Coffee
People who drink two to four cups of coffee a day die a little less often and get less type 2 diabetes, liver disease and Parkinson's than people who drink none. Almost all of that comes from watching people rather than testing coffee on them, and genetic studies often fail to back it up — but the pattern is unusually consistent, and decaf shares most of it, except possibly for the brain.
If you already enjoy coffee, two to four cups a day is a sound habit: paper-filtered if your LDL cholesterol is high, black or with a little milk rather than heavily sweetened, cooled a few minutes before drinking, and with big doses kept well away from bedtime. Stay under 400 mg of caffeine a day, or 200 mg in pregnancy. Don't take up coffee for your health if you don't like it.
Why the longevity numbers deserve both respect and doubt
The headline association is real and remarkably consistent. Pooling 40 cohort studies of roughly 3.85 million adults, risk of death was lowest at about 3.5 cups a day, about 15% lower at that point.[1] An umbrella review of 201 meta-analyses put the low point at three cups.[2] In UK Biobank, the large British health study, the lowest mortality sat at two to three cups a day for ground, instant and decaf coffee alike,[3] and unsweetened coffee tracked with up to about 29% lower mortality.[4] The curve is U-shaped: benefit flattens and fades at high intakes.
The comparison group is not neutral. "Non-drinkers" include people who stopped because they became ill. When the same British data was benchmarked against people drinking one to two cups of caffeinated coffee instead, both decaf drinkers and non-drinkers had slightly higher odds of cardiovascular disease, not equal ones.[5] In the same data, more than six cups a day went with about 22% higher odds of cardiovascular disease. Healthy-user bias, illness making people cut back, and leftover confounding by smoking all inflate apparent benefit.
Genetic studies often disagree. Mendelian randomisation uses inherited gene variants, fixed at conception, as a natural experiment to test cause and effect. Applied to coffee, 59 such studies now exist, and they are less consistent than the cohorts.[6] One testing 12 coffee-linked variants against 15 heart and circulation outcomes found no association with any of them, and suggested earlier observational studies "may have been confounded."[7] Others point the wrong way, with probable evidence for higher risk of Alzheimer's and type 2 diabetes.
Some of the research is industry-funded. Coffee-industry bodies fund or publicise part of the favourable literature. That doesn't make the big independent cohorts wrong, but it's a reason to trust them over industry summaries.
So coffee is linked to neutral or favourable outcomes almost everywhere, but not proven to extend life — and the main reviews frame their conclusions as applying to existing drinkers.
The heart: reassuring, with two caveats
Across 1.28 million adults, cardiovascular risk was lowest at three to five cups a day.[8] A larger 2026 pool was more mixed: comparing the heaviest with the lightest drinkers, it found lower risk of stroke and arrhythmia but nothing for coronary heart disease or heart failure.[9] The heart-failure benefit is modest at best — about 7% lower risk at two to four cups, on evidence the authors graded low-certainty, with a hint of higher risk at the top of the range.[10] Stroke risk was roughly 14–17% lower at two to four cups, fading back to no association by eight; the authors call the link "weakly" inverse.[11]
Heart rhythm is the one area with real trials, and they point in different directions. In the first, the DECAF trial, 200 current or recent coffee drinkers with persistent atrial fibrillation or flutter, after their rhythm had been reset, either drank at least a cup a day or gave up caffeine for six months. The irregular rhythm came back in 47% of the coffee group against 64% of the abstainers.[12] A reanalysis warns that a trial this size tends to overstate effects: the direction is convincing, the size much less so.[13] It was small, open-label and only in current or recent drinkers, so it suggests that quitting doesn't help — not that non-drinkers should start. US heart-rhythm guidelines had already concluded that caffeine abstention is of no benefit for preventing episodes, though it may still ease symptoms in people who recognise caffeine as a trigger.[14] Large cohorts agree that coffee doesn't raise atrial fibrillation risk.[15]
The second trial is the counterweight. One hundred adults wore continuous heart monitors and, on randomly assigned days, drank coffee or avoided caffeine. Extra beats from the upper chambers didn't change, but extra beats from the lower chambers (ventricles) rose by about half, mostly above one cup a day, and people slept about 36 minutes less on coffee days.[16] The ventricular finding was a secondary result and should be held loosely; the sleep loss was also secondary, but sizeable, and fits the sleep trials below.
Blood pressure rises for hours, but doesn't seem to stay raised for years. In people who already have hypertension, 200–300 mg of caffeine raised blood pressure by roughly 8 mmHg systolic and 6 diastolic for at least three hours.[17] Daily use only partly blunts it: in healthy adults taking caffeine every day, about half lost the response and the other half kept it.[18] Over the long run, 13 cohorts of about 315,000 adults found no meaningful link between coffee and developing hypertension, though the six better-quality studies pointed towards about 17% higher risk, short of statistical significance.[19] If your hypertension is severe or poorly controlled, keep coffee moderate, and don't drink it shortly before a blood-pressure reading.
