Substance + claimed effects

Industrially refined seed oils — soybean, sunflower, safflower, corn, canola (rapeseed), cottonseed, grapeseed, rice bran — are the dominant culinary fats of the modern Western food supply. They are produced by mechanical pressing followed by hexane solvent extraction, then degummed, alkali-neutralised, bleached with activated clay, and steam-deodorised at 240–260 °C: the RBD (refined, bleached, deodorised) process. The product is a bland, light, high-smoke-point liquid dominated by polyunsaturated fatty acids (PUFAs), particularly omega-6 linoleic acid (LA). Soybean oil supply in the US rose more than a thousandfold from 1909 to 1999 Blasbalg et al. 2011; subcutaneous adipose linoleic acid in US adults rose ~136% over the same window Guyenet & Carlson 2015. The claims in scope: (1) seed oils, via linoleic acid substitution for saturated fat, lower LDL-C and reduce cardiovascular events — the mainstream view enshrined by the AHA Sacks et al. 2017; (2) the substitution is harm-neutral or harmful — the contrarian view drawing on the Sydney and Minnesota trial reanalyses Ramsden et al. 2013 Ramsden et al. 2016; (3) heated seed oils generate aldehyde lipid-peroxidation products (4-HNE, MDA) that are absorbed from fried food and may drive inflammation independent of the LDL story Grootveld et al. 2020. The entry covers these consequences holistically — cardiovascular markers, longevity signal, oxidation-product exposure, the omega-6/omega-3 imbalance, and the active scientific controversy.

Evidence by addressing question

mechanism

Linoleic acid (18:2 n-6) is an 18-carbon PUFA with two double bonds; the methylene-interrupted diene structure makes the bis-allylic C—H bond weak (~50 kcal/mol vs ~75 kcal/mol for oleic acid, ~100 kcal/mol for stearic) and the molecule kinetically prone to free-radical hydrogen abstraction Grootveld et al. 2020. Two mechanism stories sit on top of this chemistry. (a) The lipid hypothesis pathway. Replacing dietary saturated fat with linoleic acid shifts hepatic fatty-acid availability, downregulates intestinal cholesterol absorption, and lowers serum LDL-C by ~10–15% on isoenergetic substitution; meta-regression puts the LDL response at roughly -1.5 mg/dL per 1% energy SFA→PUFA swap Mensink 2016 (WHO). The AHA presidential advisory Sacks et al. 2017 argues this LDL drop translates into ~30% CHD reduction, supported by the historical PUFA-substitution RCTs and primate atherosclerosis-regression data. (b) The oxidised-LA / membrane-instability pathway. Linoleic acid incorporates into membrane phospholipids and plasma cholesterol esters in proportion to dietary intake. In vivo peroxidation of membrane LA generates oxidised linoleic acid metabolites (OXLAMs: 9- and 13-HODE, 9- and 13-oxoODE) and the toxic α,β-unsaturated aldehyde 4-hydroxy-2-nonenal (4-HNE) — protein-crosslinking electrophiles found at elevated concentrations in oxidised LDL and atherosclerotic plaque DiNicolantonio & O'Keefe 2018. The competing mechanism story is that LDL-C may fall while atherogenic oxidised-LDL particles and membrane vulnerability rise — a plausible reason serum-cholesterol benefit fails to translate into mortality in the recovered-data trials.

The omega-6/omega-3 ratio is a parallel mechanism. ALA, EPA, and DHA share the Δ6-desaturase and elongase enzymes with LA; high LA intake competitively suppresses conversion of α-linolenic acid to EPA/DHA and biases eicosanoid synthesis toward 2-series prostaglandins and 4-series leukotrienes from arachidonic acid — pro-inflammatory and pro-thrombotic Simopoulos 2016. The ancestral hunter-gatherer ω-6:ω-3 ratio is estimated at ~1:1; current Western ratios run 15:1 to 20:1, driven almost entirely by the seed-oil rise Simopoulos 2016.

evidence

The cardiovascular evidence base is large, contested, and partitioned by study type.

