Five-year systemic complications in diabetic retinopathy with integrated optical coherence tomography angiography and glycated hemoglobin

Objective This study aimed to develop and validate a multimodal prediction model integrating optical coherence tomography angiography (OCTA) and glycated hemoglobin (HbA1c) for assessing the 5-year risk of severe systemic complications in patients with diabetic retinopathy (DR). Methods A total of 340 patients with type 2 diabetes and DR were retrospectively enrolled from January 2020 to December 2024. Participants were randomly allocated into training ( n = 238) and validation ( n = 102) sets at a 7:3 ratio. Univariate analysis, Least Absolute Shrinkage and Selection Operator (LASSO) regression, and multivariate logistic regression, and were applied to identify key predictors. Three models—logistic regression, gradient boosting machine, and convolutional neural network—were constructed. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) values were used to interpret the optimal model. Results Baseline characteristics were balanced between training and validation sets ( P > 0.05). The results of multivariate logistic regression analysis identified history of cardiovascular disease, duration of diabetes, HbA1c, foveal avascular zone (FAZ) area, and urinary albumin-to-creatinine ratio (UACR) as independent influencing factors for the development of systemic complications within 5 years (all P < 0.05). Among the models, the convolutional neural network exhibited superior discrimination and clinical net benefit in both training (AUC = 0.853, 95% CI: 0.797–0.909) and validation sets (AUC = 0.820, 95% CI: 0.706–0.933). SHAP analysis indicated that FAZ area contributed most to predictions, supporting model interpretability. Conclusion A multimodal prediction model incorporating OCTA and HbA1c was successfully developed and validated. The convolutional neural network demonstrated optimal predictive performance and clinical utility, offering a quantitative tool for early identification of high-risk patients and individualized management planning.