Tai Ji for stroke rehabilitation: a randomized controlled trial on improving upper limb function and cortico-muscular coupling in stroke patients — more practice, or a specific effect?

Clinical Question (PICO)

In adults aged 18–70 years with a first stroke 2 weeks to 12 months earlier, residual upper-limb movement, and sufficient balance to stand and walk independently, does adding supervised, adapted eight-form Tai Ji to conventional rehabilitation improve upper-limb motor impairment and activities of daily living after four weeks, compared with conventional rehabilitation alone?

The trial included ischemic and hemorrhagic stroke. EEG–surface EMG measurements explored directed cortex–muscle connectivity and muscle synergies. No clinical treatment-effect subgroup analysis by stroke type, chronicity, sex, or severity is reported; the healthy reference group was not a randomized treatment arm.

Bottom Line

The authors conclude that adjunctive Tai Ji is a promising, well-tolerated rehabilitation strategy. Among the 70 participants with completed assessments, FMA-UE median scores changed from 43.50 to 55.00 with Tai Ji and from 42.50 to 45.00 with control rehabilitation; the between-group comparison of individual change scores favored Tai Ji (P < 0.001). This supports the combined rehabilitation package, not a proven Tai Ji-specific effect: the intervention arm received twice the scheduled exercise time, and 14 of 84 randomized participants did not complete the study. Clinical results: Table 2.

Design

• Trial type: Parallel-group, single-center randomized trial; participants and instructors were not blinded, while outcome assessors and the final-analysis statistician were blinded.
• N randomized: 84.
  • Conventional rehabilitation plus Tai Ji: 42 randomized; 34 completed.
  • Conventional rehabilitation alone: 42 randomized; 36 completed.
• Additional reference sample: 30 healthy adults, used to describe normative muscle synergy patterns, not to estimate randomized efficacy.
• Randomization: 1:1; SAS-generated sequence prepared by an independent statistician, with sequentially numbered opaque envelopes opened by a separate research assistant. No stratification is described.
• Setting: Fuzhou Hospital, China; the acknowledgements and ethics statement identify the affiliated rehabilitation hospital of Fujian University of Traditional Chinese Medicine.
• Enrollment: Screening from March 2022 to January 2023; report published in June 2026. This is a look-back review, not a newly enrolled trial.
• Follow-up: Four-week intervention, with final assessments within one week of its completion; no longer-term follow-up or mean observation time is reported.
• Analysis: Results are presented for completers, not an explicitly reported intention-to-treat analysis. Within-group paired tests and between-group t-tests or Mann–Whitney tests were used, depending on distribution; categorical variables used chi-square tests. No adjusted clinical outcome model or missing-data imputation is described.
• Primary outcome: FMA-UE, identified as the primary outcome in the sample-size calculation, assessing upper-limb motor impairment. MBI and EEG–EMG-derived measures were also assessed.

Population

Inclusion Criteria

• First-ever stroke, with unilateral cortical lesions and/or subcortical lesions involving motor pathways.
• Age 18–70 years; onset 2 weeks to 12 months before enrollment; right-handed before stroke.
• Affected upper limb at Brunnstrom stage III–V; key upper-limb muscle strength at least MMT grade 3, and Modified Ashworth Scale below 2.
• Standing balance grade at least 2; able to stand independently for more than 5 minutes and walk independently for more than 6 meters.
• Able to tolerate EEG and surface EMG testing.

Exclusion Criteria

• Upper-limb dysfunction due to another brain disease, severe complications such as shoulder–hand syndrome, or major systemic disease/organ failure.
• Unsuitability for EEG–EMG assessment, including relevant electrode-site wounds or skin disease, implanted electronic devices, photosensitive epilepsy, or inability to tolerate testing.

