Abstract: Objective To examine the association of dietary pattern with the incidence of type 2 diabetes (T2D)in residents of Nanjing city.Methods With multi-stage cluster randomized sampling method,local residents aged over 30 years,free of diabetes and cancer in 7 communities of 2 urban districts were selected for the study.Dietary intake was measured using food frequency questionnaire (FFQ).FFQ-based dietary patterns were identified by factor analysis.The association between dietary pattern and incidence of T2D was investigated using Cox proportional hazard regression models.Results Of 3 197 eligible subjects at the baseline,2 033 completed the 3-year follow-up survey.There were 205 newly diagnosed T2D cases in the 3-year follow-up period with a cumulative incidence of 8.9%.Five dietary patterns (condiments,protein,healthy,sweet food and drinking pattern)were identified through factor analysis.After adjustment for age,sex,physical activity,smoking status,educational level,and body mass index,compared with participants in the lower tertile of the healthy pattern score (T1),the hazard ratios (HR)were 0.714 (95% confidence interval95% CI:0.514-0.993)and 0.617 (95% CI:0.429-0.848)for people with the middle and upper tertile,respectively.Contrastingly,people in the high tertile of sweet food pattern was at a significantly increased risk of developing T2D (HR=1.470,95% CI:1.043-2.072)compared to those in the low tertile.Conclusion Healthy dietary pattern might reduce,but sweet food pattern might elevate the risk of developing T2D.