Frontal plane ankle stiffness increases with axial load independent of muscle activity

INTRODUCTION

Ankle sprains are the most common musculoskeletal injury, accounting for at least 3 million hospital visits per year in the United States (Doherty et al., 2014). Nearly 85% of sprains occur due to excessive inversion of the ankle (Andersen et al., 2004). The ankle requires sufficient frontal-plane stability to avoid excessive inversion and minimize the likelihood of sprains. Both unloaded passive structures and muscle activation help maintain frontal-plane ankle stability. (Ashton-Miller et al., 1996; Leardini et al., 2000). Additionally, cadaveric studies suggest that axial-loading-sensitive mechanisms are a major contributor to ankle stability during weight-bearing conditions. This axial-loading-sensitive mechanism could be important as the ankle experiences up to 2–3 times body weight during many locomotor activities (Firminger et al., 2018). Identifying the sensitivity of ankle stability to axial loading is important to determine whether an impairment in this mechanism could be a causative factor in ankle sprains.

While frontal-plane ankle stability has been investigated during weight-bearing conditions, none have isolated the contributions to stability made by axial loading. Ankle stability has been quantified by computing the impedance, the dynamic relationship between an imposed movement and the resultant resistive torque (Kearney and Hunter, 1990). The impedance has often been modeled as a spring-mass-damper system quantified by its stiffness, inertia, and viscosity respectively (Ludvig and Kearney, 2007; Ludvig et al., 2022; Rouse et al., 2014). During standing, frontal-plane stiffness increases with weight-bearing, but is accompanied by an increase in muscle activation (Matos et al., 2021). Muscle activation is known to substantially increase ankle stiffness, (Jakubowski et al., 2022; Lee et al., 2014a; Lee et al., 2014b; Mirbagheri et al., 2000). Thus, the sensitivity of ankle stability to axial loading independent of muscle activity remains unknown.

Cadaveric studies have shown that frontal-plane ankle stability is sensitive to axial loading, but whether this sensitivity is present in vivo is unclear. Under axial load transmitted via the tibia, the ankle undergoes less rotation in all planes (McCullough and Burge, 1980; Stiehl et al., 1993; Stormont et al., 1985). This reduction in range of motion for a fixed torque is equivalent to an increase in ankle stiffness (McCullough and Burge, 1980; Stiehl et al., 1993). However, these cadaveric studies isolated motion to the talocrural joint (Stiehl et al., 1993; Watanabe et al., 2012), discounting rotation at the subtalar joint, where most of the inversion occurs under in vivo conditions (Arndt et al., 2004). Additionally, these studies remove muscle and tendon, thus altering the biomechanics of the ankle. Thus, it remains unknown whether in vivo ankle stability is sensitive to axial load.

The objective of this study was to determine the effect of passive axial load on frontal-plane ankle stiffness. We hypothesized that ankle stiffness would increase with increasing passive axial load. We had subjects seated as a load was applied to the ankle ranging from 10% to 50% body weight through the controlled application of force on the knee. Small frontal plane rotations were applied to the ankle to estimate stiffness. We tested our hypothesis by determining if there was a significant increase in ankle stiffness with increasing axial load. To ensure that any changes in stiffness could only be attributed to the passive axial load, we tested whether the activity of the surrounding muscles also increased with increasing axial load. If ankle stiffness is sensitive to passive axial loading, it would suggest that this passive-loading mechanism may substantially contribute to ankle stability during functional weight-bearing conditions.

METHODS

Participants

Twenty individuals (9 males, 11 females; age = 26.9±4.1 (mean ± standard deviation), body mass: 70±15 kg, height: 1.70±0.11 m) participated in this study. Participants were excluded if they had a known neurological disorder, connective tissue disorder, orthopeadic diagnosis, lower extremity bone fracture that required realignment, or history of significant ankle sprains in the past 5 years—defined as resulting in a loss of at least one day of desired physical activity. Each participant was right-foot dominant. Each participant’s right ankle was tested. Ethical approval was received from Northwestern University’s Institutional Review Board (STU00009204 and STU00215245). Written informed consent was obtained prior to testing.

