For years I’ve switched to decaf around three in the afternoon and felt a little smug about it. I keep the mug and the roasted bitterness and give up only the jolt that would have me staring at the ceiling at midnight. If coffee was doing my brain any favors, I assumed decaf was doing most of them too.
A study of 131,821 American nurses and health professionals, followed for up to 43 years, suggests I had that backwards. In the analysis Harvard researchers published in JAMA, the heaviest drinkers of caffeinated coffee had an 18 percent lower risk of dementia than the lightest drinkers. The heaviest decaf drinkers showed no clear dementia association at all, and they were more likely to say their memory was slipping.
The paper, by Yu Zhang, senior author Dong D. Wang and colleagues, came out in February and drew on two of the longest-running diet cohorts anywhere: the Nurses’ Health Study (86,606 women) and the Health Professionals Follow-up Study (45,215 men). Participants filled out food questionnaires every two to four years, so nobody was relying on one memory of one morning habit. Over a median of 36.8 years, 11,033 of them developed dementia, identified through death records and self-reported diagnoses. That method is why Alzheimer’s couldn’t be analyzed on its own. This week the study is back in circulation under a ScienceDaily headline promising a “dramatically lower” dementia risk.
Where “dramatically” comes from
In the quarter of participants who drank the least caffeinated coffee, dementia showed up at a rate of 330 cases per 100,000 person-years. In the quarter who drank the most, the rate was 141. Put those side by side and you’d think the risk had been cut by more than half, which is roughly where a word like “dramatically” comes from. But those are raw rates, and heavy coffee drinkers differ from light ones in plenty of ways that raw rates ignore.
Once the researchers adjusted for the measured differences between the groups, the top quarter’s hazard ratio was 0.82, which works out to the 18 percent. With more than eleven thousand cases behind it the estimate is precise, but it is still an association, and adjustment only handles the confounders someone thought to measure. Wang was more careful than the headline writers. “While our results are encouraging, it’s important to remember that the effect size is small,” he told PsyPost, adding that “there are lots of important ways to protect cognitive function as we age.”
The protective association bottomed out at around two to three cups of caffeinated coffee a day, roughly 300 mg of caffeine. In this cohort, more cups added nothing and didn’t appear to do harm either. Other data are less reassuring at the top end. A UK Biobank analysis of 17,702 people, led by Kitty Pham at the University of South Australia, reported a 53 percent higher dementia risk among people drinking more than six cups a day. That’s a different population with a different design, so I’d treat it as a reason not to go past a few cups, not as part of the same curve.
Same bean, minus one molecule
My afternoon habit doesn’t come out of this well. The JAMA team found that caffeinated coffee’s hazard ratio of 0.82 (95% CI 0.76 to 0.89) sat clearly below the line of no effect. The heaviest decaf drinkers came in at 0.97 (95% CI 0.93 to 1.01), an interval that runs right across it. That doesn’t prove decaf does nothing. It means a dataset this large, followed for this long, couldn’t pick up a dementia signal for it.
| Result | Value (HR) | Lower bound | Upper bound |
|---|---|---|---|
| Caffeinated coffee | 0.82 | 0.76 | 0.89 |
| Decaf | 0.97 | 0.93 | 1.01 |
On self-reported memory slippage the two drinks went in opposite directions. That complaint, called subjective cognitive decline, can come before a diagnosis. Heavy caffeinated-coffee drinkers were less likely to report it, 7.8 percent versus 9.5 percent in the lowest-intake group, for a prevalence ratio of 0.85. Heavy decaf drinkers were 16 percent more likely to report it (prevalence ratio 1.16, 95% CI 1.08 to 1.24).
Decaf comes from the same bean and the same roast, and it still carries much of the polyphenol chemistry that gets credited with coffee’s health halo. If those compounds were protecting brains, I’d have expected at least a faint echo of the benefit in the decaf drinkers. So why is there nothing? The authors’ leading explanation is the simplest one: caffeine itself may be the main active ingredient.
