Spinal Ependymomas. Part 1: Intramedullary Ependymomas
This study describes results for 100 patients with intramedullary ependymomas treated at Nordstadt Hospital Hannover, Germany, from 1980 to 2003 and the Christliche Krankenhaus Quakenbrück, Germany, from 2004 to 2014. Ependymomas accounted for 40% of all intramedullary tumors operated on in that 34-year period. They are the most common intramedullary tumor in adults; only 3 children presented with this tumor type in this series.
The introduction of MRI has revolutionized the diagnosis of intramedullary tumors. The neuroradiological features of ependymomas are quite variable. Although most ependymomas are midline tumors (Figs. 1, 3, and 5), some may display an eccentric growth pattern, even with an extramedullary component (Fig. 2). They may appear solid and well demarcated with homogenous contrast enhancement (Fig. 5) while others display a patchy contrast uptake (Fig. 2) or appear cystic (Fig. 3). In this series, 24.5% of ependymomas displayed evidence of tumor hemorrhages, either as hemosiderin caps on either tumor pole (Fig. 5) or as intramedullary hemorrhages into an accompanying syrinx (Fig. 3) or the tumor itself. The only sudden neurological deterioration in this series was observed with such an intratumoral hemorrhage in 1 patient. An associated syrinx was detected in 58% of patients.
The rarity of intramedullary tumors, in general, and their slow-growing nature may explain the rather long average preoperative history of almost 3 years before presentation. Due to advances in MRI, patients present with less neurological deficits and lower McCormick scores nowadays compared with the 1980s (Table 3). Prior to 1990 (Period A), the majority of patients presented with McCormick Grades II and III, with 24% of them no longer able to walk by the time of diagnosis. Since 1991, the great majority of patients are diagnosed in Grade I, with only 4%-6% of patients wheelchair dependent by the time of presentation.
Another interesting aspect is the trend for more severe neurological deficits associated with ependymomas of the thoracic cord and conus compared with cervical lesions (Table 2 and Table 8). By comparison, cervical tumors tended to produce more sensory deficits and dysesthesias. The larger size of the cervical cord compared with the thoracic cord may explain these effects, as tumors had to reach larger volumes in the cervical cord to produce severe neurological deficits.
Every operation on an intramedullary tumor should attempt a GTR. In this series, 86.3% of ependymomas underwent GTR, with a steady increase of this rate throughout the overall study period. The major reasons for the steadily increasing GTR rates appear to be the improvements in preoperative MRI as well as the intraoperative ultrasound visualization, the introduction of small ultrasonic aspirators, and increasing experience of neurosurgeons. The influence of experience on GTR rates is evident by comparing results for the 3 different groups of surgeons in this study.
The multivariate analysis disclosed a number of independent factors predicting a GTR. A low McCormick grade and operating on a de novo rather than a recurrent ependymoma were the strongest predictors for a GTR. Other independent factors favoring a GTR were absence of arachnoid adhesions between cord and dura, a short clinical history, and an experienced surgeon (Table 4). In other words, GTRs were more likely in patients with a less compromised spinal cord operated on by an experienced surgeon.
A few comments on the surgical strategy may be warranted here (Figs. 1 and 2). Obviously, monitoring of somatosensory-evoked potentials may aid the surgeon to preserve sensory functions during myelotomy and further dissection and is highly recommended. It will not only help to preserve sensory functions but also indicate how an individual tumor can and should be handled throughout the remaining operation. It was the general policy of the 2 institutions involved in the series to open the spinal cord over the entire tumor before starting any further dissection or debulking. Pial retention sutures were applied in most instances to keep the spinal cord open so that further dissection could concentrate on tumor dissection and removal without having to open the cord each time for every surgical step. All tumors were debulked first. Quite often, the tumor became better demarcated from normal cord tissue once a considerable amount was removed and the tumor partially devascularized. When the anterior surface of the tumor was reached, monitoring was switched to motor-evoked potentials. The characteristic feature of ependymomas is their main vascularization from small branches of the anterior spinal cord artery. Identification, coagulation, and transsection of these feeders were the last steps of tumor removal and required utmost care. Anesthetists were instructed to keep the mean arterial pressure high enough during this part of the surgery in particular. Excessive tension must not be applied to these feeders to avoid undue tension on the anterior spinal cord artery. Feeders were coagulated and cut close to the tumor rather than close to the artery. Finally, bipolar coagulation should be used as little as possible inside the spinal cord, applying repeated irrigation for hemostasis instead.
