Role of spinal MRI in the follow-up of children treated for medulloblastoma




The purpose of the current study was to describe the usefulness of spinal magnetic resonance imaging (MRI) in children with medulloblastoma or primitive neuroectodermal tumor (PNET) of the posterior fossa.


Children consecutively diagnosed with medulloblastoma/PNET and followed in the Hospital for Sick Children/Toronto were identified. A homogenous cohort of children treated with craniospinal irradiation as part of their initial treatment was considered. Contrast-enhanced spinal MRIs done concomitantly with cranial MRIs (doublets) were reviewed. Recurrence was defined as any new abnormal lesion (in the brain or in the spine) in symptomatic or asymptomatic patients. Doublets after the first recurrence were excluded in the final analysis. The utility of a spinal MRI in the presence of a negative cranial MRI was assessed.


In all, 73 patients (21 females and 52 males; median age, 6.6 years, median follow-up time, 4.3 years) had at least 1 evaluable doublet during the follow-up period. Since concomitant cranial and spinal MRI was introduced as the standard evaluation for medulloblastoma/PNET in 1991, 286 doublets were evaluable. Fourteen spinal MRIs and 25 cranial MRIs showed new nodular or leptomeningeal lesions. In 2 patients, repeat MRIs ruled out recurrence (false-positive). All confirmed spinal recurrences were associated with intracranial recurrence. Of 261 doublets with negative cranial MRI, no new lesion was identified on spinal MRI.


An absence of progression on cranial MRI is highly predictive of absence of progression on spinal MRI. There is little evidence that surveillance spinal MRI (in children who underwent craniospinal radiation as part of their initial treatment) improves the detection of recurrences in children with medulloblastoma. Cancer 2006. © 2006 American Cancer Society.

Standard follow-up of children treated for medulloblastoma or primitive neuroectodermal tumors (PNET) of the posterior fossa includes regular cranial and spinal magnetic resonance imaging (MRI). Over the last decade, several retrospective studies have reported conflicting results and generated an extensive debate on the clinical relevance of follow-up imaging of the brain1–5 and it is still unclear whether systematic or more intensive MRI imaging surveillance of the brain is associated with a survival benefit. Although this test is often conducted in conjunction with an MRI of the spine, information pertaining to spinal imaging is rare and there is no clear consensus regarding the frequency and the clinical value of surveillance spinal MRIs. In 1994, recommendations of the Neurology and Tumor Imaging Committees of the Children's Cancer Group6 suggested yearly spinal imaging in the follow-up of patients with medulloblastoma. Recommendations for scanning in open or recently closed international and national protocols vary from 3 to 12 monthly spinal MRIs during treatment and the first 3 years after completion of therapy. However, published reports of 2 large prospective multicenter studies7, 8 suggest that the policy regarding surveillance imaging rarely follows these recommendations.

The purposes of the present study were: 1) to review all MRI scans of the spine in a population of children with medulloblastoma of the posterior fossa treated with up-front craniospinal radiation; and 2) to assess the usefulness of surveillance spinal MRI determined by the comparison of cranial findings when the test was performed concomitantly in the follow-up of children treated with medulloblastoma/PNET.


Patients were identified via the neuro-oncology and neurosurgery databases for the period 1985–2004. Eligibility criteria included all children treated at the Hospital for Sick Children with histologically proven medulloblastoma or PNET of the posterior fossa who received craniospinal irradiation as part of their initial treatment in order to collect information on a homogeneously treated cohort. Only patients with at least 1 concomitant cranial and spinal MRI during their follow-up were eligible. Cranial and spinal MRIs done within 48 hours were labeled doublets. In our institution, imaging protocols varied over time. However, by 1991 doublets replaced CT, myelography, and asynchronous MR scanning as the standard surveillance scanning practice for children with medulloblastoma. MRI scanning of the brain and the spine was performed according to the existing protocols at the institution. The frequency of spinal MRI was therefore highly variable, depending of the protocol and the follow-up for each patient. However, since 1991 concomitant MRI of the spine was routinely performed in case of recurrence. Therefore, since 1991 most documented recurrences had a doublet available for review.

Medical records of the selected patients were reviewed to collect information such as age at diagnosis, gender, initial staging, and symptoms at the time of recurrence, treatment modalities, data of last follow-up, and survival status.

