Dietary fats
The "low-fat" decades produced a generation of confused eaters and little measurable cardiovascular benefit. The evidence that replaced them is more useful: the type of fat matters more than the amount, what replaces a removed fat largely decides the outcome, and the food a fat arrives in changes how it behaves.
Industrial trans fats are unambiguously harmful; cutting saturated fat pays off mainly when polyunsaturated fat replaces it; polyunsaturated fats — omega-3 and, contrary to a popular framing, omega-6 — look protective; olive oil anchors the best-evidenced longevity diet, on evidence weaker than its reputation; and eggs are fine at up to about one a day.
The framework, briefly
Four practical rules organise almost all of the evidence:
- Eliminate industrial trans fats. Non-negotiable.
- Don't reduce saturated fat in isolation — substitute it. The benefit comes from what you put in instead. Polyunsaturated fat has the strongest case, monounsaturated fat a probable one; refined carbohydrates don't help.
- Get adequate omega-3s. From fatty fish twice a week, or from a 1–1.5 g/day supplement of the marine omega-3s EPA (eicosapentaenoic acid) and DHA (docosahexaenoic acid) — see Omega-3.
- The food matters, not just the molecule. Saturated fat in cheese and yogurt tracks differently from saturated fat in red meat; an egg eaten with vegetables tracks differently from one eaten with bacon.
The reason amount barely makes the list is that the trials testing it came up empty on cardiovascular outcomes. The Women's Health Initiative (WHI) Dietary Modification Trial randomised 48,835 postmenopausal women to a low-fat pattern (target 20% of energy from fat) and over ~8 years found no reduction in coronary heart disease, stroke, or total cardiovascular disease.[1] Two things complicate that flat reading. The intervention arm never reached its target — by year six, fat intake had fallen about 8 percentage points of energy relative to the comparison group, not the 15 or so the design assumed — so this tested a moderate reduction, not a low-fat diet. And in a later subgroup analysis, women who entered with neither prior cardiovascular disease nor hypertension did show about 30% less coronary heart disease during the intervention period (HR 0.70, 95% CI 0.56–0.87), partly offset by an increase in ischaemic stroke (HR 1.46, 1.07–1.97); the 3.4% who entered with established disease had a higher coronary hazard ratio, but the authors judged that result uninterpretable because of post-randomisation cholesterol-lowering drug use.[2] The hazard ratios moved close to null once the intervention ended and the subgroups were not pre-registered, so this is a reason to hold the null loosely, not to overturn it.
The large PURE cohort pointed the same way, with higher total fat intake tracking lower total mortality and very-high carbohydrate intake tracking higher mortality — though PURE is observational, much of its carbohydrate signal came from low-income subsistence populations, and it drew heavy methodological criticism.[3] Neither overturns the substitution framework; both anchor the "type, not amount" reading.
Industrial trans fats — the one molecule to actually eliminate
Industrial trans fats, formed by the partial hydrogenation of vegetable oils, are the most harmful dietary fat yet characterised. The trans double bond produces a rigid, straight molecule that mimics a saturated fat structurally but behaves very differently in the body.
A 1% increase in trans-fat energy associates with roughly 6% higher all-cause mortality and 6% higher cardiovascular mortality (RR 1.06, 95% CI 1.01–1.10 and 1.06, 1.02–1.11) — pooled across 19 prospective cohorts and more than a million people, so this is observational rather than trial evidence, and both intervals sit close to the no-effect line.[4] (Throughout, the bracketed 95% confidence interval is the range where the true effect most plausibly lies; when it stays entirely on one side of 1.0 the result is unlikely to be chance, and when it crosses 1.0 the result is compatible with no effect.) Trans fats degrade the lipid profile in both directions at once — raising LDL ("bad") cholesterol while suppressing HDL ("good") cholesterol — and higher trans-fat intake tracks with higher blood levels of the inflammatory adhesion markers E-selectin, ICAM-1 and VCAM-1, and markedly higher C-reactive protein.[5]
Fat type reaches the epigenome quickly. In a seven-week overfeeding trial in healthy adults, adding 750 kcal/day of mostly saturated or mostly polyunsaturated fat produced different methylation changes in fat tissue — saturated overfeeding altered a set of genes including ADIPOQ, which encodes the insulin-sensitising hormone adiponectin, while the inflammatory signal TNF was among the much larger set that shifted under both diets.[6] That is fat quality registering epigenetically within weeks — a general result, not a trans-fat one.