Baseline Characteristics (completers; Tai Ji versus control)

• Age: mean 52.32 ± 15.11 versus 56.36 ± 13.18 years.
• Men/women: 25/9 versus 22/14.
• Time since stroke: median 2.50 months (IQR 1.00–6.00) versus 6.00 months (1.25–11.00); P = 0.067. A nonsignificant baseline test does not eliminate the clinical relevance of this recovery-stage imbalance.
• Hemorrhage/infarction: 14/20 versus 16/20.
• FMA-UE: median 43.50 (IQR 17.00–52.25) versus 42.50 (21.00–54.75).
• MBI: median 80.50 (62.00–95.00) versus 72.00 (56.00–90.50).

Only 84 of 568 screened patients entered the trial. Applicability is therefore to a selected, independently mobile population with residual arm movement, not all stroke survivors. Baseline data: Table 1.

Interventions

• Both groups: Standard medical care and health education plus therapist-delivered, task-oriented upper-limb rehabilitation, stretching, and strengthening for 60 minutes per session, five sessions per week, for four weeks: 20 scheduled hours, calculated from the protocol.
• Tai Ji arm: An additional 60-minute supervised session on the same weekly schedule: 5-minute warm-up, 50 minutes of adapted eight-form practice, and 5-minute cool-down. This adds 20 scheduled hours, yielding 40 hours total, compared with 20 in the control arm.
• Adaptations: Experienced instructors used simplified movements, reduced ranges, wider stance, chair support or hands-on assistance as needed, with supervision and support bars. The program excluded difficult deep stances and complex single-leg balances.

This was neither Tai Ji versus an equal-duration alternative exercise program nor Tai Ji replacing conventional rehabilitation.

Outcomes

Primary Outcome — FMA-UE at four weeks

Values below are median (IQR), not means. The P values compare changes between groups; the paper does not provide the individual change-score distributions or a between-group effect estimate with a 95% confidence interval. Hazard ratios, odds ratios, and risk ratios are not the appropriate summaries for this continuous endpoint.

Measure	Tai Ji: baseline → final	Control: baseline → final	Between-group change P value
FMA-UE total	43.50 (17.00–52.25) → 55.00 (27.25–64.25)	42.50 (21.00–54.75) → 45.00 (24.25–58.00)	< 0.001
Proximal/upper-limb subscore	27.00 (16.00–34.00) → 33.00 (23.00–36.00)	27.00 (18.25–33.00) → 28.50 (21.00–33.00)	< 0.001
Wrist subscore	5.50 (0.00–6.00) → 6.50 (0.00–10.00)	6.00 (1.00–9.00) → 6.00 (1.00–9.00)	0.002
Hand subscore	7.00 (1.75–10.00) → 9.00 (2.00–14.00)	9.00 (3.25–11.00) → 9.00 (4.25–13.00)	0.174
Coordination subscore	0.00 (0.00–1.00) → 3.00 (0.00–6.00)	0.50 (0.00–3.00) → 2.00 (0.00–4.00)	0.020

No prespecified clinical subgroup effect estimates are reported. In particular, the total-score result should not be restated as a significant hand-function advantage: that between-group comparison was negative.

Secondary Outcomes

• Daily activities (MBI): Median 80.50 (62.00–95.00) to 93.00 (76.00–100.00) with Tai Ji, versus 72.00 (56.00–90.50) to 76.50 (63.75–94.75) with control; between-group change P = 0.004. No corresponding effect-size confidence interval is supplied in Table 2.
• Directed connectivity: The methods describe descending motor-to-muscle and selected ascending muscle-to-sensory pathways as prespecified hypotheses. Significant comparisons included left motor cortex → middle deltoid (T = 2.053; P = 0.044), right motor cortex → triceps (T = 2.079; P = 0.041), and posterior deltoid → left sensory cortex (T = 2.188; P = 0.032). Hemispheres were standardized by channel flipping, so these labels should not be read as unmodified patient anatomy.
• Exploratory correlations: Posterior-deltoid–sensory-pathway change correlated negatively with hand improvement (r = −0.351; P = 0.042), and right-motor–triceps-pathway change correlated negatively with wrist improvement (r = −0.350; P = 0.042). These selected, unadjusted exploratory associations do not establish mediation or a causal repair mechanism.
• Muscle synergies: All four between-group change comparisons were nonsignificant: H1 P = 0.078; H2 P = 0.205; W1 P = 0.481; W2 P = 0.078. Both groups improved within group. Visual similarity to healthy patterns is not evidence of superior normalization attributable to Tai Ji. Synergy data: Table 3.

