According to the Royal College of Pathologists, 300,000 tests are performed every working day in England, with 95% of clinical pathways relying on patients having access to pathology services [1]. Laboratory data is used to support decision-making and ensure patient safety across many fields.

Laboratory tests are In Vitro Diagnostics (IVD) which are tests that analyse samples outside of the body, such as the general public might call “blood tests.” Test results are widely used for the diagnosis and monitoring of disease. Automation has significantly transformed laboratory testing over the last decade by streamlining processes at a time when workload has been rapidly increasing. The role of the diagnostic laboratory has always been more than merely the provision of an analytical result for a test that a clinician has requested. As well as ensuring the accuracy of the result, the laboratory provides service users information to help guide requesting (pre-analytical) and interpretation of the result (post-analytical).

Though a test result is mainly thought to impact patient management at that current point in time, results may be used as part of algorithmic tests in the future, for example the Acute Kidney Injury (AKI) algorithm which uses creatinine results over the last year to aid determination of the correct AKI stage [2]. Monitoring of disease requires the review of serial laboratory test measurements to look for clinically significant changes which can then be used to adjust treatment and provide individualised care. For this to be effective clinicians need to have confidence in the reliability of results. Clause 7.4.1.6 (f) of ISO 15189:2022 states that reports should identify the examination method used, where relevant, including, where possible and necessary, harmonised (electronic) identification of the measurand and measurement principle [3]. This is particularly important for the monitoring of tumour marker results especially when patients may have blood sent to different laboratories as part of their routine care.

Test results are not always about the person who is being tested, many patients are also part of clinical trials with the results impacting many more patients into the future. Clinical trials may be multicentre clinical trials with sample analysis undertaken at different laboratories which may use different analysers/methodologies. For meaningful results and conclusions there needs to be harmonisation of results between all sites. This has been recognised in a comparative study which recommended harmonisation efforts should be undertaken in multicentre trials to ensure accurate data analysis [4].

Storage of laboratory data in large datasets, on the “Cloud,” is allowing biomedical research to be upscaled and allows international collaboration. “Big data” allows integration of clinical and biomedical data from many sources. Harmonisation of results and stability of methods is crucial to allow this data to be used effectively. Without this, incorrect conclusions on data may be formed, or crucial discoveries may be missed.

The management and treatment of patients is routinely guided by local and national guidelines. The most commonly used in the UK include NICE guidelines which are evidence-based recommendations designed to improve health and social care in England and Wales with an aim to ensure a consistent approach to treatment and care, helping to reduce variation in healthcare quality [5]. The Scottish Intercollegiate Guidelines Network (SIGN) is used in Scotland [6]. NICE and SIGN work with healthcare professionals to develop guidance; however, this doesn’t always include the laboratory and doesn’t always take into consideration any manufacturer/assay related bias. Other society specific guidelines and protocols are also used which may not include manufacturer/assay related bias (likely to be due to lack of awareness) especially if looking at evidence from a small subset of laboratories that are using a single provider of analytical equipment.

It is common that guidelines involving pathology tests include “harsh” cut-offs either for classification of a patient, or as part of a diagnostic pathway. Examples of these so called “magic-numbers” that are commonly used include having a HbA1c of 48 mmol/mol for diagnosis of Diabetes as defined by the WHO in their report for the use of glycated haemoglobin (HbA1c) in diagnosis of diabetes mellitus [7]. Another example is the Fibrosis (FIB)-4 score, where although NICE recommends a single score of greater than 2.67 suggests advanced liver fibrosis; however, there is no consideration whether the ALT and AST methods used in the calculation should be those that the IFCC recommend and which do contain pyridoxal-5-phosphate or whether non-IFCC methods can be used [8]. There are significant differences between the IFCC recommended and non IFCC recommended assays for ALT and AST results in patients with liver disease which consequently does impact the FIB-4 result [9].

