中国人群发生2型糖尿病的风险预测列线图模型构建及验证  

Construction and Validation of Nomogram Model for Risk Prediction of Type 2 Diabetes in Chinese Population

作　　者：李路[1] 彭婧利[2] 郭焱雄[1] 李军[1] LI Lu;PENG Jingli;GUO Yanxiong;LI Jun(Emergency Department,South Hospital,Chenzhou First People's Hospital,Chenzhou 423099,China;Department of Ophthalmology,Chenzhou First People's Hospital,Chenzhou 423099,China)

机构地区：[1]湖南省郴州市第一人民医院南院区急诊科,423099 [2]湖南省郴州市第一人民医院眼科,423099

出　　处：《实用心脑肺血管病杂志》2025年第5期65-72,共8页Practical Journal of Cardiac Cerebral Pneumal and Vascular Disease

基　　金：湖南省科技创新计划项目(2021SK50311)。

摘　　要：目的构建中国人群发生2型糖尿病(T2DM)的风险预测列线图模型。方法本研究数据来源于2010—2016年中国15个独立医疗中心的211833例中国成年人。随机选取12个医疗中心的189501例受试者为训练集(进一步随机选取其中的2个医疗中心的42113例受试者为内部验证组),将剩余3个医疗中心的22332例受试者作为验证集。收集受试者临床资料,采用最小绝对收缩和选择算子(LASSO)回归分析、多因素Logistic回归分析探讨训练集受试者发生T2DM的影响因素;基于多因素Logistic回归分析结果,分别构建中国人群发生T2DM的风险预测列线图模型1(只纳入非侵入指标)和中国人群发生T2DM的风险预测列线图模型2(纳入全部指标);采用ROC曲线评估两个模型在训练集中的区分度,采用净重新分类指数(NRI)比较两个模型在训练集中的准确度,从而确定最佳中国人群发生T2DM的风险预测列线图模型;采用Bootstrap自抽样法重复抽样10000次,计算校正C指数以评估最佳列线图模型在训练集、内部验证组、验证集中的一致性;采用校准曲线评估最佳列线图模型在训练集中的校准度;采用决策曲线评估最佳列线图模型在验证集中的临床适用性。结果211833例受试者中,发生T2DM 4174例[训练集3776例(其中内部验证组640例),验证集398例]。训练集与验证集年龄、性别、BMI、吸烟者占比、饮酒者占比、收缩压、舒张压、空腹血糖(FPG)、总胆固醇(TC)、三酰甘油(TG)、血尿素氮(BUN)、估算肾小球滤过率(eGFR)比较,差异有统计学意义(P<0.05)。LASSO回归分析结果显示,年龄、性别、BMI、糖尿病家族史、收缩压、舒张压、FPG、TC、TG、丙氨酸氨基转移酶(ALT)、BUN、eGFR可能是训练集受试者发生T2DM的影响因素(P<0.05)。多因素Logistic回归分析结果显示,年龄、性别、BMI、糖尿病家族史、收缩压、FPG、TG、ALT是训练集受试者发生T2DM的独立影响因素(PObjective To construct the nomogram model for risk prediction of type 2 diabetes(T2DM)in Chinese population.Methods The data for this study was obtained from 211833 Chinese adults in 15 independent medical centers in China from 2010 to 2016.A total of 189501 subjects from 12 medical centers were randomly selected as the training set(42113subjects from 2 medical centers were further randomly selected as the internal validation group),and a total of 22332 subjects from the remaining 3 medical centers were selected as the validation set.The clinical data of the subjects were collected.Least absolute shrinkage and selection operator(LASSO)regression analysis and multivariate Logistic regression analysis were used to investigate the influencing factors of T2DM in the training set subjects.Based on the results of multivariate Logistic regression analysis,the nomogram model 1 for risk prediction of T2DM in Chinese population(including only non-invasive indicators)and the nomogram model 2 for risk prediction of T2DM in Chinese population(including all indicators)were constructed respectively.The ROC curve was used to evaluate the differentiation of the two models in the training set,and the net reclassification improvement(NRI)was used to compare the accuracy of the two models in the training set,so as to determine the best nomogram model for risk prediction of T2DM in Chinese population.Bootstrap self-sampling method was used to sample 10000 times,and the correction C-index was calculated to evaluate the consistency of the best nomogram model in the training set,internal validation group and validation set.Calibration curve was used to evaluate the calibration degree of the best nomogram model in the training set.The decision curve was used to evaluate the clinical applicability of the best nomogram model in the validation set.Results Among 211833 subjects,4174 developed T2DM[3776 in the training set(including 640 in the internal validation group)and 398 in the validation set].There were significant differences in age,g

关 键 词：糖尿病 2型 中国 列线图 

分 类 号：R587.1[医药卫生—内分泌]