Abstract and Introduction

Introduction

While the efficacy of surgical treatment has been well established, the comparative effectiveness of adding instrumented lumbar spinal fusion to decompressive laminectomy in lumbar spondylolisthesis and spinal stenosis is controversial. Nonrandomized, prospective comparative studies have found evidence of laminectomy with fusion being superior but there is no Class I evidence. In partial contrast, other prospective studies with at least 5 years of postsurgery follow-up demonstrate excellent outcomes with only lumbar decompression. The lack of high quality evidence hinders development of clinical practice guidelines. The SLIP trial (spinal laminectomy vs. instrumented pedicle screw) reported by Ghogawala et al^[1]reports the results at 2 years of surgical patients randomized to either decompression or decompression and fusion for lumbar degenerative spondylolisthesis.

Study Summary

The trial began with 130 patients screened of which 106 enrolled and 66 were randomized (35 decompression alone, 31 decompression and fusion). At the 2-year follow-up the group sizes were 29 and 28. At baseline, the fusion group had an short form-36 physical component summary (SF-36 PCS) score that was 3.2 points lower than in the decompression-alone group. However, there were no statistically significant differences, suggesting a successful randomization process.

At the primary endpoint of 2 years, the mean change in SF-36 PCS for the decompression-alone group was 9.5 and for the fusion group was 15.2; the difference of 5.7 was statistically significant (P = 0.046) and larger than the minimal clinically significant difference. While the difference was not statistically significant at 1 year, it was significant at 3 months, 6 months, and then at 2 years through to the end of the 4-year follow-up period. Differences in Oswestry Disability Index (ODI) did not reach statistical significance until 4 years. The fusion group had significantly higher surgical complications (blood loss, length of stay, length of procedure) but had a significantly lower rate of reoperation (14% vs. 34%).

The authors conclude that laminectomy plus fusion was associated with statistically greater and clinically meaningful improvement in physical health and function compared with laminectomy alone at 2-, 3-, and 4-year follow-up. The authors' conclusions are strongly supported by the well-designed study and analysis. Differences in low back pain disability, as measured by the ODI, were not found until 4 years after surgery. Despite its widespread use across many spinal conditions and being the primary outcome measure in many studies, the ODI is a historic measure of low back pain related disability and may not be the best measure for patients with neurogenic claudication. Surgery for lumbar spinal stenosis with spondylolisthesis is indicated to improve symptoms of neurogenic claudication, with axial lower back pain being a secondary objective if present. Therefore, the lack of difference in ODI scores noted in this study does not infer a clinically meaningful implication for patients with lumbar spinal stenosis and spondylolisthesis. On the contrary, the improvements in SF-36 PCS scores probably more accurately reflect functional improvements associated with surgery for this condition.

Methodological Review

This is a randomized, controlled trial with patients from five centers, although most (51 of 66) patients were from one site. The study plan was to enroll 100 patients, and randomly assign 64 of them. A parallel registry of the other approximately 40 was planned as an observation cohort for patients who declined randomization. Details of the final protocol are available at NEJM.org.

A detailed list of inclusion and exclusion criteria is provided. The patient flowchart is comprehensive and well-presented. A panel of 10 expert spine surgeons assessed suitability of each patient for randomization. Independent interpretation of radiologic imaging was also performed to confirm the diagnosis of degenerative lumbar canal stenosis with spondylolisthesis without disk herniation. The authors note that "This novel approach appeared to increase patient consent to undergo randomization." Patients underwent one of two surgical procedures: laminectomy or laminectomy and instrumented fusion. The details of the interventions are clearly explained and the authors insured that all surgeons had experience with at least 100 laminectomy and 100 laminectomy and instrumented fusion procedures.

The a priori primary outcome measure is the change in SF-36 physical-component (PC) at 2 years with the ODI defined as a secondary outcome measure. The choice of a generic (SF-36) rather than a disease-specific (Oswestry) outcome instrument as the primary outcome would be unconventional. Minimal clinically important differences were pre-set at five points for the SF-36 and 10 points for the ODI. Sample size calculations are based on a between-group difference of 7.5 (standard deviation 10) for the SF-36 PC and yielded a needed sample size of 64 patients. We replicated their calculations confirming appropriate power analysis was performed.

