Characterization of the New Serum Protein Reference Material Erm-da470k/ifcc: Value Assignment by Immunoassay

BACKGROUND: The availability of a suitable matrix reference material is essential for standardization of the immunoassays used to measure serum proteins. The earlier serum protein reference material ERM-DA470 (previously called CRM470), certified in 1993, has led to a high degree of harmonization of the measurement results. A new serum protein material has now been prepared and its suitability in term of homogeneity and stability has been verified; after characterization, the material has been certified as ERM-DA470k/IFCC.

Serum protein measurements are among the beststandardized protein measurements in clinical chemistry.In the homogeneous immunoassays most often used for serum protein measurements, the signal depends on factors such as antibody specificity, reaction kinetics and equilibria, multimeric state of the proteins, and matrix effects.The quantification of the proteins depends primarily on the comparison of the measurement results with those obtained with a calibrant (1 ).The EU Directive on In Vitro Diagnostic Medical Devices (directive 98/79/EC) demands the traceability of calibrants and control materials to reference measurement procedures and/or reference materials of higher order.
Manufacturers referenced their calibrators and control materials to this CRM, and the betweenlaboratory variances for assays of most serum proteins became substantially lower (3)(4)(5).In external quality assurance schemes, the variability between measure-ment results obtained with all available methods for proteins such as TTR, AAT, AAG, TRF, IgG, IgA, IgM, and C3 ranges from about 5% to 8% (3 ).
ERM-DA470k/IFCC is a new serum protein reference material prepared to replace ERM-DA470.It was spiked with CRP and ␤ 2 -microglobulin (B2M), the lat- ter being an additional protein intended to be certified.The preparation of the serum material and verification of its homogeneity and stability have been reported (6 ).
This article describes the characterization of ERM-DA470k/IFCC by use of protocols that can be considered as reference measurement procedures (7 ).We certified mass concentrations for 12 proteins (A2M, AAG, AAT, ALB, C3c, C4, HPT, IgA, IgG, IgM, TRF, and TTR).The assignment of certified values for CER and B2M requires further investigation and is ongoing.

GENERAL PRINCIPLES
The procedure for assignment of values to serum proteins in a target material using a reference preparation was developed in the late 1980s for the transfer of values to 14 serum proteins in CRM 470 (now ERM-DA470), at that time the international reference preparation for human serum proteins (3,7,8 ).The practical procedure (the transfer protocol) is based on a multiple-point value assignment obtained by several measurements a day repeated on several days, an important prerequisite being that all reconstitutions and dilutions are controlled by weighing.The more theoretical description of the procedure and the necessary mathematical equations are discussed in Blirup-Jensen et al. (7 ), whereas a practical protocol with examples is given in Blirup-Jensen et al. (9 ).
We used methods that are validated and wellestablished routine methods based on turbidimetry, nephelometry, and occasionally visible spectrometry (in the case of ALB).The concentrations of calibrant and reference material dilutions were optimized for each platform/reagent combination.
The laboratories used either of 2 approaches, called open-value transfer procedure (calibrated directly with ERM-DA470) and closed-value transfer procedure (using manufacturers' calibrants).

TRIAL RUN
The laboratories were provided with a detailed protocol and reporting sheets, 2 vials of ERM-DA470, and 2 vials of a pilot batch spiked with CRP at approximately 40 mg/L, called pilot CRP (6 ).On each of the 2 measurement days, 1 vial of each material was reconstituted with water, and 6 dilutions were prepared and measured for the 2 materials (closed-value transfer procedure).Further details can be found in the supplementary materials (which accompany the online version of this article at www.clinchem.org/content/vol56/issue12).

