Should checkpoint inhibitors be reserved for biomarker-selected pediatric brain tumors?

Abstract Purpose Pediatric brain tumors, including high-grade gliomas (HHG), medulloblastomas (MB), and ependymomas (EPN), are a leading cause of death in children. They are often immunologically “cold” with low tumor mutational burden (TMB) and very few tumor-infiltrating lymphocytes (TILs), which may limit the role of immune checkpoint inhibitors (ICI). Methods We performed a PRISMA‑guided systematic review of PubMed/MEDLINE, Embase, and Scopus from inception to September 17, 2025, for English-language studies of patients ≤ 21 years with primary CNS tumors treated with PD‑1/PD‑L1 or CTLA‑4 inhibitors. Eligible reports included prospective trials, retrospective series, and observational/case reports with extractable data on efficacy and/or toxicity. Of 479 records identified, 386 unique citations were screened, 127 underwent full‑text review, and 40 met inclusion criteria for qualitative synthesis. Results Prospective and institutional studies in biomarker-unselected diffuse midline glioma, high-grade glioma, medulloblastoma, and ependymoma showed low objective response rates (generally ≤ 6%), short median progression-free survival (1–3 months), and overall survival similar to historical controls, despite acceptable safety (grade ≥ 3 treatment-related adverse events ~ 15–25% with anti-PD-1 ± anti-CTLA-4). In contrast, across germline MMR-deficient and broader RRD/hypermutant cohorts, PD-1 blockade produced clinically meaningful and sometimes durable responses, including complete remissions in malignant glioma and approximate 2-year overall survival near 50%, often with delayed responses and pseudoprogression. Histology-specific profiling highlighted marked variation in immune contexture: pediatric gliomas segregate into immune “hot,” “altered,” and “cold” subtypes; medulloblastoma is largely PD-L1–low with prominent B7-H3 and myeloid programs; checkpoint expression is also observed in germ cell and selected sellar tumors. Evidence quality is limited by small, heterogeneous, predominantly non-comparative designs. Conclusion ICIs show manageable safety but limited efficacy in unselected pediatric CNS tumors. Durable benefit is most evident in RRD/hypermutant biology and possibly PD-L1–high niches (e.g., some low-grade gliomas and CNS-GCT). Future trials should be biomarker-driven and pair ICIs with priming combinations (e.g., radiation, epigenetic modulators, metronomic chemotherapy) to convert “cold” tumors into responders.