Altered knee joint loading in patients with chronic ankle instability during running Authors: Hyung Gyu Jeon, Kyeongtak Song, Eun Ji Hong, Subeen Wang, Sae Yong Lee, Hoon Kim Published: 2026 ABSTRACT BackgroundGiven that chronic ankle instability (CAI) changes lower extremity biomechanics, potentially leading to greater joint loading, understanding the link between these biomechanical changes and joint loading is critical to identifying early degenerative changes. We aim to compare knee joint contact force (JCF) generated during running between patients with CAI and controls and identify the lower extremity muscles that contribute to these forces.MethodsTwelve patients with CAI and 12 controls performed a running task. A musculoskeletal simulation was employed to estimate the compressive, mediolateral and anteroposterior forces on the tibiofemoral joint. Contributions of the lower extremity muscles and GRF to these forces were also analyzed. One-dimensional statistical parametric mapping and independent t-tests were performed for analysis.ResultsPatients with CAI demonstrated greater posterior and medial knee shear force than controls during running (p < 0.05). Compressive impulse was higher (p = 0.045), but peak compressive force did not differ (p > 0.05). The greater knee JCF in patients with CAI was primarily driven by greater quadriceps muscles (p < 0.05).ConclusionsCompared with the controls, the greater knee joint loading in patients with CAI results from greater quadriceps muscle contributions. These findings highlight the biomechanical link between ankle and knee joint health, suggesting that rehabilitation strategies addressing proximal muscle coordination may help reduce knee joint loading in patients with CAI.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13102-026-01797-4. Background One of the common musculoskeletal injuries is lateral ankle sprains, which is associated with a high prevalence and increasing healthcare costs [1, 2]. Chronic ankle instability (CAI), often resulting from recurrent injuries following a lateral ankle sprain, is characterized by persistent symptoms such as pain, dysfunction, and movement limitations [2, 3]. CAI has been associated with changes in osteo- and arthro-biomechanics resulting from joint laxity and impaired neuromuscular control within the ankle [4]. Previous research has reported that CAI induces an alteration in joint contact forces (JCF), and thus can lead to degenerative changes in cartilage, impairing health-related quality of life [5–8]. Neuromechanical deficits in patients with CAI are not limited to the ankle but involve the knee joint. For example, patients with CAI demonstrate reduced knee flexion and abduction during walking and running [9–12], greater knee abduction angle and moment, and greater internal rotation moment during various landing tasks [13, 14]. Alterations in neuromechanics of joints proximal to the ankle (i.e., knee and hip) in patients with CAI are frequently interpreted as proximal maladaptation because of compensatory movements to address structural/functional deficits in the ankle joint, reduced muscle forces, and altered activation patterns across the knee joint (i.e., disruptions in optimal lower extremity kinetic chain) [15–17]. Although knee joint compensatory alterations occur in patients with CAI, only a limited number of studies have investigated the biomechanical consequences for intraarticular loading [18, 19]. A significant concern is that the altered knee joint loading in patients with CAI, likely an indirect consequence of compensatory strategies for ankle joint deficits, is a risk factor for knee osteoarthritis (OA) [20]. Patients with CAI have been reported to exhibit alterations in knee joint space and tibiofemoral alignment [21, 22], which may alter the cartilage contact areas within the knee and potentially contribute to the development of knee OA [23]. Furthermore, previous research has shown that alterations in gait biomechanics are associated with elevated knee shear forces and signs of knee OA over long-term follow-up [24]. A longitudinal cohort study demonstrated that ankle pain was associated with an increased risk of incident symptomatic radiographic knee OA over 84 months of follow-up [25]. This association between ankle injury and knee joint health is further substantiated by the worse subjective dysfunction scores during both daily living and sports activities, as well as greater knee pain in patients with CAI [26]. Musculoskeletal simulation has been extensively employed as a method for estimating internal joint loads, such as JCF. In addition, it can quantify the contribution of individual muscles to the joint loads, which is crucial for determining targets for muscle strengthening and functional recovery training. For example, Li et al. [18] reported that patients with CAI exhibited higher knee shear and compressive JCF than controls during landing while the differences were not statistically significant. Similarly, Kim and Kipp [19] also demonstrated no difference in knee compression and shear JCF as well as muscle-specific contributions to these loads between patients with CAI and controls during the jump landing task. However, these studies examined single-event landing tasks, whereas running, one of the most common forms of physical activity, involves repetitive loading cycles that accumulate over thousands of steps per session. Given that cumulative joint loading is implicated in the development of knee OA [24, 27] and that patients with CAI exhibit altered lower extremity biomechanics during running [4, 11, 12], examining knee JCF during running in patients with CAI may provide more clinically