Clinical Overview

Roughly 10% of pediatric brain tumors are classified as ependymoma, and the mean age at diagnosis is 4 to 6 years.^1–5 About 5% of patients with ependymoma present with evidence of leptomeningeal dissemination at the time of diagnosis, as determined by a combination of magnetic resonance imaging (MRI) with contrast enhancement and cerebrospinal fluid (CSF) sampling.⁶ Historical 5-year survival estimates range from 50% to 64%; progression-free survival estimates are lower at 23% to 45%.^2,3,7–9 Tumor typically recurs at the site of the original tumor, but in roughly 20% of cases there is isolated metastatic recurrence. The median time to recurrence is 13 to 25 months, and most children who are disease free at 5 years are cured (although late recurrences have been described).^2,3,7–9

In childhood, 90% of ependymomas are located intracranially: one third supratentorially, and two thirds infratentorially.¹⁰ Sixty percent of supratentorial ependymomas arise contiguous with or adjacent to the ventricular system (lateral or third ventricle); the other 40% arise in areas remote to the ventricles and their ependymal surface (Fig. 200-1). Infratentorial (posterior fossa) ependymomas arise from the floor (60%), lateral aspect (30%), or roof (10%) of the fourth ventricle (Fig. 200-2).^11,12

FIGURE 200-1 Axial computed tomography without contrast (A), a sagittal T1 weighted, Gadolinium-enhanced magnetic resonance imaging (MRI) scan (B), and an axial T2-weighted non-contrasted MRI scan (C) in a 3-year-old presenting with headaches and right-sided weakness. Note the extensive calcification within the periventricular tumor. This child was diagnosed by biopsy and treated with two rounds of chemotherapy to decrease the vascularity of the tumor before definitive resection.

FIGURE 200-2 Sagittal T1-weighted, gadolinium-enhanced magnetic resonance imaging of a child presenting with headache, vomiting, and gait ataxia. The image is typical of posterior fossa ependymomas, which have a propensity to grow through the foramen of Magendie and into the cervical canal. These tumors are also likely to progress through the lateral foramina and grow out the skull base foramina.

Prognostic Factors

The two most important prognostic factors in ependymoma are the extent of disease at diagnosis (known as the M status) and the amount of tumor resected at surgery. Other clinical variables thought to be of prognostic relevance include tumor location, tumor grade, and patient age at diagnosis. There is no formally recognized staging system for ependymoma beyond the determination of whether leptomeningeal dissemination exists.

Pathology

Ependymomas are classified as glial tumors, and their diagnosis by histopathologic and immunohistochemical examination is usually straightforward. Histologically, ependymomas are moderately cellular neoplasms with monomorphic nuclei; they usually stain positively for glial fibrillary acidic protein, rarely demonstrate endothelial proliferation, and contain perivascular pseudorosettes (groups of cells arranged radially around a blood vessel) and, less commonly, true ependymal rosettes.

In the World Health Organization (WHO) classification of brain tumors, a grade II ependymoma can be diagnosed only when mitoses are rare or absent; occasional foci of palisading necrosis are allowed, as are nodules with increased cellularity and mitotic activity.³¹ Anaplastic ependymoma is a grade III lesion in the WHO classification.³² It is diagnosed in tumors with increased cellularity, brisk mitotic activity, vascular proliferation, pseudopalisading necrosis, clear ependymal differentiation, perivascular pseudorosettes, cytologic atypia, and microvascular atypia.³¹ Areas of hypercellularity may be diffuse or focal and may form well-circumscribed regions abutting those of lower cellularity. Additionally, areas of cytologic atypia, including increased nuclear-to-cytoplasmic ratios and cellular pleomorphism, may be seen. Anaplastic regions often have a higher mitotic rate, although no specific threshold for a diagnosis of anaplastic ependymoma is widely accepted. Neither focal areas of atypia nor brisk mitotic activity is sufficient to make a diagnosis of anaplastic ependymoma. It is unclear whether anaplastic ependymoma arises from the progression or malignant degeneration of classic ependymoma or whether it occurs de novo.

