{"entity":{"id":"prognostic-versus-predictive-biomarkers","kind":"term","name":"Prognostic and predictive biomarkers","aka":["prognostic versus predictive","treatment-by-biomarker interaction"],"tldr":"A marker can describe the likely course of a cancer or help identify differences in treatment benefit. Those roles require different evidence, and the same marker can have both.","summary":"A prognostic marker is associated with an outcome in a defined setting; a predictive marker identifies differences in treatment effect. Evidence for prediction needs a treatment comparison and assessment of how the treatment effect varies with the marker. A response association among people receiving one treatment can reflect prognosis or other differences. Statistical significance in one subgroup and nonsignificance in another does not by itself establish an interaction.","asOf":"2026-11-11","links":[{"label":"Prognostic and predictive biomarker methodology (2015)","url":"https://doi.org/10.1200/JCO.2015.63.3651"},{"label":"FDA: biomarkers, roles and qualification","url":"https://www.fda.gov/drugs/biomarker-qualification-program/about-biomarkers-and-qualification"},{"label":"FDA: circulating tumour DNA in early-stage solid-tumour drug development, 2024","url":"https://www.fda.gov/media/183874/download"},{"label":"CONSORT 2025 statement: Updated guideline for reporting randomised trials.","url":"https://doi.org/10.1371/journal.pmed.1004587"},{"label":"ICH E9: statistical principles for clinical trials, 1998","url":"https://www.fda.gov/media/71336/download"}],"tags":[],"related":["biomarker","subgroup-analysis","biomarker-cutpoint-selection","surrogate-validation"],"cancers":[],"sections":[],"technologies":[],"targets":[],"drugs":[],"companies":[],"institutions":[],"pathways":[],"terms":[],"trials":[],"people":[],"bottlenecks":[],"keyPapers":[],"journals":[],"dependsOn":[],"notes":["A prognostic marker is associated with a defined outcome under specified care in a defined population and time horizon. It is not automatically a description of untreated natural history. A predictive biomarker concerns the effect of a named treatment relative to a comparator: compare those treatment effects across marker groups, with an explicit effect scale and uncertainty. A marker can serve both roles. Prespecified groups and randomized treatment assignment support this comparison, although the actual design, missing specimens and applicability still matter. Keep the assay, specimen timing, population, care, endpoint and horizon with either claim.","A model that predicts response among people all receiving treatment A has not thereby measured the added effect of A versus another strategy. More responses in one marker group can support a response association while leaving that treatment-effect question unanswered. The same reasoning applies to attributing benefit to one component of a combined regimen. A post-treatment marker can reflect response rather than validate a baseline selection rule. Worse prognosis does not automatically imply a larger relative treatment benefit, and even an interaction does not establish safe withholding or escalation. Decision utility requires evidence about the marker-guided strategy and its harms.","Fictional two-year bad-outcome risks illustrate why the effect scale matters. In marker-positive people, comparator risk is 40 per cent and treatment risk 20 per cent; in marker-negative people they are 20 and 10 per cent. Comparator risks differ, while absolute risk reductions are 20 and 10 percentage points and both risk ratios are 0.5. Benefit is twice as large only on the stated absolute scale. These are exact invented quantities without sampling uncertainty, not measured trial effects or a treatment recommendation.","Significant in one subgroup and nonsignificant in another does not demonstrate different treatment effects. Assess the treatment-by-marker interaction, its uncertainty, multiplicity, prespecification, selection of available specimens and independent validation. A credible marker analysis needs enough relevant outcomes; its label alone supplies neither precision nor clinical usefulness."],"category":"Biomarkers"},"route":"/terms/prognostic-versus-predictive-biomarkers/","neighbours":{"term":[{"id":"antigen-expression-and-response","kind":"term","name":"Antigen expression and treatment response","route":"/terms/antigen-expression-and-response/"},{"id":"biomarker","kind":"term","name":"Biomarker","route":"/terms/biomarker/"},{"id":"biomarker-cutpoint-selection","kind":"term","name":"Biomarker cutoff selection and validation","route":"/terms/biomarker-cutpoint-selection/"},{"id":"measurement-to-benefit-evidence","kind":"term","name":"Evidence from measurement to patient benefit","route":"/terms/measurement-to-benefit-evidence/"},{"id":"subgroup-analysis","kind":"term","name":"Subgroup analysis (forest plots)","route":"/terms/subgroup-analysis/"},{"id":"surrogate-validation","kind":"term","name":"Surrogate endpoint validation: which stand-ins have earned trust","route":"/terms/surrogate-validation/"}]}}