Coffee and Dementia: What the 43-Year Harvard Study Found, and Why Decaf Didn't Show the Effect

“Does coffee prevent dementia” is the kind of question a headline can answer in six words and a study can only answer in six paragraphs. In February 2026, JAMA published one of the largest looks yet at it: two Harvard-affiliated cohorts, 131,821 people, followed for up to 43 years. The headline result was an 18 percent lower dementia risk for the heaviest caffeinated coffee drinkers. The part that got less airtime is the one that makes the study genuinely interesting: decaf showed no association at all.

This page walks through what the study actually measured, what the numbers are, where the benefit levelled off, why the decaf result is a puzzle rather than a footnote, and what an observational study can and cannot tell you about your own cup. It is a sourced explainer, not medical advice. If you are worried about your own cognitive health or a family history of dementia, that is a conversation for a doctor, not a coffee blog.

What the study was

The paper, “Coffee and Tea Intake, Dementia Risk, and Cognitive Function,” was published online in JAMA on February 9, 2026, with Yu Zhang as lead author and Daniel Wang as senior author, both at Mass General Brigham and Harvard. It drew on two long-running US cohorts:

• The Nurses’ Health Study, which enrolled 121,700 women in 1976 (aged 30 to 55 at the start). The analysis used 86,606 of them.
• The Health Professionals Follow-up Study, which enrolled 51,529 male health professionals starting in 1986 (aged 40 to 75). The analysis used 45,215 of them.

Follow-up ran up to 43 years, with a median of 36.8 years according to Pharmacy Times’ report of the paper. Over that time, 11,033 people developed dementia. Participants filled in questionnaires every two to four years reporting how often they consumed caffeinated beverages, which is how the researchers tracked intake across decades instead of relying on a single snapshot.

Two things matter about that design. It is a prospective cohort study: people reported their habits first and the outcomes came later, which is stronger than asking people with dementia to recall what they drank. But it is still observational. Nobody was assigned to drink coffee or not, so the study can show an association and cannot show cause. We will come back to that.

The headline numbers

Compared with people who reported little or no caffeinated coffee, those with the highest intake had:

• A hazard ratio for dementia of 0.82 (95% confidence interval 0.76 to 0.89), meaning an 18 percent lower risk.
• Dementia incidence reported as 141 versus 330 cases per 100,000 person-years across the two groups being compared, per Pharmacy Times.
• A lower prevalence of subjective cognitive decline (a prevalence ratio of 0.85), meaning people were less likely to report noticing their own thinking getting worse.

In the Nurses’ Health Study cohort, which also had cognitive testing, the top quartile of coffee intake scored 0.11 points higher on the Telephone Interview for Cognitive Status. That is a small difference, and worth keeping in proportion. Senior author Daniel Wang said as much in the Mass General Brigham release: “the effect size is small and there are lots of important ways to protect cognitive function as we age,” and he described coffee as “one piece of that puzzle.”

The association held across levels of genetic risk, with benefits consistent in people with both high and low genetic predisposition to dementia, according to the release as reported by ScienceDaily.

The sweet spot, and the plateau

The dose-response was not a straight line. Moderate intake showed the strongest association: about 2 to 3 cups of caffeinated coffee a day, or 1 to 2 cups of caffeinated tea, or roughly 300 mg of caffeine a day, per Alzheimer Europe’s summary. WBUR quoted the finding plainly: “There was no increase in benefits among those who consumed more than that.”

That plateau is worth sitting with, because it undercuts the lazy reading of the result. If more coffee were better, the curve would keep falling. It does not. The study gives no support for drinking four or five cups on the theory that you are protecting your brain, and the benefit appears to be reached at a moderate level most people already drink.

For a sense of what those cups mean in milligrams, the site’s pieces on espresso shot caffeine and cold brew caffeine content show how widely the same “one cup” varies by method and size. A 300 mg day can be three modest mugs or two very large ones.

The decaf puzzle

Here is the finding that makes this study stand apart: decaffeinated coffee showed no significant association with dementia risk or cognitive performance. Alzheimer Europe quoted the paper: “Decaffeinated coffee intake was, in contrast, not associated with dementia risk.”

Lead author Yu Zhang told WBUR the team did not see it coming: “We didn’t observe any association between decaf coffee and dementia.” The researchers’ working interpretation is that caffeine is the primary neuroprotective compound, though they acknowledged real uncertainty. Coffee and tea also contain polyphenols, which may contribute to reduced inflammation, but the decaf result suggests those alone were not enough to show up in this data.

Two caveats keep this from becoming a tidy “caffeine protects the brain” slogan:

1. Decaf drinkers are a different group. People often switch to decaf because of a health concern, pregnancy, anxiety, or poor sleep. That can scramble the comparison in ways the statistics only partly fix. This is general reasoning about observational data, not a finding reported by the authors.
2. The authors themselves call for more work. Alzheimer Europe quotes the conclusion that “further research is needed to understand the pathways by which caffeinated and decaffeinated beverages influence brain health.”

