A systematic review comparing the results of early vs delayed ligament surgeries in single anterior cruciate ligament and multiligament knee injuries

Purpose The purpose of this study was to compare clinical outcomes and incidence of concomitant injuries in patients undergoing early vs delayed surgical treatment of single anterior cruciate ligament (ACL) injury and multiligament knee injury (MLKI). Methods A literature search using PubMed, Embase, the Cochrane Library, the Cumulative Index to Nursing and Allied Health, and Scopus from their inception to April 30, 2020 was conducted. Studies with levels I to IV evidence reporting the incidence of meniscus or cartilage injury according to early vs delayed surgery in single ACL injuries and MLKIs were included. In the meta-analysis, data based on the number of meniscus and cartilage injuries were extracted and pooled. Lysholm and Tegner scores were analyzed using two-sample Z-tests to calculate the non-weighted mean difference (NMD). A meta-regression analysis was also performed to determine the effect of single ACL injury and MLKI/study design. Results Sixteen studies on single ACL injury and 14 studies on MLKI were included in this analysis. In the analysis, there were significant decreases in Lysholm score (NMD − 5.3 [95% confidence interval (CI) − 7.37 to − 3.23]) and Tegner score (NMD − 0.25 [95% CI − 0.45 to − 0.05]) and increases in risk of meniscus tear (odds ratio [OR] 1.73 [95% CI 1.1–2.73], p = 0.01) and cartilage injury (OR 2.48 [95% CI 1.46–4.2], p = 0.0007) in the delayed surgery group regardless of single ACL injury or MLKI. The result of the meta-regression analysis indicated that single ACL injury and MLKI/study design were not significant moderators of overall heterogeneity (p > 0.05). Conclusions Our study suggests that delayed ACL surgery significantly resulted in a higher risk of meniscus tear and cartilage injury and decreased Lysholm and Tegner scores compared to early ACL surgery. The Lysholm scores in the delayed MLKI surgery group were significantly decreased, but the risks of meniscus tear and cartilage injury in the delayed MLKI surgery group remained unclear. Level of evidence Level III, meta-analysis.


Introduction
Treatment of multiligament knee injury (MLKI) is still challenging for orthopedic surgeons because of its complexity and severity, and it might be more common than previously reported [1][2][3][4][5][6]. The incidence of MLKI might be underestimated because of spontaneous reductions and missed diagnoses due to combined injuries, such as vascular or nerve injury [4][5][6][7]. In cases of vascular emergencies, immediate vascular repair is necessary; thus, temporal stabilization is usually applied using an external fixator [2,5].
However, in patients without any emergencies, the timing of surgery is still controversial [4,[8][9][10]. Early surgical reconstruction was advocated in some previous studies [4,6,8], but others reported that early surgery resulted in stiffness, arthrofibrosis, and a reduced rate of return to work [1,2,6,9,11,12]. Even in patients with an anterior cruciate ligament (ACL) injury, the timing of surgery is also controversial [13][14][15][16][17][18][19][20]. Early ACL reconstruction is likely recommended [17,[21][22][23] because there is an increased risk of meniscus and cartilage injury if it is delayed [21,22]. In a study of Norwegian National Registry data, the odds of a cartilage lesion increased by nearly 1% for each month from the injury date, and the odds of cartilage lesions were nearly twice as frequent when combined with meniscal tear [24]. However, there were also studies which reported no differences between early and delayed surgery [13,18,25,26]. Thus, there is still a lack of consensus regarding the timing of surgery, regardless of the type of injury (single ACL injuries or MLKI) [6,9,[16][17][18][19][20].
A paucity of evidence on which to base treatment decisions and the lack of consensus in ACL injuries further complicate the management of MLKI. Questions are established to determine if the decision-making parameters are similar for knees with a single ACL injury or those with MLKI and if they provide strong outcomes.
This systemic review including meta-analysis aimed to compare clinical outcomes and incidence of concomitant injuries in patients undergoing early vs delayed surgical treatment of single ACL injury and MLKI. We hypothesized that early surgery would result in better clinical outcomes and less incidence of concomitant injuries compared to delayed surgical treatment.

Protocol and registration
This systematic review and meta-analysis were conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and using the PRISMA checklist [27] and registered using the PROSPERO International prospective register of systematic reviews [28] (CRD42020145204).

