Gliomatosis cerebri: 20 Years of experience at the Children's Hospital of Philadelphia


  • Gregory T. Armstrong MD,

    Corresponding author
    1. Division of Oncology, The Children's Hospital of Philadelphia, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    • Department of Epidemiology and Cancer Control, St. Jude Children's Research Hospital, 332 North Lauderdale Street, Mailstop 735, Memphis, TN 38105
    Search for more papers by this author
    • Fax: (901) 495-5845

  • Peter C. Phillips MD,

    1. Division of Oncology, The Children's Hospital of Philadelphia, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Lucy B. Rorke-Adams MD,

    1. Department of Pathology, The Children's Hospital of Philadelphia, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Alexander R. Judkins MD,

    1. Department of Pathology, The Children's Hospital of Philadelphia, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • A. Russell Localio PhD,

    1. Department of Biostatistics and Epidemiology, School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Michael J. Fisher MD

    1. Division of Oncology, The Children's Hospital of Philadelphia, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author



Gliomatosis cerebri (GC) is a rare and typically fatal glial neoplasm of the central nervous system. In this report, the authors describe the largest cohort of children to date with GC and explore relations between potential prognostic factors, treatment, and survival.


Imaging, pathologic, and outcome data were reviewed from 13 patients who were diagnosed with GC and were treated at the Children's Hospital of Philadelphia (CHOP) between 1982 and 2005. All patients had GC confirmed by biopsy. Twelve patients received cranial irradiation, and 8 of those patients received adjuvant chemotherapy. A single patient age 1 year received chemotherapy alone. A review of the literature identified 51 pediatric patients with GC.


The progression-free survival rate in this study was 13% (range, 1.5–43 months), and the overall survival (OS) rate was 64% (range, 6.5–67 months) at 2 years. OS was significantly shorter for patients who presented in the first decade of life (P = .04). The time to progression was prolonged significantly for patients who had no evidence of tumor enhancement on imaging studies (P = .03). When survival data from patients reported in the literature were combined with the CHOP cohort, treatment prolonged OS significantly (P = .003).


The outcome of pediatric patients with GC was extremely poor; however, the current results indicated that treatment may prolong OS. Age < 10 years and contrast enhancement on magnetic resonance imaging studies at diagnosis may be risk factors for shorter survival in pediatric patients with GC. Cancer 2006. © 2006 American Cancer Society.

Gliomatosis cerebri (GC) is a rare, diffusely infiltrating glial neoplasm of the central nervous system with a highly variable presentation, poorly defined clinical course, and typically fatal outcome. Nevin first used the term gliomatosis cerebri in 1938 to describe a diffuse overgrowth of neuroglial cells in wide areas of the cerebral hemispheres.1 The World Health Organization (WHO) further defined GC as a neuroepithelial neoplasm of unknown origin that involved at least 2 cortical lobes and often extended into infratentorial structures of the brain with preservation of the anatomic architecture and sparing of neurons.2 Before the introduction of magnetic resonance imaging (MRI), the diagnosis of GC typically was made postmortem.3 Modern imaging techniques combined with biopsy confirmation now allow earlier diagnosis and the opportunity to define further the natural history of the disease and its response to therapeutic interventions.4–6

More than 160 patients with GC have been reported in the literature, and the vast majority of those patients were adults. The largest pediatric case series to date included only 3 patients.7 That series suggested a relation between GC, intractable epilepsy, and developmental delay in early childhood. In this article, we report our experience at The Children's Hospital of Philadelphia (CHOP) of 13 patients who were diagnosed with GC (the CHOP cohort). To our knowledge, this is the largest case series of pediatric patients with GC from a single institution to date, and it allowed us to evaluate the impact of age, presenting radiographic features, and treatment on progression-free survival (PFS) and overall survival (OS). For purposes of comparison and discussion, we include our review of 51 pediatric patients with GC reported in the literature (the historic cohort).


