Optimal Surgical Treatment Method for Anterior Cruciate Ligament Rupture: Results from a Network Meta-Analysis

Abstract

Background

Previous studies have shown that primary repair (PR) and anterior cruciate ligament reconstruction (ACLR) can effectively treat ACL injuries. Our study aimed to compare different treatments of ACL tears, including autograft, allograft, hybrid graft ACLR, and PR, by assessing clinical outcomes and adverse events.

Material/Methods

PubMed, Cochrane Library, Embase, and CNKI databases were searched and a frequentist-framework network meta-analysis was used.

Results

Overall, PR with augmentation was superior to ACLR only for activity recovery (WMD 0.28 95%CI [0.07 to 0.49]), and there was no significant difference shown between PR without augmentation and ACLR. ACLR with irradiated allograft was a poor option for the treatment of ACL rupture, showing the weakest subjective evaluations and functional outcomes and worst safety profile. PR with or without augmentation provided fairly good postoperative efficacy results and produced less postoperative knee laxity than irradiated allograft ACLR (PR: standardized mean difference [SMD] −1.27 [−1.80 to −0.74]; ACLR: SMD −1.36 [−1.88 to −0.83]). However, PR without augmentation showed a high failure rate compared with autograft ACLR (autograft vs PR without augmentation: risk ratio 0.29 [0.10 to 0.85]).

Conclusions

For surgical treatment of ACL rupture, irradiated allograft ACLR had the worst efficacy and safety and is not recommended. PR may be an ideal treatment method in terms of efficacy but it is related to a significantly higher revision risk if without augmentation. Autograft ACLR may be the preferred method currently available for most patients requiring surgical treatment of ACL rupture.

Background

Anterior cruciate ligament (ACL) injury is a common knee sports injury that often occurs during high-intensity physical exercise [1]. In addition, it is estimated that 1 in every 120 people of working age have ACL injuries [2]. The ACL is the main structure limiting the tibia’s anterior translation and internal rotation. In most cases, a tear or rupture is caused by excessive pivoting or anterior or valgus movement of the proximal tibia, rather than by direct contact [3].

Before the popularization of arthroscopy, the most common surgical option for treating ACL tears was open primary repair (PR). This provided good short-term outcomes but was associated with long-term problems, including high re-rupture and reoperation rates and postoperative complications [4–6]. These problems together with the development of ACL reconstruction (ACLR) led to a decline in the use of PR. However, with advances in arthroscopic technology, preoperative imaging, and surgical techniques, such as suture anchors and suture augmentation, PR is once again a viable option for ACL repair. More recent studies showed that compared with ACLR, PR had a comparable failure rate [7] and provided an earlier return of range of motion [8,9], most likely resulting from the less invasive nature of the surgery. It also has no donor site morbidity because of the lack of graft collection.

ACLR is regarded as the most accepted standard surgical therapy for active and symptomatic patients with ACL injury. There are 3 main approaches for ACLR based on graft type: autograft, allograft, and hybrid graft (combination of autograft and allograft) [10,11]. Autograft is the oldest and most common ACLR technique used because it has some unique advantages, such as the lowest rejection reaction and relatively high mechanical reliability [12]. In contrast, allograft ACLR causes no donor site morbidity but has a lower graft survival rate and a higher postoperative infection rate [3,13,14]. Hybrid graft ACLR provides a graft of suitable size, with clinical results comparable to that of autografts, and offers an alternative to autograft ACLR, especially for those patients whose tendons are damaged or undersized at harvest [15].

Despite several comparative studies, it is still unclear which is the best surgical method for ACL rupture. Previous meta-analyses have compared clinical outcomes of the different types of ACLR. The study by Li et al [16] showed the performance of the autograft and hybrid graft was similar in graft failure, graft diameter, and reoperation ratio. Meanwhile, the study by Sun et al [17] also revealed no difference in long-term outcomes and failure risk after cruciate ligament reconstruction with either autograft or synthetics; however, autografts were found to be inferior to synthetics in dealing with restoring knee joint stability and were associated with more complications. Several clinical trials have reported functional outcome scores and revision rates after PR [18–20]. However, there are few randomized controlled trials (RCTs) or systematic reviews that have compared the efficacy and safety of the various types of ACLR and PR.

In summary, we speculated that PR could also bring ideal therapeutic effects and potentially be superior to ACLR. This network meta-analysis was aimed to compare the clinical outcomes and adverse events of autograft, allograft, hybrid graft ACLR, and PR in the surgical treatment of ACL tears. This study was not focused on professional athletes.

Material and Methods

Data Sources and Searches

This study was registered in the PRESPERO (registration no. CRD42021225778). The Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Network Meta-analysis (PRISMA-NMA) guidelines [21] were strictly followed throughout this study. MeSH keywords combined with a free words retrieval strategy was adopted to search the PubMed/Medline, Cochrane Library, Embase, and CNKI databases from January 1995 to July 2022. The specific query for PubMed/Medline was as follows: ((anterior cruciate ligament injury) OR (anterior cruciate ligament injury) OR (anterior cruciate ligament rupture) OR (anterior cruciate ligament tear) OR (ACL injury) OR (ACL rupture) OR (ACL tear) AND (ACL[MeSH Terms]) OR (anterior cruciate ligament[MeSH Terms]) AND (reconstruction) OR (transplantation) OR (repair) OR (primary repair) OR (suture)). The specific query for Cochrane Library was as follows: “MeSH descriptor.[Anterior Cruciate Ligament Injuries] explode all trees” OR “MeSH descriptor.[Anterior Cruciate Ligament] explode all trees” OR “(ACL injury): ti,ab,kw OR (ACL rupture): ti,ab,kw OR (ACL tear): ti,ab,kw” OR “(anterior cruciate ligament injury): ti,ab,kw OR (anterior cruciate ligament rupture): ti,ab,kw OR (anterior cruciate ligament tear): ti,ab,kw” AND “(reconstruction): ti,ab,kw OR (transplatation);: ti,ab,kw OR (repair): t,ab,kw OR (suture): ti,ab,kw OR (primary suture): ti,ab,kw”. The specific query for Embase was as follows: ((“anterior cruciate ligament injury”: ti, ab, kw OR “anterior cruciate ligament rupture”: ti, ab, kw OR “anterior cruciate ligament tear”: ti, ab, kw OR “ACL injury”: ti, ab, kw OR “ACL rupture”: ti, ab, kw OR “ACL tear”: ti, ab, kw) AND (“reconstruction”: ti, ab, kw OR “transplantation”: ti, ab, kw OR “repair”: ti, ab, kw OR “primary repair”: ti, ab, kw OR “suture”: ti, ab, kw)). The specific query for CNKI was as follows [in Chinese]: ((主题: 前交叉韧带损伤 + 前交叉韧带断裂 + 前交叉韧带撕裂 (精确) OR (主题: 前交叉韧带重建 + 前交叉韧带移植 + 前交叉韧带修复 + 前交叉韧带缝合 (精确)). A reference lists of identified articles were further reviewed to find additional eligible studies. No language restriction was set for the publication selection.

