Abstract
Introduction:
Many cohort studies have reported a protective effect of moderate alcohol use (EtOH) on the risk of adverse cardiovascular health outcomes. The exact nature of this association remains unclear in the absence of data from a randomized controlled trial.
Hypothesis:
We hypothesized that the protective effect of moderate EtOH on the risk of atherosclerotic cardiovascular disease (ASCVD) is not causal in nature and tested our hypothesis by conducting a Mendelian randomization study among European, African American, and Hispanic participants in the Women’s Health Initiative (WHI) study.
Methods:
A total of 154,643 European, African American, and Hispanic post-menopausal women participated in the WHI study. We first used logistic regression to examine the association between EtOH and ASCVD in 8,105 incident ASCVD cases and 63,094 non-cases reporting moderate use at baseline (>0 to 15 drinks / week). ASCVD outcomes included in the case group were adjudicated fatal and non-fatal myocardial infarction, coronary revascularization, angina, ischemic stroke, peripheral arterial disease, and carotid artery disease. We adjusted our analysis for baseline covariates including age, self-reported race, BMI, hypertension status, systolic blood pressure, diabetes status, smoking status, physical activity, education, and family income. We then performed an instrumental variable analysis in a subset of 1,888 cases and 9,746 non-cases with GWAS and imputed data available for rs1229984, a polymorphism in ADH1B known to be strongly associated with the degree of lifetime alcohol consumption. Lastly, we assessed whether the instrumental variable assumptions are likely to be satisfied by testing for association between rs1229984 and ASCVD as well as all risk factors for ASCVD in 3,434 WHI participants who reported never using alcohol.
Results:
Observational analysis confirmed a strong inverse linear association between the degree of moderate alcohol use and the risk of ASCVD even after adjusting for all established risk factors of ASCVD (2.2% reduction in risk per doubling of the no. of drinks/week among moderate EtOH users, p = 3.8 x 10-3). In contrast, our instrumental variable analyses among the subset of subjects with genetic data revealed that moderate alcohol use was associated with a markedly higher risk of ASCVD (47% increase in risk per doubling of EtOH use, p = 4.9 x 10-3). We found rs1229984 to be valid instrument as it was well imputed (mean r2 = 0.74), strongly associated with the degree of moderate alcohol use (F statistic = 44.4), and not associated with ASCVD or any ASCVD risk factor among never drinkers.
Conclusions:
The observational association between alcohol use and improved cardiovascular health is not causal. Our instrumental variable analysis suggests that alcohol use even among moderate users markedly increases risk of ASCVD in a multi-ethic population of post-menopausal women.