Introduction

Cervical myelopathy is a common cause of spinal cord dysfunction in older persons all over the world. The treatment options of cervical myelopathy include conservative treatment or operative management. Operative management is indicated for most of the patients with clinically evident cervical myelopathy and it is recommended for patients who have either substantial or progressive impairment of neurological function without sustained remission [1, 2]. Surgery for cervical myelopathy involving one or two motion segments may be successfully performed with a low incidence of complications, and consensus favors anterior cervical decompression with fusion and instrumentation for this kind of patients [3], while surgical treatment for multilevel cervical myelopathy (MCM) of three or more levels is associated with less predictable outcomes and a higher frequency of complications [4–6].The optimal treatment for MCM continues to be the subject of considerable debate, and controversy always exists over using anterior or posterior surgery.

Although extensive research on the two approaches has been done [7–18], yet studies on the comparison of the two approaches are limited. Furthermore, a wide variation in patient population, study design and results exists. These factors make it difficult for workers in this field to know the exact value of the two approaches. Meta-analysis represents a powerful tool to summarize findings in the literature by taking into consideration and enabling analysis of differences between studies [19, 20]. Thus, the purpose of our study is to perform a systematic review to compare the anterior surgery with the posterior surgery for MCM by comparing the surgical outcomes, radiographic changes and complications of these two approaches.

Materials and methods

Literature search

A comprehensive computer literature search of abstracts about studies in human subjects was performed to identify articles about anterior and posterior surgery for MCM. The MEDLINE, EMBASE databases, from January 1991 to November 2009, were searched with the following keywords: (“cervical spondylotic myelopathy OR cervical myelopathy OR cervical spondylosis OR cervical stenosis OR ossification of the posterior longitudinal ligament”) AND (“multisegment OR multilevel OR multi-level OR multi-segment”) AND (“anterior OR posterior OR laminoplasty OR laminectomy OR corpectomy OR discectomy OR vertebrectomy OR spondylectomy OR surgical OR surgery OR dorsal OR ventral”). No language restrictions were applied.

Other databases, such as Web of Knowledge, EBSCO, Sciencedirect, Springlink, Scopus and The Cochrane Library, were also checked for relevant articles with the same keywords. We also searched the abstracts of American Academy of Orthopaedic Surgeons Annual Meeting (2006–2008: http://www.aaos.org/education/anmeet/libscip.asp). The list of articles was supplemented with extensive cross-checking of the reference lists of all retrieved articles.

Selection of studies

Two reviewers (L.T, X.W) independently assessed potentially eligible studies. The study selection was accomplished through two levels of study screening. At level 1 screening, abstracts were reviewed for the following exclusion criteria: case reports, letters, editorial, comments, reviews and articles that did not include raw data; follow-up period <30 days; no surgical intervention. Full articles were then obtained for all studies accepted at level 1 screening and for any citations for which a determination could not be made from the abstract. If the study was not reported in full journal publications, we contacted the authors for full text or additional information needed. For level 2 screening, the inclusion criteria were any randomized or quasi-randomized controlled clinical trials of anterior versus posterior surgery for MCM in adults. When data or subsets of data were presented in more than one article, the article with the most details or the most recent article was chosen. The studies were excluded when the results were presented in combination and could not be differentiated for performance assessment.

Data extraction

The same observers independently extracted relevant data from each article by using a standardized form. Observers were not blinded with regard to the information about the journal name, the authors, the authors’ affiliation or year of publication, since this had been shown to be unnecessary [22]. To resolve disagreement between reviewers, a third reviewer (C.T.) assessed all discrepant items, and the majority opinion was used for analysis.

Common characteristics about studies: Author’s country; year of publication; type of patient’s disease; the number of patients for each group; the number of operative levels.

Study design characteristics about studies: A methodological quality assessment scheme used by Handoll [21] was used to extract relevant study design characteristics of each study. In this scheme, each item was graded either ‘Y’, ‘?’, or ‘N’, respectively indicating that the quality criteria were met for the item (“yes”), or possibly or only partially met for the item (“Possible, partial”), or not met (“No”).

Clinical characteristics about studies:

  1. 1.

    Baseline characteristics for each group: preoperative JOA score, age, duration of the symptoms, and others.

  2. 2.