relevant insight into long-term joint health. Furthermore, because muscle forces are the primary contributors to knee JCF during weight-bearing activities [28–30], identifying which muscles drive the elevated knee JCF in patients with CAI could reveal specific rehabilitation targets to reduce joint loading. Therefore, this study aimed to compare knee JCF during running between patients with and without CAI and to examine the contributions of individual lower extremity muscle groups to these forces through musculoskeletal simulation. We hypothesized that patients with CAI would demonstrate greater knee JCF compared to controls, and that specific lower extremity muscles would contribute more to knee JCF in patients with CAI, given the compensatory proximal muscle recruitment previously reported in this population [15, 16]. Methods Design and participants This cross–sectional study was conducted at a university’s motion-analysis laboratory. A priori sample size was estimated using G*Power (version 3.1) for an independent t-test, based on a previous study investigating ankle JCF in patients with CAI [7]. The analysis indicated that 10 participants per group were required (Cohen’s d = 1.21, α = 0.05, power = 0.80). To ensure sufficient statistical power and account for potential variability, 12 patients with CAI and 12 controls from a local university community participated in this study. All participants read and signed an informed consent agreement that was approved by the university’s institutional review board of Yonsei University (No. 7001988-202402-HR-2175-02). The criteria of the International Ankle Consortium were used to recruit patients with CAI [31]. Individuals with unilateral CAI were included if they met the following criteria: (1) a self-reported history of significant lateral ankle sprain with inflammatory symptoms ≥ 1 (the initial lateral ankle sprain occurred 12 months prior to study); (2) episodes of giving way in injured ankle ≥ 2 (within 6 months prior to study); (3) answering “yes” in the Ankle Instability Instrument (AII) ≥ 5; (4) Activities of Daily Living (ADL) and Sports subscale in the Foot and Ankle Ability Measure (FAAM) < 90% and 80%. If patients with CAI reported bilateral instability and symptoms, the ankle that the participant identified as more symptomatic (i.e., more frequent giving way and greater functional limitation) was selected for the analysis. Controls were recruited following criteria: (1) no history of lateral ankle sprain; (2) no “yes” answers on the AII, and (3) both the FAAM ADL and Sports scores > 99%. All participants participated in regular recreational activities more than three times per week. To ensure that any observed differences were attributable to CAI rather than other conditions affecting running biomechanics, participants were excluded following criteria: (1) a history of lower-extremity surgery or fracture; (2) any history of lower-extremity acute injury within the previous 3 months; (3) any diagnosis of ankle or knee OA during the individual’s lifetime; and (4) any balance disorders, neuropathies, diabetes, or other conditions known to affect the neuromuscular system that would limit participation in sports activities involving running and jumping. Additionally, the Western Ontario and McMaster Universities Arthritis Index (WOMAC) and the Knee Injury and Osteoarthritis Outcome Score (KOOS) were administered to assess knee-related pain and function. Instrumentation Kinematic data sampled at 200 Hz was acquired using a three-dimensional motion capture system consisting of eight cameras (MX-FX20, Vicon Oxford Metrics, UK). A force plate was employed to acquire the ground reaction force (GRF) data at 2,000 Hz (ORG-6 AMTI, Watertown, MA, USA). The muscle activities of the gluteus maximus (Gmax), gluteus medius (Gmed), vastus lateralis, semitendinosus, biceps femoris, tibialis anterior, medial gastrocnemius, peroneus longus, and soleus (SOL) were recorded on the affected limb in patients with CAI and the dominant limb (defined as the preferred kicking leg) in controls using a surface electromyography (EMG) wireless 9-channel system at a sampling frequency of 2000 Hz (Trigno, Delsys Inc., Boston, MA, USA). The location and procedures for EMG sensor placement were performed in accordance with the SENIAM guidelines [32]. All kinematic, kinetic, and EMG data were synchronized and collected using the Vicon Nexus software (version 1.8.5, Vicon Motion Systems Ltd., Oxford, UK). Procedure All participants were provided with spandex shorts, T-shirts, and running shoes (FW9975, Adidas, Herzogenaurach, Germany). Then, 58 reflective skin markers were affixed to their lower arm (styloid process of ulna and radius, and olecranon process), upper arm (anterior and posterior head of humerus, deltoid tuberosity, lateral humerus, and medial and lateral epicondyle of humerus), torso (acromion, C7, T10, sternal notch, and xiphoid process), pelvis (anterior superior iliac spine and posterior superior iliac spine), femur (greater trochanter, medial and lateral epicondyle of femur, and anterior and lateral thigh), tibia (fibula head, tibia tuberosity, and lateral tibia), and foot (1st and 3rd toe, 1st and 5th metatarsal head, and calcaneus). Before acquiring the motion capture data, a static calibration trial was conducted for each participant while they stood with the anatomical posture on the force plate. Participants performed a 5-minute self-paced walk followed by lower-extremity stretching as a warm-up. Following a familiarization period in which participants practiced running at the target speed until they achieved a natural and consistent gait pattern, participants performed