Complications such as postoperative hemorrhages, problems related to wound healing, or infections were observed at a rate of 16.7% and treatable without long-term sequelae provided they were managed appropriately and in time.
No anesthesiological complications occurred among patients operated on in the semisitting position. Risks for air embolism can be minimized by careful dissection of neck muscles and covering the lateral epidural spaces with moist cottonoids. This is particularly helpful during dura closure, as epidural veins, which are no longer under compression from the neoplasm, may start bleeding due to manipulations of the dura.
Almost all patients experienced a worse neurological status as soon as they woke up from anesthesia. This immediate postoperative worsening involved sensory functions for almost all patients. Postoperative sensory deficits are an inevitable consequence of the myelotomy required to reach an intramedullary tumor. The length of the myelotomy was found to correlate with postoperative sensory dysfunctions in several studies. The only tumors which could be removed without inflicting such sensory disturbances were those that had reached the pia mater or grown out of the cord in exophytic fashion. In such cases, the tumor could be entered directly without requiring a myelotomy (Fig. 2). During the postoperative course, some improvement of sensory symptoms did take place. However, in the great majority, this recovery was incomplete and ended within 6 months after surgery (Table 8). Apart from impaired sensations for light touch, temperature, and pain, which may render patients vulnerable to unnoticed injuries, specific problems related to impaired joint position sense are common.
For this study, surgical morbidity was defined as a postoperative worsening of motor power, gait, or sphincter functions irrespective of postoperative changes for sensory functions or pain. In this series, 67.6% of all patients experienced such surgical morbidity to some degree. This deterioration was transient for 40.1% of patients and permanent in 27.5%. Postoperative improvements were sometimes observed, but keeping the preoperative functional state was the realistic prognosis for the great majority of patients (Table 8). Indeed, the majority of patients keep their preoperative functional state, and only few patients improve on it. This remains the strongest argument to recommend surgery early. A watch-and-wait policy cannot be supported.
Further analysis showed a trend for higher rates of postoperative transient morbidity with increasing age. Elderly patients demonstrated more difficulties to overcome deficits of joint position sense in particular. Transient postoperative neurological deteriorations were more common after GTRs than partial resections and after operations by experienced surgeons of Group C, who were more likely to remove a tumor completely. The time for recovery was quite variable and required intensive, professional rehabilitation and active training by the patient. It took up to 6 months for the majority, but continued for a year or even longer in some patients.
The rather high rate of permanent surgical morbidity for intramedullary ependymomas is disturbing. This rate of 27.5% is by far the highest compared with all other intramedullary tumor types. Publications reporting permanent morbidity rates for intramedullary tumors are few, not to mention rates for separate entities. For all histological types combined, rates between 13.9% and 34.6% have been published in the literature. Fischer and Brotchi documented a postoperative increase in the McCormick grade 1 year after surgery in 19% of patients for all histological entities combined. Three previous publications on large series of ependymomas reported rates of 4.5%, 13.9%, and 35.8% for patients with a worse postoperative McCormick grade after surgery. Thus, the rate of 19.6% of patients losing a McCormick grade with surgery in this study is in the reported range. The rate of 4.5% reported by Li et al., with only 13.3% of patients experiencing new postoperative sensory deficits, appears exceptional but may, in part, be explained by inclusion of myxopapillary ependymomas as intramedullary tumors in their study.
Several factors were found to influence permanent morbidity independently (Table 7); surprisingly, the preoperative neurological state was not among them. A low spinal level of the tumor, advanced age, a long clinical history, and tumor hemorrhages had the strongest predictive power, apart from operating on a recurrent tumor. The higher permanent morbidity for thoracic ependymomas may be a consequence of their vascularization pattern mentioned above and the more vulnerable vascular supply of the thoracic cord compared with that of the cervical cord. The importance of maintaining a sufficient arterial blood pressure during tumor removal cannot be overemphasized. Furthermore, thoracic ependymomas presented with more severe neurological deficits resulting more often in a McCormick Grade II or III before surgery. Fischer and Brotchi and Stein and McCormick mentioned the conus area as a particularly dangerous region; otherwise, this increased morbidity of thoracic ependymomas has not been described previously. The increased morbidity for elderly patients may be explained by a loss of spinal cord plasticity and mental factors. Patients of advanced age often have larger difficulties to overcome deficits of joint position sense in particular. Finally, tumor hemorrhages and surgery on recurrent tumors render dissection more difficult due to formation of gliotic tissue in areas of former bleeding or surgery.
