Coffee and health: what the large cohort studies show

The largest studies of coffee drinkers point in the same direction. In cohorts of hundreds of thousands of adults in the United States, Europe, and the United Kingdom, people who drank two to five cups a day died at slightly lower rates over the following decade or more than people who drank none. The pattern holds for decaffeinated coffee, for instant coffee, and for people whose genes make them clear caffeine slowly. It also holds after accounting for the fact that coffee drinkers smoke more than non-drinkers.

The limits of this evidence matter as much as the findings. All of these are observational studies. They follow people who chose to drink coffee and compare them with people who chose not to, and the two groups differ in ways that cannot be fully measured. No large randomized trial has assigned people to drink coffee for years and counted deaths. The differences seen are also modest, in the range of 10 to 15 percent lower relative risk, and the most careful reviews describe coffee as more likely to help than harm rather than as a proven benefit.

This article is for adults who drink coffee and want to know whether the evidence gives them any reason to cut back, and for people who have been told that coffee is good for them and want to know how solid that claim is.

What a cohort study can and cannot show

A cohort study enrolls a large group of people, records their habits at the start, and follows them for years to see who develops disease or dies. The design is good at finding associations in real populations, at a scale no trial can match. Its weakness is confounding, where the habit under study travels with other habits that drive the outcome. Coffee is a clear example. In the NIH-AARP Diet and Health Study, which followed 402,260 American adults aged 50 to 71 from 1995 to 2008, coffee drinkers at first appeared to die at higher rates. They were also much more likely to smoke. After adjustment for smoking and other factors the direction reversed, and Freedman et al., 2012 reported that men drinking four or five cups a day had a 12 percent lower death rate than non-drinkers, and women a 16 percent lower rate. The authors wrote that whether this was a causal finding could not be determined from their data. That caution applies to every study described below.

What the largest cohorts found

The European Prospective Investigation into Cancer and Nutrition followed 521,330 people in 10 countries for an average of 16 years, during which 41,693 died. Gunter et al., 2017 found that people in the top quarter of coffee intake had 12 percent lower all-cause mortality among men and 7 percent lower among women than non-drinkers, with the clearest reductions in deaths from digestive diseases. The relationship did not differ between countries with very different coffee habits. In a subgroup with blood samples, higher coffee intake went with lower liver enzyme levels, which fits the lower rates of liver disease seen among coffee drinkers in other studies.

The UK Biobank added two things the earlier studies could not. Loftfield et al., 2018 followed 498,134 British adults for 10 years. Drinking two to three cups a day went with a 12 percent lower death rate, and the lower risk was still present at eight or more cups a day. Ground, instant, and decaffeinated coffee showed similar associations. The study also scored participants on genetic variants that speed or slow caffeine metabolism, and found no difference in the coffee and mortality link between fast and slow metabolizers. Both points suggest that if coffee does something protective, caffeine is probably not the main reason.

What the pooled reviews say

An umbrella review in the BMJ gathered 201 meta-analyses of observational studies covering 67 health outcomes, along with 17 meta-analyses of trials. Poole et al., 2017 found that coffee was associated with benefit more often than harm, with the largest relative reductions at three to four cups a day compared with none: 17 percent lower all-cause mortality, 19 percent lower cardiovascular mortality, and 15 percent lower cardiovascular disease. High versus low intake was linked to an 18 percent lower risk of cancer overall. Associations with harm mostly disappeared once studies adjusted properly for smoking. The exceptions were in pregnancy, where high intake was linked to low birth weight, preterm birth, and pregnancy loss, and a higher fracture risk in women but not men. The authors concluded that coffee seems generally safe at usual intakes and that randomized trials are needed to know whether any of the associations are causal.

Whether brewing method and sugar matter

Two large cohorts have looked at how coffee is prepared. In Norway, where boiled coffee was common, Tverdal et al., 2020 followed 508,747 adults for an average of 20 years. Filtered coffee was associated with 15 percent lower mortality than no coffee in both sexes, while unfiltered coffee was associated with a smaller or no reduction, and in men aged 60 and over with a borderline higher cardiovascular death rate. Unfiltered coffee contains compounds that raise cholesterol and that paper filters remove, and the gap between brewing methods widened when the authors left cholesterol out of their models, which supports that explanation.

