Medulloblastoma and Supratentorial PrimitiveNeuroectodermal TumorsAn Institutional Experience

Arnold C. Paulino, M.D.

Edward Melian, M.D.

Department of Radiotherapy, Cardinal BernardinCancer Center and Ronald McDonald Children’sHospital, Loyola University Medical Center, May-wood, Illinois.

Presented in part at the 17th Annual EuropeanSociety for Therapeutic Radiology and OncologyMeeting, Edinburgh, Scotland, United Kingdom,September 20–24, 1998.

Address for reprints: Arnold C. Paulino, M.D., De-partments of Radiology and Pediatrics, The Uni-versity of Iowa Hospitals and Clinics, 200 HawkinsDrive, Room W 189-Z, GH, Iowa City, IA 52242.

Received October 13, 1998; revision received Jan-uary 25, 1999; accepted February 15, 1999.

BACKGROUND. To the authors’ knowledge there are relatively few data concerning

supratentorial primitive neuroectodermal tumors (PNET). The authors retrospec-

tively reviewed all cases of PNET of the brain treated at the study institution to

determine whether there was a difference in presentation, overall survival, and

recurrence-free survival with regard to tumor location (supratentorium vs. poste-

rior fossa).

METHODS. Between 1977–1996 33 patients with PNET were diagnosed and treated

at 1 radiotherapy center. The median age of the patients was 9 years. The location

of the tumor was in the posterior fossa in 25 patients and the supratentorium in 8

patients. The tumor had spread to the neuraxis in six patients; four patients with

disseminated neuraxis disease had a supratentorial PNET and two had a posterior

fossa PNET. All but three patients received craniospinal irradiation. The primary

tumor received $ 5000 centigray in 27 patients and chemotherapy was employed

in 26 patients. The median follow-up was 60 months.

RESULTS. The 5-year overall and recurrence-free survival rates for all patients were

77.2% and 79.6%, respectively. The 5-year overall survival rates were 86.3% for

patients with medulloblastoma (posterior fossa PNET) and 46.9% for patients with

supratentorial PNET (P 5 0.01, log rank test). For overall survival, prognostic

factors included radiotherapy dose to the primary site, metastases (M) status, and

location of the primary tumor. The 5-year recurrence free survival rates were 89.8%

for patients with medulloblastoma and 46.9% for patients with supratentorial

PNET (P 5 0.003, log rank test). For recurrence free survival, prognostic factors

included M status and primary tumor site location; radiation dose to the primary

tumor site and patient gender were of borderline significance. In the ten patients

with inadequate posterior fossa boost fields judged by Children’s Cancer Group

criteria, there were two failures, both of which were in the original tumor bed.

CONCLUSIONS. Supratentorial PNET has a worse overall survival and recurrence

free survival than medulloblastoma. There is a suggestion that radiotherapy boosts

in medulloblastoma may not need to encompass the entire posterior fossa because

posterior fossa failures primarily are in the tumor bed. Larger studies with longer

follow-up are needed to determine whether craniospinal irradiation followed by a

boost to the tumor bed is adequate for medulloblastoma patients. Cancer 1999;86:

142– 8. © 1999 American Cancer Society.

KEYWORDS: supratentorial primitive neuroectodermal tumor, medulloblastoma, ra-diotherapy, posterior fossa boost, pineoblastoma.

Primitive neuroectodermal tumors (PNET) are a distinct smallround cell neoplasm found in the brain.1 The most common

location is in the posterior fossa (medulloblastoma), which accountsfor 20 –25% of all childhood brain tumors. PNET arising in the

