Recurrent lumbar disc herniation: a systematic review and meta-analysis of risk factors, surgical timing, and outcomes of revision discectomy versus fusion

Eligibility criteria

The inclusion and exclusion criteria for the study are listed in Table 1. The inclusion criteria were as follows: human studies, participants of any age, assessment of first-time RLDH, timeframe defining recurrence for RLDH clearly stated in the study’s inclusion criteria, surgical approach (for reoperation) reported, reoperation outcomes reported with a minimum follow-up of at least 6 months, study designs including primary research studies (experimental, observational, and descriptive), English language, and any publication date. The exclusion criteria were animal studies; assessment of second or subsequent recurrences (i.e., not limited to first-time RLDH); unclear timeframe defining recurrence for RLDH in the study’s inclusion criteria; unreported surgical approach for reoperation; reoperation outcomes not reported with a minimum follow-up of at least 6 months; study designs including review articles, editorials, and commentaries; and non-English publications. Although surgical timing between primary and revision operations was a key variable of interest, studies were not excluded based on whether this specific detail was reported.

Table 1

Inclusion and exclusion criteria

Search strategy

To identify studies for inclusion in this review, a librarian (E.A.B.) developed comprehensive search strategies for the PubMed (U.S. National Library of Medicine, National Institutes of Health), Scopus (Elsevier), and CINAHL Complete (EBSCOhost) databases. These databases were searched from inception through December 6, 2024. English language filters were applied. The search strategies utilized a combination of subject headings (e.g., MeSH in PubMed) and keywords related to surgery for RLDH. The PubMed search strategy was modified for the other two databases by replacing MeSH terms with relevant subject headings when available and maintaining similar keywords. The literature search was reported according to PRISMA-S, an extension of the PRISMA Statement for reporting literature searches in systematic reviews. Detailed search strategies for each database are provided in Appendix 1.

Screening and study selection

References were imported into the review management software (Covidence systematic review software, 2025; Veritas Health Innovation, Melbourne, Australia) for de-duplication and study selection. Two reviewers independently screened titles and abstracts to determine eligibility, with conflicts resolved by a third reviewer. Following the same process, two reviewers then independently screened full-text articles, again resolving conflicts with a third reviewer.

Quality assessment

The methodological quality of the included studies was assessed, and the risk of bias was evaluated. For case-control studies, the Newcastle-Ottawa Scale (NOS) Quality Assessment tool was used to assess the risk of bias (Table 2) [8]. For randomized controlled trials, the Cochrane Risk of Bias, version 2 (ROB-2) tool was employed (Table 3) [9].

Table 2

NOS for quality analysis of case control studiesa)

Table 3

Cochrane Risk of Bias 2 (Rob2) for randomized controlled trials

Data extraction

Data were extracted from the papers included in the systematic review by two independent reviewers using a customized data extraction form in Covidence. Conflicts were resolved by a third reviewer. The extracted data included specific details about the participants, interventions, and outcomes. When necessary, authors of the papers were contacted to request missing or additional data. For eligible studies, the extracted data comprised author names, year of publication, study design, sample size, perioperative outcomes (operation time, blood loss, and postoperative hospital stay), intraoperative complication rates (recurrence rate, dural tears, neurological deficit, instability, infection, blood transfusion, and deep vein thrombosis), clinical outcomes (Visual Analog Scale [VAS] for leg and back pain, Oswestry Disability Index [ODI], Japanese Orthopaedic Association [JOA], satisfaction). A summary of study characteristics is presented in Table 4.