Search strategy
A comprehensive literature search was conducted using several databases (PubMed, Embase, the Cochrane Library, the Cumulative Index to Nursing and Allied Health [CINAHL], and Scopus). In this study, the same included studies and some of the same extracted data were used in accordance with a previous meta-analysis comparing associated lesions in single ACL vs MLKIs. The date was restricted to all publications from the inception of these databases to April 30, 2020,  After combining the search results and removing duplicates, two authors independently screened the title and abstract for eligibility, and the agreement was assessed by kappa value. Subsequently, the same authors independently reviewed the full text of the selected studies. All references within the included studies were cross-referenced for inclusion if they were missed in the initial search. Systematic reviews and meta-analyses were excluded; however, their references were screened manually to find additional articles that were not identified in the first round. Disagreements were resolved by discussion between the two review authors or consultation with another author. noted, and (5) "multiligament" defined as disruption of at least two of four major knee ligaments. The exclusion criteria were as follows: (1) not in the English language; (2) case report, clinical opinion, or technical note; (3) emergency treatment in MLKI; and (4) concomitant fracture around the knee (Fig. 1).

Data extraction and collection
The data extraction sheet was performed based on the checklist designed by Spindler et al. [29] and the consensus of authors for variables that should be reported. The following data were extracted: (1) study type, (2) level of evidence, (3) main purpose of the study, (4) number of cases, (5) age, (6) sex, (7) combined ligament injury, (8) concomitant injuries including the meniscus, (9) cartilage injuries, (10) reported complications, (11) timing of ACL reconstruction, (12) follow-up, (13) clinical outcomes, and (14) other relevant findings including revision. The data on the timing of surgery were recorded according to the definition of each study because of the heterogeneity of the included studies. The clinical outcomes were recorded as Lysholm scores, Tegner scores, International Knee Documentation Committee (IKDC) scores, number of meniscus tears, and number of cartilage injuries. The extracted data were also cross-checked for accuracy; any disagreements were settled by discussion between the two review authors or by consultation with another author.
Grading of the quality of evidence The quality of evidence was evaluated by two authors using the guidelines of the Grading of Recommendations Assessment, Development and Evaluation (GRADE) working group [4]. The definitions of the grades of     evidence were as follows: (1) high, when further research is unlikely to change confidence in the estimate of the effect; (2) moderate, when further research is likely to have an important impact on the confidence in the estimate of the effect and may change it; (3) low, when further research is particularly likely to have an important impact on the confidence in the estimate of the effect and is likely to change it; and (4) very low, when any estimate of the effect is extremely uncertain [4]. Disagreements were resolved by discussion and assessed by kappa value.

Assessment of methodological quality
Two investigators independently assessed the methodological quality of each study using the Downs and Black quality assessment tool [30], which was developed for use in systemic reviews of both randomized and nonrandomized studies. This tool consists of 27 questions that assess the criteria for reporting, external validity, and internal validity (measurement and confounding). The highest possible score is 32. Disagreements were resolved by discussion and assessed by kappa value. For the additional graphical assessment of the risk of bias across the studies, ROBINS-I (Risk Of Bias In Nonrandomized Studies of Interventions) [31,32] was also used, which was released by the Cochrane Non-Randomized Study Group recently. The possibility of publication bias was examined by Egger's test based on Galbraith plots [33] with a random-effects model. Funnel plot asymmetry and Egger's tests were conducted to examine the possibility of publication bias. Moreover, the trim-and-fill method and calculation of a fail-safe number were also performed to evaluate the robustness of publication bias [34,35].

