Professional Documents
Culture Documents
Wirsching 2016
Wirsching 2016
Chapter 23
Glioblastoma
HANS-GEORG WIRSCHING1*, EVANTHIA GALANIS2, AND MICHAEL WELLER1
1
Department of Neurology, University Hospital Zurich, Zurich, Switzerland
2
Departments of Oncology and Molecular Medicine, Mayo Clinic, Rochester, MN, USA
Abstract
Glioblastoma is the most common and aggressive primary brain tumor in adults. Defining histopathologic
features are necrosis and endothelial proliferation, resulting in the assignment of grade IV, the highest
grade in the World Health Organization (WHO) classification of brain tumors. The classic clinical term
“secondary glioblastoma” refers to a minority of glioblastomas that evolve from previously diagnosed
WHO grade II or grade III gliomas. Specific point mutations of the genes encoding isocitrate dehydro-
genase (IDH) 1 or 2 appear to define molecularly these tumors that are associated with younger age and
more favorable outcome; the vast majority of glioblastomas are IDH wild-type. Typical molecular
changes in glioblastoma include mutations in genes regulating receptor tyrosine kinase (RTK)/rat sar-
coma (RAS)/phosphoinositide 3-kinase (PI3K), p53, and retinoblastoma protein (RB) signaling. Standard
treatment of glioblastoma includes surgery, radiotherapy, and alkylating chemotherapy. Promoter meth-
ylation of the gene encoding the DNA repair protein, O6-methylguanyl DNA methyltransferase (MGMT),
predicts benefit from alkylating chemotherapy with temozolomide and guides choice of first-line treat-
ment in elderly patients. Current developments focus on targeting the molecular characteristics that drive
the malignant phenotype, including altered signal transduction and angiogenesis, and more recently,
various approaches of immunotherapy.
*Correspondence to: Hans-Georg Wirsching, Department of Neurology, University Hospital Zurich, Frauenklinikstrasse 26,
CH-8091 Zurich, Switzerland. Tel: +41-44-255-5500, Fax: +41-44-255-4507, E-mail: hans-georg.wirsching@usz.ch
382 H-G. WIRSCHING ET AL.
gliomas of 2.6 (95% confidence interval (CI), 0.8–8.6), divide into different, more differentiated cell types).
and the excess RR for malignant tumors followed linear Finally, the capability to form secondary tumors in serial
kinetics and was 1.98/Gy (95% CI, 0.73–4.69) (Ron et al., xenotransplantation assays that resemble the original
1988; Sadetzki et al., 2005). Among long-term survivors tumors is a defining criterion of glioma-initiating cells.
who underwent high-dose irradiation (30–44.9 Gy) for Various markers of glioma-initiating cells have been
primary brain tumors in childhood, the odds ratio was described, including CD133 (prominin) (Singh et al.,
21 for gliomas (n ¼ 40, including n ¼ 9 glioblastomas) 2004), CD15 (stage-specific embryonic antigen-1) (Son
and the excess RR/Gy was 0.33 (95% CI, 0.07–1.71) et al., 2009), CD44 (Pietras et al., 2014), and integrin
(Neglia et al., 2006). The radiation dose of diagnostic alpha 6 (Lathia et al., 2010) but the abundance of any
scans does not suffice as a risk factor (Brada et al., of these markers differs strongly between tumors
1992; Minniti et al., 2005). Mobile phone use has been (Yan et al., 2013). Most progeny of glioma-initiating cells
studied extensively with respect to gliomagenesis, but have features of astrocytes, but differentiation into
no definite association has been reported (Interphone functional endothelial cells and pericytes has been noted,
Study Group, 2010; Ostrom et al., 2014b). No association too (Ricci-Vitiani et al., 2010; Wang et al., 2010; Cheng
with smoking or other cancerogenic agents has been et al., 2013). Glioma-initiating cells have been attributed
reported. Expression of cytomegalovirus (CMV) genes a crucial role in growth, angiogenesis, invasion, and
and interaction of CMV gene products with some of resistance to radiotherapy (Bao et al., 2006) and chemo-
the core pathways that drive the malignant phenotype therapy (Beier et al., 2008; Chen et al., 2012), but, despite
of glioblastoma suggest an oncomodulatory role for promising preclinical data, no drugs specifically target-
CMV, but a firm role for CMV as a glioma-initiating ing glioma-initiating cells have entered clinical practice
agent remains to be confirmed (Dziurzynski et al., to date.
