The effects of tissue flossing on neuromuscular activation in patients with chronic ankle instability: a randomized controlled trial Authors: Jianxin Ran, Yifan Long, Jingfeng Yang, Na Li, Jianxin Chen, Cheng Fang, Xiaodong Li, Yuanpeng Liao Published: 2026 ABSTRACT Chronic ankle instability (CAI) is a prevalent sequela of lateral ankle sprains, primarily characterized by impaired neuromuscular control. Although tissue flossing (TF) has demonstrated potential in acutely enhancing neuromuscular function in CAI patients, its long-term rehabilitative effects remain unclear. This study aimed to investigate the long-term neuromuscular adaptive effects of combining TF with balance training in CAI patients. Thirty-four participants with CAI were randomly assigned (using a random number remainder method) to an 8-week intervention of either balance training combined with TF (FLOSS group, n = 17) or balance training alone (CON group, n = 17).Outcome measures, assessed pre- and post-intervention, included muscle reaction time and activation level of the tibialis anterior (TA) and peroneus longus (PL) measured using surface electromyography, isometric strength was measured with a hand-held dynamometer, static balance was evaluated via a unilateral stance test (UST) on a balance system, and dynamic balance was assessed using the Y-Balance Test (YBT). Data were analyzed using a 2 (time: pre vs. post) × 2 (group: FLOSS vs. CON) repeated-measures analysis of variance. Following the intervention, a significant Time × Group interaction was found only for the RMS amplitude of the TA during the anterior reach task (F (1, 28) = 4.417, p = 0.044, ηp2 = 0.121, large effect size), with no significant between-group difference in post-hoc tests (p = 0.454). Muscle reaction time and static balance improved significantly from baseline in all participants, but the magnitude of improvement did not differ between groups. No additional between-group differences were observed in ankle muscle strength, dynamic balance, or electromyographic activity in other directions. TF combined with balance training did not provide statistically significant additional long-term benefits compared to balance training alone.Clinical trial registration: This trial was registered at the Chinese Clinical Trial Registry (Identifier: ChiCTR2400092294) (https://www.chictr.org.cn/bin/userProject?status=%E9%80%9A%E8%BF%87%E5%AE%A1%E6%A0%B8).Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-50154-5. Introduction Ankle sprains represent one of the most common injuries in sports, with a recurrence rate as high as 73%1. In the United States alone, approximately 2 million acute ankle sprains occur annually2, with about 70% of patients potentially progressing to chronic ankle instability (CAI)3. Patients with CAI are often trapped in a vicious cycle of repeated sprains and persistent instability. This condition not only significantly increases their risk of recurrent injuries and early-onset degenerative joint disease, but also leads to chronic pain and functional limitations that severely impede both daily activities and athletic performance4. Impaired neuromuscular control, muscle weakness, and deficits in postural stability are widely recognized as key factors contributing to recurrent ankle sprains in CAI patients4. Among these, impaired neuromuscular control serves as a core pathological mechanism, interacting with muscle weakness and postural instability in a vicious cycle5. This dysfunction is frequently manifested as aberrant activation of key peri-articular stabilizing muscles. Studies have indicated that CAI patients exhibit significant activation abnormalities of the tibialis anterior (TA) and peroneus longus (PL) characterized by delayed muscle onset and reduced electromyographic amplitude during functional tasks like the Y-Balance Test, forward lunge, and lateral hop6–9. This abnormal activation pattern directly compromises dynamic stability and increases the risk of re-injury. Balance training serves as the cornerstone of rehabilitation for CAI. Its primary advantage lies in effectively targeting the neuromuscular control deficits characteristic of CAI, leading to significant improvements in postural stability (both static and dynamic balance)10,11.However, its effects on muscle strength and muscle activation levels are often limited. Tissue flossing (TF) has garnered attention for its potential to improve joint range of motion and muscle function12,13. TF exerts its effects through multiple mechanisms, including fascial shear, blood flow restriction and reperfusion, stimulation of mechanoreceptors, and enhanced proprioceptive input. This intervention has been shown to effectively boost exercise performance (e.g., sprint and jump capacity) in healthy populations14,15 and aid in improving range of motion, muscle function, and pain symptoms in clinical patients12,13. Research demonstrates that applying TF to the ankle immediately enhances vertical jump height16 and produces acute beneficial effects on static balance and TA activation in CAI patients13,17. This suggests that the acute neuromuscular benefits of TF may precisely address the relative deficit of traditional balance training in elevating muscle activation levels. Nevertheless, existing evidence predominantly focuses on the immediate effects of TF and is largely derived from studies with small sample sizes or lacking control groups. A significant gap remains regarding whether TF can induce long-term, cumulative neuromuscular adaptations in CAI a chronic condition. Specifically, it is unclear if combining this acute sensory-enhancing intervention with balance training yields a synergistic effect, thereby potentially overcoming the plateau associated with single-modality training. High-quality randomized controlled trial evidence is currently lacking. To address this research gap, this study aimed to conduct an 8-week randomized controlled trial to directly compare the effects of TF combined with balance training versus balance training alone on neuromuscular function in CAI patients. The primary outcome was the change in the root mean square (RMS) amplitude of the TA surface EMG signal. Secondary outcomes included RMS amplitude of the PL, surface EMG signal muscle reaction time of the TA and PL, muscle strength, and static/dynamic balance function. We hypothesized that the combined intervention group would demonstrate significantly greater improvements in the root mean square (RMS) amplitude and reaction time of both the TA and PL compared to the balance training only group. Materials and methods Participants Sample size estimation and recruitment A priori sample size estimation was performed using G*Power (version 3.1.9.2; Heinrich Heine University Düsseldorf, Düsseldorf, Germany; available at:https://www.psychologie.hhu.de/arbeitsgruppen/allgemeine-psychologie-und-arbeitspsychologie/gpower) . Based on an effect size reported in the literature18 and with surface electromyography RMS amplitude as the primary outcome, a sample size of 28 participants (14 per group) was calculated to detect a between-group difference with a significance level (α) of 0.05 and a statistical power (1-β) of 90%. To account for an estimated 20% attrition rate, 34 volunteers were