BackgroundBreast cancer brain metastasis (BCBM) is associated with poor survival outcomes and reduced quality of life. The Graded Prognostic Assessment (GPA) score model serves as a well-established tool for predicting the prognosis of BCBM. Notably, the presence of extracranial metastasis (ECM) is considered as a significant prognostic factor in the breast GPA model. This study aims to further refine other features of ECM to enhance the prognostic prediction for BCBM.MethodsThis study included all inpatients diagnosed with BCBM at the Cancer Hospital, Chinese Academy of Medical Sciences, from January 2010 to July 2021. Baseline characteristics of patients were compared based on features of ECM, including the presence, number, location, and control status of metastases. Overall survival (OS) were compared using the Kaplan–Meier method with log-rank tests. Cox regression analyses were conducted to identify significant prognostic factors. The aforementioned ECM features were incorporated into the original Breast-GPA model to enhance its prognostic accuracy. The concordance index (C-index) and restricted mean survival time (RMST) were utilized to evaluate and compare the predictive accuracy of the updated and original survival models.Results284 patients with BCBM were included in the study. Kaplan–Meier survival curves suggested that patients without ECM when diagnosed with BCBM showed better survival (p = 0.007). In the subgroups with ECM, more than 3 organs involved, both bone and visceral metastasis and progressive ECM portended dismal OS (p = 0.003, 0.001 and <0.001). Multivariate analysis demonstrated that molecular subtype, presence of ECM, and number of brain metastasis significantly influenced OS after BCBM. By modifying the current GPA model to include more precise characteristics of ECM, the predictive accuracy was further enhanced as indicated by the C-index and RMST curve.ConclusionsMore ECM sites, both bone and visceral invasion and uncontrolled ECM were dismal prognostic factors for survival outcomes of BCBM patients. A new Breast-GPA model with better predictive effect was constructed.