根据CLSI EP09-A3文件对不同方法检测糖化血红蛋白A1c的可比性分析

Comparability analysis of glycosylated hemoglobin A1c detection by different methods based on CLSI EP09-A3 document

目的 探讨毛细管电泳法(CE)与离子交换高效液相色谱法(IE-HPLC)检测糖化血红蛋白A1c(HbA1c)的可比性及偏倚程度,为糖尿病患者、内分泌科医师及临床实验室提供准确可信的数据。方法 参照美国临床和实验室标准协会(CLSI)2013年8月的最新版本EP09-A3指南要求设计试验方法,对全自动糖化血红蛋白分析仪所用的IE-HPLC法与糖化血红蛋白电泳仪所用的CE法进行HbA1c检测比对试验。以IE-HPLC法为参考方法,CE法为候选方法,分别在上述2种HbA1c检测系统中检测65份血液标本,记录实验数据,根据EP09-A3指南,以IE-HPLC法所得结果为X轴,CE法所得结果为Y轴绘制散点图。以IE-HPLC所得结果为X轴,CE法与IE-HPLC法所得结果差值及其占IE-HPLC法结果的比例为Y轴,绘制差异偏差图和百分比差异偏差图。采用目测散点图、差异偏差图和广义极端学生化偏差法检查异常值,根据两种方法所得结果的差值变化趋势是恒定标准差(SD)或恒定变异系数(CV)选择最适合的回归模型并进行回归分析。根据回归方程计算HbA1c在两个医学决定水平0.065和0.098处的偏倚,以偏倚估计值在95%可信区间(95%CI)和偏倚小于±0.5作为临床可接受标准。结果 IE-HPLC法和CE法检测HbA1c所得结果分别为0.081(0.065,0.114)和0.078(0.064,0.113),差异无统计学意义(Z=0.222,P=0.824);目测法及广义极端学生化偏差法均未发现异常值,根据散点图、差异偏差图和百分比差异偏差图的特点,两种方法所得结果差值呈恒定SD变化,选择普通线性回归(OLR)模型进行回归分析,回归方程为Y=-0.060 92+0.993 1X,将0.065和0.098这两个HbA1c的医学决定值分别代入回归方程中,其偏倚估计值分别为-0.105 8和-0.128 5,均在偏倚估计值的95%CI范围内,也符合卫生行业标准规定的范围。结论 CE法与IE-HPLC法测定HbA1c所得结果具有可比性,偏倚也在临床可接受范围内,两种方法均能为临床提供稳定和可靠的HbA1c检测结果。

Objective To explore the comparability and bias of capillary electrophoresis(CE) and ion exchange high-performance liquid chromatography(IE-HPLC) in the detection of glycosylated hemoglobin A1c(HbA1c),and provide reliable and accurate data for diabetes patients, endocrinologists and clinical laboratories. Methods The method was designed according to the requirements of the latest version of EP09-A3 Guidelines for Method Comparison and Bias Assessment with Patient Samples: Approval Guide(third edition) issued by American Clinical and Laboratory Standards Institute(CLSI) in August 2013. The HbA1c comparison test was conducted for IE-HPLC method by automatic glycosylated hemoglobin analyzer and CE method by glycosylated hemoglobin electrophoresis instrument. The IE-HPLC method was used as reference method and CE method was used as candidate method. The 65 samples in above HbA1c detection systems were detected respectively, and the experimental data were recorded according to EP09-A3. The scatter plots were drawn with IE-HPLC results as X axis and CE results as Y axis. With the results of IE-HPLC as X axis, the difference between the results of CE method and IE-HPLC method and their percentage in the results of IE-HPLC method as Y axis, the difference deviation diagram and percentage difference deviation diagram were drawn. The visual scatter plot, difference deviation plot and generalized extreme student-centered deviation method were used to check the outliers. According to the change trend of the difference between the results of two methods was constant standard deviation(SD) or constant coefficient of variation(CV), and the most appropriate regression model was selected for regression analysis. The bias at 0.065 and 0.098 of the two medical decision levels of HbA1c was calculated according to the regression equation. The clinical acceptable standard was that the bias estimate was within its 95% confidence interval(95%CI) and the bias was less than ±0.5. Results The results of HbA1c detected by IE-HPLC and CE were 0.081(0.065, 0.114) and 0.078(0.064, 0.113) respectively, with no significant difference(Z = 0.222, P = 0.824). No abnormal value was found by visual measurement and generalized extreme student-centered deviation method. According to the characteristics of scatter plot, difference deviation plot and percentage difference deviation plot, the difference between the results of the two methods showed a constant SD change. Ordinary linear regression(OLR) model was selected for regression analysis. The regression equation was Y =-0.060 92 + 0.993 1X. The two medical decision values of HbA1c(0.065 and 0.098), were substituted into the regression equation, and the bias estimates were-0.105 8 and-0.128 5, respectively. They were within 95%CI of the bias estimate, and also conformed to the scope specified in the health industry standard. Conclusions The results of HbA1c determined by CE method and IE-HPLC method are comparable, and the bias is also within the clinically acceptable range. Both methods could provide stable and reliable HbA1c results for clinical use.