1型糖尿病临床诊断模型的建立  

Establishment of clinical diagnosis model of type 1 diabetes mellitus

作　　者：陈双[1] 付麒[1] 顾愹[1] 刘今[1] 李欣[1] 肖蕾[1] 朱婧[1] 施云[1] 陈恒[1] 孙敏[1] 顾威[2] 张梅[1] 徐湘婷[1] 王坚[3] 陈峰[4] 杨涛[1] 

机构地区：[1]南京医科大学第一附属医院内分泌科,210029 [2]南京医科大学附属儿童医院内分泌科 [3]南京军区南京总医院内分泌科 [4]南京医科大学公共卫生学院流行病与卫生统计学系

出　　处：《中华糖尿病杂志》2016年第10期598-602,共5页CHINESE JOURNAL OF DIABETES MELLITUS

基　　金：国家自然科学基金（81270897、81300668、81370939、81370922、81400813、81530026）;江苏省临床医学科技专项（BL2012026）

摘　　要：目的探讨建立1型糖尿病临床诊断模型作为量化工具，在不具备抗体精确检测条件或在抗体检测结果明确前对初诊糖尿病患者进行初步分型的诊断价值。方法选取2012年6月至2016年4月于南京医科大学第一附属医院内分泌科住院治疗的532例初诊糖尿病患者（病程1年内）为研究对象，其中1型糖尿病203例，2型糖尿病329例，按照2：1的比例随机分为建模组355例，验证组177例。采用Logistic回归分析法将常用于鉴别1型糖尿病和2型糖尿病的相关变量纳入方程建立模型。采用受试者工作特征（ROC）曲线评价模型的诊断效力。结果依据糖尿病患者初诊时年龄、体质指数（BMI）和C肽值[分别为基础（C0）、餐后120min（C120）和c肽曲线下面积（Cauc）腱立1型糖尿病临床诊断模型（CODE）：CODE-C0、CODE-C120和CODE-Cauc。在建模组人群中，三模型ROC曲线下面积分别为0．832（95％CI：0．786-0．878）、0．840（0．793-0．887）、0．846（0．799～0．892）；在验证组人群中，三模型ROC曲线下面积分别为0．825（0．753～0．897）、0．848（0．782～0．914）、0．851（0．785～0．917）。采用双切点方法，模型CODE-C0以10．80和7．30诊断1型糖尿病和2型糖尿病的特异度均可达95％，模型CODE-C120以10．9l和7．65为切点、模型CODE-Cauc以10．85和7．87为切点，诊断1型糖尿病和2型糖尿病的特异度亦可达95％。结论依据糖尿病患者初诊时的年龄、BMI和C肽值建立的1型糖尿病临床诊断模型工具，可以在暂无高敏感性胰岛自身抗体检测结果时，较好地鉴别1型糖尿病和2型糖尿病。Objective To develop a satisfactory clinical diagnosis model of type 1 diabetes mellitus （T1DM） for diagnosis and etiological classification of uncertain cases, achieve quantitative criteria for initial judgment without the standard autoantibody test results and identify specific patients needing accurate islet autoantibody assay. Methods 532 newly-diagnosed diabetic patients （diagnosed less than 1 year） were retrospectively studied, including 203 patients were diagnosed as T1DM, 329 patients were diagnosed as type 2 diabetes mellitus （T2DM）, who were hospitalized in Department of Endocrinology and Metabolism, the First Affiliated Hospital of Nanjing Medical University between Jun 2012 to Apt 2016. Eligible patients underwent randomization in a 2：1 ratio to establish（n=355） or validate （n=177） diagnosis models. Logistic regression with common variable to identify T1DM and T2DM was applied to establish clinical diagnosis model of type 1 diabetes（CODE）. The receiver operator characteristic （ROC） curves were analyzed to test the model performance. Results Three models（CODE-C0, CODE-C12o, CODE-Cauo） containing age at diagnosis time, body mass index （BMI）, C-peptide （fasting C-peptide, Co; 120 min postprandial C-peptide, C120; area of the curve of C-peptide, Cauo, respectively） were created. In the establish cohort, the area of the ROC curve（AUC） of these three models were 0.832（95% CI 0.786-0.878）, 0.840 （0.793-0.887） and 0.846（0.799-0.892）, respectively. In the validation cohort, AUC of these three models were 0.825（0.753-0.897）, 0.848（0.782-0.914） and 0.851（0.785-0.917）, respectively. With two cut off points, 10.80 and 7.30, the specificity of CODE-Co diagnosing T1DM and T2DM was 95%.In the CODE-C^2o and CODE-C,,c models, both CODE-C120〉10.91 and CODE-C,uo〉10.85 were indicative of T1DM （95% specificity）. A low CODE-C120 （〈7.65） and low CODE-Cauo （〈7.87）were indicative of T2DM （95% specificity）. Patients with seores between

关 键 词：糖尿病 1型 糖尿病 2型 模型 诊断 

分 类 号：R587.1[医药卫生—内分泌]