardiovascular events
Coronary heart disease mortality – CHD deaths
 Assessed with coronary deaths, or where these were not reported, IHD death, fatal MI or cardiac death (in that order)
Duration: range 12‐88 months
29 per 1000 26 per 1000
 (24 to 29) RR 0.90
 (0.81 to 1.00) 127,378
 (24 RCTs) ⊕⊕⊝⊝
 Lowd Increasing LCn3 fat intake may slightly reduce CHD mortality (NNTB 334, 95% CI 200 to infinity; NNTB 1000 for primary prevention; NNTB 200 for secondary prevention)
Coronary heart disease events – CHD events
 Assessed with number of participants experiencing the first outcome in this list reported for each trial: CHD or coronary events; total MI; acute coronary syndrome; or angina (stable and unstable)
Duration: range 12‐88 months
68 per 1000 62 per 1000
 (58 to 66) RR 0.91
 (0.85 to 0.97) 134,116
 (32 RCTs) ⊕⊕⊝⊝
 Lowe Increasing LCn3 fat intake may slightly reduce the risk of CHD events (NNTB 167, 95% CI 100 to 500; NNTB 200 for primary prevention; NNTB 143 for secondary prevention)
Stroke
Assessed with number of participants experiencing at least 1 fatal or non‐fatal, ischaemic or haemorrhagic stroke
Duration: range 12‐88 months
20 per 1000 21 per 1000
 (19 to 23) RR 1.02
 (0.94 to 1.12) 138,888
 (31 RCTs) ⊕⊕⊕⊝
 Moderatef LCn3 fat intake probably makes little or no difference to risk of experiencing a stroke
Arrhythmias
Assessed with number of participants experiencing fatal or nonfatal, new or recurrent arrhythmia, including atrial fibrillation, ventricular tachycardia and ventricular fibrillation
Duration: range 12‐88 months
57 per 1000 56 per 1000
 (52 to 60) RR 0.99
 (0.92 to 1.06) 77,990
 (30 RCTs) ⊕⊕⊝⊝
 Lowg Increasing LCn3 fat intake may make little or no difference to risk of arrhythmia
Harms: bleeding
Assessed with number of participants experiencing bleeding events
Duration: range 12‐72 months
16 per 1000 18 per 1000
 (14 to 22) RR 1.12
 (0.91 to 1.37) 80,147
 (11 RCTs) ⊕⊝⊝⊝
 Very lowh The effect of LCn3 fat intake on bleeding is unclear as the evidence is of very low certainty
Harms: pulmonary embolus or DVT
Assessed with number of participants experiencing pulmonary embolus or DVT
Duration: range 18‐36 months
5 per 1000 6 per 1000
 (2 to 14) RR 1.15
 (0.44 to 2.98) 3546
 (5 RCTs) ⊕⊝⊝⊝
 Very lowi The effect of LCn3 fat intake on pulmonary embolus or DVT is unclear as the evidence is of very low certainty
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
 CHD: coronary heart disease; CI: confidence interval; CVD: cardiovascular disease; DHA: docosahexaenoic acid; DVT: deep vein thrombosis; EPA: eicosapentaenoic acid; IHD: ischaemic heart disease; LCn3: long‐chain omega‐3; MI: myocardial infarction; RCT: randomised controlled trial; RR: risk ratio.
GRADE Working Group grades of evidenceHigh certainty: we are very confident that the true effect lies close to that of the estimate of the effect.
 Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.
 Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect.
 Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

aAll‐cause mortality, long‐chain omega‐3 (LCn3)

  • Risk of bias: effect size moved closer to no effect (risk ratio (RR) 1.0) when analysis was limited to trials at low summary risk of bias and low risk of compliance bias (adding weight to the suggestion of little or no effect) but did not alter with fixed‐effect meta‐analysis or results in the analysis limited to larger trials. It was further noted by the World Health Organization Nutrition Guidance Expert Advisory Group (NUGAG) Subgroup on Diet and Health that although many of the trials had issues with blinding, the tendency for lack of blinding is an overestimation of effect. This is less of a concern for this outcome, as the pooled effect was approaching null and not statistically significant. Not downgraded.
  • Inconsistency: I2 statistic was less than 60% and I2 reduced when analysis was limited to trials at low summary risk of bias. This adds weight to the suggestion of little or no effect. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with cardiovascular disease (CVD) risk factors or previous CVD as well as non‐CVD health problems. Low‐ and middle‐income countries were represented but underrepresented. Not downgraded.
  • Imprecision: tight confidence intervals, very large numbers of participants took part in long‐term randomised controlled trials (RCTs) with consistent results. Given the lack of a statistically significant effect in this very large set of participants, any effect appears too small to be individually relevant. Not downgraded.
  • Publication bias: the funnel plot suggested that some small trials with higher numbers of events in the intervention group might be missing. If such missing trials were added back in, the RR would rise. This adds weight to the suggestion of little or no effect. Not downgraded.

bCardiovascular mortality, LCn3

  • Risk of bias: effect size moved closer to no effect (RR 1.0) when analysis was limited to trials at low summary risk of bias and with fixed‐effect analysis (adding weight to the suggestion of little or no effect) but did not alter when the analyses were limited to trials at low risk of compliance bias or larger trials. It was further noted by the WHO NUGAG Subgroup on Diet and Health that although many of the RCTs had issues with blinding, the tendency for lack of blinding is an overestimation of effect. This is less of a concern for this outcome, as the pooled effect was approaching null. Not downgraded.
  • Inconsistency: I2 statistic was less than 60% and I2 reduced when analysis was limited to trials at low summary risk of bias. This adds weight to the suggestion of little or no effect. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. All trials were conducted in high‐income countries. Not downgraded.
  • Imprecision: 95% confidence intervals do not exclude important benefits. Downgraded once.
  • Publication bias: the funnel plot suggested that some small trials with higher numbers of events in the intervention group might be missing. If such missing trials were added back in, the RR would rise. This adds weight to the suggestion of little or no effect. Not downgraded.

cCardiovascular events, LCn3

  • Risk of bias: effect size moved closer to no effect (RR 1.0) when analysis was limited to trials at low summary risk of bias (adding weight to the suggestion of little or no effect) but did not alter with fixed‐effect meta‐analysis or results in the analysis limited to larger trials. It was further noted by the WHO NUGAG Subgroup on Diet and Health that although many of the RCTs had issues with blinding, the tendency for lack of blinding is an overestimation of effect. This is less of a concern for this outcome, as the pooled effect was approaching null and not statistically significant. Not downgraded.
  • Inconsistency: I2 statistic was less than 60% and I2 reduced when analysis was limited to trials at low summary risk of bias. This adds weight to the suggestion of little or no effect. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. Low‐ and middle‐income countries were represented but underrepresented. Not downgraded.
  • Imprecision: very large numbers of participants took part in long‐term RCTs with consistent results. Given the lack of an important effect in this very large set of participants, any effect appears too small to be individually relevant. The 95% confidence intervals excluded important benefits or harms. Not downgraded.
  • Publication bias: the funnel plot suggested that some small trials with higher numbers of events in the intervention group might be missing. If such missing trials were added back in, the RR would rise. This adds weight to the suggestion of little or no effect. Not downgraded.

dCoronary heart disease mortality, LCn3

  • Risk of bias: effect size more than 8% was retained when analysis was limited to trials at low summary risk of bias, low risk of compliance bias and larger trials. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. All trials were conducted in high‐income countries. Not downgraded.
  • Imprecision: large numbers of participants took part in long‐term RCTs with consistent results. As 95% confidence intervals did not exclude lack of effect we downgraded once.
  • Publication bias: the funnel plot suggested that some small trials with higher numbers of events in the intervention group might be missing. If such missing trials were added back in the RR would rise. This weakens the suggestion of an effect. Downgraded once.

eCoronary heart disease events, LCn3

  • Risk of bias: effect size moved closer to no effect (RR 1.0) when analysis was limited to trials at low summary risk of bias, but increased when limiting trials to those at low risk of compliance problems and larger trials. There was a small protective effect in the main analysis and some sensitivity analyses, but not in sensitivity analyses limiting to RCTs at low summary risk of bias or using fixed‐effect analysis. The suggestion of a dose response in meta‐regression was lost when REDUCE‐IT 2019 data were omitted. We summarised this as suggesting a true effect of around 8%. This is on the borderline of little or no effect and a more than 8% effect. Downgraded twice.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. Low‐ and middle‐income countries were represented but underrepresented. Not downgraded.
  • Imprecision: 95% confidence interval did not include the null. Not downgraded.
  • Publication bias: no suggestion from the funnel plot of publication bias. Not downgraded.

fStroke, LCn3

  • Risk of bias: effect size consistently suggested little or no effect for all sensitivity analyses. It was further noted by the WHO NUGAG Subgroup on Diet and Health that although many of the RCTs had issues with blinding, the tendency for lack of blinding is an overestimation of effect. This is less of a concern for this outcome, as the pooled effect was approaching null and not statistically significant. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. Low‐ and middle‐income countries were represented but underrepresented. Not downgraded.
  • Imprecision: very large numbers of participants took part in long‐term RCTs with consistent results. Given the lack of a statistically significant effect in this very large set of participants any effect appears too small to be individually relevant. However, as 95% confidence intervals do not exclude important harms, we downgraded once.
  • Publication bias: the funnel plot did not suggest any small trial bias. Not downgraded.

gArrhythmias, LCn3

  • Risk of bias: effect size remained similar in most sensitivity analyses, but suggested harm when limited to trials at low summary risk of bias. Downgraded once.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. Low‐ and middle‐income countries were represented but underrepresented. Not downgraded.
  • Imprecision: As 95% confidence intervals of the sensitivity analysis excluding trials at higher risk of bias included both harm and no effect, and there was a statistically significant difference in effect size between trials at low summary risk of bias and other trials, we downgraded once.
  • Publication bias: funnel plot not interpretable as trials all of a similar size and weight. Not downgraded.

hBleeding, LCn3

  • Risk of bias: effect size changed direction (from harmful to protective) when analysis limited to trials at low summary risk of bias. Downgraded once.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. Low‐ and middle‐income countries not represented. Not downgraded.
  • Imprecision: 95% confidence intervals do not exclude large and important benefits or harms. Downgraded twice.
  • Publication bias: insufficient trials for funnel plot. Not downgraded.

iPulmonary embolus or DVD, LCn3

  • Risk of bias: effect size suggested greater harm when analysis limited to trials at low summary risk of bias. Downgraded once.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. Low‐ and middle‐income countries not represented. Not downgraded.
  • Imprecision: 95% confidence intervals do not exclude large benefits or large harms. Downgraded twice.
  • Publication bias: insufficient trials for funnel plot. Not downgraded.
Summary of findings 2. High versus low alpha‐linolenic acid omega‐3 fats for preventing cardiovascular disease (primary outcomes).
High versus low alpha‐linolenic omega‐3 fats for preventing cardiovascular disease (primary outcomes)
Patient or population: adults with or without existing CVD
Setting: participants were living at home for most or all of the duration of their trials. Most trials were carried out in high‐income economies (World Bank 2018), but four were carried out in upper‐middle‐income countries (Argentina, Iran, Turkey and China). No trials took place in low‐ or low‐ to middle‐income countries.
 Intervention: higher intake of ALA
 Comparison: lower intake of ALA
The intervention was dietary supplementation, a provided diet or advice on diet. Supplementation may have been in oil or capsule form or as foodstuffs provided, to be consumed by mouth (excluding enteral and parenteral feeds and enemas). The foodstuffs or supplements must have been refined ALA: linseed (flax); canola (rapeseed); perilla; purslane; mustard seed; candlenut; stillingia; or walnut, as a food, oil, made into a spreading fat or supplementing another food (such as bread or eggs). For ALA sources the product consumed had to have an omega‐3 fat content of at least 10% of the total fat content.
Outcomes Anticipated absolute effects* (95% CI) Relative effect
 (95% CI) № of participants
 (trials) Certainty of the evidence
 (GRADE) Comments
Risk with lower ALA Risk with higher ALA
All‐cause mortality – deaths
 Assessed with number of participants dying of any cause, whether reported as an outcome or a reason for dropout
Duration: range 12‐40 months
24 per 1000 24 per 1000
 (20 to 28) RR 1.01
 (0.84 to 1.20) 19,327
 (5 RCTs) ⊕⊕⊕⊝
 Moderatea ALA intake probably makes little or no difference to risk of all‐cause mortality
Cardiovascular mortality – cardiovascular deaths
 Assessed with deaths from any cardiovascular cause. Where this was not available, we used cardiac death instead where known.
Duration: range 12‐40 months
12 per 1000 12 per 1000
 (9 to 15) RR 0.96
 (0.74 to 1.25) 18,619
 (4 RCTs) ⊕⊕⊕⊝
 Moderateb ALA intake probably makes little or no difference to risk of cardiovascular mortality
Cardiovascular events
Assessed with number of participants experiencing any cardiovascular event
Duration: range 12‐40 months
47 per 1000 45 per 1000
 (39 to 50) RR 0.95
 (0.83 to 1.07) 19,327
 (5 RCTs) ⊕⊕⊝⊝
 Lowc Increasing ALA intake may slightly reduce the risk of cardiovascular events. (NNTB 500, 95% CI 125 to −334; NNTB 500 in primary prevention; NNTB 84 in secondary prevention)
Coronary heart disease mortality – CHD deaths
 Assessed with: coronary deaths, or where these were not reported, IHD death, fatal MI or cardiac death (in that order)
Duration: range 12‐40 months
11 per 1000 10 per 1000
 (8 to 14) RR 0.95
 (0.72 to 1.26) 18,353
 (3 RCTs) ⊕⊕⊕⊝
 Moderated ALA intake probably has little or no effect on risk of CHD mortality
Coronary heart disease events
 Assessed with number of participants experiencing the first outcome in this list reported for each trial: CHD or coronary events; total MI; acute coronary syndrome; or angina (stable and unstable)
Duration: range 12‐40 months
21 per 1000 21 per 1000
 (17 to 26) RR 1.00
 (0.82 to 1.22) 19,061
 (4 RCTs) ⊕⊕⊝⊝
 Lowe ALA intake may make little or no difference to CHD events
Stroke
Assessed with: number of participants experiencing at least one fatal or non‐fatal, ischaemic or haemorrhagic stroke
Duration: range 12 to 40 months
2 per 1000 3 per 1000
 (2 to 5) RR 1.15
 (0.66 to 2.01) 19,327
 (5 RCTs) ⊕⊝⊝⊝
 Very lowf The effect of ALA intake on stroke is unclear as the evidence is of very low certainty
Arrhythmias
Assessed with number of participants experiencing fatal or nonfatal, new or recurrent arrhythmia, including atrial fibrillation, ventricular tachycardia and ventricular fibrillation
Duration: range 12‐40 months
40 per 1000 29 per 1000
 (22 to 39) RR 0.73
 (0.55 to 0.97) 4912
 (2 RCTs) ⊕⊕⊕⊝
 Moderateg ALA intake probably slightlyreduces the risk of arrhythmias. (NNTB 91, 95% CI 56 to 1000; assessment by primary or secondary prevention not possible)
Harms: bleeding
Assessed with number of participants experiencing bleeding events
The effect of ALA intake on bleeding is unclear as no trials reported this outcome.
Harms: pulmonary embolus or DVT
Assessed with number of participants experiencing pulmonary embolus or DVT
Duration: 24 months
3 per 1000 1 per 1000
 (0 to 23) RR 0.32
 (0.01 to 7.80) 708
 (1 RCT) ⊕⊝⊝⊝
 Very lowh The effect of ALA intake on pulmonary embolus or DVT is unclear as the evidence is of very low certainty
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
 ALA: alpha‐linolenic acid; CHD: coronary heart disease; CI: confidence interval; DVT: deep vein thrombosis; IHD: ischaemic heart disease; MI: myocardial infarction; RCT: randomised controlled trial; RR: risk ratio.
GRADE Working Group grades of evidenceHigh certainty: we are very confident that the true effect lies close to that of the estimate of the effect.
 Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.
 Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect.
 Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

aAll‐cause mortality, alpha‐linolenic acid (ALA)

  • Risk of bias: there was little or no effect in the main meta‐analysis or in any sensitivity analysis. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with cardiovascular disease (CVD) risk factors or previous CVD as well as non‐CVD health problems. All trials were conducted in high‐income countries. Not downgraded.
  • Imprecision: large numbers of participants took part in long‐term RCTs with consistent results. However, as 95% confidence intervals do not exclude important benefits or harms we downgraded once.
  • Publication bias: insufficient trials for funnel plot. Not downgraded.

bCardiovascular mortality, ALA

  • Risk of bias: there was little or no effect in the main analysis, or in any sensitivity analysis. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. All trials were conducted in high‐income countries. Not downgraded.
  • Imprecision: large numbers of participants took part in RCTs in long‐term trials with consistent results. However, as 95% confidence intervals do not exclude important benefits or harms we downgraded once.
  • Publication bias: insufficient trials for funnel plot. Not downgraded.

cCardiovascular events, ALA

  • Risk of bias: there was little or no effect in the main analysis, with larger trials and in fixed‐effect analysis, and a 9%‐10% reduction in risk when data were limited to RCTs at low summary risk of bias or at low risk from compliance problems. Downgraded once.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. All trials were conducted in high‐income countries. Not downgraded.
  • Imprecision: As 95% confidence intervals do not exclude important benefits we downgraded once.
  • Publication bias: insufficient trials for funnel plot. Not downgraded.

dCoronary heart disease mortality, ALA

  • Risk of bias: all sensitivity analyses suggested little or no effect. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. All trials were conducted in high‐income countries. Not downgraded.
  • Imprecision: As 95% confidence intervals do not exclude important benefits or harms we downgraded once.
  • Publication bias: insufficient trials for funnel plot. Not downgraded.

eCoronary heart disease events, ALA

  • Risk of bias: there was little or no effect in the main analyses, in fixed‐effect meta‐analysis, in larger trials or when limiting to trials at low risk of compliance bias, but some risk reduction (9%) when data were limited to RCTs at low summary risk of bias. Downgraded once.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. All trials were conducted in high‐income countries. Not downgraded.
  • Imprecision: as 95% confidence intervals do not exclude important benefits or harms we downgraded once.
  • Publication bias: insufficient trials for funnel plot. Not downgraded.

fStroke, ALA

  • Risk of bias: the main analysis, fixed‐effect analysis, and larger trials suggest increased risk of stroke with more ALA, but there was little or no effect when data were limited to RCTs at low summary risk of bias, and a suggestion of benefit when limited to trials with low risk of compliance problems. Downgraded twice.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with CVD risk factors or previous CVD as well as non‐CVD health problems. All trials were conducted in high‐income countries. Not downgraded.
  • Imprecision: only 49 participants experienced strokes in the included trials; 95% confidence intervals do not exclude important benefits or harms, downgraded once.
  • Publication bias: insufficient trials for funnel plot. Not downgraded.

gArrhythmias, ALA

  • Risk of bias: increasing ALA reduced the risk of arrhythmia in the main analysis, and also in all sensitivity analyses. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: two trials, which included adults with previous myocardial infarction or successful cardioversion in high‐income countries. Not downgraded.
  • Imprecision: as 95% confidence intervals do not exclude little or no effect we downgraded once.
  • Publication bias: insufficient trials for funnel plot. Not downgraded.

hPulmonary embolus or DVD, ALA

  • Risk of bias: the single trial was not at low summary risk of bias. Downgraded once.
  • Inconsistency: with one trial no inconsistency. Not downgraded.
  • Indirectness: healthy men and women, no participants with CVD risk factors or previous CVD; low‐ and middle‐income countries not represented. Not downgraded.
  • Imprecision: only one event included in a single trial. Downgraded twice.
  • Publication bias: insufficient trials for funnel plot. Not downgraded.
Summary of findings 3. High versus low omega‐3 fats for modification of cardiovascular disease risk factors (adiposity and lipids): key outcomes.
High versus low omega‐3 fats for modification of CVD risk factors (adiposity and lipids)
Patient or population: adults with or without existing CVD
Setting: participants were living at home for most or all of the duration of their trials. Most trials were carried out in high‐income economies (World Bank 2018), but four were carried out in upper‐middle income countries. No trials took place in low‐ or low‐middle income countries.
Intervention: higher omega‐3 intake (LCn3 or ALA)
 Comparison: lower omega‐3 intake (LCn3 or ALA)
The intervention was dietary supplementation, a provided diet or advice on diet. Supplementation may have been in oil or capsule form or as foodstuffs provided, to be consumed by mouth (excluding enteral and parenteral feeds and enemas). The foodstuffs or supplements must have been oily fish; fish oils; linseed (flax), canola (rapeseed), perilla, purslane, mustard seed, candlenut, stillingia or walnut as a food, oil, made into a spreading fat or supplementing another food (such as bread or eggs). For ALA sources the product consumed had to have an omega‐3 fat content of at least 10% of the total fat content. Refined eicosapentaenoic acid (EPA), docosahexaenoic acid (DHA) or ALAs, or concentrated fish or algal oils, were also accepted.
Outcomes
All in trials of 12 to 72 months' duration
Anticipated absolute effects* (95% CI) № of participants
 (trials) Certainty of the evidence
 (GRADE) Comments
Risk with low omega‐3 Risk with high omega‐3
Measures of adiposity, LCn3 ‐ weight, kg Mean body weight was 81.2 kg MD 0.00 kg lower
 (0.69 lower to 0.70 higher) 17,000
 (14 RCTs) ⊕⊕⊕⊕
 Higha LCn3 intake makes little or no difference to body weight
Measures of adiposity, LCn3 ‐ BMI, kg/m2 Mean BMI was 27.3 kg/m2 MD 0.06 higher
 (0.14 lower to 0.25 higher) 15,474
 (15 RCTs) ⊕⊕⊕⊕
 Highb LCn3 intake makes little or no difference to BMI
Serum total cholesterol, LCn3 – TC, mmol/L Mean TC was 5.61 mmol/L MD 0.01 lower
 (0.05 lower to 0.03 higher) 38,469
 (30 RCTs) ⊕⊕⊕⊕
 Highc LCn3 intake makes little or no difference to serum TC
Serum triglyceride,LCn3 ‐ fasting TG, mmol/L Mean TG was 1.59 mmol/L MD 0.24 lower
 (0.31 lower to 0.16 lower) 43,998
 (27 RCTs) ⊕⊕⊕⊕
 Highd Increasing LCn3 intake reduces serum TG by about 0.24 mmol/L or 15%
Serum high‐density lipoprotein, LCn3 – HDL, mmol/L Mean HDL was 1.32 mmol/L MD 0.03 higher
 (0.01 to 0.05 higher) 46,604
 (30 RCTs) ⊕⊕⊕⊕
 Highe Increasing LCn3 intake has little or no effect on serum HDL
Serum low‐density lipoprotein, LCn3 – LDL, mmol/L Mean LDL was 3.27 mmol/L MD 0.01 higher
 (0.01 lower to 0.03 higher) 43,454
 (25 RCTs) ⊕⊕⊕⊕
 Highf LCn3 intake makes little or no difference to serum LDL.
Measures of adiposity, ALA – weight, kg Mean weight was 80.9 kg MD 1.49 lower
 (4.17 lower to 1.18 higher) 664
 (4 RCTs) ⊝⊝⊝⊝
 Very lowg The effect of ALA intake on body weight is unclear as the evidence is of very low certainty
Measures of adiposity, ALA – BMI, kg/m2 Mean BMI was 27.4 kg/m2 MD 0.42 lower
 (1.53 lower to 0.69 higher) 1581
 (3 RCTs) ⊝⊝⊝⊝
 Very lowh The effect of ALA intake on BMI is unclear as the evidence is of very low certainty
Serum total cholesterol, ALA – TC, mmol/L Mean TC was 5.02 mmol/L MD 0.09 lower
 (0.23 lower to 0.05 higher) 2164
 (6 RCTs) ⊕⊕⊝⊝
 Lowi ALA intake may make little or no difference to serum TC
Serum triglyceride, ALA ‐ fasting TG, mmol/L Mean TG was 1.48 mmol/L MD 0.03 lower
 (0.11 lower to 0.05 higher) 1776
 (6 RCTs) ⊕⊕⊕⊝
 Moderatej ALA intake probably makes little or no difference to serum TG
Serum high‐density lipoprotein, ALA – HDL, mmol/L Mean HDL was 1.49 mmol/L MD 0.02 lower
 (0.08 lower to 0.03 higher) 1776
 (6 RCTs) ⊕⊕⊕⊝
 Moderatek ALA intake probably has little or no effect on serum HDL
Serum low‐density lipoprotein, ALA – LDL, mmol/L Mean LDL was 2.88 mmol/L MD 0.05 lower
 (0.15 lower to 0.04 higher) 2201
 (7 RCTs) ⊕⊕⊕⊝
 Moderatel ALA intake probably has little or no effect on serum LDL
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
 ALA: alpha‐linolenic acid; BMI: body mass index; CI: confidence interval; DHA: docosahexaenoic acid; EPA: eicosapentaenoic acid; HDL: high‐density lipoprotein; LCn3: long‐chain omega‐3 fatty acids; LDL: low‐density lipoprotein; MD: mean difference; RCT: randomised controlled trial; TC: total cholesterol; TG: triglycerides
GRADE Working Group grades of evidenceHigh certainty: we are very confident that the true effect lies close to that of the estimate of the effect.
 Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.
 Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect.
 Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

aMeasures of adiposity, weight, long‐chain omega‐3 (LCn3)

  • Risk of bias: there was little or no effect in the main analysis or in any sensitivity analysis. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were underrepresented. Not downgraded.
  • Imprecision: large numbers of participants took part in long‐term randomised controlled trials (RCTs) with consistent results; 95% confidence intervals exclude important benefits or harms. Not downgraded.
  • Publication bias: funnel plot was not interpretable, no clear small trial bias. However, we are aware of several trials whose data could not be included. Not downgraded.

bMeasures of adiposity, body mass index (BMI), LCn3

  • Risk of bias: there was little or no effect in the main analysis or in any sensitivity analysis. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were underrepresented. Not downgraded.
  • Imprecision: large numbers of participants took part in long‐term RCTs with consistent results. 95% confidence intervals exclude important benefits or harms. Not downgraded.
  • Publication bias: funnel plot was not interpretable, no clear small trial bias. However, we are aware of several trials whose data could not be included. Not downgraded.

cLipids, serum total cholesterol, LCn3

  • Risk of bias: there was little or no effect in the main analysis or in any sensitivity analysis. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were not represented. Not downgraded.
  • Imprecision: the 95% CI excluded important benefits or harms. Not downgraded.
  • Publication bias: funnel plot did not suggest clear small trial bias. However, we are aware of several trials whose data could not be included; not downgraded.

dLipids, serum triglycerides, LCn3

  • Risk of bias: there was a greater than 5% (and statistically significant) effect overall and in all sensitivity analyses, including when data were limited to RCTs at low summary risk of bias. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were not represented. Not downgraded.
  • Imprecision: large numbers of participants took part in long‐term RCTs with consistent results. 95% confidence intervals exclude harm and lack of effect. Not downgraded.
  • Publication bias: funnel plot was not interpretable, but results of fixed‐effect and random‐effects analyses were similar, suggesting little small trial bias. However, we are aware of several trials whose data could not be included. Not downgraded.

eLipids, high‐density lipoprotein (HDL), LCn3

  • Risk of bias: the suggested little or no effect (less than 5% increase) in HDL with increased LCn3 was apparent in all sensitivity analyses. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were not represented. Not downgraded.
  • Imprecision: 95% confidence intervals exclude harms or benefits. Not downgraded.
  • Publication bias: funnel plot suggested no clear small trial bias. However, we are aware of several trials whose data could not be included. Not downgraded.

fLipids, low‐density lipoprotein (LDL), LCn3

  • Risk of bias: there was little or no effect in the main analysis or in any sensitivity analysis. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were not represented. Not downgraded.
  • Imprecision: 95% confidence intervals excluded important benefits or harms. Not downgraded.
  • Publication bias: funnel plot did not suggest clear small trial bias, and results of fixed‐effect and random‐effects analyses were similar. However, we are aware of several trials whose data could not be included. Not downgraded.

gMeasures of adiposity, weight, alpha‐linolenic acid (ALA)

  • Risk of bias: no included trials were at low summary risk of bias. Downgraded once.
  • Inconsistency: I2 statistic was greater than 60%. Downgraded once.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were not represented. Not downgraded.
  • Imprecision: 95% confidence intervals included both benefits and harms. Downgraded once.
  • Publication bias: funnel plot was not interpretable, but effects in fixed‐effect and random‐effects meta‐analysis were different suggesting that small trial bias may be present. We are aware of several trials whose data could not be included. Downgraded once.

hMeasures of adiposity, BMI, ALA

  • Risk of bias: the main analysis and some sensitivity analyses suggested that ALA reduced BMI, but this was not seen when trials were limited to those at low summary risk of bias. Downgraded once.
  • Inconsistency: I2 statistic was greater than 60%. Downgraded once.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were not represented. Not downgraded.
  • Imprecision: 95% confidence intervals include benefits and harms. Downgraded once.
  • Publication bias: funnel plot was not interpretable, but effects of fixed‐effect and random‐effects analyses were distinct, suggesting some small trial bias. We are aware of several trials whose data could not be included. Downgraded once.

iLipids, serum total cholesterol, ALA

  • Risk of bias: there was little or no effect in the main analysis or in any sensitivity analysis. Not downgraded.
  • Inconsistency: I2 statistic was greater than 60%. Downgraded once.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were not represented. Not downgraded.
  • Imprecision: the main analysis included both benefits and harms. Downgraded once.
  • Publication bias: funnel plot was not interpretable, no clear small trial bias, fixed‐effect and random‐effects meta‐analysis suggested similar effects. We are aware of several trials whose data could not be included. Not downgraded.

jLipids, serum triglycerides, ALA

  • Risk of bias: there was little or no effect in the main analysis or in any sensitivity analysis. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were not represented. Not downgraded.
  • Imprecision: 95% confidence intervals included benefits. Downgraded once.
  • Publication bias: funnel plot was not interpretable, no clear small trial bias and fixed‐effect and random‐effects analysis results were similar. We are aware of several trials whose data could not be included. Not downgraded.

kLipids, HDL, ALA

  • Risk of bias: there was little or no effect in the main analysis and all sensitivity analyses. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were not represented. Not downgraded.
  • Imprecision: 95% confidence interval includes harms. Downgraded once.
  • Publication bias: funnel plot was not interpretable, effects of fixed‐effect and random‐effects meta‐analysis was very similar suggesting lack of small trial bias. We are aware of several trials whose data could not be included. Not downgraded.

lLipids, LDL, ALA

  • Risk of bias: little or no effect in main analysis and all sensitivity analyses. Not downgraded.
  • Inconsistency: I2 statistic was less than 60%. Not downgraded.
  • Indirectness: representative, generalisable adult population including men and women, including healthy participants and participants with previous CVD. However, low‐ and middle‐income countries were not represented. Not downgraded.
  • Imprecision: for main analysis 95% confidence interval included benefits. Downgraded once.
  • Publication bias: funnel plot was not interpretable, effects of fixed‐effect and random‐effects meta‐analysis were similar suggesting no small trial bias. We are aware of several trials whose data could not be included. Not downgraded.

Results

Description of studies

Results of the search

During the 2019 update, we assessed a further 2419 publications for inclusion and included another seven trials (Figure 1). We found that six previously ongoing trials had published relevant results (Broutset 2007; DREAM Asbell 2018; ENRGISE 2018; ASCEND 2018; REDUCE‐IT 2019; VITAL 2019), so we included them in the review, and we included a further trial (not previously ongoing, HEARTS 2017). We assessed 29 registry entries from ClinicalTrials.gov and 88 entries from WHO ICTRP, of which seven were newly included as ongoing trials (NCT03806426; EVAPORATE 2016; MTG 2018; NCT03784963; LO‐MAPT 2018; POSEIDON 2018; ACTRN12618000761268 2018), and one is awaiting assessment, as it appears to fulfil all inclusion criteria, except that trial duration is unclear (IRCT20100123003140N21). VITAL 2019 had a number of substudies registered with different trial registration numbers, which we included as part of the included VITAL 2019 trial (noted in the Characteristics of included studies), even though most are ongoing. This provided 86 RCTs (over 421 documents) included in this review.

1.

1

Study flow diagram

Of these 86 RCTs, we included 83 in meta‐analyses. Three trials clearly collected relevant data but did not report them in a format that we could use in meta‐analyses (Gill 2012; Ramirez‐Ramirez 2013; Reed 2014; Figure 1).

Using the data provided in Aung 2018, we updated the data for 10 trials already included in this review (AlphaOmega ‐ ALA 2010; AlphaOmega ‐ EPA+DHA 2010; AREDS2 2014; DO IT 2010; GISSI‐HF 2008; GISSI‐P 1999; JELIS 2007; OMEGA 2009; ORIGIN 2012; Risk & Prevention 2013; SU.FOL.OM3 2010). This has allowed us to include additional outcomes for some trials, and use updated numbers for others.

Included studies

The 86 included RCTs randomised 162,796 participants, adding 31% to the number of participants with this update (we included 112,059 participants in the previous version of this review, Abdelhamid 2018a). Fifteen trials (16 comparisons) randomised at least 1000 participants (AlphaOmega ‐ ALA 2010; AlphaOmega ‐ EPA+DHA 2010; ASCEND 2018; AREDS2 2014; DART 1989; DART2 2003; GISSI‐HF 2008; GISSI‐P 1999; JELIS 2007; Norwegian 1968; OMEGA 2009; ORIGIN 2012; REDUCE‐IT 2019; Risk & Prevention 2013; SU.FOL.OM3 2010; VITAL 2019), of which one was a 2x2 factorial trial, which included both interventions (AlphaOmega ‐ ALA 2010; AlphaOmega ‐ EPA+DHA 2010). Most of these larger trials assessed effects of LCn3 fats, but two trials/arms assessed effects of ALA (AlphaOmega ‐ ALA 2010; Norwegian 1968).

Participants had cardiovascular disease at baseline in 35 of the trials (secondary prevention), and the remaining 50 trials were of primary prevention, while one was mixed (REDUCE‐IT 2019).

Most trials assessed effects of LCn3 fats.

Doses of LCn3 ranged from 0.5 g/d of EPA and DHA to more than 5 g/d (19 RCTs had a dose of LCn3 < 1 g/d, 27 a dose of 1 to < 2 g/d, 12 of 2 to < 3 g/d, 19 RCTs had a dose of 3 or more g/d LCn3, 1 did not clearly state their dose).

Fewer trials assessed the effects of ALA on health outcomes.

One trial provided an intervention combining LCn3 and ALA as capsules (DIPP 2015). However, trial authors did not state the ALA dose, so we treated the trial as an LCn3 intervention.

Control groups received olive, corn, sunflower oils, other types of fats (including medium‐chain triglycerides and fat replicating the composition of an average European diet), other 'inert' or ill‐defined substances (liquid paraffin, aluminium hydroxide, 'placebo' not described), different dietary advice or foods without the omega‐3 enrichment, or no treatment/no placebo. These control groups or replacements are shown for each key outcome when subgrouping by replacement (for example Analysis 1.6 shows effects of increasing LCn3 on all‐cause mortality, grouped by replacement).

1.6. Analysis.

1.6

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 6 All‐cause mortality ‐ LCn3 ‐ subgroup by replacement.

The main trial outcome was cardiovascular in 51 trials. Twenty‐one trials (22 comparisons) aimed to measure death or cardiovascular events (AlphaOmega ‐ ALA 2010; AlphaOmega ‐ EPA+DHA 2010; ASCEND 2018; DART 1989; DART2 2003; Doi 2014; FAAT 2005; FLAX‐PAD 2013; GISSI‐HF 2008; GISSI‐P 1999; JELIS 2007; Norwegian 1968; Nye 1990; OFAMI 2001; OMEGA 2009; ORIGIN 2012; REDUCE‐IT 2019; Risk & Prevention 2013; SOFA 2006; SU.FOL.OM3 2010; THIS DIET 2008; VITAL 2019).

Thirty‐two trials aimed to measure various cardiovascular risk factors or progression of cardiovascular health.

Thirty‐five RCTs assessed effects on other health states.

Most trials took place in high‐income economies (World Bank 2018), but four were in upper‐middle‐income countries: Argentina (FORWARD 2013), Iran (Norouzi 2014), Turkey (Özaydin 2011), and China (Zhang 2017). No trials took place exclusively in low‐ or low‐ to middle‐income countries. REDUCE‐IT 2019 was a multi‐centre trial that randomised participants in the USA, Netherlands, Ukraine, Russia, South Africa, Poland, India, Romania, Australia and New Zealand.

We identified a further 25 ongoing trials, which we describe in the table of Characteristics of ongoing studies. At the time of writing this review, all of these trials appear unpublished, and some were recruiting or delivering interventions or had recently been completed, and trial authors were presumably analysing data and writing up results. Others appear overdue for publication, and their status is unclear – they may constitute missing data.

Excluded studies

We read the full texts of over 1000 papers, so the full list of excluded trials is too extensive to add to this review. The main reason for exclusion of full‐text papers was duration of less than 12 months (this was often unclear in abstracts, so we collected full‐text papers to check).

We initially included several trials into our wider data set that we later excluded (Singh 1992; Singh 1997; Singh 1997; Singh 2002). Their exclusion was due to expressions of concern published by the BMJ and The Lancet (BMJ 2005; Horton 2005; White 2005). These expressions of concern followed extensive examination of the conduct, results and publication of these trials and questioned the veracity of data behind several trials published by RB Singh. Another trial that would otherwise have been included was retracted and so not included (Matsuyama 2005).

Risk of bias in included studies

We assessed summary risk of bias as low in 28 RCTs (29 comparisons: ADCS 2010; AlphaOmega ‐ ALA 2010; AlphaOmega ‐ EPA+DHA 2010; AREDS2 2014; ASCEND 2018; Berson 2004; Caldwell 2011; Derosa 2016; DREAM Asbell 2018; EPOCH 2014; FLAX‐PAD 2013; FORWARD 2013; FOSTAR 2016; Lorenz‐Meyer 1996; MAPT 2017; MARGARIN 2002; MARINA 2011; NAT2 2013; OMEGA 2009; OPAL 2010; ORIGIN 2012; Proudman 2015; Puri 2005; Reed 2014; SCIMO 1999; SOFA 2006; SU.FOL.OM3 2010; VITAL 2019; WELCOME 2015), and we deemed it to be moderate to high in the remainder. Our definition of low summary risk of bias is in the section Assessment of risk of bias in included studies. Figure 2 itemises risk of bias by domain and trial.

2.

2

'Risk of bias' summary: review authors' judgements about each 'Risk of bias' item for each included study

Allocation

Of the 86 RCTs (87 RCT arms as AlphaOmega ‐ ALA 2010 and AlphaOmega ‐ EPA+DHA 2010 which include the same participants are represented twice) described in the 'Risk of bias' summary (Figure 2), 69 trials (70 arms) described randomisation well enough to merit an assessment of low risk (the remainder were unclear), and 49 trials (50 arms) described adequate allocation concealment (the remaining 37 were unclear).

Blinding

We considered blinding of participants and personnel to be at low risk of bias in 42 of the 86 trials (43 arms). Lack of blinding of participants put 24 trials at high risk of bias, while the remaining 20 trials were at unclear risk. Blinding of outcome assessors put 57 trials (58 arms) at low risk of detection bias and at seven at high risk; this aspect was unclear in the remainder. We found that 37 trials (38 arms) were at low risk of both performance and detection bias.

Incomplete outcome data

We found that 57 trials (58 arms) were at low risk of attrition bias, 17 at high risk, and the remaining 12 at unclear risk.

Selective reporting

We determined that 22 trials had a pre‐published trials registry entry or protocol and reported all planned outcomes appropriately so we considered them at low risk of selective reporting. Twenty‐five trials (26 arms) were at high risk of selective reporting, omitting reports on either pre‐stated outcomes or time points. We judged the remaining 39 trials to be at unclear risk of reporting bias as we could not find any protocol or prospective trial registry entry (often trials were published prior to trial registration availability).

Other potential sources of bias

We assessed risk of bias due to lack of compliance and attention bias and also noted other sources of bias. We found four trials to be at high risk of compliance bias (FAAT 2005; HERO 2009; Proudman 2015; SMART 2013), while 37 trials (38 arms) provided evidence of good compliance, and the remaining 45 trials were unclear. We noted a high risk of attention bias in three trials where intervention participants potentially had more dedicated time for dietary advice or follow‐up (DART 1989; DART2 2003; MARGARIN 2002). Eleven trials did not provide enough details to assess so we considered them to be at unclear risk of attention bias (Ahn 2016; Broutset 2007; Erdogan 2007; Gill 2012; HEARTS 2017; Kumar 2012; Kumar 2013; Risk & Prevention 2013; Sianni 2013; SMART 2013; WAHA 2016), while we thought the remaining 72 trials (73 arms) were at low risk of attention bias. We judged four trials to be at high risk of other potential biases: Ahn 2016 because it is unclear whether it was placebo‐controlled, and there was concern over reported standard deviations; DISAF 2003 because the trial stopped early; Kumar 2013 due to concerns over design; and REDUCE‐IT 2019 because some data were presented for opposite arms in different publications. Four trials were at unclear risk due to insufficient methodological detail being provided (Broutset 2007; Gill 2012; Sianni 2013; Zhang 2017).

Effects of interventions

See: Table 1; Table 2; Table 3

Primary outcomes

See Table 1 for a GRADE summary of our evidence on effects of long‐chain omega‐3 (LCn3) fats (including eicosapentaenoic acid (EPA), docosahexaenoic acid (DHA) and docosapentaenoic acid (DPA)) on our primary outcomes.

