The Bias Estimation And Comparative Analysis On The Detecting Results Of Serum Enzyme From Different Detecting Systems

Objective: To investigate comparability on the results of serumenzyme among different detecting systems through methods comparisonanalysis and bias estimation. Meanwhile, to calibrate detecting systemwithout comparability, in order to provide more evidence for traceabilityand standardization of serum enzyme measurement and medical laboratoryaccreditation. Method: (1) Based on EP9-A file of NCCLS, comparative system Xis the system which have accredited by ISO/IEC 17025 includes HITACHI7170 biochemical analyzer, reagent of ROCHE, quality controller ofROCHE and calibrator of ROCHE, the other systems without accreditationare experimental systems Y. Six serum enzymes including ALT,AST,ALP,GGT,LDH and CK are detected in both systems X and Y. Regardinghalf of the error range allowed by CLIA'88 as standard, comparability ofdifferent detecting systems are judged by relative bias (%SE) between Xand Y. (2) Regarding freeze-dried RANDOX controller as transfer, endowfreeze-dried controller with a certain calibrating value after methodcomparison in patients' fresh serum. So the controller become the steadycalibrator of self-established detecting system, then we can calibrateself-established detecting systems with this calibrator. Result: (1)Because daily CVs of different detecting systems are all less than third ofCLIA'88, precision of comparison tests sizes up and data are reliable. (2)According to the results of paired t-test, except a few results ofexperimental systems, other results of experimental systems havesignificant difference from comparative system X (P<0.05), but the relativebias of mean is less than half of the error range allowed by CLIA'88,therefore, all results are clinical acceptability. (3) Except results of ALT ofY4, high level ALP of Y1 (less than 150U/L), AST of Y3 and Y4 and highlevel GGT of Y1 and Y4, other results of experimental systems havecomparability with system X within comparative range regardingsystematic error of method comparison less than half of the error rangeallowed by CLIA'88 as judging standard of clinical acceptability accordingto statistic analysis of correlation and regress. (4) Regarding freeze-driedRANDOX controller as steady calibrator for self-established detectingsystems, bias between self-established systems and comparative system areless than quarter of CLIA'88. Conclusion:(1) When there are more than two detecting systems inone laboratory for the same test, method comparison and bias estimate areneeded for judging it's clinical acceptability, meanwhile, calibrating testand system without comparability is the guarantee of comparability ofdetecting results. (2) Method comparison must base on excellent qualitycontrol, so the daily CVs ought to be less than third of the error rangeallowed by CLIA'88 external quality assessment. (3) Both T Test andregress can be used to evaluate bias of experimental data. Regress can doquantitative estimate for bias of experimental methods within comparativerange, while T Test can only explain the bias of two methods on the level ofthe mean of experimental results. (4) The half of the error range allowed byCLIA'88 has more maneuverability and feasibility as a judging standard ofclinical acceptability, which also provide a objective guideline for internalquality management and ability validation in medical laboratoryaccreditation. (5) Regarding freeze-dried RANDOX controller as steadycalibrator for self-established detecting systems, quarter of CLIA'88 asstandard of success, traceability and long-time comparability ofself-established system can be realized.