We used the tumor registry data base of the Division of Oncology at CHOP to identify 13 patients who were diagnosed with GC from 1982 to 2005 (Table 1). To maintain comparability with recent reviews of GC in the adult literature, we limited our cohort to those patients who met criteria for the definition of primary GC.7, 8 According to WHO criteria, primary GC can be classified into 2 types: Type 1 presents with diffuse invasion of cells without an obvious tumor mass, and Type 2 involves the diffuse invasion of cells beyond the lobe of a primary mass. Patients who presented with a solitary mass that later disseminated (i.e., secondary GC) were excluded.

Table 1. The Children's Hospital of Philadelphia Cohort: Clinical Characteristics, Treatment, and Survival
PatientAge at diagnosis, YearsGenderClinical presentationTime to diagnosis, Months*Radiation dose in cGy (Field)Initial chemotherapyPFS, MonthsLOS, MonthsAlive
  • cGy indicates centigrays; PFS, progression-free survival; LOS, length of survival; IF, involved field; WB, whole brain; HD-MTX, high-dose methotrexate; VCR, vincristine; Carbo, carboplatin; PCV, procarbizine, lomustine, and vincristine; NF-1, neurofibromatosis type 1; TMZ, temozolomide.

  • *

    Time to diagnosis was measured from the onset of symptoms.

  • Survival was measured from the time of histologic diagnosis.

  • This patient received 2 cycles of HD-MTX preirradiation.

  • §

    This patient received VCR concurrent with irradiation.

  • This patient received TMZ postirradiation.

  • This patient received TMZ concurrent with radiation and postirradiation.

118MaleHemiparesis/facial weakness15500 (IF)None430No
214MaleHemiparesis77010 (IF)HD-MTX1425No
38FemaleAtaxia, limited upgaze, hemiparesis25700 (IF)None417.5No
413MaleMalaise, poor school performance, clumsiness, dizziness123960 (WB), 1440 (IF)None4.564No
516MaleDiplopia, dysarthria, lethargy34680 (WB)None1519No
611FemaleIntention tremor185580 (IF)VCR/Carbo2130No
712MaleSeizure3.55040 (WB)PCV4367No
813MaleSeizure0.253600 (WB), 1800 (IF)PCV22.527.5No
94FemaleHeadache, emesis, ataxia0.752700 (WB), progressedVCR§1.56.5No
109MaleSeizure, visual hallucinations, altered mental status0.252500 (WB), 2800 (IF)TMZ1520No
111.6MaleSeizure, developmental delay, NF-120NonePCV2222Yes
128MaleIntention tremor15800 (IF)TMZ1116Yes
1312MaleSeizure65040 (IF)TMZ57Yes

All patients had a brain lesion identified prior to biopsy by a neuroimaging study (Fig. 1). The diagnosis of GC was confirmed by biopsy in all patients. All specimens were reviewed by one neuropathologist (L.B.R-A.). Clinical data and radiographic characteristics were extracted from inpatient and outpatient medical records. The time to diagnosis was defined as the time from onset of first symptom to histopathologic confirmation of disease.

Figure 1.

These are magnetic resonance images from 2 patients with gliomatosis cerebri (GC) at the time of presentation. (A) In Patient 12, an axial fat-suppressed, fluid-attenuated, inversion recovery image shows bilateral thalamic enlargement and hyperintensity with cortical thickening and hyperintensity of the right temporal cortex. (B) In Patient 7, a coronal T2-weighted image shows right temporal lobe hyper intensity and extension of GC across the midline.

Histologic diagnosis of GC was based on the observation of diffuse, irregular, parenchymal infiltration of glial cells; perivascular and perineuronal orientation of tumor cells; and subpial growth of similar cells (Fig. 2). Neoplastic cells in the majority of patients were small, fiber-forming astrocytes that were recognized most easily by a glial fibrillary acidic protein preparation using an immunoperoxidase technique. In many instances, the cell nuclei and cell bodies were smaller than the normal resident population or reactive glial cells. A remarkable feature in some of these tumors was a prominent population of microglia in the form of rod cells. Conspicuous was the absence of neovascularization, necrosis, or mitotic activity. Instead, the overall picture was similar to that described by Scherer as a feature of gliomas and generally referred to as ‘Scherer's secondary structures.’9, 10 However, in GC, this is the primary feature and the not secondary feature. The underlying architecture was attenuated as the neoplastic cells infiltrated diffusely, but there was no mass as such. Instead, there was an overall increase in volume of the infiltrated region.