Effects of long‐chain omega‐3 fats on primary health outcomes
All‐cause mortality (LCn3)

High‐certainty evidence showed little or no effect of LCn3 on all‐cause mortality.

There was little or no effect of increasing LCn3 fats on all‐cause mortality, despite 11,297 deaths in more than 143,000 participants (RR 0.97, 95% CI 0.93 to 1.01; I2 = 5%; Analysis 1.1). The funnel plot suggested that some small trials with higher numbers of deaths in the intervention group might be missing (Figure 3), indicating small study bias, though there was almost no difference between random‐effects and fixed‐effect analysis. If any such missing trials were added back in the RR would rise slightly (towards the null value of 1.0).

1.1. Analysis.

1.1

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 1 All‐cause mortality (overall) ‐ LCn3.

3.

3

Funnel plot of comparison 1. High versus low long‐chain omega‐3 fats (primary outcomes), outcome 1.3, all‐cause mortality, sensitivity analysis by summary risk of bias

Sensitivity analyses using fixed‐effect meta‐analysis did not alter the lack of effect on all‐cause mortality (RR 0.97, 95% CI 0.93 to 1.00; Analysis 1.2). Removing RCTs not at low summary risk of bias left us with 19 RCTs involving over 75,000 participants, 5579 of whom died, suggesting no effect of LCn3 on mortality (RR 0.99, 95% CI 0.95 to 1.04; I2 = 0%; Analysis 1.3). This lack of effect was also evident in sensitivity analyses limited to trials at low risk of compliance bias and to larger trials (Analysis 1.4).

1.2. Analysis.

1.2

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 2 All‐cause mortality ‐LCn3 ‐ sensitivity analysis (SA) fixed‐effect.

1.3. Analysis.

1.3

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 3 All‐cause mortality ‐ LCn3 ‐ SA by summary risk of bias.

1.4. Analysis.

1.4

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 4 All‐cause mortality ‐ LCn3 ‐ SA by compliance and study size.

The lack of effect for LCn3 on mortality did not differ by replacement with MUFA, omega‐6 fats or other types of placebo compounds (Analysis 1.6). There was no suggestion of any dose effect for LCn3 fats on mortality (Analysis 1.5), and subgroups with RRs further away from 1.00 had wide 95% confidence intervals. The lack of effect did not differ by primary versus secondary prevention (Analysis 1.9), statin use (Analysis 1.10), mode of intervention (dietary advice, supplemental foods, or capsules, Analysis 1.7), baseline triglycerides (Analysis 1.11) or baseline diabetic status (Analysis 1.12). While there was some suggestion of a small risk reduction in total mortality with LCn3 in trials with medium to long duration (2 to < 4 years, RR 0.91, 95% CI 0.86 to 0.96) and this subgroup was clearly different from other durations (test for subgroup differences P = 0.03), the effect was not evident in shorter (1 to < 2 years) or longer trials (≥ 4 years, RR 1.00, 95% CI 0.96 to 1.05). Because of the lack of effect in longer trials, we did not assume any duration effects (Analysis 1.8).

1.5. Analysis.

1.5

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 5 All‐cause mortality ‐ LCn3 ‐ subgroup by dose.

1.9. Analysis.

1.9

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 9 All‐cause mortality ‐ LCn3 ‐ subgroup by primary or secondary prevention.

1.10. Analysis.

1.10

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 10 All‐cause mortality ‐ LCn3 ‐ subgroup by statin use.

1.7. Analysis.

1.7

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 7 All‐cause mortality ‐ LCn3 ‐ subgroup by intervention type.

1.11. Analysis.

1.11

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 11 All‐cause mortality ‐ LCn3 ‐ subgroup by baseline TG.

1.12. Analysis.

1.12

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 12 All‐cause mortality ‐ LCn3 ‐ subgroup by baseline DM.

1.8. Analysis.

1.8

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 8 All‐cause mortality ‐ LCn3 ‐ subgroup by duration.

As there was no suggestion of any effect of LCn3 fats on all‐cause mortality, we did not carry out meta‐regression.

GRADE assessment suggested that the finding of little or no effect of LCn3 on all‐cause mortality was supported by high‐certainty evidence (not downgraded, Table 1).

Cardiovascular disease mortality (LCn3)

Moderate‐certainty evidence suggests that LCn3 fat intake probably makes little or no difference to cardiovascular deaths.

Twenty‐nine trials in 117,837 participants, 5658 of whom died of cardiovascular disease, reported on cardiovascular mortality (RR 0.92, 95% CI 0.86 to 0.99; I2 = 22%; Analysis 1.13). The funnel plot suggested that some smaller trials with more cardiovascular deaths in the intervention group were missing (some small study bias, Figure 4). If this were the case then adding the missing trials would increase the RR slightly towards the null (no effect). This was supported by the fixed‐effect meta‐analysis effect being closer to the null than the random‐effects model.

1.13. Analysis.

1.13

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 13 Cardiovascular mortality (overall) ‐ LCn3.

4.

4

Funnel plot of comparison 1. High vs low long‐chain omega‐3 fats (primary outcomes), outcome 1.15, cardiovascular disease mortality, sensitivity analysis by summary risk of bias

Fixed‐effect meta‐analysis suggested little or no effect on cardiovascular disease mortality risk (RR 0.93, 95% CI 0.88 to 0.97; Analysis 1.14). Sensitivity analyses removing RCTs not at low summary risk of bias left 12 RCTs in 71,019 participants, 2266 of whom died, also suggesting little or no effect of LCn3 on cardiovascular disease mortality (RR 0.95, 95% CI 0.88 to 1.03; I2 = 0%; Analysis 1.15). Removing trials not at low risk of compliance bias and retaining only larger trials produced similar effects to the main analysis (Analysis 1.16).

1.14. Analysis.

1.14

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 14 CVD mortality ‐ LCn3 ‐ SA fixed‐effect.

1.15. Analysis.

1.15

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 15 CVD mortality ‐ LCn3 ‐ SA by summary risk of bias.

1.16. Analysis.

1.16

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 16 CVD mortality ‐ LCn3 ‐ SA by compliance and study size.

There were no statistically significant differences between subgroups and no differential effects by replacement (Analysis 1.18), mode of intervention (Analysis 1.19), duration (Analysis 1.20), primary or secondary prevention (Analysis 1.21), statin use (Analysis 1.22), omega‐3 dose (Analysis 1.17), baseline triglyceride level (Analysis 1.23) or diabetes status (Analysis 1.24). There was no suggestion of a dose‐response effect.

1.18. Analysis.

1.18

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 18 CVD mortality ‐ LCn3 ‐ subgroup by replacement.

1.19. Analysis.

1.19

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 19 CVD mortality ‐ LCn3 ‐ subgroup by intervention type.

1.20. Analysis.

1.20

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 20 CVD mortality ‐ LCn3 ‐ subgroup by duration.

1.21. Analysis.

1.21

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 21 CVD mortality ‐ LCn3 ‐ subgroup by primary or secondary prevention.

1.22. Analysis.

1.22

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 22 CVD mortality ‐ LCn3 ‐ subgroup by statin uses.

1.17. Analysis.

1.17

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 17 CVD mortality ‐ LCn3 ‐ subgroup by dose.

1.23. Analysis.

1.23

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 23 CVD mortality ‐ LCn3 ‐ subgroup by baseline TG.

1.24. Analysis.

1.24

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 24 CVD mortality ‐ LCn3 ‐ subgroup by baseline DM.

Meta‐regression to assess effects of LCn3 dose, ALA, omega‐6 and total PUFA dose, trial duration, intervention type, primary or secondary prevention and risk of bias (as well as a single multiple regression of the three factors with the smallest P values) showed no association between these factors and risk of cardiovascular mortality (all P values were > 0.2, Table 5). We saw no suggestion of dose‐response or duration effects.

2. Meta‐regression results for cardiovascular mortalitya.
Variable assessed P value
LCn3 dose 0.21
ALA dose 0.88
Omega‐6 dose 0.71
Total PUFA dose 0.78
Duration, months 0.78
Primary or secondary CVD prevention 0.82
Food or capsule 0.27
Risk of bias 0.91
Food or capsule
+ LCn3 dose
+ n6 dose
0.72
0.37
0.99
ALA: alpha‐linolenic acid; CVD: cardiovascular disease; LCn3: long‐chain omega‐3 fatty acids; PUFA: polyunsaturated fatty acids

aRandom‐effects meta‐regression exploring effects of LCn3 dose, ALA dose, omega‐6 dose, total PUFA dose, trial duration, primary or secondary prevention, food or capsule intervention, and summary risk of bias (low or moderate to high) on cardiovascular mortality. We ran the meta‐regression using all included trials that reported this outcome in this review, and its sister reviews (Hooper 2018 and Abdelhamid 2018b). For each variable the P value presented represents probability that the relationship was due to chance (as we had limited power we assumed a true relationship when P < 0.10). Meta‐regression was of each variable singly, plus a multivariate meta‐regression of the 3 single variables with lowest P values. See methods for further information.

The suggestion of a marginally protective effect disappeared in trials at low summary risk of bias, the funnel plot suggests that the true risk ratio is slightly higher than the main estimate, and there was no suggestion of dose or duration effects; thus we summarised the evidence as showing little or no effect of LCn3 on cardiovascular disease mortality. GRADE assessment suggested moderate‐certainty evidence that LCn3 fat intake probably makes little or no difference to cardiovascular deaths (downgraded once for imprecision).

Combined cardiovascular events (LCn3)

High‐certainty evidence suggests that LCn3 intake makes little or no difference to risk of cardiovascular events.

There was little or no effect of increasing LCn3 fats on cardiovascular events (RR 0.96, 95% CI 0.92 to 1.01; I2 = 44%; Analysis 1.25). Analyses included 17,619 participants with cardiovascular events in 140,482 participants in 43 trials. The funnel plot suggested that some smaller trials with more participants experiencing cardiovascular events in the intervention group were missing (some small study bias, not shown). If this were the case then adding the missing trials would increase the RR. However effect sizes were the same in fixed‐effect and random‐effects analyses suggesting any bias is limited.

1.25. Analysis.

1.25

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 25 Cardiovascular events (overall) ‐ LCn3.

Sensitivity analyses removing trials at moderate to high risk of bias left 16 trials, including 73,000 participants, 7951 of whom had cardiovascular disease events, with no suggestion of any effect of LCn3 fats (RR 0.98, 95% CI 0.95 to 1.02; I2 = 0%; Analysis 1.27). Sensitivity analyses including trials at low risk of compliance bias, at low risk of small study bias and using fixed‐effect meta‐analysis did not suggest any effect of LCn3 on cardiovascular disease events (Analysis 1.26; Analysis 1.28).

1.27. Analysis.

1.27

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 27 CVD events ‐ LCn3 ‐ SA by summary risk of bias.

1.26. Analysis.

1.26

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 26 CVD events ‐ LCn3 ‐ SA fixed effect.

1.28. Analysis.

1.28

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 28 CVD events ‐ LCn3 ‐ SA by compliance and study size.

In subgroup analysis there was no suggestion of a dose‐response effect (Analysis 1.29). Effects did not differ by replacement (Analysis 1.30), baseline cardiovascular disease risk (Analysis 1.33), type of intervention (Analysis 1.31), statin use (Analysis 1.34), LCn3 dose (Analysis 1.29), trial duration (Analysis 1.32), baseline triglycerides (Analysis 1.35) or baseline diabetic status (Analysis 1.36), and there were no important differences between subgroups.

1.29. Analysis.

1.29

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 29 CVD events ‐ LCn3 ‐ subgroup by dose.

1.30. Analysis.

1.30

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 30 CVD events ‐ LCn3 ‐ subgroup by replacement.

1.33. Analysis.

1.33

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 33 CVD events ‐ LCn3 ‐ subgroup by primary or secondary prevention.

1.31. Analysis.

1.31

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 31 CVD events ‐ LCn3 ‐ subgroup by intervention type.

1.34. Analysis.

1.34

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 34 CVD events ‐ LCn3 ‐ subgroup by statin use.

1.32. Analysis.

1.32

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 32 CVD events ‐ LCn3 ‐ subgroup by duration.

1.35. Analysis.

1.35

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 35 CVD events ‐ LCn3 ‐ subgroup by baseline TG.

1.36. Analysis.

1.36

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 36 CVD events ‐ LCn3 ‐ subgroup by baseline diabetes.

Meta‐regression suggested a negative relationship between LCn3 dose and risk of cardiovascular disease events (P = 0.02), suggesting reduction in cardiovascular disease risk at higher LCn3 doses, as would be expected from a dose response. However, when the single outlying trial REDUCE‐IT 2019 (with a large effect size and high dose) was omitted, the P value was 0.997 suggesting no relationship between LCn3 dose and risk of cardiovascular disease events. The effect was also moderated in the multiple regression (P = 0.07). Meta‐regression did not suggest associations between ALA, omega‐6 or total PUFA dose, trial duration, intervention type, primary or secondary prevention, risk of bias or the single multiple regression of the three factors with the smallest P values (Table 6).

3. Meta‐regression results for cardiovascular eventsa.
Variable assessed P value Coefficient sign where P < 0.10
LCn3 dose 0.02 Negative (lower CVD event risk with higher LCn3 dose)
ALA dose 0.67  
Omega‐6 dose 0.38  
Total PUFA dose 0.29  
Duration, months 0.16  
Primary or secondary CVD prevention 0.76  
Food or capsule 0.30  
Risk of bias 0.17  
Risk of bias
+ LCn3 dose
+ duration
0.19
0.07
0.19
Negative (lower CVD event risk with higher LCn3 dose)
LCn3 dose ‐ analysis omitting REDUCE‐IT 2019 data 0.99  
ALA: alpha‐linolenic acid; CVD: cardiovascular disease; LCn3: long‐chain omega‐3 fatty acids; PUFA: polyunsaturated fatty acids

aRandom‐effects meta‐regression exploring effects of LCn3 dose, ALA dose, omega‐6 dose, total PUFA dose, trial duration, primary or secondary prevention, food or capsule intervention, and summary risk of bias (low or moderate to high) on cardiovascular events. We ran the meta‐regression using all included trials that reported this outcome in this review, and its sister reviews (Hooper 2018 and Abdelhamid 2018b). For each variable the P value presented represents probability that the relationship was due to chance (as we had limited power we assumed a true relationship when P < 0.10). Meta‐regression was of each variable singly, plus a multivariate meta‐regression of the 3 single variables with lowest P values. See methods for further information.

GRADE assessment suggested high‐certainty evidence that LCn3 intake makes little or no difference to risk of cardiovascular events (not downgraded).

Coronary heart disease mortality (LCn3)

Low‐certainty evidence suggests that increasing LCn3 fat intake may slightly reduce coronary heart mortality (NNTB 334).

Increasing LCn3 fats reduced the risk of coronary heart mortality by 10% (RR 0.90, 95% CI 0.81 to 1.00; I2 = 35%) in 24 trials reporting 3598 events in more than 127,000 participants (Analysis 1.37). Sensitivity analyses using a fixed‐effect model suggested little or no effect on coronary heart disease mortality (RR 0.92, 95% CI 0.86 to 0.98; Analysis 1.38).

1.37. Analysis.

1.37

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 37 Coronary heart disease mortality (overall) ‐ LCn3.

1.38. Analysis.

1.38

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 38 CHD mortality ‐ LCn3 ‐ SA fixed effect.

Retaining only RCTs at low summary risk of bias, meta‐analysis of 10 trials with more than 70,000 participants and 1180 coronary heart disease deaths suggested that increasing LCn3 fats reduced coronary heart disease deaths (RR 0.89, 95% CI 0.76 to 1.04; I2 = 22%; Analysis 1.39). Sensitivity analyses retaining only trials with low risk of compliance bias suggested a 17% reduction in risk with LCn3, retaining only larger trials suggested a 13% reduction (Analysis 1.40). The funnel plot suggested that some smaller trials with higher RRs were missing, and if added back these would increase the RR. The presence of such small study bias is supported by the reduction in effect size in fixed‐effect compared to random‐effects meta‐analysis (random‐effects meta‐analysis weighs information from small trials more heavily).

1.39. Analysis.

1.39

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 39 CHD mortality ‐ LCn3 ‐ SA by summary risk of bias.

1.40. Analysis.

1.40

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 40 CHD mortality ‐ LCn3 ‐ SA by compliance and study size.

When we added this outcome we prespecified that we would use the first of the following list reported in any trial: coronary death, ischaemic heart disease death, fatal myocardial infarction and cardiac death. We used cardiac death only when no other outcomes in this category were available, and we ran a sensitivity analysis omitting cardiac death as it potentially includes other causes of death in addition to coronary heart disease, such as cardiomyopathies and congenital and valvular heart diseases (though numbers are likely to be small). Omitting cardiac death resulted in a 19% reduction in coronary heart disease deaths with LCn3 (RR 0.81, 95% CI 0.73 to 0.90, I2 = 0%, 18 trials including 106,676 participants, Analysis 1.41).

1.41. Analysis.

1.41

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 41 CHD mortality ‐ LCn3 ‐ SA omitting cardiac death.

There were no statistically significant differences between subgroups for LCn3 dose (Analysis 1.42), type of intervention (Analysis 1.44), duration (Analysis 1.45), primary or secondary prevention (Analysis 1.46), statin use (Analysis 1.47), baseline coronary artery disease status (Analysis 1.48), replacement (Analysis 1.43), baseline triglycerides (Analysis 1.49) or diabetes status (Analysis 1.50).

1.42. Analysis.

1.42

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 42 CHD mortality ‐ LCn3 ‐ subgroup by dose.

1.44. Analysis.

1.44

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 44 CHD mortality ‐ LCn3 ‐ subgroup by intervention type.

1.45. Analysis.

1.45

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 45 CHD mortality ‐ LCn3 ‐ subgroup by duration.

1.46. Analysis.

1.46

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 46 CHD mortality ‐ LCn3 ‐ subgroup by primary or secondary prevention.

1.47. Analysis.

1.47

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 47 CHD mortality ‐ LCn3 ‐ subgroup by statin use.

1.48. Analysis.

1.48

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 48 CHD mortality ‐ LCn3 ‐ subgroup by CAD history.

1.43. Analysis.

1.43

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 43 CHD mortality ‐ LCn3 ‐ subgroup by replacement.

1.49. Analysis.

1.49

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 49 CHD mortality ‐ LCn3 ‐ subgroup by baseline TG.

1.50. Analysis.

1.50

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 50 CHD mortality ‐ LCn3 ‐ subgroup by baseline DM.

Meta‐regression to assess associations between LCn3 dose and risk of coronary heart disease mortality found no relationship (P = 0.89, Table 7). Similarly we saw no relationships between ALA dose, omega‐6 dose, total PUFA dose, duration, intervention type, primary or secondary prevention, or risk of bias and coronary heart disease deaths (all P values were > 0.40, Table 7). Multiple regression of the three factors with the smallest P value found no factors associated with risk of coronary heart disease deaths. We saw no suggestion of dose or duration effects.

4. Meta‐regression results for coronary heart disease deathsa.
Variable assessed P value
LCn3 dose 0.89
ALA dose 0.94
Omega‐6 dose 0.61
Total PUFA dose 0.59
Duration, months 0.79
Primary or secondary CVD prevention 0.97
Food or capsule 0.59
Risk of bias 0.41
Risk of bias
+ Food or capsule
+ PUFA dose
0.60
0.81
0.68
ALA: alpha‐linolenic acid; CVD: cardiovascular disease; LCn3: long‐chain omega‐3 fatty acids; PUFA: polyunsaturated fatty acids

aRandom‐effects meta‐regression exploring effects of LCn3 dose, ALA dose, omega‐6 dose, total PUFA dose, trial duration, primary or secondary prevention, food or capsule intervention, and summary risk of bias (low or moderate to high) on coronary heart disease mortality. We ran the meta‐regression using all included trials that reported this outcome in this review, and its sister reviews (Hooper 2018 and Abdelhamid 2018b). For each variable the P value presented represents probability that the relationship was due to chance (as we had limited power we assumed a true relationship when P < 0.10). Meta‐regression was of each variable singly, plus a multivariate meta‐regression of the 3 single variables with lowest P values. See methods for further information.

The NNTB is 334 (95% CI 200 to infinity), so 334 people would need to increase their LCn3 intake to prevent one death from coronary heart disease. If we assess NNTB by primary or secondary prevention of cardiovascular disease, people without previous cardiovascular disease (needing primary prevention) have an NNTB of 1000 (95% CI NNTB 334 to NNTH 1000), while those with existing cardiovascular disease have an NNTB of 200 (95% CI NNTB 112 to NNTH 500).

The suggestion of a small protective effect was seen across sensitivity analyses, excepting fixed‐effect analysis. We summarised the evidence as indicating that increasing LCn3 reduces coronary heart disease mortality. GRADE assessment suggested low‐certainty evidence that LCn3 fat intake may slightly reduce coronary heart mortality (downgraded once for imprecision and once for publication bias).

Coronary heart disease events (LCn3)

Low‐certainty evidence suggests that LCn3 fat intake may slightly reduce the risk of coronary heart events (NNTB 167).

The main meta‐analysis suggested a 9% reduction in people experiencing coronary heart disease events with higher intake of LCn3 fats (RR 0.91, 95% CI 0.85 to 0.97; I2 = 37%; 32 trials, 8777 events, 134,116 participants; Analysis 1.51). The funnel plot did not suggest serious small study bias (Figure 5), but effect sizes in fixed‐effect meta‐analysis were slightly smaller.

1.51. Analysis.

1.51

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 51 Coronary heart disease events (overall) ‐ LCn3.

5.

5

Funnel plot of comparison 1. High vs low long‐chain omega‐3 fats (primary outcomes), outcome 1.51, coronary heart disease events (overall)

The size of any effect was reduced in sensitivity analyses using a fixed‐effect model (RR 0.93, 95% CI 0.89 to 0.96; Analysis 1.52) and when we limited trials to those at low summary risk of bias (RR 0.93, 95% CI 0.82 to 1.05, I2 = 38%, Analysis 1.53), but increased when analyses were limited to trials with good compliance and to larger trials (Analysis 1.54).

1.52. Analysis.

1.52

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 52 CHD events ‐ LCn3 ‐ SA fixed effect.

1.53. Analysis.

1.53

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 53 CHD events ‐ LCn3 ‐ SA by summary risk of bias.

1.54. Analysis.

1.54

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 54 CHD events ‐ LCn3 ‐ SA by compliance and study size.

There were no statistically significant differences between subgroups (Analysis 1.55; Analysis 1.56; Analysis 1.57; Analysis 1.58; Analysis 1.59; Analysis 1.60; Analysis 1.61; Analysis 1.63), except for subgrouping by baseline triglyceride, where the subgroup of trials with raised triglyceride suggested a greater reduction in coronary heart disease events (Analysis 1.62).

1.55. Analysis.

1.55

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 55 CHD events ‐ LCn3 ‐ subgroup by dose.

1.56. Analysis.

1.56

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 56 CHD events ‐ LCn3 ‐ subgroup by replacement.

1.57. Analysis.

1.57

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 57 CHD events ‐ LCn3 ‐ subgroup by intervention type.

1.58. Analysis.

1.58

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 58 CHD events ‐ LCn3 ‐ subgroup by duration.

1.59. Analysis.

1.59

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 59 CHD events ‐ LCn3 ‐ subgroup by primary or secondary prevention.

1.60. Analysis.

1.60

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 60 CHD events ‐ LCn3 ‐ subgroup by statin use.

1.61. Analysis.

1.61

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 61 CHD events ‐ LCn3 subgroup by CAD history.

1.63. Analysis.

1.63

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 63 CHD events ‐ LCn3 ‐ subgroup by baseline DM.

1.62. Analysis.

1.62

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 62 CHD events ‐ LCn3 ‐ subgroup by baseline TG.

Meta‐regression suggested a negative relationship between LCn3 dose and risk of coronary heart disease events (P = 0.02), suggesting reduction in coronary heart disease risk at higher LCn3 doses, as would be expected from a dose response. This was weakened in the single multiple regression of the three factors with the smallest P value (P = 0.06). When we omitted the single outlying trial REDUCE‐IT 2019, meta‐regression no longer suggested any relationship between LCn3 dose and risk of coronary heart disease events (P = 0.81). Meta‐regression did not suggest associations between ALA, omega‐6 or total PUFA dose, trial duration, intervention type, primary or secondary prevention, or risk of bias and risk of coronary heart disease events (Table 8).

5. Metaregression results for coronary heart disease eventsa.
Variable assessed P value Coefficient sign where P < 0.10
LCn3 dose 0.02 Negative (risk of CHD events falls as LCn3 dose increases)
ALA dose 0.18  
Omega‐6 dose 0.51  
Total PUFA dose 0.57  
Duration, months 0.16  
Primary or secondary CVD prevention 0.82  
Food or capsule 0.22  
Risk of bias 0.71  
ALA dose
+ duration
+ LCn3 dose
0.36
0.35
0.06
Negative (risk of CHD events falls as LCn3 dose increases)
LCn3 dose ‐ analysis omitting REDUCE‐IT data 0.81  
ALA: alpha‐linolenic acid; CHD: coronary heart disease; CVD: cardiovascular disease; LCn3: long‐chain omega‐3 fatty acids; PUFA: polyunsaturated fatty acids

aRandom‐effects meta‐regression exploring effects of LCn3 dose, ALA dose, omega‐6 dose, total PUFA dose, trial duration, primary or secondary prevention, food or capsule intervention, and summary risk of bias (low or moderate to high) on CHD events. We ran the meta‐regression using all included trials that reported this outcome in this review, and its sister reviews (Hooper 2018 and Abdelhamid 2018b). For each variable the P value presented represents probability that the relationship was due to chance (as we had limited power we assumed a true relationship when P < 0.10). Meta‐regression was of each variable singly, plus a multivariate meta‐regression of the 3 single variables with lowest P values. See methods for further information.

167 people would need to increase their LCn3 intake for one person to avoid a coronary heart disease event (NNTB 167, 95% CI 100 to 500). If we separate this out for people without existing cardiovascular disease the NNTB is 200 for primary prevention (95% CI NNTB 112 to NNTB infinity), and the NNTB is 143 for people with existing cardiovascular disease, secondary prevention, (95% CI NNTB 91 to NNTH 500).

While there was a small protective effect in the main analysis and some sensitivity analyses, but not in sensitivity analyses limiting to RCTs at low summary risk of bias or using fixed‐effect analysis, there was also a suggestion of a dose response. We summarised this as suggesting a true effect of around 8%. GRADE assessment suggested low‐certainty evidence that increasing LCn3 fat intake may slightly reduce risk of coronary heart events (downgraded twice for risk of bias).

Stroke (LCn3)

Moderate‐certainty evidence suggests that LCn3 fat intake probably makes little or no difference to risk of experiencing a stroke.

Increasing intake of LCn3 appears to have little effect on risk of stroke (RR 1.02, 95% CI 0.94 to 1.12; I2 = 11%; 31 trials reported 2850 strokes; Analysis 1.64), and the funnel plot did not suggest small study bias (not shown).

1.64. Analysis.

1.64

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 64 Stroke (overall) ‐ LCn3.

Sensitivity analyses removing trials not at low summary risk of bias left 14 trials with 1684 participants experiencing strokes, suggesting little or no effect of LCn3 fats on stroke (RR 0.99, 95% CI 0.90 to 1.09; I2 = 0%; Analysis 1.66). Using fixed‐effect meta‐analysis and sensitivity analysis removing trials with risk from poor compliance and smaller trials all suggested no effect of increased LCn3 (Analysis 1.66; Analysis 1.68).

1.66. Analysis.

1.66

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 66 Stroke ‐ LCn3 ‐ SA by summary risk of bias.

1.68. Analysis.

1.68

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 68 Stroke ‐ LCn3 ‐ subgroup by stroke type.

When trials reported stroke type separately, the risk of haemorrhagic stroke was increased (RR 1.23, 95% CI 0.93 to 1.64; I2 = 0%; 197 participants with events), while ischaemic stroke was unaltered (RR 0.98, 95% CI 0.79 to 1.20; I2 = 47%; 985 people with events in 10 trials; Analysis 1.68). Six trials reported only 152 participants experiencing transient ischaemic attack, suggesting a 10% increase in risk but with very wide confidence intervals (transient ischaemic attacks were not included in any other stroke categories; Analysis 1.68). Subgrouping did not suggest important differences by intervention type, replacement, statin use, trial duration, baseline triglycerides or diabetic status (Analysis 1.70; Analysis 1.71; Analysis 1.72; Analysis 1.74; Analysis 1.75; Analysis 1.76). Subgrouping by dose suggested important differences in effect at different doses, but there was no pattern to these effects so they were assumed to be spurious (Analysis 1.69). There was a suggestion of increased stroke risk in people with cardiovascular disease at baseline (RR 1.21, 95% CI 1.05 to 1.40; I2 = 0%; with differences in effect size between subgroups by primary or secondary prevention, P = 0.002; Analysis 1.73).

1.70. Analysis.

1.70

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 70 Stroke ‐ LCn3 ‐ subgroup by replacement.

1.71. Analysis.

1.71

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 71 Stroke ‐ LCn3 ‐ subgroup by intervention type.

1.72. Analysis.

1.72

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 72 Stroke ‐ LCn3 ‐ subgroup by duration.

1.74. Analysis.

1.74

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 74 Stroke ‐ LCn3 ‐ subgroup by statin use.

1.75. Analysis.

1.75

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 75 Stroke ‐ LCn3 ‐ subgroup by baseline TG.

1.76. Analysis.

1.76

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 76 Stroke ‐ LCn3 ‐ subgroup by baseline DM.

1.69. Analysis.

1.69

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 69 Stroke ‐ LCn3 ‐ subgroup by dose.

1.73. Analysis.

1.73

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 73 Stroke ‐ LCn3 ‐ subgroup by primary or secondary prevention.

Meta‐regression to assess effects of LCn3 dose did not find any clear dose response on risk of stroke (P = 0.12, Table 9). Univariate meta‐regression suggested that trials of longer duration showed smaller risk ratios (or lower risk, P = 0.03), and this effect was retained in the multivariate regression (P = 0.03). There were no clear relationships between dose of any PUFA type, risk of bias, or use of food or capsules, and only trial duration remained statistically significant in multivariate meta‐regression.

6. Metaregression results for strokea.
Variable assessed P value Coefficient sign where P < 0.10
LCn3 dose 0.12
ALA dose 0.73
Omega‐6 dose 0.23
Total PUFA dose 0.09 Negative (lower risk with higher dose)
Duration, months 0.03 Negative (smaller risk with longer duration)
Primary or secondary CVD prevention 0.44
Food or capsule 0.36
Risk of bias 0.26
Duration
+ LCn3 dose
+ total PUFA dose
0.03
0.06
0.15
ALA: alpha‐linolenic acid; CVD: cardiovascular disease; LCn3: long‐chain omega‐3 fatty acids; PUFA: polyunsaturated fatty acids

aRandom‐effects meta‐regression exploring effects of LCn3 dose, ALA dose, omega‐6 dose, total PUFA dose, trial duration, primary or secondary prevention, food or capsule intervention, and summary risk of bias (low or moderate to high) on stroke. We ran the meta‐regression using all included trials that reported this outcome in this review, and its sister reviews (Hooper 2018 and Abdelhamid 2018b). For each variable the P value presented represents probability that the relationship was due to chance (as we had limited power we assumed a true relationship when P < 0.10). Meta‐regression was of each variable singly, plus a multivariate meta‐regression of the 3 single variables with lowest P values. See methods for further information.

Analyses consistently suggested little or no effect of LCn3 on stroke risk, and there were no dose‐response relationships, though there is a possible increase in risk of haemorrhagic stroke with increased LCn3. GRADE assessment suggests moderate‐certainty evidence that LCn3 fat intake probably makes little or no difference to risk of experiencing a stroke (downgraded once for imprecision).

Arrhythmia (LCn3)

Low‐certainty evidence suggests that LCn3 fat intake may slightly increase the risk of arrhythmia (NNTH 1000).

There was little or no effect of LCn3 fats on incidence of new or recurrent (fatal and non‐fatal) arrhythmias (RR 0.99, 95% CI 0.92 to 1.06; I2 = 44%; 4586 events in more than 77,000 participants; Analysis 1.77). The funnel plot was not interpretable as trials were clustered (not shown), but the effect size moved above 1 in fixed‐effect meta‐analysis, suggesting that some trials suggesting harm from increased LCn3 may be missing.

1.77. Analysis.

1.77

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 77 Arrythmia (overall) ‐ LCn3.

Sensitivity analyses removing trials not at low summary risk of bias left 12 trials with 1602 events (> 41,000 participants), suggesting a 9% increase in risk of arrhythmia with increased LCn3 (RR 1.09, 95% CI 0.99 to 1.20, I2 = 0%, Analysis 1.79). Restricting the analysis to trials at low summary risk of bias removed heterogeneity, and there was a statistically significant difference in effect size between subgroups at low versus moderate to high risk of bias (P = 0.04, Analysis 1.79). Using fixed‐effect methodology did not alter the apparent lack of effect of LCn3 on arrhythmia (Analysis 1.78), and sensitivity analysis by compliance and trial size also suggested little or no effect of LCn3 on arrhythmia (Analysis 1.80).

1.79. Analysis.

1.79

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 79 Arrhythmia‐ LCn3 ‐ SA by summary risk of bias.

1.78. Analysis.

1.78

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 78 Arrythmia ‐ LCn3 ‐ SA fixed effects.

1.80. Analysis.

1.80

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 80 Arrhythmia‐ LCn3 ‐ SA by compliance and study size.

Subgrouping by new or recurrent arrhythmias suggested differences between subgroups, with LCn3 increasing the risk of new arrhythmias and having little or no effect on the risk of recurrent arrhythmia (Analysis 1.81). There were also statistically significant differences between subgroups by fatality, suggesting that LCn3 increases the risk of fatal arrhythmia, but protects against non‐fatal arrhythmia (Analysis 1.82), primary or secondary prevention, suggesting harm in primary prevention, little or no effect in secondary prevention (Analysis 1.87) and duration, suggesting little or no effect in trials of up to four years' duration and harm in longer trials (Analysis 1.86). Subgroup analyses by type of intervention, replacement, statin use, dose and baseline triglyceride did not suggest statistically significant differences between subgroups (Analysis 1.83; Analysis 1.84; Analysis 1.85; Analysis 1.88; Analysis 1.89). Subgrouping by diabetes status suggested significant differences between subgroups, with little or no effect of increasing LCn3 on arrhythmia for those recruited without specific diabetes risk factors, but increased risk of arrhythmia in those with diabetes risk factors and in those with diabetes at baseline (Analysis 1.90).

1.81. Analysis.

1.81

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 81 Arrhythmia ‐ LCn3 ‐ subgroup by new or recurrent.

1.82. Analysis.

1.82

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 82 Arrhythmia ‐ LCn3 ‐ subgroup by fatality.

1.87. Analysis.

1.87

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 87 Arrhythmia ‐ LCn3 ‐ subgroup by primary or secondary prevention3.

1.86. Analysis.

1.86

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 86 Arrhythmia ‐ LCn3 ‐ subgroup by duration.

1.83. Analysis.

1.83

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 83 Arrhythmia ‐ LCn3 ‐ subgroup by dose.

1.84. Analysis.

1.84

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 84 Arrhythmia ‐ LCn3 ‐ subgroup by replacement.

1.85. Analysis.

1.85

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 85 Arrhythmia ‐ LCn3 ‐ subgroup by intervention type.

1.88. Analysis.

1.88

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 88 Arrhythmia ‐ LCn3 ‐ subgroup by statin use.

1.89. Analysis.

1.89

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 89 Arrythmia ‐ LCn3 ‐ subgroup by baseline TG.

1.90. Analysis.

1.90

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 90 Arrythmia ‐ LCn3 ‐ subgroup by baseline DM.

Meta‐regression suggested a positive relationship between trial duration and risk ratio, such that longer trials appeared to increase arrhythmia risk (P = 0.03; Table 10). This relationship was lost when we controlled for primary or secondary prevention and ALA dose (P = 0.42). There was also a relationship between arrhythmia with primary versus secondary prevention, suggesting greater risk ratio (harm) in primary prevention (P = 0.03; Table 10). There were no other suggested relationships, and neither of these relationships were maintained in multivariate analysis.

7. Meta‐regression results for arrhythmiaa.
Variable assessed P value Coefficient sign where P < 0.10
LCn3 dose 0.52
ALA dose 0.45
Omega‐6 dose 0.56
Total PUFA dose 0.99
Duration, months 0.03 Positive (higher risk with longer duration)
Primary or secondary CVD prevention 0.03 Negative (greater effect with primary prevention)
Food or capsule 1.00
Risk of bias 0.65
ALA dose
+ Primary secondary prevention
+ duration
0.33
0.42
0.42
ALA: alpha‐linolenic acid; CVD: cardiovascular disease; LCn3: long‐chain omega‐3 fatty acids; PUFA: polyunsaturated fatty acids

aRandom‐effects meta‐regression exploring effects of LCn3 dose, ALA dose, omega‐6 dose, total PUFA dose, trial duration, primary or secondary prevention, food or capsule intervention, and summary risk of bias (low or moderate to high) on arrhythmia. We ran the meta‐regression using all included trials that reported this outcome in this review, and its sister reviews (Hooper 2018 and Abdelhamid 2018b). For each variable the P value presented represents probability that the relationship was due to chance (as we had limited power we assumed a true relationship when P < 0.10). Meta‐regression was of each variable singly, plus a multivariate meta‐regression of the 3 single variables with lowest P values. See methods for further information.

Overall, the trials at lower risk of bias, and the longer trials, suggested slight harm from increasing LCn3. This suggests that there may be some harm associated with increasing LCn3 on arrhythmia risk, and fatal arrhythmia, particularly in the longer term and in primary prevention. One thousand people would need to increase their LCn3 intake for one additional person to experience arrhythmia (NNTH 1000, 95% CI 200 to −334). GRADE assessment suggested low‐certainty evidence that LCn3 fat intake may slightly increase the risk of arrhythmia (downgraded once for risk of bias and once for imprecision).

Effects of alpha‐linolenic acid (ALA) on primary health outcomes

See Table 2 for a summary of our evidence on effects of ALA on our primary outcomes.

As there were fewer than 10 trials for all ALA analyses we did not create or assess funnel plots, though we did run sensitivity analyses and subgroups. We assessed ALA dose‐response and duration effects in meta‐regression of all included LCn3, ALA, omega‐6 and total PUFA trials (but not of ALA trials alone as there were too few trials to carry out meta‐regression with any reliability). None of the trials that increased ALA intakes were in participants with raised baseline triglycerides, increased risk of diabetes or diabetes at baseline, so we did not subgroup by baseline triglycerides or diabetic status.

All‐cause mortality (ALA)

Moderate‐certainty evidence suggests that ALA intake probably makes little or no difference to all‐cause mortality.

There was little or no effect of increasing ALA omega‐3 fats on all‐cause mortality, with 459 deaths in more than 19,000 participants involved in five trials (RR 1.01, 95% CI 0.84 to 1.20; I2 = 0%; Analysis 4.1).

4.1. Analysis.

4.1

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 1 All‐cause mortality (overall) ‐ ALA.

Sensitivity analyses removing trials not at low summary risk of bias left three trials with 375 deaths, again suggesting little or no effect (RR 1.02, 95% CI 0.72 to 1.45; I2 = 3%; Analysis 4.3). Other sensitivity analyses all suggested little or no effect (Analysis 4.2; Analysis 4.4).

4.3. Analysis.

4.3

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 3 All‐cause mortality ‐ ALA ‐ SA by summary risk of bias.

4.2. Analysis.

4.2

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 2 All‐cause mortality ‐ ALA ‐ sensitivity analysis (SA) fixed‐effect.

4.4. Analysis.

4.4

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 4 All‐cause mortality ‐ ALA ‐ SA by compliance and study size.

Subgrouping by ALA dose, trial duration, statin use, replacement, primary or secondary prevention, or intervention type did not result in any significant differences between subgroups (Analysis 4.5; Analysis 4.6; Analysis 4.7; Analysis 4.8; Analysis 4.9; Analysis 4.10). As there was no suggestion of effect in any subgroup, we did not carry out meta‐regression.

4.5. Analysis.

4.5

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 5 All‐cause mortality ‐ ALA ‐ subgroup by dose.

4.6. Analysis.

4.6

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 6 All‐cause mortality ‐ ALA ‐ subgroup by replacement.

4.7. Analysis.

4.7

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 7 All cause mortality ‐ ALA ‐ subgroup by intervention type.

4.8. Analysis.

4.8

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 8 All‐cause mortality ‐ ALA ‐ subgroup by duration.

4.9. Analysis.

4.9

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 9 All‐cause mortality ‐ ALA ‐ subgroup by primary or secondary prevention.

4.10. Analysis.

4.10

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 10 All‐cause mortality ‐ ALA ‐ subgroup by statin use.

GRADE assessment suggested that ALA intake probably makes little or no difference to all‐cause mortality (moderate‐certainty evidence, downgraded once for imprecision).

Cardiovascular mortality (ALA)

Moderate‐certainty evidence suggests that increasing ALA intake probably has little or no effect on cardiovascular mortality.

Four trials contributed data to this outcome. There was little or no effect of increasing ALA omega‐3 fats on cardiovascular mortality (RR 0.96, 95% CI 0.74 to 1.25; I2 = 0%; Analysis 4.11), but confidence intervals were very wide. Analyses included 219 cardiovascular disease deaths in more than 18,000 participants. All sensitivity analyses suggested little or no effect of increasing ALA (Analysis 4.12; Analysis 4.13; Analysis 4.14).