running on an 8-m pathway with the shod condition at a speed of approximately 9.66 km/h (± 5%), which was monitored using a handheld stopwatch [33]. The force plate was located at the center of the pathway to ensure data collection during the constant-speed phase. Participants were not instructed to target the force plate; instead, the starting position was adjusted so that natural foot contact occurred on the force plate without altering their stride pattern [34]. A trial was considered successful when the running speed was within the target range (± 5%), the entire foot contacted the force plate, and no visible stride adjustments were observed. Each trial yielded one stance phase cycle for analysis. Trials were repeated until five successful trials were recorded, with a 1-minute rest between trials. Musculoskeletal modelling The musculoskeletal simulation was conducted in OpenSim software version 3.3 (Stanford University, USA; Fig. 1). We employed a full-body musculoskeletal generic model with 37 degrees of freedom and a hill type of 80 muscle-tendon units in the lower extremity [35]. This model, developed by Rajagopal, was validated in a walk-run task using musculotendon parameters derived from experimental measurements of 21 cadaver lower extremities and magnetic resonance images. The knee joint in the Rajagopal model is represented as a single degree-of-freedom joint in which flexion/extension is the independent coordinate, while the tibial rotation and translation relative to the femur are prescribed as coupled functions of the knee flexion angle [35]. This constraint reduces computational complexity but assumes that frontal- and transverse-plane knee motions follow a fixed relationship with sagittal-plane motion. Fig. 1Workflow for biomechanical data acquisition and musculoskeletal modeling for estimating knee joint contact forces. Abbreviations: EMG, electromyography; GRF, ground reaction force; RRA, residual reduction algorithm Workflow for biomechanical data acquisition and musculoskeletal modeling for estimating knee joint contact forces. Abbreviations: EMG, electromyography; GRF, ground reaction force; RRA, residual reduction algorithm We used a scale tool to match each participant’s anthropometric measurements, which were obtained from the static trial marker positions and body mass. In addition to the standard scaling, we also scaled the maximum isometric muscle force in the lower extremity of the generic model using a generic (C) and participant-specific (S) multiplier (i.e., strength coefficient) based on each participant’s mass and height to estimate muscle volume [7, 36]. Component C contains all length-dependent musculoskeletal modeling parameters scaled to each participant’s static pose (Eqs. 1–3).1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F}_{\begin{array}{c}Max\:Isometric\\\:in\:subject\end{array}}\:=\:{F}_{\begin{array}{c}Max\:Isometric\\\:in\:generic\end{array}}\:\times\:\:\:(C\:\:\times\:\:\:S)$$\end{document} 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:S\:\:=\:\:\frac{{MuscleVolume}_{Subject}}{{MuscleVolume}_{Model}}$$\end{document} 3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:MuscleVolume\:\:=\:\:(47\:\:\times\:\:\:BodyMass\:\:\times\:\:\:BodyHeight)\:\:+\:\:1285$$\end{document} The joint angles were estimated using inverse kinematics, which minimized errors between the experimental and virtual markers. Using GRF data and inverse kinematic knee joint angles, inverse dynamics calculated knee flexion/extension moments during the stance phase of running. The residual reduction algorithm was applied to minimize the kinematic and kinetic profile errors. This algorithm was repeated until the residual forces, moment, and kinematic errors were less than the recommended values in the OpenSim guidelines (https://opensimconfluence.atlassian.net/wiki/spaces/OpenSim/pages/53089693/Getting+Started+with+RRA). The individual muscle excitations and forces were subsequently estimated using the computed muscle control [36]. Finally, the joint reaction analysis tool was used to calculate the three-dimensional knee (tibiofemoral) JCF [37]. To determine the contribution of each individual lower extremity muscle group and the GRF to knee JCF, we performed a series of separate joint reaction analysis simulations in which only one muscle group (or the GRF) was active while all other muscle forces were set to zero [7]. The resulting JCF from each simulation was considered the contribution of that specific muscle group or the GRF to the total knee JCF. The muscle groups examined were Gmax, Gmed, Gmin (gluteus minimus), ADD (adductor longus, brevis, and magnus), ILPSO (iliacus and psoas major), VASTI (vastus intermedius, lateralis, and medialis), REC (rectus femoris), BFEM (biceps femoris long and short head), SEMI (semitendinosus and semimembranosus), SOL, GAS (medial and lateral gastrocnemius), PFINV (tibialis posterior, flexor digitorum longus and flexor hallucis longus), PER (peroneus brevis and longus), DORSI (tibialis anterior, extensor digitorum and hallucis longus). The validity of this decomposition was confirmed by verifying that the sum of individual muscle group and GRF contributions equaled the total knee JCF. Furthermore, peak EMG activity for each muscle was normalized to the peak simulated muscle activity [35], and a visual comparison of the onset and offset timings of the simulated muscle activations with the nine lower extremity muscle activation levels from the EMG sensors validated the simulation (Fig. 2). Fig. 2Comparison of normalized muscle activations estimated from musculoskeletal simulation (gray) and measured electromyography in patients with chronic ankle