2.1

This study adheres to the PRISMA guidelines to ensure methodological rigor . We registered the study protocol prospectively on PROSPERO, accessible via registration number CRD420251273966.

2.2

We executed a thorough search across eight specialized databases. This process yielded 114 total records initially. Specific counts included 19 from CNKI, 17 from Wanfang, and 13 from PubMed. Our search was complemented by international databases, including Web of Science, Embase, and the Cochrane Library, and concluded on December 27, 2025. The strategy employed a combination of MeSH terms and free-text keywords covering key concepts: “chronic ankle instability,” “CAI,” “functional ankle instability,” “FAI,” “ankle sprain,” and “mechanical ankle instability”/”MAI” for the condition; and “Tai Chi,” “Tai Ji Quan,” along with major style-specific terms (e.g., Chen-style, Yang-style) for the intervention. To maximize the retrieval of eligible studies, we manually screened the bibliographies of all included articles to identify additional RCTs on Tai Chi for CAI, and the full PubMed search strategy is presented in Table 1 .

2.3

Our team established the study parameters according to the PICOS (Participants, Intervention, Comparison, Outcomes, and Study design) framework. In this study, Participants included individuals with Chronic Ankle Instability (CAI), which encompasses Functional Ankle Instability (FAI) and Mechanical Ankle Instability (MAI). Specifically, MAI refers to joint laxity beyond the physiological limit, often diagnosed by clinical stress tests or imaging; while FAI is characterized by the subjective feeling of joint “giving way” and recurrent sprains despite normal joint range of motion, often attributed to proprioceptive and neuromuscular deficits . (Refer to Table 2 ):

2.4

Process and Consensus: Two investigators (Tongtong Hao and Juntong Yuan) conducted the literature filtering and data acquisition independently. The findings were cross-verified to ensure accuracy. Discrepancies at this stage were resolved through team discussion or consultation with third-party experts (Ying He and Man Xu). Screening Strategy: Our team first removed irrelevant records by evaluating their titles. Subsequently, we performed an in-depth review of the remaining abstracts and full articles to confirm they met our specific eligibility criteria.External Consultation: In cases where study details remained unclear, we reached out to the primary authors via telephone or email to obtain further clarification.Data Categories: The specific information gathered from each study included: Study Metadata: The lead author’s name, the title of the paper, and the year it was published. Demographics: The age of participants and the sample size assigned to each group. Intervention Protocols: Comprehensive details regarding the exercise programs and treatments. Quality Metrics: Data necessary for evaluating methodological quality and the risk of bias. Results: Primary and secondary outcome measurements along with their corresponding statistical data.

2.5

Two independent reviewers (Tongtong Hao and Juntong Yuan) appraised the methodological quality and risk of bias of the included trials. Disagreements were arbitrated by third-party researchers (Ying He and Man Xu). Bias was specifically assessed using the Cochrane Risk of Bias 2 (RoB 2) tool, which evaluates five key domains. In parallel, the methodological rigor was quantified using the Physiotherapy Evidence Database (PEDro) scale.The PEDro assessment provides a quantitative measure of quality, with possible scores extending from a minimum of 0 to a maximum of 10 ^, . A maximum score of 10 was attainable, with one point awarded for each satisfied criterion among items 2–11. Studies were subsequently categorized as high (score ≥6), moderate (score 4–5), or low quality (score ≤3) based on the total score.

2.6

To investigate the influence of key variables—such as intervention type, program duration, session length, and exercise frequency—we conducted subgroup analyses. In cases where substantial heterogeneity remained, a sensitivity analysis was carried out to assess the stability and dependability of the combined meta-analysis findings .

2.7

To more precisely estimate the true intervention effect, the meta-analysis was conducted with STATA 15 software, utilizing the mean difference ( MD _diff ) and its standard deviation for processing the evaluation index data from the included studies .

The formula for the mean difference ( MD _diff ) is:

where MD _diff represents the difference between the means before and after the intervention; Mean _post is the post-intervention mean, and Mean _{p re} is the pre-interventiormean.

The formula for the standard deviation of the difference (SD _diff ) is:

In this analysis, SDdiff refers to the pooled standard deviation combining both pre- and post-intervention data, while SDpre and SDpost represent the baseline and post-intervention standard deviations, respectively. Additionally, the variable r represents the correlation coefficient linking the data before and after the treatment. For studies that did not explicitly report the r value, a simplified calculation model was utilized.

For the synthesis of continuous outcomes, the mean difference (MD) along with its 95 % confidence interval (CI) was utilized. The weighted mean difference (WMD) served as the main effect size metric in cases where trials employed consistent assessment tools and units. We utilized this approach to ensure standardized comparisons of the results; if the measurement methods or units differed, the standardized mean difference ( SMD ) was used for statistical analysis with 95 % CI .

The P value served to test for the presence of heterogeneity (P ≥ 0.10: absent; P < 0.10: present). The I² statistic then quantified its degree, with values below 50 % defined as low heterogeneity, for which a fixed-effects model was employed; values equal to or exceeding 50 % were defined as significant heterogeneity, for which a random-effects model was applied . Sensitivity analysis was performed by excluding individual studies one by one to assess their impact on inter-study heterogeneity . For outcomes reported in at least nine studies, publication bias was evaluated via Egger’s test, with the threshold for statistical significance set at P < 0.05.

3.1

Following the application of eligibility criteria, nine studies involving 374 participants were included. The mean participant age across these studies spanned 18.4–25.30 years, with individual study sample sizes varying between 24 and 68. All studies were recent publications, issued from 2019 to 2025. The interventions lasted 6–14 weeks, with a frequency of 2–5 sessions weekly and each session lasting 15–90 min. The Tai Chi training programs encompassed various forms, including 24-form, simplified, Chen-style, and novel simplified versions, as well as combinations with kinesio taping or Tai Chi Zhuang (standing meditation). Presented in Table 3 are the characteristics of the included studies.

3.2

Assessment with the Cochrane RoB 2 tool showed that eight studies (88.9 %) were at low risk of bias, with one study (11.1 %) raising some concerns. According to the PEDro scale, all included studies demonstrated moderate to high methodological quality, with scores ranging from 6 to 8 and an average of 6.78. This indicates high methodological quality and overall sound literature quality (see Table 4 / Fig. 2 ).