PURPOSE: Conventional wisdom has rendered patients with brain metastases ineligible for clinical trials for fear that poor survival could mask the benefit of otherwise promising treatments. Our group previously published the diagnosis-specific Graded Prognostic Assessment (GPA). Updates with larger contemporary cohorts using molecular markers and newly identified prognostic factors have been published. The purposes of this work are to present all the updated indices in a single report to guide treatment choice, stratify research, and define an eligibility quotient to expand eligibility. METHODS: A multi-institutional database of 6,984 patients with newly diagnosed brain metastases underwent multivariable analyses of prognostic factors and treatments associated with survival for each primary site. Significant factors were used to define the updated GPA. GPAs of 4.0 and 0.0 correlate with the best and worst prognoses, respectively. RESULTS: Significant prognostic factors varied by diagnosis and new prognostic factors were identified. Those factors were incorporated into the updated GPA with robust separation ( P < .01) between subgroups. Survival has improved, but varies widely by GPA for patients with non–small-cell lung, breast, melanoma, GI, and renal cancer with brain metastases from 7-47 months, 3-36 months, 5-34 months, 3-17 months, and 4-35 months, respectively. CONCLUSION: Median survival varies widely and our ability to estimate survival for patients with brain metastases has improved. The updated GPA (available free at brainmetgpa.com) provides an accurate tool with which to estimate survival, individualize treatment, and stratify clinical trials. Instead of excluding patients with brain metastases, enrollment should be encouraged and those trials should be stratified by the GPA to ensure those trials make appropriate comparisons. Furthermore, we recommend the expansion of eligibility to allow for the enrollment of patients with previously treated brain metastases who have a 50% or greater probability of an additional year of survival (eligibility quotient > 0.50).

Although machine learning (ML) has shown promise across disciplines, out-of-sample generalizability is concerning. This is currently addressed by sharing multi-site data, but such centralization is challenging/infeasible to scale due to various limitations. Federated ML (FL) provides an alternative paradigm for accurate and generalizable ML, by only sharing numerical model updates. Here we present the largest FL study to-date, involving data from 71 sites across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, reporting the largest such dataset in the literature (n = 6, 314). We demonstrate a 33% delineation improvement for the surgically targetable tumor, and 23% for the complete tumor extent, over a publicly trained model. We anticipate our study to: 1) enable more healthcare studies informed by large diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further analyses for glioblastoma by releasing our consensus model, and 3) demonstrate the FL effectiveness at such scale and task-complexity as a paradigm shift for multi-site collaborations, alleviating the need for data-sharing.