    Outcome measures: we analyzed multilevel CSM and OPLL patients separately in perioperative data, clinical outcomes, and complication outcomes. For radiographic data, we analyzed them together.

(a) Perioperative data: operative time and blood loss. (b) Clinical outcomes: JOA scale [23] and recovery rate, which is defined according to the rationale of Hirabayashi [24] as recovery rate = (ε postoperative JOA scores-ε preoperative JOA scores/17-ε preoperative JOA scores) × 100%. The results were indicated by the recover rate as follows: 75% or more (excellent), 50–74% (good), 25–49% (fair), and <25% (poor) (we calculated excellent result and good result together as ‘excellent to good’ result, and fair result and poor result together as ‘fair to poor’ result). (c) Complication outcomes. (d) Radiographic outcomes: decrease rate of ROM (range of motion) from C2 to C7, alignment deterioration, change of anterior–posterior (AP) diameter of the narrowest spinal canal, change of regional Cobb’s angle and change of overall Cobb’s angle.

Timing of outcome assessment and subgroup analysis

The results were collected for the final follow-up time. Analysis of the outcomes was divided into two subgroups according to the time of outcome assessment if possible. One was the follow-up time ≤5 years and the other was the follow-up time >5 years.

Statistical analysis

Common characteristics were summarized using basic descriptive statistics (simple counts and means). Clinical characteristics were synthesized via meta-analytic pooling of each group results. Meta-analysis was performed in line with the recommendations from the Cochrane Collaboration and the Quality of Reporting of Meta-analyses Guidelines [25–27].

All analysis were performed using Microsoft Excel 2003 (Microsoft, Seattle, Wash), SPSS 13.0 for Windows (SPSS, Chicago, III), and RevMan5.0 (Review Manager (RevMan) [Computer program]. Version 5.0. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2008). RevMan5.0 is a freeware software produced by the Cochrane Collaboration, and could be downloaded from the website “http://www.cc-ims.net/RevMan/RevMan5”.

Results

Literature search and selection of studies

After the computerized search was performed and reference lists were extensively cross-checked, about 370 abstracts were identified. Of these, 288 were rejected after level 1 screening (reviewing the abstracts).Of the remaining 82 articles, 72 relevant articles were excluded after we read the full texts or additional information of these articles because (1) the articles were not controlled clinical trails (n = 48), (2) the aim of the articles was not to compare the clinical outcome for anterior surgery with posterior surgery (n = 12), (3) the articles were about comparison for cervical myelopathy, but the clinical data of multi-level cervical myelopathy could not be extracted (n = 8), (4) the studies did not include raw data (n = 4). At last 10 articles [28–37] including 10 studies fulfilled all inclusion criteria and were selected for data extraction and analysis.

Common characteristics about studies

The studies took place in one of five countries [Japan (5), India (2), South Korea (1), Germany (1) and USA (1)]. They were mainly non-randomized controlled clinical trials except 1 cohort study. Five studies were focused on mutilevel cervical ossification of the posterior longitudinal ligament and the other five studies were focused on mutilevel cervical spondylotic myelopathy. Most operative techniques were corpectomy for anterior surgery and laminoplasty for posterior surgery. Table 1 presents specific operative techniques of anterior and posterior surgery in each study. There were total 561 patients in the selected studies and the age ranged from 30 to 75 years. Two hundred and twenty-one patients were treated by anterior surgery and 340 patients underwent posterior surgery. In five studies, the sex distribution was described: 201 patients were males and 97 patients were females. Table 2 presents the detail information about the common characteristics of the included data sets.

Table 1 Specific operative techniques of anterior and posterior surgery in each study
Full size table
Table 2 Common characteristics about studies included in the systematic review
Full size table

Study design characteristics

Most studies had a suboptimal design with regard to treatment concealment, the intention to treat analysis, and double-blind (questions 1–3 and 5–6, 100% for “no” and “?”responses to these questions). However, as for baseline characteristics, inclusion and exclusion criteria and outcome measures, etc., most studies were optimally designed (100% for “yes” responses to question 9 and 10; 80% to question 4; 70% to question 7; 60% to question 8). Actually, questions 1–3 and 5–6 were more concerned with study method and questions 4 and 7–11 were more concerned with clinical data, so questions 4 and 7–11 were more important. If the studies were ideally designed according to these questions, the clinical result would be correct and credible.

