Foods to limit or avoid

What you take out of your diet usually matters more than any superfood you add. Ultra-processed food, sugary drinks, processed meat and alcohol carry the clearest harm signals in nutrition science, with industrial trans fats and excess salt close behind — while several of the internet's favourite villains, seed oils and eggs among them, are not where the harm is.

Build most of your diet from whole or minimally processed food, drink water, tea or coffee instead of sugary drinks and juice, treat processed meat as an occasional item, keep red meat to about three servings a week, and drink as little alcohol as you can. Cook with olive oil, eat fewer packaged and restaurant meals to cut salt, and aim for a pattern you can keep rather than a perfect one.

Ultra-processed food: the biggest lever, with honest caveats

Ultra-processed food (UPF) has the broadest harm signal in nutrition research. A 2024 review of 45 pooled analyses, covering about 9.9 million participants (some cohorts counted more than once), linked high intake to 32 adverse outcomes across heart, metabolic, cancer and mental health.[1]

How big the effect is remains contested. Each extra 10% of calories from UPF tracks with somewhere between about 3% and 10% higher death rates, depending on the analysis.[2][3] These are observational studies, and people who eat a lot of UPF also smoke more, earn less and weigh more, so some of that signal is confounding.

What makes a real effect likely is controlled feeding. In a 2019 US government metabolic-ward trial, adults allowed to eat freely ate about 500 kcal a day more, and gained weight, on an ultra-processed diet than on a minimally processed one matched for sugar, fat and fibre.[4] A 2025 London trial of 55 adults pointed the same way over eight weeks, but is contested: the two diets were not matched for calories per gram, the difference was about 1 kg, and several authors have obesity-drug industry ties.[5]

Not all UPF is equal. In two large US cohorts of about 114,000 health professionals followed for about 34 years, the highest UPF intake carried only about 4% higher mortality, concentrated in processed meat and in sugary and artificially sweetened drinks; no category lowered overall death rates.[6] Whether the UPF label adds anything beyond the poor overall diet that usually comes with it is still open: in those cohorts the link largely vanished once diet quality was accounted for, but a 2026 analysis of nearly 48,000 US adults found it only partly weakened.[7]

What to do:

  • Aim for under 10–20% of calories from UPF. That is a practical target, not an official one — no authority has set a number. The average US adult is at about 53%.[8]
  • Use the kitchen test. If an ingredient list contains even one thing you would rarely find in a home kitchen (modified starches, protein isolates, hydrogenated fats) or an additive whose job is to improve looks or taste (emulsifiers, flavour enhancers, colours, sweeteners, thickeners), the product is ultra-processed.[9]

The full picture, including how foods are classified and how to read a label, is in Ultra-processed food.

Sugary drinks, juice and sweeteners

Sugary drinks are the most consistent single-food harm signal among beverages. In two large US cohorts pooled together, each daily 355 mL can tracked with roughly 7% higher death from any cause and 10% higher cardiovascular death,[10] figures a 2025 review of the wider literature closely reproduces.[11] In a cohort of Chinese adults, four or more soft drinks a week tracked with about 47% higher risk of fatty liver disease.[12] There are no randomised trials with hard outcomes; the case rests on large, consistent cohorts and a coherent mechanism.

The problem is the liquid. Fructose in a drink bypasses fullness signals and reaches the liver fast enough to be turned into fat. The same sugar inside whole fruit, slowed by fibre, behaves differently: across 27 cohort studies, each 200 g of fruit a day tracked with about 15% lower mortality.[13] Fruit juice, even 100%, is closer to the sugary drinks than to fruit, but its own evidence is thin and mixed. One pooled analysis put it at about 26% higher all-cause mortality, though only two cohorts fed that figure; unlike sugary drinks, which in the same analysis tracked with about 27% higher risk of type 2 diabetes, juice was not linked to diabetes, high blood pressure or stroke.[14] A larger Harvard analysis found no link between juice and mortality,[15] and in three US cohorts each three glasses a week went with about 8% more diabetes.[16] Why whole fruit and juice differ is on Fruit and vegetables.

The World Health Organization advises keeping free sugars under 10% of energy and ideally under 5% (about 25 g a day), and the American Heart Association caps added sugar at 25 g a day for women and 36 g for men.[17] The 2025–2030 US Dietary Guidelines dropped the percentage ceiling and state that "no amount of added sugars or non-nutritive sweeteners is recommended or considered part of a healthy or nutritious diet", with no single meal to exceed 10 g of added sugar.[18] Those thresholds rest on moderate-to-very-low-certainty evidence; the drinks data are the stronger case.