Figure 2.

These photomicrographs show representative histology in patients with gliomatosis cerebri. (A) In Patient 8, a section of cerebral cortex demonstrates perivascular and perineuronal accumulation of neoplastic glial cells characteristic of gliomatosis cerebri (hematoxylin and eosin [H&E] stain). (B) In Patient 8, a photomicrograph shows 2 neurons in the center of the field and a small vessel to the right of the lower neuron surrounded by neoplastic glial cells (H&E stain). (C) In Patient 10, a photomicrograph shows perivascular orientation of neoplastic glial cells. The immunoperoxidase method to determine the expression of glial fibrillary acidic protein (GFAP). (D) In Patient 10, a photomicrograph shows perineuronal accumulation of neoplastic astrocytes (arrow). The immunoperoxidase method was used to determine the expression of GFAP, original magnification ×200 (A); ×600 (B-D).


No national or institutional protocol existed for the systematic treatment of GC during the time of this study. All patients age >2 years (n = 12 patients) received cranial irradiation as initial therapy. Three patients received whole-brain irradiation, 3 patients received whole-brain irradiation plus a cone down to the involved field, and 6 patients received involved-field irradiation alone. The total irradiation dose ranged from 4680 centigrays (cGy) to 7010 cGy, and 70% of patients received between 5040 cGy and 5760 cGy. One patient (Patient 9) (Table 1) had apparent tumor progression during irradiation, and radiotherapy was discontinued at parental request after the administration of 2700 cGy to the whole brain. Because of the known aggressive behavior of this tumor and historically poor survival, 8 of 12 patients who were treated with irradiation received adjuvant chemotherapy. A single child (Patient 11) age 1 year received chemotherapy as the sole treatment modality. At the time of recurrence, 8 patients received systemic chemotherapy, and 2 children (Patients 8 and 13) received additional radiation to the spinal cord for tumor progression that involved that region.


All patients were followed in the Neuro-Oncology clinic at CHOP. Before the routine availability of MRI, the primary tumor site was evaluated by computed tomography (CT) with contrast enhancement. Imaging studies of the spinal cord were not obtained routinely at diagnosis or follow-up unless directed by symptoms.

Evaluation of Outcome

Clinical status at follow-up evaluation was classified retrospectively as follows: a complete response (CR) was defined as no evidence of disease on subsequent follow-up imaging studies, a partial response (PR) was defined as a decrease >50% in the area of disease involvement (in cm2), a minimal response (MR) was defined as a decrease from 25% to 50% in the area of disease involvement, and stable disease (SD) was defined as a decrease ≤25% in the area of disease involvement. Disease progression was defined by radiographic evidence of progression that prompted a discontinuation or change of in a patient's therapy.

Statistical Analysis

PFS was measured from the date of histopathologic confirmation to the date of the documentation (on central nervous system imaging studies) of progressive disease and discontinuation of previous therapy. OS was measured from the date of the initial biopsy that revealed GC to the date of death. All calculations were performed in Stata software (version 9.0; Stata Corp, College Station, TX). Distributions of PFS and OS were estimated by using the method of Kaplan and Meier.11 Comparisons of survival distributions were made by using the log-rank test.

Review of the Literature

We used the National Institutes of Health PubMed search engine to review the medical literature for cases published as GC or glioblastosis in patients age <21 years at the time of onset of symptoms. Fifty-one patients were identified, and all had histologic confirmation by either biopsy or autopsy (Table 2). By convention, OS is reported from the time of onset of symptoms until death, because large numbers of patients were undiagnosed until autopsy. Our review of the literature was consistent with this convention.

Table 2. Historic Cohort: Treatment and Survival
ReferenceAge, YearsDiagnosisTreatmentLOS, Months*Alive
  • LOS indicates length of survival; NA, not available; RT, radiation therapy; cGy, centigrays; TMZ, temozolomide; PCV, procarbazine, lomustine, and vincristine; CDDP, cisplatin; BCNU, carmustine.

