Abstract
BACKGROUND: Early detection of brain metastasis (BM) is essential for prognostic improvement in breast cancer (BC) patients. The aim was to identify predictors of BCBM in different molecular subtypes on a population-based level. METHODS: The Surveillance, Epidemiology, and End Results database was used to select BC patients diagnosed from 2010 to 2018. We evaluated the incidence and risk factors of BCBM and tested the interaction effects between molecular subtypes and other risk factors. RESULTS: Among the 527,525 selected patients, molecular subtypes significantly interacted with T stage and extracranial metastasis (ECM) patterns on the risk of BM in the whole BC population (interaction p = 0.002, <0.001, respectively) and after excluding patients with unknown states of key factors. BM development was independent of the T stage only in HR-/HER2- patients (trend p = 0.126). We selected BC patients with single-organ ECM and found a significant interaction between molecular subtypes and ECM patterns (interaction p = 0.013). The impact of ECM patterns on the risk of BM was limited to HR-/HER2- patients (trend p < 0.001), for whom using bone metastasis as a reference, lung metastasis increased the risk of BM (OR = 1.936, 95% CI: 1.300-2.882, p = 0.001). CONCLUSION: T stage and ECM patterns had different associations with BM in different molecular subtypes. HR-/HER2- BC had distinct features on BM development, manifested as a lack of tumor size effect and is associated with lung metastasis. Close surveillance for BM should be considered for HR-/HER2- BC patients.