目的：探究血糖变异系数(FPG-CV)与非增殖型糖尿病视网膜病变(NPDR)患者黄斑区形态和微循环的关系。

方法：回顾性分析2018-02/2022-06我院收治的82例82眼NPDR患者为研究对象，另选取同期82例82眼糖尿病无视网膜病变(NDR)患者作为对照组，分析两组患者的临床资料。通过多因素Logistic回归分析影响NPDR发病的危险因素，建立反向传播(BP)神经网络模型，并进行模型评价。采用Pearson相关性分析FPG-CV与患者黄斑区形态和微循环各指标的相关性。

结果：多因素Logistic回归分析结果显示，病程≥7.2a、糖化血红蛋白(HbA1c)≥7.7%、甘油三酯(TG)≥1.9mmol/L、尿微量白蛋白(MALB)≥24.5mg/L、FPG-CV≥9.8%、浅层毛细血管丛血流密度(SCP-VD)<27.6%、深层毛细血管丛血流密度(DCP-VD)<47.7%、黄斑中心凹无血管区(FAZ)面积≥0.38mm²、黄斑中心凹神经视网膜厚度(CRT)≥197.7μm以及黄斑中心凹下脉络膜厚度(SFCT)<227.7μm均为NPDR发病的危险因素(P<0.05)。隐含层节点数为5，受试者工作特征曲线(ROC)、校准曲线和临床决策曲线显示该预测模型的区分度、准确性和有效性均较好。Pearson相关性分析结果显示，FPG-CV与SCP-VD、DCP-VD以及SFCT均呈负相关(P<0.05)； FPG-CV与FAZ面积以及CRT均呈正相关(P<0.05)。

结论：病程、HbA1c、TG、MALB、FPG-CV、SCP-VD、DCP-VD、FAZ面积、CRT以及SFCT均与NPDR发病相关。随着FPG-CV升高，黄斑区形态和微循环各指标发生改变。FPG-CV与SCP-VD、DCP-VD以及SFCT呈负相关，与FAZ面积以及CRT呈正相关。

METHODS: A retrospective analysis of 82 cases(82 eyes)with NPDR admitted to our hospital from February 2018 to June 2022 was the research object, and another 82 cases(82 eyes)of non-diabetic retinopathy(NDR)patients during the same period were selected as the control group, and the clinical data of the two groups of patients were analyzed. Multivariate Logistic regression was used to analyze the risk factors affecting the incidence of NPDR, and the back propagation(BP)neural network model was established and evaluated. Pearson correlation was used to analyze the correlation among FPG-CV and macular morphology and microcirculation in patients.

RESULTS: The results of multivariate Logistic regression analysis showed that the disease duration ≥7.2a, glycated hemoglobin A1c(HbA1c)≥7.7%, triglyceride(TG)≥1.9 mmol/L, microalbuminuria(MALB)≥24.5 mg/L, FPG-CV ≥9.8%, superficial capillary plexus-vessel density(SCP-VD)<27.6%, deep capillary plexus-vessel density(DCP-VD)<47.7%, foveal avascular zone(FAZ)area ≥0.38 mm², central retinal thickness(CRT)≥197.7 μm and subfoveal choroidal thickness(SFCT)<227.7 μm were risk factors for NPDR(P<0.05). The number of hidden layer nodes is 5, and the receiver operating characteristic(ROC)curve, calibration curve and clinical decision curve show that the prediction model has good discrimination, accuracy and validity. The results of Pearson correlation analysis showed that FPG-CV was negatively correlated with SCP-VD, DCP-VD and SFCT(P<0.05); FPG-CV was positively correlated with FAZ area and CRT(P<0.05).

CONCLUSION: The course of disease, HbA1c, TG, MALB, FPG-CV, SCP-VD, DCP-VD, FAZ area, CRT and SFCT are all related to the pathogenesis of NPDR. With the increase of FPG-CV, the indexes of macular morphology and microcirculation changed. FPG-CV was negatively correlated with SCP-VD, DCP-VD and SFCT and positively correlated with FAZ area and CRT.