Study Selection

A research protocol under the PICO principle was pre-drafted as follows: (1) Population: patients underwent ACL rupture and planned surgery; (2) intervention: surgical treatment for ACL rupture; (3) comparison: different surgical methods for ACL rupture; and (4) outcomes: postoperative subjective feelings of patients, recovery of postoperative activity, functional improvement, joint laxity, and treatment failure.

Based on the PICO protocol, studies that fulfilled all of the following criteria were included: (1) Patients underwent ACLR or PR to treat ACL tears; (2) study compared 2 or more different surgery methods; (3) study used an RCT design, non-RCT design, or cohort design (CHS); and (4) the following outcomes were reported: subjective evaluation improvement, functional rehabilitation, improvement of activity, postoperative knee laxity, or the incidence of revision surgery.

The exclusion criteria were as follows: (1) The study included the revision of ACL rupture; (2) low-quality of study; (3) studies with a non-prospective design; and (4) animal or vitro basic studies, observational studies, reviews (including systematic review and meta-analyses), meeting abstracts, letters, and those with inaccessible original study data.

We would also try to contact the corresponding authors for the studies lacking complete information, and those for which no response was received were excluded. The source of data was also requested from the corresponding author for the studies presenting outcomes only in figures but not in numeric datasets, and when no response was received, 2 authors would independently try to obtain the data by the measurements shown in the figures. When there was still no access to the raw data after the above attempts, the study was excluded.

Data Extraction and Quality Assessment

The study quality was assessed by 2 authors independently. Quality of method and bias risk for RCTs was assessed by using the Cochrane risk bias assessment tools [22]. The Newcastle-Ottawa Scale (NOS) [23] was used for the evaluation of the method quality of CHSs and non-RCTs. Three main domains – the selection of the study groups, comparability among different groups, and ascertainment of either the exposure or outcome of interest – were evaluated with a score from 0 to 9. All disputes were resolved during a discussion between the 2 authors. The following information was extracted from each included study: first author, year of publication, number of participants, average age, sex ratio, traumatic mechanism, method of surgery, average follow-up time, study design, and outcome.

Outcome Measures

For the evaluation and comparison of the efficacy and safety of the surgical methods, the following outcome measurements were selected: postoperative subjective feelings of patients, recovery of postoperative activity, functional improvement, and safety endpoints.

For measuring postoperative function status, no restriction was placed on the types of questionnaires used in functional evaluation. The Lequesne Index was the first choice, followed by other functional measurement scales, including the Knee Injury and Osteoarthritis Score (KOOS) and the Western Ontario and McMaster Universities Arthritis Index (WOMAC). The standardized mean difference (SMD) was used to incorporate these different scales into the same network.

By using the Subjective International Knee Documentation Committee (IKDC) score, postoperative subjective feelings were assessed. The recovery of postoperative activity was assessed with the Tenger score. The weighted mean difference (WMD) with 95% confidence intervals (CIs) was calculated for the IKDC and Tenger scores.

The safety endpoints were joint laxity and treatment failure. The degree of laxity was the relaxation difference comparing the operated knee with the non-operated knee measured by the KT-1000 or KT-2000 scale. The SMD with a 95%CI was calculated for the laxity scores. Failure was defined as all patients who had a re-tear or recurrent laxity, regardless of whether or not they underwent revision surgery. The risk ratio (RR) and 95%CI were used for the measurement of the relative failure rate.

Statistical Analysis

Frequentist method network meta-analysis was conducted in Stata/MP (version 14.0, Stata Corp, College Station, TX, USA) with a random-effects model. The proportional variance-covariance matrix data were pooled by multivariate meta-regression with the random effect; restricted maximum-likelihood was applied in the evaluation of model fit [24].

By using global inconsistency tests and node-split tests, the inconsistencies were evaluated, and the consistency model was used under the condition of no existing significant inconsistency, or else the sensitivity analysis would be applied for the inconsistency source identification. Funnel plots and Egger’s tests were used to assess potential publication bias for each endpoint, and by using the trim and filling method, endpoints with underlying asymmetric funnel plots were estimated for whether significant publication bias existed or not [25]. We also ranked the relative efficacy and safety of different surgical methods through the surface under the cumulative ranking (SUCRA) values [26], and cluster-ranking plots for the optimal choice.

A direct pairwise meta-analysis was conducted to compare the relative efficacy of PR (with or without augmentation) with autograft using RevMan (Review Manager, Version 5.3, Copenhagen, The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). The heterogeneity across studies was tested by the Q and I² statistics. The fixed-effects model was preferred, but if there was significant heterogeneity (P<0.05 or I² >50%), the random-effects model was applied.

Although CHSs and non-RCTs can provide important data, study designs introduce unmanageable confounding factors and potential bias. Because of this, a subgroup analysis that included only RCTs was performed to reconfirm the results obtained by the main network meta-analysis. A network plot was used to graphically summarize the evidence incorporated into this network meta-analysis, in which the lines between treatment nodes indicated the direct comparisons made within the evidence and the size of nodes indicated the number of population involving in each treatment.

When the 95%CI did not cover 1 for RR or 0 for SMD and WMD, it was considered significant for differences between treatments. P<0.05 was considered statistically significant.

