Predictors of Recovery Following Lumbar Microdiscectomy for Sciatica: A Systematic Review and Meta-Analysis of Observational Studies

Chronic post-surgical pain is reported by up to 40% of patients after lumbar microdiscectomy for sciatica, a complaint associated with disability and loss of productivity. We conducted a systematic review of observational studies to explore factors associated with persistent leg pain and impairments after microdiscectomy for sciatica. We searched eligible studies in MEDLINE, Embase, and CINAHL that explored, in an adjusted model, predictors of persistent leg pain, physical impairment, or failure to return to work after microdiscectomy for sciatica. When possible, we pooled estimates of association using random-effects models using the Grading of Recommendations Assessment, Development, and Evaluation approach. Moderate-certainty evidence showed that the female sex probably has a small association with persistent post-surgical leg pain (odds ratio (OR) = 1.15, 95% confidence interval (CI) = 0.63 to 2.08; absolute risk increase (ARI) = 1.8%, 95% CI = -4.7% to 11.3%), large association with failure to return to work (OR = 2.79, 95% CI = 1.27 to 6.17; ARI = 10.6%, 95% CI = 1.8% to 25.2%), and older age is probably associated with greater postoperative disability (β = 1.47 points on the 100-point Oswestry Disability Index for every 10-year increase from age (>/=18 years), 95% CI = -4.14 to 7.28). Among factors that were not possible to pool, two factors showed promise for future study, namely, legal representation and preoperative opioid use, which showed large associations with worse outcomes after surgery. The moderate-certainty evidence showed female sex is probably associated with persistent leg pain and failure to return to work and that older age is probably associated with greater post-surgical impairment after a microdiscectomy. Future research should explore the association between legal representation and preoperative opioid use with persistent pain and impairment after microdiscectomy for sciatica.


Introduction And Background
The lifetime prevalence of sciatica in the general population ranges from 12% to 43% [1,2] and is associated with pain radiating down the leg, numbness, and motor deficits [1,3]. In the United States, the total direct and indirect costs (e.g., loss of wages and productivity) associated with sciatica exceed $50 billion annually [4][5][6]. Lumbar discectomy is an elective surgical procedure performed in approximately 10% of sciatica patients to relieve symptoms and promote functional recovery [7][8][9]; however, outcomes are variable, and up to 40% of patients report persistent post-surgical leg pain [10][11][12].
Previous systematic reviews have identified greater preoperative pain severity, comorbid mental illness, receipt of worker's compensation benefits, and higher fear avoidance as risk factors for poor outcomes following surgical decompression for sciatica [11,[13][14][15][16]. However, prior reviews have several limitations, such as outdated searches [17], language restrictions [13,15,17], and the inclusion of studies reporting predictors from unadjusted analyses [14,16,17]. We conducted a systematic review of observational studies to Studies enrolling patients with spinal stenosis or spondylolisthesis, or who underwent fusion in addition to microdiscectomy, or repeat spine surgery were not eligible for review. Eligible procedures included microdiscectomy, endoscopic microdiscectomy, microendoscopic discectomy, mini-open discectomy, and tubular microdiscectomy. We excluded randomized trials as strict eligibility criteria may exclude patients with important prognostic factors. We also excluded non-randomized studies with interventions, descriptive or qualitative studies, and letters to the editors. We excluded studies that reported only adjusted models with significant association with variables collected after baseline, as in such instances the direction of association is uncertain.

Study Selection and Data Abstraction
Trained reviewers worked in pairs to screen titles and abstracts of identified citations and full texts of all potentially eligible studies independently and in duplicate. All reviewers completed pilot exercises before screening to increase reliability. Disagreements were resolved through discussion or, when necessary, by an arbitrator (JWB).
The same pairs of reviewers independently extracted data from eligible articles, including sample size, duration of follow-up, patient characteristics, and measures of association for all factors assessed for an association with persistent leg pain, functional disability, or return to work (RTW) following lumbar discectomy. If a study reported multiple follow-up times, we captured data for the longest follow-up reported.

Risk of Bias
We used criteria from Users' Guides to the Medical Literature [22] to assess the risk of bias: (1) representativeness of the study population (low risk of bias when using random sampling, consecutive sampling, or data collected from a patient registry; high risk of bias when the source of the study population was not reported or acquired through convenience sampling); (2) validity of outcome assessment; (3) loss to follow-up (>20% was considered high risk of bias); and (4) whether predictive models were optimally adjusted (low risk of bias if adjusted, at minimum, for age, sex, and baseline pain severity).

