Fish and Seafood

People who eat fish about twice a week have less heart disease and stroke and somewhat lower death rates, and oily fish carries most of that link. The evidence is observational, the only real food trials were small and pointed in opposite directions, and frying seems to cancel the benefit. Mercury and other pollutants are a reason to choose your species, not to skip fish.

Eat fish about twice a week, roughly 225–300 g in total, and make most of it oily fish such as sardines, mackerel, herring, salmon or trout. Bake, grill, poach or steam it rather than fry it, and choose low-mercury species; if you are or might become pregnant, are breastfeeding, or are feeding children, avoid the seven highest-mercury fish completely.

What eating fish is linked to

Fish eaters have less cardiovascular disease and die somewhat less often. An umbrella review gathered 34 meta-analyses built on 298 cohort studies and rated the evidence moderate in quality. For each extra 100 g of fish a day, it found about 8% lower death rates from all causes, 25% lower cardiovascular death rates, and 12–20% less coronary heart disease, stroke, heart failure and depression. For most cancers there was no clear link.^[1] A hundred grams a day is a lot of fish, about 700 g a week, so these are slopes across a range of intakes, not a dose to aim for.

At ordinary intakes the numbers look similar. An earlier pooled analysis found that one to two servings a week went with 36% fewer coronary deaths and 17% lower overall death rates. The uncertainty range around such an estimate (the 95% confidence interval) is where the true effect most plausibly lies; for overall deaths it ran from 0% to 32%, so that half of the result only just clears no effect.^[2] For stroke, across 26 cohort studies, two to four servings a week went with about 6% less stroke than one or fewer, and five or more with about 12% less.^[3]

All of this is observational, and people who eat fish tend to differ from people who don't in income, diet and habits. Two findings show how fragile it is:

The benefit may depend on who you are. A pooled analysis covered about 190,000 people from 58 countries. Among the roughly 140,000 without heart disease, from 21 countries, eating 350 g of fish a week or more had no link with heart events or deaths (hazard ratio 0.95 for major cardiovascular events and 0.96 for deaths, both compatible with no effect). Among people who already had vascular disease, at least 175 g a week went with about 16% fewer major cardiovascular events and 18% fewer deaths.^[4]
And on where you live. In a dose-response analysis of 14 cohorts and about 911,000 people, each extra 20 g of fish a day went with 4% lower cardiovascular death rates. The overall-mortality link was only marginal, and both were significant only in Asian studies. In Western studies the curve was roughly U-shaped, lowest at about 20 g a day.^[5]

The only food trials: one success, one puzzle

Two randomised trials have tested advice to eat fish, both in men with heart disease. In the first, 2,033 men who had survived a heart attack were or were not told to eat fatty fish. Those advised to eat it had 29% lower death rates over two years, though repeat heart attacks plus coronary deaths did not differ significantly. The authors concluded that two or three portions a week may reduce mortality after a heart attack.^[6]

The follow-up did not repeat it. In 3,114 men with angina, advice to eat two portions of oily fish a week, or to take fish-oil capsules, went with 26% more cardiac deaths (hazard ratio 1.26, range 1.00–1.58), and more sudden cardiac deaths. Most of the excess was in the men given capsules, and the authors could not explain it.^[7] The large capsule trials since then are covered under Omega-3.

How much: about two servings a week

Guidance converges on about two servings a week. In heart patients risk stopped falling beyond about 175 g a week, though some cohorts show a little more benefit at higher intakes:

In people with vascular disease, the lowest risk came at 175 g a week, about two servings, with no further drop at 350 g or more.^[8]
The American Heart Association advises one to two seafood meals a week, especially in place of less healthy foods.^[9]
US dietary guidance, as quoted by the Food and Drug Administration (FDA), is at least 8 ounces (about 225 g) of seafood a week for adults, and 8–12 ounces (about 225–340 g) in pregnancy and breastfeeding.^[10]

The 2006 pooled analysis estimated that about 250 mg a day of the two long-chain omega-3 fats in fish, EPA and DHA, is enough for prevention.^[11] That is about 1.75 g a week, which two 100-g servings of canned sardines or mackerel roughly cover (see the table below).

Oily fish carries most of the signal

When cohorts split their fish by type, the link sits with oily fish. A meta-analysis of 19 reports found that the highest intakes of fatty fish, compared with the lowest, went with 8% less coronary heart disease, 17% fewer coronary deaths and 3% lower death rates overall; lean fish showed none of these links. Neither type was linked with cardiovascular disease as a whole.^[12] In the one cohort of heart patients that recorded fish types, fish rich in omega-3 fats went with less cardiovascular disease and other fish were neutral.^[13]

"Oily" means sardines, anchovies, herring, mackerel, salmon and trout. Cod, haddock, pollock, tilapia and most shellfish are lean. Lean fish is still a good protein and a source of selenium and iodine (see the canned-fish section below), and swapping it for processed meat is covered under Red and processed meat.

