Mathematical modeling of the chronic kidney disease progression rate in patients with type 2 diabetes mellitus

Abstract

Aim – to develop and provide clinical substantiation of experimental mathematical models for the chronic kidney disease (CKD) progression rate in patients with type 2 diabetes mellitus (T2DM) using the glomerular filtration rate reduction index (RI_GFR) as original diagnostic parameter. Material and methods. A cross-sectional observational study of clinical status indices was performed in a sample of patients with type 2 diabetes. The significant predictors of a high rate of CKD progression were identified by regression analysis, three variants of experimental mathematical models were developed with different combinations of arguments with an emphasis on modifiable factors. Results. The method of one-dimensional logistic regression analysis revealed the indices of clinical status having a significant impact on the rate of CKD progression on the scale of changes in RI_GFR by 1 ml/min/1.73 m2 and on the binary classification of outcomes in the groups of "slow" and "fast" decrease in kidneys filtration function with a threshold value of RI_GFR of 4.21 ml/min/1.73 m2 per year. These indices were age, body mass index (BMI), glycemia and duration of diabetes at the time of visit, history of insulin therapy, acute myocardial infarction in the anamnesis, pulse on the popliteal artery, concomitant retinopathy, hypertension risk group, treatment with sulfonylureas and calcium antagonists. Using multidimensional logistic regression, three types of experimental mathematical models were developed, including various combinations of predictors that demonstrated high values of diagnostic significance. Conclusion. The mathematical modeling of CKD progression in patients with T2DM with the RI_GFR diagnostic index allows for better understanding of the pathology's development patterns. An experimental mathematical model using modifiable drug factors that a doctor can manage during the treatment (administration of sulfonylureas and calcium channel blockers) demonstrated 55.6% sensitivity, 85.3% specificity, AUC 0.76 (0.65; 0.86), which ensured high quality prediction with an accuracy of 77.5%.