Efficacy of aerobic exercises for knee osteoarthritis: a network meta analysis of randomized clinical trials

Abstract

Background

Previous research has demonstrated the therapeutic potential of aerobic exercise in alleviating symptoms of knee osteoarthritis (KOA). Nevertheless, comparative evidence regarding the relative effectiveness of different exercise modalities remains inconclusive, and the optimal exercise protocol continues to be debated. To address this knowledge gap, we performed a systematic network meta-analysis to compare and rank the clinical efficacy of various aerobic exercise regimens for managing KOA.

Methods

A systematic literature search was conducted across five electronic databases (PubMed, Cochrane Library, Web of Science, Embase, and Scopus) from database inception through June 2024. Eligible studies included randomized controlled trials (RCTs) evaluating aerobic exercise interventions for KOA management. Two investigators independently performed study selection using predefined inclusion criteria, with discrepancies resolved through consensus or third-party adjudication. Data extraction encompassed demographic characteristics, intervention protocols, and outcome measures. Methodological quality was assessed using the Cochrane Risk of Bias Tool 2.0. Statistical analyses were performed using Stata 17.0 (Network Meta-Analysis package) under a frequentist framework, with treatment effects estimated through surface under the cumulative ranking curve (SUCRA) probabilities.

Results

The network meta-analysis included 67 randomized controlled trials comprising 4,944 patients with knee osteoarthritis (KOA), assessing 10 aerobic exercise interventions: walking (WK), weight-loss walking (LK), retro walking (RW), cycling (CY), aquatic training (AT), yoga (YG), Tai Chi (TC), Baduanjin (BD), Wuqinxi (WQ), and Pilates (PT). Surface under the cumulative ranking curve (SUCRA) probability analyses yielded the following results: Pilates (PT) demonstrated the highest probability of being optimal for WOMAC pain score (SUCRA = 0.8%), WOMAC stiffness (SUCRA = 15.7%), physical function (SUCRA = 0.0%), and total WOMAC score (SUCRA = 7.8%). Tai Chi (TC) showed the highest likelihood of efficacy for Visual Analog Scale (VAS) outcomes (SUCRA = 17.4%), while weight-loss walking (LK) ranked first for Timed Up and Go (TUG) improvement (SUCRA = 27.1%). The comprehensive efficacy ranking was PT > LK > BD > YG > AT > WK > RW > TC > WQ > CY.

Conclusion

This study demonstrates that Pilates appears to be the most effective aerobic exercise modality for managing knee osteoarthritis (KOA), particularly in enhancing overall functional outcomes. Tai Chi exhibited the greatest efficacy in reducing pain intensity, as quantified by the Visual Analog Scale (VAS). Based on these findings, Pilates and Tai Chi should be prioritized as primary therapeutic interventions for the majority of KOA patients.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13018-025-05973-z.

Introduction

Knee osteoarthritis (KOA), a chronic degenerative joint disorder, is pathologically characterized by articular cartilage degeneration, osteophyte formation, and joint space narrowing. Its global prevalence exhibits a strong age-dependent correlation, with incidence rates escalating markedly in older populations [1]. The progressive pathological cascade involves synovial inflammation, subchondral bone remodeling, and periarticular soft tissue lesions, collectively contributing to persistent nociceptive pain, mechanical stiffness, and restricted range of motion [2].Clinically, KOA not only severely impairs physical functioning in activities of daily living (e.g., ambulation and stair negotiation) but is also strongly associated with psychosocial comorbidities, including depression and social participation restrictions. Approximately 40% of patients report clinically significant quality-of-life deterioration secondary to chronic pain and functional disability [3].Moreover, KOA management imposes substantial socioeconomic burdens, encompassing direct medical expenditures (e.g., pharmacological interventions and arthroplasty procedures) and indirect costs related to workforce productivity loss. Epidemiological studies estimate the condition’s annual financial impact to account for 10–35% of patients’ disposable income [4].

Current clinical management of knee osteoarthritis (KOA) primarily revolves around pharmacological analgesia and surgical interventions, yet both approaches present considerable limitations. Oral NSAIDs and paracetamol remain first-line pharmacological agents for pain management in KOA. However, prolonged administration is associated with increased risks of gastrointestinal bleeding, cardiovascular complications, and even mortality. This safety concern is particularly pronounced in elderly populations and patients with chronic comorbidities [6, 7]. While joint replacement surgery demonstrates efficacy in restoring functionality for end-stage patients, its clinical utility is constrained by surgical invasiveness, substantial healthcare costs, and potential postoperative complications [5]. Regenerative therapeutic approaches, including stem cell-based interventions, show promising biological potential but currently lack robust clinical evidence for sustained efficacy while requiring significant financial investment [5]. These inherent limitations underscore the critical demand for developing safe, cost-effective, and sustainable non-pharmacological intervention strategies in KOA management.

