Fish Oil

What it is

Fish oil supplies omega-3 fatty acids, mainly EPA and DHA, from oily fish or capsules [1]. Prescription icosapent ethyl is a highly purified form of EPA [2].

What the research says

Several meta-analyses of placebo-controlled trials point to a small benefit for depression symptoms with EPA-rich omega-3. A 2019 review of 26 trials (2,160 people) found an overall effect of SMD -0.28, where a more negative number means more improvement [3]. Formulas with at least 60% EPA at 1 g a day or less gave SMD -0.50, while DHA-only products showed no benefit [3]. A 2023 review of 10 trials (1,426 people) found benefit with EPA at 60% or more of the EPA-plus-DHA content, and with EPA at 1 to under 2 g a day [4]. A 2025 meta-analysis of 36 trials found a significant effect on symptoms (SMD -0.26), with no difference from placebo in response or remission rates [5].

The effect is small and the evidence is low certainty. Cochrane's 2021 review of 33 placebo-controlled studies (1,848 people) found an SMD of -0.40, but said this is unlikely to be clinically meaningful and may be biased toward a positive result [6].

Prevention is different. In VITAL-DEP, a trial of 18,353 adults aged 50 and older, 1 g a day of fish oil was linked to a small rise in depression risk (HR 1.13, 95% CI 1.01 to 1.26), while mood scores did not change [7]. The trial's authors say these findings do not support omega-3 supplements to prevent depression.

Higher omega-3 intake from food is linked to lower dementia risk, but only in observational studies. A 2023 meta-analysis of 48 long-term cohort studies (103,651 people) found dietary omega-3 linked to about 20% lower risk of dementia or cognitive decline (RR 0.82), with the strongest link for DHA intake [8]. In the Alzheimer's Disease Neuroimaging Initiative cohort, long-term omega-3 supplement users had a lower Alzheimer's risk (HR 0.36, 95% CI 0.18 to 0.72) [8]. A 2022 review found Alzheimer's risk about 30% lower with up to two portions of fish a week (95% CI 0.54 to 0.89), but the all-cause dementia result was not statistically significant (95% CI 0.79 to 1.02) [9]. Cohort studies cannot show cause and effect.

Supplement trials have not shown that omega-3 prevents dementia. In MAPT, 1,525 older adults with memory complaints took omega-3 for three years, and showed no difference in cognitive decline from placebo [10].

Thinking skills show a mixed picture. A 2024 meta-analysis of 24 trials (9,660 people aged 40 and older without dementia) found a possible gain in executive function, the skills used for planning and switching tasks, with an upward trend over the first 12 months [11]. The gain was most evident above 500 mg a day of omega-3 and in people whose DHA plus EPA levels were not very low [11]. A 2024 meta-analysis of 12 studies (1,124 people) with mild cognitive impairment found a gain in global cognition (SMD 0.51), though the authors say the evidence is thin [12]. A 12-month trial of 259 older adults with memory complaints, which combined DHA-rich fish oil with cocoa flavanols, found no cognitive gain, and executive function declined in the omega-3 group [13].

In children with ADHD, a 2018 meta-analysis of 7 trials (534 youths) found benefit for ADHD symptoms (g = 0.38) [14]. A 2021 meta-analysis of 31 trials (1,755 patients) found a small parent-rated effect (SMD -0.17) but no significant teacher-rated effect (SMD -0.06), and concluded there was no clear benefit [15]. A 2026 analysis of 7 trials found benefit only among youths with low baseline omega-3 levels (SMD 0.52, 4 trials), which its authors call preliminary [16].

Eating fish is linked to lower coronary heart disease incidence in cohort studies, but cohorts cannot show cause [17].

Pooled trials of supplements show little or no effect on major cardiovascular events (risk ratio 0.96, 95% CI 0.92 to 1.01) or on all-cause death [1]. Earlier suggestions of benefit appear to come from trials at higher risk of bias [18].

Prescription EPA trials differ. In REDUCE-IT, statin-treated adults with high triglycerides had fewer composite heart events (hazard ratio 0.75) [2]. A 2024 meta-analysis found lower heart attack risk (risk ratio 0.89) [19]. In a 2026 trial in people on dialysis, serious cardiovascular events were lower (hazard ratio 0.57) [20].

