Introduction: The Cancer Genome Atlas (TCGA) has previously revealed five sub-classes of astrocytic gliomas, including four sub-classes defined by RNA expression (proneural, neural, mesenchymal, and classical), and one by isocitrate dehydrogenase mutation (IDHm). These studies demonstrated a difference in prognosis only with IDH mutation, but were limited by a lack of information on patient and clinical characteristics. Here, we determine if there is a difference in survival between each of the non-IDH mutated molecular subtypes of GBM, while accounting for patient age, Karnofsky Performance Score (KPS), or tumor grade.
Methods: We identified 1,073 patients with astrocytomas of all grades from TCGA, excluding IDHm tumors to examine the potential association between RNA expression-based subtype classifications without IDHm as a confounder. We assessed survival using univariate and multivariate Cox proportional hazards analyses adjusted for age, KPS, and tumor grade. We also used The Cancer Imaging Archive (TCIA) to examine the relationship between molecular subtype and propensity for neuroanatomic location of glioblastomas (GBM).
Results: Univariate analyses indicated improved survival with increasing KPS (HR=0.961, p<0.001), and worse survival with increasing age (HR=1.054, p<0.001) and increasing grade (HR=3.319, p=0.004 for grade 3; HR=11.432, p<0.001 for GBM; relative to grade 2). While no survival association was observed with regards to the RNA-based subtype classification in univariate analysis, in a multivariate analysis that included age, KPS, tumor grade, and RNA-based subtype classification, proneural glioblastomas are associated with worse survival (HR=1.524, p=0.012) relative to the non-proneural glioblastomas. Additionally, analysis of TCIA demonstrated that proneural glioblastomas were more likely to be located near the sub-ventricular zone (SVZ, p<0.05).
Conclusions: Our findings suggest that the RNA expression-based subtype classification has prognostic utility. We found the proneural subtype of astrocytoma to be associated with worse survival. This subtype was more likely to be located near the SVZ, suggesting potential mechanistic insights for this survival association.
Patient Care: This research will improve patient care, because it will help us to understand the prognostic importance of molecular features, as we enter the age of genomic medicine. Understanding the different prognoses associated with GBM molecular subtypes will help physicians tailor clinical management of patients.
Learning Objectives: 1) Understand if the five glioblastoma subtypes are associated with survival
2) Understand the relative survival association of glioblastoma molecular features when accounting for clinical and patient characteristics
3) Help provide more robust prognostic information for patients afflicted with glioblastoma through the use of molecular subtypes