ADIATION therapy is an effective method of treatingmany types of tumors in the CNS. Unfortunately,there are a significant number of well-established

side effects associated with this therapeutic modality. Thesereactions have been typically divided into three categories,based on when they occur: early acute reactions that occurduring radiation treatment, early delayed reactions that oc-cur days or weeks postirradiation, or late delayed reactionsthat occur months to years after irradiation.25 Early acute re-actions typically include tissue edema, skin reactions, hairloss, and fatigue. Early delayed reactions can include leth-argy, tissue edema, and focal demyelination. Late delayedreactions characteristically manifest as focal or diffuse ne-crosis (characterized by vascular changes, edema, and de-myelination), gliosis, tissue calcification, and atrophy. Inaddition, sporadic data have recently begun to indicate that

delayed tumor formation may occasionally occur after a de-lay of several years postirradiation.

We report the outcome in a long-term study of a clinical-ly relevant dose of WBRT in rhesus monkeys. The clin-ical, laboratory, radiographic, molecular, and histologicalanalyses of these animals up to 10 years after WBRT is de-scribed.

Materials and MethodsAnimal Groups

All animal experiments were performed in accordance with theNational Institutes of Health guidelines on the use of animals in re-search and were approved by the Animal Care and Use Committeeof the National Institute of Neurological Disorders and Stroke.

We used a total of 12 3-year-old male primates (Macaca mulat-ta). All animals underwent fractionated WBRT while sedated (seeWhole-Brain Radiation Therapy). Previous work in our laboratoryhad demonstrated that pentobarbital is radioprotective against theearly acute and early delayed reactions of WBRT in a rat model.37,39,40

Nonetheless, the effect of barbiturate medication on late delayed re-actions, which cause severe late radiation syndromes, was unknown.It is this late delayed toxicity that limits the cumulative dose of radi-

J Neurosurg 97:1378–1389, 2002

1378

Induction of glioblastoma multiforme in nonhumanprimates after therapeutic doses of fractionatedwhole-brain radiation therapy

RUSSELL R. LONSER, M.D., STUART WALBRIDGE, B.S., ALEXANDER O. VORTMEYER, M.D.,SVETLANA D. PACK, PH.D., TUNG T. NGUYEN, M.D., NITIN GOGATE, M.D.,JEFFERY J. OLSON, M.D., AYTAC AKBASAK, M.D., R. HUNT BOBO, M.D.,THOMAS GOFFMAN, M.D., ZHENGPING ZHUANG, M.D., PH.D.,AND EDWARD H. OLDFIELD, M.D.Surgical Neurology Branch, National Institute of Neurological Disorders and Stroke; and RadiationOncology Branch, National Cancer Institute, National Institutes of Health, Bethesda, Maryland

Object. To determine the acute and long-term effects of a therapeutic dose of brain radiation in a primate model, the au-thors studied the clinical, laboratory, neuroimaging, molecular, and histological outcomes in rhesus monkeys that had re-ceived fractionated whole-brain radiation therapy (WBRT).

Methods. Twelve 3-year-old male primates (Macaca mulatta) underwent fractionated WBRT (350 cGy for 5 days/weekfor 2 weeks, total dose 3500 cGy). Animals were followed clinically and with laboratory studies and serial magnetic reso-nance (MR) imaging. They were killed when they developed medical problems or neurological symptoms, lesions ap-peared on MR imaging, or at study completion. Gross, histological, and molecular analyses were then performed.

Nine (82%) of 11 animals that underwent long-term follow up (! 2.5 years) developed neurological symptoms and/orenhancing lesions on MR imaging, which were defined as glioblastoma multiforme (GBM), 2.9 to 8.3 years after radia-tion therapy. The GBMs were categorized as either unifocal (three) or multifocal (six), and were located in the supra-tentorial (six), infratentorial (two), or both (one) cranial regions. Histological examination revealed distant, noncontiguoustumor invasion within the white matter of all nine animals harboring GBMs. Novel interspecies comparative genomic hy-bridization (three animals) uniformly showed deletions in the GBMs that corresponded to chromosome 9 in humans.

Conclusions. The high rate of GBM formation (82%) following a therapeutic dose of WBRT in nonhuman primates in-dicates that radioinduction of these neoplasms as a late complication of this therapy may occur more frequently than is cur-rently recognized in human patients. The development of these tumors while monitoring the monkeys’ conditions withclinical and serial MR imaging studies, and access to the tumor and the entire brain for histological and molecular analy-ses offers an opportunity to gather unique insights into the nature and development of GBMs.

KEY WORDS • brain • glioblastoma multiforme • radiation-induced tumor •Macaca mulatta

R

J. Neurosurg. / Volume 97 / December, 2002

Abbreviations used in this paper: ALL = acute lymphoblastic leu-kemia; CGH = comparative genomic hybridization; CNS = centralnervous system; GBM = glioblastoma multiforme; GFAP = glial fi-brillary acidic protein; MR = magnetic resonance; WBRT = whole-brain radiation therapy.

ation and therefore tumor control. To determine the potential radio-protective effects of barbiturate agents in nonhuman primates, theanimals were divided into two groups based on whether theyreceived barbiturates during WBRT. The control group (six animals)was sedated with intramuscularly administered ketamine (25 mg/kg)during each radiation session, and the experimental group (six ani-mals) was sedated with intravenously administered pentobarbital (25mg/kg) during each radiation session. The animals were allowed toawaken at the completion of each session.

Whole-Brain Radiation TherapyFractionated WBRT was administered at 350 cGy per day for 10

days (5 days/week for 2 weeks, total dose 3500 cGy). The treatmentswere performed with a bilateral exposure, replicating treatment (60Counit) of the entire brain in humans while shielding the mouth, phar-ynx, and body at a source-to-target (midsagittal plane) distance of 50cm, and at a rate of 70 cGy/minute.

Clinical Follow UpAnimals were carefully followed for medical or neurological dif-

ficulties during the study period (up to 10 years). Animals underwentserial (every 6 months) blood analysis, including complete bloodcount, and chemistry panels, including liver function studies.

Imaging Follow UpAnimals underwent MR imaging (T1-weighted with or without

contrast enhancement and T2-weighted images, 1.5 tesla) 1 to 3months before irradiation. After completion of irradiation, the ani-mals again underwent MR imaging (T1-weighted with and withoutcontrast enhancement and T2-weighted images, 1.5 tesla) at 6-monthserial intervals for the first 2 years, and then at serial intervals of ap-proximately 6 to 12 months thereafter. Supplemental MR imagingwas also performed on development of neurological symptoms.