Randomised SFA→PUFA-substitution trials, original analyses. The Mozaffarian PLoS Medicine meta-analysis pooled eight RCTs (~13,600 participants) and reported a 10% CHD-event reduction per 5% energy SFA replaced by PUFA (RR 0.90, 95% CI 0.83–0.97) — the headline number underpinning AHA 2017 Mozaffarian et al. 2010. The Cochrane SFA-reduction review reports a 17% reduction in combined cardiovascular events on SFA reduction (RR 0.83, 95% CI 0.70–0.98, 13 RCTs, 53,758 participants, moderate-quality evidence), with the strongest signal in trials replacing SFA with PUFA. No mortality benefit was detected: all-cause mortality RR 0.96 (0.90–1.03), CVD mortality RR 0.95 (0.80–1.12) — moderate-quality Hooper et al. 2020.

Recovered-data reanalyses. Ramsden's two recovered-data papers are the contrarian camp's anchor. The Sydney Diet Heart Study (1966–73, n=458 men post-MI) replaced animal fat with safflower oil — a near-pure (~75%) linoleic-acid intervention. The intervention arm had higher all-cause mortality (17.6% vs 11.8%, HR 1.62, 95% CI 1.00–2.64), CVD mortality (HR 1.70, 1.03–2.80), and CHD mortality (HR 1.74, 1.04–2.92) Ramsden et al. 2013. The Minnesota Coronary Experiment (1968–73, n=9,423, the largest diet-heart RCT ever run) replaced SFA with corn oil and corn-oil margarine. Serum cholesterol fell ~14% in the intervention arm vs ~1% in controls; there was no mortality benefit, and post-hoc analysis suggested a 22% mortality increase per 30 mg/dL cholesterol reduction in participants over 65 Ramsden et al. 2016. Ramsden's incorporation of these recovered data into an updated meta-analysis pushed the pooled CHD-mortality estimate non-significantly upward (HR 1.33, 0.99–1.79). Both trials used near-pure LA (safflower, corn oil + margarine containing partially hydrogenated soybean — a trans-fat confound in MCE), did not mirror modern soybean/canola intake patterns, and were time-limited (3–5 years).

Prospective biomarker cohorts. Two large pooled individual-level analyses point the other way. The Marklund/FORCE consortium pooled 30 cohorts (n>68,000 with biomarkers) and found circulating linoleic acid associated with lower total CVD, ischemic stroke, and cardiovascular mortality (per interquintile range; hazard ratios in the 0.78–0.93 range) — arachidonic acid biomarker level was not associated with CVD Marklund et al. 2019. The Li et al. systematic review pooled 44 cohorts (~811,000 participants, ~170,000 all-cause deaths for dietary-intake analyses; 65,411 for biomarkers) and reported dietary LA (high vs low) associated with RR 0.87 for total mortality, 0.87 for CVD mortality, 0.89 for cancer mortality; biomarker analyses gave similar inverse associations Li et al. 2020. Confounding by overall diet quality (LA biomarker tracks plant-food-heavy diets that may also be lower in ultraprocessed food) is the main caveat the authors flag.

Dietary-pattern RCT. PREDIMED (n=7,447, median follow-up 4.8 y) compared a Mediterranean diet supplemented with extra-virgin olive oil (a monounsaturate-dominant oil, not a seed oil) or mixed nuts against a low-fat control. Major cardiovascular events were ~30% lower in the EVOO and nut arms Estruch et al. 2018. PREDIMED does not test seed oils directly; it is the strongest positive trial evidence for an oleic-acid-and-polyphenol-dominant fat pattern over a low-fat one.

oxidation products / mechanism (frying-related harm)

The bis-allylic structure that makes LA biologically essential also makes it thermally unstable. Heating linoleic-acid-rich oils above ~150 °C — well below the smoke point — initiates radical chain peroxidation; concentrations of 4-hydroxy-2-nonenal (HNE) and other α,β-unsaturated aldehydes rise measurably within hours at frying temperature. Soybean oil held at 185 °C generates appreciable HNE by 2 hours and continuing accumulation through 6 hours, with concomitant tocopherol depletion Grootveld et al. 2020. The aldehydes are absorbed by fried food: HNE has been quantified at microgram-per-serving levels in commercial fast-food fries cooked in restaurant soybean-oil fryers Csallany et al. 2015. Reused fryer oil concentrates aldehydes further. The biological case for harm is mechanistic — HNE forms Michael adducts with protein cysteines, histidines, and lysines, modifies LDL, and propagates membrane oxidation — but human RCTs of fried-food oxidation-product exposure are absent. This is the strongest contrarian mechanism case for seed-oil-specific harm distinct from the LA-substitution hypothesis: raw LA from whole foods (walnuts, sunflower seeds eaten unprocessed) may behave differently from LA delivered via deep-fried restaurant food DiNicolantonio & O'Keefe 2018.

