两个检测系统测定同种生化项目结果偏倚的评估

EVALUATION OF TWO TESTING SYSTEMS IN DETERMINING BIAS OF SAME BIOCHEMICAL ITEMS

目的:通过对两个生化分析系统测定相同结果的对比,检测偏倚。并评估两个分析系统的偏倚是否在允许范围内,是否能被临床所接受。方法:按照美国全国临床实验标准委员会(NCCLS)EP 9-A2文件,以Olympus AU2700全自动生化分析仪及其配套原装试剂、标准品组成检测系统为比较方法(x)与O lympus AU2700全自动生化分析仪和Roche试剂,c.f.a.s标准品组成的检测系统为实验方法(Y),共同测定9项生化指标,记录检验结果,计算直线回归方程和相关系数。根据直线方程计算在某医学决定水平处的预期偏倚(^Bc)和预期偏倚的95%置信区间并判断偏倚是否可以接受。结果:将CliA’88规定的总容许误差(TEA)的1/2作为可接受误差(EA)的判断标准。每个项目在3个医学决定水平(Xc)的可接受误差(EA)均落在^Bc95%置信区间或大于^Bc95%置信区间上限,并<1/2 CliA’88 TEA,说明两个检测系统检测结果的偏倚,是在各医学决定水平处均可以接受的。两个检测系统基本能得到具有一致临床意义的检测结果。结论:两个检测系统对病人样本测定结果基本一致,以实验方法代替比较方法,不会对检验结果引入明显偏倚。

Objective：To determine bias through comparing same results coming from two biochemical system, and to evaluate whether the bias of above two testing system was in the permissible limits and whether it can be accepted in clinic. Methods： According to EP 9 - A2 document of National Committee Clinical Laboratory Standards （ NC CLS）, the experimental methods （x）, including Olympus AU2700 full- automatic biochemical analyzer and originated supporting agents and c. f. a. s standard preparation ,was used to determine 9 biaochemical items concomitantly, and then made record of testing results, calculating linear regression equation and correlation coefficient. According to linear equation ,predicted bias（ Bc ）and its 95% confidence interval at certain medical determination level were calculated , meanwhile the bias was judged whether it can be accepted. Results： Half of total errors accepted （TEA）prescribed by CliA＇ 88 was thought to be judging standard of errors accepted （EA）. Every items were all within Bc 95% or exceeding demonstrate that the bias of results from two its upper limit,and lower than 1/2 CliA＇ 88 TEA,which testing systems was acceptable at every medical determination levels. Both testing systems can get testing results with uniform clinical significance. Conclusion： The testing results from samples of the same patients determined by two kinds of testing systems were fundamentally uniform, and replacing comparision methods by experimental methods will not bring obvious bias to testing results.