A high body mass index (BMI) is a major risk factor for diabetes, although little is known about the characterization of a dose-response association adjusted for potential confounders. This cross-sectional study was conducted from June 2012 to August 2012; a total of 21 435 inhabitants in Jilin Province aged between 18 and 79 years were selected randomly based on multistage, stratified cluster sampling. The estimated prevalence of diabetes was 9.1% overall, 9.4% in males and 8.9% in females. After adjusting for potential confounders, the multivariable-adjusted odds ratios for the BMI-diabetes association were 1.337 (95% confidence interval = 1.185-1.508) and 1.696 (95% confidence interval = 1.429-2.042), respectively, for overweight and obesity. Through multivariable restricted cubic spline regression, continuous variation in BMI was found to be related to diabetes in a nonlinear manner (P < .001) after adjustment for confounders in both different gender and different age groups, suggesting that there is an adjusted dose-response association between continuous BMI and diabetes, with substantial population-level effects.