CLSI EP15-A3在临床生化正确度验证中的应用  

Application of CLSI EP15-A3 in estimation for trueness of clinical biochemistry

作　　者：徐建华[1] 刘冬冬[1] 徐宁[1] 马监林 詹科海 陆丰源 林莉[1] 王建兵[1] 庄俊华[1] 黄宪章[1] XU Jianhua;LIU Dongdong;XU Ning;MA Jianlin;ZHAN Kehai;LU Fengyuan;LIN Li;WANG Jianbing;ZHUANG Junhua;HUANG Xianzhang(Department of Laboratory Medicine,Guangdong Provincial Hospital of Chinese Medicine,Guangzhou 510105,G uangdong;School of Laboratory Science,Guangdong Medical University,Dongguan 523808,Guangdong;Department of Labo-ratory Science,Bengbu Medical College,Bengbu 233000,Anhui;Department of Laboratory Medicine,School of Public Health,Xia-men University,Xiamen 361102,Fujian,China)

机构地区：[1]广东省中医院检验医学部,广州510105 [2]广东医科大学检验学院,广东东莞523808 [3]蚌埠医学院检验系,安徽蚌埠233000 [4]厦门大学公共卫生学院实验医学系,福建厦门361102

出　　处：《临床检验杂志》2018年第10期779-782,共4页Chinese Journal of Clinical Laboratory Science

基　　金：国家自然科学基金(81572088);"十二五"国家科技支撑计划(2013BAI02B04;2012BAI37B01);广东省中医院拔尖人才项目(2014-47);广东省中医院综合标准化示范项目(YN2015BZ10)

摘　　要：目的探讨美国临床和实验室标准化协会(CLSI)EP15-A3文件在正确度验证中的应用价值。方法按照EP15-A3文件5×5实验设计方案,用IFCC提供的Rela A(水平1)和Rela B(水平2)样品、美国国家标准与技术研究院(NIST)909C参考物质分别验证肌酐(Cr)、尿素(Urea)、尿酸(UA)、总胆固醇(TC)、三酰甘油(TG)、葡萄糖(Glu)的正确度。每个样品每天1批,每批重复5次,共5 d,每个样品获得25个数据。用Grubbs'法计算离群值,单因素方差分析(ANOVA)计算用户的批内不精密度(S_R)和实验室内不精密度(S_(WL)),基于S_R和S_(WL)计算结果的均值及其标准差,计算靶值(TV)及其标准误差,最后计算出TV的确认区间(VI),观察各指标检测均值是否在VI内,如通过则用户证明了候选方法的偏差可接受;若不通过则计算均值和TV的相对偏差,观察该相对偏差是否小于用户定义的可接受范围(<1/2 TEa),若是则证明候选方法的偏差可接受。否则,需查找原因或与厂家联系。结果6个生化项目的测定结果均通过Grubbs'法离群值检查,各项目无离群值。该次验证实验中,除NIST 909C参考物质样品Urea、TC,Rela样品Cr水平2均值不在VI内,但相对偏差均小于1/2 TEa外,其他均值都在VI内。结论用EP15-A3文件进行验证的6个生化项目的正确度均符合临床要求。Objective To evaluate the application of Clinical and Laboratory Standards Institute（CLSI）guideline EP15-A3 document for estimation of trueness of testing results in clinical biochemistry.Methods According to the EP15-A3 basic 5×5 precision design of experimental procedure,the bias of results of 6 common biochemistry items including creatinine（Cr）,urea,uric acid（UA）,total cholesterol（TC）,triglyceride（TG）and glucose（Glu）were verified by using IFCC Rela A（level 1）and Rela B（level 2）samples and NIST 909c reference material.The basic 5×5 design for five days and one run for each sample per day were performed,and five replicates per run should yield a total of 25 results for each sample.The outliers were calculated by Grubbs＇test.The user＇s repeatability variance（SR）and the within-laboratory variance（SWL）were calculated by one way ANOVA analysis.Both the overall mean and its standard error,and the target value（TV）and its standard error were calculated according to SRand SWL,and the verification interval（VI）for the TV was calculated at last.The mean values of each indicator were observed whether they were in VI.If the mean value was within the VI,it means that the bias of the candidate method should be acceptable,otherwise the bias between the observed mean and the TV had to be recalculated to observe if the bias was less than the user-defined allowable bias（〈1/2 TEa）.If so,it was proved that the bias of the candidate method should be acceptable.Otherwise,the reasons of bias should be found or contacting the manufacturer should be needed.Results The results of the 6 biochemistry items were all passed through Grubbs＇outliers test,and no outliers were found in each item.In this estimation for bias by two reference materials,each mean of levels was within the VI expecting the reference material of urea and TC from NIST 909c and the sample of Cr from Rela（level 2）,but the biases were all less than 1/2 TEa.Conclusion The trueness of the 6 Roche biochemistry i

关 键 词：正确度 精密度 性能验证 EP15-A3 

分 类 号：R446[医药卫生—诊断学]