Phase II study of high-dose chemotherapy before radiation in children with newly diagnosed high-grade astrocytoma

Final Analysis of Children's Cancer Group Study 9933




High-grade astrocytomas (HGA) carry a dismal prognosis and compose nearly 20% of all childhood brain tumors. The role of high-dose chemotherapy (HDCT) in the treatment of HGA remains unclear.


In a nationwide study, The Children's Cancer Group (CCG) prospectively evaluated 102 children with HGA and postoperative residual disease for efficacy and toxicity of four courses of HDCT before radiotherapy (RT). Patients were randomly assigned to one of three couplets of drugs: carboplatin/etoposide (Regimen A), ifosfamide/etoposide (Regimen B), or cyclophosphamide/etoposide (Regimen C). After HDCT, all patients were to receive local RT followed by lomustine and vincristine. Twenty-six patients were excluded after central neuroradiographic review (n = 8) or pathology review (n = 18).


Of 76 evaluable patients (median age, 11.95 yrs; range, 3–20 yrs), 30 patients relapsed during HDCT, and 11 others did not complete HDCT because of toxicity. Nonhematologic serious toxicities were common (29%), and 21% of patients did not receive RT. Objective response rates were not associated with amount of residual disease and did not statistically differ between regimens: 27% (Regimen A), 8% (Regimen B), and 29% (Regimen C). Overall survival (OS) was 24% ± 5% at 5 years and did not differ between groups. Median time to an event was longest for Regimen A (283 days compared with 83 and 91 days for Regimens B and C, respectively). The five-year, event-free survival (EFS) rate for all patients was 8% ± 3% and 14% ± 7% for Regimen A (P = 0.07).


OS and EFS were not affected by histologic grade. Patients who responded to HDCT had a nominally higher survival rate (P = 0.03 for trend). The authors conclude that these commonly used HDCT regimens provide no additional clinical benefit to conventional treatment in HGA, regardless of the amount of measurable residual tumor. Cancer 2005. © 2005 American Cancer Society.

Despite significant improvements in outcomes of most childhood central nervous system tumors, the prognosis of high-grade astrocytomas (HGA) remains poor.1 Extensive resection of HGA frequently improves neurologic function and correlates with more favorable outcome, yet it is rarely curative because of local recurrence.2, 3 The addition of postoperative involved-field radiation therapy (RT) has yielded at best a 6–18% 5-year survival rate.4–8 Although children with HGA appear to benefit from adjuvant chemotherapy with postoperative RT, the extent of this benefit remains unclear. The Children's Cancer Group (CCG) study 943 compared postoperative RT alone with postoperative RT plus adjuvant chemotherapy that consisted of prednisone, lomustine (CCNU), and vincristine (pCV regimen) in a cohort of 58 patients.8 In this study, the 5-year event-free survival (EFS) reported initially was 46% for the group that received RT and adjuvant chemotherapy compared with 18% for those with RT alone. However, in a subsequent trial (CCG-945) that compared pCV to eight drugs in one day (8-in-1), there was no significant difference in overall survival (OS) between groups (39% ± 7% compared with 29% ± 8%, respectively), and the EFS (26%) associated with pCV was substantially inferior to that noted in the original study.9 Further review of the histology of patients in CCG-943 suggested that a significant proportion (69/250, 37.7%) did not fit the central consensus definition of HGA. Nevertheless, both regimens remained superior to surgery and RT alone, suggesting that adjuvant chemotherapy has a positive effect on treatment of HGA and that results may be better with more effective chemotherapy.

One of the major obstacles in optimizing chemotherapy for HGA has been the paucity of data on efficacy of single agents or combinations, especially in untreated patients. Most Phase II studies were done with patients who were heavily pretreated, which might have conferred resistance to new agents after recurrence. Thus, testing agents before RT or other chemotherapy may help identify agents that could be given with standard treatment to improve outcome. This study tested three promising combinations of agents in the therapeutic window between surgery and RT. Patients with postoperative residual disease that measured > 1.5 cm2 were assigned randomly to receive one of three chemotherapeutic regimens: cyclophosphamide/etoposide (VP-16) (Regimen A), carboplatin/VP-16 (Regimen B), and ifosfamide/VP-16 (Regimen C), all given at maximal dose-intensity. The rationale for these agents is supported by objective response rates as high as 50% seen in patients with recurrent HGA treated with high-dose cyclophosphamide and the observed activity of high-dose carboplatin and ifosfamide in pediatric brain tumors, particularly in combination with VP-16.10–16 However, the efficacy and role of high-dose chemotherapy (HDCT) in the treatment of newly diagnosed HGA, especially in the setting of measurable residual disease, remains controversial.