Histologic grading and its clinical significance are difficult to assess, and agreement among pathologists when assigning a tumor grade to ependymoma is poor. Failure to find true ependymal rosettes or perivascular pseudorosettes is associated with a poor prognosis in children with ependymoma.¹³ Two separate groups have reported that the combination of necrosis, endothelial proliferation, and a mitotic index greater than 5 is a negative predictive factor for both overall and progression-free survival.^13,33 Others have shown an inverse correlation between mitotic rate and survival in ependymoma.³⁴ Ki-67 and MIB-1 immunolabeling methods to measure mitotic activity have also been used to show an inverse correlation between mitotic rate and outcome in two series of pediatric ependymoma.^13,21,33,35 Another group of investigators found that mitotic rate was important only in determining the prognosis of supratentorial ependymoma.³⁶

Hereditary Tumor Syndromes

The vast majority of ependymomas occur sporadically, although some well-documented associations with hereditary tumor syndromes have been described. Intramedullary spinal ependymoma is well known to occur in patients with neurofibromatosis type 2 (NF2) (Fig. 200-3). In contrast, patients with NF2 are probably not at increased risk of developing intracranial ependymomas. The same biology is found in patients with sporadic (nonfamilial) intramedullary ependymomas who have somatic mutations of the NF2 gene. Mutations of the NF2 gene are not found in intracranial ependymomas, myxopapillary ependymomas, or tanycytic ependymomas.^37,38 The NF2 gene is located on chromosome 22q. Although intracranial ependymomas are well known to show a loss of genetic material on 22q, they do not harbor NF2 mutations, suggesting the existence of another distinct ependymoma tumor suppressor gene in this region of the genome.

FIGURE 200-3 A, Sagittal T1-weighted, gadolinium-enhanced magnetic resonance imaging (MRI) of the cervical spine in a 14-year-old girl with neurofibromatosis type 2. She presented with a progressive history of pain down the right hip and leg and difficulty walking. Note the enhancing intramedullary mass in the cervical cord and the enhancing nodule in the medulla. B, Sagittal T2-weighted MRI of the same patient. This image clarifies the cystic fluid collection just rostral to the intramedullary tumor. C, Axial T1-weighted MRI of the same patient. Note the central location of the tumor, which is thought to arise from ependymal rests within the central canal and expands the spinal cord concentrically.

Ependymoma has also been reported in patients with the Li-Fraumeni familial cancer syndrome (due to a germline mutation in the p53 tumor suppressor gene).^39,40 However, somatic mutations of p53 are not commonly found in sporadic ependymomas and likely do not play a large role in their pathogenesis.^41,42 Scattered reports of patients with ependymoma and Turcot’s syndrome (brain tumors plus colonic neoplasia) are found in the literature, with some patients developing more than one ependymoma.^43–45 These patients have germline mutations in the APC gene on chromosome 5 and an overactive Wnt signaling pathway. The role of the APC gene and Wnt signaling in sporadic ependymomas has not been elucidated. Spinal ependymomas have also been reported in the context of multiple endocrine neoplasia type I due to mutation of a tumor suppressor gene on chromosome 11q13.^46,47 Other families with an increased incidence of ependymoma but no currently recognized familial tumor syndrome have also been reported.^48–50 Epidemiology studies have shown an increased incidence of breast cancer in the family members of children with ependymoma.

Cancer Genetics

There is a large body of literature on the karyotypic changes found in adult and pediatric ependymomas. As a group, ependymomas studied by G-banding karyotype show a frequent loss of genetic material on chromosomes 22q, 6q, 9q, 17p, and 11q and show a gain of genetic material on chromosome 1q.^{38,40,51–56} Loss of heterozygosity on chromosome 22q is more common in adult ependymomas and intramedullary spinal ependymomas and less common in pediatric ependymomas.⁴⁰ There is some evidence that a gain of genetic material on chromosome 1q may be an early event in the initiation of ependymoma.⁵¹

Ependymomas from different regions of the central nervous system (CNS) have distinct genetic profiles. There appear to be three genetic groups of childhood posterior fossa ependymomas. Ependymomas from very young children frequently show a balanced karyotype, with few or no regions of chromosomal gain or loss (no observed gains or losses).^57,58 A second subset of posterior fossa ependymomas has few regions of gain or loss other than gain of chromosome 1q. A third group of posterior fossa tumors shows recurrent regions of copy number gain across several autosomes. It is unclear whether there is any clinical or prognostic significance to these three groups. Supratentorial ependymomas have different regions of gain and loss, and many show homozygous deletion of the INK4A locus on chromosome 9, which is also frequently deleted in glioblastoma multiforme.⁵⁸ This key tumor suppressor locus plays a role in both TP53 and RB signaling. Recurrent tumors show more extensive karyotypic abnormalities than the original tumors; whether this is due to biologic progression or secondary to random DNA damage from radiation or chemotherapy is not clear.⁵⁹ Myxopapillary ependymomas show a high number of cytogenetic abnormalities (averaging nine per tumor) compared with other ependymomas.⁶⁰ The most common changes seen in cases of classic ependymoma are gains of chromosomal material on chromosome arms 1q or 9 and losses of 6q, 22, and the X chromosome.⁶¹ Overall, our knowledge of the genetics of ependymoma is limited by the resolution of the techniques used and the small number of tumors examined. Identification of specific genes and pathways for the development of targeted therapies will require larger studies done at higher resolution to differentiate driver genes from passenger genes.

