Introduction

Syringomyelia is defined as a cystic cavity within the spinal cord that slowly leads to chronic and sometimes irreversible myelopathy. This condition is very often associated to Chiari type I malformation (CMI), a pathologic caudal displacement of cerebellar tonsils through the foramen magnum and into the cervical canal. The mechanism of formation of the syrinx and the correlation with CMI has been poorly appreciated for a long time, and various treatments have been suggested accounting for nearly 15 different surgical approaches [1]. During the last decade, some authors elaborated theories [2–4, 6, 28–30, 32] centered on obstruction of cerebrospinal fluid (CSF) flow at the level of the foramen magnum. More recently, neuroradiological devices, such as cine-magnetic resonance imaging (MRI) for the study of CSF dynamics performed pre- and postoperatively, corroborated the surgical and pathological observations. Although no prospective randomized controlled study exists of the optimal surgical strategy for the treatment of syrinx and CMI, there is robust evidence from accumulated multicenter experience [7–11] that craniocervical decompression (CCD) represents a valid treatment. The goal of this approach is to enlarge the bony and dural posterior fossa and rebuild a new cisterna magna. While CCD has shown satisfactory results, there is still debate about the need for dura opening, the extent of the bony and dural decompression required, the need of arachnoid manipulation, and the method of dural closure. Unfortunately, many series in the literature compare patients with syringomyelia of different origin and analyze multiple surgical strategies together [12–14, 27].

Since 1996, we have operated on patients with CMI-associated syringomyelia using a standardized CCD and duroplasty with dura substitute. Here, we discuss our experience with our consecutive and uniformly treated series of patients with CMI associated syrinx, focusing on the surgical considerations and their clinical and radiological long-term results. We also performed a statistical analysis to investigate the role of preoperative predictors of outcome.

Patients and methods

We retrospectively reviewed our database concerning the period between 1996 and 2006. All of the patients with a diagnosis of syringomyelia and CMI were included. Because this analysis focused on adult patients, a lower age limit of 18 was set. Patients with associated hydrocephalus, brain tumors, history of severe head trauma, or previous cranial surgery, meningitis, and lumbo-peritoneal shunt were excluded. The data were collected by searching the hospital charts.

Preoperative and postoperative clinical evaluation

A senior neurosurgeon or neurologist performed the preoperative and postoperative neurological assessment. During the history gathering, attention was paid to the presence of other cases of syringomyelia and CMI among relatives. Because of the wide range of signs and symptoms, we preferred to summarize the clinical features using three classes: (1) posterior fossa overcrowding (PFO; headache, cervical pain, nystagmus, drop-attack, or syncope after effort); (2) long tract impairment (pain, weakness, atrophy, spasticity, sensory loss, and dysesthesias); (3) brainstem compression (BSC; cranial nerve impairment, hypoacusis, dysphagia, and sleep apnea). These symptoms were widely associated in many patients. The postoperative clinical evaluation was performed at 1, 3, and 6 months and yearly. Each patient’s clinical outcome was rated as “improved,” “unchanged”, or “worsened.” Anesthetic and early post-operative complications were recorded.

Neuroradiological preoperative and postoperative assessment

Preoperative neuroradiological assessment was achieved by 1.5 T or 1 T MRI of the entire axis. The measurement of the tonsillar herniation was accomplished using the standard basion-opisthion line as a reference for the foramen magnum plane. The degree of the tonsils’ downward displacement was classified as three types: (a) between foramen magnum and C1; (b) between C1 and C2; and (c) beyond C2. Other radiological features of CMI were also considered for diagnostic purpose: shape of tonsils (triangular or rounded); volume of the cisterna magna (reduced or absent); appearance of the perimedullary subarachnoid spaces. All of the possible associated malformations of the craniocervical region that could modify the surgical strategy were ruled out. The length of the syrinx was calculated on a T1 sagittal plane (TR 400–500 ms, TE 16–20 ms) to avoid considering edema or T2 hyperintensity related to compression of the cord as syrinx. The measurements of the syrinx were classified as cervical, cervico-thoracic, and holocord. The first postoperative MR was performed at 3 months, then at 6 months, and afterwards yearly. Syrinx modifications or absence of narrowing during follow-up were defined respectively as “collapsed” or “unchanged.” The time to narrowing of the syrinx following surgery was also recorded.

