Comparative efficacy of 5 suture configurations for arthroscopic rotator cuff tear repair: a network meta-analysis

Rotator cuff tear is one of the most common complaint with shoulder pain, disability, or dysfunction. So far, different arthroscopic techniques including single row (SR), double row (DR), modified Mason–Allen (MMA), suture bridge (SB) and transosseous (TO) have been identified to repair rotator cuff. However, no study has reported the comparative efficacy of these 5 suture configurations. The overall aim of this network meta-analysis was to analyze the clinical outcomes and healing rate with arthroscopy among SR, DR, MMA, SB and TO. A systematic literature was searched from PubMed, EBSCO-MEDLINE, Web of Science, google scholar and www.dayi100.com, and checked for the inclusion and exclusion standards. The network meta-analysis was conducted using Review Manager 5.3 and SATA 15.0 software. Thirty-four studies were eligible for inclusion, including 15 randomized controlled trials, 17 retrospective and 2 prospective cohort studies, with total 3250 shoulders. Two individual reviewers evaluated the quality of the 34 studies, the score form 5 and 9 of 10 were attained according to the Newcastle–Ottawa Scale for the 17 retrospective and 2 prospective studies. There was no significant distinction for the Constant score among 5 groups in the 16 studies with 1381 shoulders. The treatment strategies were ranked as MMA, DR, SB, SR and TO. In ASES score, 14 studies included 1464 shoulders showed that no significant differences was showed among all 5 groups after surgery. Whereas the efficacy probability was TO, MMA, DR, SB and SR according to the cumulative ranking curve. The healing rate in 25 studies include 2023 shoulders was significant in both SR versus DR [risk ratio 0.45 with 95% credible interval (0.31, 0.65)], and SR versus SB [risk ratio 0.45 (95% credible interval 0.29, 0.69)], and no significant in the other comparison, the ranking probability was MMA, SB, DR, TO and SR. Based on the clinical results, this network meta-analysis revealed that these 5 suture configurations shows no significant difference. Meanwhile, suture bridge may be the optimum treatment strategy which may improve the healing rate postoperatively, whereas the DR is a suboptimal option for arthroscopic rotator cuff repairs.

and management to this disease costs 3 billion US dollars, every year [4]. According to the. American Academy of Orthopaedic Surgeons reports that only 16% of rotator cuff tears had been managed and treated appropriately whereas 31% "may" have been appropriate, and 53% were "rarely appropriate" [5]. This situation was still a challenge for rotator cuff tear repair worldwide, with the need to promote functional recovery and increase the healing rate.
In recent years, significant development has been made in both the operative and conventional therapies of shoulder pain and pathological conditions [6]. The surgical remedy should be employed if conventional treatment fails. In arthroscopic rotator cuff repair, different  It is critical to evaluate the comparative efficacy, directly and indirectly, with the existing data using network meta-analysis, and summarize and explain the broader evidence to understand the advantages of different suture techniques. Our purpose was to prove which application in arthroscopic RCR would improve the shoulder function and tendon healing better. Therefore, this study aimed to perform the network meta-analysis for the currently available functional results and healing rate of arthroscopic RCR with SR, DR, MMA, SB and TO.

Search procedure
This network meta-analysis was conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines. PubMed, EBSCO-MEDLINE, Web of Science, google scholar and www. dayi1 00. com were searched for articles published from January 2000 to March 2020 with the following words: "rotator cuff tears; arthroscopy; rotator cuff repair (RCR); single row (SR); double row (DR); modified Mason-Allen (MMA); suture bridge (SB); transosseous (TO)". From PubMed, we used the search strategies "rotator cuff tears" AND "single row" OR "double row" OR "modified Mason-Allen" OR "suture bridge" OR "transosseous" assembled with all included literature.

Inclusion criteria
The inclusion criteria included: (1) patients diagnosed with rotator cuff injury and repaired with arthroscopy; (2) the control group was any suture configurations of 5, they were compared between two groups respectively.
(3) the studies included clinical functional outcomes and healing rate for all groups, with outcomes in accordance with Constant score system, the American shoulder or elbow surgeons score system (ASES). (4) clinical followup at least 6 months; (5) randomized controlled trial (RCT), prospective or retrospective cohort studies.

Exclusion criteria
Case report, animal experiments, and basic medicine studies were excluded. Patients who underwent shoulder surgery were also excluded.

