Same Blood, Different Verdicts: EQA Reveals Widespread Cortisol Interpretation Variation

Every working day in England, around 300,000 laboratory tests are performed, and 95 percent of clinical pathways depend on patients having access to pathology services. Yet a striking new study suggests that the number a laboratory reports, and even more importantly the meaning attached to that number, can vary dramatically depending on which analyser happens to sit in the lab and which protocol the duty biochemist follows. The research, led by Rachel Marrington and colleagues at Birmingham Quality, part of the UK NEQAS Consortium and University Hospitals Birmingham NHS Foundation Trust, and published in the British Journal of Biomedical Science, turned a routine external quality assessment exercise into a nationwide probe of how laboratories diagnose one of endocrinology’s most consequential conditions: adrenal insufficiency.

The stakes could hardly be higher. Cortisol, the glucocorticoid hormone produced by the adrenal glands, is essential to nearly every organ and tissue in the body. It drives the stress response, regulates the metabolism of carbohydrates, fats and proteins, modulates the immune system, and influences mood and cognition. When the adrenal glands fail to produce adequate cortisol, the consequences can be devastating and potentially fatal, particularly in an adrenal crisis. Diagnosing the condition reliably therefore matters enormously, and the stakes cut both ways: a missed diagnosis delays life-saving treatment, while a false one exposes patients to unnecessary anxiety and intervention.

In August 2024, the UK’s National Institute for Health and Care Excellence released guideline NG243, titled Adrenal insufficiency: identification and management, which offers a deceptively simple framework. For patients not taking oral oestrogens, an early morning serum cortisol measured between 8 am and 9 am is interpreted against three concentration bins. A result below 150 nmol/L is suspicious of adrenal insufficiency and should prompt an endocrinology referral; a result between 150 and 300 nmol/L leaves the probability uncertain and suggests repeating the test or seeking specialist advice; and a result above 300 nmol/L makes adrenal insufficiency very unlikely. But these cut-offs, the guideline notes in a rider, are only for use with modern immunoassays, a term the document never defines. That ambiguity proved fertile ground for the study.

The team harnessed the UK NEQAS for Steroid Hormones Scheme, which has provided an external quality assessment service for cortisol for more than 40 years. In December 2025, three serum specimens were dispatched to 387 participating laboratories, 82 percent in the UK and 18 percent overseas. Each specimen was presented as the same clinical scenario: an early morning cortisol request from a 25-year-old female with possible adrenal insufficiency. Critically, the specimens were unmanipulated human serum, pooled from no more than eight donations of the same blood group and sex, filtered to one micrometre and subjected to two freeze-thaw cycles. No exogenous cortisol was added, no dilutions made. The researchers argue the material is commutable with genuine patient samples, since filtering at that threshold will not remove lipoproteins and serum cortisol is known to remain stable across at least four freeze-thaw cycles.

The analytical phase of the exercise behaved exactly as expected. Roughly 370 participants returned numerical results, and the precision was reassuring, with coefficients of variation between 7.2 and 7.8 percent across the three specimens. The target values, based on the field method mass spectrometry mean of 19 results and validated against a candidate reference method, were 122 nmol/L for specimen 536A, 307 nmol/L for 536B and 260 nmol/L for 536C. Under NG243, these should map neatly onto three different interpretations: possible adrenal insufficiency, very unlikely, and uncertain, respectively. The physiology was controlled; the chemistry, it turned out, was not.

Bias between manufacturers remains a stubborn feature of cortisol measurement, as it has been for more than three decades. The Roche Cobas cortisol method, particularly the newer Cortisol II assay designed for improved specificity, tracked the mass spectrometry target closely across the 100 to 1100 nmol/L range with no appreciable difference between male and female specimens. By contrast, both the Siemens ADVIA Centaur and Siemens Atellica assays showed a much wider spread of biases and a clear difference between male and female matrices, indicating poorer precision and sex-dependent behaviour. The Abbott and Beckman assays fell in between, with wider spreads of bias than Roche that point to potential lot-to-lot variability. Siemens is currently implementing a Control System Improvement aimed at improving accuracy and lot-to-lot consistency, and the researchers hope improved EQA data will follow in the coming months.