Tissue Histological AnalysisAnimals were killed when they developed severe medical prob-

lems or neurological symptoms, or when parenchymal lesions weredemonstrated on MR imaging or at the completion of the study. Allanimals were killed by barbiturate overdose. Immediately afterdeath, the animals’ brains were removed and either frozen at !80˚Cor placed in 10% formalin. The brains were then cut coronally in10- and 40-"m-thick serial sections. These tissue sections were cutfrom areas of enhancement and/or increased T2-weighted signal in-tensity on MR images (to examine lesion pathology) and from re-gions away from obvious tumor (to examine evidence of radiationinjury and distant tumor spread not evident on MR images). Sectionswere stained with hematoxylin and eosin. Immunohistochemicalanalysis of GFAP was performed on some tumors (that is, nine tu-mors in six different animals).

Tumor CGHTo determine what genetic alterations existed between tumor and

normal tissues and how these differences corresponded with the hu-man genome, we analyzed normal brain tissues and tumor tissues inthe monkeys by using a novel interspecies CGH technique. A sum-mary of the procedure follows.

Briefly, tumor and normal brain tissues from three animals werecollected immediately after death. All tissues were then minced andsubjected to alkaline lysis, followed by alcohol extraction accordingto the manufacturer’s instructions (Qiagen, Inc., Valencia, CA). Ex-tracted normal brain DNA was labeled with digoxygenin-11–deoxy-uridine triphosphate, and tumor DNA was labeled with biotin-16–deoxyuridine triphosphate by nick translation. The labeled normalbrain and tumor DNA probes were then simultaneously hybridizedto healthy metaphase spreads from male humans. Red fluorescencelabeling of normal DNA was then performed using anti-digoxigeninFab fragments conjugated to rhodamine, and green fluorescence la-beling of tumor DNA was performed with fluorescein isothiocyanateconjugated to avidin.

Fluorescent images were then digitally captured and analyzed us-

ing commercially available software (Quips CGH software; Down-ers Grove, IL) to determine the average ratio of green (tumor) to red(normal) fluorescence. The definition of altered regions was basedon the cutoff levels of tumor to normal fluorescence ratios of 0.8 and1.2. Regions in which the ratio was lower or exceeded these thresh-old levels were considered as either a loss or a gain, respectively.Heterochromatic regions of chromosomes 1, 9, 13, 14, 15, 16, 19, 21,and 22, and the entire Y chromosome were excluded from analysis.

Statistical AnalysisStatistical analysis was performed using commercially available

software (Stat View, 5.0 version; Abacus Concepts, Berkeley, CA).Statistical tests were performed as indicated in the text.

ResultsOutcomes in Animals

Clinical Effects. All animals tolerated fractionated WBRTwithout problems and completed the course of treatmentas described. One animal died of a pulmonary abscess andsepsis 2.5 years after irradiation, thus leaving 11 animalsavailable for long-term evaluation and comparison. Therewere no detectable differences (clinical, laboratory, or oth-erwise) between animal groups (control compared withthose receiving pentobarbital) in the response to radiationduring the course of the study. Thus, all animal data wereconsidered together for the presentation of results.

There was no significant problem related to radiation ineither group of animals until 2.9 years after treatment, whenone animal (Animal 1; Table 1) developed progressive (dur-ing a period of 2–3 weeks) gait difficulties and was unableto feed itself. An MR image was obtained, revealing con-trast-enhancing lesions in the brainstem and cerebellum,and the animal was killed. The pathology at those sites wasconsistent with GBM (Fig. 1). During the course of the nextseveral years (up to 8.3 years after irradiation), eight ad-ditional animals developed contrast-enhancing cerebral le-sions that were initially found on routine serial MR imag-ing (three animals) or were demonstrated on MR imagingafter neurological deterioration (similar to Animal 1) hadprompted supplemental imaging (five animals). Neurolog-ical symptoms related to the GBMs corresponded with thelocation of the tumor on MR images, peritumoral edema,and mass effect (Table 1). Seven animals with GBMs werekilled on initial discovery of the enhancing lesions on MRimages. The remaining two animals with GBMs were fol-lowed clinically and with imaging for 1 to 2 months afterthe initial discovery of enhancing tumors on MR images(Animals 7 and 8). The mean latency period to discovery ofthe tumor in nine animals (that is, 82% of the 11 animalsthat underwent long-term follow up) that developed GBMs(either by neurological deterioration or an incidental findingon MR imaging) was 5.4 # 2.1 years (mean # standard de-viation).

The remaining two animals that did not develop MR im-aging–demonstrated lesions or neurological deficits werekilled at study completion (10 years). In these animals andthe animal that died early (2.5 years after irradiation) therewas no gross or microscopic evidence of CNS disease.

Imaging Characteristics in Animals with GBMs. Magnet-ic resonance imaging revealed unifocal (three monkeys) ormultifocal (six monkeys) contrast-enhancing lesions locat-ed in the supratentorial (six monkeys), infratentorial (two

J. Neurosurg. / Volume 97 / December, 2002

Radiation-induced glioblastoma multiforme

1379

monkeys), or both (one monkey) cranial regions (Table 1and Fig. 2). All lesions exhibited on MR imaging correlat-ed accurately with the location of the GBM on gross mor-phological study (Figs. 1 and 3). Magnetic resonance imag-es obtained in the two monkeys (Animals 7 and 8) that werefollowed after initial discovery of the enhancing lesionsrevealed rapid progression in the size of tumor and sur-rounding edema. This coincided with the development ofprogressive neurological symptoms, which prompted kill-ing the animals 1 month (Animal 8) and 2 months (Animal7) after discovery of the tumor (Figs. 4 and 5). In the nine

animals that developed GBMs, there was no evidence oftumor on MR images obtained immediately before (mean10.8 # 4.7 months before, range 6–20 months) the first MRimage that revealed tumor.

