Trends in quality assurance of metal determination in clinical chemistry† K. Byrialsen*, J. Kristiansen and J. M. Christensen National Institute of Occupational Health, Lersø Parkall�e 105, DK-2100 Copenhagen, Denmark A summary is given of the main strategies that can be used to obtain high quality results in the determination of metals in clinical chemistry. The trends in quality assurance of metal analyses are discussed. Keywords: Quality assurance; metals; clinical chemistry; human health Many decisions in a modern society are based on measurement results.For example, traded goods must meet certain quality requirements, industrial and agricultural pollution must be below certain limits, occupational exposure limits must not be exceeded, etc. Compliance with the desired quality or limit may be decided by measurement. Furthermore, actions to be taken in relation to risk assessment and human health and diseases are often related to measured changes in biomarker concentrations.Thus, measurement results have large economic, social and political impact on people and on society, and the reliability of measurement results is of crucial importance. Determination of metals is important in both the environmental and occupational health field and in clinical chemistry. Moreover, metal analysis has a long tradition in the fields mentioned and strategies for demonstrating the reliability of the results have been originated and matured within this type of analytical work.State-of-the-art Traditionally, measurement results are considered to be of high quality if the accuracy (i.e., trueness and precision) of the results is high.1 Determination of elements in trace amounts in clinical and environmental specimens poses special problems with respect to accuracy. For example, trueness may be affected by contamination or by inappropriate blank correction. Losses and contamination may occur during the sample pre-treatment procedure, and matrix interferences may significantly reduce precision and also affect trueness.Moreover, the matrix of calibration standards may not match that of the samples or the calibration standards may be subject to contamination or element loss during storage. In both cases, trueness of the measurement results will decrease. In addition, results are also affected by ‘blunders’ (gross errors). This type of error is often underestimated in daily laboratory work, and it is seldomly subject of systematic investigation.The need for laboratory personnel to be sufficiently trained and motivated is another important factor as up to 25% of all operating deficiencies are directly attributable to operating personnel.2 Laboratories aiming at high quality work must demonstrate that their measurement results meet the required level of trueness and precision. Several established quality control practices are available for the realisation of this goal and most of them have been thoroughly described in the literature and/or in standards and guidelines for quality assurance.For example, the applicability of a given analytical method must be assured on the basis of a well designed and carefully performed method validation study, as described by, e.g., Wernimont3 and Christensen et al.4 Biased measurement results have to be corrected on the basis of a method evaluation study, and inappropriate procedures may be improved resulting in reduced bias and higher precision.5 Another example of a well-established quality control practice is internal quality control which includes analysis of control samples and blank samples.Several standards describe control charts for data on control samples, for example, the ISO standards 7870,6 7873,7 79668 and 8258,9 and the IUPAC guideline on internal quality control procedures.10 The appropriate use of internal quality control serves to detect and correct large non-random changes in accuracy.Such changes will inevitably arise from changing environmental conditions, wear of the measuring instrument or parts of it, change of reagents used in the measurement process, human and technical errors, etc. The analysis of certified reference materials (CRMs) can also be used to demonstrate trueness and precision but because of the high cost of CRMs and the limited supply, the use of CRMs should be analysed less often than control samples.Because of the high degree of trust that can be put in the certified value, CRMs are particularly valuable in demonstrating the trueness of measurement results. In this context, the ISO guide 33 describes how to use CRMs.11 As the understanding and prediction of the impact of trace elements in the human body has increased, it has become evident that determination of total element concentrations may not be adequate. The determination of the abundance, distribution and toxicity of elements can be understood only in terms of trace element species.The key to successful determination of separate trace elements species is preservation of the species during all acts from sampling to the detection of the analyte. Therefore, availability of CRMs with certified values