Omega-3 Fish Oil

Fish oil is the most-taken supplement in the world for which the flagship claim has been most thoroughly disproved, and that tension is the only honest way to describe it.

The claim came from epidemiology and from two early trials. GISSI-Prevenzione gave 1 g a day of omega-3 to 11,324 people who had recently had a heart attack and found a 10–15% reduction in the combined endpoint of death, reinfarction and stroke, driven by fewer deaths [6]. JELIS added 1,800 mg of EPA to a statin in 18,645 Japanese adults with high cholesterol and found 19% fewer major coronary events over 4.6 years, with no difference in LDL cholesterol between the arms [7]. Between them, they made fish oil a cardiological recommendation and then a global supplement category.

The large modern trials did not reproduce it. VITAL randomised 25,871 US adults to 1 g a day for a median of 5.3 years: major cardiovascular events HR 0.92, not significant, and no effect on invasive cancer [3]. ASCEND randomised 15,480 people with diabetes for 7.4 years: serious vascular events RR 0.97, not significant [4]. STRENGTH gave 4 g a day of an EPA+DHA carboxylic acid preparation against corn oil to 13,078 high-risk statin-treated patients and was stopped early for futility, HR 0.99 [5]. OMEMI gave 1.8 g a day to elderly heart attack survivors for two years: HR 1.08, and more new atrial fibrillation in the treated arm [8]. The Cochrane review pooling 79 trials and 112,059 participants found high-quality evidence of little or no effect on all-cause mortality (RR 0.98) or cardiovascular events (RR 0.99), and noted that the earlier suggestions of benefit came from trials at higher risk of bias [1]. A later dose-stratified meta-analysis of 19 double-blind trials in 97,709 people reached the same place from a different direction: no significant association with all-cause mortality, cardiac death, myocardial infarction or stroke at any dose category, with one isolated exception for cardiac mortality at the equivalent of two 1 g capsules a day [9].

One large trial went the other way. REDUCE-IT gave 4 g a day of icosapent ethyl — purified EPA, no DHA — to 8,179 statin-treated patients with triglycerides of 135 to 499 mg/dL, and the primary composite endpoint occurred in 17.2% against 22.0% on placebo, HR 0.75, with cardiovascular death also reduced [10]. Why REDUCE-IT succeeded where STRENGTH, at the same 4 g dose, failed is unresolved. The candidate explanations are that EPA alone differs from EPA plus DHA, that the two trials used different comparator oils, and that the trials enrolled somewhat different populations [5][10].

What survives all of this is triglyceride lowering, which the Cochrane review rated as high-quality evidence: about 15%, dose-dependently, and the only lipid or adiposity outcome that moved at all [1]. At prescription doses of 4 g a day the reduction is much larger, and three formulations are licensed in the US for severe hypertriglyceridaemia [2][11]. There is also a reasonable case in diagnosed depression, where EPA-predominant preparations beat placebo [12], and in inflammatory arthritis at doses above 2 g a day [13]. And there are several widely believed uses with clean negative trials behind them, dry eye being the clearest [14][15].

EPA and DHA change the membrane they are put into. They are incorporated into cell membrane phospholipids in proportion to intake, which alters membrane fluidity and, more importantly, the pool of substrate available to the enzymes that make lipid signalling molecules [16][17]. That incorporation is measurable: the erythrocyte omega-3 index rises dose-dependently, with dose alone accounting for 68% of the variance in response across 0 to 1800 mg a day over five months, and dose per kilogram of body weight slightly more [17]. Adding baseline index, age, sex and physical activity brought the explained variance to 78% — which means a fixed capsule dose produces genuinely different tissue levels in different people, mostly because of body size.

The triglyceride effect is a hepatic secretion effect. Giving obese men 3.2 g a day of omega-3 ethyl esters for six weeks cut plasma triglycerides by 14% and VLDL-triglyceride by 32%, and the reason was a 32% fall in hepatic VLDL-triglyceride secretion, with no change in the rate at which VLDL was cleared [18]. The liver makes less, rather than the body clearing more. The enzyme-level account is a combination of inhibited diacylglycerol acyltransferase, increased lipoprotein lipase activity, less hepatic lipogenesis and more hepatic β-oxidation [11]. This also explains an awkward detail: DHA-containing formulations can raise LDL cholesterol while lowering triglycerides, whereas EPA-only preparations do not [11].

Inflammation is resolved, not just suppressed. EPA and DHA are the substrates for a family of lipid mediators — resolvins, protectins and maresins, collectively the specialised pro-resolving mediators — which are produced in a defined temporal sequence by resolving inflammatory exudates and actively drive the return to homeostasis, including enhanced microbial clearance [19]. This is a different pharmacology from an anti-inflammatory drug that blocks a pathway, and it is the mechanistic argument behind the arthritis results [13].