Histological AnalysisHistological analysis of all contrast-enhancing lesions on

MR imaging was consistent with GBM (nine animals, 18enhancing lesions). Diagnosis of GBM was based on char-acteristic histological findings29 including marked hyper-

R. R. Lonser, et al.

1380 J. Neurosurg. / Volume 97 / December, 2002

TABLE 1Characteristics of the nine animals that developed GBM*

LatencyAni- to Cranialmal Tumor Fo- Re-No. (yrs) cality gion Symptoms Region of Tumor†

1 2.9 M I progressive gait difficulties & inability to brainstem & cerebellumfeed during 2–3 wks

2 3.4 M S progressive visual difficulties during 10 days rt parietooccipital, lt occipital, & rt anterior temporal3 3.8 M S progressive rt arm weakness lt temporal & rt frontal4 4.3 U S progressive visual difficulties & inability to eat lt occipital pole5 4.4 M B progressive blindness & difficulties feeding bilat parietooccipital, lt parietal, rt frontal, & brainstem6 5.8 M S lt arm weakness large rt posterior temporal & smaller lt parietooccipital7 7.4 U S no symptoms initially, then increasing lt arm & rt frontal

leg weakness during 2 mos8 7.9 U S no symptoms initially, then increasing lethargy bilat frontoparietal (transcallosal)

& inability to care for itself over 1 mo9 8.3 M I asymptomatic lt & mesial cerebellum

* B = both supratentorial and infratentorial; I = infratentorial; M = multifocal; S = supratentorial; U = unifocal.† In all cases the lesions identified on MR imaging were consistent with gross findings. These imaging foci were confirmed to be GBMs

in all cases.

FIG. 1. Animal 1. a: Axial T1-weighted contrast-enhanced MR image of the pontocerebellar region, showing exten-sive enhancement of the pons and cerebellar peduncles. Imaging was performed after the animal began experiencing pro-gressive eating and gait difficulties 2.9 years after irradiation. b: Photograph of axial section obtained from the cor-responding region of the brainstem (tissue removed on the day of imaging), showing gross invasion of the pons andcerebellar peduncles (arrows) by GBM. c: Photomicrograph of tumor section obtained from the brainstem region andindicated on MR imaging as an enhancing lesion, revealing histological features consistent with GBM including pseudo-palisading cells with areas of necrosis. H & E, original magnification $ 120.

cellularity, cellular pleomorphism, neovascularity, mitoses,and pseudopalisading necrosis (Figs. 1, 3, 6, and 7). Resultsof a histological analysis of regions surrounding the tumorrevealed extensive fingerlike invasion of immediately adja-cent tissues (Fig. 6) and diffuse noncontiguous distal spreadof tumor cells particularly within the white matter (Fig. 7).Distant tumor cells were found well beyond regions thathad evidence of tumor on MR images (either T1-weighted,postcontrast, or T2-weighted imaging studies) or that couldbe seen on gross examination (Figs. 6 and 7). Histologicalanalysis of nontumorous tissues revealed no evidence ofradiation necrosis or other radiation-induced brain injury.The neoplastic cells at sites of solid GBM (nine tumors insix animals) stained uniformly positive for GFAP in threetumors, had scattered sites of positive staining in four tu-mors, and were negative for GFAP in two tumors (Fig. 8).

Comparative Genomic HybridizationTo ascertain what chromosomal differences exist be-

tween radiation-induced tumors and healthy tissue and howthese differences correspond with characteristics in the hu-man genome, we analyzed healthy brain and tumor tis-sues in primates by using an interspecies CGH technique.Successful hybridization was performed in three tumorsobtained in three separate animals (Animals 7, 8, and 9;Fig. 9).

Numerous and variable chromosome and chromosomalregion alterations were detected in all three tumors. Figure9 demonstrates the extensive chromosomal losses and gainsthat occurred in the various tumors. The loss of part orall of the genetic material corresponding to human chromo-

some 9 occurred in all tumors (Fig. 10). Deletions of thechromosomal regions corresponding to human chromo-somes 17p (p53 gene), 5q31 (epidermal growth responsefactor, interleukin-5, and interleukin-6 genes), 14 and 15(these last two are present in robertsonian fusion as chromo-some 7 in the monkey) were found in two of three tumors.Gains of the chromosomal regions corresponding to the hu-man chromosome 8q (c-myc oncogene) were detected intwo of three tumors.

DiscussionCurrent Study

Clinical Findings. The initial purpose of this study was todetermine the effects of WBRT with or without barbiturateradioprotection during treatment in a primate model. La-ter in the study, when it became evident that there wereno apparent clinical or laboratory differences between theanimal groups (barbiturate-treated compared with controls)with the use of this dosing scheme of radiation therapy andthat GBMs were developing in the animals, the goal be-came the analysis of the occurrence, distribution, and natureof radiation-induced tumors. Two animals in this study didnot develop GBMs and were killed at the completion of thestudy (10 years after irradiation). In these animals and theanimal that died early in the study from abscess and sepsis(2.5 years after irradiation) there was no imaging, gross, ormicroscopic evidence of CNS disease.

In the nine animals (82%) that developed GBMs, the cri-teria for having a radiation-induced tumor were met. Thesecriteria included the following: 1) the tumor was absent be-

J. Neurosurg. / Volume 97 / December, 2002

Radiation-induced glioblastoma multiforme

1381

FIG. 2. The majority of animals (67%) with radiation-induced GBM had multifocal disease. a–c: Axial contrast-en-hanced, T1-weighted MR images obtained in Animal 5 4.4 years after irradiation. Multifocal contrast-enhancing lesionswere located in the brainstem (a) and the cerebral hemispheres bilaterally (b and c) and were confirmed to be GBM at au-topsy. d–f: Axial contrast-enhanced, T1-weighted MR images obtained in Animal 2 3.4 years after irradiation. Multifocalcontrast-enhancing lesions were demonstrated in the cerebral hemispheres bilaterally and were confirmed to be GBM atautopsy.

fore irradiation; 2) the tumor appeared in a previously irra-diated area; 3) a latent period must have elapsed betweenthe time of irradiation and the appearance of the tumor; and4) the existence of the radiation-induced tumor must havebeen histologically proven.7,46 Preoperative MR imageswere obtained in all animals and revealed no evidence oftumor prior to irradiation. All tumors appeared within theirradiated area after a 2.9- to 8.3-year latency period. Alltumors that developed were histologically confirmed to beGBM. Moreover, the large percentage of primates in thisseries that developed radiation-induced GBMs is especiallystriking when one considers that the incidence of spontane-ous CNS malignancies in large autopsy series in primates isextremely low (0–0.004%).16,27,36

