Editorial: Glioblastoma subtype

Abstract

In this paper, the authors present a retrospective series of 302 consecutively treated patients with “primary” glioblastomas (GBMs).2 They parse the patients into 2 groups, those with labeling for neurofilament protein (NFP) of 1% or greater of tumor cells, and those with less than 1% labeling, based on immunohistochemistry. Using routine (but rigorous) statistical analysis on the 266 patients with adequate follow-up, they found that younger age at presentation, higher Karnofsky Performance Scale score, tumor not crossing the midline on imaging studies, degree of resection, use of adjuvant therapy, and positive staining for NFP were all associated with improved patient survival. Although the patients with positive NFP expression were younger and had more favorable tumor location, the presence of labeling for NFP remained a statistically significant positive prognostic variable on multivariate analysis, with those patients having positive tumor labeling surviving 13 months, compared with 7 months for those with negative tumor labeling. The authors eliminated “secondary” GBM from their analysis, suggesting that these tumors represent a different disease than “primary” tumors. They used clinical history, histopathology, and the presence of isocitrate dehydrogenase 1 (IDH1) mutations to classify and eliminate tumors as secondary. While the IDH1 mutation has been suggested as a marker of secondary GBM, the exact significance of IDH1 mutations is not clear.1,3 In a paper analyzing The Cancer Genome Atlas (TCGA) database for glioblastoma, Verhaak and coworkers4 reported that of 202 tumors, only 4 were secondary, but 12 had IDH1 mutations. Of note, the IDH1 mutations were classified almost exclusively into the proneural GBM subtype, a subtype with a better prognosis than that of the group as a whole. Were the results presented in the current paper biased by removing the IDH1 tumors (representing the good prognosis proneural group) from the NFP-negative tumors? In TCGA analysis, the GBMs were parsed into 4 tumor types based on 1740 genes consistently but variably expressed by the tumors.4 One group of these tumors, the neural subtype, was characterized by NFP expression, but only represented 16% (33/202) of the tumors investigated. Of note, the authors of the current study found that more than half of the patients (177/302, 58.6%) had positive immunohistochemical labeling for NFP, as defined by labeling of at least 1% of tumor cells. The reason for this discrepancy is not clear. Are the tumors identified in this study overlapping into subtypes of GBM other than the neural subtype, as defined by TCGA analysis? Are the 2 study populations simply very different in subgroup makeup, with the NFP-positive tumors truly representing the neural subgroup of GBMs? If the tumors identified as NFP-positive in the current study do represent the neural subgroup of GBMs, then the authors have provided evidence for the better prognosis of this group and have provided an easier way to identify the subgroup than that of gene expression. This situation would be analogous to that of IDH1 mutations; these mutations identify a group of tumors with an improved prognosis, but also label tumors as members of the proneural subgroup, which has been demonstrated to have a better prognosis than that of GBM as a whole.4 Regardless of the answers to questions raised by the data presented here, the authors have provided data that NFP labeling—defined as 1% or more of the tumor cells showing immunohistochemical labeling—is a positive prog nostic variable in the treatment of GBM. The exact meaning of this data will await further studies of patients with GBM that include both immunohistochemical labeling of the tumors for NFP and subtyping of the tumors by expression analysis, allowing for comparison and correlation between the immunohistochemical data and gene expression data. (http://thejns.org/doi/abs/10.3171/2012.1.JNS112254)