What if you could better predict whether a patient’s breast cancer will respond to a treatment? What if you could spare patients from having to deal with harsh side effects if you know that a treatment offered little impact? At the 2024 San Antonio Breast Cancer Symposium (SABCS), held December 10-13, researchers presented potential ways to get answers to questions like these. They investigated a couple of treatment options for patients with the most common subtype of breast cancer—hormone receptor (HR)-positive, human epidermal growth factor receptor 2 (HER2)-negative.
In one presentation, Pedram Razavi, MD, PhD, the scientific director of the Global Research Program at Memorial Sloan Kettering Cancer Center, looked at whether a machine learning (ML) model could help predict outcomes from the addition of CDK4/6 inhibitors to endocrine therapy as a first-line treatment for patients with HR-positive, HER2-negative metastatic breast cancer.
In another, Nan Chen, MD, assistant professor of internal medicine at the University of Chicago Medicine, used a patient’s recurrence score from the Oncotype DX genomic test to better understand which patients with stage I/II, node-negative, HR-positive, HER2-negative breast cancer would benefit from the addition of anthracyclines to adjuvant taxane-based chemotherapy.
What they found could help identify when to add these therapies to a patient’s treatment regimen.
When to Add CDK4/6 Inhibitors to Endocrine Therapy
Previous studies have found that the addition of CDK4/6 inhibitors such as palbociclib (Ibrance), abemaciclib (Verzenio), or ribociclib (Kisqali) to endocrine therapy have extended progression-free survival (PFS) and overall survival (OS) in patients with HR-positive, HER2-negative metastatic breast cancer. Razavi, however, said that responses to CDK4/6 inhibitors can vary widely.
“Some patients have progression of disease within one to two years, others do not benefit at all, and some do remarkably well and stay on these drugs for years,” he explained in an interview with Cancer Research Catalyst. “There have been major advances in treating breast cancer in recent years, offering multiple potential options for upfront escalation. This highlights a significant need in clinic to identify patients who may or may not benefit from CDK4/6 inhibitor combinations at the time of metastatic diagnosis so that we can effectively implement escalation and de-escalation strategies.”
The current method used to help determine patients who are at risk of early disease progression on first-line CDK4/6 inhibitor combinations is by examining clinical features like treatment-free interval (TFI). This looks at the time between the last dose of adjuvant endocrine therapy and the development of metastatic disease and measurable disease.
“In our opinion, we can do much better than the current clinical risk criteria that are based on very limited clinical factors,” Razavi said. “More accurate prediction of outcomes could help empower patients and their oncologist to make better informed decisions, whether to escalate care proactively or to avoid unnecessary side effects and financial toxicity from escalated upfront approaches.”
Razavi and his colleagues used an ML tool developed by Memorial Sloan Kettering called OncoCast-MPM to test whether additional clinical and genomic factors could more accurately stratify patients. They compared three models, each one incorporating factors already known to be associated with outcomes or resistance to either CDK4/6 inhibitors or endocrine therapy.
The one based on clinicopathological features (CF) looked at factors such as liver metastasis, TFI less than one year, progesterone receptor negativity, low estrogen receptor expression, and presence of visceral metastasis. The one based on genomic features (GF) examined the loss of specific genes (TP53, PTEN, and RB1), MYC amplifications, alterations in the RTK-MAPK pathway, whole genome doubling, and high proportion of loss of heterozygosity. The final model (CGF) looked at all these factors together.
“All three models performed really well, surpassing the conventional clinical risk models based on a single or a few clinical features. But the power of the analysis shone when we started combining the clinical and genomic features together,” Razavi said in a press release.
After testing in a training cohort of 761 patients with HR-positive, HER2-negative metastatic breast cancer, the CF and GF models identified a high-risk group in which patients had a median PFS of 6.3 (CF) and 9.9 (GF) months, an intermediate-risk group with median PFS of 15.2 (CF) and 18.1 (GF) months, and a low-risk group with median PFS of 24.5 (CF) and 23.1 (GF) months. The CGF model identified an additional intermediate-risk group with median PFS of 5.3, 10.7, 19.8, and 29 months from high- to low-risk groups. The superior stratification of patients with the CGF model was further supported by a 6.5-fold difference in the hazard ratio between the high- and low-risk groups compared to a three- to fold-fold difference in the CF and GF models.