Two investigators (M.S. and U.B.), blinded to patient information and clinical outcome, reviewed gadolinium (Gd-DTPA) contrast-enhanced spinal MRIs in the axial and sagittal planes. Incomplete or insufficient studies were excluded. Characteristics of spinal MRIs were described as follows: 1) no evidence of enhancement; 2) thin linear vascular enhancement or nerve root enhancement; 3) clumping of the roots; or 4) nodular lesions or diffuse leptomeningeal enhancement. Only the latter was considered a definitive pathologic feature for recurrence. Spinal recurrence was defined as any new nodular lesion or area of diffuse leptomeningeal enhancement. At the time a spinal MRI was interpreted as abnormal the concomitant cranial MRI was reviewed concurrently. In addition, all reports of cranial MRIs were reviewed and when new findings were reported, MRI images were also evaluated. Criteria for the diagnosis of recurrence in the brain on MRI included an increase in the size of residual tumor, any new area of enhancement at the site of the primary tumor, or evidence of tumor at sites previously free of disease. Changes in patterns of gadolinium enhancement alone were not considered evidence of recurrence. In addition, all immediately postoperative cranial images (computed tomography [CT] or MRI) were assessed to determine the extent of tumor resection and the initial MRI scan of the spine was reviewed for staging purposes. Staging was reported according to the criteria of Chang et al.9

Statistical Considerations

Radiologic findings were subsequently correlated with clinical data. The radiologic findings were considered false-positive when subsequent imaging studies and clinical course did not confirm progression. These false-positive doublets together with those doublets after the diagnosis of recurrence were not included in the final analysis.

After stepwise data collection, data were analyzed in order to determine the respective value of cranial and spinal MRI scanning. The presence or absence of recurrence on cranial and spinal MRI was labeled as positive or negative MRI, respectively. The added utility of a spinal MRI in the presence of a negative cranial MRI was assessed.


There were 73 patients who fulfilled the eligibility criteria and who, in addition to the baseline MRI of the brain and the spine, had at least 1 evaluable doublet (median, 3, range, 1–13) in the follow-up period. Overall, there were 286 doublets evaluated in this radiologic review. The earliest doublets included in our study are dated 1991. There were 21 females and 52 males; the median age at diagnosis was 6.6 years (range, 3–14.7 years).

According to the classification of Chang et al.,9 53 patients had M0 disease, 35 of them underwent macroscopic (gross) total resection (GTR), and 15 had a near total resection (less than 1.5 cm2 postoperative residue). Twenty (27%) patients were diagnosed with disseminated disease: 3 patients were staged as having M1 disease and 17 patients as having M2/3 disease. Only 6 of the 20 metastatic patients underwent GTR of their primary tumor and 4 underwent subtotal resection. At the time of the review the median follow-up of survivors was 4.7 years (range, 1.1–11.3 years).

All patients received postoperative craniospinal irradiation. In 64 patients, radiotherapy was combined with adjuvant chemotherapy. According to open protocols, the extent of surgical resection and initial staging, doses of craniospinal radiation varied from 23.4 Grays (Gy) to 25 Gy (19 children) and from 30 Gy to 45 Gy (53 children) with a boost to the posterior fossa up to a total dose of 54 Gy to 56 Gy (dose not available for 1 patient). Adjuvant chemotherapy consisted of ifosfamide, carboplatin, and etoposide (ICE) (n = 23) until 1998. Thereafter, average-risk patients received a combination of vincristine, lomustine, and cisplatin (n = 18)8 and high-risk patients received an etoposide, cisplatin, cyclophosphamide, and vincristine (n = 18)-based protocol (available at URL: [accessed February 28, 2006]). Five patients were treated with different chemotherapy regimens.

Tumor Recurrence

The detailed characteristics of the patients with recurrent disease and the sites of recurrence are shown in Table 1. Fourteen patients developed new spinal lesions on MRI. Two of them were subsequently reclassified as false-positive. Spinal recurrences occurred predominantly as either diffuse leptomeningeal enhancement (5 patients; see Fig. 1) or nodular enhancement (5 patients; see Fig. 2) involving the whole spine. Leptomeningeal coating limited to the posterior spine only was seen in 1 patient, whereas in the last patient the MRI showed enhancing foci in the cauda equina involving the roots.