ultimately recruited and randomly allocated to two groups (17 per group) in a 1:1 ratio. This study aimed to recruit a representative sample of the mixed clinical population with CAI commonly encountered in primary care and sports medicine settings. Therefore, eligibility was based on a combination of clinical history, symptom reports, and functional deficits, rather than on prospective classification into functional or mechanical CAI subtypes. The specific inclusion and exclusion criteria are detailed below. Eligibility criteria Participants were physically active male university students who met the criterion of regular exercise, defined as engaging in structured physical activity for more than 150 min per week, in accordance with global physical activity recommendations19.Their exercise modalities encompassed a variety of activities (e.g., ball games, running, and strength training). This inclusion criterion was based on overall activity volume rather than a specific exercise type, as the study aimed to recruit a general population with a habitual active lifestyle, not to compare the effects of different exercise modalities. Inclusion Criteria: (1) At least one significant ankle sprain within 12 months before enrollment, according to the International Ankle Consortium consensus20, with associated pain, swelling, and interruption of normal physical activity for at least one day; (2) The most recent ankle injury occurred at least 3 months prior to enrollment, with subsequent feelings of ankle instability or recurrent sprains; (3) A Cumberland Ankle Instability Tool (CAIT) score ≤ 2521; (4) No history of receiving any structured rehabilitation therapy (e.g., supervised physical therapy, proprioceptive training, or strength training protocols) for the injured ankle since in the 6 months preceding enrollment. Exclusion Criteria: (1) Acute musculoskeletal injury to other lower limb joints within 3 months before the study that impaired joint integrity and function, or an acute ankle sprain within 8 weeks before enrollment; (2) Known latex allergy; (3) Any clinically diagnosed neurological, orthopedic, dermatological, or neuromuscular conditions affecting the lower limb. To further clarify the study cohort, supplementary details regarding key clinical dimensions referenced in the eligibility criteria are provided below: Degree of Injury: The “significant ankle sprain” specified in the inclusion criteria was operationally defined in this study as an injury that, in accordance with the International Ankle Consortium consensus, resulted in pain, swelling, and interruption of normal physical activity for more than 24 h. This definition corresponds to common clinical grades I-II ligamentous sprains, excluding grade III complete ruptures typically requiring surgical intervention. Time since injury and chronicity: To ensure a chronic condition, the initial qualifying injury for all participants occurred more than 3 months prior to enrollment. Control for Prior Treatment History: To control for the confounding effect of prior rehabilitation, the inclusion criterion of “no history of receiving any structured rehabilitation therapy since the initial sprain” was implemented as follows: none of the participants had undergone any structured, supervised physical therapy or rehabilitation program for ankle instability in the 6 months preceding enrollment. This study was approved by the Ethics Committee of Chengdu Sport University (Approval No. 145–2024) and was conducted in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all participants prior to enrollment, ensuring their voluntary participation and understanding of the study procedures. Study design This study was a randomized controlled trial that was registered at the Chinese Clinical Trial Registry (Registration No. ChiCTR2400092294; Date of registration: 14 November 2024). All participants underwent initial screening to confirm eligibility. Eligible individuals provided written informed consent after receiving a detailed explanation of the study procedures (see Fig. 1). Fig. 1CONSORT flow diagram. Flowchart detailing participant progression through the randomized controlled trial phases: enrollment, allocation, follow-up, and analysis. Participants were screened, randomly allocated to the FLOSS (tissue flossing combined with BOSU-ball training) or CON (BOSU-ball training alone) groups, and assessed for eligibility. Attrition and final sample sizes for analysis are shown. CONSORT flow diagram. Flowchart detailing participant progression through the randomized controlled trial phases: enrollment, allocation, follow-up, and analysis. Participants were screened, randomly allocated to the FLOSS (tissue flossing combined with BOSU-ball training) or CON (BOSU-ball training alone) groups, and assessed for eligibility. Attrition and final sample sizes for analysis are shown. Randomization was performed using IBM SPSS Statistics (Version 27.0; IBM Corp., Armonk, NY, USA; available at: https://www.ibm.com/support/pages/downloading-ibm-spss-statistics-27). Specifically, a random number remainder method was applied to generate the allocation sequence. All eligible participants were assigned a unique identification number from 1 to N. For each participant, SPSS generated a random number, which was then divided by 2. Based on the remainder, participants were allocated to one of two groups: those with a remainder of 0 were assigned to the CON group, and those with a remainder of 1 were assigned to the FLOSS group. The allocation information for each participant was concealed in sequentially numbered, opaque, sealed envelopes. An independent third-party researcher, who was not involved in participant recruitment or assessment, prepared and safeguarded these envelopes. Following the baseline assessment, this researcher opened the envelopes in sequential order to reveal the group assignment. Subsequently, the study personnel allocated the participant to the corresponding group for the intervention. All participants formally entered the study after providing voluntary written informed consent. Both groups received the assigned intervention three times per week for 8 weeks (the FLOSS group performed balance training combined with floss band application, while the CON group performed conventional balance training alone). All outcome assessments were conducted at a dedicated motor control laboratory within the authors’ institution, at both baseline and after the 8-week intervention period. The primary outcome measure was the muscle activation level of the TA. Secondary outcomes included the muscle activation level of the PL, muscle reaction time, ankle dorsiflexion and eversion strength, reach distance in the Y-Balance Test, and center-of-pressure sway velocity during the Unilateral Stance Test. All participants were instructed to refrain from vigorous physical activity for 24 h prior to each testing session. Inventions All participants performed balance training on a BOSU ball three times per week for 8 weeks. Unstable support surfaces were chosen to directly target impaired proprioception and sensorimotor control, which are core deficits of CAI22. All training sessions were supervised by the same certified physical therapist-researcher in a dedicated