Farmed salmon is less oily than it was. The fish meal and fish oil in salmon feed have been partly replaced with land-based ingredients that lack EPA and DHA. In more than 3,000 farmed Scottish salmon from 2006 to 2015, EPA and DHA fell, so that by 2015 it took twice the portion to reach recommended intakes, though farmed salmon still supplied more than most other fish.^[14] A small sample of fillets bought in shops around Washington, DC, in 2023 held about 64% less EPA and 48% less DHA than the older food-table values.^[15] Don't let farmed salmon be your only oily fish.

The brain: consistent links, no trials of fish

Fish eaters also have less dementia in cohort studies. Across 21 studies and about 180,000 people, each extra serving of fish a week went with about 5% less dementia and 7% less Alzheimer's disease.^[16] A 2024 meta-analysis of 35 studies found the highest fish eaters had about 18% less dementia and 20% less Alzheimer's than the lowest. Its authors judged the evidence for dementia in need of further studies, and one of them works for a food company, Bolton Food.^[17]

No trial has tested eating fish against dementia. Fish is one of the components of the MIND diet, whose one randomised trial found no benefit over people's usual diet.

Frying seems to cancel the benefit

How the fish is cooked changes the picture. In about 84,500 postmenopausal women in the US, baked or grilled fish five or more times a week went with 30% less heart failure, while fried fish once a week or more went with 48% more.^[18] In nearly 107,000 women from the same study, followed until 2017, a weekly serving or more of fried fish or shellfish went with 7% higher death rates and 13% higher cardiovascular death rates than none.^[19]

These findings are observational and in women only, but they point the same way, and researchers puzzling over why fish looks better in Asian than in Western cohorts have flagged cooking methods as one thing to examine.^[20] Fried fish probably does not count toward the target.

Mercury: choose the species, don't skip the fish

Fish carry methylmercury, a nerve toxin that matters most for the developing brain of a fetus or young child. It builds up the food chain, so large, long-lived predators hold much more of it than small fish. In the FDA's testing, average levels were:^[21]

Fish	Average mercury (mg/kg)
Sardine, tilapia	0.013
Canned salmon / fresh or frozen salmon	0.014 / 0.022
Atlantic mackerel	0.05
Herring	0.078
Cod	0.111
Canned light tuna	0.126
Canned albacore ("white") tuna	0.350
Bigeye tuna, king mackerel	0.689, 0.73
Shark, swordfish	0.979, 0.995
Tilefish (Gulf of Mexico)	1.123

A swordfish steak carries about 75 times the mercury of the same weight of sardines.

The highest-mercury seven. The FDA and the Environmental Protection Agency (EPA) tell people who are or might become pregnant, people who are breastfeeding, and children aged 1–11 to avoid king mackerel, marlin, orange roughy, shark, swordfish, tilefish from the Gulf of Mexico, and bigeye tuna. Pregnant and breastfeeding people should eat two to three servings a week (a serving is 4 ounces, about 113 g) from the "best choices" list, which includes salmon, sardines, anchovies, herring, Atlantic mackerel, trout, cod, pollock, shrimp and canned light tuna, or one serving from the "good choices" list, which includes albacore tuna, yellowfin tuna, halibut and snapper. Children should have two servings a week from the best choices; a child's serving is about 28 g at ages 1–3, rising to 113 g by age 11. For fish you or friends catch locally, check local advisories; where there are none, the agencies advise one serving and no other fish that week.^[22]

In pregnancy, eating too little fish looks worse than the mercury. In about 11,900 pregnancies in a British birth cohort, children of mothers who ate no seafood had about one and a half times the odds of scoring in the lowest quarter for verbal IQ compared with children of mothers who ate more than 340 g a week, the limit US advice then set for pregnancy.^[23] Two systematic reviews using the US dietary-guidelines method, covering about 106,000 mother-child pairs, linked seafood in pregnancy with better neurodevelopment from about 4 ounces (113 g) a week upward, and found no net harm at the highest intakes despite more mercury. They rated the evidence moderate because there are few trials, and one author runs an omega-3 testing company.^[24]

For other adults, mercury at ordinary intakes shows no heart harm. In two US cohorts totalling about 173,000 adults, comparing the 3,427 who developed heart disease or stroke with matched controls, people with the highest toenail mercury had no higher risk than those with the lowest.^[25] The sensible limit for adults is at high intake: if you eat fish five or more times a week, keep the highest-mercury species to a minimum.^[26] The wider picture on metals in food is under Environmental toxins.

