Comparison of CSF Cytology and Spinal Magnetic

Resonance Imaging in the Detection of Leptomeningeal

Disease in Pediatric Medulloblastoma or Primitive
Neuroectodermal Tumor
By Maryam Fouladi, Amar Gajjar, James M. Boyett, Andrew W. Walter, Stephen J. Thompson, Thomas E. Merchant,
Jesse J. Jenkins, James W. Langston, Aiyi Liu, Larry E. Kun, and Richard L. Heideman
Purpose: Leptomeningeal disease (LMD) significantly
affects the prognosis and treatment of pediatric patients
with medulloblastoma or primitive neuroectodermal
tumor (PNET). Examination of CSF for malignant cells,
detection of LMD on spinal magnetic resonance imaging
(MRI), or both are the methods routinely used to diagnose LMD. A recent study suggested 100% correlation
between CSF and MRI findings in children with medulloblastoma. To determine the validity of this hypothesis,
we compared the rate of detection of LMD between
concurrent lumbar CSF cytology and spinal MRI performed at diagnosis in patients with medulloblastoma
or PNET.
Patients and Methods: As a part of diagnostic staging, 106 consecutive patients newly diagnosed with medulloblastoma or PNET were evaluated with concurrent
lumbar CSF cytology and spinal MRI. CSF cytology was
examined for the presence of malignant cells and spinal
MRI was reviewed independently for the presence of LMD.

Results: Thirty-four patients (32%) were diagnosed

with LMD based on CSF cytology, spinal MRI, or both.
There were 21 discordant results. Nine patients (8.5%)
with positive MRI had negative CSF cytology. Twelve
patients (11.3%) with positive CSF cytology had negative MRIs. The exact 95% upper bounds on the proportion of patients with LMD whose disease would have
gone undetected using either CSF cytology or MRI as the
only diagnostic modality were calculated at 14.4% and
17.7%, respectively.
Conclusion: With the use of either CSF cytology or
spinal MRI alone, LMD would be missed in up to 14% to
18% of patients with medulloblastoma or PNET. Thus,
both CSF cytology and spinal MRI should routinely be
used to diagnose LMD in patients with medulloblastoma
or PNET.
J Clin Oncol 17:3234-3237. r 1999 by American
Society of Clinical Oncology.

staging of metastatic disease are of great prognostic importance in childhood brain tumors and are crucial for the
selection of effective treatment.
At present, the diagnosis of LMD is based on the use of
gadolinium (Gd)-enhanced spinal magnetic resonance imaging (MRI) and the cytologic examination of CSF for
malignant cells, which were obtained 2 to 3 weeks after
surgery. In retrospective analyses of adult patients with a
variety of primary or secondary CNS tumors and positive
lumbar CSF cytology, MRI revealed abnormalities consistent with LMD in up to 70% of patients at some point in their
clinical course.9-13 Although a recent pediatric study reported 100% correlation between MRI and lumbar CSF
cytology in patients with medulloblastoma,14 the number of
patients was small and the studies were not contemporaneous. To determine the relative roles of cytology and MRI for
the diagnosis of LMD, we investigated the impact of using
only one of the two modalities to diagnose LMD in a large
cohort of patients with PNET and medulloblastoma who had
undergone both diagnostic procedures.

EPTOMENINGEAL disease (LMD) occurs in up to

32% of children with primary CNS tumors at diagnosis.1 Metastatic spread along the leptomeninges has been
reported for virtually all types of CNS tumors but is most
frequent in embryonal tumors such as medulloblastoma and
primitive neuroectodermal tumor (PNET).2-4
Patients with CNS tumors who have LMD at diagnosis
have a worse prognosis and require more aggressive
therapy.5-8 Consequently, the definitive diagnosis and careful

From the Departments of Hematology-Oncology, Radiation Oncology, Pathology and Laboratory Medicine, Biostatistics and Epidemiology, and Radiology, St Jude Childrens Research Hospital; and Departments of Pediatrics, Radiation Oncology, and Radiology, University of
Tennessee, College of Medicine, Memphis, TN.
Submitted February 2, 1999; accepted May 27, 1999.
Supported in part by Cancer Center support grants no. PA30CA
21765 and P01 CA 23099 from the National Cancer Institute and by the
American Lebanese Syrian Associated Charities (ALSAC).
Address reprint requests to Maryam Fouladi, MD, St Jude Childrens
Research Hospital; Department of Hematology-Oncology, St Jude
Childrens Research Hospital, 332 N Lauderdale, Memphis, TN 38105;
email maryam.fouladi@stjude.org.
r 1999 by American Society of Clinical Oncology.
0732-183X/99/1710-3234

3234

PATIENTS AND METHODS

Between December 1989 and August 1998, 106 consecutively
accrued pediatric patients with primary CNS PNET or medulloblastoma
had concurrent lumbar punctures and Gd-enhanced spinal MRI as part
of their initial diagnostic work-up for the presence of LMD.

