Fusion in situ versus reduction for spondylolisthesis treatment: grading the evidence through a meta-analysis

Abstract Purpose: During surgical procedure on lumbar spondylolisthesis, the role of reducing slip remains controversial. The purpose of the present study was to compare fusion in situ with reduction in clinical and radiographic outcomes. Methods: A literature research was performed at PubMed, Embase, Web of Science, and Cochrane Library. After screening by two authors, ten articles were brought into this meta-analysis finally, and the quality was evaluated by the modified Newcastle–Ottawa Scale (NOS). Isthmic, moderate, and serious spondylolisthesis were all analyzed separately. Sensitivity analyses were performed for high-quality studies, and the publication bias was evaluated by the funnel plot. Results: Most criteria did not have statistical differences between reduction and fusion in situ groups. However, in reduction group, the union rate was significantly higher (P=0.008), the slippage was much improved (P<0.001) and the hospital stay was much shorter comparing to no-reduction group (P<0.001). Subgroup analysis (containing moderate and serious slip, or isthmic spondylolisthesis) and sensitivity analysis were all consistent with original ones, and the funnel plot indicated no obvious publication bias in this meta-analysis. Conclusions: Both reduction and fusion in situ for lumbar spondylolisthesis were related with good clinical results. Reduction led to higher rate of fusion, better radiographic slippage, and shorter hospital stay. After sufficient decompression, reduction did not incur additional risk of neurologic impairment compared with fusion in situ.


Introduction
Spondylolisthesis is defined as the forward slippage of one vertebra on another. Stability reconstruction and neural decompression have been widely considered as the effective treatment for spondylolisthesis [1]. To achieve these aims, various fusion methods with different surgical approaches have been used, such as anterior lumbar interbody fusion (ALIF), posterior lumbar interbody fusion (PLIF), and transforaminal lumbar interbody fusion (TLIF) [2]. In recent decade, minimally invasive TLIF (MIS-TLIF) has been developed, this operation could result in less blood loss, less soft-tissue injury, and earlier rehabilitation [3].
The main surgical strategies for treating spondylolisthesis include spinal fusion in situ and spinal fusion with reduction. Fusion in situ means that the vertebras will be fused 'where it is' with little or no correction of the spine; while fusion with reduction suggests that the slippage will be reduced or realigned during the surgery (Figure 1). Despite the evolution of surgical treatment procedures, it still remains controversial whether the management of reducing the spondylolisthesis is necessary [1,4,5]. During surgical procedure, fusion in situ is commonly performed with stability reconstruction and effective neural decompression. Nevertheless, compared with reduction in the slipped lumbar vertebra, it leads to higher pseudarthrosis rates and progressive parameters of the deformity, especially in patients with high-grade spondylolisthesis [6].
Reduction in the spinal anatomy and disk space height to restore the sagittal spinal balance is extended with the development of surgical techniques and instrumentation [1]. However, the reduction procedure is associated with increased risks of neurological complications, loss of reduction, and prolonged treatment time [1].
Previous systematic review has compared arthrodesis and neurologic deficits in situ fusion group with reduction group for high-grade spondylolisthesis [6]. However, more and more studies are published in recent years, including surgical management with MIS-TLIF [2,3,7]. More clinical and radiographic outcomes could be reviewed. Therefore, in the present study, we conducted a meta-analysis to compare different results between in situ fusion group and reduction group for low-and high-grade spondylolisthesis.

Search strategy
A literature search using PubMed, Embase, Web of Science, and Cochrane Library was performed in May 2016 without restriction on time, nation, and publication types. The search strategy was: (fusion or Arthrodesis) AND (in situ) AND (Reduction) AND (Spondylolisthesis). In addition, the references of every selected articles were checked manually to find out if they were also related.

Inclusion and exclusion criteria
Studies satisfying the following criteria were included for review: published studies comparing the outcomes between in situ arthrodesis and reduction in spondylolisthesis; randomized controlled trial (RCT) or retrospective comparative studies (cohort or case-control studies) on humans; studies that were required to provide available data to calculate the odds ratio (OR) and the corresponding 95% confidence interval (95% CI).
Correspondingly, studies were excluded if they met the following criteria: reviews, case reports, conference abstracts, and editorials; data that overlapped with previous publications. If potentially eligible studies reported overlapped data, the most comprehensive one was included in our systematic review.

Methodological quality assessment
The methodological quality of eligible studies was evaluated following a modified 9-star system of the Newcastle-Ottawa Scale (NOS) [9]. The 'star system' was applied to judge each study on three broad perspectives: the selection of the study cases (four items), the comparability of the study populations (two items), and the ascertainment of either the exposure or outcome of interest (three items). Studies scoring≥7 stars were deemed as high quality. Disagreements between investigators were settled by discussion until consensus was reached.

