Cervical myelopathy: current diagnostic and treatment strategies☆☆☆★

Anterior cervical discectomy and fusion 

The anterior approach to the cervical spine was developed in the 1950s by Baily and Badgley [40], Smith and Robinson [41] and Cloward [42]. Since that time, anterior decompression and fusion has been applied to the treatment of cervical stenosis resulting from herniated discs, spondylosis or ossification of the PLL. When the pathologic process occurs at the level of the intervertebral disc, anterior cervical discectomy and fusion (ACDF) usually provides sufficient access to the stenotic focus. Clinical series have demonstrated successful arthrodesis in 92% to 96% of patients after single-level ACDF with satisfactory clinical results 43, 44, 45.

The extent to which decompression should include the posterior osteophytes and the PLL remains controversial. Bohlman [46] reported a series of 17 patients that underwent ACDF without removal of either the PLL or osteophytes. All but one patient experienced a good to excellent outcome. Other authors advocate the removal of osteophytes with the hope of accelerating neurologic recovery. Kadoya et al. [47] demonstrated that osteophyte removal could be safely performed in his series of 43 patients with myelopathy who underwent decompression using a microscope. The extent of decompression is dictated by the degree of stenosis present on preoperative imaging studies and local anatomic landmarks. Using microscopic visualization, posterior end plate osteophytes are thinned with a high-speed burr until only a wafer-thin cortical remnant remains. The remaining cortex and chondral fragments are then dissected off the PLL and excised with a curved micro curette.

When foraminal stenosis is present, wide foraminotomies are performed to remove uncinate spurs. The entire uncinate region is carefully cleaned of all disc material, and the lateral margin is determined by palpation with a blunt probe. The posterior portion of the uncinate region is then decompressed with a burr and micro curettes until the foramen is wide open. A micro curette or probe can be used to ascertain the patency of the foramen.

For cord compression at multiple levels, numerous anterior decompressive options exist, including multilevel ACDF, corpectomy and hybrid procedures. Although the rate of neurologic improvement remains high for multilevel ACDFs [49], the incidence of nonunion increases with the number of levels being fused. The recently reported fusion rate for a one-level ACDF using autograft iliac crest without plating (two graft–host sites) was 96% [50]. This decreased to a 75% fusion rate when a two-level ACDF was performed and to a 56% fusion rate with a three-level ACDF (six graft–host sites) [51].

Morbidity associated with autogenous bone graft harvest from the ileum and fibula spurred interest in the use of structural allograft with ACDF. Numerous studies have demonstrated the pseudarthrosis rate after single-level uninstrumented ACDF using structural allograft (2% to 20%) to be similar to that obtained with autograft. For uninstrumented multilevel ACDF, however, the incidence of nonunion is higher with allograft [45].

The introduction of anterior cervical plating has led to an improvement in the fusion rate observed after ACDF. Wang et al. [51] recently reported on 60 patients who underwent a two-level ACDF. Thirty-two patients were treated with plating and 28 without plating. Solid arthrodesis was observed in all patients with plates and 75% without. In a separate study by Wang et al. [51], the fusion rate for three-level ACDF was 82% with plating and 63% without. Improved fusion rates using plates were also demonstrated by Geck et al. [52] in their report on 205 ACDF procedures. A three-fold improvement in the fusion rate was noted (11.8% vs. 3.6%) with the plated group.

