Biomechanical Response to External Biofeedback During Functional Tasks in Individuals With Chronic Ankle Instability

We performed a crossover study to compare the real-time effects of visual and auditory external biofeedback on biomechanics during functional tasks in a cohort of physically active adults with CAI. Our independent variable was condition (baseline, visual, auditory), with baseline serving as the comparison condition. Our primary dependent variables were measures of postural control (COP location) during eyes-open and eyes-closed static balance and measures of plantar pressure (peak pressure and pressure-time integral) within the lateral foot column during step downs, lateral hops, and forward lunges. To capture a complete biomechanical profile during each task, we included the secondary variables of postural control and plantar pressure. An a priori power analysis using pilot data from our laboratory was calculated to determine sample size. We determined that a sample size of 16 was needed to obtain an α of .05, power of 0.95, and effect size of 1.

A total of 19 physically active adults with CAI (7 men, 12 women; age = 23.95 ± 5.52 years, height = 168.87 ± 6.94 cm, mass = 74.74 ± 15.41 kg) volunteered. Participants met the standards for CAI as determined by the International Ankle Consortium.²³  Briefly, they described having at least 1 significant ankle sprain that occurred at least 12 months before enrollment and the most recent sprain >3 months before enrollment. Participants self-reported foot and ankle dysfunction by scoring ≤85% on the Foot and Ankle Ability Measure–Sport subscale. Furthermore, they scored ≥11 on the Identification Functional Ankle Instability questionnaire, indicating the presence of ankle instability. Physical activity levels were determined using the International Physical Activity Questionnaire short form; participants stated that they engaged in ≥30 minutes of physical activity 3 times per week. Volunteers were excluded if they did not meet the aforementioned criteria or reported a previous ankle fracture or surgery, any underlying condition that would influence plantar pressure, or the inability to perform the tasks. All participants provided written informed consent, and the study was approved by our university's institutional review board.

Single-limb static balance was performed using an AccuSway Optimized force platform (Advanced Medical Technology, Inc) sampling at a rate of 50 Hz, and the data were processed using Balance Clinic (Advanced Medical Technology, Inc) software. Plantar pressure was collected via the pedar-x (Novel Electronics Inc, Saint Paul, MN) system sampling at a rate of 200 Hz. Calibration was conducted to ensure that plantar pressure was recorded only when the foot was in contact with the ground and excluded aerial phases.

Visual biofeedback was given via a class IIIA crossline laser diode (Calpac Lasers, Steamboat Springs, CO) powered by 2 AA batteries that has previously been used and described in detail.^20,24  Auditory biofeedback¹⁹  was supplied via a thin (14- × 25.4- × 0.203-mm) FlexiForce Load Sensor (Tekscan, Inc, South Boston, MA). The pressure sensor was connected to a FlexiForce Quickstart Board (31.75 × 31.75 mm) and a potentiometer (Tekscan, Inc) with an attached buzzer, powered by a 9-V battery. The setup of each biofeedback device is presented in Figure 1.

Under each of the 3 conditions, participants performed 4 tasks: single-limb balance, step down, lateral hop, and forward lunge. Balance trials were always conducted first because they required participants to be barefoot; however, the remaining 3 tasks were randomized for each participant by using a Latin square. After the balance trials were completed, participants were fitted with standard, neutral athletic shoes (model M680V3; New Balance Inc, Brighton, MA) with the plantar-pressure insoles placed inside. Practice trials were allowed for each task under each condition, and each task was completed under each condition before another task was started. The baseline condition for all tasks was always performed first using standard instructions. The visual and auditory biofeedback conditions were performed randomly for each task using a Latin square. All data were captured for the limb with CAI. If a participant reported a bilateral history of ankle sprains, the self-perceived worse limb was tested.

Static balance was performed while participants stood barefoot on a force plate with the uninvolved limb placed in 30° of hip flexion and 45° of knee flexion and their hands on their hips.²⁵  They were instructed to “stand as still as possible while maintaining the test position” and allowed 3 practice trials. During the baseline condition, no other instructions were provided. Participants performed 3 practice trials followed by 3 successful 10-second trials recorded with their eyes open and eyes closed. Failed trials, in which the participant moved out of the test position, were repeated. A maximum of 10 total attempts were allowed for each condition.

To perform the step-down trial, participants started from a 30-cm-tall box and were instructed to step down onto the ground with the involved limb first and continue their momentum forward for an additional few steps.⁷  Three practice trials were allowed before 10 successful step-down trials were completed and used for analysis.

