New research shows how AI may help avoid unnecessary chemotherapy for breast cancer patients
Research led by RCSI University of Medicine and Health Sciences and University College Dublin (UCD) has identified immune markers that could help doctors more accurately determine which breast cancer patients are unlikely to benefit from chemotherapy, potentially sparing some patients from unnecessary treatment.
Chemotherapy is regularly used in the treatment of early-stage, ER+HER2- breast cancer, which accounts for around 70% of all breast cancer diagnoses annually. While it remains an important treatment, the side effects can be debilitating and for a significant proportion of patients, the benefit remains uncertain, raising concerns about overtreatment for those who may have remained cancer-free without it.
New research published today in Nature Communications has found that analysing the body's own immune cells found near a tumour using new AI-based methods can provide additional insight to identify which patients can safely have chemotherapy withheld.
Currently, patients are assessed using a risk score, but the majority receive an intermediate result, meaning chemotherapy is often prescribed as a precautionary measure. To address this uncertainty, the researchers used samples from a randomised treatment trial comparing hormone-blocking therapy alone against hormone-blocking combined with chemotherapy in an Irish cohort of patients with intermediate risk scores.
Professor Darran O’Connor, research lead at RCSI School of Pharmacy and Biomolecular Sciences, explained the clinical importance of the findings: “For patients with an intermediate genomic risk, the decision around chemotherapy is often difficult and uncertainty frequently leads to treatment that may not have been necessary, impacting on quality of life.”
“Genomic testing has advanced our ability to tailor treatment for these patients, but AI-based analysis of the tumour microenvironment takes this further still. Crucially, because this approach works from tissue samples processed as standard, it has the potential to improve both the precision and the equity of treatment for most women with early-stage breast cancer, regardless of where they are treated."
The study found that a high density of cancer-targeting immune cells, called cytotoxic T-cells, in the tissue surrounding a tumour could more accurately define patient risk compared to current genomic profiling methods. Patients with high cytotoxic T-cell density were found to have poorer outcomes when treated with chemotherapy, suggesting its usefulness as a predictive marker for therapy effectiveness. RCSI and UCD have jointly filed a patent for the technology and is now seeking to commercialise the approach to support its translation into clinical practice.
Dr Zak Kinsella, first author on the study and Postdoctoral Researcher at RCSI said: "It's really encouraging to see how much additional prognostic information can be extracted from these samples using AI. The density of cytotoxic T-cells in the tumour microenvironment proved to be a remarkably strong predictor of treatment response, and that has real implications for how we approach chemotherapy decisions in this patient group."
Professor William Gallagher, Senior Author from the UCD School of Biomolecular and Biomedical Science and UCD Conway Institute, said: “Before this approach can be implemented in clinical practice, further validation in larger studies will be required. However, the findings give us a clearer picture of what drives recurrence risk of breast cancer in patients with intermediate genomic scores and brings us closer to the kind of personalised treatment decisions that could avoid unnecessary chemotherapy."
Reference:
Kinsella, Z., Jahangir, C.A., Nyarko, H.N. et al. Spatial analyses implicate high stromal tumour-infiltrating CD8+ lymphocytes as a negative predictive marker for chemotherapy in estrogen receptor-positive breast cancer. Nat Commun 17, 4863 (2026). https://doi.org/10.1038/s41467-026-73432-2
