Introduction
Knee osteoarthritis (KOA) is a highly prevalent chronic degenerative joint disease that imposes a substantial burden on global healthcare systems, particularly amidst the accelerating aging population.1,2 Clinically, KOA is characterized by progressive joint pain, stiffness, and functional limitation, which collectively lead to a significant deterioration in patients’ quality of life.3,4 Current mainstream treatments primarily rely on pharmacological interventions and intra-articular injections to manage symptoms; however, their long-term efficacy is often limited by potential adverse effects, highlighting the need for safer, non-pharmacological alternatives.5–7 In this context, exercise therapy has garnered increasing attention as an effective non-pharmacological intervention, particularly with respect to Traditional Chinese Exercises, prompting the need for comparative studies to determine their relative efficacy.8–10
Tai Chi (TC) and Baduanjin (BDJ), two representative forms of Traditional Chinese Exercises, are widely recognized as valuable non-pharmacological strategies for KOA management, with accumulating evidence supporting their efficacy in pain relief and functional recovery.11,12 Previous studies have indicated that regular practice of these exercises can yield clinical outcomes comparable to, or even superior to, standard rehabilitation in specific domains.13–16 Although both are effective, they involve distinct movement patterns and mechanisms, with TC emphasizing dynamic balance and weight shifting and BDJ focusing on stretching and breath regulation. Despite widespread clinical application of both exercises, few randomized head-to-head trials have compared TC and BDJ for KOA rehabilitation. Evidence from direct comparisons would help inform exercise selection in clinical practice.
Additionally, the absence of a HE control in many prior studies limits our understanding of their specific efficacy beyond general advice. Given that TC and BDJ produce benefits through comparable physiological pathways including muscle strengthening, balance improvement, and neuromotor control, we anticipated that both exercise interventions would yield greater clinical improvements than HE, while meaningful differences between the two exercise forms were not expected. Published trials investigating mind-body exercise for knee osteoarthritis have commonly employed training schedules of three weekly 60-minute sessions, and we adopted this well-established regimen for the present study.11,12,14 Consequently, this present three-arm randomized controlled trial compares TC, BDJ, and HE, and may provide useful evidence to inform non-pharmacological treatment strategies for KOA.
The present manuscript constitutes the final results report of a three-arm randomized controlled trial, for which the prespecified study protocol has been published previously.17 The published protocol outlined the trial design, eligibility criteria, planned interventions, and prespecified statistical plan prior to participant recruitment. In contrast, this submission presents the complete dataset of the completed trial, including intergroup comparisons of primary and secondary outcomes, safety profiles, and 48-week longitudinal follow-up findings.
Patients and Methods
Study Design
Conducted in compliance with the Declaration of Helsinki, this single-center, three-arm, open-label randomized controlled trial received approval from the Ethics Committee of Zhongshan Hospital of Traditional Chinese Medicine Affiliated to Guangzhou University of Traditional Chinese Medicine (approval number: 2024ZSZY-LL-KY-289 (00)). To ensure methodological rigor and transparent reporting, the study design aligned with the SPIRIT 2013 statement, and the final manuscript was prepared according to the CONSORT and TIDieR guidelines.18–20 Written informed consent was obtained from every participant following a comprehensive explanation of the study’s aims, procedures, and associated risks and benefits. To strictly safeguard confidentiality, all personal data will be securely managed by authorized personnel.
Participants received no financial remuneration, and all study interventions (TC, BDJ, and HE) were provided at no cost.
Trial Population
Eligible participants were recruited through multiple ways, including publicity on the hospital official website and WeChat, poster placement in clinical departments, as well as referrals from attending physicians. All candidates fulfilled the following inclusion criteria to take part in this research:5,21,22 1) Receive a knee osteoarthritis diagnosis that conforms to the 1995 clinical diagnostic standards formulated by the American College of Rheumatology; 2) Be aged between 45 and 70 years old; 3) Present Kellgren-Lawrence grade II or III degenerative changes on knee radiographic images; 4) Suffer from persistent knee pain lasting at least three months; 5) Have a knee pain score above 4 points on the 0–10 visual analogue scale (VAS); 6) Possess the capacity to sign written informed consent voluntarily. Any participant satisfying any of the following conditions were excluded from the trial: 1) Those with a prior knee arthroplasty history or scheduled knee replacement surgery; 2) Patients whose knee pain originates from rheumatoid arthritis, gout or other secondary joint disorders; 3) Individuals who underwent knee arthroscopic surgery within the past 12 months or received intra-articular injections in the previous six months; 4) People complicated with severe acute or chronic organic diseases and severe mental illnesses, as well as pregnant and breastfeeding women; 5) Those participating in other clinical trials within three months prior to recruitment; 6) Candidates unable to finish the full trial protocol due to frequent relocation, unstable mental status and other relevant factors.
Randomization and Blinding
Participants who met the inclusion criteria were allocated to the TC, BDJ, or HE groups with a 1:1:1 ratio. An independent researcher, who was not involved in patient recruitment or assessment, generated the random allocation sequence using a computer-based random number generator. The random sequence was stored in sealed, opaque envelopes to ensure allocation concealment; recruiting staff had no access to the upcoming group assignments before participant enrolment.
