Alcohol consumption and cardiovascular disease: a two-sample Mendelian randomization study with multivariable adjustment for smoking and body mass index

Abstract

Background. Observational studies of the association between alcohol consumption and cardiovascular disease may be confounded by correlated lifestyle factors. We used Mendelian randomization (MR) to assess the causal relationship between alcohol consumption and cardiovascular disease.

Methods. We conducted a two-sample MR using 35 genetic variants strongly associated (p < 5×10⁻⁸) with alcohol consumption from the GSCAN consortium, assessing seven cardiovascular outcomes: coronary artery disease (CAD), myocardial infarction (MI), atrial fibrillation (AF) and four stroke subtypes. Inverse-variance-weighted (IVW) MR with MR-Egger, weighted-median and weighted-mode sensitivity analyses was the primary analysis; multivariable MR (MVMR) was adjusted for smoking initiation and body mass index (BMI).

Results. The 35 instruments had a mean F-statistic of 76.2 (minimum 29.8), indicating strong instruments. Genetically predicted alcohol consumption showed borderline associations with CAD (OR 1.27, 95% CI 0.97–1.67, p = 0.079) and large-artery stroke (OR 1.61, 95% CI 0.96–2.71, p = 0.071) in univariate MR; CAD sensitivity analyses were significant for the weighted-median (OR 1.48, p < 0.001) and weighted-mode (OR 1.46, p < 0.001) methods. Substantial heterogeneity was detected (CAD: Q = 84.3, p < 0.001), the MR-Egger intercept indicated pleiotropy for MI (p = 0.045), and leave-one-out analysis identified rs1229984 (ADH1B) as highly influential. In MVMR using SNPs available for all three exposures (n = 28 for CAD, n = 30 for large-artery stroke), associations were attenuated for both CAD (OR 1.20, 95% CI 0.82–1.75, p = 0.355) and large-artery stroke (OR 1.26, 95% CI 0.69–2.30, p = 0.464).

Conclusions. After accounting for smoking and BMI, we found no evidence for independent causal effects of alcohol consumption on cardiovascular outcomes, despite robust associations in some sensitivity analyses. Observational associations may be explained by confounding from correlated lifestyle factors, consistent with the heterogeneity and pleiotropy seen in univariate analyses and the attenuation in MVMR.