Surveillance imaging in children with malignant CNS tumors: low yield of spine MRI
- 357 Downloads
Magnetic resonance imaging (MRI) is routinely obtained in patients with central nervous system (CNS) tumors, but few studies have been conducted to evaluate this practice. We assessed the benefits of surveillance MRI and more specifically spine MRI in a contemporary cohort. We evaluated MRI results of children diagnosed with CNS tumors from January 2000 to December 2011. Children with at least one surveillance MRI following the diagnosis of medulloblastoma (MB), atypical teratoid rhabdoid tumor (ATRT), pineoblastoma (PB), supratentorial primitive neuroectodermal tumor, supratentorial high-grade glioma (World Health Organization grade III–IV), CNS germ cell tumors or ependymoma were included. A total of 2,707 brain and 1,280 spine MRI scans were obtained in 258 patients. 97 % of all relapses occurred in the brain and 3 % were isolated to the spine. Relapse was identified in 226 (8 %) brain and 48 (4 %) spine MRI scans. The overall rate of detecting isolated spinal relapse was 9/1,000 and 7/1,000 for MB patients. MRI performed for PB showed the highest rate for detecting isolated spinal recurrence with 49/1,000. No initial isolated spinal relapse was identified in patients with glioma, supratentorial primitive neuroectodermal tumor and ATRT. Isolated spinal recurrences are infrequent in children with malignant CNS tumors and the yield of spine MRI is very low. Tailoring surveillance spine MRI to patients with higher spinal relapse risk such as PB, MB with metastatic disease and within 3 years of diagnosis could improve allocation of resources without compromising patient care.
KeywordsMRI imaging Brain neoplasms Child Childhood medulloblastoma Pineoblastoma Primitive neuroectodermal tumor
Standard care for patients with central nervous system (CNS) tumors now includes surveillance neuro-imaging and clinical evaluations at regular interval. Magnetic resonance imaging (MRI) has become the modality of choice over the last two decades due to relative high-resolution images and increased accessibility. However, there is no consensus on the optimal frequency of surveillance MRI. Furthermore, the benefits of including spine MRI as part of surveillance remains controversial, even for highly malignant tumors such as medulloblastoma (MB) . Therefore, current practice varies considerably between providers, institutions and protocols. Since most studies on surveillance imaging were conducted between 1980 and 2000, a period during which various changes were made with regards to treatment and MRI application, we evaluated the utility of surveillance MRI in children with CNS malignant tumors in a recent cohort.
After approval by the institutional review board, we retrospectively identified a cohort of children diagnosed with malignant CNS tumors from January 2000 to December 2011 at two pediatric neuro-oncology centers. The study included patients of 21 years of age or less who underwent at least one surveillance MRI. Patients with a diagnosis of MB, atypical teratoid rhabdoid tumor (ATRT), pineoblastoma (PB), supratentorial primitive neuroectodermal tumor (sPNET), supratentorial high-grade glioma (World Health Organization [WHO] grade III–IV), or CNS germ cell tumor and ependymoma (WHO II and III) were included. Patients were included regardless of their treatment. Patients with a malignant CNS tumor involving only the spine at diagnosis were excluded. Clinical characteristics, including age at diagnosis and relapse, tumor type, overall survival, relapse or progression time and site were recorded.
The total number of brain and spine MRI per patients was calculated. Brain and spine MRI studies conducted within a week to each other were considered as combined study. All MRI with report or clinical note suggesting relapses were individually reviewed to confirm the recurrence. Relapse was defined as the detection of a new lesion, new leptomeningeal disease, or a significant increase (more than 25 %) in size of a previously known lesion. When tumor recurrence was equivocal, relapse was confirmed by radiological follow-up, clinical evolution, and, when possible, pathology reports from relapse surgery. Patients were categorized as symptomatic if they had new symptoms or new clinical signs at the time of their MRI, and as asymptomatic if there was no apparent change in their clinical status.
Statistical analyses were performed using the Fisher’s exact test two-tailed and unpaired t test. Survival was calculated by the Kaplan–Meier made curves and comparisons were conducted using the log-rank (Mantel Cox) test. p values of <0.05 were considered significant for this study. SPSS Statistics, version 20.0 (IBM, Armonk, NY) was used for all analyses.
Patient demographics and total number of MRIs obtained
High grade glioma
Number of patients
N = 258
N = 34
N = 52
N = 39
N = 89
N = 25
N = 10
N = 9
Brain MRI obtained
Spine MRI obtained
Median # of brain MRI per patient (range)
Median # of spine MRI per patient (range)
Age at dx (years)
Total follow-up (months)
Detection rate of relapse
Surveillance MRI identifying relapse
High grade glioma
Identification of relapse
Spine MRI onlya
Site of relapse
Site of relapses
High grade glioma
First relapse (N)
Site of relapse (%)
All relapse (N)
Site of relapse (%)
Time to relapse (months)
>90 % of relapse (months)
Similar results were observed when considering all relapses (total 238). One patient with a known relapse of metastatic glioma had a further progression identified only in the spine; none of the patients with sPNET and ATRT had an isolated relapse occurring in the spine. Relapsed PB was associated with spinal involvement in all cases, with a quarter involving the spine only and detected by spinal MRI.