Does limiting tumor removal to a subtotal resection reduce permanent morbidity? The data from this study do not support this assumption (Table 6), bearing in mind, however, that the number of patients undergoing subtotal or partial removals was low. The factors affecting permanent morbidity for intramedullary ependymomas are not different from those for other histological types. The risks, however, appear to be higher in general. The best explanation for this appears to be the vascularization pattern of ependymomas. Applying too much tension on the anterior spinal cord artery and damaging perforating arteries during dissection are more likely to occur in this tumor entity compared with any other. This concerns thoracic tumors in particular. Paying attention to these factors, securing a sufficiently high blood pressure throughout dissection, and monitoring motor-evoked potentials reduce these risks considerably. Furthermore, thoracic ependymomas, in particular, should be brought to surgery before significant deficits have developed.
This study, as others before, has demonstrated that recurrence rates depend on the amount of tumor resection. A GTR of an ependymoma was a curative operation for 95.8% of patients in this study (Fig. 4). As one would expect for a benign tumor entity, mortality figures are low, with 92% surviving at least 20 years. Among 7 deaths in this series, only 3 were related to the ependymoma that was operated on, while 4 patients died of other tumors associated with NF2 or unrelated causes.
However, the clinical outcome does not depend on tumor recurrences and immediate postoperative morbidities alone. Even patients who underwent a GTR and made a gratifying neurological recovery may suffer from significant neuropathic pain syndromes or develop a severe myelopathy at a later stage.
In this study, 4% of patients developed a postsurgical myelopathy, which usually started a few months after surgery. This delayed complication was observed exclusively in patients with MRI evidence of postoperative tethering of the spinal cord to the dura (Fig. 5). During the overall study period, the incidence of such tetherings could be reduced significantly with pial sutures, whereas duraplasties had no protective effect in this respect. Goto et al. likewise limited postsurgical tethering by combining pial suturing with a duraplasty. Therefore, pial sutures are recommended after GTR or a subtotal removal; they never led to syrinx formation in this series. No attempt was undertaken to untether a spinal cord in patients affected by a myelopathy. Such an attempt on an often very atrophic cord carries considerable risks. Furthermore, judging from experiences on patients undergoing a second operation for arachnolysis and untethering for syringomyelia, the probability of postoperative retethering and additional permanent morbidity is high.
Neuropathic pain syndromes are the other late postoperative complication, which may appear a few weeks to months after surgery. Thirty-seven percent of patients were affected in this study. Similar to the permanent surgical morbidity rate for ependymomas, this figure for neuropathic pain syndromes was the highest among all tumor types. Fischer and Brotchi observed long-term pain syndromes in 37.4% of patients following surgery of their intramedullary tumors. Other studies provided rates between 19% and 60%. In this study, only permanent types of pain requiring medication were classified as a neuropathic pain syndrome. Most patients decribed this pain as a burning type. The wide reported range in the literature may be explained by the lack of a precise definition of this syndrome. The pathophysiological mechanism is not well established either. The higher rates of neuropathic pain after GTR and in patients with permanent surgical morbidity as determined in the multivariate analysis suggest that surgical manipulation may play a role. Li et al. found neuropathic pain syndromes to be particularly common in tumors extending over more than 10 cm, thus pointing in a similar direction. The strongest independent factor predicting this syndrome in this study was the preoperative presence of neuropathic pain. In other words, the tumor itself may also produce this problem. Postoperative cord tethering, on the other hand, had the least influence of all risk factors listed in Table 9. Therefore, this syndrome may simply be the result of a severely damaged spinal cord due to tumor expansion and surgery. The analysis does not provide any clues as to how this often debilitating problem might be avoided. Fortunately, treatment with gabapentin or pentagabalin—drugs unavailable early during the overall study period—and anticonvulsants has been moderately successful in recent years.
