Evaluating Rehabilitation Outcomes After Lumbar Fusion Surgery: A Systematic Review and Meta-Analysis

Lumbar fusion surgery (LFS) is performed to stabilize adjacent vertebral motion segments rigidly. It is often combined with surgical decompression to relieve back pain and/or neurogenic leg symptoms. ¹?³ Common indications for LFS include spondylolisthesis, disc disease, and spinal stenosis.¹ In the UK, the incidence of LFS is rising. In 2009/2010, 4,036 procedures were performed, increasing by over 60% to 6,547 in 2012/2013. A similar trend is observed in the USA, particularly among patients over 60 years of age. As projections estimate that 30% of the UK population will be over 60 by 2037,? the demand for LFS is expected to continue increasing.² Despite undergoing LFS, many patients experience persistent symptoms. Data from the Swedish National Spine Register indicate that 25% of patients report unchanged or worsening pain, while 40% remain uncertain or dissatisfied with their outcomes 12 months post-surgery. This underscores the need to explore strategies for improving postoperative recovery.³ A recent Cochrane Back Review Group (CBRG) report suggests that rehabilitation, particularly supervised active exercise, improves pain relief and functional outcomes following laminectomy for lumbar stenosis. However, its effectiveness for LFS remains unclear, with no established consensus on postoperative rehabilitation protocols.^4,5 A previous systematic review and meta-analysis found inconclusive and very low-quality evidence regarding the benefits of physiotherapy-including exercise, manual therapy, electrotherapy, and cognitive behavioral therapy (CBT)-following LFS. The study emphasized the urgent need for further research in this area.

Objectives

This review aimed to evaluate the evidence on rehabilitation strategies for adults following lumbar fusion surgery (LFS) for degenerative conditions. Eligible trials included randomized controlled designs, an appropriate comparator (e.g., usual care), and validated outcome measures assessing pain and/or disability in both the short term (<6 months) and long term (>12 months).

MATERIALS AND METHODS

Protocol and Registration

A protocol was developed based on the methods described by the Cochrane Back Review Group (CBRG) and the Cochrane Handbook. The study followed PRISMA guidelines⁶ and was registered with the International Prospective Register of Systematic Reviews (PROSPERO).

Eligibility Criteria

Studies investigating rehabilitation following LFS that met the following criteria were included in the review:

• Design: Randomized Controlled Trial (RCT)
• Participants: Adults (≥16 years) who underwent LFS for degenerative conditions
• Intervention: Rehabilitation (physical, psychological, or combined)
• Comparator: Suitable comparator (e.g., usual care)
• Publication Date: 1974 onward
• Outcome Measures: At least one validated measure of pain and/or physical function
• Follow-up Duration: Short-term (<6 months) and long-term (>12 months)
• Language: Any (translations arranged as necessary)

Search Strategy

A three-phase search strategy was employed:

1. Scoping Search: Initial searches were conducted in MEDLINE, AMED, and CINAHL using keyword combinations such as "lumbar," "fusion," and "rehabilitation."
2. Comprehensive Search Development: Keywords identified from the scoping search were refined and trialed with assistance from a librarian (K.B.).
3. Hand Searching & Grey Literature: Additional searches were conducted in key journals and grey literature sources.

Information Sources

The following databases were searched:

• CINAHL, EMBASE, MEDLINE, PEDro, PsycINFO
• Cochrane Library, Health Technology Assessment Database, NHS Economic Evaluation Database
• National Research Register, Current Controlled Trials (York)
• Cochrane Back Review Group
• Grey literature and key journal hand searches

Study Selection

Two independent reviewers (J.G., J.M.) conducted database searches, saved and pooled results, removed duplicates, and integrated findings from grey literature and hand searches.

• Titles were initially screened by one reviewer (J.G.), with irrelevant studies excluded.
• Abstracts of the remaining studies (n = 34) were obtained and assessed by two reviewers (J.G., J.M.), who classified them as eligible, ineligible, or potentially eligible based on the inclusion criteria.
• Disagreements were resolved using Cohen’s kappa statistic and mediated by a third-party expert (A.M.).
• Full texts of eligible and potentially eligible studies were independently reviewed by two authors (J.G., J.M.) for final inclusion, with inter-reviewer agreement measured using Cohen’s kappa.

Disagreements and Consensus Resolution

Disagreements between authors were resolved through a third-party mediator (A.M.) to ensure consensus.