The interpretative phase revealed the deeper problem. Around 300 laboratories, roughly 82 percent of those reporting numbers, also submitted an interpretation, choosing between three codes: the person may have adrenal insufficiency, the probability is uncertain, or adrenal insufficiency is very unlikely. For the specimen at 122 nmol/L, 68 percent correctly chose the first code. But for the specimen at 307 nmol/L, just above NICE’s 300 nmol/L threshold, responses split almost evenly between uncertain and very unlikely, and participants using six different methods reported code 1, meaning possible adrenal insufficiency, for a concentration the guideline says makes the diagnosis very unlikely. The same six methods also returned code 1 for the 260 nmol/L specimen. The Rainbow Trout plot, Birmingham Quality’s scatter plot of every result by method colour-coded by interpretation, showed disagreement not only between manufacturers but within them, with different colour codes appearing inside the same concentration bin of a single analyser platform. No clean concentration cut-off separated the interpretations, suggesting that different guidelines, protocols, or no protocols at all, are in play.

Compliance with the national guideline, even in England where NICE guidance applies directly, was sobering. Of the English laboratory responses, 56, or 29 percent, interpreted cortisol concentrations below 150 nmol/L as uncertain rather than suspicious, and five, or 3 percent, called them very unlikely. The authors acknowledge possible reasons: laboratories may have wanted more clinical context, such as whether the patient was pregnant or taking the oral contraceptive pill, which raises total cortisol through elevated oestrogens and renders normal cut-offs inapplicable, or they may follow different local protocols. Participant comments reflected this, with several noting that interpretation depends on request location, clinical details and other results, and that acutely unwell patients with hyponatraemia may warrant investigation even at higher cortisol levels. One laboratory had even seen adrenal crisis with cortisol above 400 nmol/L. Whatever the justification, the authors conclude that unwarranted variation exists, and that it ultimately threatens patient safety through delayed diagnosis, misdiagnosis, and the emotional and psychological damage both can cause, alongside a financial cost to the NHS that is rarely quantified.

The findings land in a familiar landscape of guideline-embedded magic numbers. NICE and SIGN guidelines are developed with healthcare professionals but do not always involve the laboratory or account for manufacturer and assay-related bias. The same tension surrounds the HbA1c threshold of 48 mmol/mol for diagnosing diabetes, the FIB-4 fibrosis score that assumes specific ALT and AST methods without saying so, and the renal pathways built on creatinine assays, where as of 2026 around 15 percent of UK laboratories still use the older Jaffe method despite well-documented differences from enzymatic approaches. The Endocrine Society’s guideline for adrenal insufficiency, which recommends the short synacthen test and a 500 nmol/L peak cortisol cut-off, does concede that the threshold is assay dependent, yet that nuance has not stopped legacy cut-offs, some derived from poor-specificity assays fifty years ago, from becoming engrained in clinical habit even as assays improve.

The study’s authors frame it as evidence that external quality assessment can be far more than a box-ticking route to ISO 15189 accreditation, and their post-analytical approach, which also underpins schemes for B12, ferritin, folate and vitamin D interpretation, deliberately satisfies ISO/IEC 17043 requirements for proficiency testing. They point to a 2017 editorial by J W Honour that charged EQA programmes with failing for decades to use powerful evidence of poor assay quality to pressure manufacturers into improving accuracy and specificity. Birmingham Quality and its counterparts, mostly NHS hospital departments without the legal infrastructure to confront multi-million-pound manufacturers directly, have documented these differences for years. The governance structure of UK EQA has been paused since 1 January 2026, and before that the Royal College of Pathologists was likewise not positioned to challenge manufacturers legally. What this exercise adds is a demonstration that the consequences of assay variation flow directly into clinical interpretation, and that the same sample, from the same fictional patient, can receive a reassuring verdict in one hospital and a red flag in another. The authors are not advocating that every laboratory follow NICE guidance; their point is simpler and more uncomfortable: the variation is real, it has been visible for at least thirty years, and until laboratories communicate assay limitations to clinicians and guideline writers build that variation into their cut-offs, algorithms and online calculators intended to help patients may instead quietly put them at risk.

Subject of Research: External quality assessment of cortisol assay variation and post-analytical interpretation in suspected adrenal insufficiency

Article Title: Use of EQA for assessing variation in post-analytical interpretation of results using a clinical case of ?adrenal insufficiency taking into consideration variation in analytical performance of cortisol

Article References: Marrington, R., Anderson, M., Roch, M., Davies, S., Forster, R., Williams, G., d’Oyen-Fitchett, J., & MacKenzie, F. (2026). Use of EQA for assessing variation in post-analytical interpretation of results using a clinical case of ?adrenal insufficiency taking into consideration variation in analytical performance of cortisol. British Journal of Biomedical Science, 83, Article 16559. https://doi.org/10.3389/bjbs.2026.16559

Image Credits: AI Generated

DOI: 10.3389/bjbs.2026.16559

Keywords: adrenal insufficiency, cortisol, external quality assessment, EQA, NICE guideline NG243, post-analytical interpretation, laboratory medicine, immunoassay, mass spectrometry, clinical biochemistry, patient safety, UK NEQAS