of trace metal species is of crucial importance to improve the analytical quality. However, great difficulties arise when speciation CRMs are produced as documentation of all critical parameters (e.g., homogeneity and stability) must be determined for each certified species.12 Test results should be mutually agreed across frontiers in favour of the users, and international cooperation between national accreditation bodies serves to achieve mutual recognition of test results.Therefore, test results must be comparable and participation in external quality assessment (EQA) schemes serves as an important tool for obtaining comparability. Laboratories seeking accreditation according to EN 45001 are requested to participate in EQA schemes.13 In view of use of EQA results for purposes of accreditation, methods of evaluation of laboratory performance should be comparable and wherever appropriate harmonized.Unfortunately, at present a laboratory can be judged differently by different EQA schemes as a result of differences in performance evaluation procedures. 14 The quality control practices mentioned above have been available in trace element analysis for several years.It is indisputable that, in general, implementation of good quality control practices in a laboratory lead to improvement of the † Presented at The Sixth Nordic Symposium on Trace Elements in Human Health and Disease, Roskilde, Denmark, June 29–July 3, 1997. Analyst, January 1998, Vol. 123 (7–12) 7quality of the results produced by the laboratory. However, as the needs of society are changing, so are the demands on the laboratories and so must the demands on quality control procedures. Metrological trend The most noteworthy trend in quality assurance within the field of analytical chemistry is the ongoing activities aiming at establishing a sound metrological basis for analytical measurements.This trend includes introduction of ‘new’ concepts, such as traceability and uncertainty. Quality requirements for the production, certification and use of certified reference materials are examples of practical implementation of this strategy.15,16 Another example is the International Measurement Evaluation Programme (IMEP) promoted by the Institute for Reference Materials and Measurements.17 Traceability and certified reference materials Traceability means that the result of a measurement (or a value of a calibrator) can be related to a stated reference, usually a national or international standard, through an unbroken chain of comparisons all having stated uncertainties.18 In analytical chemistry, the preferred standard of reference would be the SI base unit mole or alternatively, the SI base unit kilogram.Each step in the comparison chain involves a measurement procedure which contributes some uncertainty. Hence, the total uncertainty is a quantitative indication of the strength of the traceability chain . Certified reference materials (CRMs) play an important role in the practical establishment of traceability. The definition of a M as stated in the international vocabulary of basic and general terms in metrology (VIM)18 implies that CRMs in the context of traceability are valid standards of reference.The problems involved in application of CRMs such as the availability and representativity with respect to matrix and reference value are well known. Production of well defined CRMs with expected long-term stability often puts severe restrictions on the physical form of the material. The necessary processing of the material runs against the demand for CRMs with representative matrix and analyte form. For example, solid materials are often ground and filtered to assure homogeneity, while liquid materials sometimes are lyophilized to improve the stability. Sterile filtration is a possible alternative method to preserve biological fluids, and a CRM for trace elements in sterile filtered liquid serum is under production.19 The representativity of the CRM is not the only area for improvement. The database COMAR contains information of more than 9000 certified reference materials.20 A recent REMCO status report shows that about 1500 reference materials are under certification.21 At present, there are no general requirements to CRM producers to demonstrate their competence, and therefore it cannot be guaranteed that all CRMs in the COMAR register warrant the definition of a CRM and to the demands for strict documentation, as required by, e.g., the ISO guides 31,22 3423 and 35.24 Guidelines for production and certification of reference material issued by, e.g., the Standards, Measurement and Testing Programme (SM&T, formerly the BCR)15 and the National Institute of Standards and Technology (NIST)25 may serve as a practical reference for other minor certifying bodies.The process of certifying a BCR CRM includes strict adherence to these guidelines and scrutiny of the certification report by a certification committee consisting of a panel of expert scientists.This procedure ensures that CRMs issued by BCR are of high quality with respect to documentation and