Metastatic brain involvement in Ewing family of tumors in children

The Ewing family of tumors (EF) is a group of malignant, presumably neuroectodermally derived and histologically distinct neoplasms, ranging from the poorly differentiated Ewing's sarcoma to the more differentiated peripheral primitive neuroectodermal tumor. They arise most often from bone and less frequently from soft tissue. A common origin has been suggested on the basis of cytogenetic studies and the presence of the MIC2 cell surface protein in more than 95% of patients.^1,2 The capability of these tumors to recur locally as well as to metastasize to multiple organs is well recognized and dictates the therapeutic approach. The most predominant sites of EF metastasis are lung (38%), bone (including the spine; 31%), and bone marrow (11%).³ An important site of spread, which sometimes can be crucial to the patient's prognosis, is the brain.

The exact incidence of CNS involvement in EF is not known, especially since the introduction of the relatively new definition of EF and the advent of MRI as the major neuroimaging technique for the evaluation of affected patients.

Studies in the early 1970s suggested that the CNS is a common site of distant treatment failure; however, in the recent literature, the actual frequency of CNS involvement is considered to be less than 5%.³

In this retrospective study we evaluated the various aspects of intracranial involvement in EF to determine its incidence, clinical presentation, and implications for treatment.

Patients and methods. The charts of all patients treated in the Department of Pediatric Hematology/Oncology from 1972 through 1997 were reviewed retrospectively. Pathologic criteria were typical histologic appearance of undifferentiated small, round cell neoplasms² plus immunohistochemical¹ and ultrastructural findings consistent with the diagnosis of EF, ruling out other neoplasms such as lymphoma, neuroblastoma, or rhabdomyosarcoma. Cytogenetic studies were also performed in some patients. The diagnosis was made in all patients during the initial workup based on material that was obtained from the mass. Patients with EF and intracranial involvement documented by either brain CT or MRI were included in the study. In each of the patients with intracranial metastasis, tissue was obtained by surgery, and the diagnosis of metastatic tumor was confirmed. However, Patient 2 was too sick to undergo surgical intervention, and the diagnosis of brain metastasis was made according to the MRI appearance of the lesions and the clinical workup.

Results. During this 25-year period, 80 patients with EF were treated and followed in our center, of whom eight had intracranial involvement. Pertinent data are summarized in the table. The female-to-male ratio of the subgroup with brain involvement was 1:1 compared with 0.9:1 of the whole group. Mean age at diagnosis of the original tumor was 12.3 ± 8.4 years compared with 12.8 ± 6.2 for the whole group. Intracranial involvement was demonstrated 1.3 to 11 years from initial presentation. Median time from initial diagnosis to the finding of CNS involvement was 3 years (mean, 3.9 ± 3.1 years). Five children had only one metastasis, and three had more than one (see the table). Seven patients (8.8%) had parenchymal brain involvement that was secondary to hematogenous spread in five children and a result of direct spread from bone metastasis in two children. One patient (Patient 8) had an orbital and retro-orbital tumor without clear brain involvement. Brain metastases were hemispheric in five patients, cerebellar in one, and located in the basal ganglia in one. Clinically, CNS involvement was manifested by hemiparesis in three patients, seizures in three, and CNS bleeding in one. Two patients (Patients 3 and 5) had seizures as an early phenomenon, without any clear connection to chemotherapeutic treatment or evidence of any biochemical abnormality or infection. Imaging studies (CT and MRI) at that time did not show CNS metastasis.

In three patients the brain metastasis was completely resected, in three a biopsy was obtained, and in one no surgical intervention was possible because of the poor general condition of the patient. One patient underwent emergency surgery for bleeding within the metastatic lesion and she died soon afterward. All patients were subsequently given chemotherapy, and two underwent radiation therapy as well.

Average survival in the eight patients was 5.3 ± 4.1 years. Five-year survival was 25% compared with 51% in the whole group (see the table).

