糖尿病黄斑水肿抗VEGF治疗后视力获益早期预测模型构建与验证  

Construction and verification of an early prediction model for visual benefit of diabetic macular edema after anti-vascular endothelial growth factor treatment

作　　者：颜宇[1] 钟琴 陈妍鹏[2] 杨蕾 李港逸 李爽乐[1] YAN Yu;ZHONG Qin;CHEN Yanpeng;YANG Lei;LI Gangyi;LI Shuangle(Department of Ophthalmology,Zigong First People’s Hospital(Zigong Academy of Medical Sciences),Zigong 643000,Sichuan Province,China;Department of Ophthalmology,the First Affiliated Hospital of Hebei North University,Zhangjiakou 075000,Hebei Province,China)

机构地区：[1]自贡市第一人民医院眼科,四川省自贡市643000 [2]河北北方学院附属第一医院眼科,河北省张家口市075000

出　　处：《眼科新进展》2025年第4期298-304,共7页Recent Advances in Ophthalmology

基　　金：河北省医学科学研究课题(编号:20211086)。

摘　　要：目的基于临床资料、OCTA、血清脑组织水通道蛋白4(AQP4)mRNA、总胆红素(TBIL)构建糖尿病黄斑水肿(DME)抗血管内皮生长因子(VEGF)治疗后视力获益的早期预测模型,并进行验证。方法选取2021年10月至2024年3月自贡市第一人民医院收治的480例(480眼)DME患者,按照21分为建模集(320例)、验证集(160例)。将建模集根据抗VEGF治疗后视力获益情况分为获益组(80例)、未获益组(240例)。收集两组患者基线资料,分析DME患者抗VEGF治疗后视力获益的影响因素,构建早期预测模型,并对该模型进行内部、外部验证。结果组间比较结果显示,未获益组患者糖尿病病程长于获益组,吸烟患者占比、最佳矫正视力(BCVA)最小分辨角对数(logMAR)视力、糖化血红蛋白(HbAlc)、AQP4 mRNA均高于获益组,中心凹视网膜深层毛细血管血流密度(DCP-VD)、黄斑中央厚度(CMT)、TBIL均低于获益组(均为P<0.05)。LASSO-Logistic回归分析显示,DME抗VEGF治疗后视力获益的影响因素为CMT、BCVA(logMAR)、HbAlc、AQP4 mRNA、中心凹DCP-VD、TBIL。基于上述影响因素构建的抗VEGF治疗后视力获益的列线图模型的预测风险一致性指数为0.844。受试者工作特征曲线(ROC)显示,该模型在建模集中预测的曲线下面积(AUC)为0.844(95%CI:0.797~0.891),验证集中预测的AUC为0.898(95%CI:0.847~0.949)。决策分析曲线显示,建模集高风险阈值在0~82%,验证集高风险阈值在0~100%时,该模型可带来临床净收益。结论CMT、BCVA(logMAR)、HbAlc、AQP4 mRNA、中心凹DCP-VD、TBIL是DME抗VEGF治疗后视力获益的影响因素,基于此构建的视力获益预测模型准确性、稳定性较高,可作为临床预测治疗后视力获益的有效工具。Objective To construct and verify an early prediction model for visual benefit of diabetic macular edema(DME)after anti-vascular endothelial growth factor(VEGF)treatment based on clinical data,optical coherence tomography angiography(OCTA),serum brain tissue aquaporin-4(AQP4)mRNA and total bilirubin(TBIL)levels.Methods A total of 480 patients(480 eyes)with DME treated in the First People’s Hospital of Zigong City from October 2021 to March 2024 were selected and divided into a modeling set(320 cases)and a validation set(160 cases)at a ratio of 2 GA6FA 1.According to the visual benefit after anti-VEGF treatment,patients in the modeling set were further divided into a benefit group(80 cases)and a non-benefit group(240 cases).The baseline data of the two groups of patients were collected,and the factors influencing visual benefits in DME patients after anti-VEGF treatment were analyzed.An early prediction model was constructed and validated both internally and externally.Results The inter-group comparison results showed that the diabetes duration in the non-benefit group was longer than that in the benefit group(P<0.05).The proportion of smokers,the best corrected visual acuity(BCVA),the minimum resolution angle(logMAR)vision,hemoglobin A1c(HbAlc)and AQP4 mRNA levels were higher in the non-benefit group than those in the benefit group(all P<0.05).The foveal retinal deep capillary plexus blood flow density(DCP-VD),central macular thickness(CMT),and TBIL levels were lower in the non-benefit group than those in the benefit group(all P<0.05).The least absolute shrinkage and selection operator(LASSO)-Logistic regression analysis showed that the factors influencing visual benefit in DME patients after anti-VEGF treatment were CMT,BCVA(logMAR),HbAlc,AQP4 mRNA levels,foveal DCP-VD,and TBIL levels.The predictive risk consistency index of the nomogram model constructed based on the above-mentioned influencing factors for visual benefit prediction after anti-VEGF treatment was 0.844.The receiver operating characteristic(ROC)

关 键 词：糖尿病 黄斑水肿 LASSO-Logistic回归模型 列线图模型 危险因素 影响因素 

分 类 号：R774.5[医药卫生—眼科]