In recent years, exercise therapy has emerged as a first-line non-pharmacological intervention for knee osteoarthritis (KOA) management, endorsed by internationally recognized clinical guidelines. Its therapeutic value stems from multimodal benefits including analgesia, functional enhancement, and cardiopulmonary health promotion [4]. Contemporary aerobic exercise protocols encompass both conventional modalities (walking, cycling, aquatic exercises) and integrated mind-body disciplines (Pilates, Tai Chi, Baduanjin), with emerging evidence suggesting differential therapeutic benefits mediated through biomechanical parameters (muscle strength modulation, joint load distribution) and neuromuscular adaptation mechanisms [8–13]. However, current clinical research predominantly examines singular exercise modalities against standard care, revealing a critical paucity of head-to-head comparative efficacy data among various aerobic interventions. This evidence gap contributes to clinical decision-making challenges regarding optimal exercise prescription, particularly the absence of evidence-based selection criteria for specific patient subgroups [14]. Conducting a network meta-analysis to synthesize direct and indirect comparative evidence could establish hierarchical efficacy profiles for common aerobic exercises across key outcomes: pain intensity reduction, functional capacity improvement, and quality of life enhancement. Such methodological approach would advance precision medicine in rehabilitation by informing tailored exercise prescription protocols and optimizing personalized KOA management frameworks.

Methods

Literature search strategy

Two researchers independently conducted electronic literature searches to assess the outcomes of different aerobic exercises in treating knee osteoarthritis (KOA). As of July 2024, electronic databases including PubMed, Embase, Web of Science, and Cochrane Library were searched, limited to articles in English. Search terms included: “Osteoarthritis, Knee,” “Knee Osteoarthritis,” “Knee Osteoarthritides,” “Osteoarthritis of Knee,” “Osteoarthritis of the Knee,” “Aerobics,” “Physical Exercise,” “Aerobic Exercise,” “Aerobic Exercises,” “Physical Activity,” “walk,” “run,” “dance,” “Dance Therapy,” “swim,” “aquatic exercises,” “aquatic sports,” “bike,” “bicycle,” “treadmill,” “Treadmill exercise,” “jogging,” “Pilates,” “Tai Ji,” “Tai-ji,” “Tai Chi,” “wuqinxi,” “yoga,” “calisthenics,” “skating,” “Skateboarding,” “stair climbing,” “Climbing, Stair,” “Stair Navigation,” “Navigation, Stair,” “baduanjin,” “backward walking,” “retro walking,” etc.

Inclusion and exclusion criteria

Inclusion criteria: (1) Published randomized controlled trials (RCTs) on aerobic exercise therapy for KOA; (2) Articles limited to the English language; (3) Patients diagnosed clinically with KOA [15], without restrictions on age, gender, nationality, or disease duration; (4) Interventions included walking, low-impact walking, backward walking, cycling, aquatic exercise, yoga, Tai Chi, Baduanjin, Wuqinxi, and Pilates, with control groups involving exercise training and placebo; (5) Outcome measures (at 8 weeks post-treatment): ① Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC); ② Visual Analog Scale (VAS); ③ Timed Up and Go (TUG) test.

Exclusion criteria: (1) Studies using pharmacological treatment, surgical treatment, combined treatments, or resistance exercise therapy; (2) Duplicate studies published by the same author; (3) Reviews, meta-analyses, case reports, conference papers, etc.; (4) Studies where full-text data cannot be obtained.

Literature screening and data extraction

This study initially involved two researchers (first and second authors) who independently searched electronic databases and removed duplicate studies using EndNote software. They then screened by reading titles and abstracts, excluding irrelevant literature(Including comments, systematic reviews and meta-analyses, and studies unrelated to the topic), and finally conducted full-text screening. Any disagreements during the screening process were discussed until a consensus was reached; if no consensus was reached, a third researcher made the final decision after group discussion. Data collection involved two researchers extracting data and then pooling it, including: first author of the study, publication year, country, age of participants, sample size, intervention method, treatment duration, and outcome assessment.The proportion of missing data included in the study is relatively small, and the deletion occurs randomly, so this study adopts the direct deletion method for missing data.

Quality assessment of literature

The methodological bias and quality assessment of included RCTs were conducted according to the Cochrane Handbook for Systematic Reviews of Interventions [16, 17], assessing: sequence generation of randomization, allocation concealment, blinding of researchers and patients, blinding of outcome assessment, incomplete outcome data, selective reporting of results, and other sources of bias. Assessment options included: ‘Low risk,’ ‘High risk,’ or ‘Unclear risk.’

Statistical analysis

To conduct a comprehensive network meta-analysis, we utilized the statistical software packages “Network” and “mvmeta” within STATA 17.0 software. Continuous variables including WOMAC, VAS and TUG were analyzed using weighted mean differences (WMD) with corresponding 95% CI. When the 95% confidence interval (CI) value of WMD includes 0, it indicates that there is no statistically significant difference in the mean values between the two groups.

