Abstract
Background. Many prognostic models have been developed to predict diabetic complications, but predicting the progression of diabetic retinopathy (DR) in type 2 diabetes mellitus (T2DM) remains a challenge, as in the early stages of DR there are often no clear ophthalmologic criteria. Aim: develop a prognostic model of diabetic retinopathy progression based on the analysis of clinical data and blood glucose, glycated hemoglobin, and lipid metabolism. Materials and methods. We examined 358 patients (358 eyes) with T2DM and DR, who were divided into groups: Group 1 - with nonproliferative DR (NPDR; 189 eyes), Group 2 - with preproliferative DR (PPDR; 96 eyes) and Group 3 - with proliferative DR (PDD; 73 eyes). Patients were examined using ophthalmologic methods, fasting glucose, glycated hemoglobin (HbA1c), cholesterol, high (HDL), low (LDL) and very low (VLDL) density lipoproteins, triglycerides were determined in the blood serum by colorimetric method. The results of the study were analyzed using the EZR v.1.54 package (Austria). Conclusion. There was no significant difference between the groups in terms of age and duration of T2DM (p>0.05). Blood pressure and HbA1c gradually increased in the groups (p<0.05). The content of cholesterol, lipoprotein fractions and triglycerides in groups 2 and 3 did not differ significantly, but was 1.05-1.19 times higher than in group 1 (p<0.05). A positive effect on the risk of NPDR progression was found in the blood levels of VLDL, HbA1c, LDL, cholesterol, triglycerides, fasting glucose, as well as diastolic and systolic blood pressure (SBP). In a multivariate regression analysis, a 5-factor model of NPDR progression was built, which included age, SBP, and blood levels of HbA1c, LDL-C, and VLDL-C. The model had very good prediction quality (AUC=0.87) and high specificity (91.5%). Thus, predictors of NPDR progression were advanced age, high SBP, and increased blood HbA1c, LDL-C, and VLDL-C. The model can be recommended for confirming a high risk of NPDR progression in questionable clinical cases or as a criterion for evidence-based medical prognosis in appropriate expert systems