2012; Wick and Platten, 2014). In summary, ionizing irra-
diation of the brain is the only recognized exogenous risk
Tumor location
factor for the development of glioblastoma (Connelly
and Malkin, 2007; Bondy et al., 2008; Ostrom and There is no association of the location of glioblastoma at
Barnholtz-Sloan, 2011). diagnosis with the locations of NSPC, besides the con-
finement of most glioblastomas to the supratentorial
compartment. Reasons likely include excessive dissemi-
Cell of origin
nation of glioblastoma cells throughout the brain even at
The low frequency of glioblastoma and of primary brain early stages of the disease, but development of glioblas-
tumors in general compared to tumors arising from toma from non-NSPC cannot be excluded. Most gliomas
extraneural organs may reflect a high degree of protec- occur in the frontal (25.8%), temporal (19.7%), and pari-
tion of the brain from genotoxic stress. ATP-binding cas- etal (12.2%) lobes, whereas occipital-lobe (3.2%), cere-
sette (ABC) family transporter molecules of the blood– bellar (2.9%), brainstem (4.2%), or spinal cord (4.3%)
brain barrier restrict diffusion of chemical mutagens to gliomas are relatively rare (Ostrom et al., 2014a). Brain-
the brain (El Ali and Hermann, 2011). Furthermore, DNA stem gliomas are rare in adults, but account for the vast
is particularly sensitive to genotoxic stress during repli- majority of pediatric glioblastomas (Ostrom et al., 2015).
cation, but most brain cells have entered a definite post- Survival of patients with frontal-lobe glioblastomas was
mitotic state until adulthood. Experimental data suggest longer compared to patients with temporal- or parietal-
that adult glioblastoma may be derived from the small lobe glioblastoma (median overall survival (OS), 11.4
pool of adult neural stem and progenitor cells (NSPC), vs. 9.1 vs. 9.6 months, log rank p ¼ 0.01) in a cohort of
which are located in the subventricular zone (Sanai 645 adults who were treated within three consecutive tri-
et al., 2004), the subcortical white matter (Nunes et al., als of the Radiation Therapy Oncology Group (RTOG) in
2003), and the dentate gyrus of the hippocampus the pre-temozolomide era (Simpson et al., 1993). This is
(Eriksson et al., 1998). NSPC have retained the ability likely due to an association of frontal-lobe glioblastoma
to enter mitosis and play a major role for the plasticity with other favorable prognostic features, such as youn-
of the brain in learning and memory (Sanai et al., ger age, isocitrate dehydrogenase-1/2 (IDH-1/2) muta-
2005). A subpopulation of glioblastoma cells termed tions (Sturm et al., 2012; Ellingson et al., 2013), and
glioma-initiating (also called glioma stem-like cells) higher rates of gross total resection (Lacroix et al.,
share various features with NSPC, including the location 2001). Approximately half of all glioblastomas in adults
in perivascular and hypoxic niches (Gilbertson and Rich, infiltrate more than one lobe and approximately 5%
2007), self-renewal (i.e., the capability to divide into two grow multifocally (Djalilian et al., 1999). Leptomenin-
progeny, of which at least one resembles the initial cell) geal dissemination of glioblastoma cells is rare and
and multilineage differentiation (i.e., the capability to occurs in late stages of the disease (Herrlinger et al.,
GLIOBLASTOMA 383
2004), although a frequency of up to 14% has been in the particularly impaired elderly population (Roa
reported in one series comprising younger patients et al., 2004; Taphoorn et al., 2005; Keime-Guibert
(median age, 31 years) (Arita et al., 1994). In another et al., 2007; Gallego Perez-Larraya et al., 2011). Person-
series, 5 of 25 autopsied patients with glioblastoma ality changes and mood disorders commonly occur in
had spinal leptomeningeal metastases (Erlich and patients with frontal tumors, and these may be mistaken
Davis, 1978). Associations of leptomeningeal dissemina- for psychogenic disorders or part of the physiologic
tion with younger age, male gender, incomplete tumor aging process and thus delay diagnosis. Sensorimotor
removal, multiple resections, ventricular entry or prox- deficits are the presenting symptom in approximately
imity of the tumor to the ventricular system, and with 20% of all patients with glioblastoma, and approxi-
gains at the 1p36 chromosomal region, have been noted mately 5% of patients present with aphasia due to
(Awad et al., 1986; Grabb et al., 1992; Arita et al., 1994; tumors arising in the speech-dominant hemisphere
Lindsay et al., 2002; Korshunov et al., 2007). Distant (mostly left-sided) (Yuile et al., 2006).