Imaging and Gross Anatomical Findings. The MR imag-ing and gross anatomical findings in the animals that devel-oped GBMs in this series allow several powerful insightsinto the nature of radiation-induced tumors. First, MR im-

aging and gross tissue examinations revealed that the ma-jority of animals with GBMs had multifocal lesions (67%).This is similar to findings in a small series of three pa-tients described by Fontana, et al.,17 in which the authorssuggested that perhaps a higher rate of multifocal lesionsoccur with radiation-induced gliomas compared with thosethat occur spontaneously (approximately 20% of GBMs inhumans).44 Second, the anatomical distribution of the radi-ation-induced GBMs in our series was also unique. Typical-ly, spontaneously occurring GBMs arise in the frontopari-etal regions, but in two animals (22%) in this series GBMsdeveloped in the cerebellum, which is rarely a spontaneousoccurrence in humans (Table 1). This distribution of GBMsappears to correspond with the volume of cerebral tissueirradiated (supratentorial compared with infratentorial).Third, the serial MR imaging used to reveal the tumor ear-ly and to follow its growth in two animals confirmed theaggressive nature of these tumors, showing rapid expan-

R. R. Lonser, et al.

1382 J. Neurosurg. / Volume 97 / December, 2002

FIG. 3. Animal 3. a: Coronal contrast-enhanced, T1-weighted MR image exhibiting a large enhancing lesion in the lefttemporal region 3.8 years after the monkey had undergone WBRT. b: Coronal brain tissue section corresponding to thesame region in panel a showing gross tumor (arrows). Prior to imaging, the animal had developed progressive right armweakness. The animal was killed on the day of imaging, and histological analysis revealed the lesion to be a GBM. c andd: Photomicrographs of tumor sections obtained from the left temporal region and demonstrated on MR imaging as an en-hancing lesion revealing histological features consistent with GBM including marked endothelial proliferation (c) with ar-eas of pseudopalisading cells and necrosis (d). H & E, original magnification $ 20.

sion and surrounding edema during a short interval of time(Figs. 4 and 5). These imaging findings coincided with therapid development and progression of neurological symp-toms that prompted us to kill the animals 1 and 2 monthsafter the tumor had been discovered. Finally, in all nineanimals with GBMs, the presence of tumor was confirmedor discovered with the use of MR imaging (either after neu-rological symptoms developed or incidentally on serial im-aging). A review of MR images obtained immediately be-fore (mean 10.8 months before) initial detection of thetumor revealed no evidence of tumor, thus suggesting thatthese tumors were de novo GBMs (it seems unlikely that

they would have evolved from lower grade astrocytomas,which would have theoretically taken longer [years] to de-velop and would have been demonstrated on earlier imag-ing studies).

Histological Findings. The histological findings in the an-imals with radiation-induced GBMs allowed several criticalinsights into the nature of and what might be necessary forthe successful treatment of these and spontaneously occur-ring GBMs. First, neoplastic cells were found far beyondthe region of tumor displayed on MR images (either T1-weighted contrast enhancement or T2-weighted hyperinten-sity; Figs. 6 and 7). This confirms that MR imaging, at least

J. Neurosurg. / Volume 97 / December, 2002

Radiation-induced glioblastoma multiforme

1383

FIG. 4. Animal 7. Coronal T1-weighted, contrast-enhanced MR image showing the rapid progression of GBM in a non-human primate 7.4 years after irradiation. There was rapid growth of the ring-enhancing lesion from 2 months (a) to 1month (b) to 1 day prior to killing the animal (c). Axial T2-weighted, unenhanced MR image demonstrating progression ofperitumoral edema 2 months (d), 1 month (e), and 1 day prior to killing the animal (f). The animal exhibited progressiveweakness of the left extremities during the 2 months before death.

FIG. 5. Animal 8. a: Coronal T1-weighted, contrast-enhanced MR image displaying an enhancing tumor in the leftfrontoparietal region 1 month prior to killing the animal and 7.9 years after irradiation. b: Coronal T1-weighted, contrast-enhanced MR image demonstrating expansion of the left frontoparietal enhancing tumor through the corpus callosum andinto the right frontoparietal region on the day the animal was killed. The animal experienced progressive weakness of theright extremities in the 1-month interval between acquisition of these images.

in its current stage of development, cannot demonstrate theentire extent of tumor spread and that diffuse neoplastic in-filtration has already occurred by the time MR imaging ex-hibits tumor (Fig. 7). Second, the widespread distributionof microscopic tumor coupled with the frequent multifo-cality represented on MR images of these tumors indicatesthat either these tumors arise from a single cell clone, sub-sequently disseminating throughout the cerebrum, or that

these neoplasms arise from multiple cells in different re-gions of the CNS. The contemporaneous development ofmultifocal tumors, the frequent occurrence of similarlysized individual foci in multifocal tumors (Fig. 2), andwidespread distribution of neoplastic cells in multifocaland unifocal GBMs combined with data from previousstudies in which motility is directly proportional to tumorgrade10,11,18 indicate that these neoplasms can arise from a

R. R. Lonser, et al.

1384 J. Neurosurg. / Volume 97 / December, 2002

FIG. 6. Animal 3. Magnetic resonance images and photomicrographs of corresponding histological tumor sections dem-onstrating the extent of tumor spread and changes in cellular structure that occurred in the tumor foci 3.8 years after themonkey underwent WBRT. Axial (a and b) and coronal (c) contrast-enhanced, T1-weighted MR images obtained from var-ious regions of the cerebrum in this animal, revealing two enhancing lesions. The larger of the two enhancing lesions waslocated in the left temporal region (a and c), whereas the smaller lesion was located in the right frontal region (b). Neoplas-tic cells (arrow in d) not evident on contrast-enhanced (a–c) or T2-weighted (not shown) MR images were found directlyposterior to the cerebral aqueduct (area corresponding to white arrow in a). Tumor cells from the left temporal region (en-hanced on MR images in a and c) had a predominantly small cell structure with fingerlike invasion (arrows in e) into theimmediately adjacent white matter. In contrast, tumor cells from the right frontal lesion (enhanced region on MR image inb) had a predominantly giant cell structure (arrow in f). Stained tissue sections from various regions revealed diffuse non-contiguous tumor cell invasion along white matter tracts distant from MR imaging–demonstrated evidence of neoplasia(g–i, arrows). Neoplastic cells were found in the fibers of the corpus callosum (g; corresponds to area indicated by whitearrow in c), internal capsule (h; corresponds to area indicated by black arrow in c), and occipital regions (i; correspondsto area indicated by white arrow in b). H & E, original magnification $ 10 (e), $ 20 (d and i), and $ 40 (f–h).

