1558 / 1708
84
M. Reni et al. / Critical Reviews in Oncology/Hematology 63 (2007) 81–89
cles than in the third ventricle. Myxopapillary ependymomas
are typically and almost exclusively located in the conus –
cauda equina – filum terminale region. Rarely, they have been
observed in the upper spinal cord, in the lateral ventricles
or in the brain parenchyma. Subependymomas are typically
located in the fourth and in the lateral ventricles.
3.3. Diagnostic criteria
Ependymoma appears as a well-circumscribed lesion with
varying degrees of contrast enhancement, which is more
pronounced in anaplastic tumours and can be absent in
subependymomas, on either MRI or CT scanning. A cys-
tic component, the presence of calcium, and intra-tumoural
haemorrhage are occasionally observed, while oedema and
brain infiltration are infrequent. Surgical exploration and
biopsy are essential for the selection of appropriate treatment.
4. Staging
4.1. Staging procedures
The staging work-up should include a careful history,
physical examination and magnetic resonance imaging of the
brain and the spinal cord. Examination of the CSF for cyto-
logical evidence of malignancy is essential. The incidence of
spinal seeding is 1.6% for supratentorial tumours, 9.7% for
infratentorial lesions, 8.4–20% for high-grade tumours, and
2–4.5% for low-grade lesions
[34,35] . The highest incidence
is observed among high-grade infratentorial ependymomas.
4.2. Staging system
The UICC/AJC classification
[36]is applied to all
brain tumours and distinguishes between supratentorial and
infratentorial locations. However, this classification is rarely
used and the nodal and distant metastases categories very
rarely occur in ependymomas.
4.3. Restaging procedures
Restaging should include all the diagnostic procedures
that were positive at the time of diagnosis and of initial
staging. Spinal seeding rate is consistently different among
reported series, most likely due to different diagnostic cri-
teria and whether either clinical or pathologic seeding was
considered—the latter being almost 10 times more frequent
than the former
[37] . The most important determinants of the
risks on spinal seeding are tumour grade and localisation
[35] ;0–12.5% of patients with high-grade supratentorial lesions
developed spinal seeding, whereas 0–38%of those with high-
grade infratentorial tumours developed spinal dissemination
[33,35,38,39] .For low-grade tumours, 0–7% of patients
with supratentorial lesions developed seeding compared with
0–40% for those with infratentorial lesions
[33,35,37–39] .The incidence of spinal seeding was related directly to local
tumour control, regardless of tumour grade. The incidence
of spinal dissemination was significantly lower in locally
controlled patients than in those with uncontrolled primary
lesions (3.3% versus 9.5%)
[35] .5. Prognosis
5.1. Natural history
Grades I–II tumours, which are slowly growing gliomas,
disseminate infrequently to brain parenchyma, nerve roots,
bones and CSF; they are sometimes asymptomatic and are
found incidentally at autopsy
[40] .Anaplastic ependymo-
mas exhibit a more rapid growth pattern and are occasionally
invasive. They may occasionally be the result of malignant
progression from grade II tumours, and tend to spread into the
CSF more frequently, particularly if located in the posterior
fossa.
5.2. Prognostic factors
Most reported series of ependymomas are retrospective
and, include only a small number of patients, due to the
low incidence of this tumour type. Moreover, these studies
span several decades which hampers the interpretation of the
results due to changes in grading systems and diagnostic and
therapeutic policies, and with limited statistical power. Con-
sequently, generally accepted prognostic factors are lacking.
The prognostic significance of tumour grade is not universally
accepted, most likely due to the varying definitions of anapla-
sia
[41–43] ,to the large number (69%) of discrepancies
between local pathology diagnosis and those reported on cen-
tralised review
[44] , and to the fact that classical histological
features of anaplasia seem to be unrelated to the biologi-
cal behaviour of ependymomas
[40] .Another confounding
factor is that most series fail to distinguish patients with
malignant ependymomas from those with ependymoblas-
tomas which have an especially poor prognosis. According
to some authors, tumour grade is the most important deter-
minant of prognosis
[34,35,45–49] ,whereas others did not
find any correlation between survival and histologic grade
suggesting that the outcome is influenced by anatomical loca-
tion, which dictates resectability, rather than by pathological
features
[37,38,50–53] . The 5-year survival for patients with
low-grade tumours ranges from 55% to 87%, whereas for
anaplastic ependymomas it varies between 10% and 47%
[34,39,47,54] .A direct correlation between age and better
prognosis has been suggested. The small number of patients,
the different definitions of paediatric age among series (rang-
ing from 12 to 20 years), and the heterogeneity of histological
grade and tumour location
[34,49,54–57] between the com-
pared groups preclude reliable conclusions. Ependymomas
are uncommon in adults, and it is difficult to clearly assess
outcome in a strictly adult population as most of the published
series mainly relate to paediatric patients. Adult patients