MEDB-31. THE CLINICAL SIGNIFICANCE OF EXTENT OF RESECTION IN MEDULLOBLASTOMA

Abstract

Medulloblastoma (MB) patients determined to have a sub-total resection (STR), defined by >1.5cm2 post-surgical tumour residuum, receive intensified treatment regimes, but recently the designation of STR as a high risk feature is being questioned. We aimed to assess the clinical correlates of extent of resection (EOR) and its impact on survival, with particular consideration of EOR in relation to the four MB consensus molecular subgroups (WNT, SHH, Group 3, Group 4). We collected data from 1113 patients (n=419, UK CCLG institutions; n=694, published data) representing the largest ever combined cohort constructed to assess the impact of EOR in medulloblastoma. We performed association analyses and univariate/multivariate survival analysis using Kaplan-Meier, log-rank and Cox proportional hazard modelling, analysing overall survival (OS) cohort-wide and with reference to molecular subgroups and clinical features. Association analysis of the combined cohort evidenced that infant patients were more likely to have STR (p=0.02). In this whole-cohort analysis, EOR was significantly associated with survival in univariate analysis (HR 1.64, 95% CI 1.30-2.07, p=<0.001) but not in multivariate analysis. STR was variably prognostic in sub-cohort analyses of specific demographics and molecular subgroup; worse outcomes were observed in patients <5 years in SHH (p=0.044) and Group 4 (p=0.044). This was true for WNT patients >5 years old at diagnosis (p=0.034) although numbers were small and require validation in even larger cohorts. In this cohort of >1100 MBs, STR was significantly associated with a lower OS in univariate analysis, but this was driven by specific disease contexts (SHH and Group 4 patients <5 years old). STR was not independently prognostic overall or in any setting. We recommended that surgeons should continue to pursue maximal safe resection for all MB patients but suggest that consideration of STR as a high-risk feature should be disease context specific.