4.11. Analysis.

4.11

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 11 Cardiovascular mortality (overall) ‐ ALA.

4.12. Analysis.

4.12

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 12 CVD mortality ‐ ALA ‐ SA fixed‐effect.

4.13. Analysis.

4.13

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 13 CVD mortality ‐ ALA ‐ SA by summary risk of bias.

4.14. Analysis.

4.14

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 14 CVD mortality ‐ ALA ‐ SA by compliance and study size.

Subgrouping by ALA dose, trial duration, replacement, intervention type, statin use, or primary or secondary prevention did not suggest important differences between subgroups (Analysis 4.15; Analysis 4.16; Analysis 4.17; Analysis 4.18; Analysis 4.19; Analysis 4.20). Meta‐regression to assess for effects of ALA dose on cardiovascular mortality did not suggest dose effects (P = 0.88; Table 5).

4.15. Analysis.

4.15

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 15 CVD mortality ‐ ALA ‐ subgroup by dose.

4.16. Analysis.

4.16

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 16 CVD mortality ‐ ALA ‐ subgroup by replacement.

4.17. Analysis.

4.17

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 17 CVD mortality ‐ ALA ‐ subgroup by intervention type.

4.18. Analysis.

4.18

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 18 CVD mortality ‐ ALA ‐ subgroup by duration.

4.19. Analysis.

4.19

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 19 CVD mortality ‐ ALA ‐ subgroup by primary or secondary prevention.

4.20. Analysis.

4.20

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 20 CVD mortality ‐ ALA ‐ subgroup by statin uses.

GRADE assessment suggested that increasing ALA intake probably has little or no effect on cardiovascular mortality (moderate‐certainty evidence, downgraded once for imprecision).

Cardiovascular events (ALA)

GRADE assessment suggested low‐certainty evidence that increasing ALA intake may slightly reduce the risk of cardiovascular events (NNTB 500).

There was little or no effect on risk of cardiovascular events in five trials with increased ALA intake (RR 0.95, 95% CI 0.83 to 1.07; I2 = 0%; 884 out of > 19,000 participants experienced at least one cardiovascular event; Analysis 4.21). Sensitivity analyses removing trials at moderate to high risk of bias left three trials in which 691 of more than 5000 enrolled participants experienced at least one cardiovascular event, suggesting a 9% reduction in risk of cardiovascular disease events with higher ALA (RR 0.91, 95% CI 0.79 to 1.04; I2 = 0%; Analysis 4.23). Fixed‐effect analysis, and limiting to larger trials, suggested little or no effect on cardiovascular disease event risk (Analysis 4.22; Analysis 4.24), while trials at low risk of compliance bias suggested a 10% reduction in risk (Analysis 4.24).

4.21. Analysis.

4.21

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 21 Cardiovascular events (overall) ‐ ALA.

4.23. Analysis.

4.23

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 23 CVD events ‐ ALA ‐ SA by summary risk of bias.

4.22. Analysis.

4.22

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 22 CVD events ‐ ALA ‐ SA fixed‐effect.

4.24. Analysis.

4.24

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 24 CVD events ‐ ALA ‐ SA by compliance and study size.

Subgrouping by ALA dose, trial duration, replacement, intervention type, statin use, or primary or secondary prevention did not suggest significant differences between subgroups (Analysis 4.25; Analysis 4.26; Analysis 4.27; Analysis 4.28; Analysis 4.29; Analysis 4.30). Meta‐regression to assess for effects of ALA dose on cardiovascular events did not suggest dose effects (P = 0.67; Table 6).

4.25. Analysis.

4.25

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 25 CVD events ‐ ALA ‐ subgroup by dose.

4.26. Analysis.

4.26

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 26 CVD events ‐ ALA ‐ subgroup by replacement.

4.27. Analysis.

4.27

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 27 CVD events ‐ ALA ‐ subgroup by intervention type.

4.28. Analysis.

4.28

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 28 CVD events ‐ ALA ‐ subgroup by duration.

4.29. Analysis.

4.29

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 29 CVD events ‐ ALA ‐ subgroup by primary or secondary prevention.

4.30. Analysis.

4.30

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 30 CVD events ‐ ALA ‐ subgroup by statin use.

The NNTB was 500 (95% CI 125 to −334), so 500 people would need to increase their ALA intake to prevent one person experiencing a cardiovascular event. The NNTB for primary prevention was 500 (95% CI 125 to −112), the NNTB for secondary prevention was 84 (95% CI 35 to 143). GRADE assessment suggested that increasing ALA intake may reduce the risk of cardiovascular events by a small amount (low‐certainty evidence, downgraded once for risk of bias and once for imprecision).

Coronary heart disease mortality (ALA)

GRADE assessment suggested that increasing ALA intake probably has little or no effect on coronary heart disease mortality.

Three trials reported 193 coronary heart disease deaths in more than 18,000 participants, suggesting little or no effect on coronary heart disease mortality with increased ALA (RR 0.95, 95% CI 0.72 to 1.26; I2 = 0%; Analysis 4.31). No sensitivity analyses suggested benefits or harms (Analysis 4.32; Analysis 4.33; Analysis 4.34).

4.31. Analysis.

4.31

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 31 Coronary heart disease mortality (overall) ‐ ALA.

4.32. Analysis.

4.32

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 32 CHD mortality ‐ ALA ‐ SA fixed‐effect.

4.33. Analysis.

4.33

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 33 CHD mortality ‐ ALA ‐ SA by summary risk of bias.

4.34. Analysis.

4.34

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 34 CHD mortality ‐ ALA ‐ SA by compliance and study size.

Subgrouping by ALA dose, trial duration, replacement, intervention type, statin use, primary or secondary prevention, or previous history of coronary artery disease did not suggest important differences between subgroups (Analysis 4.35; Analysis 4.36; Analysis 4.37; Analysis 4.38; Analysis 4.39; Analysis 4.40; Analysis 4.41), and meta‐regression did not suggest dose effects (P = 0.94; Table 7).

4.35. Analysis.

4.35

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 35 CHD mortality ‐ ALA ‐ subgroup by dose.

4.36. Analysis.

4.36

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 36 CHD mortality ‐ ALA ‐ subgroup by replacement.

4.37. Analysis.

4.37

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 37 CHD mortality ‐ ALA ‐ subgroup by intervention type.

4.38. Analysis.

4.38

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 38 CHD mortality ‐ ALA ‐ subgroup by duration.

4.39. Analysis.

4.39

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 39 CHD mortality ‐ ALA ‐ subgroup by primary or secondary prevention.

4.40. Analysis.

4.40

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 40 CHD mortality ‐ ALA ‐ subgroup by statin use.

4.41. Analysis.

4.41

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 41 CHD mortality ‐ ALA ‐ subgroup by CAD history.

GRADE assessment suggested that increasing ALA intake probably has little or no effect on coronary heart disease mortality (moderate‐certainty evidence, downgraded once for imprecision).

Coronary heart disease events (ALA)

Low‐certainty evidence suggests that ALA intake may make little or no difference to coronary heart disease events.

Four trials contributed data to this outcome, with 397 of more than 19,000 participants experiencing at least one coronary heart disease event. There was little or no effect on coronary heart disease risk with increased ALA (RR 1.00, 95% CI 0.82 to 1.22 I2 = 2%; Analysis 4.42). Sensitivity analyses using fixed‐effect meta‐analysis did not alter the lack of effect (RR 1.00, 95% CI 0.82 to 1.21; Analysis 4.43). Removing trials not at low summary risk of bias left two trials with almost 5000 participants, suggesting a 9% reduction in risk of a coronary heart disease event (RR 0.91, 95% CI 0.71 to 1.15; Analysis 4.44), though little or no effect was seen when restricting analysis to larger trials or those with better compliance (Analysis 4.45).

4.42. Analysis.

4.42

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 42 Coronary heart disease events (overall) ‐ ALA.

4.43. Analysis.

4.43

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 43 CHD events ‐ ALA ‐ SA fixed‐effect.

4.44. Analysis.

4.44

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 44 CHD events ‐ ALA ‐ SA by summary risk of bias.

4.45. Analysis.

4.45

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 45 CHD events ‐ ALA ‐ SA by compliance and study size.

Subgrouping by ALA dose, trial duration, replacement, intervention type, statin use, primary or secondary prevention, or previous history of coronary artery disease did not suggest important differences between subgroups (Analysis 4.46; Analysis 4.47; Analysis 4.48; Analysis 4.49; Analysis 4.50; Analysis 4.51; Analysis 4.52). Meta‐regression did not suggest a correlation between ALA dose and coronary heart disease events (P = 0.18; Table 8).

4.46. Analysis.

4.46

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 46 CHD events ‐ ALA ‐ subgroup by dose.

4.47. Analysis.

4.47

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 47 CHD events ‐ ALA ‐ subgroup by replacement.

4.48. Analysis.

4.48

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 48 CHD events ‐ ALA ‐ subgroup by intervention type.

4.49. Analysis.

4.49

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 49 CHD events ‐ ALA ‐ subgroup by duration.

4.50. Analysis.

4.50

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 50 CHD events ‐ ALA ‐ subgroup by primary or secondary prevention.

4.51. Analysis.

4.51

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 51 CHD events ‐ ALA ‐ subgroup by statin use.

4.52. Analysis.

4.52

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 52 CHD events ‐ ALA ‐ subgroup by CAD history.

Given the differences in sensitivity analyses, GRADE assessment suggested that ALA intake may make little or no difference to coronary heart disease events (low‐certainty evidence, downgraded once for risk of bias and once for imprecision).

Stroke (ALA)

The effect of ALA intake on stroke is unclear, as the evidence is of very low certainty.

Five RCTs included more than 19,000 participants, of whom 51 experienced a stroke, suggesting a 15% increase in stroke risk with increased ALA (RR 1.15, 95% CI 0.66 to 2.01; I2 = 0%; Analysis 4.53). Sensitivity analyses removing trials not at low summary risk of bias left three trials with 27 stroke events and no suggestion of effect (Analysis 4.55). Using a fixed‐effect model suggested a 23% increased risk of stroke (Analysis 4.54), while removing trials at high risk of bias due to compliance suggested a 15% reduction in stroke risk, larger trials suggested a 15% greater stroke risk (Analysis 4.56).

4.53. Analysis.

4.53

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 53 Stroke (overall) ‐ ALA.

4.55. Analysis.

4.55

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 55 Stroke ‐ ALA ‐ SA by summary risk of bias.

4.54. Analysis.

4.54

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 54 Stroke ‐ ALA ‐ SA fixed‐effect.

4.56. Analysis.

4.56

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 56 Stroke ‐ ALA ‐ SA by compliance and study size.

Subgrouping by ALA dose, trial duration, replacement, intervention type, statin use, or primary or secondary prevention did not result in significant differences between subgroups (Analysis 4.57; Analysis 4.58; Analysis 4.59; Analysis 4.60; Analysis 4.61; Analysis 4.62). When examining data reported by type of stroke, only three trials reported on 28 ischaemic strokes, with no clear effects, and no trials reported haemorrhagic stroke (Analysis 4.63). Meta‐regression did not suggest any relationship between ALA dose and risk of stroke (P = 0.73; Table 9).

4.57. Analysis.

4.57

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 57 Stroke ‐ ALA ‐ subgroup by dose.

4.58. Analysis.

4.58

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 58 Stroke ‐ ALA ‐ subgroup by replacement.

4.59. Analysis.

4.59

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 59 Stroke ‐ ALA ‐ subgroup by intervention type.

4.60. Analysis.

4.60

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 60 Stroke ‐ ALA ‐ subgroup by duration.

4.61. Analysis.

4.61

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 61 Stroke ‐ ALA ‐ subgroup by primary or secondary prevention.

4.62. Analysis.

4.62

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 62 Stroke ‐ ALA ‐ subgroup by statin use.

4.63. Analysis.

4.63

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 63 Stroke ‐ ALA ‐ subgroup by stroke type.

The effect of ALA on stroke is unclear as the evidence is of very low certainty (downgraded twice for risk of bias and once for imprecision).

Arrhythmia (ALA)

Moderate‐certainty evidence suggested that ALA intake probably slightly reduces the risk of arrhythmias (NNTB 91).

Two trials reported effects of ALA on arrhythmia, with 173 arrhythmias in 4912 participants, suggesting a 27% reduction in arrhythmia but with wide confidence intervals (RR 0.73, 95% CI 0.55 to 0.97, I2 = 0%, Analysis 4.64). All sensitivity analyses suggested that ALA reduced arrhythmia (Analysis 4.65; Analysis 4.66; Analysis 4.67). There was no suggestion of a dose‐response relationship between ALA and arrhythmia risk in meta‐regression (P = 0.45; Table 10).

4.64. Analysis.

4.64

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 64 Arrythmia (overall) ‐ ALA.

4.65. Analysis.

4.65

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 65 Arrythmia ‐ ALA ‐ SA fixed‐effect.

4.66. Analysis.

4.66

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 66 Arrhythmia ‐ ALA ‐ SA by summary risk of bias.

4.67. Analysis.

4.67

Comparison 4 High vs low ALA omega‐3 fat (primary outcomes), Outcome 67 Arrythmia ‐ ALA ‐ SA by compliance and study size.

The NNTB was 91 (95% CI 56 to 1000) so 91 people would need to increase their ALA intake for one person to avoid experiencing arrhythmia. With only two trials it was not possible to estimate NNTB by primary or secondary prevention. GRADE assessment suggested that ALA intake probably slightly reduces the risk of arrhythmias (moderate‐certainty evidence, downgraded once for imprecision).

Secondary outcomes

See Table 3 for a summary of our evidence on effects of LCn3 fats and ALA on serum lipids and measures of adiposity.

Effects of long‐chain omega‐3 fats (EPA, DHA and DPA) on secondary health outcomes

We did not carry out sensitivity analyses or subgrouping on secondary outcomes, except for adiposity and lipids, which were key outcomes. We did carry out some post hoc sensitivity analyses to further assess effects of LCn3 on myocardial infarction, to ascertain whether the suggested protection was stable to sensitivity analyses.

Major adverse cerebrovascular or cardiovascular events (LCn3)

Five trials reported on major adverse cerebrovascular or cardiovascular events (MACCEs) in more than 34,000 participants, 4232 of whom suffered from a MACCE, suggesting little or no effect of LCn3 fats (RR 1.03, 95% CI 0.97 to 1.09; I2 = 0%; Analysis 2.1).

2.1. Analysis.

2.1

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 1 MACCEs ‐ LCn3.

Myocardial infarction (LCn3)

Twenty‐seven trials (> 133,000 participants) reported on total (fatal and non‐fatal) myocardial infarction. Meta‐analyses suggested that increasing LCn3 fats resulted in a reduction in total myocardial infarction (RR 0.88, 95% CI 0.81 to 0.96; I2 = 25%; 3992 people experiencing at least one myocardial infarction; Analysis 2.2). This was confirmed in fixed‐effect analysis (Analysis 2.3), sensitivity analyses limited to trials without compliance problems and to those that randomised at least 100 participants (Analysis 2.5), but analyses limited to trials at low summary risk of bias suggested little or no effect of LCn3 on myocardial infarction (RR 0.93, 95% CI 0.83 to 1.05; I2 = 20%; > 71,000 participants in 13 trials reporting on 1885 people experiencing at least one myocardial infarction; Analysis 2.4). This undermines a true protective effect of LCn3 on myocardial infarction.

2.2. Analysis.

2.2

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 2 Myocardial infarction (overall) ‐ LCn3.

2.3. Analysis.

2.3

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 3 Total MI ‐ LCn3 ‐ SA fixed effects.

2.5. Analysis.

2.5

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 5 Total MI ‐ LCn3 ‐ SA by compliance and study size.

2.4. Analysis.

2.4

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 4 Total MI ‐ LCn3 ‐ SA by summary risk of bias.

We ran subgroup analyses by fatality at the request of the WHO NUGAG Subgroup on Diet and Health, finding no significant difference between fatal and non‐fatal myocardial infarction subgroups (P = 0.20; Analysis 2.6).

2.6. Analysis.

2.6

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 6 Total MI ‐ LCn3 ‐ subgroup by fatality.

Sudden cardiac death (LCn3)

There was little or no effect of LCn3 fats on sudden cardiac death (RR 0.93, 95% CI 0.77 to 1.11; I2 = 41%; 1422 deaths, 14 trials in > 73,000 people; Analysis 2.7).

2.7. Analysis.

2.7

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 7 Sudden cardiac death (overall) ‐ LCn3.

Angina (LCn3)

Meta‐analysis of 13 trials involving more than 48,000 participants, 2829 of whom reported new or worsening angina, suggested little or no effect of increasing LCn3 fats (RR 0.99, 95% CI 0.92 to 1.06; I2 = 0%; Analysis 2.8).

2.8. Analysis.

2.8

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 8 Angina ‐ LCn3.

Heart failure (LCn3)

Meta‐analysis suggested little or no effect of LCn3 fatty acids on heart failure diagnosis in 17 trials with 4651 people experiencing events (RR 0.94, 95% CI 0.87 to 1.02; I2 = 23%; Analysis 2.9). Sensitivity analysis limiting to the seven trials at low summary risk of bias also suggested little or no effect (RR 0.97, 95% CI 0.89 to 1.05; I2 = 0%; 2017 participants experiencing heart failure). We concluded that there was little or no effect of LCn3 on risk of heart failure.

2.9. Analysis.

2.9

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 9 Heart failure ‐ LCn3.

Revascularisation (LCn3)

Meta‐analysis suggested little or no effect of LCn3 fats on revascularisation, all types combined (RR 0.93, 95% CI 0.86 to 1.00; I2 = 56%; 10,793 participants experiencing revascularisation; Analysis 2.10).

2.10. Analysis.

2.10

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 10 Revascularisation ‐ LCn3.

Peripheral arterial disease (LCn3)

Meta‐analysis suggested that LCn3 had little or no effect on risk of peripheral arterial disease (RR 0.95, 95% CI 0.76 to 1.18; I2 = 0%; 6 trials, 325 events in > 57,000 participants; Analysis 2.11). Sensitivity analysis, limiting to trials at low summary risk of bias, suggested that increasing LCn3 increased the risk of peripheral arterial disease by 10% (Analysis 2.13), other sensitivity analyses all suggested little or no effect (Analysis 2.12; Analysis 2.14). The data suggest that there is little or no effect of LCn3 on risk of peripheral arterial disease.

2.11. Analysis.

2.11

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 11 Peripheral arterial disease ‐ LCn3.

2.13. Analysis.

2.13

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 13 PAD ‐ LCn3 ‐ SA by summary risk of bias.

2.12. Analysis.

2.12

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 12 PAD ‐ LCn3 ‐ SA fixed effects.

2.14. Analysis.

2.14

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 14 PAD ‐ LCn3 ‐ SA compliance and study size.

Acute coronary syndrome (LCn3)

There were limited data on effects of increasing LCn3 fats on acute coronary syndrome (RR 1.19, 95% CI 0.71 to 2.00; I2 = 0%; 2 trials, 55 events in > 2000 participants; Analysis 2.15).

2.15. Analysis.

2.15

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 15 Acute coronary syndrome ‐ LCn3.

Body weight, body mass index (BMI) and other measures of adiposity (LCn3)
Body weight

High‐certainty evidence shows that increasing LCn3 intake makes little or no difference to body weight.

Fourteen trials reported on the effect of increasing LCn3 on body weight and were included in meta‐analysis, suggesting little or no effect in 17,000 participants (mean difference (MD) 0.00 kg, 95% CI −0.69 to 0.70; I2 = 42%; Analysis 2.16). Sensitivity analysis limited to trials at low summary risk of bias, low risk from compliance, larger trials or fixed‐effect analysis (not shown) did not alter this lack of effect (Analysis 2.17; Analysis 2.18; Analysis 2.19).

2.16. Analysis.

2.16

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 16 Body weight, kg ‐ LCn3.

2.17. Analysis.

2.17

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 17 Weight, kg ‐ LCn3 ‐ SA fixed effects.

2.18. Analysis.

2.18

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 18 Weight, kg ‐ LCn3 ‐ SA by summary risk of bias.

2.19. Analysis.

2.19

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 19 Weight, kg ‐ LCn3 ‐ SA by compliance and study size.

Subgroup analysis by intervention type, primary or secondary prevention, statin use and trial duration did not suggest important differences between subgroups (Analysis 2.22; Analysis 2.23; Analysis 2.24; Analysis 2.25). There was a marginally significant difference between dose subgroups (P = 0.05; Analysis 2.20) and increased body weight when participants received very high LCn3 doses (> 4.4 g/d LCn3, MD 1.51 kg, 95% CI 0.28 to 2.75; I2 = 0%; 2 trials, 261 participants; Analysis 2.20). Subgrouping by replacement suggested differences between subgroups (P = 0.001; Analysis 2.21), with reduced body weight when LCn3 replaced saturated fatty acids or carbohydrates but increased weight when LCn3 replaced nil or low LCn3 (Analysis 2.21).

2.22. Analysis.

2.22

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 22 Weight, kg ‐ LCn3 ‐ subgroup by intervention type.

2.23. Analysis.

2.23

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 23 Weight, kg ‐ LCn3 ‐ subgroup by duration.

2.24. Analysis.

2.24

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 24 Weight, kg ‐ LCn3 ‐ subgroup by primary or secondary prevention.

2.25. Analysis.

2.25

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 25 Weight, kg ‐ LCn3 ‐ subgroup by statin use.

2.20. Analysis.

2.20

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 20 Weight, kg ‐ LCn3 ‐ subgroup by dose.

2.21. Analysis.

2.21

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 21 Weight, kg ‐ LCn3 ‐ subgroup by replacement.

Several trials clearly measured body weight but did not report it in a useable way (Baldassarre 2006; Caldwell 2011; Deslypere 1992; EPE‐A 2014; MARINA 2011; Nutristroke 2009). Body weight is commonly measured in healthcare settings, so there may be considerably more missing data than this. The funnel plot was not easily interpretable but small differences in effect size between fixed‐effect and random‐effects meta‐analysis supports the possibility of a small amount of small study bias.

GRADE offers high‐certainty evidence that LCn3 intake makes little or no difference to body weight (not downgraded).

Body mass index (BMI)

High‐certainty evidence shows that LCn3 intake makes little or no difference to BMI.

Fifteen trials, 13 of which we included in meta‐analysis, reported on BMI, suggesting little or no effect of LCn3 on BMI (MD 0.06 kg/m2, 95% CI −0.14 to 0.25; I2 = 39%; > 15,000 participants; Analysis 2.26). This lack of effect was also apparent in sensitivity analyses limited to trials at low summary risk of bias (Analysis 2.28), with good compliance or with large trial size (Analysis 2.29), as well as fixed‐effect analysis (Analysis 2.27). Subgroup analyses by primary or secondary prevention, intervention type, statin use and trial duration did not suggest important differences between subgroups (Analysis 2.32; Analysis 2.33; Analysis 2.34; Analysis 2.35). There were significant differences between subgroups when subgrouped by replacement, suggesting lower BMI when LCn3 replaced saturated fatty acids or carbohydrate, and increased BMI when LCn3 replaced nil or low LCn3 (P = 0.02; Analysis 2.31). There was also a suggestion that increasing LCn3 increased BMI when the dose was above 2.4 g/d (Analysis 2.30).

2.26. Analysis.

2.26

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 26 Body mass index, kg/m² ‐ LCn3.

2.28. Analysis.

2.28

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 28 BMI, kg/m²‐ LCn3 ‐ SA by summary risk of bias.

2.29. Analysis.

2.29

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 29 BMI, kg/m²‐ LCn3 ‐ SA by compliance and study size.

2.27. Analysis.

2.27

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 27 BMI, kg/m² ‐ LCn3 ‐ SA fixed effects.

2.32. Analysis.

2.32

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 32 BMI, kg/m² ‐ LCn3 ‐ subgroup by intervention type.

2.33. Analysis.

2.33

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 33 BMI, kg/m² ‐ LCn3 ‐ subgroup by duration.

2.34. Analysis.

2.34

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 34 BMI, kg/m² ‐ LCn3 ‐ subgroup by primary or secondary prevention.

2.35. Analysis.

2.35

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 35 BMI, kg/m² ‐ LCn3 ‐ subgroup by statin use.

2.31. Analysis.

2.31

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 31 BMI, kg/m² ‐ LCn3 ‐ subgroup by replacement.

2.30. Analysis.

2.30

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 30 BMI, kg/m² ‐ LCn3 ‐ subgroup by dose.

Several trials clearly measured BMI but did not report it in a useable way (Caldwell 2011; EPE‐A 2014; Nutristroke 2009; Ramirez‐Ramirez 2013; Sofi 2010), suggesting that missing data may be an issue with this outcome. The funnel plot was not interpretable, but effect sizes from random‐effects and fixed‐effect meta‐analysis were similar suggesting minimal small study bias.

GRADE offers high‐certainty evidence that LCn3 intake makes little or no difference to BMI (not downgraded).

Other measures of adiposity

Few trials reported on other measures of adiposity (percentage body fat, percentage visceral fat, waist circumference, waist/hip ratio, abdominal circumference and hip circumference) with some suggesting higher adiposity and some lower adiposity in groups with more LCn3 (Analysis 2.36).

2.36. Analysis.

2.36

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 36 Other measures of adiposity ‐ LCn3.

Serum lipids (LCn3)

Several trials clearly measured lipids but did not report them in a way that we could include in our meta‐analyses. These included Baldassarre 2006, Gill 2012,Ramirez‐Ramirez 2013 and Reed 2014. Further trials assessed but did not report triglycerides (Ahn 2016; Caldwell 2011; Franzen 1993; Rossing 1996), or HDL and LDL cholesterol (Franzen 1993). Because of the volume of potentially missing data, small study bias may potentially bias these outcomes.

Serum total cholesterol

High‐certainty evidence shows that LCn3 intake makes little or no difference to serum total cholesterol.

Thirty trials provided data on long‐term effects of LCn3 fats on serum total cholesterol, suggesting little or no effect in more than 38,000 participants (MD −0.01 mmol/L, 95% CI −0.05 to 0.03, I2 = 14%, Analysis 2.37). Sensitivity analyses using a fixed‐effect model, or limited to trials at low summary risk of bias, low risk of compliance issues, and larger trials also suggested little or no effect of LCn3 on serum total cholesterol (Analysis 2.38; Analysis 2.39; Analysis 2.40). Subgrouping by duration, primary or secondary prevention and statin use did not suggest any differential effects of LCn3 (Analysis 2.44; Analysis 2.45; Analysis 2.46). There were almost significant differences between subgroups by dose but no logical sequence suggesting no true dose‐response effect (P = 0.05; Analysis 2.41). There were also subgroup differences for replacement and intervention type but there were no subgroups with changes of at least 5% of baseline (Analysis 2.42; Analysis 2.43).

2.37. Analysis.

2.37

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 37 Total cholesterol, serum, mmoL/L ‐ LCn3.

2.38. Analysis.

2.38

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 38 TC, mmoL/L ‐ LCn3 ‐ SA fixed effects.

2.39. Analysis.

2.39

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 39 TC, mmoL/L ‐ LCn3 ‐ SA by summary risk of bias.

2.40. Analysis.

2.40

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 40 TC, mmoL/L ‐ LCn3 ‐ SA by compliance and study size.

2.44. Analysis.

2.44

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 44 TC, mmoL/L ‐ LCn3 ‐ subgroup by duration.

2.45. Analysis.

2.45

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 45 TC, mmoL/L ‐ LCn3 ‐ subgroup by primary or secondary prevention.

2.46. Analysis.

2.46

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 46 TC, mmoL/L ‐ LCn3 ‐ subgroup by statin use.

2.41. Analysis.

2.41

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 41 TC, mmoL/L ‐ LCn3 ‐ subgroup by dose.

2.42. Analysis.

2.42

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 42 TC, mmoL/L ‐ LCn3 ‐ subgroup by replacement.

2.43. Analysis.

2.43

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 43 TC, mmoL/L ‐ LCn3 ‐ subgroup by intervention type.

GRADE assessment suggests high‐certainty evidence that LCn3 intake makes little or no difference to serum total cholesterol (not downgraded).

Serum triglycerides

High‐certainty evidence suggests that increasing LCn3 intake reduces serum triglycerides, in a dose‐dependent manner, by around 15%.

LCn3 fats significantly reduced serum triglycerides in more than 43,000 participants in 23 trials (MD −0.24 mmol/L, 95% CI −0.31 to −0.16; I2 = 48%; Analysis 2.47). This effect was not lost in sensitivity analysis excluding trials at moderate to high risk of bias, those without clear compliance or small trials, or using fixed‐effect analysis (Analysis 2.48; Analysis 2.49; Analysis 2.50). Subgrouping suggested that the reduction of serum triglycerides did not differ between subgroups by primary or secondary prevention, statin use, replacement, intervention type or trial duration (Analysis 2.52; Analysis 2.53; Analysis 2.54; Analysis 2.55; Analysis 2.56). There was a suggestion of a dose‐response relationship with greater reductions in triglycerides at higher LCn3 doses, with significant differences between subgroups (P = 0.04; Analysis 2.51).

2.47. Analysis.

2.47

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 47 Triglycerides, fasting, serum, mmoL/L ‐ LCn3.

2.48. Analysis.

2.48

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 48 TG, fasting, mmoL/L ‐ LCn3 ‐ SA fixed effects.

2.49. Analysis.

2.49

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 49 TG, fasting, mmoL/L ‐ LCn3 ‐ SA by summary risk of bias.

2.50. Analysis.

2.50

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 50 TG, fasting, mmoL/L ‐ LCn3 ‐ SA by compliance and study size.

2.52. Analysis.

2.52

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 52 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by replacement.

2.53. Analysis.

2.53

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 53 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by intervention type.

2.54. Analysis.

2.54

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 54 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by duration.

2.55. Analysis.

2.55

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 55 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by primary or secondary prevention.

2.56. Analysis.

2.56

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 56 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by statin use.

2.51. Analysis.

2.51

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 51 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by dose.

GRADE assessment suggests high‐certainty evidence that increasing LCn3 intake reduces serum triglycerides in a dose‐dependent manner (not downgraded).

HDL cholesterol

High‐certainty evidence suggests that LCn3 intake has little or no effect on HDL cholesterol.

Thirty trials including more than 46,000 participants suggested little or no effect (an increase of < 5%) in serum HDL cholesterol with increased LCn3 (MD 0.03 mmol/L, 95% CI 0.01 to 0.05; P = 0.005, I2 = 50%; Analysis 2.57). The finding of little or no effect was retained in all sensitivity analyses (Analysis 2.58; Analysis 2.59; Analysis 2.60). There were no significant differences between subgroups in any analysis and no suggestion of a dose‐response relationship (Analysis 2.61; Analysis 2.62; Analysis 2.63; Analysis 2.64; Analysis 2.65; Analysis 2.66).

2.57. Analysis.

2.57

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 57 High‐density lipoprotein, serum, mmoL/L ‐ LCn3.

2.58. Analysis.

2.58

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 58 HDL, mmoL/L ‐ LCn3 ‐ SA fixed effects.

2.59. Analysis.

2.59

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 59 HDL, mmoL/L ‐ LCn3 ‐ SA by summary risk of bias.

2.60. Analysis.

2.60

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 60 HDL, mmoL/L ‐ LCn3 ‐ SA by compliance and study size.

2.61. Analysis.

2.61

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 61 HDL, mmoL/L ‐ LCn3 ‐ subgroup by dose.

2.62. Analysis.

2.62

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 62 HDL, mmoL/L ‐ LCn3 ‐ subgroup by replacement.

2.63. Analysis.

2.63

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 63 HDL, mmoL/L ‐ LCn3 ‐ subgroup by intervention type.

2.64. Analysis.

2.64

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 64 HDL, mmoL/L ‐ LCn3 ‐ subgroup by duration.

2.65. Analysis.

2.65

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 65 HDL, mmoL/L ‐ LCn3 ‐ subgroup by primary or secondary prevention.

2.66. Analysis.

2.66

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 66 HDL, mmoL/L ‐ LCn3 ‐ subgroup by statin use.

GRADE assessment suggests high‐certainty evidence that LCn3 intake has little or no effect on HDL cholesterol (not downgraded).

LDL cholesterol

GRADE assessment suggests high‐certainty evidence that LCn3 intake makes little or no difference to LDL cholesterol.

There was little or no effect of increasing LCn3 on serum LDL cholesterol in over 43,000 participants from 25 trials (MD 0.01 mmol/L, 95% CI −0.01 to 0.03; I2 = 0%; Analysis 2.67). This lack of effect did not alter in any sensitivity analysis (Analysis 2.68; Analysis 2.69; Analysis 2.70). We saw no statistically significant differences between subgroups except for with regard to statin use, where there was a less than 5% increase in LDL cholesterol in nine trials where statin use was low (Analysis 2.71; Analysis 2.72; Analysis 2.73; Analysis 2.74; Analysis 2.75; Analysis 2.76).

2.67. Analysis.

2.67

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 67 Low‐density lipoprotein, serum, mmoL/L ‐ LCn3.

2.68. Analysis.

2.68

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 68 LDL, mmoL/L ‐ LCn3 ‐ SA fixed effects.

2.69. Analysis.

2.69

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 69 LDL, mmoL/L ‐ LCn3 ‐ SA by summary risk of bias.

2.70. Analysis.

2.70

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 70 LDL, mmoL/L ‐ LCn3 ‐ SA by compliance and study size.

2.71. Analysis.

2.71

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 71 LDL, mmoL/L ‐ LCn3 ‐ subgroup by dose.

2.72. Analysis.

2.72

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 72 LDL, mmoL/L ‐ LCn3 ‐ subgroup by replacement.

2.73. Analysis.

2.73

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 73 LDL, mmoL/L ‐ LCn3 ‐ subgroup by intervention type.

2.74. Analysis.

2.74

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 74 LDL, mmoL/L ‐ LCn3 ‐ subgroup by duration.

2.75. Analysis.

2.75

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 75 LDL, mmoL/L ‐ LCn3 ‐ subgroup by primary or secondary prevention.

2.76. Analysis.

2.76

Comparison 2 High vs low LCn3 omega‐3 fats (secondary outcomes), Outcome 76 LDL, mmoL/L ‐ LCn3 ‐ subgroup by statin use.

GRADE assessment provides high‐certainty evidence that LCn3 intake makes little or no difference to LDL cholesterol (not downgraded).

Effects of ALA on secondary health outcomes

We did not plan any sensitivity or subgroup analyses on secondary outcomes, except for the key outcomes of adiposity and lipids. As fewer than 10 ALA trials were available for these outcomes, we carried out only sensitivity analyses.

Major adverse cerebrovascular or cardiovascular events (ALA)

One trial reported on MACCEs in 110 participants, nine of whom experienced an event. There were insufficient data to suggest any effect of ALA on MACCEs (RR 1.12, 95% CI 0.32 to 3.95; Analysis 5.1).

5.1. Analysis.

5.1

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 1 MACCEs ‐ ALA.

Myocardial infarction (ALA)

Three trials reported that 333 out of more than 18,000 participants experienced a fatal or non‐fatal myocardial infarction, suggesting little or no effect of ALA on myocardial infarction (RR 1.00, 95% CI 0.76 to 1.32; I2 = 26%; Analysis 5.2). We carried out subgroup analyses by fatality at the request of the WHO NUGAG Subgroup on Diet and Health, and these suggested no significant differences between fatal and non‐fatal myocardial infarction subgroups (P = 0.36; Analysis 5.3).

5.2. Analysis.

5.2

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 2 Myocardial infarction (overall) ‐ ALA.

5.3. Analysis.

5.3

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 3 Total MI ‐ ALA ‐ subgroup by fatality.

Sudden cardiac death (ALA)

No trials assessed effects of ALA on sudden cardiac death.

Angina (ALA)

Two trials assessed the effects of increasing ALA on diagnosis of new or worsening angina (39 of > 13,000 participants experienced this). There were insufficient data to suggest any effect of ALA on angina (RR 1.41, 95% CI 0.75 to 2.64; I2 = 0%; Analysis 5.4).

5.4. Analysis.

5.4

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 4 Angina ‐ ALA.

Heart failure (ALA)

No trials assessed effects of ALA on heart failure.

Revascularisation (ALA)

Only one trial (3 events in 266 participants) reported on the effects of increased ALA on revascularisation (RR 0.72, 95% CI 0.07 to 7.84; 3 events; Analysis 5.5) or CABG specifically (RR 0.29, 95% CI 0.01 to 5.93; 2 events; Analysis 5.5). There were insufficient data to suggest any effect of ALA on revascularisation.

5.5. Analysis.

5.5

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 5 Revascularisation ‐ ALA.

Peripheral arterial disease (ALA)

Meta‐analysis suggested no clear effect of ALA on peripheral arterial disease in a single trial (RR 0.25, 95% CI 0.05 to 1.17; 10 of the > 13,000 participants experienced peripheral arterial disease; Analysis 5.6). There were insufficient data to suggest any effect of ALA on the outcome.

5.6. Analysis.

5.6

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 6 Peripheral arterial disease ‐ ALA.

Acute coronary syndrome (ALA)

No trials assessed effects of ALA on acute coronary syndrome.

Body weight, body mass index (BMI) and other measures of adiposity (ALA)

The effect of LCn3 intake on body weight and BMI is unclear as the evidence is of very low certainty.

Four trials reported on the effect of ALA on body weight in 664 participants, suggesting some weight reduction in those taking more ALA but with extremely high heterogeneity (MD −1.49 kg, 95% CI −4.17 to 1.18; I2 = 73%; Analysis 5.7). Sensitivity analysis using fixed‐effect meta‐analysis suggested little or no effect of ALA on body weight (Analysis 5.8), while no trials were at low summary risk of bias (Analysis 5.9). Retaining only trials at low risk for compliance bias or only larger trials suggested weight reduction with ALA (Analysis 5.10). There were no significant differences between subgroups by intervention type, dose, duration, replacement, statin use, or primary or secondary prevention of cardiovascular disease (Analysis 5.11; Analysis 5.12; Analysis 5.13; Analysis 5.14; Analysis 5.15; Analysis 5.16). GRADE assessment suggests that the effect of ALA intake on body weight is unclear, as the evidence is of very low certainty (downgraded once each for risk of bias, inconsistency, publication bias and imprecision).

5.7. Analysis.

5.7

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 7 Body weight, kg ‐ ALA.

5.8. Analysis.

5.8

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 8 Weight, kg ‐ ALA ‐ sensitivity analysis (SA) fixed‐effect.

5.9. Analysis.

5.9

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 9 Weight, kg ‐ ALA ‐ SA by summary risk of bias.

5.10. Analysis.

5.10

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 10 Weight, kg ‐ ALA ‐ SA by compliance and study size.

5.11. Analysis.

5.11

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 11 Weight, kg ‐ ALA ‐ subgroup by dose.

5.12. Analysis.

5.12

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 12 Weight, kg ‐ ALA ‐ subgroup by intervention type.

5.13. Analysis.

5.13

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 13 Weight, kg ‐ ALA ‐ subgroup by replacement.

5.14. Analysis.

5.14

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 14 Weight, kg ‐ ALA ‐ subgroup by duration.

5.15. Analysis.

5.15

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 15 Weight, kg ‐ ALA ‐ subgroup by statin use.

5.16. Analysis.

5.16

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 16 Weight, kg ‐ ALA ‐ subgroup by primary or secondary prevention.

Three trials reported on BMI, suggesting a reduction in BMI with increased ALA (MD −0.42 kg/m2, 95% CI −1.53 to 0.69; I2 = 65%; 1581 participants; Analysis 5.17), again with high heterogeneity. Sensitivity analyses using fixed‐effect analysis or only retaining trials at low summary risk of bias suggested a small increase in BMI with ALA (Analysis 5.18; Analysis 5.19), while limiting to trials at low risk of compliance bias or eliminating smaller trials suggested a small reduction in BMI with increased ALA (Analysis 5.20). There were no statistically significant differences between subgroups differentiated by replacement or statin use (Analysis 5.23; Analysis 5.25), but there were differences by dose; subgrouping by dose suggested greater reduction of BMI in trials giving more ALA (P = 0.03; Analysis 5.21). All included trials gave supplemental foods (Analysis 5.22). There were greater reductions in BMI in shorter trials (P = 0.02; Analysis 5.24) and in primary prevention trials (P = 0.03; Analysis 5.26), but the inclusion of Dodin 2005 in any subgroup tended to differentiate that group from the others. GRADE assessment suggests that the effect of ALA intake on BMI is unclear, as the evidence is of very low certainty (downgraded once each for risk of bias, inconsistency, publication bias and imprecision).

5.17. Analysis.

5.17

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 17 Body mass index, kg/m² ‐ ALA.

5.18. Analysis.

5.18

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 18 BMI, kg/m² ‐ ALA ‐ SA fixed‐effect.

5.19. Analysis.

5.19

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 19 BMI, kg/m² ‐ ALA ‐ SA by summary risk of bias.

5.20. Analysis.

5.20

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 20 BMI, kg/m² ‐ ALA ‐ SA by compliance and study size.

5.23. Analysis.

5.23

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 23 BMI, kg/m² ‐ ALA ‐ subgroup by replacement.

5.25. Analysis.

5.25

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 25 BMI, kg/m² ‐ ALA ‐ subgroup by statin use.

5.21. Analysis.

5.21

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 21 BMI, kg/m² ‐ ALA ‐ subgroup by dose.

5.22. Analysis.

5.22

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 22 BMI, kg/m² ‐ ALA ‐ subgroup by intervention type.

