Histopathologic Grade and Age among Malignant Astrocytomas

The Children’s Cancer Group (CCG) ^#943 study showed the median survival among children with anaplastic astrocytomas (AAST) was greater than 6 years; among GBM patients, overall survival averaged less than 15 months (P = .012).²⁷ Age and histopathologic grade jointly influenced outcome. The CCG ^#921 study found that among children less than 24 months old, the 3-year progression-free survival (3Y-PFS) was 36%. The 3Y-PFS for patients with AAST was 44%, which was dramatically better than that for GBM (0%) (Table 8-1).²⁸

Multidrug Resistance Phenotype

The PgP and MRP proteins are thought to function as ATP-dependent transmembrane transporters, which nonspecifically remove large molecular weight solutes from the brain and/or prevent hematogenously borne chemicals from entering the CNS.²⁹ PgP/MDR-1-associated resistance is hypothesized to function at the interface between the blood-brain and blood-tumor barriers as well as within the neoplastic cells themselves. Despite their association with progressive anaplasia and an adverse prognosis, there is actually little evidence that the PgP/MDR1 is a major mechanism of acquired resistance among MG.^29–31 Resistance to therapy appears to be a more complex phenomenon which disrupts the mechanisms of repair of genotoxic injury and/or induction of apoptosis.

The O6-methylguanine-DNA methyltransferase (MGMT) enzyme’s function is to counteract the cytotoxic effects of alkylating agents; its expression correlates inversely with survival among adult MG patients. Archival histopathologic specimens from 109 patients treated on the CCG Study ^#945 were analyzed, and 11% had overexpression of MGMT. The 5Y-PFS was 8.3% among these patients, in contrast to those whose tumors did not show overexpression of MGMT (5Y-PFS of 42%, P = .017). This adverse correlation was independent of age, histopathologic grade, tumor location, or extent of residual disease.³²

Hypoxia and the Abrogation of Apoptosis

Intratumoral hypoxia is known to be the most significant predictor of radiotherapeutic resistance.³³ Radiation therapy has only one third the effectiveness against hypoxic tumor cells that it does against well-oxygenated cells. It was previously thought that radiation-induced DNA damage required the presence of oxygen radicals. Instead, normal cells, when deprived of oxygen, will arrest in late G₁ and early S phase, in concert with hypophosphorylation of the retinoblastoma gene protein, pRB, and die through induction of apoptosis.³³ Arrest of the DNA-damaged cell at the G₁/S interface is associated with stabilization of intact wild-type TP53 (TP53wt) expression, inhibition of replicative DNA synthesis, and possibly the initiation of repair. TP53wt appears to achieve G₁/S arrest through transcriptional regulation of at least one critical downstream target, the Cdk inhibitor p21^WAF1/CIP1. This protein inactivates cyclin D/Cdk4, D/Cdk6, and E/Cdk2 complex activity and directly inhibits PCNA, a regulatory subunit of DNA polymerase (the principal replicative DNA polymerase).³⁴

Ionizing radiation also produces G₁ arrest in cells with TP53wt and increases TP53 protein expression.³⁵ In contrast, the hypoxic regions within the tumor actually create a reservoir for malignantly transformed cells, which do not arrest at the G₁/S interface, are more likely to have mutations in the TP53 gene (TP53 m), and become resistant to TP53-induced apoptosis (Graeber et al., 1996).³⁶ Stable cytoplasmic TP53wt expression has been shown to a favorable predictor for survival among AAST patients (P = .015). By comparison, such expression was not found among the GBM (P = .8), which by definition contain areas of necrosis.³⁷ Intratumoral hypoxia thus constitutes a key mechanism of clonal selection.

Furthermore, hypoxia has also been correlated with insensitivity to carmustine (BCNU) and cisplatin, independently of the putative drug resistance genes MDR1, MRP, O⁶MGMT, or ERCC.³⁸ Here too, the acquisition of resistance appears be mediated through TP53 m. For example, 86% of TP53wt positive MG cell lines have been shown to be responsive to currently available chemotherapy drugs. In contrast, there were no effective cytotoxic agents against 75% of the glioma cultures expressing TP53 m (P = .0006). Supporting this specificity, the resistance conferred by TP53 m is restricted to alkylators and platinators, which damage DNA, rather than against the microtubular toxins.³⁹

The Molecular Basis for Multiple Models of Glioma Progression: The Tumor Suppressor Genes

There is a large literature on the nonrandom cytogenetic aberrations observed among MGs, which include gains or losses of chromosomes, as well as rearrangements. Some of these changes, such as 10q, 9p, 11p15.5-pter and 19q may be found only among AASTs and/or GBMs.^40,41 Loss of heterozygosity (LOH) analysis has correlated 17p deletions with allelic loss of the tumor suppressor gene (TSG), TP53.⁴² Hence, it has been hypothesized that such nonrandom patterns of genetic loss can be exploited to predict the location of other as yet unknown TSGs. This in turn has generated the concept of modeling mechanisms for tumor progression.

