Individualized prediction of survival after brain metastasis development in patients with breast cancer.

BACKGROUND

Individuals with breast cancer are at high risk for developing brain and central nervous system metastases (BrM), which are aggressive and clinically challenging. This study aimed to assess the impact of clinico-genomic factors on BrM survival in breast cancer and to develop a validated nomogram for survival probability.

METHODS

Clinico-genomic data on breast cancer patients were obtained from the AACR GENIE Biopharma Collaborative. These data were split into training and testing sets by academic institution. Using data from the training set, BrM survival was evaluated through multivariable Cox proportional hazards modeling, with LASSO penalization for feature identification. Performance metrics were assessed internally on the training set through bootstrapcross-validation, and externally on the testing set. Calibration curves, AUC, and Brier scores were generated to evaluate model performance. A nomogram for 12-, 18-, and 24- month survival probability was constructed.

RESULTS

HER2+/HR+ subtype, ARID1A alterations, and a lower number of prior cancer-directed regimens were top predictor associated with improved survival after BrM development. This model showed strong discriminate capability at 12- (AUC: 79.7; Brier: 18.2), 18- (AUC: 78.5, Brier: 19.9) and 24-months (AUC 82.0, Brier: 17.9). After external validation, the model performed similarly at 12- (AUC: 73.5; Brier: 21.1), 18- (AUC: 76.2; Brier 20.1), and 24-months (AUC: 79.5; Brier: 17.9).

CONCLUSIONS

Nomograms that calculate individualized BrM survival probabilities for patients with breast cancer offer clinical value by informing scientists and healthcare professionals about risk factors for survival after BrM diagnosis. A webtool for individualized BrM survival probabilities can be found here: https://gcioffi.shinyapps.io/BrCA_nomo_brm_survival/