single high-grade clone that disseminates rapidly through-out the brain. Third, the noncontiguous spread of neoplasticcells within the white matter might make some regionaltherapeutic strategies ineffective (Figs. 6 and 7). Specifical-ly, because tumor-targeted gene therapy that depends on thedistribution of a therapeutic gene by a replicating virus re-quires cell-to-cell contact of tumor cells, the presence ofnoncontiguous neoplastic cells would render these therapiesnoncurative. Fourth, the preferential invasion of white mat-ter by tumor cells confirms previous findings that these ana-tomical regions or environment provide a route for rapidspread of these cells.31,44,52 The preferential spread of neo-plastic cells in these regions, however, may provide an idealsituation for certain regional therapies.52 Specifically, the in-creased hydraulic conductivity present in fibers of passageprovides an excellent anatomical situation to use interstitialbulk-flow delivery methods for widespread distribution ofputative therapeutic agents.4,34 From previous data one caninfer that white matter tracts provide the ideal mechanicalproperties safely, reliably, and homogeneously to distributesmall and large molecules through large volumes by usingconvection-enhanced delivery methods.33

Comparative Genomic Hybridization. Comparative ge-

nomic hybridization is a powerful cytogenetic procedurethat has been used successfully to determine genetic differ-ences between normal and tumorous tissues in humans.24

Because of the interspecies homology that exists betweenhumans and nonhuman primates, we were able to applynovel CGH methods (unpublished data) to detect and local-ize chromosomal alterations in these radiation-induced tu-mors in monkeys, and to establish the corresponding ge-nomic differences in humans. Although the large number ofchromosomal losses and gains occurring in these tumorsreveals the mutagenic potential of radiation therapy, the sin-gle difference common to all the tumors analyzed was thegenetic loss corresponding to human chromosome 9 (Fig.10). Similar losses of chromosome 9 are found in de novogliomas in humans,45 a factor that supports the theory thatthese radiation-induced GBMs are molecularly similar tospontaneously occurring GBMs in humans.

Previously, Olopade, et al.,38 found molecular evidenceof a deletion in 9p in 10 of 15 glioma-derived cell lines and13 of 35 primary gliomas. The shortest region of overlapin these deletions mapped to the interval between the cen-tromeric end of the interferon gene cluster and the methyl-thioadenosine phosphorylase locus. Bigner, et al.,3 found

J. Neurosurg. / Volume 97 / December, 2002

Radiation-induced glioblastoma multiforme

1385

FIG. 7. Animal 4. Magnetic resonance images and corresponding histological tumor sections demonstrating the extentof tumor invasion 4.3 years after the monkey underwent WBRT. Contrast-enhanced, T1-weighted (a and c) and T2-weight-ed (b) MR images, axial view, obtained from the supratentorial region, revealing only one enhancing lesion in the left oc-cipital region, with minimal surrounding edema (b). Stained tissue sections of tumor (enhanced region in a) revealed pleo-morphic neoplastic cells with glial differentiation (d). Noncontiguous neoplastic cells (arrow in e) with surroundingreactive astrocytes (pale pink cytoplasm) were found as distant as the right frontal region (corresponds to white arrow inc) within the white matter tracts. H & E, original magnification $ 40 (d and e).

chromosome abnormalities in 12 of 54 malignant gliomas.Although structural abnormalities of 9p and 19q were in-creased to a statistically significant degree, these authorsconcluded that the most frequent chromosomal changes inmalignant gliomas are gains of chromosome 7 and losses ofchromosome 10. Chromosome 10 was also implicated inGBMs by Fujimoto, et al.,19 who demonstrated loss of con-stitutional heterozygosity in tumor samples from 10 of 13patients in whom paired tumor and lymphocyte DNA sam-ples were screened. Note that only one of three GBMs inthe monkeys exhibited loss of the chromosome correspond-ing to human chromosome 10.

Of special interest is the result of the interstitial 5q31 de-letions in two of three tumors. This region was found to bedeleted in patients who had developed a secondary leuke-mia after a therapeutic radiation treatment for other cancers,suggesting the presence of a radiation-sensitive region con-taining important tumor suppressor genes (unpublished da-ta). Although acquired interstitial deletions of 5q have beenfound in a variety of secondary myeloid disorders, no sin-gle critical gene has been identified as the sole underlyingcause of these pathological changes. Thus, it is possible thatseveral genes may act as tumor suppressors in such cas-es. Among the candidate genes from the commonly delet-ed region are interleukin genes, interferon response fac-tor–1, colony-stimulating factor–1 receptor, and epidermalgrowth response factor.5

Previous Reports of GBM After WBRT

Animal Studies. Kent and Pickering28 reported the first in-stance of GBM formation in a nonhuman primate (one ani-mal) after whole-head irradiation (thermal neutron) in 1958.Since then, the development of brain tumors and other ma-lignancies after irradiation in nonhuman primates has beenobserved primarily in animals that received whole-body ir-radiation. Dalrymple and colleagues14,30,51,54 reported on alarge series of rhesus monkeys that had been treated withwhole–body surface irradiation using unfractionated sched-ules and lower doses (ionizing energies that would simulate

the space radiation spectrum) than those used in this study.The animals were followed over an extended period of time(up to 24 years). Nine of 71 animals (13%) exposed to 55-MeV proton radiation (total dose range 400–800 cGy) de-veloped GBMs with latent periods of 13 months to 20years. Likewise, Haymaker, et al.,20 found that three of 10rhesus monkeys that received single-dose, whole–body sur-face, 55-MeV proton irradiation (total dose range 600–800cGy) developed GBMs 3 to 5 years after treatment.

Data from previous reports by Caveness8 and col-leagues35,53 on nonhuman primates (Macaca mulatta) re-ceiving fractionated WBRT in the therapeutic dose rangeand routinely studied at autopsy revealed that some of thedelayed effects of radiation in the nonhuman primate cor-relate closely with observations in humans. These authorsfound dose-dependent pathological responses in healthynonhuman primate brains subjected to fractionated expo-sure of 200 cGy per day in total doses of 4000, 6000, and8000 cGy. Animals exposed to a total dose of 4000 cGy hadno adverse clinical effects and negligible pathological find-ings. In animals exposed to a total dose of 6000 cGy therewas evidence of increased intracranial pressure (papille-dema) and widely scattered necrotic lesions as early as 26weeks. Animals in the highest exposure group (8000 cGy)also developed papilledema and had profound histologicalchanges in the radiated region that included extensive focaland coalescing necrosis with significant brain destructionstarting 26 weeks after irradiation. None of the animals re-ported in this series developed tumors.

