Fig. 2.

Multivariate ROC models built based on the Random Forest Algorithm. (a) ROC curves of the models built using a cross-validation method in a training dataset that accounted for 70% of the total patients. (b) ROC curves of the models obtained in the validation performed on a testing dataset composed of patients not used to build the models (30% remaining patients not included in the training dataset). Clinical: clinical variables included in the FINDRISC and ADA scores: age, BMI, waist circumference, physical activity, dietary consumption of fruit and vegetables, use of antihypertensive medication, history of high blood glucose, family history of diabetes in addition to gestational diabetes, HDL, triglycerides and HbA1c; Indexes: OGTT-derived indexes (HOMA-IR, Homeostasis model assessment- insulin resistance; ISI, Insulin sensitivity index; IGI, Insulinogenic index; HIRI, Hepatic insulin resistance index; MISI, Muscle insulin sensitivity index; DI, Disposition index). The models were adjusted by diet and intensity of the statin treatment including these variables in all the models.