Journal of Neuro-Oncology47: 145–152, 2000.© 2000Kluwer Academic Publishers. Printed in the Netherlands.

Clinical Study

A phase II study of preradiation chemotherapy followed byexternal beam radiotherapy for the treatment of patients withnewly diagnosed glioblastoma multiforme: an Eastern CooperativeOncology Group study (E2393)

Mark Gilbert1, Anne O’Neill2, Stuart Grossman3, Margaret Grunnet4, Minesh Mehta5, Steven Jubelirer6 andRichard Hellman71Emory University, Atlanta, GA;2Dana Farber Cancer Institute, Boston, MA;3Johns Hopkins Oncology Center,Baltimore, MD;4University of Connecticut, Farmingham, CT;5University of Wisconsin, Madison, WI;6Charleston Area Medical Center, Charleston, WV;7Lawrence and Memorial Hospital, New London, CT, USA

Key words:glioblastoma multiforme, chemotherapy, radiotherapy, phase II trial

Summary

Recent publications support the use of preradiation chemotherapy in the treatment of selected primary brain tumors.In the pediatric population, this treatment strategy often delays radiotherapy and may improve the outcome inpatients. This manuscript describes the use of a preradiation chemotherapy approach for adult patients with newlydiagnosed glioblastoma multiforme. The main objective of this trial was to determine the feasibility of deliveringup to 3 monthly cycles of a 72 h continuous simultaneous intravenous infusion of BCNU (40 mg/m2/day) andcisplatin (40 mg/m2/day). Patients were evaluated for tumor response or progression after each cycle. Following thecompletion of the chemotherapy treatments or evidence of tumor progression, patients underwent external beamradiotherapy. A dose of 45 Gy was delivered to the pretreatment tumor volume plus surrounding edema and amargin of 3.0 cm. An additional 14.4 Gy was delivered to the preoperative volume plus a 2 cm margin. Fifty patientswere enrolled, 47 were eligible and analyzable. Overall, 79% of patients were able to complete at least 2 cycles oftreatment, exceeding the predefined measure of feasibility. One patient achieved a complete response, 10 patientsa partial response and 18 patients had stable disease at completion of the chemotherapy treatments. Twenty-fourpatients experienced grade 4 toxicity, mostly hematologic. All patients were able to undergo radiotherapy followingchemotherapy. These results indicate that a preradiation strategy is feasible. Although responses to the chemotherapywere seen, a phase III trial is needed to determine whether this approach provides an advantage over standardtreatment; such a phase III trial has been undertaken by ECOG.

Patients with malignant glioma have a poor prog-nosis. Adults with glioblastoma multiforme have amedian survival of 8–10 months from the time ofdiagnosis when treated with standard radiation therapyand chemotherapy [1–3]. Although glial malignanciesrarely metastasize outside the central nervous system,surgical resection is rarely curative because of perme-ation of tumor into the surrounding brain parenchymaand distant migration of tumor cells within thebrain [4–6].

Standard therapy for malignant glioma consists ofmaximal tumor resection followed by external beamradiotherapy. Most studies support the use of lim-ited field radiation to the tumor region and surround-ing edema with an additional 2–3 cm margin [7–9].Although limited-field radiotherapy will often not treatdistant microscopic disease, the use of whole brainradiotherapy does not offer an improvement in sur-vival and can cause significant neurologic toxicity.Radiation therapy improves survival when compared

146

with surgical resection and supportive care alone, butbecause radiotherapy is unable to completely eradicateall tumor cells, recurrent disease is a principal problem.

Chemotherapy, most commonly nitrosoureas (intra-venous BCNU or oral CCNU), is frequently adminis-tered to patients with brain tumors, and is usually giveneither concomitantly with radiotherapy or as adju-vant treatment after the completion of radiotherapy.Chemotherapy in this setting is of uncertain bene-fit in prolonging survival for patients with glioblas-toma multiforme. Most studies show either minimal orno survival benefit, although a metaanalysis supportsthe use of chemotherapy with external beam radiation[2,3,10]. Chemotherapy efficacy is often limited byproblems of drug delivery because of the blood–brain barrier. Attempts to overcome this problem, suchas with direct intraarterial infusion and blood–brainbarrier disruption, have had limited success but havehad significant toxicity [11].