The primary aims of this study were to determine efficacy and toxicity of four courses of HDCT before RT within a 12-week Phase II therapeutic window. After HDCT, patients were treated with involved-field RT for a total dose of 5940 cGy and weekly vincristine, then maintenance chemotherapy with CCNU and vincristine. Patients whose disease progressed during HDCT proceeded directly to RT and maintenance therapy. Thus, the secondary objectives of the study were to determine OS, EFS, and whether HDCT before RT alters tolerance of it.


The CCG-9933 study was opened in 1993 and closed to patient accrual in 1998. A total of 102 patients were registered.

Eligibility Criteria

Patients were eligible if they had newly diagnosed, nonirradiated, pathologically proven HGA and measurable residual disease ≥ 1.5 cm2 documented by contrast-enhanced magnetic resonance imagery (MRI) or computed tomography (CT). They had to be older than 3 years of age at entry and younger than 26 years of age when diagnosed originally. Patients with brain-stem tumors, primary spinal cord tumors, or oligodendroglioma were not eligible; however, those with mixed tumors that included a major component of HGA and oligodendroglioma were eligible. Patients were eligible if they had low-grade astrocytomas that were believed to have evolved to high-grade tumors without exposure to RT or chemotherapy, if the tumors were biopsy proven. Patients also were eligible who had documented disseminated disease to the cerebrospinal fluid (CSF) or spinal cord. Patients were required to begin induction chemotherapy within 12 weeks of initial surgery.


Patients were assigned randomly at entry. Randomization was stratified by glioblastoma multiforme (GBM), including GBM variant (GV), anaplastic astrocytoma (AA), and other HGAs. Patients were assigned to one of three induction chemotherapy regimens to begin after initial diagnostic resection or biopsy and before radiation.

Neuropathology Review

All tumor specimens were subjected to a primary and secondary independent neuropathology review by two separate neuropathologists. If diagnoses of primary- and secondary-review neuropathologists were discrepant, the reviewers consulted to reach a final consensus diagnosis.


Induction therapy

Induction therapy consisted of one 12-week course comprising four 3-week cycles beginning as soon as the patient had recovered sufficiently from surgery according to the neurosurgeon and neurooncologist but no later than 12 weeks from definitive surgery. Patients were assigned randomly to receive one of three induction regimens: carboplatin, VP-16 (Regimen A); ifosfamide, mesna, VP-16 (Regimen B); or cyclophosphamide, mesna, VP-16 (Regimen C). Regimen A cycles consisted of intravenous (i.v.) carboplatin at 600 mg/m2 × 2 days and i.v. VP-16 at 166 mg/m2 × 3 days. Regimen B cycles consisted of ifosfamide at 2400 mg/m2 × 5 days and i.v. VP-16 at 100 mg/m2 × 5 days. Regimen C cycles consisted of i.v. cyclophosphamide at 2100 mg/m2 × 2 days and i.v. VP-16 at 166 mg/m2 × 3 days. In Regimens B and C, mesna was given at a total dose of 60% of the total dose of ifosfamide or cyclophosphamide for bladder protection. For each regimen, granulocyte colony-stimulating factor (G-CSF) (5 μg/kg) was given subcutaneously beginning 24 hours after the last dose of VP-16 and until the absolute neutrophil count (ANC) was ≥ 10,000/mm3 to protect against severe myelosuppression. Each subsequent cycle started when patients had an absolute phagocyte count (APC) > 750/mm3 and platelet count > 100,000 mm3 without transfusion and at least 3 weeks had passed from the start of the previous cycle. If hematologic recovery was not sufficient to allow the next cycle to begin by Week 4 at the recommended dose, doses of each chemotherapeutic agent were reduced in 20% decrements with reescalation to previous dose for recovery in 3 weeks. Patients who developed signs or symptoms of tumor progression during induction were evaluated with MRI or CT, and if progression was proven conclusively, induction chemotherapy was discontinued and interim therapy was initiated.

