Patients with early-stage breast cancer can still face a risk of the cancer returning (recurring) many years after their original diagnosis and treatment. This is why it is important to look beyond the first 5 years and understand who remains at risk.

Doctors usually estimate this risk using factors such as tumor size, grade, and whether lymph nodes are involved. They look at patient age and menopausal status because lifetime exposure to hormones can drive breast cancer growth and damage DNA. Doctors also use a tumor test called Oncotype DX Breast Recurrence Score® to provide a risk score (0-100). This test assesses the activity of 21 genes that fuel tumor growth. Together, all of this information helps women and their doctors decide whether to take chemotherapy or hormone therapy after surgery to prevent cancer from coming back.  

Scientists from ECOG-ACRIN Cancer Research Group (ECOG-ACRIN) and Caris Life Sciences are working together to develop smarter, more effective ways to help women with early-stage, hormone receptor–positive (HR+) breast cancer know their recurrence risk. HR+ breast cancer is the most common breast cancer subtype. The researchers are using advanced technology from Caris and also teaching AI (artificial intelligence) to examine different types of information at once:

• What the tumor looks like under a microscope
• Basic facts about the cancer, like its size and whether it spread to nearby lymph nodes
• What's happening inside the tumor cells
• What's happening inside healthy cells, and how our immune system can inform on how fast the cancer might recur

Think of it like getting a second, third, fourth, and fifth opinion all at once—and having a computer find patterns that may be difficult to see with the human eye. This technology is designed to go beyond current standards to provide each woman with the most personalized treatment plan possible. By using AI to analyze routine tissue samples in new ways, researchers are getting closer to that goal.

At the recent San Antonio Breast Cancer Symposium in December 2025, ECOG-ACRIN researchers shared three exciting discoveries from AI-based research projects.

Predicting Late Relapses Up to 15 Years

The research team studied information from women who participated in a cancer trial called TAILORx—a National Cancer Institute-sponsored, ECOG-ACRIN-led trial with cooperative group participation across the entire National Clinical Trials Network. TAILORx was one of the largest breast cancer studies ever conducted. It involved thousands of women and years of follow-up. With the help of AI, the scientists found something interesting: different types of information were better at predicting cancer recurrence at different times.

Looking inside tumor cells (examining genes and proteins that control how the cancer behaves) was the best predictor of whether cancer would return within 5 years of diagnosis. Looking at what the tumor looks like under the microscope was best for predicting whether cancer would return after 5 years. And when the researchers combined all the information, including the extent to which immune cells attack tumor cells, they got the most accurate picture of a woman's long-term risk—up to 15 years.

Predicting Who May Benefit from Extended Endocrine Therapy

Five years of endocrine therapy after surgery is standard for patients with HR+ breast cancer, significantly reducing recurrence and mortality. Typically, premenopausal women take tamoxifen, and postmenopausal women use aromatase inhibitors. Some may need up to 10 years of treatment, but hormone therapy can cause side effects like joint pain, hot flashes, and bone loss, so knowing who truly benefits is important.

To address this problem, researchers developed an AI tool to help doctors better identify women who are more or less likely to benefit from additional hormone therapy. After successful early testing, the tool was used to study more than 4,000 TAILORx samples, where it again performed well.  

Predicting Risk in Lobular-Type Breast Cancer

Breast cancer is not one disease, but a family of related diseases. Two main groups within this large family are the ductal and lobular types. Ductal cancer forms in lumps that are easier to spot, while lobular cancer grows in a single line, making it harder to detect on mammograms. As a result, the lobular type is often diagnosed at a later stage and can also require more extensive surgery.  

A team of pathologists partnered with Case45 and Tempus/PaigeAI and combined AI tools with their usual tumor tissue slide reviews to learn more about lobular cancer. Looking at thousands of TAILORx images, they found that lobular tumors had worse long-term outcomes than ductal tumors. This was true even in cases where the Oncotype DX Breast Recurrence Score® (mentioned previously) showed that the risk of recurrence was low.

This is a very important discovery. It suggests that some patients with lobular as well as ductal cancer, who are currently advised to skip chemotherapy and/or longer endocrine therapy, may be at higher risk and could benefit from more and/or longer treatment. This finding does not change current treatment guidelines yet, but it highlights an important group of patients for future study.

The researchers also developed a new AI-based way to improve the identification of lobular cancer. In short, using AI alongside expert pathology can better personalize care and ensure higher-risk patients don’t slip through the cracks.

When Will These Tools Be Available?

These AI tools are still being tested and aren't yet ready for use in clinics. These studies looked back at data from women who were already treated and followed for many years. The results are helping researchers understand long-term patterns. Before they can be used in clinics, more trials are needed to show they improve outcomes, not just predictions.

Researchers also need to ensure these tools work equally well for people of different races, ages, and across different healthcare settings. AI models can perform poorly if trained on non-representative data.

But the early results are encouraging. For women facing a breast cancer diagnosis, these tools could eventually help to ensure that everyone gets the best treatment possible for them.

This research points toward a future where fewer women are over-treated, fewer are under-treated, and more women can move forward with confidence in their care decisions. It shows what's possible when hospitals, cancer centers, and technology companies work together to put powerful new tools in the hands of doctors and patients.

Note: These AI tools are still in development. If you or someone you love has been diagnosed with breast cancer, talk with your healthcare team about the best treatment options available today.

Featured Content:

To learn more about the research described above, please watch these three short videos featuring leading researchers at the San Antonio Breast Cancer Symposium:

• Dr. Joseph Sparano: Using AI to help predict late breast cancer relapses up to 15 years
• Dr. Eleftherios (Terry) Mamounas: An AI model helps predict who may benefit from extended endocrine therapy
• Dr. Roberto Salgado: AI helps pathologists identify high-risk lobular breast cancer features