284 - Clinicogenomic Risk Stratification Identifies Metastatic Breast Cancer Patients at High Risk for Symptomatic Brain Metastases
Presenter(s)
L. R. G. Pike1, A. Safanov1, S. Nandakumar1, D. Smith2, L. A. Boe1, E. Ferraro3, T. Erazo1, L. Bielo1, K. A. Ahmed4, K. Tsai5, I. Khatri1, J. Ah-Reum An1, J. Jee1, M. E. Robson6, A. Boire1, N. Schultz7, N. S. Moss8, W. Chatila7, and P. Razavi6; 1Memorial Sloan Kettering Cancer Center, New York, NY, 2Weill Cornell Medical College, New York, NY, United States, 3Sloan Kettering Institute, New York, NY, 4H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL, 5Carle Illinois College of Medicine, Urbana, IL, 6Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, NY, 7Department of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York, NY, 8Department of Neurosurgery, Memorial Sloan Kettering Cancer Center, New York, NY
Purpose/Objective(s): In metastatic breast cancer (MBC), brain metastases (BM) are frequently detected only after neurologic symptom onset, often necessitating craniotomy or whole-brain radiotherapy. However, routine MRI screening is not endorsed by consensus guidelines due to the low point prevalence in unselected patients. We present a novel clinicogenomic risk stratification strategy to enable targeted CNS surveillance and earlier detection of clinically significant BM.
Materials/Methods: We analyzed MBC patients without known BM at metastatic diagnosis who underwent genomic sequencing of a non-CNS specimen within one year of metastatic diagnosis. An ensemble time-dependent LASSO machine learning approach was used to develop a clinicogenomic risk model integrating baseline clinical, pathologic, and genomic features to estimate BM risk. The primary endpoint for this analysis was symptomatic brain metastasis–free survival (sBMFS), defined as time from metastatic diagnosis to first BM diagnosed in the setting of neurologic symptoms. Competing-risk and cause-specific time-to-event models were evaluated, and patients were stratified into low-, intermediate-, and high-risk groups.
Results: Among 1,594 patients with MBC, the model demonstrated robust stratification for sBMFS, with pronounced separation between risk groups. At 2 years, sBMFS was 91%, 77%, and 46% for patients classified as low-, medium-, and high-risk, respectively (log-rank p<0.001). Correspondingly, the 2-year cumulative incidence of any BM and symptomatic BM was 2.3% and 0.76%, 11% and 5.2%, and 32% and 20%, for low-, intermediate-, and high-risk patients, respectively (Gray's test p<0.001 and p<0.001, respectively). Across all risk strata, most patients who developed symptomatic brain metastases required craniotomy or whole-brain radiotherapy at initial presentation. These findings were independently validated in external cohorts, including a phase II clinical trial dataset evaluating MRI surveillance in patients with MBC.
Conclusion: Clinicogenomic risk stratification identifies patients with MBC at increased risk for symptomatic CNS progression. These findings support the development of risk-adapted CNS surveillance strategies. Earlier identification may enable intervention before the onset of neurologic compromise, when focal radiotherapy or systemic therapy alone may be sufficient, thereby reducing the need for WBRT or craniotomy.