Ependymomas continue to be one of the most challenging tumors for neurosurgeons and allied physicians. They require a multidisciplinary approach for optimal treatment outcome. They account for 8 to 14% of all pediatric central nervous system (CNS) tumors but are the third most common brain tumor, behind medulloblastoma and astrocytoma.1–3 Dissemination at the time of presentation is relatively rare, occurring in 5 to 15% of patients.4,5 Intracranial ependymomas frequently arise in the first decade of life between the ages of 4 and 6 years, with one third to one half of tumors occurring in the first 3 years of life. Ninety percent of ependymomas occur intracranially, and, of those, the ratio of infratentorial to supratentorial is 3:1 to 5:1. The current standard therapy consists of gross total resection (GTR), if possible, followed by local radiotherapy. Chemotherapy plays a less defined role in a minority of patients but may be important in younger patients (<3 years old) to delay or avoid radiotherapy. The most significant prognostic factor predicting survival in children with ependymoma is the extent of surgical resection. Although improved surgical techniques and technology have given surgeons a better chance of obtaining a GTR, there is still a significant risk of neurologic morbidity. Unfortunately, incompletely resected ependymomas almost uniformly recur, and, when this occurs, further therapies are either limited or ineffective.
Infratentorial tumors typically develop from ependymal cells lining the floor (60%) or roof (10%) of the fourth ventricle. Laterally located tumors (30%) may arise from ependyma in the lateral recess or ectopic ependymal cell rests located on the lateral aspect of the pontomedullary junction.6 Ependymomas classically extend into the surrounding subarachnoid spaces: through the foramen of Luschka and into the cerebellopontine angle (CPA) and along the anterolateral aspect of the pons and medulla, and through the foramen of Magendie and over the dorso-lateral surface of the upper spinal cord. This growth pattern, which is not unique to ependymomas but is more common in these tumors, has been termed “plastic ependymoma.”7 There are three World Health Organization (WHO) grades of ependymoma: grade I, myxopapillary and subependymoma; grade II, cellular, papillary, clear cell, tanycytic, and mixed; and grade III, anaplastic.8 In the past, ependymoblastoma was considered a grade IV tumor but is now categorized as a supratentorial primitive neuroectodermal tumor. Grade I tumors are not discussed in this chapter because they are almost never encountered in the infratentorial compartment of children.
Grade II lesions are typically well delineated and have moderate cellularity with monomorphic round-to-oval nuclei containing finely dispersed chromatin. The histological hallmarks that allow recognition of ependymomas include perivascular rosettes, ependymal rosettes, and positive labeling for glial fibrillary acidic protein (GFAP). Perivascular rosettes, usually more common than ependymal rosettes, are ependymal cell processes radially oriented toward a blood vessel that forms the center of the rosette. The nucleus is located remotely from the blood vessel. Ependymal rosettes are radial arrangements of polygonal cells with a well-defined central lumen, resembling a central canal. In grade III tumors, the nuclei are polymorphic, irregularly shaped, and hyperchromatic. These anaplastic tumors are described as having an increased cellularity (although no precise definition of “increased” has been established), brisk mitotic activity (usually 7 to 10 mitoses per 10 high-power fields), microvascular proliferation, and pseudopalisading necrosis. The histological subdivision of grade II tumors is generally not thought to be of any value ( Figs. 37.1 and 37.2; Table 37.1 ).
Published data correlating the histological grade with outcome are inconclusive. Although some authors have found no statistical difference in patient survival, most authors believe that patients with anaplastic ependymomas have a poorer outcome.9–16 For example, Merchant et al9 found that the 3-year progression-free survival (PFS) rate was 28% for patients with anaplastic ependymoma (grade III) compared with 84% for patients with differentiated ependymoma (grade II). One reason for the inconsistent association between grade and outcome may be the lack of consensus among neuropathologists regarding the diagnostic features that define and classify ependymomas discussed in the previous paragraph.17 Several authors have found that the degree of mitotic activity as measured by the MIB-1 labeling index (Ki-67) is an independent predictor of prognosis.18–22 Wolfsberger et al22 found that low (< 20.5%) or high (≥ 20.5%) Ki-67 indices predicted favorable (≥ 5 years) or unfavorable (< 5 years) patient outcome of 79% or 70%, respectively.
