Alcohol and dementia: brain scans show harm at one drink, but the death data are messier

The question of whether alcohol harms the brain has two literatures that seem to disagree. Brain imaging shows dose-dependent damage starting at very low intake. Long-run studies of who actually develops dementia show a U-shaped curve, in which people who abstain also fare worse than light drinkers. Both are real findings, and reconciling them is where the honest answer lives.

The imaging evidence: harm starts low

The clearest signal comes from brain structure. A study in Nature Communications analysed multimodal brain imaging from 36,678 generally healthy middle-aged and older adults in the UK Biobank, controlling for a long list of potential confounders [s2]. It found negative associations between alcohol intake and both the large-scale structure of the brain and its microstructure: global brain volume, regional grey matter volumes and white matter integrity all declined with higher intake [s2].

The finding that drew attention was where the decline began. The negative associations were already apparent in people consuming an average of only one to two alcohol units a day, and became stronger as intake rose [s2]. There was no flat, protected zone at the bottom of the range — the association with smaller brain volume was present at levels many people would consider unremarkable [s2].

The dementia evidence: a U-shape that misleads

The picture from long-term dementia studies is more complicated. The Whitehall II cohort followed 9,087 British civil servants, aged 35 to 55 at the start, and recorded 397 cases of dementia over a mean of 23 years [s1]. Compared with people who drank 1 to 14 units a week in midlife, those who abstained had a higher risk of dementia — a hazard ratio of 1.47 [s1].

Read carelessly, that looks like evidence that a little alcohol protects the brain. It is not, and the same study shows why. Among people drinking above 14 units a week, every additional 7 units was associated with a 17% higher dementia risk [s1]. Signs of problem drinking pointed the same way: a CAGE questionnaire score above 2 carried a hazard ratio of 2.19, and hospital admission for an alcohol-related condition carried a hazard ratio of 4.28 [s1]. Heavy drinking clearly raised risk.

The abstainer result is the part that needs care. When the researchers looked at drinking trajectories over 17 years rather than a single snapshot, long-term abstinence carried a hazard ratio of 1.74, but so did a decrease in consumption over time (1.55) — and sustained drinking above 14 units a week also raised risk (1.40), all relative to stable moderate drinking [s1]. People who abstain in midlife are a mixed group that includes former drinkers who stopped because of ill health, and that "sick-quitter" effect inflates the apparent risk of abstention [s1].

How the two literatures fit together

The imaging and dementia studies are measuring different things at different points. Brain-volume loss is a continuous, physical change that the Nature Communications study detects rising from low intake [s2]. Clinical dementia is a distant endpoint decades later, shaped by who ends up in the abstainer group and by all the other things that happen over 23 years [s1].

The most defensible synthesis is this: the evidence that heavy drinking harms the brain and raises dementia risk is strong and consistent [s1]. The evidence that light drinking measurably shrinks brain structure is real but its long-term clinical meaning is uncertain [s2]. And the apparent benefit of light drinking over abstention in dementia studies is best explained by confounding, not by a protective effect of alcohol [s1].

What it does not settle

Neither study is a randomised trial, so neither can prove causation on its own [s1][s2]. The imaging study measured brain structure, not thinking ability or dementia, and smaller volumes do not translate one-to-one into cognitive decline [s2]. The Whitehall cohort was mostly white-collar British workers, which limits how far its numbers generalise [s1]. And the dementia study's abstainer category cannot fully separate lifelong non-drinkers from people who quit when they became unwell [s1].

What a reader can take from it is that the popular idea of a brain benefit from moderate drinking does not hold up, and that the harms to brain structure appear at lower intakes than the old "only heavy drinking matters" story allowed [s1][s2]. How much a light drinker should worry about a statistical reduction in brain volume, versus the many other influences on cognition over a lifetime, is genuinely unresolved.

This article describes what the studies found. It is not medical advice, and questions about cognitive health belong with a clinician.

Sources

• Alcohol consumption and risk of dementia: 23 year follow-up of Whitehall II cohort study — The BMJ, 2018-08-01
• Associations between alcohol consumption and gray and white matter volumes in the UK Biobank — Nature Communications, 2022-03-04