“Knowing that a patient on first-line CDK4/6 inhibitors is in the high-risk group could prompt the treating oncologists to implement closer disease monitoring and utilizing liquid biopsy and tumor-derived biomarkers to inform second-line treatment options and clinical trials,” Razavi explained.
When to Add Anthracyclines to Taxane-based Chemotherapy
Chen and her colleagues examined a way to help providers make a different treatment decision. Following a patient’s primary treatment, providers often weigh the benefit of prescribing adjuvant chemotherapy to help prevent recurrence. If providers decide this is the best option for a patient, then the question becomes what type of chemotherapy.
“In the modern era, patients are generally receiving taxane-based regimens or taxane- and anthracycline-based regimens,” Chen explained to Cancer Research Catalyst. “However, there is little data to guide the use of more intensive chemotherapy via addition of anthracyclines.”
Chen added that providers generally consider various clinical data such as the size of the tumor, a patient’s age, and other risk factors when deciding whether to add anthracyclines to the treatment regimen. She also pointed out that while some previous studies have shown that anthracyclines in combination with taxane-based chemotherapy can reduce death and recurrence at higher rates, anthracyclines have also been associated with a later risk of developing leukemia. This has led to some debate about the use of anthracyclines in treating breast cancer and a need to better identify who may most benefit from them.
“Our study seeks to address this gap and provide some guidance to better individualize chemotherapy regimens for our patients,” Chen said.
Along with her colleagues, she performed a post hoc analysis of the TAILORx study, a randomized phase III trial that compared treatment with adjuvant endocrine therapy alone to combinations of endocrine therapy and chemotherapy in patients with stage I/II, node-negative, HR+/HER2- breast cancer. But for their analysis, Chen and her colleagues focused on the participants who received either taxane with anthracycline/cyclophosphamide and similar regimens (T-AC) or taxane with cyclophosphamide (TC) chemotherapy after surgery. Then, they examined the outcomes data of patients in each of these arms based on their Oncotype DX recurrence score (RS). Oncotype DX is a commonly used test to analyze the expression of 21 different genes in a tumor to arrive at a score that indicates how likely cancer is to recur.
Patients considered to be high risk (a RS of 31 or greater) who had tumors that were 2 cm or greater were found to have improved survival outcomes after five years when treated with T-AC compared to TC. This included a higher distant recurrence-free interval, distant recurrence-free survival, recurrence-free interval, and a trend towards improved recurrence-free survival and OS at five years. While most of these were dependent on tumor size, the primary endpoint of distant recurrence-free interval was higher regardless of tumor size. Further, as the RS increased above 31 so did the benefit of anthracycline therapy.
“These results are in line with current clinical practice, where we give anthracyclines more readily in tumors biologically closer to triple-negative disease,” Chen said in a press release. “While most HR-positive, HER2-negative tumors do not have RS 31 or greater, many of the highest RS tumors may have less estrogen receptor expression, higher proliferation, and are closer along the spectrum towards triple-negative disease, a subset in which the benefit of anthracyclines has been much more clearly demonstrated.”
While they did not see any trend toward benefit in patients with an RS between 26 and 30, Chen said that the low risk of recurrence in these patients could make it difficult to measure any impact from anthracyclines.
Staying Ahead of the Game
At a press conference, SABCS co-director Carlos L. Arteaga, MD, FAACR, director of the Simmons Comprehensive Cancer Center and associate dean of oncology programs at UT Southwestern Medical Center, said the technology that will help providers and patients make more-informed treatment decisions is closer to the present than the future. While the Oncotype DX test is already being used in the clinic, he explained we now have even more ways to gather data from the tumor and tools like the one Razavi described show how that data could be applied.
“We are going to have—very soon—predictive models where we can provide patients and their physicians odds of what their journey is going to be when they are on these therapies,” Arteaga said.
Both Chen and Razavi are planning to validate their findings in larger studies. Chen wants to look at more patient groups, such as those with node-positive disease, as well as examine the benefit of adding anthracyclines to other therapies for breast cancer, including CDK4/6 inhibitors.
Razavi and his team are going to test their ML model with data sets from outside of Memorial Sloan Kettering Cancer Center. The goal for his team is to eventually develop an online tool for physicians to receive patient-specific outcome predictions after inputting clinical and genomic data.
“This could put us one step closer to staying ahead of breast cancer,” Razavi said.