Experimental Set-up

To quantify changes in frontal-plane ankle stiffness due to imposed axial loading, we systematically applied axial loads on their leg and perturbed their ankle as they were seated in an adjustable chair (Biodex Medical Systems, Inc. Shirley, NY, USA) (Figure 1A). Each participant had a custom-made fiberglass cast that connected to the electric rotary motor (BSM90N-3150AF, Baldor, Fort Smith, AR, USA). The cast was attached to the foot plate so that the axis of rotation was perpendicular and passed through both the transmalleolar axis and the long axis of the tibia (Brockett and Chapman, 2016) (Figure 1B–C). Knee angle and ankle angle were set at 90° of flexion using a goniometer. Straps were tightened around the torso, waist, and right leg to limit movements. Axial loads were applied using a C-shaped pad positioned just proximal to the knee joint. Axial loads and ankle torques were determined from a six-degree-of-freedom load cell (45E15A4, JR3, Woodland, CA, USA) located underneath the participant’s foot. Ankle angle was simultaneously recorded using a 24-bit quadrature encoder card (PCI-QUAD04, Measurement Computing, Norton, MA, USA). Data acquisition and control of the rotary motor were executed using xPC target (Mathworks, Natick, MA).

Figure 1.

A. Experimental Set-Up. Participants were seated in an adjustable chair and had their ankles fixed to a load cell and rotary motor via a custom-made foot cast. External axial loading was applied via a loading device placed above the knee. This load was adjusted by the experimenter, aided visually by a feedback monitor displaying the applied axial load, as measured by the load cell. B. Apparatus between motor and foot. The participant’s cast is attached to the apparatus via aluminum posts on the side of the malleoli. In this way, the ankle’s axis of rotation is aligned with the axis of rotation of the rotary motor. C. Diagram of how the ankle’s frontal plane axis of rotation is aligned to the motor’s axis of rotation. The motor’s axis of rotation is perpendicular and passes through both the transmalleolar axis and the long axis of the tibia. The red line indicates the axis of rotation of the motor.

Electromyography (EMG) data were collected to ensure the muscle activity did not change with axial loading. EMGs were collected using surface electrodes (Bagnoli, Delsys Inc, Boston, MA) from the following muscles: tibialis anterior, lateral gastrocnemius, medial gastrocnemius, soleus, peroneus longus, and peroneus brevis, all of which can contribute to ankle torque (Brockett and Chapman, 2016). Standard skin preparation methods were used before adhering each electrode to the skin (Besomi et al., 2019; Merletti and Muceli, 2019). Palpation of the muscle belly during voluntary contraction was used to determine appropriate electrode placement. This was confirmed during isometric maximum voluntary contraction trials, after which EMGs were inspected for activity that would correspond to the contraction. Electrode placements were consistent with SENIAM standards (Hermens et al., 2000). EMG measurements were amplified (Delsys Bagnoli, Natick, MA) to maximize the range of the data acquisition system. All analog data were passed through an anti-aliasing filter (500 Hz using a 5-pole Bessel filter) and sampled at 2.5 kHz (PCI-DAS1602/16, Measurement Computing, Norton, MA, USA).

Protocol

Maximum voluntary contraction (MVC) trials were collected to normalize the EMG signals to compare across participants. Participants completed maximum voluntary contractions in 6 directions: inversion/eversion, dorsiflexion/plantarflexion, and internal/external rotation. Each MVC was repeated twice.

