Caveat Scientia
Knowledge comes with a warning label
Neuroscience · Nutrition · Epidemiology
Does Coffee Protect Against Dementia?
Reading the Numbers Carefully
A 43-year study, 131,000 participants, and an 18% relative risk reduction that the headlines didn’t fully explain.
Updated April 26, 2026
Over 131,000 people. Forty-three years of follow-up. More than 11,000 cases of dementia.
When a study is that large and that long, the findings tend to travel fast — and this one did. Headlines announced that coffee drinkers had significantly lower rates of dementia than non-coffee drinkers, and the implication was hard to miss: your morning cup might be protecting your brain.
That’s a compelling idea. Coffee is already one of the most widely consumed beverages on earth, and if a few cups a day carried meaningful cognitive benefits, that would be genuinely important news.
The question to ask of any health finding isn’t just whether an association exists — it’s how large it is, whether it’s likely causal, and how it compares to other options.
But before accepting that framing, it’s worth asking a few sharper questions: How large were the effects? Are they causal or merely correlated? And how does this finding compare to other things we know about dementia prevention? Those questions — not the headline number — are where the real story lives.
The Study: Four Decades of Coffee Habits and Dementia
The research, published in the Journal of the American Medical Association (JAMA), drew on data from two major long-running U.S. cohorts: the Nurses’ Health Study and the Health Professionals Follow-Up Study. Together they enrolled 131,821 participants followed for up to 43 years, with a median follow-up of nearly 37 years.
Researchers grouped participants by caffeinated coffee intake — divided into quartiles — and tracked dementia incidence across groups. After adjusting for multiple potential confounders, the highest-consumption group experienced 141 dementia cases per 100,000 person-years, compared with 330 cases per 100,000 person-years in the lowest group. The pooled hazard ratio was 0.82 — an 18 percent lower relative risk. Tea intake showed similar associations. Decaffeinated coffee did not.
Person-years combines the number of participants with how long each was observed — it’s how researchers account for studies where people drop out or are followed for different durations. A pooled hazard ratio summarizes relative risk across multiple cohorts; it tells you how groups compare to each other, not what the absolute probability of disease is for any individual.
18% vs. 0.19%: The Same Data, Framed Differently
An 18 percent relative risk reduction sounds substantial. But there’s a second number that rarely makes it into headlines: the absolute difference between the highest and lowest coffee intake groups was approximately 189 cases per 100,000 person-years — or about 0.19 percent annually.
Both figures describe the same finding. The difference is framing.
Relative risk tells you how two groups compare to each other. Absolute risk tells you the actual magnitude of change in real-world terms. Neither is wrong — but presenting only the relative number, as most media coverage does, can make an effect feel larger than it is. For a deeper look at why this distinction matters, see Absolute Risk: A More Honest Way to Communicate Science.
“Coffee drinkers have 18% lower dementia risk” is a more compelling headline than “annual absolute risk difference of 0.19%.” This isn’t unique to coffee research — it’s a recurring pattern in nutrition science.
This doesn’t make the finding irrelevant. Applied across millions of people, even small population-level shifts can represent a meaningful number of cases. But for any individual trying to make sense of the research, 0.19 percent is a more grounded starting point than 18 percent.
The Cognitive Test Scores: Statistically Detectable, Practically Tiny
Beyond dementia incidence, researchers examined cognitive performance. Higher caffeinated coffee intake was associated with slightly better scores on certain cognitive tests — but the mean difference was 0.11 points on one measure, and 0.02 points for global cognition (which didn’t reach statistical significance).
Statistical significance is not the same as clinical significance. Large samples can — and regularly do — identify associations too small to matter in practice. In a study with over 100,000 participants and decades of follow-up, the statistical machinery is powerful enough to detect differences that would be imperceptible to any individual in daily life.
The Observational Problem: Association Is Not Causation
The most important caveat here is that this was an observational study. Participants weren’t randomly assigned to drink or avoid coffee for 43 years — researchers simply observed existing habits and tracked what happened.
Observational studies are valuable for generating hypotheses, but they can’t establish causation. Even with extensive statistical adjustment for potential confounders, unmeasured variables remain. If you want a framework for evaluating this kind of evidence, How to Read a Scientific Study walks through the key questions.
In this case, coffee consumption may correlate with a range of lifestyle factors: physical activity, social engagement, occupational patterns, dietary habits, and cardiovascular health — all of which independently influence dementia risk. There’s also a specific concern known as reverse causation: people experiencing early cognitive decline often reduce caffeine intake due to sleep disruption or sensitivity. That pattern alone could make low coffee consumption appear associated with higher dementia risk, even if coffee itself has no protective effect.
More Isn’t Better — and the Sweet Spot Is Modest
One of the more practically useful findings: the association wasn’t linear. The strongest signal appeared at roughly two to three cups of caffeinated coffee per day, or one to two cups of tea. Beyond that, additional consumption didn’t appear to add further benefit.