The good news is policy-driven. Between 1990 and 2021, among adults aged 55 and over — the group in which the burden is concentrated and the only group this analysis covers — the age-standardised death rate attributable to high trans-fat intake fell by about 69% (from 15.5 to 4.8 per 100,000) and the disability burden by about 68%, driven by industrial-trans-fat bans in high-income regions.[7] The FDA's 2015 determination phased partially hydrogenated oils out of the US food supply between June 2018 and 1 January 2021; the EU capped industrial trans fats at 2 g per 100 g of fat from 2021. Globally the job is half done: by 2023, 53 countries had best-practice trans-fat policies covering 3.7 billion people — 46% of the world's population, up from 6% in 2018 — which the World Health Organization estimates prevents on the order of 183,000 deaths a year, while more than 4 billion people, concentrated in the African and Western Pacific regions, remain unprotected.[8] Residual exposure is concentrated in those unregulated markets and some legacy processed foods; reading ingredient labels for "partially hydrogenated oil" is still the reliable individual safeguard.
Naturally occurring (ruminant) trans fats — chiefly vaccenic acid and conjugated linoleic acid in dairy and meat — are biologically distinct from the industrial form and are not associated with cardiovascular harm at typical dietary intakes. A 2026 synthesis of 10 randomised trials and 12 cohorts found no meaningful lipid differences and no association with cardiovascular disease, cardiovascular death or type 2 diabetes across intakes from 1.3 to 13.2 g/day.[9]
Saturated fat — the substitution effect dominates
Saturated fatty acids (SFAs) are where the 1980s consensus has been most heavily revised. The honest current reading: cutting saturated fat does modestly reduce cardiovascular events, and what replaces it largely decides how much.
The strongest randomised evidence is a Cochrane review of 15 trials in ~56,675 participants; for the combined cardiovascular-events outcome specifically, 12 of those trials in 53,758 people found that reducing saturated fat for at least two years cut events by about 17% (RR 0.83, 95% CI 0.70–0.98; GRADE moderate-quality), with little or no effect on all-cause or cardiovascular mortality (RR 0.96 and 0.95).[10] Meta-regression showed that greater falls in saturated fat — and the resulting greater falls in serum cholesterol — produced greater event reductions, and the benefit was concentrated in trials replacing saturated fat with polyunsaturated fat or starchy foods, though the review found no statistically significant difference between those two replacements.
A smaller 2025 meta-analysis reads less favourably, finding essentially null effects on all-cause mortality (RR 1.01), cardiovascular mortality (RR 0.94) and coronary events (RR 0.85, not statistically significant).[11] The gap is smaller than it looks and is mostly about which trials and which endpoints. That analysis pooled 9 trials in 13,532 people, seven of them in patients who already had cardiovascular disease, and it reported no combined cardiovascular-events outcome — the outcome on which Cochrane found its benefit. On mortality the two agree. So: cutting saturated fat moves events, not deaths, and the size of the cholesterol drop tracks the size of the benefit.
What the replacement is makes the sharpest difference:
| Substitution | Effect on cardiovascular disease |
|---|---|
| SFA → polyunsaturated fat (PUFA) | ~30% risk reduction — similar effect size to statin primary prevention |
| SFA → monounsaturated fat (MUFA) | Probable, but the least certain of the unsaturated replacements — see note below |
| SFA → refined carbohydrates / added sugar | No benefit; insulin resistance and dyslipidemia worsen |
| SFA → unsaturated fat or whole-grain carbohydrate (5% isocaloric shift) | ~25% lower coronary risk if the replacement is polyunsaturated (HR 0.75), ~15% if monounsaturated (HR 0.85), ~9% if whole-grain carbohydrate (HR 0.91) |
The monounsaturated row carries the least evidence. Cochrane found the effects of replacement with monounsaturated fat "unclear" on very limited data, and the WHO's 2023 guideline makes replacement with plant monounsaturated fat a conditional recommendation while replacement with polyunsaturated fat is a strong one.[12] A 2026 pooling of 53 randomised trial comparisons in 1,690 adults found high-polyunsaturated diets beat high-monounsaturated diets on LDL cholesterol by only about 3 mg/dL (under 0.1 mmol/L) — a real but small edge.[13]
The first and third rows are the central point of the 2017 American Heart Association (AHA) Presidential Advisory;[14] the per-5%-of-energy figures in the last row come from two large US cohorts followed for decades.[15] The unifying mechanism is ApoB-containing lipoproteins: most of saturated fat's cardiovascular effect runs through its elevation of LDL (and other ApoB particles), which is why the substitution framework works and why the Cochrane cholesterol-magnitude relationship holds.