Group assignment was then conducted sequentially based on the order of enrollment.
Given the behavioral nature of the interventions, it was impractical to blind the participants and the instructors delivering the TC and BDJ programs. However, to ensure objectivity, the research personnel responsible for outcome assessments and the statistician performing the final data analysis were kept blinded to the group allocations throughout the entire study period.
Treatments
Tai Chi Intervention
The Tai Chi training sessions were conducted in a quiet, spacious, and well-ventilated room to provide a comfortable and safe environment for physical activity. Participants in this group practiced a simplified 24-form Tai Chi routine, a standardized sequence widely recognized for its safety and efficacy in rehabilitation settings. The intervention consisted of supervised group sessions held three times per week, with each session lasting approximately 60 minutes, for a total duration of 12 weeks.
All sessions were led by certified Tai Chi instructors who possessed at least five years of teaching experience and had undergone specific training on the study protocol to ensure instructional consistency.
Each 60-minute session was systematically structured into three phases: a 10-minute warm-up involving gentle joint mobilization and dynamic stretching, a 40-minute core Tai Chi practice, and a 10-minute cool-down with relaxation and breathing exercises. During the core practice, participants were guided through specific, foundational movements characteristic of the 24-form. These included “Parting the Wild Horse’s Mane” to promote weight shifting and lower-body stability, “White Crane Spreads Its Wings” to enhance upper-body coordination and balance, and “Cloud Hands”, which emphasizes continuous, flowing lateral weight transfers and core engagement.
Instructors placed particular emphasis on proper body alignment, deep diaphragmatic breathing, and mental focus throughout these sequences. To ensure adherence and safety, instructors closely monitored the participants’ biomechanics and provided individualized corrections as needed. Additionally, participants were encouraged to practice at home for at least 15 minutes daily to reinforce motor learning, although this home practice was not strictly monitored.
Baduanjin Intervention
Participants in the Baduanjin group performed the traditional Health Qigong Baduanjin routine, a widely practiced mind-body exercise consisting of eight distinct movements designed to harmonize body, breath, and mental focus. The training took place in a dedicated, well-ventilated studio to ensure a comfortable and focused atmosphere. The intervention protocol mirrored that of the Tai Chi group, comprising supervised sessions three times per week for 60 minutes each, over a continuous period of 12 weeks.
Instruction was provided by certified Qigong instructors with a minimum of five years of teaching experience, all of whom underwent standardized protocol training prior to the study to ensure instructional consistency. Each session began with a 10-minute warm-up involving gentle joint mobilization and stretching, followed by 40 minutes of core Baduanjin practice, and concluded with a 10-minute cool-down and relaxation period.
During the core practice, participants were guided through specific, representative movements. These included “Two Hands Hold Up the Heavens” to promote spinal extension and regulate the triple energizer (San Jiao); “Drawing the Bow to Shoot the Eagle” to expand the chest and enhance pulmonary function; “Single Arm Raise” to stimulate digestive function through alternating vertical stretching; and “Swaying the Head and Shaking the Tail”, which emphasizes spinal rotation and pelvic mobility to relieve internal heat. Throughout these sequences, instructors emphasized the precise coordination of movement with deep diaphragmatic breathing, providing real-time corrections to ensure proper biomechanical alignment and safety.
Participants were also encouraged to practice independently at home using provided instructional videos to reinforce motor learning, although only the supervised sessions were counted towards the primary intervention dose.
Health Education Intervention
Participants allocated to the health education control group attended weekly face-to-face sessions consistent with the visit frequency of the Tai Chi and Baduanjin groups to eliminate contact bias, but received no supervised mind-body exercise training throughout the trial.
All weekly sessions focused solely on standardized osteoarthritis self-management education delivered by professional TCM rehabilitation instructors. Core content included joint protection strategies, weight control, lifestyle adjustments, cold and damp prevention, balanced dietary advice, and simple non-pharmacological pain relief techniques. Standardized handouts and instructional videos were distributed at the baseline session for participants to review at home.
During each weekly visit, researchers only reinforced disease management knowledge, answered participants’ questions about joint discomfort, and completed regular outcome assessments. No Tai Chi or Baduanjin posture guidance, movement practice, or breathing training was provided. Subjects were instructed to maintain their usual daily physical activity and refrain from practicing structured traditional Chinese therapeutic exercises.
This group served as the control to separately compare the intervention efficacy of Tai Chi and Baduanjin for osteoarthritis.
Outcome Measurements
For participants with bilateral osteoarthritis, all outcome assessments will focus on the limb presenting more severe symptoms at baseline. Standardized evaluations are scheduled at four time points: baseline, week 12, week 24, and week 48 after intervention initiation.
The primary endpoint of this trial is the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC).23 This scale contains three independent subdomains to measure joint pain, morning stiffness, and daily physical function, with dedicated scoring ranges for each dimension to quantify osteoarthritis-related complaints.