Data Analysis
We assessed the reliability of full-text screening with the kappa statistic [23]. When possible, we pooled all factors assessed for an association with persistent pain, disability, or unemployment and reported by at least two studies. For categorical variables, we reported pooled estimates as odds ratios (ORs) and associated 95% confidence intervals (95% CIs), and for continuous variables, we reported pooled estimates as beta coefficients (β) and associated 95% CIs using DerSimonian-Laird random-effects models. To avoid overestimating the strength of association by restricting pooling to risk factors with reported associations, we imputed an OR of 1 and an associated measure of precision using the hot deck approach [24,25] for all categorical predictors that were reported as non-significant and without accompanying data. We complimented ORs with the absolute risk increase for each predictor amenable to meta-analysis. We acquired the following baseline risks from the low-risk group in the study with the largest sample size among studies eligible for our review at low risk of bias: (1) 14% for persistent post-surgical leg pain [26], (2) 20% for persistent disability [27], and (3) 7% for failure to RTW [28]. We used SPSS Statistics version 28.01.1.0 (IBM Corp., Armonk, NY, USA) for all statistical analyses; all comparisons were twp-tailed, and p-values ≤0.05 were considered statistically significant.
When pooling was not possible, we explored the consistency of the association between pooled results and studies reporting the same predictors that could not be pooled. We used the following three criteria to identify predictors that were not amenable to pooling and showed promise for future research: (1) a statistically significant association of p ≤ 0.01, (2) a large magnitude of association (OR ≥2.0 or <0.5), and (3) a sample size of ≥500.

Subgroup Analyses
We evaluated heterogeneity for all pooled estimates through visual inspection of forest plots. We generated three hypotheses to explore heterogeneity between studies, assuming larger associations with (1) a higher risk of bias on a criterion-by-criterion basis, (2) a longer duration of follow-up, and (3) a higher threshold for outcomes (e.g., moderate-to-severe persistent leg pain vs. any persistent leg pain). We only conducted subgroup analysis if there were at least two studies in each subgroup and assessed the credibility of significant subgroup effects using the modified ICEMAN criteria [29].

Sensitivity Analysis
We performed a sensitivity analysis by removing imputed data from our pooled analyses.

Certainty of Evidence
We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach to summarize the certainty of evidence for all meta-analyses [30]. With this approach, the evidence for prognostic factors begins as high certainty but can be downgraded to moderate, low, or very low based on the risk of bias, consistency, directness, precision, and publication bias [31]. Accordingly, while associations supported by high certainty evidence are presented without any qualifiers, moderate certainty evidence is preceded with the qualifier "probably" and low certainty evidence with the qualifier "may." If subgroup analysis for risk of bias did not find a significant association, we included all studies and did not rate down for risk of bias. If we found a credible subgroup effect for risk of bias, we pooled only low-risk studies [32].
On a review of baseline risks for our outcomes, clinical experts from our team (BD, NE) estimated that a 5% increase in absolute risk would be sufficient for clinicians to address modifiable risk factors, and an absolute difference in risk of 10% between groups at low and high risk for persistent pain, prolonged disability, or unemployment would be sufficient for clinicians to selectively target non-modifiable risk factors. Therefore, we rated down for imprecision if the 95% CI associated with the risk difference included 5% for modifiable risk factors or 10% for non-modifiable risk factors. For meta-analyses with at least 10 studies, we assessed publication bias by visual assessment of the asymmetry of the funnel plot and performed the Begg rank correlation test and the Egger test [33,34].

Results
Our literature search yielded 49,790 unique citations, of which 32 studies [27, with 26,876 participants were eligible for review ( Figure 1).   showed a small association between the female sex and persistent leg pain after lumbar microdiscectomy (OR = 1.15, 95% CI = 0.63 to 2.08) ( Figure 2). The absolute risk increase in persistent leg pain associated with the female sex was 1.8% (95% CI = -4.7% to 11.3%) ( Table 3).    [55], and Ziegler et al. [64] did not optimally adjust the final model for age, gender, and preoperative pain intensity. The test between subgroup homogeneity was significant (p = 0.03). Therefore, the quality of evidence was determined by the low risk of bias studies.
Quality was rated down based on imprecision because the 95% CI associated with the risk difference included the predefined threshold of 5% for modifiable factors or 10% for non-modifiable factors, which means that clinical actions based on the estimate of the lower or upper boundary may be different.

Predictors of Failure to Return to Work
A total of six studies [41,45,55,58,60] involving 2,021 patients reported the association of 39 factors with RTW after surgery, and one study [37] (n = 141 patients) explored the association with postoperative working capacity. Only sex met our criteria for meta-analysis, and low certainty evidence from five studies [41,45,55,58,60] (1,979 patients) suggested little to no association with failure to RTW after surgery (OR = 0.98, 95% CI = 0.28 to 3.41) ( Figure 5). However, we found evidence of a credible subgroup effect based on whether studies reported an optimally adjusted predictive model (Appendices: ICEMAN criteria). We found moderate certainty evidence from two studies (210 patients) that reported optimally adjusted predictive models that the female sex versus male sex was probably associated with greater odds of failure to RTW after surgery (OR = 2.79, 95% CI = 1.27 to 6.17; risk difference = 11%, 95% CI = 2% to 25%) ( Figure 6, Table 3).