For direct comparisons, a conventional meta-analysis was conducted to aggregate the results using random-effects models, serving as sensitivity analyses. The network meta-analysis employed a frequentist approach with a random-effects model to estimate both direct and indirect comparisons. The primary objective of the network meta-analysis was to assess whether any of the comparator interventions demonstrated superiority. To evaluate potential inconsistencies between indirect and direct comparisons, we employed global inconsistency, local inconsistency (using a node-splitting approach), and loop inconsistency. Statistical significance for global inconsistency was determined using P-values, with P > 0.05 indicating no significant global inconsistency. Local inconsistency was assessed through node-splitting analysis, and P > 0.05 indicated no significant local inconsistency. Heterogeneity within each closed loop was estimated using the inconsistency factor (IF), with a 95% CI (IF) value of zero signifying no statistical significance. In each pre-specified outcome, a global network diagram was employed to illustrate direct comparisons between interventions. The size of the nodes in the diagram corresponded to the number of participants receiving each treatment. Treatments subject to direct comparisons were linked by lines, and the thickness of these lines was proportional to the number of trials evaluating the specific comparison.

Within the “Results” section, the ranking probability of each intervention was presented using a cumulative probability ranking graph. The graph incorporated the Surface Under the Cumulative Ranking Curve (SUCRA) value, serving as an index that summarized the cumulative ranking probability. The SUCRA value ranged between 0 and 100%, where a larger SUCRA value indicated a higher ranking for the intervention, typically reflecting a more favorable or less favorable effect. All intervention measures were ranked based on their respective SUCRA values or the area under the curve, resulting in a comprehensive ranking of the interventions. For the handling of missing values, Due to the small and randomly distributed proportion of missing data, direct deletion was applied.

To assess the potential for publication bias, the comparison-adjusted funnel plot was utilized. This analysis aimed to determine whether there was evidence of a small sample effect or publication bias within the intervention network.

Results

Search outcomes and study characteristics

Our preliminary search across five databases returned 9849 records, with the following distribution: PubMed (2697), EMBASE (2105), Cochrane Library (1763), Web of Science (2206), and Scopus (1078). Following a rigorous screening process, we included 67 RCTs [18–84] that investigated the effects of various aerobic exercises on knee osteoarthritis (KOA), encompassing a total of 4,944 patients. No further studies were identified through review articles or the references of included studies. The two researchers were in unanimous agreement throughout the search and study selection process. The search flow is depicted in Fig. 1. The interventions in the included studies spanned 10 types of aerobic exercises: walking, low-impact walking, backward walking, cycling, aquatic exercise, yoga, Tai Chi, Baduanjin, Wuqinxi, and Pilates, with the distribution of studies across these interventions detailed in Table 1.

Fig. 1.

Flow chart of literature screening for this meta-analysis

Table 1.

