Seed oils: claims vs evidence

What it covered: 15 RCTs, over 56,000 participants. Reducing saturated fat and replacing with polyunsaturated fat.

Findings: Approximately 17% reduction in cardiovascular events with PUFA replacement. No significant effect on total mortality.

Limitations: Heterogeneity across trials; most studies from the 1960s–1980s; dietary assessment quality variable. The cardiovascular event reduction was statistically significant but moderate in absolute terms, and in low-risk populations the absolute risk difference is likely small. The presence of two discordant trials (Sydney, Minnesota) means the direction of benefit, while generally consistent, is not unequivocally established across all populations and doses.

What it covered: Synthesis of RCT and observational evidence on dietary fat and cardiovascular disease.

Findings: Replacing saturated fat with polyunsaturated vegetable oils estimated to reduce cardiovascular disease by approximately 30%.

Limitations: Advisory rather than systematic review; produced by a body with institutional positions on the topic. The estimate synthesises heterogeneous evidence and should not be read as a precision figure.

What it covered: Men with recent coronary disease randomised to replace saturated fat with safflower oil (high LA). Recovered and re-analysed decades after the original trial.

Findings: Increased cardiovascular and all-cause mortality in the safflower oil group despite LDL reduction.

Limitations: Small sample (n=458); high-dose safflower oil not typical of normal consumption; incomplete randomisation data; participants were already in a high-risk clinical population. The harm signal is real but generalising to typical dietary use requires caution.

Possible explanations (unproven): Very high-dose safflower oil may generate oxidised lipid products differently to moderate habitual use; nutrient displacement effects from replacing a large proportion of dietary fat; the post-MI population may respond differently to very high LA exposure than healthier adults.

What it covered: Institutionalised patients and nursing home residents (n=9,423) randomised to corn oil margarine replacing saturated fat. Data recovered and re-analysed from a trial conducted 1968–73.

Findings: LDL was reduced in the intervention group, but mortality was higher. Each 30 mg/dL reduction in cholesterol was associated with higher, not lower, mortality.

Limitations: Institutionalised population not representative of general adults; high corn oil dose; margarine used contained trans fats in some periods; confounding cannot be fully excluded from recovered data. The dissociation between LDL reduction and mortality outcome is a genuine finding that highlights LDL as an insufficient surrogate for clinical benefit in all contexts.

Possible explanations (unproven): Trans fats in the margarine used during some periods are now established cardiovascular risk factors and could account for part of the harm signal; the institutionalised population had high baseline morbidity; the very high-dose corn oil intervention may not be comparable to moderate habitual seed oil use in a healthier population.

What it covered: Pooled analysis of 30 prospective cohort studies using blood biomarkers of linoleic acid intake as objective exposure measures.

Findings: Higher LA biomarker levels associated with lower cardiovascular disease incidence and mortality. Also associated with lower inflammatory markers — the opposite of the inflammation hypothesis.

Limitations: Observational design; biomarker levels reflect habitual intake but confounding with overall dietary pattern cannot be excluded.

Most trials and observational studies examine populations with moderate LA intake. The long-term clinical effects of intake at the very high end of the distribution — as in ultra-processed food heavy diets — are not well characterised. Extrapolating from moderate-intake studies to very high intake is not well-supported.

The typical pattern in ultra-processed food heavy diets is high seed oil intake combined with very low oily fish or omega-3 intake. This combined exposure has not been studied with adequate power as a distinct dietary pattern. The interaction between high LA and low EPA/DHA on inflammatory pathways may behave differently to either exposure in isolation.

Laboratory studies establish that aldehydes are generated during heating of seed oils, but the absorbed dose from normally cooked food under domestic conditions has not been adequately characterised. Long-term human trials comparing health outcomes from different cooking fats under controlled conditions do not exist.

Sunflower, soybean, rapeseed, corn, and safflower oils differ considerably in fatty acid composition, minor lipid constituents, and refining methods. The evidence base rarely disaggregates these differences. Treating them as a single category for the purposes of evidence review is a simplification that may obscure meaningful heterogeneity.

High seed oil intake at the upper end of the distribution is strongly correlated with ultra-processed food consumption. People eating the most seed oils are largely consuming them via processed snacks, fast food, and ready meals rather than as cooking oils added to whole foods. Health associations in these populations may partly reflect the broader dietary pattern rather than the oils themselves. This confounding is rarely disaggregated in the literature and means observational findings at high intake levels should be interpreted with particular caution.

The Sydney Diet Heart Study and Minnesota Coronary Experiment found unexpected harm signals that have not been adequately explained by the mainstream consensus. Their methodological limitations reduce their weight but do not eliminate it. The Minnesota result in particular highlights that LDL reduction is not a sufficient surrogate for clinical benefit in all contexts. Plausible explanations include trans fat contamination, very high-dose exposure, and population-specific effects — but none has been proven.