Stem Cell Biology

It has been suggested that in many tumors only a small subset of tumor cells is capable of “regrowing” or recapitulating the tumor. These so-called cancer stem cells (CSCs) have been found in a number of tumor types, including ependymoma.⁵⁸ Ependymoma CSCs grow as neurospheres, can differentiate along several different lineages, and can recapitulate ependymoma when transplanted into immune-deficient mice. Conversely, the non-CSCs from an ependymoma do not grow when xenografted. CSCs from ependymoma samples, but not medulloblastoma CSCs, express markers of radial glial cells (i.e., Blbp, RC2, Glast).⁵⁸ These data suggest that ependymomas arise from radial glial cells, a type of CNS progenitor cell that is in fact the cell of origin of ependyma.⁶² Indeed, ependymomas from different regions of the nervous system have expression profiles similar to those of radial glial cells from those same regions, supporting a radial glial cell origin. A radial cell origin for ependymoma would certainly help explain the extraventricular location of many supratentorial ependymomas. Knowledge of the cell of origin may allow the creation of genetically modified mouse models of ependymoma by driving oncogene expression or the deletion of tumor suppressor genes in the radial glial cell compartment. These models would be welcome, because the shortage of model systems for ependymoma is undoubtedly responsible for the lack of research on the subject. There are no well-characterized ependymoma cell lines and no genetically modified mice that develop ependymoma; ependymoma xenografts have only recently been described.⁶³

Ependymoma Variants

There are a number of histologic variants of ependymoma that have a distinct appearance under the light microscope and are likely separate entities from classic ependymoma, each with its own unique biology (Table 200-1).

Neuroimaging

MRI of the CNS is the cornerstone of the diagnosis, management, and follow-up of children with ependymoma. Definition of the pathologic anatomy is critical before surgery to determine both the extent and the nature of tumor growth. For example, ependymoma has a propensity to grow out of the foramen of Luschka into the cerebellopontine angle (CPA) and out of the foramen of Magendie into the cervical spine. MRI is also used to plan radiotherapy and to judge the response to chemotherapy when it is given.

Children with ependymomas should have pre- and postoperative imaging of the brain and preoperative imaging of the spine. Postoperative imaging should be done within 72 hours of surgery; after this time, postsurgical artifacts may make it difficult for radiologists to determine the extent of resection. If this 3-day window is missed, MRI should be delayed for a couple of weeks to allow the artifacts from surgery to decrease. Leaving Gelfoam or Surgicel in the wound for hemostasis can make postoperative imaging difficult to interpret and is therefore strongly discouraged.

Although most ependymomas enhance, they often contain areas that do not enhance; thus, careful scrutiny of the preoperative films is warranted to avoid missing tumor at the time of resection. It is important to compare the pre- and postoperative films side by side to determine the location and extent of residual tumor. Most children with recurrence or progression of tumor present between 12 and 24 months after the initiation of radiation therapy. Although there is no definitive proof, it is well accepted that early detection of recurrence maximizes the opportunity for salvage therapies. Some retrospective studies have even suggested that the early detection of asymptomatic recurrence of ependymoma on surveillance imaging leads to better outcomes than diagnosis when the recurrence becomes symptomatic.^64,65

Preoperative spinal imaging is done to identify those rare patients (about 5%) who have leptomeningeal dissemination at the time of presentation of ependymoma. Surveillance imaging of the spine after surgery is problematic because of artifacts and false-positive indications of leptomeningeal disease from blood products, infection, and inflammatory changes secondary to surgery or radiation therapy. Leptomeningeal disease usually appears as a series of small nodular tumors or as thick disease coating the subarachnoid space focally. The diffuse sugarcoating seen in medulloblastoma is seldom seen in ependymoma, and it should not be diagnosed if there have been any recent invasive CNS procedures. Any spinal imaging should be performed at least 7 to 10 days after an invasive CNS procedure, including tumor resection, shunt placement or revision, and lumbar puncture. The entire spine should be imaged in at least two planes. Many children require sedation or general anesthesia to obtain high-quality images.

