Anttila · BMJ (Clinical research ed.) 2004 · prospective cohort study · n=1018

Alcohol drinking in middle age and subsequent risk of mild cognitive impairment and dementia in old age: a prospective population based study.

Cited 327 times in the scientific literature.

Level 3 - non-randomized controlled study

Prospective population-based cohort study with 23-year follow-up

PubMed 15304383 · doi:10.1136/bmj.38181.418958.BE · record verified 2026-08-31

What was done

A prospective, population-based cohort study examined the relationship between midlife alcohol drinking frequency, apolipoprotein E (APOE) genotype, and the risk of mild cognitive impairment and dementia in old age. From an initial random sample of 1,464 men and women aged 65–79 in Kuopio and Joensuu, eastern Finland, first assessed in 1972 or 1977, 1,018 individuals (70%) were re-evaluated in 1998 after an average follow-up of 23 years.

What was found

Participants who drank no alcohol in midlife and those who drank frequently were both twice as likely to develop mild cognitive impairment in old age compared with infrequent drinkers. APOE-e4 status modified the relationship between alcohol consumption and dementia (interaction p = 0.04; infrequent drinking interaction p = 0.02; frequent drinking interaction p = 0.03). Among APOE-e4 carriers, compared with non-carriers who never drank, odds ratios for dementia were 0.6 for non-drinkers, 2.3 for infrequent drinkers, and 3.6 for frequent drinkers. For APOE-e4 non-carriers, dementia odds ratios were similar across all drinking categories.

Why it matters

This study suggests that the association between midlife alcohol intake and late-life dementia risk is genetically stratified by APOE-e4 status, with increased risk observed exclusively in carriers, alongside a U-shaped association for mild cognitive impairment overall.

Limits

Of the original cohort, 30% (446 individuals) did not participate in the 23-year follow-up, introducing potential survival and attrition bias. Alcohol exposure was assessed using broad frequency categories without specific volume, grams of ethanol, or beverage types. Confidence intervals for odds ratios were not reported in the abstract.

Cited by