Diabetes and the liver: the strongest signals
Type 2 diabetes is among coffee's most consistent associations. Across 28 studies of over 1.1 million adults, the heaviest drinkers had about 30% lower risk than the lightest, and decaf did almost as well per cup as caffeinated — a sign that something other than caffeine is doing much of the work.[20] What goes in the cup matters more here than it does for mortality: each daily cup of plain coffee went with about 10% lower risk, roughly halved when sugar was added and weakened when artificial sweetener was, while cream made no difference.[21] There are no trials with hard endpoints.
The liver shows one of the most consistent dose-response patterns in nutrition research. Each extra two cups a day went with about 35% lower risk of the main form of liver cancer — 29% in the prospective cohorts alone, with the authors grading their evidence "very low" for lack of trials.[22] High intake was linked to about 65% lower risk of chronic liver disease, including cirrhosis.[23] The main worry is reverse causation, since people with liver disease often cut back. In fatty liver disease, coffee drinkers had about 35% lower odds of significant scarring, but no lower chance of developing fatty liver in the first place — and that scarring figure comes from snapshot studies.[24] The American liver society goes furthest, saying three or more cups a day "could be recommended" for fatty liver in the absence of contraindications.[25]
The brain: where caffeine itself seems to matter
Parkinson's disease has the most consistent inverse association in the coffee literature: risk was about 28% lower at around three cups a day.[26] It fits biology — caffeine blocks the same brain receptor (adenosine A2A) as several experimental Parkinson's drugs. But genetic studies found no causal effect on risk,[27] though the authors of one say they may simply have lacked the statistical power, and one found a link with a later age at onset instead.[28]
Dementia is genuinely contested. A 2026 analysis of two large US health-professional cohorts found two to three cups a day of caffeinated coffee tracking with 18% lower dementia risk.[29] But a meta-analysis of ten cohorts published weeks earlier concluded moderate coffee "does not seem to affect dementia risk," while three or more cups a day "could increase risk."[30] See dementia prevention for the wider picture.
Decaf is the exception here. In the JAMA analysis, decaf showed no dementia benefit, and heavier decaf drinkers reported slightly more cognitive decline and scored slightly worse on verbal memory. In UK Biobank, three or more cups a day of unsweetened caffeinated coffee went with lower dementia and Parkinson's risk, while decaf showed no significant link with either.[31] The decaf signal may partly reflect people switching once sleep or health is already a problem, but "decaf is neutral for the brain" overstates it. For almost everything else — mortality, diabetes, liver — decaf is linked with nearly the same benefit as regular in cohort studies.
Depression risk was about 24% lower in the heaviest drinkers, peaking around two cups a day,[32] though a newer estimate is about half that and only just clears chance.[33] Depression itself disrupts routines like coffee, so cause and effect may run backwards.
Cancer: the old scares didn't hold, and one real risk is the temperature
The World Health Organization's cancer agency reviewed roughly 1,000 studies in 2016 and downgraded coffee from "possibly carcinogenic" to "not classifiable."[34] Coffee goes with lower risk of liver and endometrial cancer — about 13% lower for endometrial overall, more at over four cups a day.[35] Prostate and pancreatic cancer show no clear effect. Breast cancer shows only a small, unsettled dip.[36]
The old bladder- and lung-cancer scares were smoking in disguise: smokers drink more coffee. The bladder link vanished after adjusting for smoking and among never-smokers,[37] and the lung link disappeared among never-smokers and once body weight was accounted for.[38]
Acrylamide, formed when coffee is roasted, causes cancer in animals, and Europe's food-safety authority considers the margin between typical intake and the animal-tumour dose uncomfortably narrow — while adding that human studies have not shown it to cause cancer in people.[39] A meta-analysis of 1.15 million people found no link at any site except lung cancer in smokers,[40] though small endometrial and ovarian signals in never-smokers "cannot be excluded."[41] Not a reason to avoid coffee.
Very hot drinks are the real one. The WHO's cancer agency rates drinking anything above 65 °C as probably carcinogenic, on limited evidence for oesophageal cancer — it is the heat, not the coffee.[42] The risk concentrates in people who habitually drink very hot drinks, but a study that measured drinking temperature directly saw it start at about 60 °C (see tea), and typical drinking temperatures sit right at that line. Let coffee cool below about 60 °C — roughly six minutes for a fresh cup — especially if you drink several a day.