CHARACTERIZATION MEASUREMENTS
Participating laboratories were provided with vials of ERM-DA470 and ERM-DA470k/IFCC, detailed protocols, and reporting sheets.The lyophilized materials were reconstituted the day before the measurements.For each of the 4 measurement days, new vials of ERM-DA470 and ERM-DA470k/IFCC were reconstituted, and new sets of 6 dilutions were prepared.
We optimized the open transfer procedure separately for each platform participating in the value assignment by using this procedure, and for each protein measured with that platform.The main parameter optimized was the dilution scheme.We took into account that dilutions done by the platforms should be avoided as much as possible (as these dilutions cannot be corrected by more accurate weighing).We grouped proteins for which similar dilution schemes could be used.Finally it was also necessary to take into account that for each material all 6 dilutions should be prepared from a single vial.For the open value transfer procedure, 3 runs were performed on each of 4 measurement days.Each run was done with new calibrations using 6 dilutions of ERM-DA470.The 6 dilutions of ERM-DA470k/IFCC and the control material (which consisted of a separate dilution of ERM-DA470) were measured as samples.Each target material dilution and the control material were measured in duplicate at each run.Laboratory number 3 used a different procedure, which had been validated previously.According to their procedure, all the dilutions are measured in triplicate, over 3 days only, resulting in a slightly higher number of measurements.

DATA ANALYSIS
The aim of the value transfer was to determine the transfer factor for each protein.A detailed description of the principles and procedures can be found in Blirup-Jensen et al. (7 ) and Zegers et al. (10 ).In all reconstitution and dilution steps, the measured masses were used for the calculations of concentrations, and not the intended volumes.
For the open-value transfer procedure, we plotted the measurement results S ijk (of the kth measurement of material i within dilution j, expressed in %) in scatter plots S ijk ϭ f(c ij ).A linear regression was performed by using the means of the S ijk as a function of the concentration c ij .The slope of the regression line is the TF.
For the closed-value transfer procedure, we plotted single measurement results S ijk in scatter plots for both ERM-DA470 and ERM-DA470k/IFCC.The ratio of the slopes of the linear regression lines for ERM-DA470k/IFCC and ERM-DA470 is the TF.
We analyzed the data for outliers (studentized residuals), homogeneity of variance, normality, and linearity and verified that the intercept Ϯ 4 times the SD includes the origin.In case of homoscedasticity, we identified outliers of the regression model according to the method of Lund (11 ).
The following general acceptance criteria were applied: the 95% CIs of the intercepts of the regressions of the calibrant and the target materials must be overlapping (closed data sets only); R 2 of the regression must be Ͼ0.97 for Immage data and Ͼ0.98 for all other platforms; data from at least 4 dilutions must be available; for each laboratory, the completeness of data, fulfilling the described quality criteria, must be at least 50% for the data for a particular protein on a particular day, and at least 2 daily value assignments must be valid.The day-today variation (relative SD, or CV) of valid data sets for a particular protein and laboratory must be Ͻ5% (8% for A2M).For the open procedure, the control values must be between 95% and 105% of the expected values.
We determined the certified concentrations in ERM-DA470k/IFCC by multiplying the mean of means of the TFs with the concentration of the protein in ERM-DA470.The relative combined standard uncertainties were calculated as the square root of the sum of squares of the individual contributions, according to Its .Individual uncertainty contributions taken into account were from the characterization (u char ), the calibrant (u cal ), the potential between-vial heterogeneity (u bb ), and the potential hidden long-term instability (u lts ).The manner in which these uncertainties were estimated is described in detail in the certification report of ERM-DA470k/IFCC (10 ).

TRIAL RUN
Twenty laboratories participated in a trial run using the closed-value transfer procedure, measuring IgG, IgA, and CRP (see also the online supplementary material).The trial run was aimed at testing that all the procedures linked to the value transfer (reconstitution, reporting of the results, etc.) were well controlled.We investigated the reasons for outlying values or high day-to-day variations, which turned out to be evaporation problems and temperature control.One analytical run can last for hours, and the samples are left open in most platforms during this time.If the sample volumes are of the order of 100 L, the evaporation of the samples can lead to concentration changes of several percent per hour.

