P11.37.B When to resect or biopsy for patients with supratentorial glioblastoma: a multivariable prediction model

Abstract Background The prospects of a patient with suspected glioblastoma may rely heavily on the indication for surgical resection versus biopsy only. Biopsy percentages vary considerably across hospitals and guidelines treatment lack criteria decision-making. To identify tumor characteristics associated decision to resect or develop validate prediction model support. Material Methods Clinical data pre-operative MRI scans were collected adults who underwent first-time surgery supratentorial from registry-based cohort study 12 Netherlands, Germany, France, Italy, United States between 1st January 2007 31st December 2011. main outcome was type procedure: Predictors patient- tumor-related characteristics. Radiological factors extracted using an automated segmentation method. A constructed multivariable logistic regression analysis. cross-validated externally validated leave-one-hospital-out approach. Results Out 1053 patients treated glioblastoma, 28% rates varied 15-40% hospitals. showed excellent discrimination average area under curve 0.86. Of patient-related characteristics, younger age more Karnofsky Performance Score 60 less biopsy. location in right hemisphere, unifocality, no midline crossing, involvement cortical spinal tract, resection, as well high expected resectability index, occipital lobe, higher percentage Schaefer’s dorsal ventral attention, limbic, default networks. External validation proved acceptable outstanding areas ranging 0.79 0.92 Conclusion is presented support glioblastoma. In this model, informative than factors. This decision-making individual patients, facilitate comparisons cohorts surgeons institutions.