Abstract
BACKGROUND: This retrospective study sought to define the incidence of brain metastases as a first site of recurrence among women with triple receptor-negative breast cancer (TNBC). METHODS: A total of 2448 patients with stage I through III TNBC who were diagnosed between 1990 and 2010 were identified. We computed the cumulative incidence of developing brain metastases as a first site of recurrence at 2 and 5 years. Cox proportional hazards models were fitted to determine factors that could predict for the development of brain metastases as a first site of recurrence. The Kaplan-Meier product limit method was used to compute survival following a diagnosis of brain metastases. RESULTS: At a median follow-up of 39 months, 115 (4.7%) patients had developed brain metastases as a first site of recurrence. The cumulative incidence at 2 and 5 years was 3.7% (95% confidence interval [CI] = 2.9%-4.5%) and 5.4% (95% CI = 4.4%-6.5%), respectively. Among patients with stage I, II, and III disease, the 2-year cumulative incidence of brain metastases was 0.8%, 3.1%, and 8%, respectively (P < .0001). The 5-year cumulative incidence was 2.8%, 4.6%, and 9.6% among patients with stage I, II, and III disease, respectively (P < .0001). In the multivariable model, patients with stage III disease had a significant increase in the risk of developing brain metastases as a first site of recurrence (hazards ratio = 3.51; 95% CI = 1.85-6.67; P = .0001) compared to patients with stage I disease. Those with stage II disease had a nonsignificant increased risk of developing brain metastases as a first site of recurrence (hazards ratio = 1.61; 95% CI = 0.92-2.81; P = .10) compared with patients with stage I disease. Median survival following a diagnosis of brain metastases was 7.2 months (range, 5.7-9.4 months). CONCLUSIONS: Patients with nonmetastatic TNBC have a high early incidence of developing brain metastases as a first site of recurrence, which is associated with subsequent poor survival. Patients with stage III TNBC in particular would be an ideal cohort in which to research preventive strategies.