In the Million Women Study, 20,070 women with hospital-recorded alcohol use disorder (AUD) were matched to 100,339 controls and followed for over 20 years. AUD carried a hazard ratio of 3.73 for all-cause dementia, highest for mixed (5.03), unspecified (4.29) and vascular dementia (3.96). Risk peaked before age 70 (HR 10.68), fell at 70-79 (3.75) and at 80 and over (2.34), and stayed elevated (2.16) with a 15-year exposure lag and after competing-mortality adjustment. Self-reported drinking quantity predicted dementia far more weakly.
The gap between diagnosed AUD and reported consumption is the key insight: dementia risk tracks the pattern and severity of dependence, not weekly units alone. Self-reported intake in older cohorts is notoriously underestimated, and a clinical AUD diagnosis likely captures binge drinking, nutritional deficiency, head injury, and vascular and liver damage that a single intake figure misses. The strong vascular and mixed dementia signal fits that picture. The early-onset gradient also matters, since a tenfold risk before 70 points to alcohol-related brain injury rather than ordinary late-life dementia. The 15-year lag argues against simple reverse causation. Caveats remain: this is a preprint, not peer reviewed; hospital-recorded AUD captures only severe, treated cases; residual confounding from smoking, deprivation and comorbidity is likely; and the cohort is women only. Practically, the result supports routine alcohol screening and earlier vascular risk management in people with AUD. Confirmatory rather than paradigm-shifting, but the effect sizes are large enough to command attention.