380 ‒ The seed oil debate: are they uniquely harmful relative to other dietary fats?

KEYPOINTS

• 01

  This was meant to be a pre-registered 'courtroom' debate; when the opposing expert withdrew, Attia steelmanned the anti-seed-oil case himself.

  Over about a year, both sides were to pre-submit every paper so nothing could be asserted that couldn't be checked live. Attia stated his lean upfront: he doesn't see much here. Norton noted his own possible bias ran the other way—he trained in a low-carb-leaning lab and was funded by the National Dairy Council, Cattlemen's Beef Association and Egg Board.

• 02

  Attia framed four arguments that seed oils are uniquely harmful.

  (1) Some old RCTs that swapped saturated fat for polyunsaturated fat lowered cholesterol but not mortality. (2) Linoleic acid makes LDL more oxidisable and fuels arachidonic-acid inflammation. (3) Industrial refining—heat, hexane—adds harm. (4) Evolutionary mismatch: linoleic-acid availability rose from under 3% of food to roughly 10%, and tissue levels more than doubled.

• 03

  The 'harmful' RCTs are confounded by trans fat from margarine.

  In the Minnesota Coronary Experiment (1966–73, ~9,000 institutionalised patients), the polyunsaturated arm used corn oil and margarine. Margarine then contained about 25–40% trans fat, which is clearly atherogenic and now effectively banned. Each 30 mg/dL fall in total cholesterol came with roughly 22% higher mortality. Results waited about 13 years for publication. Patients could also check themselves out of wards, and average exposure was only 1–2 years.

• 04

  Saturated and trans fats pack straight, making stiffer particles; cis double bonds kink and loosen them.

  Trans bonds keep chains straight like saturated fat but still carry an oxidisable double bond—'the worst of both worlds.' Membrane fluidity matters downstream for LDL-receptor recognition and for how readily particles aggregate in the artery wall.

• 05

  When trans-fat-confounded trials are removed, swapping PUFA for saturated fat reduces risk.

  Lumping all trials together gives a null effect. Ramsden's reanalysis removing omega-3-confounded trials shows a trend to harm—but still includes trans fat. A 2017 analysis excluding trans-fat trials found roughly 29–31% lower risk. Norton insists on 'logical symmetry': if you remove one side's confounders, remove the other's.

• 06

  Mendelian randomisation shows lifelong LDL exposure is causal and dose-dependent.

  About a dozen variants affecting LDL clearance, production or absorption all give the same effect. Each 1 mmol/L (~39 mg/dL) lower LDL brings about 50–55% lower cardiovascular risk, with R above 0.9. That consistency makes pleiotropy unlikely. CETP variants underdeliver because cholesterol falls more than apoB. The same genetics showed LDL has no causal link to cancer.

• 07

  Statins give ~22% per mmol/L, not 50%, because of timing, not hidden harm.

  Statins usually start around 40 or later, after decades of exposure; genetic variants act from birth. Norton uses compounding-investment and 'truck rolling downhill' analogies. He runs LDL around 150 despite a high-fibre, low-saturated-fat diet, and started a statin at 40.

• 08

  Linoleic acid doesn't measurably raise arachidonic acid or population heart disease.

  Cohorts of roughly a million people, and tissue biomarker studies that avoid recall error, link higher linoleic acid to lower cardiovascular disease. Conversion to arachidonic acid seems saturated, so eating more doesn't feed forward. Healthy-user bias would blunt a harm signal, not flip it to protection.

• 09

  Oxidation mostly happens inside the artery wall, so fewer particles getting in beats slightly more oxidisable particles.

  Under 1% of plasma LDL is oxidised; particles clear in hours and plasma is rich in antioxidants. Once apoB-particles under ~70 nm cross the endothelium and bind proteoglycans, an estimated 30–80% may oxidise over weeks. PUFA-rich LDL is fewer in number, less prone to apoB modification, and less prone to aggregation than saturated-fat-rich LDL. Saturated-fat LDL is acted on by sphingomyelinase, making ceramides that drive clumping. Antioxidant trials failing fits this picture.

• 10

  Refining residues are far below harmful doses, and refining actually lowers oxidation products.

  Hexane boils at about 69 °C and is steam-stripped under vacuum. Residues run around 0.05–0.5 ppm, often undetectable. Norton calculates about 11,340 kg of oil at once for even mild hexane effects, and hexane doesn't bioaccumulate. Refining cuts peroxides 5–10-fold and lowers aldehydes. It adds about 0.5% trans fat and some polymerised triglycerides.

• 11

  The ancestral argument cuts both ways.

  Modern beef isn't ancestral either, and cardiovascular disease rose largely because people now live long enough to get it. The Hadza, our best ancestral proxy, have LDL around 50–70 mg/dL, with only one individual found above 100. 'Natural' isn't a reliable guide to longevity.