A JAMA study reports that adults who drank the most caffeinated coffee had 141 dementia cases per 100,000 per year versus 330 among those who rarely drank it, though the finding shows association, not cause.
A long-running study of US adults published in JAMA (2026; volume 335, issue 11) reports that regular caffeinated coffee drinking is associated with a substantially lower risk of dementia. With follow-up of up to 43 years, researchers recorded 141 dementia cases per 100,000 people each year among adults who drank higher amounts of caffeinated coffee, compared with 330 per 100,000 among those who rarely or never drank it. The strongest associations appeared among people who drank two to three cups a day, according to a summary of the research distributed by the American Physiological Society.
The pattern extended beyond diagnosed dementia. Among higher caffeinated-coffee drinkers, 7.8% reported subjective cognitive decline, meaning memory or thinking changes they noticed themselves, versus 9.5% among those who rarely or never drank it. The association was apparent regardless of participants' genetic risk of dementia. Tea was also linked to better cognitive outcomes, with the strongest association at one to two cups daily, but the study did not distinguish caffeinated from non-caffeinated tea. Decaffeinated coffee did not show the same association, and drinking more than the most favourable range did not appear to add protection.
The researchers stress that these results show an association rather than proof that coffee prevents dementia, and that other differences between coffee drinkers and non-drinkers could contribute to the findings. The difference seen for decaf makes the role of caffeine worth investigating, but it does not establish caffeine as the cause. The paper, titled 'Coffee and Tea Intake, Dementia Risk, and Cognitive Function', lists Yu Zhang, Frank B. Hu, Dong D. Wang and colleagues among its authors, and more research is needed to determine whether these drinks directly influence dementia risk.
