Baseline and longitudinal joint associations of alcohol consumption and obesity with diabetes risk: evaluating...

The aim of this study was to examine whether the independent and joint effects of obesity and alcohol consumption on diabetes risk differ when exposures are assessed at baseline or as time-varying variables, and whether interaction patterns vary across multiplicative and additive scales in a Asian prospective cohort.

In total, 7817 participants aged ≥40 years without diabetes at baseline (2001-2002) were followed until 2017-2018, and the associations of obesity and alcohol consumption with incident diabetes and their interactions were assessed using Cox proportional hazards and Aalen's additive hazards models.

Obesity and high alcohol consumption (≥30 g/day) were consistently associated with incident diabetes across all analyses, with stronger associations observed in the time-varying models. Obesity (HR 1.91; 95% CI 1.75, 2.09; 21.9 additional cases per 1000 person-years; 95% CI 20.0, 23.8) and high alcohol consumption (HR 1.23; 95% CI 1.05, 1.44; 14.6 additional cases per 1000 person-years; 95% CI 11.1, 18.1) were both associated with increased diabetes risk. Associations for low-to-moderate alcohol consumption were weaker, and were statistically significant only on the additive scale. No significant multiplicative interactions were observed for the effect of baseline or time-varying obesity, current alcohol consumption or daily alcohol consumption on the risk of diabetes. However, significant antagonistic additive interactions were observed between baseline or time-varying obesity and current alcohol consumption: drinking <10 g/day with 12.8 fewer cases/1000 person-years (95% CI -17.4, -8.2), 10-29.9 g/day with 15.9 fewer cases (95% CI -22.1, -9.8), and ≥30 g/day with 14.6 fewer cases (95% CI -22.0, -7.2), with a p_interaction value <0.001. Joint analysis revealed that, in participants with obesity, the increased association with low-to-moderate alcohol consumption was minimal in terms of multiplicative and additive models, especially for time-varying exposures. The results were comparable in the sensitivity analysis that was restricted to baseline current drinkers, in which current drinkers consuming <10 g/day of alcohol were used as the reference group.

Obesity and high alcohol consumption were consistently associated with incident diabetes, and an antagonistic additive interaction was observed between obesity and all alcohol consumption levels. These findings highlight the importance of exposure modelling and scale choice, and support prioritising weight management and reducing high alcohol consumption in diabetes prevention.