Discectomy Combined with Annulus Fibrosus Repair for the Treatment of Lumbar Disc Herniation: A Bayesian analysis of Random Controlled Trials

Objective To investigate clinical effects of discectomy combined with annulus fibrosus (AF) repair for the treatment of lumbar disc herniation. Methods We searched PubMed, Embase, Web of science, CNKI and WanFang data for studies evaluating discectomy combined with AF repair and discectomy for the treatment lumbar disc herniation. Two reviewers selected studies, assessed quality and extracted data. This meta-analysis was performed to calculate weighted mean difference (WMD), risk ratios (RRs), and 95% confidence intervals (CIs).


Introduction
Lumbar disc herniation (LDH) refers to a series of clinical symptoms, such as lumbar back pain, numbness, pain and weakness of one or both lower extremities, caused by the breakthrough of annulus fibrosus (AF) in the nucleus pulposus of the intervertebral disc and the stimulation or compression of adjacent spinal nerve roots. Currently, for LDH patients with mild symptoms, conservative treatment such as medicine, bed rest and physiotherapy is often used, but some patients cannot tolerate pain in the process of conservative treatment or for patients who have failed to undergo regular conservative treatment, surgery is often needed.
Surgical interventions included discectomy, lumbar fusion and internal disc replacement.
Discectomy considered as a classic procedure for the treatment of LDH. 1 During the operation, some patients can feel the obvious relief of symptoms at the moment of being relieved of nerve compression. However, the amount of nucleus pulposus removal has been a controversial issue for clinicians. 2 Some studies have suggested that excessive removal of nucleus pulposus tissue to reduce the rate of recurrence lumbar disc herniation. 3,4 However, studies have shown that excessive removal of nucleus pulposus tissue could accelerate normal disc degeneration, reduce the height of the intervertebral disc after surgery, change the overall structure and biomechanics of the lumbar spine, and even lead to the risk of instability. 5,6 Afterwards, some scholars supported limited removal of nucleus pulposus tissue and a small amount of nucleus pulposus removal, which could reduce the operation time and speed up the recovery of patients, but the surgical 4 procedure has significantly increased the recurrence rate after surgery. 2,7 In traditional surgery, there is no solution for the ruptured AF after the removal of the nucleus pulposus tissue. Due to the poor nutritional supply in the intervertebral disc AF, the healing is slow and the anti-stress strength is low, 8 thus, there is a high risk of re-protrusion of nucleus pulposus at the original rupture site. 7 In recent years, AF repair technology has been gradually applied to repair ruptured AF after nucleus pulposus removal, and the therapeutic effect of AF repair has been evaluated and analyzed in relevant literatures. However, due to the small sample size and inconsistent conclusions of a single study, it is not enough to provide a basis for clinical treatment. Therefore, we performed this meta-analysis to investigate clinical effects of two methods for the treatment of LDH and look forward to providing evidence-based basis for clinicians.

Materials And Methods
We conducted this meta-analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. 9 The protocol for the metaanalysis was registered with the International Prospective Register of Systematic Reviews (CRD42018110414).

Search Strategy
We searched PubMed, Embase, Web of science, CNKI and WanFang data, until August, 2019 for studies evaluating the treatment effect of lumbar disc herniation by discectomy combined with AF repair and discectomy for the treatment lumbar disc herniation, without languages restrictions. Search terms included the keywords related to discectomy, annulus, lumbar, and their variants. The details of our search strategy are available in the Supporting Information (Appendix A). Additionally, we manually searched the references 5 lists of all eligible studies to identify additional potential studies.

Study selection and eligibility criteria
Inclusion criteria: (1) Type of participant: patients with lumbar disc herniation; (2) Type of exposure: discectomy combined with annulus fibrosus repair; (3) Type of study design: randomized controlled trials; (4) Type of outcome: the primary outcomes of this metaanalysis were recurrence rate, reoperation rate, postoperative reduction of VAS and ODI, postoperative complications. Secondary outcomes were operative time (minutes), intraoperative blood loss (ml) and length of surgical incision (cm); (5) Follow-up time is not less than one year.
Exclusion criteria: (1) patients with other lumbar diseases; (2) cases, reviews, abstract, and articles with incomplete summary data; (3) articles published repeatedly or using the same database.

Data extraction
We extracted the following information from each included study: author, year, sample size, age, gender, and duration of follow-up. If there are some disagreements, we resolved by discussion to reach an agreement. All extracted data were entered into a predefined standardized Excel file (Microsoft Corporation, USA). When we found duplicate reports of the same database, we extracted data from the study with most complete data set.