Although complete responses to either radiation or chemotherapy are uncommon in children with ependymoma, partial responses are often seen. After radiation, most ependymomas diminish in size over a period of years, and many show loss of contrast enhancement after only 1 year owing to a loss of vascularity. Conversely, some ependymomas have increased enhancement for several months after radiation owing to an increase in leukocyte-endothelial interactions. This should not be misinterpreted as progression of the tumor. Recurrence in the first year after adequate radiation therapy is uncommon. Similarly, responses to chemotherapy require approximately 6 months to become apparent.

Initial Surgery

Children with ependymoma most often present with localized, invasive disease. Most recurrences are local, but subarachnoid dissemination occurs rarely and is usually fatal. The surgeon plays the most important role in the treatment of ependymoma, because surgery has the greatest impact on the quality and quantity of the patient’s future life. The overriding value of gross total resection of ependymoma has been demonstrated in several institutional retrospective reviews and two prospective phase III trials.2–5,7–9^,16,19,27 Although complete resection is of paramount importance in the treatment of ependymoma, it is achieved in only 42% to 62% of patients.^3,5,8,19 Complete resection is more readily achieved in supratentorial tumors and in those that arise from the roof of the fourth ventricle. Gross total resection of tumors that invade the floor of the fourth ventricle or pass out the foramen of Luschka to involve the lower cranial nerves is much more difficult, and surgical complication rates are higher.

Sutton and colleagues⁵ retrospectively analyzed 45 patients with ependymoma and found that the 5-year survival after gross total resection or near total resection was 60%, but with subtotal resection (defined in this chapter as <90% resection) it fell to 21%. Pollack and associates⁸ reported a 5-year survival of 80% after gross total resection, compared with 22% after less than total resection. Perilongo and coworkers⁴ retrospectively evaluated 92 children with ependymoma: the 10-year survival after gross total resection was 70%, and the progression-free survival estimate was 57%. With subtotal resection the 10-year survival was 32% and the 10-year progression-free survival was 11%. Robertson and colleagues⁹ prospectively treated 32 patients on the Children’s Cancer Group Protocol 921 and found that the 5-year progression-free survival was 66% for patients with less than 1.5 cm² of residual tumor and 11% for those with more tumor. Although no randomized trial of complete versus partial resection for ependymoma has been (or will be) carried out, there is overwhelming evidence that cytoreductive surgery is beneficial to children with ependymoma. Children whose tumor burden can be decreased to minimal residual disease have the best chance at long-term survival.

Surgery without adjuvant treatment (radiation or chemotherapy) for children with ependymoma has been studied by two groups.^17,66 The results of these studies demonstrate that surgery alone is probably a reasonable therapeutic option for children with supratentorial ependymoma in the absence of pathologic features of anaplastic ependymoma and when the surgeon is able to take a rim of white matter around the tumor. Biopsies of the walls of the resection cavity should be performed if surgery is contemplated as the sole treatment modality.

Surgery for Residual or Recurrent Tumor

Except in those rare cases in which resection of the residual tumor may injure a critical CNS structure (e.g., brainstem), most children who initially receive less than a near-total resection should have additional surgery to reduce the tumor load to minimal residual disease (gross total or near-total resection). Chemotherapy before resection of residual ependymoma may decrease the vascularity of the tumor and allow surgery to be delayed, providing very young children an opportunity to grow.^22,67

Second resections for ependymoma have an acceptable morbidity.^22,67 In a series of 40 pediatric ependymoma patients referred over a 3-year period to St. Jude Children’s Research Hospital, 24 (60%) had undergone complete resection, and 16 (40%) had residual tumor at the time of referral. Of those 16 patients with residual tumor, 12 were selected to receive additional surgery based on the extent and location of disease and the child’s neurological status at the time of referral. Of those 12 patients, gross total resection was achieved in 10 and near-total resection was achieved in the other 2 during the second surgery. The overall gross or near-total resection rate for the entire cohort was 36 of 40 (90%). Successful gross total resection at second-look surgery may prolong survival time as well as allow a lower dose of radiation to be used, leading to a lower incidence of neurocognitive deficits.⁶⁷ Significant morbidity occurred in only one child at the time of second resection. Children who had a second operation within 30 days of the initial procedure had an improved performance level at 4 and 24 weeks after the second surgery and a trend toward a lower complication rate than those who had second-look surgery more than 30 days after the initial procedure.⁶⁷

Children with localized recurrent disease that is amenable to surgical resection should also undergo reoperation to reduce the tumor burden to minimal disease. In most cases, this should be followed with some form of radiotherapy, either radiosurgery or repeat conformal radiotherapy. In some cases, this aggressive treatment of recurrent local disease has led to long-term progression-free survival.