If you drink decaf for good reasons, nothing here says to stop. It says this particular study did not find a brain benefit for it. For what decaf does and does not contain, see the decaf caffeine content guide.

Why this is association, not proof

Every responsible write-up of this study says the same thing, and it deserves more than a footnote. The researchers stated that the study cannot establish causality, and that other lifestyle factors like diet and exercise may explain the differences. A few specific ways an observational result like this can mislead (again, general reasoning about study design rather than claims from the paper):

• Healthy-user effects. People who feel well, sleep well, and tolerate caffeine are the ones who keep drinking it. Someone with early memory trouble, sleep problems, or a racing heart may cut back.
• Reverse causation. Dementia develops over many years before diagnosis. If early, undiagnosed decline changes how much coffee someone drinks, the drinking looks protective when it is really a symptom. A 43-year follow-up reduces this problem but cannot eliminate it.
• Self-reported intake. Questionnaires every few years capture roughly how often people drink coffee, not the cup size, the strength, or the brew method.
• Who is in the cohorts. Nurses and other health professionals are not a random sample of the population, and the Nurses’ Health Study enrolled only women while the follow-up study enrolled only men.

None of this makes the result worthless. A consistent association across 131,821 people and decades, holding across genetic risk groups, is real evidence that moderate coffee is not harming cognition and may travel with better outcomes. It is just not a prescription.

Worked example: reading your own habit against the study

Suppose you drink one large morning mug and a second cup at lunch, and you have been wondering whether it is enough, too much, or pointless. Here is a sensible way to use what the study found without over-reading it.

1. Estimate your actual caffeine, not your cup count. Use the figures for the drinks you really have. The study’s lowest-risk band was roughly 300 mg a day, per Alzheimer Europe. If your two cups come to around that, you are already in the range.
2. Do not add cups to chase the benefit. The curve plateaued. More than the sweet spot did not buy more, and higher intake carries its own tradeoffs, covered in the site’s pieces on coffee and anxiety and coffee and cortisol.
3. Check what caffeine costs you personally. If coffee wrecks your sleep, the sleep matters more than any brain-health association. Moving your last cup earlier is a better trade than drinking more.
4. If you drink decaf, keep the reason. Nothing in this study is a reason to override a medical or personal one.
5. Treat the rest of the list as the bigger lever. Wang’s own framing was that coffee is one piece of the puzzle. Diet, exercise, and the other ways to protect cognitive function as you age carry the weight.

Common mistakes

Reading “18 percent lower risk” as “coffee prevents dementia” is the first mistake; it is an association from an observational study, and its authors say causality is unproven. Reading the plateau as a license for more is the second; benefits peaked at about 2 to 3 cups and did not grow beyond that. Switching to decaf for brain health on the strength of this study is the third; it is the one variant that showed no association. Ignoring the size of the effect is the fourth; Wang described it as small, and the cognitive test difference was 0.11 points. Treating one study as settled is the fifth; the researchers say more work is needed on the pathways, and on whether the link is causal. And dropping coffee because you are anxious or sleeping badly, then feeling guilty about brain health, is the last; that trade-off belongs between you and your doctor, not a headline.

Common questions

Does coffee prevent dementia? This study cannot say. It found that people with the highest caffeinated coffee intake had an 18 percent lower dementia risk (hazard ratio 0.82) over up to 43 years, but it was observational, so it shows association, not cause. The researchers state that more work is needed to determine whether a causal link exists.

How many cups of coffee a day were linked to the lowest dementia risk? About 2 to 3 cups of caffeinated coffee a day, or 1 to 2 cups of caffeinated tea, or roughly 300 mg of caffeine a day. WBUR reported no added benefit among those who consumed more than that.

Does decaf coffee protect against dementia? Not in this study. Decaffeinated coffee showed no significant association with dementia risk or cognitive performance. The researchers believe caffeine is the primary neuroprotective compound but acknowledge uncertainty about the pathways.

Does tea count too? Yes, with caveats. Higher intake of caffeinated tea showed similar associations with the cognitive outcomes, with 1 to 2 cups a day showing comparable protective effects.

Who was in the study? 131,821 people: 86,606 women from the Nurses’ Health Study and 45,215 men from the Health Professionals Follow-up Study, both US cohorts. Over the follow-up, 11,033 developed dementia.

Does it matter if I have a genetic risk of dementia? The benefit was consistent across participants with both high and low genetic predisposition, per the Mass General Brigham release. That is not the same as saying coffee changes a genetic risk, only that the association appeared in both groups.

Should I start drinking coffee to protect my brain? Nothing in the study supports starting a habit for that reason. The effect was small, observational, and not proven causal. If you already drink moderate coffee and tolerate it well, the study is reassuring. If caffeine disrupts your sleep or raises your anxiety, those costs are real and worth weighing yourself. For the gut side of the story, see the site’s piece on coffee and gut health.

For more sourced breakdowns like this one, delivered weekly, subscribe to The Sunday Pour.

Written by Theo Marsh · Published 18 October 2026