Aging
Associations of alcohol consumption with memory loss in European middle-aged and older adults: evaluating time-varying confounding, lagged effects, and “sick quitter” effects Shelli Vodovozov* Shelli Vodovozov Vodovozov Vodovozov Vodovozov Vodovozov Vodovozov Vodovozov Vodovozov Vodovozov Vodovozov Vodovozov Boston University School of Public Health
Estimating effects of alcohol consumption and cessation on dementia incidence is complicated by reverse causation as declines in cognition or health lead individuals to reduce or stop drinking.
Participants (n=68,379) from 20 countries in the Survey of Health, Ageing and Retirement in Europe reported alcohol consumption up to 8 times (mean=1.98, SD = 0.95). Alcohol use was categorized per sex-specific NIAAA guidelines (none, low/moderate, heavy). Memory (immediate and delayed 10-word recall; z-scored) was assessed 1–3 waves after each alcohol report.
We estimated stabilized inverse probability weights for treatment and censoring and applied generalized estimating equations to weighted data. Covariates included sociodemographics, prior cognition, health behaviors, diagnoses, and self-rated health (1 = poor to 5 = excellent). Analyses were stratified by sex. E-values assessed sensitivity to unmeasured confounding, and pooled logistic regression examined predictors of alcohol cessation to quantify reverse causation.
Memory was better among heavy drinkers (vs non-drinkers) reported one (βheavy=0.07[0.01, 0.12], two (βheavy=0.12[0.07, 0.17], or three waves prior to memory (βheavy=0.21[0.04, 0.21]. Among men, estimates were (βheavy=0.10 [0.05, 0.16] at lag one and βheavy=0.12[0.04, 0.21] at waves two and three.
E-values ranged from 1.32-1.48 among women and 1.43-1.49 among men.
Higher past cognition was associated with lower odds of alcohol cessation among women (OR=0.93[0.90, 0.96]), and each unit increase in self-rated health was associated with greater cessation among women (OR=0.79[0.76, 0.82]) and men (OR=0.78[0.75, 0.82]).
Heavy alcohol use was associated with better memory in this large European cohort. Evidence of reverse causation suggests this association reflects health and cognition-related selection into continued drinking, underscoring limitations of conventional observational designs and the need for methods that better address dynamic exposure changes.