Timing of relapse
Relapses usually occurred during the first year with a median time of 12 months (range 0.3–137 months). Ninety percent of relapses occurred within 34 months (Table 3). Median time to relapse was significantly different between tumors type (p < 0.001). Germ cell tumors had the longest median time to relapse with 38 months and ATRT the shortest with 5.5 months. Median time to relapse in patients with isolated spinal recurrence was 23 months (range 9–137). Except for the one patient with germinoma, all other patients presented their spinal relapse within 3 years.
In 14 (12 %) patients clinical status at relapse was unknown whereas in 99 the information could be retrieved. At initial relapse, 52 patients (46 %) were asymptomatic, compared with 47 patients (42 %) who were symptomatic (Appendix 1 in Electronic Supplementary Material). Only one patient out of 12 (8 %) with isolated spinal recurrence was clearly symptomatic and complained about leg pain. Nine were asymptomatic and in two patients with isolated spinal recurrence no information was available. Patients who were symptomatic did not have a longer interval since their last MRI (mean 3.9 vs 4.8 months, p = 0.14), nor did they have earlier relapses (median time to relapse 12 vs 11 months, p > 0.8) or worse outcome when compared to asymptomatic patients (median overall survival 23 vs 27 months, p > 0.3). Patients with glioma more frequently had symptoms than those with other tumor types (68 vs 38 %, respectively, p < 0.003).
After their first relapse, 67 (59 %) patients underwent new treatments: chemotherapy (29 including 4 high-dose chemotherapy with autologous stem-cell support), surgery (6), radiosurgery (2), radiation therapy (7) and multimodal therapy (23). In 18 patients the decision was made to transition toward palliative care, 12 patients continued with treatment, eight were followed with closer interval MRI, and in eight patients no clear information on management could be found.
This study demonstrates that the vast majority of relapses are detected on brain MRI and that the overall detection rate of isolated spinal relapses is extremely low. The benefits of spine surveillance MRI depend on tumor type and specific risk characteristics. Few studies on surveillance neuro-imaging of CNS tumors have been conducted. Most have limited their evaluation to MB, and included patients who underwent MRI but also CT-scans or even myelography [2, 3]. Several studies have reported that relapses were associated with symptoms [3, 4, 5]. Saunders et al.  observed that only 19.6 % of patients were asymptomatic at relapse. In our cohort, most initial recurrences were actually asymptomatic (46 %). This discrepancy could be explained by differences in imaging modalities, as all the patients in the current study underwent MRI which can detect tumor progression earlier than CT-scan and myelography before the patient becomes symptomatic .
While surveillance MRI now allows early detection of relapse, some authors have argued that the clinical benefits are marginal . In our cohort, most relapse detection led to a change in patient management and, in more than half of the cases, a direct intervention was made. Patients with relapse can also be eligible for experimental treatment (phase I or II studies) when no standard curative treatment exists.
Even though surveillance neuro-imaging with scheduled MRI is now integrated to clinical and research practice, the rate of relapse detection is generally low. Two prior studies based on a heterogeneous population of low grade CNS tumors and CT imaging reported relapse detection rates of 1.59 and 2.8 % [5, 8]. In our study, the overall detection rate of relapse was higher at 8 % for brain MRI and 4 % for spine MRI. This can partly be explained by the exclusion of WHO grade I tumors that have a lower rate of recurrence .
Since the vast majority of relapses had intracranial progression (97 %) and isolated spinal relapses were rare, the benefit of surveillance spine MRI is low. Bartels et al. conducted a study were they evaluated the yield of spinal MRI for MB and sPNET patients treated with CSI . In their cohort of 73 patients from 1985 to 2004 no isolated spinal recurrence was detected, in accordance with two previous reports [6, 9], and therefore they questioned the usefulness of spinal MRI in this population. In our study, we did observe five cases of isolated spinal relapse or progression in patients with MB. Even though MB represent 42 % of isolated spinal relapses in our study, the detection rate was very low (7/1,000) due to the large number of spine MRI exams conducted as part of their neuro-imaging surveillance. Limiting spinal surveillance MRI to the first 3 years for patients without risk factors such as spinal metastasis at the time of diagnosis could possibly increase the detection rate. Yao et al.  observed a low incidence of isolated spinal relapse (2 %) in patients without metastasis at diagnosis, in contrast to those with positive CSF fluid and/or metastasis at diagnosis (9 %). One of the patients who received low dose craniospinal radiation had an isolated spinal relapse, suggesting that patients who received dose reduced radiation might be at higher risk and may benefit from surveillance spine MRI.
The majority of high-grade gliomas relapse locally [11, 12]. Spinal metastasis is rare and has been reported to occur in 0.4–2 % of adult patients with glioblastoma. This percentage might be higher in children. Heideman et al. reported that nine (26 %) patients out of 41 high-grade glioma had disseminated disease at relapse. Of those, three (7 %) had isolated spinal recurrence . Another study observed isolated spinal relapse in 6 % of their cohort . While some experts have recommended conducting regular spine surveillance MRI in patient with high-grade glioma, data from different studies including ours does not support this practice .