Data extraction
In compliance with the predefined criteria, following information was meticulously extracted independently by two reviewers from all qualified articles: surname of the primary author; year of publication; lever of evidence; patients' number; age; Meyerding grade; comparison category and operative method. The comparable parameters were divided into primary and secondary outcomes. Primary outcomes contained clinical and radiological results; VAS score; ODI score; JOA score; patient satisfaction; union rate and complication. Patient satisfaction surveys recorded five-point Patient Subjective Outcome scores (excellent, good, fair, unchanged, worse) [8]. Radiological results included slippage, lumbar lordosis, and lumbosacral angle. Secondary outcomes contained hospital stay, blood loss, and operative time.

Statistical analysis
The data were pooled by the Cochrane Collaboration's Review Manager 5.3. Continuous results were presented as mean difference (MD) and with 95% CI while dichotomous outcomes presented as OR with 95% CI [10,11]. Heterogeneity among studies was evaluated by the Cochran's Q test and the I 2 statistic [12]. P<0.1 or I 2 > 50% was considered to be heterogeneous. The random-effects model was used if there is heterogeneity (P<0.1 or I 2 > 50%) between studies, otherwise, the fixed-effects model was used.
Sensitivity analyses were performed for high-quality studies and funnel plots were used to screen for potential publication bias. Moreover, the moderate spondylolisthesis (Meyerding grades I, II) and severe slip (Meyerding grades III, IV) were also analyzed in subgroup, and two surgical methods were also compared in the isthmic spondylolisthesis as subgroup analysis.

Search results
A literature research found 1540 potentially relevant articles from four databases ( Figure 2). A total of 1161 studies were excluded for duplication or not related to spondylolisthesis. After screening the titles and abstracts, 353 articles were excluded for irrelevant article types or no comparison. Other 16 articles were excluded for inappropriate interventions. Finally, ten articles, including two RCTs [4,5] and eight comparative studies [1][2][3]7,[13][14][15][16], were brought in the meta-analysis.
The basic characteristics of these selected articles are shown in Table 1, and the quality of included studies, which were evaluated by modified NOS, are detailed in Table 2. The quality of RCTs, which was evaluated by the Cochrane Collaboration's Handbook is shown in Figure 3 [17].

Primary outcomes
Primary outcomes were divided into clinical and radiological ones (Table 3). VAS scores, ODI scores, and JOA scores had no significant differences between reduction and fusion in situ groups (P=0.46, P=0.56, P=0.22). Expect for the JOA scores (I 2 = 83, P<0.1), there were hardly any heterogeneity in other two items (I 2 = 0, P=0.72, I 2 = 0, P=0.78). No statistical differences in patient satisfaction (P=0.73) and complication rates (0.97) were found. And the neuropathic symptom was also similar in both groups. Whereas the union rates were obviously higher in reduction groups

Secondary outcomes
Four articles including 417 patients showed that the hospital stay time was shorter in reduction group compared with fusion in situ group, and the heterogeneity was also low (P<0.001, I 2 = 1.18, P=0.76) in hospital stay. There were

Isthmic spondylolisthesis
For isthmic spondylolisthesis patients, all the primary outcomes we analyzed were consistent with the original ones. There were no significant differences in VAS score, satisfaction, and complication, and the slippage was much better in reduction group (Figures 4-7).

Subgroup analysis
The spondylolisthesis patients were divided into different subgroups according to Meyerding grade: moderate (Meyerding grades I, II) and serious slip (Meyerding grades III, IV). The analysis was performed in order to compare the primary outcomes between reduction and fusion in situ groups in each subgroups. All the items in subgroups, including VAS scores (Figure 8), satisfaction (Figure 9), complication ( Figure 10), neuropathic symptom (Figure 11), and slippage ( Figure 12), were similar to the original outcomes. In radiological outcomes, the slippage was much better in reduction group. And there were no statistical differences in all other three indexes.

Sensitivity analysis and publication bias
Two RCTs and five high-quality retrospective studies were brought into sensitivity analysis (Table 4). Different from the previous results, the lumbar lordosis was greater in reduction group (P<0.001) and the length of hospital stay was similar in two groups (P=0.2). All other items were nearly the same as the original ones. Publication bias was evaluated by the funnel plot (Figure 13), which described the complication rates between two  groups. A total of nine studies almost lie inside the 95% CI and the distribution was symmetrical, which indicated no obvious publication in this meta-analysis.