Clinical characteristics about studies

Baseline characteristics

Most studies reported that the baseline characteristics, including age, duration of the symptoms, preoperative JOA score, and follow-up time, etc. were matched for each group. The results of the meta-analysis of these baseline characteristics showed that there was no significant difference between the anterior and posterior surgery in duration of the symptoms, preoperative JOA score and follow-up time. Although the age of the anterior group was significant younger than the posterior group (p < 0.05), the mean age for the two groups were both over 50 years, and the gap between the two groups for the majority of studies was <3 years. Therefore, it would not affect the clinical outcome. The results of the meta-analysis of the baseline characteristics for anterior versus posterior surgery were summarized in Table 3.

Table 3 Results of meta-analysis of baseline characteristics for anterior versus posterior surgery
Full size table

Outcome measures

For multilevel CSM patients

(a) Perioperative outcomes: the operative time for the anterior group was significantly longer than the posterior group (p < 0.05, WMD 53.37, 95% CI [10.71, 96.04]). The operative blood loss for anterior group was significant more than the posterior group (p < 0.05, WMD 102.81 [42.46, 163.15]).

(b) Clinical outcomes: the final follow-up JOA score for the anterior group was significantly higher than the posterior group (p < 0.05, WMD 0.83 [0.24, 1.43]) in the ‘follow-up time ≤5 years’ subgroup, but had no significant differences in the ‘follow-up time >5 years’ subgroup (p > 0.05). In total, it was significantly higher for the anterior group than the posterior group too (p < 0.05, WMD 0.88 [0.31, 1.44]).

The recovery rate for the anterior group was significantly higher than the posterior group (p < 0.05, WMD 10.08 [1.39, 18.78]) in the ‘follow-up time ≤5 years’ subgroup. Because of no study reporting the recovery rate for the follow-up time >5 years, we could not analyse the recovery rate for the two groups in the ‘follow-up time >5 years’ subgroup.

Both the ‘excellent to good’ and ‘fair to poor’ result for the anterior group were not significantly different with the posterior group in the ‘follow-up time ≤5 years’ subgroup (p > 0.05). No study reporting the results for the follow-up time >5 years.

(c) Complication outcomes: the complications for the anterior group were significantly more than the posterior group in the ‘follow-up time ≤5 years’ subgroup (p < 0.05, OR 7.33 [2.96, 18.20]), but were similar to the posterior group (p > 0.05) in the ‘follow-up time >5 years’ subgroup. In total, it was significantly more than the posterior group too (p < 0.05, OR 2.97 [1.76, 5.00]).

The axial pain for the anterior group was similar to the posterior group (p > 0.05). As only two studies reported the result, we did not do subgroup analysis.

Table 4 shows the detail results of the meta-analysis of clinical and complication outcome measures for multilevel CSM patients.

Table 4 Results of meta-analysis of clinical and complication outcome measures for anterior versus posterior surgery for multilevel CSM patients
Full size table
For OPLL patients

(a) Perioperative outcomes: only one study [31] reported the perioperative outcomes. The results showed mean operative time was 302 min (range 167–470 min) and mean blood loss was 513 g (range 70–1,730 g) for the anterior group. The corresponding values for the posterior group were 177 min (range 90-395 min) and 464 g (range 50-1,800 g), respectively. No direct comparison was done.

(b) Clinical outcomes: the final follow-up JOA score and recovery rate for the anterior group were both significantly higher than the posterior group (p < 0.05, WMD 2.50 [0.16, 4.85]; p < 0.05, WMD 29.48 [29.09, 29.87], respectively) in the ‘follow-up time ≤5 years’ subgroup. One study [31] which mean follow-up time was 6 years was enrolled in the ‘follow-up time >5 years’ subgroup. The results showed there was no significant difference in final follow-up JOA score and recovery rate between anterior and posterior group for patients with occupying ratio of OPLL <60% (p > 0.05), while for patients with occupying ratio ≥60%, the final follow-up JOA score and recovery rate of anterior surgery was superior to that of posterior surgery (p < 0.05).

