Pediatric Large Cell/Anaplastic Medulloblastomas: Diagnostic Concordance and Multi-Classifier DNA Methylation...

Large cell/anaplastic medulloblastoma is a high-risk variant whose histopathological diagnosis can be challenging. We analyzed 12 pediatric cases enriched for LC/A morphology, with centralized review available for 11. In this deliberately selected, diagnostically challenging series, institutional and central LC/A versus non-LC/A assessments were discordant in 36% (4 of 11), illustrating the difficulty of applying the LC/A designation rather than a general estimate of inter-observer reproducibility. DNA methylation profiling was performed using four classifiers: Heidelberg CNS Tumor Methylation Classifier, Methylscape Analysis, EpiDiP, and the MAGIBU classifier, which applies a UMAP-based distance approach. Tumor classification was largely concordant across platforms, identifying two WNT-activated medulloblastomas, eight non-WNT/non-SHH medulloblastomas, and one tumor reclassified as an embryonal tumor with multilayered rosettes. One misclassification occurred with EpiDiP, where an SHH-activated medulloblastoma aligned with the normal cerebellar cortex reference profile. Copy number analysis revealed MYC amplification in two non-WNT/non-SHH tumors, chromosome 6 loss in both WNT-activated cases, and isochromosome 17q in one MYC-amplified non-WNT/non-SHH tumor. Adjunctive molecular data for a subset of cases were obtained from PedcBioPortal. The reclassified tumor lacked canonical molecular alterations but harbored additional CTNNB1 and TP53 variants whose contribution to its pathogenesis remains uncertain. One WNT-activated medulloblastoma also carried a CTNNB1 variant. The MAGIBU classifier correctly identified the ETMR case, classified all remaining tumors as medulloblastomas, and showed concordant subgroup assignment with the Heidelberg CNS Tumor Methylation Classifier and Methylscape Analysis, including the SHH-activated case misclassified by EpiDiP. On external validation across four independent medulloblastoma cohorts (n = 1063), MAGIBU achieved 98.5% family-level recall and 96.4% subclass-level recall. In conclusion, our findings indicate that, in these diagnostically challenging, LC/A-enriched cases, morphology-based diagnostic assessment can be inconsistent. DNA methylation profiling serves to refine tumor reclassification and ensure accurate risk stratification. Interpreting each tumor across complementary classifiers provides a triangulation that enhances diagnostic robustness.