organ metastases arising from glioblastoma have been Epilepsy may mimic aphasia or sensorimotor deficits
reported, predominantly to lung, pleura, lymph nodes, from tissue destruction postictally and occurs more fre-
bone, and liver (Schweitzer et al., 2001), but these cases quently in glioblastomas that affect the temporal lobe
are rare. The reasons for the strong brain tropism of glio- (Chaichana et al., 2009). Epilepsy is the presenting symp-
blastoma cells are not known, but likely include adapta- tom in 24–68% of glioblastomas and develops in 19–38%
tion to the metabolic and immunologic peculiarities of later during the course of the disease (Wick et al., 2005;
the brain microenvironment that precludes survival in Chaichana et al., 2009; van Breemen et al., 2009;
other organs (Fecci et al., 2014; Mashimo et al., 2014). Kerkhof et al., 2013). As a presenting symptom, epilepsy
Furthermore, spreading glioblastoma cells reactivate is associated with longer survival, probably due to an
developmental programs of neuroglial precursor cells association with younger age as well as cortical location
and may follow the same paths along which neuroglial and smaller tumor size, indicating good surgical resect-
precursor cells migrate during development, and extra- ability and possibly diagnosis earlier during the disease
neural migration of neuroglial precursor cells does not course (Yuile et al., 2006; Chaichana et al., 2009). Head-
occur. Due to the restriction of glioblastoma cell spread- aches are the presenting symptom in less than one-third
ing within the borders of the central nervous system, the of all patients with glioblastoma (Yuile et al., 2006), and
transplantation of organs from donors with glioblas- are mostly dull in nature and typically present at night or
toma is per se feasible (Watson et al., 2010; Warrens at awakening. Other signs of increased intracranial pres-
et al., 2012), but only few glioblastoma patients qualify sure, such as nausea, vomiting, dizziness, fatigue, and
as organ donors. This is because the mode of death is neurocognitive slowing, occur increasingly during the
mostly a gradual decline in a home or hospice setting, disease course, but may be present at diagnosis. Rarely,
whereas organ donation requires brain death due to a leptomeningeal dissemination may mimic painful nerve
sudden intracranial event (e.g., hemorrhage), or gradual root compression, myelitis, and other spinal diseases.
death from compromised airway protection in a hospi- Only few patients retain stable disease and remain neu-
tal setting that includes intubation to enable nonheart- rologically largely asymptomatic for years, but the
beating donation (Collignon et al., 2004). majority of patients experience severe impact on quality
of life once first-line treatments have failed (Roa et al.,
2004; Taphoorn et al., 2005; Keime-Guibert et al., 2007;
Clinical course
Gallego Perez-Larraya et al., 2011). In summary, there is
The clinical course of glioblastoma is determined by no typical clinical presentation of glioblastoma. Differ-
tumor location and dynamics of spread within the brain. ences compared to other gliomas include the more rapid
Tissue destruction, edema, and epilepsy likewise con- dynamic and somewhat lower incidence of epilepsy (see
tribute to clinical symptoms, causing rapid deteriora- Chapter 2).