Discussion
This study describes results for 100 patients with intramedullary ependymomas treated at Nordstadt Hospital Hannover, Germany, from 1980 to 2003 and the Christliche Krankenhaus Quakenbrück, Germany, from 2004 to 2014. Ependymomas accounted for 40% of all intramedullary tumors operated on in that 34-year period. They are the most common intramedullary tumor in adults; only 3 children presented with this tumor type in this series.
The introduction of MRI has revolutionized the diagnosis of intramedullary tumors. The neuroradiological features of ependymomas are quite variable. Although most ependymomas are midline tumors (Figs. 1, 3, and 5), some may display an eccentric growth pattern, even with an extramedullary component (Fig. 2). They may appear solid and well demarcated with homogenous contrast enhancement (Fig. 5) while others display a patchy contrast uptake (Fig. 2) or appear cystic (Fig. 3). In this series, 24.5% of ependymomas displayed evidence of tumor hemorrhages, either as hemosiderin caps on either tumor pole (Fig. 5) or as intramedullary hemorrhages into an accompanying syrinx (Fig. 3) or the tumor itself. The only sudden neurological deterioration in this series was observed with such an intratumoral hemorrhage in 1 patient. An associated syrinx was detected in 58% of patients.
The rarity of intramedullary tumors, in general, and their slow-growing nature may explain the rather long average preoperative history of almost 3 years before presentation. Due to advances in MRI, patients present with less neurological deficits and lower McCormick scores nowadays compared with the 1980s (Table 3). Prior to 1990 (Period A), the majority of patients presented with McCormick Grades II and III, with 24% of them no longer able to walk by the time of diagnosis. Since 1991, the great majority of patients are diagnosed in Grade I, with only 4%-6% of patients wheelchair dependent by the time of presentation.
Another interesting aspect is the trend for more severe neurological deficits associated with ependymomas of the thoracic cord and conus compared with cervical lesions (Table 2 and Table 8). By comparison, cervical tumors tended to produce more sensory deficits and dysesthesias. The larger size of the cervical cord compared with the thoracic cord may explain these effects, as tumors had to reach larger volumes in the cervical cord to produce severe neurological deficits.
Resection Results
Every operation on an intramedullary tumor should attempt a GTR. In this series, 86.3% of ependymomas underwent GTR, with a steady increase of this rate throughout the overall study period. The major reasons for the steadily increasing GTR rates appear to be the improvements in preoperative MRI as well as the intraoperative ultrasound visualization, the introduction of small ultrasonic aspirators, and increasing experience of neurosurgeons. The influence of experience on GTR rates is evident by comparing results for the 3 different groups of surgeons in this study.
The multivariate analysis disclosed a number of independent factors predicting a GTR. A low McCormick grade and operating on a de novo rather than a recurrent ependymoma were the strongest predictors for a GTR. Other independent factors favoring a GTR were absence of arachnoid adhesions between cord and dura, a short clinical history, and an experienced surgeon (Table 4). In other words, GTRs were more likely in patients with a less compromised spinal cord operated on by an experienced surgeon.
A few comments on the surgical strategy may be warranted here (Figs. 1 and 2). Obviously, monitoring of somatosensory-evoked potentials may aid the surgeon to preserve sensory functions during myelotomy and further dissection and is highly recommended. It will not only help to preserve sensory functions but also indicate how an individual tumor can and should be handled throughout the remaining operation. It was the general policy of the 2 institutions involved in the series to open the spinal cord over the entire tumor before starting any further dissection or debulking. Pial retention sutures were applied in most instances to keep the spinal cord open so that further dissection could concentrate on tumor dissection and removal without having to open the cord each time for every surgical step. All tumors were debulked first. Quite often, the tumor became better demarcated from normal cord tissue once a considerable amount was removed and the tumor partially devascularized. When the anterior surface of the tumor was reached, monitoring was switched to motor-evoked potentials. The characteristic feature of ependymomas is their main vascularization from small branches of the anterior spinal cord artery. Identification, coagulation, and transsection of these feeders were the last steps of tumor removal and required utmost care. Anesthetists were instructed to keep the mean arterial pressure high enough during this part of the surgery in particular. Excessive tension must not be applied to these feeders to avoid undue tension on the anterior spinal cord artery. Feeders were coagulated and cut close to the tumor rather than close to the artery. Finally, bipolar coagulation should be used as little as possible inside the spinal cord, applying repeated irrigation for hemostasis instead.
