新诊断2型糖尿病患者酮症倾向的危险因素及预测模型构建

Risk factors and prediction model of ketosis tendency in patients with newly diagnosed type 2 diabetes mellitus

目的探讨新诊断2型糖尿病(type 2 diabetes mellitus,T2DM)患者酮症起病的危险因素,并构建预测模型。方法回顾性分析2017年1月至2022年6月在温州市人民医院内分泌科住院的新诊断T2DM患者205例,根据是否酮症起病分为单纯起病组(125例)、酮症起病组(80例),比较两组患者的临床特征,采用Logistics回归分析酮症发生的危险因素,建立预测模型并采用受试者工作特征(receiver operator characteristic,ROC)曲线评价其对新诊断T2DM患者酮症起病的预测价值。结果与单纯起病组相比,酮症起病组患者男性多见(P=0.009)、年龄更小(P<0.001),收缩压(P=0.003)、高密度脂蛋白(high density lipoprotein,HDL)(P=0.001)、空腹及餐后2小时C肽水平(P<0.001)、促甲状腺激素(thyroid-stimulating hormone,TSH)(P=0.003)、血清游离三碘甲状腺原氨酸(free triiodothyronine,FT3)(P=0.012)更低;起病后下降体重(P<0.001)、糖化血红蛋白(glycosylated hemoglobin,HbA1c)(P<0.001)、稳态模型胰岛素抵抗指数(homeostasis model of assessment for insulin resistance index,HOMA-IR)(P<0.001)更高;多因素Logistics回归结果显示,年龄、HDL、空腹C肽水平、TSH是酮症发生的独立保护因素,HbA1c、HOMA-IR是独立危险因素;根据多因素Logistics回归分析结果建立新诊断T2DM患者酮症预测模型,该模型的曲线下面积(area under the curve,AUC)为0.880[95%CI(0.832,0.928),P<0.001],特异度80.8%、灵敏度86.3%,ROC曲线结果显示,酮症预测模型较各独立指标可更好地预测新诊断T2DM患者酮症的发生。结论年龄、HDL、空腹C肽水平、TSH是新诊断T2DM患者酮症发生的独立保护因素,HbA1c、HOMA-IR是独立危险因素,酮症预测模型对酮症倾向的发生有较好的预测作用。

Objective To explore the risk factors of ketosis onset in patients with newly diagnosed type 2 diabetes mellitus(T2DM),and to build a prediction model.Methods 205 newly diagnosed T2DM patients hospitalized in the department of endocrinology of Wenzhou People's Hospital from January 2017 to June 2022 were enrolled and divided into the simple onset group(n=125)and the ketosis onset group(n=80)according to ketosis onset or not.The clinical characteristics of the two groups of patients were compared.Logistics regression was used to analyze the risk factors of ketosis.A prediction model was established and the receiver operator characteristic(ROC)curve was used to evaluate its predictive value of the onset of ketosis in newly diagnosed T2DM patients.Results Compared with the simple onset group,patients in the ketosis onset group were more likely to be male(P=0.009),younger(P<0.001),lower systolic blood pressure(P=0.003),high density lipoprotein(HDL)(P=0.001),fasting and 2 h C peptide levels(P<0.001),thyroid-stimulation hormone(TSH)(P=0.003),free triiodothyronine(FT3)(P=0.012),higher loss of weight after the onset(P<0.001),higher glycosylated hemoglobin(HbA1c)(P<0.001),homeostasis model of assessment for insulin resistance index(HOMA-IR)(P<0.001).The multivariate Logistics regression results showed that age,HDL,fasting C peptide level,and TSH were independent protective factors for the development of ketosis,and HbA1c and HOMA-IR were independent risk factors.According to the results of multivariate Logistics regression analysis,the ketosis prediction model of newly diagnosed T2DM patients was established.The AUC of this model was 0.880(95%CI 0.832 to 0.928,P<0.001).The specificity was 80.8%and the sensitivity was 86.3%.The ROC curve results showed that the ketosis prediction model can better predict the occurrence of ketosis in newly diagnosed T2DM patients compared with the independent indicator.Conclusion Age,HDL,fasting C peptide level,and TSH are independent protective factors for the development of ketosis in newly diagnosed T2DM patients,HbA1c and HOMA-IR are independent risk factors,and the predictive model of ketosis has a good predictive effect on the occurrence of ketosis tendency.