laboratory to ensure protocol adherence and participant safety. Participants stood on the BOSU ball with hands on hips, eyes closed, and maintained balance on the affected leg. The training consisted of three stances performed in sequence: double-leg stance, single-leg stance (on the affected leg), and tandem stance. The training volume was progressively increased: during weeks 1–4, each stance was performed for 4 sets of 45-s holds, with 40-s rest intervals between sets. From weeks 5 to 8, the volume was increased to 5 sets of 45-s holds with 30-s rest intervals between sets23. FLOSS Group: A floss band was applied to the affected ankle. The wrapping started at the base of the fifth metatarsal and continued proximally under tension, with a 50% overlap per turn, until it fully covered the tibialis anterior and peroneus longus muscles, extending to the proximal tibiofibular region (Fig. 2). A Kikuhime pressure monitor (TT-Meditrade, Denmark) was placed at the proximal third of the line connecting the fibular head and the medial malleolus to ensure the compression pressure was maintained within the target range of 160 ± 3 mmHg24. Following application, participants performed the BOSU ball training. The band was removed during the 30-s rest intervals between sets to allow active ankle movements, and reapplied prior to the next set. Fig. 2Application of the floss band. Illustration of the circumferential floss band wrapping technique applied to the affected ankle, extending from the foot to the proximal tibiofibular region. Application of the floss band. Illustration of the circumferential floss band wrapping technique applied to the affected ankle, extending from the foot to the proximal tibiofibular region. CON Group: Participants performed the identical BOSU ball training protocol without the application of the floss band. All participants were instructed to maintain their usual level of activity but were prohibited from receiving any additional physical therapy, rehabilitation exercises, manual therapy, aside from the interventions prescribed by the study protocol. Any new ankle injury or treatment was required to be reported immediately, and the participant’s data would be excluded from analysis. Outcome measures The flow of data collection is summarized in Fig. 3. All assessments were conducted at baseline and immediately after the 8-week intervention period. Fig. 3Data collection and processing flowchart. Schematic overview of the assessment protocol. The process included: preparation and surface electromyography (sEMG) setup; maximum voluntary isometric contraction (MVIC) testing for strength and baseline sEMG; Y-Balance Test (YBT) for dynamic postural control; Unilateral Stance Test (UST) for static postural control; and subsequent sEMG signal processing for the analysis of muscle activation level and reaction time. Abbreviations: HHD, hand-held dynamometer; BP, band-pass filter; RMS, root mean square. Data collection and processing flowchart. Schematic overview of the assessment protocol. The process included: preparation and surface electromyography (sEMG) setup; maximum voluntary isometric contraction (MVIC) testing for strength and baseline sEMG; Y-Balance Test (YBT) for dynamic postural control; Unilateral Stance Test (UST) for static postural control; and subsequent sEMG signal processing for the analysis of muscle activation level and reaction time. Abbreviations: HHD, hand-held dynamometer; BP, band-pass filter; RMS, root mean square. Muscle strength The isometric strength of the TA and PL on the involved side was measured using a hand-held dynamometer (HHD; microFET3). Strength was determined by the peak force achieved during a maximum voluntary contraction (MVC), assessed before and after the intervention. A 10-min warm-up preceded all testing. For TA assessment, participants lay supine with the ankle in a neutral position. The HHD pad was placed on the dorsum of the foot to assess TA strength. Upon the verbal command “start,” participants performed a gradual maximal voluntary isometric dorsiflexion and held the contraction for 5 s. For PL assessment, the HHD pad was positioned laterally over the fifth metatarsal, and participants performed a gradual maximal voluntary isometric eversion, holding for 5 s25. One practice trial was conducted for each muscle group prior to formal testing to familiarize participants and ensure stability. During formal data collection, each muscle performed three 5-s MVIC trials, separated by 30-s rest intervals. An additional trial was administered if the coefficient of variation (CV) between the three trials exceeded 10% to minimize measurement variability. All measurements were performed by the same tester. The order of muscle testing (TA vs. PL) was randomized to control for potential bias. Surface electromyography (sEMG) signals from the TA and PL were recorded synchronously during the MVIC trials for subsequent normalization. Postural control Dynamic postural control was assessed using the Y-Balance Test (YBT). Participants stood on the involved leg at the center of the testing kit. With the non-stance leg, they attempted to reach as far as possible in three directions: anterior (ANT), posteromedial (PM), and posterolateral (P-L), pushing a reach indicator with their foot. Tests were performed in the order of ANT, PM, and P-L. While standing on the slightly bent involved leg, participants completed each reach following a standardized 5-s audio countdown, which guided the movement sequence of toe-off, reach, tap, and return of the non-stance foot to the platform. Three trials were performed for each direction. Throughout testing, the heel of the stance leg was required to maintain contact with the platform. The maximum reach distance in each direction was recorded. The average of the three trials for each direction was used for analysis. sEMG signals from the TA and PL were recorded synchronously during all YBT trials. Static postural control was evaluated using the Unilateral Stance Test (UST) on a Balance Manager system (NeuroCom® VSR, USA, version 9.3) The test was conducted under two conditions: UST1 (eyes open) and UST2 (eyes closed), performed in this order. For each condition, participants performed three 10-s trials of single-leg standing on the involved limb, with at least 1 min of rest between trials. Participants stood on the involved leg with hands on hips, looking straight ahead, and maintaining an upright trunk. The non-stance leg was lifted with the hip flexed at approximately 30°, the knee flexed at about 90°, and the foot elevated approximately 10 cm from the ground. The movement trajectory of the center of pressure (COP) was recorded for each trial using the manufacturer’s software. The COP sway velocity was calculated as the total path length divided by the trial duration (10 s) and expressed in degrees per second (°/s). A trial was discarded and repeated if the non-stance leg contacted the stance leg or the platform26. Muscle activation sEMG data acquisition sEMG signals were acquired using a Noraxon Ultium ESP system (sampling frequency: 2000 Hz; resolution: 16-bit). Prior to electrode placement, the skin over the target muscles was shaved and cleansed with isopropyl alcohol. With reference to