Literature screening process and results
A total of 3,794 articles were retrieved from eight databases. After removing duplicates, 2,818 articles remained. Titles and abstracts were screened, leading to the exclusion of 2,155 articles (including reviews, conference papers, and those clearly irrelevant), leaving 663 articles for further evaluation. A subsequent full-text review excluded an additional 625 articles that did not meet the inclusion criteria. Ultimately, 38 RCTs were included in the network meta-analysis.(Fig. 1).
Basic characteristics of included studies
This study included 38 articles that assessed the impact of various exercise therapies on rehabilitation outcomes following ACLR, involving a total of 1,345 patients from diverse countries, including India, Thailand, Brazil, China, several Arab countries, Turkey, Greece, the Netherlands, Tunisia, Iran, the USA, the UK, Slovenia, Japan, Germany, and Italy, thus representing a wide geographical scope. Sample sizes across the studies varied from 5 to 49 participants, with differences in gender distribution. Some studies focused exclusively on male participants, while others included both males and females. The average age of participants ranged from 20 to 35 years. The timing and duration of interventions varied widely across studies, with intervention initiation ranging from 2 weeks to 6 months post-surgery. Most interventions lasted between 6 and 12 weeks, as seen in studies by Gupta et al. (2024) [22], Tepa et al. (2024) [23], Vidmar et al. (2020) [24], and Liu et al. (2019) [25]. Some studies had shorter intervention periods, such as Nambi et al. (2020) [26], which had a 4-week intervention, while others had longer durations, such as Milandri et al. (2021) [13] and Risberg et al. (2009) [27], which lasted 24 weeks and 6 months, respectively. Exercise therapy interventions play a critical role in post-ACLR rehabilitation and encompass a wide range of training methods.
First, Eccentric Training (ET) focuses on the load during the muscle extension phase, with common exercises including squats, leg extensions, and deadlifts [28]. The muscles are extended under control to enhance their resistance to stretching. Training is performed 2–3 times per week, with each session lasting 30–45 min. The intensity is set at 60–80% of the one-repetition maximum (1RM), and the intervention period generally lasts 4–12 weeks, with the load progressively increased as recovery advances. Similarly, Isokinetic Training (IT) uses isokinetic devices (such as isokinetic machines) for muscle exercises at a constant speed, with common exercises including knee extensions, leg presses, and hip bridges. This training is performed 2–3 times per week, each session lasting 30–40 min [29]. The intensity is automatically adjusted by the device to maintain a constant speed, with the intervention period typically ranging from 6–12 weeks.In addition to strength enhancement, Motor Control Training (MCT) focuses on improving muscle coordination and stability, with common exercises including balance training, core stability exercises, and controlled squats, aimed at restoring dynamic stability to the muscles [30]. This training is done 3–4 times per week, each session lasting 20–30 min, and the intervention period typically lasts 4–8 weeks, focusing on restoring coordination in the early stages of recovery. Another common method is Progressive Training (PT), which includes both aerobic and resistance training, with exercises such as squats, leg presses, and deadlifts, aimed at progressively restoring muscle strength [31]. It is performed 2–4 times per week, with each session lasting 30–45 min, and the intervention period lasts 6–12 weeks, with intensity gradually increased as recovery progresses. Cross-Training (CT) activates different muscle groups through various forms of exercise, including running, cycling, swimming, and alternating with resistance training [31]. Training is performed 3–4 times per week, with each session lasting 30–60 min, and the intervention period typically lasts 6–8 weeks, with intensity gradually increased as physical fitness improves. Complementing cross-training, Core Stability Training (CST) strengthens the abdominal and back muscles to improve body stability [32]. Common exercises include planks, bridges, and dead bugs. This training is done 3 times per week, with each session lasting 20–30 min, and the intervention period lasts 4–6 weeks, with intensity gradually increased as the training progresses. Additionally, Water Rehabilitation Training (WR) uses the buoyancy and resistance of water to reduce joint stress and help restore muscle strength [33]. Common exercises include walking, running, and knee extensions in water. The intensity is gradually increased by adjusting the water depth and resistance, with training performed 3 times per week, each session lasting 30–60 min, and the intervention period typically lasts 4–12 weeks. Similarly, Blood Flow Restriction Training (BFRT) enhances muscle endurance and strength by applying pressure to limit blood flow [34]. Common exercises include leg presses and squats, with intensity set at 30–50% of 1RM. This training is performed 2–3 times per week, with each session lasting 20–30 min, and the intervention period typically lasts 4–8 weeks.Whole-Body Vibration Training (WBVT) uses a vibration platform for full-body exercises, with common actions including squats, lunges, and calf raises [35]. The intensity is gradually increased by adjusting the vibration frequency and load, with training performed 2–3 times per week, each session lasting 15–30 min, and the intervention period typically lasts 6–12 weeks. Finally, Multimodal Training (MT) combines various training methods, such as resistance training, aerobic exercise, and core training, aimed at comprehensively improving muscle strength, flexibility, and endurance [36]. This training is performed 3–4 times per week, with each session lasting 30–60 min, and the intervention period typically lasts 6–12 weeks, with intensity adjusted according to the recovery progress.