Statistical analysis
Data analyses were performed with Review Manager software (version 5.3; Nordic Cochrane Centre, the Cochrane Collaboration) and the R program (version 3.5.3, the R Foundation) using the "meta" and "metafor" packages. Statistical heterogeneity was assessed with I 2 statistics: I 2 > 50%, substantial heterogeneity; 20% < I 2 ≤ 50%, moderate heterogeneity; I 2 < 20%, low heterogeneity. A random-effects model was used to analyze the more robust results. Forest plots were used to show the outcome, pooled estimate of effect, and overall summary effect of each study. The treatment effects were measured by 95% confidence intervals (CIs) if the outcomes were measured on the same scale. Because of the high risk of bias due to the low level of evidence studies, the pooled mean differences were not used in clinical outcomes (Lysholm and Tegner scores). The mean synthesis and non-weighted mean differences (NMDs) of the bestevidence synthesis method were used instead to evaluate the clinical outcomes [36]. Comparisons between early and delayed values from each study were made using two-sample Z-tests using a p value < 0.05 (http://www. statskingdom.com/120MeanNormal2.html). The pooled odds ratio (OR) for the forest plot was also measured if the outcomes were collected as categorical data using the Mantel-Haenszel method. The heterogeneity of the  binary categorical data was also evaluated using the L'Abbé plot. The subgroup analysis was performed according to the single ACL injury and MLKI groups, and meta-regression analyses with a mixed-effects model were also performed to assess the effects of the potential moderators (single ACL injury vs MLKI/prospective vs retrospective studies) on the overall heterogeneity if significant heterogeneity was observed. The inter-rater reliability was assessed using kappa statistics (κ) to determine the degree of agreement in the study selection and risk assessment. Agreement was deemed fair (κ = 0.21-0.40), moderate (κ = 0.41-0.60), substantial (κ = 0.61-0.80), or almost perfect (κ = 0.81-1.00). In all analyses, a p value < 0.05 was considered significant with a two-sided tail.

Assessment of methodological quality
The results of the quality assessment in the included studies are shown in Table 2 (κ = 0.73, substantial agreement). The overall bar plot of the ROBINS-I tool is summarized in Fig. 2 and Additional file 1: Figure S1 (κ = 0.83, almost perfect agreement).
A funnel plot was used to evaluate the scores which could be obtained by weighted values (total meniscus tears and cartilage injuries). The funnel plot suggested a publication bias in the assessment of meniscus tear; the trim-and-fill method and calculation of fail-safe number were then performed to further assess the publication bias.
In the analysis of meniscus tear, evidence of asymmetry was observed (Fig. 3a), and this result was further supported by an analysis using Egger's test (p = 0.000). The adjusted funnel plot after the trim-and-fill method (Fig. 3b) indicated the absence of publication bias with eight added studies, but the observed outcome was changed to reinforce the direction of the outcome (before vs after trim and fill, OR 1.73 vs 3.42). Moreover, the fail-safe number was calculated by the Rosenthal approach [34] as 227 (p < 0.0001), which is a robust result for publication bias for this study. In the analysis of cartilage injury, there was no evidence of publication bias (p = 0.618, Additional file 2: Figure S2). The fail-safe number was calculated as 168 for the cartilage injury (p < 0.0001). The cartilage injury analysis showed a robust result for publication bias.

GRADE evidence quality of each outcome
The GRADE evidence quality of each outcome is presented in Table 3 (κ = 0.65, substantial agreement). Four outcomes were separately evaluated: one of very low quality and three of low quality. The overall results were found to have a trend of increased risk of meniscus tear and cartilage injury and decreased Lysholm and Tegner scores in the delayed surgery group. However, almost all outcomes had limitations in study design, imprecise data, and directness of the studies.
The result of the meta-regression analysis indicated that single ACL injury and MLKI/prospective and retrospective design were not significant moderators of overall heterogeneity (Fig. 4b, p = 0.255).

Risk of concomitant cartilage injury
The number of cartilage injuries according to the timing of surgery (early vs delayed) was described in 10 studies (1681 patients) regardless of single ACL injury or MLKI [25, 37-40, 43, 46-48, 57]. Only one study [57] on MLKI reported cartilage injury result according to the timing of surgery (early vs delayed); thus, the subgroup analysis was not performed. The overall risk of cartilage injury in the delayed surgery group was significantly higher than that in the early surgery group, but this showed substantial heterogeneity (OR 2.48 [95% CI 1.46-4.2], p = 0.0007; I 2 = 70%, p < 0.01), as did the L'Abbé plot (Fig. 5a, Additional file 4: Figure S4). The result of the meta-regression analysis indicated that prospective and retrospective design were not significant moderators of overall heterogeneity (Fig. 5b, p = 0.336).