PCBs, dioxins, PFAS, and farmed versus wild

Oily fish also carries fat-soluble industrial pollutants: polychlorinated biphenyls (PCBs) and dioxins, which stay in the body for years.

Farmed versus wild. A 2004 analysis of over two tonnes of salmon found organochlorine pollutants significantly higher in farmed than in wild salmon, and higher in European than in North and South American farmed fish.^[27] Levels have since fallen. In more than 2,300 Norwegian farmed salmon tested from 1999 to 2011, mercury, dioxins, dioxin-like PCBs and the pesticide DDT all decreased.^[28] Leaving other foods aside, that study put the tolerable limit at up to 1.3 kg of salmon a week, but it was calculated against the older European tolerable intake; scaled to the new limit described next, that would be under 200 g of salmon a week at 2011 levels, before any other food. In 2018 Europe's food-safety agency (EFSA) cut that limit for dioxins and dioxin-like PCBs sevenfold, from 14 to 2 picograms per kilogram of body weight a week. It found that average intakes across Europe already exceed the new limit, from all foods, with fatty fish among the main sources — up to 56% of exposure in some surveys of children, adolescents and adults.^[29]

PFAS. Fish is also a source of PFAS, the "forever chemicals". In EFSA's 2020 assessment, fish meat was one of the top contributors to European dietary exposure, together with fruit and eggs, and part of the population exceeds the safe weekly intake.^[30] In US national survey data, shellfish was the strongest dietary predictor of higher blood PFAS, while fish showed weak or inconsistent links.^[31] Shellfish is also one of the richest food sources of cadmium, another reason not to make it your main seafood.

How to weigh this against the benefits is unsettled. A Norwegian analysis of herring, mackerel and farmed salmon found that at recommended intakes both the nutrient targets and the pollutant limit are reached, and called for a formal weighing of the two.^[32] The 2006 benefit-risk review judged that PCB and dioxin levels in fish were low and outweighed by the benefits, and that they should have little effect on fish choices.^[33] The practical response is to vary your species, keep to about two servings rather than many more, and not build your diet around one source.

Canned fish: the easy way to hit the target

Canned sardines, mackerel and salmon are oily, low in mercury and need no cooking. Eaten with their soft bones, they also supply calcium that fresh fillets do not. Per 100 g, from the US food-composition database (the farmed-salmon row is the older value that recent testing has revised down):^[34]

Per 100 g	EPA+DHA (g)	Vitamin D (µg)	Calcium (mg)	Selenium (µg)
Sardines, canned in oil, with bones	0.98	4.8	382	53
Pink salmon, canned, with bones	0.99	13.7	215	33
Jack mackerel, canned	1.23	7.3	241	38
Herring, cooked	2.01	5.4	74	47
Atlantic salmon, farmed, cooked (older value)	2.15	13.1	15	41
Albacore tuna, canned in water	0.86	2.0	14	66
Light tuna, canned in water	0.27	—	11	80
Cod, cooked	0.16	1.2	14	38

Three things stand out. Bones make the difference for calcium: canned sardines carry more than 25 times the calcium of cod. Canned light tuna is low in mercury but also low in omega-3 fats, so it is a protein food, not a substitute for oily fish. And lean fish is, with milk, the main dietary source of iodine in Norway: in Norwegian testing, iodine ranged from 18 µg per 100 g in Atlantic halibut to 1,210 µg in pollack, against 12–19 µg in cow's milk.^[35]

What fish doesn't do

It doesn't prevent type 2 diabetes. Across 28 cohorts and about 956,000 people, fish intake had no link with diabetes in men; in women each extra 100 g a week went with about 2% more diabetes, a pattern found in the Americas and not elsewhere.^[36]

It isn't a cancer food. Apart from a link with less liver cancer, fish showed no clear association with most cancers, and the evidence there was rated low or very low.^[37]

What to actually do

Eat fish twice a week, about 225–300 g in total, mostly oily: sardines, mackerel, herring, anchovies, trout or salmon.
Use canned fish to make that easy and cheap; eat sardines and canned salmon with their bones.
Don't fry it. Bake, grill, poach, steam or pan-cook in a little oil.
Rotate species rather than eating the same fish every time, and don't rely on farmed salmon alone for omega-3.
Keep the highest-mercury fish occasional, and if you eat fish five or more times a week, mostly choose the low-mercury ones in the table.
If you are or might become pregnant, or are breastfeeding, don't stop eating fish: eat two to three servings a week of best-choice fish and none of the seven highest-mercury species. Children aged 1–11 should get two child-sized servings a week of best-choice fish.
If you don't eat fish, see Omega-3 for whether a supplement makes sense.