Surgical Technique

Preoperative planning and a careful review of all imaging are necessary to determine the extent of the tumor because ependymomas are notorious for extension through the foramen of Luschka and into the CPA and for extension down through the foramen magnum into the upper cervical spinal canal. Interpreting ependymoma imaging is not straightforward because one portion of an ependymoma may enhance with intravenous gadolinium while other portions of the same tumor fail to enhance. Careful comparison of the T2 and T1 images, as well as the enhanced and nonenhanced T1 images, can help the surgeon determine the true extent of tumor as well as plan the degree of resection indicated.

Childhood ependymoma can be grossly divided into supratentorial, third ventricular, posterior fossa, and spinal types. Surgical management of supratentorial ependymoma is relatively straightforward; the tumor is very distinct from the surrounding normal brain, providing for clear surgical margins. Identification of the vascular supply to the tumor on preoperative magnetic resonance angiography in conjunction with early control of the vascular supply can facilitate the operation and minimize blood loss. Frameless stereotaxy is an excellent adjunct for scalp flap localization relative to the tumor. Despite a recent trend toward minimally invasive surgery, a generous craniotomy should be performed so that all margins of the tumor are clearly visible at the time of surgery. In areas of noneloquent brain, a thin rim of normal white matter should be resected around the ependymoma to ensure complete resection and to devascularize the tumor. Ependymomas do not contain functioning neural tissue and can usually be safely resected from areas of eloquent cortex. Supratentorial ependymomas that are truly located within the ventricles are rare. Some supratentorial ependymomas abut or compress the ventricle without actually entering it. Third ventricular ependymomas are also quite rare.

Preoperative planning should consider the different surgical risks and strategies associated with posterior fossa ependymomas arising from the cerebellar hemisphere, the roof of the fourth ventricle, the floor of the fourth ventricle, or the CPA (Fig. 200-4). Fourth ventricular ependymomas are generally supplied by branches from the posterior inferior cerebellar artery. Preoperative MRI is important so that the main artery feeding the tumor can be identified ahead of time.

FIGURE 200-4 Axial T2-weighted magnetic resonance imaging of a child presenting with headaches and head tilt. This variant, referred to as cerebellopontine angle ependymoma, is thought to arise from ependymal rests at the foramen of Luschka. It grows into the fourth ventricle as well as down along the side of the brainstem, often encompassing the cranial nerves and blood vessels. These tumors often grow through the skull base foramina, such as the porus acusticus and the jugular foramen. The surgical approach demands access to both the midline and the retrosigmoid region.

The patient is positioned prone with maximal tolerated flexion so that the floor of the fourth ventricle is sloping away from the surgeon. If there is considerable hydrocephalus, a ventriculostomy may be placed before opening the posterior fossa. Care should be taken to release CSF slowly so as not to precipitate upward herniation or shifts in the tumor, with consequent intratumoral hemorrhage. Precipitous release of CSF and consequent lowering of intracranial pressure can also lead to retinal hypoperfusion with retinal infarct. Craniotomy is preferred to craniectomy at our institution because it decreases the incidence of CSF leak and pseudomeningocele as well as the incidence of chronic postoperative pain.⁶⁸ The presence of a bone flap also facilitates surgery for residual or recurrent ependymoma.

The craniotomy should be a generous size to allow easy removal of the tumor. If the tumor is large or extends down into the spinal canal, a C1 laminectomy should also be performed. If the tumor extends rostrally, the bone should be removed up to the level of the torcula without actually exposing the venous sinuses to diminish the chance of hemorrhage from sinus injury or sinus thrombosis. In all cases of posterior fossa ependymoma, the foramen magnum should be removed to increase the exposure and open up pathways for CSF flow. The dura is opened in a standard Y-shaped fashion. Extra care should be taken in very young children because the posterior fossa dura can be extremely vascular, and an imprudent surgeon may have tremendous blood loss at this point in the surgery. If the dura over the cerebellum is tight, the dura over the upper cervical cord can be incised to allow the slow egress of CSF to bring down intracranial pressure. Alternatively, CSF can be withdrawn through a previously placed ventriculostomy. After the dura is opened, the cerebellar tonsils are identified and separated to find and follow the posterior inferior cerebellar arteries. The vascular supply to the tumor invariably comes off the posterior inferior cerebellar artery, distal to the tonsillar loop. Coagulation of large feeders to the tumor from the posterior inferior cerebellar artery greatly reduces the vascularity of the tumor and diminishes intraoperative bleeding.

