JS04.6.A IMPORTANCE OF NEUROANATOMICAL SUBSITES IN DEFINING PATTERNS OF INFILTRATION AND RECURRENCE IN GLIOBLASTOMA: DEVELOPING EVIDENCE TO CHALLENGE UNIFORM MARGIN EXPANSION FOR RT TARGET VOLUME DELINEATION.

Abstract BACKGROUND RT target volume delineation protocols for Glioblastoma(GBM) utilise uniform expansion around the surgical cavity/residual tumour, without consideration of adjacent sites at more risk for infiltration. However neuroanatomical subsites may have specific regions where tumour infiltration occurs, especially related to adjacent white matter tracts(WMT). This study categorises the infiltration/relapse pattern for three neuroanatomical sites as a means for developing future alternative target volume delineation protocols. METHODS Consecutive patients managed for GBM with EORTC-NCIC Protocol(60Gy) between 2008-2023 were identified. Tumours centred on anterior temporal(TempAnt), occipital(Occ) and posterior parietal(PostPar) lobes were initially selected for analysis. Based on assessment of T1-gadolinium-enhanced/T2-weighted MRI sequences adjacent sites of tumour infiltration and subsequent relapse was assessed. These sites were categorised into zones reflecting increasing extension of tumour away from initial involved gyrus, and scored in regards to proportional rate of involvement. Adjacent WMT involvement was recorded. Primary endpoint was relapse-free survival(RFS)and rate of distant relapse as defined by >20mm from surgical cavity/residual tumour. Involvement at relapse of each defined regional zone was assessed for the three neuroanatomical sites. RESULTS 497 patients were managed with median overall survival of 17.4 months(95%CI: 15.1-19.9 months). Fifteen neuroanatomical sites were used for categorisation of which 10% were AntTemp; 9% Occ and 11% PostPar. Infiltration pattern was consistent for each site, with predictable zones identified based on initial involved gyrus. For example with AntTemp the temporal stem/insular involvement was evident in 57% for T1gd and 94% for T2 signal; however hippocampal T2 signal seen in only 21% of patients. For Occ subsite the trigone of lateral ventricle was involved in 48%; varying for medial vs lateral based tumours. PostPar subsite had initial cingulate gyrus, splenium/parahippocampal involvement in 70% at diagnosis. Similarly at relapse the patterns were predictable with consistent zonal and adjacent WMT involvement. Median RFS for whole cohort was 11.9 months(95%CI: 10.6-13.3). In AntTemp 73% relapse sites involved external/extreme capsule or frontal lobe; and 87% had evidence of relapse involving WMT. For Occ lateral tumours commonly relapsed at trigone(69%vs50%) and anterior temporal region(46%vs15%); with medial tumours involving the splenium(40%vs11%). For PostPar distant relapse was evident in 53% with distal sites along Cingulum pathway. CONCLUSION GBM infiltration and relapse pathways occurred in predictable distribution around initial neuroanatomical site. RT potentially may be individualised for neuroanatomical subsite and non-isotropic target volume expansion.