Discussion
Reduction in surgical management of patients with lumbar spondylolisthesis remains controversial. Several trials studied the correlation of clinical outcomes and reduction during surgical procedure [18]. However, few studies with sufficient sample size evaluated it. Therefore, this study assessed clinical and radiological results by meta-analysis. In this article, the quality of eight retrospective articles were evaluated by modified NOS, and other two RCTs were defined as high-quality studies directly. The modified NOS was different from original one in comparability and outcomes. It is possible that the preoperatively basic characteristics, including clinical and radiological parameters, Figure 13. Funnel plot which was illustrated by non-union rate shows the publication bias of patients affected the outcomes. Therefore, the comparability 3 stars, which contained Age, VAS score, ODI score, JOA score, Radiological parameters, and Meyerding grade was scored before surgery. The outcome and exposure were scored two points here. Considering that the outcomes not only included short-term items, such as blood loss, hospital stay, operative time, et al. but also contained much more long-term indexes, the length of follow-up was not important for our current study. So the assessment criteria that if it was enough follow-up for outcomes to occur in outcome and exposure was eliminated in our study.
Primary outcomes showed that the union rate was significantly higher in reduction group comparing to fusion in situ group, which was consistent with the previous studies [18]. These results probably suggested that reductive vertebrae could provide more solid reconstruction and contact area, leading to a better fusion. However, other clinical criteria, including VAS score, ODI score, JOA score, and satisfaction, had no statistical differences. These findings were consistent with the widely spread opinion that neurological decompression and vertebrae fusion are the main aims of surgery [4]. Surgical technique with or without reduction all achieved these goals. Considering that the pooled data in the present study were mostly from the reports describing patients for short-term follow-up, different clinical outcomes might be obtained due to the different union rate between the reduction groups and the in situ groups after long-term follow-up.
Previous study indicated that more patients suffered from neurological symptoms in reduction group, which was also considered as an important disadvantage during reductive procedure [4]. Nevertheless, in our analysis, the occurrence rate of complication was similar in two groups. This might be because the spinal canal and nerve roots far distally and laterally were decompressed before slip reduction, permitting complete visualization of the roots at all times [4]. In addition, there was no statistical differences between these two groups in blood loss and operative time. This might be due to the development of modern surgical instrumentation and techniques.
Our study demonstrated that slippage was significant improved in reduction group, while lumbar lordosis and lumbosacral angle were similar in both groups no matter before or after surgery. These results provided less meaningful evidence to sagittal plane for stable balance. Some significant spinopelvic parameters, such as pelvic incidence (PI), sacral slope (SS), pelvic tilt (PT), etc, could not be synthesized for not enough data in these articles.
The treatments for different slippage grades were different, especially moderate and severe ones. We divided spondylolisthesis into moderate (Meyerding grades I, II) and severe slippage (Meyerding grades III, IV) to do the subgroup analysis. All these results were similar to original outcomes. From these subgroup analysis, we concluded that the treatment selection was similar in slight or high-grade spondylolisthesis.
There was nearly no heterogeneity changes in subgroup analysis comparing with original outcomes. This indicated that the Meyerding grade of spondylolisthesis may not responsible for the heterogeneity. However, considering that the heterogeneity of criteria we brought into subgroup analysis were generally low, more articles were needed to compare outcomes between two surgical methods in different grade spondylolisthesis.
The most common reasons that caused spondylolisthesis were spondylolysis and degeneration. The treatments for these two types of spondylolisthesis might be different. In order to explore the treatment differences, isthmic and degenerative spondylolisthesis were analyzed separately. Due to the quantitative limitation of articles which described degenerative spondylolisthesis, only articles which investigated isthmic spondylolisthesis were extracted. The outcomes were no difference for the type of spondylolisthesis possibly. Nevertheless, more trails of degenerative spondylolisthesis should be further studied.
The sensitivity analysis was performed in seven high-quality articles (two RCT and five high-quality retrospective studies) in order to improve the credibility of our study. All these articles matched the basic characteristics of patients in two treatment groups before surgery. The results were almost consistent with original ones. Meanwhile, heterogeneity was generally lower in each item. Low-quality articles may contribute to the occurrence of heterogeneity.
Certainly, there were several limitations in our study. First, although the quality of seven articles in our meta-analysis were high, only two RCTs were brought into our analysis. Second, the heterogeneity was a little high in some indexes. Despite the heterogeneity was not significant for dichotomous outcomes, it was significant for most of the continuous variables, including studies, researched in different countries, different operative methods, different slippage degree, and measurement of outcomes. These differences might contribute to the significant heterogeneity. Pooling of data using the random-effects model might reduce the effect of heterogeneity but could not abolish it.
In conclusion, both reduction and fusion in situ for lumbar spondylolisthesis were related with good clinical results. Reduction leaded to higher rate of fusion, better radiographic slippage, and shorter hospital stay. After sufficient decompression, reduction did not incur additional risk of neurologic impairment compared with fusion in situ.