  • *

    Survival was calculated from the time of first symptoms.

Artigas et al., 1985312AutopsyNone156No
Artigas et al., 1985313AutopsyNone204No
Barth et al., 1988210AutopsyNone0.25No
Bendszus et al., 20022210BiopsyNANANA
Bendszus et al., 20022212BiopsyRT20No
Cozad et al., 19962319Biopsy5400 cGy40Yes
Cummings et al., 1999247Autopsy7265 cGy5No
del Carpio-O'Donovan et al., 19961811AutopsyNone12No
Dexter et al., 19952516BiopsyLobectomy24Yes
Dunn and Kernohan, 19572618AutopsyNone6No
Dunn and Kernohan, 19572620AutopsyNone1.5No
Elshaikh et al., 20021212Biopsy6720 cGy32.5No
Geremia et al., 1988279AutopsyCorticotropin, imuran156No
Hargrave et al., 20032815BiopsyTMZ22No
Herrlinger et al., 20022919Biopsy5400 cGy, PCV, CCNU16Yes
Horst et al., 20003012Biopsy5600 cGy21No
Horst et al., 20003012Biopsy5280 cGy24Yes
Jayawant et al., 20013110AutopsyNone3No
Jennings et al., 199570BiopsyCorticotropin13No
Jennings et al., 199570.66BiopsyCorticotropin15Yes
Jennings et al., 199576BiopsyNone10Yes
Kandler et al., 19913218AutopsyCorticotropin6No
Keene et al., 19993314Autopsy5400 cGy, BCNU/CDDP30No
Kim et al., 19981319Biopsy5580 cGy22.1No
Kim et al., 19981320Biopsy5940 cGy27Yes
Malamud et al., 19521517Autopsy5400 cGy36No
Mawrin et al., 20033417Biopsy5040 cGy, Thalidomide18No
Melignois et al., 20023514AutopsyNone1No
Onal and Bayindir, 19993613autopsyNone5No
Peretti-Viton et al., 2002416BiopsyBCNU/TMZ5Yes
Perkins et al., 20031416Biopsy6000 cGyDeadNo
Perkins et al., 20031416Resection5400 cGyDeadNo
Perkins et al., 20031416Biopsy5600 cGyDeadNo
Perkins et al., 20031416Biopsy5400 cGyAliveYes
Pyhtinen and Paakko, 19963712BiopsyNANANA
Pyhtinen 19963812Biopsy5600 cGy21No
Ross et al., 1991397AutopsyNone276No
Rippe et al., 19904016Biopsy5400 cGy12No
Sato et al., 20034115BiopsyRT, chemotherapyNANA
Scheinker and Evans, 1943428AutopsyNone24No
Shahar et al., 2002435BiopsyRT18Yes
Shahar et al., 20024310BiopsyChemotherapy24No
Shintani et al., 20004419Biopsy5000 cGy22Yes
Spagnoli et al., 1987613BiopsyNANANA
Spagnoli et al., 1987619BiopsyNANANA
Troost et al., 19874513BiopsyNone84Yes
Vates et al., 2003818BiopsyRT10No
Vates et al., 200386BiopsyRT, TMZ, Thalidomide9No
Yang et al., 20024611BiopsyNANANA
Yang et al. 20024618BiopsyNANANA
Yip et al., 2004477Autopsy4500 cGy, Carboplatin13No

Details of therapeutic intervention and length of survival were available for 40 patients. Cranial irradiation was the most common treatment modality (19 patients). Radiation doses ranged from 4500 cGy to 7265 cGy. The location and extent of radiation was reported inconsistently. Fourteen patients received radiation alone. Chemotherapy alone (3 patients) or in conjunction with radiation (5 patients) was used rarely. Four patients received symptomatic medical care only, such as corticotrophin for infantile spasms.7 This study was approved by the Institutional Review Board of CHOP for the review of retrospective data.