The food matrix is doing real work
Not all saturated fat is biologically equivalent, though the evidence is thinner than the confident version of this claim suggests. For fermented dairy (yogurt, kefir, traditionally aged cheese) the signal is neutral to mildly favourable: pooling 18 cohorts and 762,414 people, total fermented dairy fell just short of significance for stroke (RR 0.91, 95% CI 0.82–1.01), while cheese above 25 g/day reached it (0.91, 0.86–0.96).[16] Fermentation is not obviously the active ingredient — in the same analysis ordinary milk carried a small inverse association (RR 0.93 per 200 g/day, 0.88–0.98) and high-fat milk a smaller adverse one (1.04, 1.02–1.06). For red meat the substitution evidence is firmer: swapping low-fat dairy for red meat tracked with about 11% higher all-cause mortality, 13% more coronary disease and 20% more type 2 diabetes, at moderate certainty.[17] The proposed reasons are physical and biochemical:
- Dairy delivers saturated fat embedded in a calcium- and bacterial-metabolite-rich matrix; calcium binding fatty acids in the gut and the food's structure plausibly blunt the LDL rise that pure saturated fat would produce. The biomarker support is a pooled analysis of 16 prospective cohorts (63,682 participants), in which people near the top of the range for the dairy-fat fatty acids C15:0 and C17:0 had roughly 20% and 35% lower rates of new type 2 diabetes than those near the bottom (HR 0.80, 95% CI 0.73–0.87 and 0.65, 0.59–0.72).[18] But these are largely intake markers, so direct causality is unproven — and as the authors note, they "do not distinguish between different food sources of dairy fat," so they cannot tell fermented dairy from milk or butter.
- Fatty meat delivers the same molecules without the matrix advantages — and with Neu5Gc, heme iron, TMAO precursors, and (when high-heat cooked) HCAs and PAHs. See Red and processed meat.
This is why "how much saturated fat?" is less informative than "saturated fat from what?".
The epigenetic-age signal
A National Health and Nutrition Examination Survey (NHANES) epigenetic-clock analysis of 1,771 US adults aged 50+ links a doubling of total saturated-fat intake to a 0.42-year increase in GrimAge2 (95% CI 0.02–0.83), an epigenetic clock specifically trained to predict time to death.[19] Specific fatty acids vary in effect:
- Palmitic acid (16:0): +0.55 years of GrimAge2 per doubling (0.07–1.04) — the largest single saturated species.
- Stearic acid (18:0): +0.43 years per doubling (0.06–0.80).
These are suggestive, not settled. The analysis is cross-sectional, so it shows correlation on a surrogate rather than a proven aging effect, and intake came from a single 24-hour dietary recall. A doubling of monounsaturated fat was also associated with greater GrimAge2 (+0.54 years), which sits awkwardly beside the olive-oil evidence. The authors are candid that they applied no correction for multiple testing across eleven clocks — "some of our findings may be due to chance" — and in their own sensitivity analysis using a larger imputed sample both the saturated-fat and the monounsaturated-fat associations lost statistical significance. Their summary: "the most consistent and robust associations were observed for PUFAs" — the fats in the next section, not these.
The randomised tests of the broader premise are sobering. A 12-month trial randomising 144 adults to a low-fat or a low-carbohydrate diet found that changes in epigenetic age on three next-generation clocks were unrelated to changes in the trial's biomarkers and did not mediate the effects of weight loss; the authors "underscore the need for caution in interpreting epigenetic aging as a modifiable intervention target."[20] A two-year study of 899 adults found the clocks so stable within a person that they showed limited sensitivity to short-term change at all.[21]
A 2024 Mendelian randomization study — a design that uses inherited genetic variants as a natural experiment, so it is less vulnerable to confounding than a cohort — reported circulating saturated fat associated with shorter leukocyte telomeres (OR 0.80, 95% CI 0.70–0.91).[22] Two caveats. That figure comes only from the model that adjusts all the fat classes against each other; the study's primary unadjusted analysis pointed the other way, weakly associating saturated fat with longer telomeres. And telomere length is not a check on the clock findings above — the two markers are close to independent, sharing under 5% of their variance once age is accounted for, and the NHANES analysis itself found no association between any fatty acid and its own methylation-based telomere estimate.
The bottom line on saturated fat
A large 2025 cohort makes the substitution concrete: across 221,054 adults followed up to 33 years, the highest butter intake carried 15% higher total mortality (HR 1.15) while the highest plant-oil intake carried 16% lower (HR 0.84), and modelling a swap of ~10 g/day of butter for plant oils was associated with ~17% lower total and cancer mortality.[23] Observational, with the usual healthy-user caveats (higher-butter consumers smoked more and moved less) — but directionally consistent with the trial picture.
Cap saturated fat at roughly 10% of total energy (the standard public-health threshold) — but spend most of your effort on what replaces it: polyunsaturated oils, olive oil and nuts, not refined carbohydrates.
