Medical Device Link.

This article provides an overview of the development of the traceability concept, concluding with a historical and analytical discussion of the problems that once hampered the performance quality of clinical diagnostic tests and of the promise that the global adoption of reference systems grounded in credentialed traceability holds for dramatically improved measurement accuracy and consistency.

Traceability of a value attributed to a routine sample, a calibrator, or a control material is established by a series of comparative measurements using measurement procedures and reference materials in a chain of decreasing hierarchical order, as shown in Figure 2. Since each link in the traceability chain contributes to the uncertainty of the result, it is advisable to omit as many steps as possible. In metrological terms it would be ideal to omit all in-between steps of the traceability chain and to measure the routine sample directly by use of a primary reference procedure. This, of course, is not feasible.

The complete traceability chain is valid only for those measurable quantities that can have a value expressed in SI units (units traceable to the Système International d'Unités). When primary or secondary calibrators are not available, the traceability chain for many measurands in laboratory medicine ends at a lower level, for example, at the manufacturer's selected measurement procedure. In a case where a manufacturer detects a new diagnostic marker and defines the measurable quantity by establishing a measurement procedure for the marker, the manufacturer's measurement procedure will form the top of the traceability chain. Nevertheless, even in this simple situation the principles of traceability are applicable.

An inevitable precondition for establishing results traceable to calibrators and control materials is the specificity of the measurement procedures applied. Results of measurement cannot be traceable when the procedure applied partially detects components that are not consistent with the definition of the measurand.

Traceability is not really a new concept for in vitro diagnostics. Many years before traceability was mentioned in connection with general chemical metrology, reference measurement procedures and reference materials had been established in clinical chemistry. Early developments in this field in the United States, particularly those reported by scientists at the National Institute of Standards and Technology (NIST) and other independent scientists, as well as the relevant standards generated by the National Committee for Clinical Laboratory Standards probably had an important influence on the development of the concept of traceability in general chemical metrology. Some basic experimental work toward the development of reference measurement procedures and reference materials had already been undertaken in Europe.

The process of credentialing the traceability concept, its implementation, and its acceptance is not only a regional (European) or national task but also a global one. Traceability is a goal that concerns all members of the scientific community involved in the field of clinical chemical analyses, including:

For reference materials, it may not be too difficult to solve the problem. Materials that fulfill the requirement for higher-metrological-order standards are now provided by NIST and the Institute for Reference Materials and Measurements (IRMM). Additional useful materials, although usually of lower metrological order, are issued by the World Health Organization. Authorization of these materials is a question of mutual acceptance, which, in view of the "Implementing Arrangement for Cooperation in the Fields of Metrology and Measurement Standards" signed by the directors of NIST and the EU Research Directorate, has almost been achieved.

The situation regarding authorization of reference procedures and reference laboratories is somewhat more complicated, and the question of how the concept of traceability should be implemented arises. There is no simple answer, and probably there is no rule that can be applied to all situations. The new international standards dealing with traceability (prEN ISO 17511) and the requirements for reference laboratories (ISO DIS 15195), developed by CEN TC 140 and ISO TC 212, respectively, can at least give some guidance concerning the credentialing process for reference laboratories and reference procedures.^3,4

In fact, the strategy for establishing reference systems depends on the nature of the analyte.

For low-molecular-weight substances–electrolytes, organic substrates, and metabolites like cholesterol, creatinine, or steroid hormones–as well as for many drugs, the only meaningful solution is to aspire to results that are traceable to SI units. With respect to these measurands, where traceability to SI units is achievable, national metrology institutes as legal custodians of units and measurements hold the highest authority (see Figure 1 and Figure 2). The metrology institutes are connected with each other globally via the CCQM key comparisons (named for the Consultative Committee of Amount of Substance), through which they demonstrate, by means of ring trials, their ability to perform measurements of the highest available metrological level, most often by using so-called primary methods. Such key comparisons have been carried out for cholesterol, creatinine, and glucose. A group of European metrology institutes has now started a global initiative to address the problem of traceability in clinical chemical measurements in collaboration with NIST and the metrology institutes in Australia and Japan.

Figure 1. Hierarchy of laboratories.

Figure 2. Calibration hierarchy and traceability to SI in the metrological traceability chain.

In addition to the national metrology institutes there exist a number of highly specialized reference laboratories, most of which are situated in university hospitals or at manufacturers' sites. These laboratories usually have developed their own reference procedures. Some of them have long experience and perform measurements at a high metrological level.

According to the ISO standard on reference laboratories in laboratory medicine (ISO DIS 15195), their competence may be approved by the national metrology institutes, for example, by accreditation in accordance with ISO 17025. The German metrology institute Physikalisch-Technische Bundesanstalt has so far accredited two reference laboratories. One is a laboratory that mainly serves the Reference Institute of Bioanalysis, the proficiency testing organization of the German Society of Clinical Chemistry (DGKC), and also establishes target values for calibrators and control materials of commercial diagnostic kits. The second laboratory belongs to an industrial diagnostic kit manufacturer. Further accreditations will follow in due course.

The competence of reference laboratories with respect to environmental, staff, and management performance may be approved by accreditation, but this still leaves a question often asked: When is a "proposed" reference method a reference method? The beginning of the answer is that the competence of a reference laboratory should be evaluated not only according to the quality of its management as laid down in a quality manual and monitored by regular inspections of the laboratory, but also on the basis of documented reference procedures and–most importantly–on the results of parallel comparative measurements. In this way the accrediting body also approves the measurement procedures and their performance. Accreditation thus is not valid for all measurements produced by the laboratory but only for particular measurands for which agreement of results with those of the laboratory of the accrediting body, and thereby traceability to the SI, has been demonstrated in comparative measurements.