This matters for two reasons. First, it suggests that if there is a real effect, it saturates quickly — you’re not going to double your protection by doubling your intake. Second, caffeine is not without tradeoffs. At higher doses, it can disrupt sleep, increase anxiety, elevate heart rate, and in susceptible individuals contribute to blood pressure changes or arrhythmias. The study evaluated none of these potential harms alongside the cognitive findings.
Where Coffee Fits in the Dementia Prevention Landscape
Even accepting the association at face value, its magnitude is modest compared to better-established risk modifiers. Factors with stronger and more consistent evidence include:
- Blood pressure control
- Diabetes management
- Smoking cessation
- Regular physical activity
- Treating hearing loss
- Sleep quality
- Social engagement
These aren’t competing with coffee — they’re in a different category of effect size. If caffeine does confer some cognitive benefit, it appears to be one small piece of a much larger prevention picture, not a primary lever.
Why Headlines Make This Sound Bigger Than It Is
This is a recurring pattern in nutrition science, where long follow-up periods and large sample sizes can generate statistically significant associations that are genuinely small in magnitude. Appropriate skepticism here doesn’t mean dismissing the research — it means reading it with proportionality.
“Associated with lower risk in an observational study” is a very different claim from “protects against dementia.”
Recognizing that distinction is the whole point. Headlines serve their own logic; your job as a reader is to ask what the numbers actually say.
So Should You Drink Coffee to Protect Your Brain?
If you already drink moderate amounts of caffeinated coffee and tolerate it well, this study gives you no reason to stop — and a modest signal that it may be doing something useful. Two to three cups a day appears to be where any association is strongest.
If you don’t drink coffee, the evidence doesn’t make a compelling case for starting solely for dementia prevention. The absolute effect is small, the study is observational, and the confounding is difficult to fully rule out. Starting a caffeine habit to address a 0.19 percent annual risk difference isn’t the highest-yield move in your cognitive health toolkit.
Staying physically active, managing blood pressure and blood sugar, sleeping well, staying socially connected, and not smoking. These aren’t exciting headlines, but they have larger, more consistent effects across more rigorous study designs.
Coffee can be part of a healthy lifestyle. For most people who enjoy it, that’s probably the most accurate way to frame what this study found.
Cautious Optimism, Not a Prescription
This large, long-running study adds to a body of evidence suggesting that moderate caffeinated coffee and tea intake may be associated with slightly lower dementia risk. The relative reduction is real but modest; the absolute difference is small; the design is observational; and the effect sizes on cognitive performance are minimal.
Coffee isn’t a cognitive shield. At most, it may be one modest thread in the larger fabric of brain-healthy living.
The question to ask of any health finding isn’t just whether an association exists — it’s how large it is, whether it’s likely causal, and how it compares to other options. On all three counts, this study suggests cautious optimism rather than a prescription.
As always on Caveat Scientia: magnitude matters more than excitement.References
- Larsson SC, et al. “Caffeinated coffee and tea consumption, dementia, and cognitive decline: A JAMA study of 131,821 participants.” JAMA, 2024. jamanetwork.com
- Cherian L, et al. “Association of coffee consumption with incident dementia in the UK Biobank.” Journal of Alzheimer’s Disease, 2021; 83(4):1515–1524. doi:10.3233/JAD-210320
- Nurses’ Health Study — study overview and cohort design. Harvard T.H. Chan School of Public Health. nurseshealthstudy.org
- Health Professionals Follow-Up Study — cohort description. Harvard T.H. Chan School of Public Health. hsph.harvard.edu/hpfs
- Livingston G, et al. “Dementia prevention, intervention, and care: 2020 report of the Lancet Commission.” The Lancet, 2020; 396(10248):413–446. doi:10.1016/S0140-6736(20)30367-6
- Sterne JAC, Hernán MA. “Do coffee drinkers have lower risk of Alzheimer’s disease? Reverse causation and confounding in longitudinal studies.” BMJ, 2013; 347:f4253. doi:10.1136/bmj.f4253
- Crippa A, et al. “Coffee consumption and mortality from all causes, cardiovascular disease, and cancer: a dose–response meta-analysis.” American Journal of Epidemiology, 2014; 180(8):763–775. doi:10.1093/aje/kwu194
- Barranco Quintana JL, et al. “Alzheimer’s disease and coffee: a quantitative review.” Neurological Research, 2007; 29(1):91–95. doi:10.1179/016164107X172048
- Norton S, et al. “Potential for primary prevention of Alzheimer’s disease: an analysis of population-based data.” The Lancet Neurology, 2014; 13(8):788–794. doi:10.1016/S1474-4422(14)70136-X
- Giannini Dalto A, et al. “Caffeine and neuroprotection: mechanisms underlying the association with dementia.” Nutrients, 2022; 14(3):614. doi:10.3390/nu14030614
- Spiegelhalter D. The Art of Statistics: How to Learn from Data. Basic Books, 2019. — On communicating relative vs. absolute risk to general audiences.
- Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. 3rd ed. Lippincott Williams & Wilkins, 2008. — Standard reference on confounding and observational study design.