The biology at least makes that plausible. All day, adenosine pools in your brain, and the longer you’re awake the heavier it pulls you toward sleep. Caffeine sits in adenosine’s receptors so the signal can’t land. Your morning cup is literally squatting in a sleep-pressure system for a few hours, and the authors suggest that blockade, repeated for decades, could ripple out into amyloid-beta buildup, inflammation, insulin sensitivity and blood-vessel function. Those are proposed routes. Nobody has traced that chain through human brains across forty years.
The authors also raise a second, less flattering explanation: confounding by indication. Think about who drinks a lot of decaf. It’s often someone whose doctor told them to cut caffeine because of palpitations, blood pressure, anxiety or bad sleep, and some of those conditions have their own links to cognitive trouble. The researchers also concede that reverse causation can’t be fully ruled out, since people in early decline may change their habits before anyone diagnoses them. My guess is that the decaf result mixes both things: some caffeine effect, plus some of whatever sent people to decaf in the first place.
A few results make this harder to dismiss. The association didn’t vary significantly by APOE4 status or by an Alzheimer’s polygenic risk score, so carriers of the best-known risk gene showed no detectably different pattern. That doesn’t rule out a healthy-user effect, but the signal isn’t confined to the genetically lucky. It looked stronger in people 75 and younger (hazard ratio 0.65) than in those over 75 (0.81), though the paper’s main text doesn’t report a formal test of that gap. Before you generalize any of it, remember who was in the study: single-sex cohorts made up mostly of health professionals, with questionnaires that never recorded how anyone brewed their coffee.
Tea did well too, with a hazard ratio of 0.86 at the top of the intake range and the benefit leveling off at one to two cups a day. The questionnaires didn’t separate caffeinated tea from herbal, so tea can’t settle the caffeine question in either direction.
The study was funded by National Institutes of Health grants, and the funder had no role in design or analysis. The only disclosed conflict is that coauthor Frank Hu reported research support from the Analysis Group.
A molecule nobody owns
A randomized trial would settle the caffeine question, and the only one I could find is narrow. CAFCA, run by Lille University Hospital in France and registered as NCT04570085, is a Phase III placebo-controlled trial. It gives people with Alzheimer’s, from the earliest symptomatic stage through moderate dementia, a 200 mg caffeine capsule twice a day for about 30 weeks and measures memory, learning and daily function. It tests caffeine as a treatment for disease people already have. It can’t tell anyone whether two cups a day from age 40 changes what happens at 80.
Compare that with where the money goes for the same disease. When Eisai priced its anti-amyloid infusion Leqembi in January 2023, it set the U.S. list price at $26,500 a year for a drug infused every two weeks. I’m not claiming caffeine does any part of Leqembi’s job, because nobody has shown that. I’m talking about incentives. No company owns caffeine, so no sponsor stands to profit from asking whether a daily cup prevents anything. Everything we know here exists because public grants kept researchers mailing food questionnaires to nurses for decades. If the prevention trial ever happens, I’d bet public money pays for that too.
Until then, an 18 percent association from an observational study is a nudge, not a prescription. If you don’t drink coffee, this isn’t a reason to start. If caffeine leaves you jittery or awake at 2 a.m., an effect this size isn’t worth fighting your own body over. For me the change is small. I’m keeping my two morning cups fully caffeinated, moving the afternoon one earlier instead of swapping it for decaf, and I’ve stopped treating decaf as anything more than a warm drink I happen to like.
Sources
- JAMA: Zhang, Wang et al., Coffee and Tea Intake, Dementia Risk, and Cognitive Function (2026)
- PsyPost: Moderate coffee and tea consumption linked to lower risk of dementia
- ScienceDaily: Drinking 2–3 cups of coffee a day linked to dramatically lower dementia risk (November 2026)
- Alzheimer Europe: CAFCA trial of caffeine in Alzheimer’s disease, Lille University Hospital
- Eisai release via BioSpace: U.S. pricing for Leqembi (lecanemab), January 2023
- News-Medical: report on Pham, Hyppönen et al., high coffee consumption and dementia risk in UK Biobank (2021)