CHARACTERIZATION OF ERM-DA470K/IFCC
The characterization measurements were performed by 18 laboratories (Table 1).They are coded laboratory numbers 1-22, as some laboratories withdrew from the characterization process.In the plots, number 23 was used for laboratories having measured a second data set for a particular protein.Table 1 summarizes the measurements, transfer factors, and analytical platforms used by the participating laboratories.Two of the laboratories measuring with the open protocol did additional measurements using the closed protocol.The planning of the measurements was governed by the aim to obtain 12 data sets per protein, with preferably at least 1 data set for all major platform/reagent combinations with the open procedure.The open protocol cannot be applied on the Beckman Immage instrument.Therefore, data for this platform were obtained only with a closed-value transfer procedure.The measurement of A2M is supported by only a limited number of platforms and laboratories, and only 6 data sets were retained for this protein.For the other proteins, the characterization was based on between 10 and 14 valid data sets each.
Results of the characterization measurements.In total, we obtained 187 data sets.The results are summarized in Table 1, and examples are shown in Fig. 1.As part of the scrutiny of the data, all outlying values were identified, at the level of individual measurements and at the level of TFs.Outliers were not deleted automatically, but possible technical reasons for outlying values were examined with the laboratories involved.As a result the following decisions were made: • A number of typographical errors in the reported values were corrected.• For laboratory 2, data from day 1 were not used, as the data for this day were incomplete, and some issues concerning sample evaporation were still being tested.Because on days 2, 3, and 4, the dilutions were not measured in 1 analytical sequence, but separately, we accepted data giving a correlation coefficient of 0.96 (instead of 0.97).For each protein/laboratory combination, the mean of the valid daily TFs is given (for those laboratory/protein combinations for which data have been obtained).In case the data were valid, the mean of the valid daily determinations is given.In case the data for that laboratory/protein combination was not valid, the main reason is indicated by one of the following abbreviations: CI, non-overlapping confidence intervals; CC, correlation coefficient below the specified limit; DTD, day-to-day variation above specifications.
• For laboratory 8, data from measurement day 2 on AAT, HPT, IgA, IgG, and TRF were not used.There had been a delay in the measurement of certain dilutions because of technical problems with the platform, leading to partial evaporation of these dilutions.• For laboratory 5, data on C4 were withdrawn because of technical reasons leading to reproducibility problems.• For laboratory 4, the data points for the highest concentration were not used for IgG and HPT, because they were at the limit of the measuring interval.
During data analysis, the results obtained when linear regressions were forced through 0 (y ϭ bx) were compared to those when linear regressions were performed allowing for an intercept (y ϭ a ϩ bx).The mean of means of the TFs were comparable, but the SDs were higher in the second case.It was decided to process the data allowing for intercepts in the linear regression, as in that case only measured data were used, and no model other than the linearity within the actual measurement interval was assumed.
