Prognostic factors for breast cancer brain metastasis and primary construction of a prediction model

To investigate the prognostic factors of breast cancer brain metastasis（BCBM）, and to construct and validate a prediction model for prognosis of BCBM patients.

A case-control trial was designed and conducted in this study. A total of 1 133 BCBM patients were screened out from the Surveillance, Epidemiology, and End Results（SEER）database from 2016 to 2020, and were randomly assigned to a training set（n=793）and an internal validation set（n=340）at a ratio of 7∶3. The prognostic factors were determined with univariate and multivariate Cox regression analyses, and a nomogram prediction model was constructed. The discrimination, prediction and calibration of the nomograms were evaluated by C-index, receiver operating characteristic（ROC）curve and calibration curve. External validation of the nomogram prediction model was performed in 73 BCBM patients admitted to our hospital from 2004 to 2019.

Multivariate Cox regression analysis revealed that age, molecular type, surgery, radiotherapy, chemotherapy, lung metastasis and liver metastasis were identified as independent factors（P<0.05）, and a prediction model of the nomogram was thus established. The C-index of the SEER training set, internal validation set and external validation set were 0.720（95%CI: 0.674~0.766）, 0.705（95%CI: 0.633~0.777）and 0.667（95%CI: 0.610~0.724）, respectively. The ROC curve of the training set showed that our nomogram could accurately predict the survival rate of BCBM patients at 6 months（AUC=0.793）, and 1（AUC=0.759）and 2 years（AUC=0.724）. The predictive values of the calibration curve are in good agreement with the observed values.

Age, molecular type, surgery, radiotherapy, chemotherapy, and lung and liver metastases are independent prognostic factors of BCBM. Our constructed nomogram can effectively predict the 6-month, 1-year and 2-year survival probability of BCBM patients.