Recent publications, primarily examining the roleof chemotherapy in the treatment of childhood braintumors support the use of up-front or neoadjuvantchemotherapy. Most children with low grade gliomain these studies had tumor growth controlled for 1–2years with chemotherapy alone, permitting the delayof radiotherapy and its severe neurologic toxicity onthe developing nervous system [12,13]. Two phase IItrials, from the Johns Hopkins Oncology Center andthe University of Pittsburgh, utilizing preradiationneoadjuvant chemotherapy in adults with malignantglioma also showed promising results [14,15]. On thebasis of these observations, we developed a treatmentprotocol utilizing the same treatment strategy to testthe feasibility of preradiation chemotherapy in a coop-erative group setting.

Patients and methods

Adults (age≥ 18) who were newly diagnosed withglioblastoma multiforme were eligible for this trial.The pathologic diagnosis was confirmed by centralreview. Other eligibility criteria included normal bonemarrow function (WBC> 4000/mm3), normal renalfunction (serum creatinine< 1.6 mg/dl), hepatic func-tion (total bilirubin< 2.0 mg/dl) and an ECOG per-formance status of 0, 1, or 2. Patients with pulmonarysymptoms or a history of respiratory difficulty wererequired to have a carbon monoxide diffusing capacity(DLCO) ≥ 50% of predicted. Patients were required tobe on a stable or decreasing dose of steroids for the3 days preceding the initiation of treatment. Patients

who underwent stereotactic biopsy were eligible tobegin treatment after at least 10 days after the proce-dure, patients who underwent craniotomy and resectionwaited a minimum of 14 days. The diagnosis neededto have occurred within 6 weeks of entry onto thetrial. Patients were excluded if they had received anyprior chemotherapy for other malignancy or if they hadreceived any treatment except surgical resection of theirmalignant glioma.

Treatment design

Study objectivesThe primary objective of this study was to estimate theproportion of patients able to complete 2–3 cycles ofa continuous infusion regimen prior to radiotherapy.The study defined the treatment as feasible if greaterthan 70% of patients enrolled were able to complete2–3 cycles of treatment. In the Pittsburgh experience,92% of patients completed 3 cycles of chemother-apy and at Johns Hopkins, 77% completed 3 cyclesof chemotherapy. The original design was based onaccruing 40 eligible patients with a 10% allowance forineligibility and cancellation. If 34 of the 40 patientscompleted at least 2 cycles, the 95% one-sided confi-dence interval would suggest that at least 73% of thepatients would receive at least 2 cycles of chemotherapyin a future trial, indicating the feasibility of the treat-ment regimen. Secondary objectives included mea-surement of response rates and toxicities experiencedduring the chemotherapy regimen, as well as time todisease progression and overall survival.

Patients received 3 monthly cycles of BCNU(40 mg/m2/day× 3 days) and cisplatin (40 mg/m2/day× 3 days) by continuous intravenous infusion. TheBCNU was administered as a series of 12 bot-tles of chemotherapy; each infused over 6 h, whilethe cisplatin was administered as six 12 h infusions.Complete blood counts, electrolytes, BUN, serum cre-atinine, magnesium, calcium and appropriate anticon-vulsant levels were measured twice weekly during thechemotherapy. One month following the first treatment,before the second cycle of chemotherapy, neurologicexaminations and head CT or MRI were used to eval-uate patients. Patients with a stable or improving neu-rologic examination and a stable or decrease in tumorsize by imaging were eligible for a second cycle ofchemotherapy. Patients with tumor progression werereferred for radiotherapy. In addition to tumor response,criteria for retreatment included adequate bone marrowfunction (WBC> 4000/cc, platelets> 100,000/cc),

147

renal function (< 1.6 mg/ml), hepatic function (biliru-bin < 2.0 mg/dl), and pulmonary function (DLCO ≥50% of predicted). Myelosuppression resulted in adelay in the start of treatment, but there were no reduc-tions in subsequent doses of the chemotherapy.

Five weeks after completion of chemotherapy,patients began radiotherapy. The initial target volumedefined at the time of simulation and treatment planningincluded the preoperative volume of enhancement onCT, plus surrounding edema, with a margin of 3.0 cmtreated to a dose of 45 Gy in 25 fractions. Followingthis, an additional 14.4 Gy was delivered to the pre-operative volume of enhancement with a 2 cm mar-gin. Therefore, the total dose administered to the tumorbed was 59.4 Gy. Patients were then evaluated withneurologic examination and head MRI or CT scans at3 month intervals.