Discussion. In our study the rate of intracranial involvement in EF was 10% and unequivocal intraparenchymal brain involvement occurred in 8.8%. These figures are much lower than the 56% reported in 1974 by Mehta and Hendrickson⁴ in 27 patients with Ewing's sarcoma. In most of their patients the diagnosis was based on neurologic signs and symptoms of backache and blurred vision. Imaging was performed in only eight patients: five CT scans (of which four were pathologic) and three myelograms. Therefore we suspect that a clear differentiation between actual brain metastasis and neurologic complications of antitumoral treatment, extra-axial masses, or spinal involvement could not be achieved in their study.

Trigg et al.⁵ noted a 10% risk of isolated CNS in Ewing's sarcoma, with no difference between patients who received intrathecal methotrexate and cranial irradiation as CNS prophylaxis and those who did not. According to their results, patients were at greatest risk for CNS involvement within the first 2 years of diagnosis.

There are two principal modes of metastatic spread to the brain in EF. The first is direct extension from the skull, which may be the site of both primary and secondary Ewing's sarcoma to the brain.^6,7 In our study this was observed in two patients. The other is via the hematogenic route. Of the 15 patients with CNS involvement in the study of Mehta and Hendrickson,⁴ 10 patients appeared to have direct extension of the disease to the CNS, and five had hematogenous spread. Marciani et al.⁸ suggested that CNS involvement in Ewing's sarcoma is uncommon and often due to direct extension of tumor from an extra-axial origin. They described a 14-year-old girl with right temporal metastasis of Ewing's sarcoma that appeared independently from an earlier skull metastasis on the opposite side and thus probably seeded through the hematogenous route. There have been more reports of intracerebral metastases of Ewing's sarcoma⁹ as well as intracerebellar metastases.¹⁰ According to our study, the hematogenous route of brain involvement is more frequent than direct extension.

CNS involvement in EF may cause significant morbidity. Olivi et al.¹¹ described a 30-year-old woman with spontaneous intratumoral hemorrhage of a brain metastasis from Ewing's sarcoma. This also occurred in one of our patients. It should be stressed that such bleeding is life-threatening. Other complications are hemiparesis and seizure.

The earliest demonstration of CNS involvement by imaging studies in our patients was 1.8 years after diagnosis. The mean time for the whole group was 3.9 years. In 1971, Marsa and Johnson¹² described two cases of isolated CNS involvement in Ewing's sarcoma in the absence of disease in other sites. They raised the possibility that there may already be widespread subclinical metastases as early as the time of diagnosis. If these deposits occur in pharmacologically protected sites, such as the CNS, they could lead to treatment failure. The early seizures in two of our patients could represent early subclinical spread to the CNS.¹² On the other hand, seven of our eight patients had systemic metastases at the time of diagnosis of intracranial disease. If the CNS is a site for early subclinical metastases, one may expect at least a higher percentage of isolated intracranial involvement occurring before systemic spread is found.

The presence of brain metastases was associated with poorer overall prognosis, as manifested by the 5-year survival of 25% compared with 51% in the whole group of patients. Thus, in some patients the finding of CNS metastases may imply poor prognosis, either because of an already widespread tumor or because such involvement is a marker for a more aggressive or chemoresistant tumor. It is of interest that four of five patients whose CNS disease seemed to arise from hematogenous dissemination had pulmonary metastases, supporting the view that the poor prognosis is related to widespread disease. Of note is the study of Bindal et al.¹³ in patients with sarcoma metastatic to the brain showing that the complete surgical removal of all brain metastases and a preoperative Karnofsky performance score of >70 are associated with a favorable prognosis. In their study, patients with concurrent lung metastases did not worse than those with no evidence of systemic disease; thus, they concluded that the presence of lung metastases is not a contraindication for surgical excision of the brain metastases.

In our patients, however, it was impossible to determine objectively whether the resection of the metastases had any significant influence on the course of disease, except for a subjective feeling of improvement in the patients' well-being and a possible delay in disease progression in two of the three who underwent complete resection.