Attention is then turned to the superior pole of the tumor, defining its margin circumferentially and then debulking it as it becomes devascularized. This continues until the aqueduct is visualized and opened, whereupon attention is turned to the inferior pole of the tumor. The inferior pole is dissected along its lateral and posterior margins, taking care not to injure the floor of the fourth ventricle. At this point in the operation it is often possible to determine whether the tumor is arising from the roof or the floor of the fourth ventricle. If the tumor arises from the roof, complete excision is possible. If the tumor arises from the floor, a portion of the tumor is left on the floor of the ventricle. The lateral margins of the tumor are dissected out to diminish the tumor’s blood supply, followed by debulking of the tumor using either the bipolar suction technique or the ultrasonic aspirator.

Once the entire tumor has been removed, except for the portion on the floor of the fourth ventricle, the surgeon should visualize the floor of the fourth ventricle both above and below the remaining tumor and imagine a line connecting the superior floor with the inferior floor (cutting through the tumor). This line must not be transgressed by the surgeon; in fact, the tumor should probably be cauterized to within 1 to 2 mm of this line. This technique usually results in minimal morbidity while leaving an acceptable amount of residual tumor. Injury to this area through an overzealous bipolar approach can lead to cranial nerve VI and VII palsies as well as damage to the medial longitudinal fasciculus. At the end of the procedure, it is critical to look for the foramen of Luschka bilaterally to avoid missing a small piece of tumor. Care should also be taken to dissect out the cerebellar tonsils and open up the CSF flow pathways to diminish the need for a future CSF diversionary procedure. Patch grafting of the dura may also help open up the CSF pathways and reduce the need for future shunt surgery. The bone flap is replaced to reduce the incidence of pseudomeningocele.

CPA ependymomas typically originate from the lateral wall of the brainstem at the junction of the pons and the medulla (see Fig. 200-4). Their slow growth results in the encasement of multiple cranial nerves by the time of diagnosis (cranial nerves V to X are all at risk). Unilateral growth rotates the brainstem, distorting the normal anatomy and disorienting the surgeon. CPA ependymomas can grow back up the foramen of Luschka into the fourth ventricle, filling the ventricle with tumor. Like all posterior fossa ependymomas, they derive their blood supply from the posterior inferior cerebellar artery; however, they may also receive significant blood supply from the anterior inferior cerebellar artery, which runs through the CPA cistern. CPA ependymomas are often large by the time they come to clinical attention, and they occur predominantly in infants, making blood loss a major issue. In these cases, it is vital to have washed red blood cells available for transfusion to avoid high levels of potassium from the excessive transfusion of unwashed red blood cells in a very young child.

Removal of a CPA ependymoma is associated with very high morbidity and mortality and should not be attempted by inexperienced tumor surgeons. The patient is positioned prone with the head rotated so that the lateral portion of the cerebellum is highest. A curvilinear incision that goes to the midline is used, with extension into the neck if the tumor involves the cervical spine. Knowledge of vertebral artery anatomy is critical when dissecting out laterally with the patient’s head rotated. The bone is opened along the sigmoid sinus, and removal proceeds as close to the jugular foramen as possible. The dura is opened, and the intraspinal portion of the tumor is removed first, taking care not to injure the spinal roots. With the spinal portion removed, the vertebral artery and its posterior inferior cerebellar artery branch can be traced to find and remove large feeder vessels to the tumor. The tumor is then debulked laterally until cranial nerves IX and X are identified entering the jugular foramen. Cranial nerves VII and VIII are identified in the CPA; hearing is invariably lost even if nerve VIII remains in continuity. This can be significant in children who may go on to receive ototoxic chemotherapy. After identification of the cranial nerves, the portion of the tumor within the fourth ventricle is removed before the lateral portion is addressed. After the fourth ventricle has been cleaned out, the foramen of Luschka is explored, and tumor is removed. Finally, with sufficient room in the posterior fossa, the most difficult part of the operation—removal of the tumor from the lower cranial nerves—is attempted. Care must be taken not to injure perforating arteries from the vertebral artery and basilar artery as they course through the CPA cistern.