Surgical technique

The patient is placed in a prone position with the head in flexion. A midline incision from the inion to the arch of C2 is then performed. The dissection of the muscles is must be sub-periosteal, and the surgeon is careful not to cut the attachment on C2 to avoid weakening. A burr hole is placed at the level of the nuchal line on each side, and the craniectomy is then completed either by a high-speed drill or a rongeur. The craniotomy encompasses the foramen magnum and extends upward to the external occipital crest and inferior nucal line. The maximum diameter of the bone opening is 3 × 3 cm, but variations can exist according to the dimension of the posterior fossa and the degree of tonsillar descent. The posterior arch of C1 is eliminated in its most central part. Based on the position of the cerebellar tonsils, the posterior arch of C2 is sometimes also removed. The always-thickened craniocervical dural band is left in place and cut one layer during the opening of the dura to avoid accidental dural tearing and arachnoid damaging. Under the microscopic view, the dura is incised in a “Y” or “reversed L” shape (in case of asymmetrical tonsil’s descent) while trying not to cut the arachnoid. At this time, the collapsed cisterna magna is refilled by the CSF, and a normal pulsation of the tonsils ensues. Sometimes, we compressed the neck of the patient (as for a Queckenstedt test) to increase the intracranial pressure and favor the expansion of the subarachnoid space. We did not manipulate the subarachnoid space or the cerebellar structures. A patch of dura substitute is then tailored on the dimension of the dural opening and fixed with single stitches (3/0 silk or prolene).

Statistical analysis

We used the chi-square test to establish the presence of significant differences in the frequencies of preoperative tonsils’ descent and of the extension of syrinx (C, C-T, holocord) among patients grouped by clinical outcome (improved, unchanged, worsened) and radiological syrinx modification (collapsed, unchanged). Statistical significance was fixed at a p value of less than 0.05.

Results

Preoperative patients’ characteristics

During a period of 11 years (1996–2006), 39 patients were found to match the diagnosis of CMI associated syringomyelia. Two of these patients were not suitable for surgical treatment because of severe cardiovascular and liver dysfunction. One other patient refused the procedure. A total of 36 patients were treated by craniocervical osteo-dural decompression. There were 17 men and 19 women (female/male ratio 1.11) with a mean age of 40.4 (range 18–68). Long tract impairment was the most cited preoperative complaint (31 patients, 86.1%) followed by less frequent PFO (27.7%) and BSC (13.8%). Thirteen (36.1%) patients presented with combinations of symptoms. Table 1 summarizes the patients’ preoperative clinical and radiological characteristics as well as the intraoperative arachnoid tearing related complication rate. One patient presented a mild form of basilar invagination.

Table 1 Summary of the preoperative clinical and radiological characteristics (see text for abbreviation) and intraoperative arachnoid tearing related complication rate
Full size table

Surgical results

In all patients, an occipital craniectomy and C1 laminectomy was performed while paying maximal attention not to detach the muscles from C2 (Fig. 1). In two patients with a type C tonsillar position, a C2 laminectomy was also required. In some cases, a wide circular occipital sinus was encountered requiring a careful hemostasis either by cauterization or suture with 4/0 prolene. The exposure of the arachnoid plane demonstrated an opaque and thick arachnoid in some cases, but no attempt to dissect or open the arachnoid was made. The dura opening was covered with dura substitute in all cases. The mean time of the procedures was 2.3 h (range 1.5–4.4). There was no postoperative complication related to the surgical procedure (such as cerebellar contusion, hematoma of the surgical cavity, brainstem, or cranial nerves dysfunction). An accidental arachnoid tearing can occur during the opening of the dura mater. This event provokes an increase of the postoperative cervical and shoulder pain, prolonging the analgesic need and the hospital stay. Moreover, this is a predisposing factor to CSF leak that in our patients resolved with a lumbar external drainage in two cases and with surgical revision in one case. Their CSF samples did not show bacterial infections.

