1591 / 1708
studies (2, 6, 15, 18–21). We also observed that the fre-
quency of leptomeningeal seeding after therapy (9.1%) is
similar to those reported previously, ranging from to 0% to
50% in clinical and autopsy studies (4, 22–28).
Kun
et al.
(29) and Shaw
et al.
(14) reported failures
occurring shortly after treatment, with a median time to
failure of 18 months. In our study, despite the short fol-
low-up period, the majority of failures occurred after 3
years, with an estimated median time of 45 months to
disease progression. This finding might be an effect of the
long duration of combined treatment, ranging from about
half a year in the preirradiation arm to 1 year in the main-
tenance arm. Merchant
et al.
(15) reported a median time of
37 months to disease progression with 61% of the patients
receiving preirradiation or maintenance chemotherapy.
Our study results indicate that, age, sex, and tumor site do
not influence outcome, in contrast to findings in other series.
Age
In previous studies, younger children had a lower survival
rate than older patients (2, 13, 19, 24, 30, 31). However,
Salazar
et al.
(21) found a reverse trend; in their study,
patients older than 12 years fared worse than the younger
children. We, on the other hand, found no impact of age on
the survival rate, but this may be related to the fact that only
children older than 3 years of age were enrolled in our trials.
Foreman
et al.
(20) also observed no age-related effect on
survival.
Sex
Only a few studies have analyzed the prognostic influ-
ence of sex on outcome in patients with ependymomas.
Some authors have reported a worse prognosis in male
patients (3, 32). In contrast, Shaw
et al.
(14), Foreman
et al.
(20), and Zorlu
et al.
(33) did not find any significant
difference, as confirmed by our analysis.
The ratio of infratentorial to supratentorial anaplastic
ependymomas is 1:1 (15).
Tumor site
The impact of the tumor site on outcome is controversial.
For example, Foreman
et al.
(20) observed a survival ad-
vantage in patients with supratentorial ependymomas,
though the difference was not significant. Needle
et al.
(34),
on the other hand, observed a significant survival advantage
in patients with supratentorial ependymomas, but there were
only 9 such patients with anaplastic tumors in their series. A
worse prognosis for patients with supratentorial tumors was
attributed to a higher rate of anaplastic histology (10, 16,
29) or less gross total resection (15). Differences in outcome
according to tumor site have been attributed to perioperative
mortality (16, 28, 31, 35). No association of outcome with
tumor site was described by Salazar (36). Read (37) and
Vanuytsel
et al.
(3) observed no association between tumor
site and survival, but they did observe a higher rate of spinal
seeding in infratentorial tumors. Our study results showed
no impact of tumor site on treatment outcome.
Dissemination
Most studies have not analyzed the influence of tumor
dissemination at presentation. A reason for this may be that
insufficient staging techniques may very rarely identify
spread. Pollack
et al.
(2), however, surprisingly found no
correlation between the rate of tumor dissemination and
prognosis. They thought that this finding might be due to the
fact that they diagnosed spread on the basis of cytologic,
rather than gross, evidence of spread or to the fact that they
used more aggressive therapy. We also had cytologic evi-
dence of dissemination in 3 of 5 children, and all 5 children
were treated with irradiation of the neuraxis with an addi-
tional boost to the primary tumor site. However, all children
died in less than 2 years after surgery and had a significantly
worse outcome in our analysis. But it must be considered
that many children in our study did not undergo a cytologic
analysis of the CSF; it therefore is possible that some
children with disseminated disease may still be alive, which
would mean that the outcome in such patients in our series
was actually better than our data showed. In addition, neg-
ative cytology findings are no absolute proof for an absence
of tumor cells in the CSF, a possibility that could also mean
that the outcome was actually better than we observed.
Extent of resection
In the Italian Pediatric Neurooncology Group series of 93
children (38), the completeness of resection emerged as the
most significant predictor of outcome. Many studies have
shown such an influence of total resection (7, 8, 13, 15, 18,
30, 32, 39). However, some studies failed to show an
advantage for complete resection (34, 36). In the present
series, total macroscopic removal was associated with con-
siderably improved progression-free and overall survival
rates. In addition, the rate of total or gross total resections
has improved over time with total resection in half the
patients in our study versus lower rates in previous studies.
Vanuytsel
et al.
(3) described 72 incomplete resections in 93
(77.2%) children treated between 1952 and 1988. The pre-
Fig. 6. Relationship between treatment volume and estimated
progression-free survival rate for all 24 supratentorial ependymo-
mas. CSI, craniospinal irradiation; local, tumor region only.
292
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 46, Number 2, 2000