Predictive factors of brain metastasis in patients with breast cancer

Abstract

The aim of this study is to determine the risk factors associated with metastasis to the brain of primary breast cancer patients and evaluate a predictive model. The clinicopathological characteristics of 206 patients with primary breast cancer were analyzed retrospectively with a univariate and multivariate logistic regression model. A predictive model was generated, and its validity evaluated with a receiver operating characteristic (ROC) curve. Independent risk factors for brain metastasis in patients with primary breast cancer were: being younger than 35 years old at the time of diagnosis, having four or more metastatic axillary nodes, being estrogen receptor-negative, and with 24 months of metastasis-free survival. The predictive value of the brain metastasis risk model, measured as the area under the ROC curve, was 0.765 ± 0.040 (95 % CI 0.688-0.842). When 0.8 was considered the cutoff point of probability calculated by the model, the sensitivity and specificity for predicting the occurrence of brain metastases in these patients were 0.769 and 0.713, respectively. The predictive model constructed in this study can be used to forecast brain metastasis in breast cancer. Patients with a predictive level ≥0.8 could be treated preventively for brain metastases.