Fig. 1
figure 1

a Dissection of nuchal muscles is subperiosteal, and C2 attachments are left in place. Large arrow muscles attachment on C2; dashed arrow posterior arc of C1; square dashed arrow occipito-cervical band; thin arrow foramen magnum. b The arc of C1 is removed, and craniectomy is no larger than 3 cm in maximum diameters

Full size image

In two cases (5.5%), a re-operation was needed at 2 and 7 years. During the operation, both patients showed an unsatisfactory dura opening, hence a wider duroplasty was performed. Moreover, we documented that the dura manifested a very thick fibrosis around the insufficient dural openings (Fig. 2).

Fig. 2
figure 2

a Postoperative MR of a patient that experienced recrudescence of symptoms 2 years after first operation. Note the syringobulbia. b Intraoperative image. Once the previous dura patch was removed, a thick and stenosed dura was found. c Dura mater was removed, and a new and large patch was inserted. d The last MR shows rebuilding of a larger cisterna magna and the disappearance of syringobulbia

Full size image

There was no mortality in the postoperative period; and in the early (24 h) postoperative course, no anesthesiology complication was recorded.

Clinical and neuroradiological long-term outcome

During a follow-up ranging from 8 to 68 months (mean 40 months), 29 patients (80.5%) improved their symptoms. As already reported in other series (26), headache is the first symptom to disappear, sometimes as soon as the discharge. On the other hand, advanced dysesthesic pain or arm or limb weakness has the worst prognosis, and normally only a partial recovery is observed. Four patients (11.1%) presented no modification of symptoms after decompression. The two patients with a C2 laminectomy did not show any persisting cervical pain, and a plain radiographic film with dynamic test did not demonstrate instability. As described above, two patients had a re-operation. They experienced a delayed relapse of preoperative symptoms, and MR showed persistent compression of the cerebellar tonsils (Fig. 2). Symptoms eventually subsided after the second operation. Concerning our three patients who had a worsened clinical outcome, we did not find peculiar elements that could explain such an unsatisfying outcome. In those patients, postoperative MR showed a correct rebuilding of the cisterna magna, and in two the syrinx collapsed.

All of the patients performed at least three postoperative MR controls (three to six). The syringomyelia collapsed in 29 patients (80.5%; Fig. 3) and remained unchanged in seven (19.4%). We did not observe any widening of the syrinx. The mean time for the syrinx to shrink was 8 months (range 6–26 months). No patient experienced a cerebellar sagging. By using the chi-square test, we did not find any significant difference in the frequencies of preoperative syrinx extension among patients grouped by clinical outcome and syrinx postoperative modification (p = 0.551 and p = 0.574, respectively). Similarly, the preoperative degree of tonsillar descent did not show a correlation with clinical and neuroradiological postoperative evolution (p = 0.603 and p = 0.440, respectively).

Fig. 3
figure 3

a Preoperative MR showing a cervico-thoracic syringomyelia associated to a mild form of Chiari I. b Postoperative MR at 18 months: a wide cisterna magna is visible, and the syrinx dramatically narrowed

Full size image

Discussion

On average, recent surgical series demonstrate that more than two thirds of patients improve or stabilize their preoperative neurological status after CCD [7, 15–18]. Our experience is in agreement with this trend given that improved and stabilized patients accounted for 91.6%. Unlike other series [11–13, 19, 20], the present analysis derived from a consecutive and uniformly treated population in which the same surgical techniques treated only CMI related syringomyelia.

Some authors have continued to prefer the syringo-subarachnoidostomy as an adjunct to the CCD [14, 21, 31], but we think that a syringosubarachnoid shunt is disadvantageous. Spinal cord injury, accompanying myelotomy with insertion of the catheter, and shunt failure (up to 29% [14, 21, 22]) may occur, albeit not often. In addition, the syringosubarachnoid shunt cannot be indicated for patients in whom the Chiari I malformation is mainly responsible for clinical symptoms or in whom the cavity in the spinal cord is too small to insert a catheter.