Table 1. Patients with Recurrence, Characteristics and Tumor Sites
Patient no.SexAge at diagnosis, yearsStage at diagnosisExtent of resectionChemotherapyDose of radiationSite of recurrenceMonths from diagnosis to disease progression/recurrenceMonths from recurrence to deathAlive, follow-up time after recurrence
  • M indicates male; GTR, macroscopic (gross) total resection; ICE, ifosfamide/carboplatin/etoposide; F, female; CCG 9961, Children's Cancer Group protocol 9961; POG 9631, Pediatric Oncology Group protocol 9631.

  • Site of recurrence: 2 = tumor bed; 3 = cranial relapse; 4 = dissemination in brain and spine; M0, no metastasis; M1, microscopic tumor cells found in cerebrospinal fluid; M2, cranial gross nodular seeding; M3, gross nodular seeding in spinal subarachnoid space.

  • *


  • No data.

2M5.2M0GTRPOG 963136/18266 
3M5.8M0IncompleteOther35.2/18211Lost to follow-up 
5F4.6M3IncompletePOG 963136/1828 2 y
6M3.3M0GTRNo25/29.8314 6.2 y
7M6.5M0IncompleteCCG 989223.4/30.6318 1.5 y
9F8.9M3IncompletePOG 963136/18398 
10M5M3IncompletePOG 963136/183131 
16M5M3IncompletePOG 963139.6/14.4463 
19M7.3M0GTRCCG 989223.4/30.64*122 
21M4.6M0GTRPOG 996123.4/30.64*47 10 mo
22M8.1M0UnknownNoUnknown4*29Lost to follow-up 
Median in months (range)    15 (3–63)8 (1–38) 

All 12 patients with confirmed spinal recurrences had new findings on the concomitant brain MRI. Among the 2 patients with false-positive images, 1 patient had concomitant intracranial lesions. Patients with diffuse spinal leptomeningeal enhancement had diffuse cerebral leptomeningeal enhancement involving either whole brain (2) or cerebellum and brainstem (3). Nodular enhancement of the spinal cord was associated with nodular enhancement of cranial leptomeninges (5), either alone (1) or simultaneously with progression of residual tumor and a new nodule in cerebropontine angle (1), new nodular enhancement in tumor bed (2), or diffuse brain metastases (1). The patient showing leptomeningeal coating of the posterior spine had right frontal metastasis and suboccipital lesions. Finally, the patient with enhancing foci at the cauda equina had a right frontal metastasis.

Review of the radiologic reports and clinical notes revealed that 12 patients with negative spinal findings were diagnosed with progression or recurrence on a concomitant brain MRI. The radiologic review of the cranial images revealed recurrences in the primary tumor bed (n = 5) and isolated metastasis in the supratentorial region (n = 7).

Overall, none of the patients was found to have an isolated spinal recurrence and, consequently, there were 272 true-negative and 2 false-positive spinal MRIs. The median time to progression was 15 months (range, 3–63 months) after initial diagnosis. There was no recognizable pattern regarding age, staging, or treatment predisposing to a specific site of recurrence.

Table 2 shows that those patients with recurrence and those without recurrence were comparable in gender and age. Those with dissemination at the time of diagnosis were more at risk to develop recurrence. Of 20 patients diagnosed with M+ disease, 5 patients developed recurrent disease with diffuse (4 patients) or nodular (1 patient) leptomeningeal craniospinal dissemination, 3 patients with diffuse cranial leptomeningeal enhancement, and 2 patients experienced recurrence in the primary tumor bed. Ten metastatic patients did not experience recurrence after a median follow-up of 3.5 years (range, 1.1–8.6 years).

Table 2. Patient Characteristics with and without Recurrence
 Patients without recurrence (n = 49)Patients with recurrence (n = 24)
  1. GTR indicates macroscopic (gross) total resection; STR, subtotal resection; M0, no gross subarachnoid or hematogenous metastasis; M+, M1, M2, M3; M1, microscopic tumor cells found in cerebrospinal fluid; M2, cranial gross nodular seeding; M3, gross nodular seeding in spinal subarachnoid space.

