Abstract

Background: In observational studies, mainly from Western Caucasian populations, moderate alcohol use has been shown to be associated with a lower risk of diabetes. However, whether the protection is due to the attributes of alcohol or to those of moderate alcohol users is difficult to disentangle. A population with a different distribution of alcohol consumption and diabetes prevalence from Western populations was studied as a counterfactual comparison.

Methods: Baseline data from the Guangzhou Biobank Cohort Study phases 1 and 2 (2003–6) were used to examine the adjusted associations, using multivariable censored linear regression, of alcohol use with fasting blood glucose in older (⩾50 years) men (n = 5740) and women (n = 14 274) from southern China. Moderate alcohol use was defined as weekly drinking of ⩽210 g alcohol in men and ⩽140 g in women. The expected associations of alcohol with high-density lipoprotein (HDL) cholesterol and blood pressure were also checked.

Results: Moderate alcohol users had very similar fasting glucose levels to never users, adjusted for age, sex, socioeconomic status, smoking and physical activity. In contrast, moderate alcohol users had higher HDL-cholesterol by 0.05 mmol/litre (95% CI 0.02 to 0.07). Excessive (more than moderate) alcohol users had higher fasting glucose.

Conclusions: In an understudied population with a different pattern of alcohol use from the populations usually studied, the biologically expected effects of moderate alcohol use were seen, but there was little effect on fasting glucose. Although cross-sectional studies cannot be conclusive, this pattern of findings, if confirmed, suggests that moderate alcohol use may not affect fasting glucose, although excessive use may be a risk factor.