Interim therapy

Interim therapy was identical for all regimens, consisting of one 12-week course of weekly i.v. vincristine at 1.5 mg/m2 (2 mg maximum) for 8 weeks, concomitant with a 6-week course of RT, followed by a 4-week rest.

Maintenance therapy

Maintenance therapy was identical for all regimens, comprising eight 4-week cycles beginning 6 weeks after completion of RT and each subsequent course starting when patients had ANC > 750/mm3 and platelet count > 100,000 mm3 without transfusion, and at least 4 weeks had passed from the start of the previous cycle. Each cycle consisted of oral CCNU at 100 mg/m2 × 1 day and i.v. vincristine 1.5 mg/m2 (2 mg maximum) weekly for 4 weeks. CCNU dose was reduced in 25% decrements if cycles could not be given at least every 6 weeks because of delayed hematologic recovery.

Radiation therapy

Radiation therapy was initiated at the completion of the 12-week induction course, provided the patient had hematologic recovery (defined as ANC > 750/mm3 and platelet count > 75,000/mm3). Interim therapy with RT also was initiated immediately for any patient with conclusive tumor progression during induction. The total dose delivered to the original tumor site, defined by contrast-enhanced MRI or CT, plus a 2.0-cm margin, was 5400 cGy in 30 fractions of 180 cGy each. An additional boost field included the whole tumor volume after surgery or after chemotherapy, whichever was larger, plus a 1.0-cm margin and a final total tumor dose of 5940 cGy (3 fractions of 180 cGy each). The dose volume to spinal cord lesions included the documented disease, 1 vertebral body above and below the disease and 1.0 cm beyond the entire width of the vertebral body. The tumor dose in the spinal cord was 4680 cGy unless > 50% of the spinal cord was to be treated, then the dose was 4500 cGy with a boost to 4680 cGy to areas of gross disease.


The use of corticosteroids was left to clinician discretion; however, it was recommended that corticosteroids be restricted to treatment of raised intracranial pressure (ICP) and adrenal insufficiency whenever possible.

Response Criteria

Independent and blinded central radiographic review was performed by two separate neuroradiologists to evaluate response. Complete response (CR) was defined as resolution of tumor on CT or MRI scan, myelogram, or CSF examination. Continuing complete response was defined as no evidence of residual tumor in a patient who achieved CR by surgery or chemotherapy. Partial response (PR) was defined as > 50% reduction in the product of the greatest tumor diameter and its perpendicular diameter measured on CT or MRI scan. Minor response (MR) and stable disease (SD) were defined as 25–50% reduction and < 25% reduction over a 4-month interval, respectively, by the parameters described above. Progressive disease (PD) was defined as > 25% increase in tumor size by CT or MRI scan or the appearance of tumor in a previously uninvolved area.


Physical or neurologic examination was performed at least every 3 weeks during induction and interim therapy. Neuroimaging (contrast-enhanced MRI or CT) was performed at completion of induction and interim therapy, then at 1-year intervals from entry, or at the time of progressive disease, or at relapse. There were 50 (66%) patients evaluated for residual disease and tumor progression by MRI, 17 (22%) by CT, and 9 (12%) by MRI and CT.

Statistical Methods

The primary endpoint for statistical analysis was tumor response to chemotherapy. Secondary endpoints included OS, which is time from entry to death from any cause, and EFS, defined as the minimum time from entry to disease progression, relapse, a second malignant neoplasm, or death from any cause. All analyses were based on intent-to-treat, whereby patients were not censored for deviations from protocol therapy. Nonparametric estimates of EFS and OS probabilities were obtained using the product-limit estimate, with standard errors computed using the Greenwood formula.17