with respect to metrological quality, i.e., traceability and uncertainty. It has been suggested that the ISO guides and other guidelines (e.g., the BCR guide)15 could be implemented in PC-programs which may be valuable tools in the certification process, in particular for small certifying bodies.26 A PC-program specific for statistical evaluation of certified values and uncertainties on the basis of data from interlaboratory studies have been developed for that purpose. The program will generate standardized reports with full documentation of the evaluation procedure in accordance with the BCR guide.26 Uncertainty It is impossible to talk about traceability without referring to uncertainty, and vice versa.Uncertainty (of measurement) is defined as parameter, associated with the result of a measurement that characterized the dispersion of the values that could reasonably be attributed to the measurand.18 In analytical chemistry, the measurand typically is the molar (or mass) concentration of a substance in a given sample.The wording ‘dispersion of values that could reasonably be attributed to the measurand’ implies that a standard deviation based on repeated measurements does not, in general, satisfy the definition of uncertainty. A standard deviation, e.g., derived under repeatability or reproducibility conditions, is a measure of the laboratory (or a group of laboratories) ability to repeat the measurement.However, because the result of the measurement may be influenced by effects beyond the control of the laboratory (or a group of laboratories), a standard deviation is, in general, an underestimate of the uncertainty. In evaluating the uncertainty, the laboratory must assign a value to all relevant uncertainty components and combine the uncertainty components to yield an estimate of the uncertainty of the result.27 The principle is illustrated in Fig. 1, where Laboratory A does not consider the quality of the calibrators purchased from Company A or the systematic effects associated with the measurement Fig. 1 Uncertainty estimation. Laboratory A ignores uncertainty associated to the calibrator and the measurement principle and produces analytical results far from the true value. Laboratory B takes the uncertainty into account and, therefore, the true value is included in the confidence interval associated with the analytical result. 8 Analyst, January 1998, Vol. 123principle used by the laboratory when evaluating the ‘uncertainty’. As shown in the figure, the laboratory produces results far from the given reference value, and the deviation from the reference value is not accounted for by the reproducibility standard deviation reported by the laboratory. Moreover, the low reproducibility standard deviation may give the laboratory’s customers a false impression of the laboratory’s performance.On the other hand, Laboratory B is concerned with all aspects of the measurement. The laboratory concludes that the values assigned to the calibrators purchased from Company A are exact, and that bias from the measurement principle is absent. As no corrections are introduced, Laboratory B arrives at the same result of measurement as Laboratory A. However, Laboratory B estimated the calibrators values to be exact and the measurement principle to be free of bias with some uncertainty.These uncertainty components are combined with the laboratory’s reproducibility to yield a combined uncertainty of the result of measurement. Contrary to the reproducibility standard deviation, the uncertainty describes the deviation from the reference value. The results produced by Laboratory B will be in agreement with results produced by other laboratories following the same principles of uncertainty estimation.An example of uncertainty evaluation of a method for determination of cobalt in urine is shown in Fig. 2. More examples of practical applications of the uncertainty concept within the field of analytical chemistry are becoming available27,28 and more are needed. The introduction of the concept of uncertainty (as defined in the VIM18) in analytical chemistry is not a straightforward task. One important question is how to handle uncertainty components that cannot be evaluated by statistical treatment of data from repeated measurements, i.e., the so-called Type B uncertainty components.The uncertainty of defining the measurand, a significant problem in, e.g., determination of proteins, cannot easily be attacked by statistical means. Moreover, in environmental analyses the matrix may differ very much between samples, and in practice it is difficult to estimate the matrix interference for all types of samples, although such an influence is presumed to exist.The uncertainty of such components can only be evaluated by taking all available information in consideration. This information could be literature data, data from a manual, data obtained by other assays, or judgement carried out by an expert.27 This may seem forbidden to the chemist who clings to scientific objectivity, but one must recall that such ‘less objective’ procedures