Polyunsaturated fats — the geroprotective class
Polyunsaturated fatty acids (PUFAs) — both omega-3 and omega-6 — are essential nutrients with the most consistent aging-related signal among dietary fats.
Omega-3s: the strongest single epigenetic signal
The DO-HEALTH trial, a 2,157-participant, three-year randomised trial in European adults aged 70 and over, tested 1 g/day of EPA+DHA omega-3 supplementation. In an epigenetic-clock analysis, omega-3 slowed three of four next-generation DNA-methylation clocks (PhenoAge, GrimAge2, DunedinPACE), corresponding to roughly 2.9–3.8 months of biological age reduction over three years. The additive effect the trial is often quoted for applies to one clock only: the paper reports evidence of additive effects for PhenoAge and states plainly that "no additive effects across interventions were observed for GrimAge, GrimAge2 or DunedinPACE." Vitamin D and exercise on their own moved none of the clocks.[24]
That clock result is the smaller part of the trial's story. It was a post-hoc analysis of the healthier Swiss subset only (777 people), with two measurement timepoints and no single pre-specified primary outcome, while DO-HEALTH's six pre-specified primary clinical endpoints were all null for omega-3 (blood pressure, physical function, cognition, fractures, infections). In a separate report from the same trial, omega-3 modestly raised LDL cholesterol (+0.11 mmol/L, 95% CI 0.06–0.16) and total cholesterol (+0.15 mmol/L) while slightly raising HDL and lowering triglycerides — the authors' own word for the lipid picture is "mixed" — and did not reduce major cardiovascular events, though that was an exploratory endpoint the trial had no power to test.[25] The comparator was sunflower oil rather than an inert placebo, and the parent trial received capsules and support from DSM Nutritional Products, Roche, NESTEC, Pfizer and Streuli, disclosed in the paper. So: a small surrogate-endpoint signal in a trial whose hard clinical endpoints were null. The clinical-outcome trials for marine omega-3 (VITAL, ASCEND, REDUCE-IT, STRENGTH) and the atrial-fibrillation dose signal are covered under Omega-3.
The NHANES analysis extends the picture: a doubling of total PUFA intake associates with roughly 0.6–0.7 years of decreased biological age across Horvath1, Hannum, and PhenoAge, and omega-3 intake tracks with lower epigenetic age on four clocks.[26] These figures are not the same currency as the saturated-fat number above: two of the three are first-generation clocks, which a 2025 comparison across 18,859 people and 174 diseases found have "limited applications in disease settings" and are outperformed by the newer generation.[27] Alpha-linolenic acid (ALA, plant-derived 18:3) carried the largest single effect size, −0.69 years on PhenoAge — though its confidence interval (−1.31 to 0.06) crossed the null. For ALA specifically, a dose-response meta-analysis found higher intake modestly lowered all-cause mortality (RR 0.90, 95% CI 0.83–0.97) and cardiovascular mortality (0.92, 0.86–0.99), with a small increase in cancer mortality at the highest intakes (1.06, 1.02–1.11) that did not hold up in the dose-response analysis (1.03, 0.98–1.08) — a more mixed picture than the single biological-age figure implies.[28]
The proposed mechanisms run through eicosanoid signalling, membrane fluidity and mitochondrial biogenesis; polyunsaturated fats are also the substrate for the specialized pro-resolving mediators (resolvins, protectins, maresins) that actively terminate inflammation rather than merely suppressing it.
If you eat fatty fish (salmon, sardines, mackerel, herring) twice a week, you're covered. If you don't, a 1–1.5 g/day EPA+DHA supplement is reasonable; doses, freshness, oxidation, and form are covered under Omega-3.
The "seed oil myth" — what the evidence actually says
The hypothesis that omega-6 linoleic acid (LA), the dominant PUFA in sunflower, safflower, soybean, canola, and corn oils, drives systemic inflammation and cardiovascular disease has become a popular online position. The evidence does not support it.
A 2025 review pooling 150 cohort studies found higher omega-6 levels associated with lower mortality — about 12% lower for dietary intake (RR 0.88, 95% CI 0.83–0.94) — with the strongest and most consistent signals coming from blood measurements rather than diet questionnaires.[29] Its results for dietary omega-6 were null for cardiovascular disease (0.92, 0.83–1.02) and for cancer overall (1.01, 0.97–1.04), with one adverse signal — endometrial and ovarian cancer (RR 1.15, 1.01–1.30) — and certainty rated very low to moderate, so read it as no overall harm plus a mortality benefit, not a broad protective effect.
