Fax: 49 941-2080-494
Preradiation chemotherapy for pediatric patients with high-grade glioma†
Article first published online: 28 DEC 2001
Copyright © 2002 American Cancer Society
Volume 94, Issue 1, pages 264–271, 1 January 2002
How to Cite
Wolff, J. E. A., Gnekow, A. K., Kortmann, R.-D., Pietsch, T., Urban, C., Graf, N. and Kühl, J. (2002), Preradiation chemotherapy for pediatric patients with high-grade glioma. Cancer, 94: 264–271. doi: 10.1002/cncr.10114
This is a report on behalf of the Brain Tumor Study Group of The Society for Pediatric Oncology and Hematology in the Germanic Speaking Countries (Joachim Kühl, chairman and coordinator for the current study).
- Issue published online: 28 DEC 2001
- Article first published online: 28 DEC 2001
- Manuscript Accepted: 24 AUG 2001
- Manuscript Received: 5 JUN 2001
- Deutsche Leukämie Forschungs Hilfe (DLFH)
- German Pediatric Cancer Foundation
- histologic grading;
- malignant glioma;
- sandwich chemotherapy;
- total resection
To evaluate the feasibility and efficacy of intensive chemotherapy given prior to irradiation in pediatric patients with malignant glioma, the Society of Pediatric Oncology in Germany started a randomized trial in 1991. The high-grade glioma strata had to be closed because of insufficient patient accrual. The follow-up data from these patients are reported.
Fifty-two patients with World Health Organization (WHO) Grade 4 malignant glioma (n = 27 patients) or with WHO Grade 3 anaplastic astrocytoma (n = 25 patients) between the ages of 3 years and 17 years were available for analysis. The tumor locations were supratentorial in 42 patients, the cerebellum in 8 patients, and the spinal cord in 2 patients (the brainstem was excluded). Tumor surgeries were biopsy in 10 patients, partial resection in 5 patients, subtotal resection in 10 patients, and macroscopic total resection in 21 patients. Patients received either 54 grays of irradiation (n = 22 patients) followed by chemotherapy with lomustine, vincristine, and cisplatin (maintenance chemotherapy) or sandwich chemotherapy (n = 30 patients), which consisted of ifosfamide, etoposide, methotrexate, cisplatin, and cytosine arabinoside followed by irradiation.
The extent of resection was the most important prognostic factor. The median survival was 5.2 years for patients who underwent tumor resection of ≥ 90% compared with 1.3 years for patients who underwent less than complete resection (P < 0.0005). After undergoing macroscopic total resection, sandwich chemotherapy (n = 15 patients) resulted in better overall survival (median, 5.2 years) compared with the maintenance protocol (n = 16 patients; median survival, 1.9 years; P = 0.015). A Cox multivariate regression analysis showed better survival for female patients (P = 0.025), WHO Grade 3 disease (P = 0.016), tumor resection of ≥ 90% (P = 0.003), irradiation with ≥ 54 grays (P = 0.003), and sandwich chemotherapy (P = 0.006).
These data suggest that early, intensive chemotherapy increases survival rates in patients with malignant glioma who undergo complete resection. Cancer 2002;94:264–71. © 2002 American Cancer Society.
Glioblastoma multiforme (ICDO:9440) and anaplastic astrocytoma (ICDO:9401) are defined by histologic features, including poorly differentiated polymorphic cells, vascular proliferation, and necrosis. The biologic behavior is classified as Grade 4 for glioblastoma multiforme and Grade 3 for anaplastic astrocytoma.1 Although these tumors are frequent in adults, the incidence in children is low (1–2 per 1,000,000 population).2, 3 It is believed that the extent of surgical resection4–6 and radiotherapy7–9 are of prognostic relevance. The impact of adjuvant chemotherapeutic treatment is controversial. The overall prognosis for patients with these tumors remains poor, with most 5-year overall survival rates reported below 10%.2
The Children's Cancer Study Group (CCG) found a significant increase in 5-year event free survival when patients with glioblastoma multiforme and anaplastic astrocytoma were treated with lomustine (CCNU), vincristine, and prednisone after undergoing radiotherapy.10 A subsequent study compared this regimen with the eight-in-one protocol, but the differences between the regimens were not significant.11 The Pediatric Oncology Group (POG) found a significant benefit for treatment with cisplatin and carmustine (BCNU) treatment over treatment with etoposide (VP-16) and cyclophosphamide.