We assessed the quality of data by use of R 2 , which is influenced by deviation from linearity as well as  within-run and between-run imprecision.It was determined whether applying other assessments for linearity (comparing R 2 for regression models including polynomials as proposed by CLSI-Guideline EP14) or evaluating within-run and between-run imprecision by ANOVA with random effects (comparable to CLSI Guideline EP5) could provide additional information.Using these approaches, however, we most often detected cases where the nonlinearity or between-run imprecision, though sometimes clearly present, were not significant in comparison with uncertainties associated with the characterization measurements (Fig. 1A, residual plot).This is most probably because many of these assays have been optimized over decades, and any serious problems-with, for example, the linearityhave already been eliminated.
Open vs closed procedure.The open-value transfer procedure is considered to be the preferred option for assigning values to calibrants, and the closed procedure as an alternative for those situations in which it is not possible to substitute the manufacturer's calibrant with the reference material (7 ).The results obtained with the 2 procedures are compared in Fig. 2. The only protein data for which the differences are significant according to a t-test at the 95% confidence level are A2M (P ϭ 0.01) and CER (P ϭ 0.015), but only 2 data sets were measured with the open procedure for A2M.
Results for the individual proteins.The results for 12 of the 14 proteins measured show an acceptably low between-method agreement; Figs.3A and B display representative data for ALB, IgG, CER, and CRP.The relative between-laboratory variation (CVs) for the 12 proteins ranged from 1.37% to 3.02%.The transfer factors (see Table 2) are from 0.87 (A2M) to 1.07 (C4), with an average of 0.94 (SD 0.05).The TFs of the different proteins are different.This means that the concentration in ERM-DA470k/IFCC with respect to the concentration in ERM-DA470 is different for different proteins.This is not due to differences in concentrations in the starting materials (these were comparable) but to differences in the way certain proteins react with respect to the fumed silica used to delipidate the serum The bars represent the laboratory means (SD).The dotted line represents the mean of means, and the bars at the beginning and the end of the dotted line represents the mean of means (SD).The data sets are numbered according to the scheme in Table 1.(6,10 ).Small variations in the properties of the silica and the conditions of the delipidation step can cause variations in losses for certain proteins.The average TF of 0.94 indicates that generally the concentrations of the proteins in ERM-DA470/IFCC are slightly lower than those in ERM-DA470.For the 12 proteins, we calculated certified values by multiplying the individual TF obtained for each protein with the mass concentration of the protein in ERM-DA470 (Table 2).We estimated the uncertainties by combining the contributions from the characterization measurements, the calibrant, the potentially hidden between-vial heterogeneity, and the potential long-term instability.The combined standard uncertainty u c was multiplied by the coverage factor k ϭ 2 (corresponding to a level of confidence of about 95%) to give the expanded uncertainty of the certified values.In most cases, the largest contribution to the uncertainty was that from the uncertainty linked to the calibrant (ERM-DA470).For AAG, AAT, and HPT, the u cal was very small, and the largest contribution originated from the characterization measurements.For all 12 proteins, the uncertainties are sufficiently low for the intended use of the material, i.e., the calibration of homogeneous immunoassays used in clinical chemistry.
For CER, the CV of the laboratory means was 6.5% (Fig. 3C), and that for CRP was 4.5%, which were both high compared with other proteins.From the measure-ments performed for the control of the processing of the material (6 ), it was known that the CRP concentration measured in the lyophilized and reconstituted serum was considerably lower than that measured in the liquid serum directly.Those measurements were confirmed by the results from the characterization study, which showed that the TF is 0.81 for CRP (Fig. 3D).

