Model for prediction of diabetic retinopathy based on determination of transforming growth factor beta 1 and polymorphism rs1800470 of the TGFB1 gene

Background. The existence of a connection between transforming growth factor beta 1 (TGF-β1) and the development and progression of diabetic retinopathy (DR) led to an attempt to create an appropriate mathematical model for DR prognosis. Aim: to establish the possibility of predicting the stage of DR based on the determination of the content of TGF-β1 in the blood and intraocu­lar fluid (IOF), as well as the genotype of the rs1800470 polymorphism of the TGFB1 gene. Materials and methods. 102 people with type 2 diabetes were examined and divided into 3 groups according to the stages of DR: 1st — non-proliferative DR (35 people), 2nd — preproliferative (34 people) and 3rd — proliferative (33 people). The control group consisted of 61 people. The patients underwent standard ophthalmic examinations. Evaluation of TGF-β1 in blood serum and IOF was performed by enzyme immunoassay (Invitrogen Thermo Fisher Scientific, USA). Alleles rs1800470 were determined by real-time polymerase chain reaction (TaqMan Mutation Detection Assays, Life Technologies, USA). To analyze the relationship between the DR stage and risk factors, correlation analysis was performed and generalized linear models were built. Results. The duration of diabetes, content of TGF-β1 in the blood and IOF, central retinal thickness and volume had a positive correlation with the progression of DR (p < 0.05). The connection with DR was also shown with the compensation of diabetes (p = 0.002): in uncompensated diabetes, the progression of DR was significantly higher. The selection of a set of independent factors associated with the DR stage identified 5 factorial features: compensation of diabetes, the content of glycated hemoglobin and TGF-β1 in the blood, the content of TGF-β1 in IOF, as well as the rs1800470 genotype. The model predicted the DR stage with 100 % accuracy (p < 0.001). In the developed model, 3 features were of the greatest importance: compensation of diabetes, the rs1800470 genotype and the content of TGF-β1 in IOF. The model built on these three features is adequate (p < 0.0001) and had an accuracy of 97.0 %. Conclusions. The importance of TGF-β1 pathway disruption has been proven, an increase in its content corresponded to the severity of DR. The resulting generalized linear model allowed to clearly differentiate the stages of DR and quantitatively reflect its severity.