Breast cancer treatments have made great strides in recent years with significant reductions in overall mortality. However, the incidence of breast cancer has increased just slightly in recent years after a dip in the early 2000s.
“The good news is that mortality is decreasing, but it still remains high. We still have a long way to go to tackle this problem of breast cancer incidence as well as the number of deaths,” said Angela DeMichele, MD, co-leader of the breast cancer research program at the University of Pennsylvania’s Abramson Cancer Center. She participated in a session on obstacles in breast cancer treatment held in December at the San Antonio Breast Cancer Symposium. She joined other oncologists in outlining key challenges that need to be addressed to improve breast cancer diagnosis and treatment.
They highlighted six obstacles: The need for more prevention/early detection strategies; the underutilization of artificial intelligence; underuse of precision oncology such as targeted therapies; the need for innovation in clinical trials; a widening gap in cancer disparities; and the need to align incentives and funding for research collaboration, training, and retention.
Since 2012, the Food and Drug Administration has approved 20 new therapeutics to treat breast cancer. Nadia Harbeck, MD, PhD, director of the breast center at LMU University Hospital, Munich, said that the development of new therapies has in a way become a victim of its own success. Therapies and survival have improved, making it harder to compare novel therapies to the standard of care and prove a benefit. Treatment guidelines are changing so quickly that clinical trials are sometimes obsolete by the time they are published because of changes to the standard of care. That places a need on more real-world evidence that can be designed to be useful in the clinic, and AI can help here. “We need to convince regulators to act upon cleverly planned real world evidence analysis. You can randomize them, you can use registries, and you should also be able to change labels because of [new] data,” Dr. Harbeck said.
There are many risk factors that drive breast cancer, and it is very heterogeneous, said Christine Ambrosone, PhD, chair of the department of cancer prevention and control at Roswell Park Comprehensive Cancer Center, Buffalo, N.Y. She called for identifying patients who are at risk for a poor prognosis, such as patients with hormone receptor–negative breast cancer, high-grade, and triple-negative breast cancer. Otherwise there is a risk of overtreatment of low-risk tumors, which could potentially be identified with new tools in precision oncology such as liquid biopsy tests, also known as multicancer early detection tests. These tests can detect cancers long before they become symptomatic. The first such test was launched this year and many more are in clinical trials.
Regina Barzilay, PhD, professor and expert in the use of artificial intelligence in health at the Massachusetts Institute of Technology, pointed out that machine learning is used in many fields, but hardly at all in breast cancer. It could be applied to data on biomarkers and other factors collected from retrospective analyses and clinical trials. She added that machine learning is often applied to biochemistry and single cell analysis of other tumor types, but rarely in breast cancer. “It is severely underutilized,” Dr. Barzilay said. One challenge is that researchers are not necessarily familiar with the techniques of machine learning and AI. Another issue is that breast cancer data are not easy to share and may not be readily available to AI researchers. “An investment in interchangeable data is crucially important,” she said.
Artificial intelligence could assist in identifying and modeling factors that contribute to cancer risk by teasing apart complicated relationships, such as the association between pregnancy, breastfeeding, and breast cancer risk. Pregnancy reduces the risk of hormone receptor–positive disease, but increases the risk of hormone receptor–negative disease.
Another key challenge is the underuse of “omics” technologies, which measure large scale patterns in biological characteristics such as gene variation or protein expression. That has roots in the history of breast cancer being considered as a separate entity from other solid tumors such as lung or pancreatic cancer. Fabrice André, MD, PhD, an oncologist with Gustave Roussy Cancer Center, France, emphasized that breast cancer shouldn’t be considered an entity when it’s metastatic. Instead, tumors should be defined by molecular characteristics they share. He anticipates a personalized medicine future where academic and industry groups collaborate to create an individualized therapy for patients based on genetic factors.