Response criteriaComplete response was defined as absence of abnormalenhancement for greater than 1 month with the patientoff corticosteroids. Partial response was defined as agreater than 50% decrease in maximum cross-sectionalarea on a stable dose of corticosteroids for greater than1 month. Progressive disease was defined as a greaterthan 25% increase in tumor area. Patients who requiredan increase in corticosteroid dose were considered non-responders. Patients without measurable disease at thetime of entry onto the study were considered to havestable disease if no tumor growth was noted duringor after completion of the chemotherapy component.Time-to-progression was defined as time from registra-tion to tumor progression. Overall survival was definedas time from registration to death. Patients with recur-rence of tumor growth after the completion of treatmentprotocol were eligible for additional therapy at the dis-cretion of the treating physician.

Statistical methods

This phase II study had a single stage of accrual.Confidence intervals for response rates and the num-ber of cycles completed were estimated using meth-ods for exact binomial confidence intervals [16]. TheKaplan–Meier method was used to estimate progres-sion and survival distributions, and confidence intervalswere calculated using Greenwood’s formula [17].

An analysis was also performed to evaluate the asso-ciation between various covariates and response, time-to-progression, and overall survival. Fisher’s exact testwas used to determine if there was an association

between covariates and response [16]. Logistic regres-sion was used to adjust for covariates with respect toresponse [18]. Cox proportional hazards models wereused to adjust for covariates with respect to time-to-progression and survival [19]. The following covariateswere considered: age (≥ 50 vs.< 50), gender (femalevs. male), prior surgery (partial excision or gross totalresection vs. biopsy), ECOG performance status (1, 2vs. 0), and number of chemotherapy cycles received(1, 2 vs. 3). Log-rank test was used to test for differ-ences in time-to-progression and overall survival dis-tributions with respect to these covariates [20].

Results

Between April 1994 and January 1995, 50 patientswere enrolled on this trial in the Eastern CooperativeOncology Group. Eighteen institutions participated.One patient was canceled due to a duplicate registra-tion. Two patients were ineligible because of a CT scandone> 14 days before registration and entry on study< 10 days from biopsy. Therefore, 47 of 50 patientswere considered eligible and analyzable. Table 1 sum-marizes patient characteristics. All patients had histo-logic confirmation of glioblastoma by central review.

Chemotherapy treatment results

Treatment cycles completedForty-seven patients were analyzable for response.Ten completed 1 cycle of chemotherapy, 8 completed2 cycles and 29 patients completed all 3 cycles. There-fore, 37 patients (79%) were able to complete 2 or

Table 1. Patient characteristics (n = 47)

AgeMedian 57 yearsRange 34–74 years

SexMale 27Female 20

Prior surgeryBiopsy only 14Partial resection 22Gross total resection 10Shunt and resection 1

ECOG performance statusFully active (PS= 0) 13Ambulatory (PS= 1) 25In bed< 50% (PS= 2) 9

PS: Performance status.

148

3 cycles of chemotherapy. (The one-sided lower 95%confidence limit was 66%.) Progressive disease wasthe most common reason for termination after 1 cycleof treatment (7 of 10 patients, 70%). One patientencountered an early death, unrelated to treatment. Twopatients developed excessive toxicity. Among patientscompleting 2 of 3 planned chemotherapy treatments,2 of 8 developed progressive disease, 4 patients experi-enced excessive toxicity and one suffered an early deaththat was possibly related to treatment. The remainingpatient did not have a reason reported for treatment ces-sation. Three patients were found to have progressivedisease after completing the 3 planned chemotherapytreatments, determined by the scan performed beforethe initiation of radiotherapy.

Treatment responseOne patient achieved a complete response, 10 patientsa partial response, and 18 had stable disease at the com-pletion of the chemotherapy component of treatment.Twelve patients had progressive disease. The remain-ing 6 patients were unevaluable for response eitherbecause of an increase in steroid dose during treat-ment or early death. Therefore, the overall responserate (complete response+ partial response) duringthe chemotherapy portion of this protocol was 23%(11/47), with a 95% confidence interval of 12–38%.The clinical characteristics of the responders are pro-vided in Table 2.