protocol

The behavioural envelope readers can act on: (1) limit deep-fried food — the high-temperature, repeatedly heated, restaurant-fryer case is where the aldehyde-exposure argument is strongest and where everyone (mainstream and contrarian) agrees the product is worse than the oil it was made from; (2) for home cooking, prefer fats with thermal stability — extra-virgin olive oil (>70% monounsaturated, polyphenol-buffered), butter, ghee, coconut oil — for sautéing and roasting, reserving high-PUFA oils for cold use or skipping them; (3) shift the ω-6/ω-3 ratio toward 4:1 or lower via fatty fish 2× weekly, modest ALA (flax, walnuts), and reducing total seed-oil load rather than supplementing fish oil on top of high LA. Total LA intake target debated: contrarian advisories suggest the historical 1–2% energy floor; mainstream guidelines treat 5–10% energy LA as desirable. Most adults in the US currently sit at 7–8% energy LA, much of it from restaurant and ultraprocessed food rather than home cooking Blasbalg et al. 2011.

misconceptions

Three widely repeated framings deserve direct correction. (a) "Smoke point is the safety number." False — oxidation begins well below the smoke point. A high refined smoke point (~230 °C for canola, ~225 °C for sunflower) tells you when the oil visibly degrades, not when peroxidation products begin to form Grootveld et al. 2020. (b) "Seed oils are heart-healthy because they lower cholesterol." They reliably lower LDL-C; the mortality translation is what's contested. The recovered-data trials showed full LDL response with no mortality benefit, undercutting the LDL-as-sufficient-endpoint inference Ramsden et al. 2016. (c) "All seed oils are equivalent." The PUFA fraction varies sharply: safflower and grapeseed run 70–75% LA, soybean and corn 50–55%, sunflower ~65% (high-oleic varieties ~20%), canola ~20% LA + ~10% ALA + ~60% oleic. Canola sits closer to olive oil than to soybean oil on the PUFA spectrum despite being lumped with seed oils colloquially.

controversy / credibility

The active disagreement is real, not manufactured. Three camps:

• Mainstream cardiology / nutrition (AHA, AND, WHO): SFA→PUFA substitution remains a Class I recommendation; LA is treated as cardioprotective; the recovered-data trials are flagged as methodologically problematic (high attrition in MCE, trans-fat confound from the margarine, near-pure LA arms not representative of mixed PUFA intake) Sacks et al. 2017.
• Cautious revisionists (publishing in BMJ, Open Heart, the Hooper team's later analyses): SFA reduction lowers CVD events but not mortality; PUFA replacement is probably better than SFA but the case is weaker than guidelines imply; recovered-data trials warrant inclusion in meta-analyses Ramsden et al. 2016 Hooper et al. 2020.
• Contrarian / "seed oils are the driver" (DiNicolantonio, Shanahan, Knobbe, online ancestral-health community): Oxidised LA metabolites and frying-derived aldehydes drive atherosclerosis, NAFLD, and metabolic disease; the historical rise of soybean oil correlates with the obesity and CVD epidemics; LA below 2% energy is the target DiNicolantonio & O'Keefe 2018.

practicalities

Seed oils dominate the ultraprocessed-food, restaurant-fryer, and packaged-snack channels — minimising them is as much a food-choice question as a cooking-oil question. Cost: olive oil runs 5–10× the per-litre cost of soybean/canola; ghee and butter are cheap-to-moderate. Effort: home swaps are trivial once the pantry is converted; restaurant avoidance requires either asking what oil is used or accepting the soybean default. Label literacy helps — "vegetable oil" on ingredient lists is almost always soybean oil in the US, canola in Canada/EU.

