Method for predicting the risk of type 2 diabetes mellitus development in persons with visceral obesity and prediabetes

Aim. To create a mathematical model, which will predict the development of type 2 diabetes mellitus (DM 2) in individuals with visceral obesity and/or prediabetes. Materials and methods. Clinical and laboratory data of 330 patients were analyzed. Multivariate regression and cosinor analysis determined the most sensitive parameters influencing the development of DM 2. With the help of discriminant linear analysis, a mathematical model for predicting DM 2 was built, with confirmation of its quality by ROC analysis. Results. In the studied groups (DM 2), prediabetes and without carbohydrate metabolism disorders (n=110), statistically significant correlations were obtained: between basal body temperature (BBT) and daily energy value – DEV (r=0.5; p0.0001), circadian rhythm amplitude glycemia and waist circumference (r=-0.7; p=0.004), age and BBT (r=0.5; p0.001). In groups without carbohydrate metabolism disorders and prediabetes, multiple regression analysis identified significant factors influencing the development of DM 2: daily amplitude of BBT, daily amplitude of glycemia and bedtime (p=0.001), DEV and meal time (p=0.0001). Cosinor analysis of the daily model of glycemia and BBT established an amplitude-phase shift (p=0.028; p=0.012). Linear discriminant analysis yielded a predictive model: D=-16.845 + age х 0.044 + gender х 0.026 + amplitude of circadian rhythm of BBT х 1.424 + amplitude of circadian rhythm of glycemia х 11.155 + bedtime х 0.054 + DEV х 0.0001 + waist circumference х 0.022 + glycated hemoglobin х 1.19, where -16.845 – constant, 0.044, 0.026, 1.424, 11.155, 0.054, 0.0001, 0.022, 1.19 – coefficients of the linear discriminant function. At D0 no development of DM 2 is predicted, at D0 the development of DM 2 is in the near future. Sensitivity ratio – 92.5%, specificity – 79.1% (ROC analysis). Conclusion. The presented predictive model has a high (92.5%) sensitivity due to the combination of 2 mathematical analyses. Most of the applied parameters are modifiable, which makes it possible to apply this model at the preventive stage.