目的：基于超广域扫频源光学相干断层扫描血管成像(UWF-SS-OCTA)技术探讨糖尿病视网膜病变(DR)有关的危险因素，并建立临床预测模型。方法：选取2024年7月至2024年11月于徐州医科大学附属医院就诊的2型糖尿病患者共235例235眼作为研究对象，根据是否合并DR分为无DR组(NDR组)120例120眼和非增殖型DR组(NPDR组)115例115眼。收集两组患者的一般项目及实验室检查、OCTA等数据，通过单因素分析筛选出DR相关危险因素，并对相关危险因素进行Logistic回归分析，建立DR预测模型，并采用ROC曲线、校准曲线、DCA曲线评价模型的效能。结果：糖尿病病程、空腹血糖、血尿素氮(BUN)、高血压病史、脉络膜血管指数(CVI)在模型中均有差异(均P<0.05)，其中糖尿病病程、空腹血糖、BUN、高血压病史均为影响糖尿病患者发生DR的危险因素，CVI为影响糖尿病患者发生DR的保护因素。模型预测DR发生概率的曲线下面积为0.898(0.859-0.938)，诊断阈值为0.438，对应灵敏度为87.8%、特异度为78.3%，说明该模型预测DR发生的价值较高。结论：糖尿病病程、空腹血糖、BUN、高血压病史、CVI与DR显著相关，由此建立的预测模型对DR有一定的筛查作用。

AIM: To explore the risk factors associated with diabetic retinopathy(DR)based on ultra-widefield swept-source optical coherence tomography angiography(UWF-SS-OCTA), and to establish a clinical prediction model.METHODS:A total of 235 patients(235 eyes)with type 2 diabetes mellitus who were treated in the Affiliated Hospital of Xuzhou Medical University from July to November 2024 were selected as the research objects. According to the presence or absence of DR, they were divided into 120 cases(120 eyes)in non-DR group(NDR group)and 115 cases(115 eyes)in non-proliferative DR group(NPDR group). Data on general characteristics, laboratory tests, and OCTA results were collected for both groups. Univariate analysis was employed to identify DR-related risk factors. Logistic regression analysis was conducted to analyze these risk factors and to establish a DR prediction model. The efficacy of the model was evaluated using the receiver operating characteristic(ROC)curve, calibration curve, and decision curve analysis(DCA).RESULTS: The duration of diabetes, fasting blood glucose, blood urea nitrogen(BUN), history of hypertension, and the choroidal vascular index(CVI)were found to be statistically significant in the model(all P<0.05). Specifically, the duration of diabetes, fasting blood glucose, BUN, and history of hypertension were identified as risk factors for DR among diabetic patients, while CVI was recognized as a protective factor. The area under the curve for the model predicting the probability of DR was 0.898(0.859-0.938), with a diagnostic threshold of 0.438. The corresponding sensitivity and specificity were 87.8% and 78.3%, respectively, indicating that the model possesses high predictive value for the occurrence of DR.CONCLUSION: The duration of diabetes, fasting blood glucose, BUN, history of hypertension, and CVI are significantly correlated with DR. The established prediction model demonstrates a substantial screening capability for DR.

江苏省中医药科技发展计划项目(No.QN202330)