Corpectomy 

An alternate means of improving the fusion rate after multilevel decompression is the use of corpectomy. In addition to improving the fusion rate, corpectomy provides for a more extensive decompression and serves as a source of autograft. Stenosis at two adjacent levels can be decompressed with a single-level corpectomy (two graft–host interfaces) as compared with two ACDFs (four graft–host interfaces). Swank compared the fusion rate after ACDF and corpectomy. The nonunion rate in the one-level corpectomy group was 10% compared with 36% for patients undergoing two-level ACDF procedures [37]. Hilibrand reported on 131 patients undergoing multilevel ACDF and 59 patients undergoing corpectomy (one to four levels). Patients in the ACDF group had a 34% incidence of nonunion. Patients undergoing corpectomy had a 7% nonunion rate and six graft dislodgements, four requiring revisions [39]. Based on these factors, corpectomy may be considered preferable to multilevel ACDF, especially in higher-risk patients, such as recent smokers, diabetics or revision cases. If corpectomy is performed in a patient without compression behind the vertebral body, then 2 to 3 mm of the posterior cortex can be left intact. This provides a posterior buttress to prevent the graft retropulsion into the canal. For conditions with diffuse canal stenosis, corpectomy and removal of the entire posterior cortex is the only viable alternative. Examples include ossification of the PLL, malunited fractures with persistent canal stenosis and postlaminectomy kyphosis.

Static plates, buttress plates and most recently dynamic plates have been introduced with the intended purpose of decreasing the rate of nonunion and to prevent anterior strut graft dislodgement after corpectomy 53, 54, 55. The early experience with static plating with multilevel corpectomies, however, suggested that such plates may predispose to an increased incidence of strut graft dislodgment. Vaccaro et al. [38] reported a multicenter experience of 45 patients undergoing two-level (33 patients) and three-level (12 patients) corpectomies with plating. Strut graft dislodgement was observed in 9% of the two-level corpectomy group and 50% of the three-level corpectomy group, despite adjunctive halo application in 10 of 12 three-level corpectomy patients.

Foley et al. [56] evaluated the biomechanical rationale of anterior plating as an adjunct to strut grafting. They determined that with strut graft alone, cervical flexion increased the load applied to the graft, whereas extension decreased the load. With addition of an anterior static plate, a paradoxical mechanical effect was observed. Cervical extension substantially increased the load on the graft beyond that produced by extension without a plate. In contrast, cervical flexion in the presence of static plate consistently unloaded the strut graft.

Anterior cervical buttress plates were designed to prevent graft dislodgment while allowing for physiologic patterns of force application through the anterior column. Unfortunately, failure at the implant–host bone interface and subsequent strut graft dislodgement were observed after multilevel corpectomies [57]. In order to prevent these sequelae, supplementation with posterior fusion is recommended when multilevel strut graft fusion with buttress plate fixation is employed.

A new generation of “dynamic” anterior cervical plates has been developed to avoid the failures of static and buttress plates. A recent in vitro model demonstrated that these devices accommodate the normal settling of the strut graft–host bone interface, avoid stress shielding of the anterior column and provide for normal strut graft loads [58]. Critical analysis of the early experience with the dynamic plate is required in order to determine its role in multilevel anterior decompression procedures.

An alternative means of addressing the challenges of multilevel anterior decompression and fusion is to perform a combination of a corpectomy and discectomy procedures (Fig. 2). For a patient with stenosis involving three contiguous levels, the top two involved disc spaces may be addressed by a single-level corpectomy. A discectomy is then performed at the most caudad disc space. Two bone grafts are then placed to span the defects. An anterior cervical plate is then placed over the entire construct. This compromise between a two-level corpectomy and three-level discectomy preserves much of the structural integrity of the cervical spine and limits the necessary healing surfaces to four. In addition, because four screws around an ACDF anchor the inferior end of the plate, the plate is less apt to dislodge. This technique cannot be employed if there is stenosis behind the inferior vertebral body, because it is left intact. A disadvantage of such hybrid procedures is the added operative and anesthetic time resulting from the fashioning of two or more grafts. Hybrid ACDF and corpectomy procedures may also be used in patients with four-level disease. One can perform a two-level corpectomy to address the top three disc levels, along with a single-level discectomy for the lowest level. An alternative reconstructive strategy for patients with stenosis at four levels is to perform two single-level corpectomies separated by a noncorpectomy level. (Fig. 3).

View Large Image Download to PowerPoint

Fig. 2. Postlaminectomy kyphosis in an elderly woman with rheumatoid arthritis and osteoporosis was treated with corpectomies of C4 and C5, anterior cervical discectomy and fusion at C6–C7, a spanning anterior plate from C3 to C7 and lateral mass plating from C3 to C7. Corpectomy and discectomy defects were reconstructed with a combination of allograft fibula and autologous cancellous iliac crest autograft.

