对13种葡萄糖常规检验系统正确度的评价

Trueness evaluation of 13 routine measurement systems for glucose

目的评价13种葡萄糖常规检验系统的正确度；并探讨标准物质对改进常规检验系统结果-致性和正确度的作用。方法本研究为试剂评价研究。以同位素稀释液相色谱串联质谱（ID—LC／MS／MS）法为比对方法，选取瑞士罗氏、北京中生、深圳迈瑞、上海科华、四川迈克、上海丰汇、日本和光7个生产厂家的13种血糖试剂盒（葡萄糖氧化酶法或己糖激酶法）及其配套校准品在日立7170A生化分析仪上组成的常规检验系统作为待评价方法，分别测定从2011年1月1日至6月30日卫生部北京医院糖尿病门诊患者实验室检测剩余血清收集的20份新鲜冰冻血清和美国国家标准和技术研究院（NationalInstituteofStandardsandTechnology，NIST）葡萄糖标准物质SRM965a中的2、4水平中的葡萄糖浓度；将比对方法和每-待评价测定20份冰冻人血清葡萄糖的结果进行直线回归分析，拟合的直线回归方程用于估算医学决定水平上预期偏倚（Bias）；利用两个水平的SRM965a认定值与待评价方法测定值之间的数学关系，对各个待评价方法测定结果进行再校准（Recalibration），比较再校准前后13种常规方法检测20份人血清葡萄糖的方法间CV、平均偏倚及预期偏倚的变化。结果各待评价方法具有较好的精密度（CV〈1．63％）；13种待评价方法与比对方法测定20份新鲜冰冻人血清葡萄糖的结果间具有良好的相关性（R^2均大于0．998，P均〈0．01），13个评价方法测定20份患者血清葡萄糖的偏倚范围为-0．11～0．27mmoL／L（-1．4％-3．92％），在5．6、7．0及11．1mmol／L3个医学决定水平的预期偏倚范围分别为-0．08—0．23mmol／L（-1．43％-4．11％），-0．10～0．26mmol／L（-1．43％-3．71％）和-0．18—0．37mmol／L（-1．62％～3．33％）。13种待评价方法测定20份新鲜冰冻血清样本葡萄糖时，方法间CV均值2．11％（1．55％～2．39％），利用标准物质再校准后，各样本的方法间CV均有所减小，均值1．34％（0．95％～1．82％）；13种待评价方法测定20份新鲜冰冻血清样本相对偏倚平均值1．11％（1．72％-4．85％），再校准后其绝对值有所增加，平均值为-2．52％（-5．12％～1．32％）。结论13种葡萄糖常规检验系统均具有良好的正确度，其溯源性得到验证。利用标准物质再校准有助于提高不同常规方法检验结果的一致性。

Objective To evaluate the trueness of 13 routine measurement systems tor glucose anti the role of standard reference materials in improving harmonization and trueness among routine measurement systems. Methods The research is related to the reagent evaluation. The isotope dilution liquid chromatography tandem mass spectrometry （ID-LC/MS/MS） method was applied as the comparative method. 13 routine measurement systems, composed of 13 kits and calibrator for glucose measurement （ GOD method and/or HK method） from Roche, Biosino, Mindray, Kehua, Maker, Fenghui, Wako companies and Hitachi 7170A automatic analyzer, were chosen as the test methods. The comparative method and test methods were applied to 20 fresh frozen sera collected from outpatient of diabetes of Beijing hospital in 2011and two levels of NIST SRM 965a for glucose concentration. Linear regression analysis was made for the results of comparative methods and every test method to estimate the predicted bias at given medical decision levels. The mathematics relationship between the assigned values of SRM965a and results from each test method was used to recalibrate the results of 20 fresh frozen sera from 13 test methods. CV, average bias and predicted bias were compared before and after recalibration. Results All test methods demonstrated good precision （ CV 〈 1.63% ） and good correlation（ all R2 〉 0. 998, all P 〈 0. 01 ） with comparative method when measuring 20 fresh frozen sera in glucose resuhs. The bias ranged from -0. 1 lto O. 27 mmol/L when 20 fresh frozen sera measured by 13 test methods. The predicted bias ranged from - 0.08 to O. 23 （ - 1.43% to 4. 11%）, -0.10 to0.26 mmol/L（ -1.43% to 3.71%） and -0.18 to0.37 mmol/L（ -1.62% to 3.33% ） at the given medical decision levels of 5.6, 7.0 and 11.1 mmol/L respectively. After recalibration, the average of between-method CVs measured by 13 test methods decreased from 2. 11% （1.55% to 2. 39% ） to 1.34% （0. 95% to 1.82% ）. The average of relative bias increased from 1.11% （ 1.72% to 4. 85% ） to - 2. 52% （ - 5.12% to 1.32% ）. Conclusions All test methods demonstrated good trueness and their calibration traceability was verified. Recalibration of routine systems by Standard Reference Materials contributes to the harmonization among methods.