The objectives of the current study were to determine the outcome of children who were treated with chemotherapy and radiotherapy on the Children's Cancer Group (CCG) high-grade glioma protocol (CCG-945) who were diagnosed with low-grade gliomas on post hoc central pathologic review and to identify clinical and biologic features associated with prognosis.
Between 1985 and 1991, 250 children with institutionally classified high-grade gliomas were enrolled on CCG-945. Patients older than 24 months with intracranial lesions were assigned randomly to receive either lomustine, vincristine, and prednisone (control regimen) or the 8-drugs-in-1-day regimen (experimental regimen); younger patients and those with primary spinal cord tumors were assigned nonrandomly to the experimental regimen. Central independent review by 5 neuropathologists led to a reclassification of low-grade glioma in 70 patients, who were the focus of the current study.
The study involved 42 males and 28 females (median age, 7.7 years) with a median follow-up of 10.4 years. At 5 years, the progression-free survival (PFS) rate was 63% ± 6%, and the overall survival (OS) rate was 79% ± 5%, compared with a PFS rate of 19% ± 3% (P < 0.0001) and an OS rate of 22% ± 3% (P < 0.0001) in the remainder of the cohort. Significantly poorer 5-year PFS was seen in children younger than 24 months, those with fibrillary astrocytoma, and those with posterior fossa tumors. Patients demonstrated a modest improvement in PFS but no improvement in OS compared with children with low-grade gliomas who were treated with contemporary chemotherapy-alone approaches.
Low-grade gliomas account for 35% of all childhood central nervous system tumors.1 Resection remains the preferred treatment for patients with cerebellar and cerebral hemispheric tumors and frequently is curative if a macroscopic total or nearly total resection is achieved.2 However, for patients with midline and other tumors for whom extensive resection is not possible, the most appropriate management remains undefined.3–5 Although radiotherapy provides durable disease control in a significant percentage of such patients,6, 7 the neurocognitive and neuroendocrine sequelae associated with radiotherapy, especially in younger patients, have led many investigators to explore the use of chemotherapy. Agents that have been used include actinomycin D,8 vincristine,9 carmustine,10 carboplatin,11–13 and etoposide.14, 15 The most successful regimens have used carboplatin and vincristine, resulting in a 3-year progression-free survival (PFS) rate of 68% ± 7%,3 and procarbazine, 6-thioguanine, dibromodulcitol, lomustine, and vincristine, resulting in a 5-year overall survival (OS) rate of 78%.16 In the current article, we report the outcomes and clinical characteristics of 70 patients who were diagnosed initially with high-grade glioma, subsequently treated with chemotherapy and radiotherapy according to the Children's Cancer Group (CCG) high-grade glioma protocol (CCG-945), and later shown to have had low-grade glioma after central review of pathology using contemporary classification criteria.
MATERIALS AND METHODS
CCG-945 Study Cohort
Between April 1985 and April 1991, 250 patients with newly diagnosed, high-grade astrocytomas were enrolled on CCG-945. Eligibility required institutional histopathologic confirmation of a high-grade glioma arising primarily outside the brainstem. Tumors were categorized as glioblastoma multiforme (GBM), anaplastic astrocytoma (AA), or other eligible Grade 3–4 glioma, such as anaplastic oligodendroglioma and anaplastic mixed glioma. No therapy other than surgery was permitted before study entry. The protocol recommended removal of as much tumor as safely feasible. The treatment regimen for these newly diagnosed patients consisted of a combination of radiotherapy (5400 centigrays [cGy] in 180-cGy fractions) and chemotherapy with either adjuvant prednisone, lomustine (CCNU), and vincristine (pCV; the control regimen) or the 8-drugs-in-1-day (8-in-1; the experimental regimen), administered for 2 cycles before irradiation and continued after irradiation (Fig. 1). Children older than 18 months or 24 months (depending on the year of study entry) with intracranial high-grade gliomas were assigned randomly to receive either the 8-in-1 regimen or the pCV regimen. The maintenance regimen consisted of 8 cycles of either the 8-in-1 regimen or the pCV regimen given every 6 weeks. Younger patients and those with malignant spinal cord gliomas were assigned nonrandomly to receive the 8-in-1 regimen. All patients were to receive radiotherapy. In patients younger than 24 months, investigators were given the choice of involved-field radiotherapy after 2 cycles of chemotherapy or craniospinal radiotherapy after 10 cycles of chemotherapy. Most infants, however, did not receive radiotherapy, because many investigators opted to withhold irradiation.17, 18 In all patients, clinical parameters were monitored rigorously, and long-term follow-up was meticulous. No patient was censored within 6.5 years of randomization. Informed consent was obtained from patients, parents, or guardians, as appropriate, at the time of protocol enrollment.