TP53

One pathway in the evolution of the MG involves TP53 m, which occurs in 14% to 46% of adult patients and is preferentially associated with development of the glioma (often an AAST) between 18 and 45 years of age.^42,43 Clonal expansion of TP53 m cells have been observed from a small subpopulation within a primary AST to become the dominant cell type within a recurrent, or “secondary” MG.⁴⁴

Alternate means of TP53 wt inactivation exist as TP53 wt may be complexed by the MDM2 oncogene product. In 8% to 10% of GBM, the MDM2 gene is amplified and overexpressed, supporting the importance of TP53 wt in glial cell cycle dysregulation. A more recent study has identified that about 35% of gliomas demonstrate either TP53 m or methylation of p14(ARF), suggesting that TP53 wt is controlled by down-regulation of p14(ARF).⁴⁵ Inactivation of the astrocyte’s apoptotic response may therefore be effected either by deletion of 17p, mutation of TP53, MDM2 overpression, or hypermethylation of p14(ARF), thus unmasking multiple points of vulnerability along a common pathway that permit further progression towards the endpoint of malignant transformation.⁴⁵

PTEN

Another fraction of about 30% (23% to 62%) of adult GBM patients has been characterized by the association of LOH of chromosome 10 with epidermal growth factor receptor (EGFR) amplification. In contrast to the previous group, these patients (typically older individuals) appear to present de novo with a “primary” GBM. The LOH at 10q23 has been demonstrated in approximately 70% of GBMs, and therefore predicted to be the site of another glial TSG.^42,43 The “Mutated in Multiple Advanced Cancers” and “Phosphatase and Tensin” gene (MMAC/PTEN) has been localized to chromosome 10q25 and shown to function as a phosphatidylinositol 3,4,5-triphosphate phosphatase, which modulates cell growth as well as apoptosis.^46,47

Patients with GBM have demonstrated a significantly higher incidence of LOH at the MMAC/PTEN locus (72%) than did patients with AAST (29%)(P < .0001). The clinical relevance is that patients with LOH at the MMAC/PTEN locus have a more adverse prognosis (P < .0001), even when controlled for age at surgery and histologic grade.⁴⁸ Conversely, Kaplan-Meier analysis has demonstrated significantly better survival for patients whose MG exhibited high MMAC/PTEN expression.⁴⁹ These results have been interpreted to implicate alteration or loss of the MMAC/PTEN locus as a late marker of disease progression with ominous significance.

RB

About 15% to 44% of MG patients are reported to have loss of chromosome 13 as well as LOH or partial deletions of the 13q14 allele, which suggests the possibility that the RB gene is also a glioma TSG. Loss of heterozygosity at the RB 1.20 region has not been detected among low-grade gliomas, but occurred in 20% of AAST and 32% of GBM. The presence of LOH at the RB 1.20 region has also been associated with an adverse prognosis, implying that loss/inactivation of pRB contributes to disease progression.⁵⁰

Investigation suggests that aberrant function of cyclin D, Cdk4, p16^INK4A/MTS1, and/or pRB is critical for both the process of transformation and that of progression towards increasing anaplasia among glial neoplasms. Cyclin D1 was originally cloned from a human GBM cDNA library, where its transcript and 34 kD product were in abundant expression. Deletion of p16^INK4A/MTS1 and/or amplification of its target Cdk4 occur in 50% of AAST and 85% of GBM.^51,52 The coordinate deletion of both p15^INK4B/MTS2 and p16^INK4A/MTS1 among GBMs suggests that selection favors homozygous deletions of chromosome 9p21 as more efficient for the simultaneous inactivation of both Cdk inhibitors than independent intragenic mutations would be.⁵¹ This remarkable specificity suggests that knocking out p16^INK4A/MTS1’s inhibition of cyclin D1/Cdk4-induced hyperphosphorylation of pRB marks a critical event in the progression to the most advanced grade of glioma, the GBM.⁵²

Evidence is accumulating that cyclin D1 overexpression, absence of pRB expression, failure of pRB dephosphorylation, and/or pRB cleavage may also contribute to chemotherapeutic drug resistance as well, although the mechanism or mechanisms are not understood. Altered CDKN2/p16 function has been correlated with increased sensitivity to antimetabolite agents as opposed to the alkylators, topoisomerase inhibitors, and microtubular toxins.⁵³

To conclude, the genetic mechanism of glial transformation in children is not understood. In one series of 20 pediatric GBMs, 25% were associated with TP53m, the majority of which presented with a brief prodromal period and were considered “primary.” Loss of p16^{CDKN2/INK4A/MTS1} was detected in 61% of these GBMs. Overexpression of EGFR was infrequent (11%) and was coincidental with TP53 m in one case. Of the four GBMs, which progressed from lower-grade tumors, only one contained TP53 m.⁵⁴ In another series of 29 childhood MGs, 95% of cases were found to harbor an alteration in at least one member of the TP53/MDM2/p14ARF tumor suppressor pathway. Overexpression of MDM2, which downregulates TP53 transcriptional activity, was present in 65%, and loss of p14AFT, which inactivates the pathway through MDM2, was found in 10%.⁵⁵

Among pediatric MGs, the presence of MMAC/PTEN mutations has also been confirmed as a marker of an adverse outcome that is independent of patient age or histopathologic grade (III vs. IV).⁵⁶ In tumor specimens of 62 evaluable patients treated on the Children’s Cancer Group (CCG) #945 protocol for children with MG (vide infra), alteration of PTEN sequence was detected in just one, in conjunction with loss of chromosome 10. Deletions of PTEN without mutations were found in seven additional specimens. The PTEN alterations were more common among GBM than other histopathologic grades (P = .0056). Although 37% (14/38) of evaluable tumors had increased EGFR expression, only one exhibited amplification of the EGFR gene. This may be interpreted to indicate that pediatric MGs differ in the mechanisms of tumorigenesis from those of adulthood.⁵⁷

Site of Origin

It is accepted that the biology of LGGN arising within the hypothalamus, optic tectum, and cervico-medullary junction clearly differs from those arising within the pons and cerebrum. Consensus exists that gross total resection of an AST or mixed LGGN during childhood is adequate therapy and does not require immediate postoperative irradiation (Nishio et al., 1989).⁶² Univariate analysis has demonstrated that optic chiasmal origin and only partial resection are associated with a worsened prognosis.⁶³ This has influenced the controversy regarding EBRT (vide infra).