Surgical Morbidity
Complications such as postoperative hemorrhages, problems related to wound healing, or infections were observed at a rate of 16.7% and treatable without long-term sequelae provided they were managed appropriately and in time.
No anesthesiological complications occurred among patients operated on in the semisitting position. Risks for air embolism can be minimized by careful dissection of neck muscles and covering the lateral epidural spaces with moist cottonoids. This is particularly helpful during dura closure, as epidural veins, which are no longer under compression from the neoplasm, may start bleeding due to manipulations of the dura.
Almost all patients experienced a worse neurological status as soon as they woke up from anesthesia. This immediate postoperative worsening involved sensory functions for almost all patients. Postoperative sensory deficits are an inevitable consequence of the myelotomy required to reach an intramedullary tumor. The length of the myelotomy was found to correlate with postoperative sensory dysfunctions in several studies. The only tumors which could be removed without inflicting such sensory disturbances were those that had reached the pia mater or grown out of the cord in exophytic fashion. In such cases, the tumor could be entered directly without requiring a myelotomy (Fig. 2). During the postoperative course, some improvement of sensory symptoms did take place. However, in the great majority, this recovery was incomplete and ended within 6 months after surgery (Table 8). Apart from impaired sensations for light touch, temperature, and pain, which may render patients vulnerable to unnoticed injuries, specific problems related to impaired joint position sense are common.
For this study, surgical morbidity was defined as a postoperative worsening of motor power, gait, or sphincter functions irrespective of postoperative changes for sensory functions or pain. In this series, 67.6% of all patients experienced such surgical morbidity to some degree. This deterioration was transient for 40.1% of patients and permanent in 27.5%. Postoperative improvements were sometimes observed, but keeping the preoperative functional state was the realistic prognosis for the great majority of patients (Table 8). Indeed, the majority of patients keep their preoperative functional state, and only few patients improve on it. This remains the strongest argument to recommend surgery early. A watch-and-wait policy cannot be supported.
Transient Morbidity
Further analysis showed a trend for higher rates of postoperative transient morbidity with increasing age. Elderly patients demonstrated more difficulties to overcome deficits of joint position sense in particular. Transient postoperative neurological deteriorations were more common after GTRs than partial resections and after operations by experienced surgeons of Group C, who were more likely to remove a tumor completely. The time for recovery was quite variable and required intensive, professional rehabilitation and active training by the patient. It took up to 6 months for the majority, but continued for a year or even longer in some patients.
Permanent Morbidity
The rather high rate of permanent surgical morbidity for intramedullary ependymomas is disturbing. This rate of 27.5% is by far the highest compared with all other intramedullary tumor types. Publications reporting permanent morbidity rates for intramedullary tumors are few, not to mention rates for separate entities. For all histological types combined, rates between 13.9% and 34.6% have been published in the literature. Fischer and Brotchi documented a postoperative increase in the McCormick grade 1 year after surgery in 19% of patients for all histological entities combined. Three previous publications on large series of ependymomas reported rates of 4.5%, 13.9%, and 35.8% for patients with a worse postoperative McCormick grade after surgery. Thus, the rate of 19.6% of patients losing a McCormick grade with surgery in this study is in the reported range. The rate of 4.5% reported by Li et al., with only 13.3% of patients experiencing new postoperative sensory deficits, appears exceptional but may, in part, be explained by inclusion of myxopapillary ependymomas as intramedullary tumors in their study.
Several factors were found to influence permanent morbidity independently (Table 7); surprisingly, the preoperative neurological state was not among them. A low spinal level of the tumor, advanced age, a long clinical history, and tumor hemorrhages had the strongest predictive power, apart from operating on a recurrent tumor. The higher permanent morbidity for thoracic ependymomas may be a consequence of their vascularization pattern mentioned above and the more vulnerable vascular supply of the thoracic cord compared with that of the cervical cord. The importance of maintaining a sufficient arterial blood pressure during tumor removal cannot be overemphasized. Furthermore, thoracic ependymomas presented with more severe neurological deficits resulting more often in a McCormick Grade II or III before surgery. Fischer and Brotchi and Stein and McCormick mentioned the conus area as a particularly dangerous region; otherwise, this increased morbidity of thoracic ependymomas has not been described previously. The increased morbidity for elderly patients may be explained by a loss of spinal cord plasticity and mental factors. Patients of advanced age often have larger difficulties to overcome deficits of joint position sense in particular. Finally, tumor hemorrhages and surgery on recurrent tumors render dissection more difficult due to formation of gliotic tissue in areas of former bleeding or surgery.