tion in some patients. The clinical term secondary
glioblastoma refers to glioblastomas that are preceded DIAGNOSIS AND MOLECULAR
by the histologic diagnosis of WHO grade II or III CLASSIFICATION
gliomas, but this term is increasingly dispensable with
Imaging
molecular marker-based classifications of gliomas
(Weller et al., 2015a). New-onset seizures or development of neurologic defi-
Despite the invariably fatal prognosis of glioblas- cits are commonly followed by a neurologic workup that
toma, any of the standard treatment options discussed includes magnetic resonance imaging (MRI). Computed
further below may keep stable or even improve quality tomography (CT) with contrast enhancement is less sen-
of life and cognitive functioning for the time being, even sitive in detecting the typical features of glioblastoma.
384 H-G. WIRSCHING ET AL.
Its use is restricted to acute situations, e.g., when hem- blood–brain barrier disruption, central hypointensity on
orrhage is suspected, or when MRI is not available or T2-weighted images as a correlate of necrosis, and perifo-
not possible, e.g., in patients with cardiac pacemakers cal hyperintensity on T2-weighted and fluid-attenuated
or other metallic implants. Prior to biopsies, amino acid inversion recovery (FLAIR) images as a correlate of
positron emission tomography (PET) is increasingly per- edema or noncontrast-enhancing tumor (Fig. 23.1). MR
formed to guide the site of biopsy to metabolic hotspots spectroscopy, T1 contrast subtraction maps, as well as
that may represent sites of higher tumor grade diffusion/perfusion- and susceptibility-weighted MRI
(la Fougere et al., 2011); yet, PET is not part of the stan- sequences have refined radiographic diagnostics and
dard workup of glioblastoma patients (see Chapter 3). enable more reliable discrimination of glioblastoma from
On MRI, glioblastomas appear as masses with con- other contrast-enhancing lesions, including abscesses, pri-
trast enhancement at their margin as a correlate of mary central nervous system lymphomas, and metastases
Fig. 23.1. Neuroimaging features of glioblastoma. Magnetic resonance imaging (MRI), T1-weighted (A), gadolinium-enhanced
(B), T2-weighted (C), fluid-attenuated inversion recovery (FLAIR, D); (E): 18F-fluoro-ethyl-tyrosine (FET) positron emission
tomography (PET).
GLIOBLASTOMA 385
from nonprimary brain tumors (Kono et al., 2001; Development of gliosarcoma may occur secondary to
Law et al., 2003; Weber et al., 2006; Ellingson et al., glioblastoma and molecularly, mutations of the epithe-
2012; Kickingereder et al., 2014), as discussed in more lial growth factor receptor (EGFR) are infrequent com-
detail in Chapter 3. However, the appearance of glio- pared to glioblastoma. The prognosis may or may not be
blastoma on imaging scans can vary considerably, and worse than for glioblastoma, after adjustment for com-
therefore tissue-based diagnosis is indispensable (Weller mon prognostic factors (hazard ratio (HR), 1.17; 95%
et al., 2014). confidence interval (CI), 1.05–1.31) (Kozak et al., 2009).