View Large Image Download to PowerPoint

Fig. 3. Double single-level corpectomy at C4 and C6.

Anterior decompression with circumferential fusion has been advocated for cases involving decompression and fusion of three or more segments. Addition of a posterior segmental fusion has been demonstrated to increase the rate of fusion and decrease the incidence of graft and implant-related problems [59]. The addition of posterior segmental fusion to an anterior decompression and fusion procedure should be considered in cases of multilevel (three or more) spondylosis or ossification of the PLL requiring extensive decompression, multilevel spondylosis with congenital stenosis, multilevel spondylosis with degenerative kyphosis, postlaminectomy kyphosis and an S-shaped curve and posttraumatic kyphosis [60].

Complications with anterior decompression procedures 

Complications occurring with anterior decompression and fusion procedures may be related to the soft tissue structures, bone graft and the neurologic system. One of the most common unintended sequelae of anterior procedures is postoperative dysphagia. Although the incidence of esophageal perforation is very low (0.2% to 0.9%) [61], postoperative dysphagia has been observed in 2% to 48% of patients [62]. Symptoms are frequently transient, with chronic dysphagia reported in up to 12% of patients. The etiology may be multifactorial, including hematoma formation, prolonged retraction and denervation of the upper esophagus by injury to the pharyngeal plexus [63].

Temporary hoarseness is reported in 3% to 11% of patients, with permanent hoarseness occurring in 0.33% 64, 65. The etiology of postoperative dysphonia is postulated to be related to direct injury to the recurrent or superior laryngeal nerves, the length of the procedure and force of retraction. A recent study by Apfelbaum et al. [64] suggested that the incidence of dysphonia after anterior cervical spine surgery may be partly the result of injury of the recurrent laryngeal resulting from the endotracheal tube. Monitoring of endotracheal tube cuff pressure and release after retractor placement was found to dramatically decrease the incidence of dysphonia in their study of 900 consecutive patients. Historically, rates of dysphonia are reported to be higher with right-sided cervical approaches because of the varied course of the recurrent laryngeal nerve. In contrast to previous reports, a recent study by Butler et al. [66] found no difference in the incidence of recurrent laryngeal nerve injury when comparing right- and left-sided approaches. A higher incidence of nerve injury was noted with increased levels of decompression and revision surgery.

Harris et al. [67] recently presented an anatomical study on the course of the superior laryngeal nerve. They found that the sensory branch of this nerve, which innervates a portion of the larynx and vocal cords, is located most consistently between the C2 and C3 vertebra. Injury to this nerve or its branches during higher anterior cervical approaches may result in postoperative voice changes or sensory changes, leading to a diminished aspiration protective response.

Iatrogenic injury to the vertebral artery during anterior cervical procedures, although exceedingly rare, can be a devastating complication. A recent study by Burke et al. [67a] demonstrated an incidence of vertebral artery injury to be 0.03% (6 of 1,976 patients). Various treatment options have been described, including repair, ligation and tamponade. In the series of six patients, four were asymptomatic, one experienced a vertebral artery dissection with resultant lateral medullar infarct, and one died intraoperatively because of hypovolemia. An anatomic study by Thomas et al. [68] demonstrated that the incidence of brainstem infarction was 1.8% after occlusion of the right vertebral artery and 3.1% after the left was occluded.

The fate of motion segments adjacent to anterior cervical fusions remains a subject of continued investigation. It has been theorized that fused segments accelerate the degenerative process at neighboring levels. Hilibrand et al. [69] recently reported a retrospective long-term disease-free survivorship analysis in 374 patients undergoing 409 anterior cervical fusions. They found a relatively consistent incidence of adjacent segment disease of 3% per year. Degeneration was most likely at C5–C6 and C6–C7 and occurred more frequently after single-level rather than multilevel fusion.