Quality assessment
The quality of the randomized controlled trial articles was used by modified JADAD. 10 The evaluation includes: random sequence generation, randomization hiding, blinding, and withdrawal. The first three questions, each one is 2 points, the scores are based on appropriate, unclear, inappropriate, followed by 2 points, 1 point, 0 points, respectively; the last question is 1 point, depending on whether there is withdrawal reasons, give 1 point or 0 points. We assigned scores of 1 to 3, and 4 to 7, for low, and high quality of 6 studies, respectively.
Statistical analysis RevMan 5.3 (The Cochrane Collaboration, Denmark) was used to statistical analyses. The I 2 statistic is used to quantify the heterogeneity of the included studies. 11 An I 2 value greater than 50% indicates significant heterogeneity, random-effects model was used; otherwise, fixed-effects model was used. Dichotomous data was analysis by relative risks (RRs) and continuous data was used by weighted mean difference (WMD). A two-sided P value less than 0.05 was considered statistically significant. Publication bias was assessed with funnel plots.

Search results
Two reviewers undertook the searches independently, and a total of 972 articles were retrieved. Endnote X8 (version 18.0.0.10063) was used to remove 390 duplicate studies.
Besides, we deleted 454 irrelevant articles, 33 reviews, 31 conference papers, and 24 cases/comments/letters through the title and abstract. Afterwards, we deleted 13 duplicate data articles (Appendix B), 13 cohort studies (Appendix B), and 7 no control group articles through the full text. Finally, 7 articles were included. 12− 18 The study selection process was shown in Figure. 1.

Study characteristics
These included studies were published between 2013 and 2019, with sample sizes ranging from 45 to 727 subjects, and a total of 1,718 subjects. 915 subjects undergoing discectomy combined with annulus fibrosus repair surgery. All follow-up time is not less than one year. The main characteristics of the included trials are summarized in Table 1.
The average score was 4.57 (range, 4-7), suggesting that all the studies were high quality.
For the generation of random sequences, seven articles provide computer-generated random numbers or similar methods. Only one article in randomization hiding provides specific details (NCT01283438), 16 and six articles indicate the use of random number table or other random allocation scheme, but did not provide details. Two articles mentioned the article used blind, 12,16 and the other five articles do not mentioned the blind method; all articles describe the number and reasons for withdrawal. The details of the article quality are shown in Table 2.

Clinical outcomes
The primary outcomes

Recurrence rate
Six articles provided available data in recurrence rates. 13− 18 There was no significant heterogeneity (P > 0.1, I 2 = 0%), so the fixed effect model was used for analysis. The pooled outcomes demonstrated that a 2.16-fold higher risk of recurrence in the AF unrepaired group than in the repair group [RR = 2.16, 95% CI (1.50, 3.10), P < 0.00; Five studies compared the intraoperative blood loss between the two groups. 13− 15,17,18 No 9 significant heterogeneity was detected between the studies (P > 0.1, I 2 = 0%), a fixed effect model was used. The pooled outcomes indicated that there was no significant difference in intraoperative blood loss between the two groups [WMD = -1.23, 95% CI (-4.46, 2.00), P = 0.45; Fig. 7].

Operative time (minutes)
Five studies provided available data in operative time. 13− 15,17,18 A significant heterogeneity was detected between the studies (P < 0.1, I 2 = 67%). Therefore, the random-effects model was used to do analysis. The pooled outcomes suggested that the operation time of the AF repaired group was significantly longer than that of the unrepaired group at 6.73 min [WMD = 6.73, 95% CI (2.80, 10.66), P < 0.01; Fig. 8].

Length of surgical incision (cm)
Two studies compared the length of surgical incision between the two groups. 14,15 No significant heterogeneity was detected between the studies (P > 0.1, I 2 = 0%), a fixed effect model was used. The pooled outcomes indicated that there was no significant difference in length of surgical incision between the two groups [WMD = 0.08, 95% CI (-0.01, 0.18), P = 0.10; Fig. 9].

Publication bias
Considering the limited number of studies (less than 10 studies) in all outcomes, so we did not use the Begg's test and the Egger's test for quantitative publication bias. The funnel plot was used by Reveman5.3, and the symmetric characteristic of the pooled plot indicated that no significant publication bias (shown in Fig. 10).