Patient Characteristics

Of 102 patients registered on CCG-9933, 26 were ineligible after review. Eight did not have measurable postoperative residual disease documented by enhanced MRI neuroimaging, and 18 had ineligible or wrong diagnoses. Of the remaining 76 patients, 23 were assigned randomly to Regimen A, 27 to Regimen B, and 26 to Regimen C. The overall median age on study for the 76 eligible patients was 11.95 years (range, 3–20 yrs); the median age by regimen was 12.4 years for A, 11.1 for B, and 12.3 for C. The percentage of males was 47.4%; by regimen, it was 43.5% males in A, 59.3% in B, and 38.5% in C. For the entire eligible group, there were 69.7% Whites, 14.5% Hispanics, 10.5% Blacks, and 5.3% Other or Unknown; by regimen, there were 60.9% Whites in A, 77.8% in B, and 69.2% in C. Most patients (86.8%) had supratentorial tumors based on all sites of tumor involvement; of the 66 patients with supratentorial tumors, 39 primarily were in the cerebral hemisphere. The remaining 13.2% included 6 infratentorial (4 with cerebellar involvement, 1 fourth ventricle, 1 with other infratentorial site), 1 with both supra- and infratentorial (parietal lobe and ventricles including fourth), and 3 unknown site. Five percent of the 76 eligible patients had metastatic disease. The percentages of supratentorial tumors for regimens A, B, and C were 91%, 85%, and 85%, respectively.

Patient Histology

All patients were to have histologic verification of high-grade astrocytoma. Table 1 gives the breakdown by regimen of the composite diagnosis based on the reviewer pathology when available (for 73 patients) or the institutional pathology (for 3 patients).

Table 1. Pathology for Eligible Patients (N = 76)
A no. (%)B no. (%)C no. (%)Total no. (%)
  1. GBM: glioblastoma multiforme; GV: GBM variant; AA: anaplastic astrocytoma.

GBM/GV10 (43)14 (52)16 (62)40 (53)
AA 9 (39)11 (41)10 (38)30 (39)
Other 4 (17) 2 (7) 0 6 (8)

Residual Disease

Patients were to have measurable postoperative residual disease, documented by enhanced MRI or CT neuroimaging. Postoperative residual disease based on the reviewer or institutional neuroradiology (NR) data was ≥ 1.5 cm2 for 72 of 76 patients. Although data were missing for four patients, they were deemed eligible based on available imaging reports, but they were not evaluable for response.

Objective Response to Chemotherapy

The median number of days on chemotherapy for the 76 eligible patients was 88 (range, 1–141 days); the medians for regimens A, B, and C were 91, 85, and 89.5 days, respectively. Table 2 shows the status of the 72 evaluable patients by regimen by the end of chemotherapy based on institutional data. Thirty patients had relapsed or had progressive disease; 11 went off protocol therapy during or at the end of chemotherapy as follows: 1 developed toxicity (poor creatinine clearance), 7 were withdrawn by physicians or parents, 1 failed to respond, 2 died. Nine of 30 patients who had progressive disease also were withdrawn from protocol therapy. On the basis of overall response at the end of induction, postoperative chemotherapy with carboplatin/VP-16 (Regimen A) had the lowest rate of progressive disease (32%) compared with ifosfamide/VP-16 (Regimen B) (50%) and compared with cyclophosphamide/VP-16 (Regimen C) (42%). The rate of objective response (CR, PR, MR) for the three regimens was 27% for A, 8% for B, and 29% for C; there were no CRs. There was no significant difference in the rate of objective response between the regimens by Fisher exact test (P = 0.10), comparing those with or without objective response. Also, there was no significant association between regimens and failure during induction (P = 0.48), comparing those whose treatment failed (defined as PD, withdrawal, death, or removal due to toxicity) versus those whose treatment did not fail (PR/MR/SD); the percentage of those who had failed induction was 45% for A, 62% for B, and 63% for C. The status of one patient in Regimen B was unknown because films or forms were unavailable; because he completed chemotherapy and went to RT, we assumed he was to be in the group whose treatment did not fail. Also, there was no significant association between response and amount of residual tumor (1.5–3.0 vs. > 3.0 cm2) by Fisher exact test (P = 0.72); the percentage of patients with tumors that exceeded 3.0 cm2 was 87% for PR/MR, 73% for SD, and 73% for PD.