Higher blood linoleic acid is not merely neutral but favourable: pooling 30 cohorts across 13 countries, 68,659 people and 15,198 events, the highest-versus-lowest range carried about 7% lower total cardiovascular disease (HR 0.93, 95% CI 0.88–0.99), 22% lower cardiovascular mortality (0.78, 0.70–0.85) and 12% lower ischaemic stroke (0.88, 0.79–0.98) — the authors conclude the results "support a favorable role for LA in CVD prevention."[30] And linoleic acid does not raise circulating inflammatory markers in controlled trials: a systematic review of 15 randomised trials found no effect on C-reactive protein, fibrinogen, cytokines, adhesion molecules or TNF-α.[31]
The AHA and most national authorities classify replacing saturated fat with polyunsaturated fat as cardioprotective, and the Cochrane review of saturated-fat reduction supports it. Cochrane's separate review of trials increasing omega-6 — which usually displaced saturated or monounsaturated fat — is more guarded: across its 19 trials in 6,461 people it found little or no difference in mortality (RR 1.00, 95% CI 0.88–1.12) or cardiovascular events (0.97, 0.81–1.15) at low certainty, concluding there is no evidence that increasing omega-6 reduces cardiovascular outcomes other than heart attack.[32] The strongest case for seed oils is as a substitute for saturated fat.
The remaining valid critiques of "seed oils":
- Industrial deep-frying — commercial fryers running for hours — produces oxidised lipid species with measurable effects in humans, though the evidence is about blood-vessel function rather than inflammation. In a controlled test in 10 men, a meal cooked in fat taken from a commercial fast-food fryer abolished the arteries' ability to dilate, while the identical meal cooked in fresh fat did not.[33] The pro-inflammatory characterisation rests on animal and cell studies; no chronic human exposure trial exists. The issue is cooking practice, not the molecule.
- Refined oils are calorie-dense and show up alongside refined sugar, refined flour, and additives inside ultra-processed foods — and that surrounding matrix is the more plausible driver of harm than linoleic acid itself. See Ultra-processed food.
The fixed reading: avoid ultra-processed foods and commercial deep-fried food; don't avoid linoleic acid in a home kitchen using fresh oil at moderate heat.
The omega-6 : omega-3 ratio
The two families of polyunsaturated fat compete for the same desaturase enzymes, especially Δ6-desaturase, and the popular version of the argument runs from there to a pro-inflammatory baseline. That step does not survive the human data. Pooling 36 human trials spanning everything from a 90% cut to a six-fold increase in dietary linoleic acid, changes in intake showed no relationship with arachidonic acid — the omega-6 the inflammatory signalling molecules are actually built from — in blood.[34] Absolute amounts, not the ratio, appear to drive conversion: in a 29-person feeding trial, two diets with the same 1:7 ratio of plant omega-3 to linoleic acid produced different conversion of plant omega-3 into its longer-chain forms.[35] The measured Western ratio is also lower than usually quoted: US survey estimates run around 8:1 to 10:1, rather than the 15:1 or 20:1 figure that circulates.[36]
The ratio does predict outcomes. A UK Biobank analysis of 85,425 adults found the circulating omega-6:omega-3 ratio to be an independent mortality predictor: the highest-ratio fifth of participants had about 26% higher all-cause mortality than the lowest (HR 1.26, 95% CI 1.15–1.38) and 31% higher cardiovascular mortality (1.31, 1.10–1.55); the cancer figure of 14% higher is the weakest, with an interval that reaches down to no effect (1.14, 1.00–1.31). The same paper reports something the ratio framing tends to hide: taken on its own, higher omega-6 was also protective — about 23% lower all-cause mortality and 20% lower cancer mortality in the highest fifth versus the lowest (HR 0.77, 0.70–0.85 and 0.80, 0.68–0.92).[37] As its plain-language summary puts it, "both are protective", and the ratio predicts harm mainly because omega-3 protects more strongly — an argument for adding omega-3, not for subtracting linoleic acid. A separate and larger UK Biobank analysis added the ratio to the SCORE2 cardiovascular risk model and found it improved prediction, but barely, moving the model's discrimination from 0.742 to 0.747 on a scale where 0.5 is a coin flip and 1.0 is perfect — an improvement its own authors call "modest from a clinical standpoint."[38]
There is no authoritative ratio target. A joint FAO/WHO expert consultation concluded there is "no rational for a specific recommendation for n-6 to n-3 ratio" provided absolute intakes of each are adequate.[39] The Nordic Nutrition Recommendations, the European Food Safety Authority, the US Institute of Medicine, Japan and China all likewise set absolute amounts and no ratio. The 1:1 to 4:1 figure that circulates online traces to a single review, which gave disease-specific ratios (4:1 for secondary prevention of cardiovascular disease, 2–3:1 in rheumatoid arthritis, 5:1 in asthma) rather than one target, not to any guideline.[40]
Monounsaturated fats and extra-virgin olive oil
Monounsaturated fatty acids (MUFAs), characterised by a single double bond and dominated dietarily by oleic acid, are the lipid backbone of the Mediterranean diet — the most-evidenced longevity dietary pattern by a comfortable margin. See Dietary patterns.