The proportion of monkeys developing GBMs (82%)in our study is much higher than in previous animal stud-

R. R. Lonser, et al.

1386 J. Neurosurg. / Volume 97 / December, 2002

FIG. 8. Immunohistochemical analysis with GFAP was per-formed on nine tumors from six separate animals. These tumorswere uniformly positive for GFAP, four tumors showed scatteredpositive cells, and two tumors were negative for GFAP. Shown hereis a photomicrograph of a tumor tissue section from Animal 2 stain-ing positive for GFAP. Original magnification $ 40.

FIG. 9. Using CGH techniques, tumor DNA from a radiation-in-duced GBM (green fluorescence) and healthy brain tissue (red flu-orescence) from a monkey (Animal 7) was successfully hybridizedto corresponding human chromosomes. The 4,6-diamidino-2-phen-ylindole image of the normal human metaphase spread used as thehybridization template (a), hybridization of the animal’s tumorDNA (fluorescein isothiocyanate image; b), the animal’s normalreference DNA (rhodamine image) on the normal human meta-phase spread (c), and the three fluorochrome colors merged (d).

ies, regardless of whether they received whole-body, single-dose exposure (13–30% of animals)7,54 or underwent frac-tionated WBRT (0%).8,35,53 One likely reason for the lowerpercentage of GBMs in animals treated with whole-body,single-dose irradiation is because of the direct relationshipbetween total radiation dose and tumor formation. Animalsin those studies were irradiated with considerably lower to-tal doses of radiation (total dose range 400–800 cGy) com-pared with doses in the current study (total dose 3500 cGy).The difference in the tumor-induction rate in this series,compared with that in the Caveness series is probably relat-ed to the length of follow up. All animals were killed at 2years or sooner after radiation therapy in the Caveness se-ries. The first animal in the present series developed a GBM2.9 years after WBRT.

Studies in Humans. A number of sporadic reports inthe literature implicate radiation in the induction of high-grade glial CNS tumors in humans.6,13,17,23,26,32,42,43,47–49,55,56 In a1990 report, Cavin, et al.,9 summarized 56 previously pub-lished cases and added four cases from their institution ofCNS tumor induction following radiotherapy. They foundthat patients who had undergone cranial radiotherapy forALL (mean 2389 cGy), CNS neoplasms (4544 cGy), or be-nign disease (313 cGy) were at a significantly higher risk(relative risk 125 in the ALL group) for developing intraax-

ial brain tumors. These radiation-induced tumors were iden-tified as GBMs (73%), astrocytomas (22%), ependymomas(3%), or others (2%). The mean latency period to tumor in-duction for ALL, CNS neoplasm, and benign disease radi-ation groups was 7.5, 11, and 9.9 years, respectively.

The rate of GBM formation in our series of monkeys ismuch higher than would be expected after reviewing litera-ture on radiation therapy in humans. Several factors mayexplain this difference. First, there might exist an interspe-cies variation in which nonhuman primates are more sus-ceptible to the tumor-induction effects of radiation. Second,it is possible that these animals were infected with knownoncogenic viruses endemic to the nonhuman primate popu-lation, such as simian virus 40 (associated with simian im-munodeficiency virus–infected monkeys) or JC virus.12,22 Asubsequent infection by these viruses could predispose thestudy primates to develop gliomas; however, this seems un-likely to be the predisposing cause of tumor development inthis study, because in none of the animals was there sero-logical evidence of the simian immunodeficiency virus or acompromised immune system. Third, animals in this studyunderwent WBRT, which may be associated with a high-er incidence of tumor development because of the largervolume of tissue exposed to radiation compared with morefocused radiation therapies such as stereotactic radiosur-

J. Neurosurg. / Volume 97 / December, 2002

Radiation-induced glioblastoma multiforme

1387

FIG. 10. Summary of the genetic changes detected in the three GBMs from three nonhuman primates (Animals 7,8, and 9) by interspecies CGH projected to the homologous human chromosomes. The extent and specific chromosomalalterations are demonstrated by colored lines next to the affected chromosome. Lines to the left of the chromosomes rep-resent chromosomal or chromosome region losses. Lines to the right of the chromosomes represent chromosomal or chro-mosome region gains. Specifically, the green lines represent genomic alterations in the tumor of Animal 7, the red linesrepresent alterations in the tumor from Animal 8, and the blue lines represent alterations in the tumor from Animal 9.

gery. Recently, however, radiation-induced malignant cra-nial tumors have been reported after stereotactic radiosur-gery,23,50,55 thus suggesting that, despite the reduction of ra-diation-exposed tissue, tumor formation is still a potentialcomplication of this form of radiation therapy. The induc-tion of tumors following stereotactic radiosurgery couldresult from either malignant progression of a lower-gradetumor or de novo tumor formation in the rim of irradiatedtissue at the tumor margin. Fourth, because of the long la-tency period to tumor formation, many patients with poor-grade CNS neoplasms who have undergone radiation ther-apy die before they develop radiation-induced neoplasms.Alternatively, as the number of young patients who under-go radiation therapy for benign cranial tumors (for example,vestibular schwannomas or meningiomas) and other dis-ease processes (for example, Parkinson disease, seizures,trigeminal neuralgia, or intractable pain) increases, therewill be a large enough cohort with prolonged follow up (10years or longer) to get a better estimate of the true incidenceof radiation-induced tumors in humans (including thosewho have undergone stereotactic radiosurgery). Fifth, pa-tients undergoing radiotherapy for glial neoplasms who la-ter develop a higher-grade tumor in the radiated field maybe occasionally misclassified as having a recurrence ratherthan a radiation-induced tumor. In any case, our findingsand those from an increasing number of published clinicalreports indicate that the induction of neoplasms, particularlyhigh-grade tumors, should be considered a possible long-term complication of therapeutic radiation, and patientsshould be monitored for tumor induction during long laten-cy periods.

Potential Mechanisms of Radiation-Induced Tumorigenesis

The mutagenic effects of radiation in these animals seemto be the most plausible explanation for the developmentof these tumors. Recently, Holland, et al.,21 discovered evi-dence that genetic alterations of pluripotent stem cells, notdifferentiated glial cells, might be the underlying precursorin GBM formation. These researchers induced GBM for-mation in mice after tissue-specific transfer of activated tu-morigenic genes into neuroprogenitor cells, but were un-able to do so in terminally differentiated astrocytes. Thedevelopment of GBMs in our monkeys might have resultedfrom radiation-induced genetic alterations or the radiation-induced impairment of DNA repair mechanisms within oneor more CNS neuroprogenitor cells that trigger the transfor-mation and subsequent development of the cells into neo-plasms.