established guidelines, electrodes for the TA were placed at the proximal one-third of the line connecting the fibular head and the medial malleolus; electrodes for the PL were positioned at the proximal one-quarter of the line from the fibular head to the lateral malleolus9. General sEMG data processing All raw EMG data were processed offline using a custom MATLAB script (R2023a; The MathWorks, Inc., Natick, MA, USA; available at: https://www.mathworks.com/products/matlab.html). The standard processing pipeline, applied to data from all tasks (MVIC, YBT, Reaction Time test), was executed in the following order: Band-pass Filtering: The raw signals were filtered using a 4th-order zero-lag Butterworth filter with a passband of 20–450 Hz to remove movement artifacts and high-frequency noise. A 50 Hz notch filter was applied to eliminate power line interference. Smoothing: For subsequent amplitude and onset analysis, the filtered signals were full-wave rectified and smoothed using an RMS algorithm with a 100-ms moving window. Muscle Activation Level (%MVIC). MVIC Reference Value: During the MVIC tests, sEMG signals from the TA and PL were recorded. For each muscle, a stable 2-s epoch of peak activity was manually selected from the recorded trial. The RMS amplitude over this epoch was calculated. The trial with the maximum RMS amplitude among the three repetitions was used to establish the MVC reference value for each muscle. YBT Activation: During the YBT, sEMG signals were synchronously recorded. For each reach direction (ANT, PM, P-L), a stable 2-s epoch during the reaching phase was selected. The peak RMS amplitude within this epoch was extracted. Normalization: Muscle activation level for each YBT direction was expressed as a percentage of the MVIC (%MVIC), calculated by: (Peak RMS during YBT epoch / MVC reference RMS) × 10027. Muscle reaction time Reaction time was assessed using the “UP” perturbation of the Weight-Bearing Squat protocol on a Balance Manager system. Participants stood barefoot, wearing opaque glasses and headphones to eliminate anticipatory cues. Three trials were performed, and the average was used for analysis. Definition: Reaction time was defined as the time interval from the onset of dynamic platform rotation to the onset of the corresponding muscle’s EMG burst. Baseline Activity Calculation: The mean and standard deviation (SD) of the EMG signal for threshold determination were calculated from a 100-ms time window immediately preceding the platform perturbation onset. Onset Detection: The onset of muscle activity was automatically detected from the smoothed EMG signal (processed as described in the General sEMG Data Processing section). The onset was identified when the signal amplitude exceeded a threshold of the mean baseline activity plus 2 SDs for a minimum duration of 10 ms (20 data points at 2000 Hz). A search window of 1 s post-perturbation was used. If no activation was detected within this window, the trial was excluded from analysis for that muscle. Reliability and validity of outcome measures The reliability and validity of the instruments or methods used for all key outcome measures were evaluated. Detailed data are provided in the online Supplementary Material S1. A brief summary is as follows: Muscle Strength: MVIC was assessed using a HHD (microFET3). This testing protocol has demonstrated good test–retest reliability for evaluating peri-ankle muscle strength (ICC range = 0.31–0.79)28. YBT: Its test–retest reliability and construct validity for assessing dynamic postural control in individuals with CAI have been extensively validated (ICC range = 0.85–0.99)29. UST: Sway velocity, calculated from the center of pressure trajectory recorded via the NeuroCom system, is an objective and repeatable measure (ICC range = 0.81–0.93)30. Muscle Activation Level: sEMG was used to synchronously record the activation of the TA and PL during the YBT, following the protocol established by Jaber et al.9. The intra-rater reliability of the sEMG data processing pipeline in this study was excellent (ICC > 0.90 for all key outcomes). Muscle Reaction Time Test: This newly developed test demonstrated acceptable test–retest reliability, good content validity, and was able to effectively discriminate between CAI patients and healthy controls (known-groups validity). Statistical analysis All statistical analyses were performed using SPSS version 27.0. The normality of the data distribution for each group was assessed using the Shapiro–Wilk test. Given the mixed experimental design of 2 (Group: FLOSS vs. CON) × 2 (Time: Pre- vs. Post-intervention), a repeated-measures analysis of variance (RM-ANOVA) was employed for all continuous outcome measures to simultaneously examine the main effect of time, the main effect of group, and, crucially, the Group × Time interaction. If a significant (p < 0.05) or a trending (p < 0.10) Group × Time interaction was detected, simple effect analyses were conducted to delineate the specific patterns of change over time within each group. The significance level (alpha) was set at p < 0.05. Effect sizes were calculated as partial eta squared (ηp2) and interpreted as small (0.01), medium (0.06), or large (0.14)31. Results A total of 30 participants completed the study (Table 1). One participant in the FLOSS group and one in the CON group were excluded due to intervention discontinuation. Additionally, one participant in the FLOSS group dropped out due to scheduling conflicts preventing post-test assessment, and one in the CON group withdrew after sustaining a knee injury during daily activities unrelated to the study. No intervention-related adverse events (such as numbness, persistent pain, skin allergy, bruising, or discoloration) were reported in either the FLOSS or CON groups throughout the 8-week intervention period. No significant differences were observed between the two groups in terms of demographic characteristics or baseline performance measures (p = 0.119–0.701). Table 1Demographic characteristics of the participants (n = 30).GroupFLOSS (n = 15)CON (n = 15)p valueAge, years21.41 ± 1.9722.11 ± 1.660.119Height, cm173.64 ± 3.71175.23 ± 4.210.251Weight, kg71.58 ± 2.2071.99 ± 3.300.701Number of ankle sprains, times2.9 ± 1.143.1 ± 1.280.563CAIT score20.17 ± 2.2420.82 ± 2.480.431CAIT: Cumberland Ankle Instability Tool. Table 1Demographic characteristics of the participants (n = 30).GroupFLOSS (n = 15)CON (n = 15)p valueAge, years21.41 ± 1.9722.11 ± 1.660.119Height, cm173.64 ± 3.71175.23 ± 4.210.251Weight, kg71.58 ± 2.2071.99 ± 3.300.701Number of ankle sprains, times2.9 ± 1.143.1 ± 1.280.563CAIT score20.17 ± 2.2420.82 ± 2.480.431CAIT: Cumberland Ankle Instability Tool. Demographic characteristics of the participants (n = 30). CAIT: Cumberland Ankle Instability Tool. Following the 8-week intervention, the changes in primary and secondary outcome measures are summarized below (detailed data are presented in Tables 2 and 3 and Figs. 4 and 5). All analyses were conducted using repeated-measures ANOVA, with effect sizes reported as partial η2 (small: 0.01, medium: 0.06, large: 0.14).Table 2Descriptive statistics (mean ± SD) for all outcome measures at pre- and post-intervention.PrePostChangeMVIC (N)DorsiflexorsFLOSS254.301 ± 