Discussion
The main finding of this meta-analysis demonstrated that discectomy combined with AF repair is superior to discectomy regarding postoperative recurrence rate, reoperation rate and postoperative pain relief, but the operation time is slightly longer than the latter.
There is no significant difference in intraoperative blood loss, length of surgical incision, postoperative complication rate and reduction in ODI between the two surgical procedures. Although all studies included in this meta-analysis were RCTs and the heterogeneity was low, the number of included studies and the total sample size is limited. Therefore, we extracted the data from 13 cohort studies, and analyzed it according to the statistical analysis method of the meta-analysis. The pooled outcomes were similar to ours, and further prove the stability of our research results. The pooled results of the cohort study are shown in Appendix S2.
The surgical approach of LDH should aim to maximize the restoration of the stability of the intervertebral disc structure and better balance of forces. Discectomy is an effective surgical procedure for the treatment of LDH, but the postoperative recurrence rate is 10-15%, 19,20 and similar to the total recurrence rate of AF unrepaired group in this metaanalysis (12.5%). There are many factors affecting the recurrence of LDH, such as postoperative defect of the annulus fibrosus, the degree of disc degeneration, obesity, smoking and other factors. 4,7,21− 25 Lebow et al. found that recurrence after nucleus pulposus removal occurred mostly in the original nucleus. 26 Rupture of AF around with high level of phospholipase A 2 , which aggravates the hyperplasia of scar tissue, 27 hyperplastic scar tissue can compress nerve root or dural sac and increase postoperative recurrence rate. Therefore, the surgeon repairs the ruptured AF, reducing the stimulation of nerve roots. 28 The study of Carrage has shown that a significant correlation between the integrity of AF and postoperative recurrence. 29 By comparing the area of the AF rupture, McGirt et al. found that the incidence of reoperation required for patients with an AF rupture greater than 54 mm 2 was approximately 18%, whereas for patients with a diameter less than 36 mm 2 , the incidence of reoperation was 4.7%. 5 Bostelmann et al. performed biomechanical analysis of human corpses and found that the pressure of the intervertebral disc after AF repair was basically close to the preoperative level. 30 Our meta-analysis showed that the AF-repaired group significantly reduced postoperative recurrence rates, and the same conclusion was obtained in 11 cohort studies [RR = 3.10, 95% CI (1.80, 5.33), P < 0.01, Appendix B].
The rLDH patients with mild symptoms can be treated conservatively. However, some patients who cannot tolerate symptoms such as pain and numbness need to be surgery again. Lebow et al. showed that surgery was used for the rLDH patients, the postoperative treatment was not as effective as the initial postoperative rehabilitation, 26 and the reoperation would lead to biomechanical changes of the intervertebral disc and accelerate the degeneration of adjacent discs. 31 The study of Klassen has shown that patients undergoing second surgery for LDH have a postoperative ODI value of 2.9-folds, 3.6-folds incidence of lower extremity pain, and 1.4-folds incidence of back pain that of primary surgery. 32 Mainly because the second operation increased the degree of ruptured AF scar adhesion, and the removal of the nucleus pulposus tissue again caused the height of the intervertebral disc to be significantly lower than that primary surgery. In addition, with the height of the intervertebral disc changes, the overall structure of the lumbar spine changes as well, accelerates the degeneration of the facet joints and even the instability of the lumbar spine in their remainder lifetime. 26,33 Therefore, reoperation is not an ideal choice; the important is how to reduce the recurrence rate and reoperation rate of patients after initial surgery. The AF repair technique can significantly reduce the incidence of secondary surgery in our meta-analysis, and the pooled outcomes of the seven cohort studies also showed that the incidence of reoperation in the AF unrepaired group was 3.8 times that of the repair group (shown in Appendix B).
For the secondary outcomes, there was no statistically significant difference in the intraoperative blood loss and the length of the surgical incision, and it was consistent with the results obtained by the cohort study, as shown in Appendix B. However, the metaanalysis showed that the AF repair group had more operative time than the unrepaired group (6.73 min), and the combined results of the nine cohort studies showed that the AF repair group had more operative time than the unrepaired group 3.12 min. The AF repair group can remove limited nucleus pulposus tissue than the unrepaired group, which may save some operation time, but it will increase the operation time during the repaired process. The results of the combination of RCTs and the cohort studies showed that the increased operation time of the AF repaired group within 10 minutes, which may not have clinical significance for the entire operation time.

Call for Future Studies
Materials used for AF repair is a polymer polyethylene material, which is an inert biological material and not easily degraded in the body. It is unclear that the suture knot as a foreign matter increases the formation of scar adhesion. (2) The Barricaid produced in the United States is another repair material for AF. Can the filled material be blended with the surrounding tissue in the late stage? In addition, the bone anchor fixator needs to be fixed in the endplate, which causes the endplate to deform. 34

Conclusions
Discectomy combined with AF repair is superior to discectomy regarding postoperative recurrence rate, reoperation rate and postoperative pain relief, but the operation time is slightly longer than the latter. There were no significant difference in intraoperative blood loss, length of surgical incision, postoperative complication rate and reduction in ODI between the two surgical procedures. However, due to the small number of studies and the small sample size, we expect a multi-center, large sample, high quality, double-blind RCT study to further update our conclusions.  Figure 1 The flow chart of studies selecting.
21 Figure 2 The forest plot for recurrence rate between the two groups. The forest plot for reoperation rate between the two groups. The forest plot for postoperativse reduction of VAS between the two groups.
22 Figure 5 The forest plot for postoperative reduction of ODI between the two groups.
23 Figure 6 The forest plot for postoperative complications between the two groups.
24 Figure 7 The forest plot for intraoperative blood loss (ml) between the two groups.

Figure 8
The forest plot for operative time (minutes) between the two groups. The forest plot for length of surgical incision (cm) between the two groups.
25 Figure 10 The funnel plot for publication bias