history

Industrial seed oils are a 20th-century phenomenon. Cottonseed oil entered the US food supply as Crisco in 1911 (Procter & Gamble, marketed as a lard substitute). Soybean oil scaled with US soybean cultivation post-1940. The AHA's 1961 guidance to substitute polyunsaturated for saturated fat — drafted with input from Ancel Keys — institutionalised the substitution. The Sydney (1966–73) and Minnesota (1968–73) trials were run during this window and largely went unpublished or were published with incomplete mortality data until Ramsden's recovered-data work in 2013 and 2016 surfaced the negative results Ramsden et al. 2013 Ramsden et al. 2016. The contrarian historical frame: the dietary substitution preceded the obesity-CVD epidemic by ~30 years and tracks it closely; the mainstream rebuttal: so did sugar, refined grain, ultraprocessed food, sedentary lifestyle, and screen time.

stakes

For the typical Western adult, current exposure puts adipose LA roughly double the 1960s level Guyenet & Carlson 2015 — the tissue-incorporation half-life of LA is ~2 years, so the body composition reflects multi-year intake. The mainstream evidence base suggests this is approximately neutral-to-favourable for cardiovascular mortality. The contrarian case suggests the exposure pattern — restaurant-fried, repeatedly heated, polyphenol-stripped LA — is delivering an aldehyde load that is invisible to LDL-C tracking. The honest reader-facing forecast is uncertain: if mainstream is right, ignoring seed oils costs nothing; if contrarian is right, the cost is decades of low-grade vascular and metabolic damage that doesn't show on a standard lipid panel.

payoff

Replacing the deep-fried-restaurant channel with home cooking in olive oil or butter delivers an immediate, measurable change in dietary aldehyde exposure and a plausible (not proven) reduction in oxidative-stress markers. Mediterranean-pattern eating around EVOO and nuts has the strongest RCT evidence base for cardiovascular event reduction Estruch et al. 2018; what fraction of that effect comes from removing seed oils versus adding olive oil and whole plants is unidentifiable from the trial.

out-of-scope

Topical/dermatological use of seed oils (skincare formulations) is a different exposure pattern. Fish oil and omega-3 supplementation deserves its own entry; here LA and ω-3 enter as inputs to the ratio framing only. The metabolic-syndrome / NAFLD specific case for LA — Sanders' soybean-oil mouse models, the lipogenic-induction question — is large enough to warrant separate treatment. Industrial trans fats (partially hydrogenated soybean oil) are a separate substance; modern refined seed oils contain trace (~1%) trans isomers from deodorisation but are not the same product.

The credibility range

The optimist case (mainstream)

Linoleic acid is an essential fatty acid; the body cannot make it. Modern seed oils are a cheap, scalable, LDL-lowering substitute for animal fat. The largest pooled biomarker analyses — Marklund (n>68,000, 30 cohorts) and Li (n~811,000, 44 cohorts) — show inverse associations between LA biomarkers and cardiovascular mortality, all-cause mortality, and cancer mortality Marklund et al. 2019 Li et al. 2020. Cochrane's most recent meta-analysis finds a 17% reduction in cardiovascular events with SFA reduction, strongest with PUFA replacement Hooper et al. 2020. The mechanism — LDL-C reduction, anti-inflammatory eicosanoid metabolism of dietary LA — is consistent across primate and human work. The recovered-data trials had near-pure-LA arms not representative of population intake, used margarines now known to contain harmful trans fats, and showed wide confidence intervals; meta-analytic re-inclusion did not push effects to statistical significance. The historical-correlation argument is weak: many things changed alongside seed-oil consumption. Conclusion: seed oils are a net cardiovascular positive, the AHA recommendation stands, the moral panic is a reaction to ultraprocessed food being miscoded as a fat problem.

The skeptic case (contrarian)

The two RCTs with full data and pure LA arms — Sydney and Minnesota — both showed mortality going the wrong way after the cholesterol response was fully realised Ramsden et al. 2013 Ramsden et al. 2016. The decoupling of LDL drop from mortality outcome is the single most important fact in this debate, and it undermines the LDL-as-surrogate-endpoint inference at the core of AHA guidance. The biomarker cohorts conflate dietary pattern with LA exposure — people with high circulating LA are generally health-conscious, eating walnuts, fish, and vegetables, not deep-frying in soybean oil. The mechanism case for harm is concrete: linoleic acid's structure makes it peroxidisable; oxidised LA metabolites accumulate in atherosclerotic plaque; heated LA generates HNE that gets eaten from restaurant fries at microgram-per-serving doses Csallany et al. 2015 Grootveld et al. 2020. The ω-6/ω-3 ratio has risen 5–10× from ancestral baseline, and dietary patterns with high LA + low ω-3 show elevated inflammatory tone Simopoulos 2016. The historical correlation is hard to dismiss in full: the dietary fat composition of the US food supply changed dramatically; the chronic-disease profile changed dramatically; the two are at least worth disentangling rather than assumed coincident.