Chemotherapy toxicityThe incidences of severe and life-threatening toxic-ities are listed in Table 3. Two patients died while

Table 2. Summary of responders

Response Age Gender ECOG TTP OverallPS (mons.) survival

CR 65 Female 2 17.8 25.1PR 62 Female 1 2.8 7.5PR 43 Male 2 11.9 15.4PR 43 Male 0 12.4+ 12.4+PR 39 Male 0 47+ 47+PR 70 Male 1 21.7 23.8PR 43 Male 2 18.2 37.4PR 48 Male 0 6.9 19.8PR 59 Male 1 5.2 10.9PR 55 Male 1 16.8 23.7PR 47 Female 2 6.6 10.5

ECOG PS: ECOG performance status; TTP: time-to-progression; CR: complete response; PR: partial response.

Table 3. Toxicities during chemotherapy

Toxicity Severe Life-threatening(grade 3) (grade 4)

Leukopenia 13 5Granulocytopenia 15 16Thrombocytopenia 11 17Anemia 18 3Infection 6 1Nausea 5 —Vomiting 2 —Liver 2 —Cardiac — 1Pulmonary — 2Phlebitis (DVT or PE) 2 1Neurologic worsening 5 4Metabolic 7 —Coagulation 2 —

undergoing treatment and their deaths were thoughtto be possibly treatment-related. Twenty-four patientsexperienced life-threatening toxicities (grade 4);mostly thrombocytopenia and granulocytopenia. Six-teen patients experienced, at worst, severe toxicities(grade 3), with anemia, granulocytopenia, thrombocy-topenia and leukopenia most common.

Radiotherapy treatment results

Treatment responseThe patient who achieved a complete response dur-ing the chemotherapy phase maintained this responsebeyond 2 years. Four patients who achieved a partialresponse had evidence of tumor progression at the com-pletion of radiation. Of the 18 patients with stable dis-ease following chemotherapy, one achieved a partialresponse with radiation, 3 patients maintained stabledisease, 12 patients had progressive disease within4 months of completing radiotherapy, 1 patient wasunevaluable and 1 had and unknown response. Three of12 patients with progressive disease during chemother-apy had disease stabilization during radiotherapy. Fourpatients had further tumor progression. One patient wasunevaluable because of a major variation in the radio-therapy regimen. For the remaining 4 patients, responseto radiotherapy is not known. One of the 6 patients whowas not evaluable for response to the chemotherapycomponent achieved a partial response with radiother-apy. One patient had disease stabilization and 1 patienthad tumor progression. Response of the remaining 3 of6 patients is unknown.

149

Progression within 2 months of completion of radi-ation was seen in 3 of 10 patients achieving a par-tial response during the chemotherapy regimen andearly progression was seen in 10 of 18 patients withstable disease during the chemotherapy. Of the 12patients with tumor progression during chemotherapy,one died before receiving radiotherapy, 4 showed con-tinued tumor progression and 4 were unevaluable forresponse either due to major variation in treatment ortreatment at a non-ECOG institution.

Radiation treatment toxicityToxicity data is available on 39 patients. Two patientshad life-threatening toxicities; neurologic deteriorationand dehydration. Five patients had severe toxicitieswith leukopenia, anemia and neurologic worsening.

Time-to-progression and survivalMedian time to progression was 5.2 months and rangedfrom 2 days to 47+ months. Median survival was9.3 months, ranging from 2 days to 47+ months (seeFigure 1). The proportion of patients surviving attwelve months by Kaplan-Meier estimate is 38% witha 95% confidence interval of 24-52%. Two patientsremain alive at the time of this report, 12+ and 47+months after study entry.

Covariate analysesA covariate analysis was performed on the 47 analyz-able patients using the statistical methods describedabove. For the response, time-to-progression, and sur-vival analyses, all 5 variables (age, gender, prior

Figure 1. Kaplan–Meier survival curve depicting the fraction ofpatients surviving from the time of diagnosis.

Table 4. Covariate analysis for response to chemotherapy

Covariate Odds 0.95 CI P -valueratio

Age (≥ 50 vs.< 50) 0.12 0.02,0.87 0.04Gender (F vs. M) 0.39 0.07,2.27 0.30Surgery (resection vs. biopsy) 1.39 0.21,9.14 0.72ECOG PS (1, 2 vs. 0) 4.54 0.47,43.4 0.18Chemotherapy cycles (1, 2 vs. 3) 0.08 0.01,0.91 0.04

ECOG PS: ECOG performance status.