Accreditation of a reference laboratory for a particular measurable quantity carries with it an approval of the measurement procedure, which includes the measurement principle (such as IDMS as a primary method for the measurement of cholesterol), the complete standard operating procedure (SOP), and the uncertainty of results. This may serve as an answer to the question of when a proposed reference method is a reference method.

The concept of traceability has been promoted in Europe during the past 15 years by the organizers of external quality assessment schemes. In the German proficiency testing system in particular, the use of reference measurement procedures for several measurands has been prescribed by legislation since 1988. Consequently, the Reference Institute of Bioanalysis has established reference measurement procedures for electrolytes, metabolites and substrates, enzymes, hormones, and drugs (see Table I). Reference methods for 13 of the 30 analytes listed in the table have been developed in the reference laboratories of the DGKC using the analytical principle of the IDMS primary method. These include creatinine, urea, cholesterol, total glycerol, uric acid, and glucose, as well as the steroid hormones and thyroxine. The reference methods are now applied regularly in setting up target values in the control samples of internal and external quality assessments and in certifying matrix reference materials of the IRMM.

The system is shared with partners in Portugal, the Czech Republic, and, occasionally, Denmark by means of an exchange of samples for external quality assessment with reference procedure target values.

Variability of Results is a Problem. A list of routine method target values for creatinine, uric acid, total cholesterol, and total glycerol in the control material of one manufacturer that was issued before 1988 shows a large scatter of up to 30% among methods and test kits (see Table II). The fact that only one value for creatinine concentration in serum can be the "true" one made this situation particularly untenable. Obviously, any progress toward improving the comparability of analytical results from different laboratories is hindered as long as methods with a known–or even unknown–bias are accepted.

This unsatisfactory situation became apparent in external quality assessment as well, as evidenced by results from a ring trial. Two different samples were distributed in a 1987 DGKC routine ring trial for cholesterol to about 1300 laboratories, and the results obtained were displayed in a Youden diagram (see Figure 3). Each dot in the diagram represents the two results from one laboratory, that for sample A being read from the abscissa and that for sample B from the ordinate. A laboratory whose performance dot falls in the middle of the screen is in full agreement with the target value, which here is the reference method value certified by isotope dilution mass spectrometry.

Figure 3. A Youden diagram of a collaborative survey for cholesterol conducted in 1987. The three broken-line squares show the method-dependent evaluation limits for the Liebermann-Burchard method (upper right), the CHOD-PAP methods (middle), and the CHOD-iodide method (lower left). The solid-line square in the center shows the acceptance limits based on the target of the IDMS reference method value.

Results from the survey clearly show that three different groups of data have been reported, representing three different methods of cholesterol determination. Survey participants reporting relatively high cholesterol results had used the Liebermann-Burchard procedure, which was still in use in 1987. The group obtaining low cholesterol values had applied the cholesterol oxidase iodide (CHOD-iodide) method. The data from laboratories using the CHOD-PAP method are situated in the middle of the screen. Until 1988, ring trial participants' results were evaluated by comparison with the mean of each peer group, according to the different methodological principles used. Differences up to 50% among the peer group target values could be observed for cholesterol measurements. In view of the fact that there can be only one true cholesterol concentration value in a serum, this situation, too, was untenable.

After introducing reference procedure values for cholesterol that are based on IDMS measurements, the divergent peer group target values have now been replaced by reference method values that, in the case depicted, are represented as the exact middle of the screen. The corresponding limits of acceptance are shown as the solid-line square. As a consequence, methods displaying inherent systematic error like the Liebermann-Burchard and the CHOD-iodide disappeared from the market, so that today only methods that are within the limits of acceptance with the reference method values established by IDMS exist.

Twelve years ago there was unacceptably wide scatter in method-dependent target values for many clinical chemical parameters. To improve accuracy in clinical chemistry it was essential to replace these method-dependent values with reference method values.

The measurement of hormone concentrations in human body fluids has proved to be a valuable diagnostic tool in the field of clinical endocrinology. Thyroxine and the various steroids, the most commonly determined hormones, are usually measured by radioimmunoassay or enzyme immunoassay with a fairly high degree of sensitivity. However, a manufacturer's list, issued in 1987, of aldosterone, cortisol, progesterone, and estradiol-17ß target concentrations in a commercial serum pool indicates that, given the same sample and using immunoassay, assigned values varied considerably from one test kit to another (see Table III). For cortisol and aldosterone the range of results was between 100% and 200%, and for progesterone and estradiol-17ß determinations the results differed by a factor of 7. This was probably due to variations in quality among the antibodies and reagents used in the several commercial kits. What could a consensus value mean in such a context? A target value based on a consensus mean or median was of little use in judging test kits that gave such variable results.

Using method-dependent assigned values for external quality control means having many different target values for the same analyte in the same control serum–a highly impractical and, from a theoretical standpoint, unsatisfactory procedure that generates different results for a substance of known molecular weight and with a defined number of molecules.

Here is another illustration of the utility of having target values based on reference methods. In the early years of performing external quality control, the accuracy of unconjugated estriol in serum proved to be astonishingly high. This changed dramatically toward the end of 1981 (see Figure 6). Especially when the control samples contained conjugated estriol, the ranges of collaborative survey participants' results were significantly higher than the mass spectrometric target values.

Figure 6. Results of collaborative surveys for estriol. Columns indicate the range of distribution between the 16th and 84th percentiles of each survey.