Median age, y (range)7.6 (3–14)5.8 (3.3–14.7)
 GTR + STR3812
 >1.5 cm2 residue10
 No data 2
 GTR + STR21
 >1.5 cm2 residue00
>1.5 cm2 residue28
 Radiation only72
 Radiation and chemotherapy4222

Sixteen recurrences were detected on surveillance scans without any symptoms (no information in 1 patient). Seven of 24 patients with recurrent disease had new symptoms leading to imaging. The symptoms consisted of headache and vomiting (4 patients) with double vision (2 patients). One patient had a sudden onset of confusion. Two patients had spinal symptoms: both had back pain associated with gait disturbances in 1 and urinary retention in the second. Both had extensive spinal recurrence associated with multiple intracranial lesions. All but 1 symptomatic patient had disseminated recurrence. As noted in previous reports,10 the mean survival time of patients after diagnosis of recurrence was shorter in the symptomatic group as compared with the asymptomatic patients (2 months vs. 10 months).


Treatment of recurrence varied based on open protocols and shared decision-making between physicians and families from palliative treatment (2 patients) to chemotherapy (14 patients) and high-dose chemotherapy followed by autologous stem cell transplantation (4 patients) combined with resurgery (2 patients) or focal radiation (2 patients). One patient progressed rapidly after resurgery without further treatment and another patient was lost to follow-up after diagnosis of recurrence.

Eighteen patients died of disease progression. Two patients who were receiving palliative care at the time of the last visit were lost to follow-up. At the time of last follow-up, 4 patients were alive, including 3 in second remissions. Two of the surviving patients had isolated supratentorial metastasis 14 months and 18 months, respectively, after initial diagnosis of M0 medulloblastoma. Both were treated with high-dose chemotherapy followed by autologous transplant and focal radiation. They were alive 6.5 years and 2 years after recurrence, respectively.


One patient at the end of treatment developed a new hemorrhagic lesion within the fourth ventricle associated with a nodular lesion in the cervical spine. Simultaneously, the patient was found to have a systemic fungal infection and was treated accordingly. The patient was discharged on palliative care with the presumptive diagnosis of an early recurrence, but her condition eventually improved and she remained alive at last follow-up, 4.9 years after this episode. In retrospect, the craniospinal imaging was mimicking recurrence. However, even after a careful retrospective examination of the spinal and cranial scans, the reviewers were unable to distinguish between recurrence and infection/inflammation.

In another patient a nodular enhancing lesion was detected on the axial slide only (not on the sagittal slide) 2 years after diagnosis. The child did not receive any treatment; however, this lesion disappeared on follow-up. Even in retrospect, the detailed reevaluation of this MRI could not provide an explanation for this false-positive finding: it was neither a vessel nor a recognizable artifact.


The purpose of our study was to evaluate the role of systematic surveillance scanning of the spine during the follow-up of patients with medulloblastoma treated with craniospinal irradiation. We found that among 286 simultaneous cranial and spinal surveillance MRIs among 73 children, no isolated spinal recurrences were detected. Of the 12 true-positive spinal recurrences, all were associated with concurrent cranial recurrences demonstrated on concomitant brain MRI.1, 2

Figure 1.

Diffuse leptomeningeal recurrence (doublet at the time of recurrence of Patient 16 in Table 1). (A) Cranial magnetic resonance imaging (MRI), axial, T1 plus contrast. (B) Spinal MRI, sagittal, T1 plus contrast.

Figure 2.

Nodular leptomeningeal recurrence (doublet at the time of recurrence of Patient 24 in Table 1). (A) Cranial magnetic resonance imaging (MRI), sagittal, T1 plus contrast. (B) Spinal MRI, sagittal, T1 plus contrast.

It is important to emphasize that, due to the eligibility criteria of this study, all patients were treated with craniospinal radiation (±chemotherapy) after initial surgery at the time of diagnosis. Therefore, the results of this study may not be applicable to the population of younger children treated without radiation or with radiation limited to the posterior fossa.

There has been extensive debate on the role of surveillance scanning in medulloblastoma patients. To our knowledge, no study to date has addressed the role of systematic surveillance scanning of the spine. One may assume that spinal imaging complements routine MRI scan of the brain in order to increase the rate of detection of recurrent disease. This practice was introduced as the access to MRI scans was improving and most recent protocols recommend routine surveillance MRI of the spine, with wide variations in the recommendations between protocols. However, whether this practice is appropriate and based on clinical evidence is still uncertain. Anecdotal reports mention the occurrence of isolated spinal metastasis. Some of these reports precede the MRI era and may have missed radiologic findings, which would have been more obvious on modern MRI.11, 12 Other reports are poorly documented with respect to the techniques of detection used.13 A study from Wootton-Gorges et al.14 describes a 3.5-year-old patient with metastatic medulloblastoma (M3), treated with low-dose craniospinal radiation (23.4 Gy), who was alive without evidence of disease 5 years after an isolated spinal recurrence. This unusual combination may allow questions concerning the accuracy of the diagnosis of isolated spinal recurrence.