Table 4

Studies included (n=10); study characteristics, interventions, demographics, and surgical timing

Statistical analysis

All statistical analyses were conducted using R Studio software ver. 4.4.3 (R Foundation for Statistical Computing, Vienna, Austria). Continuous variables, including patient age, body mass index (BMI), operation time, blood loss, postoperative hospital stay, and patient-reported pain and disability scores (VAS and ODI), were summarized as mean±standard deviation or as ranges where applicable. Categorical variables, such as sex, smoking status, alcohol consumption, and prevalence of diabetes mellitus (DM) and hypertension (HTN), were summarized using frequencies and percentages. Differences between groups for continuous variables were assessed using analysis of variance, with Tukey’s honest significant difference test for post-hoc analysis when significant differences were detected. Chi-square and Kruskal-Wallis tests were used to evaluate differences in categorical and non-parametric variables across surgical groups. Wilcoxon signed-rank tests assessed differences in VAS and ODI scores preoperatively and 12 months postoperatively within groups. Additionally, Z-score analysis was performed to identify significant outliers among mean changes in VAS and ODI scores. In a subgroup meta-analysis, independent t-tests and Wilcoxon rank-sum tests were used to assess differences in 12-month improvement in VAS and ODI scores between fusion and nonfusion groups. Statistical significance was set at an alpha of 0.05 for all tests.

Study inclusion

The number of articles retrieved, screened, excluded, and extracted is presented in the PRISMA flow diagram (Fig. 1). A total of 1,290 records were identified through database searches. After the removal of 569 duplicates, 721 eligible studies remained for title and abstract screening. Of these, 637 were excluded for not meeting eligibility criteria. The remaining 84 studies underwent full-text review, resulting in the exclusion of 74 studies. Ultimately, 10 studies met the inclusion criteria and were included in this systematic review. A summary of the study characteristics is presented in Table 4 [10–19].

Fig. 1

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) diagram.

Demographics and risk factors

The demographic and clinical characteristics of patients in the included studies varied widely (Table 4). Age ranged from a mean of 40.87 to 57.2 years, with no significant variation between studies (p=0.166). Sex distribution varied significantly across studies (p=0.034), showing a significant predominance of male patients (42%–80%). BMI did not differ significantly across studies (p=0.056), with values ranging from 23.73 to 29.5 kg/m². However, Ahsan et al. [11] in 2021 reported that 75% of patients in the nonfusion group and 80% in the fusion group had a BMI >30 kg/m². Although inconsistently reported, the prevalence of smoking, alcohol consumption, DM, and HTN showed significant intergroup variability (p<0.001) (Appendix 2).

Surgical timing

These studies were evaluated to determine the importance of surgical timing. Seven out of 10 studies included did not report data on the time interval between the index and revision procedures (Table 4). For the studies that provided this information, the reported timeframes were inconsistent and could not be synthesized. Ahsan et al. [11] in 2021 was the only study to report the mean pain-free interval (in months) and mean symptom duration prior to the second surgery (in months). Two studies reported the interval between surgeries but did not comment on symptom duration prior to the revision surgery [12,15].

Perioperative outcomes

Perioperative outcomes notably differed between surgical groups (Table 5). Fusion techniques were associated with significantly higher blood loss (range, 245–660 mL) compared to RD alone (range, 17.75–306.67 mL; p=0.039). Procedures involving fusion had longer operation times (range, 148.2–232.5 minutes) compared to RD alone (range, 72.8–125.3 minutes); however, this finding was not statistically significant (p=0.067). Similarly, postoperative hospital stays were generally longer following fusion procedures, although this difference did not reach statistical significance (p=0.277).

Table 5

Perioperative outcomesa)

Intraoperative complications

Overall complication rates varied substantially, ranging from 1.72% in the minimally invasive surgery (MIS)-transforaminal lumbar interbody fusion (TLIF) groups to as high as 86.66% in open fusion groups (TLIF/posterolateral fusion [PLF]). However, no statistically significant difference in complication rates was observed across all procedures (p=0.204). The recurrence rates of LDH did not differ significantly among surgical methods (p=0.262), although a trend toward lower recurrence was noted in fusion groups. Notably, the incidence of dural tears was highest among patients undergoing open fusion procedures (reaching up to 53.33%). Neurological deficits and postoperative instability were relatively uncommon, and their incidence did not differ significantly across procedures. Frequencies of commonly reported complication rates are presented in Fig. 2, with additional reported complications outlined in Appendix 3.

