Memorial Sloan Kettering Physician-Scientists Develop Innovative Multimodal Machine Learning Model That Improves Prediction of Metastatic Breast Cancer Treatment Options

New research presented during the 2024 San Antonio Breast Cancer Symposium (SABCS) reveals a new machine learning model that could change the way metastatic breast cancer is treated in the future. By combining clinical and genomic data, physician-scientists from Memorial Sloan Kettering Cancer Center (MSK) developed a tool that could help improve predictions of how people with hormone receptor-positive, HER2-negative (HR+/HER2-) metastatic breast cancer respond to CDK4/6 inhibitors, a class of oral medications that control cell division and are often prescribed in combination with hormone therapy to treat this subset of patients.

CDK4/6 inhibitors have changed first-line treatment for HR+/HER2- metastatic breast cancer when combined with endocrine therapy. However, current clinical tools often fall short in identifying patients most likely to benefit. The new model, developed using MSK's OncoCast-MPM platform, integrates detailed clinical and pathologic characteristics and genomic data to better stratify patients before they start treatment.

"Treatment for breast cancer is constantly evolving, but not all patients benefit equally from CDK4/6 inhibitors," said Pedram Razavi, MD, PhD, the study's senior author and Scientific Director of the Global Research Program at MSK. "While some patients experience remarkable responses, others face resistance and limited benefit."

The new MSK model offers a more precise way to predict outcomes, potentially enabling the implementation of tailored escalation and de-escalation strategies at the time of metastatic diagnosis for people with HR+/HER2- breast cancer. "Our risk adaptive approach aims to improve the outcomes for high-risk patients while sparing low-risk patients unnecessary side effects, fostering more informed decision-making between patients and their physicians," explained Dr. Razavi.

In the study, the MSK team, led by members of the Breast Translational Program, compared three predictive models using data from 1,078 patients (761 patients in the training cohort and 326 patients in the test cohort) who received hormone therapy with CDK4/6 inhibitors combinations. A targeted tumor-sequencing test called MSK-IMPACT^®, which detects mutations and other critical changes in the genes of both rare and common cancers, was used before treatment commenced or within two months of the start of treatment.

Key findings:

• Clinicopathological Features (CF) including tumor grade, location of metastases, and prior treatments: Identified three risk groups with median progression-free survival (PFS) ranging from 6.3 to 24.5 months.
• Genomic Features (GF) including tumor mutational burden and gene mutations such TP53, RB1, PTEN, among others: Achieved stratification in three groups with PFS range of 9.9 to 23.1 months.
• Combined Clinical and Genomic Model (CGF) including a combination of both: Delivered results with four risk categories and a PFS range of 5.3 to 29 months, showing the power of this integrated analysis.

Unlike currently available clinical tools, the CGF model provided nuanced predictions, helping physicians make informed decisions. "The tool doesn't just refine risk categories-it introduces the possibility of proactive adjustments to patients' treatment plans," explained Dr. Razavi. For example, people flagged as high-risk can receive closer monitoring and quicker transitions to second line therapies or considered for clinical trials of escalated care as first line treatment, potentially improving survival outcomes.

While the study's retrospective nature and single-institution design are noted limitations, its findings mark a critical step forward in predictive oncology. The MSK team is now working toward validating the model with external datasets with the goal to develop an online individualized prediction tool based on this model - often referred to as a nomogram - for oncologists worldwide. "This innovation could redefine how we treat newly diagnosed metastatic breast cancer, moving us closer to personalized medicine," said Dr. Razavi.

The study was partially and indirectly supported by the National Institutes of Health, Department of Defense, Susan G. Komen for the Cure, Breast Cancer Research Foundation, AstraZeneca, Sophia Genetics, Novartis, and Tempus.