instability (CAI, red) and controls (blue). Data are expressed as the mean (solid) and standard deviation (shaded area) Comparison of normalized muscle activations estimated from musculoskeletal simulation (gray) and measured electromyography in patients with chronic ankle instability (CAI, red) and controls (blue). Data are expressed as the mean (solid) and standard deviation (shaded area) Data reduction Data processing was conducted using the Matlab software (R2021b, Mathworks Inc., Natick, MA, USA). The marker trajectory and GRF data were filtered using a 4th order Butterworth filter with a cut-off frequency of 6 Hz [38]. The EMG data were first bandpass-filtered with a 4th order Butterworth filter at cutoff frequencies of 20 and 450 Hz. Subsequently, we refined the EMG data using a low-pass 4th-order Butterworth filter at a frequency of 10 Hz. We defined the stance phase onset and offset as the points when the vertical GRF exceeded or fell below 20 N, respectively [39]. The stance phase was time-normalized to 100% for each trial. Each of the five running trials per participant was processed individually through the musculoskeletal simulation pipeline, and the resulting knee JCF, individual muscle force contributions, and GRF were averaged across trials for subsequent analysis. The knee JCF, contributions of individual muscle forces, and GRF were normalized based on each participant’s body weight (BW). Three-dimensional knee JCF variables (peak force, time to peak, loading rate, and impulse) were calculated over the entire stance phase according to the following operational definitions. For the compressive force, the peak was defined as the maximum value of the curve. For the anteroposterior and mediolateral shear forces, the peak was defined as the maximum absolute value (positive or negative) of the curve. The time to peak was expressed as a percentage of the stance phase (% stance). The loading rate (BW/s) was defined as the slope from the first data point of the stance phase to the peak force, representing how rapidly the knee joint experiences mechanical loading. The impulse (BW·s) was calculated as the area under the force-time curve over the entire stance phase using the trapezoidal method, representing the total cumulative loading during stance. For muscle force contributions, the value at the instant of peak knee JCF was extracted from each trial and averaged across trials. Statistical analysis All statistical analyses were performed using Matlab software and one-dimensional statistical parametric mapping (SPM). A SPM independent t-test was employed to compare knee angle and moment, knee JCF, and muscle contribution to JCF data in time series between patients with CAI and controls. The normality of all dependent variables was verified using the Kolmogorov-Smirnov test [40], and no significant deviations from normality were observed (p > 0.05). Demographics, knee JCF profile (peak force, time to peak, loading rate, and impulse), and muscle contribution to peak JCF were compared between groups using the independent t-test. Effect sizes were calculated using Cohen’s d, with values of 0.2, 0.5, and 0.8 interpreted as small, medium, and large, respectively. The statistical significance level was set at 0.05. Results Participant characteristics There were no significant differences between patients with CAI and controls in age, body mass, height, International Physical Activity Questionnaire (IPAQ), or Tegner activity level (p > 0.05; Table 1). As expected, patients with CAI reported significantly higher AII scores and lower FAAM-ADL and FAAM-Sports scores than controls (p < 0.001). Patients with CAI also demonstrated worse WOMAC (p = 0.025) and KOOS (p = 0.011) scores compared to controls. Among the KOOS subscales, patients with CAI reported significantly worse scores in pain (p = 0.047), other symptoms (p = 0.019), sports/recreation function (p = 0.035), and knee-related quality of life (p = 0.020), whereas ADL function did not differ between groups (p = 0.113). Table 1Study participant demographicsSexaPatients with CAI(n = 12)Controls(n = 12)P valueM = 5, W = 7M = 5, W = 71.0Age25.92 ± 6.1923.25 ± 4.200.230Body Mass (kg)72.94 ± 18.4764.07 ± 10.710.164Height (cm)168.95 ± 11.83166.13 ± 7.780.497AII7.33 ± 1.440 ± 0< 0.001FAAM-ADL83.73 ± 3.64100 ± 0< 0.001FAAM-Sports66.42 ± 11.54100 ± 0< 0.001WOMAC11.58 ± 13.391.50 ± 2.320.025KOOS overall81.28 ± 12.9393.53 ± 7.590.011KOOS subscaleKOOS pain85.20 ± 13.9394.90 ± 7.840.047KOOS other symptoms72.00 ± 20.0289.30 ± 12.650.019KOOS function in ADL92.20 ± 8.4097.40 ± 7.180.113KOOS function in sports/recreation83.30 ± 16.5695.40 ± 8.650.035KOOS knee related QoL73.40 ± 20.6690.60 ± 11.770.020IPAQ (METs)3482.25 ± 2493.723778.67 ± 1915.610.747Tegner Activity Scale6.67 ± 1.786.58 ± 1.510.902a This variable was subjected to chi-square analysisAbbreviations: ADL activity of daily life, AII Ankle Instability Instrument, CAI chronic ankle instability, FAAM Foot and Ankle Ability Measure, IPAQ International Physical Activity Questionnaire, KOOS Knee Injury and Osteoarthritis Outcome Score, METs metabolic equivalents of task, QoL quality of life, WOMAC Western Ontario and McMaster Universities Arthritis Index Table 1Study participant demographicsSexaPatients with CAI(n = 12)Controls(n = 12)P valueM = 5, W = 7M = 5, W = 71.0Age25.92 ± 6.1923.25 ± 4.200.230Body Mass (kg)72.94 ± 18.4764.07 ± 10.710.164Height (cm)168.95 ± 11.83166.13 ± 7.780.497AII7.33 ± 1.440 ± 0< 0.001FAAM-ADL83.73 ± 3.64100 ± 0< 0.001FAAM-Sports66.42 ± 11.54100 ± 0< 0.001WOMAC11.58 ± 13.391.50 ± 2.320.025KOOS overall81.28 ± 12.9393.53 ± 7.590.011KOOS subscaleKOOS pain85.20 ± 13.9394.90 ± 7.840.047KOOS other