Calibration for a Country: Matching Diagnostic Validation to the UK's Population Mix

A new analyser arrives in a UK laboratory with a thick file of evidence, a certificate of conformity and a set of performance claims. The file will usually describe analytical precision, limit of detection, interference studies and method comparison against a predicate device. What it will rarely describe in adequate detail is the population from which the comparison samples were drawn. That omission is not a regulatory failure. It is a structural feature of how in-vitro diagnostic devices are approved, and it places a specific obligation on the laboratory that intends to use them.

The CE mark is a statement about conformity with regulatory requirements, not a statement that an assay has been calibrated for the population it will now serve. In a country as demographically varied as the UK, that distinction has practical consequences for haemoglobinopathy screening, renal function estimation and several tumour markers. The task for 2026 is not to distrust the certificate, but to surround it with local evidence.

What the Certificate Does and Does Not Say

Manufacturers design validation cohorts to support claims, not to represent every intended-use population. A study may include sufficient numbers of pregnant women, renal patients or neonates to satisfy a performance requirement while remaining thin on the ancestral groups that make up a substantial share of several UK cities. Where an analyte is affected by genetic variation, renal handling, body composition or co-morbidity, a cohort that is not stratified can produce average performance figures that conceal divergent behaviour at the margins.

This matters most at decision thresholds. A method comparison may show acceptable agreement across a wide range, yet the same assay can behave differently in the narrow band where a clinician acts. Laboratories are familiar with this idea in relation to analytical imprecision; they are less accustomed to applying it to population fit.

Three Assays Where Population Fit Matters

Haemoglobinopathy screening

The NHS Sickle Cell and Thalassaemia Screening Programme depends on the reliable separation and identification of variant haemoglobins. High-performance liquid chromatography and capillary electrophoresis both rely on retention or migration windows that were established in particular populations. A device validated predominantly in one ancestral group may perform well for HbS and HbC while leaving the detection of less common variants, or the interpretation of borderline windows, less well characterised in the groups most likely to carry them. Local validation should therefore include samples with confirmed variant haemoglobins drawn from the local screening population, and should examine whether the assay's windows remain discriminative when co-inherited conditions are present.

Renal function estimation

The 2021 adoption in England of a creatinine-based equation without an ethnicity coefficient was a deliberate attempt to remove one source of population-dependent bias from estimated glomerular filtration rate. It did not remove all of them. Creatinine generation varies with muscle mass, diet and body habitus, and cystatin C is influenced by adiposity, inflammation and thyroid status. An assay's calibration traceability matters here as much as its equation. Laboratories introducing a new creatinine or cystatin C method should verify agreement against the current local method across the range that spans the reporting thresholds for chronic kidney disease staging, and should inspect discordance by age, sex and body mass index rather than in aggregate.

Tumour markers

Reference limits for markers such as CA-125, prostate-specific antigen and alpha-fetoprotein are frequently derived from populations that may not reflect local case mix. Benign conditions that elevate a marker, such as endometriosis or liver disease, vary in prevalence between groups. A cut-off that performs acceptably in a manufacturer's cohort may yield different positive predictive value in a local population with a different spectrum of comorbidity. This is an argument for reviewing thresholds, not for abandoning them.

A Local Validation Protocol for 2026

A workable protocol need not be elaborate, but it should be explicit and documented.

1. Reconstruct the intended-use population. Use local laboratory data and public health profiles to describe the age, sex, ancestry and relevant clinical subgroups that the assay will actually encounter. Write this down before selecting samples.
2. Stratify the sample set. Select residual or prospectively collected samples so that the main subgroups are represented in numbers sufficient to detect meaningful discordance, not to prove equivalence. For most laboratories this means tens of samples per subgroup rather than hundreds.
3. Run a bias review against the manufacturer's data. Compare the demographic composition of the manufacturer's study with the local intended-use population. Where a group is absent or sparse, record that as a known limitation and plan targeted confirmation.
4. Re-check decision thresholds. Compare the new method with the current method or a reference method in the window around each clinical cut-off. Report the direction and magnitude of disagreement, and state whether any result would have changed clinical management.
5. Record the residual uncertainty. Validation should end with a documented statement of what has been demonstrated, what has not, and what monitoring will follow after go-live.

The Governance Question

ISO 15189:2022 requires validation of non-standard methods and verification of standard ones, and UKAS assessments already examine whether laboratories understand the limits of the evidence they hold. Population fit belongs within that framework. It is a verification activity, not a research project, and it should be signed off by the clinical lead for the service rather than left to the quality manager alone.

As the UK's device regulatory transition continues, many laboratories will keep receiving CE-marked equipment whose evidence base was assembled elsewhere. That is workable, provided the local file tells the truth about the local population. A certificate is a starting point for validation, not a substitute for it.

A Routine Obligation

The strongest argument for doing this work is that it is unremarkable. Laboratories already verify precision, linearity and interference. Adding a stratified sample review and a threshold check extends an existing habit rather than inventing a new one. Done consistently, it turns a regulatory formality into a defensible account of how an assay performs in the patients who will actually depend on it.

The certificate tells you the device met its requirements. The local validation tells you whether it meets yours.