The study was designed to determine the effect of passive load on ankle stiffness. Participants were instructed to remain relaxed during each trial. Axial loads were applied up to a maximum of 50% body weight (BW). The resting condition refers to the participant being seated without the external load applied; thereby including only the weight of the foot and shank. Across all participants, this load was 10.9±1.3% body weight (mean ± standard deviation). The loading conditions were resting, 20% BW, 30% BW, 40% BW, 50% BW. We set the upper limit of the tested load to 50% of the participant’s body weight to minimize discomfort in compressing the soft tissues above the knee. Subjects did not report any discomfort at the ankle, as the tested loads were well below what is experienced during activities involving single-legged stance. Each loading condition was repeated twice. Following data collection, we ensured that horizontal forces remained nearly constant with axial loading; both the anterior-posterior force and the medial-lateral force were less than 3% of axial load. The order of the experimental trials was randomized. To estimate the frontal-plane ankle stiffness, we applied rotational perturbations in the frontal plane. The perturbations were a pseudorandom binary sequence of amplitude 0.03 radians, velocity 3.0 rad/s, and 0.150 switching time (Matos et al., 2021). Each trial lasted 65 seconds. We also estimated the impedance of the equipment and cast with a trial in which only the cast was perturbed. All measures of impedance were normalized by the participant’s weight. To determine whether the axial-load stiffness relationship was sensitive to perturbation amplitude, we completed a secondary experiment in a subset of participants (N = 8). We tested perturbation amplitudes between 0.02 and 0.06 radians in increments of 0.01 radians at resting load, 30% body weight, and 50% body weight.

Quantifying Ankle Impedance

We quantified the stability of the ankle in the frontal plane by computing its impedance. First, we decimated the data to 50 Hz. Then, we characterized the impedance of the ankle by computing a non-parametric impulse response function (IRF) between the measured ankle angle and the measured torque in the frontal plane (Figure 2) (Matos et al., 2021). The stiffness was computed by integrating the IRF (Kearney and Hunter, 1990). Damping and inertia were estimated by fitting the impedance to a 2^nd-order model using a constrained optimization such that the stiffness of the 2^nd-order model matched the stiffness as measured by integrating the IRF (Kearney and Hunter, 1990). The constrained optimization was performed using the Levenburg-Marquardt nonlinear least squares algorithm (Kearney et al., 1997; Mirbagheri et al., 2000; Seber, 2003). Separately, we computed the stiffness, viscosity, and inertia of the cast and apparatus from the cast-only trial, and removed these measures from the respective measures of the experimental trials. For the experimental trials, the IRF accounted for 94.0±3.6% of the measured torque (Figure 3). The parametric 2^nd-order model accounted for 92.2 ± 7.7% (mean ± standard deviation) of the measured torque. This decrease in variance accounted for between the IRF and the parametric method was small but statistically significant (Δ = 1.8% VAF, t₂₀₈ = −5.26, p < 0.0001). Importantly, the parametric fit was a good predictor of the measured torque across all axial loading conditions (Figure 3D). We performed all data processing and analysis in MATLAB (Mathworks, Natick, MA, USA).

Figure 2. Quantifying Ankle Stiffness.

The raw data (A), including the controlled input, ankle angle, and the measured output, frontal-plane torque, were used to compute the impulse response function defining the ankle’s impedance. (B) This impedance was inverted to the admittance (solid line), allowing for it to be fit to a 2^nd order system (dashed line). Stiffness (C) was computed from this parametric fit. The green and red stiffness values in (C) correspond to their respective green and red parametric fits in (B). (D) Raw data from the peroneus brevis is shown, which is then rectified to obtain (E), after which, the mean is taken. (F) The mean for all trials of this representative subject is shown.

Figure 3. Evaluating the performance of the models used to quantify ankle stiffness.

(A) Measured rotational perturbations at the ankle for 5 seconds of an experimental trial. (B) Both the torque predicted from the impulse response function (blue) and the spring-mass-damper parametric fit (red) were good fits to the measured resultant torque (black). (C) This goodness-of-fit was supported by the high variance accounted for (%VAF) of the IRF in predicting the measured torque, as well as a high %VAF between the admittance IRF and the parametric fit IRF seen in Figure 2B. (D) Due to the high %VAF of both the impedance IRF and the parametric fit, the parametric fit was a good predictor of the measured torque across all axial loading conditions. The boxplots in (C) and (D) show the distribution of across all experimental trials (approximately 15 per participant) across all participants.

