Introduction

Lumbar spinal stenosis (LSS) is characterized by narrowing of the central vertebral canal, lateral recesses, and vertebral foramina. LSS causes patients significant long-term symptoms (e.g., intermittent neurogenic claudication, radicular back and leg pain), and patients often do not respond well to conservative treatments such as physiotherapy, analgesics, and steroids [1–4]. Decompression surgery can achieve neural compromise and improve the associated pain by relieving lumbar canal stricture and removing redundant tissue. Laminectomy, as one technique of decompression surgery, is a recommended surgical approach for LSS [5]. Unilateral laminotomy with bilateral decompression (ULBD), a less invasive surgical technique proposed by Chang et al. was demonstrated to immediately and substantially improve the physical score and bodily pain score in patients with LSS [6]. However, many post-operative complications and other lumbar degenerative diseases, especially spondylolisthesis, were associated with decompression. Mardjetko et al. reported a high incidence (31%) of slip progression in laminectomy alone for LSS with degenerative spondylolisthesis (DS) [7].

Fusion can more appropriately treat the defect of progressive lumbar instability after decompression. In the last few decades, decompression plus fusion had become very popular and was regarded as a gold standard for LSS. However, the debate over its use was never settled, due to spinal fusion being a more traumatic procedure, requiring a longer operative time, causing more blood loss, and exhibiting an increased complication rate in extremely elderly patients [8–11]. Many authors reported similar treatment outcomes between decompression alone and decompression plus fusion among the majority of patients with LSS and DS, and they sought to uphold the view of there being no absolute need for additional fusion [12–16].

Because there is a paucity of evidence—particularly in primary evidence—supporting either argument, we conducted a meta-analysis aiming to compare the efficacy of decompression alone versus decompression plus fusion for LSS. We also discussed the optimal indications for performing additional surgical fusions to treat this condition.

Methods

Literature search and evaluation

An online search was performed in the Web of Science, PubMed, Embase, and Springer databases, from January 1970 to May 2016. Relevant references were selected and the included studies were manually reviewed. The search strategy is detailed in Fig. 1. To facilitate future updates of this systematic review and meta-analysis, we present the search strategy as follows: (laminotomy OR laminectomy OR fenestration OR hemilaminectomy OR decompression) AND (lumbar spondylolisthesis OR lumbar spinal stenosis OR lumbar canal stenosis OR degenerative lumbar spondylolisthesis) AND (fusion OR arthrodesis).

Fig. 1
figure 1

Flow diagram of studies included in the systematic review

Full size image

Eligibility criteria

Included studies fulfilled the following criteria: (1) they were randomized controlled trials (RCTs) or clinical cohort studies written in English; (2) the studies assessed the comparison between decompression alone versus fusion plus decompression surgery for LSS; (3) the LSS was with or without grade I DS; and (4) the studies reported the means and standard deviations of intra- and post-operative assessments with sample size between decompression alone group and decompression plus fusion group, as well as the reported number and total number of major complications, reoperations, and clinically excellent and good rates between the two groups [or alternately the studies provided sufficient data to construct those contingency tables (the difference between pre- and post-operative back and leg pain, EQ-5D, and ODI in Försth et al. [40] and Ghogawala et al. [17])] were constructed by Graphpad instat 3.0, which was developed by GraphPad software company in America.

Exclusion criteria

Studies were excluded if they were: (1) non-English-language articles, case reports, duplicate papers, or conference reports; (2) original articles without a controlled group or with a small sample size (<16 patients); (3) studies investigating mixed lumbar degenerative diseases (LDD), tumors, fractures, osteoporosis, or other non-degenerative diseases; (4) studies not specifically concerning decompression alone and fusion plus decompression surgery; (5) studies mainly evaluating a surgical approach, or new/non-standardized surgical techniques or instruments; or, finally, (6) studies with incomplete or unacceptable information for a comparison.