Histopathology
GIANT CELL GLIOBLASTOMA
The defining histopathologic features of glioblastoma
are necrosis and microvascular proliferation, and these Giant cell glioblastoma is characterized by multinu-
qualify glioblastoma for the highest grade in the WHO cleated cells with a diameter of more than 500 mm and
classification of primary brain tumors, grade IV lymphocytic infiltration. Giant cell glioblastomas com-
(Louis et al., 2007). Other signs of malignancy that are prise approximately 1% of all glioblastomas. The inci-
present in glioblastoma include anaplasia, high mitotic dence may be higher in young adults (median age at
rates, and invasiveness, but these features are also pre- diagnosis 51 years). The prognosis for patients with giant
sent in anaplastic gliomas, which are assigned WHO cell glioblastoma is better than for glioblastoma, after
grade III (Louis et al., 2007). In addition, immunohisto- adjustment for common prognostic factors (HR, 0.76;
chemical markers are commonly assessed to ascertain 95% CI, 0.59–0.97) (Kozak and Moody, 2009).
the diagnosis of glioblastoma, including glial fibrillary Other histopathologic variants have been suggested,
acidic protein expression to confirm astrocytic lineage including fibrillary glioblastoma, small cell astrocytoma,
differentiation and MIB-1/Ki-67 to aid quantification glioblastoma with oligodendroglial component, glioblas-
of proliferation. An antibody that specifically detects toma with primitive neuroectodermal tumor, and granu-
mutant IDH-1R132H (Capper et al., 2009) has been inte- lar cell astrocytoma (Karsy et al., 2012), but none of
grated into the standard diagnostic workup of suspected these have been accepted as designated variants within
glioblastoma, but it must be recognized that this anti- the current WHO classification of brain tumors because
body does not detect other IDH-1 or IDH-2 mutations. of a lack of distinct molecular and clinical features
A more detailed overview on histopathologic grading (Louis et al., 2007).
of gliomas is provided in Chapter 5. We emphasize
that the traditional histopathologic definition of glio- Molecular markers
blastoma (and other gliomas) is likely to be comple-
MGMT promoter methylation predicts benefit from
mented by molecular approaches with the next edition
temozolomide chemotherapy in newly diagnosed and
of the WHO classification of brain tumors, due to be
probably also recurrent glioblastoma (Hegi et al., 2005;
released in 2016.
Malmstrom et al., 2012; Wick et al., 2012; Weller et al.,
Glioblastoma variants 2015b). MGMT is a DNA repair protein that counteracts
DNA alkylation by chemotherapy; hypermethylation of
Two rare histopathologic variants of glioblastoma are the MGMT promoter results in gene silencing (Weller
defined in the WHO classification of primary brain et al., 2010). Since the discovery of its role in the resistance
tumors, gliosarcoma and giant cell glioblastoma (Louis of glioblastoma to alkylating chemotherapy, MGMT has
et al., 2007). remained the biomarker with the strongest impact on clin-
ical decision making, particularly in elderly glioblastoma
GLIOSARCOMA
patients (Fig. 23.2) (Weller et al., 2014).
Gliosarcoma is characterized by a metaplastic mesenchy- IDH mutations occur in approximately 5–10% of all
mal component that stains for reticulin and may show glioblastomas, are associated with younger age and bet-
signs of fibroblastic, cartilaginous, osseous, smooth ter outcome (Parsons et al., 2008), and rarely occur in
and striated muscle, or adipose cell lineage, and these patients aged 65 years or older. Mutant IDH produces
features are present in addition to characteristics of glio- the oncometabolite 2-hydroxyglutarate (Turcan et al.,
blastoma (Han et al., 2010). Gliosarcomas comprise 2012), which has been linked to a distinct epigenetic pat-
approximately 2% of all patients diagnosed with glio- tern designated glioma CpG island methylator pheno-
blastoma or gliosarcoma and clinical features include type (G-CIMP) (Noushmehr et al., 2010). In contrast,
predilection for location in the temporal lobe, a IDH wild-type glioblastoma is associated with older
meningioma-like macroscopic appearance at surgery, age and poor prognosis; TERT promoter mutation marks
and frequent reports of extracranial metastases. particularly poor outcome in these patients (Labussiere
386 H-G. WIRSCHING ET AL.
MRI
Contrast enhancing, diffusely infiltrating mass
Surgery/Biopsy
Gross total resection, if safely possible
Glioblastoma
Repeat RT
Mostly stereotactic No firstline RT
1st RT >6 months Recurrent glioblastoma
Young age
Small tumors
Good KPS
Systemic therapies Surgery
contraindicated or Only if gross total resection is safely possible
failed
Combined?