EMG Analysis

For each trial, the EMG signals were notch filtered to remove 60-Hz power line noise with a 2^nd-order Butterworth filter, with cut-off frequencies at 59 and 61 Hz. Secondly, we subtracted the mean from the signal. We then full-wave rectified the signals. Finally, we took the mean of the signal and normalized to the MVC value from the same muscle. The EMG voltage in each 7-second MVC trial was processed the same as all other trials. The data were notch filtered to remove 60-Hz noise, the mean removed, full-wave rectified, and then digitally zero-phase filtered using a 2^nd-order lowpass Butterworth filter with cut-off frequency of 1 Hz (Besomi et al., 2020). To obtain the MVC voltage, we took the maximum value of the resulting signal. This analysis resulted in estimates of muscle activity as a percent of MVC for each trial, allowing us to determine if there was an axial-load-dependent relationship with muscle activity.

Statistics

Our primary hypothesis was that ankle stiffness would increase with increasing axial load. We used a linear mixed-effects model to model ankle stiffness as the dependent variable, axial load as a continuous factor, sex as a fixed factor, and subject as a random factor. The same model structure was used to estimate the effect of axial load on the viscosity, inertia, and damping ratio. To ensure that muscle activity did not contribute to changes in stiffness with load, we computed separate linear mixed-effects models for each muscle, where EMG was the dependent variable, axial load was a continuous factor, and subject was a random factor. To determine the effect of perturbation amplitude on the axial-load stiffness slope, we used a linear mixed effects model with stiffness as the dependent variable, load as a continuous factor, perturbation amplitude as a continuous factor, and subject as a random factor. We estimated the parameters for these models using a restricted maximum likelihood method and Satterthwaite approximations for the degrees-of-freedom (Luke, 2017). All metrics are reported as mean estimate ± standard error unless otherwise noted. For all hypothesis tests, significance was set to 0.05. Quality of the models were assessed by computing the R², following removal of the subject-specific intercepts.

RESULTS

Frontal-plane ankle stiffness increased with axial load for all subjects. One representative subject is shown in Figure 4. There was a linear increase in stiffness normalized by body weight (BW) with axial load from the resting load (about 10% of body weight) to 50% of the subject’s body weight (R² = 0.98). Across all subjects (Figure 5), stiffness increased linearly (R² = 0.96) at a rate of 8.4±0.5 x10⁻⁴ Nm/rad/N/%BW, (t₁₃ = 15.2, p < 0.0001). As a result, at 50% BW, the ankle is approximately 3 times stiffer compared to 0% BW.

Figure 4. Ankle stiffness increased with axial load in a representative subject.

This subject demonstrated a linear increase in stiffness across the entire range of applied load that was tested (R² = 0.98). Each point shows the stiffness estimated in a trial, with the line representing a least-squares fit to the data.

Figure 5. Ankle stiffness increased with axial load across all subjects.

The stiffness increased linearly with axial load across all subjects (R² = 0.96). The solid line shows the group average from the linear mixed-effects model, with the shaded area showing the 95% confidence intervals of the fit.

The effect of axial load on ankle stiffness was not due to an axial-load-dependent change in muscle activity. Figure 6 shows the level of activation for the 6 muscles recorded during the experiment for one representative subject. There was no muscle that surpassed an activation level of 5% MVC. More importantly, activation did not change with the axial load applied to the ankle for any muscle. Across all subjects, there was little effect of axial load on muscle activity, with the EMG-axial-load slope ranging from −0.018 to 0.034 %MVC/%BW, and none reaching significance (all p > 0.10, Table 1).

Figure 6. Muscle activation did not increase with load.

The mean EMG is shown from the same representative subject as in Figure 4. The lines represent a least-squares fit to the data. The grey areas represent the 95% confidence interval. All muscles remained below 5% MVC throughout the range of applied load. Muscle activation did not increase with load in any muscle. These results were consistent across the entire group, as demonstrated by the results of the linear mixed-effects model shown in Table 1.

