Graded Prognostic Assessment (GPA) for brain metastases

What is measured: expected survival after a diagnosis of brain metastases, by primary tumour. How: age, Karnofsky performance status, the number of metastases on contrast MRI, extracranial disease and, in the molecular versions, the tumour's driver status: EGFR and ALK (and PD-L1 in the 2021 update) for lung adenocarcinoma, BRAF for melanoma and receptor subtype for breast cancer, scored in half-points to a maximum of 4. In lung adenocarcinoma a score of 3.5 to 4 predicts a median survival near four years against about seven months for 0 to 1 (Sperduto and colleagues). Brain metastasis velocity (new lesions per year after radiosurgery) predicts the need for salvage, and the older RPA classes are the ancestor. What a result changes: a poor score (under 1.5) argues for best supportive care or whole-brain radiotherapy, which in the QUARTZ trial did not lengthen life over steroids alone in poor-prognosis lung cancer, and against surgery; a good score supports surgery for large or symptomatic lesions with radiosurgery to the cavity, radiosurgery rather than whole-brain treatment for up to ten or more metastases to preserve cognition, hippocampal-avoidance whole-brain radiotherapy with memantine when the whole brain must be treated, and, in asymptomatic patients with a brain-active drug available (osimertinib, lorlatinib, tucatinib, trastuzumab deruxtecan, ipilimumab with nivolumab), systemic therapy first; it also sets trial eligibility. Where it matters: secondary brain tumours and HER2-positive breast cancer brain metastases.