The ‘excellent to good’ result for the anterior group was significantly better than the posterior group in the ‘follow-up time ≤5 years’ subgroup (p < 0.05 OR 3.38 [1.06, 10.84]), but there were no significant differences between the two groups in the ‘follow-up time >5 years’ subgroup and in total (p > 0.05) too. The ‘fair to poor’ result for the anterior group was not significantly different with the posterior group in the ‘follow-up time ≤5 years’ subgroup, in the ‘follow-up time > 5 years’ subgroup and in total (p > 0.05).

(c) Complication outcomes: the complications for the anterior group were significantly more than the posterior group in the ‘follow-up time ≤5 years’ subgroup (p < 0.05, OR 4.44 [1.80, 10.98]), but were similar to the posterior group (p > 0.05) in the ‘follow-up time >5 years’ subgroup. In total, it was significantly more than the posterior group too (p < 0.05, OR 3.12 [1.60, 6.07]).

Table 5 shows the detail results of the meta-analysis of clinical and complication outcome measures for multilevel OPLL patients.

Table 5 Results of meta-analysis of clinical and complication outcome measures for anterior versus posterior surgery for multilevel OPLL patients
Full size table

The main complications for the anterior group were graft complications, dysphagia and dysphonia for CSM patients and CSF leakage and graft complications for OPLL patients. The complications for the posterior group were similar between OPLL patients and CSM patients, which were mainly C5 root palsy (Table 6) displays complications for anterior and posterior surgery in each study.

Table 6 Complications for anterior and posterior surgery in each study
Full size table

(d) Radiographic outcomes: three studies [29, 34, 35] reported the mean decrease rate of ROM. However, as they all did not report the standard deviation, we could not do meta-analysis for it. Two studies [29, 34] which mean follow-up time were <50 months both showed that the decrease rate of ROM for the anterior group was more than the posterior group, while they also reported there were no statistic different (p > 0.05). However, the other study [35] which mean follow-up time was longer than 11 years showed that the decrease rate of ROM for the anterior group was significantly less than the posterior group (p < 0.05). Table 7 shows the detail result of mean decrease rate of ROM.

Table 7 Mean decrease rate of ROM for anterior versus posterior surgery
Full size table

The alignment deterioration for the anterior group was similar to the posterior group (p > 0.05). However as only two studies [31, 35] reported the result, we did not do subgroup analysis.

Only one study [29], which follow-up time <50 months reported the mean change of anterior–posterior (AP) diameter of the narrowest spinal canal and mean change of regional Cobb’s angle for the two groups. The mean change in the AP diameter was 9.1 mm for anterior surgery and 4.11 mm for posterior surgery (p < 0.05). The mean change in the regional Cobb’s angle was 1.7° and −3.1°, respectively (p > 0.05). Two studies [28, 29] reported the mean change in the overall Cobb’s angle was similar for the two groups (p > 0.05).

Discussion

The results of this systematic review have shown that for both multilevel CSM and OPLL patients compared with posterior surgery, anterior surgery had longer operative time, more operative blood loss. Anterior surgery had higher final follow-up JOA score and higher recovery rate with more complications in “follow-up time ≤5 years” subgroup. In "follow-up time >5 years" subgroup, complications were similar between the two groups. Anterior surgery had similar final follow-up JOA score with posterior surgery for CSM patients and OPLL patients with occupying ratio of OPLL <60%, while in OPLL patients with occupying ratio ≥60%, anterior surgery had superior final follow-up JOA score and recovery rate to posterior surgery.

In this systematic review, no language restrictions were applied and we chose lots of keywords and searched as more databases as possible to search all relevant articles in the world to avoid selection bias. To minimize bias in the selection of studies and data extraction, reviewers independently selected articles on the basis of inclusion criteria and extracted characteristics by using a standardized form. Any disagreement was resolved by discussion.

To do an ideal systematic review, we would better enroll randomized controlled trials (RCT) with homogeneity. However, in practice, RCT is very rare, especially for surgery [38, 39]. In this systematic review, all the studies selected were not RCT, while it did not influence the credibility of the results. Because almost all the studies reported the baseline characteristics were matched for each group and the results for the meta-analysis of baseline characteristics also showed no significant difference between the two groups.