Results

Literature Selection

Thirty-six studies [27–62] were identified through systematic screening (Figure 1). Six different surgical methods were identified and analyzed: autograft (Autograft), allograft with irradiation (irAllograft), allograft without irradiation (nirAllograft), hybrid graft (Hybrid), PR with augmentation (waRepair), and PR without augmentation (nwaRepair). The irAllograft group was chosen as the standard control group because in the pre-analysis it had the worst performance for efficacy and safety.

Figure 1.

PRISMA Flow Diagram.

Study Characteristics

The network plot for main and subgroup results is presented in Figure 2. The main network included 24 RCTs, 5 non-RCTs, and 7 CHs, and a total of 3231 patients. A gap of evidence was found between the Hybrid and nwaRepair groups in the main and subgroup networks. A total of 1587, 267, 650, 362, 201, 164 patients were included in the Autograft, irAllograft, nirAllograft, Hybrid, waRepair, and nwaRepair groups, respectively, for the main network, and 1106, 267, 183, 179, and 59 patients were included in the Autograft, irAllograft, nirAllograft, Hybrid, waRepair, and nwaRepair groups for the subgroup network. The average age was 28.12±5.04 (years, mean±SD), and the proportion of male patients was 64% (range 30–87%) (Table 1).

Figure 2.

Structure of network formed by interventions. The lines between treatment nodes indicate the direct comparisons made with evidences, the size of nodes indicate the number of participants involved in each treatment. Numbers (n/n) with a blue frame near the line indicate ‘number of trials/number of participants’ of the related treatment group, numbers (n/n) with a red frame near the line indicate ‘number of trials/number of participants’ of the related comparisons. (A) Main network meta-analysis. (B) Subgroup analysis. (Made with Stata/MP, version 14.0, manufacturer Stata Corp.).

Table 1.

Baseline characteristics of included studies.

Author	No.	Year	Number of patients	Mean age	Male/Female	Traumatic mechanism	Surgery method	Mean follow-up (months)	Study design	Control intervention I	Control intervention II	Control intervention III
Gagliardi AG et al [31]	1	2019	179	15.48	87/92	Low-energy trauma 164 High-energy trauma 0	Arthroscopy	36.0	Cohort study	Autograft	Repair with augmentation	/
Meunier A et al [39]	2	2007	100	21.44	68/32	NR	Arthroscopy	180.0	Randomized clinical trial	Repair with augmentation	Repair without augmentation	Autograft
Hoogeslag RAG et al [19]	3	2019	48	21.50	13820	NR	Arthroscopy	24.0	Randomized clinical trial	Repair with augmentation	Autograft	/
Achtnich A et al [27]	4	2016	40	31.80	NR	NR	Arthroscopy	28.0	Non-randomized clinical trial	Repair without augmentation	Autograft	/
Bottoni CR et al [29]	5	2015	97	29.05	84/13	Low-energy trauma 98 High-energy trauma 2	Arthroscopy	120.0	Randomized clinical trial	Autograft	Non-irradiated allograft	/
Sporsheim AN et al [44]	6	2019	150	29.00	36/28	NR	Arthroscopy	360.0	Randomized clinical trial	Repair without augmentation	Repair with augmentation	Autograft
Sun K et al [46]	7	2009	99	30.55	70/29	Low-energy trauma 88 High-energy trauma 10	Arthroscopy	31.0	Randomized clinical trial	Autograft	Non-irradiated allograft	Irradiated allograft
Murray MM et al [41]	8	2020	100	17.00	44/56	NR	Arthroscopy	24.0	Randomized clinical trial	Repair without augmentation	Autograft	/
Schliemann B et al [43]	9	2017	60	28.65	23/37	NR	Arthroscopy	12.0	Randomized clinical trial	Repair with augmentation	Autograft	/
Wang HD et al [52]	10	2017	57	32.70	19/38	NR	Arthroscopy	36.0	Cohort study	Hybrid	Autograft	/
Murray MM et al [40]	11	2016	20	24.35	6/14	Low-energy trauma 19 High-energy trauma 1	Arthroscopy	3.0	Cohort study	Repair without augmentation	Autograft	/
Edgar CM et al [30]	12	2008	83	29.22	46/37	NR	Arthroscopy	50.0	Cohort study	Autograft	Non-irradiated allograft	/
Li J et al [37]	13	2016	80	31.37	50/30	NR	Arthroscopy	67.2	Randomized clinical trial	Autograft	Irradiated allograft	Hybrid
Kösters C et al [34]	14	2020	85	28.16	56/29	NR	Arthroscopy	24.0	Randomized clinical trial	Repair with augmentation	Autograft	/
Sun K et al [60]	15	2009	156	32.26	124/32	Low-energy trauma 154 High-energy trauma 2	Arthroscopy	67.2	Randomized clinical trial	Autograft	Non-irradiated allograft	/
Sun K et al [47]	16	2009	65	24.89	17/48	Low-energy trauma 61 High-energy trauma 4	Arthroscopy	31.0	Randomized clinical trial	Autograft	Irradiated allograft	/
Sun K et al [48]	17	2011	67	30.60	15/52	Low-energy trauma 61 High-energy trauma 4	Arthroscopy	42.2	Randomized clinical trial	Autograft	Irradiated allograft	/
Sun K et al [58]	18	2011	186	30.42	149/37	Low-energy trauma 61 High-energy trauma 8	Arthroscopy	93.0	Randomized clinical trial	Autograft	Non-irradiated allograft	/
Jia YH et al [33]	19	2015	106	29.50	54/52	NR	Arthroscopy	81.0	Randomized clinical trial	Non-irradiated allograft	Autograft	/
Tian S et al [50]	20	2016	121	30.21	96/25	Low-energy trauma 113 High-energy trauma 8	Arthroscopy	55.2	Randomized clinical trial	Autograft	Non-irradiated allograft	/
Tian S et al [49]	21	2016	83	28.89	66/17	Low-energy trauma 77 High-energy trauma 6	Arthroscopy	82.8	Randomized clinical trial	Autograft	Irradiated allograft	/
Yoo SH et al [54]	22	2015	132	27.09	120/12	NR	Arthroscopy	33.6	Randomized clinical trial	Autograft	Non-irradiated allograft	/
Tian S et al [51]	23	2010	69	31.25	56/13	Low-energy trauma 60 High-energy trauma 9	Arthroscopy	38.5	Randomized clinical trial	Autograft	Irradiated allograft	/
Bi HY et al [28]	24	2013	86	32.00	60/26	NR	Arthroscopy	38.5	Randomized clinical trial	Autograft	Non-irradiated allograft	/
Sun K et al [45]	25	2004	53	32.34	44/9	NR	Arthroscopy	19.0	Randomized clinical trial	Autograft	Non-irradiated allograft	/
Gorschewsky O et al [32]	26	2002	201	32.97	150/51	NR	Arthroscopy	24.0	Randomized clinical trial	Autograft	Non-irradiated allograft	/
Li J et al [38]	27	2015	95	30.62	50/45	Low-energy trauma 87 High-energy trauma 8	Arthroscopy	71.2	Randomized clinical trial	Autograft	Irradiated allograft	Hybrid
Murray MM et al [18]	28	2019	20	24.35	6/14	Low-energy trauma 19 High-energy trauma 1	Arthroscopy	24.0	Cohort study	Repair without augmentation	Autograft	/
Darnley JE et al [56]	29	2016	54	20.90	34/20	NR	Arthroscopy	24.0	Cohort study	Autograft	Hybrid	/
Zheng X et al [55]	30	2019	97	30.77	69/28	NR	Arthroscopy	20.1	Randomized clinical trial	Autograft	Hybrid	Irradiated allograft
Kraeutler MJ et al [35]	31	2017	148	33.78	NR	NR	Arthroscopy	50.1	Randomized clinical trial	Autograft	Hybrid	/
Pennock AT et al [42]	32	2016	40	15.65	15/25	NR	Arthroscopy	24.0	Non-randomized clinical trial	Hybrid	Autograft	/
Leo BM et al [36]	33	2016	95	27.20	60/35	NR	Arthroscopy	24.0	Non-randomized clinical trial	Autograft	Hybrid	/
Xu H et al [59]	34	2018	76	35.66	52/24	Low-energy trauma 67 High-energy trauma 9	Arthroscopy	27.0	Cohort study	Autograft	Hybrid	/
Burrus MT et al [57]	35	2015	58	26.90	20/38	NR	Arthroscopy	46.2	Non-randomized clinical trial	Hybrid	Autograft	/
Xu H et al [53]	36	2017	68	33.90	43/25	NR	Arthroscopy	24.0	Non-randomized clinical trial	Autograft	Hybrid	/