21.003258.575 ± 23.3284.271 ± 8.407CON251.164 ± 21.457255.946 ± 22.7904.771 ± 11.033EvertorsFLOSS170.174 ± 18.226171.822 ± 16.6701.652 ± 3.955CON164.110 ± 14.791166.521 ± 15.2442.412 ± 4.957YBT (cm)ANTFLOSS55.852 ± 6.92956.823 ± 6.7480.977 ± 1.508CON59.211 ± 7.95859.764 ± 7.8390.556 ± 1.378P-LFLOSS105.699 ± 9.034105.937 ± 9.6120.238 ± 1.747CON103.169 ± 11.388103.577 ± 11.3790.408 ± 1.537PMFLOSS100.876 ± 9.151101.107 ± 8.2190.231 ± 1.782CON104.817 ± 8.299105.228 ± 8.5090.411 ± 1.798UST↓ (°/s)UST1FLOSS1.309 ± 0.2511.181 ± 0.2810.128 ± 0.109CON1.259 ± 0.1711.171 ± 0.1910.088 ± 0.081UST2FLOSS3.012 ± 0.6712.753 ± 0.6770.249 ± 0.092CON2.781 ± 0.7192.638 ± 0.7020.131 ± 0.029RT↓ (ms)TAFLOSS457.882 ± 71.297349.411 ± 63.198108.408 ± 67.138CON445.411 ± 116.918376.470 ± 76.60865.876 ± 47.572PLFLOSS466.517 ± 70.974442.426 ± 104.30824.123 ± 93.752CON397.605 ± 94.096352.407 ± 62.09845.292 ± 89.238sEMG Amplitude (%MVIC)ANT(TA)FLOSS52.941 ± 13.51258.647 ± 14.7145.706 ± 3.980CON51.705 ± 15.43454.882 ± 14.2343.176 ± 2.91P-L(TA)FLOSS57.352 ± 16.38159.411 ± 15.5202.058 ± 6.138CON65.117 ± 13.53165.764 ± 11.9030.647 ± 3.75PM(TA)FLOSS49.705 ± 14.70250.235 ± 14.8970.529 ± 1.328CON55.588 ± 11.77956.235 ± 11.6270.647 ± 1.587ANT(PL)FLOSS46.588 ± 17.22349.882 ± 18.1483.294 ± 9.439CON42.588 ± 12.80144.058 ± 10.6851.470 ± 5.948P-L(PL)FLOSS49.470 ± 11.35854.235 ± 19.1624.764 ± 13.511CON45.941 ± 12.81348.941 ± 12.8373.000 ± 8.580PM(PL)FLOSS45.470 ± 9.19350.823 ± 16.2915.352 ± 11.586CON41.058 ± 10.23142.764 ± 12.9831.705 ± 8.651Data are presented as mean ± standard deviation. Change scores (Post – Pre) are calculated so that a positive value indicates improvement for MVIC, YBT, and IEMG, and a negative value indicates improvement for UST and RT (↓ indicates that a lower raw value is better).MVIC: maximum voluntary isometric contraction; YBT: Y-Balance test; ANT: anterior; P-L: posterolateral; PM: posteromedial; UST: unilateral stance test (UST1: eyes open; UST2: eyes closed); RT: reaction time; TA: tibialis anterior; PL: peroneus longus.Normalized Muscle Activationwas expressed as a percentage of the maximum voluntary isometric contraction (%MVIC), calculated from the root mean square amplitude of the sEMG signal. Table 2Descriptive statistics (mean ± SD) for all outcome measures at pre- and post-intervention.PrePostChangeMVIC (N)DorsiflexorsFLOSS254.301 ± 21.003258.575 ± 23.3284.271 ± 8.407CON251.164 ± 21.457255.946 ± 22.7904.771 ± 11.033EvertorsFLOSS170.174 ± 18.226171.822 ± 16.6701.652 ± 3.955CON164.110 ± 14.791166.521 ± 15.2442.412 ± 4.957YBT (cm)ANTFLOSS55.852 ± 6.92956.823 ± 6.7480.977 ± 1.508CON59.211 ± 7.95859.764 ± 7.8390.556 ± 1.378P-LFLOSS105.699 ± 9.034105.937 ± 9.6120.238 ± 1.747CON103.169 ± 11.388103.577 ± 11.3790.408 ± 1.537PMFLOSS100.876 ± 9.151101.107 ± 8.2190.231 ± 1.782CON104.817 ± 8.299105.228 ± 8.5090.411 ± 1.798UST↓ (°/s)UST1FLOSS1.309 ± 0.2511.181 ± 0.2810.128 ± 0.109CON1.259 ± 0.1711.171 ± 0.1910.088 ± 0.081UST2FLOSS3.012 ± 0.6712.753 ± 0.6770.249 ± 0.092CON2.781 ± 0.7192.638 ± 0.7020.131 ± 0.029RT↓ (ms)TAFLOSS457.882 ± 71.297349.411 ± 63.198108.408 ± 67.138CON445.411 ± 116.918376.470 ± 76.60865.876 ± 47.572PLFLOSS466.517 ± 70.974442.426 ± 104.30824.123 ± 93.752CON397.605 ± 94.096352.407 ± 62.09845.292 ± 89.238sEMG Amplitude (%MVIC)ANT(TA)FLOSS52.941 ± 13.51258.647 ± 14.7145.706 ± 3.980CON51.705 ± 15.43454.882 ± 14.2343.176 ± 2.91P-L(TA)FLOSS57.352 ± 16.38159.411 ± 15.5202.058 ± 6.138CON65.117 ± 13.53165.764 ± 11.9030.647 ± 3.75PM(TA)FLOSS49.705 ± 14.70250.235 ± 14.8970.529 ± 1.328CON55.588 ± 11.77956.235 ± 11.6270.647 ± 1.587ANT(PL)FLOSS46.588 ± 17.22349.882 ± 18.1483.294 ± 9.439CON42.588 ± 12.80144.058 ± 10.6851.470 ± 5.948P-L(PL)FLOSS49.470 ± 11.35854.235 ± 19.1624.764 ± 13.511CON45.941 ± 12.81348.941 ± 12.8373.000 ± 8.580PM(PL)FLOSS45.470 ± 9.19350.823 ± 16.2915.352 ± 11.586CON41.058 ± 10.23142.764 ± 12.9831.705 ± 8.651Data are presented as mean ± standard deviation. Change scores (Post – Pre) are calculated so that a positive value indicates improvement for MVIC, YBT, and IEMG, and a negative value indicates improvement for UST and RT (↓ indicates that a lower raw value is better).MVIC: maximum voluntary isometric contraction; YBT: Y-Balance test; ANT: anterior; P-L: posterolateral; PM: posteromedial; UST: unilateral stance test (UST1: eyes open; UST2: eyes closed); RT: reaction time; TA: tibialis anterior; PL: peroneus longus.Normalized Muscle Activationwas expressed as a percentage of the maximum voluntary isometric contraction (%MVIC), calculated from the root mean square amplitude of the sEMG signal. Descriptive statistics (mean ± SD) for all outcome measures at pre- and post-intervention. Data are presented as mean ± standard deviation. Change scores (Post – Pre) are calculated so that a positive value indicates improvement for MVIC, YBT, and IEMG, and a negative value indicates improvement for UST and RT (↓ indicates that a lower raw value is better). MVIC: maximum voluntary isometric contraction; YBT: Y-Balance test; ANT: anterior; P-L: posterolateral; PM: posteromedial; UST: unilateral stance test (UST1: eyes open; UST2: eyes closed); RT: reaction time; TA: tibialis anterior; PL: peroneus longus. Normalized Muscle Activationwas expressed as a percentage of the maximum voluntary isometric contraction (%MVIC), calculated from the root mean square amplitude of the sEMG signal. Table 3Results of repeated-measures ANOVA.Dependent variableEffectF-valueP-valuePartial η2MVIC (N)DorsiflexorsMain Effect of TimeF (1, 28) = 7.2350.011*0.184Main Effect of GroupF (1, 28) = 0.0230.8810.001Time × Group InteractionF (1, 28) = 0.1510.7000.005EvertorsMain Effect of TimeF (1, 28) = 7.0040.013*0.180Main Effect of GroupF (1, 28) = 0.2470.6220.008Time × Group InteractionF (1, 28) = 1.0520.3130.032YBT (cm)ANTMain Effect of TimeF (1, 28) = 9.4230.004*0.227Main Effect of GroupF (1, 28) = 0.7080.4060.022Time × Group InteractionF (1, 28) = 1.5570.2210.046P-LMain Effect of TimeF (1, 28) = 1.3060.2620.039Main Effect of GroupF (1, 28) = 0.0970.7570.003Time × Group InteractionF (1, 28) = 0.4700.4980.014PMMain Effect of TimeF (1, 28) = 1.1040.3010.033Main Effect of GroupF (1, 28) = 0.0820.7760.003Time × Group InteractionF (1, 28) = 1.9050.1770.056UST (°/s)UST1Main Effect of TimeF (1, 28) = 44.000 < 0.001*0.579Main Effect of GroupF (1, 28) = 1.7420.1960.052Time × Group InteractionF (1, 28) = 0.1890.6670.006UST2Main Effect of TimeF (1, 28) = 23.656 < 0.001*0.425Main Effect of GroupF (1, 28) = 4.6340.039*0.126Time × Group InteractionF (1, 28) = 0.2740.6050.008RT (ms)TAMain Effect of TimeF (1, 28) = 78.694 < 0.001*0.711Main Effect of GroupF (1, 28) = 0.0720.790.002Time × Group InteractionF (1, 28) = 3.9070.0570.109PLMain Effect of TimeF (1, 28) = 4.8890.034*0.133Main Effect of GroupF (1, 28) = 10.5790.003*0.248Time × Group InteractionF (1, 28) = 0.4550.5050.014sEMG Amplitude (%MVIC)TA-ANTMain Effect of TimeF (1, 28) = 54.466 < 0.001*0.630Main Effect of GroupF (1, 28) = 0.2570.6160.008Time × Group InteractionF (1, 28) = 4.4170.044*0.121TA-P-LMain Effect of TimeF (1, 28) = 2.4030.1310.070Main Effect of GroupF (1, 28) = 2.0960.1570.061Time × Group InteractionF (1, 28) = 0.6540.4250.020TA-PMMain Effect of TimeF (1, 28) = 5.5270.025*0.147Main Effect of GroupF (1, 28) = 1.6900.2030.050Time × Group InteractionF (1, 28) = 0.0550.8160.002PL-ANTMain Effect of TimeF (1, 28) = 5.0280.032*0.136Main Effect of GroupF (1, 28) = 0.0600.8090.002Time × Group InteractionF (1, 28) = 0.3450.5610.011PL-P-LMain Effect of TimeF (1, 28) = 4.0010.0540.111Main Effect of GroupF (1, 28) = 0.9490.3370.029Time × Group InteractionF (1, 28) = 0.2070.6520.006PL-PMMain Effect of TimeF (1, 28) = 4.0510.0530.112Main Effect of GroupF (1, 28) = 1.0810.3060.033Time × Group InteractionF (1, 28) = 2.5490.1200.074This table presents the main results of the 2 (Time: Pre, Post) × 2 (Group: FLOSS, CON) repeated-measures ANOVA. For each dependent variable, the main effect of Time, the main effect of Group, and the Time × Group interaction are reported. The interaction is the key statistic for evaluating whether the improvement trends differed between groups. Effect sizes are reported as partial η2. * Indicates a statistically significant effect (p < 0.05).MVIC: maximum voluntary isometric contraction; YBT: Y-Balance test; ANT: anterior; P-L: posterolateral; PM: posteromedial; UST: unilateral stance test (UST1, eyes open; UST2, eyes closed); RT: reaction time; TA: tibialis anterior; PL: peroneus longus.Normalized Muscle Activation was expressed as a percentage of the maximum voluntary isometric contraction (%MVIC), calculated from the root mean square amplitude of the sEMG signal. Table 3Results of repeated-measures ANOVA.Dependent variableEffectF-valueP-valuePartial η2MVIC (N)DorsiflexorsMain Effect of TimeF (1, 28) = 7.2350.011*0.184Main Effect of GroupF (1, 28) = 0.0230.8810.001Time × Group InteractionF (1, 28) = 0.1510.7000.005EvertorsMain Effect of TimeF (1, 28) = 7.0040.013*0.180Main Effect of GroupF (1, 28) = 0.2470.6220.008Time × Group InteractionF (1, 28) = 1.0520.3130.032YBT (cm)ANTMain Effect of TimeF (1, 28) = 9.4230.004*0.227Main Effect of GroupF (1, 28) = 0.7080.4060.022Time × Group InteractionF (1, 28) = 1.5570.2210.046P-LMain Effect of TimeF (1, 28) = 1.3060.2620.039Main Effect of GroupF (1, 28) = 0.0970.7570.003Time × Group InteractionF (1, 28) = 0.4700.4980.014PMMain Effect of TimeF (1, 28) = 1.1040.3010.033Main Effect of GroupF (1, 28) = 0.0820.7760.003Time × Group InteractionF (1, 28) = 1.9050.1770.056UST (°/s)UST1Main Effect of TimeF (1, 28) = 44.000 < 0.001*0.579Main Effect of GroupF (1, 28) = 1.7420.1960.052Time × Group InteractionF (1, 28) = 0.1890.6670.006UST2Main Effect of TimeF (1, 28) = 23.656 < 0.001*0.425Main Effect of GroupF (1, 28) = 4.6340.039*0.126Time × Group InteractionF (1, 28) = 0.2740.6050.008RT (ms)TAMain Effect of TimeF (1, 28) = 78.694 < 0.001*0.711Main Effect of GroupF (1, 28) = 0.0720.790.002Time × Group InteractionF (1, 28) = 3.9070.0570.109PLMain Effect of TimeF (1, 28) = 4.8890.034*0.133Main Effect of GroupF (1, 28) = 10.5790.003*0.248Time × Group InteractionF (1, 28) = 0.4550.5050.014sEMG Amplitude (%MVIC)TA-ANTMain Effect of TimeF (1, 28) = 54.466 < 0.001*0.630Main Effect of GroupF (1, 28) = 0.2570.6160.008Time × Group InteractionF (1, 28) = 4.4170.044*0.121TA-P-LMain Effect of TimeF (1, 28) = 2.4030.1310.070Main Effect of GroupF (1, 28) = 2.0960.1570.061Time × Group InteractionF (1, 28) = 0.6540.4250.020TA-PMMain Effect of TimeF (1, 28) = 5.5270.025*0.147Main Effect of GroupF (1, 28) = 1.6900.2030.050Time × Group InteractionF (1, 28) = 0.0550.8160.002PL-ANTMain Effect of TimeF (1, 28) = 5.0280.032*0.136Main Effect of GroupF (1, 28) = 0.0600.8090.002Time × Group InteractionF (1, 28) = 0.3450.5610.011PL-P-LMain Effect of TimeF (1, 28) = 4.0010.0540.111Main Effect of GroupF (1, 28) = 0.9490.3370.029Time × Group InteractionF (1, 28) = 0.2070.6520.006PL-PMMain Effect of TimeF (1, 28) = 4.0510.0530.112Main Effect of GroupF (1, 28) = 1.0810.3060.033Time × Group InteractionF (1, 28) = 2.5490.1200.074This table presents the main results of the 2 (Time: Pre, Post) × 2 (Group: FLOSS, CON) repeated-measures ANOVA. For each dependent variable, the main effect of Time, the main effect of Group, and the Time × Group interaction are reported. The interaction is the key statistic for evaluating whether the improvement trends differed between groups. Effect sizes are reported as partial η2. * Indicates a statistically significant effect (p < 0.05).MVIC: maximum voluntary isometric contraction; YBT: Y-Balance test; ANT: anterior; P-L: posterolateral; PM: posteromedial; UST: unilateral stance test (UST1, eyes open; UST2, eyes closed); RT: reaction time; TA: tibialis anterior; PL: peroneus longus.Normalized Muscle Activation was expressed as a percentage of the maximum voluntary isometric contraction (%MVIC), calculated from the root mean square amplitude of the sEMG signal. Results of repeated-measures ANOVA. This table presents the main results of the 2 (Time: Pre, Post) × 2 (Group: FLOSS, CON) repeated-measures ANOVA. For each dependent variable, the main effect of Time, the main effect of Group, and the Time × Group interaction are reported. The interaction is the key statistic for evaluating whether the improvement trends differed between groups. Effect sizes are reported as partial η2. * Indicates a statistically significant effect (p < 0.05). MVIC: maximum voluntary isometric contraction; YBT: Y-Balance test; ANT: anterior; P-L: posterolateral; PM: posteromedial; UST: unilateral stance test (UST1, eyes open; UST2, eyes closed); RT: reaction time; TA: tibialis anterior; PL: peroneus longus. Normalized Muscle Activation was expressed as a percentage of the maximum voluntary isometric contraction (%MVIC), calculated from the root mean square amplitude of the sEMG signal. Fig. 4White circles represent the mean, and boxes (whiskers) represent the SD. TA, tibialis anterior. White circles represent the mean, and boxes (whiskers) represent the SD. TA, tibialis anterior. Fig. 5White circles represent the mean, and boxes (whiskers) represent the SD., TA, tibialis anterior. White circles represent the mean, and boxes (whiskers) represent the SD., TA, tibialis anterior. For the primary outcome, a significant Time × Group interaction was found for the RMS amplitude of the TA during the ANT reach task (F (1,28) = 4.417, p = 0.044, ηp2 = 0.121, large effect size). Simple effect analysis indicated that the RMS amplitude of the TA increased from pre- to post-intervention in both groups, but the between-group difference at post-intervention was not statistically significant (p = 0.454, see Fig. 4). No significant interactions were observed for the RMS amplitude of the TA in the P-L and PM directions (all p > 0.05). However, a significant main effect of Time was found for the TA in the PM direction (F (1,28) = 5.527, p = 0.025, ηp2 = 0.147, large effect size), indicating an overall improvement in TA activation over time for all participants in that direction. Regarding secondary outcomes, TA reaction time shortened significantly in all participants (main effect of Time: F (1,28) = 78.694, p < 0.001, ηp2 = 0.711, large effect size), but the difference in improvement trends between groups was not significant (interaction: p = 0.057, ηp2 = 0.109, medium effect size) (Fig. 5). PL reaction time showed overall improvement (main effect of Time: F (1,28) = 4.889, p = 0.034, ηp2 = 0.133, large effect size), and a significant main effect of Group was observed (F (1,28) = 10.579, p = 0.003, ηp2 = 0.248, large