Diet drinks are a lesser problem, not a free pass. In a long-running Massachusetts community study, a roughly threefold higher rate of ischaemic stroke (the kind caused by a blocked artery) belonged to daily artificially sweetened drinks, not sugary ones — a finding routinely misreported — while its dementia link did not survive full adjustment or replicate in a later pooling of six cohorts.[19][20] In about 12,800 Brazilian adults, the highest sweetener intake tracked with slightly faster cognitive decline — but the effects were very small, appeared only in people under 60, and the paper has been corrected twice.[21]

The sugar alcohols erythritol and xylitol, common in "keto" and "sugar-free" products, deserve more caution. Among patients referred for heart tests at US hospitals (and, for erythritol, a German one), those with the highest blood levels had more heart attacks, strokes and deaths over the next three years — roughly double for erythritol and about 60% more for xylitol — and lab work showed both make platelets clot more readily.[22][23] It is observational, but enough not to treat them as harmless. Water, unsweetened tea, sparkling water and unsweetened coffee are better defaults. The detail is in Sugar substitutes and Sweeteners.

Processed meat causes cancer; plain red meat is a smaller, murkier risk

Processed meat — anything smoked, cured, salted, fermented or chemically preserved, such as bacon, sausages, hot dogs, ham, salami, deli meats and jerky — is a known human carcinogen. The World Health Organization's cancer agency (the International Agency for Research on Cancer, IARC) puts it in the same category as tobacco, though the risk is far smaller: each 50 g a day raises colorectal cancer risk by about 18%,[24] which moves a US adult's lifetime risk from about 3.9% to roughly 4.6%.

The other harms are observational but consistent. Per 50 g a day, type 2 diabetes risk is about 15% higher, in 1.97 million adults across 31 cohorts.[25] Per daily serving, death rates run about 15% higher.[26] Per 50 g a day, cardiovascular risk is about 26% higher.[27] In over 133,000 US adults, eating a quarter-serving a day or more, compared with under a tenth, tracked with 13% more dementia and 14% more self-reported cognitive decline.[28]

Processed meat carries about four times the sodium of fresh red meat, and sodium alone is estimated to account for about two-thirds of the gap in coronary heart disease risk between processed and fresh red meat.[29] Suspected causes of the cancer risk include compounds formed from nitrite preservatives, smoking and high-heat cooking. The World Cancer Research Fund (WCRF) advises eating "very little, if any": no level of intake has been shown to be free of colorectal-cancer risk.[30]

Unprocessed red meat is rated "probably carcinogenic", one step down.[31] Its risks are smaller: about 11% higher cardiovascular risk and about 10% higher type 2 diabetes risk per 100 g a day,[32][33] and about 15% higher colorectal cancer risk in a 2025 pooling of 60 studies, against about 21% for processed meat — for colorectal cancer as a whole; for colon and rectal cancer taken separately, red meat's estimates run slightly higher.[34] The signal is less consistent too: it appears in US cohorts but often not in European and Asian ones, and genetic studies that use inherited variation as a stand-in for lifelong intake have so far found no causal link with heart disease, stroke or type 2 diabetes.[35]

That weaker evidence fuels a real dispute. A 2019 guideline panel argued it was too uncertain to tell people to cut down,[36] though its own analysis still found "a very small reduction in risk" with less meat,[37] and IARC and thirteen other organisations rejected the recommendation.[38] The most conservative re-analyses rate the evidence "weak" for both: for plain red meat, weak for four outcomes and absent for stroke; for processed meat, weak — though the 2025 authors still say limiting intake is justified.[39][40]

What to do: up to three servings of red meat a week is a defensible cap — the WCRF's ceiling, about 350–500 g cooked weight a week, with one serving about 120–170 g cooked. Replacing red meat with nuts or legumes tracked with about 16% lower coronary heart disease, eggs 14%, poultry 12% and dairy 10%; fish showed no benefit.[41] If you eat red meat, choose lean cuts and cook gently — stew, braise, slow-roast — rather than charring. Mechanisms, cooking and the rest of the debate are in Red and processed meat.

