Primary spinal cord tumors (PSCTs) are one of the rarest categories of tumors, accounting for 4–8% of all central nervous system (CNS) tumors [13,17,26]. The incidence of pediatric PSCT (0.26 per 100000 person-years) is even lower than of the incidence among adults (0.74 per 100000 person-years) [28]. In addition to the rare incidence, the pathological presentation of pediatric PSCT varies widely. Because of these obstacles, there exists only literature comprised of a small series of patients [5,8,32] or a larger series of patients in multicenter studies of specific tumor types [15,24].

In the present study, we report a series of 126 pediatric patients with PSCTs who underwent surgery in a single institution. A descriptive analysis regarding symptomatology, tumor level, anatomical compartment, pathological diagnosis, and surgical outcomes was performed. In addition, we conducted detailed survival analysis according to particular anatomical compartments.

This study was approved by the Institutional Review Board of Seoul National University Hospital (H-1910-168-1074). Between October 1985 and December 2017, a total of 183 consecutive pediatric patients with spinal cord tumors were surgically treated, and their charts were retrospectively reviewed. The clinical presentation, radiographic imaging characteristics, surgical outcomes and pathological results were evaluated. Follow-up was conducted for up to 10 years. Patients with metastatic lesions in the spine were excluded from this study. Patients with multiple craniospinal tumors, usually associated with genetic syndromes such as neurofibromatosis (NF) or von Hippel-Lindau syndrome, were also excluded. Among patients with multiple lesions, most patients (26/27) had NF (NF-1 in nine patients, NF-2 in 14 patients, and undetermined type in three patients). Nonetheless, patients who had solitary spinal lesions in neuraxis were included in the study despite having genetic syndromes (n=5). Following medical chart and radiographic image review, 57 patients were excluded (metastasis in 25 patients, multiple craniospinal tumors in 27 patients, and incomplete data in five patients). Finally, a total of 126 patients were evaluated for the present study.

Magnetic resonance imaging (MRI) was adopted at our institution in September 1987. Before the adoption of MRI, PSCTs were diagnosed using computed tomography-myelography. The number of involved segments, involved spinal level, and anatomical compartment of the tumor were analyzed. The level of tumor involvement was classified into cervical, cervico-thoracic, thoracic, thoraco-lumbar, lumbar, lumbosacral, and sacral. Tumors involving almost the whole spinal cord were particularly classified as holocord tumors. The anatomical compartment was categorized into intramedullary (IM), intradural extramedullary (IDEM), and extradural (ED) compartments. Some ED tumors with extension to the paravertebral space were also categorized as ED tumors. Tumors extending beyond one anatomical compartment were described by including the whole area of the tumors (IDEM/ED, for example).

The extent of tumor removal was categorized into gross total removal (GTR; >95% of tumor removal), subtotal removal (STR; >80% of tumor removal), partial removal (PR; <80% of tumor removal), and biopsy [11,33]. The mode of bony removal was categorized into laminoplasty, laminectomy, (partial) hemilaminectomy, and corpectomy.

All tumor specimens were inspected by a neuropathologist. The classification of tumors was based on the 2016 World Health Organization (WHO) classification of tumors of the CNS [18]. Tumor biological behaviour codes /0 and /1 were classified as nonmalignant tumors, whereas tumors coded /3 were classified as malignant tumors. ID tumors were classified by a grading system. The old pathologic specimens were reviewed, and a molecular study was performed if possible.

The patient’s neurological status was assessed preoperatively and postoperatively at three months using the modified McCormick scale. The modified McCormick scale is as follows : grade I, neurologically intact, ambulates normally, may have minimal dysesthesia; grade II, mild motor or sensory deficit, patient maintains functional independence; grade III, moderate deficit, limitation of function, independent of external aid; grade IV, severe motor or sensory deficit, limitation of function with a dependent patient; and grade V, paraplegia or quadriplegia, even if there is flickering movement [20]. Progression-free survival (PFS), defined as the absence of any clinical or radiographic sign of recurrence of the tumors, and overall survival (OS) were estimated. OS and PFS were evaluated for total PSCT patients as well as for the IM and ED groups. Tumors spanning more than one anatomical compartment (such as IM/IDEM and IDEM/ED) were not analyzed separately for survival statistics.

To assess survival data for patients who were lost to follow-up, we requested and collected death registry data from the National Statistical Office.

In the present study, we reviewed a 32-year experience of surgical management of pediatric PSCT in a single institution. Although several studies comprising a large number of pediatric patients with IM PSCTs exist, studies regarding tumors involving entire spinal columns are quite rare [9,32]. To the best of our knowledge, the present study is the largest series comprising all primary neoplasms of the spinal axis with an evaluation of detailed survival analysis.

