山西省2型糖尿病患者非增殖性糖尿病视网膜病变现况调查及预测模型的构建

A survey of the current status of non-proliferative diabetic retinopathy in patients with type 2 diabetes mellitus in Shanxi province and the construction of the predictive model

目的调查山西省2型糖尿病(T2DM)患者糖尿病视网膜病变(DR)的现况,并构建T2DM非增殖性DR(NPDR)发病风险的预测模型。方法依托中国慢性病及其危险因素监测系统,采用多阶段分层随机抽样方法,自2019年6至8月选取山西省2个区和2个县1891例糖尿病患者进行现况调查,符合本研究纳入排除标准且资料完整的T2DM患者共1276例。收集T2DM患者的一般资料和实验室数据,包括T2DM病程、收缩压、糖化血红蛋白(HbA_(1c))、总胆固醇(TC)、尿微量白蛋白/肌酐比值(UACR)、血肌酐(Scr)、估算的肾小球滤过率(eGFR)。根据眼底镜检查结果将T2DM患者分为非DR(NDR)组(1119例)和NPDR组(157例),并采用随机数字表法按6∶4比例随机分为训练集(766例)和验证集(510例)。采用多因素logistic逐步回归法筛选模型变量,构建T2DM患者NPDR发病风险的列线图模型,利用受试者工作特征曲线、拟合优度曲线和临床决策曲线综合评估模型的预测价值、校准度和临床实用性。结果山西省T2DM患者中NPDR患病率为12.3%(157/1276)。在训练集中,与NDR组相比,NPDR组的T2DM病程、收缩压、HbA_(1c)、TC、UACR和Scr升高,eGFR下降,差异均具有统计学意义(P<0.05)。二元logistic回归分析结果显示,HbA_(1c)≥7%(OR=3.254)、SBP≥140 mmHg(1 mmHg=0.133 kPa)(OR=1.739)、UACR≥30 mg/g(OR=1.892)和Scr升高(OR=1.015)是T2DM伴发NPDR的危险因素,年龄升高(OR=0.971)是T2DM伴发NPDR的保护因素(均P<0.05)。训练集和验证集的受试者工作特征曲线下面积分别为0.756(95%CI 0.705~0.807)和0.739(95%CI 0.674~0.803),Hosmer-Lemeshow检验显示模型拟合度良好,预测结果接近于实际,临床决策曲线提示本模型具有临床获益。结论山西省T2DM患者中NPDR患病率为12.3%。本研究构建了预测T2DM患者NPDR发病风险的列线图模型,对早期甄别NPDR高风险人群具有一定的临床意义。

Objective To investigate the current status of diabetic retinopathy(DR)in type 2 diabetes mellitus(T2DM)in Shanxi province and to construct a predictive model for the risk of developing non-proliferative DR(NPDR)in T2DM.Methods Based on the Chinese Chronic Disease and its Risk Factor Surveillance System,a multi-stage stratified random sampling method was used to select 1891 patients with diabetes mellitus in two districts and two counties in Shanxi province from June,2019 to August,2019 for a status survey.A total of 1276 patients with T2DM who met the nadir criteria and had complete data for this study were selected.The general information and laboratory data were collected,including T2DM course,systolic blood pressure,glycated hemoglobin A_(1c)(HbA_(1c)),total cholesterol(TC),urinary microalbumin-creatinine ratio(UACR),serum creatinine(Scr),and estimated glomerular filtration rate(eGFR).According to the results of fundus microscopy,T2DM patients were divided into non-DR(NDR,1119 cases)and NPDR group(157 cases),and T2DM patients were randomly divided into training set(766 cases)and validation set(510 cases)by random number table method in a ratio of 6∶4.The multifactor logistic stepwise regression method was used to screen the model variables.The study constructed a column line graph model for the risk of developing NPDR in T2DM patients,and the predictive value,calibration and clinical utility of the model were comprehensively evaluated using subject working characteristic curves,goodness-of-fit curves and clinical decision curves.Results The prevalence of NPDR among T2DM patients in Shanxi province was 12.3%(157/1276).In the training set,compared with the NDR group,the duration of T2DM,systolic blood pressure,HbA_(1c),TC,UACR and Scr in the NPDR group were elevated and eGFR was decreased,and the differences were all statistically significant(P<0.05).Binary logistic regression analysis suggested that HbA_(1c)≥7%(OR=3.254),systolic blood pressure≥140 mmHg(1 mmHg=0.133 kPa)(OR=1.739),UACR≥30 mg/g(OR=1.892)and elevated Scr(OR=1.015)were risk factors for T2DM with NPDR,and advanced age(OR=0.971)was a protective factor for T2DM with NPDR(all P<0.05).The area under the working characteristic curve for subjects in the training and validation sets was 0.755(95%CI 0.700-0.806)and 0.737(95%CI 0.670-0.801),respectively.TheHosmer-Lemeshow test showed that the model fit was good and the prediction results were close to the actual,clinical decision curve,suggesting that the model had clinical benefit.Conclusion The prevalence of NPDR among T2DM patients in Shanxi province was 12.3%.In this study,we constructed a nomogram mode to predict the risk of NPDR in T2DM patients,which has clinical significance for early screening of people at high risk of NPDR.