In a study on sPNET without metastasis at diagnosis, Hong et al.  reported a treatment failure rate of 48 %, with 42 % involving the primary site and only 3 % with isolated CSF and/or spinal relapse. Timmermann et al.  reported similar findings, in which sPNET patients had 35 (92 %) relapses involving the primary site. Only four patients developed isolated distant metastases, all of which had metastatic disease at time of diagnosis. Those reports are similar to our observations which suggest that sPNET relapse locally and the benefit of conducting regular spine MRI is low. Limited data are available with regards to relapses pattern of PB. Spinal relapses have been described but most studies reported patients who did not undergo spine MRI at the time of recurrence [18, 19]. Despite the small number of patients with PB, our study supports the use of spine MRI in patients with PB since all had involvement of the spine at relapse and 40 % had isolated spinal recurrence.
Among our small cohort of ATRT patients, we observed that half failed locally and half had a diffuse leptomeningeal relapse. Similarly, Chi et al.  reported that out of eight patients, three failed locally, two had distant metastases and three had disseminated disease. In a meta-analysis including ATRT patients, 42 patients (58 %) had diffuse leptomeningeal relapse involving the spine and three (4 %) appears to have presented isolated spinal relapse . However, due to a heterogeneous population and imaging modalities conclusion on the benefice of spine MRI is limited for ATRT patient. As observed previously and demonstrated in our study, ATRT relapsed within 3 years . Therefore, spine surveillance MRI after 3 years may not be useful.
Most ependymomas and germ cell tumors relapse locally, and as such, we observed a low detection rate of isolated spinal relapse (1 and 2 %, respectively). The benefits therefore, appear low but based on the current literature, the usefulness of surveillance spine MRI in this population remains unclear. Messahel et al. studied more than 100 cases of relapsed pediatric ependymomas and reported that 84 % had local progression at time of relapse. In their study, isolated spinal relapses represent 5 % of treatment failure. It is, however, unknown if those patients had metastatic disease or positive cerebrospinal fluid cytology at diagnosis . Spinal relapse in germ cell tumors is rare but has been described. In a study including 60 patients, one had an isolated spinal relapse . Kamoshima et al.  reported a series of 25 patients with relapsed germ cell tumors, six had isolated spinal relapse and the latest relapse occurred 109 months after diagnosis. They, therefore, recommended performing yearly brain and spine for at least 10 years.
Given the rarity of ATRT and PB, our sample size is small and limits the extend of our conclusion. Despite this, our study offers valuable information. In the context where detection rate of isolated spinal relapse is very low, systematic use of brain and spine MRI as part of surveillance neuro-imaging may be carefully considered. Financial impacts on health care system and individual are not negligible as a brain combined with a spine MRI can cost up to three times more than a brain MRI. The cost of a brain and spine MRI can vary from 1,000 USD to 30,000 USD depending on the institution not including anesthesia. Furthermore, long scan times (90 min) of combined brain and spinal MRI could increase risk of patient motion, particularly during the contrast-portion that occurs at the end of the study. Risks associated with sedation during an MRI study is generally low, but complication range from drowsiness in 20 % of children in the following days to more serious complications such as cardio-pulmonary arrest . Concerns have also been raised with regards to sedation and possible long-term cognitive sequelae in young children undergoing anesthesia . Based on our study, selecting high-risk patient groups could improve the yield of surveillance spine MRI without compromise in identifying relapse. Patients with PB, patients with metastasis at diagnosis, and patients within 3 years of tumor diagnosis could benefit from surveillance spine MRI. However, blanket spine surveillance MRI in all patients with CNS tumors, particularly those with a diagnosis of high-grade gliomas and sPNET, may not be efficacious. In these cases, a brain MRI could be conducted followed by a spine MRI if an intracranial recurrence is identified.
Our study demonstrates the utility of surveillance MRI in detecting isolated spinal relapse in asymptomatic patients is low. Based on our findings, a subset of patients with PB and MB with metastatic disease and within 3 years of initial tumor diagnosis may benefit most from routine surveillance spine MRI. However, routine spine surveillance MRI of all patients with CNS tumors, and in particular, those with high-grade glioma and sPNET, may not be warranted. In this group, spinal MRI could be considered in the event an intracranial recurrence is detected by surveillance brain MRI. Tailoring surveillance spine MRI to patients with higher spinal relapse risk could improve allocation of resources and reduce iatrogenic risks without compromising patient care.
Sébastien Perreault is a Beverly and Bernard Wolfe Pediatric Neuro-Oncology fellow at Lucile Packard Children’s Hospital at Stanford University. He received Grants from Justine Lacoste Fundation and Fonds de Recherche en Santé du Québec (FRSQ) (Bourses de formation en recherche post-diplôme professionnel/Fellowship).
Conflict of interest
The authors declare that they have no conflict of interest.