Neurofibromatosis

The 5Y-OS and 10Y-OS for patients with optic nerve gliomas and NF1 has been reported to be 93% and 81%, respectively. This compares favorably to survival rates of 83% and 76% at 5 and 10 years, respectively, for those children without evidence of NF1. The mean time to disease progression (TDP) was significantly different between the two groups (P < .01) in favor of those with NF1, although this did not impact OS.⁶⁴ However, this relative advantage does not appear to apply to patients with von Recklinghausen disease who suffer from gliomas of the cerebrum.⁶⁵

Histopathologic Grade and Malignant Progression among Low-Grade Astrocytomas

The process of malignant transformation was studied among 11 of 65 children with LGGN. The median latency was 5.1 years, and the 15-year cumulative incidence estimate was 6.7 %. Overexpression of TP53 was more common after progression to higher histopathologic grade. Deletions of RB1 and/or CDKN2A were observed in 71% of the LGGN and 90% of their malignant successors. There were PTEN pathway abnormalities found in 76% of patients.⁶⁶

Pathologic Studies

A series of 330 cases studied by the Pediatric Oncology Group (POG) found significant anaplasia among 24%, which was strongly associated with an unfortunate outcome. Diffuse or extensive anaplasia was worse than focal involvement. In contrast, extensive nodularity conferred a more favorable course.⁷¹ Reevaluation of 347 MBL biopsies treated under the SIOP II trial confirmed that severe anaplasia conferred an adverse prognostic effect on 5-year progression-free survival (49.5%) relative to mild-moderate change (65.4%) (P = .001). This effect was magnified when the presence or absence of extensive apoptosis (P = .00216) was factored in.⁷²

Drug Resistance Genes

There is evidence that the expression of the MDR1, MRP1, LRP and BCRP genes does not correlate with overall outcome among patients with MBL/PNET.⁷³

Biologic Determinants of Survival

Established chromosome abnormalities include amplication of isochromosome 17q, novel amplicons, and losses as well as gains of chromosomes; these are known to occur in greater than 20% of MBL. Copy number abnormalities have been identified in specific regions of chromosomes 1, 8p, 10q, 11p, and 16q, which occur frequently among MBLs and may identify distinct subsets.⁷⁴ Furthermore, differences in DNA copy number at chromosome regions 1p12-22.1, 9p21, 19p, and chromosome 17 have been used to segregate between MBL and supratentorial PNET. The 9p21 deletions correlated with loss of CDKN2A protein expression more frequently among the PNET (P < .001). Gains of 19p were also more evident among the PNET (P = .02), whereas gains of 17q were more common among the MBL (P = .02).⁷⁵

Pathologic Grading

The prognostic utility of pathologic grading of ependymomas has engendered a long-standing controversy, as there has been no clear distinction in 5Y-OS rates between “benign” and “malignant” EPD. However, one recent blinded review demonstrated that the PFS at 2 years following postoperative irradiation was 84% for differentiated EPD, and 32% for specimens considered anaplastic.⁸⁷

Site of Origin

There is a correlation between site of origin and histopathologic grade of malignancy. Approximately two thirds of supratentorial EPDs are high-grade at diagnosis, whereas those arising in the IVth ventricle tend to have a better prognosis.⁸⁸ A series of 49 patients found that 78% originated infratentorially and exhibited low grade histology.⁸⁹ Those arising in the spine and cauda equina are usually low-grade and/or myopapillary.

Extent of Surgical Resection

The CCG ^#921 study accessioned 20 EPD and 12 anaplastic EPD patients, of whom only three had metastatic disease at diagnosis. The prognostic variables of significance included the extent of resection (gross total vs. other, P = .0001) and postoperative residual disease of < 1.5 cm² (P < .0001), in contrast to age, staging, or treatment, emphasizing the importance of local disease control.⁹⁰

The French Society of Pediatric Oncology has studied the efficacy of postoperative chemotherapy, with the intent of avoiding EBRT, among 73 children less than 5 years of age with intracranial EPD. The favorable prognostic variables included supratentorial origin (P = .0004) and complete resection (P = .0009). Patients with gross total resection demonstrated a 4Y-OS of 74% in contrast to those with residual disease (35%).⁹¹

Univariate and Multivariate Analysis

Three prognostic factors have been identified to have a significant correlation with the 5Y-PFS. These included the extent of surgical resection (P < .0001), age at diagnosis (P = .003), and the prediagnosis symptomatic interval (P = .02).²³ Another multicenter retrospective study of 83 children with EPD confirmed that age of less than 3 years, identifiable postoperative residual disease, and Grade III histology were significant adverse factors for PFS in both univariate and multivariate analysis.⁹²

Biologic Determinants of Survival

Overexpression of specific genes (YAP and LOC374491) and downregulation of others such as SULT4A1, NF-[kappa]B2, and PLEK have been implicated in determining the age of onset, relapse potential, and tumor location of pediatric EPD.^93,94 The catalytic subunit of telomerase, the human telomere reverse transcriptase (hTERT), which aids in uncontrolled cell proliferation, has been shown to correlate with prognosis among 65 children with EPD. Study of 87 tumors from these patients found the presence of hTERT to be adversely correlated with 5Y-OS (41% vs. 84% in its absence).⁹⁵