Secondary evaluation indicators include the 36-item Short-Form Health Survey (SF-36), Berg Balance Scale (BBS), Visual Analog Scale (VAS), and six-minute walk test (6MWT).24–28 The SF-36 questionnaire evaluates participants’ physical and mental health-related quality of life across multiple dimensions. The Berg Balance Scale assesses static and dynamic balance function, with total scores ranging from 0 to 56; higher scores correspond to better balance performance. The six-minute walk test objectively reflects functional exercise endurance by recording the total continuous walking distance completed within six minutes.
Any adverse events occurring over the whole trial will be closely tracked and thoroughly recorded at each follow-up session to enable comprehensive safety monitoring.
Statistical Analysis
The null hypothesis (H0) posits that no statistically significant intergroup differences in primary outcome changes exist across the TC, BDJ, and HE arms. This study is designed as an exploratory randomized controlled trial. No formal prospective sample-size calculation was performed for this trial.
We aimed to recruit 40 participants per group (total planned sample of 120 participants) to obtain sufficient data for exploratory analyses. Primary statistical analyses were performed under the modified intention-to-treat (mITT) framework (n=116; TC n=38, BDJ n=39, HE n=39), which included all randomized participants with at least one post-baseline measurement. Chi-square tests were adopted to compare categorical baseline characteristics across the TC, BDJ and HE groups.
Besides, mixed-effects model for repeated measures (MMRM) was applied to analyze longitudinal changes in continuous outcome indicators including WOMAC subscale scores, SF-36 scores, Berg Balance Scale results and six-minute walk test distances over multiple follow-up time points. Baseline measurements were incorporated as covariates. The mixed-effects model for repeated measures was fitted using restricted maximum-likelihood (REML) estimation to handle missing longitudinal data under the missing-at-random assumption; mean imputation was not applied for the primary efficacy analysis.
Intervention grouping, assessment time point, and the interaction term of grouping×time were specified as fixed effects, and individual participants were designated as random effects within the mixed-effects model for repeated measures. This identical analytical framework was uniformly adopted to process all aforementioned continuous measurement data. The per-protocol dataset (n=110), consisting of participants completing full 48-week follow-up, was used for sensitivity analysis.
To compare the incidence of exercise-related adverse reactions across groups, the chi-square test was utilized. All participants who received at least one standardized intervention session were included in the safety analysis. All statistical computations were performed with SPSS 26.0 software for Windows, and GraphPad Prism 9.5 was used to generate all outcome visualization figures.
Bonferroni correction was applied for the three pairwise group-to-group comparisons of the primary outcome. All reported p-values for these pairwise comparisons are adjusted after Bonferroni correction.
Results
Patients
Between June 1, 2025 and June 30, 2026, 167 eligible subjects completed trial screening procedures. After screening evaluation, 120 participants were selected to take part in the present study and randomly allocated to three independent groups (see Figure 1). In detail, the three arms consisted of 40 patients in the TC group, 40 in the BDJ group, and 40 in the HE group.
Every participant underwent no less than one treatment session and was included for safety outcome analysis. After 48 weeks of observation, 110 participants completed the entire trial, which was used as the per-protocol sensitivity analysis population. Ten participants withdrew prior to 48-week assessment; six of these withdrawn participants had at least one post-baseline measurement and were therefore retained in the mITT analysis, while four withdrawn participants had no post-baseline efficacy measurements and were excluded from efficacy analyses.
The primary modified intention-to-treat (mITT) analysis set included 116 participants (TC n=38, BDJ n=39, HE n=39), comprising randomized subjects with at least one post-baseline assessment; four randomized participants without any post-baseline measurement were excluded from efficacy analysis. Key baseline information of all mITT subjects is listed in Table 1. Baseline demographic and clinical characteristics of the 116 mITT participants are summarized in Figures 2 and 3.
