Real-world performance of the breast GPA in breast cancer brain metastases in a resource limited setting

The Breast Graded Prognostic Assessment (B-GPA) is a validated prognostic score used to estimate survival in patients with breast cancer brain metastases (BCBM). However, real-world evidence assessing its prognostic accuracy, particularly in low- and middle-income countries (LMICs), remains limited. This study aimed to externally evaluate the performance of B-GPA by comparing predicted survival estimates with observed outcomes in a real-world cohort. This retrospective cohort study included patients with invasive breast cancer and BCBM who underwent neurosurgical resection at the Brazilian National Cancer Institute between January 2015 and December 2022. The performance of B-GPA in predicting overall survival (OS) was evaluated alongside sociodemographic and clinicopathological variables. Among 85 eligible patients, 17 (20%) had B-GPA scores of 0.0–1.0, 29 (34%) of 1.5–2.0, 30 (35%) of 2.5–3.0, and 9 (11%) of 3.5–4.0. The observed median OS across these subgroups was 6.0 months (95% CI, 2.8–11.3), 6.0 months (95% CI, 3.1–14.0), 11.9 months (95% CI, 9.8–20.4), and 21.0 months (95% CI, 18.4-not reached), respectively. For the entire cohort, the median predicted OS according to the B-GPA was 12.9 months (95% CI, 12.9–23.5), whereas the observed post-resection OS was 9.9 months (95% CI, 6.6–12.1). Age < 60 years, diabetes, HER2-negative status, and high Ki-67 status were associated with worse prognosis. In this real-world cohort, survival after BCBM was shorter than that predicted by B-GPA. These findings suggest that prognostic estimates derived from high-income settings may not fully reflect outcomes observed in some LMIC real-world populations, underscoring the importance of further evaluation of factors that may influence model performance across different healthcare systems.