Adverse Events / Safety

The discussion states that no adverse events occurred, and the protocol describes exercise logs recording events. However, a detailed arm-specific safety table, falls analysis, and explicit adverse-event ascertainment results are not provided. Eight Tai Ji participants and six control participants withdrew because of incomplete intervention or refusal of repeat EEG; the report does not break those reasons down by arm. A small, short, selected trial cannot establish rare-event safety or safety in patients unable to stand independently.

Figures

The original figures below are embedded unchanged from the verified PMC CDN. Captions are paraphrased for this educational review; the original images and their contents have not been altered. Attribution: Xie et al., 2026, under CC BY-NC-ND 4.0. Click a thumbnail to enlarge without JavaScript.

×

Source: Xie et al., PMC13560274, original Figure 1; CC BY-NC-ND 4.0. Image unchanged; caption paraphrased. 12984_2026_2054_Fig1_HTML.webp. Click image to expand.

×

Source: Xie et al., PMC13560274, original Figure 2; CC BY-NC-ND 4.0. Image unchanged; caption paraphrased. 12984_2026_2054_Fig2_HTML.webp. Click image to expand.

×

Source: Xie et al., PMC13560274, original Figure 3; CC BY-NC-ND 4.0. Image unchanged; caption paraphrased. 12984_2026_2054_Fig3_HTML.webp. Click image to expand.

×

Source: Xie et al., PMC13560274, original Figure 4; CC BY-NC-ND 4.0. Image unchanged; caption paraphrased. 12984_2026_2054_Fig4_HTML.webp. Click image to expand.

Criticisms

• Dose and attention are inseparable from treatment identity. The Tai Ji group was offered 40 hours of structured exercise versus 20 hours in control. This answers whether adding the package helps, not whether Tai Ji is better per hour than another supervised exercise.
• Missing outcomes matter. Fourteen of 84 randomized participants, or 16.7% calculated, were absent from the completed analysis. An intention-to-treat analysis, reason-specific missingness assessment, and sensitivity analyses are not reported.
• Recovery stage may matter. Median stroke chronicity was 2.5 versus 6 months among completers. No adjusted model addresses that imbalance; P = 0.067 does not certify equivalent recovery potential.
• Abstract versus tables. Broad language about motor subscales and synergy normalization runs ahead of the hand-subscore result (P = 0.174) and all four negative between-group synergy comparisons.
• Mechanism remains exploratory. Numerous pathway and correlation tests are discussed without a reported multiplicity correction. PTE directionality is a property of the signal model, not an experimental demonstration of direct anatomical causation or clinical mediation.
• Effect magnitude is incompletely characterized. Table 2 gives medians and P values, not a clinical treatment-effect confidence interval. The discussion quotes Cohen’s d ≈ 1.84 without enough accompanying change-score detail to independently reconstruct it from the table; do not treat it as a precise Tai Ji-specific effect.
• External validity and durability are limited. One center, relatively young completers, independent mobility, residual muscle strength, and four-week assessment leave substantial uncertainty about older, severely impaired, nonambulatory, or long-term populations.

Funding

National Natural Science Foundation of China grants 82305357 and 62373108, and Natural Science Foundation of Fujian Province grant 2025J01948. The funding paragraph does not specify the funders’ role in trial design, conduct, analysis, or publication. The authors declare no competing interests. Trial registration: ChiCTR2200061376; this is a Chinese Clinical Trial Registry identifier, not an NCT number.

The paper

• Authors. Qiurong Xie et al.
• Title. Tai Ji for stroke rehabilitation: a randomized controlled trial on improving upper limb function and cortico-muscular coupling in stroke patients.
• Journal. Journal of NeuroEngineering and Rehabilitation.
• Year. 2026; published June 25, 2026.
• DOI. 10.1186/s12984-026-02054-y
• PMCID. PMC13560274
• PMID. 42351241