As shown in Figure 2, pairwise comparison revealed a statistically significant inter-group difference in BMI only between the TC and BDJ groups, whereas age and symptom duration were comparable among the TC, BDJ, and HE groups (all P > 0.05). Figure 3 further demonstrates that gender distribution and Kellgren-Lawrence (KL) grade were well balanced across the three groups (all P > 0.05). Although a baseline discrepancy in BMI was noted, it was considered clinically acceptable and would be adjusted for in subsequent mixed-effects models for repeated measures.Table 1Baseline CharacteristicsCharacteristicTC (n=38)BDJ (n=39)HE (n=39)Gender, n (%)Male9(23.7)9 (23.7)11 (28.2)Female29 (76.3)28 (71.8)28 (71.8)Age, mean ± SD, yr60.9 ±7.061.4 ±6.659.8 ±6.8BMI, mean ± SD25.7 ±4.724.3 ±3.324.4 ±2.4Symptom duration, mean ± SD, yr11.3 ±9.813.7 ±10.412.1 ±8.0KL grade, n (%)II22 (57.9)25 (64.1)23 (59.0)III16 (42.1)14 (35.9)16 (41.0)WOMAC Index, mean ± SD43.7 ±7.642.8 ±4.042.6 ±4.0Pain subscale, mean ± SD9.6 ±2.59.8 ±2.09.7 ±1.9Stiffness subscale, mean ± SD3.9 ±1.54.0 ±1.14.0 ±1.1Function subscale, mean ± SD30.2 ±7.529.0 ±3.429.0 ±3.0VAS, mean ± SD6.1 ±1.96.3 ±2.06.5 ±2.1SF-36, mean ± SDPF Physical Functioning44.2 ±4.443.9 ±5.044.0 ±5.9RP Role-Physical37.5 ±3.837.3 ±4.436.8 ±4.8BP Bodily Pain39.2 ±4.438.7 ±4.538.6 ±5.4GH General Health37.4 ±4.837.6 ±5.238.2 ±5.3VT Vitality45.3 ±4.344.8 ±4.745.1 ±5.2SF Social Functioning51.7 ±2.952.5 ±3.852.4 ±4.2RE Role-Emotional58.1 ±3.657.5 ±4.357.0 ±4.7MH Mental Health55.9 ±3.655.9 ±4.556.3 ±4.6Berg Balance Scale,mean ± SD39.3 ±4.839.9 ±3.639.7 ±3.7Six-Minute Walk Test,mean ± SD391.2 ±56.5398.6 ±44.5397.1 ±43.5Notes: Data are presented as number (%) for categorical variables and mean ±standard deviation(SD) for continuous variables.Abbreviations: TC, Tai Chi group; BDJ, Baduanjin group; HE, health education control group; KL, Kellgren-Lawrence; BMI, body mass index.Figure 1Participant flow diagram.
Of 167 screened patients, 120 were randomized. Four participants without post-baseline measurements were excluded. The primary mITT analysis set comprised 116 participants (TC = 38, BDJ = 39, HE = 39), including six with partial follow-up data.
The bottom box shows the per-protocol sensitivity set (n = 110) with complete 48-week follow-up.A participant flowchart from screening through randomization, follow up and per protocol analysis.Flowchart sequence: 167 patients screened. Exclusions: 47 not meeting inclusion criteria, including 14 VAS score of knee pain less than 4 points, 13 joint pain induced by other primary diseases, 7 age fails to meet trial enrollment criteria, 5 complicated with other severe systemic diseases, 3 simultaneously enrolled in other clinical trials and 5 cognitive impairment, unable to finish questionnaires independently. Randomized, n equals 120.
Three allocation branches: Allocated to TC, n equals 40, received allocated intervention; Allocated to BDJ, n equals 40, received allocated intervention; Allocated to HE, n equals 40, received allocated intervention. TC branch notes: 1 relocated out of town, 1 missed assessments due to work conflicts; 38 included in the intention to treat set; 1 quit due to poor symptom improvement, 1 lost long term follow up contact; 36 completed 48 week follow up. BDJ branch notes: 1 subject missed all post baseline evaluations due to long distance travel; 39 included in the intention to treat set; 2 withdrew before 48 week follow up finished; 37 completed 48 week follow up.
HE branch notes: 1 subject could not complete post baseline tests for transportation inconvenience; 39 included in the intention to treat set; 2 withdrew before 48 week endpoint; 37 completed 48 week follow up. Bottom box: Per protocol sensitivity analysis, n equals 110.Figure 2Comparison of baseline demographic and clinical characteristics among the three groups. TC, Tai Chi group; BDJ, Baduanjin group; HE, health-education group. *: indicates statistically significant difference; ns: non significant difference.A bar graph showing age, body mass index and symptom duration across three study groups.A grouped bar graph compares three measures across three groups labeled in a legend as TC group, BDJ group and HE group.
Horizontal axis label: Age, BMI, Symptom duration. No unit is shown on the horizontal axis. Vertical axis: numeric scale from 0 to 100, with labeled ticks at 0, 20, 40, 60, 80 and 100.
No vertical axis label or unit is shown. Age group of bars: three bars with error bars. All three bars are near 60 on the vertical scale.
Brackets above the bars show pairwise comparisons labeled ns, ns and ns and a larger bracket spanning the group also labeled ns. BMI group of bars: three bars with error bars. The TC group bar is near 25.
The BDJ group bar is near 23. The HE group bar is near 23. Brackets above the bars show a comparison labeled asterisk between TC group and BDJ group, a comparison labeled ns between BDJ group and HE group and a larger bracket spanning TC group and HE group labeled asterisk.
Symptom duration group of bars: three bars with error bars. The TC group bar is near 9. The BDJ group bar is near 9.
The HE group bar is near 8. Brackets above the bars show pairwise comparisons labeled ns, ns and ns and a larger bracket spanning the group also labeled ns. Two vertical dashed divider lines separate the three bar groups.Figure 3Comparison of gender and Kellgren-Lawrence grade distribution among the three groups.
Intergroup comparisons were performed using the Chi-square test. No significant differences were observed among the three groups (P>0.05). ns: non significant difference.A grouped bar graph showing participant counts by gender and Kellgren Lawrence grade across three groups.A grouped bar graph comparing participant distributions across three groups for gender and Kellgren Lawrence grade categories. The y axis label is Number of participants.