Table 1:

Load Parameters for the Linear Mixed Effects Model for Each EMG

Muscle	EMG - Load Slope [%MVC/%BW]	EMG - Load Slope Standard Error [%MVC/%BW]	t-statistic	Degrees of Freedom	p-value
Tibialis Anterior	0.0033	0.0032	1.03	35	0.31
Lateral Gastrocnemius	0.034	0.033	1.02	37	0.32
Medial Gastrocnemius	0.005	1.41	0.76	14	0.46
Soleus	−0.018	0.030	−0.61	56	0.54
Peroneus Longus	0.0073	0.0043	1.71	13	0.11
Peroneus Brevis	0.014	0.010	1.38	14	0.19

The effect of load on stiffness did not differ between men and women. Though stiffness was 37% reduced in females (−0.0067 +/− 0.0032 Nm/rad/N, t₁₃ = −2.07, p = 0.058) at 0% BW, there was no difference in the rate of increase of stiffness with axial load between men and women (−4.2 ±10.6 x10⁻⁴ Nm/rad/N/%BW, t₁₃ = −0.40, p = 0.70).

While there was an increase in viscosity with axial loading, the ankle remained substantially underdamped throughout the range of axial loads. There was a linear increase in viscosity with axial loading (R² = 0.88) at a rate of 4.5±0.8 x10⁻⁶ Nm/rad/s/N/%BW (t₁₃ = 5.39, p = 0.0001, Figure 7A). Inertia remained unchanged with axial loading 1.1±0.7 x10⁻⁸ kgm²/N/%BW (t₁₃ = 1.34, p = 0.20, Figure 7B). Combining the estimates of stiffness, viscosity, and inertia, we found that the ankle remained substantially underdamped, ranging from 0.05 at 0% BW to 0.21 at 50% BW (Figure 7C).

Figure 7. Viscosity increased with axial load, but system remained substantially underdamped at all load levels.

(A) The viscosity increased linearly with load across the group of subjects (R² = 0.88). (B) As expected, the inertia remained unchanged with increasing axial load (R² = 0.32). (C) The damping remained low at all levels of load, demonstrating the minimal effect viscosity has on the ankle impedance under passive-axial-loaded conditions. The solid line in each plot shows the group average fit from the linear mixed-effects model, with the shaded area showing the 95% confidence intervals of the fit.

We observed a significant effect of perturbation amplitude on the axial-load-dependence of frontal-plane ankle stiffness (Figure 8). While stiffness increased with axial load for all the perturbation amplitudes tested, the stiffness-load slope significantly decreased with an increase in perturbation amplitude (t₁₀₉ = −2.58, p = 0.011, Figure 8B).

Figure 8.

(A) Normalized stiffness vs. axial load for the tested perturbation amplitudes (N = 8) For clarity, the predictions for 0.03 rad and 0.05 radians have been omitted. (B) The effect of perturbation amplitude on the slope between normalized stiffness and axial load. Shaded areas for both plots are 95% confidence intervals.

DISCUSSION

The objective of this study was to determine the effect of passive axial loading on frontal-plane ankle stiffness. We accomplished this by estimating ankle stiffness in a seated configuration as axial load was applied via pressure on the knee. We tested the hypothesis that ankle stiffness would increase linearly with axial load. Our hypothesis was supported as we demonstrated a significant effect of axial load on ankle stiffness. We did not see any axial-load-dependent relationship with muscle activity, indicating that the changes in ankle stiffness were due solely to the changes in axial load. These results demonstrate that axial loading is a significant contributor to maintaining frontal-plane ankle stability. Disruptions to this mechanism may result in ankle instability.