The headline trial: PREDIMED, a 7,447-participant Spanish randomised trial in adults at high cardiovascular risk, showed a ~30% reduction in major cardiovascular events in both the Mediterranean-plus-olive-oil arm (HR 0.69, 95% CI 0.53–0.91) and the Mediterranean-plus-nuts arm (0.72, 0.54–0.95) versus a low-fat control, over a median 4.8 years.[41] Those are the corrected figures: the original 2013 paper was retracted in 2018 and republished after a statistical audit found randomisation deviations affecting ~21% of participants (1,588 of 7,447), including a clinic that misused the randomisation table and whole households co-assigned to one arm. The reanalysis showed roughly the same benefit, so the conclusion survived. One conflict belongs on the record, narrower than usually stated: the funding was public, from the Spanish health ministry's research institute and related bodies, but the olive oil and nuts given to participants were donated by producers — Spanish olive-oil, almond and hazelnut firms and the California Walnut Commission — none of which, the paper states, had any role in the study's design, analysis or reporting.
An umbrella review of 31 meta-analyses found olive oil consumption associated with about 15% less cardiovascular disease (RR 0.85, 95% CI 0.77–0.93) and 16% less type 2 diabetes (0.84, 0.77–0.92), comparing the highest intakes with the lowest.[42] Two honest qualifications: cancer overall was exactly null (1.00, 0.96–1.04), and the review rated 20 of its 31 included analyses as critically low quality, with only three high. The cardiovascular and diabetes signals are consistent; the evidence underneath them is not strong.
The benefit of olive oil is largest when it replaces something worse, and there is a direct measurement of this: across 92,383 US adults followed 28 years, swapping 10 g/day of butter, margarine, mayonnaise or dairy fat for olive oil tracked with 8% to 34% lower mortality depending on the outcome — while "no significant associations were observed when olive oil was compared with other vegetable oils combined."[43]
The same Mendelian randomization analysis discussed above associated circulating monounsaturated fat with longer telomeres (OR 1.15, 95% CI 1.07–1.24) — again, only in the model that adjusts the fat classes against each other; the unadjusted estimate is far smaller and barely clears no effect (OR 1.019, 1.006–1.033).[44]
Extra-virgin olive oil also delivers phenolic compounds (oleocanthal, hydroxytyrosol, oleuropein) that inhibit LDL oxidation independently of its fatty-acid content. That effect is dose-dependent and regulated: European authorities permit the health claim only for oils carrying at least 5 mg of hydroxytyrosol and its derivatives per 20 g of oil, at 20 g a day. The practical catch is that most oil on the shelf does not reach it — an analytical survey found no more than about 10% of bottled oils have enough phenolic content for the claim to apply.[45]
One small trial is a curiosity rather than a guide: healthy young women took 40 g/day of either high-phenolic extra-virgin olive oil or sunflower oil for 8 weeks, and within the olive-oil arm LDL cholesterol rose slightly (+5.1 mg/dL, about 0.13 mmol/L) while fasting insulin fell.[46] Those are within-group changes, with no head-to-head lipid comparison reported, so what the sunflower arm did is unknown. Twelve people completed the active arm; it is not a reason to switch oils.
Avocado oil and nut oils are reasonable alternatives — with one caveat specific to avocado oil, which has no legal identity or grade standard in the US. Repeated analyses from the same laboratory have found most retail samples rancid or adulterated with cheaper oils: 82% of bottled samples in 2020, around 70% of private-label bottles in 2023, and 89% of avocado-oil-labelled processed foods in 2026, with some bottles labelled "pure" or "extra virgin" almost entirely soybean oil.[47] Buy on brand reputation rather than label wording. Other plant oils — canola, soybean — sit in the neutral middle: balanced MUFA/PUFA profiles, no harm signal in cohort data, fewer phenolics than extra-virgin olive oil.
Coconut oil is not a health food. A meta-analysis of 16 trials found that, versus non-tropical vegetable oils, coconut oil raised LDL cholesterol by 10.5 mg/dL (95% CI 3.0–17.9; about 0.27 mmol/L) and HDL by 4.0 mg/dL, with no effect on measures of blood sugar, inflammation or body fatness.[48] The interval is wide and the trials disagree substantially, so treat 10.5 as the centre of a broad range. The "medium-chain triglyceride" framing is misleading: coconut oil's dominant fat is lauric acid (C12), which is absorbed via chylomicrons and behaves metabolically like a long-chain saturated fat, raising LDL rather than acting like a true MCT.