Salt: most of it is already in the food you buy

Cutting salt lowers blood pressure in randomised trials. In one, 412 adults were assigned to three sodium levels: going from high to low lowered systolic pressure by about 7 mm Hg on a typical US diet — a substantial fraction of what a single first-line blood-pressure drug delivers.[42] The effect is larger in salt-sensitive people, who are more common with age, in African ancestry, and among people with hypertension or chronic kidney disease.[43]

The World Health Organization target is under 2 g of sodium a day (about 5 g of salt); the US average is around 3.4 g. In one US study, about 70% of sodium was added during commercial processing and restaurant cooking, not at the table,[44] so eating fewer packaged and restaurant meals is the lever that matters. The sodium-to-potassium ratio may predict cardiovascular outcomes better than sodium alone,[45] a reason to favour potassium-rich whole foods. If you have high blood pressure or are at risk of it, a potassium salt substitute is a reasonable swap for the salt you add at home, unless you have kidney disease or take a drug that holds on to potassium.

The claim that too little salt is dangerous rests mainly on a 17-country study of about 102,000 people that reported excess risk at low as well as high intakes,[46] but it estimated sodium from a single morning urine sample, a method known to manufacture that J-shaped curve; studies using repeated full 24-hour collections find risk rising steadily with intake, and part of the apparent low-salt harm is simply that people who are already ill eat less.[47]

Salt is also linked to stomach cancer: each extra 5 g a day tracks with about 12% higher gastric cancer risk, thought to act through damage to the stomach lining and an interaction with H. pylori infection.[48]

Alcohol: the cancer risk starts with the first drink

Alcohol is an IARC Group 1 carcinogen, and its cancer risk has no threshold. Even light drinking — up to one drink a day — raises the risk of mouth and throat cancer by about 13%, of one common type of oesophageal cancer by about 26%, and of breast cancer by about 4%; heavy drinking (over 50 g of alcohol a day) raises both mouth-and-throat and oesophageal cancer roughly five-fold.[49] Breast-cancer risk rises about 7% for every 10 g of alcohol a day, roughly one drink.[50] About 4% of all new cancers in 2020 were attributable to alcohol, around 103,000 of them to moderate drinking under 20 g a day.[51]

The old idea that a drink a day protects the heart has largely collapsed; much of the apparent benefit came from counting sick ex-drinkers as non-drinkers. Genetic studies, which sidestep that problem, find alcohol "not protective for any disease outcome", with each extra daily drink linked to about 2.65 mm Hg higher systolic blood pressure.[52] A second genetic study found higher risk of high blood pressure and coronary artery disease, rising more steeply at higher intakes.[53]

Whether moderate drinking changes overall death rates is still genuinely disputed; the cancer risk is not part of that dispute. The Global Burden of Disease collaboration first put the healthiest level at zero, and named alcohol the leading risk factor for death and disability at ages 15–49,[54] then in 2022 kept it near zero under 40 but found a J-shaped curve in older adults, varying by region — a statement about an age group, not a promise to any individual.[55] A 2025 US National Academies review concluded with moderate certainty that moderate drinking is associated with lower all-cause mortality, while finding breast-cancer risk rising about 5% for every 10–14 g a day at every level of intake.[56] A competing US federal review the same year put meaningful risk above roughly six to nine drinks a week,[57] WHO Europe holds there is no safe level, and the 2025–2030 US Dietary Guidelines dropped numeric limits in favour of simply advising people to drink less.[58] Wine is no exception: same risks, no special benefit despite the resveratrol marketing. See Alcohol.

Fats: trans fats are bad, seed oils are not the villain

Industrial trans fats carry the clearest harm signal of any dietary fat. Each 1% of energy from trans fat tracks with roughly 6% higher all-cause and cardiovascular mortality.[59] Restrictions are paying off: between 1990 and 2021, the age-adjusted rate of deaths blamed on trans fat among adults aged 55 and over fell by about 69%, most steeply in rich countries that banned it,[60] but WHO's 2024 report found more than 4 billion people still lived without best-practice trans-fat policies,[61] and WHO attributes more than 278,000 deaths a year to industrial trans fat.[62] Where it is not regulated, check labels for "partially hydrogenated oil". The natural trans fats in dairy and meat show no link with heart disease at normal intakes — no evidence of harm, though not proof of none.[63]

The claim that seed oils are toxic does not hold up. Their main fat, linoleic acid (an omega-6), is often called inflammatory, but the heart data point the other way: across 30 cohorts, people with the most of it in their blood or fat tissue had about 22% lower cardiovascular mortality than those with the least.[64] A 2025 review pooling 150 cohorts found more omega-6 in the diet linked to about 12% lower death rates, though the diet data showed no link either way to heart disease or cancer overall, and one signal ran the other way: a 15% higher risk of endometrial and ovarian cancer. These are observational links, not proof of benefit.[65]

The plausible real problem is oil reused for hours in commercial fryers: in one small human test, a single meal made with commercial fryer fat cut blood-vessel dilation from 5.9% to 0.8% within hours — but that was ten men and one meal, with no long-term trial.[66] Otherwise, the harm lies mostly in the ultra-processed food that hides refined oil alongside sugar, refined flour and additives, not in the oil itself.