The quality and bias-risk assessments of all studies are presented in Tables 2 and 3. Two main sources of bias were found. Considering the specificity of the surgical intervention in this study, blinding participants and surgeons was almost impossible; meanwhile, it was also difficult to conceal the allocations. It caused a relatively high and inevitable risk of performance bias. The funnel plots and Egger’s tests did not indicate publication bias in any network (Figure 3). Cluster-rank plots are presented in Figure 4. The league plots, which showed the relative effects between different groups, are presented in Tables 4–8.

Table 2.

Methodological quality and risk of bias evaluation of randomized controlled studies.

Author	No.	1. Sequence generation	2. Allocation concealment	3. Blinding	4. Incomplete outcome data	5. Selective outcome reporting	6. Other source of bias
Meunier A et al [39]	2	H	U	U	L	L	L
Hoogeslag RAG et al [19]	3	L	U	U	L	L	L
Bottoni CR et al [29]	5	L	H	L	L	L	L
Sporsheim AN et al [44]	6	L	H	L	L	L	L
Sun K et al [46]	7	L	H	U	L	L	L
Murray MM et al [41]	8	L	H	U	L	L	L
Schliemann B et al [43]	9	L	U	L	L	L	L
Li J et al [37]	13	L	U	U	L	L	L
Kösters C et al [34]	14	L	L	U	L	L	L
Sun K et al [60]	15	L	U	U	L	L	L
Sun K et al [47]	16	L	U	U	L	L	L
Sun K et al [48]	17	L	U	H	L	L	L
Sun K et al [58]	18	L	U	H	L	L	L
Jia YH et al [33]	19	L	L	H	L	L	L
Tian S et al [50]	20	L	H	H	L	L	L
Tian S et al [49]	21	L	H	H	L	L	L
Yoo SH et al [54]	22	L	L	L	L	L	U
Tian S et al [51]	23	L	U	U	L	L	U
Bi HY et al [28]	24	L	H	H	L	L	L
Sun K et al [45]	25	L	H	H	L	L	L
Gorschewsky O et al [32]	26	L	L	U	L	L	L
Li J et al [38]	27	L	U	U	L	L	L
Zheng X et al [55]	30	L	H	H	L	L	L
Kraeutler MJ et al [35]	31	L	H	H	L	L	L

L – low risk of bias; U – unclear risk of bias; H – high risk of bias.

Table 3.

Methodological quality and risk of bias evaluation of non-randomized control and cohort design studies.

Author	No.	Selection				Comparability	Outcome
		Representativeness of the exposed cohort	Selection of the non-exposed cohort	Ascertainment of exposure	Demonstration that outcome of interest was not present at start of study	Comparability of cohorts on the basis of the design or analysis	Assessment of outcome	Was follow-up long enough for outcomes to occur	Adequacy of follow up of cohorts
Gagliardi AG et al [31]	1	1	1	1	0	2	1	1	1
Achtnich A et al [27]	4	1	1	1	1	1	1	1	1
Wang HD et al [52]	10	1	1	1	1	2	1	1	1
Murray MM et al [40]	11	1	1	1	1	2	1	0	0
Edgar CM et al [30]	12	1	1	1	1	2	1	1	1
Murray MM et al [18]	28	1	1	1	1	2	1	1	1
Darnley JE et al [56]	29	1	1	1	1	1	1	1	1
Pennock AT et al [42]	32	1	1	1	1	1	0	1	1
Leo BM et al [36]	33	1	1	1	1	2	0	1	1
Xu H et al [59]	34	1	1	1	1	2	1	1	1
Burrus MT et al [57]	35	1	1	1	1	2	0	1	1
Xu H et al [53]	36	1	1	1	1	1	1	1	1

Figure 3.