Table 5. Covariate analysis for time-to-progression

Covariate Hazard 0.95 CI P -valueratio

Age (≥ 50 vs.< 50) 1.7 0.68,4.09 0.25Gender (F vs. M) 1.3 0.69,2.63 0.37Surgery (resection vs. biopsy) 0.71 0.32,1.54 0.38ECOG PS (1, 2 vs. 0) 0.77 0.33,1.76 0.53Chemotherapy cycles (1, 2 vs. 3) 4.0 1.76,8.84< 0.01

ECOG PS: ECOG performance status.

Table 6. Covariate analysis for survival

Covariate Hazard 0.95 CI P -valueratio

Age (≥ 50 vs.< 50) 1.1 0.47,2.58 0.80Gender (F vs. M) 1.5 0.79,2.84 0.21Surgery (resection vs. biopsy) 0.60 0.29,1.23 0.16ECOG PS (1, 2 vs. 0) 1.0 0.48,2.14 0.95Chemotherapy cycles (1, 2 vs. 3) 5.1 2.28,11.45< 0.01

ECOG PS: ECOG performance status.

surgery, performance status and number of chemother-apy cycles) were considered together. The results forthe covariate analyses determined that for response,time-to-progression and overall survival, only the num-ber of chemotherapy cycles administered was statisti-cally significant (Tables 4–6, respectively).

Discussion

Malignant gliomas remain a therapeutic challenge.Although these tumors rarely metastasize outsidethe central nervous system, local therapies are oftenof limited benefit. Microscopic infiltration of brainparenchyma, often extending from one cerebral hemi-sphere to the other via the corpus callosum limits thebenefit of surgical resection and other regional treat-ments such as brachytherapy and radiosurgery [4–6].Recently, limited field external beam radiotherapy hasbeen determined to be as effective as whole brain

150

treatment and the limited field treatment has a lowerincidence of neurotoxicity [7,21]. Local recurrence oftumor is the most common site of tumor regrowth, butthis may reflect the limited ability of radiotherapy toprovide long-term tumor control for malignant glioma.

Chemotherapy is often administered concurrentlywith external beam radiotherapy and as an adjuvanttreatment. Large prospective series and metaanalysiscomparing adjuvant chemotherapy following radio-therapy with radiation treatment alone generally showonly a marginal increase in survival [1,3,10]. The rea-sons for the poor response to adjuvant chemotherapyare many. Possible explanations include poor deliveryof drug to the tumor because of radiation-induced dam-age to the tumor vascular supply and a decrease inthe cytotoxic effects of chemotherapy on dormant orinjured tumor cells in the immediate post-radiotherapyperiod [22,23]. Conversely, radiation treatment mayfurther impair the blood–brain barrier in the peritu-moral region, improving drug penetration and thereis some data to suggest that in some cases there israpid tumor repopulation, making these cells poten-tially maximally chemotherapy-sensitive.

Neoadjuvant (preradiation) chemotherapy has beenwidely used in the treatment of pediatric brain tumors.The primary goals for these treatment regimens are tocontrol tumors while delaying the potentially severeneurotoxic effects of radiotherapy on the developingcentral nervous system. These trials have shown thatneoadjuvant chemotherapy has a very high responserate and can often delay the institution of radiotherapyfor several years.

Neoadjuvant chemotherapy for malignant gliomasin adults has undergone only limited evaluation in thepast. An early trial using VM-26 and CCNU as neoad-juvant treatment showed a poor response with manypatients showing disease progression while receivingthe chemotherapy [24]. Grossman et al. treated poorprognosis patients with malignant gliomas initiallywith chemotherapy [25]. A continuous infusion regi-men of BCNU and cisplatin was administered over 3days on a monthly schedule. A maximum of 8 cycleswas given, although only a few patients received theentire treatment regimen. Most patients stopped treat-ment either because of tumor progression or treatmenttoxicity. Tumor response was determined after eachchemotherapy cycle, demonstrating maximal responseafter 3 or 4 treatment cycles, before most serioustoxicity developed. Grossman et al. then began a trialof neoadjuvant chemotherapy followed by external

beam radiotherapy [14]. Adults with newly diagnosedglioblastoma multiforme or anaplastic astrocytomawere treated with 3 cycles of the continuous infusionBCNU and cisplatin regimen, followed by externalbeam radiotherapy. The development of venous throm-boembolic disease was the most common reason forpatients not completing the full course of chemother-apy. In this study, 59% of patients demonstrated anobjective response (> 50% reduction in tumor volume)to the chemotherapy component of treatment.