are already in use in other scientific fields, e.g., in risk analysis of chemicals, which involves extrapolation between species and application of a variety of safety factors in the process. The scientist must recognize, that the doubt about the exactness of a measurement result (i.e., the uncertainty) cannot always be quantified with the same scientific objectivity as the result itself.This situation may be acceptable as long as the evaluation of uncertainty is clearly documented. In the discussion of uncertainty it must be remembered that the measurement result is the primary outcome of a measurement and the uncertainty is a secondary (but not insignificant) outcome.International measurement evaluation programme Demonstrating the need for reference measurements, the Institute for Reference Materials and Measurement (IRMM) in 1988 set up the International Measurement Evaluation Programme (IMEP). In IMEP the results from the participating laboratories are compared with a reference range that contains a value that is traceable to SI system of measurements in order to give an objective picture of the state-of-the-art practice of measurements in field.The reference values are either ‘certified’ or ‘assigned’, and to qualify as a certified reference value measurements must have been performed with a primary method of measurements yielding reference values traceable to international system of units (SI). All reference values in IMEP are accompanied by combined uncertainties taking into account all known sources of uncertainty.With the reference values traceable to SI, a transparent realisation of international comparability of results has been performed and presently IRMM is making preparations for a large study scheduled for 1997–1998. The starting point is to establish traceable values for as many of the following inorganic components as possible: Ca, Cu, Fe, K, Li, Mg, Na, Se, Zn, in human liquid serum. Furthermore, future rounds of IMEP are planned to evaluate measurements of Pb and Cd in whole blood, Cd in urine and Al in serum.17 Trends in standardization In 1997 the guidance document EAL-G25 was released by European Cooperation for Accreditation of Laboratories (EAL).30 The document is a supplement to ISO/IEC Guide 2521 with special emphasis on the accreditation of medical laboratories.As in the ISO/IEC Guide 25 validation of analytical methods is required. It is stressed that standard or published methods shall not be taken for granted and their implementation in the laboratory shall be validated.Validation of new, original or in-house methods and procedures shall necessary be more comprehensive. For a measurement procedure, validation shall include, as appropriate, calibration function, analytical specificity, and sensitivity. It shall also include investigation of useable calibration materials (e.g., their traceability), estimation of the uncertainty of measurements (after corrections for known systematic deviations and based on within and between run variations) and suitable control systems (control materials, internal quality control and external quality assessment).Moreover, the laboratory will have to demonstrate the usefulness of the results, e.g., by establishing reference intervals, by participation in EQA schemes based on human samples and by establishing the clinical sensitivity and specificity. The total uncertainty of measurements (including sampling) should be in accordance with the clinical requirements for accuracy.30 Recognizing the value of intercomparison studies in external quality control, both ISO and IUPAC have issued guidelines focusing on interlaboratory studies.32,33 A fertile ground for harmonization and collaboration among EQA scheme organizers already exists: in 1996, the European Committee for EQA Programmes in Laboratory Medicine (EQALM), was established. EQALM is a forum for cooperation and exchange of knowledge about quality-related matters, especially with regard to external quality assessment.Among other activities, EQALM intends to organise meetings with scientific or practical themes Fig. 2 Example of uncertainty evaluation of a method for the analysis of cobalt in urine. The relation between the total uncertainty, RSDtot, and the single uncertainty components, RSDi, is calculated using the equation (RSDtot )2 = S (RSDi )2 . Major contributions to the total uncertainty in this example are recovery and matrix effects.Analyst, January 1998, Vol. 123 9for members and other interested parties, to establish working groups related to specific scientific matters and to issue scientific publications. Only organizations or regional EQA schemes with more than 100 participants can become full members.34 Serious errors may occur at the state of sampling. Therefore, sample collection guidelines (technical report) for trace elements in blood and urine have been issued by IUPAC.35 The guidelines include the most important elements measured in occupational and clinical chemistry and embraces harmonized procedures for collection, preparation, analysis and quality