Statistical analysis includes standard comparisons of two groups, using independent-sample t tests for continuous variables and chi-square tests or Fisher's exact tests for categorical variables. To compare groups with respect to the changes from baseline in SF-36 and ODI scores, mixed-effects models for repeated measures were used. Fixed effects were treatment type (the key variable under study), site, time since randomization, and time-by-treatment interaction. Finally, changes in SF-36 PC were dichotomized as over or under five points (the minimal clinically important difference) at 2 years, and a logistic regression carried out using the same fixed effects as in the mixed-effects models.

Analysis was on an intent-to-treat basis (i.e., according to a patient's original randomized treatment assignment) for all patients with follow-up assessments. Details of the statistical analysis plan are also available in the study protocol (NEJM.org); the study design, presentation, and statistical analysis are strong and appropriate for the design and nature of the investigational question.

Recommendation Regarding Impact on Clinical Practice

This paper demonstrates, with high quality and a high level of evidence, that lumbar laminectomy with fusion leads to a greater (statistically significant and clinically meaningful) improvement in physical health-related quality of life than laminectomy alone, over a follow-up period of 4 years. While the short-term complications were lower in the laminectomy alone group, reoperation rates were significantly higher. Therefore, we offer a strong recommendation that laminectomy plus fusion is the better treatment option for patients with lumbar spinal stenosis and spondylolisthesis. While there is a role for laminectomy alone, the subset patient population in whom this procedure will prove durable remains to be identified.

Surgery to treat lumbar spinal stenosis has become more prevalent over recent decades such that it is the most common indication for spine surgery at present. Surgical options may include decompression or decompression with fusion. When stenosis is present with a spondylolisthesis many have advocated fusion, in addition to decompression, in an effort to prevent worsening of the spondylolisthesis and recurrent or persistent symptoms. This current study investigates whether or not the addition of fusion to surgical decompression, in those with spinal stenosis, results in better clinical outcomes at 2 years. Neurogenic claudication caused by degenerative lumbar spinal stenosis is common. High-quality evidence to guide treatment decisions is lacking; specifically, there is controversy as to the benefit of adding fusion to surgical decompression. In an attempt to address these ongoing issues, the authors performed the present study which was intended to assess the utility of performing an adjunctive fusion for patients with symptomatic spinal stenosis.^[2]

Study Summary

Forsth et al^[2] have conducted a prospective randomized trial investigating the effectiveness of fusion as an adjunct to decompression in treatment of lumbar spinal stenosis. Clinical inclusion criteria included patients between the ages of 50 and 80 with more than 6 months of neurogenic claudication in one or both legs and back pain (visual analogue scale [VAS] >30) while radiographic criteria limited the population to patients with 1 to 2 level disease with thecal cross-sectional area of ≤75 mm². Patients were additionally stratified according to presence or absence of degenerative spondylolisthesis defined as ≥3 mm of anterolisthesis on plain lateral X-ray (not flexion-extension dynamic views). A total of 247 patients were enrolled between 2006 and 2012 (123 assigned to decompression and fusion and 124 to decompression alone). Of these, 14 patients did not receive the prescribed intervention and 5 were lost to follow-up, yielding 228 patients who were included in the per-protocol analysis. Baseline characteristics and preoperative variables are reported and were similar across all groups.

Specific surgical procedures performed were left to the discretion of the operating surgeons. As might be expected, there was a statistically significant difference between the treatment groups for intraoperative variables with longer operating times and greater estimated blood loss in the fusion groups. There was no significant difference in either intraoperative or postoperative complications. While there was no difference in reoperation rates or postoperative resource utilization between the two groups, patients in the fusion group stayed in the hospital longer (7.4 days for fusion vs. 4.1 days for decompression, P < 0.001). The mean operative costs were higher in the fusion group ($12,000 USD vs. $5,400 USD).