Discussion
The CRM ERM-DA470k/IFCC has been produced to replace the widely accepted ERM-DA470.To maintain the high degree of comparability of measurement results for serum proteins, there were stringent requirements on the properties of the new material.The detailed information available on the preparation of ERM-DA470 (2 ) made it possible to produce a reference material with similar properties.This is not trivial, as the final processing procedure applied for ERM-DA470 was the result of several years of work of a large team of people, and even the smallest details had been optimized.
For clinical CER measurements, the use of ERM-DA470 had not led to the same degree of harmonization of measurement results (12 ).This protein is known to exist in different forms, with possible differences in the number of copper ions bound to the protein, and to have a high susceptibility to proteolysis; the a The certified values are valid when the material is reconstituted according to the specified procedure.U CRM , expanded uncertainty of the CRM with a coverage factor k equal to 2; u bb , relative standard uncertainty related to the between-bottle heterogeneity; u c , combined relative standard uncertainty; u cal , relative standard uncertainty of the calibrant; u char , relative standard uncertainty related to the characterization; u lts , relative standard uncertainty related to the long-term stability of the material; u sts , relative standard uncertainty related to the short-term stability of the material.
predominant form in serum is believed to be a 132-kDa monomer (12)(13)(14)(15).These different protein forms are also not stable over time in stored serum, and Beetham et al. (12 ) speculated that, during the aging of serum, the CER in the serum becomes more comparable to that found in ERM-DA470.It could be concluded from the characterization data that the use of certain methods (Beckman Immage, Architect) leads to higher TF values, whereas the Dako, Roche, and Olympus methods provide lower TF values on the same material.The variation in the results could be understood in terms of different method specificities, but would require further investigations of the properties of CER in the 2 materials.Because the reasons for the between-method variation have not been clarified, no values were assigned for CER in ERM-DA470k/IFCC.For CRP, the concentration found in ERM-DA470k/IFCC is approximately 20% below what would be expected on the basis of the amount of CRP added to the material, showing a similar reduction to that measured on the pilot batches.It was later found that lyophilization can induce changes in the oligomeric structure of CRP (16 ).Therefore, CRP was not certified in ERM-DA470/IFCC, and a separate liquid frozen material was produced (ERM-DA472/IFCC) and found to be suitable.
The reference material ERM-DA470k/IFCC was spiked with B2M, and this protein is expected to be certified at a later stage.The B2M in the reference material has been found to be stable and commutable (6 ).
It was possible to establish certified values for 12 proteins in ERM-DA470k/IFCC.The concentrations in ERM-DA470k/IFCC are comparable to those in ERM-DA470, although slightly lower.The proteins in the material have been shown to be stable over time (6 ), and the material has the required properties in terms of optical clarity.Here we have demonstrated that it has also been possible to assign robust values for the 12 serum proteins to ERM-DA470k/IFCC, by using the value transfer procedures developed originally for ERM-DA470.These procedures were developed in the 1990s (7 ) and are listed by the Joint Committee for Traceability in Laboratory Medicine as reference procedures.The characterization measurements were performed with a wide range of immunoassays, with different principles (nephelometry, turbidimetry), different antibody selectivities, etc.The procedures have been further optimized for the use with the current immunoassays for serum proteins.The closed-value transfer procedure has now also been optimized for all the platforms participating in the study and provided results equivalent to those obtained with the openvalue transfer procedure.The between-method varia-tion was small for results from the characterization measurements for all 12 proteins, particularly taking into consideration the large variation in assay design and antibodies used, as well as the complexity of the measurands, with often a wide mixture of isoforms being measured.This demonstrates that the set of developed procedures makes it possible for the major serum proteins to transfer values reliably from one material to another.These value transfer protocols can also be used for the assignment of values to manufacturers' calibrators or control materials (9 ).Therefore it can be expected that the state of standardization for these 12 proteins will be maintained with the use of ERM-DA470k/IFCC.ticipation of the IFCC Committee for Plasma Proteins (C-PP), and the support of the Scientific Division of the IFCC.We are thankful to Lars-Olle Hansson, Anders Larsson, and Tom Ledue for thoughtful discussions.The contribution of the following laboratories participating in the value assignment has been highly appreciated: Abbott Diagnostics, Irving (US); Centro Interdipartimentale sulla Riferibilita `Metrologica in Medicina di Laboratorio (CIRME), Universita degli Studi di Milano (Italy); Beckman Coulter, Brea (US); Dade Behring Marburg GmbH-A Siemens Company, Marburg (Germany); Dako Denmark, Glostrup (Denmark); Denka Seiken, Tokyo (Japan); Eiken Chemical Co., Tokyo (Japan); Foundation for Blood Research (FBR), Scarborough (US); Klinisk Kemi Malmö, Lunds Universitet (Sweden); Kreiskliniken Altoetting-Burghausen (Germany); Laboratoire d'immunologie, Centre Hospitalier Lyon-Sud (France); Laboratorio di Biotecnologie e Laboratorio Centrale, Policlinico San Matteo, Pavia (Italy); Medical Biological Laboratories.Co., Nagano-Ken (Japan); Mitsubishi Kagaku Patron Ins., Shinjuku-ku (Japan); Nitto Boseki, Fukushima (Japan); Odense Universitetshospital (Denmark); Olympus Life and Materials Science, Clare (Ireland); Protein Reference Unit, St. Georges Hospital, London (UK); Roche Diagnostics GmbH, Penzberg (Germany).

Fig. 1 .
Fig. 1.Examples of the measurement results obtained.(A), Example of results obtained with the open procedure.Data are shown from laboratory 5 on day 2 for AAG.Left, individual measurement results plotted as a function of the relative concentration of the material calculated from the dilutions (R 2 for the linear regression 0.99).The slope corresponds to the TF.Right top, Y-residuals plotted; bottom right, Q-Q plot, which is a graphical method for comparing 2 probability distributions by plotting their quartiles against each other.The linearity of the data indicates a normal distribution.(B), Example of results obtained with the closed procedure.Data are shown from laboratory 19 on day 2 for TTR.Left, measurement results for ERM-DA470; right, for ERM-DA470k/IFCC.The ratio of the 2 slopes corresponds to the TF.

Fig. 2 .
Fig. 2. Comparison of TFs obtained with open (light bars) and closed (dark bars) procedures.The bars correspond to the average values per protein.SDs are represented by error bars.