By contrast, recent retrospective reviews of large medulloblastoma series with a focus on the pattern of recurrence have not reported any occurrence of spinal recurrences in the absence of cranial disease.5, 15, 16 Data from prospective protocols are more difficult to interpret, as some do not provide detailed information regarding the sites of recurrence. A study from the Children's Oncology Group (COG)8 including 65 children with nondisseminated medulloblastoma reported 3 (of 14) patients with recurrences outside the primary tumor site without any further specification. A randomized study from the International Society of Paediatric Oncology (SIOP) and United Kingdom Children's Cancer Study Group (UKCCSG)7 including 179 eligible children with medulloblastoma M0-1 found 16 recurrences in the neuroaxis without evidence of recurrence in the posterior fossa. One patient was diagnosed with a recurrence based on positive cerebrospinal fluid cytology. Apart from this information, there is no further description of the neuroaxis recurrences. The predominant sites of failures in the German HIT-9117 study were distant within the central nervous system and described as either leptomeningeal spread or brain metastases. In the German HIT-91 study including children with disseminated medulloblastoma at the time of diagnosis, no isolated spinal spread was observed.

Single-institution series may benefit from higher detection rates of metastatic disease due to more robust and consistent investigations at the time of recurrence compared with reports of multi-institutional trials. They also allow a careful radiology review, whereas information on recurrences in cooperative clinical trials most often rely on voluntary reporting and do not include central radiology review. However, retrospective single-institution studies on surveillance imaging have usually included MRI and CT scanning and myelograms.1, 18 To the best of our knowledge, the current study is the first including concomitant MRIs only to evaluate the value of spinal MRI and the site of recurrence. MRI assessments without gadolinium or CT scans can easily miss leptomeningeal seeding at the level of the brain. Studies including follow-up imaging using different techniques or performed at a different time can also be misleading. Their findings with regard to primary site of recurrence may be questionable and conclusions have to be cautious. One may consider that there was an imbalance in the manner in which data were reviewed in our study, as all spinal imaging was systematically reviewed, whereas only imaging with abnormal reports were considered for cranial imaging review. This was due to the significant number of equivocal reports on spinal findings, such as “clumping of the roots” or “linear vascular enhancement.” By contrast, equivocal report on cranial MRI were rare. In our study, we found 2 false-positive spinal MRIs. One patient had concomitant findings in the brain. The second patient had an isolated spinal lesion detected on an axial view. A repeat MRI did not confirm this finding, and the diagnosis of recurrence was ruled out. False-positive findings have been described in the early postoperative period, either in the operation cavity or at the spinal level.19–21 This is not a valid explanation in our cases because the false-positive lesions in our patients were observed 21 and 22 months after initial surgery. False-positive findings can occur in patients suffering from infectious complications and this is applicable to 1 of the false-positive patients, whose imaging was mimicking recurrence.22 There is no clear explanation for the other false-positive in this experience. The radiologic characteristics are in agreement with recently described radiation-induced white matter changes, which have been described in the brain; however, no such finding has been reported at the spinal level.23

The implications of our findings are numerous. They question the relevance of a common practice, which does not seem to translate into a higher detection rate or into a benefit for the patient her/himself. The psychological stress on patients and families during follow-up is not insignificant, and follow-up tests that do not provide a real benefit should be reconsidered. Moreover, the addition of a spinal MRI significantly prolongs the exam, with the necessity of different surface coils to allow imaging the entire neuroaxis. To provide a reliable study, sedation or general anesthesia is often needed, adding a substantial risk to an otherwise safe examination.24 It also requires prolongation of sedation or general anesthesia for young children.

In summary, we found that a negative cranial MRI is highly predictive of a negative spinal MRI, and the rationale for routine surveillance spinal MRIs is not supported by our findings in patients with medulloblastoma treated with craniospinal irradiation. Our study has limitations, the first of which is the small number of patients. These results should be confirmed in larger prospective clinical trials, which should include a systematic central review of the scans at the time of recurrence to answer this important question. This is particularly important at a time where the newer average-risk medulloblastoma protocols are decreasing the dose of radiation to the spine from 24 Gy to 18 Gy.25