| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Endocrinology and Metabolism |
1 Laboratory for Analytical Chemistry, Faculty of Pharmaceutical Sciences, Ghent University, Gent, Belgium.
2 Laboratory of the Government Chemist, Teddington, United Kingdom.
3 Institute for Clinical Biochemistry, University of Bonn, Bonn, Germany.
4 National Institute of Standards and Technology, Gaithersburg, MD.
aAddress correspondence to this author at: Laboratory for Analytical Chemistry, Faculty of Pharmaceutical Sciences, Ghent University, Harelbekestraat 72, 9000 Gent, Belgium. Fax 32-9-264-8198; e-mail linda.thienpont{at}ugent.be.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: Four candidate RMLs (cRMLs) developed/implemented variants of a candidate reference measurement procedure (cRMP) based on isotope dilution–liquid chromatography–mass spectrometry. The sole constraint implemented was calibration with a common thyroxine primary calibrator. The RMPs were externally validated and assessed for comparability in round-robin trials using common samples, i.e., 5 lyophilized and 33 frozen native sera. At the same time, the performance of the cRMLs organized in a network was assessed. For uniform external quality assessment, common performance specifications were agreed on.
Results: All cRMLs performed the cRMPs with fulfillment of the predefined specifications: total and between-laboratory CVs 
2.0% and 2.5%, respectively, and a systematic deviation 0.9%, estimated with a target assigned from the mean of means obtained by the cRMLs. The mean expanded uncertainty for value assignment to the native sera was 2.1%.
Conclusions: A network of cRMLs, with externally conformed competence to properly perform RMPs, has been established. Performance specifications were defined and will form the basis for admittance of new network members. A serum panel, successfully targeted during the validation process, is available for split-sample measurements with commercial routine measurement procedures. The model can now be used for other measurands for which traceability to the Système International d’Unités is needed.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Although three other EN/ISO standards, i.e., 15193, 15194, and 15195 (12)(13)(14), cover the essential requirements of the elements of such a RMS, the diagnostics industry was initially stymied by how to identify them for each specified measurand; in other words, the legislation was in place but not the tools by which to comply. In this respect, two initiatives were taken. In the first initiative, the International Committee of Weights and Measures, the IFCC, and the International Laboratory Accreditation Cooperation agreed to cooperate toward the establishment of a Joint Committee for Traceability in Laboratory Medicine (JCTLM) (15), with the mission statement to categorize higher order reference materials and RMPs and to establish criteria and processes assessing the competencies required for laboratories performing the RMPs. In the second initiative, the European Commission funded a project with the reference G6RD-CT-2001-00587, aiming at a RMS for thyroid hormone measurements as a model (16). The project members consisted of expert laboratories in Belgium (Ghent University), the United Kingdom (Laboratory of the Government Chemist), Germany (University of Bonn), and the United States (NIST). The project’s topic was chosen with great care via a worldwide survey conducted by the Scientific Division of IFCC to identify "measurands of priority" in laboratory medicine. As a result, it was considered that establishing metrologically traceable measurements of thyroid hormones would be of great support to the correct diagnosis and monitoring of therapeutic measures. With respect to the routine measurement of thyroxine (T4), the free rather than the total hormone fraction was promoted as a prime diagnostic measurand (17). However, the controversial analytical principle of some of the commercial immunoassays for free T4 seems to have undermined the anticipated diagnostic potential (18)(19)(20), with the consequence that even the most recently developed automated test systems still include total T4. It was therefore decided that establishing SI traceability of the measurement of total T4 would be the first part of the European project.
Here we describe our investigation of the feasibility of developing such a RMS. Although a primary T4 calibrator was lacking initially, such a calibrator has since been established. A comprehensive description of the certification process for this calibrator according to the EN/ISO 15194 guidelines will be given elsewhere. We therefore report here only the development of RMPs and reference materials. We also deal with the project’s efforts to organize the candidate reference measurement laboratories (cRMLs) in a network, which required establishment of and agreement on a process to assess proper performance of the members of the project group at the inception of the network and of candidate members subsequently applying for admittance.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
measurement protocol and performance specifications
Before starting the implementation and/or development of the cRMPs, the project laboratories agreed on a validation strategy. They adopted the specifications for the precision of a total T4 RMP as published previously by a European Working Group, i.e., a maximum total CV of 2.0% for a measurement protocol consisting of duplicate analyses on three independent occasions (28). In addition, it was specified that the difference between duplicates in a measurement series should not exceed 2.5% and that the within-run CV should be 1.5%. During the meetings held to evaluate progress of the work, the laboratories were required to disclose all encountered problems and to show representative reconstructed ion chromatograms. A laboratory could claim to have a cRMP available only after sufficient internal assessment to show that the predefined performance specifications were met. At that stage, assessment of the comparability among the different cRMPs could be started.