control. Recently, it has been decided that the coming revison of ISO/ IEC Guide 25 will be upgraded to international standard and, in addition to this, the coming revision of EN 45001 will be in concordance with the revised ISO/IEC Guide 25.36 However, standardization by itself will not result in improvements of the quality of output.The classical quality management is focused on corrective actions to be made in order to secure continuous conformity to the proper standards. A new trend is gradually introduced in the field of analytical chemistry: total quality management (TQM) which focuses on continuous quality improvements. The motivation for improvements will not arise from standard-regulated activities. Improvements occur when the analysts are constantly on the lookout for problems, and have sufficient energy and vitality to improve the quality of their work.In this context, TQM puts greater weight on attitudes towards quality than does classical quality management. As a consequence of the increasing use of laboratory management systems (LIMS) the U.S. Environmental Protection Agency, EPA, has developed Good Automated Laboratory Practices (GALP) which provide the users with principles and guidance for regulations to assure data integrity in automated laboratory operations.37 The document includes a guidance for implementation.The GALPs are developed to assure the reliability of data prepared by laboratories using LIMS to acquire, record, manipulate, store and archive data. Discussion The reliability and usefulness of any test result depends critically on the competence of the laboratory carrying out the test. Laboratory accreditation provides the laboratory with an independent evaluation and recognition of technical competence to perform specific tests.The evaluation covers all aspects of the laboratory operations and is made by technical experts and quality professionals. In the near future, accreditation will provide users of test results with a network of serious and well recognized clinical laboratories where quality control is a matter of daily routine. Furthermore, accreditation will increasingly become a demand from legal authorities as a great advantage from accreditation is the mutual recognition of test results. ISO 900138 or ISO 900239 certification of A laboratories and certification of personnel are not expected to attract laboratories performing routine analyses, but it may be an interesting alternative for laboratories working with research (ISO 9001).Nomination and accreditation of national reference laboratories in the field of clinical chemistry may serve as an enforcing factor to introduce a greater extent of traceability.Reference laboratories can provide traceability by participation in international EQA schemes, the IMEP programme and by production of CRMs with matrices relevant for clinical laboratories. The production of speciation CRMs may be another challenge for the national reference laboratories. Today most EQA schemes covering metal analyses are not operated on a metrological basis. Most of the schemes provide the participants with an evaluation based on consensus values, trimmed means, etc., which can be sensitive to group bias among participants.The consensus mean value (if necessary after exclusion of outliers) may be deemed very reliable if the EQA scheme involves many laboratories and with representation of many different methods including definitive methods. However, the use of an assigned value obtained from concurrent results from a number of expert laboratories is preferable. Only few schemes offer traceability together with the assigned value.This problem should not be solved by using CRMs as control samples in EQA schemes as stocks of the CRMs soon would run out. However, the concept of the IMEP programme is a valuable possibility for obtaining traceability in a EQA scheme and further initiatives must be encouraged. The expression of the uncertainty of a result allows comparison of results with reference values given in specifications and standards. The accreditation standard EN 45001 and the ISO Guide 25 call for uncertainty to be reported for measurement results.Although this demand represents a significant progress because it recognizes that the quality of measurement results is important for the use of the data, this progress is not reflected, in general, in the design of external quality control schemes or in the guidelines for holding interlaboratory studies. In other words, the performance of a laboratory is still judged on the basis of the distance between the result reported by the laboratory and the target value (calculated by the scheme organizer).The distance may be deemed acceptable or not acceptable according to some limits common to all laboratories. However, these limits do not reflect the individual uncertainties of the laboratories and it may happen that a result is denoted ‘unacceptable’ although it is satisfactory viewed from the perspective of the laboratory (see Fig. 3). Introduction of measurement uncertainty in EQA schemes will allow a