5.24. Analysis.

5.24

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 24 BMI, kg/m² ‐ ALA ‐ subgroup by duration.

5.26. Analysis.

5.26

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 26 BMI, kg/m² ‐ ALA ‐ subgroup by primary or secondary preventionA.

One trial reported on visceral adipose tissue, suggesting no clear effect. Three trials reported on waist circumference. Meta‐analysis of two of these suggested that increasing ALA resulted in reduced weight circumference (MD −1.59 cm, 95% CI −3.10 to −0.07; I2 = 0%; Analysis 5.27). However, the single trial that we could not include in the meta‐analysis due to lack of information on variance suggested effects in the opposite direction. Sensitivity analyses (only retaining trials at low summary risk of bias, not shown) removed all trials.

5.27. Analysis.

5.27

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 27 Other measures of adiposity ‐ ALA.

Serum lipids (ALA)
Serum total cholesterol

Low‐certainty evidence suggests that ALA intake may make little or no difference to serum total cholesterol.

Six trials provided data on the long‐term effects of ALA on serum total cholesterol, suggesting that increased ALA intake has little or no effect on total cholesterol, with high heterogeneity (MD −0.09 mmol/L, 95% CI −0.23 to 0.05; I2 = 63%; in > 2000 participants; Analysis 5.28). The suggestion of little or no effect did not alter in any sensitivity analyses (Analysis 5.29; Analysis 5.30; Analysis 5.31). All trials provided food supplements (Analysis 5.33), but subgroup analyses suggested greater (though still small) reductions in total cholesterol in trials of shorter duration (P = 0.02; Analysis 5.35). Other differences between subgroups resulted from effect groups where ALA replacement or statin use was 'unclear' (Analysis 5.34; Analysis 5.36), or there were no differences (Analysis 5.32; Analysis 5.37). GRADE assessment suggests low‐certainty evidence that ALA intake may make little or no difference to serum total cholesterol (downgraded once each for imprecision and inconsistency).

5.28. Analysis.

5.28

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 28 Total cholesterol, serum, mmoL/L ‐ ALA.

5.29. Analysis.

5.29

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 29 TC, mmoL/L ‐ ALA ‐ SA fixed‐effect.

5.30. Analysis.

5.30

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 30 TC, mmoL/L ‐ ALA ‐ SA by summary risk of bias.

5.31. Analysis.

5.31

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 31 TC, mmoL/L ‐ ALA ‐ SA by compliance and study size.

5.33. Analysis.

5.33

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 33 TC, mmoL/L ‐ ALA ‐ subgroup by intervention type.

5.35. Analysis.

5.35

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 35 TC, mmoL/L ‐ ALA ‐ subgroup by duration.

5.34. Analysis.

5.34

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 34 TC, mmoL/L ‐ ALA ‐ subgroup by replacement.

5.36. Analysis.

5.36

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 36 TC, mmoL/L ‐ ALA ‐ subgroup by statin use.

5.32. Analysis.

5.32

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 32 TC, mmoL/L ‐ ALA ‐ subgroup by dose.

5.37. Analysis.

5.37

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 37 TC, mmoL/L ‐ ALA ‐ subgroup by primary or secondary prevention.

Serum triglycerides

Moderate‐certainty evidence suggests that increasing ALA intake probably makes little or no difference to serum triglycerides.

There was little or no effect of ALA on serum triglycerides in 1776 participants in six trials (MD −0.03 mmol/L, 95% CI −0.11 to 0.05; I2 = 0%; Analysis 5.38). There was little or no effect of ALA in sensitivity analysis removing trials of moderate to high risk of bias (Analysis 5.40), in fixed‐effect meta‐analysis (Analysis 5.39), or limiting by compliance bias or trial size (Analysis 5.41). Subgrouping suggested no important differential effects by dose, duration, replacement, intervention type, statin use, or primary or secondary prevention (Analysis 5.42; Analysis 5.43; Analysis 5.44; Analysis 5.45; Analysis 5.46; Analysis 5.47). GRADE assessment suggests moderate‐certainty evidence that ALA intake probably makes little or no difference to serum triglycerides (downgraded once for imprecision).

5.38. Analysis.

5.38

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 38 Triglycerides, fasting, serum, mmoL/L ‐ ALA.

5.40. Analysis.

5.40

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 40 TG, fasting, mmoL/L‐ ALA ‐ SA by summary risk of bias.

5.39. Analysis.

5.39

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 39 TG, fasting, mmoL/L ‐ ALA ‐ SA fixed‐effect.

5.41. Analysis.

5.41

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 41 TG, fasting, mmoL/L‐ ALA ‐ SA by compliance and study size.

5.42. Analysis.

5.42

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 42 TG, fasting, mmoL/L ‐ ALA ‐ subgroup by dose.

5.43. Analysis.

5.43

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 43 TG, fasting, mmoL/L‐ ALA ‐ subgroup by intervention type.

5.44. Analysis.

5.44

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 44 TG, fasting, mmoL/L‐AL ‐ subgroup by replacement.

5.45. Analysis.

5.45

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 45 TG, fasting, mmoL/L‐ ALA ‐ subgroup by duration.

5.46. Analysis.

5.46

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 46 TG, fasting, mmoL/L ‐ ALA ‐ subgroup by statin use.

5.47. Analysis.

5.47

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 47 TG, fasting, mmoL/L‐ ALA ‐ subgroup by primary or secondary prevention.

HDL cholesterol

Moderate‐certainty evidence suggests that ALA probably has little or no effect on HDL cholesterol.

There was little or no effect of ALA on HDL cholesterol in 1776 participants of six trials (MD −0.02 mmol/L, 95% CI −0.08 to 0.03; I2 = 53%; Analysis 5.48), or in any sensitivity analyses (Analysis 5.49; Analysis 5.50; Analysis 5.51). A further trial, WAHA 2016, also measured HDL but did not provide data in a useable format for meta‐analysis. There were no statistically significant differences between subgroups (Analysis 5.52; Analysis 5.53; Analysis 5.54; Analysis 5.55; Analysis 5.56; Analysis 5.57). GRADE assessment suggests moderate‐certainty evidence that ALA probably has little or no effect on HDL cholesterol (downgraded once for imprecision).

5.48. Analysis.

5.48

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 48 High‐density lipoprotein, serum, mmoL/L ‐ ALA.

5.49. Analysis.

5.49

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 49 HDL, mmoL/L ‐ ALA ‐ SA fixed‐effect.

5.50. Analysis.

5.50

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 50 HDL, mmoL/L ‐ ALA ‐ SA by summary risk of bias.

5.51. Analysis.

5.51

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 51 HDL, mmoL/L ‐ ALA ‐ SA by compliance and study size.

5.52. Analysis.

5.52

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 52 HDL, mmoL/L ‐ ALA ‐ subgroup by dose.

5.53. Analysis.

5.53

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 53 HDL, mmoL/L ‐ ALA ‐ subgroup by intervention type.

5.54. Analysis.

5.54

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 54 HDL, mmoL/L ‐ ALA ‐ subgroup by replacement.

5.55. Analysis.

5.55

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 55 HDL, mmoL/L ‐ ALA ‐ subgroup by duration.

5.56. Analysis.

5.56

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 56 HDL, mmoL/L ‐ ALA ‐ subgroup by statin use.

5.57. Analysis.

5.57

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 57 HDL, mmoL/L ‐ ALA ‐ subgroup by primary or secondary prevention.

LDL cholesterol

Moderate‐certainty evidence suggests that ALA intake probably makes little or no difference to LDL cholesterol.

There was little or no effect of increasing ALA on LDL cholesterol in 2201 participants of seven trials (MD −0.05 mmol/L, 95% CI −0.15 to 0.04; I2 = 46%; Analysis 5.58), with similar effects in all sensitivity analyses (Analysis 5.59; Analysis 5.60; Analysis 5.61). Subgrouping suggested no differences in effect by ALA dose or primary or secondary prevention (Analysis 5.62; Analysis 5.63; Analysis 5.64; Analysis 5.66; Analysis 5.67), but there was a statistically significant difference between trials of longer and shorter duration, though little or no effect in both groups (Analysis 5.65). GRADE assessment suggests moderate‐certainty evidence that ALA intake may make little or no difference to LDL cholesterol (downgraded once for imprecision).

5.58. Analysis.

5.58

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 58 Low‐density lipoprotein, serum, mmoL/L ‐ ALA.

5.59. Analysis.

5.59

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 59 LDL, mmoL/L ‐ ALA ‐ SA fixed‐effect.

5.60. Analysis.

5.60

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 60 LDL, mmoL/L ‐ ALA ‐ SA by summary risk of bias.

5.61. Analysis.

5.61

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 61 LDL, mmoL/L ‐ ALA ‐ SA by compliance and study size.

5.62. Analysis.

5.62

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 62 LDL, mmoL/L ‐ ALA ‐ subgroup by dose.

5.63. Analysis.

5.63

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 63 LDL, mmoL/L ‐ ALA ‐ subgroup by intervention type.

5.64. Analysis.

5.64

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 64 LDL, mmoL/L ‐ ALA ‐ subgroup by replacement.

5.66. Analysis.

5.66

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 66 LDL, mmoL/L ‐ ALA ‐ subgroup by statin use.

5.67. Analysis.

5.67

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 67 LDL, mmoL/L ‐ ALA ‐ subgroup by primary or secondary prevention.

5.65. Analysis.

5.65

Comparison 5 High vs low ALA omega‐3 fat (secondary outcomes), Outcome 65 LDL, mmoL/L ‐ ALA ‐ subgroup by duration.

Tertiary outcomes

Effects of long‐chain omega‐3 fats (EPA, DHA and DPA) on tertiary health outcomes

We extracted these outcomes from trials that we included for other outcomes, so we did not assess them completely or systematically. We did not carry out sensitivity analyses or subgrouping for these outcomes. We are aware of missing data for some of these outcomes, including blood pressure in Ramirez‐Ramirez 2013.

Blood pressure (LCn3)

Seventeen included trials (> 35,000 participants) contributed data on effects of LCn3 fats on blood pressure. Meta‐analysis suggested little or no effect of LCn3 on systolic (MD 0.01 mmHg, 95% CI −0.31 to 0.34; I2 = 0%; Analysis 3.1) or diastolic (MD −0.02 mmHg, 95% CI −0.22 to 0.17; I2 = 0%; Analysis 3.1) blood pressure in trials of at least one year.

3.1. Analysis.

3.1

Comparison 3 High vs low LCn3 omega‐3 fats (tertiary outcomes), Outcome 1 Blood pressure, mmHg ‐ LCn3.

Serious adverse effects (LCn3)

As part of the larger set of reviews, we formally systematically reviewed effects of omega‐3 fats on type 2 diabetes diagnoses, measures of glucose metabolism (Brown 2019), cancers including breast cancer (Hanson 2019), neurocognitive outcomes such as dementia (Brainard 2019), irritable bowel disease (IBD) and inflammatory factors (Thorpe 2017), depression and anxiety (Deane 2019), and functional outcomes (Abdelhamid 2019), so we do not present these outcomes here.

We collected data on the following potentially important health outcomes (Analysis 3.2).

3.2. Analysis.

3.2

Comparison 3 High vs low LCn3 omega‐3 fats (tertiary outcomes), Outcome 2 Serious adverse events ‐ LCn3.

  • Any serious adverse event (RR 1.00, 95% CI 0.94 to 1.06; I2 = 0%; 3 trials, > 9000 participants, 2668 events).

  • Bleeding (RR 1.12, 95% CI 0.91 to 1.37; I2 = 44%; 11 trials, > 80,000 participants, 1324 events). Assessed as being very low‐certainty evidence, so the effect of LCn3 on bleeding is unclear (Table 1).

  • Serious gastrointestinal events (RR 1.34, 95% CI 0.64 to 2.80; I2 22%; 3 trials, 774 participants, 49 events).

  • Pulmonary embolus or DVT (RR 1.15, 95% CI 0.44 to 2.98; I2 = 0%; 5 trials, > 3000 participants, 20 events). Assessed as being very low‐certainty evidence, so the effect of LCn3 on pulmonary embolus or DVT is unclear (Table 1).

  • Progression to advanced age‐related macular degeneration (RR 0.96, 95% CI 0.90 to 1.02; 1 trial, > 4000 participants, 2049 events).

  • Thrombophlebitis: no data identified

  • Urolithiasis: no data identified

Side effects (non‐serious, LCn3)

To assess side effects we collected data on the following potential side effects (Analysis 3.3).

3.3. Analysis.

3.3

Comparison 3 High vs low LCn3 omega‐3 fats (tertiary outcomes), Outcome 3 Side effects ‐ LCn3.

  • Withdrawal due to side effects: the data suggest more participants taking LCn3 fats dropped out because of side effects (RR 1.16, 95% CI 0.99 to 1.36; I2 = 1%; 23 trials, > 16,000 participants, 620 dropouts).

  • Increased abdominal pain or discomfort: data suggest an association with higher LCn3 (RR 1.05, 95% CI 0.91 to 1.20; I2 = 16%; 9 trials, > 41,000 participants, 10,040 events).

  • Diarrhoea: the data suggested an increased risk with increased LCn3 (RR 1.02, 95% CI 0.87 to 1.19; I2 = 49%; 13 trials, > 37,000 participants, 12,303 events).

  • Nausea: risk increased with LCn3 (RR 1.20, 95% CI 0.96 to 1.49; I2 = 54%; 8 trials, > 35,000 participants, 7639 events).

  • Any gastrointestinal side effect: risk also appeared to increase with LCn3, albeit with very high heterogeneity (RR 1.10, 95% CI 0.97 to 1.26; I2 = 74%; 33 trials, > 895,000 participants, 6651 events).

  • Skin problems, including itching or rashes: these were not affected by LCn3 in a meta‐analysis with high heterogeneity (RR 1.11, 95% CI 0.52 to 2.37; I2 = 68%; 9 trials, > 36,000 participants, 293 events).

  • Headache or worsening migraine: there were limited data on this outcome (RR 0.85, 95% CI 0.51 to 1.40; I2 = 0%; 4 trials, 1526 participants, 60 events).

  • Reflux: there were limited data (RR 1.23, 95% CI 0.79 to 1.91; I2 = 32%; 3 trials, > 8000 participants, 282 events).

  • Joint, lumbar and muscle pain: meta‐analysis of data from three trials suggested that LCn3 had little or no effect on such pain (RR 0.95, 95% CI 0.74 to 1.23; > 27,000 participants, 989 people experienced pain).

  • All side effects: there was no suggestion that LCn3 increased or decreased all side effects combined in a meta‐analysis with very high heterogeneity (RR 1.01, 95% CI 0.95 to 1.08; I2 = 79%; 14 trials, > 39,000 participants, 9863 people with at least one side effect).

Dropouts (LCn3)

Included trials reported 6643 dropouts in over 56,000 participants in 35 trials, suggesting no difference in dropout rates between intervention and control arms (RR 0.97, 95% CI 0.90 to 1.04; I2 = 28%; Analysis 3.4).

3.4. Analysis.

3.4

Comparison 3 High vs low LCn3 omega‐3 fats (tertiary outcomes), Outcome 4 Dropouts ‐ LCn3.

Quality of life, economic costs (LCn3)

We found no data on quality‐of‐life outcomes or economic costs.

Effects of ALA on tertiary health outcomes

We extracted these outcomes from trials that we included for other outcomes, so we did not assess them completely or systematically. We did not carry out sensitivity analyses or subgrouping for these outcomes.

Blood pressure (ALA)

Four included trials (1671 participants) contributed data on effects of ALA on blood pressure. Meta‐analysis suggested little or no effect of ALA on systolic (MD −0.87 mmHg, 95% CI −4.48 to 2.75; I2 = 58%; Analysis 6.1) or diastolic (MD −1.42 mmHg, 95% CI −4.40 to 1.57; I2 = 74%; Analysis 6.1) blood pressure in trials of at least one year. The heterogeneity in these results reflect a single trial, FLAX‐PAD 2013, that showed large diastolic and systolic blood pressure effects. The other (larger) trials did not suggest such effects.

6.1. Analysis.

6.1

Comparison 6 High vs low ALA omega‐3 fats (tertiary outcomes), Outcome 1 Blood pressure, mmHg ‐ ALA.

Serious adverse effects (ALA)

As part of the larger set of reviews we formally systematically reviewed effects of omega‐3 fats on type 2 diabetes diagnoses and measures of glucose metabolism (Brown 2019), cancers including breast cancer (Hanson 2019), neurocognitive outcomes such as dementia (Brainard 2019), irritable bowel disease (IBD) and inflammatory factors (Thorpe 2017), depression and anxiety (Deane 2019), and functional outcomes (Abdelhamid 2019), so we do not present these outcomes here.

We collected data on the following potentially important health outcomes (Analysis 6.2).

6.2. Analysis.

6.2

Comparison 6 High vs low ALA omega‐3 fats (tertiary outcomes), Outcome 2 Serious adverse events ‐ ALA.

  • Any serious adverse event: no data identified

  • Bleeding: no data identified

  • Serious gastrointestinal effects: no data identified

  • Pulmonary embolus or DVT: only one event was identified in a single trial, so there were insufficient data to assess effects. GRADE assessment suggested very low‐certainty evidence (downgraded once for risk of bias and twice for imprecision (Table 2).

  • Progression to advanced age‐related macular degeneration: no data identified

  • Thrombophlebitis: there were insufficient data to assess effects (RR 1.59, 95% CI 0.72 to 3.51; 1 trial, > 13,000 participants, 26 events).

  • Urolithiasis: there were insufficient data to assess effects (RR 0.80, 95% CI 0.47 to 1.36; 1 trial, > 13,000 participants, 54 events).

Side effects (non‐serious, ALA)

To assess potential side effects, we collected data on the following (Analysis 6.3).

6.3. Analysis.

6.3

Comparison 6 High vs low ALA omega‐3 fats (tertiary outcomes), Outcome 3 Side effects ‐ ALA.

  • Dropouts due to side effects: data suggested that ALA increased the risk of withdrawal, although there was high heterogeneity (RR 2.10, 95% CI 0.66 to 6.71; I2 = 62%; 5 trials, > 3000 participants, 68 events).

  • Abdominal pain or discomfort: no data identified

  • Diarrhoea: a single trial identified 10 participants with diarrhoea, suggesting a higher risk of diarrhoea with greater ALA intake (RR 3.82, 95% CI 0.82 to 17.88).

  • Nausea: there were insufficient data to assess effects of ALA (RR 6.29, 95% CI 0.33 to 118.93; 1 trial, 110 participants, 3 events).

  • Any gastrointestinal side effect: there were insufficient data to assess effects of ALA (RR 2.06, 95% CI 0.62 to 6.80; I2 = 58%; 4 trials, > 3000 participants, 49 events). The high heterogeneity suggests that gastrointestinal side effects may be collected in different ways in different trials.

  • Skin problems, including itching or rashes: no data identified

  • Headache or worsening migraine: no data identified

  • Reflux: no data identified

  • All side effects combined: no data identified

Dropouts (ALA)

Included trials reported 558 dropouts in over 3000 participants in six trials, suggesting slightly similar dropout rates in participants taking higher and lower ALA (RR 1.08, 95% CI 0.92 to 1.25; I2 = 0%; Analysis 6.4).

6.4. Analysis.

6.4

Comparison 6 High vs low ALA omega‐3 fats (tertiary outcomes), Outcome 4 Dropouts ‐ ALA.

Quality of life, economic costs (ALA)

We found no data on quality of life outcomes or economic costs.

Discussion

Summary of main results

We included 86 randomised controlled trials (162,796 participants), of which 28 were at low summary risk of bias (randomisation, allocation concealment, selection and detection bias all at low risk for supplementation trials; randomisation, allocation concealment and detection bias all at low risk for dietary advice trials). This compares to 79 RCTs including 112,059 participants, of which 25 were at low summary risk of bias in the previous review (Abdelhamid 2018a). Trials of 12 to 88 months' duration included adults at varying levels of cardiovascular risk, mainly in high‐income countries. Most trials assessed LCn3 supplementation with capsules, but some used LCn3‐ or ALA‐rich or enriched foods or dietary advice compared to placebo or usual diet.

Pooled trial results suggested there is probably little or no effect of increasing LCn3 fats on risk of all‐cause mortality, cardiovascular deaths, cardiovascular events, stroke or arrhythmias (moderate‐ and high‐certainty evidence). But in this update we found low‐certainty evidence suggesting that increasing LCn3 intake may reduce coronary heart disease mortality (NNTB 334) and coronary heart disease events (NNTB 167). These are small effects and have not been apparent in the more limited data sets of the past.

Limiting LCn3 analyses to trials at low summary risk of bias moved the effect size towards 1.0 (the null value) for most primary outcomes except coronary heart disease mortality and arrhythmia. We found no suggestion of dose responses in subgrouping but there were suggestions of dose response by LCn3 dose in meta‐regression for cardiovascular disease events and coronary heart disease events. These results apply to supplemental LCn3 intake. We did not see important differences in LCn3 trials between those providing oily fish (dietary source) or EPA/DHA capsules (supplemental source), but as few trials provided whole fish health effects may differ.

Increasing ALA intake suggested moderate‐ and low‐certainty evidence of little or no effect on all‐cause mortality, cardiovascular mortality, coronary heart disease mortality or events, but low‐certainty evidence suggested a small protection from cardiovascular events (NNTB 500) and moderate‐certainty evidence, protection from arrhythmia (NNTB 91). Effects on stroke were unclear. Data were more limited than for LCn3, and there were too few trials for informative funnel plots or subgroup analyses. These suggested that benefits of ALA need to be considered with caution, as effects were small, and few trials addressed the outcomes.

Meta‐analyses suggested little or no effect of increasing LCn3 fats intake on secondary outcomes: major adverse cerebrovascular or cardiovascular events, fatal or non‐fatal myocardial infarction, or both, sudden cardiac death, new or worsening angina, heart failure, revascularisation, peripheral arterial disease or acute coronary syndrome. There were very limited data on effects of ALA on these outcomes.

High‐certainty evidence suggested that increasing LCn3 has little or no effect on measures of adiposity (body weight or BMI), but effects of ALA on measures of adiposity were unclear as the evidence was of very low‐certainty. High‐certainty evidence shows that increasing LCn3 reduces serum triglycerides by ˜15% in a dose‐dependent manner, but moderate‐certainty evidence suggests little or no effect of ALA on triglycerides. High‐certainty evidence showed no effect of increasing LCn3 on total, HDL or LDL cholesterol in these long‐term trials, while moderate‐ and low‐certainty evidence suggested little or no effect of increasing ALA on these lipids. Within the included trials we assessed effects on blood pressure, serious adverse effects, side effects and dropouts. There was no suggestion that blood pressure or risk of adverse events such as bleeding differed by LCn3 or ALA intake. Thus, proposed mechanisms for omega‐3 activity, including lowering of blood pressure, reduced thrombotic tendency and anti‐arrhythmic effects are not apparent in long‐term trials of adult humans, but LCn3 does lower serum triglyceride levels.

The review has provided some answers for its secondary questions.

  • If omega‐3 fatty acids confer protection:

    • does protection occur equally in those at low and at high risk of cardiovascular disease? There is no evidence of differential effects on mortality or cardiovascular health by primary or secondary cardiovascular disease prevention, except in the case of LCn3 and arrhythmia, where there is a suggestion of harm in primary prevention, little or no effect in secondary prevention. However, because the underlying risk of an event is different in primary and secondary prevention the NNTB differs in these two groups. For example, in assessing the effects of LCn3 on coronary heart disease mortality the NNTB is 334 (95% CI 200 to infinity), so 334 people would need to increase their LCn3 intake to prevent one death from coronary heart disease. If we assess NNTB by primary or secondary prevention of cardiovascular disease, people without previous cardiovascular disease (needing primary prevention) have an NNTB of 1000 (95% CI NNTB 334 to NNTH 1000), while those with existing cardiovascular disease have an NNTB of 200 (NNTB 91 to NNTH 500). When assessing effects of LCn3 on coronary heart disease events, the NNTB overall is 167 (95% CI 100 to 500), while the NNTB is 200 for primary prevention (95% CI NNTB 112 to NNTB infinity) and 143 for people with existing cardiovascular disease, secondary prevention (NNTB 143, 95% CI NNTB 91 to NNTH 500).

    • does protection depend on the dose of omega‐3 fats taken per day? We ran subgroup analyses for primary and key outcomes and meta‐regression for primary outcomes but found no evidence of differential effects by LCn3 or ALA dose on any outcomes except LCn3 on serum triglycerides, cardiovascular disease events and coronary heart disease events, where there were statistically significantly greater reductions with higher LCn3 dose.

    • do effects differ between dietary and supplemental omega‐3 sources? We assessed this question by looking for statistically significant differences between subgroups but found no evidence of differential effects by dietary or supplemental LCn3 or ALA sources. However, few of the LCn3 trials advised or gave fish, most gave supplemental fish oils, so our ability to assess effects of eating more oily whole fish are limited.

    • does protection depend on trial summary risk of bias? Some analyses suggested a protective effect of LCn3 fats, but these effects disappeared when analyses were limited to trials at low summary risk of bias. The stronger trials with higher internal validity suggested few or no effects of LCn3 on mortality or most cardiovascular disease outcomes, but effects of LCn3 were greater for coronary heart disease mortality and arrhythmia when limited to the trials at low summary risk of bias. On the other hand, for all‐cause mortality, cardiovascular disease mortality and events and coronary heart disease mortality and events, ALA trials at low summary risk of bias suggested greater protection with higher ALA than in the main analysis (including trials of all levels of summary risk of bias).

  • Is protection or harm stronger with longer trial duration? In subgroup analyses for primary and key outcomes and in meta‐regression for primary outcomes, there was no evidence that longer trials increased the effect of LCn3 or ALA. The exceptions were effects of LCn3 on stroke (where trials of longer duration showed smaller risk ratios, or lower risk) and arrhythmia (where the risk of arrhythmia was higher in longer trials).

  • Are effects of omega‐3 fatty acids dependent on baseline triglyceride levels or diabetic status? In subgroup analyses for primary outcomes we found no evidence that increasing LCn3 had different effects in trials with higher baseline triglycerides or with diabetes or diabetes risk factors, however few trials had raised triglycerides or diabetes at baseline so the assessment was not well powered. The exception was when assessing coronary heart disease events, where the subgroup of three trials with raised triglyceride suggested a greater reduction in coronary heart disease events (Analysis 1.62), this effect relied on REDUCE‐IT 2019. No trials of increased ALA had raised baseline triglycerides or diabetes, so effects could not be assessed for ALA.

Overall completeness and applicability of evidence

We searched very carefully to find all trials relevant to this review and located 86 trials randomising 162,796 participants to higher and lower omega‐3 fats (LCn3 or ALA) for at least 12 months.

To reduce selection bias, we contacted authors of trials that appeared to have randomised appropriate participants to appropriate intervention and comparator but may not have published relevant outcomes (to April 2017). If trial authors had assessed any of our outcomes, we requested data and included the trial. This enabled us to include several additional trials. We also contacted authors of all included trials that randomised at least 100 participants (and most smaller trials to April 2017) to request data on any further outcomes (as well as on methodological issues) that may have been recorded but not reported. We tried to contact 72 of the 86 included trials (all except Baldassarre 2006; HERO 2009; Mita 2007; Nutristroke 2009; Özaydin 2011; Shinto 2014; Sofi 2010; and the newly added trials ASCEND 2018; Broutset 2007; DREAM Asbell 2018; ENRGISE 2018; HEARTS 2017; REDUCE‐IT 2019; VITAL 2019). This allowed us to collect useful additional data on outcomes such as deaths and cardiovascular events that we would not have had access to otherwise. For all trials we carefully searched out and data extracted trials registry entries, protocols, supplementary materials, letters, conference abstracts and additional publications to help us locate additional data.

Most of our newly included trials were previously ongoing. We have detailed 25 remaining ongoing trials, of which seven are newly ongoing, that appear to be unpublished at the time of writing (Characteristics of ongoing studies). We have labelled these trials as ongoing, although some appear overdue for publication, and their status is unclear – they may constitute missing data. We tried to contact authors of all 'overdue' ongoing trials, and some stated that publications are forthcoming; others did not reply. We suspect that if trial authors have not published outcomes, it is likely that any protective health effects did not reach statistical significance. Given that existing trials suggest no effects of omega‐3 fats on cardiovascular health outcomes, any missing data may not affect outcomes greatly; however, for completeness we would prefer to include all available data.

Post hoc, we followed advice to assess differences in effects between EPA and DHA within the review. However, most LCn3 trials provided or advised changes resulting in increased intakes of both EPA and DHA (as in natural fish oil), though in different ratios. Only three trials provided data on DHA only (ADCS 2010; Berson 2004; Zhang 2017), and six provided data on EPA only (Doi 2014; JELIS 2007; Mita 2007; Nye 1990; Puri 2005; REDUCE‐IT 2019). Unfortunately, for any single outcome, only two or three of these trials were represented, so our ability to assess differential effects of the DHA‐only and EPA‐only interventions was very limited, and we have not presented these analyses or attempted to draw any information from them.

Quality of the evidence

Figure 2 displays risk of bias of included trials. Of the 86 RCTs, 28 were at low summary risk of bias (at low risk of selection bias, performance bias and detection bias, plus low risk of performance bias in supplemental trials). We assessed the validity of evidence in meta‐analyses by running sensitivity analyses that removed trials not at low summary risk of bias. Funnel plots for LCn3 trials suggested that there may be missing trials for all primary outcomes except stroke and arrhythmia, and in all cases adding such trials back in would move effect sizes closer to no effect (RR 1.0). This lack of effect in the trials at lowest risk of bias (with suggestions of effect in trials at moderate to high risk of bias) supported our interpretation of lack of effect of LCn3 fats on many of our primary outcomes.

However, the increased numbers of trials, of longer trials and of greater numbers of people randomised within included trials (up by 31% in this review from the previous iteration, Abdelhamid 2018a), has increased our power to see small effects of increasing LCn3. We pre‐stated that effect sizes needed to be greater than 8% (RR < 0.92 or > 1.08) to suggest an important effect. Effect sizes of 9% or 10% are still small, and for the outcomes that have reached this level of effectiveness in this review NNTBs are still very large. The effects of LCn3 and ALA we are seeing on cardiovascular diseases and triglycerides are small.

Potential biases in the review process

Potential adverse effects include cancers and neurological problems associated with polychlorinated biphenyls (PCBs) or mercury in fish oils, and bleeds associated with reductions in clotting (see How the intervention might work). We have collated any data on bleeds, including haemorrhagic stroke, in this review, though we did not ask trial authors specifically for additional data on these outcomes. Unfortunately there were insufficient data on serious harms (bleeding and pulmonary embolism or deep vein thrombosis) to assess these potential harms. We have not collated data on cancers (Hanson 2019), and neurological problems (Brainard 2019), within this review but have formally systematically reviewed them elsewhere. This approach is preferable to including data on these outcomes from within included trials, which would be incomplete and potentially underpowered to show important effects.

One problem with cardiovascular disease outcomes is that they are collected together in a variety of ways, depending on the trial. For example, in assessing coronary heart disease mortality, we prespecified that we would include the first of the following list reported in any trial: coronary death, ischaemic heart disease death, fatal myocardial infarction, cardiac death. Each included trial collates outcome data in its own way, so we had to adapt to this in our analysis. One way to get around this problem would be to ask trial authors to go back to their original datasets to assess outcomes in a single prespecified way. This was done by a recent meta‐analysis that included only 10 trials but was able to formulate their outcome data to match between trials (Aung 2018). In doing this they were able to include data on outcomes that we were not able to access. For example, numbers of coronary heart disease deaths for GISSI‐HF 2008 have risen considerably in this version of the review, as we were able to include data on coronary heart disease deaths from a greater range of sources than were previously published: sudden cardiac deaths, deaths due to ventricular arrhythmias, and heart failure in patients with coronary heart disease, myocardial infarction, or deaths occurring after coronary revascularisation or heart transplant. Their data are also richer with regard to summing people with events (rather than numbers of events). Numbers of events are relatively easy to extract from published papers, but are not additive across composite endpoints such as cardiovascular disease events or coronary heart disease events (as a single individual may have experienced a stroke and a heart attack as well as onset of angina, but should be counted once only within any one composite outcome ‐ counting events would lead to their inclusion twice in coronary heart disease events and three times for cardiovascular disease events). For that reason previous versions of this review have been conservative in the data we have used, but use of the Aung 2018 data has allowed more data inclusion. The next section discusses similarities and differences between this review and Aung 2018, but their findings were highly similar. For example, effects of LCn3 on coronary heart disease mortality, our meta‐analytic estimate of effect of LCn3 was RR 0.90, (95% CI 0.81 to 1.00; I2 = 35%) in 24 trials reporting 3598 coronary heart disease deaths (Analysis 1.37), while theirs was RR 0.93 (95% CI 0.83 to 1.03) in 10 trials reporting 2695 coronary heart disease deaths.

While we tried hard to locate all available trials and collect additional outcome data where possible, there was evidence from funnel plots of some small study bias. Some smaller trials showing increased risk of cardiovascular disease outcomes with omega‐3 fats may be missing from some of the meta‐analyses. If these trials were replaced they would tend to increase risk ratios. This suggests that there is some underlying small study bias within the review.

Underlying risks vary from trial to trial. For example, the risk of death in GISSI‐HF 2008 was 28% in people with heart failure at baseline, but less than 4% in the healthy population recruited to VITAL 2019. Despite this there was little suggestion of heterogeneity between trials (I2 = 5%; Analysis 1.1). Where there have been suggestions of effects we have translated these into benefits or harms in both higher‐ and lower‐risk populations to clarify.

Given that the many LCn3 trials at moderate to high risk of bias appear to be inflating any protective effects, and that small trial bias is also inflating any protective effects, it is justified to view with skepticism the occasional suggestion of a protective effect. We do see some protective effects, but they are small. Given the very large number of subgroup analyses we carried out, it is also important to treat the occasional subgroup analysis that throws up a statistically significant difference between subgroups very cautiously.

A secondary question asked by this review was about differential effects of dietary and supplemental LCn3 fats. LCn3 interventions included dietary advice (advice to eat more oily fish), supplemental foods (LCn3 fats incorporated into other foods such as margarine) and supplements or capsules (by far the greatest proportion of trials). Dietary fish is likely to have different health effects, as it may take the place of less healthy foods in the diet (leading to reduced saturated fat intake, for example) and provides many nutrients in addition to omega‐3 fats (such as protein, selenium, iodine, calcium, magnesium, etc.). There were only four LCn3 dietary advice trials with event data (DART 1989; DART2 2003; DISAF 2003; THIS DIET 2008), and two of these also provided fish oil capsules when participants did not want to eat more fish (DART 1989; DART2 2003). We found no statistically significant differences between dietary advice subgroups and supplemental foods or capsule subgroups for primary outcomes. This may mean that health effects between the two types of intervention are not different, but it is likely that our analysis was underpowered to see any such differences if they exist.

Population LCn3 status varies widely across the world, from over 8% of fatty acids in Japan, Scandinavia and other areas with non‐Westernised dietary patterns to less than 4% in North, South and Central America; Europe; the Middle East; Southeast Asia; and Africa (Stark 2016). We hypothesised that additional LCn3 might have greater health effects in people whose usual LCn3 intake was relatively low, but unfortunately we were not able to ascertain baseline LCn3 intake or status for most of our included trials. However, most of the included trials were carried out in areas of the world with lower LCn3 status, so we would expect to see effects of increasing LCn3 in most included trials if such effects exist.

Agreements and disagreements with other studies or reviews

One potential difference between the findings of this review and some other trials and reviews is our running sensitivity analyses assessing effects exclusively in trials at low summary risk of bias. This clarified the lack of effect of LCn3 fats on stroke, which otherwise appeared slightly harmful. On the other hand, these sensitivity analyses suggested protective effects of LCn3 on coronary heart disease mortality and ALA on cardiovascular events. The effectiveness or lack of effectiveness of LCn3 on coronary heart disease events is harder to call. The main analysis suggests a 9% reduction in coronary heart disease events, but sensitivity analyses limiting trials to those at low summary risk of bias, or fixed‐effect analysis suggested 7% and 8% reductions, while limiting to trials with low risk of compliance problems or to larger trials suggested reductions of 16% and 11% respectively. We have suggested that this implies that LCn3 does slightly reduce risk of coronary heart disease events, but the interpretation that there was little or no effect is also logical.

There was no suggestion that blood pressure or risk of adverse events such as bleeding differed by LCn3 or ALA intake. This suggests that possible mechanisms for omega‐3 activity, including lowering of blood pressure, reduced thrombotic tendency and anti‐arrhythmic effects are not important in most adults, though LCn3 does lower serum triglyceride levels. We did not systematically review blood pressure data so may have missed a few long‐term trials (though not many) – missing data from included trials is likely to be a bigger issue. Of the 15 included trials that reported blood pressure outcomes, nine reported numbers of hypertensive participants at baseline, ranging from 5% in MARINA 2011 to 79% of participants in ORIGIN 2012. Effects did not differ by proportions of hypertensive participants (I2 was 0% for both systolic and diastolic blood pressure; Analysis 3.1).

Nearly 20 years ago, the GISSI‐P 1999 trial suggested that LCn3 had its primary effects in reducing sudden cardiac death. However, the forest plot clearly shows that subsequent trials have not seen this effect individually or in aggregate (Analysis 2.7).

The scope of this review is similar to that of the extensive Agency for Healthcare Research and Quality review (Balk 2016), so we have compared our results with theirs. Given that our review included 86 RCTs randomising more than 162,000 participants, who experienced over 11,000 deaths and more than 5000 cardiovascular disease deaths, we were surprised to read that Balk and colleagues characterised the body of evidence as having "limited data ... from RCTs on the effect of n‐3 FA on clinical CVD [cardiovascular disease] outcomes" (Balk 2016). This appears to be because the Balk review excluded RCTs of people with non‐cardiovascular disease and non‐diabetes‐related diseases at baseline, while we included them. While Balk 2016 excluded some trials that we included, it did not include any trials providing all‐cause mortality data that we excluded. This meant that in analysing effects on all‐cause mortality, Balk 2016 included 18 RCTs randomising 81,027 participants experiencing 8480 deaths, while we included more than 143,000 participants randomised to high or low LCn3 or ALA experiencing 11,297 deaths. Balk 2016 excluded trials that we included, such as AREDS2 2014, a high‐quality trial with 368 deaths in more than 4000 participants with age‐related macular degeneration. This sort of population appeared ideal to us for assessment of omega‐3 fats on primary prevention of cardiovascular disease, as these people tend to be elderly but there is no clear reason why omega‐3 fats would affect cardiovascular disease differently in this population than in other older adults at usual cardiovascular disease risk.

Despite these slight differences in approach, we obtained very similar effect estimates to Balk 2016. We meta‐analysed effects of LCn3 and ALA trials, finding an RR for all‐cause mortality of 0.97 (95% CI 0.93 to 1.01; I2 = 5%), compared to a pooled RR for all‐cause mortality of 0.97 (95% CI 0.92 to 1.03) in Balk 2016. While that review seldom pooled their results, we can compare our results with theirs where they did. Despite our slightly different inclusion criteria, our results are very comparable (Table 11).

8. Comparison of the results of long‐chain omega‐3 interventions in this review with other major recent reviewsa.

Systematic review Balk 2016 Aung 2018 Hu 2019 This review
Outcome Number of people experiencing events RR (95% CI) Number of people experiencing events RR (95% CI) Number of people experiencing events RR (95% CI) Number of people experiencing events RR (95% CI)
All‐cause mortality 8480 0.97 (0.92 to 1.03) Not assessed Not assessed 11,297 0.97 (0.93 to 1.01)
Cardiovascular deaths 3799 0.92 (0.82 to 1.02) Not assessed 4630 0.93 (0.88 to 0.99) 5658 0.92 (0.86 to 0.99)
CVD events (MACCEs in Balk 2016) 8085 0.96 (0.91 to 1.02) 12,001 0.97 (0.93 to 1.01) 14,694 0.97 (0.94 to 0.99) 17,619 0.96 (0.92 to 1.01)
CHD deaths Not pooled 2695 0.93, (0.83 to 1.03) 2934 0.92 (0.86 to 0.98) 3598 0.90 (0.81 to 1.00)
CHD events Not assessed 6273 0.96, (0.90 to 1.01) 7536 0.95 (0.91 to 0.99) 8791 0.91 (0.85 to 0.97)
Stroke 1467 0.98 (0.88 to 1.09) 1713 1.03 (0.93 to 1.13) 2459 1.05 (0.98 to 1.14) 2850 1.02 (0.94, 1.12)
Arrhythmia Not pooled Not assessed Not assessed 4586 0.99 (0.92 to 1.06)
CHD: coronary heart disease; CI: confidence interval; CVD: cardiovascular disease; MACCE: major adverse cerebrovascular or cardiovascular event; RCT: randomised controlled trial; RR: risk ratio

aMeta‐analysis of effects of long‐chain omega‐3 in three recent systematic reviews, Balk 2016, Aung 2018 and Hu 2019, comparing their findings with our findings for our primary outcomes.

