Precision oncology: what the famous reality checks actually found
SHIVA and NCI-MATCH did not show that all biomarker-matched therapy fails. They showed how steep the route from sequencing to treatment can be and how weak-evidence matches in late-line disease often underdeliver.
Written retrospectively and published on August 3, 2026.
Background reading, not clinical guidance. This describes published research; it is not a recommendation for any individual case, and the cited studies may have been superseded.
The promise of precision oncology is intuitive: find a molecular alteration and match it to a drug. The difficult part is that an alteration may be biologically real while still being a weak predictor of benefit from one particular drug in one particular cancer.
In the randomised SHIVA trial, 195 people with refractory metastatic solid tumours received either an off-label targeted drug selected by a pre-specified molecular algorithm or treatment of physician’s choice. Median progression-free survival was 2.3 versus 2.0 months, with no statistically significant difference.
A match is also an operational funnel
NCI-MATCH sequenced 5,540 successful tumour samples. An alteration actionable for a study arm was found in 37.6%, but only 17.8% were assigned to an arm after molecular and clinical exclusions. About 70% of assigned participants started treatment. The gap is part biology, part evidence quality, and part trial logistics.
The more useful conclusion is conditional. In SAFIR02-BREAST, matched treatment improved progression-free survival for high-evidence ESCAT tier I or II alterations, but not for unselected alterations. “Matched” is therefore not a single evidence category.
Comments
Threads are public and indexed — please keep personal medical data out of them. Found a clinical error? Open an issue instead.
Comments are not switched on yet — the backend still needs to be configured.