Is Coffee Good for You? What the Evidence Actually Shows

For most healthy adults, moderate coffee drinking appears safe and is associated with a lower risk of several diseases, including type 2 diabetes, liver disease and early death. The benefits in the research peak at around three to four cups a day.

That is the short answer, and it is accurate. It also leaves out the most important thing about the evidence: almost all of it is observational, and the scientists who reviewed it rated most of it low or very low quality. Coffee is very probably not harming you. Whether it is actively helping is a harder question than headlines suggest.

The biggest review

The most comprehensive look at coffee and health is a 2017 umbrella review in the BMJ by Robin Poole and colleagues. Rather than run a new study, they gathered existing ones: 201 meta-analyses of observational research covering 67 health outcomes, and 17 meta-analyses of clinical trials covering 9.

Its main findings:

• Death from any cause. The largest reduction in relative risk was at three cups a day, about 17% lower than drinking none.
• Liver. The most striking associations. Drinking any coffee, compared with none, was linked to a 29% lower risk of non-alcoholic fatty liver disease and a 39% lower risk of cirrhosis.
• Type 2 diabetes. High consumption, compared with low, was linked to about 30% lower risk.
• Overall. Coffee was "more likely to benefit health than harm" at usual levels of intake, with the largest benefits at three to four cups a day.

The review also found associations with harm. Some — lung cancer, for example — largely disappeared once smoking was taken into account, because coffee drinkers in older studies were more likely to smoke. Two did not: risks in pregnancy, and fracture risk in women.

Why "associated with" matters

Every one of those numbers describes an association. People who drink coffee had these outcomes less often. That is not the same as coffee causing the difference, and the authors were careful to say so.

The review graded the quality of its evidence using the standard GRADE system. About a quarter was rated "low" and three-quarters "very low." That is not a criticism of the researchers; it is the nature of studying a habit you cannot easily randomise people into for thirty years.

Three problems run through observational coffee research:

• Confounding. Coffee drinkers differ from non-drinkers in many ways — smoking, alcohol, income, diet, work. Statistical adjustment catches some of this and misses some. The review specifically warned that leftover confounding by smoking can hide a benefit or exaggerate a harm.
• The sick-quitter effect. People who become unwell often give up coffee. That makes the non-drinking group look sicker for reasons that have nothing to do with coffee.
• Healthy-user effects. People who can comfortably drink several cups a day may be, on average, healthier to begin with.

None of this means the associations are false. The liver findings in particular are consistent across many studies. It means the honest reading is "safe, and possibly beneficial," not "a health drink."

It is not only the caffeine

One telling result comes from a 2022 study in the European Journal of Preventive Cardiology. David Chieng and colleagues followed 449,563 UK Biobank participants for around twelve and a half years and compared three kinds of coffee.

At two to three cups a day, all three were associated with lower all-cause mortality than drinking no coffee — ground coffee most strongly, but decaffeinated coffee too, with a hazard ratio of 0.86.

If the association held only for caffeinated coffee, caffeine would be the obvious suspect. That decaf shares it points to other compounds in coffee, such as its polyphenols, or to something about coffee drinkers themselves. The study cannot say which.

Who should be careful

The broadly reassuring picture has real exceptions.

Pregnancy. The European Food Safety Authority advises keeping caffeine to 200 mg a day or less during pregnancy, and the 2017 review found associations between higher intake and pregnancy loss, low birth weight and preterm birth. This is the clearest case for limiting coffee.

Women at high risk of fracture. The same review found an association between coffee and fracture risk in women that did not appear in men.

Unfiltered coffee and cholesterol. Coffee oil contains cafestol, which raises LDL cholesterol — the cholesterol guide explains which brewing methods carry it. Paper filters hold most of it back; French press, boiled coffee and some other methods do not. A Norwegian study of more than half a million people found unfiltered coffee was associated with higher mortality than filtered coffee. If your cholesterol is a concern, make paper-filtered coffee your daily habit — the French press guide explains the trade-off.

Sleep and anxiety. Caffeine's effects on sleep and on anxiety vary enormously between people — the coffee and sleep guide explains why. If coffee leaves you jittery or wakeful, the population averages above do not describe you.

How much is too much

The European Food Safety Authority's 2015 safety assessment is the most widely used benchmark:

The catch is that "a cup" is not a unit of caffeine. Depending on the coffee, the brew method and the size, one cup can hold anything from about 30 mg for a weak instant to 200 mg or more for a large mug of strong filter coffee. The caffeine guide has figures by drink.

What this does not mean

• Coffee is not medicine. There is no good reason to start drinking coffee for its health effects if you do not enjoy it or it disagrees with you.
• Additions change the picture. The research mostly cannot separate black coffee from a large sweetened drink with syrup and cream. Nutritionally they are very different things.
• More is not better. The associations level off at moderate intake, and the safety limits above still apply.

The fair summary

If you enjoy coffee and drink a few cups a day, the best available evidence says you can relax about it, and that it may even be doing you some good — particularly your liver. If you are pregnant, at high risk of fracture, sensitive to caffeine, or relying heavily on unfiltered coffee, the picture is different and worth taking seriously.

What the evidence cannot yet say is that coffee itself makes people healthier. For that, the research would need the kind of long-term randomised trials that nobody has run, and probably nobody will.