Implementation of metrological traceability in laboratory medicine: where we are and what is missing

Background

Very often, only a brief description of metrological traceability is provided with commercial calibrators. The information is frequently limited to the name of the higher-order reference material (RM) and/or reference measurement procedure (RMP) to which the assay calibration is traceable, without any description of implementation steps. Information such as the applied calibration hierarchy, the measurement uncertainty (MU) associated with the calibrator, and the employed acceptable uncertainty limits is often partly available [1]. On the other hand, accumulated experience shows that standardization projects not only have to address metrological traceability but should also consider the efficacy of its implementation [2]. Previous analyses highlighted how strongly MU may be dependent on the type of traceability chain adopted by the in vitro diagnostics (IVD) manufacturers to implement the traceability of commercial calibrators [1], [3], [4]. The selection of different types of traceability chains (no matter if they all employ certified RMs and approved RMPs) may lead to different combined uncertainties at the level of commercial calibrators, not always permitting to fulfill the suitable uncertainty budget at the level of clinical sample measurements [1], [3], [4], [5]. Therefore, to aid IVD manufacturers in the traceability implementation, the identification and definition of available reference measurement systems (RMS) and of metrological traceability chains in their entirety and not just in their main components (i.e. RMs and RMPs) can be extremely helpful. The Joint Committee on Traceability in Laboratory Medicine (JCTLM) recently created the Task Force on Reference Measurement System Implementation (TF-RMSI) with the aim to provide guidance on traceability implementation to the IVD community. The objectives of the Task Force are: (a) to identify and describe available RMS and traceability chains in their entirety, based on the information present in the JCTLM database; (b) to illustrate the evolution of MU through the entire metrological traceability chains; and (c) to identify those measurands for which further advancements to existing RMS are needed or some RMS components are lacking. Here, using creatinine as an example, we illustrate details of the current situation of what is listed in the JCTLM database to enable metrological traceability of this measurement, highlighting merits and some limitations.

Data extraction from the JCTLM database

We accessed the JCTLM database (https://www.bipm.org/jctlm/) on July 1, 2019, searching for RMs and RMPs for creatinine. In addition to two primary RMs (National Institute of Standards and Technology [NIST] SRM 914a and National Metrology Institute of Japan CRM6005-a), four secondary (matrixed) RMs and three reference method principles were identified (Table 1). According to the metrological traceability theory, to transfer trueness from higher-order references to commercial calibrators, IVD manufacturers have two possibilities: (a) directly calibrating their internal procedure with a suitable secondary RM or (b) aligning their internal procedure to an RMP by a comparison study [6].

Fulfillment of goals for suitable uncertainty

In addition to the correct transfer of trueness from higher-order references to IVD calibrators, we would like to emphasize that the definition of appropriate analytical performance specifications (APS) for the total uncertainty budget that should be fulfilled at the level of patient results is essential to ensure that laboratory measurements are clinically usable [3], [10]. For MU we believe that the relevant goal that should be fulfilled is that related to the allowable random variability of patient results, as the correct trueness transfer along the metrological traceability chain should allow the achievement of unbiased (or negligibly biased) results [11].

According to concepts for deriving APS established at its 2014 Strategic Conference [12], the European Federation of Clinical Chemistry and Laboratory Medicine has worked on the criteria for assigning laboratory measurands to different models [13]. Serum creatinine was assigned to the biological variation-based model that should be applied when the measurand has stable concentrations, i.e. is in a “steady state” status when a subject is in good health, and deviations from these concentrations will not in itself cause symptoms [14]. By using published data about the intra-individual biological variation (CV_I) of serum creatinine (4.4%) [15] and the classical formulae for deriving APS for random variability [11], APS for standard MU of serum creatinine measurement on clinical samples are 3.3% (≤0.75CV_I, minimum quality), 2.2% (≤0.50CV_I, desirable quality), and 1.1% (≤0.25CV_I, optimum quality).

The higher-order references represent the first contribution to the overall MU budget and, therefore, the MU of references may significantly affect the MU of the patient’s results. Due to error propagation in the calibration hierarchy, it is intuitive that this contribution in terms of MU should be sufficiently small to allow to fulfil APS for MU at the clinical sample level, when IVD calibrator and random uncertainties have been added [3], [10]. Accordingly, it has been recommended to turn the approach upside down by focusing first on the established APS of the field measurement results and then to defining the goal for MU of the RM for a given measurand by the performance needs of the clinical assays for that measurand [5]. We conventionally recommended that no more than one third of the total uncertainty budget should be consumed by higher-order references [3]. Therefore, using APS for standard MU of serum creatinine measurement on clinical samples mentioned above, the standard MU associated to higher-order references should be ≤1.10%, ≤0.73%, and ≤0.37% at minimum, desirable, and optimum quality levels, respectively.

Two out of four secondary RMs listed in the JCTLM database would allow at least the minimum quality level for APS related to MU to be fulfilled (Table 1). As expected, fulfillment of minimum or desirable APS was also dependent on the creatinine concentration levels in the same type of RM, the MU being higher when the measurand concentration becomes lower. The LGC ERM-DA materials display relative MU far greater (an order of magnitude) than other materials, and while certificates of the materials indicate that the uncertainty contribution added to account for long-term stability of the material is by far the largest component, the uncertainty from characterization is not negligible [16]. It is not generally expected that such a large difference in uncertainties should be evident for very similar materials, and as such general conclusions should not be drawn based on these materials. Regarding the listed RMPs, the MU associated with ID/GC/MS and ID/LC/MS should allow APS for clinical measurements to be fulfilled, while ID/SERS displays higher MU that would make it difficult for IVD manufacturers using this RMP as a higher-order reference to produce measuring systems able to fulfill MU APS at the level of clinical samples.

Conclusions

In this brief commentary, we focused on an example of what the JCTLM database is offering in terms of practical help to IVD manufacturers for implementing traceability to higher-order references and to end-users in terms of MU of different options for their clinical use. Using creatinine as the example, the overall characteristics reported in Table 1 show that the most recently listed RM, LNE CRM Bio 101a, which was developed and reviewed against ISO 15194:2009 requirements, appears to be suitable for correctly implementing traceability to SI (with commutability explicitly assessed) and has suitably small MU, considering the impact of the error propagation in the calibration hierarchy, to allow APS to be met for measurements on patient samples. Splitting clinical samples with a laboratory performing ID/GC/MS or ID/LC/MS provides an alternative route to establishing a calibration hierarchy, but also requires all components of uncertainty to be assessed. Using APS to judge whether reference measurement system components are fit for purpose certainly represents a novel approach that should be more widely tested. In our paper, it was only applied for creatinine, so the conclusions about the general status of the traceability implementation could be limited. The JCTLM TF-RMSI is planning a review of a greater number of measurands, chosen based on their frequency of measurement in clinical laboratories, for providing more robust information about the state of the art of available reference measurement systems and their impact on the ability of clinical measurements to meet APS.