In 1987, the Society for Pediatric Oncology and Hematology (GPOH) started a pilot trial with intensive chemotherapy given prior to irradiation (Protocol S).12, 13 After pilot data were encouraging, a randomized, Phase III study (Hirntumor 1991[HIT-91]) was started. This study was open for patients age > 3 years with malignant brain tumors located outside of the brainstem. It compared the intensive preradiation sandwich chemotherapy protocol with the vincristine, cisplatin, and CCNU protocol (Protocol M)14,15 given after irradiation. However, the accrual of patients with malignant glioma did not meet the expected numbers for a statistical power of 0.8, and a preliminary analysis did not show any difference between the two arms. Therefore, that stratum had to be closed. It was substituted by a prospective series of single-arm trials that started in 1995.16, 17 Here, we report a retrospective follow-up analysis of the patients with malignant glioma patients who were treated originally according to the Phase III study.
MATERIALS AND METHODS
Pediatric patients with histologically proven (by a local neuropathologist) high-grade glioma were enrolled in this multicenter Internal Review Board-approved study after informed consent was obtained. The treatment was randomized without stratification for surgical resection to either the sandwich chemotherapy arm or the maintenance chemotherapy arm, as described below. Central neuropathologic review was offered but was not mandatory. Response to treatment was documented in five categories: A complete response (CR) was defined as the disappearance of all tumor manifestations, a partial response (PR) was defined as shrinkage of residual tumor by > 50% of the size measured in two dimensions (greatest axial cross-sectional area), a minor response (MR) was defined as tumor shrinkage between 25% and 50%, stable disease (SD) was defined as a change in tumor size of < 25%, and progressive disease (PD) was defined as tumor progression > 25% of the pretreatment size. Survival was defined as the time between histologic diagnosis and either death or the end of observation. The definition of event free survival included death, tumor progression, metastases, and second malignancies as events.
The statistical analysis reported here was done retrospectively. Therefore, P values are descriptive and do not reflect the originally planned, hypothesis-based testing. Response to treatment was compared in patients with measurable disease after surgery using the chi-square test or the Fisher exact test for small numbers. Survival and event free survival were compared first using Kaplan–Meier curves and a log-rank test. Groups were defined first using one of the following parameters: year of diagnosis, gender, age at diagnosis, tumor location, World Health Organization (WHO) grade, extent of surgery, irradiation, and treatment arm. For quantitative parameters, such as age at diagnosis, various cut-off levels were used to define the groups. Subsequently, subgroup analysis was done using combinations of those parameters. Next, a Cox regression model was used, employing the parameters that showed differences in the univariate analysis or the subgroup analysis. When entering quantitative parameters, the cut-off levels were used that had discriminated two groups best in the previous analysis. All statistical analyses were done using the commercially available program package SPSS® (base 9.0; SPSS Inc. Chicago, IL).
The HIT sandwich chemotherapy protocol (Protocol S) was started in Week 3 postsurgery with the first block including ifosfamide (3 g/m2) as a continuous infusion over 24 hours on Days 1–3 and VP-16 (150 mg/m2) infused over 1 hour on Days 4–6. The next block was given twice in Weeks 5 and 6. It included methotrexate (5 g/m2) over 24 hours, with citrovorum factor (CF) rescue starting after 42 hours with 6 × 15 mg/m2. After hematologic recovery in Week 7 or 8, the fourth block was given, which included cisplatin (40 mg/m2 per day) as a 1-hour infusion on Days 1–3 combined with cytarabine (400 mg/m2 per day) as 30-minute infusion on Days 1–3. Tumor status was reevaluated 1 week later, and, for those patients with either no evidence of tumor or with a response to chemotherapy, the four blocks were repeated. At Week 17 after surgery, radiotherapy was started. For patients with high-grade glioma, a total of 54 grays (Gy) were given to the original tumor location with 2-cm margins.