Data extraction and quality assessment
The title and abstracts of all the searched literatures were accessed, the duplicates and animal trials were removed. Time of publication, study type, first author, patient information, surgical technique, clinical outcomes, and healing rate was listed into the standard form to compare. All of them were extracted by two independent authors (Peng Li and Hui Kang). Another author (Yibin Meng) crosschecked all the included and excluded studies for any discrepant opinion. The quality of randomized controlled trials (RCT) was evaluated with Collaboration tool [7]. The judgment standard included six indexes: sequence generation, allocation hiding, blindness, incomplete result data, selective result reporting, and other "bias", "low risk", "high risk", or "unclear" were the grading standard for each index of the included studies. According to the Newcastle-Ottawa Scale (NOS) [8], the quality of prospective and retrospective cohort studies was evaluated. The two authors (Hongchuan Li and Jian Li) independently assessed the quality of these literatures.

Data analysis
Revman 5.3 software was employed for all conventional meta-analysis. The weighted mean difference (MD) and standard deviation (SD) were used to analyze the continuous variables (Constant, ASES), and relative risk was used to appraise the dichotomous variables (healing rate). Values were considered as statistically significant when P value < 0.05, including 95% CI. The I 2 statistical was selected to test the heterogeneity of included studies (significance, I 2 > 50%). For the pool outcomes of comparable studies, I 2 > 50% was considered the significant heterogeneity and belong to the random-effect model. The network meta-analysis was based on a frequent framework with indirect and direct comparing. Stata software (version 15.0) was used to perform network, forest, and predictive interval plots [9,10]. The rank of the five suture configurations for arthroscopic repair in the aspect of shoulder function and healing rate was assessed with the SUCRA [11]. The surface indicates the treatment efficacy, and the more surface shows, the better result. The inconsistencies were estimated with network side-split. The publication bias was judged with the funnel plot.

Study identification and assessment
One thousand ninety-five studies were identified, studies not fulfilling the inclusion criteria were excluded.

Constant score
Sixteen studies [12, 14, 15, 17, 19-21, 24, 27, 30, 31, 33, 35, 36, 39, 43], including 1381 shoulders assess, the clinical functions using Constant score, showed postoperatively difference between two groups in this network meta-analysis. The conventional meta-analysis is presented in Fig. 3a-c (MD with 95% CI). The network plot between the five techniques, and the network metaanalysis is shown in Fig. 3d-g. In Constant score, direct and indirect comparison by conventional and network meta-analysis illustrated no significant differences among SR, DR, SB, MMA, and TO (Fig. 3e, f ). According to the SUCRA (Fig. 3g), the ranking probability of the treatment efficacy of each suture configuration for Constant score was MMA, DR, SB, SR, and TO.

ASES score
Regarding the ASES score, 14 studies [12, 15, 20, 21, 23-25, 27, 29, 30, 33, 39, 41, 43], including 1464 shoulders assess, the clinical function between the different two groups postoperatively in this network meta-analysis. The conventional meta-analysis is shown in Fig. 4a-c (MD with 95% CI). The network plot between the five techniques and network meta-analysis is shown in Fig. 4d-g. In the ASES score, it was no significant between any two sutures configurations in the 14 studies (Fig. 4e, f ) with direct and indirect comparison by both conventional and network meta-analysis. On the basis of the SUCRA (Fig. 4g), the ranking probability of the treatment efficacy of each suture configuration for ASES score was TO, MMA, DR, SB and SR.

Healing rate
Twenty-five studies [13-16, 18-22, 24-26, 28-30, 33, 34, 36, 37, 39, 41-45], including 2023 shoulders assess, the healing rate between differentiate two groups postoperatively in this network meta-analysis. The conventional meta-analysis is presented in Fig. 5a-d (RR with 95% CI). The network plot between the 5 techniques and network meta-analysis is shown in Fig. 5e-h. Regarding the healing rate, there was significant differences both SR versus DR and SR versus SB in the network meta-analysis (Fig. 5g), and no significant in the other comparison ( Fig. 5f, g). Judging from the SUCRA (Fig. 5h), the ranking probability of the treatment efficacy of each method for healing rate was MMA, SB, DR, TO and SR.

Inconsistency and bias of publication analysis
The outcomes of pair-wise meta-analysis and the network meta-analysis matched significantly. In the study, there was no inconsistency for each result between the direct and indirect comparison (Table 2). Moreover, no visual evidence of bias of publication for each outcome was demonstrated from the funnel plots (Fig. 6), and it was similarly balanced on both sides of the funnel.