A potentially lethal complication of anterior cervical surgery is postoperative airway obstruction caused by edema or hematoma formation. Emery et al. [70] recommend that all patients with severe myelopathy be kept intubated postoperatively. A recent study involving a large number of patients who underwent anterior cervical procedures [71] suggests that the operative time and numbers of levels of decompression are the most important factors in determining whether the patient should be kept intubated postoperatively. The authors advocate extubating patients when the total cervical retraction time is less than 2.5 hours or when the total operative time is less than 3 hours. For patients who are difficult to intubate, the authors recommend extubation if the operative time is less than 90 minutes. Using these parameters, the authors found that airway difficulties resulting from edema were not encountered. Respiratory distress resulting from hematoma formation was encountered in 4 of 256 patients 30 to 48 hours postoperatively. Each of the 4 patients was each reintubated and underwent hematoma evacuation.

If a patient is kept intubated postoperatively, a cuff-leak test should be performed before extubation. The cuff-leak test is performed by occluding the endotracheal tube lumen with the cuff balloon deflated. If the patient can breathe around the tube, then there is adequate space between the tube and the larynx. If the patient cannot breathe around the tube, then they should be kept intubated and the cuff leak test repeated in 24 hours.

Laminectomy 

Historically, laminectomy has been regarded as the standard posterior procedure for the treatment of multilevel cervical myelopathy. Laminectomy is a straightforward technique that provides for an extensile posterior decompression and excellent visualization of the neural elements. A recent study by Kaptain et al. [80] evaluated the success of laminectomy in the management of cervical spondylotic myelopathy in 46 patients. With an average of 4 years of follow-up, the function for 29% of patients was improved compared with preoperative levels, in 42% it was unchanged, and in 29% it was worse. The presence of a kyphotic sagittal alignment was present in 9% of patients before surgery and in 28% of patients at latest follow-up. Overall, good to excellent results have been reported in 40% to 85% of patients after laminectomy. In a subset of patients, however, the results tend to worsen over time 81, 82, 83. The deterioration in results after laminectomy has been attributed to multiple causes, including the development of a scar membrane around the dura, segmental instability and kyphosis [84]. Postoperative cervical kyphosis after laminectomy is directly related to the amount of facet joint resection. Raynor et al.[85] have shown that a 50% facetectomy allows a visualization of 3 to 5 mm of nerve root, whereas a 70% resection allows visualization of 8 to 10 mm. These authors concluded that resection of greater than 50% of the facet joint significantly compromises facet strength . Zdeblick and Ducker [45] noted that segmental hypermobility of the cervical spine results if a foraminotomy involves resection of more than 50% of the facet. A biomechanical study by Nowinski et al. [86] concluded that as little as a 25% facetectomy affects stability after multilevel laminectomy. Hypermobility produced by a decompressive laminectomy may result in segmental listhesis and angular kyphosis. In addition to the production of a postural deformity and axial discomfort, a hypermobile cervical spine with a kyphotic alignment may result in recurrent cord dysfunction with neurologic deterioration.

If postlaminectomy kyphosis occurs, an extensive reconstruction is generally necessary to halt the progression of both the sagittal deformity and myelopathy. In postlaminectomy patients, corpectomy and fusion alone as a reconstructive option should be avoided, because it results in complete disassociation of the lateral pillars and results in severe instability (Fig. 4). Surgical correction of a postlaminectomy kyphosis, therefore, requires anterior strut graft placement to re-establish a more normal sagittal alignment followed by posterior stabilization (360-degree fusion).

View Large Image Download to PowerPoint

Fig. 4. Computed tomography myelogram after cervical corpectomy was performed for the treatment of postlaminectomy kyphosis. Note the complete disassociation of the two lateral masses.