Most if not all renal pathways incorporate serum creatinine into either (i) an equation — estimated glomerular filtration rate (eGFR) and Kidney Failure Risk Equation (KFRE) for Chronic Kidney Disease (CKD) [10] — or (ii) or part of an algorithm — AKI [11]. It is well known that there are differences between the performance of the Jaffe creatinine and Enzymatic creatinine assay, and even differences exist between manufacturers within each method principle; however, as of 2026 ∼15% of the UK are still using the Jaffe creatinine assay (data from the UK NEQAS for Acute and Chronic Disease Scheme January 2026) [12].

“Adrenal insufficiency: identification and management” is the example that is used in this paper. NG243 covers identifying and managing adrenal insufficiency (hypoadrenalism) in babies, children, young people and adults [13]. Table 1 in NG243 provides cortisol cut-offs for the interpretation of serum cortisol levels from an 8 am–9 am test. There is a rider that the “cut-offs are only for use with modern immunoassays. Local guidelines may need to be followed if alternative assays are used.” There is no definition of what a modern immunoassay is. The majority of laboratories are likely to be correct in thinking that using an automated analyser is using a modern immunoassay rather than a manual solvent extraction method or even the historical fluorometric Mattingly assay [14]. NICE Guidelines do apply to the UK, but specifically England. Decisions on how NICE Guidelines are implemented in the devolved nations are up to relevant government ministers. It is not mandatory to apply the recommendations, and the guideline does not override the responsibility to make decisions appropriate to the circumstances of the individual.

There are no equivalent Scottish Intercollegiate Guidelines Network (SIGN) for adrenal insufficiency; however, the Scottish Paediatric Endocrine Group (SPEG) does provide guidance for assessment of adrenocortical function [15]. Interpretation is based on the cortisol concentration at 60 min (>430 nmol/L Abbott assay). We are not aware of specific adrenal insufficiency guidance for Wales and Northern Ireland; however, local protocols may exist.

Result interpretation for adrenal insufficiency has been covered by one similar exercise (Distribution 527 (March 2025) — which focused on short synacthen test interpretation in adult males. Target audits for short synacthen tests have also looked at variation in practice across the UK [16].

Section 7.3 of ISO 15189:2022 focuses on Examination processes and there is one key area which is important for how laboratory data is used in the context discussed thus far [3]. Section 7.3.7.4 looks at the comparability of examination results and ensuring that a process is in place within a given laboratory. The aim is to assess the comparability of results for examinations (analysis) that occurs on different methods or equipment, at different sites. If differences are observed the impact of these differences on biological reference intervals and clinical decision limits should be evaluated and acted upon with users informed of any clinically significant differences in the comparability of results. In reality this is probably only limited to different methods and/or different equipment and/or on multiple sites. However, service users including clinicians need to be aware of all limitations for all methods that could impact comparability and interpretation of results for any protocols and guidelines that they are working to, which is especially important if clinicians are using data as part of research to develop new protocols.

This is where EQA is of significant value. EQA is the retrospective assessment of the performance of an assay system usually undertaken by the analysis of physical clinical specimens. By careful scheme design a good EQA scheme can probe many areas of an assay system, not only the analytical performance of an assay. EQA specimens should be as close in matrix as possible to a patient sample which ensures that it mimics the characteristics of patient material when assayed. Laboratories like to compare themselves with other users of the same method; however, this is not good practice and is not always appropriate. The Target or Assigned Value should always be the “best estimate of the truth.” This may be from a Reference Method, or from a validated target value within the data in the Scheme. Though it is likely that an assay in an individual laboratory can only perform as well as other users of the method, this may not be good enough for the patient [17–23].

Clause 7.3.7.3 (d) of ISO 15189:2022 states that “The EQA programme(s) selected by the laboratory shall, to the extent possible: 1) have the effect of checking pre-examination, examination, and post-examination processes; 2) provide samples that mimic patient samples for clinically relevant challenges; fulfil ISO/IEC 17043 requirements [24].