142

© 1999 American Cancer Society

supratentorium is less common and comprises ap-proximately 2.5% of childhood brain neoplasms.Treatment for medulloblastoma traditionally has in-cluded maximal surgical debulking and craniospinalirradiation with a posterior fossa boost with or withoutchemotherapy. Currently, chemotherapy is employedin high risk medulloblastoma patients in combinationwith radiotherapy for better survival outcome; it also isused in low risk medulloblastoma patients to substi-tute for a lower craniospinal dose in children.2,3 Forsupratentorial PNET, maximal surgical debulking fol-lowed by chemotherapy and radiotherapy is em-ployed. To our knowledge information regarding thepresentation of these patients and survival outcome islimited. In a previous Children’s Cancer Group (CCG)report, 3-year overall survival and progression freesurvival rates for 55 children with supratentorial PNETwere 57% and 45%, respectively.4 Children were ran-domized to receive craniospinal irradiation (CSI) fol-lowed by eight cycles of lomustine, vincristine, andprednisone or two cycles of eight-in-one chemother-apy followed by CSI and eight additional cycles ofeight in one chemotherapy. No significant differencein survival or progression free survival was observedbetween the two treatment groups. Pineal site of in-volvement and M0 status were found to be indepen-dent prognostic factors for an improved outcome. Inanother report from the Hospital for Sick Children inToronto, the median survival for 36 children with su-pratentorial PNETs was 23 months; the 5-year overallsurvival rate was 18%.5 Pineal region tumors werefound in 10 of 36 cases (27.8%). All the survivors re-ceived CSI; there was a trend toward better survival inchildren whose surgery was comprised of macro-scopic total resection. The purpose of this study was toreview our institutional experience with PNET of thebrain and to identify prognostic factors in relation tothe overall survival and recurrence free survival rates.In addition, we analyzed the patterns of failure in

relation to radiotherapy fields to obtain a better un-derstanding of these tumors.

MATERIALS AND METHODSPatient and Tumor CharacteristicsBetween 1977 and 1996, 33 patients at 1 institutionwere diagnosed and treated for PNET of the brain.There were 22 males and 11 females with a median ageof 9 years (range, 12 months–33 years). The location ofthe tumor was in the posterior fossa in 25 patients andin the supratentorium in 8 patients. The median timefrom the onset of symptoms to diagnosis was 1 month.Symptoms included nausea and emesis in 22 patients(66.7%), headache in 22 patients (66.7%), ataxia in 15patients (45.5%), visual disturbances in 15 patients(45.5%), clumsiness in 3 patients (9.1%), and back painin 2 patients (6.1%). Prior to 1985, computed tomog-raphy (CT) of the brain with contrast, cerebrospinalfluid (CSF) cytology, and myelography were used todocument neuraxis dissemination. After 1985 all pa-tients underwent magnetic resonance imaging (MRI)of the brain and spine and CSF cytology. Dissemina-tion to the neuraxis was categorized according to theM classification proposed by Chang et al.6; 27 patientswere classified as M0, 2 patients were classified as M1,and 4 patients were classified as M3.

Table 1 lists the patient, tumor, and treatmentcharacteristics of the eight patients with supratento-rial PNET. There was no gender preponderance, andthe ages of the patients ranged from 3–32 years. Half ofthe lesions were in the pineal gland and half of thepatients had neuraxis dissemination at the time ofinitial presentation.

TreatmentEighteen of 33 patients with intracranial PNET under-went shunt placement for the management of hydro-cephalus. Macroscopic total resection of the primarylesion was performed in 15 patients, subtotal resection

TABLE 1Patient, Tumor, and Treatment Characteristics in Supratentorial Primitive Neuroectodermal Tumors and Outcome

Patientno.

Age(yrs) Gender

Location oftumor

Mstatus Treatment Outcome

1 9 M R parietal lobe M0 S 1 CSRT 1 chemo Alive at 34 mos with nonprimary brain recurrence2 4 F Pineal gland M3 Bx 1 CSRT 1 chemo Dead at 2 mos with progression of tumor3 3 F Suprasellar region M3 S 1 CSRT 1 chemo Dead at 8 mos with spine and nonprimary brain recurrence4 32 M Pineal gland M0 S 1 CSRT Alive at 61 mos5 14 M L temporal lobe M0 S 1 CrRT 1 chemo Dead at 12 mos with recurrence at primary tumor site6 6 F R frontal lobe M0 S 1 CSRT 1 chemo Alive at 99 mos7 10 F Pineal gland M1 S 1 CSRT 1 chemo Alive at 24 mos8 13 M Pineal gland M3 Bx 1 CSRT 1 chemo Dead at 17 mos with recurrence in primary tumor site, brain, and spine

M status: metastases status; M: male; R: right; S: surgery; CSRT: craniospinal radiation therapy; chemo: chemotherapy; F: female; Bx: biopsy; L: left; CrRT: cranial radiation therapy.