Laminectomy and fusion 

The frequency of instability and loss of lordosis after laminectomy has prompted some authors to advocate prophylactic fusion of decompressed cervical levels 73, 74, 75, 87. Concurrent lateral mass plating has the advantages of stabilizing the decompressed segment in a lordotic posture and preventing segmental instability (Fig. 5). Fusion also permits a more expansive laminectomy and foraminal decompression without jeopardizing stability. Instrumentation is used in all but the stiffest spines to maintain segmental stability and lordosis while the arthrodesis takes place. Various facet-wiring techniques have been employed for three decades. Presently, lateral mass plating is most commonly used for more rigid internal fixation. Arthrodesis typically spans the decompressed levels but may be extended to the subjacent level at the cervicothoracic junction. The few published reports describing the results of laminectomy and fusion reflect a high rate of functional improvement. Limitations of the procedure relate principally to attempts at fusion: nonunion, hardware failure, adjacent segment degeneration, loss of lordosis and autograft harvest site discomfort [88]. While potential problems related to posterior plating with screws include injury of the vertebral artery or cervical nerve roots, such complications are fortunately uncommon 88, 89.

View Large Image Download to PowerPoint

Fig. 5. Computed tomography myelogram after laminectomy and instrumented fusion with lateral mass plates. (Courtesy of Gerald Rodts, MD.)

Laminoplasty 

In Japan, poor outcomes associated with cervical laminectomy spawned the evolution of a different strategy for achieving posterior decompression. In 1973 Oyama [90] introduced a canal-expansive laminoplasty procedure. Many modifications of the procedure have been reported since then 78, 91, 92. Laminoplasty strategies have taken two forms: eccentric expansion with a unilateral hinge and symmetric expansion with bilateral hinges. The common purpose of all laminoplasty procedures is the increase in canal area through reconfiguring of the posterior bony arch (Fig. 6). Fusion of decompressed levels should be avoided except in those cases in which segmental instability is noted preoperatively. Decompression of neuroforamina may also be achieved by performing a laminoforaminotomy before expansion of the posterior arch or through direct decompression of the neuroforamina after the expansion of the posterior arch.

View Large Image Download to PowerPoint

Fig. 6. Lateral radiograph after open-door laminoplasty from C3 to C7 demonstrating an expanded canal and suture anchors in the lateral masses.

Success of various laminoplasty procedures in decompressing the spinal cord and producing a functional improvement in patients with cervical myelopathy has been well documented 72, 77, 78, 93. A recent study by Kawaguchi et al. [48] evaluated the results of laminoplasty in 133 patients with a minimum 10-year follow-up after laminoplasty. They found that the functional recovery rate of 58.3% was well maintained over time (55.1%). Laminoplasty has also been demonstrated to be successful in the management of cervical spondylotic myelopathy in the elderly [94], diabetics [95] and patients on dialysis [96]. Radiographic and anatomic studies have demonstrated laminoplasty's ability to expand the cross-sectional area of the canal and decompress the spinal cord 93, 97, 98. Cord decompression occurs as the spinal cord migrates dorsally away from impinging anterior structures. The success of laminoplasty in decompressing the cord, however, is contingent on the presence of a neutral to lordotic sagittal alignment [99]. Laminoplasty performed on patients with cervical kyphosis has been shown to result in less cord enlargement and functional recovery than those patients with a lordotic alignment. Additional factors associated with inferior outcomes are cord atrophy, long symptom duration, advanced age, severe cord compression and radiculopathy 33, 100. Although laminoplasty is not considered a fusion procedure, most patients experience a significant loss of subaxial motion after the procedure [93]. Satomi et al. [77] reported a 47% decrease in cervical motion in the sagittal plane at 1 year, which increased to 56% at final follow-up (7.8 years), after open-door laminoplasty. Edwards et al. [93] reported a 38% decrease in cervical motion after spinous process–splitting laminoplasty in North American patients with mean 2-year follow-up. Spontaneous segmental fusion at laminoplasty hinge sites is believed to be the causative factor in most cases. Our current practice is to apply bone wax to the exposed cancellous surface at the hinge site to prevent intersegmental bone formation and subsequent ankylosis. Bone graft is not placed at the hinge site, as has been previously described. Early motion is encouraged, starting the day of surgery.