Table 2. Status of Evaluable Patients at the End of Chemotherapy
StatusA no.B no.C no.Total
 (%)(%)(%)no. (%)
Partial response PR 2 (9) 0 5 (21) 7 (10)
Minor response MR 4 (18) 2 (8) 2 (8) 8 (11)
Stable disease SD 6 (27) 7 (27) 2 (8)15 (21)
Progressive disease PD 7 (32)13 (50)10 (42)30 (42)
Withdrawn/Died/Toxicity 3 (14) 3 (12) 5 (21)11 (15)
Unknown 0 1 (4) 0 1 (1)


After chemotherapy or progression, 52 of 76 (68%) eligible patients went to radiotherapy, as prescribed by protocol. Eight additional patients had nonprotocol radiotherapy, seven after progression and withdrawal from protocol therapy, and one after removal due to toxicity. Sixteen patients did not have radiotherapy; 11 went off protocol at physician or patient request, toxicity, or death, and 5 were withdrawn, 4 after progression and 1 after minor response. The cranial doses delivered ranged from 900 cGy to 6600 cGy with median and mode of 5940 cGy. Two patients had additional RT to the spine for leptomeningeal dissemination. Table 3 summarizes the numbers by regimen.

Table 3. Radiotherapy by Regimen
 ABCTotal no. (%)
Protocol RT after chemotherapy or PD17181752 (68)
Nonprotocol RT after PD & w/d or toxicity242 8 (11)
No known RT– off protocol therapy45716 (21)

Of 52 patients who had protocol radiotherapy, 22 had RT after progression on chemotherapy (median 8 days), 21 had RT before progression (median time to PD 147 days), and 9 had no progression. By regimen, 7 (41%) patients in Regimen A had RT before PD compared with 9 (50%) in B, and 5 (29%) in C; 5 (29%) in Regimen A had RT after PD compared with 8 (44%) in B, and 9 (53%) in C. The median time from surgery to beginning of RT was 114 days (range, 27–154 days). The median and most common total dose was 5940 cGy.


During induction, 22 of 76 (29%) patients had common toxicity criteria Grade 3 or 4 nonhematologic toxicity, including 8 in Regimen A, 9 in B, and 5 in C. Of 43 Grade 3 or 4 toxicities, 11 were in the CNS. For the 11 patients with CNS toxicity, 6 had seizures as a symptom, 6 had motor weakness, 6 had adventitial movements, and 4 had ataxia; several had multiple symptoms or signs. Table 4 lists the sites that had at least 2 occurrences; there were 8 other toxicities that occurred only once.

Table 4. Occurrence of Grade 3 or 4 Nonhematologic Toxicity during Chemotherapy
Site– measureABCTotal
Nervous system– central35311
Gastrointestinal– nausea/vomiting0303
Renal/Genitourinary– hematuria0303
Nervous system– peripheral1203
Renal/Genitourinary– BP diastolic1012
Electrolytes– Na0022
Performance (Karnofsky %)1102
Other toxicity4228

During the radiotherapy phase, 10 of 52 (19%) patients had Grade 3 or 4 toxicity; 3 patients were in Regimen A, 5 in B, and 2 in C. Of 20 Grade 3 or 4 toxicities, the CNS was, again, the most common site with 4, followed by infections, which numbered 3. During the chemotherapy phase, 58 of 76 (76%) patients were given red blood cell transfusions. Of 18 patients who did not have transfusions, 9 (50%) were in Regimen B. The median number of transfusions was 2 (range, 0–5) for Regimen A, 2 for Regimen B (range, 0–6), and 3.5 for Regimen C (range, 0–14). Also, 48 of 76 (63%) patients were given platelet transfusions. Of 28 patients who did not need platelet transfusions, 23 (82%) were in Regimen B. The range of platelet transfusions for Regimen B was 0–5 compared with 0–17 for Regimens A and C; the median was 3 for Regimen A and 3.5 for C. The median number of days that G-CSF was given was 45 (range, 5–72 days) for Regimen A, 37.5 (0–55 days) for B, and 43 (0–70 days) for C. Fifteen patients on Regimen C had fever and neutropenia compared with 5 on A and 7 on B. The median days hospitalized during this course were 13, 21, and 25.5 for Regimens A, B, and C, respectively.

Survival and EFS

Of 76 eligible patients, 56 (74%) died; 52 deaths were primarily disease related, 1 was due to infection, 2 to hemorrhage, and 1 to another reason (Ara-C pulmonary toxicity for a patient who developed acute myeloid leukemia). The 5-year OS rate based on the 76 eligible patients was 24% ± 5% (Fig. 1); the survival rate for Regimens A, B, and C was 18% ± 8%, 39% ± 10%, and 16% ± 7%, respectively (P = 0.23) (Fig. 2).