Most patients with cervical myelopathy have pre-existing cervical spinal stenosis and then are affected by progressive cervical spondylosis or ossification of the posterior longitudinal ligament which compress the spinal cord anteriorly. When the cervical spinal canal diameter is reduced, the amount of room available for the spinal cord decreased. Then, when the cervical spine is in hyperextension, the result is buckling of the ligamentum flavum posteriorly, and the cord may become compressed against an anterior osteophyte or ossification lesion.

The advantage for anterior surgery is that it is more radical than posterior surgery in decompressing the nerve tissue by directly removing all of the anterior pathogenic structures such as protruded discs, osteophyte or ossification lesion. With grafting, immediate stability of cervical spine can be achieved. In addition, with anterior stabilization, there is no longer concern over buckling of ligamentum flavum posteriorly. In our systematic review, it was shown that during the early stage, the anterior surgery was significantly better than posterior in the final follow-up JOA score and the recovery rate. Anterior surgery had better clinical outcomes. However, we also found that during the late stage, there was no statistically difference in the final follow-up JOA score, “excellent to good” result, and “fair to poor” result between anterior surgery and posterior surgery except OPLL patients with occupying ratio ≥60%. It indicated that these two procedures were generally identical with respect to long-term clinical outcomes. That is possibly because late deterioration secondary to degenerative changes at the adjacent levels after anterior surgery [40, 41].

Another important disadvantage of anterior surgery is it leads to more surgical complications than posterior surgery at early stage after operation. For CSM patients, the main complications for anterior group were complications related to bone grafting, causing major concern during the postoperative course, dysphagia and dysphonia. For OPLL patients, the main complications for anterior group were CSF leakage and also complications related to bone grafting. The main complication for the posterior group between OPLL patients and CSM patients was both C5 root palsy. The cause of this complication maybe tethering of the nerve roots by fibrosis or spondylotic changes at the root canal and such tethering plus ventral or dorsal shift of the spinal cord after decompression would exert traction on the root or rootlet leading to intractable pain and paralysis [42].

As for posterior surgery, it is more concerned about “total decompression effect” resulting from posterior shift of spinal cord. It decompresses the nerve tissue indirectly. From this systematic review, we can see posterior surgery had identical long-term clinical outcomes to anterior surgery for CSM patients and OPLL patients with occupying ratio of OPLL <60%. While, in patients with occupying ratio ≥ 60%, it had inferior outcome. The results indicated the indirectly decompression was effective for most cervical myelopathy patients, but if anterior pathogenic structures were too big, the effectiveness was limited. In this condition, anterior surgery or combination of the two approaches is better. Preoperative cervical alignment is another important factor for selection of posterior surgery. In general, cervical kyphosis patients are not suitable for posterior surgery.

Besides anterior surgery and posterior surgery, anteroposterior surgery combining the anterior and posterior approaches is another option for MCM. Sometimes, it is optimal correction of deformity and fusion. The most appropriate use of this combined approach may be in the setting of severe multilevel spondylosis and stenosis requiring more than three inter-vertebral decompressions or two vertebrectomies, multilevel spondylosis and substantial kyphosis, and post-traumatic or post-laminectomy kyphosis [3]. As no controlled study is concerned with anteroposterior surgery for MCM, anteroposterior surgery was not included in this systematic review.

The limitation of this study is relatively small number of qualified studies, especially the studies with long follow-up time (>10 years). Another limitation is most studies did not report preoperative cervical alignment (kyphosis or lordosis) very clearly except Iwasaki [31], in whose study anterior group included 15 lordotic, 7 straight and 5 kyphotic patients and posterior group included 32 lordotic, 29 straight and 5 kyphotic patients. The other limitation of this study is although we used lots of methods to avoid selection bias, 5 of the 10 papers enrolled were from Japan, only 1 paper from the USA and 1 paper from Germany. Thus, this analysis may have regional bias.

On the basis of the results of this systematic review, we can conclude that anterior surgery had better clinical outcomes and more complications at the early stage after operation for both multilevel CSM and OPLL patients. At the late stage, posterior surgery had similar clinical outcomes and complications to anterior surgery for CSM patients, and OPLL patients with occupying ratio of OPLL <60%, while for OPLL patients with occupying ratio ≥60%, anterior surgery had superior clinical outcome to posterior surgery.