Biologic Determinants of Survival

In several studies of LGGN, the classic histology of the OLG has been strongly associated with 1p deletion (P = .002), loss of 19q (P < .0001), or loss of both (P < .0001).⁹⁷ Deletion of 9p was found in 36% (8/22), always in association with tumor necrosis and/or microvascular proliferation. In addition, epigenetic alterations of CDKN2A were observed in 71% of these 1p/19q/9p deleted OLG, suggesting that it may have a role in microvascular proliferation.⁹⁸ A collaborative study of 162 diffuse gliomas (52 OLG, 79 AST, and 31 AST-OLG) has demonstrated that the combined loss of 1p/19q is a statistically significant predictor (P < .0001) of prolonged survival among patients with pure OLG, even after adjusting for patient age and tumor grade (P < .01).⁹⁷

Allelic deletions on the short arm of chromosome 1 have been correlated with chemosensitivity and a better prognosis for patients with high-grade OLG. The p18^INK4C gene is considered to be a good candidate for the putative TSG located at the chromosome 1p32 locus. The incidence of LOH at 1p is 50% among primary OLG; mutations in the gene have been found among recurrent tumors, implicating it as a possible progression factor as well.⁹⁹ The 1p loss has been shown to be inversely related to deletions of the CDKN2A gene on 9p, which encodes a key cell cycle regulatory molecule p16^INK41.¹⁰⁰ The 19q13.3 locus is deleted in 50% to 80% of OLG. This region codes for the p190RhoGAP gene, which appears to suppress gliomagenesis by inducing a differentiated glial phenotype.¹⁰¹

Histopathologic diagnosis

Matsutani et al.¹⁰² performed a clinical analysis of 153 cases treated between 1963 and 1994 at the University of Tokyo Hospital, which revealed pathologic diagnosis to be the dominant prognostic determinant. The 10-year overall survival rate (10Y-OS) for germinomas was 93%. The 10Y-OS rates for mature and “malignant” teratoma were 93% and 71%, respectively. The diagnoses of embryonal carcinoma, endodermal sinus tumor, and choriocarcinoma had a 3Y-OS of 27%.¹⁰²

Biomarker Expression

The First International Germ Cell Tumor Study observed, in a series of 71 patients, that elevated β-HCG was associated with increased risk of disease progression (P = .06), but did not affect OS.¹⁰³ In a Japanese study, 33 GCTs (16 germinomas, 11 β-HCG positive germinomas, 3 mixed teratoma-germinoma and 3 NG-GCT) were treated with preradiation chemotherapy. Patients with “pure germinomas” had a 86% 5Y-PFS, while germinoma patients with measurable β-HCG expression exhibited only a 44% 5Y-PFS.¹⁰⁴

Surgical Intervention

Among newly diagnosed pediatric MG patients, the CCG ^#943 study demonstrated the impact of the following factors on progression free-survival. Deeply seated diencephalic MG demonstrated a median survival being less than 10 months. Patients receiving only a biopsy experienced a median survival of less than 8 months regardless of subsequent therapy. Partial resection and gross total resection were associated with better but similar long-term outcomes (Table 8-1).²⁷ Among children with MG treated on the CCG phase III study ^#945, radical surgical resection doubled the 5Y-PFS rate (P = .006) (Table 8-1).¹⁰⁶ Unfortunately, it has only been possible to achieve minimal postoperative residual tumor burden for 33% (19% to 45%) of newly diagnosed pediatric MG patients treated in contemporary series.^106–109

Chemotherapy Trials: Phase III and Phase II, Dose Intensification of Chemotherapy with Hematopoietic Support, and Chemo-radiotherapy

Surgical Intervention

A de facto consensus has emerged within the CCG and POG that a biopsy is not warranted in cases of diffuse pontine glioma (DPG) with a “classical” presentation and a diagnostic MRI appearance, as the therapeutic approach employed by current multicenter protocols has not been altered by a specific pathologic diagnosis.¹²⁶

Radiotherapy

Historically, “standard therapy” for the DPG constituted a radiotherapeutic prescription of 45 to 55 Gy, delivered in single daily fractions of 180 to 200 cGy. Unfortunately, the median time of disease progression (TDP) was only 5 to 7 months with an expected survival of 9 to 13 months. Hyperfractionation of the radiotherapy prescription should lessen the neurotoxicity. However, these hyperfractionated radiation therapy (HFEBRT) trials have used doses of 64.8, 66, 70.2, 72, and 78 Gy without significantly altering the TDP.