Cranial nerve morbidity is high following the removal of a CPA ependymoma. Children are kept intubated until the vocal cords can be checked by an otolaryngologist. Vocal cord paralysis, when present, should be unilateral. Complete unilateral vocal cord paralysis in children who cannot handle their secretions mandates a tracheostomy. If there is some vocal cord movement, most children can be safely extubated and closely observed. Within 48 hours of extubation we perform a swallowing test. It is not uncommon for children to be able to manage their secretions (and not require a tracheostomy) but still require a gastrostomy. Although many children have VI, VII, IX, or X nerve palsies postoperatively, more than 90% experience significant recovery within 3 to 6 months. Most children who initially require a tracheostomy or a gastrostomy can have it removed during their convalescence. One additional complication of surgery for CPA ependymoma is the presence of residual tumor on postoperative imaging. In almost all cases, this should precipitate another operation to achieve gross total resection. Although it is a morbid procedure, surgery for CPA ependymoma offers affected children their only meaningful chance of survival.

Treatment of Hydrocephalus

Many children with ependymoma have some element of hydrocephalus at the time of presentation, particularly those with posterior fossa tumors. Currently, most neurosurgeons treat such children with corticosteroids rather than with ventriculoperitoneal shunts. This dramatically improves symptoms associated with pressure and edema and allows time for surgery to be arranged. Preoperative shunting is avoided when possible to avoid making children shunt dependent postoperatively. Children who present in extremis may be treated with a ventriculostomy followed by more definitive surgery to address both the tumor and the hydrocephalus.

Hydrocephalus is much less common in patients with supratentorial ependymomas. Rarely, hydrocephalus is seen in the setting of a spinal ependymoma owing to high levels of CSF protein secreted by the tumor. Hydrocephalus secondary to a posterior fossa tumor is usually noncommunicating in nature owing to occlusion of the fourth ventricle or the aqueduct. In most children, removing the fourth ventricular tumor, opening up the aqueduct, and ensuring good CSF flow at the level of the foramen magnum will restore CSF dynamics, obviating the need for postoperative CSF diversion.

Postoperative hydrocephalus usually develops slowly over days or even weeks and has an incidence of 25% to 50%. The typical presentation is a pseudomeningocele at the site of the posterior fossa incision, but hydrocephalus may also present in a more traditional manner with headache, nausea, vomiting, and lethargy. Ventriculomegaly that is stable and asymptomatic after fourth ventricular tumor excision should be observed rather than treated. Patients undergoing a second surgery have a statistically significantly higher incidence of requiring a shunt compared with those treated by one surgery.⁶⁹ When a child has progressive and symptomatic hydrocephalus, treatment should involve either a CSF diversionary device (ventriculoperitoneal shunt) or an endoscopic third ventriculostomy.⁷⁰

Radiation Therapy

Postoperative radiation is part of the standard of care for ependymoma patients. Mork and Loken²⁸ demonstrated that survival was roughly 17% for patients who underwent resection alone, compared with 40% for those who underwent resection and radiation. It is acceptable to withhold radiation therapy in children with supratentorial ependymomas if gross total resection with adequate margins has been achieved and the pathology is classic ependymoma (not anaplastic ependymoma).^17,66 Additionally, radiation may not be necessary in some children after resection of an intramedullary spinal cord ependymoma.

A great deal of literature has focused on the danger of radiation therapy in ependymoma patients younger than 3 years. Numerous strategies to avoid or delay radiation have been proposed in this patient cohort in an attempt to minimize neurological, endocrinologic, and cognitive morbidity. Although the side effects associated with irradiating very young patients with ependymoma have not been well documented, they are assumed to resemble those seen in infants with medulloblastoma. A Pediatric Oncology Group study showed that young children with completely resected ependymoma in whom radiation was delayed for 2 years had a worse outcome (5-year survival 38%) than those in whom radiation was delayed only 1 year (5-year survival 88%).²⁷ These results suggest that delaying radiation for the treatment of focal ependymoma of infancy is not advisable.

Less than 5% of children with ependymoma have disseminated (leptomeningeal) disease at the time of diagnosis. Risk factors for dissemination include younger age, less than gross total resection, high proliferative index, and high-grade histology.⁷¹ Most children with ependymoma who relapse do so at the primary site, regardless of the location of the primary tumor or its grade.^2,4,5 Multiple retrospective trials have failed to show the benefit of prophylactic craniospinal irradiation in the treatment of localized ependymoma.^3,12,30,72 In light of these facts, the current recommendation for nondisseminated ependymoma is localized radiotherapy. Craniospinal radiotherapy is reserved for those children with leptomeningeal dissemination.