effect size), indicating that the CON group had consistently shorter PL reaction times overall across the study period compared to the FLOSS group. No differential improvement between groups was found (interaction: p = 0.505, ηp2 = 0.012, small effect size). Significant main effects of Time were found for both ankle dorsiflexor and evertor MVIC (dorsiflexors: F (1,28) = 7.235, p = 0.011, ηp2 = 0.184, large effect size; evertors: F (1, 28) = 7.004, p = 0.013, ηp2 = 0.180, large effect size), indicating that strength improved in all participants post-intervention. However, no significant main effects of Group or Time × Group interactions were found for strength measures (all p > 0.05), meaning the degree of improvement did not differ between groups. For static balance, sway velocity during the UST decreased significantly under both eyes-open and eyes-closed conditions (UST1: main effect of Time, p < 0.001, ηp2 = 0.579, large effect size; UST2: main effect of Time, p < 0.001, ηp2 = 0.425, large effect size). Furthermore, a significant main effect of Group was found for UST2 (F (1, 28) = 4.634, p = 0.039, ηp2 = 0.126, large effect size), indicating that the CON group demonstrated overall better static balance across the study period compared to the FLOSS group. The Time × Group interactions for both conditions were non-significant (p > 0.05), indicating that the magnitude of improvement did not differ between the two intervention modalities. For dynamic balance, a significant main effect of Time was found for ANT reach distance in the Y-Balance Test (F (1, 28) = 9.423 p = 0.004, ηp2 = 0.227, large effect size), A significant main effect of Time was also found for the TA in the PM direction (F (1, 28) = 5.527, p = 0.025, ηp2 = 0.147, large effect size). Critically, no significant Time × Group interactions were observed for any reach directions of the YBT (all p > 0.05). Discussion To our knowledge, this randomized controlled trial is among the first to investigate the long-term effects of TF on neuromuscular function around the ankle in patients with CAI. The results showed that, for the vast majority of outcome measures, the TF group did not demonstrate a statistically significant superiority over the balance training-only group. This suggests that adding TF to an 8 weeks balance training regimen did not provide additional benefits. This study was designed to specifically evaluate whether TF could offer additive rehabilitative value. Therefore, balance training was employed as the sole intervention for the CON group. This design was based on two key considerations. First, balance training is an internationally recommended, evidence-based key component of CAI rehabilitation, with its efficacy in improving neuromuscular control and postural stability being well-documented10 . Consequently, the value of any novel adjunctive technique should be judged by its ability to provide additional gain on this solid foundation. Second, this design ensured that the training content was identical between the two groups, with the application (or not) of TF as the sole independent variable. This allows for a more reliable attribution of any between-group differences to the technique itself. A significant Time × Group interaction was observed solely for the TA activation during the ANT reach task. However, post-hoc tests indicated that the magnitude of improvement in this measure did not differ significantly between the two groups. Therefore, the initial hypothesis that the combined intervention would be superior to balance training alone in improving TA and PL function was not supported. Previous studies have reported that TF can induce an immediate facilitatory effect on TA activation in patients with CAI13,17. This acute effect is more likely related to neuromodulatory mechanisms rather than to purely fascial mechanical changes. Available evidence supports that its action aligns with the “gate control theory of pain”: mechanical pressure applied to the skin and fascia stimulates large-diameter Aβ afferent fibers, thereby activating inhibitory interneurons in the dorsal horn of the spinal cord. This inhibits nociceptive signal transmission, reduces pain perception, and alleviates reflex muscle tension32. This mechanism helps improve tissue tolerance to stretch. For patients with CAI, who often exhibit impaired periankle mechanoreceptor function, enhanced peripheral sensory input may promote sensorimotor integration32. It has been confirmed that a single intervention can provide approximately 45 min of sustained improvement in neuromuscular control13. However, the present study did not observe a long-term cumulative effect of this modulatory response. This may suggest that patients with CAI develop rapid tolerance to neuromodulatory interventions, and that the acute benefits may not translate into long-term advantages through simple repeated application. Another potential reason for the lack of significant difference may lie in the intervention modality used in this study. Previous research demonstrating acute benefits applied TF during dynamic, active movements13. In contrast, the present study employed TF during static balance training on a BOSU ball. This posture-stabilization-dominant task mode may have been insufficient to elicit the “fascial shear” effects theorized to occur during dynamic activities. Consequently, the intervention’s effectiveness may have relied more on a local compression mechanism, analogous to blood flow restriction training18, which could have limited its neuromodulatory potential. No significant interaction effects were observed for TA in either the P-L or PM directions. This finding is consistent with existing research indicating that CAI patients exhibit significantly lower TA activation in the ANT direction compared to healthy individuals, with no such difference in the PM direction9, suggesting the existence of direction-specific activation strategies. The YBT-ANT is an “ankle strategy”-dominant task that increases reach distance through ankle dorsiflexion and knee flexion. In contrast, the P-L/PM directions rely more on a “hip strategy” core, depending on pelvic rotation and contralateral leg flexion–extension/adduction-abduction movements. The execution of these movements is governed more by peri-hip muscle groups and trunk-pelvic coordination33. This fundamental difference in task modes may be a key reason why the intervention effects did not manifest in these directions. Furthermore, characteristics of the testing paradigm itself must be considered. On one hand, studies have indicated that in the Star Excursion Balance Test, there is no significant difference in PL activation levels between healthy individuals and CAI patients7,9,34. On the other hand, CAI-related PL activation abnormalities are more frequently observed during dynamic tasks (such as running, landing, and jumping), manifesting as higher activation amplitude, longer duration, or earlier timing35–37. The relatively static YBT paradigm used in this study may not have been sensitive enough to capture the abnormal activation patterns of the PL in CAI patients, which could also be a reason for the lack of significant between-group interaction effects. The significant main effect of Time