paediatrics Brussels 17

Articles

http://oncology.thelancet.com

Vol 8 August 2007

703

precludes strong conclusions, although immaturity and

tumour biology are likely to affect outcome.

Comparison of event-free survival in cohorts receiving

primary radiotherapy with overall survival in those

receivingprimary chemotherapy anddelayed radiotherapy

when necessary can be justified since radiotherapy is

used in both, but delayed or avoided in some depending

on the efficacy of the primary surgery and chemotherapy.

2,23

The 3-year overall survival of 79·3% of this UK study was

higher than the St Jude study (3-year progression-free

survival 69·5%)

and higher than other primary

radiotherapy studies.

Critical to the interpretation of

this data is the proportion receiving radiotherapy and the

number of patients who were not irradiated despite

progression, through parental or physician choice. The

approach used here for calculating cumulative incidence

of radiotherapy, using the competing risks methodology,

reflects more accurately the need for radiotherapy,

justifying our conclusion that 42% of patients studied are

true radiotherapy-free survivors.

Comparisons between these relatively small studies

highlight inconsistencies in methods of reporting, and

the effect of tumour and patient factors such as age,

histological grading, and surgical resection upon primary

outcomes, making direct comparison problematic. The

current international effort using meta-analysis to arrive

at a clinical and scientific consensus on the optimum

stratification of patients for the next era of clinical trials

in this diagnostic group is therefore justified as the most

important next step for testing of the next generation of

treatments.

Consistent application of histological grading of

malignant ependymomas using the WHO 2000

classification,

identifies classic ependymoma (grade II)

and anaplastic ependymoma (grade III). Differentiation

between these categories requires the recognition and

interpretation of a spectrum of pathological features. In

this cohort, 59 (66%) of 89 had WHO grade II histology,

whilst the SFOP reported 60 (88%) of 68 as grade III.

Some reports, including ours, did not identify histological

grade as prognostic,

3,26–30

whilst others did.

2,4,31–33

international consensus on the interpretationof histological

grade and its true value as a prognostic factor is required.

Nine (10%) of 89 children in this cohort had metastatic

disease, and all nine progressed. Metastatic disease has

been reported in 7–12%of patientswith ependymoma.

22,26,27

Although two studies found no effect of metastatic

disease on outcome,

22,27

four studies (including ours) did

detect an effect.

5,26,34

Compliance with CSF cytological examination in this

study was relatively poor, with results in just over 40% of

cases. However, recent evidence suggests that this

investigation was not helpful in predicting those patients

who would subsequently have a relapse in the spinal

cord.

The use of CSF cytology in determining outcome

in ependymoma seems therefore to be limited, although

a larger multicentre assessment is needed to clarify this.

The lack of CSF cytology to determine M1 status in our

study, although a deficiency, is probably of lesser

importance than the detection of leptomeningeal deposits

on MRI scanning in terms of patient outcome.

This study was mainly devised to investigate whether

chemotherapy can avoid or delay radiation in young

children with malignant brain tumours. Assessing

tumour response to chemotherapy was not a principal

component of the study. It is now accepted that assessing

chemotherapy response in ependymoma is a considerable

challenge. Several factors can confound radiological

review, for example, changes in the contrast uptake by

the tumour, the use of surgicell as an adjunct to stopping

tumour-bed haemorrhage, the timing of the scan, and

residual anatomical abnormality after surgery. The role

of more sophisticated imaging methods such as PET or

magnetic resonance spectroscopy now need to be

investigated.

The selection of drugs in this study was determined by a

combination of factors predictive of tumour sensitivity

and myelotoxicity to create a time-intensive schedule.

Methotrexate was selected because of the high-level folate

receptor expression in ependymoma cells, thereby

providing a mechanism to maximise drug accumulation

into the tumour.

The Children’s Cancer and Leukaemia

Group have just initiated a phase II study investigating the

single-agent activity of high-dose methotrexate in children

with incompletely resected ependymoma. The group

continues to use this chemotherapy protocol for children

with completely resected disease, but now recommend

conformal radiotherapy using multiple fields and a dose

of 59·4 Gy to the tumour bed. Cisplatin doses in this study

were higher than those in the SFOP protocol and

vincristine was used in three of four courses of each cycle.

Cisplatin ototoxicity is well recognised, but the incidence

of grade IV ototoxicity in this study was reassuringly low.

Future strategies to improve outcome could include the

use of granulocyte-colony stimulating factor to maintain

dose intensity and etoposide in combination with either

cisplatin or carboplatin. Scheduling and sequencing of

chemotherapy drugs could also be important, since the

14-day schedule might have acted through antiangiogenic

mechanisms and the predicted cytotoxic effects.

We delayed radiotherapy to avoid damaging the

developing CNS at a crucial point in its maturation. This

protocol began in an era in which the longitudinal

follow-up of patients was not considered as crucial as it

now is, and neuropsychological assessment was not an

intrinsic part of this study as this cohort was recruited as

part of a study seeking to establish the role of

chemotherapy for children younger than 3 years with a

variety of brain tumours. However, data on neuro-

psychological outcome of nine children, none of whom

had been irradiated, from a single UK centre treated on

this protocol have recently been reported.

At a mean age

at diagnosis of 22 months

and a mean time from

diagnosis of 75 months, all children had full-scale IQ