P07.10.B INTER-RATER AGREEMENT FROM A MULTI-NATIONAL EXPERT PANEL ON RESECTABILITY OF RECURRENT GLIOBLASTOMA: A COMPLEMENTARY POST-HOC ANALYSIS OF THE PROSPECTIVE, RANDOMIZED DIRECTOR TRIAL

Abstract BACKGROUND The role of re-resection in recurrent glioblastoma remains controversial. Associations between higher extents of re-resection and prolonged survival might be explained by the assumption that tumors in less functional (surgically more accessible) brain areas identify with an inherently better prognosis. We used the DIRECTOR cohort on recurrent glioblastoma to (I) assess heterogeneity in the expected ‘resectability’ among surgical experts and (II) explore whether ‘resectability’ or residual tumor is associated with survival. MATERIAL AND METHODS DIRECTOR was a randomized, prospective multicenter trial assessing two temozolomide regimens for first glioblastoma relapse. Patients could have a re-resection per centers’ preference. As no outcome differences between the two study arms were encountered, we pooled the arms to achieve a homogenously treated cohort. For each IDH-wildtype glioblastoma with baseline MRI available, eleven surgical neuro-oncologists rated tumor resectability. Cohen’s kappa estimated inter-rater agreement. Ratings on resectability and actual postoperative tumor were correlated with survival. RESULTS We studied 69 patients with first glioblastoma recurrence, including 40 individuals who underwent re-resection prior to temozolomide re-challenge. In patients who received a re-resection, a ‘meaningful resection’ was deemed feasible by 0-50% of the raters in 4/40 cases (10.0%), by 51-80% in 6/40 cases (15.0%), and by 81-100% in 30/40 cases (75.0%). Among patients without post-operative contrast-enhancing tumor, pre-operative agreement that tumors were resectable was high (agreement by >81-100% of raters: 23/24 cases, 95.8%). In patients who were managed without re-resection, in only 3/29 patients (10.4%) all raters agreed that tumors were not resectable; while 1-50% of the raters judged tumors as resectable in 14/29 patients (48.3%), 51-80% in 7/29 patients (24.1%), and 81-100% in 5/29 patients (17.2%). Knowledge of clinical factors virtually never changed the initial judgment based on MRI alone. Surgical decision-making varied between raters, ranging from 29 to 59/69 cases being classified as ‘resectable’ (kappa: 0.405). In the overall cohort, consensus on resectability by >80% of the raters was not associated with favourable post-recurrence survival (10.4 ± 1.8 vs 10.0 ± 3.0 months, p = 0.924) or progression-free survival (2.0 ± 0.5 vs 1.9 ± 0.2 months, p = 0.757). Compared to submaximal resection or no re-resection, patients with complete resection of the enhancement had longer survival, particularly in MGMT unmethylated tumors. CONCLUSION Assessment of feasibility for re-resections in recurrent glioblastoma substantially varies between surgical experts. An objective consensus on resectability remains undefined, with major implications for practice and trial stratification. ACKNOWLEDGMENT DIRECTOR investigators.