The general date of the included studies

Inclusion study	Yr	Country	N	Intervention measure	Period	Age	Outcome Index
Hiyama [18]	2012	Japan	20/20	WK/PB	8wk	71.9 ± 5.2 73.8 ± 5.7	③
Drummen [19]	2023	Australia	17/16	WK/PB	24wk	65.1 ± 8.9 67.4 ± 8.3	①②
Arvin [20]	2020	Malaysia	17/17	RW/WK	12wk	60.12 ± 6.1 56.71 ± 5.39	②③
Lund [21]	2008	Denmark	27 27 25	AT/PB/ET	8wk,12wk	65 ± 12.6 70 ± 9.9 68 ± 9.5	②
Ettinger [22]	1997	America	127 133 132	WK/ET/PB	12wk	69 ± 6 68 ± 6 69 ± 6	②
Wallias [23]	2017	Australia	22/23	WK/PB	12wk	68 ± 8 67 ± 7	①②
Bingchen [24]	2008	China	11/10	BD/PB	8wk	65.4 ± 8.2 64.6 ± 6.7	①
Kabiri [25]	2018	China	24/23/23	WK/CY/ET	8wk	56.92 ± 1.37 55.74 ± 1.43 60.72 ± 6.37	②③
Foley [26]	2023	China	35/35/35	AT/ET/PB	8wk	73 ± 8.2 69.8 ± 9.2 69.8 ± 9	①
Alghadir [27]	2019	Saudi Arabia	21/20/18	RW/WK/ET	8wk	45–66 45–66 47–65	①②③
Watanabe [28]	2013	Japan	15 15	WK/LK	8wk	75 ± 7.6 80 ± 5.9	②③
Ebnezar [29]	2012	India	125 125	YG/PB	8wk,12wk	59.56 ± 9.54 59.42 ± 10.66	②
Saleem [30]	2022	Pakistan	20 20	PT/PB	8wk	57.6 ± 6.34 55.65 ± 7.28	①②
Zhu [31]	2017	China	23 23	TC/PB	24wk	64.6 ± 3.4 64.5 ± 3.4	②
Deepeshwar [32]	2018	India	31/35	YG/PB	8wk	61.83 ± 9.1 60.13 ± 8.6	③
Evcik [33]	2002	Turkey	27 28 26	ET/PB/WK	12wk	56.3 ± 6.1 56.9 ± 6.5 55.8 ± 6.9	①②
Lopez [34]	2022	Spain	20 20	AT/ET	8wk	60.29 ± 9.91 65.18 ± 9.04	①③
Lim [35]	2010	Korea	26/25/24	AT/ET/PB	8wk	65.7 ± 8.9 67.7 ± 7.7 63.3 ± 5.3	①
Taglietti [36]	2018	Brazil	31/29	AT/PB	8wk,24wk	67.3 ± 5.9 68.7 ± 6.7	①②③
Ha [37]	2018	Korea	9 8	AT/PB	12wk	60.89 ± 5.06 61.25 ± 1.91	①
Jsus [38]	2017	Spain	17 17	AT/ET	8wk,12wk	65.62 ± 7.15 66 ± 6.35	①
Bokaeian [39]	2021	Iran	22 19 18	YG/ET/WK	8wk	54.9 ± 5 57 ± 4.9 56.7 ± 4.7	①②
John [40]	2012	India	118/117	YG/PB	12wk	59.56 ± 8.18 59.42 ± 10.66	②
Khruakhorn [41]	2021	Thailand	17/17	AT/ET	8wk	64.88 ± 7.44 57.88 ± 7.75	①
Chen [42]	2021	China	8 10	LK/WK	8wk	59.63 ± 834 58.3 ± 8.54	①②
Tiago [43]	2023	Brazil	8/9	PT/PB	8wk	65.75 ± 2.76 64.78 ± 2.17	①
Zhu [44]	2016	China	23 23	TC/PB	24wk	64.61 ± 3.4 64.53 ± 3.43	①②
Hu [45]	2019	China	52 40	TC/PB	24wk	66.32 ± 4.16 65.54 ± 3.59	①②
Yong [46]	2020	China	45 40	WQ/PB	24wk	70.7 ± 9.36 70.2 ± 9.35	①③
Kuntz [47]	2018	Canada	10 10 10	YG/ET/PB	8wk	65.5 ± 5.6 63.7 ± 8.9 71.1 ± 9.3	②
Alonso [48]	2021	Spain	59/56	ET/AT	8wk	69.7 ± 11.2 69.3 ± 5.8	①
Mei [49]	2020	China	34 34	WQ/PB	8wk,12wk	70.7 ± 9.36 70.2 ± 10.35	①③
Kang [50]	2022	China	12 15	TC/PB	8wk	63.4 ± 4.6 64.7 ± 6.1	①
Brismee [51]	2007	America	22 19	TC/PB	8wk,12wk	70.8 ± 9.8 68.8 ± 8.9	①②
Dias [52]	2017	Brazil	33/32	AT/PB	8wk	70.8 ± 5 71 ± 5.2	①
Silva [53]	2008	Brazil	32/32	AT/ET	8wk	59 ± 7.6 59 ± 6.09	①②
Oliveira [54]	2012	Brazil	50 50	CY/PB	8wk	61.5 ± 6.94 58.78 ± 9.6	①③
Chen [55]	2021	China	36 32	TC/PB	8wk	77.4 ± 5.9 75.4 ± 6.4	③
Alkatan [56]	2016	America	24 24	AT/CY	8wk	59 ± 2 61 ± 1	①
Vilai [57]	2018	Thailand	33/37	AT/ET	8wk	62.1 ± 6.4 61.7 ± 6.9	②
Keogh [58]	2018	Australia	9/8	ET/CY	8wk	59.1 ± 6.7 66.1 ± 8.8	①
Ye [59]	2020	China	28 28	BD/PB	8wk,12wk	65.11 ± 6.57 63.61 ± 2.63	①
Kuptniratsaikul [60]	2022	Thailand	34/40	LK/AT	8wk	63.9 ± 5.5 66.2 ± 6.4	①②
Cochrane [61]	2005	Britain	59/39	AT/PB	12wk	68.8 ± 6.37 69.9 ± 5.16	①
Arun [62]	2013	India	15/15	ET/RW	8wk	63.43 ± 6.2 63.43 ± 6.2	②
Kovar [63]	1992	America	47 45	WK/PB	8wk	70.38 ± 9.11 68.48 ± 11.32	②
Li [64]	2019	China	54 53	TC/ET	12wk	69.6 ± 4.3 68.5 ± 3.5	①
Ni [65]	2010	China	14 15	TC/PB	24wk	62.89 ± 2.79 63.47 ± 2.85	①
Lee [66]	2009	Korea	29/15	TC/PB	8wk	70.2 ± 4.8 66.9 ± 6	①
Mazloum [67]	2017	Iran	17/16/16	PT/ET/PB	8wk	55 ± 8.2 50.3 ± 8.3 50.8 ± 9.9	①
Kilic [68]	2020	Turkey	25 25	WK/PB	8wk	59.52 ± 8.57 60.48 ± 7.43	①②
Samut [69]	2015	Turkey	15/14/13	ET/WK/PB	8wk	62.46 ± 7.71 57.57 ± 5.79 60.92 ± 8.85	①②
Jiulong [70]	2022	China	20/20	TC/PB	12wk,24wk	64.15 ± 8.56 64.15 ± 8.56	①③
Assar [71]	2020	Iran	12/12/12	AT/ET/PB	12wk	57.5 ± 6.9 55.9 ± 8.6 63.8 ± 7.4	①②
Xiao [72]	2021	China	132 134	WQ/PB	8wk,24wk	71 ± 2.92 69 ± 3.72	①
Kim [73]	2012	Korea	35 35	AT/PB	12wk	/	②
Wyatt [74]	2001	America	21/21	AT/ET	8wk	45–70 45–70	②
JiaJia [75]	2020	China	25/25	BD/PB	8wk,12wk	64.48 ± 7.81 63.08 ± 3.65	①
Salacinski [76]	2015	Canada	19 18	CY/PB	8wk,12wk	55.1 ± 10.5 60.6 ± 8.4	①②
Kunduracilar [77]	2018	Turkey	30 29	AT/PB	8wk	63.2 ± 7.59 58.23 ± 7.55	①②
Wang [78]	2021	China	43/41	ET/BD	12wk	65.7 ± 3.5 64.74 ± 2.8	①
Zehua [79]	2021	China	16 16	RW/PB	8wk	60.31 ± 7.85 60.94 ± 6.89	①②
Britt [80]	2023	Norvay	54/53/54	ET/CY/PB	12wk	57.6 ± 6.6 57.3 ± 7.1 57.8 ± 7.4	②
Brosseau [81]	2012	Canada	43 41	WK/PB	24wk	/	①
Parisa [82]	2019	Iran	15/15	WK/PB	8wk	53.8 ± 7.43 59.6 ± 7.43	②
Cheung [83]	2014	America	18 18	YG/PB	8wk	69.3–74.6 69–75	①
Moonaz [84]	2015	America	25/28	YG/PB	8wk	49.2 ± 3.2 55.9 ± 8.9	②