Clinical characteristics of the 13 patients from our series who were diagnosed with GC are detailed in Table 1. The median age at the time of diagnosis was 12 years (range, from 20 months to 18 years). Eight of 13 patients (62%) were diagnosed within the second decade of life. Only 2 patients were age <8 years. In general, the age at diagnosis of our cohort clusters in the early portion of the second decade. Ten of the 13 patients were male (77%).

There was a wide range of clinical presentations within the CHOP cohort. The most common presenting complaint was seizure (5 patients; 38%): All 5 of those patients had seizures that were difficult to control throughout the course of their disease and required complex antiepileptic drug regimens. One child (Patient 11) presented at age 6 weeks with infantile spasms and no radiographic evidence of GC by MRI. At age 20 months, repeat MRI studies revealed thickening of the cortex; widespread enlargement of the parietal, temporal, and occipital lobes; and increased T2-weighted signal, all consistent with GC. It is noteworthy that this patient met the clinical criteria for a diagnosis of neurofibromatosis Type 1. Hemiparesis was the second most common presenting symptom (23%). Lethargy, altered mental status, cranial nerve deficits, and intention tremor were seen at presentation in 2 patients each. Only 1 patient reported either headache or emesis. The median time to diagnosis from the first report of symptoms was 3 months (range, from 1 week to 20 months) (Table 1).

Initial location of the primary lesion was documented by CT in 3 patients and MRI in 10 patients. Similar to adults with GC, children with GC had the classic MRI findings of hyperintensity on T2-weighted and fluid-attenuated inversion recovery images. Eleven patients (85%) had involvement of the cortex or subcortical white matter, and 2 patients had disease that was limited to the diencephalon and brain stem (Table 3). The temporal lobe (7 patients; 54%), frontal lobe (5 patients; 38%), and parietal lobe (5 patients; 38%) were the most commonly involved sites. It is noteworthy that involvement of the thalamus was common (5 patients). Extension across the corpus callosum into the opposite hemisphere occurred in 3 patients. The cerebellum (2 patients) and the occipital lobe (1 patient) rarely were involved. Evidence of mass effect was seen in 8 patients (62%), but obvious ventriculomegaly consistent with hydrocephalus or evidence for herniation was rare (2 patients). Three patients had lesions that enhanced after intravenous iodopaque (CT) or gadolinium (MRI) at presentation.

Table 3. The Children's Hospital of Philadelphia Cohort: Radiographic Findings
Patient No.LocationHydrocehalus/HerniationMass effectEnhancement
  1. R indicates right; B, bilateral; L, left.

1R frontal, R parietalNoNoYes
2R parietalNoYesYes
3B thalamus, upper brain stemNoNoNo
4Corpus callosum, subcortical white matter, cerebellumNoYesNo
5Midbrain, pons, cerebellar pedunclesNoNoNo
6B temporal, B thalamusNoNoNo
7R frontal, R parietal, R temporal, corpus callosumNoYesNo
8R frontal, R temporalNoYesNo
9B temporal, B thalamus, corpus callosum, ponsYesYesYes
10R frontal, R temporalNoYesNo
11L parietal, L temporal, L occipitalNoNoNo
12R parietal, R temporal, B thalamus, basal gangliaYesYesNo
13B frontal, R thalamus, corpus callosumNoYesNo

At the time of the current report, 3 of 13 patients were alive at 7 months, 16 months, and 22 months after diagnosis. All patient deaths in our cohort were caused by disease progression. The 1-year, 2-year, and 5-year PFS rates were 54%, 13%, and 0%, respectively (range, 1.5–43 months); and the 1-year, 2-year, and 5-year OS rates were 92%, 64%, and 21%, respectively (range, 6.5–67 months) (Fig. 3, Table 4). No patient survived more than 6 years. Objective evidence of treatment response was poor. No patients achieved a CR or PR during therapy. One child (Patient 12) had an MR that lasted 11 months after receiving involved-field radiation with concurrent and adjuvant temozolomide. All other patients had SD until tumor progression. Neither the radiation field (whole brain vs. involved field) nor the use of adjuvant chemotherapy correlated with PFS or OS.

Figure 3.

