Clinical Validation of Biomarkers: What the Evidence Must Show

A claim that a biomarker is “clinically validated” should identify its purpose, the patients studied and the evidence behind it. Biomarkers are measurable characteristics indicating a biological process, disease process or response to an intervention. Validation establishes what researchers can reasonably conclude from that measurement, whether they are investigating diagnosis, disease progression or treatment benefit.

Three different kinds of evidence

Analytical validity concerns whether an assay measures the marker accurately and consistently under specified conditions. Clinical validity concerns whether the measurement reliably relates to the clinical condition or outcome of interest. Clinical utility asks whether using the result to guide care brings benefit.

The clinical validation of biomarkers connects laboratory findings to a defined clinical application. A dependable assay provides the foundation; studies involving relevant patients evaluate its clinical performance.

Archived blood or tissue can allow researchers to evaluate a marker against known clinical outcomes. For these studies, decisions about Samples Storage should reflect the intended assay’s requirements. Reliable linked clinical records and appropriate permissions for reuse are also needed.

Does the study match the intended use?

The intended use should name the patient population, setting and endpoint. Detecting disease in symptomatic patients is a different task from estimating future progression or identifying differences in treatment benefit. Each purpose requires evidence tailored to that question.

Consider a hypothetical diagnostic marker assessed using samples from people with advanced disease and healthy volunteers. It might distinguish those groups, yet perform less well among patients whose similar symptoms arise from different conditions. Those are the patients a clinician needs to distinguish.

Age, disease stage, other conditions and treatment history can also affect how well findings transfer to practice.

For archived-specimen studies, examine who had material available and who was excluded. A convenient subset may differ from the wider patient population. The study also needs an appropriate reference standard to establish clinical status separately from the marker being evaluated.

Test beyond the discovery dataset

Discovery data help researchers select promising markers and develop models. Assessing performance only in those same data risks rewarding patterns specific to the original study.

Suppose researchers try several cut-offs and choose whichever best separates their groups. Measuring that cut-off’s performance in the same patients can produce an optimistic estimate. The threshold, model and analysis plan should be fixed before independent evaluation.

Independent testing examines whether the finding holds beyond the development data. Internal cross-validation helps assess performance within the development dataset; external validation tests the selected marker or model in a separate cohort. That cohort still needs to reflect the intended patient population.

Performance measures must suit the claim. For a diagnostic marker, sensitivity is the proportion of people with the condition correctly identified; specificity is the proportion without it correctly classified as negative. Predictive values describe how likely positive or negative results are to be correct. They depend partly on disease prevalence in the population being tested.

Confidence intervals indicate uncertainty around performance estimates, giving readers more information than a headline figure alone. Blinding laboratory assessment to clinical status, where appropriate, also helps prevent knowledge of the outcome from influencing measurements or interpretation.

Does using the result improve care?

Clinical utility asks what happens when clinicians act on the result. A marker’s clinical validity and the benefit of using it are separate evidence questions.

For a diagnostic test, false positives may prompt unnecessary investigations, while false negatives may delay appropriate care. Evaluation should therefore consider the actions triggered by the result and their consequences, compared with existing practice.

Association with an outcome and prediction of differences in treatment benefit also require distinct evidence. To use a treatment-related change in a marker as a substitute for improved symptoms, functioning or survival, researchers need evidence that it reliably captures the treatment effect on those outcomes in that context.

Reading a validation claim

When assessing a claim that a biomarker is clinically validated, check:

• Its precise intended purpose and clinical setting.
• The patients included, comparison groups and relevant exclusions.
• The clinical endpoint or reference standard.
• Whether thresholds and models were fixed before independent evaluation.
• Performance estimates, their uncertainty and relevance to practice.
• Whether the evidence establishes clinical validity, clinical utility or both.

A credible validation claim is specific: this marker, for this purpose, in these patients, supported by this evidence.