Frontal-plane ankle stiffness increased with axial load while seated. While no studies have quantified frontal-plane ankle stiffness with applied axial loading while seated, studies have looked at frontal-plane ankle stiffness during weight bearing while standing. Our measures of ankle stiffness at 50% BW (0.055±0.004 Nm/rad/N) are within the range of other studies that quantified frontal-plane ankle stiffness at 50% BW [.044–0.112 Nm/rad/N] (A. Ribeiro et al., 2018; Matos et al., 2021; Nalam et al., 2021). Our results also agree with studies showing that frontal-plane ankle stiffness increased with increasing weight bearing (Matos et al., 2021; Nalam et al., 2021), though none of these studies reliably isolated the effect of axial loading from that of muscle activation. Our estimate of the increase in stiffness with axial load is as large as those during standing, suggesting that much of the weight-bearing stiffness is due to axial loading of the ankle. Interestingly, our estimates were higher than others where muscle activation was present. One possible explanation is that we had the leg constrained, by the ankle connected to the motor and the knee by the loading device. During standing, the ankle was perturbed while the leg was free to move (Matos et al., 2021). Any body sway in response to the perturbation would influence ankle stiffness estimates. This is similar to how elbow stiffness decreases when the trunk and shoulder are free to move compared to when they are constrained (Perreault et al., 2000). Alternatively, postural sway has been shown to induce large changes in sagittal plane ankle stiffness (Amiri and Kearney, 2019) and sway in the frontal plane can be nearly as high (Yamamoto et al., 2015). It is therefore possible that the higher stiffness observed in our study compared to standing studies is a result of previous studies not controlling for the effect of sway in addition to the effect of load.

Most mechanisms that contribute to ankle stiffness would not be sensitive to passive compressive axial loading. Ligaments, muscles and tendons all contribute to ankle stiffness (Leardini et al., 2000; Sakanaka et al., 2018), but the stiffness of each of these only increases with tensile load rather than compressive load (Fung, 1993; Martin et al., 1998). Cadaveric studies suggest that the articular surfaces may be the mechanism responsible for the increase in ankle stiffness with passive axial loading (Stiehl et al., 1993; Stormont et al., 1985; Watanabe et al., 2012). Under axially-loaded-conditions, the ligaments provide negligible resistance to imposed rotations in the frontal plane, implying that nearly 100% of the resistance torque to the imposed rotations is due to articular surfaces (Stormont et al., 1985). Furthermore, this resistance torque, a similar measure to stiffness, increases with axial loading (Stiehl et al., 1993; Stormont et al., 1985). This increase in resistance torque may be due to articular surfaces having greater articular congruity (Stiehl et al., 1993) or increased contact stress (Tochigi et al., 2006) in the load-bearing state. Articular surface loading, either through weight-bearing or muscle activation would increase stiffness independently from the activation-dependent increases in muscle and tendon stiffness (Cui et al., 2008; Jakubowski et al., 2022). Contributions from articular surface compression and increases in muscle and tendon stiffness would add, as these elements are mechanically connected in parallel. Our results suggest that the stiffness contributions from the articular surface provide an important mechanism for increasing stiffness since it is not limited by tendon stiffness, which restricts the ability of the muscles to increase stiffness in the sagittal plane (Jakubowski et al., 2022), and presumably also the frontal plane. While our study does not directly isolate a mechanism responsible for the axial load effect, our results are in line with cadaveric studies pointing to greater contribution of articular surfaces. While our results suggest that the articular surface would bear the greatest amounts of stress during excessive inversion, it does not preclude other tissues from being injured prior to the articular surface as tendons, muscles and ligaments all experience injury at lower stresses than the articular surface (Kuthe et al., 2014; Maganaris and Paul, 1999; Verteramo and Seedhom, 2004; Zens et al., 2015).

We found an increase in viscosity with loading. Our estimate of viscosity (0.5x10⁻⁴ (Nm s/rad)/(N) at resting load), was lower than another study that computed frontal-plane impedance (3x10⁻⁴–8x10⁻⁴ (Nm s/rad)/(N) (Lee et al., 2014a), however that study had muscle activations up to 35% MVC. This lower damping during passive conditions is consistent with the positive relationship between damping and activation demonstrated in the sagittal plane (Mirbagheri et al., 2000). Most importantly, this axial-loading sensitive viscosity likely plays little role in maintaining ankle stability considering that throughout the range of loading, the ankle was substantially underdamped (no more than 0.21).