EQA schemes are now available that either look at pre-analytical or post-analytical elements alone or as part of an analytical EQA Scheme. These pre- and post-analytical elements can also be accredited to ISO/IEC 17043:2023 and participants must participate in the return of results if they are claiming ISO 15189:2022 accreditation in such areas [24]. Birmingham Quality asks for the interpretation of analytical results in a number of EQA Schemes, for example interpretation of the numerical B12, Ferritin and Folate in the UK NEQAS for Haematinic Assays and Vitamin D in the UK NEQAS for Vitamin D. Other examples of post analytical EQA schemes include the use of the analytical result reported by the laboratory within equations, for example eGFR or FIB4 and within algorithm pathways, AKI.

The types of EQA schemes discussed above describe elements that are present at every distribution. Educational EQA schemes also have the flexibility to include ad-hoc scenarios or surveys where the information reported is not used for any scoring elements but can be used to gain an understanding of current laboratory practices to provide evidence to laboratories and other stakeholders. This is the type of EQA that is presented in the current paper.

Adrenal insufficiency occurs when the adrenal glands fail to produce adequate concentrations of primarily cortisol, and sometimes aldosterone. Cortisol is a glucocorticoid and is an essential hormone that affects almost every organ and tissue in the body. Cortisol is important as a stress response, to keep the body alert and ready to respond to danger. Cortisol also regulates the metabolism—how carbohydrates, fats and proteins are used; it modulates the immune system and plays a significant role in mood and cognitive function.

Cortisol is routinely measured by immunoassay and is available 24 h a day, 7 days a week by most, if not all laboratories. Cortisol is usually requested in cases of suspected adrenal insufficiency or cortisol excess (Cushings Disease). In both cases dynamic function tests, for example a short synacthen test (SST) (to stimulate the release of cortisol) or a dexamethasone suppression test (to suppress the release of cortisol) is performed with the cortisol results interpreted against local or standard guidelines.

Cortisol deficiency or insufficiency can have serious consequences and can be a medical emergency. Therefore, it is important to be able to reliably diagnose patients so that management plans can be put in place to allow them to live as normal a life as possible.

In August 2024, NICE released Guideline NG243 — “Adrenal insufficiency: identification and management” which provided guidance on interpretation of a single cortisol result in the context of suspected adrenal insufficiency [13].

The UK NEQAS for Steroid Hormones Scheme has provided an EQA service for cortisol for over 40 years. Serum specimens are provided for the analysis of only cortisol with the distribution of three specimens, notionally monthly to participants in the UK and worldwide. This Scheme design allows predominantly endogenous concentrations of cortisol to be distributed to laboratories. Hydrocortisone is added to some specimens to assess assay recovery and in some cases other steroids are added as part of interference studies, for example Prednisolone [25].

The aim of the current study was to look at the impact of both within- and between manufacturer/analyser variation when interpreting a single early morning serum cortisol against NICE guidelines — in this case NICE Guideline NG243 — “Adrenal insufficiency: identification and management” [13]. The EQA Scheme has been used to probe both the analytical and post-analytical aspects of this process based on a provided scenario with participants analysing three specimens designed to mimic potential adrenal insufficiency. All specimens contained endogenous concentrations of cortisol. The only manipulation was the pooling of donations to obtain the volume required, filtering of serum and two freeze-thaw cycles (preparation of specimens and distribution of specimens). Formal commutability studies have not been undertaken; however, we would assert that the material is commutable with patient material because filtering to 1 µm will not remove lipoproteins and serum cortisol has been shown to be very stable during at least four freeze thaw cycles [26]. There were no additions of exogenous cortisol, or dilutions made.

Three pools (C663, C668 and C667) of serum were prepared using serum sourced as off-the-clot serum from NHS Blood and Transplant, Filton, UK. Serum was received frozen and used as “pooled material” with serum donations from donors with the same blood group and sex of donor being mixed and then filtered to 1 μm, using a glass fibre filter paper, Merck, Gillingham, UK. The maximum number of donations in each pool was eight. No exogenous material was added.

All three pools had previously been distributed through the EQA Scheme and the field method mass spectrometry mean was used to guide selection for Distribution 536 to align with required cut-offs for the proposed clinical scenario.

Each pool was pipetted into labelled 2 mL Sarstedt tubes (Nümbrecht, Germany) as 0.5 mL aliquots and stored frozen until dispatch where they were then transported at room temperature (the majority of specimens arrive within 24 h) 387 participants (82% UK and 18% overseas). Participants are advised to analyse immediately as if they were from a patient.

