糖化白蛋白临床诊断切点的选择和评估

Selection and evaluation of clinical diagnostic cut point of glycated albumin

目的观察GA与HbA1c的相关性，评估GA诊断preDM和DM的切点，评价各糖代谢标志物的临床应用性能。方法病例对照研究。评估组选取2011年在中山医院同时检测FPG、HbA1c和GA的门诊患者，除外严重肾损伤和严重肝损伤。最终人组研究对象共20491例，其中男12407例，女8084例，年龄20～90岁，中位年龄64岁，观察GA与HbA1c的相关性并分析GA最佳诊断切点；对照组为DM疑似和初诊患者，共731例（男424例，女307例，年龄13～92岁，中位年龄57岁），检测OGTr、GA和HbA1c对照组再次分为NGT组101例（男54例，女47例，年龄14—82岁，中位年龄26岁）、preDM组159例（男87例，女72例，年龄13—86岁，中位年龄56岁）和DM组471例（男283例，女188例，年龄15—92岁，中位年龄59岁），通过ROC分析观察各糖代谢指标的敏感性、特异性、漏诊率、准确度、阳性预测值和阴性预测值。结果GA与HbA1c呈显著正相关（R2=0．737，P〈0．001）。在评估组人群中，GA=14．3％作为评估preDM的切点时敏感性和特异性分别为62．68％和60．14％，AUC为0．66；GA=16．3％为评估DM的切点时敏感性和特异性分别为81．54％和84．28％，AUC为0．9051。在对照组的preDM人群中，评估各糖代谢标志物的ROC曲线分析：OGTr2hPG的AUC最高（0．89），GA最低为0．65。OGTT2hPG、HbA1c和FPG的特异性均高达100％，OGTT2hPG的诊断敏感性最高（69．81％），GA取14．3％时的特异性为87．13％，敏感性为34．59％。在对照组的DM人群中评估各标志物的ROC曲线分析中：OGTr2hPG的AUC仍为最高（0．97），FPG最低（0．75），GA为0．87。OGTT2hPG、HbA1c和FPG的诊断特异性均高达100％，GA为90．38％，OGTT2hPG的诊断敏感性最高（91．93％），FPG的敏感性最低（43．10％）。结论在诊断DM时，OGTT 2hPG的诊断性能最佳，其次为HbA1c而作为诊断指标单独使用GA并无更好的临床意义（AUC显著小于OGTr2hPG和HbA。P〈0．01）。但作为OGTT2hPG和HbA1c的补充，当OGTI＇2hPG和HbA1c检测结果可能受影响时，与FPG相比较，GA仍具有较高的临床价值。本文得出的诊断切点相比厂商声明的参考区间上限能降低漏诊率。

Objective To investigate the relationship between GA and HbAI1c, to identify the optimal GA cut-off values used to diagnose preDM and DM, and to verify the diagnostic significance of blood glucose biomarkers. Methods A total of 20 491 samples from Chinese people were involved in the correlation analysis. The participants were expected to complete the test of FPG, HbA1c and GA. The final dataset included 12 407 men and 8084 women aged from 20 to 90 （median 64）. ROC curves were used to identify the optimal GA cut-off values to diagnose PreDM and DM. In order to verify the diagnostic significance of GA, 731 Patients with no history of DM were involved as the control group. The participants were grouped by the Standards of Medical Care in Diabetes -2011 from American Diabetes Association. Results GA had a strong positive correlation with HbA1c （R2 = 0. 735, P 〈 0. 001 ）. Among participants forthe correlation analysis, the sensitivity and specificity were 62. 68%, 60. 14% respectively for detecting preDM when GA threshold was set to 14. 3% （AUC = 0. 66）. The sensitivity and specificity were 81.54%, 84. 28% respectively when GA threshold was set to 16. 3% for detecting DM （AUC = 0. 9051 ）. When analyzing the ROC of detecting preDM in the control group, we found that 2hPG had the highest AUC （0. 89） while GA had the lowest AUC （0. 65）. What＇s more, the specificities of 2hPG from OGTT, HbAlo, or FPG were all 100% at their respective cut-off points. The sensitivity of 2hPG from OGTT was the highest （69. 81%）. The specificity and sensitivity of GA （ 14. 3%） were 87. 13% and 34. 59% respectively. When analyzing the ROC of detecting DM, we found that 2hPG also had the highest AUC （0. 97） , and then HbAlc（0. 92） , GA （0. 87）, and FPG （0. 75）. 2hPG from OGTT, HbA1c and FPG all had the specificity of 100%, while GA was 90. 38%. 2hPG from OGTI＂ had the highest sensitivity （91.93%）, while the lowest is FPG （43.10%）. Conclusions When diagnosing DM, 2hPG from OGTT has the best performance and then HbAIo. It is no good to set GA as a single diagnostic biomarker to detect PreDM or DM. However, as the supplement of 2hPG and HbAxc, GA is more efficacious to diagnose DM than FPG when the results of 2hPG and HbAlc are disturbed. GA cut-off value defined at 14. 3% can decrease the misdiagnosed rate of PreDM than the upper limit of reagent instruction.