more harmonized and up-to-date comparison of the test result obtained by the laboratories with the target values and acceptability limits: if the confidence interval of the laboratory result overlaps the acceptance interval in the EQA evaluation, the performance should be considered as acceptable.On the other hand, if the interval is not overlapping, a discrepancy exists which must be investigated and explained by the laboratory. The benefit gained from interlaboratory studies like EQA schemes depends on the design of the scheme.Critical subjects are estimation of the target value, sample distribution, communication and feedback among participants and scheme organizer and the internal politics of the participating laboratory (Are the results reviewed and discussed? If necessary, are improvements initiated?).40 Participation of the laboratories in EQA schemes is required by the accreditation bodies. In this context, it is important that scoring results from participation on EQA schemes are transparent to the laboratories and the accreditation bodies. Poor performance must be subject to a careful Fig. 3 Performance evaluation in external quality assessment. When the laboratory result is outside the acceptance range (shaded area) the laboratory performance will traditionally be deemed ‘unacceptable’. However, if the EQA evaluation takes notice of the uncertainty of the laboratory result (illustrated by the normal distribution) the laboratory performance should rightfully be deemed ‘acceptable’. 10 Analyst, January 1998, Vol. 123examination by the laboratory in order to explain what caused the bad result. The importance of harmonization of existing schemes with respect to performance score must be stressed as this factor imposes a major instigation for improvement of laboratory performance. Further steps in the harmonization of EQA schemes covering trace elements in clinical chemistry should involve harmonization of acceptability limits for results and expected standard deviation for analytes of major concern, i.e., those for which biological exposure indices have been established.In view of the importance of EQA schemes in relation to the accreditation process, the need for organizers of EQA schemes to adhere to an internationally accepted protocol and to become either accredited or certified will be adressed in the future. Today numerous national EQA schemes exist covering e.g., the parameter lead in blood.41 Collaboration among EQA scheme organizers would possibly reduce the number of EQA schemes with a small number of participants leaving resources for EQA schemes to cover less common parameters like different element species.The visible effect of the quality work applied to analytical chemistry is unfortunately affected by the unacceptable high number of human errors. In a recent interlaboratory study we experienced that approximately 5% out of 2462 reported results were affected by human errors, mostly simple calculation errors and reporting of wrong units of measurement.Blunders were reported by 25% of the participating laboratories. In this context, the underlying philosophy of continuous quality improvement from TQM seems to be a valuable step forward. It may be expected that the number of blunders occurring in analytical measurements will decrease if the analytical laboratories implement TQM Conclusion Future demands of laboratories working in the field of clinical chemistry are implementation of traceability and uncertainty, the two indispensable concepts in metrology.Measurement results are no longer acceptable without a statement of a standard uncertainty. Participation in EQA is required in accreditation and, therefore, EQA scheme organizers should adhere to an internationally recognized and harmonized protocol thus providing more transparency in the results to the laboratories and accreditation bodies.With the implementation of traceability and uncertainty and with revised evaluation of laboratory performance in EQA schemes, the field of clinical chemistry will be entering a new era based on sound metrology rather than just comparability. References 1 International Standards Organisation, Accuracy (Trueness and Precision) of Measurement Methods and Results. Part 1: General Principles and Definitions, ISO Standard 5725-1, ISO, Geneva, 1994. 2 Broderick, B. E., Cofino, W. P., Cornelis, R., Heydorn, K., Horwitz, W., Hunt, D. T. E., Hutton, R. C., Kongston, H. M., Muntau, H., Baudo, R., Rossi, D., vaan Raaphorst, J. G., Lub, T. T., Schramel, P., Smyth, F. T., Wells, D. E., and Kelly, A. G., Mikrochim. Acta, 1991, II, 523. 3 Wernimont, G. T., Use of Statistics to Develop and Evaluate Analytical Methods, ed. Spindley, W., Association of Official Analytical Chemists, Arlington, VA, 1985. 4 Christensen, J. M., Poulsen, O. M., and Anglov, T., in Handbook on Metals in Clinical and Analytical Chemistry, ed. Seiler, H. G., Sigel, A., and Sigel, H., Marcel Dekker, New York, 1994, pp. 45–46. 5 Christensen, J. M., Mikrochim. Acta, 1996, 123, 231. 6 International Standards Organisation, Control Charts - General Guide and Introduction, ISO 7870, ISO, Geneva, 1993. 