①WOMAC;②VAS;③TUG; BD = baduanjin; PB = control group; WK = walk; PT = pilates train; ET = exercise therapy; WQ=Wuqinxi

RW = retro walking; AT = aquatic train; TC = tai chi; CY = cycing; YG = yoga; LK = weighe-loss walking;

Risk of Bias and quality assessment

The quality assessment of the included 67 RCTs was conducted using the Cochrane Collaboration’s “Risk of Bias” tool. The risk of bias assessment for the included studies is presented in Fig. 2.

Fig. 2.

Risk of bias assessment summary of this meta-analysis

Network Meta-Analysis results

Network relationship diagram

The included literature involved 12 distinct intervention modalities: walk, weight-loss walking, retro walking, cycling, aquatic training, yoga, Tai Chi, Baduanjin, Wuqinxi, Pilates, exercise therapy and placebo. Specifically, 23 studies assessed the WOMAC Knee pain score, creating 6 closed loops; 21 studies evaluated the WOMAC Knee stiffness score, with 3 closed loops; and another 21 studies examined the WOMAC Knee function score, also forming 3 closed loops. The Total WOMAC score was reported in 22 studies, resulting in 6 closed loops. VAS scores were documented in 24 studies, yielding 11 closed loops, while 20 studies reported TUG results, forming 5 closed loops. Figure 3 illustrates the network relationship.

Fig. 3.

Network analysis of eligible comparison for (A) WOMAC pain score, (B) WOMAC function score, (C) WOMAC stiffness score, (D) WOMAC total score, (E) VAS, (F)TUG. The size of each node represents the number of participants, while the thickness of the line represents the number of studies directly comparing the two interventions

Consistency assessment

Consistency assessments were conducted on the closed loops for the six outcome indicators extracted from the literature. The findings revealed that for the six closed loops related to the WOMAC Knee pain score, the three closed loops related to the WOMAC Knee stiffness score, the three closed loops related to the WOMAC Knee function score, the six closed loops related to the WOMAC Total score, the eleven closed loops related to VAS scores, and the five closed loops related to TUG, the lower limits of the 95% CIs were above zero for all. Collectively, this indicates good consistency within the closed loops for each of the five indicators, justifying the use of a consistency model in the network meta-analysis. Figure 4 illustrates these findings.

Fig. 4.

Inconsistency plot of eligible comparison for (A) WOMAC pain score, (B) WOMAC function score, (C) WOMAC stiffness score, (D) WOMAC total score, (E) VAS, (F)TUG

Network Meta-Analysis results for included studies’ outcomes

Network forest plots were constructed for each outcome measure, and the results are as follows: WOMAC Pain Score: PT was significantly more effective than PB, RW, LK, WQ, CY (P < 0.05).BD outperformed PB/WQ, CY outperformed BD, and ET outperformed PB.WOMAC Stiffness Score: BD was significantly more effective than PB, PT, YG, WQ, CY (P < 0.05).PT outperformed WQ/CY.WOMAC Function Score: PT was significantly more effective than WK, RW, TC, YG, BD, CY, PB, and WQ (P < 0.05).ET outperformed PB.WOMAC Total Score: CY outperformed PB, RW, AT, and ET outperformed CY, PB.VAS Score: TC was significantly more effective than RW, PB.TUG Test: YG was significantly more effective than WK, TC, PB.with all other comparisons showing no statistical significance (P > 0.05), see Fig. 5.