How you brew it and what you put in it
Unfiltered coffee raises LDL cholesterol. Coffee oil contains cafestol, which blunts the liver's main route for clearing cholesterol. Across six randomised trials, about six cups a day of unfiltered coffee raised LDL by about 0.39 mmol/L (15 mg/dL) compared with filtered, while paper-filtered coffee didn't raise it at all.[43] French press, Turkish, boiled and moka-pot coffee keep the oil, and espresso can carry the most of all, though it varies enormously from sample to sample; office self-serve machines sit between paper-filtered and boiled, so don't count them as filtered.[44] In a 20-year Norwegian study of over 500,000 people, filtered coffee had a better heart-death profile than unfiltered, and than no coffee.[45] If your LDL is high or heart disease runs in your family, use a paper filter. See cholesterol.
Cold brew is a taste choice, not an upgrade. Hot brewing extracts more antioxidants,[46] the "low-acid" claim is mostly perception, and long steeps can carry more caffeine than you expect. Going by extraction chemistry, cold brew probably carries little cafestol, which helps if your LDL is high.[47] No study links either method to a health outcome.
Sugar matters most for diabetes. In UK Biobank, about a teaspoon of sugar per cup tracked with much the same mortality benefit as black coffee,[48] but in three large US cohorts, adding sugar roughly halved coffee's link with lower diabetes risk, and artificial sweetener weakened it too.[49] Heavily sweetened café drinks, which can carry 200–500 kcal, probably cancel the benefit. A little milk or cream appears harmless. Butter-and-oil "bulletproof" coffee has no demonstrated benefit and adds a lot of saturated fat.
Timing: sleep, mornings and iron
The cut-off depends on the dose. Caffeine is cleared by one liver enzyme whose speed varies several-fold between people: half-life is usually 3–6 hours but ranges from about 2 to 10. In one trial, 400 mg taken six hours before bed still cut sleep by more than an hour, and people didn't notice.[50] In another, 100 mg (about one cup) had no measurable effect even four hours before bed, while 400 mg disturbed sleep when taken within 12 hours.[51] So a single afternoon cup is fine for most people, but a heavy total wants 10–12 hours before bed — stricter still if caffeine lingers in you or you sleep badly.
Keep coffee in the morning, but don't wait for cortisol. Drinking coffee mainly in the morning was associated with lower mortality than drinking it all day,[52] plausibly because it protects sleep. The popular rule to delay your first cup 60–90 minutes after waking rests on blog-level reasoning, not trials. Do it if you like it. See circadian rhythms, and sleep inertia for caffeine taken just before a nap.
Coffee before iron, not after. Coffee with a meal cut iron absorption by about 39% (tea: 64%). Coffee an hour before the meal caused no loss, but coffee an hour after inhibited it just as much.[53] This matters only if your iron is marginal — menstruating women, plant-based eaters, people on iron tablets. If you take iron, have the coffee first and the tablet an hour later. See iron and aging.
Who should be careful
Caffeine limits. Europe's food-safety authority and the US Food and Drug Administration both put the safe daily amount for healthy adults at 400 mg; the European authority adds no more than 200 mg in one go.[54] A brewed cup has 75–100 mg, an espresso 60–80 mg, decaf 2–15 mg. Above 400 mg, anxiety, insomnia, palpitations and stomach upset become more likely.
Pregnancy. Keep to 200 mg a day from all sources, including tea, cola, chocolate and energy drinks — about one to two cups. Each extra 100 mg a day was linked to about 14% higher risk of miscarriage, 19% of stillbirth, 10% of a small-for-dates baby and 7% of low birth weight, with no threshold below which the link disappeared, though the authors note bias could explain some of this.[55] The only trial evidence is too thin to settle it,[56] so 200 mg is a precaution, not a proven safe line. Caffeine clearance roughly halves late in pregnancy, and the fetus can barely clear it. Decaf keeps the ritual.
Breastfeeding. Up to 200–300 mg a day is generally considered compatible with nursing; infant fussiness and poor sleep have been reported at very high intakes, equivalent to about ten or more cups a day.[57] Newborns clear caffeine extremely slowly, with a half-life of about 100 hours,[58] and reach adult clearance only by 3–5 months.[59]
Medicines. Caffeine and several drugs share the same liver enzyme.
- Clozapine: coffee itself barely moves the drug level,[60] but smoking does, and coffee tracks with smoking. Tell your prescriber before you start, stop or cut down smoking.
- Fluvoxamine, ciprofloxacin, oral contraceptives, verapamil and cimetidine slow caffeine clearance, so the same cup hits harder. Theophylline adds to caffeine's stimulant effect.
- Levothyroxine and alendronate absorb less well with coffee — separate by 30–60 minutes.