Previous Combination Therapy Trials

Adjuvant chemotherapy with nitrosoureas has not improved the outcome among children with BSG receiving EBRT. However, pilot studies with neo-adjuvant multiagent chemotherapy did achieve responses in terms of tumor reduction.^126,127 The POG-8833 study of preirradiation chemotherapy and HFEBRT for BSG treated 32 evaluable, newly diagnosed patients with four induction cycles of cisplatin and cyclophosphamide followed by HFEBRT (66 Gy). While clinical responses were reported in 65%, the 1-year PFS was 16%; the median survival of 9 months was not significantly different from the previous POG experience with HFEBRT alone.¹²⁶

The CCG Phase II trial ^#9941 evaluated the efficacy and toxicity of two arms of induction chemotherapy administered prior to HFEBRT. Neither the difference in postinduction response rate nor the difference in postinduction/HFEBRT response rate was statistically significant. The PFS was 17% at one year and 6% at two years. Hyperfractionated radiotherapy was also not found to consolidate the response to induction chemotherapy among those who did achieve a partial or minor response.¹²⁸

Surgical Intervention

Operative resection is the accepted standard of care for actively symptomatic patients or those with progressive disease. Analysis of 70 children with LGGN treated in a single institution demonstrated that extent of surgical resection was the dominant predictor for PFS (P = .015) and OS (P = .013). Actuarial PFS rates were 88%, 79%, and 76% at 5, 10, and 20 years, respectively, following complete removal.¹²⁹

About 80% of infratentorial and only 18% of supratentorial LGGN are accessible enough to allow such radical resection.⁵⁹ A retrospective review of cerebellar AST demonstrated postoperative residual disease in as many as 35%.¹³⁰ Realistically, a significant percentage of children with LGGN have postoperative residual disease and are at risk for disease progression or recurrence.

This raises the related controversy of “early vs. late” surgery. This issue has been difficult to answer in the setting of multi-institutional group trials due to the selection bias toward operating on readily accessible lesions. What data exists would suggest that patients presenting with epilepsy as their sole symptom have a favorable prognosis, which is not adversely affected by waiting until disease progression.¹³¹

Radiotherapy

While EBRT has been the “accepted” treatment for unresectable and/or progressive disease, a quantitative determination of its contribution to disease control remains problematic.^129,132 Postoperative irradiation for adults with incompletely resected LGGN has been associated with 5Y-OS of approximately 48% and 10Y-OS of 26%, which are twice the 26% and 12% rates, respectively, reported for patients receiving surgery alone.¹³³ The addition of EBRT has been found to favorably influence disease control among patients older than 35 years (P = .008). However, this association between younger age at diagnosis with a lower grade of glioma and better prognosis has long been known and compromises interpretation of this data.

Recent analysis of the literature has concluded that for gliomas of the visual pathway, local tumor control with stable or improved visual function is achieved with EBRT in approximately 90% of cases. There is a consensus to employ radiotherapy in older children with progressive disease, regardless of location or histopathologic subtype.¹³⁴ Given the absence of compelling data for the use of EBRT among children, there has been increasing interest in the application of chemotherapy for unresectable, residual, or progressive disease.

Chemotherapy

Among children, single chemotherapy agents have largely been considered ineffective. The Washington group piloted a 10-week induction regimen of carboplatinum-vincristine for children with LGGN. Patients with either SD or a cytoreductive response were additionally treated with twelve cycles of carboplatinum-vincristine. Among 37 newly diagnosed patients, the overall response rate was 62%. In a subsequent report of 70 children less than 5 years of age who were treated with this combination, 57% experienced objective responses, with a PFS rate of 85% at one year and 76% at two years.¹³⁵

The San Francisco group has proposed the TPCV combination (thioguanine, procarbazine, lomustine, and vincristine) in order to defer EBRT among younger children with chiasmal and hypothalamic AST. Among 41 children with a median age of 5.2 years, the median TDP was delayed to 30 months with a 5Y-OS of 83%.¹³⁶

Surgical Resection

The studies reviewed in Table 8-2 have demonstrated the advantage of aggressive surgical debulking to achieve minimal residual disease among children with MBL. Several of these reach statistical significance.^67–70,137 Jakacki et al.¹³⁸ and Albright et al.¹³⁹ investigated the results of the CCG ^#921 study for children with pineal (Figure 8-5) and other supratentorial PNETs. Among these other supratentorial PNETs, postoperative tumor burden of greater than 1.5 cm² was associated with a 4Y-OS of only 13%. This was much worse than that of children with a residual of less than 1.5 cm², 40% of whom were alive after 4 years.¹³⁹

Figure 8-5 Axial (A) and sagittal (B) T1 contrast-enhanced MRI showing densely enhancing pineoblastoma.

(Slides courtesy Susan Chi, M.D., Dana-Farber Cancer Institute and Boston Children’s Hospital.)

Radiotherapy

For children older than 3 to 4 years of age with minimal postoperative residual disease and no evidence of dissemination, treatment with surgical resection and conventional craniospinal EBRT (tumor dose: 54 Gy; whole brain and spinal dose 36 Gy) has achieved a 5Y-PFS rate of approximately 63% (±5) (Table 8-7).⁷⁰

Lower Craniospinal EBRT Dosimetry for Standard-Risk Therapy

There has been increasing concern regarding the neurologic and endocrinologic toxicity of neuraxis irradiation for younger children with standard risk MBL, and attempts have been made to reduce the extratumoral dosage. The SIOP sponsored a study that compared 36 Gy vs. 24 Gy craniospinal EBRT and found the 5Y-PFS was actually better for the reduced dose arm, although some patients did receive preradiation chemotherapy (Table 8-7).¹⁴⁰

A joint CCG-POG trial was performed between 1985 and 1991 to determine whether the craniospinal EBRT prescription could safely be reduced from 36 Gy to 23.4 Gy. The study was halted when preliminary information suggested that the reduced dose arm was associated with a higher early relapse and “distant failure” rate. However, subsequent follow-up studies have found that the 5Y-PFS for the standard radiation prescription arm was greater but not statistically significant when compared to the reduced dose arm (P = .08). These results remained durable at 8 years following treatment, as well.¹⁴¹