Figure 1.

Overall survival for eligible patients.

Figure 2.

Overall survival by regimen.

The 5-year EFS rate was 8% ± 3% (Fig. 3). The EFS rate for Regimens A, B, and C was 14% ± 7%, 4% ± 4%, and 8% ± 6%, respectively (P = 0.07) (Fig. 4). Regimen A had a modest increase in patients with “other” histology compared with Regimes B and C. If patients with “other” type histology in Regimen A are removed from analysis, there is less difference in EFS between the regimens, but it is not significant. Similarly, comparing EFS by histology only for Regimen A, although not significantly different, the “other” histology EFS is nominally higher. The median time to relapse, progression, or death based on eligible patients was 105 days. Regimen A had the longest (283 days) compared with B (83 days) and C (91 days). Although Regimen B had the lowest 5-year EFS rate (4% ± 4%), it had the highest OS rate (39% ± 10%), but the difference in OS between regimens was not significant. Of 13 evaluable patients in Regimen B who progressed during induction, 10 (77%) had nonprotocol anticancer treatment later. Two of 13 patients and 2 other patients in Regimen B had second surgery; all 4 were alive at last contact. Three of 7 (43%) patients in Regimen A and 7 of 10 (70%) in Regimen C progressed during induction and had nonprotocol treatment. One patient in Regimen A and two in Regimen C had second surgery, but all died.

Figure 3.

Event-free survival for eligible patients.

Figure 4.

Event-free survival by regimen.

The 5-year survival rate after relapse for Regimen B was 38% ± 10% compared with 0% for A and 11% ± 7% for C (P = 0.13). The 5-year survival rate from the start of any RT (protocol or nonprotocol) for Regimen B was 43% ± 11% compared with 14% ± 8% for A and 21% ± 9% for C (P = 0.40). The 8 patients who had nonprotocol RT died within the first 2 years. Survival 5 years from the start of protocol RT for Regimen B was 53% ± 12% compared with 16% ± 9% for A and 24% ± 10% for C (P = 0.21). By composite pathology for the eligible patients, 5-year OS was 25% ± 8% for AA, 22% ± 7% for GBM, and 40% ± 22% for other eligible (P = 0.47) (Fig. 5). The 5-year EFS rate was 7% ± 5% for AA, 8% ± 4% for GBM, and 21% ± 18% for other eligible (P = 0.28) (Fig. 6). By institutional response for evaluable patients, the OS rate at 5 years was 57% ± 19% for PRs, 38% ± 17% for MRs, 21% ± 11% for SDs, and 20% ± 7% for PDs (P = 0.10 for homogeneity and P = 0.03 for trend) (Fig. 7).

Figure 5.

Overall survival by composite pathology.

Figure 6.

Event-free survival by composite pathology.

Figure 7.

Overall survival by institutional response.


Within the pediatric neurooncology community, there continues to be controversy concerning the efficacy and role of HDCT for treatment of primary HGA, particularly in patients with evidence of residual disease. In this prospective study of 3 different HDCT regimens in 76 pediatric patients with confirmed newly diagnosed HGA and postoperative residual disease, we conclude that, in comparison to conventional treatment regimens, HDCT does not significantly improve survival and does not provide additional clinical benefit, regardless of the amount of residual tumor present.

Finlay et al. first reported a 60% response rate with myeloablative alkylator-based chemotherapy in 10 pediatric patients with recurrent or primary HGA, suggesting that HDCT may be effective for previously untreated primary HGA.18 Subsequently, only a few small studies have been performed using HDCT in children with primary HGA. Although most of these studies have used similar alkylator-based regimens, the results have been mixed. Conclusions drawn from these studies are substantially limited by the relatively small number of patients investigated and, in most cases, a lack of a centralized data review. It is also unclear from these earlier studies whether the amount of postoperative residual disease has a significant impact on HDCT efficacy. Heideman et al. initially reported no improvement in the survival of 11 pediatric patients with newly diagnosed HGA who had bulky residual disease treated with high-dose thiotepa and cyclophosphamide followed by autologous bone marrow rescue.19 Kedar et al. also reported that survival of three patients with newly diagnosed HGA treated with a similar HDCT regimen was no better than that seen with conventional therapy.20 Likewise, Abrahams et al. concluded from a study of 15 patients that dose-intensified cyclophosphamide was inactive against HGA.11 More recently, however, two separate reports have suggested modest activity and acceptable toxicity of high-dose cyclophosphamide in newly diagnosed HGA in cohorts of 10 and 14 pediatric patients, respectively.13, 21 Finally, Grovas et al. reported the feasibility of HDCT in 11 patients treated with carmustine (BCNU), thiotepa, and etoposide for newly diagnosed HGA.22

