Prognostic versus predictive biomarkers

In one sentence

A prognostic biomarker is associated with outcome, while a predictive biomarker identifies differences in the effect of a particular treatment.

The intuition

A weather forecast tells you the risk of rain. Evidence that an umbrella helps differently under particular conditions answers another question. In cancer, a marker of higher recurrence risk does not automatically identify which extra drug will reduce it.

How it works

A prognostic study relates a marker to an outcome in a defined population and care context. A predictive claim concerns a treatment effect: does the difference between treatment and comparator depend on the marker? A marker can be both, one or neither.

Simply finding a significant treatment effect in one subgroup and a nonsignificant effect in another is not a valid proof that effects differ. Interaction analysis and an appropriate study design address the difference more directly. Small subgroups and many tests can generate unstable findings.

ctDNA (circulating tumor DNA) positivity after treatment can be strongly associated with later recurrence. That supports a prognostic claim. To establish a useful treatment-selection rule, studies must test whether a specified action after that result improves outcomes. A negative result likewise does not establish that standard therapy can safely be omitted.

The same separation applies to RCB (residual cancer burden), TILs and molecular scars. A cohort showing favorable outcomes with high TILs does not show that vaccination will reproduce the difference. An HRD (homologous recombination deficiency)-associated response pattern does not establish eligibility for a drug label using a different biomarker. Clinical utility asks whether using the information improves a meaningful decision and outcome.

Why it matters in cancer

It separates information about outcome from evidence that a marker identifies who benefits from a particular treatment.

Worked example

Residual-disease patients with a negative marker do better in a cohort. Without a comparison testing omission, the result cannot prove they should skip radiation. Without a matched drug comparison, it cannot identify the best additional medicine either.

Common confusions

High risk is not proof of larger benefit from every treatment.
Predicting response statistically is not always the same as a predictive treatment biomarker.
A validated assay and an actionable strategy are separate achievements.

RCB and actionability.

Sources and scope

General teaching, source-checked October 8, 2026. The examples are hypothetical unless a study is explicitly named. Expert and learner review remain pending.