Variables Not Amenable to Meta-Analysis
Tables 4-9 present the associations with persistent pain, persistent disability, and failure to RTW (38 variables) for approximately 50 variables factors that were not amenable to meta-analysis. Two of these factors, opioid use before surgery and legal representation at the time of surgery, met our criteria as promising for future investigations (       One study out of the two studies showed a significant association between passive coping with functional limitations/disability  [55] One out of four studies showed higher preoperative functional disability was associated with less probability of returning to work after the lumbar microdiscectomy. *: Combined score for pain and disability Work-related stress was a significant factor for failure to RTW in one out of two studies

Discussion
We found moderate certainty evidence that the female sex was probably associated with a small increased risk (2%) of persistent post-surgical leg pain and a large increased risk (11%) of failure to RTW after microdiscectomy for sciatica. Moderate certainty evidence also showed that older age was probably associated with a small increased risk for persistent disability after decompression surgery. Studies have tested approximately 50 predictors that could not be pooled, of which opioid use before surgery and legal representation at the time of surgery warrant additional investigation.
The key strengths of our review are that methodologically, our review was more rigorous as we accounted for non-significant variables and imputed 1 for excluded variables due to a non-significant association in the univariable analysis. Carrying only significant predictors to the multivariable analysis increases the risk of overestimation in the final analysis model. We presented our results with an absolute measure of association, such as risk difference. Compared to the relative measure of association, such as OR, and RR, the absolute measure of association, is essential to guide clinical decision-making. We performed subgroup analysis based on the risk of bias and further assessed the credibility of subgroup effects using the ICEMAN criteria.
Our review also suffered from a few limitations. Many predictors were only reported by a single study, due to which we could not perform a meta-analysis. Another limiting factor that precluded us from a meta-analysis of most variables was incomplete data reporting, such as many studies only reported p-values [38,46,57] or SE [36] or OR without 95% CI [44] or only beta-coefficient [49,53].
Compared to the previous systematic reviews, we identified more studies that previous reviews did not include [14,15,17,21]. Den Boer et al. [17] included 13 out of 15 (>85%) studies that reported composite scores. Our rationale for excluding studies with composite scores such as patient satisfaction and medication use was because composite scores can obscure the vital information specific to outcomes [67,68]. Composite scores are more useful when the outcome is rare, and combining multiple outcomes such as pain, disability, work capacity, doctor visits, analgesic use, sleep disturbances, patient's opinion, or clinical examination [69] can reduce the type I error, but combining variables reduces meaningful information and makes the interpretation difficult [67,69].
Furthermore, den Boer et al. [17] included studies that analyzed outcomes data with unadjusted analysis and reported the positive and negative association based on the number of studies reporting a variable.
Analyzing the association of baseline variables with the outcome in a multivariable-adjusted analysis accounts for the effect of potential known variables that can affect the outcome. Previous reviews included heterogeneous study designs such as RCT [21] and studies that analyze the association of various baseline variables with the outcomes in unadjusted analyses [14,15,17].
Interim of conducting our systematic review, potential new studies were published. Mehendiratta et al. [70] reported a significant association of younger age, males, and non-smokers, with symptom duration fewer than six weeks, and with disc herniation at L3 to L4 with a postoperative disability after the lumbar microdiscectomy. The study analyzed the predictive association of baseline variables in unadjusted analysis with postoperative disability. The adjusted multivariable analysis allows us to account for the effect modification and relationship between various baseline risk factors. Future studies should analyze the association of various baseline variables with postoperative pain, disability, and RTW in large sample-size studies and optimally adjusted models.

Conclusions
Our review found moderate certainty evidence that the female sex had a higher probability of persistent leg pain and failure to RTW after a microdiscectomy for sciatica and that older age is probably associated with greater post-surgical impairment. We also identified the limitations in the current published literature such as heterogeneous reporting of the results, small study samples, and not consistently adjusting final models for important variables such as age, sex, and preoperative sciatica pain severity, which have shown significant association with postoperative outcomes after lumbar microdiscectomy. We also identified two important variables such as legal representation and preoperative opioid use that were not amenable to pooling but met our criteria for potential variables that may have a significant association with postoperative outcomes after lumbar microdiscectomy. Future research should explore the association between legal representation, preoperative opioid use, and persistent pain and impairment after microdiscectomy for sciatica in a large sample and methodologically rigorous studies.
Outcome: Failure to return to work (RTW), Predictor: Female sex (reference: male)

Conflicts of interest:
In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: Jason Busse declare(s) a grant from Canadian Institutes of Health Research (CIHR) . JWB is funded, in part, by a Canadian Institutes of Health Research (CIHR) Canada Research Chair in the prevention and management of chronic pain. CIHR has no role in designing, conceptualization, data analysis, and manuscript draft. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work.