observed for both TA and PL reaction times in this study is most likely attributable to the effects of the balance training itself. Training on the unstable surface of the BOSU ball provides a high-intensity, continuous stimulus to the periankle proprioceptors, thereby effectively improving inter-muscular coordination and the accuracy of sensory feedback38–40. The primary reason for the non-significant interaction effect for PL reaction time may lie in the testing paradigm itself. The “UP” task elicits postural adjustment through a sudden tilting of the dynamic force plate. This action is primarily driven by concentric contraction of the TA. In this dorsiflexion-dominant mode, the PL, as a plantar flexor-evertor, is reciprocally inhibited. This results in a substantially compressed activation window, making its onset time difficult to capture effectively. Regarding dynamic balance, the results of the present study differ from the findings reported by Hassan et al.41. This discrepancy may stem from differences in the movement-pattern similarity between the training and testing tasks. The training employed by Hassan et al. (e.g., squats, lunges) involved multi-joint coordinated force production, which has higher motor pattern transferability to the hip-knee-ankle coordinated extension required in the YBT, particularly in the P-L and PM directions. In contrast, the present study primarily utilized single-plane ankle stability training, which may have a more limited transfer effect to dynamic tasks like the YBT that demand whole-body coordination and proximal stability. In terms of static postural control, patients with CAI often exhibit impaired sensory reweighting ability due to decreased reliability of ankle proprioceptive input, manifesting as increased center-of-pressure sway velocity, especially under eyes-closed conditions (depriving visual compensation)42. Although both interventions in this study effectively enhanced TA activation, this “local neuromuscular gain” was insufficient to translate into an overall postural control advantage. More notably, blood flow restriction training is often associated with an “immediate post-exercise fatigue-induced increase in postural sway”43. Since TF similarly applies extrinsic compression, it may induce a comparable fatigue response, thereby exerting a certain negative influence on the participants’ balance performance. Compared to blood flow restriction training, the combination of TF and balance training did not demonstrate superiority in improving the overall balance function of CAI patients18. When contrasted with studies on whole-body vibration training, which, although reporting balance improvements, generally show small effect sizes23, the present findings suggest that the therapeutic efficacy of different peripheral sensory enhancement techniques for CAI patients may vary, and their overall benefit may be limited. Previous studies have shown that TF combined with bodyweight resistance training can significantly improve ankle strength44. The potential mechanisms may include TF’s stimulation of mechanoreceptors and its effect on motor unit recruitment14,45. However, the present study employed TF in conjunction with balance training alone, which lacked the stimulus necessary to induce significant structural and metabolic muscular adaptations. This could be a primary reason for the absence of strength improvement. The findings of this study offer clear guidance for clinical rehabilitation practice in CAI. The results indicate that adding TF to a standard 8-week balance training regimen did not yield statistically significant additional functional benefits. This suggests that clinicians should prioritize ensuring the full and correct implementation of evidence-based balance training when planning rehabilitation for patients with CAI, rather than routinely adding TF. Given the current evidence, concentrating limited therapy time and resources on the core intervention (balance training) that has been robustly proven effective, is likely a more efficient approach. These findings can assist clinicians in avoiding the unnecessary application of adjunctive techniques whose long-term additive value has not been substantiated, thereby promoting more streamlined, focused, and cost-effective rehabilitation protocols. This study has several limitations. First, the participants with CAI were not sub-classified. Although the inclusion criteria were based on medical history, symptoms, and functional scores, which aligns with common clinical practice, this led to considerable sample heterogeneity. Specifically, we did not differentiate between functional and mechanical instability, nor did we perform subgroup analysis based on the primary direction of instability. Given that different CAI subtypes involve distinct pathological mechanisms and may respond differently to interventions, the present findings represent an average effect across a mixed population. Therefore, the conclusions are primarily applicable to a broad, unselected CAI cohort. Furthermore, a control group for placebo effect was not included, and the intervention protocol had certain constraints. Neither group received systematic strength restoration training, and the BOSU training used may have been too intense, potentially hindering the improvement of muscular activation patterns. Moreover, only male participants were enrolled, the sample size was relatively small, and the intervention duration was limited to 8 weeks. These factors may have resulted in insufficient statistical power to detect between-group differences of small effect sizes. Finally, balance assessment relied heavily on the Balance Master “UP” task, which is primarily an ankle dorsiflexion-dominated task and places relatively low demand on the plantarflexion-eversion function of the peroneus longus muscle. This may have underestimated its activation level. Additionally, ankle range of motion was not assessed during the Y-balance test, which could affect the interpretability of the results. Due to experimental constraints, surface electromyography signals were not synchronously collected during unstable surface balance tasks, thus preventing real-time analysis of muscle recruitment strategies in participants with CAI. Based on the findings and limitations of this study, future research is recommended to integrate imaging and standardized clinical examinations for the precise phenotyping of CAI patients, aiming to clarify the differential responses of various subtypes to specific interventions, such as the TF technique. Furthermore, experimental designs should incorporate synchronized electromyographic data collection and ensure sufficient sample size, appropriate gender balance, and adequate intervention duration to comprehensively enhance the clinical applicability and statistical power of the research. Conclusion Balance training serves as an effective approach for improving neuromuscular function in patients with CAI, while superimposing TF does not demonstrate long-term additional benefits. Therefore, in the rehabilitation of CAI patients, clinical practice should prioritize the full implementation of treatment protocols centered on balance training. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1