A recent individual meta‐analysis of 10 large trials in almost 78,000 people at high risk of cardiovascular disease found no associations of LCn3 with coronary heart disease mortality (RR 0.93, 99% CI 0.83 to 1.03), nonfatal myocardial infarction (RR 0.97, 99% CI 0.87 to 1.08), coronary heart disease events (RR 0.96, 95% CI 0.90 to 1.01) or major vascular events (RR 0.97, 95% CI 0.93 to 1.01; Aung 2018). Aung 2018 included individual patient data from the participants of large, long trials (randomising at least 500 participants and following them for at least one year) (AlphaOmega ‐ EPA+DHA 2010; AREDS2 2014; DO IT 2010; GISSI‐HF 2008; GISSI‐P 1999; JELIS 2007; OMEGA 2009; ORIGIN 2012; Risk & Prevention 2013; SU.FOL.OM3 2010), and this review includes all these trials. Their review had the advantage of being able to ensure that they had complete and equivalent data for all of their key outcomes from all the trials, reducing the risk of publication bias (and this review has incorporated some of their data to improve and extend our data set), but the disadvantage of missing all the data from many large LCn3 trials such as DART 1989, DART2 2003, FORWARD 2013, MAPT 2017, SHOT 1996 and SOFA 2006, all LCn3 trials randomising at least 500 participants. It also missed large trials of LCn3 in lower‐risk participants such as OPAL 2010, and large trials of ALA such as MARGARIN 2002, Norwegian 1968 and WAHA 2016, as well as all the smaller trials. However, though taking different approaches, the numerical results of this review and Aung 2018 are also very similar (Table 11). Aung 2018 has been updated with data from the three largest recent trials (ASCEND 2018; REDUCE‐IT 2019; VITAL 2019), published as Hu 2019. This review, Balk 2016, Aung 2018 and Hu 2019 addressed the analysis of the data in slightly different ways, creating sensitivity analyses for each other. The fact that they came to similar numerical conclusions reassures us that our conclusions are solidly based. However, the risk ratios for coronary heart disease mortality and events have fallen in this update of the review and crossed the line to suggest small beneficial effects of LCn3.

Cardiovascular diseases and cancers are both major non‐communicable killers, so overall assessment of the utility of increasing omega‐3 fats will depend on effects on both sets of outcomes. Figure 6 summarises the information on absolute effects of LCn3 and ALA on cardiovascular and cancer outcomes (focusing on outcomes where risk ratios were < 0.92 or > 1.08). The cancer data come from a sister review (Hanson 2019). Combining data from the two reviews suggests that for every 1000 men increasing their intake of LCn3, three will avoid dying from coronary heart disease, six will avoid a coronary heart disease event, one will avoid the diagnosis of arrhythmia and three additional men will be diagnosed with prostate cancer. Similarly, for every 1000 men increasing their intake of ALA two will avoid a cardiovascular disease event and 11 will avoid diagnosis of arrhythmia but three additional men will be diagnosed with prostate cancer (Figure 6). These small negative effects on cancer outcomes for men attenuate the small beneficial effects of LCn3 and ALA on cardiovascular outcomes.

6.

6

Assessment of health effects across cancers and cardiovascular outcomes. Bars above zero suggest the number of people who would benefit out of 1000 people consuming more long‐chain omega‐3 (LCn3) or alpha‐linolenic acid (ALA), bars below zero suggest the number of people who would be harmed out of 1000 people consuming more LCn3 or ALA. Cardiovascular disease (CVD) data are from this review, cancer data from a sister review (Hanson 2019). CHD: coronary heart disease.

Authors' conclusions

Implications for practice.

We found high‐certainty evidence that long‐chain omega‐3 fats (LCn3) do not have important positive or negative effects on mortality or cardiovascular events and moderate‐certainty evidence that they have little or no effect on cardiovascular disease mortality, stroke or arrhythmia in primary or secondary prevention. However, we found low‐certainty evidence that LCn3 slightly reduces risk of coronary heart disease mortality (number needed to treat for an additional beneficial outcome (NNTB) 334, NNTB 200 in secondary prevention, NNTB 1000 in primary prevention), and coronary heart disease events (NNTB 167, NNTB 143 in secondary prevention, NNTB 200 in primary prevention). As these effects are very small (and numbers needed to treat for an additional benefit high), supplemental LCn3 fats are probably not useful for preventing or treating cardiovascular disease. LCn3 fats can help to reduce serum triglycerides, though they do not appear to affect body fatness or other lipid fractions.

An NNTB of 167 means that 167 people will need to take LCn3 supplements for around four years each so that one of those people avoids a coronary heart disease event. The other 166 people receive no benefit. Similarly, an NNTB of 334 means that 334 people need to take LCn3 supplements for around four years each for one person to avoid death from coronary heart disease, the other 333 people do not benefit.

How does an NNTB of 143 or 500 compare with effective medications in cardiovascular disease prevention? Could we compare, for example, with the use of statins in secondary prevention of myocardial infarction or ezetimibe added to statins after acute coronary syndrome? In the 4S trial (Scandinavian Simvastatin Survival Study Group 1994), 8% of the participants taking simvastatin died, and 12% of those taking the placebo died, a difference of 4%, so the NNTB was 25. Twenty‐five people needed to take simvastatin for around five years to prevent one person dying. Most of us decide to take statins post‐myocardial infarction. Ten‐year NNTBs for statins in primary prevention when used according to appropriate guidance are around 30, to prevent one case of atherosclerotic cardiovascular disease (Mortensen 2019). The IMPROVE‐IT trial showed that giving ezetimibe in addition to statin to 50 people with acute coronary syndrome for seven years each prevents one person from having a cardiovascular event (Turgeon 2015).

Individuals may make differing decisions about whether an NNTB of 143 is useful. Perhaps NNTBs should be noted on fish oil supplement packages: “If 143 people with existing cardiovascular disease take this supplement for 5 years each then one of those 143 people will avoid a coronary event (such as a heart attack). If 1000 people without existing cardiovascular disease take this supplement for 5 years each then one person will avoid dying from coronary heart disease”. Then each of us could decide whether this is the way we wish to spend our money, on omega‐3 supplements or a different cardiovascular treat such as a pair of running shoes, healthy fruit or fish, or a relaxation class.

Most evidence on LCn3 fats came from trials of capsules of fish oil or eicosapentaenoic acid (EPA)/docosahexaenoic acid (DHA) mixtures. While we did not see important differences between trials of supplemental capsules and trials of oily fish, there were few trials of oily fish. Fish and seafood are nutrient‐dense and rich in a variety of other nutrients (such as vitamin D, calcium, iodine, selenium) so are useful foods even without cardiovascular benefits. In light of the evidence in this review it would be appropriate to review official recommendations supporting supplemental LCn3 fatty acid intake.

We found low‐certainty evidence that increasing ALA may slightly reduce risk of cardiovascular disease events (NNTB 500) and moderate‐certainty evidence that increasing ALA probably reduces risk of arrhythmia (NNTB 91). However, there is probably little or no effect on all‐cause or cardiovascular mortality, coronary heart disease mortality or coronary heart disease events (low‐ and moderate‐certainty evidence). As with LCn3, effects of ALA were very small; 91 people would need to increase their ALA intake to prevent one person developing arrhythmia, and 500 would need to take more ALA to prevent one person experiencing a cardiovascular disease event. We found no evidence that ALA affected adiposity or serum lipids. Trials of ALA gave ALA‐rich or enriched foods such as walnuts or ALA‐enriched margarine.

Implications for research.

There are several large ongoing trials of supplemental LCn3 fats (see Characteristics of ongoing studies, and large parts of VITAL 2019 are ongoing). We suggest that given the minimal protective effects suggested for omega‐3 fats in the large number of trials to date, no further trials should be initiated until the ongoing trials have reported. Ongoing and completed trials should make data on baseline LCn3 intake, and details of deaths, cardiovascular outcomes, lipids, adiposity and blood pressure available, as well as other key health outcomes, regardless of their primary outcomes.

Further large and high‐certainty trials of ALA carried out in lower‐ and higher‐income countries, and that assess baseline ALA intake and use biomarkers to assess compliance would be helpful to clarify the cardiovascular effects of ALA. Similarly trials of dietary fish would be helpful.

Feedback

Interpretation of effect estimates, 18 July 2018

Summary

I am not clear how the quoted RRs and CIs in the abstract support statements of no effect in one part but statements of effect in another part. It seems that throughout the CIs comprehensively span unity. For example, how is a statement of 'probably reduces risk of CHD mortality' supported by the metrics '(1.1% to 1.0%, RR 0.95, 95% CI 0.72 to 1.26, 18,353 participants; 193 CHD deaths, 3 RCTs'? That seems like an entirely null result.

Reply

Thank you for your query.

We described the process of deciding whether there was “little or no effect”, or a positive or negative effect, in the methods section of the review (under “Summary of findings table”).

It was agreed with the World Health Organization (WHO) Nutrition Guidance Expert Advisory Group (NUGAG) Subgroup on Diet and Health, who commissioned this review as part of a wider set that an effect size of 8% (either way, so RR > 1.08 or < 0.92) in the point estimate of the risk ratio would suggest benefit or harm. Presence or absence of effect is decided on the basis of the pre‐stated outcome measure, here RR. This certainty of this finding was assessed through the GRADE process (which is represented in the Summary of Findings tables).

  • Wide confidence intervals lead to downgrading for imprecision,

  • changes in results in sensitivity analyses lead to downgrading for risk of bias,

  • skewed funnel plots or knowledge of quantities of missing data lead to downgrading for publication bias,

  • heterogeneity of results (high I2) lead to downgrading for inconsistency, and

  • limited representativeness of included populations lead to downgrading for indirectness.

We used absolute risk or NNTB to describe the scale of effect where an effect was suggested – this could be large or small (all in this review were very small).

You are absolutely correct in pointing out that this means that ALA intake probably makes little or no difference to CHD mortality. We apologise for this confusion, which resulted from us using an earlier cut‐off of 7% to consider effectiveness. The effect for CV events is still “ALA intake may reduce the risk of cardiovascular events but by a very small amount (from 4.8 to 4.7%)” even though the effect is RR 0.95 (95% CI 0.83 to 1.07) as the sensitivity analysis limiting to studies at low summary risk of bias suggested a 9% reduction in risk (RR 0.91, 95% CI 0.79 to 1.04, I2 = 0%). Effects of ALA on arrhythmia are clearer (main analysis suggests RR 0.79, 95% CI 0.57 to 1.10, moderate‐certainty evidence).

This finding has now been corrected in the review.

Thank you for your keen eye! We hope this clarifies how decisions were made and effects expressed within the review (and the review series).

Contributors

Feedback submitted by: Bruce Neal

Response by Lee Hooper, contact author of review, and Bill Cayley, feedback editor of Cochrane Heart

Dosing and conclusions, 19 July 2018

Summary

Most of the CVD OR ranges listed for lcN3 data showed zones of significant OR benefit. No dosing information was included. I question the study's "conclusions", and believe that a more sensitive analysis of the data could easily show benefit in terms of CVD risk reduction.

Reply

Thank you for your comments, and your attention to the question of dosing. While with any intervention it certainly might seem plausible that a higher dose would be more likely to show benefit than a lower dose, this was not borne out in the studies that met inclusion criteria for this Review. As outlined in the Summary of Findings Tables, and summarized in the Abstract, the authors “found no evidence of dose‐response or duration effects for any primary outcome, but there was a suggestion of greater protection in participants with lower baseline omega‐6 intake across outcomes.”

Contributors

Feedback submitted by: TR Morris

Response by Lee Hooper, contact author of review, and Bill Cayley, feedback editor of Cochrane Heart

Effect of study duration, 26 November 2019

Summary

I think it's incorrect to determine mortality as the percentage of patients who died during the considered studies (e.g., 9% for patients taking long chain omega 3, according to the article). Since those studies have different lengths (from 12 to 72 months), patients from long studies are expected to die with much higher rate, than patients from short studies. Long studies will contribute toward increasing the mortality metric, while short studies will decrease it. Instead, I suggest normalizing mortality by study duration. For example, let's take 1 month as a unit of duration. Then, if a study lasted 24 months and had mortality rate 3%, that is living rate of 97%, or 0.97, you calculate average living rate per month as 0.97^(1/24) ≈ 0.99875 = 99.875%, and you use 100% ‐ 99.875% = 0.125% as average mortality rate per month in this study.

Reply

Thank you for your comments on the duration of our included trials. Outcome risk does differ between included trials and depends on baseline risk as well as trial duration, but this is not necessarily true of relative risk, our outcome measure. To assess duration effects, as well as effects of dose, baseline risk, intervention type, replacement, statin use, baseline triglyceride levels or diabetic status, we carried out extensive sub‐grouping and meta‐regression. We found no suggestion of consistent duration effects in either sub‐grouping or meta‐regression for either LCn3 or ALA except for effects of LCn3 on stroke, where longer trials suggested a smaller effect size in meta‐regression (this was not clear in sub‐grouping), and longer trials showed increased risk of arrhythmia with increased LCn3 in sub‐grouping and meta‐regression. These sub‐grouping and meta‐regression results are all discussed in the results section of the review. We believe that this accounts appropriately and fully for duration effects within the systematic review.

Contributors

Feedback submitted by: Philip Blagoveschensky, Data Science Masters student, Skoltech

Response by Lee Hooper, contact author of review, and Bill Cayley, feedback editor of Cochrane Heart

What's new

Date Event Description
11 October 2019 New citation required and conclusions have changed Seven new trials (six of LCn3, one of ALA) added to included trials (of which six were moved from ongoing trials), and seven new ongoing trials added. Addition of these included trials boosts numbers of included participants by over 30%. The 86 included RCTs randomised 162,796 participants to trials of at least 12 months' duration.
The updated evidence suggests that increasing LCn3 slightly reduces CHD mortality and CHD events (previously the evidence suggested little or no effect). Our understanding of effects of LCn3 on other outcomes, and of ALA on all outcomes, has not altered.
21 September 2019 New search has been performed Electronic searches updated to February 2019, trials registry searches to July 2019.

History

Protocol first published: Issue 2, 1999
 Review first published: Issue 4, 2004

Date Event Description
28 November 2018 New citation required but conclusions have not changed The amendments did not change our conclusions.
28 November 2018 Feedback has been incorporated We have responded to feedback by two parties.
28 November 2018 Amended Effects of alpha‐linolenic acid on coronary heart disease mortality now correctly interpreted as "little or no effect" as effect size was < 8%.
Effects of long‐chain omega‐3 and alpha‐linolenic acid on serum high‐density lipoprotein reinterpreted as "little or no effect" as changes were < 5% of baseline.
Study flow corrected.
13 March 2018 New citation required and conclusions have changed This update now reports arrhythmia (atrial fibrillation) and cardiovascular mortality data. Data now included from 79 RCTs (112,059 participants) lasting at least one year, of which 25 were at low summary risk of bias.
We added the following outcomes to the list of primary outcomes upon the request of World Health Organization (WHO) Nutrition Guidance Expert Advisory Group (NUGAG) Subgroup on Diet and Health.
  1. Cardiovascular mortality.

  2. Arrhyhtmia (new and recurrent).


We altered inclusion criteria to include only RCTs of at least 12 months' duration (rather than 6 months as previously), and we excluded cohort studies.
We are assessing effects of long‐chain omega‐3 fats separately from effects of alpha‐linolenic acid (as planned in the previously published version).
27 April 2017 New search has been performed Electronic searches updated to 27 April 2017
14 March 2012 Amended Additional tables re‐numbered
16 October 2011 New search has been performed Searches updated to July 2011.
Cohort studies not included in this update, and previously included cohort studies and related text have been removed.
Previously included trials where we know that no deaths or primary or secondary health events occurred were removed.
New secondary outcomes added (fatal and non‐fatal arrhythmias, and diabetes)
Cardiovascular mortality added as a primary outcome.
9 September 2008 Amended Converted to new review format.
1 August 2004 New citation required and conclusions have changed Substantive amendment

Acknowledgements

The wider set of reviews, including this one, were carried out by the Polyunsaturated Fats and Health (PUFAH) group. PUFAH includes Asmaa Abdelhamid, Zoya Ahmed, Sarah MA Ajabnoor, Fai K AlAbdulghafoor, Lena Al‐Khudairy, Priti Biswas, Julii Suzanne Brainard, Charlene Bridges, Tracey J Brown, Katherine HO Deane, Daisy H Donaldson, Sarah Hanson, Lee Hooper, Oluseyi F Jimoh, Nicole Martin, Katie Maas, Helen J Moore, Alex T O’Brien, Karen Rees, Ruksana Sivakaran, Fujian Song, Carolyn D Summerbell,, Gabrielle C Thorpe, Xia Wang , Ailsa Welch, Lauren Winstanley, and Helen V Worthington.

The review authors thank all of the authors of primary trials who kindly provided us with the best set of data available, including: D Kromhout, Wageningen University (AlphaOmega ‐ ALA 2010; AlphaOmega ‐ EPA+DHA 2010); Emily Chew, NIH (AREDS2 2014); J Brox, University Hospital of North Norway (Brox 2001); C Argo, University of Virginia (Caldwell 2011); ML Burr, University of Wales and A Ness, University of Bristol (DART 1989; DART2 2003); G Derosa and P Maffioli, University of Pavia (Derosa 2016); PNM Demacker, University Hospital Nijmegen (Deslypere 1992); S Tokudome, National Institue of Health and Nutrition, Japan (DIPP 2015); R Harrison, University of Manchester, UK (DISAF 2003); G Einvik, Akershus University Hospital and H Arnesen, Oslo University Hospital (DO IT 2010); S Dodin, Universite Laval (Dodin 2005); D Schoenfeld, Harvard Medical School (FAAT 2005); A Macchia, Fundacio'n GESICA (FORWARD 2013); C Hill, the Queen Elizabeth Hospital (FOSTAR 2016); D Franzen, Universitat zu Koln (Franzen 1993); P Bauer, Medical University of Vienna (Lorenz‐Meyer 1996); WJ Bemelmans, National Institute for Public Health and the Environment, Bilthoven (MARGARIN 2002); T Sanders, King's College, London (MARINA 2011); E Souied, Hôpital intercommunal de Créteil, France (NAT2 2013); D Nilsen, University of Bergen, Norway (OFAMI 2001); S Schneider, Institut für Herzinfarktforschung, Germany (OMEGA 2009); A Dangour, London School of Hygiene & Tropical Medicine (OPAL 2010); M James, Royal Adelaide Hospital, Australia (Proudman 2015); R Zurier, University of Massachusetts Medical School (Reed 2014); C Roncaglioni and I Marzona, IRCCS‐Istituto di Ricerche Farmacologiche Mario Negri, Italy (Risk & Prevention 2013); A Manni, Penn State College of Medicine, USA (Sandhu 2016); C von Schacky, Ludwig Maximilians University, Munich, Germany (SCIMO 1999); L Tapsell, M Batterham and E Neale, University of Wollongong, Australia (SMART 2013); P Galan, Université Paris (SU.FOL.OM3 2010); K Tande, Calanus AS, Norway (Tande 2016); K Tuttle, Sacred Heart Medical Center, Spokane (THIS DIET 2008); J Sabaté, Loma Linda University, California (WAHA 2016); E Scorletti, University of Southampton, UK (WELCOME 2015).

Thanks also to the authors who replied but were not able to provide further details or confirmed no relevant outcomes, including: S Jebb, University of Oxford (FISH 2012); R Holman, University of Oxford (AFORRD 2010); J Quinn, Oregon Health & Science University (ADCS 2010); D Bates, Royal Victoria Infirmary, Newcastle on Tyne and R Dworkin, University of Rochester, UK (Bates 1989); E Berson, Harvard Medical School, USA (Berson 2004); A Sanyal, Virginia Commonwealth University, USA (EPE‐A 2014); G Pierce, St. Boniface Hospital Research Centre, Canada (FLAX‐PAD 2013); H Gerstein, McMaster University, Canada (ORIGIN 2012); B Puri, Imperial College London (Puri 2005); M Raitt, Oregon Health & Science University, USA (Raitt 2005); J Eritsland, Oslo University Hospital, Norway (SHOT 1996); and I Brouwer, Vrije Universiteit Amsterdam, the Netherlands (SOFA 2006).

Warm thanks and appreciation to the authors of the first review who did not continue as authors in the review update, including Roger A Harrison, Andrew Ness, Nigel E Capps, George Davey Smith, Shah BJ Ebrahim and Rudolph A Riemersma. Acknowledgements for the earlier version of this systematic review, which we have built on to create the current review: our thanks to Julian Higgins, who was involved in the design of the review, along with Theresa Moore and Margaret Burke from Cochrane Heart. Also to Charlotte C Hammer for help with German translation for Franzen 1993.

Appendices

Appendix 1. Search strategy 2019

CENTRAL

#1 MeSH descriptor: [Fish Oils] explode all trees

#2 MeSH descriptor: [Linseed Oil] this term only

#3 MeSH descriptor: [Linolenic Acids] this term only

#4 MeSH descriptor: [Fatty Acids, Omega‐3] explode all trees

#5 (fish near/3 oil*)

#6 (oil* near/3 (cod* or marin*))

#7 (omega‐3 or omega3 or (omega* near/5 fat*))

#8 eicosapentaen*

#9 docosahexaen*

#10 (oil* near/3 (flax* or rapeseed* or canola*))

#11 (Linolen* or alpha‐linolen* or alphalinolen*)

#12 (perilla* or linseed* or maxepa*)

#13 (oil* near/3 (rape or colza))

#14 (marin* near/3 lipid*)

#15 (naudicelle* or herring* or sild)

#16 (clupe* near/3 hareng*)

#17 (whitebait or sardine* or sardina* or pilchard* or sprat* or brisling*)

#18 (salmo* near/3 trut*)

#19 (trout or bloater or kipper* or salmon or mackerel* or scomb* or conger* or tuna or tunny or tunafish or tuna‐fish)

#20 (thunnus* or swordfish* or xiphias* or dogfish or scyliorrhinus*)

#21 (crab or crabs or (cancer pagarus))

#22 (DHA or EPA)

#23 MeSH descriptor: [Salmoniformes] explode all trees

#24 MeSH descriptor: [Tuna] this term only

#25 MeSH descriptor: [alpha‐Linolenic Acid] this term only

#26 MeSH descriptor: [Flax] this term only

#27 (fish near/3 (diet* or capsul* or nutrit* or supplement*))

#28 (icosapentaen* or docosapentaen*)

#29 (oil* near/3 (purslane or mustard* or candlenut* or stillingia or walnut*))

#30 (laks or lax)

#31 (ALA or DPA)

#32 (algal near oil*)

#33 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 or #18 or #19 or #20 or #21 or #22 or #23 or #24 or #25 or #26 or #27 or #28 or #29 or #30 or #31 or #32 Publication Year from 2017 to 2019

MEDLINE Ovid

1. exp Fish Oils/

2. Linseed Oil/

3. linolenic acids/ or alpha‐linolenic acid/

4. Flax/

5. exp Fatty Acids, Omega‐3/

6. (fish adj3 (diet* or nutrit* or oil* or supplement*)).ti,ab.

7. (oil* adj3 (cod* or marin*)).ti,ab.

8. (omega‐3 or omega3 or (omega* adj5 fat*)).ti,ab.

9. eicosapentaen*.ti,ab.

10. docosahexaen*.ti,ab.

11. (oil* adj3 (flax* or rapeseed* or canola*)).ti,ab.

12. (Linolen* or alpha‐linolen* or alphalinolen*).ti,ab.

13. (perilla* or linseed* or maxepa*).ti,ab.

14. (oil* adj3 (rape or colza)).ti,ab.

15. (marin* adj3 lipid*).ti,ab.

16. (naudicelle* or herring* or sild).ti,ab.

17. (clupe* adj3 hareng*).ti,ab.

18. (whitebait or sardine* or sardina* or pilchard* or sprat* or brisling*).ti,ab.

19. (salmo* adj3 trut*).ti,ab.

20. (trout or bloater or kipper* or salmon or mackerel* or scomb* or conger* or tuna or tunny or tunafish or tuna‐fish).ti,ab.

21. (thunnus* or swordfish* or xiphias* or dogfish or scyliorrhinus* or laks or lax).ti,ab.

22. (crab or crabs or cancer pagarus).ti,ab.

23. exp salmoniformes/ or tuna/

24. (fish adj3 capsul*).ti,ab.

25. icosapentaen*.ti,ab.

26. docosapentaen*.ti,ab.

27. (oil* adj3 (purslane or mustard* or candlenut* or stillingia or walnut*)).ti,ab.

28. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23 or 24 or 25 or 26 or 27

29. randomized controlled trial.pt.

30. controlled clinical trial.pt.

31. randomized.ab.

32. placebo.ab.

33. clinical trials as topic.sh.

34. randomly.ab.

35. trial.ti.

36. 29 or 30 or 31 or 32 or 33 or 34 or 35

37. exp animals/ not humans.sh.

38. 36 not 37

39. 28 and 38

40. limit 39 to ed=20170427‐20190213

41. 39 not (1* or 2*).ed.

42. 40 or 41

Embase Ovid

1. exp salmoniformes/ or tuna/

2. fish oil/

3. linseed oil/

4. linolenic acid/

5. Flax/

6. omega 3 fatty acid/

7. (fish adj3 (diet* or nutrit* or oil* or supplement*)).ti,ab.

8. (oil* adj3 (cod* or marin*)).ti,ab.

9. (omega‐3 or omega3 or (omega* adj5 fat*)).ti,ab.

10. (eicosapentaen* or icosapentaen*).ti,ab.

11. docosahexaen*.ti,ab.

12. (oil* adj3 (flax* or rapeseed* or canola*)).ti,ab.

13. (Linolen* or alpha‐linolen* or alphalinolen*).ti,ab.

14. (perilla* or linseed* or maxepa*).ti,ab.

15. (marin* adj3 lipid*).ti,ab.

16. (naudicelle* or herring* or sild).ti,ab.

17. (clupe* adj3 hareng*).ti,ab.

18. (whitebait or sardine* or sardina* or pilchard* or sprat* or brisling*).ti,ab.

19. (salmo* adj3 trut*).ti,ab.

20. (trout or bloater or kipper* or salmon or mackerel* or scomb* or conger* or tuna or tunny or tunafish or tuna‐fish).ti,ab.

21. (thunnus* or swordfish* or xiphias* or dogfish or scyliorrhinus* or laks or lax).ti,ab.

22. (crab or crabs or (cancer adj3 pagarus)).ti,ab.

23. exp salmonine/

24. (fish adj3 capsul*).ti,ab.

25. docosapentaen*.ti,ab.

26. (ALA or DHA or DPA or EPA).ti,ab.

27. (algal adj oil*).ti,ab.

28. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23 or 24 or 25 or 26 or 27

29. random$.tw.

30. factorial$.tw.

31. crossover$.tw.

32. cross over$.tw.

33. cross‐over$.tw.

34. placebo$.tw.

35. (doubl$ adj blind$).tw.

36. (singl$ adj blind$).tw.

37. assign$.tw.

38. allocat$.tw.

39. volunteer$.tw.

40. crossover procedure/

41. double blind procedure/

42. randomized controlled trial/

43. single blind procedure/

44. 29 or 30 or 31 or 32 or 33 or 34 or 35 or 36 or 37 or 38 or 39 or 40 or 41 or 42 or 43

45. (animal/ or nonhuman/) not human/

46. 44 not 45

47. 28 and 46

48. limit 47 to dd=20160721‐20170427

ClinicalTrials.com

1. Recruiting, Not yet recruiting, Active, not recruiting, Completed, Enrolling by invitation, Suspended, Terminated, Withdrawn, Unknown status Studies | Interventional Studies | omega‐3 or "omega 3" OR EPA OR DHA OR eicosapentaen* OR docosahexaen* OR docosapentaen* | Adult, Older Adult | Studies that accept healthy volunteers | Phase 1, 2, 3, 4 | First posted from 09/20/2016 to 08/01/2019.

2. This search was re‐run using the term "eicosapentaenoic" in place of the omega‐3 string.

3. This search was re‐run using the term "fish oil*" in place of the omega‐3 string.

ICTRP

(omega‐3 or "omega 3" OR EPA OR DHA OR eicosapentaen* OR docosahexaen* OR docosapentaen* OR fish oil*) for intervention (title and condition not limited, including any recruitment status, all phases though run separately, limited to date of registration 20th Sept 2016 to 1st August 2019).

Appendix 2. Medline search strategy 2002 (for the previous version of this review)

1 exp Fish Oils/
 2 exp Linseed Oil/
 3 linolenic acids/ or exp alpha‐linolenic acid/
 4 exp Fatty Acids, Omega‐3/
 5 (fish adj5 (diet$ or nutrit$ or oil$ or supplement$)).tw.
 6 (oil$ adj3 (cod$ or marin$ or rapeseed$ or canola$)).tw.
 7 (omega‐3 or omega3).tw.
 8 (eicosapentaen$ or icosapentaen$).tw.
 9 docosahexaen$.tw.
 10 (Linolen$ or alpha‐linolen$ or alphalinolen$).tw.
 11 (maxepa$ or omacor$).tw.
 12 (trout or kipper$ or salmon or mackerel$ or tuna or tunafish or sardine$ or pilchard$ or herring$).tw.
 13 flax$.tw.
 14 rapeseed$.tw.
 15 canola$.tw.
 16 alphalinolen$.tw.
 17 perilla$.tw.
 18 linolen$.tw.
 19 linseed$.tw.
 20 maxepa$.tw.
 21 (oil$ adj3 colza).tw.
 22 (marin$ adj3 (lipid$ or oil$)).tw.
 23 naudicelle$.tw.
 24 sild.tw.
 25 (clupe$ adj3 hareng$).tw.
 26 whitebait$.tw.
 27 sprat$.tw.
 28 brisling$.tw.
 29 (salmo adj3 trut$).tw.
 30 bloater.tw.
 31 scomb$.tw.
 32 conger$.tw.
 33 tunny.tw.
 34 tuna‐fish.tw.
 35 thunnus$.tw.
 36 swordfish$.tw.
 37 xiphias$.tw.
 38 dogfish.tw.
 39 scyliorhinus$.tw.
 40 (crab or crabs).tw.
 41 (cancer adj3 pagurus).tw.
 42 (laks or lax).tw.
 43 exp Flax/
 44 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23 or 24 or 25 or 26 or 27 or 28 or 29 or 30 or 31 or 32 or 33 or 34 or 35 or 36 or 37 or 38 or 39 or 40 or 41 or 42 or 43
 45 randomized controlled trial.pt.
 46 controlled clinical trial.pt.
 47 randomized.ab.
 48 placebo.ab.
 49 clinical trials as topic.sh.
 50 randomly.ab.
 51 trial.ti.
 52 50 or 47 or 51 or 46 or 45 or 48 or 49
 53 (animals not (human and animals)).sh.
 54 52 not 53
 55 44 and 54
 56 (20$ not (2000$ or 2001$)).ed.
 57 55 and 56

Appendix 3. Search strategy 2016 (to update the omega‐3 review)

CENTRAL

#1 MeSH descriptor: [Fish Oils] explode all trees

#2 MeSH descriptor: [Linseed Oil] this term only

#3 MeSH descriptor: [Linolenic Acids] this term only

#4 MeSH descriptor: [Fatty Acids, Omega‐3] explode all trees

#5 (fish near/3 oil*)

#6 (oil* near/3 (cod* or marin*))

#7 (omega‐3 or omega3 or (omega* near/5 fat*))

#8 eicosapentaen*

#9 docosahexaen*

#10 (oil* near/3 (flax* or rapeseed* or canola*))

#11 (Linolen* or alpha‐linolen* or alphalinolen*)

#12 (perilla* or linseed* or maxepa*)

#13 (oil* near/3 (rape or colza))

#14 (marin* near/3 lipid*)

#15 (naudicelle* or herring* or sild)

#16 (clupe* near/3 hareng*)

#17 (whitebait or sardine* or sardina* or pilchard* or sprat* or brisling*)

#18 (salmo* near/3 trut*)

#19 (trout or bloater or kipper* or salmon or mackerel* or scomb* or conger* or tuna or tunny or tunafish or tuna‐fish)

#20 (thunnus* or swordfish* or xiphias* or dogfish or scyliorrhinus*)

#21 (crab or crabs or (cancer pagarus))

#22 (DHA or EPA)

#23 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 or #18 or #19 or #20 or #21 or #22 Publication Year from 2002 to 2016

#24 MeSH descriptor: [Salmoniformes] explode all trees

#25 MeSH descriptor: [Tuna] this term only

#26 MeSH descriptor: [alpha‐Linolenic Acid] this term only

#27 MeSH descriptor: [Flax] this term only

#28 (fish near/3 (diet* or capsul* or nutrit* or supplement*))

#29 (icosapentaen* or docosapentaen*)

#30 (oil* near/3 (purslane or mustard* or candlenut* or stillingia or walnut*))

#31 (laks or lax)

#32 (ALA or DPA)

#33 (algal near oil*)

#34 #24 or #25 or #26 or #27 or #28 or #29 or #30 or #31 or #32 or #33

#35 #23 or #34

MEDLINE Ovid

1. exp Fish Oils/

2. Linseed Oil/

3. linolenic acids/ or alpha‐linolenic acid/

4. Flax/

5. exp Fatty Acids, Omega‐3/

6. (fish adj3 (diet* or nutrit* or oil* or supplement*)).ti,ab.

7. (oil* adj3 (cod* or marin*)).ti,ab.

8. (omega‐3 or omega3 or (omega* adj5 fat*)).ti,ab.

9. eicosapentaen*.ti,ab.

10. docosahexaen*.ti,ab.

11. (oil* adj3 (flax* or rapeseed* or canola*)).ti,ab.

12. (Linolen* or alpha‐linolen* or alphalinolen*).ti,ab.

13. (perilla* or linseed* or maxepa*).ti,ab.

14. (oil* adj3 (rape or colza)).ti,ab.

15. (marin* adj3 lipid*).ti,ab.

16. (naudicelle* or herring* or sild).ti,ab.

17. (clupe* adj3 hareng*).ti,ab.

18. (whitebait or sardine* or sardina* or pilchard* or sprat* or brisling*).ti,ab.

19. (salmo* adj3 trut*).ti,ab.

20. (trout or bloater or kipper* or salmon or mackerel* or scomb* or conger* or tuna or tunny or tunafish or tuna‐fish).ti,ab.

21. (thunnus* or swordfish* or xiphias* or dogfish or scyliorrhinus* or laks or lax).ti,ab.

22. (crab or crabs or cancer pagarus).ti,ab.

23. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22

24. randomized controlled trial.pt.

25. controlled clinical trial.pt.

26. randomized.ab.

27. placebo.ab.

28. clinical trials as topic.sh.

29. randomly.ab.

30. trial.ti.

31. 24 or 25 or 26 or 27 or 28 or 29 or 30

32. exp animals/ not humans.sh.

33. 31 not 32

34. 23 and 33

35. limit 34 to ed=20020201‐20160721

36. exp salmoniformes/ or tuna/

37. (fish adj3 capsul*).ti,ab.

38. icosapentaen*.ti,ab.

39. docosapentaen*.ti,ab.

40. (oil* adj3 (purslane or mustard* or candlenut* or stillingia or walnut*)).ti,ab.

41. 36 or 37 or 38 or 39 or 40

42. 33 and 41

43. 35 or 42

Embase Ovid

1. exp salmoniformes/ or tuna/

2. fish oil/

3. linseed oil/

4. linolenic acid/

5. Flax/

6. omega 3 fatty acid/

7. (fish adj3 (diet* or nutrit* or oil* or supplement*)).ti,ab.

8. (oil* adj3 (cod* or marin*)).ti,ab.

9. (omega‐3 or omega3 or (omega* adj5 fat*)).ti,ab.

10. (eicosapentaen* or icosapentaen*).ti,ab.

11. docosahexaen*.ti,ab.

12. (oil* adj3 (flax* or rapeseed* or canola*)).ti,ab.

13. (Linolen* or alpha‐linolen* or alphalinolen*).ti,ab.

14. (perilla* or linseed* or maxepa*).ti,ab.

15. (marin* adj3 lipid*).ti,ab.

16. (naudicelle* or herring* or sild).ti,ab.

17. (clupe* adj3 hareng*).ti,ab.

18. (whitebait or sardine* or sardina* or pilchard* or sprat* or brisling*).ti,ab.

19. (salmo* adj3 trut*).ti,ab.

20. (trout or bloater or kipper* or salmon or mackerel* or scomb* or conger* or tuna or tunny or tunafish or tuna‐fish).ti,ab.

21. (thunnus* or swordfish* or xiphias* or dogfish or scyliorrhinus* or laks or lax).ti,ab.

22. (crab or crabs or (cancer adj3 pagarus)).ti,ab.

23. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22

24. random$.tw.

25. placebo$.tw.

26. (doubl$ adj blind$).tw.

27. (singl$ adj blind$).tw.

28. double blind procedure/

29. randomized controlled trial/

30. single blind procedure/

31. 24 or 25 or 26 or 27 or 28 or 29 or 30

32. (animal/ or nonhuman/) not human/

33. 31 not 32

34. 23 and 33

35. (2002* or 2003* or 2004* or 2005* or 2006* or 2007* or 2008* or 2009* or 2010* or 2011* or 2012* or 2013* or 2014* or 2015* or 2016*).dd,em.

36. 34 and 35

37. exp salmonine/

38. (fish adj3 capsul*).ti,ab.

39. docosapentaen*.ti,ab.

40. (ALA or DHA or DPA or EPA).ti,ab.

41. (algal adj oil*).ti,ab.

42. 37 or 38 or 39 or 40 or 41

43. 33 and 42

44. 36 or 43

Appendix 4. Search strategy 2017 (for allied reviews)

These searches were developed and run to collect relevant trials for the systematic reviews on omega‐6 fats (the update of Hooper 2018) and on total polyunsaturated fatty acid (PUFA) fats (Abdelhamid 2018b) on health. They are shown here as these searches were run with the searches for this review, the identified titles and abstracts de‐duplicated and combined, so that we assessed titles and abstracts for all three reviews together. These searches were each run from database inception, due to the widening of the inclusion criteria, then de‐duplicated with each other. The RCT filter for MEDLINE is the Cochrane sensitivity and precision‐maximising RCT filter, and for Embase, we applied terms as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Lefebvre 2011).

CENTRAL

#1 MeSH descriptor: [Fatty Acids, Essential] explode all trees

#2 MeSH descriptor: [Fatty Acids, Unsaturated] this term only

#3 ((polyunsaturat* or poly‐unsaturat*) near/3 fat*)

#4 (poly* adj4 unsat* near/4 fatty acid*)

#5 PUFA

#6 MeSH descriptor: [Fatty Acids, Omega‐6] explode all trees

#7 omega‐6

#8 (n‐6 near/4 acid*) or ("n 6" near/4 acid*)

#9 linoleic acid*

#10 MeSH descriptor: [Corn Oil] this term only

#11 MeSH descriptor: [Cottonseed Oil] this term only

#12 MeSH descriptor: [Olive Oil] this term only

#13 MeSH descriptor: [Safflower Oil] this term only

#14 MeSH descriptor: [Sesame Oil] this term only

#15 MeSH descriptor: [Soybean Oil] this term only

#16 ((corn or maize or mazola) near/4 oil*)

#17 (cottonseed* or (cotton next seed*))

#18 (olive near/4 oil*)

#19 (safflower near/4 oil*)

#20 (sesame near/4 oil*)

#21 ((soy bean or soybean) near/4 (oil* or fat*))

#22 (so?a near/4 oil*)

#23 so?aoil*

#24 (soy near/4 oil*)

#25 (sunflower near/4 oil*)

#26 helianth*

#27 (grapeseed near/4 oil*)

#28 (canola near/4 oil*)

#29 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 or #18 or #19 or #20 or #21 or #22 or #23 or #24 or #25 or #26 or #27 or #28

MEDLINE Ovid

1. exp fatty acids, essential/

2. fatty acids, unsaturated/

3. ((polyunsaturat* or poly‐unsaturat*) adj3 fat*).ti,ab.

4. (poly* adj4 unsat* adj4 fatty acid*).ti,ab.

5. PUFA.ti,ab.

6. exp fatty acids, omega‐6/

7. omega‐6.ti,ab.

8. (n‐6 adj4 acid*).ti,ab.

9. linoleic acid*.ti,ab.

10. corn oil/ or cottonseed oil/ or olive oil/ or safflower oil/ or sesame oil/ or soybean oil/

11. ((corn or maize or mazola) adj4 oil*).ti,ab.

12. (cottonseed* or (cotton adj seed*)).ti,ab.

13. (olive adj4 oil*).ti,ab.

14. (safflower adj4 oil*).ti,ab.

15. (sesame adj4 oil*).ti,ab.

16. ((soy bean or soybean) adj4 (oil* or fat*)).ti,ab.

17. (so?a adj4 oil*).ti,ab.

18. so?aoil*.ti,ab.

19. (soy adj4 oil*).ti,ab.

20. (sunflower adj4 oil*).ti,ab.

21. helianth*.ti,ab.

22. (grapeseed adj4 oil*).ti,ab.

23. (canola adj4 oil*).ti,ab.

24. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23

25. randomized controlled trial.pt.

26. controlled clinical trial.pt.

27. randomized.ab.

28. placebo.ab.

29. clinical trials as topic.sh.

30. randomly.ab.

31. trial.ti.

32. 25 or 26 or 27 or 28 or 29 or 30 or 31

33. exp animals/ not humans.sh.

34. 32 not 33

35. 24 and 34

Embase Ovid

1. exp essential fatty acid/

2. unsaturated fatty acid/ or docosapentaenoic acid/ or omega 6 fatty acid/ or polyunsaturated fatty acid/

3. ((polyunsaturat* or poly‐unsaturat*) adj3 fat*).ti,ab.

4. (poly* adj4 unsat* adj4 fatty acid*).ti,ab.

5. PUFA.ti,ab.

6. omega‐6.ti,ab.

7. (n‐6 adj4 acid*).ti,ab.

8. linoleic acid*.ti,ab.