Sugar is a separate question. Liu et al., 2022 used UK Biobank diet records from 171,616 adults followed for seven years. Unsweetened coffee at one and a half to four and a half cups a day was associated with 16 to 29 percent lower mortality. Coffee with added sugar showed a similar pattern at moderate intakes but not at the highest, and the results for artificial sweeteners were inconsistent. The follow-up was short and the number of deaths small, so these estimates are less stable than those from the larger mortality studies.

How much is reasonable and who should be careful

The cohort data suggest that the lowest risk sits somewhere between two and five cups a day, with no clear extra benefit above that and no clear harm up to about eight cups in the UK Biobank. The practical ceiling is usually set by caffeine rather than by coffee itself. The US Food and Drug Administration cites 400 milligrams of caffeine a day as an amount not generally associated with harmful effects in healthy adults, and lists a 12-ounce cup of brewed coffee at 113 to 247 milligrams, so that ceiling is reached at about two to four cups depending on strength. Decaffeinated coffee still contains 2 to 15 milligrams per cup.

Pregnancy is the main exception in the evidence. The NHS advises no more than 200 milligrams of caffeine a day during pregnancy, roughly two cups of instant coffee, because higher intakes have been linked to miscarriage and stillbirth. People with heart rhythm problems, uncontrolled high blood pressure, anxiety disorders, reflux, or trouble sleeping, and anyone taking medicines that interact with caffeine, should ask their doctor what amount is sensible for them. Caffeine sensitivity varies widely between individuals, and a dose that one person does not notice can leave another jittery or awake at night.

What the evidence adds up to

For most adults, the large cohort studies give no reason to stop drinking coffee and some reason to think moderate intake is harmless or mildly beneficial. They do not show that coffee prevents disease, because the people who drink it differ from those who do not in ways that no statistical adjustment fully captures, and because no trial has tested the question. Starting to drink coffee in order to live longer is not supported by this evidence. Continuing to drink it, filtered and without much sugar, at an amount that does not disturb sleep, is consistent with it.

Sources

• Association of Coffee Drinking with Total and Cause-Specific Mortality (Freedman et al., New England Journal of Medicine, 2012) — prospective cohort, NIH-AARP Diet and Health Study, 402,260 US adults aged 50 to 71, followed 1995 to 2008
• Coffee Drinking and Mortality in 10 European Countries: A Multinational Cohort Study (Gunter et al., Annals of Internal Medicine, 2017) — prospective cohort, EPIC, 521,330 participants, mean follow-up 16.4 years, 41,693 deaths
• Association of Coffee Drinking With Mortality by Genetic Variation in Caffeine Metabolism: Findings From the UK Biobank (Loftfield et al., JAMA Internal Medicine, 2018) — prospective cohort, 498,134 UK adults, 10 years of follow-up, stratified by genetic caffeine metabolism score
• Coffee consumption and health: umbrella review of meta-analyses of multiple health outcomes (Poole et al., BMJ, 2017) — umbrella review of 201 meta-analyses of observational studies and 17 of interventional studies, 67 health outcomes
• Coffee consumption and mortality from cardiovascular diseases and total mortality: Does the brewing method matter? (Tverdal et al., European Journal of Preventive Cardiology, 2020) — prospective cohort, 508,747 Norwegian adults aged 20 to 79, average 20 years of follow-up, filtered vs unfiltered coffee
• Association of Sugar-Sweetened, Artificially Sweetened, and Unsweetened Coffee Consumption With All-Cause and Cause-Specific Mortality (Liu et al., Annals of Internal Medicine, 2022) — prospective cohort, UK Biobank, 171,616 adults, median 7 years of follow-up, 3,177 deaths
• Spilling the Beans: How Much Caffeine Is Too Much? (US Food and Drug Administration) — consumer guidance on caffeine amounts and the 400 milligram daily figure for healthy adults
• Foods to avoid in pregnancy (NHS) — public health guidance including the 200 milligram daily caffeine limit in pregnancy