NEW RESEARCH presented at the San Antonio Breast Cancer Symposium (SABCS) has shown that an advanced machine learning model that integrates clinical and genomic data to predict how patients with HR+/HER2- metastatic breast cancer respond to CDK4/6 inhibitors combined with endocrine therapy. This innovation addresses a significant challenge in oncology: the varied responses among patients, with some experiencing early resistance to treatment.
While CDK4/6 inhibitors have significantly improved outcomes for many patients, a subset remains resistant to therapy. To better predict treatment success, researchers analyzed data from 535 patients with HR+/HER2- metastatic breast cancer. Using the OncoCast-MPM machine learning framework, the team trained their model on 370 patients and validated it on 165. The goal was to stratify patients into risk groups and predict progression-free survival.
The model identified three distinct risk groups based on the likelihood of responding to CDK4/6 inhibitors. Patients in the best-performing group had a median progression-free survival of 31.3 months, while those in the poorest-performing group had a median progression-free survival of just 7.9 months. The integration of clinical and genomic data proved more accurate than models relying on either type of data alone. Key factors influencing predictions included tumor mutational burden, liver metastases, primary tumor grade, and specific genomic alterations.
This development has significant implications for cancer care. By accurately identifying patients unlikely to benefit from standard therapy, oncologists can explore alternative treatments or implement more intensive monitoring strategies. The model’s precise stratification capabilities also enhance clinical trial designs, ensuring treatments are tested on appropriately matched patient groups.
Looking ahead, the research team plans to refine the model further and validate its effectiveness across broader clinical settings. This work underscores the transformative potential of AI in delivering personalized cancer care, improving outcomes, and optimizing treatment strategies.
Reference
Razavi P. Multimodal integration of real world clinical and genomic data for the prediction of CDK4/6 inhibitors outcomes in patients with HR+/HER2- metastatic breast cancer. Abstract GS3-09. SABCS 2024, 10-13 December, 2024.
Aleksandra Zurowska | AMJ
