Previous studies have described the incidence of spinal dumbbell tumors as 13–18%1−3. In our study, the rate was 14.45%, which is consistent with these reports. Therefore, dumbbell tumors are not unusual. Moreover, many studies have reported that dumbbell tumors occur most commonly in the cervical spine and that most are neurilemmoma1,7−9. Our research showed that 63.51% of tumors were neurilemmoma, but there was no significant difference in the location of tumors between the cervical and thoracic segments. Of the 74 tumors, the following 4 were malignant (5.40%): hemangiopericytoma, hemangioblastoma, multiple myeloma and chondrosarcoma. This incidence was slighter lower than the incidence found in previous studies1,10. Moreover, the clinical symptoms depended on the location of the tumor and there was no obvious specificity, with symptoms including pain and weakness of the lower limbs.
In 1941, Eden was the first to classify spinal dumbbell tumors and this classification was widely recognized. However, due to the backwardness of medical technology at that time, this classification system did not clearly help with preoperative judgment of the relationship between tumors and adjacent structures, especially in the formulation of surgical strategies. In 2004, Asazuma et al. proposed a new classification system with nine categories called Toyama typing5. This approach was the first classification based on three-dimensional computed topography (CT) or magnetic resonance imaging (MRI). Because this classification is more explicit and convenient than Eden’s classification for our study, we adopted this classification for our research. Moreover, there have been other classifications proposed for dumbbell tumors. Jiang et al. developed a new classification system that included 7categories (types 1–7) and 2 foraminal modifiers for determining the surgical approach4. In addition, Liu et al. recommended a novel classification of spinal dumbbell tumors based on the characteristics of the surgical approach8. Their study aimed to help the surgeon select a suitable surgical approach. Regretfully, this was a retrospective study of a single surgeon’s experience and the validity has not been confirmed by large-scale samples or other scholars. To date, operative approaches to dumbbell tumors are still controversial, but the choices are mostly based on the surgeon’s familiarity with regional anatomy and om personal preference.
An ideal surgical strategy for dumbbell tumors should be able to reduce the recurrence rate and simultaneously avoid spinal deformities. All kinds of surgical approaches for dumbbell tumors including total resection of spinal dumbbell tumors have been implemented, for instance, a single posterolateral or anterolateral approach, a combined posterior and anterior approach and an anterolateral approach with the use of 2 stages. We believe that most dumbbell tumors can be satisfactorily resected with the single posterior approach; McCormick has also advocated that dumbbell tumors can be effectively managed with a single posterior approach combined with laminectomy11. Some scholars have reported good results with the use of an anterior approach for the total removal of cervical dumbbell tumors12. Lot and George used a lateral approach and vertebral artery ligation for complete resection13. In addition, Jiang et al. treated patients with an anterior followed by a posterior approach for gross total resection, but found a high rate of compliacations4. However, most surgeons are unfamiliar with the anatomy associated with these approaches. More importantly, tumor removal with the use of these approaches is associated with a risk of injuring the vertebral artery (VA) and the accessory or hypoglossal nerves. Yong huang et al. suggested that a combined anterior and posterior approaches only be performed when there is residual tumor, but this suggestion was not widely accepted by patients and required a long surgical time14. However, the posterior approach is popular and safe and has a low potential for damaging important structures. As long as the VA can be visualized and protected, a one-stage posterior approach can be performed and even extended to the ventral side of the vertebral artery14. For these reasons, we prefer the posterior approach for our patients.
In addition, for cervical dumbbell tumors, attention should be paid to the protection of the vertebral artery. There is usually a capsular layer between the neurilemmoma and the vertebral artery. As long as the tumors do not completely encapsulate the vertebral artery, most of them can be completely separated from the vertebral artery. In addition, when approaching the vertebral artery, venous hemorrhage is often obvious. Most of these cases are hemorrhage of the venous plexus around vertebral artery. For thoracic dumbbell tumors, most scholars have suggested a combined approach followed by thoracoscopic removal of the intrathoracic component is the standard procedure15,16. However, Y. Li et al. demonstrated that the posterior approach achieved better results and fewer complications compared to the combined surgical approach and they thought the posterior approach allowed for the interaction with familiar structures throughout the operation, requiring only a single incision and avoiding the need for a postoperative chest tube, which diminish potential complications such as postoperative pain, pulmonary dysfunction and infection17. In our study, 9 patients underwent second-stage thoracoscopic surgery because of large paravertebral tumor and none experienced complications after surgery.
Generally, GTR of spinal dumbbell tumors is ideal because it reduces the chance of recurrence. Some studies have reported an increased recurrence rate after STR and have strongly suggested complete resection1,4,18. Moreover, Klekamp et al. showed a recurrence rate of 10.7% at 5 years and 28.2% after 10 to 15 years in patients19. However, Jiang et al. suggested the recurrence rate of patients with GTR was only 2.8% with an average follow-up of 5years4. One of the reasons for the high recurrence rate is the tumor capsule. As Benzel et al. reported removing nerves from the tumor origin rather than nenucleation, can avoid the recurrence of tumors20. The other reason is the subtotal resection rate. Sometimes it is difficult to remove the tumor without damaging the entire nerve root, and it is even sometimes impossible to achieve radical resection. Previous studies have noted that GTR can be achieved in 86–95% of patients5,18,21. However, the risk of injury is sometimes high for radical resection and 20% of patients present with radicular deficits4.
In our study, 52.7% of patients underwent GTR, and the reasons for STR were as follows: 10 patients’ tumors extended too far into the intervertebral foramen and thus caused incomplete exposure; 6 patients’ tumors were too close to or enveloped the vertebral artery to avoid unnecessary bleeding; 6 patients’ tumors adhered tightly to nerve roots and were unresected; vertebral venous plexus hemorrhage resulted in unclear intraoperative visual field in 6 patients; the artery that provided blood supply to the tumor was bleeding in 1 patient; 2 patients’ tumors were located close to the thoracic or descending aorta; 3 patients’ tumors were widespread basement tumors or were widely distributed; and 1 malignant tumor invaded the surrounding tissues. However, only 17.57% of patients had tumor recurrence after an average 10-year follow-up, and this result was lower than the 28.2% recurrence rate reported by Klekamp et al. after 10 to 15 years and the 20% recurrence rate reported by Sung Mo Ryu et al. after 35-month follow-up19,22. In addition, only 9 patients had complications, but no major complications were observed and good neurologic outcomes were achieved. In a recent study, Sung Mo Ryu et al. noted that the overall neurologic outcomes were better in the STR group than in the GTR group22. Thus, we aimed to maximize tumor removal with functional preservation.
The PFS of our study was 82.43%, Based on the JOA score (50–100%), 85.13% of patients had a good outcome, which was slighter higher than the outcomes reported by K. Ito et al. and Sung Mo Ryu et al.22,23. Thus, it is significant to understand the risk factors for tumor recurrence and prognosis. The univariate analysis showed that recurrence and poor outcome were related to STR and tumor location, respectively (P < 0.05). However, multivariate regression analyses did not show any positive results. The difference between the two analyses may be due to the limitation of the small sample size in our study.