The mean duration of symptoms was 9.3 months in the present study. This is similar to most previous studies ranging from 6 to 12 months [8,11,12,15,16,29]. As Houten and Weiner [15] described, the symptom duration was significantly shorter in patients with malignant tumors in our study. It is noteworthy that the symptom duration was much shorter in patients with ED tumors (2.2 months) and longer in patients with IM tumors (17.5 months) than in the total study population. These differences might result from the high proportion of malignancy in ED tumors (70.2%) and the predominance of lower grade tumors in IM tumors (74.4%).

The majority of children diagnosed with PSCT presented with some type of symptom. Among the symptoms, motor weakness (80.2%), including gait disturbance, was most common. This is in contrast to some previous studies that described that the most common presenting symptomatology was neck or back pain [12,15,29,30]. However, others described motor weakness as the most common symptom [7,8,11,16,21]. Despite the discrepancy in dominant symptoms, the sensory symptoms in pediatric patients with spinal cord tumors are not as predominant as those in adult patients [8,23]. This may be because pediatric patients often cannot be aware or complain of sensory change, while adolescents or adults can [8]. Among the patients presenting with spinal deformity, most of them (7/10) had IM tumors. The prevalence of spinal deformity in patients with IM tumors (7/43, 16.3%) was similar to that in previous literature [1,32,33]. As described earlier, spinal deformity was more evident in patients with IM tumors than in patients with tumors in other compartments because of the inequality between signalling to both sides of the spine and paraspinal muscles [3]. There have been increasing issues related to postoperative progressive spinal deformity following laminectomy and irradiation [27,32,33]. Post-laminectomy kyphoscoliosis is beyond the scope of the present study; however, further studies are necessary. In ED tumors, the proportion of incidental masses on radiographic examination was significantly higher than that in IM tumors (26.3% vs. 0%, p<0.001). The high incidence of peripheral neuroblastic tumors among ED tumors characterized by large mediastinal or retroperitoneal masses could explain the incidental findings.

In contrast to the adult type, pediatric PSCTs tend to extend to multiple levels [22,30]. In the present study, approximately half (56.3%) of patients had tumors involving more than three levels, and nine patients (7.1%) revealed holocord tumors, which was similar to the previous literature [30]. Regarding the anatomical compartment, ED tumors were most common (45.2%), followed by IM tumors (34.1%) and IDEM tumors (12.7%) in the present study. The anatomical distribution of PSCTs was quite similar to those described in previous studies. Spacca et al. [30] reported in their study consisting of 134 patients that the anatomical compartment was ED in 39.5% and IM in 34.3% of patients. Wetjen and Raffel [31] reported ED tumors accounted for 34.5% and IM tumors accounted for 29.7% by collecting and assessing 10 studies. Although some studies reported that the IM compartment was most common, it seems obvious that, unlike adult counterparts, pediatric IDEM tumors constitute a small portion of PSCTs [3,12].

Regarding pathologic presentation, malignant tumors comprise half of all tumors. However, the proportion of malignancy significantly differs in terms of anatomical compartment. Although malignant tumors account for approximately half of IM tumors (tumors’ biologic behaviour code /3 in WHO classification of CNS tumors), approximately 70% of ED tumors are malignant. There was no malignant tumor in the IDEM location. These findings imply that PSCTs have distinct behavioural characteristics according to anatomical compartment and should be assessed separately. Similar to previous studies, the most common types of tumors were nerve sheath tumors (n=23), low-grade gliomas (n=20), and neuroblastomas (n=14). Astrocytic tumors such as pilocytic astrocytomas, diffuse astrocytomas, anaplastic astrocytomas, and glioblastomas were dominant among IM tumors (46.5%) [7,11,30].

Notably, the pathologic composition of pediatric spinal cord tumors is far different from that of adult tumors [19]. First, the majority of spinal cord tumors in adults are metastatic tumors. In the present study, however, approximately one-quarter (36/126) of all tumors were embryonal tumors (CNS or peripheral origin), and there were only 25 patients out of 183 patients on registry (before exclusion) with metastatic lesions (13.7%). Moreover, the majority of cases of metastatic lesions occurred from metastasis from the brain, not from solid organs. Second, meningiomas are encountered very rarely, in contrast to in adults, in whom meningiomas are one of the most common tumors [28]. There was no case of meningioma in the present study after exclusion of one patient with atypical meningioma because of multiple neuraxis tumors associated with NF-2. Furthermore, summing all cases of meningiomas in the literature about pediatric PSCT, there were only three cases among 255 patients (1.2%) [8,12,30,32]. Although the old research by Fortuna et al. [14] stated that meningiomas comprised 4.3% of childhood spinal tumors (sporadic in 21 cases, NF-associated in five cases), it seems much rarer than previously reported. Third, the histologic proportion in IM tumors also differs from that in adult IM tumors. Astrocytomas predominate over ependymomas, especially in younger children; gangliogliomas are more prevalent; and hemangioblatomas are found very rarely in the pediatric population [10,15,25]. Our study regarding IM tumors was also consistent with previous literature. Nonetheless, the proportion of ependymomas (including IM myxopapillary type) in IM tumors was very low (n=3, 7%) in the present study. Spinal cord ependymomas are known to be associated with NF-2 in the pediatric population, and mutations in the NF-2 transcript are frequently found even in sporadic cases [3,6]. Before exclusion, there were eight cases of IM ependymomas, which were the second most common IM tumors, as previously reported [3,11,12,30]. However, five cases of ependymomas were excluded from the study because of multiple lesions associated with NF-2 (n=4) and incomplete data (n=1).