Journal of Clinical Oncology, Vol 17, No 10 (October), 1999: pp 3234-3237

Downloaded from jco.ascopubs.org on September 26, 2013. For personal use only. No other uses without permission.
Copyright 1999 American Society of Clinical Oncology. All rights reserved.

3235

LEPTOMENINGEAL DISEASE IN MEDULLOBLASTOMA/PNET

Patients eligible for the study were younger than 21 years of age, with
a histologically proven diagnosis of medulloblastoma or PNET. Two to
three weeks after the initial surgery, CSF cytology by lumbar puncture
was obtained up to 48 hours postGd-enhanced MRI as part of the
initial staging evaluation.
CSF samples were mixed with one drop of 22% bovine albumin
(Organon Teknika, Durham, NC), placed in a cytospin sample chamber,
and centrifuged at 1,000 revolutions per minute for 5 minutes (Shandon
centrifuge; Shandon, Cheshire, United Kingdom). A pathologist reviewed all slides for the detection of malignant cells. Both cytologic
studies and spinal MRIs were interpreted independently of each other.
Spinal MRI examinations were performed on three different Siemens
MRI machines over the 10-year period of this study, with software
provided by the same vendor. The software underwent a slow evolution
over this period. During the first 9 years, the examinations relied
primarily on postcontrast imaging, with contiguous 4- or 5-mm sagittal
images of the entire spine followed by routine transverse imaging below
the conus. During the last year, the examination consisted of 3- or 4-mm
overlapping sagittal images at 1- and 1.5- or 2-mm intervals, without
routine transverse images.

DISCUSSION

Statistical Considerations
Exact test and 95% confidence upper bounds, for the proportion of
patients in whom LMD would be missed using either MRI or CSF as the
only modality for diagnosis of LMD, were calculated using StatXact 3
(Statistical software for exact nonparametric Inference; CYTEL Software Corporation, Cambridge, MA). Logistic regression was used to
investigate the relationship between CSF positivity and volume of the
CSF test sample. SAS Release 6.12 (SAS Institute, Cary, NC) was used
for this analysis.

RESULTS

Among the 106 patients, there were 68 males and 38

females. The median age was 7.3 years (range, 0.3 to 19.8
years). Twenty-five patients had positive CSF cytology for
malignant cells at diagnosis, and 81 had negative cytology.
There were 22 patients with MRI evidence of LMD and 84
without. As listed in Table 1, 34 children (32%) were
diagnosed with LMD based on CSF cytology, spinal MRI, or
both. Overall, among the 106 patients, there were 21
discordant results. Nine patients (8.5%) with positive MRI
had negative CSF cytology. The exact 95% upper bound on
the proportion of patients with LMD who would have gone
undetected using CSF cytology alone is 14.4%. Twelve
patients (11.3%) with positive CSF cytology had negative
Table 1. Lumbar CSF Cytology and Spinal MRI Correlation in 106 Patients
With PNET/Medulloblastoma at Diagnosis

Positive CSF
Negative CSF
Total

MRIs. The exact 95% upper bound on the proportion of

patients who would have gone undetected using only MRI of
the spine is 17.7%.
CSF volume was available in 103 of the 106 patients in
the study; three patients did not have CSF volume data
available. Only three patients had CSF volumes less than 0.5
mL. The median CSF volume was 1.3 mL. The quartile
range for CSF volume (ie, 25% to 75%) was 1 mL to 2 mL.
There was no evidence of an association between the
volume of the CSF sample and the probability that the
sample was positive (P .7). Furthermore, the discordant
rates between CSF cytology and MRI did not differ significantly when CSF volume was greater than 0.5 mL and less
than 2.0 mL versus a CSF volume greater than or equal to
2.0 mL.

Positive
MRI

Negative
MRI

Total
No. of
Patients

13
9
22

12
72
84

25
81
106

NOTE. Total number of patients diagnosed with LMD by MRI only, CSF
cytology only, or both: 9 12 13 34.