Encouraging results with adjuvant chemotherapy had been reported around this time by Packer et al.¹⁴² for the treatment of “high-risk” MBL patients. The addition of eight cycles of lomustine-vincristine-cisplatinum increased the 5Y-PFS to 85%. This exceeded the survival rates being reported among “standard-risk” patients treated with EBRT alone, at the time. Therefore, the CCG initiated trial ^#9892 for children without significant postoperative or metastatic disease, aged 3 to 10 years, using the same adjuvant chemotherapy but with a decrease in the neuraxis radiation prescription (23.4 Gy). The 5Y-PFS of this regimen was a remarkable 78% (Table 8-7).¹⁴³

A series of trials then examined the same reduced EBRT dosage and randomized for adjuvant chemotherapy. These results are summarized in Table 8-7, but suffice it to say that the 5Y-PFS has remained at a plateau of 80% (±6). In one study, patients with average-risk medulloblastoma were treated with 23.4 Gy of CSRT, 55.8 Gy of posterior fossa RT, plus one of two adjuvant chemotherapy regimens: lomustine (CCNU), cisplatin, and vincristine; or cyclophosphamide, cisplatin, and vincristine. Five-year event-free survival and survival was 81% ± 2.1% and 86% ± 9%, respectively. These results were comparable to those obtained with higher doses of irradiation, suggesting that reduced-dose craniospinal radiation may be feasible in children with average-risk medulloblastomas. ¹⁴⁷

High-Risk Medulloblastoma

This cancer would appear the prototypical target for chemotherapy, as it is characterized by a high mitotic rate, has known radiosensitivity, and has a relatively high rate of meningeal dissemination (36% to 43%) at time of diagnosis.¹⁹

Phase III Trials regarding the Role of Chemotherapy in Treatment of Medulloblastoma

Two parallel-designed studies conducted by the European (SIOP-1) and North American CCG (#942) have been discussed (see Prognostic Variables). The SIOP-I enrolled patients between 0 and 16 years during the years 1975 through 1979, and randomized between surgery and irradiation vs. surgery, EBRT, and CCNU-vincristine. Analysis at 54 months demonstrated a statistically significant advantage in outcome for the combination therapy arm (P = .005) (Table 8-8).⁶⁷ The CCG ^#942 accessioned patients between 2 and 16 years, during the years 1975 through 1981. Comparison was made between surgery and EBRT vs. surgery, irradiation, and adjuvant chemotherapy with CCNU-vincristine-prednisone. Among the high-risk patients, the PFS was significantly better for the EBRT-chemotherapy treated patients (P = .006) (Table 8-8).⁶⁸ The third randomized trial was a POG study comparing conventional neuraxis irradiation with/without MOPP chemotherapy (1979 to 1986), which demonstrated no advantage to adjunctive pharmacotherapy among children less than 5 years of age (Table 8-8).¹⁵⁰

Multiagent Combination Chemotherapy Regimens

Preliminary studies of the Eight in One combination chemotherapy regimen had been encouraging in terms of response rates among MBL/PNET patients (Table 8-9). The CCG ^#921 protocol was designed to compare the relative efficacy of adjuvant CCNU-vincristine-prednisone for high-risk patients with this more aggressive chemotherapy regimen before and after EBRT. This study demonstrated no significant advantage for the Eight in One polypharmacy combination (P= .06) (Table 8-9).⁷⁰

Synergistic Chemotherapy Studies

Platinators and topoisomerase antagonists have become increasingly popular by virtue of a putative synergistic interaction in treatment of a number of pediatric brain tumors (Table 8-9). The German HIT’91 trial demonstrated that high dose, multiagent chemotherapy could be as effective in achieving a complete response (42-57% of patients) as irradiation (57-63%) among newly diagnosed children with MBL (Table 8-9).^157,158

Dose Intensive Chemotherapy with Autologous Bone Marrow Transplant or Peripheral Stem Cell Rescue

Several groups have examined the feasibility, toxicity, and efficacy of dose-intensification using this approach. A collaborative study treated 53 children with newly diagnosed MBL/PNET, 19 of whom were high-risk, and 34 with average-risk disease. All patients were treated with craniospinal EBRT. Postradiotherapy chemotherapy consisted of four courses of high-dose cyclophosphamide-cisplatin-vincristine administered with PBSC or ABMT hematopoietic support. The high-risk patients additionally received topotecan in a 6-week phase II window between EBRT and this chemotherapy regimen. Early outcome analysis revealed a 2Y-PFS of 93.6% among the average-risk patients and 73.7% among high-risk subjects.¹⁶⁰

The New York University Group employed five courses of cisplatin-etoposide-cyclophosphamide-methotrexate with leucovorin rescue for MBL patients with disseminated disease. Following this induction, eligible patients were treated with a single myeloablative course of chemotherapy and PBSC. The 81% complete response rate and 49% 3Y-PFS were considered encouraging enough to warrant additional trials with methotrexate.¹⁶¹

Surgical Management

Historically, meaningful surgical debulking has been possible in about 42% to 62% of patients, usually those with tumors originating supratentorially or in the roof of the IVth ventricle. Not surprisingly, the 5Y-PFS rates have been poor, on the order of 23% to 45%. In recent studies, aggressive surgical resection has improved the 5Y-PFS estimates to 51% to 75%.^23,89,90,92

Radiotherapy

There has been no study to establish the optimal dose or appropriate treatment volume for EBRT among pediatric or adult patients with EPD. Comparison between patients treated with surgery versus surgery plus EBRT has shown long-term survival rates of 17% and 40%, respectively.¹⁶² A dose-response relationship was documented.