Fig. 2

Intraoperative complications heatmap. Kruskal-Wallis test (chi-square=19.216, degrees of freedom=15, p=0.2041), no statistically significant difference in complication rates across procedure groups. Turkey honest significant difference test (from analysis of variance [ANOVA], p=0.0485), showed largest differences from revision discectomy with fusion (highest) compared to minimally invasive transforaminal lumbar interbody fusion (MIS-TLIF) and percutaneous endoscopic lumbar discectomy (PELD) (lowest complication rates). ANOVA testing found no statistical differences in recurrence rate between procedure groups (p=0.262). Z-score analysis identified Mosaad et al. [10] (2023) fusion (TLIF/PLF) as an outlier for overall complication rate (Z=+3.136). RD, revision or repeat discectomy (unspecified); MD, microdiscectomy; MED, microendoscopic discectomy; PLF, posterolateral fusion; TLIF, transforaminal lumbar interbody fusion.

Postoperative outcomes

At 12 months, the mean change in postoperative pain (VAS back, VAS leg) and disability (ODI) scores did not differ significantly across all procedures (VAS back p=0.97, VAS leg p=0.89, ODI p=0.31). However, clinically relevant improvements were observed across all procedures (Figs. 3–5). In a study by Mossad et al. [8] in 2023, the RD with fusion (TLIF/PLF) group exhibited a larger reduction in ODI scores at 12 months compared to other fusion and nonfusion decompressive procedures (Z-score=−2.26); however, this difference seemed attributable to a higher baseline disability score. Overall, equipoise remained in the direct comparison of fusion versus nonfusion procedures within each study, as shown in Fig. 6A–C, with only Mosaad et al. [10] in 2023 and Yao et al. [13] in 2017 (March) indicating a preference for nonfusion in the weighted mean difference in VAS back scores (Fig. 6A), and Mosaad et al. [10] in 2023 showing a preference for nonfusion in the weighted mean difference in ODI (Fig. 6C). A subgroup analysis comparing fusion and nonfusion groups confirmed these results, demonstrating no significant differences between the groups for VAS or ODI improvement at 12 months (all p>0.3). Finally, clinical assessment via the JOA scores significantly improved across all surgical groups postoperatively. Patient satisfaction was high (>80%) across all surgical interventions, with the highest satisfaction ratings observed in patients undergoing TLIF, MIS-TLIF, and PELD (Table 6).

Fig. 3

Forrest plot, mean change in Visual Analog Scale (VAS) back (12 mo–preoperative) by procedure. CI, confidence interval; MD, microdiscectomy; TLIF, transforaminal lumbar interbody fusion; PLF, posterolateral fusion; MIS-TLIF, minimally invasive transforaminal lumbar interbody fusion; PELD, percutaneous endoscopic lumbar discectomy; MED, microendoscopic discectomy; RD, revision or repeat discectomy (unspecified); RE, random effect.

Fig. 4

Forrest plot, mean change in Visual Analog Scale (VAS) leg (12 mo–preoperative). CI, confidence interval; MD, microdiscectomy; TLIF, transforaminal lumbar interbody fusion; PLF, posterolateral fusion; MIS-TLIF, minimally invasive transforaminal lumbar interbody fusion; PELD, percutaneous endoscopic lumbar discectomy; MED, microendoscopic discectomy; RD, revision or repeat discectomy (unspecified); RE, random effect.

Fig. 5

Forrest plot, mean change in Oswestry Disability Index (ODI) (12 mo–preoperative). Z-score analysis (|Z|>2) detected Mosaad et al. [10] (2023) 2 as an outlier (Z-score=–2.26), indicating significantly larger mean change compared to other procedures. CI, confidence interval; MD, microdiscectomy; TLIF, transforaminal lumbar interbody fusion; PLF, posterolateral fusion; MIS-TLIF, minimally invasive transforaminal lumbar interbody fusion; PELD, percutaneous endoscopic lumbar discectomy; MED, microendoscopic discectomy; RD, revision or repeat discectomy (unspecified); RE, random effect.

Fig. 6

(A–C) Forrest plots. weighted mean difference in mean change in Visual Analog Scale (VAS)/Oswestry Disability Index (ODI) scores (12 mo–preoperative) for fusion vs. nonfusion groups by study. CI, confidence interval.