The y axis range is 0 to 40 participants. The x axis label is not legible and no unit is shown. The x axis categories are female, male, KL Grade 2 and KLGrade 3.
Bars are vertical and grouped by category with three bars per category. The legend entries are TC group, BDJ group and HE group. Female category values are TC group 29 participants, BDJ group 28 participants and HE group 28 participants.
Male category values are TC group 9 participants, BDJ group 11 participants and HE group 11 participants. KL Grade 2 category values are TC group 22 participants, BDJ group 25 participants and HE group 23 participants. KLGrade 3 category values are TC group 16 participants, BDJ group 14 participants and HE group 16 participants.
Bracketed significance annotations appear above each category for pairwise and grouped comparisons and each annotation reads non significant difference.
Primary Outcome
The mean scores for WOMAC pain, stiffness, and physical function at week 12,24 and 48 varied among the treatment groups (Table 2). The MMRM model demonstrated a significant group-by-time interaction for WOMAC pain subscale (P < 0.001), indicating that the trajectory of pain relief differed significantly across the three groups over the study period. Post-hoc pairwise comparisons showed that both intervention groups achieved significantly greater reductions in pain compared with the HE group.
The estimated marginal mean difference (MD) between TC and HE was −3.22 (95% CI: −3.96 to −2.48; P < 0.001), while the MD between BDJ and HE was −3.04 (95% CI: −3.78 to −2.30; P < 0.001). No significant difference in pain reduction was observed between the TC and BDJ groups (P = 0.618). Similar trends were observed for the WOMAC stiffness and physical function subscales (Table 2).
Furthermore, by week 24 and week 48, the TC and BDJ groups demonstrated sustained and significantly greater improvements compared to the HE group across all three subscales (P < 0.001 for all comparisons). For instance, at week 48, the mean WOMAC physical function score was 9.5 ± 6.6 in the TC group and 6.4 ± 5.0 in the BDJ group, significantly lower than the 19.8 ± 3.6 observed in the HE group. Over the course of the study, there were no notable disparities in efficacy between the TC and BDJ groups at any follow-up time point.
As observed in Figure 2, BMI exhibited marginal imbalance across groups at baseline. We adjusted for baseline BMI within the MMRM model, and this adjustment did not materially change the observed between-group differences or statistical conclusions for the primary outcomes. The temporal progression of these scores is graphically represented in Figure 4.Table 2Primary OutcomesTime PointTai Chi (TC, n=38) Mean ± SDBaduanjin (BDJ, n=39) Mean ± SDHealth Education (HE, n=39) Mean ± SDp valueWOMAC Pain SubscaleBaseline9.6 ± 2.59.8 ± 2.09.7 ± 1.90.94112w5.6 ± 1.75.7 ± 1.38.6 ± 1.8<0.00124w3.5 ± 1.43.6 ± 1.18.5 ± 1.8<0.00148w1.9 ± 1.31.8 ± 0.97.5 ± 1.9<0.001WOMAC Stiffness SubscaleBaseline3.9 ± 1.54.0 ± 1.14.0 ± 1.10.94712w3.0 ± 1.42.9 ± 1.23.7 ± 1.30.01324w2.3 ± 1.52.1 ± 1.23.5 ± 1.4<0.00148w1.7 ± 1.51.3 ± 1.02.9 ± 1.2<0.001WOMAC Function SubscaleBaseline30.2 ± 7.529.0 ± 3.429.0 ± 3.00.55012w20.3 ± 7.517.7 ± 5.225.6 ± 3.5<0.00124w13.9 ± 7.912.3 ± 5.023.8 ± 3.5<0.00148w9.5 ± 6.66.4 ± 5.019.8 ± 3.6<0.001Abbreviations: WOMAC, Western Ontario and McMaster Universities Osteoarthritis Index; TC, Tai Chi group; BDJ, Baduanjin group; HE, health education control group.Figure 4Temporal changes in WOMAC subscale scores across the three treatment groups over the 48-week study period. abcd: represent the point-in-time differences comparison using the letter notation method.A set of three line graphs showing WOMAC pain, stiffness and function subscale scores over time.A) Line graph of WOMAC Pain Subscale across baseline, week 12, week 24 and week 48 for three groups with error bars and letter annotations.
Horizontal axis label: baseline, week 12, week 24, week 48. Unit: not shown. Vertical axis label: WOMAC Pain Subscale.
Unit: not shown. Vertical axis range: 0 to 12. Legend entries: TC group, BDJ group, HE group.
TC group coordinate pairs: baseline: approximately 9.5, week 12: approximately 5.5, week 24: approximately 3.5, week 48: approximately 1.8. BDJ group coordinate pairs: baseline: approximately 9.6, week 12: approximately 5.6, week 24: approximately 3.6, week 48: approximately 1.7. HE group coordinate pairs: baseline: approximately 9.6, week 12: approximately 8.5, week 24: approximately 8.4, week 48: approximately 7.4.