The HIT maintenance chemotherapy protocol (Protocol M) started with radiotherapy. Fifty-four grays were given in 27 fractions of 2.0 Gy 5 days per week (standard fractionation). Vincristine (1.5 mg/m2) given intravenously once per week was added to the radiotherapy. After irradiation, maintenance chemotherapy was started, including cycles of oral CCNU (75 mg/m2) on Day 1; cisplatin (70 mg/m2) infused over 6 hours on Day 1; and vincristine (1.5 mg/m2) given intravenously on Days 1, 8, and 15. Eight of those cycles were given every 4 weeks (Day 1 to Day 1). Supportive care included dexamethasone and ondansetron for patients on both chemotherapeutic arms.
Data from 58 patients were available. Six patients had to be excluded from the analysis for the following reasons: location in the brainstem in one patient, age < 3 years in one patient, reference histology ependymoma in one patient, and lack of crucial data in three patients (type of surgery and postsurgical imaging data in one patient, follow-up data in one patient, and both in one patient 1). The analysis below refers to the remaining 52 patients. All patients and parents had signed informed consent for the study.
Thirty patients enrolled in Protocol S were available for analysis (Table 1). The male:female ratio was 14:16, and the median age at diagnosis was 8 years (range, 3–15 years) (Fig 1). Tumor locations were supratentorial in 21 patients, the cerebellum in 5 patients, and the spinal cord in 2 patients. Two patients had metastatic disease (one patient had a supratentorial tumor with a solid supratentorial metastasis, and one patient had an infratentorial tumor with diffuse meningeal infiltration). Tumor surgeries (Table 2) were classified by the neurosurgeons as stereotactic biopsy in six patients, open biopsy in one patient, minor resection (less than half of the tumor) in four patients, partial resection (50–90% of tumor mass removed) in four patients, subtotal resection (> 90% tumor tissue removed) in six patients, and macroscopic total resection in nine patients (Table 2). Postoperative imaging showed measurable disease in 19 patients. The histology was classified as WHO Grade 3 in 16 patients and Grade 4 in 14 patients. The histology was reviewed by the Reference Center allocated for this study (Neuropathology Department, University of Bonn) in 17 patients. With the exception of the previously excluded ependymoma patient, there was no disconcordance in the tumor grading or diagnosis.
|No. of patients||30||22|
|Median age in yrs at diagnosis (range)||8 (3–15)||10 (4–17)|
|WHO grade (III:IV)||16:14||11:11|
|Surgery||Treatment arm (no. of patients)|
|Macroscopic total resection||9||12|
There were 22 patients enrolled in Protocol M (Table 1). The male:female ratio was 6:16, with a median age of 10 years (range, 4–17 years). Tumor locations were supratentorial in 21 patients and cerebellar in 3 patients. The extent of resection (Table 2) was classified by the neurosurgeons as stereotactic biopsy in 1 patient, open biopsy in 2 patients, minor resection in 2 patients, partial resection in 1 patient, subtotal resection in 4 patients, and macroscopic total resection in 12 patients. Tumor grading was Grade 3 in 11 patients and Grade 4 in 11 patients. Histologic review was available for 10 patients, with no disconcordant results.
Response after 8 weeks of treatment for all 29 patients who were available for response analysis (both treatment arms) was CR in 2 patients, PR in 4 patients, MR in 6 patients, SD in 10 patients, and PD in 7 patients (Table 3). The overall survival of patients who responded was better than that of patients who did not respond. In the maintenance arm, radioimaging after surgery showed measurable disease in 10 patients. The response rates were CR in one patient, PR in two patients, MR in two patients, SD in five patients, and PD in 0 patients. Tumor status after the complete treatment was CR in two patients, PR in one patient, MR in one patient, SD in one patient, and PD in five patients. In the sandwich arm, 19 patients were available to evaluate response. After the first cycle (3 months), the response was CR in one patient, PR in two patients, MR in four patients, SD in five patients, and PD in seven patients. Treatment was discontinued in the patients with PD. Two of the patients with SD skipped the next cycle and began with irradiation. The three other patients continued with the second cycle. One of those patients had PD on this cycle, and treatment was discontinued. The two other patients remained in SD throughout subsequent cycles of chemotherapy and irradiation. The tumor size in two patients with MR did not change during the next cycle or subsequent irradiation. The tumors in the two other patients with MR and the in two patients with PR continued to shrink, resulting finally in a CR after irradiation. After the complete treatment, including irradiation, the tumor status for the total group was CR in five patients, PR in no patients, MR in two patients, SD in four patients, and PD in eight patients.