Data Extraction

A data extraction form, based on the "Characteristics of Included Studies" table from the Cochrane Handbook,⁷ was developed in parallel with the search strategy and adapted to suit the needs of this review. Two independent reviewers (J.G., J.M.) performed data extraction, while a third reviewer (M.H.) verified the form against selected studies to ensure accuracy. Authors of the included studies were contacted for raw data, which was successfully obtained from one study.⁸

Extracted Data Items

The extracted data included study design, participant characteristics (including surgical indications), intervention details, comparators, and both primary and secondary outcome measures. These outcomes were assessed at short-term and long-term follow-ups and covered aspects such as disability, pain, mental health, and fear-avoidance behavior. No assumptions or data simplifications were made.

Summary Measures and Data Synthesis

The review protocol allowed only for the inclusion of studies with comparable participants, interventions, comparators, and outcome measures. Short-term and long-term outcomes for disability, pain, mental health, and fear-avoidance behavior were identified as suitable for pooled analysis. A meta-analysis was conducted using RevMan⁹ software, applying the inverse variance model for continuous data (change in mean values from baseline). The DerSimonian and Laird¹⁰ random-effects model was used to account for variations across studies while assuming related intervention effects.

• Confidence intervals (CIs): Set at 95%.
• Statistical analysis: Mean change from baseline scores analyzed using the standardized mean difference.¹¹
• Standard deviation (SD): Available for one study, while for another, raw data provided by the corresponding author were used to calculate SD.¹²

Risk of Bias Assessment

Within Individual Studies

The Cochrane Risk of Bias Assessment Tool was used to evaluate internal validity and potential sources of systematic error.

Across Studies

A formal risk-of-bias assessment across studies was not conducted due to the limited number of studies. Funnel plots were deemed unnecessary. The GRADE¹³ criteria were used to assess the quality of evidence.

Additional Analyses

Due to the small number of studies, additional analyses were not feasible.

RESULTS

Study Selection Process

Database searches were conducted on October 13 and October 20, 2014, by J.G. and J.M., respectively. This process identified 1,006 studies, which were screened by title, leading to the removal of 972 irrelevant or duplicate papers. The remaining 34 abstracts were retrieved and reviewed by J.G. and J.M., resulting in 12 papers classified as eligible or potentially eligible for full-text review. The inter-reviewer reliability was strong (Cohen’s k = 0.78). After full-text review, five papers reporting data from three original studies met the inclusion criteria. The agreement between authors was very strong (Cohen’s k = 0.88). (See Figure 1 for study selection flowchart).

Figure 1.  A Flowchart Depicting the Study Selection Process Along with Reasons For Rejection

Three papers met the eligibility criteria for inclusion in this review.¹⁴ All three studies compared usual care with innovative forms of rehabilitation. Christensen et al.¹⁵ compared usual care with a "back café" group and a physical training group, Abbott et al.¹⁶ with "psychomotor therapy," and  Monticone et al.¹? with a combination of exercise and cognitive behavioral therapy (CBT). Rehabilitation sessions lasted 60 minutes, twice per week, for four weeks. Rehabilitation commenced after LFS; however, the exact timing was not well described. The primary outcome measured was post-rehabilitation change in the Oswestry Disability Index (ODI) score. Secondary outcomes included the Tampa Scale of Kinesiophobia (TSK), pain levels (Numerical Rating Scale - NRS), and quality of life (QoL), including mental health (SF-36). Outcomes were recorded pre-treatment, immediately post-rehabilitation, and 12 months after LFS.

Study Details

• Christensen et al. (2016);¹⁵

This study included 90 participants (mean age: 45 years) randomized into three intervention groups:

1. Usual care group – Patients received a video demonstration and a single physiotherapy session explaining exercises.
2. Physical training group – Patients attended twice-weekly physiotherapy sessions (90 minutes each) for eight weeks, focusing on supervised exercises.
3. Back café group – In addition to usual care, these patients attended three 90-minute group sessions facilitated by a physiotherapist. These sessions aimed to help patients share experiences about pain, disability, rehabilitation concerns, and coping strategies.

No primary outcome measure was explicitly identified, but the Low Back Pain Rating (LBPR) Scale²³ was used for evaluations at 3, 6, 12, and 24 months post-LFS.