Letter annotations near points: baseline shows a near all three groups; week 12 shows b near all three groups; week 24 shows b near HE group and c near TC group and BDJ group; week 48 shows c near HE group and d near TC group and BDJ group. B) Line graph of WOMAC stiffness Subscale across base, week 12, week 24 and week 48 for three groups with error bars and letter annotations. Horizontal axis label: base, week 12, week 24, week 48.
Unit: not shown. Vertical axis label: WOMAC stiffness Subscale. Unit: not shown.
Vertical axis range: 0 to 6. Legend entries: TC group, BDJ group, HE group. TC group coordinate pairs: base: approximately 3.9, week 12: approximately 2.9, week 24: approximately 2.2, week 48: approximately 1.6.
BDJ group coordinate pairs: base: approximately 4.0, week 12: approximately 2.8, week 24: approximately 2.0, week 48: approximately 1.2. HE group coordinate pairs: base: approximately 3.9, week 12: approximately 3.6, week 24: approximately 3.4, week 48: approximately 2.8. Letter annotations near points: base shows a near all three groups; week 12 shows b near all three groups; week 24 shows b near HE group and c near TC group and BDJ group; week 48 shows c near HE group and d near TC group and BDJ group.
C) Line graph of WOMAC Function Subscale across base, week 12, week 24 and week 48 for three groups with error bars and letter annotations. Horizontal axis label: base, week 12, week 24, week 48. Unit: not shown.
Vertical axis label: WOMAC Function Subscale. Unit: not shown. Vertical axis range: 0 to 40.
Legend entries: TC group, BDJ group, HE group. TC group coordinate pairs: base: approximately 30, week 12: approximately 20, week 24: approximately 14, week 48: approximately 9. BDJ group coordinate pairs: base: approximately 29, week 12: approximately 18, week 24: approximately 12, week 48: approximately 6.
HE group coordinate pairs: base: approximately 29, week 12: approximately 26, week 24: approximately 24, week 48: approximately 20. Letter annotations near points: base shows a near all three groups; week 12 shows b near all three groups; week 24 shows b near HE group and c near TC group and BDJ group; week 48 shows c near HE group and d near TC group and BDJ group.
Secondary Outcomes
Figure 5 presents the longitudinal changing trajectories of all secondary outcome indicators via line graphs. All secondary indicators were measured at baseline, week 12, week 24 and week 48 for intergroup and intra-group comparisons.Figure 5Trends in secondary outcome measures during the intervention period and follow-up. (A) Change in the curve of VAS pain score within 48 weeks.
(B) Change in the curve of Berg balance scale within 48 weeks. (C) Change in the curve of 6-minute walk test within 48 weeks. (D) Change in the curve of SF-36 Physical Component Summary within 48 weeks.
(E) Change in the curve of SF-36 Mental Component Summary within 48 weeks. abcd: represent the point-in-time differences comparison using the letter notation method.Line graphs showing secondary outcome changes over time for TC group, BDJ group and HE group.A) Line graph with error bars and three series labeled TC group, BDJ group and HE group. The x-axis label is baseline, week 12, week 24, week 48. The y-axis label is VAS pain score, unit not shown, with a range from 0 to 10.
TC group values: baseline about 6.2, week 12 about 4.2, week 24 about 3.2, week 48 about 3.0. BDJ group values: baseline about 6.2, week 12 about 4.3, week 24 about 3.2, week 48 about 3.1. HE group values: baseline about 6.3, week 12 about 6.0, week 24 about 5.6, week 48 about 5.5.
Letter annotations near points include a, b, c, d. B) Line graph with error bars and three series labeled TC group, BDJ group and HE group. The x-axis label is baseline, week 12, week 24, week 48.
The y-axis label is Berg balance scale, unit not shown, with a range from 30 to 60. TC group values: baseline about 39, week 12 about 46, week 24 about 49, week 48 about 51. BDJ group values: baseline about 40, week 12 about 47, week 24 about 51, week 48 about 52.
HE group values: baseline about 40, week 12 about 41, week 24 about 42, week 48 about 44. Letter annotations near points include a, b, c, d. C) Line graph with error bars and three series labeled TC group, BDJ group and HE group.
The x-axis label is baseline, week 12, week 24, week 48. The y-axis label is 6-minute walk test, unit not shown, with a range from 300 to 600. TC group values: baseline about 390, week 12 about 460, week 24 about 500, week 48 about 520.
BDJ group values: baseline about 395, week 12 about 470, week 24 about 510, week 48 about 530. HE group values: baseline about 395, week 12 about 405, week 24 about 415, week 48 about 425. Letter annotations near points include a, b, c, d.
D) Line graph with error bars and three series labeled TC group, BDJ group and HE group. The x-axis label is baseline, week 12, week 24, week 48. The y-axis label is Physical Component Summary of SF-36, unit not shown, with a range from 25 to 50.
TC group values: baseline about 32, week 12 about 36.5, week 24 about 39, week 48 about 41.5. BDJ group values: baseline about 32, week 12 about 37, week 24 about 41, week 48 about 44. HE group values: baseline about 32.5, week 12 about 33, week 24 about 33.5, week 48 about 34.