Central Review Process for the Current Cohort
The initial published analyses of the CCG-945 study results incorporated a central review by a single neuropathologist in addition to the institutional pathology review.19, 20 However, both pathology assessments predated the publication of the revised World Health Organization classification criteria,21 which established more stringent and widely accepted criteria for distinguishing low-grade gliomas from high-grade gliomas. Recognition of these new criteria prompted concern that, based on contemporary classification guidelines, a number of patients who were included in the original clinical cohort may have had diagnoses other than high-grade glioma. To address this issue, each case was masked and independently reviewed by five nationally respected neuropathologists. The Pediatric Branch of the Cooperative Human Tissue Network (PCHTN) coordinated tissue accrual and distribution for each patient. Specimens were coded so that the study investigators remained masked to clinical and outcome results as well as the classifications of other neuropathologists. Tumors were classified as GBM, AA, other eligible high-grade glioma, and discordant (not high-grade glioma); among tumors with discordant results, a revised diagnosis was provided.
Specimens for which at least three of five pathologists reached a common histologic diagnosis of low-grade glioma were identified for the current analysis. These specimens were reviewed, because the postoperative management of such patients with protocol-directed chemotherapy and radiotherapy was substantially more aggressive compared with the typical treatment for patients with low-grade tumors. The outcome results in these children would provide insight into the advisability of therapeutic intensification for pediatric patients with low-grade gliomas as a strategy to improve long-term disease control. The clinical and biologic characteristics and outcome of these patients are the primary focus of this report.
Biologic Correlates of Outcome
Because the status of TP53 mutations, p53 expression, and the tumor proliferation index each has been associated with outcome in children with high-grade gliomas within the CCG-945 cohort,22 we asked whether those biologic parameters also were associated with outcome in the children who had tumors that were reclassified as low-grade gliomas. The PCHTN coded and provided paraffin-embedded tumor specimens for all analyses. Slides were reviewed, and blocks that contained representative tumor specimens were sectioned at a thickness of 4 μm and stained immunohistochemically or subjected to microdissection-based p53 genotyping. Assessable specimens were available from 37 patients.
For immunohistochemical analyses, sections were processed and subjected to microwave antigen enhancement as described previously.22 For assessment of proliferation indices, sections were incubated overnight at 4 °C with anti-MIB-1 antibody (Immunotech, Westbrook, ME; 1:100 dilution) in common antibody dilutent (BioGenex, San Ramon, CA). Control sections were treated with diluent alone. Slides were then rinsed with phosphate-buffered saline and incubated with biotinylated horse antimouse immunoglobulin G (1:200 dilution; Vector Laboratories, Burlingame, CA) for 30 minutes. Antibody binding was visualized using a Vector Elite avidin-biotin complex (ABC) kit and the substrate 3,3′-diaminobenzidine (Sigma Chemical Company, St. Louis, MO), according to the manufacturer's protocol. The slides were then counterstained with Mayer hematoxylin, dehydrated through graded concentrations of ethanol, cleared in xylene, mounted, and examined using a light microscope as described previously.23 The percent MIB-1 labeling was quantitated by counting stained and unstained cells in 5–10 high-power fields (at least 2000 cells) that included the highest levels of MIB-1 labeling.