PNET and Medulloblastoma/Paulino and Melian 143

in 15 patients, and biopsy only in 3 patients. All butthree patients received CSI. For CSI, the patient wasimmobilized in a prone position and treated with twolateral brain fields abutting a posterior spinal field. Inolder children and young adults, two spinal fields wereemployed. The cranial fields were collimated to matchthe divergence from the spinal field; likewise thecouch was angled to match the divergence from thecranial fields. The craniospinal junction was movedtwo to three times during the entire treatment courseto prevent overdosage or underdosage at the cervicalspinal cord in this region. Beam energies of 1.25- to6-megavolt photons were used at a fraction size of 180centigray (cGy) per day. Median doses to the cranialand spinal fields were 3960 cGy (range, 1800 – 4500cGy) and 3600 cGy (range, 1800 – 4500 cGy), respec-tively. The variability of the doses to the brain andspine was secondary to the radiation oncologist’s pref-erence, with younger patients usually receiving thelower doses. All but 6 patients received a dose of .5000 cGy to the primary tumor site. The total dose tothe primary tumor site ranged from 4500 –5580 cGy. Ofthe three patients who received less than CSI, tworeceived posterior fossa irradiation only and the otherreceived whole brain irradiation followed by a poste-rior fossa boost. For the 25 patients with posteriorfossa PNETs, 22 had evaluable preoperative diagnosticscans of the primary site and radiation ports. Twelveof 22 patients with medulloblastoma had adequateposterior fossa fields based on CCG standards (Fig. 1).Of the 10 patients who had inadequate posterior fossaboost fields, 9 had the tumor or surgical bed with a2-cm margin included in the boost field. The reasonsfor inadequacy of posterior fossa boost fields areshown in Table 2. Chemotherapy was employed in 26patients. The most common regimen was vincristine,lomustine, and prednisone for 8 cycles in 16 patients.

Statistics and Follow-UpEstimates of overall and recurrence free survival werecalculated using the Kaplan–Meier method.7 Overalland recurrence free survivals were calculated from thetime of initial diagnosis to the event of interest. Foroverall survival, the event of interest was deathwhereas for recurrence free survival the event of in-terest was any recurrence. Univariate analysis of pos-sible prognostic factors was performed using the Pear-son chi-square test. For multivariate analysis ofprognostic factors, the Cox proportional hazardsmodel was used.8 The log rank test was employed tocompare survival and recurrence free survival curves.The median follow-up time was 60 months (range,24 –240 months).For patients diagnosed after 1985,MRI of the brain and spine was obtained every 3

months during the first 2 years of follow-up. This wasrepeated every 6 months until 5 years from diagnosis.For patients diagnosed prior to 1985, CT scans of thebrain were obtained every 3 months for the first 2years followed by a scan every 6 months until 5 yearsof follow-up. Myelography was not performed rou-tinely except for symptomatic patients.

RESULTSSurvival and Recurrence Free SurvivalThe 5-year overall and recurrence free survival ratesfor all patients were 77.2% and 79.6%, respectively.The 5-year overall survival rates for patients withPNETs of the posterior fossa and supratentorium were86.3% and 46.9%, respectively, as demonstrated inFigure 2 (P 5 0.01); the 5-year recurrence free survivalrates were 89.8% and 46.9%, respectively, for patients

FIGURE 1. Example of adequate boost to the posterior fossa. The anterior

border is at the posterior clinoid, the inferior border is at the C1-C2 interspace,

the posterior border is located beyond the skull, and the superior border is

above the tentorium cerebelli.

TABLE 2Reasons for Inadequacy in the 10 Patients with Inadequate PosteriorFossa Boost Fields

Reason Frequency

Inadequate superior margin 6/10 (60%)Inadequate anterior margin 7/10 (70%)Inadequate posterior margin 4/10 (40%)Inadequate inferior margin 6/10 (60%)

144 CANCER July 1, 1999 / Volume 86 / Number 1

with PNETs of the posterior fossa and supratentorium(P 5 0.003). The 5-year overall survival rates for pa-tients without neuraxis dissemination (M0) and thosewith metastatic spread (M1) were 87.2% and 25.0%,respectively, as demonstrated in Figure 3 (P 5 0.004).The 5-year recurrence free survival rates for M0 andM1 patients were 90.5% and 25.0%, respectively (P 50.001).

Analysis of Prognostic FactorsThe most important prognostic factors for overallsurvival in decreasing importance were dose to theprimary tumor site, metastases (M) status, and thelocation of the primary tumor (Table 3). On multi-variate analysis, both dose to the primary tumor siteand M status were significant. For recurrence freesurvival, the most important prognostic factors indecreasing importance were M status and loca-tion of the primary tumor site; dose to the primarytumor site and gender were of borderline signifi-cance (Table 4). On multivariate analysis, both Mstatus and location of the primary tumor site weresignificant prognostic factors for recurrence freesurvival.

