Statistical model to predict brain metastasis risk in patients with early-stage breast cancer

Abstract

Background: Breast cancer (BC) brain metastases (BCBM) are emerging as a major factor of morbidity and poor overall survival. We hypothesized that a clinical risk prediction model could predict BCBM and allow the selection of an enriched early-stage BC patient population, who are at higher risk for BCBM, for preventive trials. Method(s): Electronic medical records were retrospectively reviewed for patients diagnosed and treated for early-stage BC atMDAnderson Cancer Center between 1997 and 2014 under a study approved by the institutional review board. The clinicopathologic prognostic features selected for analysis are: age, HER-2 receptor status and hormone receptor (HR) status, tumor histology, grade, and stage, menopausal status, vascular and lymphatic invasion. A multivariate Cox proportional hazards regression analysis was conducted. Result(s): A total of 15164 patients with complete data for key variables were studied. Patients were randomly split 2:1 into training and validation sets. Of the 10026 patients in the training set, 317 developed BCBM and of the 5138 in the validation set, 133 developed BCBM. The 10-year estimated risk of brain metastasis was 4.2% (95% CI, 3.7% to 4.7%) in the training set. Younger age, HER-2 negative and HR-negative receptor status, higher tumor stage and grade, were all significantly and independently associated with BCBM. The risk prediction model had an estimated Harrell's concordance index of 81% (95%CI, 77% to 86%) in the validation set. In the 10% of validation set patients predicted to have the highest risk, the 10-year risk of brain metastasis was 15% (95% CI, 11% to 18%). Conclusion(s): This risk prediction model for brain metastasis risk in early BC allows us to: (1) Identify the significant clinical risk factors for BCBM, (2) use these risk factors to develop an individualized risk score for BCBM, and (3) use this score to select patients at higher risk of BCBM to be prioritized for preventive trials.