Dietary cholesterol and eggs
Eggs are the most-litigated single food in modern nutrition science. The evidence points to small effects that depend heavily on where, and with what, eggs are eaten.
The geographic paradox
Cohort outcomes for egg consumption differ by population, in a way that maps onto the rest of the dietary pattern around them. The cleanest comparison comes from a meta-analysis that reported its cardiovascular estimates separately by region, per 50 g of egg per day — about one egg:[49]
| Population | Association with cardiovascular disease, per ~1 egg/day |
|---|---|
| US cohorts | About 8% higher risk (RR 1.08, 95% CI 1.02–1.14) |
| European cohorts | Borderline, and the range still includes no effect (RR 1.05, 0.98–1.14) |
| Asian cohorts | No association (RR 0.96, 0.87–1.06) |
The Asian row is the one most often overstated: it shows an absence of harm, not a protective effect. A separate pooling of three large US cohorts alongside 28 others puts the Asian estimate slightly on the protective side at about 8% lower risk (RR 0.92, 0.85–0.99), so the direction is uncertain and the size small either way.[50]
A plausible explanation is the companion foods. In Western diets, an egg is typically eaten with bacon, sausage, refined-flour toast, and butter; in a traditional Japanese diet, the same egg is eaten with rice, miso, fish, and vegetables. The egg is identical; the meal is not.
The dose-response
The dose-response for eggs and mortality is close to linear, and modest. A meta-analysis pooling cohorts by eggs per week tested for a threshold effect on six outcomes and found none — the curve does not bend at any particular intake.[51] Two results stand out. The one clearly positive overall finding is cancer mortality, about 20% higher at the highest intakes versus the lowest (RR 1.20, 95% CI 1.04–1.39, 13 studies) — roughly 4% per additional egg per week. And the association is concentrated in US cohorts, where all-cause mortality was about 18% higher (RR 1.18, 1.10–1.26) and cardiovascular mortality about 20% higher (1.20, 1.10–1.31), against an inverse all-cause association in Asian cohorts (0.89, 0.80–0.99) — the same geographic split as above, in a second independent analysis. The American Heart Association's 2019 advisory sets no numeric limit on dietary cholesterol at all, framing the question around the overall dietary pattern; it is generally read as permitting up to one whole egg a day for healthy adults, with more caution in diabetes and in familial hypercholesterolaemia.[52]
Why the lipid panel barely moves
When healthy adults eat extra eggs, the liver compensates by down-regulating its own cholesterol synthesis. Feeding studies summarised in that egg dose-response meta-analysis show the lipid response is real but small: each extra 100 mg/day of dietary cholesterol raises LDL cholesterol by roughly 1.9 mg/dL (about 0.05 mmol/L) and HDL cholesterol by roughly 0.4 mg/dL.[53] The common add-on claim that the LDL-to-HDL balance stays put is not settled: the pooled trial analysis behind the most-quoted figure reports the total-cholesterol-to-HDL ratio rising by about 0.02 units per 100 mg/day.[54] A small movement, but a movement, not a wash. Two small Japanese intervention studies point the other way, reporting higher HDL cholesterol and less oxidised LDL after a month of one egg a day; both come from the research group behind a 2026 review of eggs and health, and neither is large enough to settle it.[55]
Phosphatidylcholine and the brain
Eggs are also the densest dietary source of phosphatidylcholine, which makes up roughly a third of the brain's phospholipids and serves as a precursor to acetylcholine, the neurotransmitter most damaged in Alzheimer's. In a crossover trial in 18 adults, choline delivered as egg-yolk phospholipid produced a roughly 4× higher plasma choline response than the same dose of synthetic choline bitartrate — though the dose tested was roughly 20 times the choline in one egg, and the authors note it may have saturated the absorption pathway.[56] Whether that translates into anything is less clear. The nearest evidence is a Finnish cohort followed 22 years in which men in the highest quarter of dietary phosphatidylcholine intake had about 28% lower dementia risk — a finding whose confidence interval runs from 1% to 48%, so it is real but barely.[57]
The bottom line on eggs
Whole eggs at up to about one a day are compatible with most longevity-oriented dietary patterns for healthy adults. Any risk is small, rises gradually rather than past a threshold, and is concentrated in the cohorts where eggs arrive alongside bacon, white toast, and processed cheese. People with type 2 diabetes may warrant more caution; some cohort data show a stronger cardiovascular association with eggs in established diabetes specifically (HR 1.69, 95% CI 1.09–2.62, at one egg a day or more versus less than one a week). The same meta-analysis linked that intake to about 42% more incident type 2 diabetes across five Western general-population cohorts of 69,297 people (HR 1.42, 1.09–1.86) — observational, with diabetes a secondary outcome in most of the studies, and the authors caution that it may not justify changing dietary advice.[58]
Practical guidance
A defensible, evidence-aligned framework for fat quality:
Eliminate industrial trans fats. Read ingredient labels for "partially hydrogenated oil". In regulated markets (US, EU, UK) residual exposure is now low; in unregulated markets it can still be substantial.