Fig. 5.

Forest plots for (A) WOMAC pain score, (B) WOMAC function score, (C) WOMAC stiffness score, (D) WOMAC total score, (E) VAS, (F)TUG

Ranking results of network Meta-Analysis for included studies’ outcomes

Cumulative probability rankograms were plotted for comparisons of different intervention measures in each outcome indicator, and SUCRA values were calculated for ranking based on these values.

1. WOMAC (Knee pain score): PT (0.8%) > DB (16.9%) > YG (28%) > LK (36.4%) > WK (42.6%) > AT (45.3%) > ET (48%) > TC (59.8%) > RW (69.8%) > CY (80.5%) > WQ (84.7%) > PB (87.2%).

2. WOMAC (Knee stiffness score): BD (0.1%) > PT (15.7%) > YG (23.5%) > WK (29.7%) > AT (41.6%) > ET (52.1%) > CY (67.2%) > WQ (75.1%) > TC (80.2%) > PB (81.9%) > RW (82.9%).

3. WOMAC (Knee function score): PT (0%) > ET (22.3%) > AT (22.6%) > WK (45.6%) > YG (49.8%) > RW (51.2%) > TC (55%) > CY (57.7%) > BD (73.1%) > PB (84.2%) > WQ (88.5%).

4. WOMAC (Total score): PT (7.8%) > LK (22.5%) > RW (35.3%) > AT (41.1%) > BD (41.9%) > YG (44.7%) > WK (49.5%) > TC (61.5%) > ET (65.6%) > PB (87.5%) > CY (92.5%).

5. VAS: TC (17.4%) > LK (25.2%) > PT (32.7%) > RW (33.6%) > AT (33.8%) > ET (53.1%) > WK (57%) > CY (78.5%) > YG (79.1%) > PB (89.6%).

6. TUG: LK (27.1%) > YG (27.9%) > RW (29.2%) > WK (37.2%) > ET (39%) > TC (49.5%) > CY (58.1%) > AT (59%) > WQ (81.1%) > PB (92%).

As shown in Fig. 6, the comprehensive ranking of aerobic measures for treating knee osteoarthritis based on various assessment indicators is: PT > LK > BD > YG > AT > WK > RW > TC > WQ > CY.

Fig. 6.

Surface under the cumulative ranking (SUCRA) for (A) WOMAC pain score, (B) WOMAC function score, (C) WOMAC stiffness score, (D) WOMAC total score, (E) VAS, (F)TUG

Evaluation of publication Bias in included studies

Funnel plots with adjustment for comparison were created to evaluate publication bias for WOMAC (Knee pain score), WOMAC (Knee stiffness score), WOMAC (Knee function score), WOMAC (Total score), VAS scores, and TUG (Fig. 7). The plotted points are predominantly within the funnel’s boundaries, and they exhibit a roughly symmetrical distribution around the line of symmetry at X = 0, indicating a low probability of publication bias or small sample size effects.

Fig. 7.

Funnel plots of (A) WOMAC pain score, (B) WOMAC function score, (C) WOMAC stiffness score, (D) WOMAC total score, (E) VAS, (F>) TUG

Discussions

While extensive systematic reviews and meta-analyses have established the favorable clinical outcomes of various aerobic exercises in knee osteoarthritis (KOA) management [85–93], consensus regarding optimal exercise modalities remains elusive. This study presents the first network meta-analysis comparing therapeutic efficacy between traditional Chinese exercises (Tai Chi, Baduanjin, Wuqinxi) and Western-derived regimens (Pilates, weight-loss walking, yoga) in KOA. Our findings reveal that Pilates demonstrated superior composite functional outcomes (WOMAC function, TUG) compared to traditional modalities, whereas Tai Chi exhibited optimal performance in pain alleviation (VAS), aligning with conclusions from Pacheco-Barrios [94] and Nunez-Cortes [95]. This divergence may originate from distinct biomechanical and neurophysiological mechanisms underlying these exercise paradigms.Pilates enhances knee valgus correction [98] and hip-knee kinematic coordination [99] via core muscle activation and dynamic stability training, directly addressing KOA-related biomechanical dysfunction. Its standardized protocol (e.g., 45-minute sessions thrice weekly) facilitates reproducibility and may prolong functional benefits through consistent neuromuscular adaptation [96]. In contrast, Tai Chi modulates pain pathways through vagus nerve-mediated neuroimmunological effects, downregulating pro-inflammatory cytokines (IL-6, TNF-α) [105] and amplifying T lymphocyte-dependent anti-inflammatory responses [106, 108]. While these mechanisms underpin its analgesic superiority, its low-intensity nature may constrain biomechanical adaptation potential [115–121].