Publication bias and Egger’s test for main networks. (A) Subjective improvement. (B) Functional improvement. (C) Activity recovery. (D) Postoperative laxity. (E) Failure rate.

Figure 4.

Cluster-rank plots. (A) The cluster-rank plot of subjective improvement and failure rate from main network analyses. (B) The cluster-rank plot of functional improvement and failure rate from main network analyses. (C) The cluster-rank plot of activity recovery and failure rate from main network analyses. (D) The cluster-rank plot of subjective improvement and failure rate from subgroup analyses. (E) The cluster-rank plot of functional improvement and failure rate from subgroup analyses. (F) The cluster-rank plot of activity recovery and failure rate from subgroup analyses. (The cluster-rank value is the product of the abscissa and ordinate of each treatment.)

Table 4.

The league plots of subjective improvement. Main network analysis (red) and subgroup analysis (blue). (From the top left to the bottom right, higher comparator vs lower comparator, WMD with 95% CI).

nwRepair	−2.53 (−9.40, 4.34)	−3.58 (−9.92, 2.76)	−4.10 (−10.17, 1.97)	−4.72 (−11.06, 1.62)	−7.93 (−14.26, −1.59)
0.76 (−4.76, 6.28)	waRepair	−1.05 (−4.70, 2.60)	−1.57 (−4.78, 1.64)	−2.19 (−5.89, 1.51)	−5.40 (−9.08, −1.72)
0.69 (−4.30, 5.68)	−0.08 (−3.55, 3.40)	nirAllograft	−0.52 (−2.34, 1.31)	−1.14 (−3.70, 1.42)	−4.35 (−6.82, −1.87)
1.01 (−3.63, 5.66)	0.25 (−2.75, 3.26)	0.33 (−1.52, 2.18)	Autograft	−0.62 (−2.47, 1.23)	−3.83 (−5.64, −2.01)
1.33 (−3.55, 6.20)	0.56 (−2.75, 3.88)	0.64 (−1.68, 2.96)	0.31 (−1.20, 1.82)	Hybrid	−3.21 (−5.38, −1.04)
4.77 (−0.23, 9.78)	4.01 (0.49, 7.53)	4.09 (1.56, 6.62)	3.76 (1.88, 5.64)	3.45 (1.34, 5.56)	irAllograft

Table 5.

The league plots of functional improvement. Main network analysis (red) and subgroup analysis (blue). (From the top left to the bottom right, higher comparator vs lower comparator, SMD with 95% CI).

nwaRepair	−0.95 (−2.05, 0.16)	−0.95 (−2.05, 0.15)	−1.12 (−2.33, 0.09)	−1.25 (−2.54, 0.05)	−1.35 (−2.58, −0.11)
0.76 (−0.09, 1.60)	Autograft	−0.01 (−0.74, 0.72)	−0.18 (−0.67, 0.32)	−0.30 (−0.98, 0.38)	−0.40 (−0.95, 0.15)
0.80 (−0.09, 1.69)	0.05 (−0.54, 0.63)	waRepair	−0.17 (−1.05, 0.71)	−0.29 (−1.29, 0.71)	−0.40 (−1.31, 0.52)
0.90 (−0.04, 1.84)	0.14 (−0.28, 0.56)	0.10 (−0.62, 0.82)	nirAllograft	−0.13 (−0.96, 0.71)	−0.23 (−0.94, 0.49)
1.01 (0.06, 1.96)	0.25 (−0.19, 0.70)	0.21 (−0.53, 0.94)	0.11 (−0.50, 0.72)	Hybrid	−0.10 (−0.85, 0.65)
1.14 (0.17, 2.12)	0.39 (−0.09, 0.87)	0.34 (−0.42, 1.10)	0.25 (−0.37, 0.87)	0.14 (−0.45, 0.72)	irAllograft

Table 6.

The league plots of activity recovery. Main network analysis (red) and subgroup analysis (blue). (From the top left to the bottom right, higher comparator vs lower comparator, WMD with 95% CI).

waRepair	−0.20 (−0.75, 0.34)	−0.22 (−0.54, 0.11)	−0.30 (−0.73, 0.12)	−0.38 (−0.76, −0.01)	−0.77 (−1.18, −0.36)
0.23 (−0.31, 0.78)	nwaRepair	−0.02 (−0.58, 0.55)	−0.10 (−0.74, 0.53)	−0.18 (−0.78, 0.41)	−0.57 (−1.19, 0.05)
0.30 (0.01, 0.60)	0.07 (−0.49, 0.63)	Autograft	−0.09 (−0.36, 0.19)	−0.17 (−0.34, 0.01)	−0.55 (−0.80, −0.31)
0.31 (−0.06, 0.69)	0.08 (−0.53, 0.68)	0.01 (−0.22, 0.23)	Hybrid	−0.08 (−0.40, 0.25)	−0.47 (−0.80, −0.13)
0.45 (0.10, 0.79)	0.21 (−0.37, 0.80)	0.14 (−0.02, 0.31)	0.13 (−0.15, 0.41)	nirAllograft	−0.39 (−0.68, −0.09)
0.84 (0.46, 1.23)	0.61 (0.00, 1.22)	0.54 (0.30, 0.78)	0.53 (0.23, 0.84)	0.40 (0.11, 0.69)	irAllograft

Table 7.

The league plots of postoperative laxity. Main network analysis (red) and subgroup analysis (blue). (From the top left to the bottom right, higher comparator vs lower comparator, SMD with 95% CI).