Tegner score
The Tegner scores according to the timing of surgery (early vs delayed) were described in 9 studies (728 patients) on single ACL [23, 25, 38-40, 43, 44] and five studies (171 patients) on MLKI [30,33,51,55,59]. In the study by Herbst et al. [23], the researchers reported the results as separate groups according to the meniscus injury; thus, we analyzed the results as two different studies. The overall Tegner scores in the delayed surgery group were significantly decreased (early vs delayed, 5.4 ± 1.05 vs 5.1 ± 1.01; p < 0.001; NMD − 0.25 [95% CI − 0.45 to − 0.05]) compared to those in the early surgery group (Fig. 7). In the subgroup analysis according to the type of injury (single ACL injury and MLKI), the Tegner score of the delayed surgery single ACL injury group was significantly lower than that of the early surgery single ACL injury group (early vs delayed, 5.8 ± 0.97 vs 5.5 ± 0.97; p < 0.001; NMD − 0.3 [95% CI − 0.51 to − 0.15]), but the Tegner score in the MLKI group was not significant, regardless of the timing of surgery (early vs

Sensitivity analysis
We performed a meta-analysis with the results of meniscus tear and cartilage injury, after removing a single study, for the sensitivity analysis (Additional file 5: Figure S5 and Additional file 6: Figure S6). The results of the sensitivity analysis were similar to those of the initial analysis.

Discussion
The most important finding of this meta-analysis was that the delayed ligament surgery group was significantly found to have a higher risk of meniscus tear and cartilage injury and decreased Lysholm and Tegner scores compared to the early ligament surgery group. In single ACL injury, a high risk of meniscus tear and decreased Lysholm and Tegner scores were found in the delayed surgery group. In MLKI, only the Lysholm score was decreased in the delayed surgery group. However, the high risk of bias due to the low level of evidence studies was also affected by the results of clinical outcomes (Lysholm and Tegner scores), and the clinical relevances of these results are still questionable.
Several previous systemic reviews and meta-analyses reported that the timing of the ACL reconstruction would not affect the outcomes [34,44,45], but other meta-analyses with high levels of evidence reported similar but somewhat superior results for early ACL reconstruction compared to delayed surgery [25,60]. Recent studies, other than meta-analyses, reported that early ACL reconstruction showed better clinical results due to rapid restoration of stability and function and less risk of meniscus and cartilage injury than delayed ACL reconstruction [24,41,42,61,62]. In a recent meta-analysis of MLKI [4], early ligament reconstruction was recommended because of superior patient-reported and clinical outcomes. Of all patients undergoing early surgery, 31% had a normal or near-normal knee, compared to only 15% of patients undergoing delayed reconstruction [4]. However, Mook et al. [1] found worse outcomes in terms of stiffness, anterior stability, and clinical outcomes in the early surgery group.
The studies on MLKI are extremely heterogeneous, so the results might change according to the inclusion Fig. 5 a Forest plot of odds ratio with 95% confidence intervals in cartilage injury. The gray squares represent the results of each study. Ends of the horizontal bars represent 95% confidence intervals. Dark gray diamonds show the overall results of all studies. Random-effects models were used. b Plot of the meta-regression analysis indicated that prospective and retrospective design were not significant moderators of overall heterogeneity (p = 0.336) criteria. Nonetheless, all relevant articles were focused on the clinical or stability outcomes, not the incidence of meniscus tear and cartilage injury in early vs delayed surgery and the differences between single ACL injury and MLKI. Although Ferguson et al. [20] performed a meta-analysis on ACL injuries, including meniscus tear and cartilage injury, only six studies with various study designs were included. Hohmann et al. [11] performed a meta-analysis on MLKI but did not assess the outcomes of meniscus or cartilage injury and comparison with single ACL injury, which could be a control group with well-known results. Thus, it was useful to perform a meta-analysis and systemic review with studies including meniscus tear and cartilage injury results to assess the differences between early and delayed ligament reconstructions and the subgroup differences between MLKI and single ACL injury.
This systemic review including meta-analysis revealed that early ligament reconstruction could result in better Lysholm and Tegner scores and lesser risk of meniscus and cartilage injury in overall ligament injuries. The overall risk of meniscus tear and cartilage injury in the delayed surgery group was significantly increased compared to that in the early surgery group (meniscus, OR 1.73, p = 0.015; cartilage, OR 2.48, p = 0.0007), and this trend was found mostly in patients with a single ACL injury (Figs. 4a and 5a). These results are similar to those in studies on ACL injury indicating that chronic ACL injuries increase the risk of meniscus and cartilage injuries, which were recognized as predictors of osteoarthritis in the long-term follow-up [24,41,42,[61][62][63].
However, there were also studies that showed no differences between early and delayed surgeries [34,44,45]. In the recent meta-analysis by Ferguson et al. [20], the risk of meniscus tear or cartilage injury was not significant between the early and delayed surgery groups, although the result for cartilage injury was borderline significant (p = 0.06). This difference might originate from the difference in the number of included studies due to the longer inclusion period of this study. Moreover, the Tegner scores in the meta-analysis by Ferguson et al. [20] were 0.39 point greater in the early surgery group than in the delayed surgery group, which was similar to the results of our analysis (− 0.25 [95% CI Fig. 6 Forest plot of mean difference with 95% confidence intervals in Lysholm scores. The gray squares represent the results of each study. Ends of the horizontal bars represent 95% confidence intervals. Due to the heterogeneity, the non-weighted mean differences were used to assess overall results by the best-evidence synthesis; they were not shown in this plot  [34,45,64], based on the results and wider included literature in this study, early intervention would be recommended to decrease the risk of developing meniscal/ cartilage lesions and potentially reduce the subsequent risk of osteoarthritis and low functional outcomes. In MLKI, the timing of surgery did not significantly affect the incidence of meniscus tears and postoperative Tegner scores. The Lysholm scores were higher in the early surgery group than in the delayed surgery group (Fig. 6). Because all published studies on MLKI had low levels of evidence in study designs and showed heterogeneity, it is possible that future publications may change the trend of this meta-analysis by either confirming the outcome of this analysis or reversing these observed outcomes. Despite the limitations of the included study, the results of this meta-analysis are also in line with those of previous studies, which showed favorable results for early surgery [4,30,32,33,49,65]. According to Levy [65] and Hohmann [11], higher Lysholm and IKDC scores and satisfactory final range of motion (ROM) were found in the early surgery group. McKee et al. [66] and Vicenti et al. [60] also suggested the general consensus and results of early surgery in MLKI, within the first 3 weeks, and found greater ROM in the early surgery group than in the delayed surgery group [60]. The results of our analysis, including the Lysholm score, were similar to those of previous studies [4,65]. Although other previous studies reported that a high incidence of arthrofibrosis was found in the early surgery group [26,[67][68][69], and good clinical outcomes were found in the delayed surgery group [12,55,70,71], the results of this analysis and previous systemic reviews and meta-analyses [4,60,65,66] suggest that early surgery of MLKI yields higher Lysholm scores with similar incidence of concomitant injuries and improved functional outcomes compared to delayed surgery.