9. edible oil/ or canola oil/ or corn oil/ or cotton seed oil/ or olive oil/ or safflower oil/ or safflower oil plus soybean oil/ or sesame seed oil/ or soybean oil/ or sunflower oil/

10. ((corn or maize or mazola) adj4 oil*).ti,ab.

11. (cottonseed* or (cotton adj seed*)).ti,ab.

12. (olive adj4 oil*).ti,ab.

13. (safflower adj4 oil*).ti,ab.

14. (sesame adj4 oil*).ti,ab.

15. ((soy bean or soybean) adj4 (oil* or fat*)).ti,ab.

16. (so?a adj4 oil*).ti,ab.

17. so?aoil*.ti,ab.

18. (soy adj4 oil*).ti,ab.

19. (sunflower adj4 oil*).ti,ab.

20. helianth*.ti,ab.

21. (grapeseed adj4 oil*).ti,ab.

22. (canola adj4 oil*).ti,ab.

23. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22

24. double blind procedure/

25. single blind procedure/

26. randomized controlled trial/

27. ((double* or single*) adj blind*).ti,ab.

28. (random* or placebo*).ti,ab.

29. 24 or 25 or 26 or 27 or 28

30. (animal/ or nonhuman/) not human/

31. 29 not 30

32. 23 and 31

Data and analyses

Comparison 1. High vs low LCn3 omega‐3 fats (primary outcomes).

Outcome or subgroup title No. of studies No. of participants Statistical method Effect size
1 All‐cause mortality (overall) ‐ LCn3 45 143693 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.93, 1.01]
2 All‐cause mortality ‐LCn3 ‐ sensitivity analysis (SA) fixed‐effect 45 143693 Risk Ratio (M‐H, Fixed, 95% CI) 0.97 [0.93, 1.00]
3 All‐cause mortality ‐ LCn3 ‐ SA by summary risk of bias 45   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
3.1 Low summary risk of bias 19 75741 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.95, 1.04]
3.2 Moderate or high summary risk of bias 26 67952 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.87, 1.01]
4 All‐cause mortality ‐ LCn3 ‐ SA by compliance and study size 44   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
4.1 SA ‐ low risk of compliance bias 22 50929 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.90, 1.05]
4.2 SA ‐ 100+ randomised 41 143437 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.93, 1.01]
5 All‐cause mortality ‐ LCn3 ‐ subgroup by dose 45 143693 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.93, 1.01]
5.1 LCn3 ≤ 150 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
5.2 LCn3 > 150 ≤ 250 mg/d 1 407 Risk Ratio (M‐H, Random, 95% CI) 0.77 [0.27, 2.18]
5.3 LCn3 > 250 ≤ 400 mg/d 1 2033 Risk Ratio (M‐H, Random, 95% CI) 0.72 [0.56, 0.92]
5.4 LCn3 > 400 ≤ 2400 mg/d 31 129505 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.94, 1.03]
5.5 LCn3 > 2.4 ≤ 4.4 g/d 10 11466 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.77, 1.04]
5.6 LCn3 > 4.4 g/d 2 282 Risk Ratio (M‐H, Random, 95% CI) 0.33 [0.03, 3.08]
6 All‐cause mortality ‐ LCn3 ‐ subgroup by replacement 45   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
6.1 LCn3 replacing SFA 6 3988 Risk Ratio (M‐H, Random, 95% CI) 0.71 [0.56, 0.91]
6.2 LCn3 replacing MUFA 18 88062 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.93, 1.01]
6.3 LCn3 replacing N‐6 8 2601 Risk Ratio (M‐H, Random, 95% CI) 0.75 [0.51, 1.10]
6.4 LCn3 replacing CHO 1 281 Risk Ratio (M‐H, Random, 95% CI) 0.52 [0.05, 5.65]
6.5 LCn3 replacing nil/low n‐3 placebo 11 47844 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.86, 1.08]
6.6 LCn3 replacement unclear 5 4000 Risk Ratio (M‐H, Random, 95% CI) 1.05 [0.77, 1.43]
7 All‐cause mortality ‐ LCn3 ‐ subgroup by intervention type 45 143693 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.93, 1.01]
7.1 Dietary advice 3 5554 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.60, 1.35]
7.2 Supplemental foods 2 5039 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.84, 1.24]
7.3 Supplements (capsule) 39 132895 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.93, 1.00]
7.4 Any combination 1 205 Risk Ratio (M‐H, Random, 95% CI) 0.65 [0.11, 3.79]
8 All‐cause mortality ‐ LCn3 ‐ subgroup by duration 45 143693 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.93, 1.01]
8.1 Medium duration 1 to < 2 years in study 20 10981 Risk Ratio (M‐H, Random, 95% CI) 1.03 [0.82, 1.29]
8.2 Medium‐long duration: 2 to < 4 years in study 15 29519 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.86, 0.96]
8.3 Long duration: ≥ 4 years in study 10 103193 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.96, 1.05]
9 All‐cause mortality ‐ LCn3 ‐ subgroup by primary or secondary prevention 45 143693 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.93, 1.01]
9.1 Primary CVD prevention 21 83797 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.95, 1.05]
9.2 Secondary CVD prevention 23 51736 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.88, 1.03]
9.3 Mixed primary & secondary CVD prevention 1 8160 Risk Ratio (M‐H, Random, 95% CI) 0.88 [0.75, 1.03]
10 All‐cause mortality ‐ LCn3 ‐ subgroup by statin use 45 143693 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.93, 1.01]
10.1 LCn3 ≥ 50% of control group on statins 11 66834 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.92, 1.04]
10.2 LCn3 < 50% of control group on statins 29 73719 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.91, 1.03]
10.3 LCn3 ‐ use of statins unclear 5 3140 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.58, 1.63]
11 All‐cause mortality ‐ LCn3 ‐ subgroup by baseline TG 45 143693 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.93, 1.01]
11.1 Baseline TG not restricted 43 135413 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.94, 1.01]
11.2 Baseline TG raised 2 8280 Risk Ratio (M‐H, Random, 95% CI) 0.85 [0.54, 1.36]
12 All‐cause mortality ‐ LCn3 ‐ subgroup by baseline DM 45 143693 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.93, 1.01]
12.1 Normal diabetes risk 40 94761 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.92, 1.04]
12.2 Diabetes risk factors at baseline 2 12787 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.90, 1.06]
12.3 Diabetes at baseline 3 36145 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.89, 1.03]
13 Cardiovascular mortality (overall) ‐ LCn3 29 117837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.86, 0.99]
14 CVD mortality ‐ LCn3 ‐ SA fixed‐effect 29 117837 Risk Ratio (M‐H, Fixed, 95% CI) 0.93 [0.88, 0.97]
15 CVD mortality ‐ LCn3 ‐ SA by summary risk of bias 29   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
15.1 Low risk of bias 12 71019 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.88, 1.03]
15.2 Moderate/high risk of bias 17 46818 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.80, 1.02]
16 CVD mortality ‐ LCn3 ‐ SA by compliance and study size 28   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
16.1 SA ‐ low risk of compliance bias 15 47829 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.82, 1.04]
16.2 SA ‐ 100+ randomised 24 91710 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.85, 1.00]
17 CVD mortality ‐ LCn3 ‐ subgroup by dose 30 117938 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.86, 0.99]
17.1 LCn3 ≤ 150 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
17.2 LCn3 > 150 ≤ 250 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
17.3 LCn3 > 250 ≤ 400 mg/d 1 2033 Risk Ratio (M‐H, Random, 95% CI) 0.70 [0.53, 0.91]
17.4 LCn3 > 400 ≤ 2400 mg/d 21 105477 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.88, 1.03]
17.5 LCn3 > 2.4 ≤ 4.4 g/d 6 10146 Risk Ratio (M‐H, Random, 95% CI) 0.84 [0.70, 1.01]
17.6 LCn3 > 4.4 g/d 2 282 Risk Ratio (M‐H, Random, 95% CI) 0.33 [0.03, 3.08]
18 CVD mortality ‐ LCn3 ‐ subgroup by replacement 30   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
18.1 N‐3 replacing SFA 3 2537 Risk Ratio (M‐H, Random, 95% CI) 0.69 [0.53, 0.90]
18.2 N‐3 replacing MUFA 15 86128 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.86, 1.00]
18.3 N‐3 replacing N‐6 4 1435 Risk Ratio (M‐H, Random, 95% CI) 0.69 [0.41, 1.19]
18.4 N‐3 replacing carbohydrates/sugars 1 281 Risk Ratio (M‐H, Random, 95% CI) 0.69 [0.12, 4.07]
18.5 N‐3 replacing nil/low n‐3 placebo 8 27252 Risk Ratio (M‐H, Random, 95% CI) 0.83 [0.74, 0.93]
18.6 Replacement unclear 3 3388 Risk Ratio (M‐H, Random, 95% CI) 0.61 [0.13, 2.88]
19 CVD mortality ‐ LCn3 ‐ subgroup by intervention type 29 117837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.86, 0.99]
19.1 Dietary advice 2 5147 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.52, 1.71]
19.2 Supplemental foods 2 5039 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.72, 1.32]
19.3 Supplements (capsule) 25 107651 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.87, 0.97]
19.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
20 CVD mortality ‐ LCn3 ‐ subgroup by duration 29 117837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.86, 0.99]
20.1 Medium duration 1 to < 2 years in study 11 6712 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.64, 1.37]
20.2 Medium‐long duration: 2 to < 4 years in study 10 26736 Risk Ratio (M‐H, Random, 95% CI) 0.89 [0.82, 0.95]
20.3 Long duration: ≥ 4 years in study 8 84389 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.86, 1.07]
21 CVD mortality ‐ LCn3 ‐ subgroup by primary or secondary prevention 29 117837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.86, 0.99]
21.1 Primary prevention 10 59817 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.86, 1.02]
21.2 Secondary prevention 18 49841 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.83, 1.06]
21.3 Mixed primary & secondary prevention 1 8179 Risk Ratio (M‐H, Random, 95% CI) 0.82 [0.67, 0.99]
22 CVD mortality ‐ LCn3 ‐ subgroup by statin uses 29 117837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.86, 0.99]
22.1 LCn3 ‐ ≥ 50% of control group on statins 8 47653 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.82, 1.04]
22.2 LCn3 ‐ < 50% of control group on statins 18 69296 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.84, 1.03]
22.3 LCn3‐ Use of statins unclear 3 888 Risk Ratio (M‐H, Random, 95% CI) 0.52 [0.13, 2.05]
23 CVD mortality ‐ LCn3 ‐ subgroup by baseline TG 29 117837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.86, 0.99]
23.1 Baseline TG not restricted 27 109538 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.87, 1.01]
23.2 Baseline TG raised 2 8299 Risk Ratio (M‐H, Random, 95% CI) 0.81 [0.67, 0.99]
24 CVD mortality ‐ LCn3 ‐ subgroup by baseline DM 29 117837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.86, 0.99]
24.1 Normal diabetes risk 24 68856 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.84, 1.03]
24.2 Diabetes risk factors at baseline 2 12817 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.88, 1.10]
24.3 Diabetes at baseline 3 36164 Risk Ratio (M‐H, Random, 95% CI) 0.87 [0.75, 1.01]
25 Cardiovascular events (overall) ‐ LCn3 43 140482 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.92, 1.01]
26 CVD events ‐ LCn3 ‐ SA fixed effect 43 140482 Risk Ratio (M‐H, Fixed, 95% CI) 0.96 [0.94, 0.99]
27 CVD events ‐ LCn3 ‐ SA by summary risk of bias 43   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
27.1 Low summary risk of bias 16 73000 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.95, 1.02]
27.2 Moderate or high summary risk of bias 27 67482 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.86, 1.02]
28 CVD events ‐ LCn3 ‐ SA by compliance and study size 42   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
28.1 SA ‐ low risk of compliance bias 18 47699 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.81, 1.06]
28.2 SA ‐ 100+ randomised 38 140162 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.92, 1.02]
29 CVD events ‐ LCn3 ‐ subgroup by dose 43   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
29.1 LCn3 ≤ 150 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
29.2 LCn3 > 150 ≤ 250 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
29.3 LCn3 > 250 ≤ 400 mg/d 1 2033 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.88, 1.05]
29.4 LCn3 > 400 ≤ 2400 mg/d 31 127458 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.93, 1.03]
29.5 LCn3 > 2.4 ≤ 4.4 g/d 9 10644 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.73, 1.14]
29.6 LCn3 > 4.4 g/d 3 422 Risk Ratio (M‐H, Random, 95% CI) 1.09 [0.65, 1.81]
30 CVD events ‐ LCn3 ‐ subgroup by replacement 43   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
30.1 N‐3 replacing SFA 4 2888 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.87, 1.04]
30.2 N‐3 replacing MUFA 18 86416 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.94, 1.00]
30.3 N‐3 replacing n‐6 7 2180 Risk Ratio (M‐H, Random, 95% CI) 1.07 [0.86, 1.32]
30.4 N‐3 replacing carbohydrates/sugars 1 258 Risk Ratio (M‐H, Random, 95% CI) 0.68 [0.12, 3.98]
30.5 N‐3 replacing nil/low n‐3 placebo 13 48169 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.84, 1.06]
30.6 Replacement unclear 4 3631 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.50, 1.64]
31 CVD events ‐ LCn3 ‐ subgroup by intervention type 43 140482 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.92, 1.01]
31.1 Dietary advice 3 5248 Risk Ratio (M‐H, Random, 95% CI) 1.13 [0.86, 1.49]
31.2 Supplemental foods 2 5039 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.89, 1.17]
31.3 Supplements (capsule) 38 130195 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.90, 1.00]
31.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
32 CVD events ‐ LCn3 ‐ subgroup by duration 43 140482 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.92, 1.01]
32.1 Medium duration 1 to < 2 years in study 19 8396 Risk Ratio (M‐H, Random, 95% CI) 0.86 [0.69, 1.07]
32.2 Medium‐long duration: 2 to < 4 years in study 15 29052 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.95, 1.03]
32.3 Long duration: ≥ 4 years in study 9 103034 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.88, 1.03]
33 CVD events ‐ LCn3 ‐ subgroup by primary or secondary prevention 43 140482 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.92, 1.01]
33.1 Primary prevention of CVD 19 81391 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.91, 1.02]
33.2 Secondary prevention 23 50912 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.93, 1.05]
33.3 Mixed primary & secondary prevention 1 8179 Risk Ratio (M‐H, Random, 95% CI) 0.76 [0.68, 0.85]
34 CVD events ‐ LCn3 ‐ subgroup by statin use 43 140482 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.92, 1.01]
34.1 LCn3 ‐ ≥ 50% of control group on statins 11 66333 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.86, 1.01]
34.2 LCn3 ‐ < 50% of control group on statins 25 71031 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.93, 1.04]
34.3 LCn3 ‐ use of statins unclear 7 3118 Risk Ratio (M‐H, Random, 95% CI) 0.85 [0.51, 1.43]
35 CVD events ‐ LCn3 ‐ subgroup by baseline TG 43 140478 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.91, 1.00]
35.1 Baseline TG not restricted 41 135344 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.92, 1.01]
35.2 Baseline TG raised 3 5134 Risk Ratio (M‐H, Random, 95% CI) 0.78 [0.67, 0.90]
36 CVD events ‐ LCn3 ‐ subgroup by baseline diabetes 43 140479 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.91, 1.00]
36.1 Normal diabetes risk 37 91281 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.93, 1.04]
36.2 Diabetes risk factors at baseline 4 16426 Risk Ratio (M‐H, Random, 95% CI) 0.88 [0.72, 1.07]
36.3 Diabetes at baseline 3 32772 Risk Ratio (M‐H, Random, 95% CI) 0.89 [0.76, 1.05]
37 Coronary heart disease mortality (overall) ‐ LCn3 24 127378 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.81, 1.00]
38 CHD mortality ‐ LCn3 ‐ SA fixed effect 24 127378 Risk Ratio (M‐H, Fixed, 95% CI) 0.92 [0.86, 0.98]
39 CHD mortality ‐ LCn3 ‐ SA by summary risk of bias 24   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
39.1 Low summary risk of bias 10 70259 Risk Ratio (M‐H, Random, 95% CI) 0.89 [0.76, 1.04]
39.2 Moderate or high summary risk of bias 14 57119 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.79, 1.06]
40 CHD mortality ‐ LCn3 ‐ SA by compliance and study size 23   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
40.1 SA ‐ low risk of compliance bias 10 38809 Risk Ratio (M‐H, Random, 95% CI) 0.83 [0.67, 1.03]
40.2 SA ‐ 100+ randomised 22 114762 Risk Ratio (M‐H, Random, 95% CI) 0.87 [0.77, 0.98]
41 CHD mortality ‐ LCn3 ‐ SA omitting cardiac death 18 106676 Risk Ratio (M‐H, Random, 95% CI) 0.81 [0.73, 0.90]
41.1 Low risk of bias 7 53373 Risk Ratio (M‐H, Random, 95% CI) 0.81 [0.68, 0.97]
41.2 Moderate/high risk of bias 11 53303 Risk Ratio (M‐H, Random, 95% CI) 0.82 [0.71, 0.94]
42 CHD mortality ‐ LCn3 ‐ subgroup by dose 24 127378 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.81, 1.00]
42.1 LCn3 ≤ 250 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
42.2 LCn3 > 250 ≤ 400 mg/d 2 5147 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.50, 1.74]
42.3 LCn3 > 400 ≤ 2400 mg/d 18 121329 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.83, 0.99]
42.4 LCn3 > 2.4 ≤ 4.4g/d 3 822 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.49, 1.78]
42.5 LCn3 > 4.4g/d 1 80 Risk Ratio (M‐H, Random, 95% CI) 0.32 [0.01, 7.57]
43 CHD mortality ‐ LCn3 ‐ subgroup by replacement 23   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
43.1 N‐3 replacing SFA 3 2514 Risk Ratio (M‐H, Random, 95% CI) 0.67 [0.51, 0.88]
43.2 N‐3 replacing MUFA 13 85492 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.81, 1.00]
43.3 N‐3 replacing n‐6 3 1409 Risk Ratio (M‐H, Random, 95% CI) 0.60 [0.33, 1.09]
43.4 N‐3 replacing carbohydrates/sugars 1 258 Risk Ratio (M‐H, Random, 95% CI) 0.34 [0.01, 8.23]
43.5 N‐3 replacing nil/low n‐3 placebo 7 37651 Risk Ratio (M‐H, Random, 95% CI) 0.81 [0.70, 0.94]
44 CHD mortality ‐ LCn3 ‐ subgroup by intervention type 24 127378 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.81, 1.00]
44.1 Dietary advice 2 5147 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.50, 1.74]
44.2 Supplemental foods 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.69, 1.33]
44.3 Supplements (capsule) 21 117394 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.83, 0.98]
44.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
45 CHD mortality ‐ LCn3 ‐ subgroup by duration 24 127378 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.81, 1.00]
45.1 Medium duration 1 to < 2 years in study 7 5978 Risk Ratio (M‐H, Random, 95% CI) 0.85 [0.58, 1.23]
45.2 Medium‐long duration: 2 to < 4 years in study 9 26545 Risk Ratio (M‐H, Random, 95% CI) 0.85 [0.77, 0.94]
45.3 Long duration: ≥ 4 years in study 8 94855 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.83, 1.16]
46 CHD mortality ‐ LCn3 ‐ subgroup by primary or secondary prevention 24 127378 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.81, 1.00]
46.1 Primary prevention 8 77676 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.75, 1.07]
46.2 secondary prevention 16 49702 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.79, 1.04]
47 CHD mortality ‐ LCn3 ‐ subgroup by statin use 24 127378 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.81, 1.00]
47.1 LCn3 ‐ ≥ 50% of control group on statins 7 58041 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.78, 1.09]
47.2 LCn3 ‐ < 50% of control group on statins 16 69079 Risk Ratio (M‐H, Random, 95% CI) 0.89 [0.77, 1.03]
47.3 LCn3 ‐ use of statins unclear 1 258 Risk Ratio (M‐H, Random, 95% CI) 0.34 [0.01, 8.23]
48 CHD mortality ‐ LCn3 ‐ subgroup by CAD history 24 127378 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.81, 1.00]
48.1 Previous CAD 11 29074 Risk Ratio (M‐H, Random, 95% CI) 0.88 [0.71, 1.10]
48.2 No previous CAD 13 98304 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.84, 1.03]
49 CHD mortality ‐ LCn3 ‐ subgroup by baseline TG 24 127378 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.81, 1.00]
49.1 Baseline TG not restricted 23 127258 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.81, 1.00]
49.2 Baseline TG raised 1 120 Risk Ratio (M‐H, Random, 95% CI) 0.17 [0.01, 4.05]
50 CHD mortality ‐ LCn3 ‐ subgroup by baseline DM 24 127378 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.81, 1.00]
50.1 Normal diabetes risk 20 86599 Risk Ratio (M‐H, Random, 95% CI) 0.87 [0.77, 0.99]
50.2 Diabetes risk factors at baseline 2 12794 Risk Ratio (M‐H, Random, 95% CI) 1.07 [0.92, 1.25]
50.3 Diabetes at baseline 2 27985 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.67, 1.25]
51 Coronary heart disease events (overall) ‐ LCn3 32 134116 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.85, 0.97]
52 CHD events ‐ LCn3 ‐ SA fixed effect 32 134116 Risk Ratio (M‐H, Fixed, 95% CI) 0.93 [0.89, 0.96]
53 CHD events ‐ LCn3 ‐ SA by summary risk of bias 32   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
53.1 Low summary risk of bias 14 71578 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.82, 1.05]
53.2 Moderate or high summary risk of bias 18 62538 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.83, 0.97]
54 CHD events ‐ LCn3 ‐ SA by compliance and study size 31   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
54.1 SA ‐ low risk of compliance bias 14 47497 Risk Ratio (M‐H, Random, 95% CI) 0.84 [0.74, 0.95]
54.2 SA ‐ 100+ randomised 29 133899 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.84, 0.95]
55 CHD events ‐ LCn3 ‐ subgroup by dose 32   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
55.1 LCn3 ≤ 150 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
55.2 LCn3 > 150 ≤ 250 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
55.3 LCn3 > 250 ≤ 400 mg/d 1 2033 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.82, 1.04]
55.4 LCn3 > 400 ≤ 2400 mg/d 23 122081 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.87, 0.99]
55.5 LCn3 > 2.4 ≤ 4.4 g/d 6 9655 Risk Ratio (M‐H, Random, 95% CI) 0.88 [0.63, 1.22]
55.6 LCn3 > 4.4 g/d 3 422 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.54, 1.85]
56 CHD events ‐ LCn3 ‐ subgroup by replacement 32   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
56.1 N‐3 replacing SFA 3 2514 Risk Ratio (M‐H, Random, 95% CI) 0.53 [0.16, 1.75]
56.2 N‐3 replacing MUFA 17 86305 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.89, 1.02]
56.3 N‐3 replacing n‐6 4 1549 Risk Ratio (M‐H, Random, 95% CI) 1.09 [0.78, 1.53]
56.4 N‐3 replacing carbohydrates/sugars 1 258 Risk Ratio (M‐H, Random, 95% CI) 0.11 [0.01, 2.07]
56.5 N‐3 replacing nil/low n‐3 placebo 9 46105 Risk Ratio (M‐H, Random, 95% CI) 0.81 [0.72, 0.92]
56.6 Replacement unclear 2 445 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.45, 2.17]
57 CHD events ‐ LCn3 ‐ subgroup by intervention type 32 134116 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.85, 0.97]
57.1 Dietary advice 2 2134 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.67, 1.52]
57.2 Supplemental foods 2 5039 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.75, 1.20]
57.3 Supplements (capsule) 28 126943 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.83, 0.98]
57.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
58 CHD events ‐ LCn3 ‐ subgroup by duration 32 134116 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.85, 0.97]
58.1 Medium duration 1 to < 2 years in study 11 7009 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.64, 1.25]
58.2 Medium‐long duration: 2 to < 4 years in study 13 27187 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.90, 0.99]
58.3 Long duration: ≥ 4 years in study 8 99920 Risk Ratio (M‐H, Random, 95% CI) 0.85 [0.74, 0.98]
59 CHD events ‐ LCn3 ‐ subgroup by primary or secondary prevention 32 134116 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.85, 0.97]
59.1 Primary prevention of CVD 13 78716 Risk Ratio (M‐H, Random, 95% CI) 0.87 [0.75, 1.02]
59.2 Secondary prevention of CVD 18 47221 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.91, 0.99]
59.3 Mixed primary & secondary prevention 1 8179 Risk Ratio (M‐H, Random, 95% CI) 0.70 [0.60, 0.82]
60 CHD events ‐ LCn3 ‐ subgroup by statin use 32 134116 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.85, 0.97]
60.1 LCn3 ‐ ≥ 50% of control group on statins 11 66679 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.80, 1.07]
60.2 LCn3 ‐ < 50% of control group on statins 18 66545 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.89, 0.97]
60.3 LCn3 ‐ use of statins unclear 3 892 Risk Ratio (M‐H, Random, 95% CI) 0.65 [0.11, 3.83]
61 CHD events ‐ LCn3 subgroup by CAD history 32 134116 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.85, 0.97]
61.1 Previous CAD 13 26409 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.87, 1.02]
61.2 No previous CAD 18 99528 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.86, 1.00]
61.3 Mixed ‐ some previous CAD 1 8179 Risk Ratio (M‐H, Random, 95% CI) 0.70 [0.60, 0.82]
62 CHD events ‐ LCn3 ‐ subgroup by baseline TG 32 134116 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.85, 0.97]
62.1 Baseline TG not restricted 29 125753 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.89, 0.99]
62.2 Baseline TG raised 3 8363 Risk Ratio (M‐H, Random, 95% CI) 0.70 [0.60, 0.82]
63 CHD events ‐ LCn3 ‐ subgroup by baseline DM 32 134116 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.85, 0.97]
63.1 Normal diabetes risk 26 84915 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.88, 0.97]
63.2 Diabetes risk factors at baseline 3 13037 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.45, 1.95]
63.3 Diabetes at baseline 3 36164 Risk Ratio (M‐H, Random, 95% CI) 0.84 [0.70, 1.01]
64 Stroke (overall) ‐ LCn3 31 138888 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.94, 1.12]
65 Stroke ‐ LCn3 ‐ SA fixed effect 31 138888 Risk Ratio (M‐H, Fixed, 95% CI) 1.01 [0.94, 1.09]
66 Stroke ‐ LCn3 ‐ SA by summary risk of bias 31   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
66.1 Low risk of bias 14 73390 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.90, 1.09]
66.2 Moderate/high risk of bias 17 65498 Risk Ratio (M‐H, Random, 95% CI) 1.05 [0.87, 1.26]
67 Stroke ‐ LCn3 ‐ SA by compliance and study size 30   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
67.1 SA ‐ low risk of compliance bias 14 48501 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.80, 1.19]
67.2 SA ‐ 100+ randomised 29 138761 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.93, 1.12]
68 Stroke ‐ LCn3 ‐ subgroup by stroke type 15   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
68.1 Ischaemic stroke ‐ LCn3 10 69090 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.79, 1.20]
68.2 Haemorrhagic stroke ‐ LCn3 10 70695 Risk Ratio (M‐H, Random, 95% CI) 1.23 [0.93, 1.64]
68.3 Transient ischaemic attack (TIA) 6 13211 Risk Ratio (M‐H, Random, 95% CI) 1.10 [0.76, 1.60]
69 Stroke ‐ LCn3 ‐ subgroup by dose 31 138888 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.94, 1.12]
69.1 LCn3 ≤ 150 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
69.2 LCn3 > 150 ≤ 250 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
69.3 LCn3 > 250 ≤ 400 mg/d 1 2033 Risk Ratio (M‐H, Random, 95% CI) 0.45 [0.14, 1.44]
69.4 LCn3 > 400 ≤ 2400 mg/d 26 127686 Risk Ratio (M‐H, Random, 95% CI) 1.04 [0.96, 1.12]
69.5 LCn3 > 2.4 ≤ 4.4 g/d 2 8789 Risk Ratio (M‐H, Random, 95% CI) 0.73 [0.57, 0.94]
69.6 LCn3 > 4.4 g/d 2 380 Risk Ratio (M‐H, Random, 95% CI) 6.58 [0.78, 55.16]
70 Stroke ‐ LCn3 ‐ subgroup by replacement 31   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
70.1 N‐3 replacing SFA 3 2514 Risk Ratio (M‐H, Random, 95% CI) 0.53 [0.19, 1.50]
70.2 N‐3 replacing MUFA 16 86603 Risk Ratio (M‐H, Random, 95% CI) 1.03 [0.94, 1.14]
70.3 N‐3 replacing n‐6 3 1179 Risk Ratio (M‐H, Random, 95% CI) 2.08 [0.18, 24.31]
70.4 N‐3 replacing carbohydrates/sugars 1 258 Risk Ratio (M‐H, Random, 95% CI) 0.34 [0.01, 8.23]
70.5 N‐3 replacing nil/low n‐3 placebo 9 47398 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.80, 1.17]
70.6 Replacement unclear 2 3450 Risk Ratio (M‐H, Random, 95% CI) 1.21 [0.61, 2.43]
71 Stroke ‐ LCn3 ‐ subgroup by intervention type 31 138888 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.94, 1.12]
71.1 Dietary advice 3 5248 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.42, 2.05]
71.2 Supplemental foods 1 4837 Risk Ratio (M‐H, Random, 95% CI) 1.11 [0.47, 2.62]
71.3 Supplements (capsule) 27 128803 Risk Ratio (M‐H, Random, 95% CI) 1.03 [0.93, 1.13]
71.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
72 Stroke ‐ LCn3 ‐ subgroup by duration 31 138888 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.94, 1.12]
72.1 Medium duration 1 to < 2 years in study 11 7467 Risk Ratio (M‐H, Random, 95% CI) 1.35 [0.86, 2.12]
72.2 Medium‐long duration: 2 to < 4 years in study 11 28387 Risk Ratio (M‐H, Random, 95% CI) 1.15 [0.93, 1.41]
72.3 Long duration: ≥ 4 years in study 9 103034 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.89, 1.08]
73 Stroke ‐ LCn3 ‐ subgroup by primary or secondary prevention 31 138888 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.94, 1.12]
73.1 Primary prevention of CVD 11 80683 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.90, 1.07]
73.2 Secondary prevention of CVD 19 50026 Risk Ratio (M‐H, Random, 95% CI) 1.21 [1.05, 1.40]
73.3 Mixed primary & secondary prevention 1 8179 Risk Ratio (M‐H, Random, 95% CI) 0.73 [0.57, 0.95]
74 Stroke ‐ LCn3 ‐ subgroup by statin use 31 138888 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.94, 1.12]
74.1 LCn3 ‐ ≥ 50% of control group on statins 10 66621 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.84, 1.10]
74.2 LCn3 ‐ < 50% of control group on statins 18 70870 Risk Ratio (M‐H, Random, 95% CI) 1.14 [1.01, 1.29]
74.3 LCn3 ‐ use of statins unclear 3 1397 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.38, 2.34]
75 Stroke ‐ LCn3 ‐ subgroup by baseline TG 31 138888 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.94, 1.12]
75.1 Baseline TG not restricted 29 130373 Risk Ratio (M‐H, Random, 95% CI) 1.04 [0.96, 1.12]
75.2 Baseline TG raised 2 8515 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.25, 4.00]
76 Stroke ‐ LCn3 ‐ subgroup by baseline DM 31 138888 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.94, 1.12]
76.1 Normal diabetes risk 25 89594 Risk Ratio (M‐H, Random, 95% CI) 1.10 [0.98, 1.22]
76.2 Diabetes risk factors at baseline 3 13130 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.80, 1.08]
76.3 Diabetes at baseline 3 36164 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.73, 1.30]
77 Arrythmia (overall) ‐ LCn3 30 77990 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.92, 1.06]
78 Arrythmia ‐ LCn3 ‐ SA fixed effects 30 77990 Risk Ratio (M‐H, Fixed, 95% CI) 1.03 [0.98, 1.09]
79 Arrhythmia‐ LCn3 ‐ SA by summary risk of bias 30 77990 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.92, 1.06]
79.1 Low summary risk of bias 12 41816 Risk Ratio (M‐H, Random, 95% CI) 1.09 [0.99, 1.20]
79.2 Moderate to high risk of bias 18 36174 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.86, 1.04]
80 Arrhythmia‐ LCn3 ‐ SA by compliance and study size 29   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
80.1 SA ‐ low risk of compliance bias 12 21628 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.88, 1.12]
80.2 SA ‐ 100+ randomised 29 77943 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.91, 1.06]
81 Arrhythmia ‐ LCn3 ‐ subgroup by new or recurrent 30   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
81.1 New arrhythmia 19 74111 Risk Ratio (M‐H, Random, 95% CI) 1.09 [1.01, 1.17]
81.2 Recurrent arrhythmia 12 4425 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.84, 1.03]
82 Arrhythmia ‐ LCn3 ‐ subgroup by fatality 9   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
82.1 Fatal arrhythmias ‐ LCn3 2 12938 Risk Ratio (M‐H, Random, 95% CI) 1.11 [0.95, 1.31]
82.2 Non‐fatal arrhythmias ‐ LCn3 8 2079 Risk Ratio (M‐H, Random, 95% CI) 0.74 [0.57, 0.96]
83 Arrhythmia ‐ LCn3 ‐ subgroup by dose 30 77990 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.92, 1.06]
83.1 LCn3 ≤ 150 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
83.2 LCn3 > 150 ≤ 250 mg/d 1 407 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.90, 1.12]
83.3 LCn3 > 250 ≤ 400 mg/d 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
83.4 LCn3 > 400 ≤ 2400 mg/d 20 67015 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.90, 1.08]
83.5 LCn3 > 2.4 ≤ 4.4 g/d 5 9790 Risk Ratio (M‐H, Random, 95% CI) 0.88 [0.58, 1.32]
83.6 LCn3 > 4.4 g/d 2 342 Risk Ratio (M‐H, Random, 95% CI) 1.10 [0.32, 3.83]
83.7 Unclear LCn3 dose 2 436 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.76, 1.28]
84 Arrhythmia ‐ LCn3 ‐ subgroup by replacement 30   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
84.1 N‐3 replacing SFA 2 632 Risk Ratio (M‐H, Random, 95% CI) 0.74 [0.10, 5.67]
84.2 N‐3 replacing MUFA 15 58652 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.91, 1.11]
84.3 N‐3 replacing n‐6 4 1302 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.86, 1.16]
84.4 N‐3 replacing carbohydrates/sugars 1 258 Risk Ratio (M‐H, Random, 95% CI) 0.34 [0.04, 3.21]
84.5 N‐3 replacing nil/low n‐3 placebo 5 16569 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.70, 1.30]
84.6 Replacement unclear 5 1381 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.91, 1.08]
85 Arrhythmia ‐ LCn3 ‐ subgroup by intervention type 30 77990 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.92, 1.06]
85.1 Dietary advice 2 508 Risk Ratio (M‐H, Random, 95% CI) 0.87 [0.44, 1.72]
85.2 Supplemental foods 2 5039 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.67, 1.26]
85.3 Supplements (capsule) 26 72443 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.91, 1.08]
85.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
86 Arrhythmia ‐ LCn3 ‐ subgroup by duration 30 77990 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.92, 1.06]
86.1 Medium duration 1 to < 2 years in study 18 9088 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.84, 1.04]
86.2 Medium‐long duration: 2 to < 4 years in study 7 17701 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.84, 1.10]
86.3 Long duration: ≥ 4 years in study 5 51201 Risk Ratio (M‐H, Random, 95% CI) 1.13 [1.03, 1.25]
87 Arrhythmia ‐ LCn3 ‐ subgroup by primary or secondary prevention3 30 77990 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.92, 1.06]
87.1 Primary prevention 10 30580 Risk Ratio (M‐H, Random, 95% CI) 1.10 [0.98, 1.22]
87.2 Secondary prevention 19 39231 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.87, 1.03]
87.3 Mixed primary & secondary prevention 1 8179 Risk Ratio (M‐H, Random, 95% CI) 1.22 [0.99, 1.50]
88 Arrhythmia ‐ LCn3 ‐ subgroup by statin use 30 77990 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.92, 1.06]
88.1 LCn3 ‐ ≥ 50% of control group on statins 7 47438 Risk Ratio (M‐H, Random, 95% CI) 1.10 [1.00, 1.21]
88.2 LCn3 ‐ < 50% of control group on statins 19 29467 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.85, 1.04]
88.3 LCn3 ‐ use of statins unclear 4 1085 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.80, 1.18]
89 Arrythmia ‐ LCn3 ‐ subgroup by baseline TG 30 77990 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.92, 1.06]
89.1 Baseline TG not restricted 28 69475 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.90, 1.05]
89.2 Baseline TG raised 2 8515 Risk Ratio (M‐H, Random, 95% CI) 1.21 [0.99, 1.50]
90 Arrythmia ‐ LCn3 ‐ subgroup by baseline DM 30 77990 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.92, 1.06]
90.1 Normal diabetes risk 24 28696 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.87, 1.03]
90.2 Diabetes risk factors at baseline 3 13130 Risk Ratio (M‐H, Random, 95% CI) 1.10 [0.93, 1.29]
90.3 Diabetes at baseline 3 36164 Risk Ratio (M‐H, Random, 95% CI) 1.16 [1.02, 1.31]

1.65. Analysis.

1.65

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 65 Stroke ‐ LCn3 ‐ SA fixed effect.

1.67. Analysis.

1.67

Comparison 1 High vs low LCn3 omega‐3 fats (primary outcomes), Outcome 67 Stroke ‐ LCn3 ‐ SA by compliance and study size.

Comparison 2. High vs low LCn3 omega‐3 fats (secondary outcomes).