Autograft	−0.29 (−1.61, 1.04)	0.03 (−0.90, 0.96)	0.31 (−0.31, 0.93)	0.81 (0.07, 1.55)	2.32 (1.74, 2.89)
−0.05 (−0.46, 0.36)	nwaRepair	0.32 (−1.12, 1.75)	0.60 (−0.86, 2.07)	1.10 (−0.42, 2.61)	2.60 (1.15, 4.06)
−0.14 (−0.55, 0.27)	−0.09 (−0.60, 0.43)	waRepair	0.29 (−0.84, 1.41)	0.78 (−0.41, 1.97)	2.29 (1.18, 3.40)
−0.22 (−0.56, 0.12)	−0.17 (−0.70, 0.36)	−0.08 (−0.61, 0.45)	nirAllograft	0.49 (−0.46, 1.45)	2.00 (1.18, 2.82)
−0.33 (−0.70, 0.04)	−0.28 (−0.83, 0.28)	−0.19 (−0.75, 0.36)	−0.11 (−0.61, 0.39)	Hybrid	1.51 (0.73, 2.28)
−1.41 (−1.74, −1.08)	−1.36 (−1.88, −0.83)	−1.27 (−1.80, −0.74)	−1.19 (−1.64, −0.73)	−1.08 (−1.51, −0.65)	irAllograft

Table 8.

The league plots of failure rate. Main network analysis (red) and subgroup analysis (blue). (From the top left to the bottom right, higher comparator vs lower comparator, RR with 95% CI).

Autograft	1.37 (0.74, 2.52)	1.06 (0.41, 2.72)	0.96 (0.43, 2.14)	1.94 (0.79, 4.78)	3.52 (1.83, 6.76)
0.85 (0.38, 1.91)	nirAllograft	0.77 (0.25, 2.35)	0.70 (0.26, 1.91)	1.42 (0.47, 4.30)	2.58 (1.17, 5.69)
0.60 (0.26, 1.39)	0.70 (0.22, 2.24)	Hybrid	0.91 (0.26, 3.15)	1.84 (0.51, 6.68)	3.33 (1.16, 9.53)
0.47 (0.19, 1.19)	0.56 (0.16, 1.89)	0.79 (0.22, 2.79)	waRepair	2.02 (0.86, 4.71)	3.66 (1.32, 10.13)
0.29 (0.10, 0.85)	0.35 (0.09, 1.32)	0.49 (0.13, 1.92)	0.62 (0.20, 1.98)	nwaRepair	1.81 (0.58, 5.70)
0.30 (0.13, 0.73)	0.36 (0.12, 1.07)	0.51 (0.17, 1.55)	0.64 (0.18, 2.27)	1.03 (0.26, 4.07)	irAllograft

Network Meta-Analysis

Subjective Evaluation Improvement

There were 28 trials with 2403 patients included in the final analysis. No inconsistency was detected and the consistency model was used.

The SUCRA results showed that the nwaRepair group had the highest postoperative subjective evaluation improvement (SUCRA=70.9%), followed by nirAllograft (SUCRA=65.2%) and waRepair (SUCRA=61.8%), while the lowest was irAllograft (SUCRA=1.0%). Paradoxically, all groups except nwaRepair (WMD 4.77, 95% CI [−0.23 to 9.78]), were significantly superior to irAllograft. The interpretation of the results should be done cautiously.

Functional Improvement

A total of 32 trials with 2976 patients were included in this network. No significant inconsistency was detected and the consistency model was used.

The nwaRepair group had the highest probability of having the best postoperative functional improvement (SMD 1.14, 95%CI [0.17 to 2.12] SUCRA=97.2%), followed by Autograft (SMD 0.39, 95%CI [−0.09 to 0.87], SUCRA=62.9%) and waRepair (SMD 0.34, 95%CI [−0.42 to 1.10], SUCRA=52.3%), with nirAllograft being the lowest (SUCRA=16.0%). Based on the SMD, only nwaRepair was significantly better than nirAllograft.

Activity Recovery

A total of 25 trials with 2330 patients were included in this network. No significant in consistency was detected, and the consistency model was used. All groups were significantly better than the nirAllograft group for postoperative activity status. Based on the SUCRA ranking, the best groups for activity recovery were waRepair (WMD 0.84, 95%CI [0.46 to 1.23], SUCRA=94.5%), nwaRepair (WMD 0.61, 95%CI [0.00 to 1.22], SUCRA=62.5%), and Autograft (WMD 0.54, 95%CI [0.30 to 0.78], SUCRA=58.5%).

Safety Outcomes

A total of 26 trials (2241 patients) reporting the degree of postoperative knee laxity and 28 trials (2727 patients) reporting the failure rate were assessed in these 2 networks, respectively. No significant inconsistencies were detected, and the consistency model was used for both networks.

All groups had significantly less joint laxity than the irAllograft group. The Autograft group had the lowest degree of laxity (SMD −1.41, 95%CI [−1.74 to −1.08], SUCRA=84.0%), followed by the nwaRepair group (SMD −1.36, 95%CI [−1.88 to −0.83], SUCRA=72.1%) and waRepair group (SMD −1.27, 95%CI [−1.80 to −0.74], SUCRA=60.2%). There was no significant difference among the Autograft, nwaRepair, repair, nirAllograft, and Hybrid groups.

The Autograft group also had the lowest rate of failure (RR 0.30, 95%CI [0.13 to 0.73], SUCRA=89.5%), followed by nirAllograft (RR 0.36, 95%CI [0.12 to 1.07], SUCRA=76.0%) and Hybrid graft (RR 0.51, 95%CI [0.17 to 1.55], SUCRA=54.7%). No significant difference was shown between nwaRepair and irAllograft (RR 1.03, 95%CI [0.26 to 4.07], SUCRA=19.1%). It is worth noting that nwaRepair had significantly higher failure rates than Autograft (RR 3.40, 95%CI [1.18 to 9.82]). Detailed SUCRA values of the main network analyses are presented in Table 9, and the forest plots are presented in Figures 5 and 6.

Table 9.

Detailed results of main network analysis.