Although the theoretical bases of the CCD are widely accepted (namely the re-building of a new cisterna magna), there is still debate about several surgical aspects. We advise performing a craniectomy that does not exceed 3 cm of maximum diameter so that cerebellar sagging is avoided. Some authors [7, 10, 11, 26], believing that the decompression of the PF is insufficient to restore a normal CSF flow, support the coartation of tonsils or dissection of the arachnoid. Even though their results in terms of neurological outcome are convincing, the extra-arachnoidal approach is safer. In series where decompression with intrarachnoidal manipulation was used, postoperative complications such as dysphagia (3.4%), hearing loss (5%), mesencephalic dysfunction (2.2%) have non-negligible rates [11, 16]. In our series, no patients had complications related to direct surgical damage.

Several observations dictate the choice to open and replace the dura. The first criterion is whether the thickened occipito-cervical dural band strongly contributes to the narrowing. Some authors [7, 12, 16, 23] reported an increased rate of syrinx resolution when suboccipital decompression is combined with duroplasty. The second criterion, as already reported [5], is whether this band shows an increase in collagen fibers, hyalinous nodules, and calcifications. In fact, in our two patients, that underwent a second operation, a thicker and more fibrous band was found. In his experience, Tubbs and collaborators [18] observed thickening and scarring of the occipito-cervical membrane in a patient re-operated on for the recurrence of symptoms.

Opening the dura somewhat enhances the risk of accidental arachnoid tearing. In three (8.3%) cases, we recorded CSF leaks that eventually resolved without more severe complications. This percentage is lower than those series reporting patients undergoing CCD with arachnoid manipulation that showed up to 16% of external CSF leaks [11, 16]. Hoffman and Souweidane had a 0% CSF leak rate [24] in a series of patients operated on by CCD and duroplasty. On the other hand, Perrini et al. [17], in describing a CCD with dura opening but without duroplasty, reported a 4% rate of CSF leaks with need of reoperation. They prefer not to perform duroplasty because they believe this maneuver involves a higher risk of arachnoid tearing. In our experience, accidental arachnoid opening mainly happens during the opening of the dura rather than during its closure, so duroplasty is useful to prevent CSF leaks.

Our results demonstrate that preoperative extension of syrinx and the type of tonsillar descent does not correlate statistically to postoperative evolution of symptoms and MR modification after surgery. These results are in agreement with other investigators [4, 15], leading to our conclusion that surgeons cannot use these MR findings to predict outcome. Hence, as in other pathologies of the spinal cord, the first feature to consider when counseling a patient with CMI associated syringomyelia for surgery is the severity of the preoperative neurological status. Moreover, referring to the work of Attal et al. [15], the most important predictors of surgical outcome is the interval between onset of symptoms and CCD. An early surgical approach in symptomatic patients should be preferred. Nevertheless, most of the “improved” patients continue to present minor symptoms (such as dysesthesias or paresthesias) even when their syrinx shrinks. Indeed, the slow and progressive cavitation of the spinal cord produces a lesion that can be only partially reversed, even after a satisfying CCD. Recently, Wetjen et al. [25] published a prospective study that investigated the neuroradiological and clinical effect of CCD on syringomyelia during a long period of observation. They confirmed previous results that CCD progressively reduces syrinx volume and improves symptoms even in those cases where the narrowing is only partial. The persistence of symptoms does not mean a failure of the procedure but rather a confirmation of a chronic spinal damage. In our routine practice, if a patient presents persistence or worsening of symptoms after surgery, a MR of the posterior fossa is advised to primarily rule out an insufficient CCD. In the same way, if symptoms recur after surgery, a new MR is necessary to detect a re-stenosis (as described in our cases) probably due to insufficient duroplasty.

Conclusions

Treating syringomyelia associated with Chiari I malformation through craniocervical osteo-dural decompression with duroplasty results in a large percentage of patients with satisfying results and no irreversible complications. Although our experience is limited by a relatively small number of patients and the lack of prospective observation, it seems that our data agrees with larger studies. Particularly, the extension of the syrinx does not seem to predict clinical and MR results; hence, preoperative clinical conditions are the most reliable predictors of outcome. Consequently, a symptomatic patient must be identified early for this procedure. Further research is needed in order to better comprehend the pathophysiology of the syrinx in Chiari I malformation so that a larger percentage of patients can benefit from this procedure.