Observational studies suggest a “U-shaped” relationship between alcohol use and dementia, in which light drinking appears to be protective. This study combined large-scale cohort analyses with Mendelian randomization to investigate this association. Data were drawn from the US Million Veteran Program and the UK Biobank (n=559,559 adults aged 56–72; >90 percent reporting alcohol use at baseline), with follow-up periods of 4–12 years. Researchers used summary data from 45 genome-wide association studies cohorts to assess genetic variants linked to self-reported weekly alcohol use and alcohol use disorder (AUD) diagnoses coded in medical records.

• During follow-up, 14,540 individuals (3 percent) developed dementia.
• Observational analyses demonstrated a U-shaped association, with higher dementia risk among abstainers, people with heavy drinking (>40 drinks per week), and individuals with AUD; the lowest risk was observed among people with light drinking (<7 drinks per week).
• However, Mendelian randomization analyses demonstrated a monotonic increase in dementia risk with greater genetically proxied alcohol consumption. A one-standard deviation increase in drinks per week was associated with 15 percent higher odds of dementia. A twofold increase in the genetic liability for AUD was associated with a 16 percent higher dementia risk.
• Longitudinal analyses suggested that individuals with early cognitive decline tend to reduce alcohol consumption, likely creating the illusion of benefit among people with light drinking in observational studies.

Comments: These findings challenge long-standing beliefs that light-to-moderate drinking may be neuroprotective. Instead, this analysis supports a dose-dependent harmful effect, suggesting there is no safe level of alcohol consumption for brain health. For clinicians, this reframes counselling for older adults and underscores the importance of identifying and treating AUD as a dementia-prevention strategy. Public health messaging may need to evolve as well, as the perceived benefits of “moderate” drinking are increasingly inconsistent with causal evidence.

John Fomeche, MD* and Darius A. Rastegar, MD *2025–26 Rich Saitz Editorial Intern & Addiction Medicine Fellow, Yale University

Reference: Topiwala A, Levey DF, Zhou H, et al. Alcohol use and risk of dementia in diverse populations: evidence from cohort, case-control and Mendelian randomisation approaches. BMJ Evid Based Med. 2026;31(1):13–22.