We observed a perturbation-amplitude-dependent effect on the sensitivity of frontal-plane ankle stiffness to axial loading. A perturbation-amplitude-dependent effect has been observed with sagittal-plane ankle stiffness (Loram et al., 2007), which has been attributed to muscle’s short-range stiffness (Kearney and Hunter, 1990; Loram et al., 2009; Rack and Westbury, 1974). This study was focused on contributions from other tissues, as muscles were passive. We observed a modest decrease in the stiffness-axial-load slope with perturbation amplitude, approximately 35% over a range of 0.07 radians in amplitude. This is similar to the perturbation amplitude dependence on sagittal plane ankle impedance with muscle activation (Kearney and Hunter, 1982). Therefore, the perturbation sensitivity in axial-loading increases in ankle stiffness is like the perturbation sensitivity seen during active conditions.

Frontal-plane ankle stiffness likely plays a larger role in preventing ankle sprains than in maintaining upright stance. Our stiffness estimates are much less than would be required to maintain single-legged stance (~500 Nm/rad) (Loram et al., 2007). However, single-legged quiet stance is not a normal mode of operation. In contrast, passive ankle stiffness may help prevent excessive ankle rotations that would lead to injuries. Ankle sprains tend to occur when the ankle is inverted past 35 degrees (Markolf et al., 1989; Wright et al., 2000). We estimated the average subject’s stiffness to be 72 Nm/rad at 100% body weight. The ankle would therefore need to be exposed to greater than 44 Nm of inversion torque before exceeding 35 degrees inversion. Ankle sprains also tend to occur during active muscle conditions which will increase stiffness, and thus the ankle will be able to withstand torques greater than our estimate of 44 Nm. Furthermore, as the ankle is inverted the stiffness likely increases (Birmingham et al., 1997). Both factors suggest that our estimate of torque required to sprain an ankle is an underestimate.

One limitation of this study is that activation of the tibialis posterior, a major invertor of the ankle, was not measured because it is a deep muscle that could not be easily recorded with surface EMG. However, we did record activation from the tibialis anterior, another inverter of the ankle and one that exhibits similar activation levels during isometric ankle supination (Hagen et al., 2016). During walking, the tibialis posterior behaves similarly to the peroneus longus and the medial gastrocnemius (Murley et al., 2014). Based on the similarities, it is unlikely that the tibialis posterior experienced a load-dependent increase in activation while the other muscles did not. Another limitation is the alignment of the motor’s axis of rotation to the ankle’s axis of frontal plane motion. Any misalignment could cause the perturbations to move the tibia in the vertical direction, which could create torques that lead to errors in our impedance estimates. First, we determined the relationship between the mean axial load and the mean frontal plane torque on a subject-by-subject basis, which represented an effective moment arm between ankle rotation and the point of load application. Next, we modeled the relationship between the perturbations and the axial load, as this relationship would be strong if there were to be misalignment. We scaled this relationship by the effective moment arm to represent the stiffness due to misalignment. The perturbations induced only small amounts of torque through this mechanism. Across all trials, this stiffness due to varying load was less than 3% of the measured ankle stiffness (95% CI = −2.7% to 2.1%). However, in certain trials, this stiffness increased the measured ankle stiffness and in other cases it decreased it. Thus, we saw no consistent effect of misalignment on our measured ankle stiffness (5.88 x 10⁻³ Nm/rad).

In conclusion, we found that axial loading increases frontal-plane ankle stiffness by about 3-fold from 0% BW to 50% BW. Importantly, this is independent of muscle activation, indicating this is a passive mechanism of stabilizing the ankle. These results suggest that passive axial loading of the ankle joint is an important contributor to maintaining ankle stability. Deficits in this mechanism could lead to ankle instability.