symptoms72.00 ± 20.0289.30 ± 12.650.019KOOS function in ADL92.20 ± 8.4097.40 ± 7.180.113KOOS function in sports/recreation83.30 ± 16.5695.40 ± 8.650.035KOOS knee related QoL73.40 ± 20.6690.60 ± 11.770.020IPAQ (METs)3482.25 ± 2493.723778.67 ± 1915.610.747Tegner Activity Scale6.67 ± 1.786.58 ± 1.510.902a This variable was subjected to chi-square analysisAbbreviations: ADL activity of daily life, AII Ankle Instability Instrument, CAI chronic ankle instability, FAAM Foot and Ankle Ability Measure, IPAQ International Physical Activity Questionnaire, KOOS Knee Injury and Osteoarthritis Outcome Score, METs metabolic equivalents of task, QoL quality of life, WOMAC Western Ontario and McMaster Universities Arthritis Index Study participant demographics a This variable was subjected to chi-square analysis Abbreviations: ADL activity of daily life, AII Ankle Instability Instrument, CAI chronic ankle instability, FAAM Foot and Ankle Ability Measure, IPAQ International Physical Activity Questionnaire, KOOS Knee Injury and Osteoarthritis Outcome Score, METs metabolic equivalents of task, QoL quality of life, WOMAC Western Ontario and McMaster Universities Arthritis Index Knee joint angles and moments Compared to controls, patients with CAI demonstrated greater knee flexion angle in the time series (27%–72% of the stance phase, p < 0.001) and peak knee flexion angle (p = 0.001; Supplementary 1). However, the time to peak knee flexion did not vary between the groups (p = 0.469). The groups exhibited comparable knee extension moments in the time series (p > 0.05) and peak knee extension moment (p = 0.062) (Supplementary 2). Patients with CAI exhibited a slower time to peak knee extension moment than controls (p = 0.041). Knee joint contact forces in time series Patients with CAI exhibited greater medial force data in the time series than controls (26%–82% of the stance phase, p = 0.002; Figure 3A). Although patients with CAI demonstrated greater anterior (3%–7%, p = 0.040) and posterior (76%–77%, p = 0.049; and 89%–100%, p = 0.017) force data in the time series during the stance phase than controls, the magnitude of these differences was not substantial. Knee joint contact force parameters In the anteroposterior direction, patients with CAI exhibited a higher peak posterior force (p = 0.008), loading rate (p = 0.022), and impulse (p = 0.008) during the stance phase than controls (Table 2). Although there was no significant difference in the time-series of the compressive force between the two groups (p> 0.05), patients with CAI exhibited a greater compressive impulse (p = 0.045) than controls. For the mediolateral direction, patients with CAI demonstrated a higher peak medial force (p = 0.002), faster time to peak (p = 0.007), and greater loading rate (p = 0.004) and impulse (p = 0.004) than controls.Table 2Profile of knee joint contact force parameters in patients with chronic ankle instability and controlsVariablesPatients with CAIControlst valuep valueESa(95% CI)Anteroposterior shear loadsPeak force (BW)−6.35 ± 1.35−5.06 ± 0.74−2.8950.008−1.18(− 2.04, − 0.30)Time to peak(% stance)63.83 ± 2.7962.0 ± 2.521.6880.1050.69(− 0.14, 1.51)Loading rate (BW/s)−0.10 ± 0.02−0.08 ± 0.01−2.4690.022−1.01(− 1.85, − 0.14)Impulse(BW∙s)−344.33 ± 70.88−275.47 ± 39.78−2.9350.008−1.20(− 2.06, − 0.31)CompressiveloadsPeak force (BW)9.15 ± 1.718.13 ± 0.881.8320.0810.75(− 0.12, 1.59)Time to peak(% stance)43.83 ± 6.9743.0 ± 2.300.3940.6980.16(− 0.65, 0.96)Loading rate (BW/s)0.21 ± 0.040.19 ± 0.021.6560.1170.67(− 0.18, 1.51)Impulse(BW∙s)609.50 ± 110.81534.67 ± 50.972.1250.0450.87(− 0.03, 1.73)Mediolateral shear loadsPeak force (BW)−1.79 ± 0.80−0.91 ± 0.30−3.5180.002−1.44(− 2.33, − 0.52)Time to peak(% stance)42.33 ± 2.9338.75 ± 2.962.9800.0071.22(0.33, 2.08)Loading rate (BW/s)−0.04 ± 0.02−0.02 ± 0.01−3.2390.004−1.32(− 2.20, − 0.42)Impulse(BW∙s)−99.41 ± 56.61−43.76 ± 22.33−3.1680.004−1.29(− 2.17, − 0.40)a effect size was calculated as Cohen’s dAbbreviations: BW body weight, CAI chronic ankle instability, CI confidence interval Table 2Profile of knee joint contact force parameters in patients with chronic ankle instability and controlsVariablesPatients with CAIControlst valuep valueESa(95% CI)Anteroposterior shear loadsPeak force (BW)−6.35 ± 1.35−5.06 ± 0.74−2.8950.008−1.18(− 2.04, − 0.30)Time to peak(% stance)63.83 ± 2.7962.0 ± 2.521.6880.1050.69(− 0.14, 1.51)Loading rate (BW/s)−0.10 ± 0.02−0.08 ± 0.01−2.4690.022−1.01(− 1.85, − 0.14)Impulse(BW∙s)−344.33 ± 70.88−275.47 ± 39.78−2.9350.008−1.20(− 2.06, − 0.31)CompressiveloadsPeak force (BW)9.15 ± 1.718.13 ± 0.881.8320.0810.75(− 0.12, 1.59)Time to peak(% stance)43.83 ± 6.9743.0 ± 2.300.3940.6980.16(− 0.65, 0.96)Loading rate (BW/s)0.21 ± 0.040.19 ± 0.021.6560.1170.67(− 0.18, 1.51)Impulse(BW∙s)609.50 ± 110.81534.67 ± 50.972.1250.0450.87(− 0.03, 1.73)Mediolateral shear loadsPeak force (BW)−1.79 ± 0.80−0.91 ± 0.30−3.5180.002−1.44(− 2.33, − 0.52)Time to peak(% stance)42.33 ± 2.9338.75 ± 2.962.9800.0071.22(0.33, 2.08)Loading rate (BW/s)−0.04 ± 0.02−0.02 ± 0.01−3.2390.004−1.32(− 2.20, − 0.42)Impulse(BW∙s)−99.41 ± 56.61−43.76 ± 22.33−3.1680.004−1.29(− 2.17, − 0.40)a effect size was calculated as Cohen’s dAbbreviations: BW body weight, CAI chronic ankle instability, CI confidence interval Profile of knee joint contact force parameters in patients with chronic ankle instability and controls −1.18 (− 2.04, − 0.30) Time to peak (% stance) 0.69 (− 0.14, 1.51) −1.01 (− 1.85, − 0.14) Impulse (BW∙s) −1.20 (− 2.06, − 0.31) Compressive loads 0.75 (− 0.12, 1.59) Time to peak (% stance) 0.16 (− 0.65, 0.96) 0.67 (− 0.18, 1.51) Impulse (BW∙s) 0.87 (− 0.03, 1.73) −1.44 (− 2.33, − 0.52) Time to peak (% stance) 1.22 (0.33, 2.08) −1.32 (− 