The requirement for craniospinal EBRT among children with anaplastic EPD has been controversial. Goldwein et al.¹⁶³ have shown that the 2Y-OS for children receiving neuroaxis irradiation was 52%, while focal irradiation was resulted in a 40% survival rate over the same period. From the perspective of long-term survival, five out of 11 children treated with craniospinal EBRT were alive at 6 years, compared to none of those treated with local EBRT.¹⁶³ Paulino et al.⁸⁹ cited the 10Y-OS rates to be 64% for patients receiving craniospinal EBRT, 60% for whole brain irradiation, and 65% for local field radiotherapy (P = 0.88).

Chemotherapy

In retrospective analysis of 83 children with EPD, adjuvant chemotherapy did not significantly alter the PFS among patients older than 3 years of age, nor did craniospinal EBRT improve the PFS for M₀ patients.⁹² Duffner et al.¹⁰⁷ demonstrated responses and durable survival with two cycles of cyclophosphamide-vincristine among 25 infants less than 3 years of age. The VETOPEC regimen achieved an impressive initial response rate (Table 8-10).¹⁰⁹ The French Society of Pediatric Oncology has studied the efficacy of postoperative chemotherapy, in hopes of avoiding EBRT among 73 children less than 5 years of age with intracranial EPD. The regimen consisted of seven cycles of three courses, alternating two drugs at each course (procarbazine–carboplatin, etoposide–cisplatin, cyclophosphamide−vincristine) over 1½ years. Unfortunately, there were no complete or partial responses observed. While the 4Y-OS was 59%, only 23% of those surviving evaded irradiation (Table 8-10).⁹¹ Disappointingly, dose intensification using PBSC with cyclophosphamide–etoposide−vincristine, with or without cisplatinum, has achieved response rates of only 16% among children with EPD less than 6 years of age.¹²¹

Surgical Intervention

In Japan, it has been the practice to first irradiate a pineal tumor thought to be a germinoma with 20 Gy (“the radiation test”), and then, if the tumor regressed, to continue radiotherapy. If there was no reduction in tumor size, only then was surgical excision to be considered.¹⁶⁷ Over time, aggressive surgical resection has proven itself to be an important determinant of survival, as evidenced in the largest single series of patients (number 153) reported by the University of Tokyo.¹⁰²

The pendulum has swung back again, because of the increasing effectiveness of nonsurgical therapies. If the appropriate neuroradiographic findings and biomarker serologic studies are present, the Japanese and Korean Societies for Pediatric Neurosurgery advocate a minimally invasive surgical procedure, such as neuroendoscopic or stereotactic biopsy, to be followed by platinum-based chemotherapy and/or targeted EBRT.¹⁶⁷ It is only among cases demonstrating a poor response to “trial therapy” that surgical intervention is then reconsidered to debulk the tumor and clarify the pathologic diagnosis.¹⁶⁸

External Beam Radiotherapy

Among patients with known germinomas, combined surgical resection and radiotherapeutic intervention improved 10Y-OS rates from 69% to 93%.¹⁰² The typical radiotherapy prescription today consists of a primary tumor dose of 50 to 55 Gy, with 36 Gy to the neuraxis. Subsequent controversy regarding the treatment of germinomas has centered on the issues of dosimetry and the indication for craniospinal radiotherapy. As germinomas of the CNS proved to be as radiosensitive as those of the testis, it was shown that those without evident CSF dissemination could be controlled with involved field radiotherapy alone.^102,169,170 In contrast, there is no debate regarding the necessity of treating NG-GCT with a full radiotherapy prescription to the tumor (50 to 55 Gy) and craniospinal axis (36 to 40 Gy) due to their significantly higher rate of metastasis and recurrence.^8,168,169 Unfortunately, the NG-GCTs remain refractory to radiation prescriptions even greater than 50 Gy.

Chemotherapy

A number of reports investigating the role of chemotherapy in the management of GCT have been encouraging (Table 8-11).^{103,171–174}

The practice of the University of Tokyo has been to rank GCT patients by relative risk: (i) mixed germinoma and teratoma, (ii) mixed GCT with predominance of germinoma or teratoma with some “pure malignant tumor” (embryonal carcinomas, endodermal sinus tumors, and choriocarcinomas), and (iii) mixed tumors with predominance of “pure malignant tumor.” Surgery and radiotherapy produce a 10Y-OS rate of 91.7% among germinoma patients. In the intermediate prognostic group, combination chemotherapy (cisplatin-vinblastine-bleomycin, cisplatin-etoposide, or carboplatin-etoposide) and EBRT has been shown to significantly reduce the risk of disease recurrence when compared to irradiation alone (P = .049). The high-risk patients did relatively better with chemotherapy than with only EBRT, although the difference was not statistically significant.¹⁰²

The First International Germ Cell Tumor Study proposed a chemotherapy-only regimen of carboplatin, etoposide, and bleomycin. This study accessioned 45 patients with germinoma and 26 with NG-GCT, of whom 68 were considered evaluable. The protocol for germinoma patients, who achieved a complete response after four induction courses, prescribed two additional cycles. Those subjects with a lesser response were treated with a chemotherapy regimen fortified by cyclophosphamide then followed by EBRT. A complete response was achieved in 57% of patients after four induction courses of chemotherapy, and an additional 24% were left with no evident disease after intensified chemotherapy or “second-look surgery.” Thus, a significant majority were rendered disease-free without irradiation (Table 8-11). Thirty-nine percent of patients remained in complete remission over a median follow-up of 31 months.¹⁰³

