Hormone receptor-positive, HER2-negative (HR+/HER2–) breast cancer is the most common type of breast cancer.
Over half of distant recurrence cases occur five years or more after completing treatment for HR+/HER2– node-negative breast cancer.
According to the authors of a study published in August in npj breast cancer, HR+/HER2- breast cancer “accounts for about 60% of all breast cancer deaths due to high prevalence and persistent recurrence risk beyond 5 years.”
The 21-gene Recurrence Score (Oncotype DX) is an existing test used by healthcare professionals to predict the likelihood of the cancer returning and assess whether the patient would benefit from adjuvant chemotherapy.
However, the authors of the study published in npj breast cancer have created an artificial intelligence (AI) tool to provide more accurate predictions regarding recurrence risk in people with early stage HR+/HER2– node-negative breast cancer.
Read more:Heart disease: Sleeping in a darker room may help reduce risk
According to the study authors, the 21-gene Recurrence Score is most effective at predicting recurrence that develops within 5 years after treatment. They designed the new AI model, called IICM+, to hopefully better predict recurrence that develops beyond the 5-year mark.
Researchers used the C-index metric, which measures performance on predicting recurrence, to determine how good IICM+ was at predicting overall distant recurrence compared with the ability of the 21-gene Recurrence Score.
The researchers developed the IICM+ AI model using data from 2,808 patients from the TAILORx breast cancer trial. They then tested the model on data from a separate group of 1,621 patients.
The researchers found that the performance of IICM+ was particularly notable when predicting breast cancer recurrence more than five years after diagnosis. The AI model had a C-index of 0.735 for predicting overall distant recurrence. compared with 0.578 for the 21-gene Recurrence Score.
Debra Patt, MD, PhD, MBA, practicing medical oncologist and Executive Vice President of Policy and Strategy at Texas Oncology, spoke with Medical News Today about the significance of this difference.
Read more:New FDA-cleared blood tests could help identify Alzheimer’s earlier
“This is a very meaningful difference. For reference, a C-index value of 1 has perfect ranking, and at 0.5 it performs better than random guessing,” Patt, who was not involved in the study, said.
“With the IICM+ having a C-index of 0.735, the model shows good discrimination. Practically, that means it is a better estimate of risk — or a more reliable crystal ball,” she explained.
Richard Reitherman, MD, PhD, board certified radiologist and medical director of breast imaging at MemorialCare Breast Center at Orange Coast Medical Center in Fountain Valley, CA, agreed that “the difference is potentially quite meaningful.”
Reitherman, who was also not involved in the study, told MNT, “It suggests that IICM+ may be better able to distinguish which patients are at higher risk for distant recurrence than the 21-gene Recurrence Score alone.”
However, Reitherman added that, “before we change clinical practice, we need to see these results validated in larger, independent patient populations and determine whether using IICM+ improves treatment decisions and patient outcomes.”
Read more:Sullivan, 16, starts for USMNT; youngest to earn cap since Adu in ’06
HR+ breast cancer, including HR+/HER2– node-negative breast cancer, may lead to distant recurrence 10 years or more after receiving treatment.
The 21-gene Recurrence Score test is more efficient at predicting recurrence within the 5-year mark. As such, the researchers focused on the ability of IICM+ to predict recurrence after 5 years.
“Late recurrence is one of the important challenges in hormone receptor-positive breast cancer. Cancer cells can leave the original tumor, travel to another organ, and remain dormant for years before becoming clinically apparent,” Reitherman said.
“If we can better identify patients who are at risk for recurrence 5, 10, or even 15 years after diagnosis, we may be able to tailor long-term treatment more appropriately. IICM+ is promising in this regard because it appears to provide information about both early and late distant recurrence, but that potential needs to be confirmed in additional studies,” he explained.
Patt added that “having information about that risk can inform doctors and patients about what steps are necessary today to reduce the risk optimally.”
The study authors suggest that the IICM+ tool may help in identifying patients with a 21-gene Recurrence Score of 0 to 25 who would usually decline adjuvant chemotherapy but are at a higher risk of recurrence as indicated by IICM+.
A high IICM+ risk assessment could lead to patients considering opting in to appropriate adjuvant treatment.
“This could be quite significant because IICM+ may add another layer of information to the Oncotype DX result. A patient who otherwise appears to have a relatively low risk of recurrence could potentially be identified as having a higher risk based on additional biological characteristics of the tumor,” said Reitherman.
He added, “That could lead to a more individualized discussion about whether the potential benefits of additional treatment outweigh the risks. The important point is that IICM+ would inform the conversation; it would not replace clinical judgment or shared decision making.”
While the results of the study are promising, more research is necessary before IICM+ could be implemented in routine clinical practice.
“I would want to see the results reproduced in large, independent, and diverse patient populations. We also need to know whether IICM+ provides information beyond what we already obtain from established genomic tests and clinical factors, and, most importantly, whether using it improves patient outcomes,” Reitherman explained.
“Oncotype DX took years of research and clinical experience before it became widely accepted. IICM+ will need to go through a similar process of validation and clinical adoption,” he added.
Reitherman also noted that “[t]he technology itself may be quite practical, but clinical implementation requires more than simply having an AI algorithm. We need to establish that the test is reproducible, accessible, cost-effective, and easy for clinicians to incorporate into existing workflows.”
He concluded, “We also need to make sure it performs consistently across different patient populations and healthcare settings. Those questions will need to be answered before widespread clinical adoption.”
