Villejuif, 21 september 2021

ESMO 2021

Artificial Intelligence predicts the risk of recurrence for women with the most common breast cancer

The RACE AI study conducted by Gustave Roussy and the startup Owkin, as part of the AI for Health Challenge organized by the Ile-de-France Region in 2019, was presented as a proffered paper at ESMO (European Society of Medical Oncology). This study shows that thanks to deep learning analysis applied to digitized pathology slides, artificial intelligence can classify patients with localized breast cancer between high risk and low risk of metastatic relapse in the next five years . This AI could thus become an aid to therapeutic decision making and avoid unnecessary chemotherapy and its impact on personal, professional and social lives for low risk women. This is one of the first proofs of concept illustrating the power of an AI model for identifying parameters associated with relapse that the human brain could not detect.

With 59,000 new cases per year, breast cancer ranks first among cancers in women, clearly ahead of lung cancer and colorectal cancer. It is also the cancer that causes the greatest number of deaths in women, with 14%1 of female cancer deaths in 2018,. 80%1 of breast cancers are said to be hormone-sensitive or hormone-dependent. But these cancers are extremely heterogeneous and about 20% of patients will relapse with distant metastasis.

RACE AI is a retrospective study that was conducted on a cohort of 1400 patients managed at Gustave-Roussy between 2005 and 2013 for localized hormone-sensitive (HR+, HER2-) breast cancer. These women were treated with surgery, radiotherapy, hormone therapy, and sometimes chemotherapy to reduce the risk of distant relapse.

Chemotherapy is not routinely administered because not all women will benefit from it due to a naturally favorable prognosis. The practitioner's choice is based on clinico-pathological criteria (age of the patient, size and aggressiveness of the tumor, lymph node invasion, etc.) and the decision to administer or not adjuvant chemotherapy varies between oncology centers. Genomic signatures exist today to help identify women who benefit from chemotherapy, but they are not recommended by the French National Authority for Health and are not reimbursed by the French National Health Insurance (although they are included on the RIHN reimbursement list), which makes their access and use heterogeneous in France.

Gustave Roussy and Owkin have taken up the challenge of proposing a new method that is simple, inexpensive and easy to use in all oncology centers as a therapeutic decision-making tool. Ultimately, the goal is to direct patients identified as being at high risk towards new innovative therapies and to avoid unnecessary chemotherapy for low-risk patients.

In the RACE AI study, Owkin's Data Scientists, guided by Gustave Roussy's research physicians, developed an AI model capable of reliably assessing the risk of relapse with an AUC of 81% to help the practitioner determine the benefit/risk balance of chemotherapy. This calculation is based on the patient's clinical data combined with the analysis of stained and digitized histological slides of the tumor. These slides, used daily in pathology departments by anatomo-pathologists, contain very rich and decisive information for the management of cancer. It is not necessary to develop a new technique or to equip a specific technical platform. The only essential equipment is a slide scanner, which is a common piece of equipment in laboratories. Like an office scanner that digitizes text, this scanner digitizes the morphological information present on the slide.

The results of this first study by the Owkin and Gustave Roussy teams open up strong prospects and next steps include prospectively validating the model on an independent cohort of patients treated outside Gustave Roussy. If the results are confirmed, through providing reliable information to clinicians, this AI tool will prove to be a valuable aid to therapeutic decisions.

Source
ESMO 2021 – Oral Session
Proffered paper : Translational research
Prediction of distant relapse in patients with invasive breast cancer from deep learning models
applied to digital pathology slides
Présentation n° 1124O – Channel 5 – 14h20-14h30 Sunday 19th Septembre 2021
Speaker : Ingrid J. Garberis, Gustave Roussy

Media Contacts

GUSTAVE ROUSSY :
Claire Parisel – Tél. 01 42 11 50 59 – 06 17 66 00 26 – claire.parisel@gustaveroussy.fr

OWKIN :
Talia Lliteras – Tél. +33 (0)7 87 21 81 90 – talia.lliteras@owkin.com


 1 - Institut national du cancer (France):
https://www.e-cancer.fr/Professionnels-de-sante/Les-chiffres-du-cancer-en-France/Epidemiologie-des-cancers/Les-cancers-les-plus-frequents/Cancer-du-sein
https://www.e-cancer.fr/Patients-et-proches/Les-cancers/Cancer-du-sein/Hormonotherapie

Icône PDF ► Read the press release in PDF

Catégorie de la page: