A prospective cohort study published online February 9, 2026, in JAMA followed 131,821 participants in the Nurses’ Health Study and Health Professionals Follow-up Study for up to 43 years (11,033 dementia cases identified via death records and physician diagnoses) and found higher caffeinated coffee intake was associated with lower dementia risk (HR 0.82; 95% CI, 0.76–0.89; highest vs. lowest quartile). Tea showed similar associations. Caffeine appears to have a correlation in the data. Decaffeinated coffee did not have the same relationship, a design detail that turns a familiar “coffee and cognition” signal into something closer to observational target validation for caffeine-linked biology. To be clear, however, the data is observational.
But in any case, the caffeinated–decaf split is consistent with caffeine itself playing a key role though it cannot exclude other beverage differences or residual confounding between the two groups. Caffeine is a nonselective adenosine receptor antagonist, and the A₂A receptor subtype has been implicated in neuroinflammation, Aβ-related synaptic toxicity and tau phosphorylation. All are core pathological processes in Alzheimer’s disease. The adenosine pathway is already a validated drug target in neurodegeneration: istradefylline (Nourianz), a selective A₂A antagonist, has been approved in the U.S. and Japan for Parkinson’s disease since 2019. A 131,000-person, four-decade dataset showing a consistent pattern. That is, caffeinated exposure associated with lower risk, decaffeinated not. This adds population-scale weight to the question of whether A₂A modulation could matter in dementia, too.
The study also assessed cognitive function directly, though objective testing was limited to the Nurses’ Health Study cohort and conducted via telephone-based neuropsychological measures. Higher caffeinated coffee intake was associated with a modest improvement on the Telephone Interview for Cognitive Status (mean difference, 0.11; 95% CI, 0.01–0.21) and lower prevalence of subjective cognitive decline (7.8% vs. 9.5%; prevalence ratio, 0.85). However, the association with a global cognition composite score did not reach statistical significance (P = .06) — a reminder that the cognitive signal, while consistent in direction, is small and not uniform across measures.
The pharma comparison, imperfect but hard to ignore
The 18% risk reduction observed here invites a pointed, if imperfect, comparison with the anti-amyloid antibodies that have consumed billions in development. In the Clarity AD trial, lecanemab (Leqembi) slowed decline on the CDR-SB scale by 0.45 points over 18 months: a 27% reduction, but on a different outcome measure, in a different study design, over a far shorter time horizon. That modest clinical benefit came with a meaningful safety burden: ARIA-E (brain edema) occurred in roughly 13% of treated patients vs. 2% on placebo, with symptomatic cases in about 3%.
The metrics are genuinely not comparable, an observational hazard ratio across decades versus a randomized clinical endpoint at 18 months. But the juxtaposition is difficult for the field to dismiss entirely: a cheap, widely consumed, nonselective adenosine antagonist showing population-level signal in the same disease space where the most expensive drug programs in history are delivering incremental gains with nontrivial risks.
Caveats and what comes next
Apart from the observational flag, other standard concerns apply. Reverse causation, people in preclinical cognitive decline may reduce their coffee intake years before diagnosis, is another concern. The long follow-up period and repeated dietary assessments every 2 to 4 years help mitigate this, but they don’t eliminate it. Residual confounding is also possible; decaf drinkers may differ from caffeinated-coffee drinkers in ways the models don’t fully capture.
A dose-response analysis showed a nonlinear relationship, with the most pronounced associations at approximately 2 to 3 cups per day of caffeinated coffee or 1 to 2 cups of tea. That is consistent with receptor-saturation kinetics and a potential ceiling on benefit.
Zhang Y, Liu Y, Li Y, et al. Coffee and Tea Intake, Dementia Risk, and Cognitive Function. JAMA. Published online February 9, 2026. doi:10.1001/jama.2025.27259
Filed Under: Neurological Disease