At 2 years, there was no significant difference in the primary outcome (i.e., ODI) between the two groups. A subgroup analysis showed no significant difference in outcome between patients with spondylolisthesis undergoing decompression alone and those undergoing decompression with fusion. Secondary measures included EuroQol-Five Dimensions, VAS-back, VAS-leg, Zurich Claudication Questionnaire score, 6-minute walk test, and patient-reported satisfaction. There were no differences in any of these secondary outcomes measures between patients undergoing decompression and those undergoing decompression and fusion. Moreover, this same finding applied for patients with spondylolisthesis. Of the 228 included patients, 144 were available for the 5-year follow-up and 138 provided outcome information. Once again, there were no differences in any of the primary or secondary outcomes measures between the fusion group and the decompression group at 5 years.

On the basis of the above data, the authors conclude that the addition of fusion as an adjunct to decompression for patients with lumbar spinal stenosis does not confer any significant clinical benefit but does increase hospital length of stay, blood loss, and direct costs. Most interestingly, the authors conclude that the presence of spondylolisthesis does not change this finding: decompression and fusion for spondylolisthesis with stenosis is no better than decompression alone for patients presenting with symptoms of neurogenic claudication and back pain. Moreover, in their population, fusion did not offer protection from the risk of reoperation even in the setting of spondylolisthesis.

Methodological Review

The authors clearly state that the aim of this study was to investigate whether the addition of fusion to decompression surgery resulted in better clinical outcomes at 2 years. There was no clear statement of study hypothesis and the authors did not commit to a stated hypothesis a priori. By examining the inclusion/exclusion criteria as well as the heterogeneity of surgical procedures permissible this present study does not seek to define the efficacy of fusion as an adjunct but rather the effectiveness of fusion as an adjunct to decompression.

The authors have used a prospective randomized trial design—patients with lumbar spinal stenosis were randomly assigned to undergo decompression plus fusion or decompression surgery alone. Patients were stratified according to the presence or absence of spondylolisthesis. Using this prognostic stratification ensured an equal distribution of spondylolisthesis patients in both treatment groups. Patients were not blinded to treatment received.

A clear description of the population from which the study sample was drawn was lacking, however the authors did mention that the study sample was derived from seven Swedish hospitals. A detailed list of appropriate inclusion and exclusion criteria were provided; stenosis was defined as a cross-sectional area of the dural sac of ≤75 mm². A detailed patient flow diagram was provided and in so doing the authors accounted for all eligible patients. Baseline comparability of the treatment and control groups were documented in terms of age, sex, smoking status, ASA score, various quality-of-life measures, and a functional test.

It was clearly stated that the surgical interventions were at the discretion of the treating surgeon. The type of decompression or fusion was not stipulated as part of the trial design. Postoperative cointervention with other treatments was not controlled for, again in keeping theme with this trials effectiveness design. All patients who entered the study were accounted for. Patients who did not receive the intervention to which they were allocated were excluded from the analysis. This is of concern in that the excluded patients may have been systematically different from those who were included; however, the number excluded was small and would not likely have changed the results. All the outcome measures utilized were appropriate and relevant to the questions asked.

The statistical tests were simple and applied appropriately. Sample size was considered prior to the study and the study sample was large enough to detect important differences should they have existed.

This study demonstrated no significant difference between groups in the mean ODI score at 2 years. This was the primary outcome. The results were the same in those patients with or without spondylolisthesis. Patients who underwent fusion were hospitalized longer and with greater associated operative costs. The recorded adverse events and reoperation rate were not substantially different between groups.

It is likely that the patients entered into this study are sufficiently representative that the results can be generalized to other settings. Similarly, the heterogeneity of surgical intervention, in terms of the type of decompression and fusion, also allows for generalizability. Generic, disease specific and functional outcome measures all suggested that the addition of fusion did not result in a superior result.

Recommendation Regarding Impact on Clinical Practice

Forsth et al have conducted a very well-designed and executed clinical trial addressing an exceedingly common pathology encountered by spine surgeons. The results of this study should cause the reader to seriously question the addition of fusion to surgical decompression in the symptomatic patient with central/lateral recess lumbar stenosis secondary to degenerative changes, particularly without clear evidence of segmental instability. Nevertheless, given the methodological limitations discussed previously, we believe that these findings support only a weak recommendation to incorporate these findings into clinical practice and no changes are warranted for patients with degenerative spondylolisthesis.

