Analytical Validity, Clinical Validity and Clinical Utility

Three questions, three different answers

A test can be technically excellent and still be useless to you. Evaluators separate the evidence into links in a chain. The U.S. EGAPP working group, set up to review genetic and other genomic tests, describes the chain as starting with the technical performance of the test (analytic validity), then the strength of the link between the result and the condition (clinical validity), and finally the evidence that results can change patient management and improve net health outcomes (clinical utility). [1] EGAPP adopted aspects of the earlier ACCE model, which uses the same ideas. [1] The framework was built for genomic tests; we use it here as a general way of thinking, which is our own application and not a rule stated by that group.

The three links in plain words

Why the chain matters

• Evidence for one link is not evidence for the next. EGAPP notes that direct evidence on a test's overall effectiveness is rarely available, so reviewers build a chain of evidence from the separate links. [1] A company that proves its assay is precise has shown the first link only.
• Accuracy is not benefit. The same logic applies to wearables and home kits. See Device vs Clinical Reference: How Validation Works for devices.
• Clinical validity depends on who is tested. Predictive values change with how common the condition is. In one worked example, a test with a 95% positive predictive value in a 50%-prevalence study sample had a 15% positive predictive value at 1% prevalence, with the same sensitivity and specificity. [3] See False Positives, False Negatives and Test Accuracy.
• A biomarker is not an outcome. Moving a number in a good direction does not prove you feel or live better; see Biomarker Changes vs Meaningful Health Outcomes.

Questions to ask about any test

1. What exactly is measured, and in what sample?
2. Who validated it, in what kind of people, against what reference method? (analytical and clinical validity)
3. What happens to people who are told "abnormal" and act on it, compared with people who are not tested? (utility)
4. What could go wrong: false alarms, extra procedures, unnecessary restriction? See Multiple Testing and Overdiagnosis.
5. Is the laboratory-developed test reviewed by a regulator? In the US, an FDA final rule that would have added laboratory-made tests to the definition of in vitro diagnostic products was vacated by a federal district court on March 31, 2025, and the FDA reverted the text on September 19, 2025. [4] This is not legal advice; other countries differ and we did not verify them.