The author's call

Land cautiously on the revisionist middle. The mainstream case for population-level CVD-event reduction via SFA→PUFA substitution is real but smaller than guidelines imply, and the mortality benefit is genuinely unproven in the Cochrane data. The contrarian case for seed-oils-as-cause-of-modern-chronic-disease overreaches the evidence — the biomarker cohorts are too consistent in pointing the other direction to be wholly confounded — but the specific case against industrially heated, repeatedly used restaurant fryer oils is mechanistically strong and worth acting on. The honest reader call: home cooking with olive oil and butter is a defensible default; the LA fraction of unprocessed soybean/sunflower/canola in moderate home use is probably not the lever; the deep-fried restaurant and ultraprocessed-snack channels are the lever, and reducing them happens to reduce seed-oil exposure as a side effect. The aggressive "all seed oils are poison" framing is unsupported. The dismissive "the debate is settled, eat your canola" framing is also unsupported. Score the controversy high (4/5); score the evidence moderate (3/5) reflecting the genuine partitioning of the literature.

Stakeholder + incentive map

• Commercial pro-seed-oil: US soybean producers (USB), Canola Council, ADM, Cargill, Bunge — multibillion-dollar industries whose primary product is soy oil and meal. The American Soybean Association funds nutrition research and CME content.
• Professional pro-seed-oil: AHA, AND (Academy of Nutrition and Dietetics), USDA Dietary Guidelines committee — institutional positions long predate the recovered-data trials and have inertia. Sacks/Lichtenstein co-authored both the 2017 advisory and editorials critical of the Ramsden reanalyses Sacks et al. 2017.
• Commercial contrarian: grass-fed meat, butter/ghee, tallow, coconut oil, and EVOO producers; supplement industry (omega-3, polyphenol products). Direct-to-consumer brands have monetised the "seed oils bad" framing.
• Cultural / community contrarian: ancestral-health and carnivore communities (Shanahan, Saladino, Knobbe), low-carb / keto practitioners, paleo writers, parts of the functional-medicine field. Online attention skews strongly contrarian.
• Independent investigators: Ramsden (NIH), Hibbeln (formerly NIH), Hooper (Cochrane group, UEA) — published the recovered-data trials and the meta-analyses; no obvious commercial alignment.

Population variability

• Secondary-prevention populations (post-MI): the Sydney trial population. Highest signal for potential LA harm in the recovered data; arguably most relevant to a high-risk reader.
• Older adults (>65): the MCE subgroup analysis suggests stronger negative cholesterol-mortality association in this group Ramsden et al. 2016; biological plausibility (cholesterol-mortality reversal in elderly cohorts is well-documented) supports treating older adults as a higher-stakes subgroup.
• Already low-LA diets (Mediterranean, traditional Asian, vegan-Mediterranean): incremental change from current LA likely small.
• Pregnancy / breastfeeding: LA and ALA needs rise; maternal seed-oil-heavy diet shifts breast-milk PUFA composition. Specific population not covered well in the recovered-data trials.
• Frequent restaurant / fast-food eaters: the aldehyde-exposure case is concentrated here. Largest expected payoff from change.

Knowledge gaps

What's not known and what would change the author's call: (1) a modern, well-powered RCT of mixed-PUFA seed-oil intake (not pure LA, not 1970s margarines with trans fat) on hard cardiovascular endpoints — the trial that would replace Sydney and MCE has not been run and is unlikely to be funded. (2) Human dose-response data on dietary aldehyde exposure and oxidative-stress biomarkers — the in-vitro and animal mechanistic case is strong, the human exposure-effect quantification is missing. (3) Whether unheated seed oils (cold-pressed, used as salad dressing, in baked goods at moderate temperature) carry the same risk profile as repeatedly-heated restaurant frying oil — the question that would let the entry distinguish "home soybean oil" from "deep-fryer soybean oil." (4) Replication of the older-adult cholesterol-mortality reversal under modern care patterns. (5) Long-term ratio-correction trials: does shifting ω-6:ω-3 from 15:1 to 4:1 over years move hard endpoints?