Posterior Fossa BoostOf the 25 patients with medulloblastoma, 22 hadevaluable preoperative imaging of the primary tu-mor and radiation portals. Twelve of 22 posteriorfossa boost fields were considered adequate by CCGprotocol standards. None of the patients had poste-

rior fossa failure with adequate posterior fossafields. Of the ten posterior fossa fields judged to beinadequate, there were two posterior fossa failures,both in the tumor bed. One failure most likelywas secondary to the exclusion of the whole tumorbed in the boost field; this patient also had failurein the spine. On statistical analysis, the adequacyof posterior fossa boost fields was not related tooverall survival (P 5 0.99) or recurrence free survival(P 5 0.81).

FIGURE 2. Overall survival of patients according to tumor location.

FIGURE 3. Overall survival of patients according to metastases (M) status.

TABLE 3Patient, Tumor, and Treatment Characteristics in Relation to OverallSurvival for Primitive Neuroectodermal Tumor of the Brain

Characteristic Deaths (%)Univariateanalysis

Multivariateanalysis

Age (yrs) P 5 0.25 —, 3 2/5 (40.0)$ 3 5/28 (17.9)Gender P 5 0.56 —Male 4/22 (18.2)Female 3/11 (27.3)Primary tumor site P 5 0.03 P 5 NSPosterior fossa 3/25 (12.0)Supratentorium 4/8 (50.0)Size of primary tumor (cm) P 5 0.58 —# 3 cm 2/8 (25.0). 3 cm 5/25 (20.0)M status P 5 0.003 P 5 0.012M0 3/27 (11.1)M1 4/6 (66.7)Brainstem involvement P 5 0.99 —Yes 0/2 (0.0)No 7/31 (22.6)Shunt placement P 5 0.10 —Yes 6/18 (33.3)No 1/15 (6.7)Dose to primary tumor site

(cGy) P 5 0.002 P 5 0.0002, 5000 4/6 (66.7)$ 5000 3/27 (11.1)Time from surgery to RT

(weeks) P 5 0.32 —# 6 5/26 (19.2). 6 2/7 (28.6)Adequacy of CSRT fields P 5 0.24 —Adequate 2/14 (14.3)Inadequate 5/19 (26.3)Chemotherapy P 5 0.57 —Yes 6/26 (23.0)No 1/7 (14.3)Type of surgery P 5 0.95 —Macroscopic total resection 3/15 (20.0)Biopsy or subtotal

resection 4/18 (22.2)

NS: not significant; M status: metastases status; cGy: centigrays; RT: radiation therapy; CSRT: cranio-

spinal radiation therapy.

PNET and Medulloblastoma/Paulino and Melian 145

Patterns of FailureTable 5 lists the patterns of failure among the ninepatients in the current study who developed a recur-rence. The most common location of a recurrence wasin the primary tumor site (supratentorium: four pa-tients; posterior fossa: two patients). Both recurrencesin medulloblastoma patients were in the tumor bed.Nonprimary site brain and the spine were involved infive and four recurrent cases, respectively. Failure inthe primary tumor site alone was found in three pa-tients; all initially were diagnosed with M0 disease.

DISCUSSIONPNET can occur anywhere in the brain. The mostcommon location is in the posterior fossa, but PNETalso can occur in supratentorial locations such as thecerebral cortex and the pineal gland region (pine-oblastoma).

In this report the overall survival for patients withmedulloblastoma was high compared with other se-ries.9-13 One explanation for the 5-year survival rate of86.3% reported in the current study is the predomi-nance of M0 disease in our series; only 2 of 25 patients(8.0%) had neuraxis dissemination at the time of initialpresentation. This is supported by the better survivalof patients with M0 tumors compared with patientswith M1 tumors. Our findings also are consistent withprevious reports showing that survival is better inpatients whose radiation dose to the posterior fossadose is . 5000 cGy. At the Mayo Clinic, Garton et al.reported a 72% survival rate for children who re-ceived . 5000 cGy to the posterior fossa comparedwith 24% for children whose radiation doses were ,5000 cGy.9 Silverman and Simpson reported theWashington University experience that showed sur-vival rates of 80% and 38%, respectively, for radiationdoses to the posterior fossa of . 5000 cGy and , 5000cGy.13 Khafaga et al. studied 149 patients and showeda survival rate of 54% for patients who received a doseof . 5000 cGy to the posterior fossa versus 18% forthose who received a dose , 5000 cGy.14