In atherosclerosis the mechanistic case is plausible and the outcome case is contested. EPA incorporates into plaques and has reported effects on endothelial function, oxidative stress, foam cell formation, platelet aggregation, thrombus formation and plaque rupture, along with the triglyceride reduction that does not raise LDL [16]. All of that was assembled as the rationale for REDUCE-IT, and it was written before the trial read out [16]. The trial was positive; the mechanistically similar STRENGTH, at the same dose, was not [5][10]. Mechanistic plausibility and outcome evidence have parted company here, and the plausibility should not be used to argue past the trials.

The form matters more for fish oil than for almost any other supplement, because it determines how much of the label dose actually gets absorbed.

The difference is not subtle. In a four-way crossover in 54 overweight adults, a single 4 g dose of an omega-3 free fatty acid preparation gave 4.0 times the EPA+DHA exposure of a prescription ethyl ester during a low-fat eating period, and 1.3 times during a high-fat one. During the low-fat period, 59% of people on the free fatty acid form retained at least half of their high-fat exposure — against 6% on the ethyl ester [20]. A review of the three US prescription products puts the carboxylic acid form's bioavailability at up to four-fold that of the ethyl esters, precisely because it does not depend on pancreatic lipase hydrolysis [11]. At least one reformulated ethyl ester, using an excipient system that forms micelles independently of bile salts, removed the food effect for DHA and reduced it for EPA [21]. Whether the chemical form also changes how much ends up deposited in tissue, rather than merely how much appears in plasma, is still an open question [22].

The practical consequence: if a product is an ethyl ester — which most concentrated, high-EPA products are — taking it with a low-fat breakfast can waste most of the dose.

Quality is the other formulation issue, and the evidence is genuinely mixed rather than uniformly bad. A 2015 analysis of all 32 encapsulated fish oil supplements sold in New Zealand found only 3 contained at least their labelled EPA and DHA, 69% contained less than two-thirds of the label claim, 83% exceeded recommended peroxide values, 50% exceeded recommended total oxidation values, and only 8% met all the international quality indices. Best-before date, price, country of origin and "exclusivity" were all poor predictors of which products were good [23]. A follow-up in 2016 covering Australian and New Zealand products reached the opposite conclusion — products met their omega-3 label claims and were not oxidised [24]. The two analyses used different product sets and methods; the reasonable reading is that quality varies by market and by year, and that the front-of-bottle number is not a guarantee.

Fish oil at supplement doses is well tolerated. The commonest complaints are gastrointestinal — reflux, fishy burps, loose stools — and no excess of serious adverse events or bleeding appeared in VITAL over 5.3 years [3]. The Cochrane review found no evidence that increasing EPA and DHA altered serious adverse events [1].

The bleeding warning is largely wrong. Patients are routinely told to stop fish oil before surgery. A randomised trial in 1,516 cardiac surgery patients gave 8–10 g of EPA+DHA for two to five days before the operation and 2 g a day afterwards, which is a far larger dose than anyone takes as a supplement. Major bleeding occurred in 6.1% overall, and the odds ratio against placebo was 0.81 — no increase. Transfusion requirements were lower in the fish oil group, and higher achieved plasma omega-3 levels were associated with less bleeding, with roughly a two-thirds reduction in the top quartiles compared with the lowest [26].

The atrial fibrillation signal is real but dose- and risk-dependent. REDUCE-IT reported more hospitalisation for atrial fibrillation or flutter on 4 g a day of icosapent ethyl, 3.1% against 2.1% [10], and OMEMI found new atrial fibrillation in 7.2% on 1.8 g a day against 4.0% on placebo, which fell just short of significance [8]. An earlier meta-analysis of cardiovascular outcome trials found an overall association [27]. The most complete analysis, covering 35 trials and 114,592 people and incorporating unpublished adverse-event data, found the risk is concentrated rather than general: only high-cardiovascular-risk patients taking more than 1,500 mg a day showed a significant increase, odds ratio 1.43 with an absolute risk difference of 0.8%. Low-dose EPA+DHA did not raise atrial fibrillation risk even in high-risk populations, and neither did high doses in low-risk ones [25]. Consistently, VITAL's 1 g a day produced no effect on incident atrial fibrillation in 25,119 participants [28].

DHA-containing preparations can raise LDL cholesterol while lowering triglycerides, which is one reason the prescription EPA-only product exists; the accompanying non-HDL cholesterol does not rise, and non-HDL is thought to be the better risk indicator in this population [11].

Oxidation is the quality issue rather than a toxicity issue in the strict sense, but a supplement that exceeds peroxide and total oxidation limits is not delivering what it claims, and most products in one national survey did [23].