Posterior arch closure after laminoplasty has been reported [93] but may be avoided using a combination of techniques. For open-door laminoplasty, we prefer to keep the posterior arch open with a nonresorbable braided suture attached to a suture-anchor placed into the hinge-side lateral mass. Before the posterior arch is hinged open, a transverse hole is made at the base of the spinous process as for a spinous process wire. The sutures are passed through this hole and tied down (Fig. 6). When spinous process–splitting laminoplasty is performed, immediate stability is achieved by suturing the bone graft strut in place between the two hinged laminae.

Complications with posterior decompression procedures 

Neurologic complications observed after posterior decompressive procedures include iatrogenic cord or root trauma, nerve root dysfunction and late neurologic deterioration. Laminectomy in the past featured significant complications resulting from direct damage to the cord by a rongeur introduced into the spinal canal where stenosis is severe. Fortunately, with the introduction of en block laminectomy techniques, such devastating iatrogenic complications are rare [101]. Nerve root palsy, especially C5, has been observed to develop in the first few days after both laminectomy and laminoplasty. Dai et al. [102] reported on a series of 287 consecutive patients undergoing laminectomy in which 12.9% developed postoperative radiculopathy, principally at C5 and C6. Large laminoplasty series have documented a prevalence of nerve palsy after laminoplasty ranging from 3% to 8%, with the C5 root being most commonly involved 48, 78, 103. In a series of 203 laminoplasty procedures reported by Yonenobu et al. [104], 12 patients developed fifth or sixth nerve root palsy, 3 patients had seventh nerve root involvement, and 1 had an eighth root complication. Patients often have a normal examination immediately after surgery. The onset of motor weakness generally develops within 1 to 5 days postoperatively. Sensation typically remains intact. Motor weakness for most patients resolves substantially or completely within one year. Explanations concerning the apparent predisposition of the C5 nerve root to dysfunction after posterior decompression principally center around three factors: the C5 segment is usually the midpoint of decompression and the extent of shifting at this segment is greater than that at other segments, and the C5 root is shorter than those of other segments. Moreover, because the deltoid receives sole C5 innervation, C5 palsy is more likely to be clinically evident than roots with shared motor functions [105].

A decrease in cervical lordosis has been reported after each of these posterior decompressive procedures. With laminectomy, nominal soft tissue capsule removal around the facet and less than 50% facetectomy minimizes destabilization. Recent reports suggest that the incidence of kyphotic deformity after laminectomy currently ranges from 21% to 33% 80, 106. At greatest risk are children, adults with a preexisting neutral to kyphotic alignment and those patients who have undergone a wide laminectomy without surgical stabilization 105, 107, 108. A decrease in lordosis after laminectomy and fusion or laminoplasty is less common than after laminectomy and tends to occur principally at the upper and lower extents of decompression where the ligamentum flavum and interspinous ligaments have been disrupted or are left unreconstructed 88, 109.

Patients commonly report axial symptoms, such as neck pain, stiffness, fatigue or shoulder discomfort, after posterior decompressive procedures. The causes of these symptoms are not well understood. Hosono et al. [110] reported a prevalence of axial symptoms in 60% of 72 patients after laminoplasty. Symptoms were of such severity as to require some form of analgesics in 13% of patients. Fortunately, symptoms resolved in a majority of patients within 1 year. In a matched cohort study of laminoplasty and laminectomy with fusion, the incidence of axial symptoms (69%) and the need for postoperative analgesics were identical for both procedures [88]. The prevalence of axial symptoms experienced after laminectomy was reported by Yonenobu et al. [103] to be 27.2%. The causes of such symptoms after posterior decompressive procedures are not well understood and merit further investigation. A study presented by Asano et al. [111] investigated the effect of early motion without immobilization after spinous-splitting laminoplasty. The authors found that early motion without immobilization resulted in markedly less axial pain with significantly better maintenance of lordosis as compared with a control group. Our current practice after laminoplasty is to place patients in a soft collar for use on an as-needed basis. Patients are directed to remove the collar several times per day for cervical range of motion exercises beginning on postoperative day 1.