Fig. 23.2. Therapeutic approach to glioblastoma. MRI, magnetic resonance imaging; MGMT, O6-methylguanyl DNA methyl-
transferase gene promoter; TMZ, temozolomide 150–200 mg/m2 on 5/28 days; TMZ/RT, 30 2 Gy ¼ 60 Gy with daily concom-
itant temozolomide at 75 mg/m2; hypofractionated RT, radiotherapy at 15 2.66 Gy ¼ 40 Gy; BEV, bevacizumab 10 mg/kg body
weight every 2 weeks; KPS, Karnofsky performance score; PD, progressive disease.
et al., 2014). Despite identical histopathologic appear- EGFR to drive cellular transformation and the glioma-
ance, IDH wild-type and mutant glioblastomas consti- initiating cell phenotype via signal transducer and activa-
tute two distinct molecular and prognostic entities that tor of transcription (STAT) 3 signaling (Fan et al., 2013),
are likely to be separated in future classifications of and the EGFRvIII deletion mutation may confer poor
brain tumors, as outlined in more detail in Chapter 5. prognosis compared to EGFR-amplified glioblastoma
About half of all glioblastomas with EGFR amplifi- without co-occurrence of EGFRvIII (Shinojima et al.,
cation (i.e., 20–25% of all IDH wild-type glioblastomas) 2003; Heimberger et al., 2005). Promising results with
harbor a deletion within the extracellular ligand-binding vaccination targeting EGFRvIII in phase II clinical trials
domain of EGFR (designated delta-EGFR or EGFRvIII) have been reported (Sampson et al., 2010; Hegi et al.,
that yields ligand-independent receptor activity (Aldape 2012; Reardon et al., 2015).
et al., 2004). EGFRvIII is commonly expressed only in
a subset of EGFR-amplified cells (Nishikawa et al.,
Molecular signatures
2004), but preclinical and clinical data support a role
for EGFRvIII-positive cells as drivers of disease pro- Integrated large-scale analyses of genetic, epigenetic,
gression. The EGFRvIII protein interacts with wild-type and expression data increasingly complement the
GLIOBLASTOMA 387
understanding of the biology and yield continuous proneural glioblastoma, whereas G34, K27, and RTK
refinement of the subclassification of glioblastoma I are associated with poor prognosis that is comparable
beyond histologic grading. In 2006 three gene expression to the two nonproneural subtypes (Sturm et al., 2012).
subtypes were defined by cluster analysis of 35 genes Meanwhile, the gene expression-based glioblastoma sub-
that were correlated the strongest with survival in types have also been detected on the single cell level
76 patients with newly diagnosed anaplastic astrocytoma within the same tumors, and an association of these gene
or glioblastoma and termed proneural, mesenchymal, expression profiles with stemness-related genes and
and proliferative (Phillips et al., 2006). The proneural glioma-initiating cells was suggested (Patel et al.,
subtype was associated with prolonged survival, youn- 2014). To date no therapeutic implications for clinical
ger age, and anaplastic histology and lacked alterations routine are based on these global characterizations of
of phosphatase and tensin homolog on chromosome the molecular background of individual tumors. The
10 (PTEN) or EGFR. The proliferative and mesenchymal molecular subclassification of glioblastoma is discussed
subtypes expressed genes that were associated with pro- in more detail in Chapter 6.
