Discrepancies in Inter-Rater Agreement on Resectability of Recurrent Glioblastoma: Complementary Post-hoc Analysis of the Prospective, Randomized DIRECTOR Trial

BACKGROUND: Tumor resection is a prerequisite in many studies of new glioblastoma therapeutics; however, no clear parameters for "resectability" exist. We evaluated inter-rater variability in assessing tumor resectability and potential associations between resectability and survival in a trial cohort of glioblastoma recurrence. METHODS: DIRECTOR (NCT00941460; 9/2009-6/2012) evaluated two dose-dense temozolomide regimens for first recurrent glioblastoma, yielding similar outcomes between arms. Re-resection was allowed before initiation of systemic therapy by institutional decision. Eleven surgical neuro-oncologists (blinded to final outcomes) rated whether a 'meaningful resection' was achievable for each recurrent IDH-wildtype glioblastoma based on imaging and clinical data. RESULTS: MRI scans from 69 patients were available (median age:58.2 ± 1.1 years, median survival:10.0 months). 40 patients underwent re-resection (median age:56.4 ± 1.7 years, median survival:10.8 months). Surgical decision-making markedly varied between raters, ranging from 30-58 of 69 cases being classified as 'resectable' (κ = 0.405). In patients who received re-resection, a 'meaningful resection' was deemed feasible by > 80% of raters in 30/40 cases (75.0%). For patients without re-resection, unanimous agreement on non-resectability occurred in only 3/29 cases (10.4%); and 5/29 tumors (17.2%) were considered resectable by > 80% of raters. Knowledge of additional clinical factors virtually never changed MRI-based judgments. While patients who had a complete resection of contrast-enhancing tumor had favorable outcomes, a consensus on resectability by > 80% of raters was not associated with prolonged overall survival. DISCUSSION: Feasibility assessment for re-resection is heterogenous among neurosurgeons, challenging single-surgeon evaluation of "resectability". Those findings are limited by the number of surgical raters and the size of the DIRECTOR cohort.