The hypothesis of a neuroprogenitor precursor is intrigu-ing because it could explain some findings in the presentstudy and previous studies, and provide direction for fur-ther investigation. First, naturally occurring neuroprogeni-tor cell populations are found in nonhuman primates andhumans in amounts inversely proportional to age.41 Thisage-related relationship may explain why these relativelyyoung animals (3 years of age at the time of irradiation,which roughly corresponds to a 9- to 13-year-old human)developed GBMs with such frequency, and why childrenundergoing low-dose scalp radiation therapy (for tinea capi-tis) have a statistically significant increase in glioma forma-tion later in life.43 Second, the rapid and widespread distri-

bution of implanted neural stem cells is similar to that ofinfiltrative glioma cells.1,2,15 Thus, increased understandingof the mechanisms involved in the migration of neuropro-genitor cells could result in a better understanding of gliomacell infiltration and vice versa. Third, if neuroprogenitorcells are the precursor cells in gliomagenesis, radiation ap-plied to these cell populations with an appropriate dos-ing scheme could provide a means of developing a reliablelower animal model of an invasive high-grade glioma for-mation.

ConclusionsWe found that a high percent (82%) of animals devel-

oped radiation-induced GBMs in the late delayed period af-ter radiation therapy (latency 2.9–8.3 years), suggesting thatthe induction of high-grade glial neoplasms in irradiatedCNS tissue in primates may be a frequent late complicationof this therapy. The clinical, serial MR imaging, histologi-cal, and molecular analyses in these primates allow uniqueinsights into the nature and development of GBMs.

References

1. Aboody KS, Brown A, Rainov NG, et al: From the cover: neuralstem cells display extensive tropism for pathology in adult brain:evidence from intracranial gliomas. Proc Natl Acad Sci USA 97:12846–12851, 2000

2. Benedetti S, Pirola B, Pollo B, et al: Gene therapy of experimentalbrain tumors using neural progenitor cells. Nat Med 6:447–450,2000

3. Bigner SH, Mark J, Burger PC, et al: Specific chromosomal abnor-malities in malignant human gliomas. Cancer Res 48:405–411,1988

4. Bobo RH, Laske DW, Akbasak A, et al: Convection-enhanced de-livery of macromolecules in the brain. Proc Natl Acad Sci USA91:2076–2080, 1994

5. Boultwood J, Fidler C, Lewis S, et al: Allelic loss of IRF1 inmyelodysplasia and acute myeloid leukemia: retention of IRF1on the 5q-chromosome in some patients with the 5q-syndrome.Blood 82:2611–2616, 1993

6. Brat DJ, James CD, Jedlicka AE, et al: Molecular genetic alter-ations in radiation-induced astrocytomas. Am J Pathol 154:1431–1438, 1999

7. Cahan WG, Woodard HQ, Higinbotham NL, et al: Sarcoma aris-ing in irradiated bone: report of eleven cases. Cancer 82:8–34,1998

8. Caveness WF: Pathology of radiation damage to the normal brainof the monkey. Natl Cancer Inst Monogr 46:57–76, 1977

9. Cavin LW, Dalrymple GV, McGuire EL, et al: CNS tumor induc-tion by radiotherapy: a report of four new cases and estimate ofdose required. Int J Radiat Oncol Biol Phys 18:399–406, 1990

10. Chicoine MR, Silbergeld DL: Assessment of brain tumor cell mo-tility in vivo and in vitro. J Neurosurg 82:615–622, 1995

11. Chicoine MR, Silbergeld DL: The in vitro motility of human gli-omas increases with increasing grade of malignancy. Cancer 75:2904–2909, 1995

12. Chretien F, Boche D, Lorin de la Grandmaison G, et al: Progres-sive multifocal leukoencephalopathy and oligodendroglioma in amonkey co-infected by simian immunodeficiency virus and simi-an virus 40. Acta Neuropathol 100:332–336, 2000

13. Chung CK, Stryker JA, Cruse R, et al: Glioblastoma multiformefollowing prophylactic cranial irradiation and intrathecal metho-trexate in a child with acute lymphocytic leukemia. Cancer 47:2563–2566, 1981

14. Dalrymple GV, Nagle WA, Moss AJ, et al: The protons of space

R. R. Lonser, et al.

1388 J. Neurosurg. / Volume 97 / December, 2002

and brain tumors: I. Clinical and dosimetric considerations, inRestor AC, Trombka JI (eds): High-Energy Radiation Back-ground in Space. AIP Conference Proceedings. New York:American Institute of Physics, 1989, Vol 186, pp 407–411

15. Dirks PB: Glioma migration: clues from the biology of neural pro-genitor cells and embryonic CNS cell migration. J Neurooncol53:203–212, 2001

16. Fairbrother RW, Hurst EW: Spontaneous diseases observed in 600monkeys. J Pathol Bacterial 35:867–873, 1932

17. Fontana M, Stanton C, Pompili A, et al: Late multifocal gliomasin adolescents previously treated for acute lymphoblastic leuke-mia. Cancer 60:1510–1518, 1987

18. Friedlander DR, Zagzag D, Shiff B, et al: Migration of brain tumorcells on extracellular matrix proteins in vitro correlates with tumortype and grade and involves alphaV and beta1 integrins. CancerRes 56:1939–1947, 1996

19. Fujimoto M, Fults DW, Thomas GA, et al: Loss of heterozygosityon chromosome 10 in human glioblastoma multiforme. Genom-ics 4:210–214, 1989

20. Haymaker W, Rubinstein LJ, Miquel J: Brain tumors in irradiatedmonkeys. Acta Neuropath 20:267–277, 1972

21. Holland EC, Celestino J, Dai C, et al: Combined activation of Rasand Akt in neural progenitors induces glioblastoma formation inmice. Nat Genet 25:55–57, 2000

22. Hurley JP, Ilyinskii PO, Horvath CJ, et al: A malignant astrocy-toma containing simian virus 40 DNA in a macaque infected withsimian immunodeficiency virus. J Med Primatol 26:172–180,1997

