Abstract:Objective: To investigate the risk factors for dry eye syndrome secondary to type 2 diabetes mellitus (T2DM) and to develop a nomogram model for early risk prediction.Methods: A total of 347 T2DM patients treated in our hospital between March 2020 and April 2024 were enrolled and randomly divided into training and validation sets at a 7:3 ratio. Demographic data, glycemic parameters, and clinical treatments were compared between non-dry eye syndrome (control group) and dry eye syndrome to type 2 diabetes mellitus (dry eye group) in the training set. Statistically significant indicators were incorporated into multivariate logistic regression to identify independent predictor for secondary dry eye. These factors were then used to construct a nomogram model using R software, which was subsequently validated using the validation set. Results: The percentage of patients with secondary dry eye syndrome in 242 cases of T2DM was 64.46% (156/242). Multifactorial logistic regression revealed that blood glucose variability, glycosylated serum protein, retinopathy, meibomian gland functional status, duration of T2DM, and meibomian gland opening blockage belonged to the independent predictor for secondary dry eye (OR > 1, P < 0.05). A column-line graph prediction model was constructed based on the above six indicators, and the area under the ROC curve was verified to be 0.994 (95% CI:0.989~0.999) and 0.990 (95% CI:0.977~0.999), respectively. The slopes of the calibration curves were similar, as tested by the Hosmer-Lemeshow test χ2 = 1.461, 1.566, P = 0.993, 0.992. The column-line graphical model could provide good utility for clinical decision-making.Conclusion: Glycemic variability, glycated serum protein, retinopathy, meibomian gland dysfunction, T2DM duration, and meibomian gland orifice obstruction significantly increase the risk of dry eye secondary to T2DM. The constructed nomogram model serves as a valuable tool for early risk assessment and intervention, potentially improving patient outcomes. Future studies should validate this model across diverse populations and clinical settings.