Discussion
The network meta-analysis revealed the comparative efficacy of 5 suture configurations for rotator cuff tear in terms of the Constant score, ASES score, and healing rate for patients who underwent arthroscopic repair with MMA, SB, SR, DR, and TO. This study showed the following: (1) there was no significant differences among the five suture configurations in term of Constant score, and the overall ranking was MMA, DR, SB, SR, and TO; (2) there was no significant difference in ASES score, and the overall ranking was TO, MMA, DR, SB, and SR. (3) SR leading to a lower healing rate than DR and SB, and the ranking for healing rate was MMA, SB, DR, TO, and SR. The constant score is a critical criterion for shoulder treatment including shoulder function, range of motion, pain and strength [46]. Gerhardt et al. [14] found that clinical results after MMA and SB techniques do not demonstrate significant differences in a matched patient cohort. Moreover, Hantes et al. [13] and his co-worker found no difference in Constant scores between the SR and DR techniques in a 46 months follow-up study having 66 patients. Furthermore, Garofalo et al. [43] reported that MMA repair provides comparable clinical results to SR repair in Constant score with arthroscopy. Zafra et al. [36] suggested that there were no differences in Constant score between SR and SB techniques. The network metaanalysis compared the difference among the 5 techniques combined direct and indirect evidences for rotator cuff repair with quantitative way, which illustrated no significant difference among 5 suture configurations. The SUCRA was used to assess the slight differences among MMA, SB, SR, DR and TO. In order to achieve better Constant score, the techniques can be arranged as follows: MMA, DR, SB, SR and TO.
The ASES score is essential for evaluating the therapeutic effect of these five arthroscopic techniques. No difference among the MMA, SB, SR, DR, and TO has been analyzed by the previous evidence-based study. Khalil et al. [12] previously reported that MMA provides comparable functional results to the SB repair technique.
McCormick et al. [20] considered that using SR, DR, or SB techniques, yielded a clinical improvement and revealed no statistically significant difference for ASES score. Garofalo et al. [43] reported no statistically significant difference between SR and TO for the rotator cuff repair in the comparative analysis of ASES scores. No significant differences among MMA, SB, SR, DR and TO repair was found in terms of ASES score from this network meta-analysis. Furthermore, TO technique provided a greater ASES score than MMA, DR, SB and SR techniques according to the SUCRA.
As we all know that the critical point of RCT requires the repaired site tend-to-bone surface healing [47,48]. One of the most common reasons for the failure of an RCR is the retear because of nonhealing of the primary repair [49]. A study by Park et al. [27] showed that approximately 50% of repaired rotator cuffs do not heal completely, though the surgical techniques were used. Hantes et al. [13] found that significant superior healing rate was potentially provided with the DR rather than the SR technique, which may due to the contact surface of tendon and bone. So Franceschi et al. [34] suggested that in selected patients with required accelerated postoperative rehabilitation, double-row repair lowered the risk of retear, while maintaining a low rate of stiffness. Tudisco et al. [19] reported that the healing rates after arthroscopic rotator cuff repair were 89.2% and 95.3%, respectively, for the SR and SB techniques, which was statistically significant. According to the SUCRA, the treatments efficacy was ranked as MMA, SB, DR, TO and SR repair based on their healing rate. Our study has several advantages. Firstly, except for only direct groups compare, this network meta-analysis assesses five treatments simultaneously indirectly. As to our knowledge, this is the first time of comparison of MMA, SB, SR, DR, and TO techniques for arthroscopic rotator cuff repair. We compared the five different methods and supplied the SUCRA indirectly with a frequentist framework for network meta-analysis when no headto-head compare existed by combining directly [50,51]. Secondly, we avoided selection bias by synthesizing much more studies rather than a conventional meta-analysis. Additionally, this study could gain more precise effect assessments for the five techniques with an updated statistical approach of network meta-analysis.
However, there are also some limitations of the network meta-analysis: (1) some low-quality RCTs and two prospective and 17 retrospective cohort studies, which may impair the significance of the conclusions, but according to the NOS, the mostly score were more than 7. (2) The outcomes were incomplete in some included studies, imputed data were used in the analysis, as we used the same imputation method for the same treatment, the outcomes should still supply effective evaluation. (3) We did not perform a meta-analysis on tear size because the results were reported rarely in 2 studies. In our study, we compared the overall treatment efficacy for all types of rotator cuff repair. (4) Some potential publication biases in the study, it was similarly balanced on both sides of the funnel and demonstrated no visual evidence of publication bias for each result.

Conclusion
Our network meta-analysis revealed that no significant difference was found for the functional outcomes among the five suture configurations. SB repairs might be the optimum treatment and improve the healing rate postoperatively. Meanwhile, the DR is a suboptimal option for arthroscopic rotator cuff repairs, which may help and guide clinicians on the appropriate operative program.