The current study tests the treatment protocol ofGrossman et al. for patients with newly diagnosedglioblastoma multiforme in a cooperative group set-ting. The primary goal was to determine the feasibilityof this treatment approach in a wide variety of insti-tutions. The results, with 79% of patients receiving2–3 cycles of chemotherapy, suggest that this treat-ment regimen can be successfully administered ina wide variety of institutions. The overall responserate to the chemotherapy component was modest, butthere were durable responses and 1 patient achieveda complete radiographic response. Additional patientsdemonstrated a response after the completion of theradiotherapy. The lower response rate in this trial com-pared with the one reported by Grossman may reflectdifferences in the criteria for response; Grossman uti-lized computer generated tumor volumes whereas thecurrent study used cross-sectional tumor area. Further-more, the individual investigator at each participatinginstitution made response determination. Estimation ofresponse was a secondary endpoint of the current trial,hence the finding of stable disease or response weresufficient to continue treatment to evaluate feasibility,the primary endpoint.

Covariate analysis evaluating potential factorsthat predict improvement in response, time-to-tumorprogression and overall survival indicate that onlythe number of chemotherapy cycles administered wassignificantly associated with an improved outcome.Although these data are suggestive of a positive rolefor the chemotherapy regimen, alternative explana-tions need to be considered. It is possible that theability to administer a full course of 3 cycles ofchemotherapy indicates that the biologic behavior ofthe tumor is more favorable. Therefore, the prognosisfor these patients would be better, regardless of whetherchemotherapy was a component of their treatment. Thelack of statistical significance of surgical procedure(biopsy vs. resection) and age may be a consequenceof the small sample size in the study.

151

The results of this trial indicate that it is feasi-ble to administer an intensive chemotherapy regimento patients with newly diagnosed glioblastoma. Thesefindings have led to further testing of neoadjuvantchemotherapy in the treatment of malignant gliomas.As a phase II trial, however, conclusions regardingthe overall benefit to patients when compared to stan-dard treatment regimens can be questioned. Althoughthe majority completed all 3 cycles of treatment, mostpatients did display some treatment toxicity. Althoughthe severe toxicities were predominantly hematologicand reversible, these effects are likely to impact onpatient’s sense of well being and quality of life. There-fore, a phase III trial is being performed to comparethis regimen with standard therapy not only by out-come but also compare risk/benefit and the toxicity andmorbidity associated with each treatment regimen. Fur-thermore, the questions raised by the covariate analysisregarding the impact of the chemotherapy on outcomecan be also be addressed in the phase III trial.

Acknowledgements

This study was conducted by the Eastern Coopera-tive Oncology Group (Robert L. Comis, MD, Chair)and supported in part by Public Health Service grantsCA23318, CA73590, CA21076, CA13650, CA66636,CA21115 from the National Cancer Institute, NationalInstitutes of Health and the Department of Health andHuman Services. Its contents are solely the responsi-bility of the authors and do not necessarily representthe official views of the National Cancer Institute.

References

1. Walker MD, Green SB, Byar DP: Randomized compar-isons of radiotherapy and nitrosoureas for the managementof malignant glioma after surgery. N Engl J Med 303:1323–1329, 1980

2. Walker MD, Alexander EJ, Hunt WE: Evaluation of BCNUand/or radiotherapy in the treatment of anaplastic gliomas.A cooperative clinical trial. J Neurosurg 49: 333–343, 1978

3. Kristiansen K, Hagen S, Kollwvoid T: Combined modalitytherapy of operated astrocytomas Grade III and IV. Confir-mation of the value of postoperative irradiation and lack ofpotentiation of bleomycin on survival time: a prospectivemulticenter trial of the Scandinavian glioblastoma studygroup. Cancer 47: 649–652, 1981

4. Halperin EC, Burger PC, Bullard DE: The fallacy of thelocalized supratentorial malignant glioma. Int J RadiatOncol Biol Phys 15: 505–509, 1976

5. Salazar OM, Rubin P: The spread of glioblastoma as a deter-mining factor in the radiation treated volume. Int J RadiatOncol Biol Phys 1: 627–637, 1996