Letter annotations near points include a, b, c, d. E) Line graph with error bars and three series labeled TC group, BDJ group and HE group. The x-axis label is baseline, week 12, week 24, week 48.
The y-axis label is the Mental Component Summary of SF-36, unit not shown, with a range from 35 to 65. TC group values: baseline about 44, week 12 about 50, week 24 about 53, week 48 about 57. BDJ group values: baseline about 44, week 12 about 48, week 24 about 52, week 48 about 56.
HE group values: baseline about 44, week 12 about 45, week 24 about 45.5, week 48 about 45. Letter annotations near points include a, b, c, d.
Compared with baseline, VAS scores were markedly reduced at week 12 in both the TC and BDJ groups (all p < 0.001), accompanied by sustained pain relief up to week 48. A small but statistically significant decrease in VAS was also observed in the HE group at week 12 (MD = −0.52, 95% CI −0.94 to −0.10, p = 0.013). No significant differences in VAS scores were found between the TC and BDJ groups at any time point (all p > 0.05).
From week 12 through week 48, both TC and BDJ groups demonstrated suggested lower values VAS scores than the HE group (all p < 0.05). Post-hoc pairwise comparisons were performed using the LSD method.
Compared with baseline, Berg Balance Scale scores increased significantly at week 12 in both the TC and BDJ groups, with progressive improvements sustained up to week 48 (all p < 0.001). The HE group also exhibited a slight but statistically significant elevation in balance performance at week 12 (p = 0.003), although the magnitude of improvement was far smaller than the two exercise groups. No significant intergroup difference was detected between the TC and BDJ groups at any follow-up time point (all p > 0.05).
Starting from week 12, both TC and BDJ groups achieved suggested higher values Berg scores relative to the HE group at week 12, week 24 and week 48 (all p < 0.001). Post-hoc pairwise comparisons were performed using the LSD method.
Longitudinal changes in six-minute walk test (6MWT) distance differed significantly among the three groups. Substantial improvements in walking endurance were observed in the TC and BDJ groups starting at week 12, with progressive gains maintained until week 48 (all p < 0.001). Although the HE group achieved a modest increase in walking distance at week 12 (p<0.001), no further meaningful improvement was detected from week 12 to week 24 (p = 0.121).
Walking capacity was comparable between the TC and BDJ groups at all time points (all p > 0.05). From week 12 through week 48, both exercise groups exhibited significantly longer 6-minute walking distances relative to the HE group (all p < 0.001). Post-hoc pairwise comparisons were performed using the LSD method.
For the Physical Component Summary of SF-36, significant within-group improvements from baseline were observed in both the TC and BDJ groups at week 12 (TC: MD, 4.07; 95% CI, 3.54 to 4.60; p<0.001; BDJ: MD, 5.11; 95% CI, 4.58 to 5.64; p<0.001). Conversely, the HE group did not exhibit a statistically significant change at this initial time point (MD, 0.47; 95% CI, −0.02 to 0.96; p=0.108). Although the difference between the two exercise interventions was negligible at week 12 (p=0.086), both TC and BDJ participants achieved significantly higher physical health scores than the HE group at weeks 24 and 48 (all p<0.001).
These data indicate that the physical benefits of mind-body exercises were not only superior to health education but also sustained throughout the follow-up period.
Regarding the Mental Component Summary of SF-36, both the TC and BDJ groups demonstrated significant improvements compared to baseline at week 12 (TC: MD, 4.91; 95% CI, 2.82 to 7.00; p<0.001; BDJ: MD, 4.28; 95% CI, 2.21 to 6.35; p<0.001). The HE group also showed a statistically significant improvement from baseline at week 12 (MD, 1.53; 95% CI, 0.34 to 2.72; p=0.006), although the magnitude of change was smaller than that of the exercise groups. Notably, while there was no significant difference between the TC and BDJ groups at week 12 (p=0.859), both groups exhibited significantly higher scores compared to the HE group by week 24 and week 48 (p<0.001 for both comparisons).
Furthermore, the improvements in mental health observed in the TC and BDJ groups were sustained throughout the 48-week follow-up period, whereas the HE group’s scores plateaued after week 12 with no further significant changes.
Adverse Events
Adverse events associated with traditional Chinese mind-body exercise primarily manifested as transient muscle soreness and slight knee discomfort. No severe adverse events including knee sprain, falls, persistent dizziness or cardiovascular discomfort were detected in any group. The incidence of mild adverse events was 7.5% (3/40) in the Tai Chi group and 7.5% (3/40) in the Baduanjin group.
No exercise-related adverse reactions arose in the health education group. These mild symptoms were fully relieved within 1–3 days through rest or local hot compress. All six mild adverse events were assessed by the investigator as possibly related to the mind-body exercise intervention.
Discussion
Our findings indicated that regular TC and BDJ interventions effectively alleviate pain and improve physical function among patients with KOA, and these favorable therapeutic effects were sustained across the intervention and follow-up periods. Notably, both mind-body exercises produced superior clinical benefits relative to health education alone. Of the three groups enrolled, the TC and BDJ groups attained significantly better overall efficacy than the HE group.