Current evidence suggests Pilates may offer substantial therapeutic benefits for knee osteoarthritis management. As a multidimensional mind-body intervention applicable across rehabilitation and fitness contexts, Pilates incorporates dynamic stretching protocols that appear to enhance cardiopulmonary efficiency and aerobic capacity [96, 97]. González et al. [98] reported measurable biomechanical improvements through a 12-week Pilates program, including a 5.1 cm reduction in dynamic knee valgus displacement. Complementary findings by Laws et al. [99] indicate potential benefits in lower extremity joint alignment through enhanced hip-knee kinematic coordination during functional movements. While the precise mechanisms underlying these observed effects require further elucidation, available data point to several plausible therapeutic pathways: neuromuscular re-education supporting dynamic joint stabilization, development of controlled movement patterns, and adaptive loading strategies that account for degenerative joint pathophysiology.

Weight-loss walking, a therapeutic gait modification strategy utilizing suspension systems (e.g., unloading vests or harnesses), partially offsets gravitational loading on lower extremities to facilitate ambulatory training within biomechanically safe parameters. Emerging evidence positions weight-loss walking as demonstrating secondary efficacy to Pilates in comprehensive knee osteoarthritis (KOA) management, while its analgesic outcomes approximate those of Tai Chi. This modality achieves systematic joint unloading during gait cycles, mechanistically distinct from aquatic therapy by eliminating hydrodynamic resistance requirements while enhancing pain modulation and gait velocity in KOA populations [60]. Peeler et al. [100] validated its clinical applicability for maintaining ambulatory frequency without exacerbating joint stress, with observed benefits including optimized quadriceps neuromuscular recruitment. Further investigations corroborate its therapeutic multifactoriality: redistribution of plantar pressure vectors, attenuation of peak tibiofemoral compressive forces, and sustainable elevation of daily activity thresholds—critical biomarkers for KOA rehabilitation [101, 102].

Tai Chi demonstrates superior efficacy in pain reduction as quantified by Visual Analog Scale (VAS) measurements among knee osteoarthritis (KOA) interventions. This traditional Chinese mind-body practice employs coordinated movement sequences featuring controlled postural transitions, diaphragmatic breathing synchronization, and meditative relaxation, offering dual benefits for psychosomatic health and cardiovascular adaptation. Its clinical implementation as a non-pharmacological analgesic modality in KOA management has gained substantial recognition [103, 104]. Mechanistically, Du et al. [105] have identified immunomodulatory effects through vagus nerve-mediated anti-inflammatory pathways, involving downregulation of IL-6 and TNF-α cytokines alongside tissue regeneration potentiation. Furthermore, controlled trials reveal Tai Chi’s capacity to enhance erythrocytic immune competence and lymphocyte (T-cell/B-cell) proliferation kinetics in postmenopausal populations, establishing a physiological foundation for its localized anti-inflammatory and analgesic outcomes [106].

Tai Chi and weight-loss walking demonstrate distinct therapeutic advantages in pain management through dual mechanisms involving central sensitization modulation (via endogenous opioid system activation [112]) and peripheral nociceptive signal regulation (through biomechanical load redistribution during gait cycles [60]). This mechanistic alignment provides empirical support for Pacheco-Barrios’ hypothesis postulating an inverse relationship between exercise intensity and analgesic efficacy in knee osteoarthritis management [94].

Baduanjin and yoga demonstrate clinically meaningful efficacy in knee osteoarthritis (KOA) management, both categorized as low-to-moderate intensity aerobic interventions. Baduanjin, a traditional Chinese Qigong modality, employs eight biomechanically structured movement sequences that induce systemic physiological regulation through targeted musculoskeletal engagement [107]. When evaluated by oxygen consumption (VO2)– the criterion standard for exercise intensity quantification– Baduanjin meets the threshold for moderate-intensity aerobic activity (mean energy expenditure: 23.3 ± 4.4 kcal/session), confirming its safety and therapeutic validity in KOA populations [108]. These practices, alongside conventional aerobic regimens, constitute foundational rehabilitation strategies for KOA, demonstrating multidimensional benefits encompassing pain modulation, joint preservation, and functional restoration [32]. Yoga’s therapeutic paradigm integrates controlled respiratory patterns, low-impact postures (asanas), and meditative focus, exhibiting particular efficacy in mitigating KOA-associated psychological comorbidities such as depression and anxiety [110]. Biomechanical analyses reveal yoga’s clinical advantage through reduced mean knee adduction torque relative to gait-cycle peaks (18.2% reduction), correlating with patient-reported improvements in symptom severity, functional mobility, and activity tolerance [111]. Comparative evidence suggests these low-impact modalities share therapeutic advantages over traditional aerobic exercises (walking/cycling/aquatics), including optimized joint protection through load distribution and enhanced neuromuscular coordination via proprioceptive refinement, while avoiding prolonged high-intensity joint stress. Emerging mechanistic evidence highlights neuromuscular control as the primary driver of functional recovery. Both Pilates and Baduanjin exhibit superior capacity to counteract KOA-related functional deterioration through quadriceps eccentric strengthening (28.7% improvement vs. baseline [78]) and enhanced proprioceptive acuity (41.3% error reduction in joint position sense [109]). These findings align with Nunez-Cortes’ proposed rehabilitation paradigm emphasizing neuromuscular control as the critical determinant of therapeutic outcomes over isolated strength gains in KOA management [95].

