Are Seed Oils Bad for You?

The “Hateful Eight” framing groups canola, corn, cottonseed, soybean, sunflower, safflower, grapeseed and rice bran oil into one suspect category. The argument begins with a true observation: vegetable-oil intake in the West rose steeply across the twentieth century, and so did obesity and metabolic disease. That is a correlation. Essentially everything about the food supply changed at the same time — portion sizes, refined carbohydrate, added sugar, snacking frequency and the ultra-processing of nearly everything. A trend line that two things share is a hypothesis, not a finding.

The 2025–2030 Dietary Guidelines for Americans, released in January 2026, keep the long-standing limit of less than 10% of daily calories from saturated fat — roughly 22 g on a 2,000-calorie diet — but their presentation of cooking fats lists butter and beef tallow alongside plant oils and no longer singles out seed oils as the preferred replacement for saturated fat.^[15] The American Heart Association responded on 7 January 2026 by urging the public to “limit high-fat animal products including red meat, butter, lard and tallow, which are linked to increased cardiovascular risk.”^[16] This is a disagreement about emphasis between serious bodies, and it is a reason to look at the evidence rather than at the messaging.

A seed oil is oil pressed or solvent-extracted from the seed of a plant rather than from its fruit. Olive and avocado oil are fruit oils, which is the only real sense in which they are a different category. Expeller or cold pressing uses mechanical force; industrial extraction may use hexane, followed by refining, bleaching and deodorising.

Soybean, corn, sunflower and safflower oils are dominated by linoleic acid, the main dietary omega-6. Canola is mostly monounsaturated and contains a meaningful amount of the omega-3 alpha-linolenic acid. “Seed oil” therefore lumps together oils with quite different fatty-acid profiles, which is itself a reason to be suspicious of blanket claims.

Hexane is used as an extraction solvent and almost all of it is removed during refining. The EU sets a maximum residue level of 1 mg/kg in oils, and EFSA — whose last full evaluation dates from 1994 — has acknowledged that residue and exposure data are scarce and outdated and has been mandated to reassess it.^[17] The current limits are very low and there is no evidence of harm at them, but “we should have better data than a 1994 assessment” is a fair criticism. Cold-pressed oils avoid the question entirely if it bothers you.

The mechanistic claim is reasonable on paper. Linoleic acid is the metabolic precursor of arachidonic acid, which is the substrate for pro-inflammatory eicosanoids; therefore more omega-6 in the diet should mean more arachidonic acid in tissue and more inflammation. It has been tested directly, repeatedly, and it does not hold.

A meta-analysis of 30 randomised controlled trials in 1,377 adults found that increasing dietary linoleic acid had no significant effect on any measured inflammatory marker: C-reactive protein SMD 0.09 (95% CI −0.05 to 0.24, 16 studies), interleukin-6 0.11 (−0.07 to 0.29), tumour necrosis factor −0.01 (−0.19 to 0.17), monocyte chemoattractant protein-1 0.14 (−0.33 to 0.60) and adiponectin 0.17 (−0.17 to 0.50). The authors' own caveat matters: subgroup and meta-regression analysis suggested that in subjects with a very large increase in linoleic acid intake, CRP might rise.^[8]

An earlier systematic review of 15 randomised trials in healthy people reached the same conclusion: none reported significant effects on CRP, fibrinogen, PAI-1, cytokines, soluble adhesion molecules or TNF-α.^[9] The step before inflammation also fails. A systematic review of human trials found that decreasing dietary linoleic acid by up to 90%, or increasing it up to six-fold, produced no significant correlated change in arachidonic acid in plasma, serum or red-cell phospholipids (p = 0.39 and p = 0.72 respectively) in adults eating Western-type diets.^[10] The conversion is tightly regulated; you do not load your membranes with arachidonic acid by cooking in sunflower oil.

The mechanism that the entire “seed oils are inflammatory” claim rests on has been measured in humans and does not behave as the claim requires. That is about as clean a refutation as nutrition science produces.

The largest and most recent observational evidence followed 221,054 US adults in the Nurses' Health Study, Nurses' Health Study II and Health Professionals Follow-up Study for up to 33 years, during which 50,932 people died. Highest versus lowest butter intake was associated with 15% higher total mortality (HR 1.15, 95% CI 1.08–1.22); highest versus lowest total plant-oil intake with 16% lower total mortality (HR 0.84, 0.79–0.90). Per 5 g/day, the hazard ratios were canola 0.85 (0.78–0.92), soybean 0.94 (0.91–0.96) and olive 0.92 (0.91–0.94). Substituting 10 g/day of butter with plant oils was associated with 17% lower total mortality (HR 0.83, 0.79–0.86) and 17% lower cancer mortality.^[1] This is observational evidence, vulnerable to residual confounding by overall diet quality.

Biomarker cohorts remove the food-frequency-questionnaire problem. A pooled individual-level analysis of 30 prospective studies from 13 countries, with 68,659 participants and 15,198 cardiovascular events, found higher circulating and tissue linoleic acid associated with lower risk: total CVD HR 0.93 (0.88–0.99), cardiovascular mortality 0.78 (0.70–0.85), and ischaemic stroke 0.88 (0.79–0.98). Arachidonic acid was not associated with higher risk either; comparing extreme quintiles, higher levels went with lower total CVD (0.92, 0.86–0.99).^[2] It is still association, not causation — but it runs opposite to the toxicity claim.