In this study, more than half of our eligible patients relapsed during HDCT or did not complete HDCT because of toxicity. Nonhematologic Grade 3 or 4 toxicities were common, and only 68% of all eligible patients received protocol-directed RT, and 21% did not receive any RT. The most common reasons for patients not receiving RT were progressive disease or patient sickness (too ill). Of those who went on to RT, 19% had Grade 3 or 4 nonhematologic toxicity, with CNS the most common site. Compared with similar strategies in adults, the toxicities in this study were more severe and included more neurologic toxicity.23, 24 On the basis of the final analysis, it appears that HDCT before RT markedly affected the number of patients who received RT and their ability to complete treatment. The use of less HDCT cycles before RT, or less toxic approaches (i.e., stem cell support), may reduce HDCT treatment-related toxicity.

Although Regimen A and Regimen C achieved objective tumor response rates of 27% and 29%, respectively, there was no statistical difference among the three regimens, and response was not affected by amount of residual disease. Thus, no regimen was clearly superior to conventional treatment. Response rates for Regimens A and C compared favorably with the 11% objective response rate reported in children 36 months of age or older who were enrolled in CCG-945.25 The median time to an event was longest for Regimen A, and 5-year EFS for all patients was 8% ± 3% compared with 14% ± 7% for Regimen A, suggesting that of the three regimens carboplatin/VP-16 may have more activity in HGA. However, OS did not differ among treatment groups, regardless of histologic grade.

Notably, one fourth of 102 patients registered were later excluded from the analysis because of absence of postoperative residual disease > 1.5 cm2 or a diagnosis other than HGA. The relatively high exclusion rate underscores inherent problems in attempting accurate measurement of postoperative residual tumor by neuroimaging and histologic diagnosis of HGA. Because HGA frequently has hemorrhage, necrosis, and edema on pre- and postoperative imaging, assigning a clear demarcation of residual disease compared with inflammatory changes can present a great challenge to even the most experienced neuroradiologist. In addition to measurements of T1 or T2-weighted MRI signals, newer imaging techniques with fluid attenuated inversion recovery (FLAIR), magnetic resonance spectroscopy, and positron emission tomography (PET) may provide more robust assessment of residual disease status.26–28 In future studies, it may be important to incorporate neuroimaging guidelines that can be used for standardizing tumor measurement. Likewise, because of the heterogeneous appearance of HGA, expedited central neuropathology review and inclusion of specific genetic and biologic attributes, such as the MIB-1 labeling index, may be helpful for purposes of determining eligibility or stratification in future clinical trials.29, 30 Indeed, a similar conclusion was recently drawn from study CCG-945.31 These problems may partly explain the mixed results that have been previously reported for HDCT in treatment of HGA in which central histologic and radiologic review were lacking.

Finally, a slight, but significant, trend in survival was found in patients who responded to any of the regimens. Although the sample was too small to produce a definitive conclusion, within each arm there might have been patients with chemotherapy-responsive HGA who warranted additional investigation. Because each regimen contained an alkylating agent, it may be worthwhile to determine whether patients who had objective responses and prolonged EFS had tumors with relatively low levels of the DNA alkylator repair enzyme, O6-alkylquanine DNA-alkyltransferase (AGT). If this is the case, it may be better to use strategies that deplete AGT levels rather than intensify or modify the type of alkylator used.32–35 This also may allow for dose-reduction and limit toxicity.

In conclusion, we found no significant clinical advantage of HDCT regimens containing the most commonly used agents for the treatment of pediatric solid tumors over conventional therapeutic regimens in patients with newly diagnosed HGA, regardless of the amount of postoperative residual disease. Alternative strategies and/or newer cytotoxic or cytostatic agents specifically targeting HGA biology should be investigated for the treatment of primary HGA in children.