Aging
Estimated effects of recent alcohol use on memory in the Health and Retirement Study: using marginal structural models to account for time-varying confounding and censoring Scott C. Zimmerman* Scott C. Zimmerman Zimmerman Zimmerman Zimmerman Zimmerman Zimmerman Zimmerman Zimmerman Zimmerman Zimmerman Zimmerman Zimmerman Boston University & UCSF
Estimating effects of current alcohol use on memory is difficult due to time-varying confounding (e.g. “sick quitting” effect) and informative censoring.
US Health and Retirement Study participants from waves 3–14 (1996-2018) born after 1957 were included (n=7,147 women, 6,332 men). We coded alcohol use in the last 30 days as: none; low-to-moderate (LtoM; women: ≤7, men: ≤14 weekly drinks and no heavy episodic drinking [HED; 6+ drinks/occasion]); or heavy (women: >7, men: >14 weekly drinks or any HED). Memory score was the sum of correctly recalled words from 10-item immediate and delayed word recall tests (range 0-20).
To account for time-varying confounding, we calculated stabilized exposure (multinomial regression) and censoring (logistic regression) weights, conditional on past cognitive, behavioral, health and sociodemographic predictors. We fit adjusted sex-stratified weighted linear mixed effects models of memory level and change on alcohol use reported at the prior wave.
A 1-point higher memory score in the previous wave was associated with lower odds of current abstinence compared to LtoM drinking (women: 0.970 [0.961,0.980]; men: 0.975 [0.964,0.986]), but similar odds of current heavy drinking. Among LtoM drinkers, average memory declined over time (women=0.8 & men=0.9 points/decade). In contrast, non-drinkers of both sexes had higher scores at age 65 (women: 0.087 points, 95%CI[0.034, 0.140], men: 0.04 [-0.011, 0.091]) and slower decline (women: 0.291 [0.150, 0.432], men: 0.259 [0.128, 0.390]). For heavy compared to LtoM drinkers, women had similar scores at 65 (-0.046 [-0.169, 0.077]) and similar decline (0.053 [-0.002, 0.108]), while men had higher scores at age 65 (0.125 [0.019, 0.231]), but faster decline (-0.07 [-0.127, -0.013]).
Abstinence may confer memory benefits relative to light-to-moderate drinking for both sexes, and heavy drinking may lead to faster memory decline for men. Sex differences may reflect different alcohol category definitions.
