Coffee and Longevity

Studies on Coffee and Lifespan

Most longevity research follows groups of people over time, records their coffee habits and compares subsequent deaths. These are prospective cohort studies: researchers observe existing behaviour rather than randomly assigning people to drink coffee.

That distinction matters. A cohort can establish that coffee drinking came before a recorded death, but it cannot make coffee drinkers and non-drinkers otherwise identical. Our coffee and health guide explains how this evidence fits alongside trials and other research.

What the major studies found

In 2015, Ding and colleagues analysed three large US cohorts. Consumption of 1 to 5 cups a day was associated with lower mortality than drinking none; above 5 cups, the overall association was not statistically significant. Among people who had never smoked, however, the pattern was more consistently inverse, illustrating how smoking can change the apparent relationship.

A 2018 UK Biobank analysis by Loftfield and colleagues included 498,134 participants. Lower mortality was associated with drinking ground, instant and decaffeinated coffee. Associations were also seen across genetic scores related to caffeine metabolism, although this remained observational research, not a genetic experiment proving that coffee extends life.

Kim and colleagues’ 2019 meta-analysis combined 40 studies involving approximately 3.85 million participants. A meta-analysis pools results statistically. Its lowest estimated all-cause mortality risk occurred at about 3.5 cups a day, with a relative risk of 0.85 compared with no coffee consumption.

That means approximately 15% lower relative risk, not 15% longer life.

More recent work has examined drinking patterns. A 2025 study of 40,725 US adults found that morning coffee drinking was associated with lower all-cause mortality than non-drinking, whereas the all-day pattern was not significantly associated with lower mortality. It did not establish that moving a coffee to the morning prolongs life.

Evidence summary

The strength descriptions below are editorial assessments of what each finding supports, not formal evidence grades.

What U-Shaped Associations Really Mean

A U-shaped association describes a curve in which the measured risk falls as consumption increases from none to a moderate amount, then rises again at higher consumption. A J-shaped curve is similar but asymmetrical. These labels describe the results, not a biological explanation.

Crucially, the rising arm may remain below the risk recorded among non-drinkers. “Less favourable than moderate consumption” does not automatically mean “more dangerous than drinking none”. Some studies instead show a flattening curve, with little additional association beyond moderate intake.

There is no universal coffee curve. Ding’s results changed when the analysis was restricted to never-smokers, while the UK Biobank analysis found inverse associations even in its highest consumption category. Differences in participants, measurements and statistical adjustment can produce different shapes.

Nor does a mathematical minimum identify your ideal intake. The 3.5-cup estimate in Kim’s meta-analysis is a point on a fitted population curve, not a prescription precise to half a cup.

Study cups are not standard caffeine doses. Serving sizes, preparation and strength differ. Treat cup counts as descriptions of the research, not instructions to fill a particular mug several times.

Confounding: Why Coffee May Get Too Much Credit or Blame

Confounding happens when another factor is related both to coffee drinking and to the outcome being studied. It can make coffee appear beneficial or harmful even when some of that relationship comes from elsewhere.

Smoking and other habits

Smoking is especially important because it has often accompanied heavier coffee consumption. Ding’s never-smoker analysis demonstrates why simply comparing all coffee drinkers with all non-drinkers can mislead.

Researchers also account for factors such as alcohol, physical activity, diet and socioeconomic circumstances. Statistical adjustment helps, but only as far as those factors have been measured accurately. A questionnaire cannot capture every relevant difference between people.

Any remaining influence is called residual confounding. A very large sample makes estimates more precise, but does not automatically remove it.

Illness can change coffee drinking

The direction of influence can also run backwards. Someone developing an illness may reduce or stop coffee before the illness is diagnosed. If that person subsequently dies, their low intake may appear to predict death even though declining health caused the change in intake.

This is reverse causation. Researchers can investigate it by excluding early deaths or participants with known disease, but early, undiagnosed illness remains difficult to rule out completely.

Coffee habits are measured imperfectly

Self-reported intake may not represent years of drinking habits. In the 2025 life-expectancy analysis by Yan and colleagues, coffee intake came from short dietary recalls collected at baseline. The authors acknowledged that these may not adequately capture long-term consumption.

Habits can also change during follow-up. Measurements that miss those changes can place participants in categories that no longer describe their behaviour.

Finally, the number of papers is not the number of independent populations. The two 2025 analyses discussed here both used US National Health and Nutrition Examination Survey data. They ask different questions, but should not be mistaken for wholly separate replications.

What Can We Conclude About Coffee and Longevity?

The defensible conclusion is that coffee consumption is repeatedly associated with lower mortality, particularly at moderate intakes. Establishing how much, if any, of that association coffee causes is harder.

A 2017 umbrella review, which brought together existing meta-analyses, called for robust randomised trials to clarify causality. Pooling observational studies does not remove the limitations of their underlying designs.

A plausible mechanism is not proof

Coffee contains compounds other than caffeine. Researchers have proposed effects involving inflammation and insulin sensitivity as possible explanations for favourable associations. These remain possible mechanisms, not a demonstrated chain from drinking coffee to living longer.

Decaffeinated findings are useful here: they make a caffeine-only explanation less convincing. However, decaf drinkers also have habits and health histories that can influence results. For the chemistry rather than the lifespan question, see antioxidants in coffee.

Likewise, evidence concerning a particular disease cannot simply be converted into a claim about extra years of life. Coffee’s role in disease examines those condition-specific findings separately.

Mortality risk is not years gained

Most cohort studies compare the rate or risk of death during a defined follow-up period. They do not observe everyone’s complete lifespan.

Some researchers go further by modelling life expectancy. Yan and colleagues’ 2025 analysis used life tables and adjusted mortality estimates to calculate differences between consumption groups. Those are modelled estimates based on observational associations, not years demonstrably added by prescribing coffee.

A useful reading rule is to ask: was lifespan measured, estimated or merely implied by the headline?

Relative figures also need an absolute context. As a purely hypothetical illustration, reducing a risk from 10% to 8.5% is a 15% relative reduction, but a 1.5 percentage-point absolute difference. That arithmetic is not a prediction for an individual coffee drinker, and it does not establish causation.

How to Use the Evidence Without Chasing a Dose

For someone who already enjoys coffee, these findings offer reassurance rather than a target. There is no established longevity dose to pursue, and the observational evidence does not justify starting coffee solely to extend life.

Keep the distinction between coffee consumption and caffeine safety clear:

• Count caffeine from all sources. EFSA’s 2015 assessment found that up to 400 mg a day, spread throughout the day, raised no safety concerns for healthy, non-pregnant adults.
• Consider individual servings. EFSA found that single doses up to 200 mg raised no safety concerns for most healthy adults.
• Protect sleep. Even 100 mg close to bedtime can affect sleep in some adults.
• Apply pregnancy guidance separately. The NHS advises no more than 200 mg a day from all sources during pregnancy.

These are population safety benchmarks, not recommended intakes or guarantees of personal tolerance. They do not override symptoms or advice relating to medicines and medical conditions.

If coffee causes unwanted effects, a favourable population association is not a reason to persist. Consult a GP or pharmacist about personal concerns, or a midwife during pregnancy. Coffee and health risks covers the cautions in more detail.

The practical value of coffee and longevity research is therefore modest but useful: it helps put an enjoyable habit in perspective without turning it into treatment.