ROC曲线分析糖化血红蛋白作为糖尿病诊断标准的应用

摘要

目的 评价糖尿病高风险人群中糖化血红蛋白(HbA1c)诊断糖尿病的价值,并探讨糖基化生物学变异对HbA1c作为糖尿病诊断标准的影响.方法 对2010年12月～2011年10月筛选出的205例糖尿病高风险人群行葡萄糖耐量试验(OGTT),并测定HbA1c、高密度脂蛋白胆固醇(HDL)、三酰甘油(TG),依据WHO(1999年)糖尿病诊断标准分为糖耐量正常组(NGT),血糖调节受损组(IGR)和糖尿病组(DM);利用ROC曲线分析HbA1c诊断糖尿病和糖尿病前期的最佳切点,并评价诊断效果.对依据该研究HbA1c最佳诊断切点漏诊和误诊糖尿病病例行毛细血管血糖(CBG)监测,获得平均血糖水平,分析糖基化个体差异对HbA1c作为诊断标准的影响.结果 糖尿病高风险人群检出率为35.6%;HbA1c诊断糖尿病的ROC曲线下面积(AUC)为0.891(95%CI:0.856～0.926),与空腹血糖相比,差异无统计学意义(Z=1.98,P>0.05),在最佳诊断切点6.05%,敏感度、特异度分别为78%和87.9%;HbA1c诊断糖尿病前期的AUC为0.567(95%CI:0.492～0.641),低于以空腹血糖单独诊断糖尿病前期的AUC(0.772,95%CI:0.712～0.831)(Z=3.56,P<0.05);Bland-Altman分析表明,依据该研究HbA1c最佳诊断切点漏诊患者的HbA1c实测值和理论值存在临床不可接受的偏差,糖基化的个体差异不可忽视.结论 HbA1c诊断糖尿病的最佳诊断切点为6.05%,诊断效果与空腹血糖相当,但诊断糖尿病前期的能力弱;个体间糖基化差异可能是影响诊断效果的重要原因.%Objective To evaluate the value of glycated hemoglobin (HbA1c) for diagnosis of diabetes mellitus (DM) and the influence of glycated biological variation in diagnosis of DM using HbA1c.Methods All enrolled subjects underwent oral glucose tolerance test (OGTT) and whole blood HbA1c,serum high density lipoprotein (HDL) and triglycerides (TG) were detected.The optimal cut-points of HbA1c in diagnosis of DM and pre-diabetes were acquired using a receiver operating characteristic curve (ROC curve).In missed-diagnosed and misdiagnosed subjects the level of capillary whole blood glucose (CBG) was measured 13 times one day.The influence of glycated biological variation in diagnosis of DM using HbA1c was evaluated by analyzing the relationship between HbA1c and mean blood glucose (MBG).Results The ROC curve showed that the area under the curve(AUC) of HbA1c for diagnosing DM was 0.891 (95% confidence interval:0.856 to 0.926) , which was similar to that of fasting plasma glucose (FPG) (Z=1.98 , P>0.05) , the optimal cut-point of HbA1c for diagnosing DM was 6.05% , with sensitivity of 78% ,specificity of 87.9%.The optimal cut-point of HbA1c for diagnosing pre-diabetes was 5.65% and AUC was 0.567 (95% confidence interval:0.492 to 0.641) , which was significantly lower than that of FPG (AUC 0.772,95% CI:0.712 to 0.831, Z=3.56, P<0.05).Bland-Altman analysis indicated that in missed-diagnosed subjects diagnosed by the optimal cut-point of HbA1c 6.05% there was significant bias between measured HbA1c and theoretical value based on MBG.Conclusion The capability of HbA1c for diagnosing DM was similar to that of FPG.And based on the ROC analysis,6.05% was the optimal cut-point of HbA1c for diagnosing DM.But the value of HbA1c for diagnosing pre-diabetes was low.The glycated biological variation would influence the value of HbA1c in diagnosing DM.