Eyes. In AREDS2, adding DHA and EPA (350 mg and 650 mg a day) to eye-health supplements did not slow progression to advanced macular degeneration over a median of 5 years, and the null result held at 10 years [21] [22]. For dry eye, reviews disagree: a meta-analysis of 17 trials (3,363 people) found omega-3 improved symptoms and signs [23], while Cochrane's review of 34 trials found little to no symptom relief and called the evidence uncertain [24]. A 2026 review of 27 trials found benefits only for dry eye linked to screen use or contact lenses, not for meibomian-gland-related or unspecified dry eye [25], and in computer users omega-3 improved dry-eye symptoms over 45 days to 3 months in two small trials [26].

Doses used in studies

Cochrane trials ranged from 0.5 g a day to more than 5 g a day of long-chain omega-3, lasting 12 to 88 months [1]. VITAL tested 1 g a day of marine omega-3 [27]; its atrial fibrillation substudy tested 460 mg of EPA and 380 mg of DHA a day [28]. REDUCE-IT gave 4 g a day of icosapent ethyl [2], and the dialysis trial gave 4 g a day of fish oil (1.6 g EPA and 0.8 g DHA) [20].

In depression studies, the effective EPA formulations were at most 1 g a day [3], and the 2023 review found benefit at EPA doses of 1 to under 2 g a day [4]. In VITAL-DEP, participants took 1 g a day of fish oil (465 mg EPA and 375 mg DHA) [7]. A 2026 trial in children with depression gave 1.5 g a day (1 g EPA and 0.5 g DHA) [29].

In thinking-skills studies, MAPT gave 800 mg DHA and 225 mg EPA a day [10], and CANN gave 1.1 g DHA and 0.4 g EPA a day with cocoa flavanols [13]. The 2024 meta-analysis found the executive function gain above 500 mg a day of omega-3, with up to 420 mg of EPA [11].

These are study doses, not recommendations.

Eye-health trials: AREDS2 gave 350 mg of DHA and 650 mg of EPA a day [21].

Safety and interactions

Common side effects: omega-3 supplements did not increase gastrointestinal problems overall [30]. Bleeding: serious bleeding was 2.7% versus 2.1% in REDUCE-IT, not statistically significant (P=0.06) [2], and a 2024 meta-analysis describes prescription EPA ethyl ester as carrying a high bleeding risk, though its pooled bleeding result was not statistically significant [30]. Heart rhythm: a 2021 meta-analysis found atrial fibrillation risk rose (hazard ratio 1.25), and more above 1 g a day [31].

Upper limits and toxic doses: some trials in the 2020 Cochrane review gave more than 5 g a day of long-chain omega-3 [1]; no safe upper limit was verified for this revision. Medicine interactions: the trials read did not report interactions with blood-thinning medicines, so people who take them should ask a doctor first. Pregnancy and breastfeeding: the studies read for this revision did not report these groups. Kidney disease: in people on dialysis, adverse events did not differ meaningfully between groups [20].

Mood and brain studies: in adults with depression, adverse events were similar or slightly more common with omega-3 than with placebo (OR 1.27, 95% CI 0.99 to 1.64; very low certainty) [6]. In the 2026 trial of 257 children with depression, serious adverse events numbered 45 in the omega-3 group and 31 on placebo, including 28 suicide attempts across both groups; none were judged caused by the study product [29].

Caveats and open questions

Ordinary supplements and high-dose prescription EPA give different results, and REDUCE-IT was funded by Amarin Pharma [2]. Whether the benefit and the rhythm risk depend on dose and form is not settled [31][1]. RESPECT-EPA was open-label and did not reach statistical significance on its primary end point [32]. The dialysis trial is a single study and needs replication [20].

Brain results are mixed. Large trials in older adults have not shown that omega-3 slows memory decline [10], and the 12-month CANN trial found no cognitive gain [13]. The depression evidence is low certainty overall [6]. In children, a 2024 Cochrane review of five small trials is very uncertain about depression symptoms [33], and the 2026 trial found no benefit over placebo [29]. The ADHD meta-analyses disagree [14][15]. Dementia links come from observational studies, which cannot show cause [8]. The dry-eye evidence is inconsistent, and more than half of the trials in the Cochrane review had a high risk of bias in at least one area [24] [23].