Does limiting tumor removal to a subtotal resection reduce permanent morbidity? The data from this study do not support this assumption (Table 6), bearing in mind, however, that the number of patients undergoing subtotal or partial removals was low. The factors affecting permanent morbidity for intramedullary ependymomas are not different from those for other histological types. The risks, however, appear to be higher in general. The best explanation for this appears to be the vascularization pattern of ependymomas. Applying too much tension on the anterior spinal cord artery and damaging perforating arteries during dissection are more likely to occur in this tumor entity compared with any other. This concerns thoracic tumors in particular. Paying attention to these factors, securing a sufficiently high blood pressure throughout dissection, and monitoring motor-evoked potentials reduce these risks considerably. Furthermore, thoracic ependymomas, in particular, should be brought to surgery before significant deficits have developed.
Long-term Outcome
This study, as others before, has demonstrated that recurrence rates depend on the amount of tumor resection. A GTR of an ependymoma was a curative operation for 95.8% of patients in this study (Fig. 4). As one would expect for a benign tumor entity, mortality figures are low, with 92% surviving at least 20 years. Among 7 deaths in this series, only 3 were related to the ependymoma that was operated on, while 4 patients died of other tumors associated with NF2 or unrelated causes.
However, the clinical outcome does not depend on tumor recurrences and immediate postoperative morbidities alone. Even patients who underwent a GTR and made a gratifying neurological recovery may suffer from significant neuropathic pain syndromes or develop a severe myelopathy at a later stage.
In this study, 4% of patients developed a postsurgical myelopathy, which usually started a few months after surgery. This delayed complication was observed exclusively in patients with MRI evidence of postoperative tethering of the spinal cord to the dura (Fig. 5). During the overall study period, the incidence of such tetherings could be reduced significantly with pial sutures, whereas duraplasties had no protective effect in this respect. Goto et al. likewise limited postsurgical tethering by combining pial suturing with a duraplasty. Therefore, pial sutures are recommended after GTR or a subtotal removal; they never led to syrinx formation in this series. No attempt was undertaken to untether a spinal cord in patients affected by a myelopathy. Such an attempt on an often very atrophic cord carries considerable risks. Furthermore, judging from experiences on patients undergoing a second operation for arachnolysis and untethering for syringomyelia, the probability of postoperative retethering and additional permanent morbidity is high.
Neuropathic pain syndromes are the other late postoperative complication, which may appear a few weeks to months after surgery. Thirty-seven percent of patients were affected in this study. Similar to the permanent surgical morbidity rate for ependymomas, this figure for neuropathic pain syndromes was the highest among all tumor types. Fischer and Brotchi observed long-term pain syndromes in 37.4% of patients following surgery of their intramedullary tumors. Other studies provided rates between 19% and 60%. In this study, only permanent types of pain requiring medication were classified as a neuropathic pain syndrome. Most patients decribed this pain as a burning type. The wide reported range in the literature may be explained by the lack of a precise definition of this syndrome. The pathophysiological mechanism is not well established either. The higher rates of neuropathic pain after GTR and in patients with permanent surgical morbidity as determined in the multivariate analysis suggest that surgical manipulation may play a role. Li et al. found neuropathic pain syndromes to be particularly common in tumors extending over more than 10 cm, thus pointing in a similar direction. The strongest independent factor predicting this syndrome in this study was the preoperative presence of neuropathic pain. In other words, the tumor itself may also produce this problem. Postoperative cord tethering, on the other hand, had the least influence of all risk factors listed in Table 9. Therefore, this syndrome may simply be the result of a severely damaged spinal cord due to tumor expansion and surgery. The analysis does not provide any clues as to how this often debilitating problem might be avoided. Fortunately, treatment with gabapentin or pentagabalin—drugs unavailable early during the overall study period—and anticonvulsants has been moderately successful in recent years.
SHARE