Table 6

Postoperative outcomes, clinical assessment, and patient satisfactiona)

Summary of key findings

RLDH remains a considerable clinical challenge and is characterized by substantial variability in patient outcomes and surgical complexity, depending on the strategy employed. This systematic review highlighted notable differences in perioperative outcomes between fusion and nonfusion procedures, particularly the higher blood loss associated with fusion techniques (p=0.039), which aligns with prior findings in the spine surgery literature [20,21]. Although postoperative pain and disability scores were similar across procedures, minimally invasive techniques such as MIS-TLIF and PELD demonstrated significantly better complication profiles, shorter hospital stays, and higher patient satisfaction rates, supporting the growing preference for minimally invasive approaches in spinal surgery [22,23].

Risk factor assessment and stratification

This systematic review reaffirmed key risk factors associated with RLDH, including male sex, obesity (BMI >30 kg/m²), smoking, DM, and larger annular defects [24,25]. These factors have been consistently identified in the literature, suggesting that targeted preoperative risk stratification could significantly enhance patient selection, thereby optimizing surgical decision-making and improving outcomes [24–26]. For instance, advanced risk modeling that incorporates demographic, anatomical, and other clinically relevant patient variables has shown predictive utility for recurrence and complication risk in spinal surgery [28–30]. Future studies should focus on developing accurate, validated predictive algorithms to guide surgical decision-making based on risk.

Timing of intervention

Our review revealed a substantial knowledge gap regarding the underreported time interval between the onset of recurrent symptoms and the subsequent revision procedure for RLDH. This information is essential for optimizing treatment timing and improving patient outcomes, yet it remains largely absent from existing literature. It is unclear whether clinical decision-making for revision mirrors strategies used in primary disc herniation. Previous studies suggest that early intervention may reduce the risk of permanent nerve damage and improve recovery trajectories; however, even in the context of primary disc herniation, the evidence remains mixed, and optimal timing is unclear [31,32]. Furthermore, the definition of RLDH varied between studies. For example, Yao et al. [13,14] in 2017 and Wang and Yu [16] in 2020 defined RLDH as symptom recurrence after at least 1 month of pain relief following primary lumbar disc surgery, while other studies required a minimum of 6 months. Standardizing definitions for RLDH, as well as for “early” versus “late” interventions, is crucial for meaningful comparative analysis in future investigations of ideal surgical timing. Additionally, incorporating psychological assessments, such as mental health scores, could provide valuable data on how timing impacts patient recovery [33,34].

Surgical strategy: fusion versus revision microdiscectomy

Both fusion and nonfusion methods have proven effective for managing RLDH, although significant distinctions in their risk-benefit profiles exist. Fusion procedures appear advantageous for reducing the risk of recurrence; however, these benefits come at the cost of higher perioperative burdens, including increased blood loss and potentially greater costs and resource utilization. This finding aligns with previous studies that report lower revision rates in fusion cohorts, albeit at the expense of increased perioperative morbidity and resource use [35,36]. Conversely, nonfusion techniques, particularly minimally invasive ones, offer substantial benefits in patient recovery, reduced complication rates, and shorter hospitalizations. Recent randomized controlled trials and meta-analyses strongly support minimally invasive nonfusion methods for appropriately selected, low-risk patients with limited instability or smaller annular defects [18,37,38]. However, the literature remains divided on whether these approaches are universally applicable, and the advantages of open fusion surgeries—such as lower recurrence rates and more complete disc removal—remain valid points of discussion. Several studies continue to show no significant long-term differences between minimally invasive and open techniques [6,39]. Therefore, surgical decision-making should be guided by patient-specific risk profiles and individual clinical presentations, as previously proposed.