- Adenosine heart stress tests: caffeine blocks the test drug and can make the scan miss a problem. Stop all caffeine for at least 12 hours beforehand, be wary even of decaf, and follow your centre's instructions.[61]
Children and teenagers. The American Academy of Pediatrics says caffeine and other stimulants in energy drinks have "no place" in the diet of children and teenagers.[62] The often-quoted limits of 100 mg a day for teenagers and 2.5 mg per kg for younger children come from a separate review.[63]
Anxiety and panic. Caffeine often worsens symptoms, and 400 mg or more can trigger panic attacks in susceptible people. A common gene variant in caffeine's target receptor makes some people far more anxious on it, regardless of how fast they clear it.[64]
Arrhythmias. If you have hard-to-control arrhythmias or an extra electrical pathway in the heart (Wolff-Parkinson-White syndrome), get individual advice — the reassuring trial covered atrial fibrillation and flutter only, and the other found more ventricular extra beats on coffee days.
Heavy drinkers' bones. At ordinary intakes, coffee drinkers had about 21% lower odds of osteoporosis.[65] But in women over 65, more than five cups a day may track with lower bone density, especially alongside more alcohol.[66] See bone density.
Quitting. Withdrawal gave about half of people a headache and about one in eight real distress, starting 12–24 hours after the last dose and lasting 2–9 days.[67] Taper instead: cut about 25% every 3–4 days.
Smaller questions, quick answers
- Genetic "slow metaboliser" tests. An early study linked four or more cups a day to heart attacks in slow metabolisers,[68] but that hasn't held up: in nearly 500,000 people the mortality benefit held regardless of genes.[69] Testing wouldn't change your advice.
- Mould toxins. Even at four cups a day, intake of the main toxin, ochratoxin A, was about 2% of the tolerable limit.[70]
- Reflux. One large cohort linked unsweetened coffee to less gut disease, reflux included;[71] a meta-analysis found reflux symptoms slightly more common in drinkers, of "unclear clinical significance."[72] If coffee gives you reflux, treat it as your trigger; there's no blanket rule.
- Gut bacteria. One gut bacterium of the kind linked to a healthy colon lining rises with coffee intake, decaf included.[73] Interesting; not a health outcome.
- Exercise. A real effect, shown in randomised trials: 3–6 mg of caffeine per kg about an hour beforehand improves endurance by roughly 2–3%.[74]
- A newly found longevity mechanism. A 2026 lab study found coffee compounds, caffeine included, binding a cellular stress sensor called NR4A1. It was cell work only, with no human outcome data, and one author's university coffee-research centre supplied the coffee.[75]
- "Five years younger" telomeres. That came from patients with severe mental illness, and doesn't transfer to healthy adults.[76]
Further reading
- Poole R et al. Coffee consumption and health: umbrella review. BMJ 2017.[77]
- van Dam RM, Hu FB, Willett WC. Coffee, Caffeine, and Health. NEJM 2020.[78]
- Dewland TA, van Dam RM, Marcus GM. Coffee, Caffeine, and CVD: State-of-the-Art Review. Eur Heart J 2025.[79]
- Wong CX, … Marcus GM. Caffeinated Coffee Consumption or Abstinence to Reduce Atrial Fibrillation: The DECAF Randomized Clinical Trial. JAMA 2026;335:317–325 (n=200, open-label).[80]
- Marcus GM et al. Acute Effects of Coffee Consumption on Health among Ambulatory Adults (CRAVE). NEJM 2023 (n=100, randomized case-crossover).[81]
- Liu D et al. Coffee with/without sugar and mortality. Ann Intern Med 2022.[82]
- Tverdal A et al. Coffee consumption and mortality from cardiovascular diseases and total mortality: Does the brewing method matter? Eur J Prev Cardiol 2020.[83]
- Zhang Y et al. Coffee and Tea Intake, Dementia Risk, and Cognitive Function. JAMA 2026 — 2–3 cups/day caffeinated coffee associated with 18% lower dementia risk; tea ~14%; decaf null.[84]
- Loomis D et al. Carcinogenicity of drinking coffee, mate, and very hot beverages. Lancet Oncol 2016.[85]
- EFSA Scientific Opinion on the safety of caffeine.[86]
- Greenwood DC et al. Caffeine intake during pregnancy and adverse birth outcomes: a systematic review and dose-response meta-analysis. Eur J Epidemiol 2014.[87]
- Gardiner CL et al. Dose and timing effects of caffeine on subsequent sleep: a randomized clinical crossover trial. Sleep 2025.[88]
- Morck TA, Lynch SR, Cook JD. Inhibition of food iron absorption by coffee. Am J Clin Nutr 1983.[89]
- Juliano LM, Griffiths RR. A critical review of caffeine withdrawal. Psychopharmacology 2004.[90]