糖尿病黄斑水肿患者康柏西普疗效诺莫图预测模型的构建  

Construction of a Nomogram prediction model for the efficacy of Conbercept in treating diabetic macular edema

作　　者：刘淼[1] 金昱[1] 元芳秀 王玲 吴磊 LIU Miao;JIN Yu;YUAN Fangxiu;WANG Ling;WU Lei(Department of Ophthalmology,Nanchang First Hospital,Nanchang 330038,Jiangxi Province,China;Department of Ophthalmology,Jiangsu Provincial Hospital of Traditional Chinese Medicine,Affiliated Hospital of Nanjing University of Traditional Chinese Medicine,Nanjing 210000,Jiangsu Province,China)

机构地区：[1]南昌市第一医院眼科,江西省南昌市330038 [2]南京中医药大学附属医院,江苏省中医院眼科,江苏省南京市210000

出　　处：《眼科新进展》2024年第9期702-706,共5页Recent Advances in Ophthalmology

基　　金：江西省南昌市科技支撑计划项目重点项目(编号:洪科字[2021]129号-2)。

摘　　要：目的 探讨糖尿病黄斑水肿(DME)康柏西普疗效的预警因素,并基于预警因素构建疗效的诺莫图预测模型。方法 选取2021年1月至2023年3月南昌市第一医院收治的经康柏西普治疗的DME患者269例(269眼),根据治疗后3个月疗效分为有效组、无效组。对康柏西普疗效进行单因素分析,以随机森林法对康柏西普疗效特征变量进行筛选与降维,以Logistic回归分析康柏西普疗效的相关因素,R语言绘制康柏西普疗效的诺莫图预测模型,绘制决策分析曲线(DCA曲线)评价诺莫图预测模型的临床效用。结果 无效组患者糖尿病病程、饮酒史患者占比、空腹血糖、餐后2 h血糖、糖化血红蛋白均高于有效组,黄斑中心视网膜厚度、中心凹视网膜深层毛细血管丛血流密度均低于有效组(均为P<0.05)。随机森林算法重要性排序前5的康柏西普疗效的预测因素为糖化血红蛋白、黄斑中心视网膜厚度、空腹血糖、餐后2 h血糖、中心凹视网膜深层毛细血管丛血流密度。Logistic回归分析结果显示,糖化血红蛋白(OR=5.012)、空腹血糖(OR=3.877)、餐后2 h血糖(OR=4.231)是康柏西普疗效的相关危险因素,黄斑中心视网膜厚度(OR=0.409)、中心凹视网膜深层毛细血管丛血流密度(OR=0.410)是康柏西普疗效的相关保护因素(P<0.05)。诺莫图显示,其C-index为0.900(95%CI:0.859~0.941),预测敏感度为90.58%,特异度为75.64%;DCA曲线显示,采用该诺莫图预测模型预测康柏西普疗效能取得临床正向净获益,提示其具有一定临床效用。结论 DME患者的康柏西普疗效受多种因素影响,涉及糖化血红蛋白、空腹血糖、餐后2 h血糖、黄斑中心视网膜厚度、中心凹视网膜深层毛细血管丛血流密度,基于以上因素构建的诺莫图模型或可作为早期预测患者治疗应答的一个方法,为临床决策提供循证依据。Objective To investigate the early warning factors for the efficacy of Conbercept in treating diabetic macular edema(DME)and build a Nomograph prediction model based on the early warning factors.Methods A total of 269 DME patients(269 eyes)treated with Conbercept at Nanchang First Hospital from January 2021 to March 2023 were selected and divided into an effective group and an ineffective group according to the therapeutic effect at 3 months after treatment.Single factor analysis was made on the efficacy of Conbercept.The random forest method was used to screen and reduce the dimension of the characteristic variables on the efficacy of Conbercept,Logistic regression was used to analyze the relevant factors affecting the efficacy of Conbercept,and R language was used to draw the Nomograph prediction model on the efficacy of Conbercept.The decision curve analysis(DCA)was made to evaluate the clinical effectiveness of the Nomograph prediction model.Results The duration of diabetes,drinking history,fasting blood glucose,2-hour postprandial blood glucose,and glycosylated hemoglobin of patients in the ineffective group were higher than those in the effective group,while the macular central retinal thickness and vessel density in the foveal retinal deep capillary plexus were lower than those in the effective group(all P<0.05).According to the random forest algorithm,the top five predictive factors for the efficacy of Conbercept were glycosylated hemoglobin,macular central retinal thickness,fasting blood glucose,2-hour postprandial blood glucose,and vessel density in the foveal retinal deep capillary plexus.Logistic regression analysis showed that glycosylated hemoglobin(OR=5.012),fasting blood glucose(OR=3.877),and 2-hour postprandial blood glucose(OR=4.231)were risk factors for the efficacy of Conbercept,while the macular central retinal thickness(OR=0.409)and vessel density in the foveal retinal deep capillary plexus(OR=0.410)were protective factors for the efficacy of Conbercept(all P<0.05).The Nomograph showed that

关 键 词：糖尿病 黄斑水肿 康柏西普 预测模型 诺莫图 

分 类 号：R774.5[医药卫生—眼科]