Data collection and methodological quality

Two authors independently sorted and reviewed all abstracts or full texts of the retrieved articles based on eligibility and exclusion criteria. Disagreements were resolved by consensus between two senior orthopedics. Data were extracted and summarized as follows: (1) the basic characteristics of each study: author, publication year, country, age, sex ratio, research period, comorbidities, surgery type, and follow-ups (within a 2-year period) were reported; (2) the primary outcomes of back pain and leg pain (when comparison was shown between back and leg pain, BP > LP were selected), walking ability, major quality-of-life EuroQol-5 Dimensions (EQ-5D), Oswestry Disability Index (ODI) scores, and Japanese Orthopaedic Association (JOA) scores (as many articles did not offer total scores in these questionnaires, partial sub-items that met our measures were abstracted instead) were reported; (3) secondary outcomes that included intraoperative surgical data (blood loss and duration of operation) and patient-reported outcomes (length of stay, complications, reoperations, and clinical satisfaction) were reported. All studies were assessed for quality evaluation according to the classic Newcastle–Ottawa Scale (NOS). Scale scores range from zero to nine points, with higher scores indicating better quality.

Data analysis

Statistical analyses were performed using the Cochrane Collaboration’s Review Manager 5.2 software. Pooled weighted mean differences (WMDs) with 95% confidence intervals (CIs) for continuous variables and pooled odds ratios (ORs) with 95% CIs for enumeration data were calculated. A Z test was performed to determine the overall effects. If the heterogeneity between studies was statistically significant (I 2 ≥ 50%), a random effects model was used for further sensitivity analysis. Otherwise, a fixed effects model was selected (I 2 < 50%). Influential analysis was examined by removing one individual study at one time to check heterogeneity that biased the overall estimate. Two-sided P ≤ 0.05 was considered statistically significant.

Results

A total of 4442 studies in the Web of Science (1063), PubMed (1643), Embase (732), and Springer (1004) databases were reviewed. After perusing the titles and abstracts, 53 relevant studies were identified, and, with further meticulous examination, 15 studies were finally selected that met our eligibility criteria and were transferred to data abstraction for statistical analysis (Fig. 1). A description of the main characteristics of all studies is listed in Table 1.

Table 1 Description of the main characteristics of the studies included in the meta-analysis
Full size table

Results of meta-analysis

Basic characteristics

A total of 17,785 cases in 15 articles were finally enrolled in our study, including those from five controlled trials. Overall, 12,417 patients with LSS received decompression surgery alone, compared to 5368 patients who underwent decompression plus fusion surgery. The number of females and males was available in 12 studies containing 6549 females and 3187 males, with a sex ratio (F/M) of 2.05. Overall, the average age of patients with LSS could not be calculated, because most studies could be separately described according to sex or surgical strategy. While the likely dominant age was approximately 65 years, females were higher in numbers and average onset age than males.

Primary measures

Back pain

Eight studies reported changes in back pain between the two subgroups, including two based on JOA and six based on VAS. With regard to the overall results of heterogeneity testing, there was a statistically significant difference between the two groups (P = 0.003, I 2 = 67%), and a random effects model was applied for meta-analysis (Fig. 2). No statistically significant difference was found in the changes between pre- and post-operative back pain evaluated according to JOA and VAS between the two groups [JOA subgroup, MD = 0.01, 95% CI (–0.41, 0.42), z = 0.03, P = 0.98; VAS subgroup, MD = 0.01, 95% CI (–0.52, 0.54), z = 0.04, P = 0.97].

Fig. 2
figure 2

Forest plot of mean difference for back pain of patients with LSS within 2 years follow-up

Full size image

Leg pain

Seven studies reported changes in leg pain between the two subgroups, including six based on VAS and one according to JOA. A heterogeneity test indicated a statistically significant difference between the two groups (P < 0.00001, I 2 = 96%), and a random effects model was applied for meta-analysis (Fig. 3). In the VAS subgroup, MD = 0.79, 95% CI (–0.47, 2.05), z = 1.23, and P = 0.22; in the JOA subgroup, MD = 0.10, 95% CI (–0.23, 0.43), z = 0.59, and P = 0.55. These results demonstrated that the differences in pre- and post-operative leg pain were not significantly different between the two groups.