Up to 32% of patients with PNET or medulloblastoma

have metastatic disease at diagnosis.1 The diagnosis of LMD
generally requires the detection of malignant cells in lumbar
CSF or the presence of metastatic disease by Gd-enhanced
MRI or computed tomography (CT) myelogram. The sensitivity and specificity of each diagnostic method is difficult to
estimate because there is no diagnostic standard short of
detection of LMD at autopsy.
Some series have suggested that CSF cytology obtained
from cisternal or shunt taps may be more sensitive than CSF
obtained from lumbar punctures in detecting LMD.15,16 In
contrast to largely retrospective studies of adults with a
variety of malignancies, we recently completed a prospective study that compared the findings of cytologic examinations of CSF obtained from concurrent lumbar and ventriculoperitoneal shunt taps in 52 pediatric patients with primary
CNS tumors as part of an initial diagnostic work-up or
during follow-up testing to detect LMD.17 A total of 90
paired shunt and lumbar CSF samples were analyzed.
Among the 90 paired samples in that study, malignant cells
were detected at a significantly higher rate in lumbar CSF
than in shunt CSF (P .0018, one-sided). When the analysis
was confined to only those patients with embryonal tumors,
malignant cells continued to be detected at a higher rate in
lumbar CSF samples than in shunt samples (P .0008,
one-sided). Thus, we concluded that lumbar CSF should
remain the specimen of choice for the routine cytologic
detection of malignant cells in the CSF of children with
LMD.
In recent studies, Gd-enhanced MRI has been found to be
more sensitive than CT myelogram in the detection of
LMD.18,19 Heinz et al18 reported the results of a prospective

3236

FOULADI ET AL

study of 33 patients with primary CNS tumors who had

Gd-enhanced MRI within 2 weeks of a CT myelogram.
Seven patients were found to have metastatic disease.
Although both MRI and CT myelogram detected metastases
in patients, the number of lesions detected by MRI was 24
compared with 15 detected by CT myelogram. These results
lead to the conclusion that MRI may be more sensitive than
CT myelogram in detecting LMD.18 This data has also been
confirmed in a larger study of 61 patients by Chamberlain.19
Most studies of the role of neuroimaging and CSF
cytology in the diagnosis of LMD have been retrospective
and have primarily reported the incidence of MRI or CT
abnormalities in patients with positive CSF cytologic results. In these early studies, which involved adults with a
variety of metastatic solid tumors, MRI revealed abnormalities in 23% to 71% of patients with positive results on CSF
cytology.9-13
The only study to address the correlation between CSF
cytology and demonstration of spinal metastases by Gdenhanced MRI in medulloblastomas is by Harrison et al.14
Of 26 patients with medulloblastoma, 10 patients had been
investigated with both CSF cytology and MRI of the spine
up to 8 months after diagnosis. Seven of these samples were
lumbar, seven were obtained intraoperatively from cisterna
magna, and three were from both. When comparing lumbar
CSF with spinal MRI, three cases demonstrated involvement
by both routes, and four were negative by both routes. The
authors concluded that there was a 100% correlation between lumbar CSF and spinal MRI. A significant limitation
to Harrisons study is that the sample size is too small to
draw a meaningful conclusion regarding the sensitivity of
these two modalities; among the 26 patients, only seven had
both lumbar CSF cytology and spinal MRI. Furthermore, the
CSF samples and spinal MRIs were not concurrent. The
interval between the two studies ranged from 6 to 198 days
(average, 65 days). Finally, the timing of the MRI was not
uniform, ranging from preoperative to postoperative (range,
8 days to 8 months).

In contrast, the current study involved 106 consecutive

patients with PNETs or medulloblastomas who had both
CSF cytology and Gd-enhanced MRI of the spine performed
as part of their initial staging evaluation in the early
postoperative period. There was no selection of patients by
symptoms of leptomeningeal disease, and the two diagnostic
methods were performed concurrently (within 48 hours of
each other), 2 to 3 weeks postsurgery. Based on our findings,
if CSF cytology had been the only diagnostic procedure
used, we estimate that up to 14.4% of patients with LMD
would have been missed. If only MRI of the spine had been
used to diagnose LMD, up to 17.7% of patients with LMD
would have been missed. Thus, the use of either diagnostic
modality alone could result in missing the diagnosis of LMD
in up to 14% to 18% of patients, which, in turn, could lead to
the potential under-treatment of a patient.
There was no association noted in this study between the
volume of CSF and the probability that the sample was
positive (P .7). Furthermore, the discordant rates between
CSF cytology and MRI did not differ significantly when
CSF volume was greater than 0.5 mL and less than 2.0 mL
compared with volumes greater than or equal to 2.0 mL.
However, it must be pointed out that the CSF volume was
not standardized because the data were collected retrospectively. A prospective study that uses larger volumes of CSF
needs to be performed to assess this question more accurately and in a more standardized fashion.
In conclusion, this study strongly suggests that neither
spinal MRI nor CSF cytology alone is sufficient for the
definitive diagnosis of LMD. We conclude that both CSF
cytology and spinal Gd-enhanced MRI are necessary to
maximize the potential identification of the presence of
LMD in patients with medulloblastoma or PNET.
ACKNOWLEDGMENT
The authors thank Jennifer Havens and Annemarie Fraga for
assistance in data collection and Wei Wang for assistance with the
statistical analysis.

