Coffee: what a daily habit actually does to your health, according to the data

Coffee is the most-studied dietary compound in medicine. Here is the surprisingly good news — and the legitimate caveats around timing and preparation.

Coffee spent decades on the list of things you were supposed to feel slightly bad about, which makes the current state of the evidence mildly comic: it is now among the better-supported things in an ordinary diet. The caveats that survive are real, but they are narrower and more specific than the ones people carry around — and one of them is about your filter rather than your dose.

What the umbrella review found

The most useful single document in this field is a 2017 umbrella review published in the BMJ, which pooled more than 200 meta-analyses covering a very large number of health outcomes. Its summary was that coffee consumption is more often associated with benefit than harm, with the largest reductions in risk appearing at around three to four cups a day.

Also read

Supplement Reviews

Hypnozan 8H Program review: 30 days with a two-part evening formula

1 Sep 2026

Supplement Reviews

Pure Marine Collagen review: the best-labelled collagen here, and the least collagen in it

28 Aug 2026

At that intake, the review reported lower all-cause mortality and cardiovascular mortality, along with reduced risk of type 2 diabetes, several liver conditions including cirrhosis and liver cancer, Parkinson’s disease, and some other cancers. The dose-response curve was generally U-shaped: benefit rising to a few cups, then flattening or reversing.

The standard caveat applies with force. This is overwhelmingly observational evidence, and coffee drinkers differ from non-drinkers in ways that are hard to fully adjust for — historically including smoking, which confounded much of the older research and made coffee look worse than it was. What strengthens the case is the consistency across many independent outcomes and populations, and the dose-response pattern.

Where the harms actually are

The same review identified two consistent signals in the other direction, and both are relevant to this readership.

Pregnancy. Higher coffee intake was associated with low birth weight, preterm birth, and pregnancy loss. Standard guidance limits caffeine to around 200 mg daily in pregnancy — roughly two mugs of instant or one to two of brewed coffee, depending considerably on strength. Some researchers argue the evidence supports a stricter position; the safest reading is that this is the one context where caution is clearly warranted.

Fracture risk in women. Higher intake showed an association with fracture risk in women, though not in men. The mechanism is not well established and the effect is modest, but it is one of the few signals that has appeared consistently.

A third caveat is individual rather than universal. People vary substantially in how quickly they metabolise caffeine, largely through variation in a single liver enzyme. Research has found that slow metabolisers may face increased cardiovascular risk at high intakes, where fast metabolisers do not — which is part of why population averages describe nobody precisely.

The filter question

This is the most actionable thing in the article and the least known. Coffee contains diterpenes — cafestol and kahweol — which raise LDL cholesterol. A paper filter removes most of them. Methods that do not use one, including French press, Turkish, boiled, and to a lesser extent espresso, leave them in the cup.

The effect is real enough that a large Norwegian cohort study found different mortality outcomes by brewing method, with filtered coffee associated with better results than unfiltered. If you drink several cups of cafetière coffee daily and have raised cholesterol, switching to filtered is a genuine intervention rather than a fussy detail.

Timing: what holds up and what does not

Caffeine’s half-life is around five hours in most people, meaning half the dose from a 4pm coffee is still circulating at 9pm. A controlled trial gave participants 400 mg of caffeine at bedtime, three hours before, and six hours before sleep. The six-hour dose still reduced total sleep time by around an hour — and participants did not subjectively notice the disruption they were measurably experiencing.

That last detail is the important one. “Coffee doesn’t affect my sleep” is a report about perception, not about sleep architecture, and the two come apart.

The other timing claim in circulation has far less behind it: the advice to delay your first coffee by 90 minutes after waking, to avoid interfering with the morning cortisol rise. It is a plausible-sounding extrapolation from what is known about the cortisol awakening response, but there is no trial demonstrating that delayed coffee produces better alertness, energy, or any other outcome. Treat it as a hypothesis someone found compelling, not a finding.

The alertness illusion

Regular coffee drinkers develop tolerance, and overnight abstinence produces mild withdrawal. Research on this has argued that much of the perceived morning benefit in habitual consumers is the reversal of that withdrawal rather than a net gain over someone who does not drink coffee at all.

This is not an argument for quitting — returning to baseline is a perfectly good reason to drink something you enjoy. But it explains why the first coffee feels transformative and the third does not, and why people who cut back for a fortnight often report feeling roughly the same as they did before, once the headaches pass.

Two things worth knowing if you are a woman

Beyond the pregnancy and fracture signals, one practical interaction is worth flagging: coffee inhibits the absorption of non-haem iron, the form found in plant foods, when consumed with or near a meal. For anyone managing low iron — which disproportionately affects menstruating women, and particularly those eating little or no meat — moving coffee an hour away from iron-containing meals is a small change with a real effect.

And decaf is not a null option. Several of the associations in the observational literature, notably for type 2 diabetes and liver outcomes, appear in decaffeinated coffee too, which suggests the polyphenols rather than the caffeine are doing part of the work.

The practical takeaway

If you enjoy coffee and sleep well, the evidence gives you no reason to stop and a few reasons to feel fine about three or four cups. The specific things worth acting on are narrower: use a paper filter if your cholesterol is raised, stop caffeine by early afternoon regardless of whether you think it affects you, keep it away from iron-rich meals if your iron is low, and respect the 200 mg limit in pregnancy.

What none of this justifies is treating coffee as a health intervention. The research says it is not harming you and may be associated with some good things. That is a licence to enjoy it, not a reason to drink more of it than you want.