Fig. 3
figure 3

Forest plot of mean difference for leg pain in patients with LSS within 2 years follow-up

Full size image

ODI

Five studies reported ODI in the two groups. The heterogeneity test showed no statistically significant difference between the two groups (P = 0.25, I 2 = 25%), and a fixed effects model was applied for meta-analysis (Fig. 4). There was no significant difference in ODI between the decompression group and the decompression plus fusion group (MD = 0.43, 95% CI (–1.15, 2.00), z = 0.53, P = 0.59).

Fig. 4
figure 4

Forest plot of mean difference for ODI in patients with LSS within 2 years follow-up

Full size image

EQ-5D

Two studies reported EQ-5D in the two groups. There was no statistically significant heterogeneity between the two groups (P = 0.34, I 2 = 0%), and a fixed effects model was applied for meta-analysis (Fig. 5). No statistically significant difference was identified between the two groups.

Fig. 5
figure 5

Forest plot of mean difference for EQ-5D in patients with LSS within 2 years follow-up

Full size image

Duration of operation

Three studies reported the duration of operation in the two groups. There was statistically significant heterogeneity between the two groups (P < 0.00001, I 2 = 96%). A random effects model was applied for meta-analysis (Fig. 6), which indicated that the decompression plus fusion group underwent more operative time than the decompression alone group.

Fig. 6
figure 6

Forest plot of standard mean difference for duration of operation in patients with LSS within 2 years follow-up

Full size image

Intraoperative blood loss

Three studies reported intraoperative blood loss in the two groups. Statistically significant heterogeneity was found between the two groups (P < 0.00001, I 2 = 97%). A random effects model was applied for meta-analysis (Fig. 7), which demonstrated that the blood loss in the decompression alone group was significantly less than in the decompression plus fusion group.

Fig. 7
figure 7

Forest plot of standard mean difference for intraoperative blood loss in patients with LSS with in 2 years follow-up

Full size image

Length of hospital stay

Two studies reported the length of hospital stay in the two groups. There was statistically significant heterogeneity between the two groups (P = 0.006, I 2 = 87%), and a random effects model was applied for meta-analysis (Fig. 8). The length of hospital stay was not found to be statistically different between the decompression alone group and decompression plus fusion group.

Fig. 8
figure 8

Forest plot of mean difference for length of hospital stay in patients with LSS within 2 years follow-up

Full size image

Major complications

Five studies reported major complications in the two groups with a heterogeneity test showing relatively lower statistically significant heterogeneity between groups (P = 0.30, I 2 = 18%). A fixed effects model was applied for meta-analysis (Fig. 9), which indicated that the decompression plus fusion group has approximately 1.4 times higher risk of sustaining major complications than the decompression alone group.

Fig. 9
figure 9

Forest plot of odds ratio for risk of major complications in patients with LSS within 2 years follow-up

Full size image

Walking ability

Two studies reported walking ability evaluated according to JOA in the two groups. No statistically significant heterogeneity was found between the two groups (P = 0.58, I 2 = 0%), and a random effects model was applied for meta-analysis (Fig. 10). The walking ability recovery among patients receiving decompression plus fusion surgery was not better than those in the decompression alone group.

Fig. 10
figure 10

Forest plot of mean difference for walking ability in patients with LSS within 2 years follow-up

Full size image

Number of reoperation

Three studies reported reoperation numbers in the two groups. Extremely low heterogeneity was identified between the two groups (P = 0.48, I 2 = 0%), and a fixed effects model was applied for meta-analysis (Fig. 11). There was no statistically significant difference between the two groups [OR = 1.04, 95% CI (0.90, 1.19), z = 0.52, P = 0.60].

Fig. 11
figure 11

Forest plot of odds ratio for number of reoperations in patients with LSS within 2 years follow-up

Full size image

Clinically excellent and good rates

Four studies reported the clinically excellent and good rates in the two groups. A significantly different heterogeneity was found between the two groups (P = 0.010, I 2 = 74%), and a random effects model was applied for meta-analysis (Fig. 12). The clinical satisfaction in the decompression plus fusion group was not better than in the decompression alone group [OR = 0.31, 95% CI (0.06, 1.59), z = 1.40, P = 0.16].

