Moderate alcohol use and health

Abstract

Background: Many western observational studies suggest moderate alcohol use is associated with better health including lower risk of cardiovascular diseases, diabetes and cognitive decline. However, the apparent benefit is susceptible to confounding by healthier attributes in moderate users. Randomized controlled trials of moderate alcohol use are infeasible. To assess the validity of these association for causal inference, I examined these associations in a setting (Southern China) with a different social patterning of alcohol use from more commonly studied western populations and using a Mendelian randomization design.

Objectives: This thesis utilized two large Southern Chinese cohorts, the Guangzhou Biobank Cohort Study (GBCS) (n=30,499) and the Elderly Health Centre (EHC) Cohort (n=64,353) to examine sex-specific association of moderate alcohol use with cognitive function using observational designs. I also examined systematic differences between alcohol users and the credibility of alcohol-metabolizing genes as instruments for Mendelian randomization in GBCS. Mendelian randomization was used to examine the effect of alcohol use on cognitive function and cardiovascular risk factors and morbidity among men in GBCS.

Methods: I used multivariable linear regression to examine the adjusted association of alcohol use categories (never, occasional, social weekly (EHC only), moderate, heavy and former) with cognitive function, measured by delayed 10-word recall test (phases 1-3 of GBCS), Mini-Mental State Examination (phase 3 of GBCS) and Abbreviated Mental Test (EHC), stratified by sex and age. I used multinomial logistic regression to examine the sex-specific systematic difference by alcohol category in GBCS. I used multivariable linear regression to examine the genetic association of ALDH2 with different cardiovascular risk factors and morbidities, cognitive outcomes and liver enzymes and to assess if alcohol phenotypes mediated any apparent genetic association in men. I used 2 stage least squares (2SLS) regression to examine the association of alcohol units (10g ethanol/day) with cognitive function and cardiovascular risk factors (blood pressure, lipids and fasting glucose) and morbidities (self reported cardiovascular disease and ischemic heart disease) in men in GBCS.

Results: Occasional alcohol use, rather than moderate alcohol use, was consistently associated with higher cognitive function in both studies. Systematic differences among alcohol users were present. Occasional alcohol users had better health attributes while moderate users had slightly poorer attributes compared to never users. Aldehyde dehydrogenase 2 (ALDH2) was a credible instrument for Mendelian randomization. From Mendelian randomization, low to moderate alcohol use was not associated with cognitive function in men. However, it was positively associated with HDL cholesterol and diastolic blood pressure but not with fasting glucose or cardiovascular morbidity in men.

Conclusions: Moderate alcohol use was not associated with cognitive function, suggesting that previous positive studies could be confounded by better health attributes in moderate users. The lack of association of alcohol use with cardiovascular morbidity despite raising HDL cholesterol is consistent with non-observational studies showing the non-causal role of HDL cholesterol in cardiovascular disease. These may suggest the apparent cardioprotection of alcohol is confounded although it remains possible that cardioprotection is population-specific via pathways other than HDL cholesterol, which require further investigations.