Intended for healthcare professionals


Relation between alcohol consumption in midlife and dementia in late life

BMJ 2018; 362 doi: (Published 01 August 2018) Cite this as: BMJ 2018;362:k3164
  1. Sevil Yasar, associate professor
  1. Departments of Medicine and Neurology, Johns Hopkins School of Medicine, Baltimore, MD 21224, USA
  1. syasar1{at}

It’s complicated

Alzheimer’s disease is a rapidly growing clinical and public health problem with currently no disease modifying therapies to treat or prevent the disease.1 Research suggests that pathological changes precede clinical symptoms by decades,2 resulting in an increased interest in modifiable lifestyle and health related risk factors in midlife.3

Interest in the health effects of alcohol, including risk of dementia, has been ongoing for a long time. Epidemiological studies reported a J-shaped relation, in which abstaining from alcohol (<1 unit/week) or consuming large amounts (>21-28 units/week) was associated with an increased risk of dementia compared with light to moderate (7-21 units/week) consumption.4 However, these studies had relatively short follow-ups of 2-9 years.4 Only a few studies had longer follow-up or assessed the role of ethnicity or sex or types of alcohol consumed, and none evaluated the effects of change in consumption behaviour. Light to moderate alcohol use has also been associated with a cardioprotective effect through anti-inflammatory or antioxidant pathways, or both,5 and a reduction in mortality.6

In a linked article, Sabia and colleagues (doi:10.1136/bmj.k2927) attempt to tackle some of these issues.7 They evaluated the effects of midlife alcohol consumption on risk of dementia in 9087 participants of the Whitehall II study over a mean follow-up of 23 years. Participants were stratified as abstainers; moderate alcohol drinkers, consuming 1-14 units/week (this group also served as the reference group); or excessive drinkers consuming 14 units/week or more—a definition based on recently revised UK guidelines.8

Excessive use

Long term consumption of >14 units/week was associated with an increased risk of dementia. Although the link was not evident in all analyses, the authors show clearly that with every 7 unit/week increase, there was a significant 17% increase in dementia risk. Risk was also significantly increased among participants with at least one alcohol related hospital admission, and presence of alcohol dependence. These findings are consistent with previous research, also from the Whitehall II cohort, reporting >14 units/week alcohol consumption is associated with hippocampal atrophy and faster cognitive decline.9,10

Mortality was found to be statistically significantly higher (31%) among excessive alcohol users and significantly increased (12%) with every 7 unit/week increase but was not found to be a competing risk for dementia. Dementia risk among participants with or without cardiometabolic disease was not statistically different, but with every 7 unit/week increase the risk of cardiometabolic disease (7%) and dementia risk in participants with cardiometabolic disease (20%) significantly increased, suggesting potential mediating effects.


Abstinence from alcohol in midlife, long term abstinence, and decrease in consumption were associated with a significantly higher risk of dementia of 45%, 67%, and 50%, respectively, compared with consumption of 1-14 units/week, but only among participants who reported abstinence from wine. Abstinence was not associated with increased mortality, but it was associated with a significantly increased incidence of cardiometabolic disease (14%). Dementia risk among abstinent participants with cardiometabolic disease was non-significantly increased, whereas dementia risk in participants free of cardiometabolic disease was not altered, suggesting a possible mediating role of cardiometabolic disease in dementia risk among abstainers.

The most intriguing finding from this study was the significantly increased risk of dementia among abstainers, including long term abstainers and participants who became abstainers, and that association was only present in those who abstained from wine.

Abstainers were mainly women, had lower education and physical activity, were obese, and had a higher prevalence of cardiometabolic risk factors, all associated with an increased risk of dementia,11,12 which could explain the differences; however, adjustment for confounding factors did not alter the findings.

Dementia risk was mediated by the presence of cardiometabolic disease, which was non-significantly increased among abstainers, raising the question of a possible protective effect from moderate alcohol consumption. This is further supported by findings of an increased risk of dementia observed only in those who abstained from wine. Wine, in addition to alcohol, contains polyphenolic compounds, which have been associated with neuroprotective effects on both neurodegenerative and vascular pathways, and with cardioprotective effects through inflammation reduction, inhibition of platelet aggregation, and alteration of lipid profile.13

This study is important since it fills gaps in knowledge, but we should remain cautious and not change current recommendations on alcohol use based solely on epidemiological studies. The next steps should include confirmation of findings in other long term cohort studies and ideally a randomised clinical trial, to answer pressing questions about the possible protective effects of light to moderate alcohol use on risk of dementia and the mediating role of cardiovascular disease. Such a trial would be of long duration and ethically challenging. It would have to be funded exclusively by government agencies to avoid bias. In. summary, alcohol consumption of 1-14 units/week may benefit brain health; however, alcohol choices must take into account all associated risks, including liver disease and cancer.


  • Competing interests: I have read and understood the BMJ policy on declaration of interests and declare the following interests: none.

  • Provenance and peer review: Commissioned; not peer reviewed.


Log in

Log in through your institution


* For online subscription