Scope vs brief. The topic brief named soybean, sunflower, and canola explicitly and asked for linoleic acid intake, omega-3/omega-6 ratio, oxidation products, cardiovascular markers, and the active debate. All five are covered. The article additionally distinguishes canola (~20% LA, chemically closer to olive oil) from the higher-LA seed oils in misconceptions — the colloquial "seed oils" shorthand collapses real chemical differences and treating them as one is a misframing.

Hard scoping calls.

• Cautious-revisionist landing rather than either camp. The dossier's credibility range explicitly aired both the mainstream (Sacks/AHA, Hooper/Cochrane, Marklund/Li biomarker cohorts) and contrarian (Ramsden recovered-data trials, DiNicolantonio's oxidised-LA hypothesis, ω-6/ω-3 ratio case) positions. The article lands on "the deep-fried restaurant channel is the strongest case; home use of moderate-LA oils is honestly uncertain." This is the editorial call and a reviewer should know it: an alternative entry could land harder either way.
• No alarmism on home cooking. The contrarian online discourse pushes "all seed oils are poison" framing. The literature does not support that on home-use moderate-LA oils. We chose to be explicit (in misconceptions: "the weakest case is against canola in moderate home use") rather than let the dek's hedge carry the work.
• Aldehyde mechanism kept central despite thin human RCT evidence. The mechanism story (4-HNE, Csallany fast-food fries measurement, Grootveld review) is in-vitro/observational, but it is the one part of the contrarian case that even mainstream sources acknowledge. We anchored it with a science callout in mechanism rather than burying it, because it's where the actionable advice lands.

Excluded — flagged for separate-entry candidates.

• Omega-3 supplementation — the other half of the ratio question. Mentioned in protocol and out-of-scope; deserves its own entry covering EPA/DHA dose, fish-oil vs algae-oil, ALA conversion limits, the REDUCE-IT vs STRENGTH controversy.
• Industrial trans fats — co-occur with seed oils in the food supply but are a chemically distinct substance with settled-bad evidence. Future entry candidate.
• ApoB / lipid panel literacy — what cholesterol numbers do and don't tell you about heart risk. Surfaced in this entry's discussion of why "LDL down" ≠ "longer life" but warrants its own treatment.
• NAFLD and metabolic-syndrome-specific LA case — the soybean-oil mouse-model work (Sanders, Deol et al.) and human NAFLD associations are substantial enough for their own entry.
• Topical / dermatological use of seed oils — different exposure route and evidence base. Belongs in the skin category if at all.

Rating difficulties.

• Longevity (scored 2). The single hardest call in this entry. Defensible range is 1–3: a strict reading of Cochrane's no-mortality-effect finding pushes toward 1; the Marklund/Li biomarker cohorts taken at face value push toward 3; the recovered-data trial weight pushes toward 1–2. Settled on 2 as honest middle, reflecting the small benefit available from removing restaurant deep-fried exposure without overclaiming the broader case.
• Evidence (scored 2). Hovered between 2 and 3. The literature is large (3+ would be defensible by volume), but the partitioning into camps with diverging conclusions on the same trials drops the de-facto strength to 2 (sparse / contested / mechanism plausible but trials mixed). The Cochrane review is moderate-quality on events and finds null on mortality, which is the cleanest summary.
• Controversy (scored 4). No real hesitation. AHA presidential advisory vs BMJ recovered-data reanalyses vs Open Heart oxidised-LA hypothesis — multiple credible camps with foundational disagreement, no near-term resolution.
• Short-term health and beauty-cumulative (both 1). Pushed by community signal and Mediterranean-pattern data; not by trial evidence specific to seed-oil removal. Honest 1s; would not defend 2.
• Energy / focus / sleep / mood (all 0). Considered non-zero for energy on the "less inflammation → less fatigue" community line, but the dossier doesn't carry trial backing. Conservative 0s.

Future-link candidates. When entries are written, link from this one: omega-3 fish-oil supplementation, ApoB and the lipid panel, ultraprocessed food, industrial trans fats, the Mediterranean dietary pattern.