2.20, − 0.42) Impulse (BW∙s) −1.29 (− 2.17, − 0.40) a effect size was calculated as Cohen’s d Abbreviations: BW body weight, CAI chronic ankle instability, CI confidence interval Muscle contributions to knee joint contact force Figure 3C and Figure 3D depict the muscle contributions to knee JCF in three planes. Patients with CAI displayed a significantly greater contribution to the anterior-posterior shear force from the VASTI muscle group (p< 0.001) than controls (Figure 4). Patients with CAI also exhibited a significantly greater contribution of the Gmin (p = 0.045), VASTI (p = 0.001), and BFEM (p = 0.025) to the knee compressive force than controls, but less contribution of ADD (p = 0.010). Patients with CAI exhibited a significantly greater contribution to the knee medial force from the Gmax (p = 0.041) and VASTI (p < 0.001) muscle groups. Discussion This study examined knee JCF between patients with CAI and controls during the stance phase of running. Our primary findings support our hypothesis that patients with CAI would exhibit higher JCF patterns and altered joint loading parameters compared to controls. Our findings also suggest that individual lower-extremity muscle contributions to peak knee JCF would differ between groups. Last, the results demonstrate that the greater knee JCF in patients with CAI was driven primarily by muscle forces, especially those of the quadriceps muscles. These findings provide preliminary evidence for a biomechanical pathway by which repetitive loading during running may contribute to long-term knee joint health concerns in patients with CAI. We discovered that patients with CAI experienced greater mediolateral and anteroposterior shear forces in the knee compared to controls (see Fig. 3A). Additionally, the knee compressive impulse was greater in patients with CAI than in controls. These altered knee loading profiles showed in patients with CAI might result from knee flexion angle and moment changes. In our simulation, greater knee flexion angle and slower time to peak knee extension moment were found in patients with CAI than controls during the middle stance phase of running (see Supplementary 1 and 2). These knee biomechanics, characterized by greater eccentric muscle contractions, may allow for greater absorption and control of GRF; however, they may also contribute to greater knee JCF [41]. Recent studies also supported this notion by reporting that patients with CAI exhibit altered knee kinematics and kinetics and rely more on the knee joint for force absorption during dynamic tasks [9, 42, 43]. Fig. 3Comparison of knee joint contact forces and muscle-specific contributions during the stance phase of running in patients with chronic ankle instability (CAI) and controls. A Group mean time-series curves for anteroposterior shear, compressive, and mediolateral shear knee joint contact forces. Red lines indicate patients with CAI, and blue lines indicate controls. Shaded areas represent standard deviations. B Statistical parametric mapping results comparing knee joint contact force waveforms between groups. C, D Contributions of individual lower-extremity muscle groups and ground reaction force to knee joint contact force in patients with CAI and controls, respectively. Force data were time-normalized to 100% of the stance phase and normalized to body weight. Abbreviations: ADD, adductor longus, brevis, and magnus; BFEM, biceps femoris long and short head; BW, body weight; CAI, chronic ankle instability; DORSI, tibialis anterior, extensor digitorum, and hallucis longus; GAS, medial and lateral gastrocnemius; Gmax, gluteus maximus; Gmed, gluteus medius; Gmin, gluteus minimus; GRF, ground reaction force; ILPSO, iliacus and psoas major; JCF, joint contact force; PER, peroneus brevis and longus; PFINV, tibialis posterior, flexor digitorum longus, and flexor hallucis longus; REC, rectus femoris; SEMI, semitendinosus and semimembranosus; SOL, soleus; SPM, statistical parametric mapping; VASTI, vastus intermedius, lateralis, and medialis Comparison of knee joint contact forces and muscle-specific contributions during the stance phase of running in patients with chronic ankle instability (CAI) and controls. A Group mean time-series curves for anteroposterior shear, compressive, and mediolateral shear knee joint contact forces. Red lines indicate patients with CAI, and blue lines indicate controls. Shaded areas represent standard deviations. B Statistical parametric mapping results comparing knee joint contact force waveforms between groups. C, D Contributions of individual lower-extremity muscle groups and ground reaction force to knee joint contact force in patients with CAI and controls, respectively. Force data were time-normalized to 100% of the stance phase and normalized to body weight. Abbreviations: ADD, adductor longus, brevis, and magnus; BFEM, biceps femoris long and short head; BW, body weight; CAI, chronic ankle instability; DORSI, tibialis anterior, extensor digitorum, and hallucis longus; GAS, medial and lateral gastrocnemius; Gmax, gluteus maximus; Gmed, gluteus medius; Gmin, gluteus minimus; GRF, ground reaction force; ILPSO, iliacus and psoas major; JCF, joint contact force; PER, peroneus brevis and longus; PFINV, tibialis posterior, flexor digitorum longus, and flexor hallucis longus; REC, rectus femoris; SEMI, semitendinosus and semimembranosus; SOL, soleus; SPM, statistical parametric mapping; VASTI, vastus intermedius, lateralis, and medialis The greater knee JCF observed in patients with CAI warrants consideration in the context of long-term knee joint health, as cumulative joint loading has been proposed to contribute to the initiation and progression of knee