Abstract
BACKGROUND: Throughout the past decade, molecular tools have revolutionised ependymoma classification, yet their application has been limited in clinical trial cohorts. METHODS: We present the molecular analysis of paediatric ependymoma patients from the E-HIT2000 trial (n = 165), or subsequent HIT2000-interim and I-HIT-MED registries (n = 79) based on DNA-methylation profiling with multi-level classification using the Heidelberg Brain Tumor Classifier v12.5 and the analysis of chromosome 1q, 6q, and CDKN2A copy number status. Prior to further analysis, cases without ependymoma group prediction were excluded (n = 16/244). RESULTS: 5-year PFS and OS rates varied by group: EPN-PFA (n=146/228) 45±4% and 76±4%; EPN-PFB (n=19/228) 90±7% and 100%; EPN-ZFTA (n=59/228) 63±7% and 87±5%; and EPN-YAP1 (n=4/228) 50±25% and 100%. EPN-PFA was subclassified into nine subtypes. Poor median PFS was observed for PFA1c (n=22, 5-yr-PFS 32±10%/5-yr-OS 77±9%), PFA1d (n=13, 35±14%/72±14%), PFA1e (n=15, 47±13%/59±13%), and PFA2a (n=16, 27±12%/50±14%). Presence of 1q gain was associated with inferior survival (n = 39/146, 29±8%/65±8%). 6q loss impacted PFS, but not OS (n=8, 25±14%/100%). Among 59 EPN-ZFTA cases, 46 clustered with EPN ZFTA-RELA fused, and 12 clustered within recently described subgroups of EPN-ZFTA (EPN-ZFTA alt., Cluster 1: 1/12, Cluster 2: 11/12). Strikingly, 6/11 EPN ZFTA-alt. (Cluster 2) cases showed infratentorial (4/6) or supratentorial midline location (2/6). Homozygous deletion of CDKN2A (CDKN2A-/-; n=8) was only observed in EPN ZFTA-RELA. Presence of EPN ZFTA-alt. (cluster 2) (5-yr-PFS: 36±15%; p<0.001), and CDKN2A-/- (19±16%; p<0.0001) was associated with inferior PFS compared to EPN ZFTA-RELA CDKN2A-wt (81±6%). OS was only impacted for patients with EPN ZFTA-RELA CDKN2A-/- (5-yr-OS: 38±20%, p>0.0001). CONCLUSIONS: We present strong evidence supporting the inclusion of molecular criteria into the next generation of clinical trials for children with ependymoma.