Data is also shown from other cortisol EQA specimens over the last year. All specimens are prepared according to the same protocol of using either single donations, or pooling together a number of donations, from the same sex donor, with the only exception being the addition of exogenous hydrocortisone (Merck Life Science Ltd., Gillingham, UK) as part of recovery experiments. These specimens are clearly identified to the participant as having been manipulated. No other additions are made unless intentional as part of an interference study — this data has not been included.

Data has been classified on whether the material is from a male or female donor and whether used endogenously (endog) or whether cortisol has been added (exog).

A clinical scenario was provided with all three specimens:

The three Serum Cortisol specimens at Distribution 536 (A, B and C) comprise 3 different early morning (08:30 am) serum cortisol requests with a scenario ?Adrenal insufficiency.

You can assume the patients are 25-year-old females.

Please select the interpretation code from the dropdown:

Code 1 is Person may have adrenal insufficiency

For each specimen, participants reported back online (manually or automatically), to the Scheme at Birmingham Quality, the cortisol concentration measured. Participants are also able to check the method that they are using and can change if required to ensure that they are registered correctly.

Participants were also asked to interpret their cortisol result based on the clinical scenario and could choose one of 5 options from a drop-down menu.

Outliers in EQA schemes offered by Birmingham Quality are removed by Healy trimming with is applied to all data [27]. This essentially leads to the highest and lowest 5% of specimen level data being removed from statistical calculations of means and %CVS. A method mean, standard deviation and coefficient of variation is calculated for each manufacturer/analyser combination. The target value for every laboratory irrespective of the method that they are using is the field method mass spectrometry mean, which is calculated as the method mean of mass spectrometry users (which for Distribution 536 was based on 19 individual results). The field method mass spectrometry mean is routinely used as the target value for cortisol and has been validated by comparison of a number of pools with a candidate reference method for cortisol [28].

Each participant interpreted their measured cortisol result in terms of the clinical scenario provided. Data is presented for each specimen as (i) Rainbow Trout plot — a scatter plot for each method, coloured by interpretation, (ii) Pie chart — based on all reported cortisol interpretations and (iii) Bar chart — split by method and presented both as number of reported results and percentage of each method. The target value for each interpretation is based on the interpretation of the field method mass spectrometry cortisol mean in relation to the cut-offs provided within NICE Guideline NG243 — “Adrenal insufficiency: identification and management” [13].

Data has also been reviewed and is presented for each of the four devolved nations within the UK.

In December 2025 results, cortisol results were returned by ∼370 participants in the UK NEQAS for Steroid Hormones Scheme (Distribution 536). The breakdown by manufacturer/analyser is shown in Figure 1 with the Roche Cobas and the Abbott Alinity being the two most used methods. Figure 1 shows histograms and associated tabulated data for the three specimens at Distribution 536. The target value is the field method mass spectrometry mean which for Specimen 536A is 122 nmol/L (SD 9 nmol/L, %CV 7.5%), Specimen 536B is 307 nmol/L (SD 24 nmol/L, %CV 7.8%) and Specimen 536C is 260 nmol/L (SD 19 nmol/L, %CV 7.2%). These three specimens have behaved exactly as expected at this point in time and there are no concerns regarding specimen homogeneity or stability, as seen by the %CVs (∼10%) which are similar to distributions of previously distributed material at similar concentrations (Figure 2). Stability and homogeneity are checked on every specimen, at every distribution. Degradation of a specimen during the preparation stage, or whilst in transit, would show a wider spread of reported results which consequently would impact the standard deviation and corresponding %CV.

Commutability with a patient sample has not been proven, but there is nothing to suggest that using off-the clot serum with no manipulation other than the pooling of a number of donations, filtering to 1 µm and two freeze thaw cycles would cause commutability issues. Single donations would be preferred; however, this does not allow for repeat testing which is a valuable tool in the assessment of whether the performance of a method has changed. There are different relative method biases across the three specimens, at different cortisol concentrations; however, these are as expected and are consistent with what has previously been seen — see Figure 3.