These various exercise therapy interventions provide a comprehensive strategy to help patients restore muscle strength and joint function, enabling personalized rehabilitation plans tailored to specific exercise content, frequency, intensity, and duration. For example, GUPTA et al. (2024) [22] and TEPA et al. (2024) [23] examined the effects of isokinetic training and resistance training on post-surgical rehabilitation, demonstrating significant improvements in quadriceps strength and the International Knee Documentation Committee (IKDC) score. VIDMAR et al. (2020) [24] and LIU et al. (2019) [30] found that isokinetic training effectively enhanced the strength of both the quadriceps and hamstrings. Regarding outcome indicators, 31 studies [13, 14, 22,23,24,25, 27, 37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59] reported changes in quadriceps strength, emphasizing the importance of quadriceps strength as a key parameter in ACL rehabilitation. Additionally, 18 studies [13, 14, 25, 27, 37, 38, 42, 43, 45, 46, 48, 50, 53,54,55, 57, 58, 60] reported improvements in hamstring strength, highlighting the critical role of hamstring strength in post-surgical rehabilitation. Moreover, 18 studies assessed knee function scores [22,23,24,25,26,27, 37, 39, 49, 56,57,58, 61,62,63,64,65,66], including 7 studies on Lysholm scores [24, 25, 37, 57, 62, 64, 65], 8 studies on the International Knee Documentation Committee score [22, 23, 39, 49, 56, 58, 64, 66], 3 studies on the Cincinnati score [27, 61, 63], and one study on the Western Ontario and McMaster Universities Osteoarthritis Index [26]. Higher scores in the Lysholm, International Knee Documentation Committee, and Cincinnati scales indicate better knee function, [67] while a lower score on the Western Ontario and McMaster scale suggests better knee function [68]. Thus, knee function outcomes can be compared across studies (Table 2).
Quality assessment results of included studies
A total of 38 studies were included in this research, and their quality was thoroughly assessed using the Cochrane Risk of Bias Tool. Overall, the majority of studies exhibited strong performance in key areas such as randomization, allocation concealment, handling of missing data, and selective reporting, with only a few studies showing significant risk of bias. However, several studies raised concerns about blinding and other potential sources of bias, warranting further scrutiny. In terms of the randomization process, approximately 70% of the studies (e.g., OHTA [38], 2003; LIU [54], 2003) were rated as low risk, as they employed clear randomization methods and proper execution. Similarly, regarding allocation concealment, most studies (e.g., TAKARADA [40], 2004; COOPER [61], 2005) demonstrated a low risk of bias, indicating that appropriate measures were implemented to ensure proper allocation concealment during the study design. In relation to the handling of missing data, the majority of studies (e.g., NAMBI [26], 2020; VIDMAR [24], 2020) employed effective strategies for managing and supplementing missing data, minimizing their impact on the results and exhibiting a low risk of bias. Selective reporting was not a significant issue in most studies (e.g., OHTA, [38] 2003; KASMI [65], 2023), contributing to the transparency of the findings. However, certain studies (e.g., TOVIN [37], 1994; ZARGI [51], 2018) displayed more pronounced risks of bias, particularly in relation to blinding and unclear study design descriptions, which may have compromised the reliability of their results. See Table 3.
In summary, while most of the included studies demonstrated strong performance in areas such as randomization, allocation concealment, handling of missing data, and selective reporting, a small number of studies still exhibited risks in the implementation of blinding and reporting transparency. These issues should be addressed in future research. Despite these limitations, the overall quality of the studies provides a solid evidence base for this network meta-analysis.
Meta-analysis results
Network evidence plot
In the network evidence plot (Fig. 2), the size of each node is positively correlated with the sample size, while the thickness of the connecting lines between nodes reflects the frequency with which the two exercise methods are compared in the studies. A thicker line indicates that more studies are related to that comparison.
Evidence network diagram of included studies
Inconsistency test
The inconsistency analysis of outcome measures was performed using loop inconsistency tests, inconsistency models, and node-splitting methods. The loop inconsistency test revealed that in the RR-ET-MT loop for knee joint outcomes, the inconsistency factor (IF) was 1.538, with a p-value of 0.015. This indicates significant inconsistency and suggests differences in efficacy between the pathways. The results of the inconsistency model test showed that the p-values for all outcome measures exceeded 0.05, indicating that the inconsistency was not significant, and therefore, a consistency model was appropriate for further analysis. Additionally, the node-splitting method demonstrated that the direct and indirect evidence for each outcome measure was consistent (P > 0.05), further supporting the reliability of the results.