Fig. 12
figure 12

Forest plot of odds ratio for clinically excellent and good rates in patients with LSS within 2 years follow-up

Full size image

Sensitivity analysis

When the random effects models were applied, the overall effect MD (95% CI) of the difference in pre- and post-operative back pain and leg pain, operative time, intraoperative blood loss, and length of hospital stay were 0.04 (–0.36, 0.44), 0.69 (−0.38, 1.76), −2.04 (−3.12, −0.96), −3.96 (−6.64, −1.27) and −4.21 (−10.03, 1.62) (z = 0.18, 1.26, 3.71, 2.92 and 1.41, respectively; P = 0.86, 0.55, 0.0002, 0.004 and 0.16, respectively). When the fixed effects models were applied, the overall effect MD (95% CI) of ODI, EQ-5D and walking ability were 0.43 (−1.15, 2.00), 0.01 (−0.01, 0.03) and 0.04 (−0.49, 0.57) (z = 0.52, 1.16 and 0.15, respectively; P = 0.59, 0.24 and 0.88, respectively). The overall effect OR (95% CI) of major complications, number of reoperations, and clinically excellent and good rates between the two groups were 0.70 (0.60, 0.81), 1.04 (0.90, 1.19) and 0.31 (0.06, 1.59) (z = 4.63, 0.53 and 1.40, respectively; P < 0.00001, 0.60 and 0.16, respectively). Except for blood loss, the residual results were consistent between the random and fixed effects models, suggesting that our findings were reliable.

Publication bias

Considering the small sample size (<10) in our meta-analysis, funnel plot analysis was not applicable for publication bias.

Discussion

The limited number of included publications, the relatively low quality, and the inconsistent descriptions of some measures in the 15 selected studies greatly affected the quality of our meta-analysis. Also, the converted parameters possibly impaired the stability of our final outcomes. In our meta-analysis, there was no difference in the effectiveness between decompression alone versus decompression plus fusion; however, patients in the decompression plus fusion group lost more blood, experienced prolonged operation time, and suffered more major complications.

Recently, two published RCTs concentrated on whether decompression plus fusion surgery yielded better clinical results than decompression alone for LSS. In the study by Zoher Ghogawala et al., a slight improvement in overall physical health-related quality-of-life was observed with spinal fusion and laminectomy for LSS, with or without grade I DS [17]. Forsth et al. then concluded that patients with LSS who underwent fusion plus decompression surgery received no better clinical outcomes than patients who received single decompression surgery at long-term follow-ups [18]. Carreon et al. [8], Glassman et al. [9] and Cassinelli et al. [10] demonstrated that posterior spinal fusion following decompression led to longer operative time, more blood loss, and a higher complication rate in extremely elderly patients. To our knowledge, no meta-analysis has compared the effectiveness of decompression versus decompression plus fusion in patients with LSS.

The debate on efficacy of additional fusion for LSS is ongoing. Decompression without arthrodesis was recommended for typical LSS with no history of previous lumbar spine operation, no spinal instability, and DLS ≤20 [19]. Decompression alone has been demonstrated to be significantly less invasive than decompression combined with spinal fusion [20]. Although decompression seems to be the logical procedure that has the potential to give the patient immediate relief, instability of the spine is a potential consequence that needs to be considered [21, 22]. Yone et al. reported that decompression alone for LSS obtained good results only if patients did not present with concomitant spinal instability; otherwise, isolated decompression surgery cannot guarantee satisfactory clinical outcomes [23]. Herkowitz et al. reported that one-third of patients receiving isolated decompression were not satisfied with the outcomes, especially in cases presenting with concomitant lumbar instability, which might lead to a higher reoperation rate [24].