We analyzed p53 expression using similar techniques, except that sections were incubated overnight at 4 °C with anti-p53 antibody (DO-7; Dako, Carpenteria, CA) at a 1:300 dilution in common antibody diluent followed by ABC-based labeling and detection. As in the analysis of proliferation labeling, an observer who was unaware of the histologic diagnoses, outcomes, or clinical features assessed p53 immunoreactivity. Only cells with dense nuclear staining were interpreted as positive. Tumors were categorized as expressing little or no p53 (Grade 0 or 1), similar to normal brain tissue, or overexpressing p53 with staining observed in 25–50% of cells (Grade 2), 50–75% of cells (Grade 3), or > 75% of cells (Grade 4) in the high-power field in areas with maximal staining.22
For TP53 mutational analysis, tissue specimens were removed from tumor sections using microdissection techniques, as described previously.24–27 Exons 5, 6, 7, and 8 were examined specifically because they encompass most TP53 mutations detected in astrocytic tumors28 and nonastrocytic tumors.29 Individual exons were amplified by polymerase chain reaction (PCR), as described previously.22, 24 The PCR products were isolated and directly sequenced by dideoxy chain termination using 35S-labeled deoxyadenosine triphosphate.22, 24 Sequences were read from autoradiograms of 6% polyacrylamide gels.
Kaplan–Meier30 estimates were computed for the OS and PFS of the comparison subgroups for each demographic, clinical, and molecular factor. Survival curves and tables were generated based on these estimates. The exact log-rank test based on the Mantel–Haenszel method31 was used to compare survival distributions between and among subgroups. Multivariate analysis of prognostic variables was conducted using the Cox32 proportional hazards model.
Exact tests were done using PROC StatXact in SAS software (Version 8.0; SAS Institute, Cary, NC). A specially developed SAS macro was used to generate survival tables and curves. Multivariate analyses were run in SAS software using PROC PHREG. To assess the association between biologic parameters and outcome, tumors were subdivided based on whether they exhibited p53 mutations or overexpression, and MIB-1 labeling was applied both as a continuous variable and as a dichotomous variable, depending on whether the index was greater than or less than the median.
Histopathologic material was reviewed from the 250 patients with institutionally diagnosed high-grade gliomas who were treated on the CCG-945 protocol. It was found that 70 of those patients (28%) had low-grade gliomas by consensus of 3 of 5 neuropathologists on post hoc central review. Demographic and clinical characteristics are summarized in Table 1.
Table 1. Patient Characteristics
No. of patients (%)
Age at diagnosis (yrs)
< 2 yrs
≥ 2 yrs
Extent of resection
Biopsy (< 10%)
Total (> 90%)
There were 34 patients with fibrillary astrocytomas (FA), 19 patients with juvenile pilocytic astrocytomas (JPA), 14 patients with other low-grade gliomas (6 patients with oligodendroglioma or mixed oligoastrocytoma, 4 patients with pleomorphic xanthoastrocytoma, 2 patients with ganglioglioma, 1 patient with astroblastoma, and 1 patient with desmoplastic infantile gangliocytoma), and 3 patients with low-grade gliomas with no subtype specified. Of the 70 tumors, 44 were classified institutionally as anaplastic astrocytoma (AA) in the CCG-945 study, 6 tumors were classified as GBM, and 20 tumors were classified as other malignant gliomas. These numbers represent 37% of patients with AA enrolled on the CCG-945 study (n = 119), 8% of patients with GBM (n = 76), and 34% of patients with other eligible tumors (n = 44). The rate of discordance did not differ significantly as a factor of age or tumor location.
Treatment and Outcome
Of 70 patients in this cohort, 45 were treated with the 8-in-1 regimen, and 25 were assigned randomly to the control regimen. The median follow-up was 10.4 years. The 5-year PFS rate was 63% ± 6%, and the OS rate was 79% ± 4% for the entire group. For the 8-in-1 group, the PFS rate was 60% ± 7%, and the OS rate was 80% ± 6%. For the control group, the PFS rate was 68% ± 9% (P = 0.13), and the OS rate was 76% ± 8% (P = 0.83).
Fifty-seven patients received radiotherapy as well as chemotherapy as part of their initial therapy, and 13 patients received chemotherapy only. The PFS rate for patients who received chemoradiotherapy was 68% ± 6%, compared with 38% ± 12% (P = 0.04) for patients who received chemotherapy only. The 5-year OS rate for the chemoradiotherapy group was 79% ± 3%, compared with 77% ± 10% for the chemotherapy-only group (P = 0.34). All patients enrolled on this protocol were to have received radiotherapy; thus, those who received chemotherapy alone represented protocol violations. However, some patients received successful salvage radiotherapy, leading to a similar OS rate in both groups.