6. Burger PC, Heinz ER, Shibata T, Kleihues P: Topographicanatomy and CT correlations in the untreated glioblastomamultiforme. J Neurosurg 68: 698–704, 1988

7. Sheline GE: Radiotherapy for high-grade gliomas. Int JRadiat Oncol Biol Phys 18: 793–803, 1990

8. Wallner KE, Galicich JH, Krol G, Arbit E, Malkin MG: Pat-terns of failure following treatment for glioblastoma multi-forme and anaplastic astrocytoma. Int J Radiat Oncol BiolPhys 16: 1405–1409, 1989

9. Marks JE: Ionizing radiation. In: Salcman M (ed) Neuro-biology of Brain Tumors. Williams & Wilkins, Baltimore,1991, pp 299–320

10. Fine HA, Dear KBG, Loeffler JS, Black PM, Canellos GP:Meta-analysis of radiation therapy with and without adju-vant chemotherapy for malignant gliomas in adults. Cancer71: 2585–2597, 1993

11. Greenberg HS, Ensminger WD, Chandler WF: IntraarterialBCNU chemotherapy in the treatment of malignant gliomasof the central nervous system. J Neurosurg 61: 423–429,1984

12. Duffner PK, Horowitz ME, Krishner JP et al: Post-operativechemotherapy and delayed radiation in children less thanthree years of age with malignant brain tumors. N Engl JMed 328: 1725–1731, 1993

13. Hanauske AR: New approaches to chemotherapy of braintumors with or without radiation. Crit Rev Neurosurg 3:45–49, 1993

14. Grossman SA, Wharam M, Sheidler V, Kleinberg L,Zeltzman M, Yue N et al: Phase II study of continuous infu-sion carmustine and cisplatin followed by cranial irradia-tion in adults with newly diagnosed high-grade astrocytoma.J Clin Oncol 15: 2596–2603, 1997

15. Gilbert MR, Lunsford LD, Kondziolka D: A phase II studyof continuous infusion chemotherapy, external beam radio-therapy and local boost radiotherapy for malignant gliomas.(Abstract) Proc Amer Soc Clin Oncol 12: 176, 1993

16. Mehta CR, Patel NR: A network algorithm for the exacttreatment of Fisher’s exact test in rxc contingency tables.J Amer Stat Assoc 78: 427–434, 1983

17. Kaplan EL, Meier P: Nonparametric estimation fromincomplete observations. J Amer Stat Assoc 53: 457–481,1958

18. Cox DR, Snell EJ: Analysis of Binary Data. 2nd edn.,Chapman and Hall, London, 1970, pp 66–68

19. Cox DR: Regression models and life tables (with discus-sion). JR Stat Soc B (Methodol) 34: 187–220, 1972

20. Mantel N: Evaluation of survival data and two new rankorder statistics arising in its consideration. Cancer ChemothRep 50: 163–170, 1966

21. Shapiro WR, Green SB, Burger PC, Mahaley Jr MS,Selker RG, VanGilder JC et al: Randomized trial of threechemotherapy regimens and two radiotherapy regimensand two radiotherapy regimens in postoperative treatmentof malignant glioma. Brain Tumor Cooperative GroupTrial 8001. J Neurosurg 71: 1–9, 1989

152

22. Rutten EHJM: Radiation injury to the brain. In:Karim ABMF, Laws ERJ (eds) Glioma. Springer-Verlag,Heidelberg, 1991, pp 171–178

23. Volk PE, Dillon WP: Radiation injury of the brain. Amer JNeurol Radiol 12: 45–62, 1991

24. Poisson M, Pouillart P, Bataini JP, Mashaly R, Pertuisset BF,Metzger J: Malignant gliomas treated after surgery bycombination chemotherapy and delayed irradiation. ActaNeurochir 51: 15, 1979

25. Grossman SA, Sheidler V, Weissman D, Ahn H, Gilbert MR:Continuous infusion therapy for primary brain tumors: Anew approach with a high response rate. (Abstract) ProcAmer Soc Clin Oncol 6: 72, 1987

Address for offprints: Mark Gilbert, Emory University, 1365Clifton Rd., Atlanta, GA 30322, USA; Tel.: (404)778-5770; Fax:(404)778-2191; E-mail: markgilbert@emory.org