Subtraction MRI-based residual tumor volumetry outperforms extent of resection and enhances radiologic–molecular prognostication in glioblastoma

Extent of resection (EOR) is a traditional prognostic indicator in glioblastoma (GB), yet evidence suggests that residual tumor volume (RTV) may provide superior predictive value. Accurate RTV quantification is challenging due to postoperative imaging artifacts. This study evaluated subtraction MRI–based RTV as a prognostic biomarker compared with EOR and examined its interaction with MGMT promoter methylation status (MGMT_ms). We retrospectively analyzed 83 patients with IDH-wildtype GB who underwent maximal safe resection followed by radiochemotherapy. RTV was quantified on postoperative MRI using a standardized T1-weighted subtraction workflow minimizing artifactual enhancement. Prognostic performance of RTV and EOR was compared through multivariable Cox analysis adjusted for age, performance status, and MGMT_ms. Bootstrap validation and analyses of long-term survivors (overall survival [OS] ≥ 28 months) were performed. Interobserver agreement for RTV was excellent (ICC = 0.91). RTV and EOR were significant predictors of OS in separate multivariable models, but only RTV retained significance when modeled together. The RTV-based model demonstrated superior discrimination (C-index 0.77) compared with the EOR-based model (C-index 0.62). An RTV threshold of 1 cm 3 stratified patients into distinct prognostic groups (median OS 24 vs. 14 months, p < 0.001). MGMT_ms conferred a substantial survival benefit, particularly in patients with RTV ≤ 1 cm 3 . Among long-term survivors no covariates retained significance. Subtraction MRI–based RTV quantification is superior to EOR for survival prediction in GB. Combining RTV with MGMT_ms enhances prognostic stratification, although current models do not fully explain long-term survival, highlighting the need for radiogenomic approaches.