|No. of patients||19||10|
Comparing response rates showed greater numbers of patients with early PD in the sandwich arm compared with the maintenance arm (7 of 19 patients vs. 0 of 10 patients, respectively). This was of marginal statistical significance (Fisher exact one-sided test; P = 0.032). However, because patients in the maintenance arm more frequently developed PD later, the frequency of PD in patients in the sandwich arm was even less at the end of treatment compared with the maintenance arm (8 of 19 patients vs. 5 of 10 patients, respectively; P = 0.68; not significant). There was no significant difference in tumor status between patients in the two treatment arms.
The prognosis was correlated with the extent of resection in all of our analyses. For instance, when using all categories documented by the neurosurgeons, the null hypothesis that the survival distribution was equal among the categories was rejected with a P value of 0.0001 (log-rank test). The best cut-off value for creating only two groups was found by grouping patients who underwent macroscopic totally and subtotally resected tumors together and comparing them with patients who underwent partial resections and biopsies. In this comparison, the respective median survivals were 1.25 years ± 0.25 years versus 5.17 years ± 2.2 years (P < 0.0005). Comparable results were found when event free survival was used as the end point (0.72 years ± 0.18 years vs. 2.15 years ± 0.35 years; P < 0.0005). The phenomenon also was confirmed in a Cox regression analysis using both overall survival (P = 0.003; Table 4) and event free survival (risk ratio [RR], 4.12; 95% confidence interval [95%CI], 1.70–9.96; P = 0.0017).
|No. of patients||19||10|
|Factor||No. of patients||RR||95% CI||P value|
|Male gender||20 of 52||3.156||1.54–8.627||0.025|
|Age ≤ 9 years||27 of 52||0.521||0.194–1.399||0.196|
|WHO Grade 3||27 of 52||0.228||0.069–0.755||0.016|
|Resection < 90%||21 of 52||5.272||1.762–15.775||0.003|
|Residual tumor > 2 cm||5 of 44||4.95||1.076–22.745||0.040|
|Irradiation ≥ 54 Gy||35 of 52||0.127||0.033–0.487||0.003|
|Sandwich treatment arm||30 of 52||0.215||0.072–0.6445||0.006|
Postoperative imaging was documented in addition to the impression of the neurosurgeon. Of the 52 patients, 15 patients had no radiologically detectable postoperative tumor, 29 patients had residual tumor, and documentation was missing or inconclusive for 8 patients. Aside from judging the presence and/or absence of tumor, the neuroradiologists had been asked to document two perpendicular tumor dimensions in cm (to the closest complete cm). In 14 patients, these two measures were documented at > 0 cm. In those patients, the average ± standard deviation greatest tumor dimension was 2.64 cm ± 1.86 cm (range, 1–7 cm), and the smaller dimension was 2.29 cm ± 1.64 cm (range, 1–6 cm). The distribution of the tumor sizes among the two treatment groups was unequal. Patients in the sandwich arm had more of the larger tumors compared with patients in the maintenance arm (average product of greatest dimensions: 11.0 cm2 ± 14.2 cm2 vs. 2.5 cm2 ± 1.0 cm2, respectively; P = 0.033; t test). Despite the small number of patients available for analysis, the radiologic measures were related significantly to survival. The best cut-off value was 2 cm (single dimension > 2 cm compared with ≤ 2 cm; P = 0.0049). This parameter was entered in the Cox regression model, and the relevance was confirmed for overall survival (P = 0.04, Table 4) but not for event free survival (RR, 0.83; 95%CI, 0.22–3.1; P = 0.78). Compared with the surgical opinion, the one-dimensional radiologic measure added further relevant information for predicting overall survival.
