The number that matters: 0.67 against 0.61. Those are the concordance scores, known as C-index, that New York University researchers reported when they set their artificial intelligence test against Oncotype DX, the 21-gene assay, in 858 breast cancer patients who had taken both. The AI came out ahead. The confidence intervals overlap, though, so it is an edge, not a knockout.
The study, led by Krzysztof J. Geras, was published in Nature Communications. Geras is affiliated with NYU through its data science center and its Grossman medical school.
What goes in
The test reads digitized pathology slides, the same stained tumor images pathologists look at anyway, and combines them with ordinary clinical data: tumor stage, age, estrogen and progesterone receptor status, HER2 status and histologic subtype. Nothing new has to be ordered. “Our AI test can read the same tumor slides pathologists already examine and, combined with basic clinical details, accurately estimate how likely a patient’s cancer is to return,” Geras said in a news release.
Developers trained and tested it on data from 8161 patients in 15 cohorts across 7 countries. They then checked it against 3502 patients in 5 separate cohorts. The pooled C-index there was about 0.71, and patients the model labeled high risk were far more likely to see recurrence, with a hazard ratio of 3.63 compared with the low-risk group.
The cost of the incumbent
Genomic assays like Oncotype DX guide adjuvant treatment for hormone receptor-positive, HER2-negative disease. They are costly, need specialized lab processing, can take weeks to come back and use up tissue that might serve later molecular work. Because the AI works from slides already on file, the researchers say answers could arrive in hours and at lower expense.
The reclassification numbers are the part investors in diagnostics will want to read twice. Patients who got an intermediate Oncotype DX score, the murky middle, were sorted by the AI: roughly 80% landed in the low-risk group and nearly 20% in the high-risk group. If that holds up, it could sharpen chemotherapy decisions for exactly the patients doctors find hardest to call.
Where it stops
The model also held up in triple-negative and HER2-positive cancers, two subtypes with no guideline-recommended genomic recurrence assay. Evaluation in completed randomized clinical trials is still needed before anyone uses the tool to guide treatment. Patients should talk to their own doctor about what it means for them.
Sources
- AI Test Predicts Breast Cancer Recurrence Using Standard Pathology Slides – Pharmacy Times
- New AI Test Delivers Rapid Breast Cancer Recurrence Predictions – LabMedica