Figure 3 shows the individual specimen % biases compared to the target value, the field method mass spectrometry mean for all data for the major manufacturers between January and December 2025. The Roche Cobas cortisol method is minimally biased with respect to the target value and there appears to be no difference in bias between male and female specimens. In comparison, both the Siemens ADVIA Centaur and Siemens Atellica cortisol assays show a much wider spread of biases and there is a clear difference between male and female matrices.

Participants were asked to interpret their measured cortisol result with the scenario that the sample had been taken from a 25-year-old female at 08:30 am. The scenario did not make reference to guidelines. This was intentional because the aim was to assess how laboratories interpret based on their usual procedures, not to report a response based on available guidelines. Interpretations were returned by ∼300 participants (82% of the total number of participants reporting numerical cortisol results). Figures 4, 5 show the same cortisol interpretation data but presented in a different way. Figure 4 is the Birmingham Quality trademark Rainbow Trout plot (scatter plot) which shows each individual cortisol result (y-axis) by method (x-axis) colour coded based on the interpretation provided by the Laboratory itself (see legend). This data presentation shows quite clearly the spread of results, by method, and that some method/manufacturers, for example Roche (BO5) for Specimen 536A has a smaller spread of user results than for example Beckman (SF1), for the same specimen. Also, the Siemens Atellica (SM20) not only has a similar spread of results to Beckman, but nearly all results are greater in value than Roche. Data is presented as a scatterplot where up to 8 identical values are plotted on a single row then on another row above for a further 8. This visualisation nicely shows that not only is there variation in interpretation between manufacturers, but also within a manufacturer. This can be seen by the different colour filled circles within the same concentration bin.

Figure 5 shows a summary of all data, in a pie chart, for all three specimens broken down by interpretation. Data is only included for categories 1, 2 and 3 because those are the only codes relevant for this exercise. The bar charts below the pie charts break down interpretation by method/manufacturer. The bottom bar chart is the same data as the top bar chart normalised as a percentage. For Specimen 536A, which had a cortisol of 122 nmol/L, 68% of responses were Code 1 with a significant proportion reported by all manufacturers. The mode was taken as the Target Value. However, Specimen 536B, which had a cortisol of 307 nmol/L had a more even split of interpretations (between code 2 and code 3). For Specimen 536B the cortisol concentration is known to be 307 nmol/L based on the field method mass spectrometry mean and has been interpreted based on Nice Guideline NG243. This cortisol concentration corresponds to code 3, which is what has been set as the Target Value, not the mode, which was code 2. A number of participants, from six methods, reported Specimen 536B as code 1. The majority of interpretations were split between codes 2 and 3. For Specimen 536C, which had a cortisol of 260 nmol/L, 67% of responses were code 2. In this case the mode was the assigned Target Value. Interestingly, participants from the same six methods as Specimen 536B, reported a Code 1 for the interpretation of Specimen 536C.

Data has been broken down by each laboratory in each devolved nation within the UK (Figure 6). Pie charts are shown for each specimen, for each data set, as previously discussed. Data has been further broken down by cortisol concentration bin <150 nmol/L, 150–300 nmol/L and >300 nmol/L which corresponds to the cortisol concentration bins used in Nice Guideline NG243. There are no obvious trends within a region or for a specific specimen.

Participants are able to add comments to all their EQA returns and a number were made specifically for cortisol at this distribution. A number are detailed in Table 1 and relate to two themes (i) requirement for further information and (ii) comments applied.

This study required participants to analyse and interpret cortisol result as part of their routine EQA. Neither element is mandatory. Conclusions have been formed on the best data set that we have available; however, we know that because this is EQA, and not a patient sample, it may have been different personnel who were interpreting results than those who would a patient sample.

Cortisol data has been linked to manufacturer/method information. We try to ensure that it is correct; however, we are completely reliant on the Participant informing us of any changes to the system that they are using.

EQA has always shown variation in cortisol assays — concentration dependent and sex related biases

Post-analytical EQA has been used to show the impact of the variation in cortisol assays across different manufacturers