It is clear that ependymomas are a clinical and molecularly heterogeneous group of tumors, although tumors from the same location (e.g., infratentorial) or from different locations (e.g., infratentorial versus supratentorial) are histologically indistinguishable. There is currently an effort by many researchers to better define and classify brain tumors based on their molecular “fingerprint” relative to the WHO classification.23,24 This effort has been difficult with ependymomas because up to half of those in an infratentorial location demonstrate a balanced genomic profile; the genetic alterations that have been reported are numerous, and, until recently, the cellular origin of ependymomas has been unknown.25 The most frequently observed chromosomal abnormalities include losses on chromosomes 6q, 17p, and 22q and gains on 1q and 9q. Less frequent genetic alterations involve losses on chromosomes 7p, 9p, 10q, 13q, and 16q and gains on 4q, 5, 7, 8, 12q, 15q, 18, and 20.26 Chromosome 22 loss has been reported to be the most common copy number alteration in intracranial ependymomas.25 The group at St. Jude Children′s Research Hospital has made significant advances in understanding the molecular basis of ependymomas. They used messenger RNA (mRNA) profiles to segregate ependymomas by central nervous system (CNS) location and unmasked previously unknown subgroups among supratentorial, posterior fossa, and spinal ependymomas, classifying ependymomas into nine distinct subgroups (subgroups A to I).27 They identified potential ependymoma oncogenes, which included several regulators of stem cell proliferation, pluripotency, and neural differentiation. These include THAP11, PSPH, EPHB2, ten genes within the PCDH cluster, KCNN1, RAB3A, PTPRN2, and NOTCH1. They also discovered that radial glial cells, which are neural stem cells, are likely the cell of origin for all ependymomas, and hypothesized that distinct populations of radial glial cells with specific genetic alterations will give rise to certain ependymomas.28 This finding enabled the researchers to create the first ependymoma mouse model. The specific combination of embryonic cerebral radial glial cells, deletion of Ink4a/Arf, and amplification of EphB2 generated supratentorial ependymomas27 ( Tables 37.2 and 37.3 ).
Cranial computed tomography (CT) is typically the first-line imaging study in many patients when they present to the local emergency room or pediatrician. CT scans are particularly useful to search for hydrocephalus, intratumoral cysts, and calcifications. All patients require high-quality, multiplanar preoperative craniospinal magnetic resonance imaging (MRI) that details the location and extent of the tumor. Ependymomas are usually well-demarcated tumors, hypointense on T1-weighted images and hyperintense on T2-weighted and fluid-attenuated inversion recovery (FLAIR) images with a mixed pattern of enhancement.29 Careful evaluation of the images is critical to determine areas of nonenhancing tumor. Some ependymomas can be quite vascular, as evidenced by intratumoral flow voids. Areas of cystic degeneration and calcifications are not uncommon. The growth of tumor into the posterior fossa subarachnoid spaces, particularly the cisterna magna, cervical subarachnoid space, and cerebellopontine and cerebellomedullary angles, is the radiological hallmark of this neoplasm. In comparison with ependymomas that arise from within the fourth ventricle, those that originate in the lateral recess preferentially expand into the cerebellopontine and cerebellomedullary cisterns with little expansion into the fourth ventricle. Complete, biplanar, contrast-enhanced spinal imaging is best done before surgery, but if it is done postoperatively, it should be delayed for 1 to 2 weeks because postoperative inflammation caused by surgery can mimic metastatic disease ( Table 37.4; Figs. 37.3 and 37.4 ).
Losses on 6q, 17p, 22q
Gains on 1q, 9q
Postoperative changes, seen as enhancement around the re-section cavity, may begin within 24 hours of surgery. Therefore, it is recommended that imaging after surgery should be done within the first 48 hours. New contrast enhancement can occur months to years after surgery, and the question is whether it represents true tumor recurrence or delayed changes due to radiochemotherapy, so-called pseudoprogression.30 Adjunct imaging studies such as positron emission tomography (PET), proton magnetic resonance (MR) spectroscopy, and perfusion MRI can help to differentiate tumor recurrence from posttreatment changes.29 Patients must be monitored with serial imaging at a frequency that is determined by the fact that recurrence or progression typically occurs 12 to 24 months after treatment. It is generally believed that the earlier tumor progression is discovered, the more likely salvage treatments will have some benefit.