OA [24, 27]. The knee mediolateral shear and compressive loads induce cartilage damage and may potentially expedite the progression of knee OA [24, 44]. Lynn et al. [24] supported this notion by demonstrating that alterations in gait biomechanics can elevate mediolateral knee shear forces, potentially leading to OA symptoms over long-term follow-up. Similarly, Fong et al. [45] reported that runners with knee OA exhibited greater total knee JCF than healthy controls during running, further supporting the association between elevated knee joint loading and OA. Therefore, if patients with CAI continue to engage in physical activity without any interventions, the cumulative loading from elevated knee loading across daily steps over time may lead to earlier onset of knee joint degeneration compared to healthy individuals. Our patient-oriented outcomes, including Western Ontario and McMaster Universities Arthritis Index and Knee Injury and Osteoarthritis Outcome Score scores, further supported these biomechanical findings by indicating that patients with CAI experienced greater knee pain and functional limitations compared to controls. To confirm this hypothesis, future prospective cohort studies should investigate whether a history of ankle injury increases the incidence of knee osteoarthritis and other degenerative changes. The higher knee JCF in patients with CAI compared to the controls may be developed from the greater contribution of individual muscles. Mainly, the VASTI muscles contributed to greater knee JCF, in line with previous findings showing that quadriceps muscle forces were the primary contributors to knee JCF during weight-bearing activities [28, 29]. The greater contribution from proximal muscles to the knee JCF in patients with CAI may be because of the compensatory strategy of recruiting neighboring muscles more to overcome the CAI-associated neuromuscular deficits of distal muscles [46]. This compensatory proximal muscle recruitment may be triggered by altered ankle joint kinematics adopted to protect against recurrent sprains [4, 16, 47], which may limit the force-generating capacity of distal muscles and shift the demand to proximal muscle groups. In previous studies, patients with CAI with lower SOL and PER volume and impaired function exhibited greater quadriceps recruitment, which the authors interpreted as a compensatory strategy for distal muscle deficits [16, 48]. However, to understand the mechanisms by which greater proximal muscle recruitment elevates JCF, additional studies are warranted to investigate the direction of the line of pull of individual muscles, as both the magnitude and vector of muscle force influence their contribution to JCF. The contributions of lower extremity muscles to knee JCF were substantially greater than those of the GRF in both groups (see Fig. 4). This finding supports previous studies demonstrating that muscle force plays a dominant role in knee joint loads, especially during more dynamic tasks [49]. However, GRF should also be considered as a factor in reducing the elevated knee joint loads in patients with CAI. The previous works demonstrating greater GRF in patients with CAI compared to controls during walking and running corroborate our findings of greater knee joint loads [9, 11]. Typically, during walking or running, patients with CAI contact the ground with stiffened joints, requiring movement strategies that protect against recurrent sprains (e.g., reduced plantar flexion angle) [47]. Therefore, implementing rehabilitation programs aimed at improving shock absorption movement (e.g., reprogramming) in patients with CAI could be beneficial for mitigating knee joint loads and preserving overall lower extremity joint health. Fig. 4Statistical comparison of lower extremity muscle contributions to peak knee joint anterior-posterior shear force, compressive force, and medio-lateral shear force between patients with chronic ankle instability (CAI, red) and controls (blue).* p < .05; ** p < .01; *** p < .001. Abbreviations: ADD, adductor longus, brevis, and magnus; BFEM, biceps femoris long and short head; BW, body weight; CAI, chronic ankle instability; DORSI, tibialis anterior, extensor digitorum, and hallucis longus; GAS, medial and lateral gastrocnemius; Gmax, gluteus maximus; Gmed, gluteus medius; Gmin, gluteus minimus; GRF, ground reaction force; ILPSO, iliacus and psoas major; JCF, joint contact force; PER, peroneus brevis and longus; PFINV, tibialis posterior, flexor digitorum longus, and flexor hallucis longus; REC, rectus femoris; SEMI, semitendinosus and semimembranosus; SOL, soleus; and VASTI, vastus intermedius, lateralis, and medialis Statistical comparison of lower extremity muscle contributions to peak knee joint anterior-posterior shear force, compressive force, and medio-lateral shear force between patients with chronic ankle instability (CAI, red) and controls (blue).