Background: Moderate alcohol consumption has been reported to be associated with lower risk for diabetes with some studies showing a U-shaped association. Whether and how the association might differ by gender or obesity status is controversial. Objective: To evaluate the prospective association between alcohol consumption and the long-term risk of diabetes in the Atherosclerosis Risk in Communities (ARIC) Study. Methods: A prospective analysis of 11,263 ARIC participants without prevalent diabetes (55% women, 81%white, mean age 54 years). Alcohol consumption was assessed at visit 1 (1987-1989). Participants were followed-up for incident diabetes defined by fasting glucose more than 126 mg/dL, non-fasting glucose more than 200 mg/dL, self-reported diagnosis of diabetes or use of diabetic medication. We used Cox models to estimate hazard ratios of diabetes risk by drinking categories in women and men, respectively. Results: During a median follow-up of 21 years, there were 3518 incident diabetes cases. In the fully adjusted model, compared to never drinkers, among women, 7-14 drinks/week was associated with a significantly lower risk of diabetes; whereas among men, 14-21 drinks/week was associated with a significantly lower risk ( Table ). There was a significant interaction between drinking categories and smoking status or between drinking categories and body mass index in women. Among women, a U-shaped association was mainly present among non-smokers, and significant decreasing risk is only found among normal-weight and overweight participants, but not obese participants. Conclusion: Low levels of alcohol intake (1-2 drinks per day for women and 2-3 drinks per day for men) are inversely associated with diabetes risk. The association is modified by smoking and body mass index in women.