Outcome or subgroup title No. of studies No. of participants Statistical method Effect size
1 MACCEs ‐ LCn3 5 34730 Risk Ratio (M‐H, Random, 95% CI) 1.03 [0.97, 1.09]
2 Myocardial infarction (overall) ‐ LCn3 27 133012 Risk Ratio (M‐H, Random, 95% CI) 0.88 [0.81, 0.96]
3 Total MI ‐ LCn3 ‐ SA fixed effects 27 133012 Risk Ratio (M‐H, Fixed, 95% CI) 0.89 [0.83, 0.94]
4 Total MI ‐ LCn3 ‐ SA by summary risk of bias 27   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
4.1 Low summary risk of bias 13 71376 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.83, 1.05]
4.2 Moderate or high summary risk of bias 14 61636 Risk Ratio (M‐H, Random, 95% CI) 0.84 [0.76, 0.94]
5 Total MI ‐ LCn3 ‐ SA by compliance and study size 26   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
5.1 SA ‐ low risk of compliance bias 12 47052 Risk Ratio (M‐H, Random, 95% CI) 0.81 [0.70, 0.94]
5.2 SA ‐ 100+ randomised 24 121545 Risk Ratio (M‐H, Random, 95% CI) 0.87 [0.80, 0.96]
6 Total MI ‐ LCn3 ‐ subgroup by fatality 27   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
6.1 Fatal MI 16 86342 Risk Ratio (M‐H, Random, 95% CI) 0.73 [0.52, 1.03]
6.2 Non‐fatal MI 25 132443 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.84, 1.02]
7 Sudden cardiac death (overall) ‐ LCn3 15 73183 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.77, 1.11]
8 Angina ‐ LCn3 13 48621 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.92, 1.06]
8.1 Low summary risk of bias 6 21355 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.90, 1.08]
8.2 Moderate or high risk of bias 7 27266 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.88, 1.10]
9 Heart failure ‐ LCn3 17 73303 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.87, 1.02]
9.1 Low summary risk of bias 7 39656 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.89, 1.05]
9.2 Moderate to high risk of bias 10 33647 Risk Ratio (M‐H, Random, 95% CI) 0.85 [0.69, 1.04]
10 Revascularisation ‐ LCn3 24 129382 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.86, 1.00]
10.1 Low summary risk of bias 8 64545 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.91, 1.01]
10.2 Moderate to high risk of bias 16 64837 Risk Ratio (M‐H, Random, 95% CI) 0.88 [0.78, 1.00]
11 Peripheral arterial disease ‐ LCn3 7 57214 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.76, 1.18]
12 PAD ‐ LCn3 ‐ SA fixed effects 7 57214 Risk Ratio (M‐H, Fixed, 95% CI) 0.95 [0.76, 1.18]
13 PAD ‐ LCn3 ‐ SA by summary risk of bias 7   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
13.1 Low summary risk of bias 2 12738 Risk Ratio (M‐H, Random, 95% CI) 1.10 [0.75, 1.62]
13.2 Moderate or high risk of bias 5 44476 Risk Ratio (M‐H, Random, 95% CI) 0.89 [0.68, 1.15]
14 PAD ‐ LCn3 ‐ SA compliance and study size 7   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
14.1 SA compliance 2 8381 Risk Ratio (M‐H, Random, 95% CI) 1.05 [0.58, 1.87]
14.2 SA study size 100+ 7 57214 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.76, 1.18]
15 Acute coronary syndrome ‐ LCn3 2 2703 Risk Ratio (M‐H, Random, 95% CI) 1.19 [0.71, 2.00]
15.1 LCn3 2 2703 Risk Ratio (M‐H, Random, 95% CI) 1.19 [0.71, 2.00]
16 Body weight, kg ‐ LCn3 14 17000 Mean Difference (IV, Random, 95% CI) 0.00 [‐0.69, 0.70]
17 Weight, kg ‐ LCn3 ‐ SA fixed effects 14 17000 Mean Difference (IV, Fixed, 95% CI) 0.19 [‐0.19, 0.57]
18 Weight, kg ‐ LCn3 ‐ SA by summary risk of bias 14   Mean Difference (IV, Random, 95% CI) Subtotals only
18.1 Low summary risk of bias 8 16406 Mean Difference (IV, Random, 95% CI) ‐0.10 [‐0.94, 0.75]
18.2 Moderate or high risk of bias 6 594 Mean Difference (IV, Random, 95% CI) 0.27 [‐1.08, 1.63]
19 Weight, kg ‐ LCn3 ‐ SA by compliance and study size 12   Mean Difference (IV, Random, 95% CI) Subtotals only
19.1 SA ‐ low risk of compliance bias 7 828 Mean Difference (IV, Random, 95% CI) 0.58 [‐0.52, 1.69]
19.2 SA ‐ 100+ randomised 9 16733 Mean Difference (IV, Random, 95% CI) 0.06 [‐0.67, 0.79]
20 Weight, kg ‐ LCn3 ‐ subgroup by dose 14   Mean Difference (IV, Random, 95% CI) Subtotals only
20.1 LCn3 ≤ 150 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
20.2 LCn3 > 150 ≤ 250 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
20.3 LCn3 > 250 ≤ 400 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
20.4 LCn3 > 400 ≤ 2400 mg/d 9 16368 Mean Difference (IV, Random, 95% CI) ‐0.34 [‐1.14, 0.46]
20.5 LCn3 > 2.4 ≤ 4.4 g/d 4 481 Mean Difference (IV, Random, 95% CI) 0.11 [‐2.58, 2.80]
20.6 LCn3 > 4.4 g/d 2 261 Mean Difference (IV, Random, 95% CI) 1.51 [0.28, 2.75]
21 Weight, kg ‐ LCn3 ‐ subgroup by replacement 14   Mean Difference (IV, Random, 95% CI) Subtotals only
21.1 N‐3 replacing SFA 2 433 Mean Difference (IV, Random, 95% CI) ‐2.51 [‐4.30, ‐0.72]
21.2 N‐3 replacing MUFA 8 16036 Mean Difference (IV, Random, 95% CI) 0.17 [‐0.34, 0.68]
21.3 N‐3 replacing n‐6 1 41 Mean Difference (IV, Random, 95% CI) ‐1.3 [‐3.83, 1.23]
21.4 N‐3 replacing carbs/sugars 1 258 Mean Difference (IV, Random, 95% CI) ‐2.70 [‐4.75, ‐0.65]
21.5 N‐3 replacing nil/low n‐3 placebo 1 240 Mean Difference (IV, Random, 95% CI) 0.44 [‐1.11, 1.99]
21.6 Replacement unclear 3 425 Mean Difference (IV, Random, 95% CI) 1.46 [0.23, 2.68]
22 Weight, kg ‐ LCn3 ‐ subgroup by intervention type 14 17000 Mean Difference (IV, Random, 95% CI) 0.00 [‐0.69, 0.70]
22.1 Dietary advice 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
22.2 Supplemental foods 1 202 Mean Difference (IV, Random, 95% CI) 1.5 [0.25, 2.75]
22.3 Supplement (capsule) 11 16726 Mean Difference (IV, Random, 95% CI) ‐0.14 [‐0.82, 0.53]
22.4 Any combination 2 72 Mean Difference (IV, Random, 95% CI) ‐0.43 [‐6.47, 5.61]
23 Weight, kg ‐ LCn3 ‐ subgroup by duration 14   Mean Difference (IV, Random, 95% CI) Subtotals only
23.1 Medium duration 1 to < 2 years in study 8 840 Mean Difference (IV, Random, 95% CI) ‐0.54 [‐2.21, 1.12]
23.2 Medium‐long duration: 2 to < 4 years in study 4 676 Mean Difference (IV, Random, 95% CI) 0.81 [‐0.30, 1.92]
23.3 Long duration ≥ 4 years in study 2 15484 Mean Difference (IV, Random, 95% CI) 0.03 [‐0.53, 0.59]
24 Weight, kg ‐ LCn3 ‐ subgroup by primary or secondary prevention 14   Mean Difference (IV, Random, 95% CI) Subtotals only
24.1 Primary CVD prevention 11 16526 Mean Difference (IV, Random, 95% CI) ‐0.04 [‐0.87, 0.78]
24.2 Secondary CVD prevention 3 474 Mean Difference (IV, Random, 95% CI) 0.16 [‐1.24, 1.56]
25 Weight, kg ‐ LCn3 ‐ subgroup by statin use 14   Mean Difference (IV, Random, 95% CI) Subtotals only
25.1 LCn3 ‐ ≥ 50% of control group on statins 4 15819 Mean Difference (IV, Random, 95% CI) 0.11 [‐0.42, 0.63]
25.2 LCn3 ‐ < 50% of control group on statins 5 614 Mean Difference (IV, Random, 95% CI) 0.47 [‐0.66, 1.60]
25.3 LCn3 ‐ use of statins unclear 5 567 Mean Difference (IV, Random, 95% CI) ‐1.51 [‐3.30, 0.27]
26 Body mass index, kg/m² ‐ LCn3 15 15474 Mean Difference (IV, Random, 95% CI) 0.06 [‐0.14, 0.25]
27 BMI, kg/m² ‐ LCn3 ‐ SA fixed effects 15 15474 Mean Difference (IV, Random, 95% CI) 0.06 [‐0.14, 0.25]
28 BMI, kg/m²‐ LCn3 ‐ SA by summary risk of bias 15   Mean Difference (IV, Random, 95% CI) Subtotals only
28.1 Low risk of bias 5 14190 Mean Difference (IV, Random, 95% CI) ‐0.01 [‐0.36, 0.33]
28.2 Moderate/high risk of bias 10 1284 Mean Difference (IV, Random, 95% CI) 0.05 [‐0.12, 0.21]
29 BMI, kg/m²‐ LCn3 ‐ SA by compliance and study size 11   Mean Difference (IV, Random, 95% CI) Subtotals only
29.1 SA ‐ low risk of compliance bias 5 1848 Mean Difference (IV, Random, 95% CI) 0.09 [‐0.21, 0.38]
29.2 SA ‐ 100+ randomised 10 15222 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.12, 0.15]
30 BMI, kg/m² ‐ LCn3 ‐ subgroup by dose 15   Mean Difference (IV, Random, 95% CI) Subtotals only
30.1 LCn3 > 150 ≤ 250 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
30.2 LCn3 ≤ 150 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
30.3 LCn3 > 250 ≤ 400 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
30.4 LCn3 > 400 ≤ 2400 mg/d 11 14789 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.11, 0.13]
30.5 LCn3 > 2.4 ≤ 4.4 g/d 4 685 Mean Difference (IV, Random, 95% CI) 1.08 [0.07, 2.10]
30.6 LCn3 > 4.4 g/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
31 BMI, kg/m² ‐ LCn3 ‐ subgroup by replacement 15   Mean Difference (IV, Random, 95% CI) Subtotals only
31.1 N‐3 replacing SFA 1 258 Mean Difference (IV, Random, 95% CI) ‐0.60 [‐1.14, ‐0.06]
31.2 N‐3 replacing MUFA 7 14180 Mean Difference (IV, Random, 95% CI) 0.08 [‐0.12, 0.28]
31.3 N‐3 replacing n‐6 3 513 Mean Difference (IV, Random, 95% CI) 0.18 [‐0.46, 0.81]
31.4 N‐3 replacing carbs/sugars 1 258 Mean Difference (IV, Random, 95% CI) ‐0.60 [‐1.14, ‐0.06]
31.5 N‐3 replacing nil/low n‐3 placebo 2 300 Mean Difference (IV, Random, 95% CI) 0.86 [‐0.30, 2.01]
31.6 Replacement unclear 2 223 Mean Difference (IV, Random, 95% CI) 0.58 [‐1.17, 2.33]
32 BMI, kg/m² ‐ LCn3 ‐ subgroup by intervention type 15 15474 Mean Difference (IV, Random, 95% CI) 0.06 [‐0.14, 0.25]
32.1 Dietary advice 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
32.2 Supplemental foods 1 1260 Mean Difference (IV, Random, 95% CI) 0.1 [‐0.10, 0.30]
32.3 Supplement (capsule) 13 14169 Mean Difference (IV, Random, 95% CI) 0.04 [‐0.22, 0.29]
32.4 Any combination 1 45 Mean Difference (IV, Random, 95% CI) 1.60 [‐0.43, 3.63]
33 BMI, kg/m² ‐ LCn3 ‐ subgroup by duration 15   Mean Difference (IV, Random, 95% CI) Subtotals only
33.1 Medium duration 1 to < 2 years in study 9 906 Mean Difference (IV, Random, 95% CI) 0.24 [‐0.40, 0.88]
33.2 Medium‐long duration: 2 to < 4 years in study 5 2032 Mean Difference (IV, Random, 95% CI) 0.13 [‐0.06, 0.32]
33.3 Long duration ≥ 4 years in study 1 12536 Mean Difference (IV, Random, 95% CI) 0.0 [‐0.20, 0.20]
34 BMI, kg/m² ‐ LCn3 ‐ subgroup by primary or secondary prevention 15   Mean Difference (IV, Random, 95% CI) Subtotals only
34.1 Primary CVD prevention 11 13610 Mean Difference (IV, Random, 95% CI) 0.15 [‐0.36, 0.66]
34.2 Secondary CVD prevention 4 1864 Mean Difference (IV, Random, 95% CI) 0.06 [‐0.07, 0.20]
35 BMI, kg/m² ‐ LCn3 ‐ subgroup by statin use 15   Mean Difference (IV, Random, 95% CI) Subtotals only
35.1 LCn3 ‐ ≥ 50% of control group on statins 4 14131 Mean Difference (IV, Random, 95% CI) 0.17 [‐0.17, 0.50]
35.2 LCn3 ‐ < 50% of control group on statins 4 665 Mean Difference (IV, Random, 95% CI) 0.02 [‐0.15, 0.19]
35.3 LCn3 ‐ use of statins unclear 7 678 Mean Difference (IV, Random, 95% CI) 0.06 [‐0.86, 0.97]
36 Other measures of adiposity ‐ LCn3 7   Mean Difference (IV, Random, 95% CI) Subtotals only
36.1 Percentage body fat 2 127 Mean Difference (IV, Random, 95% CI) 0.85 [‐6.87, 8.57]
36.2 Percentage visceral fat 1 95 Mean Difference (IV, Random, 95% CI) ‐1.80 [‐15.03, 11.43]
36.3 Waist circumference, cm 4 916 Mean Difference (IV, Random, 95% CI) 0.72 [‐0.17, 1.62]
36.4 Waist‐hip ratio 1 100 Mean Difference (IV, Random, 95% CI) 0.0 [‐0.01, 0.01]
36.5 Abdominal circumference, cm 1 256 Mean Difference (IV, Random, 95% CI) ‐0.70 [‐8.78, 7.38]
36.6 Hip circumference, cm 1 258 Mean Difference (IV, Random, 95% CI) ‐2.40 [‐9.80, 5.00]
37 Total cholesterol, serum, mmoL/L ‐ LCn3 30 38469 Mean Difference (IV, Random, 95% CI) ‐0.01 [‐0.05, 0.03]
38 TC, mmoL/L ‐ LCn3 ‐ SA fixed effects 30 38469 Mean Difference (IV, Fixed, 95% CI) ‐0.04 [‐0.06, ‐0.02]
39 TC, mmoL/L ‐ LCn3 ‐ SA by summary risk of bias 30   Mean Difference (IV, Random, 95% CI) Subtotals only
39.1 Low summary risk of bias 10 15878 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.04, 0.06]
39.2 Moderate or high risk of bias 20 22591 Mean Difference (IV, Random, 95% CI) ‐0.03 [‐0.08, 0.02]
40 TC, mmoL/L ‐ LCn3 ‐ SA by compliance and study size 22   Mean Difference (IV, Random, 95% CI) Subtotals only
40.1 SA ‐ low risk of compliance bias 14 3341 Mean Difference (IV, Random, 95% CI) 0.02 [‐0.05, 0.09]
40.2 SA ‐ 100+ randomised 17 37810 Mean Difference (IV, Random, 95% CI) ‐0.00 [‐0.05, 0.05]
41 TC, mmoL/L ‐ LCn3 ‐ subgroup by dose 30   Mean Difference (IV, Random, 95% CI) Subtotals only
41.1 LCn3 ≤ 150 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
41.2 LCn3 > 150 ≤ 250 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
41.3 LCn3 > 250 ≤ 400 mg/d 1 1715 Mean Difference (IV, Random, 95% CI) 0.10 [‐0.01, 0.21]
41.4 LCn3 > 400 ≤ 2400 mg/d 19 35210 Mean Difference (IV, Random, 95% CI) ‐0.04 [‐0.06, ‐0.01]
41.5 LCn3 > 2.4 ≤ 4.4 g/d 8 1456 Mean Difference (IV, Random, 95% CI) ‐0.10 [‐0.22, 0.01]
41.6 LCn3 > 4.4 g/d 2 88 Mean Difference (IV, Random, 95% CI) 0.08 [‐0.28, 0.45]
42 TC, mmoL/L ‐ LCn3 ‐ subgroup by replacement 30   Mean Difference (IV, Random, 95% CI) Subtotals only
42.1 N‐3 replacing SFA 3 2148 Mean Difference (IV, Random, 95% CI) 0.10 [‐0.01, 0.20]
42.2 N‐3 replacing MUFA 16 17452 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.04, 0.05]
42.3 N‐3 replacing n‐6 4 729 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.34, 0.31]
42.4 N‐3 replacing carbs/sugars 1 258 Mean Difference (IV, Random, 95% CI) ‐0.20 [‐1.03, 0.63]
42.5 N‐3 replacing nil/low n‐3 placebo 7 19745 Mean Difference (IV, Random, 95% CI) ‐0.05 [‐0.07, ‐0.03]
42.6 Replacement unclear 2 285 Mean Difference (IV, Random, 95% CI) 0.12 [‐0.14, 0.38]
43 TC, mmoL/L ‐ LCn3 ‐ subgroup by intervention type 30   Mean Difference (IV, Random, 95% CI) Subtotals only
43.1 Dietary advice 1 1715 Mean Difference (IV, Random, 95% CI) 0.10 [‐0.01, 0.21]
43.2 Supplemental foods 1 1210 Mean Difference (IV, Random, 95% CI) 0.02 [‐0.09, 0.13]
43.3 Supplement (capsule) 26 35333 Mean Difference (IV, Random, 95% CI) ‐0.04 [‐0.07, ‐0.02]
43.4 Any combination 2 211 Mean Difference (IV, Random, 95% CI) 0.13 [‐0.10, 0.37]
44 TC, mmoL/L ‐ LCn3 ‐ subgroup by duration 30   Mean Difference (IV, Random, 95% CI) Subtotals only
44.1 Medium duration 1 to < 2 years in study 15 1661 Mean Difference (IV, Random, 95% CI) ‐0.06 [‐0.16, 0.04]
44.2 Medium‐long duration: 2 to < 4 years in study 11 4471 Mean Difference (IV, Random, 95% CI) 0.03 [‐0.04, 0.09]
44.3 Long duration ≥ 4 years in study 4 32337 Mean Difference (IV, Random, 95% CI) ‐0.01 [‐0.08, 0.07]
45 TC, mmoL/L ‐ LCn3 ‐ subgroup by primary or secondary prevention 30   Mean Difference (IV, Random, 95% CI) Subtotals only
45.1 Primary prevention 18 33744 Mean Difference (IV, Random, 95% CI) ‐0.04 [‐0.06, ‐0.02]
45.2 Secondary prevention 12 4725 Mean Difference (IV, Random, 95% CI) ‐0.00 [‐0.08, 0.07]
46 TC, mmoL/L ‐ LCn3 ‐ subgroup by statin use 30   Mean Difference (IV, Random, 95% CI) Subtotals only
46.1 LCn3 ‐ ≥ 50% of control group on statins 8 34011 Mean Difference (IV, Random, 95% CI) ‐0.04 [‐0.06, ‐0.02]
46.2 LCn3 ‐ < 50% of control group on statins 15 3871 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.08, 0.10]
46.3 LCn3 ‐ use of statins unclear 7 587 Mean Difference (IV, Random, 95% CI) ‐0.03 [‐0.27, 0.22]
47 Triglycerides, fasting, serum, mmoL/L ‐ LCn3 27 43998 Mean Difference (IV, Random, 95% CI) ‐0.24 [‐0.31, ‐0.16]
48 TG, fasting, mmoL/L ‐ LCn3 ‐ SA fixed effects 27 43998 Mean Difference (IV, Fixed, 95% CI) ‐0.21 [‐0.26, ‐0.17]
49 TG, fasting, mmoL/L ‐ LCn3 ‐ SA by summary risk of bias 27   Mean Difference (IV, Random, 95% CI) Subtotals only
49.1 Low summary risk of bias 8 14654 Mean Difference (IV, Random, 95% CI) ‐0.19 [‐0.28, ‐0.10]
49.2 Moderate or high risk of bias 19 29344 Mean Difference (IV, Random, 95% CI) ‐0.25 [‐0.35, ‐0.15]
50 TG, fasting, mmoL/L ‐ LCn3 ‐ SA by compliance and study size 20   Mean Difference (IV, Random, 95% CI) Subtotals only
50.1 SA ‐ low risk of compliance bias 13 11485 Mean Difference (IV, Random, 95% CI) ‐0.26 [‐0.36, ‐0.16]
50.2 SA ‐ 100+ randomised 17 43484 Mean Difference (IV, Random, 95% CI) ‐0.24 [‐0.32, ‐0.15]
51 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by dose 27   Mean Difference (IV, Random, 95% CI) Subtotals only
51.1 LCn3 ≤ 150 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
51.2 LCn3 > 150 ≤ 250 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
51.3 LCn3 > 250 ≤ 400 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
51.4 LCn3 > 400 ≤ 2400 mg/d 18 34388 Mean Difference (IV, Random, 95% CI) ‐0.18 [‐0.25, ‐0.11]
51.5 LCn3 > 2.4 ≤ 4.4 g/d 7 9526 Mean Difference (IV, Random, 95% CI) ‐0.36 [‐0.53, ‐0.20]
51.6 LCn3 > 4.4 g/d 2 84 Mean Difference (IV, Random, 95% CI) ‐0.41 [‐0.68, ‐0.14]
52 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by replacement 27   Mean Difference (IV, Random, 95% CI) Subtotals only
52.1 N‐3 replacing SFA 2 429 Mean Difference (IV, Random, 95% CI) ‐0.27 [‐0.59, 0.04]
52.2 N‐3 replacing MUFA 13 14634 Mean Difference (IV, Random, 95% CI) ‐0.18 [‐0.25, ‐0.10]
52.3 N‐3 replacing n‐6 4 715 Mean Difference (IV, Random, 95% CI) ‐0.33 [‐0.50, ‐0.16]
52.4 N‐3 replacing carbs/sugars 1 258 Mean Difference (IV, Random, 95% CI) ‐0.5 [‐1.49, 0.49]
52.5 N‐3 replacing nil/low n‐3 placebo 6 27776 Mean Difference (IV, Random, 95% CI) ‐0.18 [‐0.51, 0.14]
52.6 Replacement unclear 3 615 Mean Difference (IV, Random, 95% CI) ‐0.20 [‐0.58, 0.18]
53 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by intervention type 27   Mean Difference (IV, Random, 95% CI) Subtotals only
53.1 Dietary advice 1 71 Mean Difference (IV, Random, 95% CI) 0.02 [‐0.36, 0.40]
53.2 Supplemental foods 1 1210 Mean Difference (IV, Random, 95% CI) ‐0.03 [‐0.15, 0.09]
53.3 Supplement (capsule) 24 42556 Mean Difference (IV, Random, 95% CI) ‐0.19 [‐0.38, ‐0.00]
53.4 Any combination 1 161 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.28, 0.30]
54 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by duration 27   Mean Difference (IV, Random, 95% CI) Subtotals only
54.1 Medium duration 1 to < 2 years in study 13 1880 Mean Difference (IV, Random, 95% CI) ‐0.27 [‐0.36, ‐0.19]
54.2 Medium‐long duration: 2 to < 4 years in study 10 2550 Mean Difference (IV, Random, 95% CI) ‐0.16 [‐0.31, ‐0.02]
54.3 Long duration ≥ 4 years in study 4 39568 Mean Difference (IV, Random, 95% CI) ‐0.20 [‐0.32, ‐0.07]
55 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by primary or secondary prevention 27   Mean Difference (IV, Random, 95% CI) Subtotals only
55.1 Primary prevention 17 33114 Mean Difference (IV, Random, 95% CI) ‐0.20 [‐0.26, ‐0.14]
55.2 Secondary prevention 9 2705 Mean Difference (IV, Random, 95% CI) ‐0.27 [‐0.44, ‐0.10]
55.3 Mixed primary & secondary prevention 1 8179 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
56 TG, fasting, mmoL/L ‐ LCn3 ‐ subgroup by statin use 27   Mean Difference (IV, Random, 95% CI) Subtotals only
56.1 LCn3 ‐ ≥ 50% of control group on statins 7 40976 Mean Difference (IV, Random, 95% CI) ‐0.11 [‐0.21, ‐0.01]
56.2 LCn3 ‐ < 50% of control group on statins 14 2414 Mean Difference (IV, Random, 95% CI) ‐0.27 [‐0.36, ‐0.18]
56.3 LCn3 ‐ use of statins unclear 6 608 Mean Difference (IV, Random, 95% CI) ‐0.23 [‐0.38, ‐0.08]
57 High‐density lipoprotein, serum, mmoL/L ‐ LCn3 30 46604 Mean Difference (IV, Random, 95% CI) 0.03 [0.01, 0.05]
58 HDL, mmoL/L ‐ LCn3 ‐ SA fixed effects 30 46604 Mean Difference (IV, Fixed, 95% CI) 0.01 [0.00, 0.02]
59 HDL, mmoL/L ‐ LCn3 ‐ SA by summary risk of bias 30   Mean Difference (IV, Random, 95% CI) Subtotals only
59.1 Low summary risk of bias 9 15840 Mean Difference (IV, Random, 95% CI) 0.02 [‐0.01, 0.05]
59.2 Moderate or high risk of bias 21 30764 Mean Difference (IV, Random, 95% CI) 0.03 [0.00, 0.06]
60 HDL, mmoL/L ‐ LCn3 ‐ SA by compliance and study size 23   Mean Difference (IV, Random, 95% CI) Subtotals only
60.1 SA ‐ low risk of compliance bias 14 11381 Mean Difference (IV, Random, 95% CI) 0.05 [0.01, 0.10]
60.2 SA ‐ 100+ randomised 18 45940 Mean Difference (IV, Random, 95% CI) 0.03 [0.01, 0.05]
61 HDL, mmoL/L ‐ LCn3 ‐ subgroup by dose 30   Mean Difference (IV, Random, 95% CI) Subtotals only
61.1 LCn3 ≤ 150 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
61.2 LCn3 > 150 ≤ 250 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
61.3 LCn3 > 250 ≤ 400 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
61.4 LCn3 > 400 ≤ 2400 mg/d 20 36920 Mean Difference (IV, Random, 95% CI) 0.02 [0.00, 0.04]
61.5 LCn3 > 2.4 ≤ 4.4 g/d 9 9625 Mean Difference (IV, Random, 95% CI) 0.06 [0.02, 0.10]
61.6 LCn3 > 4.4 g/d 1 59 Mean Difference (IV, Random, 95% CI) 0.0 [‐0.16, 0.16]
62 HDL, mmoL/L ‐ LCn3 ‐ subgroup by replacement 30   Mean Difference (IV, Random, 95% CI) Subtotals only
62.1 N‐3 replacing SFA 3 2143 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.10, 0.07]
62.2 N‐3 replacing MUFA 16 17453 Mean Difference (IV, Random, 95% CI) 0.03 [‐0.00, 0.06]
62.3 N‐3 replacing n‐6 3 688 Mean Difference (IV, Random, 95% CI) 0.06 [0.00, 0.11]
62.4 N‐3 replacing carbs/sugars 1 258 Mean Difference (IV, Random, 95% CI) 0.10 [‐0.17, 0.37]
62.5 N‐3 replacing nil/low n‐3 placebo 8 27924 Mean Difference (IV, Random, 95% CI) 0.04 [‐0.01, 0.09]
62.6 Replacement unclear 2 281 Mean Difference (IV, Random, 95% CI) ‐0.03 [‐0.14, 0.08]
63 HDL, mmoL/L ‐ LCn3 ‐ subgroup by intervention type 30   Mean Difference (IV, Random, 95% CI) Subtotals only
63.1 Dietary advice 2 1785 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.02, 0.04]
63.2 Supplemental foods 1 1210 Mean Difference (IV, Random, 95% CI) 0.03 [0.00, 0.06]
63.3 Supplement (capsule) 24 43375 Mean Difference (IV, Random, 95% CI) 0.04 [0.01, 0.06]
63.4 Any combination 3 234 Mean Difference (IV, Random, 95% CI) 0.10 [‐0.10, 0.31]
64 HDL, mmoL/L ‐ LCn3 ‐ subgroup by duration 30   Mean Difference (IV, Random, 95% CI) Subtotals only
64.1 Medium duration 1 to < 2 years in study 13 1562 Mean Difference (IV, Random, 95% CI) 0.08 [0.01, 0.14]
64.2 Medium‐long duration: 2 to < 4 years in study 12 4526 Mean Difference (IV, Random, 95% CI) 0.02 [‐0.00, 0.04]
64.3 Long duration ≥ 4 years in study 5 40516 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.01, 0.04]
65 HDL, mmoL/L ‐ LCn3 ‐ subgroup by primary or secondary prevention 29   Mean Difference (IV, Random, 95% CI) Subtotals only
65.1 Primary prevention 18 33804 Mean Difference (IV, Random, 95% CI) 0.03 [0.00, 0.06]
65.2 Secondary prevention 10 4547 Mean Difference (IV, Random, 95% CI) 0.03 [‐0.00, 0.06]
65.3 Mixed primary & secondary prevention 1 8179 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
66 HDL, mmoL/L ‐ LCn3 ‐ subgroup by statin use 30   Mean Difference (IV, Random, 95% CI) Subtotals only
66.1 LCn3 ‐ ≥ 50% of control group on statins 10 42261 Mean Difference (IV, Random, 95% CI) 0.02 [0.00, 0.04]
66.2 LCn3 ‐ < 50% of control group on statins 13 3690 Mean Difference (IV, Random, 95% CI) 0.04 [‐0.00, 0.08]
66.3 LCn3 ‐ use of statins unclear 7 653 Mean Difference (IV, Random, 95% CI) 0.07 [‐0.07, 0.21]
67 Low‐density lipoprotein, serum, mmoL/L ‐ LCn3 25 43454 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.01, 0.03]
68 LDL, mmoL/L ‐ LCn3 ‐ SA fixed effects 25 43454 Mean Difference (IV, Fixed, 95% CI) 0.01 [‐0.01, 0.03]
69 LDL, mmoL/L ‐ LCn3 ‐ SA by summary risk of bias 25   Mean Difference (IV, Random, 95% CI) Subtotals only
69.1 Low summary risk of bias 9 14840 Mean Difference (IV, Random, 95% CI) 0.02 [‐0.03, 0.07]
69.2 Moderate or high risk of bias 16 28614 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.02, 0.03]
70 LDL, mmoL/L ‐ LCn3 ‐ SA by compliance and study size 19   Mean Difference (IV, Random, 95% CI) Subtotals only
70.1 SA ‐ low risk of compliance bias 14 11344 Mean Difference (IV, Random, 95% CI) 0.05 [‐0.02, 0.11]
70.2 SA ‐ 100+ randomised 14 42885 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.01, 0.03]
71 LDL, mmoL/L ‐ LCn3 ‐ subgroup by dose 25   Mean Difference (IV, Random, 95% CI) Subtotals only
71.1 LCn3 ≤ 150 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
71.2 LCn3 > 150 ≤ 250 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
71.3 LCn3 > 250 ≤ 400 mg/d 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
71.4 LCn3 > 400 ≤ 2400 mg/d 16 34054 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.01, 0.02]
71.5 LCn3 > 2.4 ≤ 4.4 g/d 7 9312 Mean Difference (IV, Random, 95% CI) 0.03 [‐0.08, 0.15]
71.6 LCn3 > 4.4 g/d 2 88 Mean Difference (IV, Random, 95% CI) 0.22 [‐0.09, 0.54]
72 LDL, mmoL/L ‐ LCn3 ‐ subgroup by replacement 25   Mean Difference (IV, Random, 95% CI) Subtotals only
72.1 N‐3 replacing SFA 2 429 Mean Difference (IV, Random, 95% CI) 0.17 [‐0.14, 0.47]
72.2 N‐3 replacing MUFA 14 14710 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.03, 0.05]
72.3 N‐3 replacing n‐6 2 242 Mean Difference (IV, Random, 95% CI) 0.14 [‐0.26, 0.55]
72.4 N‐3 replacing carbs/sugars 1 258 Mean Difference (IV, Random, 95% CI) 0.20 [‐0.51, 0.91]
72.5 N‐3 replacing nil/low n‐3 placebo 6 27790 Mean Difference (IV, Random, 95% CI) 0.00 [‐0.02, 0.02]
72.6 Replacement unclear 2 454 Mean Difference (IV, Random, 95% CI) 0.09 [‐0.04, 0.23]
73 LDL, mmoL/L ‐ LCn3 ‐ subgroup by intervention type 25   Mean Difference (IV, Random, 95% CI) Subtotals only
73.1 Dietary advice 1 71 Mean Difference (IV, Random, 95% CI) 0.08 [‐0.22, 0.38]
73.2 Supplemental foods 1 1124 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.10, 0.06]
73.3 Supplement (capsule) 21 42187 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.01, 0.03]
73.4 Any combination 2 72 Mean Difference (IV, Random, 95% CI) 0.08 [‐0.44, 0.61]
74 LDL, mmoL/L ‐ LCn3 ‐ subgroup by duration 25   Mean Difference (IV, Random, 95% CI) Subtotals only
74.1 Medium duration 1 to < 2 years in study 14 1862 Mean Difference (IV, Random, 95% CI) 0.06 [‐0.03, 0.14]
74.2 Medium‐long duration: 2 to < 4 years in study 7 2024 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.06, 0.07]
74.3 Long duration ≥ 4 years in study 4 39568 Mean Difference (IV, Random, 95% CI) 0.03 [‐0.04, 0.10]
75 LDL, mmoL/L ‐ LCn3 ‐ subgroup by primary or secondary prevention 25   Mean Difference (IV, Random, 95% CI) Subtotals only
75.1 Primary prevention 16 32717 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.01, 0.03]
75.2 Secondary prevention 8 2558 Mean Difference (IV, Random, 95% CI) 0.02 [‐0.04, 0.09]
75.3 Mixed primary & secondary prevention 1 8179 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
76 LDL, mmoL/L ‐ LCn3 ‐ subgroup by statin use 25   Mean Difference (IV, Random, 95% CI) Subtotals only
76.1 LCn3 ‐ ≥ 50% of control group on statins 9 41227 Mean Difference (IV, Random, 95% CI) 0.00 [‐0.02, 0.02]
76.2 LCn3 ‐ < 50% of control group on statins 9 1564 Mean Difference (IV, Random, 95% CI) 0.12 [0.03, 0.21]
76.3 LCn3 ‐ use of statins unclear 7 663 Mean Difference (IV, Random, 95% CI) ‐0.01 [‐0.17, 0.14]

Comparison 3. High vs low LCn3 omega‐3 fats (tertiary outcomes).

Outcome or subgroup title No. of studies No. of participants Statistical method Effect size
1 Blood pressure, mmHg ‐ LCn3 17   Mean Difference (IV, Random, 95% CI) Subtotals only
1.1 Systolic BP ‐ LCn3 17 35601 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.31, 0.34]
1.2 Diastolic BP ‐ LCn3 15 35626 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.22, 0.17]
2 Serious adverse events ‐ LCn3 19   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
2.1 Any serious adverse events 3 9116 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.94, 1.06]
2.2 Bleeding 11 80147 Risk Ratio (M‐H, Random, 95% CI) 1.12 [0.91, 1.37]
2.3 Serious GI events 3 774 Risk Ratio (M‐H, Random, 95% CI) 1.34 [0.64, 2.80]
2.4 Pulmonary embolus or DVT ‐ LCn3 5 3546 Risk Ratio (M‐H, Random, 95% CI) 1.15 [0.44, 2.98]
2.5 Progression to advanced AMD (age‐related macular degeneration) 1 4203 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.90, 1.02]
3 Side effects ‐ LCn3 38   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
3.1 Dropouts due to side effects 23 16755 Risk Ratio (M‐H, Random, 95% CI) 1.16 [0.99, 1.36]
3.2 Abdominal pain or discomfort 9 41056 Risk Ratio (M‐H, Random, 95% CI) 1.05 [0.91, 1.20]
3.3 Diarrhoea 13 37013 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.87, 1.19]
3.4 Nausea 8 35819 Risk Ratio (M‐H, Random, 95% CI) 1.20 [0.96, 1.49]
3.5 Any gastrointestinal side effect ‐ LCn3 fats 33 89668 Risk Ratio (M‐H, Random, 95% CI) 1.10 [0.97, 1.26]
3.6 Skin problems (itching, rashes) 9 36721 Risk Ratio (M‐H, Random, 95% CI) 1.11 [0.52, 2.37]
3.7 Headache or worsening migraine 4 1526 Risk Ratio (M‐H, Random, 95% CI) 0.85 [0.51, 1.40]
3.8 Reflux 3 8916 Risk Ratio (M‐H, Random, 95% CI) 1.23 [0.79, 1.91]
3.9 Pain (joint, lumbar, muscle pain) 3 27359 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.74, 1.23]
3.10 All side effects combined 14 39439 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.95, 1.08]
4 Dropouts ‐ LCn3 35 56089 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.90, 1.04]

Comparison 4. High vs low ALA omega‐3 fat (primary outcomes).