liferation or angiogenesis, respectively, and survival was
poor. These analyses were extended to larger datasets TREATMENT AND FOLLOW-UP
and complemented by mutation analyses in subsequent
Surgery
years. In 2008, three altered key signaling pathways were
determined by The Cancer Genome Atlas project (2008), Tissue is required to establish the diagnosis of glioblas-
namely receptor RTK/RAS/PI3K (88%), p53 (87%), and toma, as outlined above. It can be obtained by stereotac-
retinoblastoma protein (78%), and in the same year, tic or open biopsy, or by microsurgical resection of the
IDH-1/2 mutations were discovered as a marker that tumor. Biopsies are mostly performed as stereotactic
is strongly associated with secondary glioblastoma, biopsies in patients with multifocal disease and in tumors
younger age, and better outcome (Parsons et al., 2008; deemed unresectable due to their location in so-called
Yan et al., 2009). In 2010, the initially suggested gene “eloquent” areas, i.e., areas that cannot be resected with-
expression-based subclassification of glioblastoma was out causing major disability. However, biopsies may not
further refined using 601 genes and sequence data yield enough tissue for molecular analyses and bear the
from 91 patients, yielding four subtypes designated pro- risk of sampling errors. The therapeutic value of micro-
neural, neural, classic, and mesenchymal, based on sim- surgical resection has been demonstrated best in a ran-
ilarities with known gene expression profiles (Verhaak domized trial exploring the use of the fluorescent
et al., 2010). label, 5-aminolevulinc acid, to facilitate gross total resec-
IDH-mutant glioblastomas constitute a subgroup of tion (Stummer et al., 2006). It is further supported by
proneural glioblastomas associated with G-CIMP numerous retrospective cohort studies, establishing
(Noushmehr et al., 2010). Of note, there is no association gross total resection as standard of care whenever
of gene expression-based subtypes of G-CIMP with deemed feasible (Fig. 23.2). (Sanai et al., 2011; Weller
MGMT promoter methylation (Bady et al., 2012). By et al., 2014) Feasibility of this concept has also been dem-
2012, epigenetic profiling yielded a classification of onstrated for elderly patients aged 65 years or older:
six subtypes, four of which clustered within the pro- Extent of resection was a prespecified, independent
neural gene expression subtype: two distinct histone prognostic factor in a survival model that controlled
H3 (H3F3A) mutations, G34 and K27, were detected for age, extent of resection, histology, MGMT status,
in a large proportion of pediatric glioblastomas and these and study treatment among 371 elderly patients with ana-
mutations as well as IDH mutations were mutually plastic astrocytoma (n ¼ 40) or glioblastoma (n ¼ 331)
exclusive. These three subtypes share an association with randomized to receive postoperative radiotherapy or
mutations of the tumor suppressor gene TP53. A fourth temozolomide chemotherapy within the NOA-08 trial
proneural subgroup, designated RTK I, was character- (Wick et al., 2012). Further, among 342 patients with
ized by upregulation of platelet-derived growth factor newly diagnosed glioblastoma aged 60 or older who
receptor alpha (PDGFRA) and frequent deletions of were randomized to two different radiotherapy regi-
the CDKN2A gene encoding cyclin-dependent kinase mens or temozolomide, biopsy was associated with infe-
inhibitor 2A (Sturm et al., 2012). The nonproneural sub- rior OS compared to surgical resection (HR, 1.50; 95%
types were designated RTK II/classic, which is charac- CI, 1.17–1.92) in multivariate analyses controlling for
terized by frequent EGFR amplifications and age, type of surgery, study treatment, and WHO perfor-
CDKN2A deletions, and the mesenchymal subtype. mance score (Malmstrom et al., 2012). The use of
These six subclasses were associated with distinct tumor 5-aminolevulinic acid (see above) (Stummer et al.,
locations, age distributions, and prognosis. Only IDH- 2006, 2008) or intraoperative imaging by ultrasound
mutant glioblastomas retain the good prognosis of or MRI may aid the surgeon to obtain the goal of an
388 H-G. WIRSCHING ET AL.
optimal resection (Rygh et al., 2008; Kubben et al., 2011), radiotherapy was likely that standard radiotherapy was
while nuclear imaging with 18F-DOPA is currently in clin- not completed by a substantial fraction of patients
ical testing (NCT02020720). (Malmstrom et al., 2012).