老年糖尿病视网膜病变预测列线图的建立

Establishment of predictive nomogram for elderly diabetic retinopathy

目的建立老年糖尿病视网膜病变(DR)的临床预测列线图。方法纳入我院门诊及住院部232例老年糖尿病(DM)患者,根据是否患有DR将其分为DM组(不合并DR,n=116)和DR组(合并DR,n=116)。比较两组的一般资料;通过多因素二分类Logistic回归模型判断DR发病风险;建立DR预测模型,绘制列线图,并通过绘制受试者工作特征(ROC)曲线、Calibration曲线以及决策曲线分析(DCA)评价该模型的区分度、校准度及临床运用价值。结果DR组的DM病程长于DM组,脂肪肝、吸烟、饮酒占比及收缩压、腰臀比、空腹血糖、空腹胰岛素、糖化血红蛋白水平高于DM组,体力活动量少于DM组(P<0.05)。通过最佳子集法(基于赤池信息准则)构建DR发病的预测模型,绘制列线图,DM长病程、饮酒及空腹血糖、空腹胰岛素水平升高是DR发病的独立危险因素(P<0.05)。针对模型绘制预判DR发生的ROC曲线,曲线下面积为0.833,特异度为0.732,灵敏度0.897。模型的校准度:Calibration曲线U指数为-0.009(P=0.726),Brier评分为0.136。DCA显示,患病概率介于10%~90%的患者可从该模型获益。结论该研究建立了基于DM病程、饮酒、空腹血糖及空腹胰岛素的DR诊断预测列线图,为老年DR临床筛查提供了新的工具。

Objective To establish the clinical predictive nomogram of elderly diabetic retinopathy(DR).Methods A total of 232 elderly diabetes mellitus(DM)patients in the outpatient and inpatient departments of our hospital were included and the patients were divided into DM group(without DR,n=116)and DR group(with DR,n=116)according to whether they had DR.The general data of the two groups were compared;the risk of DR was judged by multivariate binary Logistic regression model;DR prediction model was established,nomogram was drawn,and the discrimination,calibration and clinical application value of the model were evaluated by drawing receiver operating characteristic(ROC)curve,Calibration curve and decision curve analysis(DCA).Results The duration of DM in the DR group was longer than that in the DM group,the proportions of fatty liver,smoking,drinking,and systolic blood pressure,waist hip ratio,fasting blood glucose,fasting insulin,glycosylated hemoglobin levels were higher than those in the DM group,and the amount of physical activity was less than that in the DM group(P<0.05).The prediction model of DR was constructed by the best subset method(based on Akaike information criterion),and the nomogram was drawn,long duration of DM,drinking,elevated fasting blood glucose and fasting insulin levels were independent risk factors for DR(P<0.05).The ROC curve for predicting DR was drawn base on the model,the area under the curve was 0.833,the specificity was 0.732,and the sensitivity was 0.897.The calibration of the model showed that the U index of the Calibration curve was-0.009(P=0.726),and the Brier score was 0.136.DCA showed that patients with a prevalence rate between 10%and 90%could benefit from the model.Conclusion This study established a DR diagnostic prediction nomogram based on DM duration,drinking,fasting blood glucose and fasting insulin,providing a new tool for clinical screening of DR in the elderly.