{"entity":{"id":"idea-tr2-ai-cdx-change-control","kind":"idea","name":"Version control and locked reference sets for AI algorithms used as companion diagnostics","aka":[],"tldr":"AI is starting to decide which patients get which cancer drug. Every change to the software should be tested against a fixed public set of cases before it is used on patients.","summary":"AI-based scoring of HER2, PD-L1 and other markers is entering clinical use. Software updates can shift positivity rates silently. Regulators (FDA's predetermined change control plans, EU AI Act) are building frameworks. A concrete requirement for oncology companion diagnostic algorithms: every version must report performance on a locked public reference set, changes must be logged with effect on positivity rates, and laboratories must record the algorithm version in each patient report.","asOf":"2026-09-08","links":[{"label":"Bottleneck evidence (Biomarkers are not validated or standardised): Fernandez et al., Examination of low ERBB2 protein expression in breast cancer tissue (JAMA Oncology 2022)","url":"https://doi.org/10.1001/jamaoncol.2021.7239"}],"tags":[],"related":["paige","pathai","idea-ai-her2-low-scoring","idea-tr2-open-cdx-validation-sets"],"cancers":[],"sections":[],"technologies":["digital-pathology-ai","pathology-foundation-model"],"targets":[],"drugs":[],"companies":[],"institutions":[],"pathways":[],"terms":[],"trials":[],"people":[],"bottlenecks":["b-biomarker-validation","b-ai-validation"],"keyPapers":["paper-fernandez-jama-oncol"],"journals":[],"dependsOn":[],"notes":[],"hypothesis":"Version reporting will reveal at least one clinically meaningful drift (positivity change above five percentage points) in a deployed algorithm within two years, which would otherwise have gone undetected.","rationale":"Software versioning is routine in engineering and absent in diagnostic pathology reporting; drift has been documented in deployed medical AI.","test":"Implement version logging and reference-set testing for two deployed pathology algorithms across ten laboratories; monitor positivity rates by version.","maturity":"early-clinical","actor":"regulator","cost":"small","horizonYears":2},"route":"/ideas/idea-tr2-ai-cdx-change-control/","neighbours":{"company":[{"id":"paige","kind":"company","name":"Paige AI","route":"/companies/paige/"},{"id":"pathai","kind":"company","name":"PathAI","route":"/companies/pathai/"}],"idea":[{"id":"idea-ai-her2-low-scoring","kind":"idea","name":"AI quantification of HER2-low and HER2-ultralow","route":"/ideas/idea-ai-her2-low-scoring/"},{"id":"idea-tr2-positivity-rate-surveillance","kind":"idea","name":"Monitor biomarker positivity rates across labs in real time to catch assay drift","route":"/ideas/idea-tr2-positivity-rate-surveillance/"},{"id":"idea-tr2-open-cdx-validation-sets","kind":"idea","name":"Public gold-standard datasets for validating every cancer biomarker test","route":"/ideas/idea-tr2-open-cdx-validation-sets/"}],"technology":[{"id":"digital-pathology-ai","kind":"technology","name":"Digital pathology & AI","route":"/technologies/digital-pathology-ai/"},{"id":"pathology-foundation-model","kind":"technology","name":"Pathology & radiology foundation models","route":"/technologies/pathology-foundation-model/"}],"bottleneck":[{"id":"b-ai-validation","kind":"bottleneck","name":"AI that is built but not validated or deployed","route":"/bottlenecks/b-ai-validation/"},{"id":"b-biomarker-validation","kind":"bottleneck","name":"Biomarkers are not validated or standardised","route":"/bottlenecks/b-biomarker-validation/"}],"paper":[{"id":"paper-fernandez-jama-oncol","kind":"paper","name":"Examination of Low ERBB2 Protein Expression in Breast Cancer Tissue","route":"/key-papers/paper-fernandez-jama-oncol/"}]}}