7 International Standards Organisation, Control Charts for Arithmetic Means and Warning Limits, ISO 7873, ISO, Geneva, 1993. 8 International Standards Organisation, Acceptance Control Charts, ISO 7966, ISO, Geneva, 1993. 9 International Standards Organisation, Shewhart Control Charts, ISO 8258, Geneva, 1991. 10 Thompson, M., and Wood, R., Pure Appl. Chem., 1995, 67, 649. 11 International Standards Organisation, Uses of Certified Reference Materials, ISO Guide 33, ISO, Geneva, 1989. 12 Cornelis, R., Mikrochim. Acta, 1991, III, 37. 13 General criteria for the operation of testing laboratories, Committ�e Europ�een de Normalisation/Committ�e Europ�een de Normalisation Electrotechnique, Brussels, EN 45001, 1989. 14 Christensen, J. M., and Olsen, E., Ann. Ist. Super. Sanit`a, 1996, 32, 285. 15 European Commission, Guidelines for the Production and Certification of BCR Reference Materials, Doc. BCR/48/93, Commission of the European Community, Brussels, 1994. 16 Taylor, B. N., and Kuyatt, C. E., Guidelines for evaluating and expressing the uncertainty of NIST measurement results.NIST Technical Note 1297, 1994 edn., United States Department of Commerce, National Institute of Standards and Metrology, Gaithersburg, MD. 17 Lamberty, A., Van Nevel, L., Moody, J. R., and De Bievre, P., Accred. Qual. Assur., 1996, 1, 71. 18 Bureau International des Poids et Mesures, International Electrochemical Commision, International Federation of Clinical Chemistry, International Organization for Standardization, International Organization for Pure and Applied Chemistry, International Organization for Pure and Applied Physics and International Organization of Legal Metrology, International Vocabulary of Basic and General Terms in Metrology, ISO, Geneva, 1993. 19 Christensen, J. M., Kristiansen, J., Heydorn, K., Damsgaard, E., and Cornelis, R., IAEA meeting report, in the press. 20 Klich, H., in Quality Assurance and TQM for Analytical Laborato. Parkany, M., The Royal Society of Chemistry, Cambridge, 1995, pp. 80–85. 21 Rasberry, S., ISO/REMCO, ISO, Geneva, 1995. 22 International Standards Organisation, Contents of Certificates of Reference Materials, ISO Guide 31, ISO, Geneva, 1981. 23 International Standards Organisation, International Organization for Standardization, ISO Guide 34, ISO, Geneva, 1996. 24 International Standards Organisation, Certification of Reference Materials—General and Statistical Principles, ISO Guide 35, ISO, Geneva, 1989. 25 Taylor, J. K., Handbook for SRM Users, ed. Trahey, N. M, United States Department of Commerce, National Institute of Standards and Metrology, Gaithersburg, MD, 1993. 26 Kristiansen, J., Christensen, J. M., Lillemark, L., Linde, S. A., Merry, J., Nyeland, B., and Petersen, O., Fresenius’ J. Anal. Chem., 1995, 352, 157. 27 Bureau International des Poids et Mesures, International Electrochemical Commission, International Federation of Clinical Chemistry, International Organization of Standardization, International Union of Pure and Applied Chemistry, International Union of Pure and Applied Physics and International Organization of Legal Metrology, Guide to the Expression of Uncertainty in Measurement, ISO, Geneva, 1993. 28 Kristiansen, J., Christensen, J. M., and Nielsen, J. L., Mikrochim. Acta, 1996, 123, 241. 29 Kristiansen, J., and Christensen, J. M., Ann. Clin. Biochem., in the press. 30 European cooperation for Accreditation of Laboratories, European Confederation of Laboratory Medicine, Publication EAL-G25, Accreditation for Medical Laboratories, EAL/ECLM, 1997. 31 International Standards Organisastion, General Requirements for the Competence of Testing and Calibration laboratories, ISO/IEC Guide 25, ISO, Geneva, 1990. 32 Proficiency Testing by Interlaboratory Intercomparisons—Part 1: Development and Operation of Proficiency Testing Schemes—Part 2: Selection and Use of Proficiency Testing Schemes by Laboratory Accreditation Bodies, ISO, Geneva, 1996. 33 Thompson, M., and Wood, R., Pure Appl. Chem., 1993, 65, 2123. 34 Uldall, A., EQAnews, 1996, 8, 3. Analyst, January 1998, Vol. 123 1135 Cornelis, R., Heinzow, B., Herber, R. F. M., Molin Christensen, J., Poulsen, O. M., Sabbioni, E., Templeton, D. M., Thomassen, Y., Vahter, M., and Vesterberg, O., Pure Appl. Chem., 1995, 67, 1575. 36 van de Leemput, P. J. H. A. M., Accred. Qual. Assur., 1997, 2, 263. 37 Good Automated Laboratory Practices, U.S. Environmental Protection Agency, North Carolina, 1995. 38 International Standards Organisation, Quality Systems—Model for Quality Assurance in Design, Development, Production, Installation and Servicing, ISO 9001, ISO, Geneva, 1994. 39 International Standards Organisation, Quality Systems—Model for Quality Assurance in Production, Installation and Servicing, ISO 9002, ISO, Geneva, 1994. 40 Taylor, A., Mikrochim. Acta, 1996, 123, 251. 41 Taylor, A., Patriarca, M., Menditto, A., and Morisi, G., Ann. Ist. Super Sanit`a, 1996, 32, 295. Paper 7/06358J Received September 1, 1997 Accepted October 29, 1997 12 Analyst, January 1998, Vol. 123