Prognostic Stratification of Brain Metastases from Nonanaplastic Follicular Cell–Derived Thyroid Carcinoma: Results of an International Multicenter Retrospective Study

Background: Brain metastases (BMs) from nonanaplastic follicular cell–derived thyroid carcinoma (hereinafter referred to as non-ATC) are rare and not well characterized, with limited evidence guiding treatment and prognostic stratification. We aimed to describe clinical features, treatment patterns, and outcomes of patients with BMs from non-ATC and to develop and validate a disease-specific prognostic score. Methods: This retrospective international multicenter study included adults with histologically confirmed non-ATC and radiologically documented BMs treated at 19 tertiary referral centers. Clinical, radiological, treatment, and outcome data were collected. Independent predictors of progression-free survival (PFS) and overall survival (OS) were analyzed using Cox regression, and variables identified were used to develop a disease-specific Graded Prognostic Assessment (non-ATC GPA), exploratorily assessed in an independent SEER cohort. Results: A total of 189 patients were included; median age at BM diagnosis was 59 years (interquartile range [IQR] 48–67). First-line BM treatments included stereotactic radiosurgery (27.0%), neurosurgery (23.8%), whole-brain radiotherapy (18.5%), and best supportive care (10.6%). Median PFS was 7.0 months (IQR: 2.9–22.4), and median OS was 15.0 months (IQR: 4.0–38.4). Independent predictors of shorter OS were age >75 years, Eastern Cooperative Oncology Group performance status ≥2, papillary histology, extracranial metastases, and ≥4 BM. The GPA stratified patients into four prognostic groups with median OS ranging from 72.6 months in the highest-score group to 3.7 months in the lowest. An exploratory SEER-based assessment showed survival differences across adapted prognostic groups ( p = 0.036), although this analysis was limited by the absence of all key GPA variables. Conclusions: BMs from non-ATC are associated with poor survival and marked prognostic heterogeneity. The proposed GPA may provide a useful framework for prognostic stratification, pending further validation in independent datasets to support individualized management.