A meta-analysis of 8 trials, 13,614 participants and 1,042 coronary events found that replacing saturated fat with polyunsaturated fat reduced coronary heart disease events by 19% (RR 0.81, 95% CI 0.70–0.95, p=0.008), roughly 10% per 5% of energy (RR 0.90, 0.83–0.97). Average PUFA intake was 14.9% of energy in intervention groups versus 5.0% in controls.^[3] The 2020 Cochrane review of 15 trials in 56,675 participants found that reducing saturated fat cut combined cardiovascular events by 17% (RR 0.83, 0.70–0.98, moderate-quality evidence, NNT 56 over about four years in primary prevention) with no effect on all-cause mortality (RR 0.96, 0.90–1.03), and found replacement with polyunsaturated fat or carbohydrate about equally useful.^[4]

Now the other side, without softening it. A separate 2018 Cochrane review asked what happens when omega-6 specifically is increased. Across 19 trials in 6,461 people, low-quality evidence suggested little or no effect on all-cause mortality (RR 1.00, 0.88–1.12) or cardiovascular events (RR 0.97, 0.81–1.15), uncertainty about cardiovascular mortality, a possible reduction in myocardial infarction (RR 0.88, 0.76–1.02, about 53 people needing to increase intake to prevent one MI), and high-quality evidence that omega-6 lowers total cholesterol by about 0.33 mmol/L.^[5] The benefit of PUFA replacing saturated fat is better established than the benefit of omega-6 considered on its own.

Recovered data from the Sydney Diet Heart Study — 458 men with recent coronary events studied from 1966 to 1973 — found higher mortality in the group given safflower oil in place of saturated fat: all-cause mortality 17.6% versus 11.8% (HR 1.62, 1.00–2.64, p=0.05), cardiovascular mortality 1.70 (1.03–2.80), and coronary mortality 1.74 (1.04–2.92).^[6] Recovered data from the Minnesota Coronary Experiment — 9,423 participants from 1968 to 1973 — found serum cholesterol fell substantially more on the corn-oil diet (−13.8% versus −1.0%, p<0.001) with no mortality benefit. An updated meta-analysis of five such trials (n=10,808) showed no benefit on coronary heart disease mortality (1.13, 0.83–1.54) or all-cause mortality (1.07, 0.90–1.27).^[7]

The criticisms are legitimate: these were institutional populations, used trans-fat-containing margarines of the era, and the recovered data were analysed decades later. They should not be dismissed. The honest synthesis is that the trial evidence that specifically swapped in large amounts of omega-6 oil, in populations of the 1960s and 70s, is genuinely mixed; the modern observational and biomarker evidence is consistently favourable; nothing in either body of work supports “toxic.”

Most seed-oil intake in a Western diet is not a spoonful of canola in a home pan; it is the frying medium and fat phase of ultra-processed food. A meta-analysis of observational studies found that comparing highest with lowest fried-food intake was associated with a 28% higher risk of major cardiovascular events (RR 1.28, 95% CI 1.15–1.43, 17 studies), 22% higher coronary heart disease risk (1.22, 1.07–1.40), and 37% higher heart failure risk (1.37, 1.07–1.75). There was notably no significant association with cardiovascular mortality (1.02, 0.93–1.14) or all-cause mortality (1.03, 0.96–1.12), and heterogeneity was high.^[11]

A 2025 analysis of two Chinese cohorts — 6,637 and 3,466 participants, with sibling comparison to control for family confounding — found a fried-food-related gut microbiota index associated with higher BMI, more central fat, and higher incidence of diabetes (HR 1.16, 1.07–1.27) and major adverse cardiovascular events (1.16, 1.06–1.26).^[12]

Prolonged and repeated heating degrades oil into polar compounds, polymers, free fatty acids and aldehydes. That is a recognised food-science problem, not a fringe claim.^[14] In animal work, feeding repeatedly heated vegetable oil raises blood pressure and produces measurable vascular damage — intima-media thickening, raised thiobarbituric acid reactive substances, raised thromboxane and angiotensin-converting enzyme — with effects scaling with the number of heatings.^[13] These are rodent studies at exposures well above normal home cooking, and there is no randomised human trial of reheated frying oil. But fresh oil in your pan and the oil in a commercial fryer on its fourth day are not the same exposure, and both camps routinely treat them as if they were.

Extra-virgin olive oil is the default for most cooking, with the strongest observational mortality data behind it. Canola or another neutral plant oil is reasonable for higher-heat cooking where olive oil's flavour or smoke point does not suit. Butter and tallow are fine in the amounts people actually use them for flavour; the evidence argues against making them the bulk of your fat intake.

The decision that matters far more than the brand of oil is how much of your food is deep-fried or ultra-processed. Peak Nutrition lets you see what your actual fat intake looks like rather than guessing. And when omega-3 intake comes up, start with what EPA, DHA and DPA actually do, because absolute omega-3 intake matters more than chasing a ratio.