The Cooperative German/Italian Study accepted the classic neuroradiographic appearance with elevated biomarkers (β-HCG and/or α-FP) as diagnostic entry criteria. Nineteen patients (16 males) were placed into the study. The therapeutic design consisted of two induction courses of PEI (cisplatinum–etoposide-ifosfamide); patients responding to chemotherapy were to receive an additional two courses. Nonresponders and those with progressive disease were to be advanced to surgical resection, if feasible, prior to craniospinal EBRT (30 Gy with a tumor boost of 24 Gy). Thirteen of 16 with elevated α-FP and/or β-HCG had normalization of biomarker levels following the second course of PEI induction chemotherapy, which paralleled objective neuroradiographic evidence of a cytoreductive effect in ten. Increasing response rates were seen with further chemotherapy (Table 8-11). Seventeen of the patients survived, of whom 81% remain in remission over a median follow-up interval of 11 months (range 7 to 39 months).¹⁷²

The Second International CNS Germ Cell Study Group employed two courses of cisplatin–etoposide–cyclophosphamide–bleomycin to assess chemosensitivity in a group of 20 patients with NG-GCT. The study design planned that participants achieving a complete response would receive two additional courses of carboplatin–etoposide–bleomycin and another cycle of the original treatment regimen. Those not reaching or sustaining this underwent “second-look surgery” and/or EBRT. A greater than 50% tumor cytoreduction was achieved in a remarkable majority of patients (Table 8-11). The median PFS for subjects with a complete response was 62 months compared with patients with lesser responses, who demonstrated a mean PFS of 23 months. Regrettably, 69% of evaluable patients suffered from progressive disease during or following chemotherapy.¹⁷³

Studies of infants with Malignant Astrocytomas

The “Baby POG” protocol design consisted of alternating cycles of cyclophosphamide-vincristine and cisplatinum-etoposide.¹⁰⁷ This combination yielded “very encouraging” results among infants with MA and DPG, which exceeded the responses seen at the time among older children treated with postoperative EBRT alone (PFS 20%), irradiation with CCNU–vincristine−prednisone, or the Eight in One regimen before and after EBRT (Table 8-4).^27,107 Experience with Eight in One combination, the VETOPEC regimen (vincristine-etoposide-cyclophosphamide for four courses, followed by sequential administration of cyclophosphamide-vincristine, cisplatin-etoposide, and carboplatin-etoposide), and the more recently reported BBSFOP Protocol (seven cycles of three drug pairs: carboplatin-procarbazine, cisplatin-etoposide, cyclophosphamide-vincristine) is shown in Table 8-12.^28,109,177 None of these has emerged as clearly superior.

Infants with Standard and High-Risk Medulloblastoma, Primitive Neuroectodermal Tumor and Ependymoma

The “Baby POG” study of alternating cyclophosphamide-vincristine and cisplatinum-etoposide yielded both responses and durable remissions for young children with MBL (Table 8-12).¹⁰⁷ The CCG study ^#921 administered the Eight in One regimen; many of the infants, with different tumor types, were able to avoid irradiation (Table 8-12).¹⁵¹ The concurrently-administered combination of cisplatin-cyclophosphamide-etoposide-vincristine has been considered very active in infants with newly diagnosed CNS tumors, particularly supratentorial PNETs, with a response rate of approximately 80% in this high risk subgroup (Table 8-12).¹⁵⁵ More recent studies have acknowledged 3Y-PFS rates of less than 50% with the Eight in One regimen and the VETOPEC regimen among babies with MBL and PNET.^109,159 The addition of intrathecal and intravenous methotrexate to the platinator-cyclophosphamide-etoposide-vincristine structure appeared to make a difference among the younger infants with identifiable postoperative disease, however (Table 8-12).¹⁷⁸

The POG study ^#9233 was the first to compare conventionally dosed versus dose-intensive induction chemotherapy among infants with newly diagnosed MBL (101 patients) and EPD (84 eligible patients). Induction consisted of cyclophosphamide-vincristine and cisplatin-etoposide, with the investigational arm receiving a relative dose intensification of 1.8. The MBL patients showed no difference in PFS (P = .67) between the two arms. The subjects with EPD found that dose intensification was advantageous in terms of PFS (P = .003).¹⁷⁹

Dose-Intensive Chemotherapy with Autologous Stem Cell Rescue

A series of 15 children less than 38 months of age with newly diagnosed malignant CNS tumors was reported using high-dose chemotherapy with PBSC followed by selective EBRT. Three courses of cisplatin-cyclophosphamide-etoposide-vincristine were succeeded by three cycles of carboplatin-thiotepa with reinfusion of PBSC. Ten patients are disease-free at a median of 18 months; five received EBRT (Table 8-12). There were no toxicity-related deaths.¹⁸⁰

The “Head Start” I and II protocols administered postoperative induction chemotherapy with vincristine-cisplatin-cyclophosphamide-etoposide for five cycles. Myeloablative chemotherapy with carboplatin-thiotepa-etoposide was supported with PBSC. Irradiation was used only at relapse. A series of 21 children with nonmetastatic MBL were studied. The 5Y-PFS rates are shown in Table 8-12; gross total resection and desmoplastic morphology were associated with especially favorable results. Most survivors (71%) and 52% of the whole cohort avoided EBRT. Mean intellectual function among these patients remained within the average range for the majority tested.¹⁸¹