Introduction

Cervical spondylotic myelopathy (CSM) is an age-related disorder associated with degeneration of intervertebral discs and adjacent vertebral structures. It is the most common spinal cord disorder in elderly patients, resulting in progressive spinal canal narrowing and subsequent nerve root compression [1, 2]. Early surgical treatment can alter the natural history of CSM as well as improve prognosis in selected patients [3].

For single-level CSM, an anterior surgical approach is typically employed as it confers greater ability to address compressing lesions while minimizing trauma and muscle loss [3, 4]. The anterior approach has also been associated with superior ability to provide decompression over kyphotic segments. However, anterior cervical fusion for multi-level CSM represents a more complex procedure and may be associated with longer operative times as well as complications such as graft dislodgement, hoarseness, dysphagia and trigeminal nerve palsy [2].

With regard to posterior surgical approaches, laminectomy alone has been the historical treatment standard for CSM decompression [5]. More recently, laminectomy has been performed in conjunction with lateral mass fixation or fusion to reduce the incidence of post-operative segmental instability and kyphosis [6–8]. Laminoplasty represents an alternative posterior approach for CSM that may enable better preservation of cervical motion, adjacent structural integrity and natural lordosis [9]. However, some studies have reported less effective and less extensive cord decompression for laminoplasty compared to laminectomy and fusion.

To address limitations in the current literature, the present meta-analysis was conducted to systematically compare the safety and efficacy of the two posterior approaches for multi-level CSM (LF vs EL) with regards to post-operative patient-rated scores and complication rates.

Methods

Literature search strategy

The present review was conducted according to the PRISMA guidelines [10, 11]. Electronic searches were performed using Ovid Medline, PubMed, Cochrane Central Register of Controlled Trials (CCTR), Cochrane Database of Systematic Reviews (CDSR), ACP Journal Club and Database of Abstracts of Review of Effectiveness (DARE) from their dates of inception to January 2016. To achieve maximum sensitivity of the search strategy and to identify all studies, we combined the terms: “cervical”, “laminectomy”, and “laminoplasty” as either keywords or MeSH terms. The reference lists of all retrieved articles were reviewed for further identification of potentially relevant studies. All identified articles were systematically assessed using the inclusion and exclusion criteria.

Selection criteria

Eligible comparative studies for the present systematic review and meta-analysis included: (1) patients undergoing extended laminoplasty (EL) or laminectomy and fusion (LF) and (2) patients with CSM caused by multi-segment spinal stenosis. When institutions published duplicate studies with accumulating numbers of patients or increased lengths of follow-up, only the most complete reports were included for quantitative assessment at each time interval. All publications were limited to those involving human subjects and in the English language. Abstracts, case reports, conference presentations, editorials and expert opinions were excluded. Review articles were omitted because of potential publication bias and duplication of results.

Data extraction and critical appraisal

All data were extracted from article texts, tables and figures. Two investigators independently reviewed each retrieved article (K.P. and V.L.). From the studies, extracted data included: (1) study design and description including country, study period, length of follow-up, patient distribution, (2) patient baseline demographics and comorbidities, (3) mean Japanese Orthopaedic Association (JOA) grade before and after posterior surgeries, (4) mean visual analogue scale (VAS) before and after surgery, (5) mean cervical curvature index (CCI) before and after surgery, and (6) complications including reoperations and nerve palsies. Discrepancies between the two reviewers were resolved by discussion and consensus with a third reviewer (R.J.M). Assessment of risk of bias for each selected study was performed according to the most updated Cochrane statement. Discrepancies between the two reviewers were resolved by discussion and consensus. The final results were reviewed by the senior investigator (R.J.M).