* p < .05; ** p < .01; *** p < .001. Abbreviations: ADD, adductor longus, brevis, and magnus; BFEM, biceps femoris long and short head; BW, body weight; CAI, chronic ankle instability; DORSI, tibialis anterior, extensor digitorum, and hallucis longus; GAS, medial and lateral gastrocnemius; Gmax, gluteus maximus; Gmed, gluteus medius; Gmin, gluteus minimus; GRF, ground reaction force; ILPSO, iliacus and psoas major; JCF, joint contact force; PER, peroneus brevis and longus; PFINV, tibialis posterior, flexor digitorum longus, and flexor hallucis longus; REC, rectus femoris; SEMI, semitendinosus and semimembranosus; SOL, soleus; and VASTI, vastus intermedius, lateralis, and medialis Clinically, our findings recommend reducing knee loads in patients with CAI. Adapting muscle coordination patterns to shift load-sharing from the knee extensors to functionally overlapping muscles, such as the hip extensors and abductors that also contribute to body weight support during running [50], could be an effective intervention for reducing knee joint loading [29]. During the initial phase of running, the quadriceps provide half of the BW support, while the Gmax, Gmed, and ADD provide the remaining half [50]. Since the VASTI muscles contribute more to the knee JCF in all directions than the hip extensors and abductors (see Fig. 4), shifting load-sharing toward the hip musculature could potentially help reduce knee joint loading. Previous studies have reported that patients with CAI tend to exhibit reduced hip muscle strength and altered hip muscle activation patterns compared to healthy controls [51, 52]. Although hip muscle strength was not directly measured in the present study, these findings suggest that the hip musculature may be a modifiable target in this population. Therefore, hip strengthening and/or neuromuscular retraining may encourage greater recruitment of the hip extensors and abductors, thereby reducing reliance on the VASTI muscles and limiting knee joint loading. Gmed strengthening may also prevent the contralateral pelvic drop, thereby limiting the medial shift of the GRF moment arm and medial knee loading [53]. Moreover, hip-focused exercise programs have been shown to reduce symptoms and knee load in patients with knee OA [54, 55], supporting the potential transfer of this approach to CAI populations at risk of knee joint degeneration. In addition, hip-focused interventions in patients with CAI have been shown to improve static and dynamic balance and self-reported ankle function [56, 57], suggesting that such programs may benefit overall lower extremity joint health. Therefore, incorporating hip strengthening exercises into a traditional ankle rehabilitation program may help patients with CAI to enhance lower extremity joint health. Simulation studies comparing knee JCF during running between individuals with and without CAI are scarce. However, knee JCF levels are comparable to those in studies evaluating the same task in healthy participants [28]. Specifically, the JCF in the time series and peak values observed in our control group were similar to those verified in previous studies conducted on healthy males during running (e.g., a compressive force of 8 BW occurred during around 40% of the stance phase) [28]. Consequently, our simulation results provide reasonable estimates of knee joint loading trends in patients with CAI, although the absolute values should be interpreted with caution given the modeling constraints described above. These findings identify altered knee JCF as a potential biomechanical pathway by which CAI may influence long-term knee joint health, although direct evidence linking CAI to greater knee OA risk remains to be established in prospective studies. However, this study has several limitations. First, the knee joint of the model is represented as a single degree-of-freedom joint in which frontal- and transverse-plane motions are prescribed as coupled functions of the knee flexion angle [35]. As the estimation of joint contact forces is sensitive to the joint kinematic definition [35], the mediolateral and anteroposterior shear forces reported in this study should be interpreted with this constraint in mind. Because abnormal running parameters in the frontal and transverse planes are associated with OA development [24], future studies employing a model incorporating independent abduction/adduction and internal/external rotation degrees of freedom may provide more detailed information on knee loads [49]. Second, we standardized running speed to ensure that between-group differences in knee JCF were not confounded by variations in running velocity [58]. However, this approach may have constrained individual natural running strategies and reduced ecological validity. Third, we estimated the loading at a single point on the tibiofemoral joint without considering the distribution across different compartments. Given that knee joint loads are disproportionately distributed between the medial and lateral compartments, examining individual compartmental loads may offer valuable insights into OA-related joint mechanics [59]. Fourth, although maximum isometric forces were estimated using body mass and height-based scaling coefficients [7, 36], we did not directly measure individual muscle forces or volumes. Furthermore, because this scaling was applied uniformly to all lower extremity muscles, the between-muscle maximum isometric force ratios remained those of the generic model. This approach may not capture individual muscle architecture or CAI-specific adaptations such as intrinsic and extrinsic foot and ankle muscle atrophy [48]. Fifth, mechanical ankle joint laxity was not directly measured; therefore, variability in the degree of laxity among patients with CAI may have influenced the observed knee JCF. Conclusions In conclusion, the anteroposterior and mediolateral knee shear forces were substantially higher in patients with CAI than in controls during the stance phase of running. The knee compressive force was comparable between the groups, while the knee compressive impulse was higher in patients with CAI. The quadriceps muscles may have contributed more to the greater knee JCF in patients with CAI. These altered knee loading patterns may represent a potential biomechanical pathway warranting investigation in prospective studies of long-term knee joint health in this population. Supplementary Information Supplementary Material 1. Supplementary Material 1.