The aim of the present study was to build a predictive model of nonalcoholic fatty liver disease (NAFLD) in obese children. Fifty-six obese 10-year-old children underwent blood tests for biochemical measures and magnetic resonance imaging for NAFLD diagnosis. A model combining waist-to-height ratio, homeostasis model assessment of insulin resistance, adiponectin, and alanine aminotransferase was accurate in predicting NAFLD (AUROC = 0.94 [95% confidence interval 0.89-0.99], P < 10). When adiponectin was not included in the model, the discrimination accuracy was still good (AUROC = 0.88 [95% confidence interval 0.79-0.97], P < 10). In conclusion, a predictive equation combining routinely available variables may allow physicians to identify obese children at the highest risk of NAFLD.