Cervical spondylotic myelopathy (CSM) is a degenerative spinal condition and the most common cause of spinal cord dysfunction worldwide. Although surgery may be beneficial, it is still a challenge to accurately predict improvement after surgical treatment. A clinical prediction rule would: (1) help appropriately manage patient expectations, and, therefore, satisfaction; (2) give decision-making support to surgeons; (3) provide a quantitative tool to discuss prognostic information during the consent conversation; and (4) align surgeons' perceptions of outcomes across hospitals, regions, and countries.

This study is the third of a three-part study, and reports on a new analysis of previously collected data, combining data from a CSM-North America study of 278 patients, and a CSM-International study of 479 patients (from 16 global sites). The first study (North America) led to a clinical prediction model to predict postoperative functional outcomes with prospective data.^[3] The model distinguished between patients with an mJOA score of 16+ versus <16 1-year postoperatively and reported predictive factors of age, duration of symptoms, severity, smoking, depression, bipolar disorder, and impaired gait. The model from the second study led to similar predictors as in the first model, but with some differences, most importantly, psychiatric disorders (relevant in North America but not Internationally).^[4] Other limitations noted included observing that the model did better in predicting outcome in moderate (mJOA = 12–14) to mild (mJOA = 15–18) myelopathy. The objective of the current (third) part was to address these limitations and refine the original model to increase global validity by using the combined data set.^[5]

Methodological Review

The two studies were both prospective multicenter cohort studies done to compare preoperative and postoperative neurological status and quality of life in CSM patients. The secondary objective was to develop a clinical prediction rule with the most significant predictors of surgical outcome. Inclusion criteria were age 18+, symptomatic CSM, evidence of spinal cord compression on imaging, and no previous spine surgery. Exclusion criteria were asymptomatic, active infection, neoplastic disease, rheumatoid arthritis, ankylosing spondylitis, or concomitant lumbar stenosis.

All patients underwent surgical decompression of the spinal cord and details of the surgical intervention are provided. Data were collected at baseline and 12 months postoperatively on demographics, symptomatology, imaging and clinical assessment, medical history, previous conservative treatment, functional status and health related quality of life (mJOA, neck disability index [NDI], SF-36, 30-meter walking test).

The primary outcome measure was the 18-point mJOA. The authors note that the reliability of the mJOA has not yet been established, but the original JOA has high inter- and intra-observer reliability. To develop prediction models, a cut-off of 16 was used for one model and 12 for a second model. Statistical methods used were univariate log-binomial regression (to model the relationship between clinical factors and primary outcome measure, and for relative risk estimation). Multi-collinearity was assessed to guard against over-specification of the model. Modified Poisson regression with robust error variance was used to create the final multivariate model and calculate relative risk for each predictor. Logistic regression was carried out on the final model to obtain an receiver operating characteristic (ROC) curve and to compute the area under the curve (AUC) as a one-number summary of the predictive performance of the model.

Odds ratios and relative risks are similar for rare events, but when events are not rare, odds ratios are not good estimates of relative risk. Hence, log-binomial regression, which estimates relative risk directly, was used here rather than the more familiar logistic regression. Both methods are appropriate for binary (dichotomous) outcomes. Poisson regression is an appropriate approach for analyzing rare events when subjects are followed for a variable length of time, but when used with binary data, the error for the estimated relative risk will be overestimated. Modified Poisson regression with a robust error variance procedure (known as sandwich estimation) fixes this problem. Logistic regression is the appropriate method to get the ROC and AUC. Hence, all the modeling techniques appear to be appropriately used.

The primary weakness of the study is the lack of a final, unifying predictive model. The authors call the paper a clinical prediction rule, but the findings are presented as the reduction in relative risk based on which category of a predictor variable a patient is in. There is, however, no overall model presented that can be used to compute a probability of a positive outcome. A logistic regression model apparently was developed to get the ROC and AUC, but the coefficients are not presented, only the individual relative risk (RRs). Unfortunately, the authors do not discuss how a clinician may combine the various RRs for the predictors in the final model.