Statistical analysis

The weighted mean difference (WMD) and relative risk (RR) was used as a summary statistic. In the present study, both fixed- and random-effect models were tested. In the fixed-effects model, it was assumed that treatment effect in each study was the same, whereas in a random-effects model, it was assumed that there were variations between studies. χ2 tests were used to study heterogeneity between trials. I 2 statistic was used to estimate the percentage of total variation across studies, owing to heterogeneity rather than chance, with values greater than 50 % considered as substantial heterogeneity. I 2 can be calculated as: I 2 = 100 % × (Q − df)/Q, with Q defined as Cochrane’s heterogeneity statistics and df defined as degree of freedom. If there was substantial heterogeneity, the possible clinical and methodological reasons for this were explored qualitatively using remove-one-study sensitivity analysis, as well as sub-group analysis comparing RCT vs non-RCT studies. In the present meta-analysis, the results using the random-effects model were presented to take into account the possible clinical diversity and methodological variation between studies. Specific analyses considering confounding factors were not possible because raw data were not available. All P values were 2-sided. All statistical analysis was conducted with Review Manager Version 5.3 (Cochrane Collaboration, Software Update, Oxford, UK).

Quality of evidence and publication bias

Interstudy risk of publication bias was assessed using funnel plot methodology. The funnel plot represented the estimate (proportion of an event) in the x-axis vs its precision (inverse of the standard error of the estimate) in the y-axis. Significant asymmetry indicates potential publication bias, which may have affected the validity of presented results.

The checklist by Furlan et al. [12] was used to evaluate methodological quality of randomized controlled studies. Risk of bias assessment was performed using the checklist proposed by Cowley et al. [13] for non-randomized studies. The items were scored with “yes”, “no”, or “unclear”. A Furlan score of 6 or more out of a possible 12, or a Cowley score of 9 or more out of a possible 17, was considered to reflect “high methodological quality”. These studies were independently assessed by two reviewers, and any discrepancies were resolved by discussion and consensus.

Results

Quality of studies

A total of 1035 references were identified through six electronic database searches. After exclusion of duplicate or irrelevant references, 1021 potentially relevant articles were retrieved. After detailed evaluation of these articles, 47 studies remained for detailed assessment. After applying selection criteria, 10 studies [5, 14–22] were selected for analysis (Fig. 1). The study characteristics of these trials are summarized in Table 1. From these 10 included studies, there were 335 patients who underwent laminectomy and fusion compared to 320 patients who underwent extended laminoplasty. Of these, two studies were prospective randomized trials [17, 22] whilst the remainder [5, 14–16, 18–21] were retrospective observational studies.

Fig. 1
figure 1

PRISMA flow chart of systematic review and meta-analysis comparing laminectomy and fusion vs extended laminoplasty

Full size image
Table 1 Characteristics and demographics of studies included in the present systematic review and meta-analysis
Full size table

The mean age in the LF group ranged from 52.6 to 69.2 years, compared to the EL group, which ranged from 46.3 to 66.1 years. Overall, there was no significant difference in age between the groups (P = 0.08). The proportion of males was found to be similar between LF and EL groups (P = 0.25). Baseline JOA was found to be similar between the two groups (WMD −0.28, 95 % CI −0.96, 0.39, I 2 = 89 %, P = 0.41). Likewise, baseline VAS scores were similar between LF and EL groups (WMD 0.65, 95 % CI −1.15, 2.45, I 2 = 83 %, P = 0.48), as well as CCI. The average follow-up ranged from 12 to 110 months.

Assessment of JOA and VAS scores

Postoperative JOA score was reported in nine included studies. There was no significant difference found postoperatively between LF and EL groups (WMD −0.36, 95 % CI −1.19, 0.47, I 2 = 96 %, P = 0.39) (Fig. 2).

Fig. 2
figure 2

Forest plot of patient clinical outcomes. a Postoperative JOA, b postoperative VAS neck pain, c postoperative CCI, d postoperative Nurich grade

Full size image

Postoperative VAS scores were reported in five studies. No significant difference in postoperative VAS score was found between LF and EL cohorts (WMD 0.05, 95 % CI −1.00, 1.10, I 2 = 89 %, P = 0.93).

Assessment of CCI

Postoperative CCI was reported in four studies. No significant difference was found between LF and EL groups (WMD 0.31, 95 % CI −0.30, 0.93, I 2 = 50 %, P = 0.32).

