Comparability Of Measurement Procedures And Commutability Of Reference Materials For Serum Protein Measurements

ObjectiveTo investigate thecomparabilityof measurement results for serum proteins and to provide the scientific evidence for effective use of the measurement results and for standardization of serum protein measurements; To evaluate the commutability of possiblereference materials and to select candidatereference materials for EQA scheme and standardization programs.MethodSerum panels, which consisted of50or70individual patient specimens, were collected at the clinical laboratoriesof two Beijing Hospitals. Possible reference materials were randomly interspersed among thepatient specimens, and all of which were aliquoted into5-9vials andthen unified labeled. Each specimen was analyzed in duplicate or triplicate based on the measurand. The possible reference materials included:(1)Fresh frozen serum pools(HSP), pooled byleftover sera of patient samples.(2) International standard materials, prepared following manufacturer’s instructions.(3)EQA materials, prepared following our EQA instructions.(4) Purified materials, purchased from YangtzeDeltaRegionlnstituteofTsinghuaUniversity.(5)Animal sera, collected from an abattoir or model animals.(6) Pleural effusion, collected at department of clinical laboratory of xx hospital. A total of22measurands(ie, IgA, IgG, IgM, C3,1C4, PA, β2M, Cystatin C, RF, ASO, TRF, CRP/HsCRP, cTnI, CK-MB, MYO, TSH, PSA, AFP, CEA, CA125, CA153and CA199) and27analytical reagents (ie, imported A1-A4, B1-B3, Cl, C2, D1-D3, E, F, G, H, I and J, and domestic reagents A1, A2,B,C,D,E,F,G andH) were evaluated.For eachmeasurand, Correlations and comparabilities between/among methods were analyzed after outliers and incomplete data were excluded. Correlations were evaluated based on scatter plots and the differences between the ratio [R,2times of residual standard deviation(Sy.x) to median of the serum panel]and minimum allowable total error (mTEa) derived from biological variations. Measurands were divided into3groups and correlations were analyzed accordingly. If R<mTEa formost the methods, correlations were regarded to be "good"and methods withR> mTEa were deleted from correlation analysis. If R was slightly higher than mTEa, correlations were designated to be "average" in which methods with extremely high R values were excluded. Correlations were"poor" if most of the methods had R values significantly higher than mTEa, and in this case, only methods with relatively low R were analyzed. Furthermore, comparability analysis was done for all the above methods thatwere included in the correlation analysis.Method comparabilities were evaluated based on the difference between2times of CV derived fromthe mean of assays and mTEa. If2CV was higher than mTEa, the results among assays were regarded to be consistent. In addition, commutabilities of possible reference materials were evaluated according to C53-A protocol. Deming regression was applied directly for measurands with narrow measurement ranges (such as, IgA, IgG, C3, C4, PA and TRF) and performed after the data were logarithmic transformed for those with wide ranges (IgM, β2M, Cystatin C, RF, ASO, CRP/HsCRP, cTnI, CK-MB, MYO, TSH, PSA, AFP, CEA, CA125,CA153and CA199).Results1. Correlation and comparabilityBased on the correlation results and scatter plots, the measurands were divided into3groups:(1) Measurands with good correlations(R<mTEa) among most of the methods, includingIgA, IgG, IgM, C3, C4, PA, β2M, Cystatin C, TRF, CRP/HsCRP, CK-MBand MYO, in which IgA, IgM, PA and CRP/HsCRP had consistent results among different methods. The rest7measurands showed systematic biases in one or more methods and the results were inconsistent.(2)Measurands with R> mTEa for most of the methods, includingASO, cTnI, CK-MB, TSH, PSA, AFP, CEA, CA125and CA153, in which CK-MB and PSA, but not the rest of the7measurands, had consistent results among methods.(3) Measurands with poor correlations among most of the methods (R> mTEa),including RF and CA199. However, for RF, correlations were desirable among imported B2, D3and domestic Dalthough results were inconsistent. For measurands CA199, correlations and comparabilities were desirable amongimported B4, D2, E and domestic Alssays.2. Commutability of potential reference materials Human fresh frozen serum pools(HSP) were commutablefor all pair comparisons for22evaluated measurands. International reference materials,ERM470,471and474, were commutableamong all pair comparisonsfor IgA, IgG, IgM, C4, PA, Cystatin C, TRF, andCRP, but not for C3. However, SRM2921only showed commutability for certain combinations of methods.NIBSC81/565wascommutable when diluted with saline and bovine serum albumin,but lack of commutability when diluted with human or swine serum. PSA reference materialNIBSC96/670, diluted with saline and bovine serum albumin, did not showcommutability for some combinations. EQA materials were commutableinall pair comparisonsforlgM, PA, B2M, TRF, CRP, MYO, TSH, AFP, CEA and CA153assays. Purified materials were commutable for PA, β2M and CA125but not for PSA. Swine sera collected from an abattoir or model swineswith acute myocardial infarction (AMI)were incommutable for IgG, IgM, cTnI, CK-MB and MYO. Moreover, swine serashowed no reactivity to IgA, C3, C4and PA assay kits. Rat sera collected form AMI rats showed markedly decreasedcommutability for cTnI and MYO among certain combinations, and it showed no reactivity to all CK-MB kits. Highconcentration human sera diluted with swine sera were commutable only for β2M and RF, but not forPA, Cystatin C, CRP/HsCRP, TSH, PSA andCA125.Pleural effusion diluted with human or swine sera showed commutability for CA125.ConclusionsThe comparabilities of protein results were variable depending on different measurands. Some proteins hadconsistent results among most of the methods, some showed significant calibration biases, while some others had analytical methods that not so specific.Different reference materials showed different commutabilities. HSP was commutable for all the evaluated measurands and methods. Other prepared materials and animal sera were lack of commutability in varying degrees. Therefore, according to different situations, selections of proper reference materials, and establishmentof effective quality monitoring and improving mechanisms, will be important tasks in the standardization of protein measurements in the future.