The advice was sensible at the time. Mid-century epidemiology had linked dietary saturated fat to elevated serum cholesterol, and serum cholesterol to coronary heart disease. The straightforward intervention: substitute saturated fats with polyunsaturated ones. Cholesterol drops, heart disease drops. That logic underwrote the American Heart Association’s dietary guidance for the next half-century and reshaped the global food supply — replacing animal fats and tropical oils with soybean, corn, canola, and sunflower oils on a planetary scale.
The mid-century trials that anchored this recommendation, however, mostly stayed unpublished. The 1968–1973 Minnesota Coronary Experiment was one of them — a randomized, double-blind, controlled trial of 9,423 institutionalized patients, comparing a saturated-fat diet against one in which corn oil substituted for animal fat. It was, by some distance, the largest such trial ever conducted. The serum-cholesterol results came out as expected. The clinical outcomes — mortality, autopsy-confirmed atherosclerosis, myocardial infarction — never published.
In 2016, a team led by NIH researcher Christopher Ramsden recovered the original tapes and re-ran the analysis.1 The findings, in The BMJ:
Serum cholesterol fell as predicted. Mortality went the wrong way. Each 30 mg/dL drop in serum cholesterol was associated with a 22% higher risk of death in the corn-oil group. No reduction in atherosclerosis at autopsy. No reduction in myocardial infarction. A pooled meta-analysis of five RCTs (n=10,808) added to the recovery paper showed no mortality benefit from PUFA substitution.1
The Sydney file.
Three years earlier, Ramsden had run a similar recovery on the Sydney Diet Heart Study — a 1966–1973 RCT of 458 men with prior coronary events.2 Half received safflower-oil-based margarine and shortening (high in linoleic acid); half received their habitual diet. The original investigators reported no significant differences. Ramsden’s recovery, published in The BMJ in 2013, found something else:
All-cause mortality was 62% higher in the linoleic-acid arm. CHD mortality was 74% higher. The original trial was small (n=458, men with prior CHD events) and the findings approached but did not always reach statistical significance — but the consistent direction across endpoints, in a trial that had been cited for decades as null-result evidence, mattered.
A separate Ramsden meta-analysis the same year, in the British Journal of Nutrition, pulled apart studies that had pooled n-3 and n-6 polyunsaturated interventions: n-6-specific PUFA replacement raised CHD events by 13% across the trial set, while mixed n-3/n-6 replacement lowered them by 22%.3 The historical “PUFA is protective” meta-analyses had conflated the two.
The mechanism.
DiNicolantonio & O’Keefe, writing in Open Heart in 2018,4 formalized the oxidized-linoleic-acid hypothesis. Linoleic acid (C18:2, two double bonds, polyunsaturated) is structurally vulnerable to oxidation. Under sustained heat (frying) or once incorporated into LDL particles in the bloodstream, it produces OXLAMs — oxidized linoleic acid metabolites including HNE, 9-HODE, and 13-HODE. These are documented drivers of foam-cell formation, atherosclerotic plaque, and inflammatory signaling.
The same paper documents that U.S. adipose-tissue linoleic acid concentration rose from 9.1% to 21.5% between 1959 and 2008 — a roughly 2.5-fold increase, tracking with the substitution of seed oils for animal fats in the food supply. The mechanism is not in serious dispute. The dose-response and population-level magnitude of the harm is.
What this actually tells you.
The Ramsden recovery papers do not prove that linoleic acid kills people. They do show, with peer-reviewed published data, that the strongest historical RCT evidence for the “replace saturated fat with PUFA” recommendation does not support the recommendation. The serum-cholesterol intermediate behaves as predicted. The clinical outcomes don’t follow.
For a procurement team specifying industrial cooking oil at scale, that’s a defensible position to anchor to. You don’t have to claim seed oils are uniquely harmful — you only have to claim that the trial evidence underwriting their default status is weaker than commonly assumed, and that a lower-PUFA, higher-oleic alternative is a reasonable specification choice.
The mainstream guidance hasn’t moved.
The American Heart Association and most national bodies still recommend substituting saturated fat with polyunsaturated fat, citing Mensink et al. 2003 and other meta-analyses. The Ramsden recovery papers are persuasive but they have not overturned consensus guidelines. The mechanism (OXLAMs, oxidized LDL) is increasingly accepted; the population-scale magnitude of harm remains contested. TOI is not arguing seed oils are toxic. We’re arguing the procurement-grade case for a low-PUFA, high-oleic alternative is stronger than the published consensus suggests.
