Background: We aimed to determine the association between nonalcoholic fatty liver disease (NAFLD) and diabetes risk among body size phenotypes which was based on cross-classification of body mass index (BMI) categories (normal or overweight/obesity) and metabolic status (metabolically health or metabolically at-risk). Methods: We conducted a cross-sectional analysis using a cohort of 10,761 apparently healthy Chinese adults who underwent comprehensive health checkups including abdominal ultrasonography. Subjects were classified as metabolically at-risk by having any two of the following, consistent with the Adult Treatment Panel-III metabolic syndrome definition: (1) systolic/diastolic blood pressure >= 130/85 mmHg, (2) triglycerides >= 1.7 mmol/L, (3) fasting blood glucose >= 5.6 mmol/L, (4) HDL-cholesterol >= 1.0/1.3 mmol/L for men/women. Results: Among participants without metabolically at-risk, multivariate-adjusted odds ratios (ORs) for diabetes from NAFLD compared with those without NAFLD in the normal-weight (BMI <23 kg/m(2)) and overweight/obese (BMI >= 23 kg/m(2)) group were 2.10 (1.85-3.93) and 1.85 (1.35-2.53), respectively. Among participants with metabolically at-risk, the significant association between NAFLD and diabetes was lost, regardless of obesity status. There were only 27.1 % subjects with the presence of the three factors (overweight/obesity, NAFLD, and metabolically at-risk) occurring together, while the three factors occurring together was common (56.16 %) in diabetic individuals. The multivariate-adjusted ORs for diabetes were 1.1 (0.61-1.98) for overweight/obesity, 2.23 (1.05-5.14) for NAFLD, and 8.04 (5.0-12.09) for metabolically at-risk. The OR for the presence of all the three factors occurring together was 23.22 (13.96-38.63). Conclusions: NAFLD was associated with diabetes risk among participants without metabolically at-risk. The clustering of overweight/obesity, NAFLD, and metabolically at-risk is common in diabetic subjects and strikingly and markedly increases the diabetes risk.