Assessment of Nurich grade

The postoperative Nurich grade was reported in two studies. There was no significant difference found between LF and EL cohorts (WMD −0.29, 95 % CI −1.00, 0.42, I 2 = 83 %, P = 0.42). Remove-one-study sensitivity analysis as well as RCT vs non-RCT subgroup analysis for assessment of JOA, VAS, CCI and Nurich outcomes did not demonstrate any significant change to the trends of the results presented.

Assessment of complications

From seven studies, the total complication rate was pooled (Fig. 3). A significantly higher complication rate was found for LF compared to EL (26.4 vs 15.4 %, RR 1.77, 95 % CI 1.10, 2.85, I 2 = 34 %, P = 0.02). Reoperation rate was reported in three studies. This rate was found to be similar between LF and EL groups (5.7 vs 4.9 %, RR 1.34, 95 % CI 0.55, 3.30, I 2 = 0 %, P = 0.52). For nerve palsies, a significantly higher pooled rate was found in the LF group compared to EL (9.9 vs 3.7 %, RR 2.76, 95 % CI 1.10, 6.92, I 2 = 26 %, P = 0.03).

Fig. 3
figure 3

Forest plot of a total complications, b reoperation rate, and c nerve palsies

Full size image

Assessment of operative time and intraoperative blood loss

From 3 studies, no significant difference in operative time was determined (WMD 1.95 min, 95 % CI −39.95, 40.85, I 2 = 95 %, P = 0.92). Intraoperative blood loss was also similar between LF and EL cohorts (WMD −29.58 mL, 95 % CI −120.57, 61.42, I 2 = 98 %, P = 0.52).

Assessment of evidence quality

Publication bias was assessed using funnel plots (Fig. 4). Significant asymmetry was noted and detected, suggesting that publication bias was an influencing factor for the results of the present study.

Fig. 4
figure 4

Funnel plots for a postoperative JOA, b postoperative VAS, c postoperative CCI, d postoperative Nurich grade, e total complication rates, f reoperations, and g nerve palsy

Full size image

In terms of risk of bias, the Furlan scores for the two included RCTs ranged from 6 to 8 points out of 12 (Table 2). All RCTs received Furlan scores of 6 or higher, indicating overall lower risk of bias in the randomized trials. The most notable shortcomings was unclear descriptions of patient blinding to intervention, care provide blinding to intervention, outcome assessor blinding to intervention, and compliance. The Cowley scores for the non-randomized studies ranged from 10 to 14 out of 17 (Table 3). All studies scored Cowley scores >9, demonstrating high methodological quality.

Table 2 Risk of assessment of randomized controlled studies using the checklist by Furlan et al. [12]
Full size table
Table 3 Risk of assessment of non-randomized controlled studies using the checklist by Cowley et al. [13]
Full size table

Discussion

Our findings demonstrate that LF and EF had similar postoperative patient-rated disability outcomes. There was no significant difference in JOA and VAS scores, with the assessment of CCI revealing comparable outcomes between the two surgical approaches. These results were consistent with a prior review and meta-analysis [23], which concluded that LF and EL lead to equivalent clinical improvement and loss of lordosis. The authors included that there is no evidence to support EL over LF in the treatment of multi-level CSM; however, the relative complication rates between the two approaches was not clearly established. In the present study, we demonstrated that the total complication rate in LF was around twofold higher than that of EF. In particular, LF was associated with significantly higher nerve palsy complications and trended towards higher reoperation rates.

Laminectomy is a widely used surgical procedure for CSM as it allows extensive decompression of the spinal cord. Overall, it is considered to be a safe and effective treatment method, with high overall rates of success ranging from 62 to 70 % [24]. However, laminectomy alone has been shown to compromise spinal column flexibility and stability especially at the lower level of laminectomy [25]. Subsequently, repetitive spinal cord microtrauma can result from this segmental instability, detrimentally affecting neurological results and long-term stability [26]. These clinical findings are supported by biomechanical studies on animals, which found that laminectomy resulted in a significant increase in forward sagittal angulation along with a decrease in spine stability along the sagittal plane [27]. Laminectomy is also not optimal in patients with congenitally short pedicles, due to the formation of post-laminectomy membranes [28, 29]. For these patients, while the procedure may initially decompress the spinal cord, the scar that forms over the dura mater can recompress the spinal cord as it adheres to the lateral masses and contracts [25].

