Validation and Development of a Modified Breast Graded Prognostic Assessment As a Tool for Survival in Patients With Breast Cancer and Brain Metastases

Abstract

Purpose: Several indices have been developed to predict overall survival (OS) in patients with breast cancer with brain metastases, including the breast graded prognostic assessment (breast-GPA), comprising age, tumor subtype, and Karnofsky performance score. However, number of brain metastases-a highly relevant clinical variable-is less often incorporated into the final model. We sought to validate the existing breast-GPA in an independent larger cohort and refine it integrating number of brain metastases.

Patients and methods: Data were retrospectively gathered from a prospectively maintained institutional database. Patients with newly diagnosed brain metastases from 1996 to 2013 were identified. After validating the breast-GPA, multivariable Cox regression and recursive partitioning analysis led to the development of the modified breast-GPA. The performances of the breast-GPA and modified breast-GPA were compared using the concordance index.

Results: In our cohort of 1,552 patients, the breast-GPA was validated as a prognostic tool for OS (P < .001). In multivariable analysis of the breast-GPA and number of brain metastases (> three v ≤ three), both were independent predictors of OS. We therefore developed the modified breast-GPA integrating a fourth clinical parameter. Recursive partitioning analysis reinforced the prognostic significance of these four factors. Concordance indices were 0.78 (95% CI, 0.77 to 0.80) and 0.84 (95% CI, 0.83 to 0.85) for the breast-GPA and modified breast-GPA, respectively (P < .001).

Conclusion: The modified breast-GPA incorporates four simple clinical parameters of high prognostic significance. This index has an immediate role in the clinic as a formative part of the clinician's discussion of prognosis and direction of care and as a potential patient selection tool for clinical trials.

Fig 1.

Kaplan-Meier overall survival (OS) curves for breast graded prognostic assessment (breast GPA) in MD Anderson validation cohort. Table lists 1-, 2-, and 3-year OS estimates and median OS in months by breast GPA group (concordance index, 0.78; 95% CI, 0.77 to 0.80).

Fig 2.

Recursive partitioning analysis model for modified breast graded prognostic assessment. Internal nodes presented in circles; terminal nodes presented in rectangles. Top number within each node is number of patients in node; bottom number is median overall survival (OS) in months and 95% CI. Table lists 1-, 2-, and 3-year OS estimates and median OS in months in terminal nodes. HER2, human epidermal growth factor receptor 2; KPS, Karnofsky performance score.

Fig 3.

Kaplan-Meier overall survival (OS) curves for modified breast graded prognostic assessment (modified breast-GPA) in MD Anderson cohort. Table lists 1-, 2-, and 3-year OS estimates and median OS in months by breast GPA group (concordance index, 0.84; 95% CI, 0.83 to 0.85).