reference materials
Two types of candidate reference materials were investigated to conduct the external validation of the cRMPs (see below): 5 lyophilized serum materials with different T4 concentrations, purchased from the German Society of Clinical Chemistry and Laboratory Medicine (Bonn, Germany; 
500 vials containing 3 mL of serum for each concentration) and 33 frozen off-the-clot sera, each obtained from a single blood donation by an apparently healthy male or female donor. The frozen sera were purchased from Scantibodies Laboratory, Inc. All blood collections were carried out according to accepted protocols at blood banks regulated by the US Food and Drug Administration. One hour after collection and coagulation of the blood at room temperature, serum was isolated by centrifugation. Only those units found negative for the presence of antibodies to HIV I/II, syphilis, and hepatitis B surface antigen were shipped to Scantibodies (refrigerated between 2 and 8 °C) for further processing. T4 standard material was added to 3 of the units to achieve hyperthyroid concentrations; the remaining 30 units, however, were native. No preservatives were added, but the sera were filtered through 0.22 µm filters to assure sterility and were fractionated into 1-mL portions in polypropylene vials. For each blood donation, a maximum of 150 aliquots were available. After aliquoting, they were immediately stored at –70 °C and shipped to the participating laboratories on dry ice with continued storage at –70 °C.
stability study of the reference materials
The candidate reference materials were subjected to a stability study. The protocol was from the IRMM and consisted of a short- and long-term isochronous study for the lyophilized sera, i.e., storage for 2 weeks at 40 and 4 °C, with –20 °C as the reference temperature, and for 12 and 18 months at 4 and –20 °C, with –70 °C as the reference temperature. The frozen sera were tested after long-term storage at –70 °C (also during 12 and 18 months) vs –160 °C as the reference point. In each study, two measurements per time point were to be performed by ID-MS. A one-way ANOVA was used to evaluate the data.
external validation of the performance of the CRMPS
External validation of the cRMPs was done in several round-robin trials with blind measurement of common samples. In the first and second intercomparisons, the laboratories were asked to analyze the above-described lyophilized sera in duplicate on three independent occasions. Results from all individual measurements were assessed for fulfillment of the predefined specifications. Subsequently, for each set of results the arithmetic mean, SD, and CV were calculated, and afterward, the mean of means and the within- and between-laboratory CVs were calculated. The latter was done via a one-way ANOVA. In deciding whether there was sufficient comparability of the RMPs, the limit for the between-laboratory CV was predefined as 2.5%. For the final validation of the cRMPs, the panel of 33 frozen sera was measured. The measurements were divided among the cRMLs in such a way that finally three sets of duplicate results would be available per serum. Note that for these measurements, duplicates meant measurement of each serum in two independent measurement series.
validation of the performance of the CRMLS in a network
The round-robin trials were also used to assess the performance of the cRMLs in a network. The assessment criteria were the above-mentioned precision specifications (total and between-laboratory CV) extended with a limit for systematic deviation of 0.9% (28). The latter was estimated from repeated measurements of the two lyophilized serum materials targeted in the second round-robin trial by the four cRMLs. In addition, the capability to deliver measurement results before a deadline was a criterion to judge adequate performance of a network member.
calculation of the uncertainty of values assigned to frozen off-the-clot sera
The measurements of the frozen samples by the four cRMLs enabled the assignment of total T4 target values and the estimation of individual measurement uncertainties. The latter was done in compliance with the Guide to the Expression of Uncertainty in Measurement (GUM) (29). Because the uncertainty of the purity of the calibrator was not yet known, it was provisionally set to zero; hence, only the uncertainty attributable to measurement was considered. A single-factor ANOVA gave evidence for the fact that the latter consisted of both the within-laboratory imprecision and the between-laboratory variation. The former was estimated for each cRML as the within-group SD calculated from the individual serum measurement values (duplicates per serum) in a single-factor ANOVA. The within-laboratory SD, divided by the square root of 2, allowed calculation of the uncertainty of each cRML’s reported mean serum value (uncertainty expressed in percentage compared with the mean of the mean values for each serum). Combination of these uncertainties for the measurements by the three (occasionally two) laboratories that targeted the same serum (via the square root of the summed squares) gave the combined uncertainty of the serum’s assigned value (the mean of the three mean values) attributable to the within-laboratory variation. The uncertainty attributable to the between-laboratory variation was estimated for each serum from the deviation between the highest and lowest reported mean values (also expressed as a percentage). The mean of all 33 deviations was then used to estimate the between-laboratory uncertainty according to the rectangular distribution; hence, a constant value was assumed. The final combined uncertainty was estimated from combination of the within- and between-laboratory uncertainties (quadratic addition and square root). The expanded uncertainty was calculated with a coverage factor based on the Student t-distribution (95% confidence level).