In this study, the 5-year overall survival and re-currence free survival rates were both 46.9% for su-pratentorial PNET. To our knowledge, information re-garding supratentorial PNET is limited. Our figures areconsistent with two previous reports that showed aworse outcome for patients with PNET originating inthe supratentorium compared with the posteriorfossa.4,5 On univariate analysis, patients with posteriorfossa tumors had a better overall survival than those

TABLE 4Patient, Tumor, and Treatment Characteristics in Relation toRecurrence Free Survival

CharacteristicRecurrence(%)

Univariateanalysis

Multivariateanalysis

Age (yrs) P 5 0.50 —, 3 2/5 (40.0)$ 3 8/28 (28.6)Gender P 5 0.058 P 5 NSMale 4/22 (18.2)Female 6/11 (54.5)Primary tumor site P 5 0.002 P 5 0.024Posterior Fossa 4/25 (16.0)Supratentorium 6/8 (75.0)Size of primary tumor (cm) P 5 0.87 —# 3 2/8 (25.0). 3 8/25 (32.0)M status P 5 0.0003 P 5 0.0001M0 6/27 (22.2)M1 4/6 (66.7)Brainstem involvement P 5 0.99 —Yes 0/2 (0.0)No 10/31 (32.3)Shunt placement P 5 0.19 —Yes 3/18 (16.7)No 7/15 (46.7)Dose to primary tumor site

(cGy) P 5 0.052 P 5 NS, 5000 3/6 (50.0)$ 5000 7/27 (25.9)Time from surgery to RT

(weeks) P 5 0.76 —# 6 8/26 (30.8). 6 2/7 (28.6)Adequacy of CSRT fields P 5 0.19 —Adequate 3/14 (21.4)Inadequate 7/19 (36.8)Chemotherapy P 5 0.42 —Yes 9/26 (34.6)No 1/7 (14.3)Type of surgery P 5 0.79 —Macroscopic total resection 4/15 (26.7)Biopsy or subtotal

resection 6/18 (33.3)

NS: not significant; M status: metastases status; cGy: centigrays; RT: radiation therapy; CSRT: cranio-

spinal radiation therapy.

TABLE 5Patterns of Failure According to M Status

Location of recurrence

Frequency

M0 M1

Primary tumor site alone 3 —Braina alone 1 —Spine alone — —Primary tumor site 1 braina — 1Primary tumor site 1 spine 1 —Braina 1 spine 1 1Primary tumor site 1 braina 1 spine — 1

M status: metastases status.a Represents nonprimary tumor site brain.

146 CANCER July 1, 1999 / Volume 86 / Number 1

with tumors located in the supratentorium. On mul-tivariate analysis, patients with medulloblastoma hada better recurrence free survival compared with thosewith supratentorial PNET. Like medulloblastoma, su-pratentorial PNET has a tendency to spread via theneuraxis; four of the eight patients had neuraxis dis-semination at the time of initial presentation. Initialradiotherapeutic management of this disease there-fore should include craniospinal irradiation.

In this study, both pineal and cortical PNET wereanalyzed together as supratentorial PNET. Rorke pre-viously proposed a simplified classification for smallround cell neoplasms of the brain or PNET that in-cluded both of these sites.1 However, the nomencla-ture and overall concept of PNETs is not accepteduniversally. The CCG study regarding supratentorialPNET likewise included both pineal and cortical sites;it is interesting to note that pineal tumors had a betteroutcome in the CCG study. The 3-year progressionfree survival rates were 61% and 33% for pineal andcortical PNET, respectively.4 Our data are based on asmall number of patients and show no difference insurvival outcome between the two groups; however,three of four patients with pineoblastoma presentedwith M positive disease compared with one of fourpatients with cortical PNET.