The radiation dose response level for ependymoma ranges from 54 to 59.4 Gy, but the optimal dose remains unclear.^73,74 Children with subtotal resection (who cannot benefit from further surgery) may benefit from dose escalation or hyperfractionation, or both, although recent reports have shown that hyperfractionation does not improve outcome. Conformal radiation therapy (CRT) limits the highest dose of radiation to the primary tumor site (as opposed to irradiating the entire posterior fossa, for example) and decreases the dose received by normal tissues (especially the inferior portion of the temporal lobes).¹⁵ The role of CRT in the treatment of children with ependymoma was tested in a phase II trial from 1997 to 2003 at St. Jude.¹⁵ Patients with localized ependymoma were evaluated before and after CRT to determine its neurological, endocrine, and cognitive effects. A total of 88 children (median age, 2.8 years) received irradiation therapy: 73 patients received 59.4 Gy, and 15 patients (age younger than 18 months and with gross total resection) received 54.0 Gy. Gross total resection was achieved through one or more craniotomies in 74 patients before irradiation; near-total resection was obtained in 6 patients, and subtotal resection in 8. This study showed excellent results at 38 months, with a 3-year progression-free survival estimate of 74.7% ± 5.7%. These impressive results show the value of gross total resection of ependymoma and that effective tumor targeting can be achieved using CRT techniques, without an increase in the relapse rate at the margin of the treatment volume. Children treated with CRT had their IQs evaluated before and after radiation; surprisingly, their IQs were normal and did not change significantly during the study period (there was trend toward improvement). The study included 48 children younger than 3 years at the time of irradiation. These results suggest that CRT is the preferred radiotherapy modality in children with localized ependymoma.¹⁵

Radiosurgery has been used to treat a number of children with ependymoma, with good local control and no excessive neurotoxicity.^75–77 It is unclear when stereotactic radiosurgery should be used instead of open resection, but it is probably appropriate for children at high risk for anesthesia and for very small tumors in locations associated with high surgical risks. Children who present with recurrent ependymoma within a previously irradiated field are treated with maximal safe surgical resection of the recurrence, followed by repeat irradiation in the form of fractionated external beam irradiation or radiosurgery as some of their options. Disease control in patients with recurrent ependymoma treated with surgery and a second course of radiation therapy has recently been reported.⁷⁸

Chemotherapy

The role of chemotherapy in the treatment of childhood ependymoma is unclear; it is not part of standard care. The response rate of ependymoma to single agents is roughly 11%, with less than 5% showing a complete response.⁷⁹ Of the chemotherapeutic agents tried in the treatment of ependymoma, cisplatin seems to be the most active.⁷⁹ In those rare children with ependymoma that responds to chemotherapy, there is no clear increase in survival. Several retrospective reviews failed to show a benefit of cytotoxic chemotherapy in the treatment of newly diagnosed ependymoma.^{2–5,8,30,73} The Children’s Cancer Group conducted a randomized controlled trial comparing children with ependymoma who received irradiation alone with those who received radiation and chemotherapy: there was no survival benefit in the arm that received chemotherapy.¹ Similarly, a Pediatric Oncology Group trial did not show any overall survival benefit for infants with ependymoma treated with dose-intensive chemotherapy. A Societé Internationale d’Oncologie Pediatrique study treated 73 children with chemotherapy for 16 to 18 months after maximal resection; the children did not receive radiotherapy. Progression-free survival estimates at 2 and 4 years were 33% and 22%, respectively.⁸⁰ These results are clearly worse than historical controls treated with radiotherapy. Despite the best efforts of the neuro-oncology community, there is no clear role for chemotherapy in the treatment of children with localized ependymoma.

Chemotherapy may have a role in avoiding or delaying radiotherapy in children younger than 3 years with intracranial ependymoma. The United Kingdom Children’s Cancer Study Group–Societé Internationale d’Oncologie Pediatrique prospective trial demonstrated that the use of adjuvant chemotherapy to delay or avoid radiation in children younger than 3 years was feasible, without affecting survival in a negative manner.⁸¹ Chemotherapy may also have a role in the treatment of children with residual tumor after initial tumor resection. For example, some authors suggest that the administration of chemotherapy between the first and second craniotomies makes the tumor borders better defined and easier to dissect.²² Chemotherapy may also have a role in the treatment of children who need to delay surgery or radiation treatment because of illness or neurological complications of the disease or its treatment.

Conclusion

Ependymoma is a neurosurgical disease, with local control being the most important prognostic factor. Failure to observe gross total resection on postoperative imaging should usually precipitate a second craniotomy to improve the likelihood of long-term disease control. CRT should be given to all children with localized ependymoma, except for a small minority with classic supratentorial tumors that are completely resected. Although cytotoxic chemotherapy currently has little role in the treatment of ependymoma, future therapeutics based on the biology of the disease offer hope for more effective, less toxic treatments.