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
.
|
In the second round-robin trial, the four sets of results could again be accepted for the calculation of the mean of means (± 95% confidence interval), i.e., 96.74 ± 1.06 and 130.0 ± 1.2 nmol/L. In this trial, the maximum within-laboratory CV was estimated to be 0.5%, and the maximum between-laboratory CV was estimated to be 0.6% (Fig. 2
).
|
In the final round-robin trial, the cRMPs were validated for measurement of the frozen, off-the-clot sera. For internal accuracy control of these measurements, the two lyophilized sera of the second round-robin trial and their target values (mean of means) were used. On average, 7–10 measurement series were needed by each cRML to complete the duplicate measurements of the frozen sera. The systematic deviation estimated from these repeated measurements ranged from –0.2% to –0.8%. Note that on the basis of the described internal accuracy control measures, one laboratory decided to withdraw its results for four sera. Through this final validation trial, each serum was assigned a target value from the mean of three sets of duplicates (except for the above-mentioned sera, for which only two sets of duplicates were available). Because it is the intention to use these sera in split-sample measurements with routine immunoassays, the individual concentrations, covering a total T4 concentration range between 50 and 300 nmol/L, cannot be published here. The within-laboratory SDs for measurement of the sera by the cRMLs, as estimated from a single-factor ANOVA, were 1.18, 0.96, 0.61, and 0.64 nmol/L, respectively. The mean between-laboratory uncertainty was estimated to be 0.55%. Use of these values in the estimation of the expanded relative uncertainty gave a mean of 2.1% (range, 1.5–2.9%).
One-way ANOVA of the data for the short- and long-term stability assessments of the lyophilized and frozen sera did not reveal a significant difference between the storage conditions (P >0.05).
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
reference materials
It is obvious from this study that serum-based matrix reference materials are an important tool in the process of validation of newly developed/implemented measurement procedures and/or of internal accuracy assessment for cRMPs. They also play a prime role in establishing the relationship of routine measurement procedures with the SI unit (11). Whereas their commutability is not an issue in combination with matrix-independent RMPs, it is of utmost importance with regard to routine measurement procedures [see, e.g., Ref. (32)]. The European project members, representing the diagnostics industry, therefore urged that reference materials with matrices closely matching patient sera be included in this feasibility study. Use of such materials in the traceability chain at the level of the manufacturer’s calibrators is not only permissible by the EN/ISO 17511, but is even recommended for validation of metrologically traceable calibration. From this perspective, the project selected authentic, single-donation, native off-the-clot sera with concentrations as evenly distributed as practicable over the whole of the measuring interval for serum total T4. Because it was considered unethical to collect blood from individuals suffering from hyperthyroidism, T4 was added to three donations to obtain materials with increased T4 concentrations. The only manipulations that all serum donations underwent were sterile filtration to avoid bacterial contamination and freezing for logistic reasons.
For the purpose of validation and internal accuracy control of the cRMPs, the project used lyophilized sera. The advantage of this type of material lies in the fact that it is available in large batches and is prepared to last for several years. The rather restricted number of vials purchased for this project (500 per concentration) should not be seen to be in conflict with the latter statement because the current project was only a pilot study investigating the feasibility of establishing a RMS for total T4. The preparation of a large batch of commonly available lyophilized reference materials may be a future task for institutes such as the IRMM. On the other hand, with respect to the number of 1-mL aliquots of the single-donation sera, the diagnostics industry should realize that these materials are available only in limited stock sizes (the certification process with the cRMLs and the stability study consumed 90 of the 150 mL that was the maximum obtained per donation).
The isochronous stability study performed for both types of serum-based matrix reference materials allowed us to infer that the sera remain sufficiently stable to last for several years after certification, if stored under appropriate conditions.
validation of the CRMPS
The results of the external validations (Figs. 1
and 2
) demonstrate not only that all laboratories were successful in fulfilling the predefined total CV requirements, but that the results obtained with the four cRMPs agreed well (see the between-laboratory CVs for the first and second round-robin trials). Nevertheless, the progress in intercomparability from the first to the second trial was remarkable (the deviation between the lowest and highest sets of results was reduced from 4.7% to 1.6%). This was most probably attributable to an increase in skill in performing the cRMPs. The outcome of the internal accuracy control further shows that the cRMLs were able to maintain the accuracy of their performance over a longer period with fulfillment of the 0.9% limit for systematic deviation. With respect to the assignment of target values by different cRMLs, the expanded relative uncertainties calculated for the frozen sera demonstrate that this is possible with sufficient reliability. Nevertheless, a significant ANOVA suggested that both the within- and between-laboratory variation contributed to the measurement uncertainty. Generally speaking, they were in the same order of magnitude, but there were sera for which the contribution by the within-laboratory imprecision was negligible compared with the between-laboratory imprecision, whereas the converse was never the case.
