RADT-47. EXPLORING A PERSONALIZED CLINICAL TARGET VOLUME MARGIN IN GLIOBLASTOMA RADIOTHERAPY USING RECURRENCE PATTERNS IN A NATIONWIDE COHORT

Abstract BACKGROUND This nationwide recurrence pattern study aims to investigate the potential of clinical target volume (CTV) margin differentiation in radiotherapy for newly diagnosed glioblastoma. METHODS All adults newly-diagnosed with glioblastoma between 2015-2019 in Denmark, planned to start long-course chemoradiotherapy, with a dedicated radiotherapy planning-MRI, were included (n=756). Patients with a confirmed IDH mutation were excluded (n=24). In total, 492 patients were available for analyses (treatment response not-evaluable n=82, no-radiographic progression n=62, CSF/meningeal/hematological dissemination n=28, technical error n=68). The gross tumor volume (GTV) and contrast-enhancing recurrence (following RANO-criteria) were automatically segmented on the respective MRIs using AI. Each segmentation was manually checked and adapted if needed. Imaging and radiotherapy dose distribution were rigidly co-registered. With respect to the 95% radiotherapy isodose of the clinical plans, recurrences were classified as central/in-field/marginal/distant in 414/22/21/35 patients, respectively. The GTV was expanded in 1mm steps to simulate CTVs. The CTV expansion needed to cover 95% of the most distant recurrence volume was compared between patient subgroups using the Kruskal-Wallis and Mann-Whitney-U test where appropriate. RESULTS A 9/25/59mm simulated CTV margin covered the most distant recurrence lesion in 50/75/95% of the patients, respectively. Comparing patients with a uni- (n=379) to multifocal GTV (n=113), the simulated CTV margin to cover the recurrences in 75% of the patients was 17 vs, 40mm (p<.001). Simulated CTV margins were also significantly different for extent of resection, however, no longer after stratifying for uni- and multifocal GTV. Patient age, sex, & performance status, tumor location & crossing midline, and MGMT methylation status did not result in significantly different simulated GTV-to-CTV margins, also not after uni- and multifocal GTV stratification. CONCLUSIONS There was a substantial and significant difference in simulated CTV margins when comparing uni- vs. multifocal GTVs, but not for MGMT promotor methylation status or extent of surgery.