In practice:

  • Extra-virgin olive oil has the best evidence; make it the default for cooking and dressing.
  • Avocado and nut oils are fine.
  • Canola, sunflower and safflower are neutral in normal use.
  • Avoid food from repeated commercial deep-frying — fast-food fryers, not your home pan.

Coconut oil is the mirror-image myth, sold as a health food. Compared with non-tropical vegetable oils it raises LDL cholesterol by about 0.27 mmol/L (10.5 mg/dL) — an average over trials that disagreed considerably — with no weight or blood-sugar benefit.[67] It is not uniquely toxic either; use it sparingly. More in Dietary fats.

Smaller worries: refined grains, cheese, cooking and contaminants

Refined grains. The message is to limit white bread, sugary cereals and potato products, not to avoid carbohydrates. Glycemic load — how much carbohydrate a food delivers and how fast it raises blood sugar — is the better measure. High glycemic load tracks with about 5% higher coronary heart disease risk per 50-unit increment,[68] though a dedicated analysis found no link with death from any cause or from cardiovascular disease.[69] The largest pooling of mega-cohorts found about 8% higher all-cause mortality and more diabetes-related cancer with high glycemic index, and about 15% higher type 2 diabetes and cardiovascular disease with high glycemic load.[70] In one of those cohorts, the link between high glycemic index and cardiovascular events was steeper in people who already had cardiovascular disease.[71] See Glycemic index; whether refined grains are harmful in themselves or just crowd out whole grains is covered on Whole grains.

Cheese. The brain-focused MIND diet says under one serving a week, but its authors chose that from literature reviews, not from a cheese effect in their own cohort.[72] Pooled cohorts show no excess stroke risk with cheese,[73] and in trials 135 g a day of hard cheese lowered LDL cholesterol compared with the same fat eaten as butter.[74] One to three servings a week fits a Mediterranean pattern; large daily amounts, especially of processed cheese with added oils and salt, are worth limiting.

Cooking compounds. The evidence is mostly from animals and lab work. Acrylamide forms in fries, crisps, coffee and bread crust and is a probable carcinogen, but human studies are largely null, apart from lung cancer in smokers.[75] The loose ends are in women: higher ovarian and endometrial cancer risk in never-smokers, and in premenopausal women breast-cancer risk rising from about 20 µg a day, roughly the average intake.[76] Chemicals formed when meat is charred or grilled (heterocyclic amines and polycyclic aromatic hydrocarbons) are a plausible part of the meat–colorectal link; avoid char and direct flame. Diets low in advanced glycation end-products, which form in dry high-heat cooking, modestly improved insulin resistance and LDL cholesterol in thirteen small trials, mostly in people with type 2 diabetes or obesity, with no hard outcomes.[77] A reason to favour steaming, boiling and stewing, not a proven longevity lever.

Contaminants. The rule is minimise and diversify, not avoid — fish and whole grains are net beneficial.

  • Mercury concentrates in large predatory fish: swordfish, shark, king mackerel, marlin, orange roughy, Gulf of Mexico tilefish, bigeye tuna (see Fish and seafood). Eating fish one to two times a week is linked to about 36% lower coronary death, so choose species rather than skipping fish. Pregnant or breastfeeding women and young children are most sensitive, and adults eating fish five or more times a week should also favour lower-mercury species.[78]
  • Arsenic builds up in rice, about ten times more than in other grains, and more in brown than white. It is a Group 1 carcinogen with no defined safe level, though the added risk at low exposure is modest. Rinse rice, cook it in plenty of water (about six parts to one) and drain, which removes a substantial share at the cost of some minerals and, in enriched rice, some added B vitamins; and vary your grains. Infant rice cereal is the sensitive case — the US action level is 100 µg/kg.
  • Cadmium is most concentrated in shellfish and organ meats, but most of it comes from staple cereals, potatoes and vegetables, so vary the staples. Aflatoxins, from badly stored nuts, grains and corn, cause liver cancer and act together with hepatitis B and C infection.[79]

