The melanoma-specific graded prognostic assessment does not adequately discriminate prognosis in a modern population with brain metastases from malignant melanoma

Background: The melanoma-specific graded prognostic assessment (msGPA) assigns patients with brain metastases from malignant melanoma to 1 of 4 prognostic groups. It was largely derived using clinical data from patients treated in the era that preceded the development of newer therapies such as BRAF, MEK and immune checkpoint inhibitors. Therefore, its current relevance to patients diagnosed with brain metastases from malignant melanoma is unclear. This study is an external validation of the msGPA in two temporally distinct British populations. Methods: Performance of the msGPA was assessed in Cohort I (1997-2008, n = 231) and Cohort II (2008-2013, n = 162) using Kaplan-Meier methods and Harrell's c-index of concordance. Cox regression was used to explore additional factors that may have prognostic relevance. Results: The msGPA does not perform well as a prognostic score outside of the derivation cohort, with suboptimal statistical calibration and discrimination, particularly in those patients with an intermediate prognosis. Extra-cerebral metastases, leptomeningeal disease, age and potential use of novel targeted agents after brain metastases are diagnosed, should be incorporated into future prognostic models. Conclusions: An improved prognostic score is required to underpin high-quality randomised controlled trials in an area with a wide disparity in clinical care.