Outcome or subgroup title No. of studies No. of participants Statistical method Effect size
1 All‐cause mortality (overall) ‐ ALA 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.84, 1.20]
2 All‐cause mortality ‐ ALA ‐ sensitivity analysis (SA) fixed‐effect 5 19327 Risk Ratio (M‐H, Fixed, 95% CI) 1.01 [0.85, 1.21]
3 All‐cause mortality ‐ ALA ‐ SA by summary risk of bias 5   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
3.1 Low risk of bias 3 5213 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.72, 1.45]
3.2 Moderate/high risk of bias 2 14114 Risk Ratio (M‐H, Random, 95% CI) 1.09 [0.71, 1.67]
4 All‐cause mortality ‐ ALA ‐ SA by compliance and study size 5   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
4.1 SA ‐ low risk of compliance bias 3 5811 Risk Ratio (M‐H, Random, 95% CI) 1.05 [0.68, 1.63]
4.2 SA ‐ 100+ randomised 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.84, 1.20]
5 All‐cause mortality ‐ ALA ‐ subgroup by dose 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.84, 1.20]
5.1 ALA low < 5 g/d 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.80, 1.19]
5.2 ALA high ≥ 5 g/d 4 14490 Risk Ratio (M‐H, Random, 95% CI) 1.16 [0.77, 1.75]
6 All‐cause mortality ‐ ALA ‐ subgroup by replacement 5   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
6.1 ALA replacing SFA 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
6.2 ALA replacing MUFA 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.80, 1.19]
6.3 ALA replacing n‐6 2 13672 Risk Ratio (M‐H, Random, 95% CI) 1.37 [0.48, 3.86]
6.4 ALA replacing CHO 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
6.5 ALA replacing nil/low n‐3 placebo 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
6.6 ALA replacement unclear 2 818 Risk Ratio (M‐H, Random, 95% CI) 2.78 [0.29, 26.49]
7 All cause mortality ‐ ALA ‐ subgroup by intervention type 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.84, 1.20]
7.1 Dietary advice 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
7.2 Supplemental foods 4 5921 Risk Ratio (M‐H, Random, 95% CI) 0.99 [0.82, 1.21]
7.3 Supplements (capsule) 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.07 [0.70, 1.64]
7.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
8 All‐cause mortality ‐ ALA ‐ subgroup by duration 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.84, 1.20]
8.1 Medium duration 1 to < 2 years in study 2 13516 Risk Ratio (M‐H, Random, 95% CI) 1.09 [0.71, 1.67]
8.2 Medium‐long duration: 2 to < 4 years in study 3 5811 Risk Ratio (M‐H, Random, 95% CI) 1.05 [0.68, 1.63]
8.3 Long duration: ≥ 4 years in study 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
9 All‐cause mortality ‐ ALA ‐ subgroup by primary or secondary prevention 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.84, 1.20]
9.1 Primary CVD prevention 3 14380 Risk Ratio (M‐H, Random, 95% CI) 1.14 [0.75, 1.74]
9.2 Secondary CVD prevention 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.81, 1.19]
10 All‐cause mortality ‐ ALA ‐ subgroup by statin use 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.84, 1.20]
10.1 ALA ‐ ≥ 50% of control group on statins 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.98 [0.81, 1.19]
10.2 ALA ‐ < 50% of control group on statins 3 14380 Risk Ratio (M‐H, Random, 95% CI) 1.14 [0.75, 1.74]
11 Cardiovascular mortality (overall) ‐ ALA 4 18619 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.74, 1.25]
12 CVD mortality ‐ ALA ‐ SA fixed‐effect 4 18619 Risk Ratio (M‐H, Fixed, 95% CI) 0.96 [0.74, 1.25]
13 CVD mortality ‐ ALA ‐ SA by summary risk of bias 4   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
13.1 Low risk of bias 3 5213 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.70, 1.28]
13.2 Moderate/high risk of bias 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.58, 1.70]
14 CVD mortality ‐ ALA ‐ SA by compliance and study size 4   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
14.1 SA ‐ low risk of compliance bias 2 5103 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.70, 1.27]
14.2 SA ‐ 100+ randomised 4 18619 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.74, 1.25]
15 CVD mortality ‐ ALA ‐ subgroup by dose 4 18619 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.74, 1.25]
15.1 ALA low < 5 g/d 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.69, 1.27]
15.2 ALA high ≥ 5 g/d 3 13782 Risk Ratio (M‐H, Random, 95% CI) 1.04 [0.62, 1.73]
16 CVD mortality ‐ ALA ‐ subgroup by replacement 4   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
16.1 N‐3 replacing SFA 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
16.2 N‐3 replacing MUFA 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.69, 1.27]
16.3 N‐3 replacing n‐6 2 13672 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.60, 1.70]
16.4 N‐3 replacing carbohydrates/sugars 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
16.5 N‐3 replacing nil/low n‐3 placebo 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
16.6 Replacement unclear 1 110 Risk Ratio (M‐H, Random, 95% CI) 2.69 [0.11, 64.74]
17 CVD mortality ‐ ALA ‐ subgroup by intervention type 4 18619 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.74, 1.25]
17.1 Dietary advice 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
17.2 Supplemental foods 3 5213 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.70, 1.28]
17.3 Supplements (capsule) 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.58, 1.70]
17.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
18 CVD mortality ‐ ALA ‐ subgroup by duration 4 18619 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.74, 1.25]
18.1 Medium duration 1 to < 2 years in study 2 13516 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.61, 1.73]
18.2 Medium‐long duration: 2 to < 4 years in study 2 5103 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.70, 1.27]
18.3 Long duration: ≥ 4 years in study 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
19 CVD mortality ‐ ALA ‐ subgroup by primary or secondary prevention 4 18619 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.74, 1.25]
19.1 Primary prevention 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.58, 1.70]
19.2 Secondary prevention 3 5213 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.70, 1.28]
20 CVD mortality ‐ ALA ‐ subgroup by statin uses 4 18619 Risk Ratio (M‐H, Random, 95% CI) 0.96 [0.74, 1.25]
20.1 ALA ‐ ≥ 50% of control group on statins 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.70, 1.28]
20.2 ALA ‐ < 50% of control group on statins 2 13672 Risk Ratio (M‐H, Random, 95% CI) 1.01 [0.60, 1.70]
21 Cardiovascular events (overall) ‐ ALA 5 19327 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.83, 1.07]
22 CVD events ‐ ALA ‐ SA fixed‐effect 5 19327 Risk Ratio (M‐H, Fixed, 95% CI) 0.95 [0.84, 1.07]
23 CVD events ‐ ALA ‐ SA by summary risk of bias 5   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
23.1 Low risk of bias 3 5213 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.79, 1.04]
23.2 Moderate/high risk of bias 2 14114 Risk Ratio (M‐H, Random, 95% CI) 1.12 [0.84, 1.48]
24 CVD events ‐ ALA ‐ SA by compliance and study size 5   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
24.1 SA ‐ low risk of compliance bias 3 5811 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.79, 1.04]
24.2 SA ‐ 100+ randomised 5 19327 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.83, 1.07]
25 CVD events ‐ ALA ‐ subgroup by dose 5 19327 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.83, 1.07]
25.1 ALA low < 5 g/d 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.79, 1.05]
25.2 ALA high ≥ 5 g/d 4 14490 Risk Ratio (M‐H, Random, 95% CI) 1.07 [0.82, 1.40]
26 CVD events ‐ ALA ‐ subgroup by replacement 5   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
26.1 N‐3 replacing SFA 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
26.2 N‐3 replacing MUFA 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.79, 1.05]
26.3 N‐3 replacing n‐6 2 13672 Risk Ratio (M‐H, Random, 95% CI) 0.76 [0.24, 2.41]
26.4 N‐3 replacing carbohydrates/sugars 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
26.5 N‐3 replacing nil/low n‐3 placebo 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
26.6 Replacement unclear 2 818 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.36, 2.43]
27 CVD events ‐ ALA ‐ subgroup by intervention type 5 19327 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.83, 1.07]
27.1 Dietary advice 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
27.2 Supplemental foods 4 5921 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.79, 1.04]
27.3 Supplements (capsule) 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.13 [0.85, 1.51]
27.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
28 CVD events ‐ ALA ‐ subgroup by duration 5 19327 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.83, 1.07]
28.1 Medium duration 1 to < 2 years in study 2 13516 Risk Ratio (M‐H, Random, 95% CI) 1.13 [0.86, 1.50]
28.2 Medium‐long duration: 2 to < 4 years in study 3 5811 Risk Ratio (M‐H, Random, 95% CI) 0.90 [0.79, 1.04]
28.3 Long duration: ≥ 4 years in study 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
29 CVD events ‐ ALA ‐ subgroup by primary or secondary prevention 5 19327 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.83, 1.07]
29.1 Primary prevention 3 14380 Risk Ratio (M‐H, Random, 95% CI) 0.87 [0.46, 1.67]
29.2 Secondary prevention 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.80, 1.05]
30 CVD events ‐ ALA ‐ subgroup by statin use 5 19327 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.83, 1.07]
30.1 ALA ‐ ≥ 50% of control group on statins 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.80, 1.05]
30.2 ALA ‐ < 50% of control group on statins 3 14380 Risk Ratio (M‐H, Random, 95% CI) 0.87 [0.46, 1.67]
31 Coronary heart disease mortality (overall) ‐ ALA 3 18353 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.72, 1.26]
32 CHD mortality ‐ ALA ‐ SA fixed‐effect 3 18353 Risk Ratio (M‐H, Fixed, 95% CI) 0.95 [0.72, 1.26]
33 CHD mortality ‐ ALA ‐ SA by summary risk of bias 3   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
33.1 Low risk of bias 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.67, 1.30]
33.2 Moderate/high risk of bias 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.58, 1.70]
34 CHD mortality ‐ ALA ‐ SA by compliance and study size 3   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
34.1 SA ‐ low risk of compliance bias 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.66, 1.28]
34.2 SA ‐ 100+ randomised 3 18353 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.72, 1.26]
35 CHD mortality ‐ ALA ‐ subgroup by dose 3 18353 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.72, 1.26]
35.1 ALA low < 5 g/d 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.66, 1.28]
35.2 ALA high ≥ 5 g/d 2 13516 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.61, 1.73]
36 CHD mortality ‐ ALA ‐ subgroup by replacement 3   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
36.1 N‐3 replacing SFA 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
36.2 Coronary heart mortality‐ ALA 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
36.3 N‐3 replacing MUFA 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.66, 1.28]
36.4 N‐3 replacing n‐6 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.58, 1.70]
36.5 N‐3 replacing carbohydrates/sugars 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
36.6 N‐3 replacing nil/low n‐3 placebo 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
36.7 Replacement unclear 1 110 Risk Ratio (M‐H, Random, 95% CI) 2.69 [0.11, 64.74]
37 CHD mortality ‐ ALA ‐ subgroup by intervention type 3 18353 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.72, 1.26]
37.1 Dietary advice 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
37.2 Supplemental foods 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.67, 1.30]
37.3 Supplements (capsule) 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.58, 1.70]
37.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
38 CHD mortality ‐ ALA ‐ subgroup by duration 3 18353 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.72, 1.26]
38.1 Medium duration 1 to < 2 years in study 2 13516 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.61, 1.73]
38.2 Medium‐long duration: 2 to < 4 years in study 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.66, 1.28]
38.3 Long duration: ≥ 4 years in study 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
39 CHD mortality ‐ ALA ‐ subgroup by primary or secondary prevention 3 18353 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.72, 1.26]
39.1 Primary prevention of CVD 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.58, 1.70]
39.2 Secondary prevention of CVD 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.67, 1.30]
40 CHD mortality ‐ ALA ‐ subgroup by statin use 3 18353 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.72, 1.26]
40.1 ALA ‐ ≥ 50% of control group on statins 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.93 [0.67, 1.30]
40.2 ALA ‐ < 50% of control group on statins 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.58, 1.70]
41 CHD mortality ‐ ALA ‐ subgroup by CAD history 3 18353 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.72, 1.26]
41.1 Previous CAD 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.66, 1.28]
41.2 No previous CAD 2 13516 Risk Ratio (M‐H, Random, 95% CI) 1.02 [0.61, 1.73]
42 Coronary heart disease events (overall) ‐ ALA 4 19061 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.82, 1.22]
43 CHD events ‐ ALA ‐ SA fixed‐effect 4 19061 Risk Ratio (M‐H, Fixed, 95% CI) 1.00 [0.82, 1.21]
44 CHD events ‐ ALA ‐ SA by summary risk of bias 4   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
44.1 Low risk of bias 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.71, 1.15]
44.2 Moderate/high risk of bias 2 14114 Risk Ratio (M‐H, Random, 95% CI) 1.20 [0.86, 1.67]
45 CHD events ‐ ALA ‐ SA by compliance and study size 4   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
45.1 SA ‐ low risk of compliance bias 2 5545 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.73, 1.17]
45.2 SA ‐ 100+ randomised 4 19061 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.82, 1.22]
46 CHD events ‐ ALA ‐ subgroup by dose 4 19061 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.82, 1.22]
46.1 ALA low < 5 g/d 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.72, 1.17]
46.2 ALA high ≥ 5 g/d 3 14224 Risk Ratio (M‐H, Random, 95% CI) 1.16 [0.84, 1.61]
47 CHD events ‐ ALA ‐ subgroup by replacement 4   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
47.1 N‐3 replacing SFA 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
47.2 N‐3 replacing MUFA 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.72, 1.17]
47.3 N‐3 replacing n‐6 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.19 [0.85, 1.65]
47.4 N‐3 replacing carbohydrates/sugars 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
47.5 N‐3 replacing nil/low n‐3 placebo 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
47.6 Replacement unclear 2 818 Risk Ratio (M‐H, Random, 95% CI) 0.69 [0.08, 5.81]
48 CHD events ‐ ALA ‐ subgroup by intervention type 4 19061 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.82, 1.22]
48.1 Dietary advice 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
48.2 Supplemental foods 3 5655 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.72, 1.16]
48.3 Supplements (capsule) 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.19 [0.85, 1.65]
48.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
49 CHD events ‐ ALA ‐ subgroup by duration 4 19061 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.82, 1.22]
49.1 Medium duration 1 to < 2 years in study 2 13516 Risk Ratio (M‐H, Random, 95% CI) 0.94 [0.34, 2.58]
49.2 Medium‐long duration: 2 to < 4 years in study 2 5545 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.73, 1.17]
49.3 Long duration: ≥ 4 years in study 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
50 CHD events ‐ ALA ‐ subgroup by primary or secondary prevention 4 19061 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.82, 1.22]
50.1 Primary prevention 2 14114 Risk Ratio (M‐H, Random, 95% CI) 1.20 [0.86, 1.67]
50.2 Secondary prevention 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.71, 1.15]
51 CHD events ‐ ALA ‐ subgroup by statin use 4 19061 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.82, 1.22]
51.1 ALA ‐ ≥ 50% of control group on statins 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.91 [0.71, 1.15]
51.2 ALA ‐ < 50% of control group on statins 2 14114 Risk Ratio (M‐H, Random, 95% CI) 1.20 [0.86, 1.67]
52 CHD events ‐ ALA ‐ subgroup by CAD history 4 19061 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.82, 1.22]
52.1 Previous CAD 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.72, 1.17]
52.2 No previous CAD 3 14224 Risk Ratio (M‐H, Random, 95% CI) 1.16 [0.84, 1.61]
53 Stroke (overall) ‐ ALA 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.15 [0.66, 2.01]
54 Stroke ‐ ALA ‐ SA fixed‐effect 5 19327 Risk Ratio (M‐H, Fixed, 95% CI) 1.23 [0.71, 2.13]
55 Stroke ‐ ALA ‐ SA by summary risk of bias 5   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
55.1 Low risk of bias 3 5213 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.45, 2.09]
55.2 Moderate/high risk of bias 2 14114 Risk Ratio (M‐H, Random, 95% CI) 1.39 [0.62, 3.13]
56 Stroke ‐ ALA ‐ SA by compliance and study size 5   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
56.1 SA ‐ low risk of compliance bias 3 5811 Risk Ratio (M‐H, Random, 95% CI) 0.85 [0.39, 1.87]
56.2 SA ‐ 100+ randomised 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.15 [0.66, 2.01]
57 Stroke ‐ ALA ‐ subgroup by dose 5   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
57.1 ALA low < 5 g/d 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.39, 2.15]
57.2 ALA high ≥ 5 g/d 4 14490 Risk Ratio (M‐H, Random, 95% CI) 1.36 [0.65, 2.85]
58 Stroke ‐ ALA ‐ subgroup by replacement 5   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
58.1 N‐3 replacing SFA 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
58.2 N‐3 replacing MUFA 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.92 [0.39, 2.15]
58.3 N‐3 replacing n‐6 2 13672 Risk Ratio (M‐H, Random, 95% CI) 1.26 [0.53, 3.01]
58.4 N‐3 replacing carbohydrates/sugars 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
58.5 N‐3 replacing nil/low n‐3 placebo 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
58.6 Replacement unclear 2 818 Risk Ratio (M‐H, Random, 95% CI) 1.79 [0.31, 10.17]
59 Stroke ‐ ALA ‐ subgroup by intervention type 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.15 [0.66, 2.01]
59.1 Dietary advice 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
59.2 Supplemental foods 4 5921 Risk Ratio (M‐H, Random, 95% CI) 0.97 [0.46, 2.03]
59.3 Supplements (capsule) 1 13406 Risk Ratio (M‐H, Random, 95% CI) 1.44 [0.62, 3.36]
59.4 Any combination 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
60 Stroke ‐ ALA ‐ subgroup by duration 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.15 [0.66, 2.01]
60.1 Medium duration 1 to < 2 years in study 2 13516 Risk Ratio (M‐H, Random, 95% CI) 1.56 [0.70, 3.44]
60.2 Medium‐long duration: 2 to < 4 years in study 3 5811 Risk Ratio (M‐H, Random, 95% CI) 0.85 [0.39, 1.87]
60.3 Long duration: ≥ 4 years in study 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
61 Stroke ‐ ALA ‐ subgroup by primary or secondary prevention 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.15 [0.66, 2.01]
61.1 Primary prevention 3 14380 Risk Ratio (M‐H, Random, 95% CI) 1.25 [0.57, 2.74]
61.2 Secondary prevention 2 4947 Risk Ratio (M‐H, Random, 95% CI) 1.05 [0.47, 2.34]
62 Stroke ‐ ALA ‐ subgroup by statin use 5 19327 Risk Ratio (M‐H, Random, 95% CI) 1.25 [0.71, 2.18]
62.1 ALA ‐ ≥ 50% of control group on statins 2 4947 Risk Ratio (M‐H, Random, 95% CI) 1.25 [0.56, 2.77]
62.2 ALA ‐ < 50% of control group on statins 3 14380 Risk Ratio (M‐H, Random, 95% CI) 1.25 [0.57, 2.74]
63 Stroke ‐ ALA ‐ subgroup by stroke type 3 13782 Risk Ratio (M‐H, Random, 95% CI) 1.40 [0.65, 3.01]
63.1 Ischaemic stroke ‐ ALA 3 13782 Risk Ratio (M‐H, Random, 95% CI) 1.40 [0.65, 3.01]
63.2 Haemorrhagic stroke ‐ ALA 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
64 Arrythmia (overall) ‐ ALA 2 4912 Risk Ratio (M‐H, Random, 95% CI) 0.73 [0.55, 0.97]
64.1 ALA ‐ new arrhythmias 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.79 [0.57, 1.10]
64.2 ALA ‐ recurrent arrhythmias 1 75 Risk Ratio (M‐H, Random, 95% CI) 0.60 [0.35, 1.03]
65 Arrythmia ‐ ALA ‐ SA fixed‐effect 2 4912 Risk Ratio (M‐H, Fixed, 95% CI) 0.75 [0.57, 1.00]
66 Arrhythmia ‐ ALA ‐ SA by summary risk of bias 2   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
66.1 Low risk of bias 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.79 [0.57, 1.10]
66.2 Moderate/high risk of bias 1 75 Risk Ratio (M‐H, Random, 95% CI) 0.60 [0.35, 1.03]
67 Arrythmia ‐ ALA ‐ SA by compliance and study size 1 9674 Risk Ratio (M‐H, Random, 95% CI) 0.79 [0.63, 1.00]
67.1 SA ‐ low risk of compliance bias 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.79 [0.57, 1.10]
67.2 SA ‐ 100+ randomised 1 4837 Risk Ratio (M‐H, Random, 95% CI) 0.79 [0.57, 1.10]

Comparison 5. High vs low ALA omega‐3 fat (secondary outcomes).

Outcome or subgroup title No. of studies No. of participants Statistical method Effect size
1 MACCEs ‐ ALA 1 110 Risk Ratio (M‐H, Random, 95% CI) 1.12 [0.32, 3.95]
2 Myocardial infarction (overall) ‐ ALA 3 18353 Risk Ratio (M‐H, Random, 95% CI) 1.00 [0.76, 1.32]
3 Total MI ‐ ALA ‐ subgroup by fatality 3   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
3.1 Fatal MI 2 4947 Risk Ratio (M‐H, Random, 95% CI) 0.95 [0.62, 1.46]
3.2 Non‐fatal MI 3 5213 Risk Ratio (M‐H, Random, 95% CI) 0.52 [0.15, 1.77]
4 Angina ‐ ALA 2 13516 Risk Ratio (M‐H, Random, 95% CI) 1.41 [0.75, 2.64]
5 Revascularisation ‐ ALA 1   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
5.1 CABG ‐ ALA 1 266 Risk Ratio (M‐H, Random, 95% CI) 0.29 [0.01, 5.93]
5.2 Angioplasty ‐ ALA 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
5.3 Any revascularisation ‐ ALA 1 266 Risk Ratio (M‐H, Random, 95% CI) 0.72 [0.07, 7.84]
6 Peripheral arterial disease ‐ ALA 1   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
7 Body weight, kg ‐ ALA 4 664 Mean Difference (IV, Random, 95% CI) ‐1.49 [‐4.17, 1.18]
8 Weight, kg ‐ ALA ‐ sensitivity analysis (SA) fixed‐effect 4 664 Mean Difference (IV, Fixed, 95% CI) 0.17 [‐0.61, 0.96]
9 Weight, kg ‐ ALA ‐ SA by summary risk of bias 4   Mean Difference (IV, Random, 95% CI) Subtotals only
9.1 Low risk of bias 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
9.2 Moderate/high risk of bias 4 664 Mean Difference (IV, Random, 95% CI) ‐1.49 [‐4.17, 1.18]
10 Weight, kg ‐ ALA ‐ SA by compliance and study size 3   Mean Difference (IV, Random, 95% CI) Subtotals only
10.1 SA ‐ low risk of compliance bias 3 629 Mean Difference (IV, Random, 95% CI) ‐1.59 [‐4.47, 1.30]
10.2 SA ‐ 100+ randomised 3 629 Mean Difference (IV, Random, 95% CI) ‐1.59 [‐4.47, 1.30]
11 Weight, kg ‐ ALA ‐ subgroup by dose 4   Mean Difference (IV, Random, 95% CI) Subtotals only
11.1 ALA low < 5 g/d 3 485 Mean Difference (IV, Random, 95% CI) ‐0.71 [‐3.31, 1.90]
11.2 ALA high > 5 g/d 1 179 Mean Difference (IV, Random, 95% CI) ‐4.20 [‐7.61, ‐0.79]
12 Weight, kg ‐ ALA ‐ subgroup by intervention type 4   Mean Difference (IV, Random, 95% CI) Subtotals only
12.1 Dietary advice 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
12.2 Supplemental foods 3 526 Mean Difference (IV, Random, 95% CI) ‐1.23 [‐5.27, 2.80]
12.3 Supplement (capsule) 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
12.4 Any combination 1 138 Mean Difference (IV, Random, 95% CI) ‐1.98 [‐5.89, 1.92]
13 Weight, kg ‐ ALA ‐ subgroup by replacement 4   Mean Difference (IV, Random, 95% CI) Subtotals only
13.1 ALA replacing SFA 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
13.2 ALA replacing MUFA 1 138 Mean Difference (IV, Random, 95% CI) ‐1.98 [‐5.89, 1.92]
13.3 ALA replacing n‐6 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
13.4 ALA replacing carbs/sugars 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
13.5 ALA replacing nil/low n‐3 placebo 1 35 Mean Difference (IV, Random, 95% CI) ‐0.30 [‐10.57, 9.97]
13.6 Replacement unclear 2 491 Mean Difference (IV, Random, 95% CI) ‐1.43 [‐6.26, 3.39]
14 Weight, kg ‐ ALA ‐ subgroup by duration 4   Mean Difference (IV, Random, 95% CI) Subtotals only
14.1 Medium duration 1 to < 2 years in study 4 664 Mean Difference (IV, Random, 95% CI) ‐1.49 [‐4.17, 1.18]
14.2 Medium‐long duration: 2 to < 4 years in study 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
14.3 Long duration ≥ 4 years in study 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
15 Weight, kg ‐ ALA ‐ subgroup by statin use 4   Mean Difference (IV, Random, 95% CI) Subtotals only
15.1 ALA ‐ ≥ 50% of control group on statins 1 35 Mean Difference (IV, Random, 95% CI) ‐0.30 [‐10.57, 9.97]
15.2 ALA ‐ < 50% of control group on statins 1 138 Mean Difference (IV, Random, 95% CI) ‐1.98 [‐5.89, 1.92]
15.3 ALA ‐ use of statins unclear 2 491 Mean Difference (IV, Random, 95% CI) ‐1.43 [‐6.26, 3.39]
16 Weight, kg ‐ ALA ‐ subgroup by primary or secondary prevention 4   Mean Difference (IV, Random, 95% CI) Subtotals only
16.1 Low CVD risk 3 629 Mean Difference (IV, Random, 95% CI) ‐1.59 [‐4.47, 1.30]
16.2 Moderate CVD risk 1 35 Mean Difference (IV, Random, 95% CI) ‐0.30 [‐10.57, 9.97]
16.3 High CVD risk 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
17 Body mass index, kg/m² ‐ ALA 3 1581 Mean Difference (IV, Random, 95% CI) ‐0.42 [‐1.53, 0.69]
18 BMI, kg/m² ‐ ALA ‐ SA fixed‐effect 3 1581 Mean Difference (IV, Fixed, 95% CI) 0.12 [‐0.06, 0.30]
19 BMI, kg/m² ‐ ALA ‐ SA by summary risk of bias 3   Mean Difference (IV, Random, 95% CI) Subtotals only
19.1 Low risk of bias 2 1402 Mean Difference (IV, Random, 95% CI) 0.15 [‐0.04, 0.33]
19.2 Moderate/high risk of bias 1 179 Mean Difference (IV, Random, 95% CI) ‐1.5 [‐2.86, ‐0.14]
20 BMI, kg/m² ‐ ALA ‐ SA by compliance and study size 3   Mean Difference (IV, Random, 95% CI) Subtotals only
20.1 SA ‐ low risk of compliance bias 3 1581 Mean Difference (IV, Random, 95% CI) ‐0.42 [‐1.53, 0.69]
20.2 SA ‐ 100+ randomised 3 1581 Mean Difference (IV, Random, 95% CI) ‐0.42 [‐1.53, 0.69]
21 BMI, kg/m² ‐ ALA ‐ subgroup by dose 3   Mean Difference (IV, Random, 95% CI) Subtotals only
21.1 ALA low < 5 g/d 1 1260 Mean Difference (IV, Random, 95% CI) 0.15 [‐0.03, 0.33]
21.2 ALA high > 5 g/d 2 321 Mean Difference (IV, Random, 95% CI) ‐1.12 [‐2.24, 0.01]
22 BMI, kg/m² ‐ ALA ‐ subgroup by intervention type 3   Mean Difference (IV, Random, 95% CI) Subtotals only
22.1 Dietary advice 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
22.2 Supplemental foods 3 1581 Mean Difference (IV, Random, 95% CI) ‐0.42 [‐1.53, 0.69]
22.3 Supplement (capsule) 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
22.4 Any combination 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
23 BMI, kg/m² ‐ ALA ‐ subgroup by replacement 3   Mean Difference (IV, Random, 95% CI) Subtotals only
23.1 ALA replacing SFA 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
23.2 ALA replacing MUFA 1 1260 Mean Difference (IV, Random, 95% CI) 0.15 [‐0.03, 0.33]
23.3 ALA replacing n‐6 1 142 Mean Difference (IV, Random, 95% CI) ‐0.3 [‐2.29, 1.69]
23.4 ALA replacing carbs/sugars 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
23.5 ALA replacing nil/low n‐3 placebo 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
23.6 Replacement unclear 1 179 Mean Difference (IV, Random, 95% CI) ‐1.5 [‐2.86, ‐0.14]
24 BMI, kg/m² ‐ ALA ‐ subgroup by duration 3   Mean Difference (IV, Random, 95% CI) Subtotals only
24.1 Medium duration 1 to < 2 years in study 1 179 Mean Difference (IV, Random, 95% CI) ‐1.5 [‐2.86, ‐0.14]
24.2 Medium‐long duration: 2 to < 4 years in study 2 1402 Mean Difference (IV, Random, 95% CI) 0.15 [‐0.04, 0.33]
24.3 Long duration ≥ 4 years in study 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
25 BMI, kg/m² ‐ ALA ‐ subgroup by statin use 3   Mean Difference (IV, Random, 95% CI) Subtotals only
25.1 ALA ‐ ≥ 50% of control group on statins 1 1260 Mean Difference (IV, Random, 95% CI) 0.15 [‐0.03, 0.33]
25.2 ALA ‐ < 50% of control group on statins 1 142 Mean Difference (IV, Random, 95% CI) ‐0.3 [‐2.29, 1.69]
25.3 ALA ‐ use of statins unclear 1 179 Mean Difference (IV, Random, 95% CI) ‐1.5 [‐2.86, ‐0.14]
26 BMI, kg/m² ‐ ALA ‐ subgroup by primary or secondary preventionA 3   Mean Difference (IV, Random, 95% CI) Subtotals only
26.1 Primary prevention of CVD 2 321 Mean Difference (IV, Random, 95% CI) ‐1.12 [‐2.24, 0.01]
26.2 Secondary prevention of CVD 1 1260 Mean Difference (IV, Random, 95% CI) 0.15 [‐0.03, 0.33]
27 Other measures of adiposity ‐ ALA 4   Mean Difference (IV, Fixed, 95% CI) Subtotals only
27.1 Visceral adipose tissue, cm² 1 35 Mean Difference (IV, Fixed, 95% CI) 27.0 [‐21.28, 75.28]
27.2 Subcutaneous adipose tissue, cm² 1 35 Mean Difference (IV, Fixed, 95% CI) 0.0 [0.0, 0.0]
27.3 Waist circumference, cm 3 629 Mean Difference (IV, Fixed, 95% CI) ‐1.59 [‐3.10, ‐0.07]
28 Total cholesterol, serum, mmoL/L ‐ ALA 6 2164 Mean Difference (IV, Random, 95% CI) ‐0.09 [‐0.23, 0.05]
29 TC, mmoL/L ‐ ALA ‐ SA fixed‐effect 6 2164 Mean Difference (IV, Fixed, 95% CI) ‐0.10 [‐0.17, ‐0.03]
30 TC, mmoL/L ‐ ALA ‐ SA by summary risk of bias 6   Mean Difference (IV, Random, 95% CI) Subtotals only
30.1 Low risk of bias 3 1436 Mean Difference (IV, Random, 95% CI) 0.00 [‐0.13, 0.14]
30.2 Moderate/high risk of bias 3 728 Mean Difference (IV, Random, 95% CI) ‐0.19 [‐0.36, ‐0.01]
31 TC, mmoL/L ‐ ALA ‐ SA by compliance and study size 4   Mean Difference (IV, Random, 95% CI) Subtotals only
31.1 SA ‐ low risk of compliance bias 4 2045 Mean Difference (IV, Random, 95% CI) ‐0.10 [‐0.25, 0.05]
31.2 SA ‐ 100+ randomised 4 2045 Mean Difference (IV, Random, 95% CI) ‐0.10 [‐0.25, 0.05]
32 TC, mmoL/L ‐ ALA ‐ subgroup by dose 6   Mean Difference (IV, Random, 95% CI) Subtotals only
32.1 ALA low < 5 g/d 3 1759 Mean Difference (IV, Random, 95% CI) ‐0.07 [‐0.24, 0.09]
32.2 ALA high > 5 g/d 3 405 Mean Difference (IV, Random, 95% CI) ‐0.13 [‐0.47, 0.21]
33 TC, mmoL/L ‐ ALA ‐ subgroup by intervention type 6   Mean Difference (IV, Random, 95% CI) Subtotals only
33.1 Dietary advice 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
33.2 Supplemental foods 6 2164 Mean Difference (IV, Random, 95% CI) ‐0.09 [‐0.23, 0.05]
33.3 Supplement (capsule) 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
33.4 Any combination 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
34 TC, mmoL/L ‐ ALA ‐ subgroup by replacement 6   Mean Difference (IV, Random, 95% CI) Subtotals only
34.1 ALA replacing SFA 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
34.2 ALA replacing MUFA 1 1210 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.13, 0.09]
34.3 ALA replacing n‐6 1 142 Mean Difference (IV, Random, 95% CI) 0.14 [‐0.10, 0.38]
34.4 ALA replacing carbs/sugars 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
34.5 ALA replacing nil/low n‐3 placebo 1 35 Mean Difference (IV, Random, 95% CI) 0.30 [‐0.30, 0.90]
34.6 Replacement unclear 3 777 Mean Difference (IV, Random, 95% CI) ‐0.21 [‐0.31, ‐0.11]
35 TC, mmoL/L ‐ ALA ‐ subgroup by duration 6   Mean Difference (IV, Random, 95% CI) Subtotals only
35.1 Medium duration 1 to < 2 years in study 4 812 Mean Difference (IV, Random, 95% CI) ‐0.20 [‐0.33, ‐0.07]
35.2 Medium‐long duration: 2 to < 4 years in study 2 1352 Mean Difference (IV, Random, 95% CI) 0.02 [‐0.12, 0.16]
35.3 Long duration ≥ 4 years in study 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
36 TC, mmoL/L ‐ ALA ‐ subgroup by statin use 6   Mean Difference (IV, Random, 95% CI) Subtotals only
36.1 ALA ‐ ≥ 50% of control group on statins 3 1329 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.15, 0.11]
36.2 ALA ‐ < 50% of control group on statins 1 142 Mean Difference (IV, Random, 95% CI) 0.14 [‐0.10, 0.38]
36.3 ALA ‐ use of statins unclear 2 693 Mean Difference (IV, Random, 95% CI) ‐0.21 [‐0.30, ‐0.11]
37 TC, mmoL/L ‐ ALA ‐ subgroup by primary or secondary prevention 6   Mean Difference (IV, Random, 95% CI) Subtotals only
37.1 Primary prevention of CVD 4 870 Mean Difference (IV, Random, 95% CI) ‐0.09 [‐0.30, 0.12]
37.2 Secondary prevention of CVD 2 1294 Mean Difference (IV, Random, 95% CI) ‐0.03 [‐0.14, 0.08]
38 Triglycerides, fasting, serum, mmoL/L ‐ ALA 6 1776 Mean Difference (IV, Random, 95% CI) ‐0.03 [‐0.11, 0.05]
39 TG, fasting, mmoL/L ‐ ALA ‐ SA fixed‐effect 6 1776 Mean Difference (IV, Fixed, 95% CI) ‐0.03 [‐0.11, 0.05]
40 TG, fasting, mmoL/L‐ ALA ‐ SA by summary risk of bias 6   Mean Difference (IV, Random, 95% CI) Subtotals only
40.1 Low risk of bias 3 1436 Mean Difference (IV, Random, 95% CI) 0.03 [‐0.13, 0.19]
40.2 Moderate/high risk of bias 3 340 Mean Difference (IV, Random, 95% CI) ‐0.05 [‐0.18, 0.09]
41 TG, fasting, mmoL/L‐ ALA ‐ SA by compliance and study size 4   Mean Difference (IV, Random, 95% CI) Subtotals only
41.1 SA ‐ low risk of compliance bias 4 1657 Mean Difference (IV, Random, 95% CI) ‐0.05 [‐0.13, 0.04]
41.2 SA ‐ 100+ randomised 4 1657 Mean Difference (IV, Random, 95% CI) ‐0.05 [‐0.13, 0.04]
42 TG, fasting, mmoL/L ‐ ALA ‐ subgroup by dose 6   Mean Difference (IV, Random, 95% CI) Subtotals only
42.1 ALA low < 5 g/d 3 1371 Mean Difference (IV, Random, 95% CI) ‐0.07 [‐0.16, 0.03]
42.2 ALA high > 5 g/d 3 405 Mean Difference (IV, Random, 95% CI) 0.05 [‐0.09, 0.19]
43 TG, fasting, mmoL/L‐ ALA ‐ subgroup by intervention type 6   Mean Difference (IV, Random, 95% CI) Subtotals only
43.1 Dietary advice 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
43.2 Supplemental foods 5 1650 Mean Difference (IV, Random, 95% CI) ‐0.01 [‐0.10, 0.07]
43.3 Supplement (capsule) 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
43.4 Any combination 1 126 Mean Difference (IV, Random, 95% CI) ‐0.12 [‐0.33, 0.09]
44 TG, fasting, mmoL/L‐AL ‐ subgroup by replacement 6   Mean Difference (IV, Random, 95% CI) Subtotals only
44.1 ALA replacing SFA 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
44.2 ALA replacing MUFA 2 1336 Mean Difference (IV, Random, 95% CI) ‐0.07 [‐0.17, 0.02]
44.3 ALA replacing n‐6 1 142 Mean Difference (IV, Random, 95% CI) 0.13 [‐0.16, 0.42]
44.4 ALA replacing carbs/sugars 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
44.5 ALA replacing nil/low n‐3 placebo 1 35 Mean Difference (IV, Random, 95% CI) 0.30 [‐0.39, 0.99]
44.6 Replacement unclear 2 263 Mean Difference (IV, Random, 95% CI) 0.04 [‐0.15, 0.23]
45 TG, fasting, mmoL/L‐ ALA ‐ subgroup by duration 6   Mean Difference (IV, Random, 95% CI) Subtotals only
45.1 Medium duration 1 to < 2 years in study 4 424 Mean Difference (IV, Random, 95% CI) ‐0.01 [‐0.15, 0.12]
45.2 Medium‐long duration: 2 to < 4 years in study 2 1352 Mean Difference (IV, Random, 95% CI) ‐0.01 [‐0.17, 0.15]
45.3 Long duration ≥ 4 years in study 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
46 TG, fasting, mmoL/L ‐ ALA ‐ subgroup by statin use 6   Mean Difference (IV, Random, 95% CI) Subtotals only
46.1 ALA ‐ ≥ 50% of control group on statins 3 1329 Mean Difference (IV, Random, 95% CI) 0.03 [‐0.17, 0.23]
46.2 ALA ‐ < 50% of control group on statins 2 268 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.26, 0.23]
46.3 ALA ‐ use of statins unclear 1 179 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.20, 0.16]
47 TG, fasting, mmoL/L‐ ALA ‐ subgroup by primary or secondary prevention 6   Mean Difference (IV, Random, 95% CI) Subtotals only
47.1 Primary prevention 4 482 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.14, 0.11]
47.2 Secondary prevention 2 1294 Mean Difference (IV, Random, 95% CI) 0.02 [‐0.22, 0.25]
48 High‐density lipoprotein, serum, mmoL/L ‐ ALA 6 1776 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.08, 0.03]
49 HDL, mmoL/L ‐ ALA ‐ SA fixed‐effect 6 1776 Mean Difference (IV, Fixed, 95% CI) ‐0.02 [‐0.05, 0.00]
50 HDL, mmoL/L ‐ ALA ‐ SA by summary risk of bias 6   Mean Difference (IV, Random, 95% CI) Subtotals only
50.1 Low risk of bias 3 1436 Mean Difference (IV, Random, 95% CI) ‐0.03 [‐0.06, 0.00]
50.2 Moderate/high risk of bias 3 340 Mean Difference (IV, Random, 95% CI) 0.04 [‐0.14, 0.22]
51 HDL, mmoL/L ‐ ALA ‐ SA by compliance and study size 4   Mean Difference (IV, Random, 95% CI) Subtotals only
51.1 SA ‐ low risk of compliance bias 4 1657 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.08, 0.04]
51.2 SA ‐ 100+ randomised 4 1657 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.08, 0.04]
52 HDL, mmoL/L ‐ ALA ‐ subgroup by dose 6   Mean Difference (IV, Random, 95% CI) Subtotals only
52.1 ALA low < 5 g/d 3 1371 Mean Difference (IV, Random, 95% CI) 0.06 [‐0.08, 0.19]
52.2 ALA high > 5 g/d 3 405 Mean Difference (IV, Random, 95% CI) ‐0.07 [‐0.12, ‐0.01]
53 HDL, mmoL/L ‐ ALA ‐ subgroup by intervention type 6   Mean Difference (IV, Random, 95% CI) Subtotals only
53.1 Dietary advice 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
53.2 Supplemental foods 5 1650 Mean Difference (IV, Random, 95% CI) ‐0.03 [‐0.06, ‐0.00]
53.3 Supplement (capsule) 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
53.4 Any combination 1 126 Mean Difference (IV, Random, 95% CI) 0.15 [0.01, 0.29]
54 HDL, mmoL/L ‐ ALA ‐ subgroup by replacement 6   Mean Difference (IV, Random, 95% CI) Subtotals only
54.1 ALA replacing SFA 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
54.2 ALA replacing MUFA 2 1336 Mean Difference (IV, Random, 95% CI) 0.05 [‐0.11, 0.22]
54.3 ALA replacing n‐6 1 142 Mean Difference (IV, Random, 95% CI) ‐0.04 [‐0.11, 0.03]
54.4 ALA replacing carbs/sugars 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
54.5 ALA replacing nil/low n‐3 placebo 1 35 Mean Difference (IV, Random, 95% CI) 0.10 [‐0.17, 0.37]
54.6 Replacement unclear 2 263 Mean Difference (IV, Random, 95% CI) ‐0.09 [‐0.17, ‐0.02]
55 HDL, mmoL/L ‐ ALA ‐ subgroup by duration 6   Mean Difference (IV, Random, 95% CI) Subtotals only
55.1 Medium duration 1 to < 2 years in study 4 424 Mean Difference (IV, Random, 95% CI) ‐0.00 [‐0.13, 0.13]
55.2 Medium‐long duration: 2 to < 4 years in study 2 1352 Mean Difference (IV, Random, 95% CI) ‐0.02 [‐0.05, 0.00]
55.3 Long duration ≥ 4 years in study 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
56 HDL, mmoL/L ‐ ALA ‐ subgroup by statin use 6   Mean Difference (IV, Random, 95% CI) Subtotals only
56.1 ALA ‐ ≥ 50% of control group on statins 3 1329 Mean Difference (IV, Random, 95% CI) ‐0.03 [‐0.09, 0.03]
56.2 ALA ‐ < 50% of control group on statins 2 268 Mean Difference (IV, Random, 95% CI) 0.05 [‐0.14, 0.23]
56.3 ALA ‐ use of statins unclear 1 179 Mean Difference (IV, Random, 95% CI) ‐0.09 [‐0.20, 0.02]
57 HDL, mmoL/L ‐ ALA ‐ subgroup by primary or secondary prevention 6   Mean Difference (IV, Random, 95% CI) Subtotals only
57.1 Low CVD risk 2 305 Mean Difference (IV, Random, 95% CI) 0.03 [‐0.21, 0.26]
57.2 Moderate CVD risk 2 177 Mean Difference (IV, Random, 95% CI) ‐0.03 [‐0.10, 0.04]
57.3 High CVD risk 2 1294 Mean Difference (IV, Random, 95% CI) ‐0.04 [‐0.11, 0.03]
58 Low‐density lipoprotein, serum, mmoL/L ‐ ALA 7 2201 Mean Difference (IV, Random, 95% CI) ‐0.05 [‐0.15, 0.04]
59 LDL, mmoL/L ‐ ALA ‐ SA fixed‐effect 7 2201 Mean Difference (IV, Fixed, 95% CI) ‐0.05 [‐0.11, 0.00]
60 LDL, mmoL/L ‐ ALA ‐ SA by summary risk of bias 7   Mean Difference (IV, Random, 95% CI) Subtotals only
60.1 Low risk of bias 3 1350 Mean Difference (IV, Random, 95% CI) 0.02 [‐0.05, 0.10]
60.2 Moderate/high risk of bias 4 851 Mean Difference (IV, Random, 95% CI) ‐0.14 [‐0.22, ‐0.06]
61 LDL, mmoL/L ‐ ALA ‐ SA by compliance and study size 5   Mean Difference (IV, Random, 95% CI) Subtotals only
61.1 SA ‐ low risk of compliance bias 5 2085 Mean Difference (IV, Random, 95% CI) ‐0.05 [‐0.16, 0.06]
61.2 SA ‐ 100+ randomised 5 2085 Mean Difference (IV, Random, 95% CI) ‐0.05 [‐0.16, 0.06]
62 LDL, mmoL/L ‐ ALA ‐ subgroup by dose 7   Mean Difference (IV, Random, 95% CI) Subtotals only
62.1 ALA low < 5 g/d 4 1796 Mean Difference (IV, Random, 95% CI) ‐0.06 [‐0.17, 0.05]
62.2 ALA high > 5 g/d 3 405 Mean Difference (IV, Random, 95% CI) ‐0.04 [‐0.28, 0.19]
63 LDL, mmoL/L ‐ ALA ‐ subgroup by intervention type 7   Mean Difference (IV, Random, 95% CI) Subtotals only
63.1 Dietary advice 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
63.2 Supplemental foods 6 2075 Mean Difference (IV, Random, 95% CI) ‐0.06 [‐0.17, 0.05]
63.3 Supplement (capsule) 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
63.4 Any combination 1 126 Mean Difference (IV, Random, 95% CI) 0.0 [‐0.25, 0.25]
64 LDL, mmoL/L ‐ ALA ‐ subgroup by replacement 4   Mean Difference (IV, Random, 95% CI) Subtotals only
64.1 ALA replacing SFA 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
64.2 ALA replacing MUFA 2 1250 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.07, 0.09]
64.3 ALA replacing n‐6 1 142 Mean Difference (IV, Random, 95% CI) 0.14 [‐0.08, 0.36]
64.4 ALA replacing carbs/sugars 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
64.5 ALA replacing nil/low n‐3 placebo 1 32 Mean Difference (IV, Random, 95% CI) ‐0.10 [‐0.59, 0.39]
65 LDL, mmoL/L ‐ ALA ‐ subgroup by duration 7   Mean Difference (IV, Random, 95% CI) Subtotals only
65.1 Medium duration 1 to < 2 years in study 5 935 Mean Difference (IV, Random, 95% CI) ‐0.14 [‐0.22, ‐0.06]
65.2 Medium‐long duration: 2 to < 4 years in study 2 1266 Mean Difference (IV, Random, 95% CI) 0.03 [‐0.06, 0.13]
65.3 Long duration ≥ 4 years in study 0 0 Mean Difference (IV, Random, 95% CI) 0.0 [0.0, 0.0]
66 LDL, mmoL/L ‐ ALA ‐ subgroup by statin use 7   Mean Difference (IV, Random, 95% CI) Subtotals only
66.1 ALA ‐ ≥ 50% of control group on statins 3 1240 Mean Difference (IV, Random, 95% CI) 0.00 [‐0.08, 0.08]
66.2 ALA ‐ < 50% of control group on statins 2 268 Mean Difference (IV, Random, 95% CI) 0.08 [‐0.09, 0.24]
66.3 ALA ‐ use of statins unclear 2 693 Mean Difference (IV, Random, 95% CI) ‐0.16 [‐0.25, ‐0.07]
67 LDL, mmoL/L ‐ ALA ‐ subgroup by primary or secondary prevention 7   Mean Difference (IV, Random, 95% CI) Subtotals only
67.1 Primary prevention of CVD 5 993 Mean Difference (IV, Random, 95% CI) ‐0.08 [‐0.20, 0.05]
67.2 Secondary prevention of CVD 2 1208 Mean Difference (IV, Random, 95% CI) 0.01 [‐0.08, 0.09]

Comparison 6. High vs low ALA omega‐3 fats (tertiary outcomes).

Outcome or subgroup title No. of studies No. of participants Statistical method Effect size
1 Blood pressure, mmHg ‐ ALA 4   Mean Difference (IV, Random, 95% CI) Subtotals only
1.1 Systolic BP ‐ ALA 4 1671 Mean Difference (IV, Random, 95% CI) ‐0.87 [‐4.48, 2.75]
1.2 Diastolic BP ‐ ALA 4 1671 Mean Difference (IV, Random, 95% CI) ‐1.42 [‐4.40, 1.57]
2 Serious adverse events ‐ ALA 2   Risk Ratio (M‐H, Random, 95% CI) Subtotals only
2.1 Any serious adverse events 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
2.2 Bleeding 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
2.3 Serious GI effects 0 0 Risk Ratio (M‐H, Random, 95% CI) 0.0 [0.0, 0.0]
2.4 Pulmonary embolus or DVT 1 708 Risk Ratio (M‐H, Random, 95% CI) 0.32 [0.01, 7.80]
2.5 Thrombophleibitis 1 13406 Risk Ratio (M‐H,