Thirty-seven tumors had adequate tissue for evaluation of p53 expression status. Among these, there were 33 patients with a p53 score of 0 or 1 and 4 patients with a p53 score of 2 or 3, a much lower frequency than had been observed in the review-confirmed, high-grade glioma cohort from CCG-945.22 Thirty-four tumors were assessable for MIB-1 labeling indices, with at least 2000 countable nuclei. Indices ranged from 1.6 to 25.7, with a median of 7.1, also significantly lower than the indices observed in the review-confirmed subgroup.23 Thirty tumors had sufficient specimens to permit amplification of p53 exons 5–8. Among this subgroup, p53 mutations were present in 7 patients and absent in 23 patients (Table 2).
Table 2. Molecular Characteristics of Tumors from the 70 Patients on Study
Age at diagnosis and gender did not affect OS. However, in univariate analysis, younger age (P = 0.09) and male gender (P = 0.07) were factors trending toward a worse PFS (Fig. 2).
Extent of resection
In the current series, macroscopic total resection was defined as a resection > 90%. Patients with low-grade glioma who underwent macroscopic total resection had a 5-year OS rate of 86% ± 6%, compared with 73% ± 7% (P = 0.13) for patients who underwent less than complete resection. The PFS rate for patients who underwent macroscopic total resection was 72% ± 8%, compared with 56% ± 8% (P = 0.21) for patients who underwent less than complete resection. The extent of resection did not affect PFS or OS based on histology (FA vs. JPA).
For all patients with low-grade gliomas, the PFS rate was 63% ± 6%, and the OS rate was 79% ± 5%, with a median time to progression (TTP) of 2.2 years. These results were significantly more favorable compared with the results from patients with consensus-confirmed, high-grade gliomas who had a PFS rate of 19% ± 3% (P < 0.0001), an OS rate of 22% ± 3% (P < 0.0001), and a median TTP of 0.7 years (Fig. 3). Within the low-grade glioma cohort, there were significant differences in PFS (P = 0.04) and OS (P = 0.006) between patients with JPA and FA. Patients with JPA had a PFS rate of 68% ± 10% and an OS rate of 95% ± 5% at 5 years, whereas patients with FA had a PFS rate of 52% ± 8 % and an OS rate of 64% ± 8% at 5 years (Fig. 4). The median TTP for patients with FA was 1.2 years, compared with 3.8 years for patients with JPA.
The 5-year OS rate was 91% ± 5% for patients with hemispheric tumors, 68% ± 10% for patients with midline tumors, and 50% ± 16% for patients with tumors of the posterior fossa, with significantly better survival for patients who had cerebral hemispheric tumors compared with patients who had tumors of the posterior fossa (P = 0.0096). Among the 8 patients with tumors of the posterior fossa, tumors were located in the pons (n = 2), the medulla (n = 1), the cerebellar hemisphere (n = 1), and the cerebellar peduncle (n = 4). There also was a significantly better PFS for patients who had hemispheric tumors compared with patients who had tumors of the posterior fossa (P = 0.039). No significant differences in PFS or OS were noted between patients with tumors in other sites (Fig. 5).
There were no significant differences in PFS (P = 0.13) or OS (P = 0.83) between patients who were treated with the 8-in-1 regimen and those who were treated with the control regimen. Among patients who were treated with the 8-in-1 regimen, patients with JPA had significantly better 5-year OS (P = 0.014) and PFS (P = 0.04). In contrast, among patients who received the control regimen, there was no significant difference in PFS (P = 0.33) or OS (P = 0.39) between patients with JPA and patients with FA.
Molecular factors and proliferation index
There were no differences in PFS and OS for patients based on the level of p53 expression, the presence of p53 mutations, or the MIB-1 labeling index.
Multiple Regression Analysis
The 4 variables that had prognostic significance (P < 0.05) or borderline significance (0.05 < P < 0.1) as determined by univariate analysis were included in the multivariate analysis. The 4 significant variables were age (younger than 24 months vs. older than 24 months), gender, diagnosis (specifically, JPA vs. FA), and site (limited to hemispheric, midline, or posterior fossa). A stepwise selection procedure was used to include significant variables one at a time, with the option of removal if a variable proved to be insignificant upon inclusion of other significant variables. A stepwise multiple regression analysis was performed on the subset of 46 patients after eliminating patients who had tumors with pathologies or sites other than those mentioned above. The results of the multivariate analysis identified age younger than 24 months (P = 0.0041), FA (P = 0.047), and posterior fossa location (P = 0.027) as prognostic factors for poor PFS. No differences were noted in OS.
