Tumor Location of GBM Predicts for Survival

Abstract

Factors such as age, performance status (PS) and resection extent all affect prognosis of glioblastoma (GBM). Studies have divided patients into those with tumors in functionally silent or eloquent areas based on functional MRI (fMRI) scans, the latter having poorer overall survival (OS). We aimed to develop a prognostic classification based on tumor location in patient’s diagnostic MRI scans, without requiring costlier fMRI. We hypothesized that patients with tumors in non-eloquent areas would have better OS than those in near-eloquent and eloquent areas. This single institutional study involved 68 patients with histological diagnosis of GBM from Nov 2007-Nov 2016. All patients received concurrent temozolomide-RT. IMRT was given to a dose of 60gy/30#. Extent of resection, neurological deficits, and PS were noted. All patients had pre-treatment MRIs and were followed-up with 3-6 monthly MRIs until progression and/or death. Eloquent areas were defined as cortical areas that if removed, cause loss of sensory processing, linguistic ability, or paralysis – i.e.: motor and sensory cortices, visual and speech centers, internal capsule, basal ganglia, and brainstem. Near-eloquent areas were defined as those <8 mm from eloquent areas. Tumor location was confirmed by two radiation oncologists. Primary outcome was OS and secondary outcome was progression-free survival (PFS), analyzed using log-rank test. 23.6% of patients had tumors in non-eloquent areas, 35.3% in near-eloquent areas, and 41.1% in eloquent areas. 72% of patients had PS of 0 or 1. Median OS was 22.0mo, 15.8mo and 12.9 mos. Compared to those in non-eloquent areas, those in near-eloquent areas had poorer OS (hazard ratio: 0.449[0.197-1.02], p=0.056), likewise for those in eloquent areas (HR 0.435[0.189-1.00], p=0.050). One possible reason for the differences is that tumors in eloquent/near-eloquent areas had more limited resection. However, high proportion of the tumors in non-eloquent areas also had limited resection - biopsy (27.3%) or partial resection (18.7%), and this subset of patients also had improved OS (median 20.2 mo). There was no significant difference in median PFS (9.0 mo overall) between non-eloquent, near-eloquent and eloquent groups. The location of a GBM in a non-eloquent area predicts for better OS compared to those in eloquent or near-eloquent areas regardless of resection extent. There was no significant difference in PFS, showing equal efficacy of treatment in disease control between groups. However, the processes of surgery or chemoradiotherapy to tumors in/near-eloquent areas may have led to side effects such as speech or motor deficits which then led to poorer OS. With our classification system, patients’ existing MRIs can be used, with no extra investigations required. We aim to incorporate other factors such as PS to create a prognostic scoring system, and to develop MRI reporting guidelines to specify relation of a tumor to an eloquent area.