Comparing the Chemotherapy Treatment Arms
The patient populations entering the two treatment arms were unequal. There were more patients with macroscopic totally resected tumors entering the maintenance treatment arm than the sandwich treatment arm (12 of 22 patients [55%] vs. 9 of 30 patients [30%], respectively). In addition, patients in the sandwich arm had larger tumors. All five patients with tumors measuring > 2 cm had been treated on the sandwich arm, and two patients with metastatic disease also were enrolled in the sandwich arm. There were no notable differences with respect to gender, age at diagnosis, year of diagnosis, tumor location, or tumor grade.
There was no difference in event free survival or overall survival when patients in the two treatment arms were compared in an univariate analysis. However, this comparison was of limited value, because the prognostic importance of resection was distributed unequally. When the analysis was repeated in subgroups, the picture changed. Because the best way to group patients was by surgical classification, separating patients who underwent complete and subtotal resection and patients who underwent partial resection and biopsy, this was used for the further subanalysis. The 15 patients who were treated on the sandwich arm after undergoing complete or subtotal resection had a significantly better overall survival compared with the 16 patients with the same surgical results who were treated on the maintenance arm (P = 0.0154) (Fig. 2). The median overall survival was 5.17 years ± 2.18 years for patients on the sandwich arm compared with only 1.94 years ± 0.36 years for patients on the maintenance arm. The Cox regression model, which included all of these parameters in the total group of patients, confirmed the superior survival after Protocol S. This was only a trend for event free survival (RR, 0.597; 95%CI, 0.27–1.29; P = 0.194) but was significant for overall survival (P = 0.06) (Table 4, Fig. 3).
Further parameters with prognostic relevance included female patients, who had a tendency toward better survival in the univariate analysis (median overall survival, 1.49 years ± 0.50 vs. > 6 years; P = 0.06; log-rank test) and in the Cox regression analysis (P = 0.025; Table 4). Younger patients had a better prognosis. The best cut-off age was ≤ 9 years compared with age > 9 years (median overall survival ± standard error: 5.17 years ± 1.47 years vs. 1.49 years ± 0.15; P = 0.0044; log-rank test). Histologic grading (WHO Grade 3 vs. Grade 4) did not make a difference in univariate analysis of the total group. However, in the subgroup analysis, a tendency for better survival was noted in patients with Grade 3 tumors after undergoing complete resection compared with patients who had Grade 4 tumors (P = 0.078) (Fig. 4). In the Cox multiple regression analysis, grading was significant when using overall survival (P = 0.016) (Table 4) as an end point. When event free survival was used as the end point, none of the three parameters was significant. The irradiation dose was documented accurately in 47 patients. Among those 47 patients, the dose was 0 Gy in 7 patients, 0–54 Gy in 5 patients, exactly 54 Gy in 27 patients, and > 54 Gy in 8 patients. The lower doses reflect treatment discontinuations when the disease was progressive. Irradiation with 54 Gy was related to significantly better survival compared with survival when there was no irradiation or irradiation with reduced dose was used. This remained true when those patients with irradiation doses of 0 Gy were excluded from the analysis, regardless of whether overall survival (P = 0.029; log rank test) or event free survival (P = 0.011) was chosen as the end point. In contrast, tumor location and year of diagnosis had no prognostic relevance in our data.
The data from this study confirmed the important role of resection in the multimodal treatment of pediatric patients with high-grade glioma. When a biopsy shows high-grade glioma, radical surgical resection frequently is avoided, because the tumor is viewed pessimistically as not curable, and, in these patients, surgical morbidity appears less acceptable than in patients with tumors that may be cured by surgery alone. The data presented here do not support this approach. The extent of resection was the most important prognostic factor found here. This confirms the results of the largest North American study in patients with childhood malignant glioma.18, 6 In contrast, in the adult population, the difference that surgery makes appears to be either less striking19, 20 or not significant at all.21 This may reflect biologically different diseases.22 Based on the pediatric data, a fatalistic approach avoiding maximal possible resection is not justified.