Replace saturated fat rather than just cutting it. Cap it near ~10% of energy, and swap butter and fatty red meat for polyunsaturated oils, olive oil and nuts — polyunsaturated fat has the strongest trial evidence. Replacing saturated fat with refined carbohydrates produces no benefit.
Use extra-virgin olive oil as the default cooking and dressing oil — ideally a high-phenolic one, since most bottles fall short of the level behind the health claim. Canola, sunflower, and safflower oils are sound alternatives; if you use avocado oil, buy a reputable brand.
Get omega-3s. Fatty fish twice a week or 1–1.5 g/day EPA+DHA. The marine-omega-3 outcome trials and dose caveats live under Omega-3.
Don't fear linoleic acid in a home kitchen. Avoid ultra-processed food and commercial deep-fried food, not the oil — and don't chase an omega-6:omega-3 ratio, which no major authority sets.
Treat dairy and red meat differently. Fermented dairy is roughly neutral; red meat in volume is the saturated-fat source most worth limiting.
Eggs: up to one a day is fine for most healthy adults. Pair them with vegetables and whole grains rather than bacon and refined-flour bread. People with diabetes or familial hypercholesterolaemia warrant a more conservative approach.
Further reading
- Sacks FM et al. Dietary fats and cardiovascular disease — AHA Presidential Advisory. Circulation 2017.[59]
- Marklund M et al. Biomarkers of dietary omega-6 fatty acids and incident cardiovascular disease and mortality. Circulation 2019.[60]
- Sadeghi R et al. Dietary and circulating omega-6 fatty acids and cardiovascular disease, cancer, and mortality — global meta-analysis of 150 cohorts. J Transl Med 2025.[61]
- de Goede J et al. Dairy consumption and risk of stroke — systematic review and updated dose-response meta-analysis. JAHA 2016.[62]
- Estruch R et al. Primary prevention of cardiovascular disease with a Mediterranean diet supplemented with extra-virgin olive oil or nuts — PREDIMED. NEJM 2018.[63]
- Bischoff-Ferrari HA et al. Individual and additive effects of vitamin D, omega-3 and exercise on DNA methylation clocks of biological aging in older adults from the DO-HEALTH trial. Nature Aging 2025.[64]
- Bozack AK et al. Dietary fatty acids and epigenetic aging in US adults: results from the National Health and Nutrition Examination Survey. npj Aging 2025.[65]
- Yamada S et al. Saturated fat restriction for cardiovascular disease prevention — meta-analysis of RCTs. JMA Journal 2025.[66]
- Hooper L et al. Reduction in saturated fat intake for cardiovascular disease. Cochrane Database of Systematic Reviews 2020.[67]
- Zhang Y et al. Butter and plant-based oils intake and mortality. JAMA Internal Medicine 2025.[68]
- Imamura F et al. Fatty acid biomarkers of dairy fat consumption and incidence of type 2 diabetes. PLOS Medicine 2018.[69]
- Neelakantan N et al. The effect of coconut oil consumption on cardiovascular risk factors. Circulation 2020.[70]
- Howard BV et al. Low-fat dietary pattern and risk of cardiovascular disease — WHI Dietary Modification Trial. JAMA 2006.[71]
- Dehghan M et al. Associations of fats and carbohydrate intake with cardiovascular disease and mortality — PURE. Lancet 2017.[72]
- Naghshi S et al. Dietary intake and biomarkers of alpha linolenic acid and risk of all cause, cardiovascular, and cancer mortality. BMJ 2021.[73]
- Mousavi SM et al. Consumption of eggs and risk of all-cause and cause-specific mortality — dose-response meta-analysis. Advances in Nutrition 2022.[74]
- Zhao B et al. Associations of dietary cholesterol, serum cholesterol, and egg consumption with mortality — systematic review and meta-analysis. Circulation 2022.[75]
- Chen Y et al. Relationship between fatty acid intake and aging — a Mendelian randomization study. Aging (Albany NY) 2024.[76]
- Ye J et al. Impact of diets high in trans-fatty acids on cardiovascular diseases in adults aged 55 and older. Front Nutr 2026.[77]