More recently, laminectomy has been performed in conjunction with lateral mass fixation or fusion to address the post-operative instability that results from posterior lamina removal. The outcomes of laminectomy with posterior fusion or fixation have been promising, with lower detected levels of kyphosis, segmental instability and neurological deterioration [8]. However, cervical spine fusion compromises the natural biomechanics of spinal movement, causing axial and rotational forces to become unevenly distributed over adjacent spinal structures [20]. As a result of this, fusion procedures have been implicated with elevated levels of degeneration in adjacent spinal segments [30]. Screw loosening, screw avulsion and broken plates have also been reported in a few cases of laminectomy with cervical lateral mass screws/plates for fixation/fusion [9, 31].

To address the issues of segmental instability, late kyphosis, and neurological deficits from a combined laminectomy and fusion approach, laminoplasty was developed as to surgically treat multilevel CSM with lower levels of late deformity and neurological defects. In laminoplasty, the main advantage stems from the fact that the laminae are still available for load bearing and attachment of paraspinous muscles [32]. In fact, animal models showed that a spine treated with laminoplasty was very similar to that of an intact spine [27]. Thus, the natural biomechanics of spinal movements are preserved following laminoplasty, reducing the risk of spinal segment deterioration, daily microtrauma and instability [25]. Along with that, laminae retention in laminoplasty also prevents the formation of a post-laminectomy membrane, which can recompress the spinal cord [29]. Complications of laminoplasty have been reported to include neck spasms and shoulder pain. Another potential complication occurring in 5–11 % of laminoplasty patients is motor dominant C5 root palsy [33]. It is hypothesized that laminoplasty can cause posterior drift at the level of C5, placing the C5 nerve root under stretch and causing the palsy [34]. Mechanical tethering of the nerve root in the foramina with concurrent posterior cord migration may stretch the C5 nerve and further exacerbate the palsy. This theory; however, does not provide an explanation as to why C5 palsies still occur following anterior decompression. In contrast for posterior reconstruction surgery, Kurakawa et al. [35] demonstrated correction angle and preoperative diameter of the C4/5 intervertebral foramen as predictors of C5 palsy, following multivariate adjustment for confounders. Ultimately, while surgical management of CSM has advanced greatly over the years, all our current treatment options still present various potential complications.

The analysis of our current study is limited by a few factors. The studies included, consisted of both randomized and non-randomized observational studies. This presents the issue of selection bias, which limits the internal validity of our findings. There were also underlying baseline differences in the various studies, with some patients already with preoperative kyphosis and segmental stability. This can contribute to the underestimation or overestimation of the actual effect of a surgical procedure delivered. Across centers, there were also variations in the technical nuances of the surgical procedure, with laminoplasty being performed with different techniques such as French door and open door. Along with that unstandardized reporting for quality and satisfaction of outcomes was also a limiting factor. Recent studies have also suggested that sagittal alignment may affect surgical outcome in degenerative cervical myelopathy for different approaches [36, 37]. Based on a comparative analysis of three techniques, Koda et al. [36] suggested that posterior or anterior decompression and fusion but not laminoplasty may be used for K-line cervical ossification of the posterior longitudinal ligament. The impact of preoperative sagittal alignment on postoperative outcomes in the present meta-analysis could not be assessed due to the nature of data reported by the included studies. A lack of heterogeneity in regards to follow-up period also makes it difficult to make definitive conclusions about the long-term efficacy of the two surgical approaches. For example, late neurological deterioration is a known complication of laminoplasty for cervical spondylotic myelopathy [38]. These factors all add uncontrollable bias to our findings. Ideally, to minimize bias the operative technique, defined outcomes and follow-up period should be standardized across all cases.

Conclusions

From the available low-quality evidence, LF and EL approaches for CSM demonstrates similar clinical improvement and loss of lordosis. However, a twofold higher complication was found in LF group, including significantly higher nerve palsy complications. This requires further validation and investigation in larger sample-size prospective and randomized studies.