Central Pathologic Review
CCG-945, as one of the few pediatric studies to examine chemotherapy efficacy in patients with high-grade glioma in a randomized controlled trial, demonstrates the importance of central pathologic review in studies of children with gliomas. The criteria for the diagnosis of low-grade astrocytomas have been appreciated more widely during the past decade with the publication and acceptance of the revised classifications of the World Health Organization.21 The more uniform and widely accepted guidelines for diagnosing low-grade gliomas provided an impetus for examining whether a subset of the original cohort on CCG 945 may have been misdiagnosed on initial institutional review. To ensure consistency in the review process, we used an independent consensus review process involving an expert panel of neuropathologists. Among the 250 patients with an institutional diagnosis of high-grade glioma who were enrolled on the current study, it was noted that 28% had low-grade gliomas on central review. Patients with AA, GBM, and other eligible malignant gliomas were reclassified with low-grade gliomas in 37%, 8%, and 34% of patients, respectively. If central pathologic review had not taken place, then the efficacy of the proposed therapy in patients with high-grade gliomas would have been overestimated because of the significantly better OS (P < 0.0001) and PFS (P < 0.0001) in patients with low-grade glioma. Furthermore, a slightly higher rate of discordance was noted in patients younger than 24 months and in patients with midline tumors, making these groups of patients particularly susceptible to misdiagnosis and inappropriate, often overaggressive therapy.
Although the OS rate for patients with low-grade glioma (79% ± 4.9%) was significantly better than the OS rate for similarly treated patients with malignant gliomas (22% ± 3.3%; P < 0.0001), neither the 8-in-1 regimen (OS, 80% ± 6%) nor the control regimen (OS, 76% ± 8%; P = 0.83) extended the OS of patients with low-grade gliomas compared with standard therapy.3 However, it is difficult to compare this select group of patients who had clinical, radiographic, or pathologic characteristics that led to a diagnosis of high-grade glioma by the treating institution with other series of patients who had more typical low-grade gliomas.
Many aspects of therapy for patients with low-grade gliomas remain controversial and depend on tumor type, location, and patient age. For patients with low-grade gliomas who can safely undergo resection, macroscopic total resection remains the preferred treatment.2, 33, 34 For patients who have tumors that are not amenable to resection, the choice of therapy is more controversial. Although radiotherapy provides durable disease control with 5-year OS rates of 75–90% in patients with hypothalamic-optic pathway gliomas,6, 35–37 concerns for the long-term neurocognitive and neuroendocrine sequelae of radiotherapy have led investigators to explore the use of chemotherapy, especially in younger patients. Furthermore, a recent study demonstrated that among children with subtotally resected low-grade gliomas, there was no statistically significant difference in OS between patients who received radiotherapy initially and patients who were observed.5
The current study is unique, in that the majority of patients (81%) received both chemotherapy and irradiation as their primary therapy. However, the use of this aggressive, combined-modality approach did not appear to improve the 5-year OS rate (79% ± 5%). In contrast, the 5-year PFS rate for the group of patients who received combination chemoradiotherapy (68% ± 6%) was significantly better compared with the cohort of patients who received chemotherapy only (38% ± 12%; P = 0.04). However, the younger age of the latter group makes interpretation of these results difficult, because tumors in younger patients may be more aggressive than in older patients.38–41 Similarly, the 5-year PFS rate of 68% in the chemoradiotherapy group was superior to the most successful chemotherapeutic regimens reported in the literature, including carboplatin alone (5-year failure-free survival rate, 61%13; 3-year PFS rate, 64%12); carboplatin and vincristine (3-year PFS rate, 68% ± 7 %3; 5-year PFS rate, < 50%); and procarbazine, 6-thioguanine, dibromodulcitol, CCNU, and vincristine (3-year PFS rate, 45%).16 No improvement in OS was observed. It is noteworthy that compared with the current study cohort, patients in studies by both Packer et al. and Prados et al. had a considerably younger median age, had predominantly midline tumors rather than hemispheric tumors, and had a much shorter median follow-up period (approximately 3 years).
