Data from the Women's Health Initiative Memory Study, involving 6,467 postmenopausal women (65+) who reported some level of caffeine consumption, has found that those who consumed above average amounts of coffee had a lower risk of developing dementia.
Caffeine intake was estimated from a questionnaire. The median intake was 172 mg per day (an 8-ounce cup of brewed coffee contains 95mg of caffeine, 8-ounces of brewed black tea contains 47mg, so slightly less than 2 cups of coffee or less than 4 cups of tea). The women were cognitively assessed annually.
Over ten years, 388 were diagnosed with probable dementia (209) or MCI (179). Those who consumed above the median amount of caffeine had a 36% reduction in risk. The average intake in this group was 261 mg (3 cups of coffee), while the average intake for those below the median was 64 mg per day (less than one cup).
Risk factors such as hormone therapy, age, race, education, body mass index, sleep quality, depression, hypertension, prior cardiovascular disease, diabetes, smoking, and alcohol consumption, were taken into account.
The findings are consistent with other research finding a benefit for older women. It should not be assumed that the findings apply to men. It also appears that there may be a difference depending on education level. This sample had a high proportion of college-educated women.
It should also be noted that there was no clear dose-response effect — we could put more weight on the results if there was a clear relationship between amount of caffeine and benefit. Part of the problem here, however, is that it’s difficult to accurately assess the amount of caffeine, given that it’s based on self-report intake of coffee and tea, and the amount of caffeine in different beverages varies significantly.
Moreover, we do have a couple of mechanisms for caffeine to help fight age-related cognitive decline.
A recent study using rats modified to have impaired receptors for the adenosine A2A produced rats showing typical characteristics of an aging brain. In humans, too, age-related cognitive decline has been associated with over-activation of these receptors and dysfunction in glucocorticoid receptors.
The rat study shows that over-activation of the adenosine A2A receptors reduces the levels of glucocorticoid receptors in the hippocampus, which in turn impairs synaptic plasticity and cognition. In other words, it is the over-activation of the adenosine receptors that triggers a process that ends with cognitive impairment.
The point of all this is that caffeine inhibits the adenosine A2A receptors, and when the rats were given a caffeine analogue, their memory deficits returned to normal.
Another more recent study has found that caffeine increases the production of an enzyme that helps prevent tau tangles.
Building on previous research finding that an enzyme called NMNAT2 not only protects neurons from stress, but also helps prevent misfolded tau proteins (linked to Alzheimer’s, and other neurodegenerative disorders), the study identified 24 compounds (out of 1,280 tested) as having potential to increase the production of NMNAT2. One of the most effective of these was caffeine.
When caffeine was given to mice modified to produce lower levels of NMNAT2, the mice began to produce the same levels of the enzyme as normal mice.
 . Relationships Between Caffeine Intake and Risk for Probable Dementia or Global Cognitive Impairment: The Women’s Health Initiative Memory Study. The Journals of Gerontology: Series A [Internet]. 2016 ;71(12):1596 - 1602. Available from: https://academic.oup.com/biomedgerontology/article/71/12/1596/2513764/Relationships-Between-Caffeine-Intake-and-Risk-for
 The caffeine-binding adenosine A2A receptor induces age-like HPA-axis dysfunction by targeting glucocorticoid receptor function. Scientific Reports [Internet]. 2016 ;6:31493. Available from: http://www.nature.com/srep/2016/160811/srep31493/full/srep31493.html
 . Screening with an NMNAT2-MSD platform identifies small molecules that modulate NMNAT2 levels in cortical neurons. Scientific Reports [Internet]. 2017 ;7:43846. Available from: http://www.nature.com/srep/2017/170307/srep43846/full/srep43846.html