Ummon Healthtech™

Breast cancer (BC) is the leading cause of cancer-related deaths for women in Europe and worldwide. Adjuvant or neoadjuvant systemic chemotherapy is a standard treatment for high-risk early breast cancer (eBC) and consists of performing chemotherapy after or before surgery, respectively. Chemotherapy is recommended for the vast majority of patients with HER2-amplified (HER2+) or triple-negative (TN) eBC, as well as for patients with ER+/HER2- tumors who have the most significant risk of metastatic relapse. The decision to administer systemic chemotherapy, either in an adjuvant or neoadjuvant setting, is made by the oncologist based on each individual patient’s risk factors for relapse.

Neoadjuvant chemotherapy (NAC) can be proposed to patients to reduce the size of the primary tumor before surgery, to facilitate conservative surgery, but also to administer systemic treatment very early on, with a view to treating any micrometastatic disease as quickly as possible. In the Early Breast Cancer Trialists’ Collaborative Group (EBCTCG) meta-analysis, there was no significant difference between patients treated with a neoadjuvant scheme in terms of distant recurrence, breast cancer mortality, or death from any other cause, compared to the same standard adjuvant chemotherapies.

An advantage of NAC is the ability to evaluate the chemosensitivity of each patient’s tumor to the standard chemotherapy on the surgical specimen. Thus, patients with complete pathological response (pCR, defined as no residual invasive cancer cells either in the breast or axillary lymph nodes: pT0 and pN0) after NAC have significantly better relapse-free, and overall survival compared to patients with residual disease (RD) on the surgical specimen (especially for HER2-amplified and TN breast cancer). For these two BC subtypes, patients with RD can be selected after surgery for post-operative, adjuvant treatment intensification.

For the ER+/HER2- eBC subtype, achieving pCR is significantly less common and has a reduced prognostic value. Nevertheless, there is a small subset of chemosensitive ER+/HER2- tumors that achieve pCR, and for which chemotherapy is probably of great interest in the systemic treatment plan. Unfortunately, there is currently no effective predictive marker for identifying these tumors prior to treatment. Therefore, predicting tumor chemosensitivity (in terms of pCR) to conventional chemotherapy regimens right from the initial diagnosis would be highly beneficial.

Such prediction would enable medical oncologists and surgeons to identify patients who are likely to achieve pCR with standard treatment, and who may therefore not require intensification of systemic therapy. Conversely, they would also be able to identify those with tumors unlikely to achieve pCR after standard treatment, indicating a potential need for treatment intensification early on, or for consideration of alternative therapies. In this context, our study analyzed two extensive French cohorts of eBC patients undergoing standard NAC to develop and validate a deep learning model that uses whole slide images (WSI) from initial tumor biopsies to predict the likelihood of pCR following NAC.