The question of whether a tumor bed boost canreplace a posterior fossa boost cannot be answered inthis study. Twelve of 22 evaluable ports were classifiedas adequate per CCG recommendations. The tradi-tional posterior fossa boost guidelines include an an-terior border at the posterior clinoid, an inferior bor-der at the C1-C2 interspace, a posterior borderflashing the skull, and a superior border above thetentorium cerebelli which usually is 1 cm above halfthe distance from the base of skull to the vertex. Of theten patients who had inadequate posterior fossaboosts, there were two posterior fossa failures, both ofwhich were in the tumor bed. Only one patient had anisolated posterior fossa failure. One can argue thatthese two failures may have occurred with a tumorbed boost only because failures were in the tumor bedand not elsewhere in the parenchyma of the posteriorfossa. There was no difference in survival or recur-rence free survival regardless of whether the posteriorfossa was classified as adequate or inadequate by CCGstandards. Two European studies previously reportedthat treatment of the entire posterior fossa in theboost field may not be necessary.15,16 Miralbell et al.found no correlation between local control and treat-ment of the entire posterior fossa. For the 16 posteriorfossa fields with acceptable margins, the 5-year pro-gression free survival rate was 69% whereas it was 70%for the 56 posterior fossa fields with unacceptable

margins.15 Carrie et al. noted no posterior fossa recur-rences in 20 children with reduced posterior fossaboost fields and concluded that it may be possible toboost only the tumor defined preoperatively.16 Shouldone abandon the practice of boosting the entire pos-terior fossa? Although our data suggest that volumesless than the entire posterior fossa may be adequate,we continue to boost the posterior fossa compartmentin children with medulloblastoma because of the lim-itations of our study. First, there are few patients inour study, and this may have biased the results. Sec-ond, the median follow-up in the current study wasonly 5 years. Approximately 15% of all recurrences inchildren with medulloblastoma occur after 5 years.17

Third, one cannot conclude that a tumor bed with asafety margin is the appropriate target volume for theboost from our study and the two European reports.Volumes more than the tumor bed with a safety mar-gin but less than the entire posterior fossa have beentreated in majority of patients.

Tomita and McClone previously have shown pos-itive biopsies from visibly normal areas of the cisternalarachnoid separate from the primary tumor in four offive patients.18 Data from a combined series from theUniversity of Michigan and Children’s Hospital ofPhiladelphia demonstrate that local failure within theposterior fossa but outside the tumor bed occurred in11 of 27 of all failures (41%).19 Of these 11 failures inthe posterior fossa, 7 failed in the leptomeninges, 3 inthe posterior fossa parenchyma, and 1 in the brain-stem. In a recent study of 19 children who underwentmacroscopic or subtotal resection of medulloblas-toma, 10 children (52.6%) demonstrated microscopiclocal leptomeningeal invasion adjacent to the primarytumor.20 With this high incidence of microscopic localleptomeningeal invasion and approximately 40% rateof failures in the posterior fossa outside the tumorbed, it is difficult to justify not including the entireposterior fossa in the irradiated field. A careful patternof failure study with long term follow-up in patientstreated with a tumor bed and margin boost is neededto determine whether we can eliminate the posteriorfossa boost.

There was a suggestion in this study that patientswith M0 disease were more likely to fail in the primarytumor site alone compared with patients with M1disease, in whom the predominant pattern of failureusually involves the neuraxis (three of six M0 patientsvs. none of three M1 patients). This is consistent witha CCG report that showed that primary tumor sitefailure alone is more likely in M0 patients, whereas theentire neuraxis, especially the spine, was at high riskfor failure in M1 patients.21

PNETs arising in the supratentorium have a worse

PNET and Medulloblastoma/Paulino and Melian 147

prognosis compared with those arising in the poste-rior fossa. A large pattern of failure analysis with longterm follow-up is needed to determine whether tumorbed boost can replace the posterior fossa boost inpatients with medulloblastoma.

REFERENCES1. Rorke LB. The cerebellar medulloblastoma and its relation-

ship to primitive neuroectodermal tumors. J NeuropatholExp Neurol 1983;42:1–15.

2. Goldwein JW, Radcliffe J, Johnson J, Moshang T, Packer RJ,Sutton LN, et al. Updated results of a pilot study of low dosecraniospinal irradiation plus chemotherapy for children un-der five with cerebellar primitive neuroectodermal tumors(medulloblastoma). Int J Radiat Oncol Biol Phys 1996;34:899 –904.

3. Halberg FE, Wara WM, Fippin LF, Edwards MSB, Levin VA,Davis RL, et al. Low-dose craniospinal radiation therapy formedulloblastoma. Int J Radiat Oncol Biol Phys 1991;20:651– 4.