Treatment	SMD (95% CI) for subjective improvement	SURCA for subjective improvement, %	SMD (95%CI) for Functional improvement	SURCA for functional improvement, %	SMD (95% CI) for activity improvement	SURCA for activity improvement, %	SMD (95% CI) for laxity	SURCA for laxity, %	RR (95% CI) for failure	SURCA for failure, %
irAllograft	Reference	1.0	Reference	16.0	Reference	0.5	Reference	0.0	Reference	18.4
Autograft	3.76 (1.88, 5.64)	55.0	0.39 (−0.09, 0.87)	62.9	0.54 (0.30, 0.78)	58.5	−1.41 (−1.74, −1.08)	84.0	0.30 (0.13, 0.73)	89.5
nirAllograft	4.09 (1.56, 6.62)	65.2	0.25 (−0.37, 0.87)	42.1	0.40 (0.11, 0.69)	29.5	−1.19 (−1.64, −0.73)	47.8	0.36 (0.12, 1.07)	76.0
Hybrid	3.45 (1.34, 5.56)	45.7	0.14 (−0.45, 0.72)	29.5	0.53 (0.23, 0.84)	54.5	−1.08 (−1.51, −0.65)	35.9	0.51 (0.17, 1.55)	54.7
waRepair	4.01 (0.49, 7.53)	61.8	0.34 (−0.42, 1.10)	52.3	0.84 (0.46, 1.23)	94.5	−1.27 (−1.80, −0.74)	60.2	0.64 (0.18, 2.27)	42.4
nwaRepair	4.77 (−0.23, 9.78)	70.9	1.14 (0.17, 2.12)	97.2	0.61 (0.00, 1.22)	62.5	−1.36 (−1.88, −0.83)	72.1	1.03 (0.26, 4.07)	19.1

Figure 5.

Forest plots of network comparisons of main network meta-analysis (made with Microsoft Excel, version 2019MSO).

Figure 6.

Forest plots incorporated direct comparisons and indirect comparisons of main network meta-analysis. (A) Subjective improvement. (B) Functional improvement. (C) Activity recovery. (D) Postoperative laxity. (E) Failure rate. (A or 1: Autograft; B or 2: Allograft with irradiation; C or 3: Allograft without irradiation; D or 4: Hybrid graft; E or 5: Repair with augmentation; F or 6: Repair without augmentation).

Subgroup Analysis of RCTs

All 24 RCTs (2541 patients) were included in this subgroup analysis. No inconsistencies were reported, and the consistency model was used for all outcomes. The only difference from the full analysis was in activity recovery. In the RCTs subgroup analysis, no significant difference was shown between nwaRepair and irAllograft (WMD 0.57, 95%CI [−0.05 to 1.19]) in activity recovery (Table 10). Forest plots for results of subgroup analysis are shown in Figures 7 and 8.

Table 10.

Detailed results of subgroup analysis.

Treatment	SMD (95% CI) for subjective improvement	SURCA for subjective improvement, %	SMD (95%CI) for functional improvement	SURCA for functional improvement, %	SMD (95% CI) for activity improvement	SURCA for activity improvement, %	SMD (95% CI) for laxity	SURCA for laxity, %	RR (95% CI) for failure	SURCA for failure, %
irAllograft	Reference	0.3	Reference	19.1	Reference	0.9	Reference	0.0	Reference	3.8
Autograft	3.83 (2.01, 5.64)	45.6	0.40 (−0.15, 0.95)	60.5	0.55 (0.31, 0.80)	65.6	−2.32 (−2.89, −1.74)	76.8	0.28 (0.15, 0.55)	75.6
nirAllograft	4.35 (1.87, 6.82)	59.3	0.23 (−0.49, 0.94)	40.0	0.39 (0.09, 0.68)	32.8	−2.00 (−2.82, −1.18)	47.2	0.39 (0.18, 0.86)	48.8
Hybrid	3.21 (1.04, 5.38)	32.7	0.10 (−0.65, 0.85)	29.6	0.47 (0.13, 0.80)	47.9	−1.51 (−2.28, −0.73)	29.9	0.30 (0.10, 0.86)	68.2
waRepair	5.40 (1.72, 9.08)	73.0	0.40 (−0.52, 1.31)	54.2	0.77 (0.36, 1.18)	91.1	−2.29 (−3.40, −1.18)	66.5	0.27 (0.10, 0.76)	75.7
nwaRepair	7.93 (1.59, 14.26)	89.1	1.35 (0.11, 2.58)	96.7	0.57 (−0.05, 1.19)	61.7	−2.60 (−4.06, −1.15)	79.5	0.55 (0.18, 1.73)	27.8

Figure 7.

Forest plots of network comparisons of subgroup network meta-analysis (made with Microsoft Excel, version 2019MSO).

Figure 8.

Forest plots incorporated direct comparisons and indirect comparisons of subgroup analysis. (A) Subjective improvement. (B) Functional improvement. (C) Activity recovery. (D) Postoperative laxity. (E) Failure rate. (A or 1: Autograft; B or 2: Allograft with irradiation; C or 3: Allograft without irradiation; D or 4: Hybrid graft; E or 5: Repair with augmentation; F or 6: Repair without augmentation.).

Direct Pairwise Meta-Analysis

Direct pairwise comparisons of the postoperative efficacy and safety of PR and ACLR showed significant heterogeneity in all outcomes. Therefore, based on a pre-analysis, the direct comparisons were separated into 2: waRepair vs ACLR, and nwaRepair vs ACLR (Tables 11, 12). Sensitivity analysis was used to minimize heterogeneity.

Table 11.

The detailed results of direct pair-wise meta-analyses between primary repair (PR) with augmentation and anterior cruciate ligament reconstruction (ACLR).

Comparison (PR with augmentation vs ACLR)	No. of trials	No. of patients	Heterogeneity, I2	Effect index	Effect size
Subjective improvement	3	193	4.50%	WMD (95% CI)	0.157 (−0.127 to 0.440)
Functional improvement	4	335	0.00%	SMD (95% CI)	−0.042 (−0.302 to 0.218)
Activity improvement	5	461	67.50%	SMD (95% CI)	0.278 (0.068 to 0.488)
Laxity	5	461	0.00%	WMD (95% CI)	0.18 (−0.081 to 0.442)
Failure	3	200	0.00%	RR (95% CI)	0.535 (0.221 to 1.296)

Table 12.

The detailed results of direct pair-wise meta-analyses between primary repair (PR) without augmentation and anterior cruciate ligament reconstruction (ACLR).