Meta-analysis of effect sizes and ranking results
In the analysis of quadriceps strength, compared to the control group, WBVT significantly improved quadriceps strength [SMD = 1.66, 95% CI (0.44, 2.87), P < 0.05], and WR also showed positive effects [SMD = 1.66, 95% CI (0.44, 2.87), P < 0.05]. Additionally, progressive training (PT) [SMD = 0.75, 95% CI (−0.42, 1.91), P > 0.05] led to some improvement in quadriceps strength, but it was less effective than both WBVT and WR. When compared to CT, WR resulted in a more significant improvement in quadriceps strength [SMD = 1.45, 95% CI (0.27, 3.18), P < 0.05] (Fig. 3). The Surface Under the Cumulative Ranking (SUCRA) probability ranking indicated that WR had the highest probability of being the most effective intervention (SUCRA = 83.6), followed by the control group (SUCRA = 73.1) and MCT (SUCRA = 61.6), with WBVT showing the lowest probability (SUCRA = 9.0) (Fig. 4).
Network meta-analysis results for quadriceps
Area under the cumulative probability ranking curve for quadriceps
In the analysis of hamstring strength, compared to the control group, WR significantly improved hamstring strength [SMD = 2.05, 95% CI (0.32, 3.79),P < 0.01], while MT also showed positive effects [SMD = 1.71, 95% CI (−0.21, 3.64), P = 0.04]. Furthermore, ET [SMD = 1.18, 95% CI (−0.61, 2.97), P > 0.05] resulted in a moderate increase in hamstring strength, but was less effective than both WR and MT. When compared to PT, WR demonstrated a more significant improvement in hamstring strength [SMD = 0.87, 95% CI (−0.24, 1.99), P < 0.05] (Fig. 5). The Surface Under the Cumulative Ranking (SUCRA) probability ranking showed that WR had the highest probability of being the most effective intervention (SUCRA = 89), followed by MT (SUCRA = 70) and the control group (SUCRA = 64.2), with IT having the lowest probability (SUCRA = 6.3) (Fig. 6).
Network meta-analysis results for hamstrings
Area under the cumulative probability ranking curve for hamstrings
In the analysis of knee joint strength, BFRT significantly improved knee strength compared to the control group [SMD = 2.00, 95% CI (−0.65, 4.65), P < 0.05], while MT also showed positive effects [SMD = 1.52, 95% CI (−0.50, 3.54), P > 0.05]. Additionally, IT [SMD = 0.47, 95% CI (−0.85, 1.78), P > 0.05] provided some improvement in knee strength, but was less effective than both BFRT and MT. When compared to PT, BFRT demonstrated a more significant improvement in knee strength [SMD = 2.83, 95% CI (0.05, 5.60), P < 0.05] (Fig. 7). The Surface Under the Cumulative Ranking (SUCRA) probability ranking revealed that BFRT had the highest probability of being the most effective intervention (SUCRA = 93.9), followed by WR (SUCRA = 71.9) and motor control training (MCT) (SUCRA = 67), with MT having the lowest probability (SUCRA = 15.2) (Fig. 8).
Network meta-analysis results for knee joint
Area under the cumulative probability ranking curve for knee joint
Test for publication bias or small sample effect
Publication bias for the outcome measures was assessed using a funnel plot. The results indicated that the plot exhibited good symmetry, suggesting minimal influence from publication bias or small sample effects (Fig. 9).
Publication bias test plots for included studies
Evidence strength rating of the included literature indicators
In accordance with the GRADE system for assessing evidence quality, the strength of the study’s findings may be downgraded based on five factors: (1) Risk of bias: most studies lacked clear descriptions of their randomization methods, allocation concealment, and blinding procedures; (2) Inconsistency: moderate to high levels of heterogeneity among results; (3) Indirectness: outcomes not directly linked to the research objectives; (4) Imprecision: wide confidence intervals; and (5) Publication bias: asymmetry in the funnel plot [69, 70]. All included studies were randomized controlled trials. However, due to insufficient clarity regarding randomization, allocation concealment, and blinding methods in the majority of studies, all three outcome measures were initially downgraded. Consequently, the quadriceps strength outcome was further downgraded due to high heterogeneity, resulting in a “low” level of evidence. The hamstring strength outcome, which was only influenced by risk of bias, remained at a “moderate” level. Knee function, which was affected by a combination of risk of bias, inconsistency, and indirectness, was ultimately downgraded to a “very low” level of evidence. Detailed results are presented in Table 4.
link