• Abbott et al. (2010)¹⁶

This study randomized 107 participants (mean age: 51 years) into two groups:

1. Usual care group (n = 54) – Patients received a single 20-minute session of exercise advice from a hospital physiotherapist.
2. Psychomotor therapy group (n = 53) – Patients received usual care plus three 90-minute hospital outpatient appointments (at postoperative weeks 3, 6, and 9). These sessions involved:
   • Core stability exercises
   • Cognitive coping strategies
   • Relaxation techniques
   • Motivational goal setting
   • Assistance in managing setbacks

This rehabilitation approach, combining physical therapy (based on Richardson and Hides²?) and CBT (based on Linton²?), was termed "psychomotor therapy". Rehabilitation began within three weeks of discharge following LFS. The primary outcome measure was the Oswestry Disability Index (ODI), assessed at 3, 6, 12, and 24–36 months post-LFS. Secondary outcomes included:

• Pain levels (Visual Analog Scale - VAS)
• Quality of life (EQ-5D and SF-36 mental health subscale)
• Fear-avoidance behavior (Tampa Scale of Kinesiophobia - TSK)

• Monticone et al. (2014)¹⁷

This study randomized 130 participants (mean age: 57 years) into two groups:

1. Usual care group – Patients participated in 90-minute supervised exercise sessions, five times per week, for four weeks.
2. Experimental group – Patients received the same exercise program plus CBT.

Risk of Bias Assessment

• Christensen et al.¹⁵ had a mix of high and unclear risk-of-bias domains. Subsequent publications assessing long-term primary healthcare demands²? and economic analysis²? did not alter the overall risk assessment.
• Abbott et al¹⁶ and Monticone et al.¹⁷ had lower overall risk of bias, with one unclear and one high-risk domain, while all other domains were low risk (Fig. 2).
• Blinding was problematic due to the nature of the interventions, making it a high-risk domain across all three studies. However, this is unlikely to have significantly impacted the results.

Inter-reviewer agreement on risk-of-bias assessment was good (Cohen’s k = 0.72).

Synthesis of Results

• The Christensen et al.¹⁵ study reported median and range (nonparametric) data, making it ineligible for meta-analysis.
• The Abbott et al.¹⁶ and Monticone et al.¹⁷ studies each included:
  • A usual care (exercise) group
  • An experimental group (exercise + CBT)

For the purpose of this review, the combined approach of exercise + CBT was termed "complex rehabilitation." Data from both studies were pooled to compare complex rehabilitation versus usual care across comparable outcomes.

Consensus among review authors led to pooling data for:

• Disability (ODI)
• Back pain (VAS and NRS)
• Mental health (SF-36 mental health subscale)
• Fear-avoidance behavior (TSK)

Pooled Results from Abbott et al.¹⁶ and Monticone et al.¹⁷

• Participants: n = 237 (62% female, mean age = 55 years)
• Short-term follow-up:
  • Abbott et al.¹⁶: 3 months post-LFS
  • Monticone et al.¹⁷: Immediately after the 4-week rehabilitation program
• Long-term follow-up: 1 year after study entry

Key Findings

• In the short term, one study (Abbott et al.)¹⁶ showed significant improvements in disability, back pain, and fear-avoidance behavior.
• The other study (Monticone et al.)¹⁷ reported significant improvements in disability, pain (low back and leg), fear-avoidance behavior, and mental health.
• Both studies favored complex rehabilitation over usual care.

Meta-Analysis Results

Pooled data from Abbott et al.¹⁶ and Monticone et al.¹⁷ indicated significant short-term effects favoring complex rehabilitation:

• Disability (ODI):
  • Effect size: 0.85
  • 95% confidence interval (CI): 1.41 to 0.29
• Fear-avoidance behavior (TSK):
  • Effect size: -1.07
  • 95% CI: -1.33 to -0.80

These findings suggest that complex rehabilitation (exercise + CBT) significantly improves disability and fear-avoidance behavior compared to usual care.

Pooled Analysis and Heterogeneity

The pooled analysis for short-term low back pain (LBP) narrowly failed to reach statistical significance (0.71, 95% CI: -1.44 to 0.01). Heterogeneity (I²) was high in the pooled analysis for disability (77%) and LBP (87%), which may have contributed to the lack of observed effect. In contrast, heterogeneity for fear-avoidance behavior was low (0%). In the long term (12 months), one study reported significant improvements in disability and fear-avoidance behavior, while another study found significant improvements in disability, pain (both back and leg), fear-avoidance behavior, and mental health. In all cases, the results favored "complex rehabilitation." Pooled analysis demonstrated statistically significant effects for disability (effect size: 0.84, 95% CI: -1.11 to -0.58) and fear-avoidance behavior (-1.40, 95% CI: -1.69 to -1.12), both in favor of "complex rehabilitation." Heterogeneity was within acceptable limits for both meta-analyses. However, the long-term meta-analysis for LBP did not support any positive effect of "complex rehabilitation" over usual care.

Risk of Bias Across Studies

Two studies were included in the meta-analysis.^18,19 Both had one high-risk domain related to blinding participants, while one study also had an unclear risk of bias concerning blinding of outcome assessment. Overall, the majority of data comes from studies with a low or unclear risk of bias, making the overall risk of bias across studies uncertain (Fig. 2).