Factors Affecting Outcome
Although this cohort shared the diagnosis of low-grade glioma, its heterogeneity in location and histology may be important distinguishing characteristics that affect outcome. Furthermore, this cohort includes a select group of patients with low-grade glioma who had clinical, radiographic, and pathologic features at presentation that led to the diagnosis of malignant glioma at the treating institution. Similar to other studies,33 patients younger than 24 months had a poorer PFS compared with older patients in univariate analysis (P = 0.09) and multivariate analysis (P = 0.004). OS was independent of age.
In univariate analysis, patients who had JPA had significantly better PFS and OS compared with patients who had FA. Multivariate analysis confirmed the better PFS in patients with JPA. Similarly, a number of studies have indicated that patients who have JPA have a survival advantage compared with patients who have FA.34, 42–46 In contrast, in a study of 142 children with low-grade astrocytoma, Gajjar et al.33 found no relation between tumor histology (JPA vs. astrocytoma, not otherwise diagnosed) and PFS in an analysis that controlled for age and anatomic site.
Primary Tumor Site
Patients who had cerebral hemispheric tumors had better PFS (P = 0.04) and OS (P = 0.001) compared with patients who had tumors of the posterior fossa. Although a difference was noted between patients with cerebral hemispheric tumors and midline tumors in PFS (79% hemispheric vs. 50% midline) and OS (91% hemispheric vs. 68% midline), the difference was not statistically significant. In contrast, Gajjar et al.33 reported that patients who had cerebral hemispheric and cerebellar tumors had significantly better PFS compared with patients who had tumors in all other sites combined (P = 0.0006). The difference between the two series may reflect the larger sample size of the series of Gajjar et al.; the higher percentage of macroscopic total resections and near total resection (51% vs. 41%),33 cerebellar hemispheric tumors (35% vs. 9%), and JPAs (54% vs. 27%) in that series; and the fact that the patients in the current cohort did not have radiologically, pathologically, or clinically typical low-grade gliomas.
Extent of Resection
Although patients who underwent macroscopic total resection had better OS and PFS compared with patients who underwent less extensive resection, the difference was not statistically significant, even when patients with JPA were analyzed separately. These results may reflect the relatively small patient subset numbers in this series and the less stringent definition of macroscopic total resection used in CCG-945 (> 90% resection), rather than the more conventional definition, which requires removal of all visible tumor. Similarly, Gajjar et al.33 did not report a significant difference in PFS for patients who underwent macroscopic total resection or near total resection (4-year PFS rate, 89%) compared with patients who underwent incomplete resection (4-year PFS rate, 77%). However, Pollack et al.34 reported that among 71 patients with low-grade gliomas of the cerebral hemispheres who underwent resection, the extent of resection was the factor associated most strongly with PFS (P = 0.015) and OS (P = 0.013). Similarly, in the preliminary analysis of the CCG-9891/Pediatric Oncology Group 9130 intergroup low-grade glioma study, Wisoff et al. reported that the impact of resection extent on outcome was significant, specifically between patients who underwent macroscopic total resection (no visible disease postoperatively on imaging) and patients with any amount of residual disease (P < 0.0001); patients who underwent near total resection (90–99%) fared no better than patients who underwent subtotal resection (50–90%).47
Pollack et al.22 recently reported that overexpression of p53 in pediatric malignant gliomas was associated strongly with adverse outcome, independent of clinical prognostic factors and histologic findings. However, such an association has not been established in adult or pediatric low-grade gliomas. Nakamura et al.48 reported that in adults with low-grade gliomas, neither p53 nor MIB-1 staining was associated with prognosis. Similarly, Machen et al.49 reported that MIB-1 staining in 48 pilocytic astrocytomas was negative in 66.7% of patients and that outcome was not correlated with MIB-1. The results of the current series concur with previous reports in that no difference in PFS or OS was noted for patients with low-grade gliomas based on their p53 expression status, the presence of p53 mutations, or the MIB-1 labeling index.
The authors thank their collaborators, Judith Burnham, Ronald L. Hamilton, and Sydney D. Finkelstein (University of Pittsburgh, Pittsburgh, PA), for the biologic analyses of the study cohort.