The precise mechanisms underlying traditional forward-walking aerobic exercise for pain mitigation in knee osteoarthritis (KOA) remain incompletely characterized, though hypothesized to involve neuromechanical gait modulation.These neurophysiological adaptations typically require sustained implementation to manifest clinically measurable analgesia in KOA populations [112]. Paradoxically, acute exercise responses reveal transient elevation of central pain sensitization following single 30-minute moderate-intensity walking sessions, a phenomenon potentially compromising long-term therapeutic adherence [113].Retroambulation (backward walking) presents biomechanical advantages through patellofemoral joint offloading while maintaining cardiovascular stimulus.Protocol optimization studies suggest retroambulation at 60–70% maximal heart rate achieves significant pain reduction without exacerbating joint stress when integrated into structured aerobic conditioning programs [114]. Arun et al. [62] further demonstrate enhanced clinical outcomes when combining retroambulation with conventional forward-walking protocols compared to unimodal approaches. Current evidence positions both ambulatory modalities as moderately effective interventions within KOA therapeutic hierarchies, though their efficacy rankings remain provisional due to insufficient high-quality randomized trials. While mechanistically plausible, the clinical utility of forward or retrograde walking as monotherapy requires rigorous validation through multicenter longitudinal studies with standardized outcome metrics.

This investigation adopted a tripartite outcome framework comprising the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), Visual Analog Scale (VAS), and Timed Up and Go (TUG) test to standardize cross-trial comparisons and ensure methodological coherence. While secondary assessment metrics were not systematically integrated into the analytical matrix, this methodological decision does not negate their clinical relevance in comprehensive KOA evaluation. Across all comparative analyses, structured aerobic exercise interventions demonstrated consistent superiority over control conditions in functional and symptomatic outcomes. Furthermore, the study’s incorporation of resistance training elements within comparator protocols aligns with Goh et al.‘s [4] methodological paradigm for exercise trials in degenerative joint disease, reinforcing the validity of observed therapeutic hierarchies.

Based on the SUCRA ranking and safety data, for the majority of KOA patients, Pilates and Tai Chi exercises are the preferred choices. Alternative options include weight-loss walking, yoga and Baduanjin.

Limitations

Abundant meta-analyses exist on the treatment of Knee Osteoarthritis (KOA) with exercise therapy, yet a gap remains in directly comparing different aerobic exercises for KOA treatment. The aim of this study was to perform a comprehensive network meta-analysis of all available randomized controlled trials on aerobic exercise treatments for KOA to identify the optimal therapeutic approach and provide evidence-based medical evidence for clinical decision-making. Despite these efforts, the study has its limitations.First, significant clinical heterogeneity was observed across studies, particularly in exercise protocols, Although random-effects models and subgroup analyses were used to address this variability, residual heterogeneity may still affect the pooled estimates.Second, the methodological quality of included trials was variable. While sensitivity analyses excluding high-risk-of-bias studies yielded consistent results the overall strength of evidence remains constrained by these limitations. These limitations highlight the need for future large-scale RCTs with standardized exercise protocols, stratified randomization by disease severity, and long-term follow-up. Individual participant data meta-analysis (IPDMA) could further elucidate how patient-level factors mediate treatment responses.

Conclusion

This study demonstrates that Pilates appears to be the most effective aerobic exercise modality for managing knee osteoarthritis (KOA), particularly in enhancing overall functional outcomes. Tai Chi exhibited the greatest efficacy in reducing pain intensity, as quantified by the VAS. Based on these findings, Pilates and Tai Chi should be prioritized as primary therapeutic interventions for the majority of KOA patients. In the future, we will supplement and expand the data to enhance the reliability of the research.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: YL, XC. Data curation: YL，XC，HB-G，LC，LY-Z, SQ-L. Formal analysis: YL，XC，HB-G，LC，LY-Z, SQ-L. Funding acquisition: YL, XC.Investigation: YL，XC，HB-G，LC，LY-Z, SQ-L.Methodology: YL, XC. Project administration: YL，XC，HB-G，LC，LY-Z, SQ-L. Resources: YL，XC，HB-G，LC，LY-Z, SQ-L.Software: YL, XC.Supervision: YL，XC，HB-G，LC，LY-Z, SQ-L. Validation: YL，XC，HB-G，LC，LY-Z, SQ-L. Visualization: YL，XC，HB-G，LC，LY-Z, SQ-L. Writing– original draft: YL, XC.Writing– review & editing: YL, XC.

Funding

No funding.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethical approval

Not applicable.

Consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.