Survival for The Children's Hospital of Philadelphia cohort from the time of histologic diagnosis. (A) Progression-free survival (PFS) is illustrated. (B) Overall survival (OS) is illustrated.

Table 4. Overall Survival Comparison
Length of survival, ySurvival rate (%)
CHOP cohort(n = 13)Historic cohort (n = 40)Combined cohort (n = 53)
  • CHOP indicates Children's Hospital of Philadelphia; PFS, progression-free survival; OS, overall survival.

  • *

    OS was calculated from the time of biopsy.

  • OS was calculated from the onset of symptoms.


Tumor enhancement on the initial imaging study correlated significantly with PFS. For patients who had no evidence of tumor contrast enhancement, the 6-month PFS rate was 70% (range, 4–43 months). By contrast, the 6-month PFS rate was only 33% (range, 1.5–14 months) for patients who had evidence of tumor contrast enhancement (P = .034). In addition, the presence of seizure at diagnosis was correlated with a longer PFS (60% vs. 12.5% at 18 months). This difference cannot be explained by lead time bias caused by an early diagnosis, because there was no significant relation between seizures and the time to diagnosis (P = .79).

Age at diagnosis was a significant predictor of OS (P = .040). Children who were diagnosed in the first decade of life had a 2-year survival rate of 19% (range, 6.5–22 months) compared with 86% for children who were diagnosed in the second decade (range, 19–67 months). There was no effect of age on PFS. Though only 2 patients had evidence of hydrocephalus or herniation at diagnosis, their OS was significantly shorter (P = .019, data not shown) than patients without these findings (18-month survival: 50% [range, 6.5–16 months] vs. 90% [range, 17.5–67 months], respectively).

The median age at diagnosis for the historic cohort was 13 years (range, from birth to 20 years). Forty of 51 patients (78%) presented in the second decade of life, and only 3 of those patients (6%) were age <1 year. In 40 patients who had reports that were evaluable for survival (Table 4), the 1-year, 2-year, and 5-year OS rates from the time of initial symptom(s) were 74%, 45%, and 25%, respectively (range, 0.25–276 months). It is noteworthy that 3 patients survived longer than 155 months after the onset of symptoms. These prolonged survivals constitute extreme outliers and raise the possibility that, although they eventually met the diagnostic criteria for GC, their initial reported symptoms may not have been caused by GC.

When they were combined for the purposes of survival assessment, our series together with the 40 patients from the literature review yielded 53 evaluable patients with childhood and adolescent GC. The OS for this combined cohort, measured from the onset of symptoms, was 79%, 51%, and 26% (range, 0.25–276 months) at 1 year, 2 years, and 5 years, respectively, from the time of initial symptoms. No differences in PFS or OS were observed between different treatment modalities. However, survival for all patients who received treatment was prolonged significantly (P = .0029) compared with patients who received no treatment (Fig. 4).

Figure 4.

Overall survival (OS) is illustrated by treatment status for the combined cohort of 53 patients who were evaluable for survival.


Despite aggressive and often multimodal therapeutic intervention, survival rates for adult and childhood patients with GC are extremely poor. Forty-seven percent of all children in our series had disease progression within 1 year, and no one survived beyond 6 years after diagnosis. Recent case series of biopsy-proven GC in adults who received radiotherapy reported a median survival that ranged from 11.4 months to 38.4 months, similar to our CHOP cohort (median survival, 27 months).8, 12–14 Thus, our data reveal that GC in children is as aggressive and has the same poor prognosis as GC in adults.

Early case series of patients who received radiation for GC reported stabilization of clinical symptoms and occasional transient improvement.15 More recently, larger cohorts of adult patients who received radiation have been studied. Perkins et al. reported a 33% partial radiographic response rate to radiation, but this improvement remained stable at 10 months in only 50% of patients.14 In contrast, no response to initial radiation was observed in 11 of 12 patients in our series. These observations suggest that objective responses to radiation occur rarely and, when obtained, do not predict longer survival rates.