REFERENCES
1. Packer RJ, Siegel KR, Sutton LN, et al: Leptomeningeal dissemination of primary central nervous system in childhood. Ann Neurol
18:217-221, 1985
2. Deutsch M, Riegel DH: The value of myelography in the
management of childhood medulloblastoma. Cancer 45:2194-2197,
1980
3. Dorwat RH, Wara WM, Norman D, et al: Complete myelographic
evaluation of spinal metastases from medulloblastoma. Radiology
139:403-408, 1981
4. Packer R, Sutton L, Goldwein J, et al: Improved survival with the
use of adjuvant chemotherapy in the treatment of medulloblastoma. J
Neurosurg 74:433-440, 1991

5. Packer RJ, Sutton LN, Elterman R, et al: Outcome of children

with medulloblastoma treated with radiation, cisplatin, CCNU, and
vincristine chemotherapy. J Neurosurg 81:690-698, 1994
6. Boyett J, Zetlzer P, Finlay J, et al: Progression free survival and
risk factors for primitive neuroectodermal tumors (PNET) of the
posterior fossa (PF)[medulloblastoma] in children: Report of the
Childrens Cancer Study Group (CCG) randomized trial, CCG 921.
Proc Am Soc Clin Oncol 14:147, 1995 (abstr 283)
7. Kuhl J, Berthold F, Bode U, et al: Preradiation chemotherapy of
children with poor prognosis medulloblastoma: Response rate and
toxicity of the ifosfamide-containing multidrug regimen HIT 88/89.
Am J Pediatr Hematol Oncol 15:567-571, 1993 (suppl A)

LEPTOMENINGEAL DISEASE IN MEDULLOBLASTOMA/PNET

8. Bailey C, Gnekow A, Wellek S, et al: Prospective randomised trial
of chemotherapy given before radiotherapy in childhood medulloblastoma: International Society of Pediatric Oncology (SIOP) and the
German Society of Pediatric Oncology (GPO)SIOP II. Med Pediatr
Oncol 25:166-178, 1995
9. Freilich RJ, Krol G, DeAngelis LM: Neuroimaging and cerebrospinal fluid cytology in the diagnosis of leptomeningeal metastasis. Ann
Neurol 38:51-57, 1995
10. Krol G, Sze G, Malkin M et al: MR of cranial and spinal
meningeal carcinomatosis: Comparison with CT and myelography.
AJNR Am J Neuroradiol 9:709-714, 1988
11. Chamberlain MC, Sandy AD, Press GA: Leptomeningeal metastasisa comparison of Gadolinium enhanced MR and contrast enhanced
CT of the brain. Neurology 40:435-438, 1990
12. Davis PC, Friedman NC, Fry SM, et al: Leptomeningeal
metastases: MR imaging. Radiology 163:449-454, 1987
13. Youssem DM, Patrone PM, Grossman RI: Leptomeningeal
metastases: MR evaluation. J Comput Assist Tomogr 14:255-261, 1990

3237
14. Harrison SK, Ditchfield MR, Walters K: Correlation of MRI and
CSF cytology in the diagnosis of medulloblastoma spinal metastsases.
Pediatr Radiol 28:571-574, 1998
15. Wasserman WR, Glass JP, Posner JB: Diagnosis and treatment of
leptomeningeal metastases from solid tumors: experience with 90
patients. Cancer 49:759-772, 1982.
16. Rogers LR, Duchesneau PM, Nunez C, et al: Comparison of
cisternal and lumbar CSF examination in leptomeningeal metastases.
Neurology 42:1239-1241, 1992
17. Gajjar A, Fouladi M, Walter A: Comparison of lumbar and shunt
cerebrospinal fluid specimens for cytologic detection of leptomeningeal disease in pediatric patients with brain tumors. J Clin Oncol 17:
1825-1828, 1999
18. Heinz R, Wiener D, Friedman H, et al: Detection of cerebrospinal fluid metastasis: CT myelography or MR? AJNR Am J Neuroradiol16:1147-1151, 1995
19. Chamberlain MC: Comparative spine imaging in leptomeningeal
metastases J Neurooncol 23:233-238, 1995