Prediction of the risk of developing diabetic retinopathy based on blood molecules and retinal features

Background: Diabetic retinopathy (DR) has rapidly become the leading blinding eye disease threatening the working population. We aimed to explore molecular biomarkers and retinal features and build prediction models of DR. Methods: Participants from UK Biobank were recruited from 2006 to 2010, and prospectively followed up until 2021. We divided the enrolled population according to the full-course DR into 5 groups: no diabetes mellitus (no DM), prediabetes (pre DM), diabetes mellitus (DM), non-proliferating diabetic retinopathy (NPDR), and proliferating diabetic retinopathy (PDR). The molecular biomarkers evaluated at baseline includes 7 lipids and 8 proteins, while the retinal features were measured by Optical coherence tomography (OCT). The associations between molecular biomarkers and retinal features were performed by correlation analysis. A predictive model of DR was constructed using both retinal features and molecular biomarkers. Result: The study included 3953 participants (2095 [53.0%] female), with a mean age (SD) of47.3 (5.8) years. Apo A, Apo B, HDL, LDL were associated with retinal nerve fiber layer (RNFL), ganglion cell layer (GCL), and inner nuclear layer (INL) in full-course DR (all r range from 0.5 to 1 0.5, p< 0.05). Protein biomarkers, including albumin, total protein, creatine, showed significant correlation with GCL and RNFL in pre DM, DM, NPDR, and PDR groups compared to the no DM group (all p< 0.05). The Area Under Curve (AUC) of the DR prediction model based on the combination of molecular biomarkers and retinal features is 0.790 (95%CI:0.711-0.847), p <0.01, which is higher than the prediction models based on molecular biomarkers or retinal features alone. Conclusion: Molecular biomarkers were associated with retinal features during the full-course DR. DR prediction model based on the combination of molecular biomarkers and retinal features presented a higher AUC, suggesting a possible strategy for early diagnosis of DR.