Abstract
BACKGROUND AND AIM: An algorithm for predicting Type 2 diabetes (DM) risk in a population with prevalent metabolic syndrome (MetS) is needed since ethnicity influences the pathogenesis of DM. MATERIAL AND METHODS: The 8- yr risk of DM was estimated in 2261 middle-aged Turkish adults free of DM at baseline who were followed for over 7.6 yr. DM newly developed in 212 subjects. Cox proportional hazard regression and 15 variables were used to predict DM. Discrimination was assessed with area under receiver operating characteristics curve (AROC). RESULTS: In multivariable analysis, height, family income brackets, systolic blood pressure, smoking status, alcohol usage, and HDL-cholesterol levels were not predictive in either sex. In addition to sex, family history of DM, fasting glucose, and waist circumference were predictors, in men, age and non-HDL-cholesterol, while in women physical inactivity and serum C-reactive protein were so. AROC of the final model was 0.783 in men, 0.772 in women (p