Biomarker Discovery Research

Start here

A biomarker is a measurable sign, such as a blood level, scan feature or gene pattern. Finding one is the start of a long road, not proof that it helps anyone. This page explains the methods and why so few discoveries become useful tests. It is education, not advice about any particular test.

What a biomarker is

The FDA-NIH BEST glossary, as described by the FDA, defines a biomarker as a defined characteristic measured as an indicator of normal biological processes, pathogenic processes, or responses to an exposure or intervention. Molecular, histologic, radiographic or physiologic characteristics all count, and a biomarker is not an assessment of how an individual feels, functions or survives. [1] That last point matters: a biomarker change is not a clinical outcome. See biomarkers vs clinical outcomes and surrogate vs patient-important outcomes.

BEST defines seven categories: susceptibility/risk, diagnostic, monitoring, prognostic, predictive, pharmacodynamic/response and safety. [1] A biomarker may be a single measurement or a panel. [1] The same marker can fill different roles, and each role needs its own evidence.

The pipeline from idea to use

Ioannidis and Bossuyt describe the pipeline as: discovery, validation, clinical translation, evaluation, and implementation (with deimplementation if a biomarker proves not useful). [2]

1. Discovery. Researchers compare samples, often with large-scale "omics" measurements, to find differences between groups. See multi-omics explained. Discovery studies can suffer from poor methods and incomplete, selective reporting. [2]
2. Validation. The finding must be tested in independent samples. External independent validation is limited. [2]
3. Clinical translation and evaluation. Does the test change decisions or outcomes? Evaluation is sparse, and the clinical utility of many biomarkers remains unknown. [2]
4. Implementation. Even then, a test may be used without strong evidence, and removing an entrenched but unhelpful biomarker can meet resistance. [2]

Evidence label for the field's track record: mixed. The same authors note an expanding literature with almost ubiquitous significant results, yet few discovered biomarkers used in routine care. [2] The paper is a review and opinion piece, so this is an expert assessment rather than a count of every biomarker.

What good discovery research looks like

The authors propose more stringent methods, better reporting, larger collaborative studies, careful external validation, preregistration, systematic reviews, pivotal randomized trials and implementation studies. [2] For readers, this translates into questions: how many people, from where, was it tested in a separate group, and does it add information beyond what clinicians already use? See predictive accuracy and external validation and analytical validity.