What's not on the list

Eggs. Up to about one a day is not associated with cardiovascular disease for most adults.[80] US cohorts show a small positive signal and Asian cohorts none,[81] and dietary cholesterol does nudge LDL up slightly.[82] An earlier meta-analysis linked an egg a day or more to about 42% more type 2 diabetes in Western cohorts — observational, and its authors caution it may not justify changing advice.[83] People with type 2 diabetes may want a lower cap: in that group each extra daily egg carried about 25% higher cardiovascular risk, just short of statistical significance.[84]

Saturated fat from whole foods, in moderation — yogurt, cheese, eggs, occasional butter — whole-food sources, which are not the same as saturated fat from processed foods. Trials of cutting saturated fat found no effect on cardiovascular or overall death, across nine trials and 13,532 people.[85] Where cutting it helped, it was on combined cardiovascular events (heart attacks, strokes and similar) rather than deaths: pooled trials found about 17% fewer such events overall, with the benefit seen where saturated fat was replaced by polyunsaturated fat or starchy foods, though death rates barely moved;[86] in the trials the American Heart Association pooled, swapping it for polyunsaturated vegetable oil cut cardiovascular disease by about 30%, while swapping it for refined carbohydrate and sugar did not.[87]

Coffee and tea have generally favourable data.

Gluten, dairy, legumes and nightshades. For adults without a confirmed allergy, coeliac disease or specific sensitivity, there is no evidence of harm — an absence of evidence rather than proof of safety, and for nightshades there is essentially no research either way.

The 80/20 rule

The goal is the pattern, not perfection. Get 80% or more of your food from whole or minimally processed foods — vegetables, legumes, fish, nuts, fruit, whole grains, olive oil, fermented dairy — and keep the categories on this page under 20%. That split is a practical target, not a measured threshold; no analysis has found the benefit levelling off at a fifth. The reason not to push for 95/5 is that a pattern you keep beats a stricter one you abandon.

Further reading

  • Lane MM et al. Ultra-processed food exposure and adverse health outcomes — umbrella review of epidemiological meta-analyses. BMJ 2024.[88]
  • Hall KD et al. Ultra-processed diets cause excess calorie intake and weight gain — NIH metabolic-ward trial. Cell Metab 2019.[89]
  • Fang Z et al. Association of ultra-processed food consumption with all-cause and cause-specific mortality — population-based cohort study. BMJ 2024.[90]
  • Malik VS, Hu FB. Sugar-sweetened beverages and cardiometabolic health — pooled analyses. Circulation 2019.[91]
  • Witkowski M et al. The artificial sweetener erythritol and cardiovascular event risk. Nature Medicine 2023.[92]
  • Bouvard V et al. Carcinogenicity of consumption of red and processed meat (IARC). Lancet Oncol 2015.[93]
  • Shi W et al. Red meat consumption, cardiovascular diseases, and diabetes — systematic review and meta-analysis. European Heart Journal 2023.[94]
  • Lescinsky H et al. Burden of Proof — health effects of unprocessed red meat consumption. Nature Medicine 2022.[95]
  • Sacks FM et al. Effects on blood pressure of reduced dietary sodium and the DASH diet (DASH-Sodium). NEJM 2001.[96]
  • Bagnardi V et al. Alcohol consumption and site-specific cancer risk — comprehensive dose-response meta-analysis. British Journal of Cancer 2015.[97]
  • Marklund M et al. Biomarkers of dietary omega-6 fatty acids and incident cardiovascular disease and mortality. Circulation 2019.[98]
  • Sacks FM et al. Dietary fats and cardiovascular disease — AHA Presidential Advisory. Circulation 2017.[99]
  • Jenkins DJA et al. Association of glycaemic index and glycaemic load with type 2 diabetes, cardiovascular disease, cancer, and all-cause mortality — meta-analysis of mega cohorts of more than 100,000 participants. Lancet Diabetes & Endocrinology 2024.[100]
  • Zhao B et al. Associations of dietary cholesterol, serum cholesterol, and egg consumption with mortality. Circulation 2022.[101]

For the deep dives that sit underneath this survey:

  • Red and processed meat — Neu5Gc, TMAO, cooking-method mitigation, NutriRECS, dementia, APOE4
  • Dietary fats — trans, saturated, PUFA, MUFA, eggs, omega-6:omega-3 ratio, epigenetic-clock signals
  • Ultra-processed food — NOVA, the UCL crossover, biological-age effects
  • Sweeteners — sugar, fructose, high-fructose corn syrup
  • Sugar substitutes — non-sugar sweeteners, erythritol/xylitol, allulose
  • Alcohol — Mendelian randomization, J-curve collapse

— § —