23. Kaido T, Hoshida T, Uranishi R, et al: Radiosurgery-induced braintumor. Case report. J Neurosurg 95:710–713, 2001

24. Kallioniemi A, Kallioniemi OP, Sudar D, et al: Comparative ge-nomic hybridization for molecular cytogenetic analysis of solidtumors. Science 258:818–821, 1992

25. Karim ABMF: Radiation therapy and radiosurgery for brain tu-mors, in Kaye AH, Laws ER Jr (eds): Brain Tumors: An Ency-clopedic Approach. Edinburgh: Churchill Livingstone, 1995, pp331–348

26. Kaschten B, Flandroy P, Reznik M, et al: Radiation-induced glio-sarcoma: case report and review of the literature. J Neurosurg 83:154–162, 1995

27. Kennard MA: Abnormal findings in 246 consecutive autopsies onmonkeys. Yale J Biol Med 13:701–712, 1941

28. Kent SP, Pickering JE: Neoplasms in monkeys (Macaca mulatta)spontaneous and irradiation induced. Cancer 11:138–147, 1958

29. Kleihues P, Cavenee WK: Pathology and Genetics of Tumoursof the Nervous System. Lyon: IARC Press, 2000

30. Krupp JH: Nine-year mortality experience in proton-exposed Ma-caca mulatta. Radiat Res 67:244–251, 1976

31. Laws ER Jr, Goldberg WJ, Bernstein JJ: Migration of human ma-lignant astrocytoma cells in the mammalian brain: Scherer revis-ited. Int J Dev Neurosci 11:691–697, 1993

32. Liwnicz BH, Berger TS, Liwnicz RG, et al: Radiation-associatedgliomas: a report of four cases and analysis of postradiation tu-mors of the central nervous system. Neurosurgery 17:436–445,1985

33. Lonser RR, Gogate N, Morrison PF, et al: Direct convective de-livery of macromolecules to the spinal cord. J Neurosurg 89:616–622, 1998

34. Morrison PF, Laske DW, Bobo H, et al: High-flow microinfusion:tissue penetration and pharmacodynamics. Am J Physiol 266:R292–R305, 1994

35. Nakagaki H, Brunhart G, Kemper TL, et al: Monkey brain dam-age from radiation in the therapeutic range. J Neurosurg 44:3–11, 1976

36. O’Connor GT: Using primates in medical research. 2. Recent

comparative research. Cancer—a general review. Primates Med3:9–22, 1969

37. Oldfield EH, Friedman R, Kinsella T, et al: Reduction in radiation-induced brain injury by use of pentobarbital and lidocaine protec-tion. J Neurosurg 72:737–744, 1990

38. Olopade OI, Jenkins RB, Ransom DT, et al: Molecular analysis ofdeletions of the short arm of chromosome 9 in human gliomas.Cancer Res 52:2523–2529, 1992

39. Olson JJ, Friedman R, Orr K, et al: Enhancement of the efficacyof x-irradiation by pentobarbital in a rodent brain-tumor model. JNeurosurg 72:745–748, 1990

40. Olson JJ, Shelley C, Orr K, et al: The cerebral radioprotective ef-fect of alternative barbiturates to pentobarbital. Neurosurgery30:720–723, 1992

41. Palmer TD, Schwartz PH, Taupin P, et al: Cell culture. Progenitorcells from human brain after death. Nature 411:42–43, 2001

42. Rimm IJ, Li FC, Tarbell NJ, et al: Brain tumors after cranial irradi-ation for childhood acute lymphoblastic leukemia. A 13-year ex-perience from the Dana-Farber Cancer Institute and the Children’sHospital. Cancer 59:1506–1508, 1987

43. Ron E, Modan B, Boice JD Jr, et al: Tumors of the brain and ner-vous system after radiotherapy in childhood. N Engl J Med 319:1033–1039, 1988

44. Scherer HJ: The forms of growth in gliomas and their practical sig-nificance. Brain 63:1–35, 1940

45. Schlegel J, Scherthan H, Arens N, et al: Detection of complexgenetic alterations in human glioblastoma multiforme using com-parative genomic hybridization. J Neuropathol Exp Neurol 55:81–87, 1996

46. Schrantz JL, Araoz CA: Radiation induced meningeal fibrosarco-ma. Arch Pathol 93:26–31, 1972

47. Shamisa A, Bance M, Nag S, et al: Glioblastoma multiforme oc-curring in a patient treated with gamma knife surgery. Case reportand review of the literature. J Neurosurg 94:816–821, 2001

48. Simmons NE, Laws ER Jr: Glioma occurrence after sellar irradia-tion: case report and review. Neurosurgery 42:172–178, 1998

49. Soffer D, Gomori JM, Pomeranz S, et al: Gliomas following low-dose irradiation to the head report of three cases. J Neurooncol8:67–72, 1990

50. Thomsen J, Mirz F, Wetke R, et al: Intracranial sarcoma in a pa-tient with neurofibromatosis type 2 treated with gamma knife ra-diosurgery for vestibular schwannoma. Am J Otol 21:364–370,2000

51. Traynor JE, Casey HW: Five-year follow-up of primates exposedto 55 MeV protons. Radiat Res 47:143–148, 1971

52. Tysnes BB, Mahesparan R: Biological mechanisms of gliomainvasion and potential therapeutic targets. J Neurooncol 53:129–147, 2001

53. Wakisaka S, O’Neill RR, Kemper TL, et al: Delayed brain dam-age in adult monkeys from radiation in the therapeutic range. Ra-diat Res 80:277–291, 1979

54. Wood DH: Long-term mortality and cancer risk in irradiated rhe-sus monkeys. Radiat Res 126:132–140, 1991

55. Yu JS, Yong WH, Wilson D, et al: Glioblastoma induction afterradiosurgery for meningioma. Lancet 356:1576–1577, 2000

56. Zuccarello M, Sawaya R, deCourten-Meyers G: Glioblastoma oc-curring after radiation therapy for meningioma: case report and re-view of literature. Neurosurgery 19:114–119, 1986

Manuscript received April 25, 2002.Accepted in final form August 7, 2002.Address reprint requests to: Edward H. Oldfield, M.D., Surgical

Neurology Branch, National Institute of Neurological Disorders andStroke, National Institutes of Health, Building 10, Room 5D37, Be-thesda, Maryland 20892-1414. email: eo10d@nih.gov.

J. Neurosurg. / Volume 97 / December, 2002

Radiation-induced glioblastoma multiforme

1389