Additionally, our review found no significant difference in patient-reported outcomes, including VAS and ODI scores, between fusion and nonfusion interventions, leaving it unclear which approach offers superior clinical improvements. This challenges the commonly held belief that fusion provides better symptomatic relief for patients experiencing significant axial back and radicular leg pain. While the Mosaad et al. [10] in 2023 TLIF/PLF group demonstrated significant improvements in ODI, the fusion group also had higher baseline ODI scores, raising questions about whether the greater postoperative improvement reflects true superiority or merely regression to the mean. Further research is needed to clarify the relationship between baseline VAS/ODI scores and surgical outcomes for fusion versus nonfusion procedures. Stratifying patients by baseline ODI scores along with long-term clinical outcome data could help identify whether specific patient subgroups benefit more from fusion or nonfusion strategies. Such data would enable more nuanced, evidence-based recommendations tailored to each patient’s baseline disability and symptom severity.

Although previous meta-analyses generally favored fusion approaches, the reported clinical outcomes remain inconsistent [6,40]. Lei et al. [6] found a significant improvement in JOA scores with fusion, whereas Santos et al. [40] reported that fusion was associated with significantly reduced VAS scores but without consistent improvement in functional recovery or other pain scores. These discrepancies likely reflect differences in statistical methodologies, follow-up durations, and heterogeneous inclusion criteria across an already variable pool of fusion versus nonfusion techniques. Given this lack of consensus, our finding of no significant difference in clinical outcomes between fusion and nonfusion procedures appears more appropriate. Moreover, both studies associated fusion with lower postoperative instability and reduced recurrence rates, which are important advantages in the context of RLDH management. However, the clinical relevance of segmental stability is difficult to interpret in the absence of long-term data on fusion durability and adjacent segment disease; thus, we excluded this from our analysis.

In the absence of consistent, statistically significant differences in clinical outcome measures between fusion and nonfusion approaches, our findings suggest that RD is preferable due to its lower perioperative burden. However, fusion may be warranted for high-risk patients, particularly those with large annular defects or biomechanical instability, where the potential to reduce recurrence may justify the added risks.

Limitations

A major limitation in our review was the lack of standardized reporting on the time interval between index surgery, recurrent symptom onset, and revision procedures for RLDH. This greatly restricted comprehensive meta-analytic synthesis and prevented conclusions on the optimal timing of revision surgery. Significant heterogeneity across the included studies, in terms of study design, patient population, and surgical technique, further limited direct compatibility. For example, the basic type of disc herniation probably plays a significant role in postoperative outcomes, yet this kind of granularity was completely absent from these data [41]. Many studies lacked standardization in defining surgical techniques, patient selection criteria, complications reported, and the definition of RLDH. Additionally, selection criteria for intervention types were often based on surgeon preference, introducing potential bias driven by institutional practices and other undisclosed criteria [42].

Our comparison of fusion versus nonfusion procedures reveals limitations due to considerable heterogeneity within each group. While this grouping was intended to enhance the depth of our meta-analysis, future research should stratify studies by surgical technique to better identify optimal approaches. Lastly, all included studies reported outcomes only up to 12 months postoperatively, limiting conclusions to the short-term postoperative period. Data on long-term outcomes are necessary to assess the durability of symptom relief, recurrence rates, adjacent segment disease, and the relative longevity or failure of fusion constructs. Extending follow-up intervals and establishing consistent reporting standards for RLDH studies is critical for future comparative research.

Future directions

There remains an urgent need to understand the time from recurrent symptom onset following index surgery to revision surgery in order to identify the optimal timing of treatment for patients with RLDH. However, our review demonstrates that this timeframe is vastly underreported, and clinical outcomes for both fusion and nonfusion interventions remain inconsistently characterized. Prospective studies with clearly standardized definitions of RLDH, consistent reporting of timing intervals from recurrent symptom onset to revision operation, and a comprehensive investigation of complications and patient risk factors would be ideal for identifying optimal management strategies for RLDH based on specific clinical circumstances. Admittedly, a study of this nature would be challenging to conduct. At a minimum, as a spine community, the depth and granularity of our registry data need to be improved to better address these types of questions. The integration of anatomical, demographic, functional, and physiological risk factors into stratified risk assessment models could significantly enhance decision-making and patient outcomes [43]. Further investigation of patient satisfaction and long-term quality of life outcomes associated with each approach could also refine surgical recommendations, ultimately improving patient-centered treatment of RLDH.