Spinal fusion was initially used by Harms and Rolinger [25]. Suk et al. reported that spinal fusion after a complete decompression can alleviate future back and leg pain [26]. Kleinstueck et al. and Martin et al. both found better results when fusion was added to decompression in patients with LSS and DS [27, 28]. However, as an invasive procedure, fusion has many uncertainties that can greatly influence the final outcomes of LSS. The altered biomechanical function of the spine, such as loss of motion at the fused levels, was compensated for by increased motion at the unfused segments. This process caused certain mechanical stresses, which then accelerated adjacent lumbar level fusion problems and produced back pain and leg pain [26, 29]. Ekman et al. also confirmed this pathological change in a long-term RCT [30]. However, the dilemma of fusion was also inherent in the surgical indication for LSS. Patients with long-standing preoperative symptoms and concomitant diseases often had poor results and less satisfaction in clinical outcomes [15]. Mixed lumbar degenerative diseases and other degenerative changes, such as osteophyte formation, decreased disc height, and calcified ligaments, strengthened the lumbar stability and thereby reduced the demand of fusion. The most important disadvantage of fusion was the co-existence of other complications, higher reoperation rates, and heavier financial costs. Hallett et al. revealed a cost difference of approximately USD $6290 per patient for an additional fusion implant [31]. Dailey et al. [32] suggested that neither technique (lumbar fusion or instrumentation and nonfusion) was associated with an increased reoperation rate at the surgical level or adjacent lumbar levels. He noted that the only specific risk factor for reoperation between lumbar fusion and nonfusion was a duration of pretreatment symptoms of more than 12 months (i.e., the natural history of spinal degenerative disease). He also reported a 13% reoperation rate, similar to the 8.0% in our meta-analysis. Brodke et al. summarized that the common reason for reoperation in patients treated with laminectomy and fusion was due to symptomatic adjacent segment pathology; he also found that additional fusion had no superior survival curve, improved clinical outcomes, or improved patient satisfaction rates over laminectomy alone [33].

Therefore, surgeons should exercise great caution while performing spinal fusion in patients with LSS. A stratification of carefully screened patients on the basis of age, gender, comorbidities, with or without preoperative spondylolisthesis, intraoperative evaluation of slippage possibility, and other considerations should be completed, and the ultimate goal of treating LSS need always focus on the balance between decompression of the compressed nerve and adequate bone retention for spinal mechanical stability [16]. McCullen proposed that women and patients with preoperative spondylolisthesis may require changes in decompression without fusion modality to improve outcomes or alterations in long-term expectations for LSS [34]. Brown et al. affirmed intraoperative spinal stiffness measurements did not predict clinical results after lumbar spine surgery [35].

The proper indications for fusion remained unclear, but after searching the literature, some authors’ personal experiences may offer us some guidelines. Matsudaira reported better clinical results in patients with grade IDS by preserving the posterior elements of the spinal canal roof [36]. Radcliff recommended decompression and spinoplasty to preserve posterior ligament complex integrity for multilevel lumbar canal stenosis [37]. Yone suggested Posner’s method to define instability for fusion treatment, and fusion with instrumentation should be performed on elderly patients with instability after decompression [23]. Lawhorne preferred artificial ligamentous bands to decrease the flexion instability [29]. Tuli et al. concluded that the best alternative was an adequate decompressive laminectomy with a nonfusion technique of preserving the posterior ligament complex integrity [38].

The limitations of our study were: (1) the relatively low quality of the 15 included trials, which used multimodal decompression types as well as fusion treatments and made the results less effective; (2) insufficient data in EQ-5D, walking ability, and length of stay; (3) the various complications and nonconformity of assessment criteria in clinical satisfaction, which shared some inner inconsistencies that may have contributed to risk bias; (4) useful objective indicators such as cost-utility, post-operative walking distance, and SF-36, for example, were lacking; (5) our study follow-up period was less than 2 years. A longer-term analysis, including more comparative trials with moderate and high grade evidence, would be expected to improve the validity and reliability of our outcome.

Conclusions

Decompression plus fusion yielded no better clinical results than decompression alone in treating LSS, while resulting in a longer duration of operation, more blood loss, and a higher risk of complications. We believe decompression alone to be a sound choice for LSS, and we expect more controlled trials, prospective studies, and multi-center studies to further testify the long-term outcomes of additional fusion. More research is required to delineate the precise surgical protocol for LSS.