Comparison (PR without augmentation vs ACLR)	No. of trials	No. of patients	Heterogeneity, I2	Effect index	Effect size
Subjective improvement	3	140	58.80%	WMD (95% CI)	0.074 (−0.574 to 0.722)
Functional improvement	3	146	92.30%	SMD (95% CI)	0.195 (−0.437 to 0.827)
Activity improvement	2	126	0.00%	WMD (95% CI)	−0.109 (−0.461 to 0.242)
Laxity	6	306	36.50%	SMD (95% CI)	−0.03 (−0.341 to 0.280)
Failure	4	243	43.10%	RR (95% CI)	1.638 (0.658 to 4.078)

Direct comparison of waRepair with ACLR showed significant heterogeneity in activity improvement (I²=67.5%) and consequently, a random-effects model was used for this outcome, while a fixed-effects model was used for other outcomes. The only significant difference between waRepair and ACLR was in activity improvement outcome (WMD 0.28 95%CI [0.07 to 0.49]). Considering the heterogeneity, this result should be viewed with caution.

Direct comparison of nwaRepair with ACLR showed significant heterogeneity in subjective evaluation improvement (I²=58.80%) and functional improvement (I²=92.30%), and a random-effects model was used for these 2 outcomes. No significant differences were found between nwaPR and ACLR.

Discussion

This is the first network meta based on high-quality studies to compare the functional recovery and adverse effects of PR with ACLR in the surgical treatment of ACL rupture. The important observations from this analysis are: (1) waRepair and nwaRepair ranked the best for postoperative efficacy in activity recovery and subjective and functional improvement, while ACLR with irAllograft was a poor option for the surgical treatment of ACL rupture, with the weakest efficacy and the worst safety profile; (2) ACLR with allograft without irradiation produced a similar improvement in subjective evaluation improvement when compared with PR, but less functional improvement and activity recovery; (3) PR produced less postoperative knee laxity than irradiated allograft ACLR but had a higher failure rate than ACLR with Autograft, if without an augmentation. This suggests that PR may have other potentially serious complications that could necessitate revision surgery; (4) ACLR with autograft and hybrid graft yielded good results for efficacy and safety, and both were good choices for surgery; and (5) ACLR with autograft was the safest and most stable surgery method according to the results of postoperative knee laxity and revision rate and the cluster-rank analysis.

In a study by Sun et al [46], autograft and non-irradiated allograft ACLR had comparable outcomes for postoperative symptoms, functional improvement, and activity level, and both had a lower incidence of graft failure than irradiated allograft ACLR. Separately, Curran et al [61] studied the effects of irradiated allograft ACLR and showed that the low dose of irradiation could weaken the strength and stiffness of the allograft, result in altered graft function, and affect the clinical outcomes of ACLR. Our findings are consistence with these results. In addition, the network meta-analysis by Yang et al [62] compared the long-term outcomes of different grafts in ACLR and recommended double-bundle hamstring autograft as the best choice for its good prognosis. Another study [63] compared different tendon grafts for ACLR and also revealed that autografts (especially quadricep tendon autografts) rather than artificial ligaments were suitable for primary ACLR because artificial ligaments could increase the risk of knee laxity. Compared with the previous studies, the present study pointed out that PR could be an ideal surgical method in terms of efficacy but is related to a significantly higher revision risk, and autograft ACLR may be the optimal strategy for surgical treatment of ACLR.

In this analysis, the PR technique showed similar and even better clinical outcomes than autograft and hybrid graft ACLR. This most likely resulted from the lack of harvesting of the graft tissue and avoiding donor site complications, leading to better activity recovery and subjective and functional improvements. However, it should be noted that while PR was shown to produce less knee laxity, it was found to lead to a relatively high graft failure rate compared with that of autograft and non-irradiated allograft. One explanation for this is that after PR surgery the scar formed for healing of the ruptured ACL leads to a contracture of ACL or/and limited knee movement (Hoogeslag et al [19]).

There were still several limitations in this study. First, some factors can affect outcomes but cannot be adjusted or removed by statistical methods, such as the skill of the surgeons and the quality of the postoperative rehabilitation. Second, an RCT is more sensitive to complications, with a high incidence and a short period, while observational studies can more effectively assess the complications with a low incidence and a long period. Given the important role of observational studies, such as cohort studies, in exploring the long-term efficacy and safety of ACL rupture surgery and although CHSs and non-RCTs potentially have confounding factors that can bias the results, we still included them in the analysis and used a subgroup analysis to examine the impact of observational studies on the results. The subgroup analysis on the RCTs showed only 1 difference from the main network analysis, suggesting that any biases from CHSs and non-RCTs were unlikely to be a factor in this analysis. Third, studies with no events in both treatment arms were inevitably included in the failure rate network. Omitting studies with rare events was recommended by the Cochrane Handbook, but this is still controversial as it can alter the biased evaluations and the accuracy of the combined estimation. Therefore, we included such trials and used a 0.5 zero-cell correction. However, the results of the failure rates should be interpreted with caution. Fourth, while funnel plots and Egger’s tests showed no significant publication bias or small study effects, only a limited number of trials were included, and therefore more high-quality trials are warranted.

Conclusions

For surgical treatment of ACLR, irradiated allograft ACLR had the worst efficacy and safety and is not recommended. PR may be an ideal treatment method in terms of efficacy but it is related to a significantly higher revision risk if done without augmentation. Autograft ACLR may be the optimal method currently available for most patients requiring surgical treatment of ACL rupture.

Abbreviations

ACL

anterior cruciate ligament

ACLR

ACL reconstruction

Autograft

autograft reconstruction

CHS

cohort study

Hybrid

hybrid graft reconstruction

IKDC

International Knee Documentation Committee

irAllograft

allograft with irradiation reconstruction

KOOS

Knee Injury and Osteoarthritis Score

nirAllograft

allograft without irradiation reconstruction

NOS

Newcastle-Ottawa Scale

nwaRepair

primary repair without augmentation

PR

primary repair

PRISMA-NMA

Systematic Reviews and Meta-Analyses for Network Meta-analysis

RCTs

randomized controlled trials

RR

risk ratio

SMD

standardized mean difference

SUCRA

surface under the cumulative ranking

waRepair

primary repair with augmentation

WMD

weighted mean difference

WOMAC

Western Ontario and McMaster Universities Arthritis Index