Patients on the two treatment protocols used in this study had different outcomes. After correcting for the unequal distribution of patients with residual tumors, the sandwich protocol was superior to the maintenance protocol. This was true in both the univariate subgroup analysis and the Cox regression model. In the subgroup analysis, patients with tumors resected by > 90% had significantly better overall survival when they received the sandwich protocol. In the Cox model, the relative risk for death after randomization to the sandwich arm was 0.215, indicating a four-fold better chance of survival when patients were treated on this protocol. It is hard to pinpoint exactly which of the differing elements among the two protocols may have caused the survival difference. Both protocols included cisplatin, vincristine, and irradiation. Drugs in the successful sandwich protocol that were missing in the inferior maintenance protocol included ifosfamide, VP-16, high-dose methotrexate, and cytosine arabinoside. Narrowing this list down, the GPOH group presently includes ifosfamide, cisplatin, and VP-16 but not methotrexate or cytosine arabinoside in their current study (HIT-GBM-C). The outcome of that study will be helpful in determining whether the right choices were made. The results from the two protocol arms in patients with high-grade glioma came were opposite to the results obtained in patients with medulloblastoma who were treated on the same study, which will be reported separately. The better result for patients with medulloblastoma on the maintenance arm may reflect a larger impact of irradiation in patients with this primitive embryonal tumor, which may have translated into a larger disadvantage when irradiation was delayed.
Phase II Window Studies for Pediatric Patients with High-Grade Glioma
Drug specific information remains rare, because most glioma studies include multiple-agent treatment. To make matters more complicated, irradiation may induce chemotherapy resistance,23, 24 and this may suggest incorrectly a failure of chemotherapeutic protocols. Therefore, the window studies have been designed to give the chemotherapy in a time window prior to irradiation. These studies frequently use tumor response as the primary end point and enroll only patients with residual tumor postsurgery into single-arm protocols. The data presented here allow critical analysis of this concept. For the window (sandwich) chemotherapy used here, two relevant observations were made: 1) the incidence of early progressive disease was higher, and 2) the survival was better in the final analysis for patients who underwent macroscopic total or subtotal resection. These data suggest that there is enough time for a window study in patients who underwent complete resection. After macroscopic total resection, response cannot be analyzed, which means that deviating from a classical Phase II study concept is necessary, using survival time as an end point in such a concept. For pediatric patients who undergo less than 90% resection, however, our data suggest that irradiation should not be postponed. This is different from the approach in adult patients, in whom window studies evaluating response do not appear to have a negative impact on survival.25 However, to start exclusively with either irradiation or chemotherapy is not the only possible choice. Instead, concurrent radiochemotherapy26 can combine the benefits of the two other approaches. Both, the German and the North American cooperative pediatric groups chose to give chemotherapy simultaneously with irradiation in the recent protocols. The data presented here confirm the validity this approach.
In summary, the extent of tumor resection is an important factor for survival in children with malignant glioma. With respect to chemotherapy, three different cooperative groups confirmed that certain protocols can improve survival. Successful regimens included prednisone, CCNU, and vincristine (CCG); cisplatin and BCNU (POG); and ifosfamide, VP-16, methotrexate, cisplatin, and cytosine arabinoside (GPOH). In addition, this study adds evidence that chemotherapy may be given safely prior to irradiation when macroscopic total resection can be achieved.
- 1Histological typing of tumor of the central nervous system. In: ScheithauerBW, editor. World Health Organization international histology classification of tumors, 2nd ed. Berlin: Springer, 1993., , .
- 13Postoperative neoadjuvant chemotherapy before radiotherapy as compared to immediate radiotherapy followed by maintenance chemotherapy in the treatment of medulloblastoma in childhood: results of the German prospective randomized trial HIT '91. Int J Radiat Oncol Biol Phys 2000; 15: 269–79., , , , , , et al.
- 14Chemotherapy for childhood medulloblastoma and primitive neuroectodermal tumors. Oncologist 1996; 1: 381–93., .
- 19[The effect of extent of tumor resection on the outcome of combined therapy in patients with glioblastoma multiforme]. Srp Arh Celok Lek 1997; 125: 93–8., , , , , , et al.