Limitations
This study has some limitations. First, the standard of timing was different according to the studies: the Fig. 7 Forest plot of mean difference with 95% confidence intervals in Tegner scores. The gray squares represent the results of each study. Ends of the horizontal bars represent 95% confidence intervals. Due to the heterogeneity, the non-weighted mean differences were used to assess overall results by the best-evidence synthesis; they were not shown in this plot definition of early ranged broadly up to 5 months and that of delayed ranged from 10 weeks to 24 months. Therefore, we included studies based on the author's definition of early and delayed rather than separate them as arbitrary time frames of early and delayed reconstructions by another definition. Second, all studies were included in the meta-analysis even if they did not report early vs delayed surgery as a primary outcome, especially in studies on MLKI; thus, the results were found to have substantial heterogeneity. It is extremely difficult to evaluate the clinical results on MLKI as prospective, comparative studies because of the heterogeneous nature of the injuries. Thus, we included all possible outcomes in this study and compared the results of single ACL injury as a control group. Third, relatively few studies with lower levels of evidence and small sample sizes were also major limitations, especially in studies on MLKI. Due to the searching strategy, the studies including early vs delayed and meniscus/cartilage injuries were selected for the systemic review, and the number of studies seems to be low. Moreover, in MLKIs, the different injury mechanisms and irregular knee ligament involvements and any possible combined traumas (vascular, nerve, fractures, etc.) contribute to the difficulties of analysis and obtaining consensus of treatment. The high risk of bias and heterogeneous publication of low level of evidence studies make the pooling results difficult to interpret as it is. However, in the meta-regression analysis according to the study design, there was no significant effect of the heterogeneity (Figs. 4b and 5b). Thus, we focused on the prevalence of concomitant injuries in the meta-analysis rather than the clinical outcomes, although the clinical outcomes were also reported as non-weighted means.

Conclusions
Our study suggests that delayed ACL surgery significantly resulted in higher risk of meniscus tear and cartilage injury and decreased Lysholm and Tegner scores compared to early ACL surgery. The Lysholm scores in the delayed MLKI surgery group were significantly decreased, but the risks of meniscus tear and cartilage injury in the delayed MLKI surgery group remained unclear.