Extent of resection has consistently and repeatedly been demonstrated to be the most important determinant of recurrence and survival. 6,9,10,14,15,20,21,31–39 Every effort must be made to obtain a GTR because the predominant pattern of failure is local. GTR is defined as absence of tumor on the postoperative MRI, even if there is a thin carpet of tumor left on the floor of the fourth ventricle or adherent to cranial nerves (and visualized with the operating microscope at the time of surgery). Subtotal resection (STR) is diagnosed when there is gross residual tumor remaining after surgery that was visible to the surgeon intraoperatively or is visible on postoperative MRI. A near-total resection (NTR) has been arbitrarily defined as < 1.5 cm3 on postoperative imaging (similar to the value quoted for medulloblastoma40) and/or 0.5-cm residual thickness of tumor bed enhancement.
Complete resection is more difficult to obtain in infratentorial than in supratentorial ependymomas, because it may be possible in the latter to obtain a margin of normal tissue surrounding the tumor. With infratentorial tumors, GTR is often difficult at best or impossible at worst because of the proximity of and adherence to three critical structures: the floor of the fourth ventricle; the cerebellar peduncles (typically inferior and middle, rarely the superior); and the middle to lower cranial nerves. In our experience, CPA ependymomas, unlike medulloblastomas, encase rather than displace the cranial nerves and rotate and elevate the lower brainstem. The cranial roots and rootlets in infants are extremely thin and vulnerable. As mentioned previously, this type of ependymoma is thought to arise from the lateral aspect of the brainstem, at the pontomedullary junction where the tumor is frequently found to be densely attached.6 Families must be counseled preoperatively of the risk of neurologic impairment, particularly various cranial neuropathies such as facial palsy, hearing loss, dysphonia, and dysphagia leading to the need of a tracheostomy and gastrostomy.41 Post-operative neurologic deficits are maximal immediately after surgery, but as long as the continuity of the cranial nerves is preserved, recovery is usually the rule. The one major exception is hearing loss, which is often irreversible.42
The importance of making GTR the highest priority cannot be overstated. Trying to minimize the risk of violating the “first do no harm” rule by leaving ependymoma behind can, in fact, cause more harm in the future by virtually sealing that child to a dismal fate. Although ependymoma is relatively radiosensitive, as discussed in the next section, that fact is not enough to eradicate the extremely negative prognosis carried by the presence of residual tumor. In the past, it was thought that NTRs were essentially equivalent to GTR in terms of outcome. More recent data, however, have shown that patients with NTR have a survival that is more like that of patients in the STR group. Thus, a neurosurgeon who believes, a priori, that a complete resection is not achievable should strongly consider having the child transferred to a facility with more experience. With improvement in surgical techniques, we have seen our operative time, blood loss, and neurologic morbidity decrease, our GTR rate increase to 96%, and our 5-year progression-free survival (PFS) and overall survival (OS) rates increase to 54% and 64%, respectively.6
Intratumoral cysts/Ca+ common
Well demarcated, mixed enhancement
Hypointense on T1
Hyperintense on T2/FLAIR
Cystic degeneration, calcification common
Grow into subarachnoid spaces/cerebellopontine angle
If postoperative imaging demonstrates more than 1 mL3 of residual tumor, repeat or second-look surgery should be considered if it is thought that the child can tolerate further surgery and the residual tumor is accessible and potentially resectable. The timing of second-look surgery is of some debate. In some situations, the oncologist may recommend chemotherapy between the initial and second resection to make the second surgery easier by decreasing the tumor volume or vascularity.43 For patients with recurrent disease, as with the initial occurrence, radical resection has been shown to be the most important factor in survival.38,44–47 The same issues are present when counseling patients and their parents regarding surgery for recurrent disease as with initial diagnosis, with one major exception. Arachnoid scarring resulting in adhesions within tissue planes makes dissection, identification of normal structures, and resection of recurrent tumor more challenging. Therefore, the aggressiveness of surgery, the associated neurologic risks, and the preservation of an acceptable quality of life must be openly evaluated to determine the most appropriate course of treatment.