列线图预测中国成人3年后进展为2型糖尿病风险的研究

A nomogram for risk estimation in progression to type 2 diabetes mellitus after a follow-up of 3 years in Chinese adults

目的探讨个体化预测中国成人3年后进展为2型糖尿病(T2DM)风险的列线图模型。方法回顾性分析于2012年2月至2015年8月完成的中国T2DM患者肿瘤发生风险的流行病学研究(REACTION)的流行病学资料,选取2788名基线无糖尿病且完成3年随访的社区居民为研究对象。采用简单随机化方法将研究对象分为建模队列(75%,2094例)与验证队列(25%,694例)。采用logistic回归分析构建T2DM风险预测模型,选取赤池信息准则(AIC)信息量最小的logistic回归模型为最终预测模型,并绘制列线图。采用受试者工作特征(ROC)曲线比较列线图与糖尿病前期对进展为T2DM风险的预测价值,采用决策曲线分析(DCA)评价列线图的临床实用性。结果随访3年,2788名受试者中有210例(7.53%)进展为T2DM,据此估算年发病率为2.51%。在建模队列中,多因素logistic回归分析结果显示,年龄、空腹血糖、糖负荷后2 h血糖、糖化血红蛋白是T2DM发病的独立危险因素(优势比为1.03、1.13、1.01、1.17,均P<0.05),并以上述4个因素为自变量构建列线图。列线图在建模队列中预测进展为T2DM风险的ROC曲线下面积(AUC)为0.825(95%可信区间0.787~0.864),在验证队列中为0.781(95%可信区间0.716~0.846),其预测价值均优于糖尿病前期(AUC分别为0.693、0.628,P<0.01,Delong法)。DCA显示列线图具有良好的临床应用价值。结论本研究中构建的列线图可个体化预测中国成人3年后进展为T2DM的风险,有助于甄别糖尿病高风险人群并制定个体化的干预措施。

Objective To establish a nomogram for predicting the possibility of progression to type 2 diabetes mellitus(T2DM)in 3 years among Chinese adults.Methods This retrospective study was one component of the baseline and 3-year follow-up surveys conducted for the Risk Evaluation of cAncers in Chinese diabeTic Individuals:a lONgitudinal study(REACTION)from February 2012 to August 2015.In all,2788 subjects who diagnosed without diabetes at baseline and completed the 3-year survey were taken into this analysis.All participants were divided into a modeling cohort(75%,2094 cases)or a validation cohort(25%,694 cases)via simple random sampling.According to the Akaike information criterion,a logistic regression model was adopted to construct the nomogram that predicted the risk of T2DM.Receiver operating characteristic(ROC)curve was utilized to evaluate the predictive value of this nomogram.Decision curve analysis was employed to assess the clinical utility of the predictive model.Results After a follow-up of 3 years,210 cases(7.53%)of the 2788 non-diabetic participants developed diabetes,with the annual incidence estimated to be 2.51%.In the modeling cohort,multivariate logistic regression analysis exhibited that age,fasting plasma glucose,2-hour postprandial plasma glucose,glycated hemoglobin A1c(HbA1c)were independent risk factors for T2DM(odds ratio were 1.03,1.13,1.01,1.17,respectively,all P<0.05).Therefore,we used these four factors as independent variables to construct our nomogram.The area under the curve(AUC)of the nomogram for predicting the possibility of progression to T2DM in modeling cohort was 0.825(95%confidence interval 0.787-0.864),and the AUC of the nomogram established above in the validation cohort was 0.781(95%confidence interval 0.716-0.846).The predictive value of such nomogram is better than that of prediabetes(P<0.01,Delong′s test).Decision curve analysis suggested that the nomogram which was established in this study exhibited clinical application value.Conclusions The nomogram constructed in this study can effectively predict the risk of progression to T2DM after a follow-up of 3 years,which can help to identify patients with high-risk of diabetes and develop individualized interventions.