4. Cohen BH, Zeltzer PM, Boyett JM, Geyer JR, Allen JC, FinlayJL, et al. Prognostic factors and treatment results for supra-tentorial primitive neuroectodermal tumors in children us-ing radiation and chemotherapy: a Children’s Cancer GroupRandomized Trial. J Clin Oncol 1995;13:1687–96.

5. Dirks PB, Harris L, Hoffman HJ, Humphreys RP, Drake JM,Rutka JT. Supratentorial primitive neuroectodermal tumorsin children. J Neurooncol 1996;29:75– 84.

6. Chang CH, Housepian EM, Herbert C. An operative stagingsystem and a megavoltage radiotherapeutic technique forcerebellar medulloblastoma. Radiology 1969;93:1351–9.

7. Kaplan EL, Meier P. Nonparametric estimation from incom-plete observations. J Am Stat Assoc 1958;53:457– 81.

8. Cox DR. Regression models and life tables. J R Stat Soc (B)1972;34:187–220.

9. Garton GR, Schomberg PJ, Scheithauer BW, Shaw EG, Il-strup DM, Blackwell CR, et al. Medulloblastoma – prognos-tic factors and outcome of treatment: review of the MayoClinic experience. Mayo Clin Proc 1990;65:1077– 86.

10. Hershatter BW, Halperin EC, Cox EB. Medulloblastoma: theDuke University Medical Center experience. Int J RadiatOncol Biol Phys 1986;12:1771–7.

11. Hughes EN, Shillito J, Sallan SE, Loeffler JS, Cassady JR,

Tarbel NJ. Medulloblastoma at the Joint Center for Radia-tion Therapy between 1968 and 1984. The influence of ra-diation dose on the patterns of failure and survival. Cancer1988;61:1992– 8.

12. Merchant TE, Wang MH, Haida T, Lindsley KL, Finlay J,Dunkel IJ, et al. Medulloblastoma: long-term results forpatients treated with definitive radiation therapy during thecomputed tomography era. Int J Radiat Oncol Biol Phys1996;36:29 –35.

13. Silverman CL, Simpson JR. Cerebellar medulloblastoma: theimportance of posterior fossa dose to survival and patternsof failure. Int J Radiat Oncol Biol Phys 1982;8:1869 –76.

14. Khafaga Y, Kandil AE, Jamshed A, Hassounah M, DeVol E,Gray AJ. Treatment results for 149 medulloblastoma pa-tients from one institution. Int J Radiat Oncol Biol Phys1996;35:501– 6.

15. Miralbell R, Bleher A, Huguenin P, Ries G, Kann R, Miri-manoff RO, et al. Pediatric medulloblastoma: radiationtreatment technique and patterns of failure. Int J RadiatOncol Biol Phys 1997;37:523–9.

16. Carrie C, Alapetite C, Mere P, Aimard L, Pons A, Kolodie H,et al. Quality control of radiotherapeutic treatment of me-dulloblastoma in a multicentric study: the contribution ofradiotherapy technique to tumour relapse. Radiother Oncol1992;24:77– 81.

17. Belza MG, Donaldson SS, Steinberg GK, Cox RS, Cogen PH.Medulloblastoma: freedom from relapse longer than 8 years– a therapeutic cure ? J Neurosurg 1991;75:575– 82.

18. Tomita T, McClone DG. Spontaneous seeding of medullo-blastoma: results of cerebrospinal fluid cytology and arach-noid biopsy from the cisterna magna. Neurosurgery 1983;12:265–7.

19. Fukanaga-Johnson N, Lee JH, Robertson P, Sandler HM,McNeil E, Goldwein JW. Patterns of failure following treat-ment for medulloblastoma: is it necessary to treat the entireposterior fossa? Int J Radiat Oncol Biol Phys 1997;39S:144.

20. Ayan I, Kebudi R, Bayindir C, Darendeliler E. Microscopiclocal leptomeningeal invasion at diagnosis of medulloblas-toma. Int J Radiat Oncol Biol Phys 1997;39:461– 6.

21. Yao MS, Mehta MP, Boyett JM, Li H, Donahue B, Rorke LB,et al. The effect of M-Stage on patterns of failure in posteriorfossa primitive neuroectodermal tumors treated on CCG-921: a phase III study in a high-risk patient population. IntJ Radiat Oncol Biol Phys 1997;38:469 –76.

148 CANCER July 1, 1999 / Volume 86 / Number 1