The assumption that the within-laboratory SD is constant for the measurement of all sera is justified from the fact that for each serum the same absolute amount of analyte is taken through analysis. Which contribution was the highest depended on the serum: the higher the concentration, the more dominant the between-laboratory effect became. This is because the latter was assumed constant. Although the between-laboratory variation might again be explained by small differences in the skill levels in the individual cRMLs, the most probable cause lies in small systematic differences in the measurement procedures (e.g., degree of sample purification, chromatographic, and MS resolution). The latter effects come to light only when many samples, each with its individual matrix, are measured.
organization of the CRMLS in a network
A premise to the success of the concept of ensuring/demonstrating SI traceability via RMSs is not only the availability of validated RMP(s) but also of competent RMLs. As explained earlier, it is not the manufacturer who should be left with the decision about the competence of a RML. Therefore, the JCTLM established objective criteria and processes to do this in a sufficiently transparent way through the publication of a database (15). This is one mechanism applicable to existing RMLs. However, because it is foreseeable that the needed RML capacity will be high in the near future, a second mechanism is required to ensure that different cRMLs perform at the same level of analytical quality. Only if this can be guaranteed will a manufacturer be confident in the analytical services provided by whichever cRML available at the time of SI-traceable value assignment is needed. In this respect, it has long been recommended that the best mechanism consists of organizing cRMLs in a network (28)(33) where they are regularly subjected to external performance assessments. Because at present only a few networks exist, e.g., the IFCC hemoglobin A1c and the Cholesterol Network (31)(34), it became an additional aim of the European project to organize the four cRMLs into a network for measurement of serum total T4. The underlying idea was that the network would not only be a tool to assess proper performance of the cRMLs in the time span of the project, but should be maintained after the project. Hence, it would create an opportunity for increased RML capacity by setting a system in place to assess laboratories that apply for admittance in the near future.
As can be inferred from the results presented in this study, a network with regular organization of external quality assessment is an excellent mechanism to show that existing network members or potential new members perform the cRMPs at the same level of analytical quality. It also cannot be emphasized enough that common internal accuracy control materials with reliable target values are the key to proper performance of a cRMP, even in an experienced cRML, as well as to maintaining the analytical level in a RML network. The fact that in this study one of the laboratories decided to withdraw its results on the basis of the internal accuracy measures can be used as an argument for this statement. Hence, for the future of the total T4 network, it will be necessary to provide the cRMLs with new materials in time to perform overlapping runs with the old material.
In conclusion, the study conducted in the framework of European project G6RD-CT-2001-00587 focused on the development of an RMS for total T4 in serum as a model measurand. Different variants of ID-MS cRMPs are now available and are performed with predefined analytical specifications in four cRMLs organized in a network. The next phase in the project will be the development of a technical implementation plan. This will comprise nomination of the cRMPs and cRMLs for review by the JCTLM and potential inclusion in the database. In this way, the diagnostics industry will become acquainted with their availability. The network activities also will need to be continued according to the mechanism developed here. However, in view of the expected need for increased RML capacity, it is hoped that other laboratories will show interest in network membership. The decisive step to demonstrate that the model RMS works toward establishing SI traceability of routine measurement procedures will be the organization of split-sample measurements with routine commercial total T4 procedures. This step recently has been initiated. The invitation to participate was not restricted to European in vitro diagnostic companies but was launched worldwide because SI traceability is an issue of interest to all continents. At least 15 companies accepted the invitation and received aliquots of the native serum panel. The data from these parallel measurements will be interpreted in a method-comparison study by graphical and statistical techniques. Last but not least, continuation of this project toward the development of a RMS for free T4 would be desirable.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  K. Van Uytfanghe, D. Stockl, H A. Ross, and L. M. Thienpont Use of Frozen Sera for FT4 Standardization: Investigation by Equilibrium Dialysis Combined with Isotope Dilution-Mass Spectrometry and Immunoassay Clin. Chem., September 1, 2006; 52(9): 1817 - 1821. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. M. Thienpont, K. Van Uytfanghe, J. Marriott, P. Stokes, L. Siekmann, A. Kessler, D. Bunk, and S. Tai Feasibility Study of the Use of Frozen Human Sera in Split-Sample Comparison of Immunoassays with Candidate Reference Measurement Procedures for Total Thyroxine and Total Triiodothyronine Measurements Clin. Chem., December 1, 2005; 51(12): 2303 - 2311. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