Analysis of our combined cohort of 53 patients suggests that treatment of any kind prolongs survival in pediatric patients with GC. This effect likely is attributable to the use of radiation, because it was the most common therapeutic modality (78%) employed among treated patients. Adult studies have suggested that radiation may be more effective in younger patients. Perkins et al. reported that the median time to disease progression and the median survival were prolonged significantly for patients age <40 years with GC who received radiation.14 However, it is unclear whether this improved survival was because of a better response to radiation in younger patients or simply because younger patients have improved survival regardless of treatment, as demonstrated in other glial neoplasms.16 The current study was limited both by its small size and by the potential biases that may occur in a study of patients who are treated at different institutions at different times, as in any study of a rare disease process. Nevertheless, this report suggesting a treatment effect supports the current treatment of GC with radiation and, thus, should be validated by investigation into patient series at other pediatric institutions.

Patients with GC inevitably experience tumor progression. No patient in the CHOP cohort achieved long-term, disease-free status. After first progression, the time to death is short (<7 months) despite aggressive use of salvage therapy in this population. Similar to other glial tumors, both malignant and benign, the location of tumor progression typically is at the primary site.17 Two patients in our series, however, had evidence of spinal cord involvement at first recurrence. Thus, in patients with recurrent disease, there should be a heightened awareness of the possibility of leptomeningeal tumor spread. However, the incidence of leptomeningeal tumor dissemination is not sufficiently high to warrant routine spinal cord imaging at initial diagnosis in the absence of relevant localizing signs and symptoms.

Prognostic factors for childhood GC largely remain uncharacterized. The presence of histologic features that are seen more typically in low-grade astrocytomas belies its aggressive clinical course. Unlike patients with brain stem gliomas, in which an early response to radiation therapy may predict longer OS, we did not observe a sufficient number of objective responses in patients with childhood GC to suggest a potential utility of the response to radiation as a prognostic factor. Conversely, age at diagnosis was a significant prognostic factor for OS in our series. Patients in the second decade of life had a significantly prolonged survival compared with younger children. However, this correlation was not attributable to the avoidance of radiation therapy in young children. In fact, in the current series, all but 1 patient in the first decade of life received radiation, and the single patient who did not remains alive at the time of this report.

In this study, tumor enhancement after contrast administration occurred in 3 of 13 patients and was associated strongly with a shorter PFS. This is not the first suggestion that enhancement may predict aggressive GC. Vates et al., in a review of 22 largely adult patients, noted that the absence of enhancement was correlated with an improved length of survival.8 To date, no other series has evaluated GC radiographic characteristics at presentation as prognostic variables for survival; however, these findings suggest that further investigation is warranted.

The clinical presentation of GC is determined by the anatomic site involved. Nevin's initial series described 3 patients who had seizures, memory deficits, personality changes, and symptoms of increased intracranial pressure.1 In larger studies among patients of all age groups, it was noted that the most common neurologic findings at presentation were headache, seizures, corticospinal deficits, and dementia.3, 7, 8, 13, 18 Our data support a range of neurologic presentations with no single symptom that occurred in the majority of patients. However, seizure was the most common finding in our cohort.

The age at presentation for children with GC has not been defined previously. Artigas et al. described a bimodal distribution of age at initial GC diagnosis, noting peaks in the second and fifth decades of life.3, 19 In the current study, we observed that 62% of patients with childhood GC were diagnosed in the second decade of life, and the median age at presentation was 12 years. This confirms an increase in the incidence of GC during the second decade and suggests that GC, like malignant supratentorial astrocytomas (median age at presentation, 9–10 years) is a malignancy of older childhood.20

Evidence that treatment improves OS for children with GC supports the rationale for radiation therapy as a foundation for current treatment and warrants further investigation of adjuvant treatments. The molecular phenotypes of GC are understood poorly. Such studies will be of critical importance to identify relevant therapeutic targets and to inform the selection of appropriate therapies. Future clinical studies may be aided by prognostic observations, such as the absence of enhancement on initial neuroimaging studies and a diagnosis of GC during the second decade of life, both of which predicted longer survival in our series. The importance of these findings is underscored further by their statistical significance despite a relatively small sample size. Because of the rarity of this diagnosis, future questions regarding therapeutic strategies must be addressed by prospective, multiinstitutional investigations.