Specifically, pain relief and daily living capacity improved to a markedly greater extent in the two exercise groups at all assessment time points, while no statistically significant difference in therapeutic outcomes was detected between TC and BDJ.
Previous studies have validated the symptom-relieving effects of TC and BDJ for patients with knee osteoarthritis.29–33 However, most existing trials lack direct head-to-head comparisons between these two traditional Chinese mind-body exercises, and many combined auxiliary physical therapy or analgesic medication, making it difficult to isolate their independent rehabilitative effects. Furthermore, the majority of prior investigations only adopted short-term 12-week intervention windows, with limited evidence regarding long-term sustained therapeutic effects. The present trial adopted a rigorous three-arm design with standardized unified training protocols, without any additional adjuvant treatments throughout the full 48-week observation period.
The added health education group also allowed us to separate exercise-specific improvements from general disease self-management benefits. Consistent and prominent reductions in knee pain, joint stiffness, and functional impairment were observed across all follow-up assessments for both exercise groups. These lasting therapeutic gains may partly be explained by the consistent muscle-strengthening training embedded within both exercise regimens.
Unlike general casual daily activity, both TC and BDJ involve repeated, controlled movements that steadily strengthen the quadriceps femoris and surrounding knee soft tissues. Stronger periarticular muscle groups provide better joint support, reduce abnormal mechanical stress on the degenerative knee, and thereby mitigate core KOA symptoms.34,35 Although TC and BDJ feature distinct movement sequences and training focuses, both place equal emphasis on lower limb muscle conditioning and balance practice, which ultimately leads to similar overall rehabilitation outcomes.
This study combined subjective symptom questionnaires and objective functional assessments to systematically evaluate rehabilitative effects. Knee pain, stiffness and impaired physical function are the dominant symptoms of knee osteoarthritis, which greatly restrict patients’ daily walking and household activities. Serial follow-up assessments showed that compared with health education, both TC and BDJ significantly alleviated knee pain measured by the VAS scale and lowered total WOMAC scores throughout the 48-week observation period, findings with important implications for clinical rehabilitation practice.
Besides joint symptoms, we further adopted the Berg Balance Scale and 6-minute walking test to objectively reflect lower-limb motor capacity. Both exercise interventions steadily improved balance performance and walking endurance over time, indicating that regular mind-body exercise strengthens knee peripheral muscle stability and mitigates functional decline. Meanwhile, elevated SF-36 scores revealed additional holistic benefits in physical and mental quality of life brought by long-term standardized training.
No obvious disparity was observed between TC and BDJ across these key indicators.
According to our research, middle-aged and elderly patients with knee osteoarthritis can obtain clear rehabilitative benefits from both TC and BDJ training. The two traditional mind-body exercises both yielded steady improvements in knee pain, joint stiffness, physical function, balance capacity and walking endurance, with favorable effects sustained throughout the entire 48-week follow-up period. Although TC and BDJ differ greatly in movement routines and training styles, no statistically significant difference in rehabilitative outcomes was detected between the two interventions for patients with KOA.
Meanwhile, long-term standardized practice of both regimens also effectively lifted participants’ physical and mental quality of life beyond mere joint symptom relief. Compared with routine health education alone, structured traditional Chinese exercise delivers more persistent gains in alleviating degenerative knee discomfort and preserving daily mobility. Collectively, this study verifies that both TC and BDJ are reliable, feasible non-pharmacological rehabilitation alternatives for patients with knee osteoarthritis in clinical practice.
Notably, between-group differences may be partially influenced by unequal weekly contact frequency and therapist attention across study arms, given that the health-education group functioned as an active attention control instead of a no-treatment condition. When choosing between TC and BDJ for individual patients, clinicians should also take patient preference, local resource availability, and personal functional limitations into consideration.
Study Limitations
There are several limitations inherent to the present study that deserve acknowledgment. Firstly, owing to the characteristics of mind-body exercise intervention, double-blinding design could not be implemented; both participants and instructors were aware of the assigned group, which might introduce subjective reporting bias. Secondly, this trial was conducted within a single clinical center, and the participant population was geographically confined, potentially limiting the generalizability of our findings; future multicenter trials with broader recruitment will be required to replicate and verify these outcomes.
Additionally, although we extended the follow-up period to 48 weeks, longer-term observation beyond one year is still lacking, which cannot fully clarify the sustained maintenance effect and long-term adherence trajectory of the two exercise regimens. Furthermore, we did not set a dedicated placebo exercise group, so we were unable to quantitatively distinguish the specific therapeutic benefits of standardized TC and BDJ from general psychological and placebo effects. Moreover, our sample included patients aged 45–70 years with KL grade II–III knee osteoarthritis; therefore our findings may not be generalizable to patients with advanced KL grade IV KOA or individuals with substantial multiple comorbidities.
Finally, this exploratory trial lacked a formal prospective sample-size calculation, which constitutes an additional methodological limitation. While our findings provide useful exploratory data and effect-size references for designing future confirmatory studies, firm definitive conclusions about clinical efficacy will require larger, adequately powered randomized controlled trials.