An algorithm to predict risk of Type 2 diabetes in Turkish adults: Contribution of C-reactive protein

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<0.001 each). An algorithm using the stated 7 variables was developed separately for each sex. Men and women in the top quintile of risk score were, respectively, 20 and 50 times and significantly more likely to develop DM than those in the bottom quintile. The predictive value of the algorithm was validated in 2 split samples. Conclusions: A marker of low grade inflammation provides useful predictive ability beyond other simple predictors in a female population with MetS prevailing. The derived simple algorithm may be useful in estimating the 8-yr risk of DM among middle-aged Turkish men and women.