Alcohol's stroke risk climbs steeply past light drinking

The relationship between alcohol and stroke has long looked J-shaped in observational data, with small amounts appearing benign or even favourable and larger amounts clearly harmful. Stroke is not one disease, though: ischaemic stroke, caused by a blockage, and haemorrhagic stroke, caused by a bleed, may respond to alcohol very differently. Much of the evidence behind drinking guidance was generated in Western populations, whose drinking patterns and genetics differ. This review set out to describe the dose-response relationship between alcohol intake and both stroke and all-cause mortality specifically in East Asian men.

The authors searched Pubmed, OVID, Embase and ISI Web of Knowledge using subject headings covering alcohol drinking, ethanol, stroke, cerebrovascular disease and mortality, restricted to Korean, Japanese or Chinese populations. Seventeen prospective cohort studies met the inclusion criteria. A standardised protocol was used to extract study design, participant characteristics, amount of alcohol consumed, stroke and mortality outcomes, control for confounding, and risk estimates. Relative risks and 95% confidence intervals were pooled using fixed or random effects models depending on heterogeneity testing. Cohorts ranged in size from 1322 to 108,461 subjects; data available for women were too limited to pool, so the analysis covers men only.

Compared with non-drinkers, men drinking up to 20 g a day had a relative risk of ischaemic stroke of 0.85, with no significant difference at 21-40 g and non-significant elevations at 41-60 g and above 60 g. For haemorrhagic stroke the pattern was different: no significant change at lower intakes, but a relative risk of 1.74 above 60 g a day. All-cause mortality was lower at up to 20 g (0.83) and at 21-40 g (0.93), sat at 1.01 for 41-60 g, and rose to 1.32 above 60 g. So the apparent advantage was confined to the lowest intake band and disappeared as consumption increased.

The findings echo the J-shaped curves reported in Western cohorts while adding a clear split by stroke subtype, with bleeding strokes the outcome most sharply linked to heavy intake. Because every included study is observational, the low-intake associations are vulnerable to the usual problems: people who drink lightly may differ from non-drinkers in many ways that no statistical adjustment fully removes. The absence of usable data on women is a substantial gap, and the review's search reaches only to the late 2000s. Whether the lower risk at light intake reflects alcohol itself or the company it keeps remains open.

The consistent and least ambiguous part of this picture is the upper end: heavier drinking tracked with substantially higher risk of haemorrhagic stroke and of dying from any cause. The apparently lower ischaemic stroke risk among the lightest drinkers is an association measured in East Asian men, not a demonstrated benefit, and it does not transfer neatly to women or to other populations. Reading a graded pattern like this, the useful signal is the direction of travel as intake rises rather than any single reassuring number at the bottom. Anyone weighing their own drinking has other health considerations in play that a stroke-only analysis cannot capture.