Regarding clinical outcomes, 36 patients (28.6%) suffered disease progression, and 22 patients (17.5%) died from the disease. According to previous literature, progression rates and mortality rates of pediatric PSCTs range from 18.5–37.4% and 8.8–40.0%, respectively [5,8,29,30]. Regarding IM tumors, the 5-year OS rates (82%) and PFS rates (73%) in the present study were quite similar to those of a previous study reported by Constantini et al. [11] (OS rates, 76%; PFS rates, 71%). Consequently, the clinical outcomes of our study were in general agreement with those of previous studies, although literature with detailed survival statistics was hardly found due to the great diversity of spinal cord pathology. In the present study, we conducted survival analysis for the total group and subgroup analysis for IM and ED tumors separately. The major finding in the survival analysis was that patients with benign pathology who achieved GTR of tumors showed significantly better OS rates in the total group and in IM tumors and ED tumors separately.

The benefit of benign pathology on PFS rates was also obvious for total PSCTs, IM tumors, and ED tumors. Similarly, regarding the extent of resection, GTR of tumors showed significantly better PFS rates in the total group and the ED tumor group. For IM tumors, however, the PFS curve demonstrated a different pattern. The beneficial effect of GTR, which was statistically insignificant, completely disappeared after longterm follow-up in the IM tumor group. According to previous literature, the pathological composition is the main determinant for patient survival and tumor progression for pediatric IM tumors [11]. However, the effect on OS and PFS of radical resection of IM tumors seems less clear [11,15,16], although some researchers advocated beneficial effects of total excision of tumors [2,4,8,12]. Constantini et al. [11] reported that long-term PFS did not differ between groups with GTR and STR beyond 3 years postoperatively. These findings were similar to those of the present study. Therefore, for IM tumors, tumor grade seems to be a main determinant of disease progression, whereas the extent of removal had little effect.

It is noteworthy to consider the role of radical excision of tumors in patients with IM tumors. Although the role of GTR was limited in terms of tumor progression, GTR resulted in better OS rates in patients with IM tumors. If the GTR results in profound neurological deterioration after surgery, the attempt to achieve GTR may be questionable. In the present study, however, patients with IM tumors who achieved GTR did not show worse functional outcomes than patients who underwent moderate excision. Consequently, radical excision of IM tumors could be a reasonable option if feasible.

There are some limitations in the present study. First, the main limitation is the heterogeneity of pathology. This heterogeneity is inherent to pediatric PSCTs due to the rarity of this type of tumor, which is the main obstacle to the establishment of a treatment protocol. To overcome this limitation, we categorized and evaluated the tumors according to a particular anatomical compartment. However, we think that future multicenter studies with a larger number of patients are needed. Second, a number of patients were excluded from the study. Representatively, we excluded patients with multiple lesions along the craniospinal axis because the assessment of recurrence or cause of death may be confusing in those patients. We suppose these patients would have worse clinical outcomes than patients with solitary lesions. It would be beneficial to evaluate syndromic patients with multiple lesions separately afterward. Third, there are missing data due to the loss of medical charts or radiographic images before the installation of electronic medical records, picture archiving and communication systems in our institution. The lost data could result in bias in the evaluation. Fourth, because of some old data from before the era of practical use of MRI, in-depth radiographic analysis was not possible. This resulted from the enrolment of an earlier registry to evaluate the long-term outcomes of PSCTs.

PSCTs are uncommon in pediatric patients. The most common symptomatology was motor weakness, and the symptom duration was shorter in patients with malignant tumors. Malignancy was found in approximately half of all tumors, the proportion of which is different according to the anatomical compartment. The most common anatomical compartment was ED, followed by IM and IDEM. The most common tumors were schwannomas and neuroblastomas. Both the pathology and extent of resection had a decisive effect on OS. The extent of removal also showed a beneficial effect on PFS in the total group and ED tumors. In IM tumors, however, pathology seems to be a main determinant of PFS, whereas the extent of removal had little effect. Radical excision of IM tumors could be a viable option for survival improvement without an increased risk of worse functional outcomes.