Participants performed lateral hops²⁵  over a piece of athletic tape that was placed along the floor and continued up the wall in front of them. They were instructed to hop laterally over the tape while focusing forward on the wall and to use the tape as a guide for hopping over the tape on the ground. All participants started with the tape toward the outside of their involved limb so that every beginning jump was lateral. A successful trial consisted of a lateral hop, balance maintained on landing, and a hop back to the starting position without removing their hands from their hips or taking extra steps. Based on our pilot testing, we were unable to find a consistent hop rate that could be performed successfully during all 3 testing conditions; therefore, we chose to standardize the hopping distance rather than the hopping speed using a metronome. Participants completed 3 practice trials before 10 continuous successful trials were recorded for analysis.

Forward lunges were performed from a neutral stance with the hands on the hips.²⁵  Participants lunged with the involved limb forward into a 90°/90° position of the hip and knee and touched the back (uninvolved) knee to the ground before returning to the starting position. Three practice trials were allowed, and then a total of 10 forward lunges were performed and used for analysis.

Visual biofeedback was provided by the crossline laser device that was secured to the dorsum of the foot using a strap (Figures 1A and B).^20,24  The laser remained fastened to the foot during every trial of every task to eliminate differences in COP or plantar-pressure data distribution during all trials but was turned off during the nonvisual-feedback conditions. During the visual-biofeedback conditions, the crossline laser was turned on and was visible to participants on a wall directly in front of them. Before each task was conducted, the laser was adjusted to a neutral stance and starting position. A piece of white athletic tape served as a reference for the starting point of each task. Specific instructions were given before each visual-biofeedback condition of each task, with the general instruction to “perform the task as naturally as possible while keeping the vertical line of the laser in line with the tape and limit the amount of rotation of the crossline.” Participants performed 3 practice trials using visual biofeedback before we collected 10 trials for analysis.

Auditory external focus-of-attention biofeedback was provided by the auditory device and calibrated for each participant before each task (Figures 1B and C). During single-limb balance, the sensor was taped to the force plate under the head of the fifth metatarsal, which ensured consistent placement of the foot on the force plate. Laboratory shoes were cut to allow the sensor to be taped to the insole of the shoe under the fifth metatarsal while maintaining the integrity of the shoe.¹⁹  Participants were instructed to shift all of their weight onto the sensor, leaning in an anterolateral direction. The potentiometer was then adjusted to the first point when noise was heard. During nonauditory-biofeedback conditions, the auditory instrument remained in place under the fifth metatarsal but was turned off by disconnecting the battery. Specific instructions were given before each auditory-biofeedback condition of each task with the general instruction to “perform the task as naturally as possible without making the buzzer elicit a noise.” Participants performed 3 practice trials using auditory biofeedback before we collected 10 trials for analysis.

During each static-balance trial, a time series of 500 COP data points (10 seconds × 50 Hz) was generated, and a custom MATLAB code (version R2019a; The MathWorks, Inc, Natick, MA) was used to determine the location of each data point in 4 quadrants of the foot (anteromedial, anterolateral, posteromedial, and posterolateral).¹⁷  More data points equated to more loading in the quadrant.

The peak pressure (kPa) and pressure-time integral (kPa·s) were calculated from the 10 steps performed during each task using Database Pro software (Novel Electronics Inc). A standard mask was applied to divide the foot into 9 regions; our primary regions of interest were the lateral heel, lateral midfoot, and lateral forefoot.

The COP 95% confidence eclipse area (cm²) and mean velocity (cm/s) were calculated using Balance Clinic software with a fourth order, zero-lag, low-pass filter with a cutoff frequency of 5 Hz. Time-to-boundary variables (absolute minima and standard deviation of the minima) were calculated in the anteroposterior and mediolateral directions using a custom MATLAB code. Smaller area, velocity, and time-to-boundary values indicated worse postural control.

Additional plantar-pressure measures were extracted and consisted of contact area (cm²), contact time (milliseconds), maximum force (N), and force-time integral (N·s). The remaining regions created from the applied mask were the medial heel, the medial midfoot, medial forefoot, central forefoot, great toe, and lesser toes. We also included a total foot region.

Separate within-factor repeated-measures analyses of variance were used (SPSS version 26; IBM Corp) to compare the means for each dependent variable across the 3 conditions (baseline, visual, auditory). We report only the comparisons of baseline with each biofeedback condition and do not report comparisons between the biofeedback conditions. We set the α level a priori at .05. In accordance with modern statistical recommendations,²⁶  we did not control for multiple comparisons. Rather, we calculated Hedges g effect sizes (ESs) and associated 95% CIs and interpreted the results as different if P ≤ .05 and the ESs were moderate to large with 95% CIs that did not cross 0. Effect sizes were considered large (≥0.80), moderate (0.50–0.79), or small (0.20–0.49) and were calculated using Excel (version 2016; Microsoft Corp).

All participant characteristics are provided in Table 1. Results for the primary outcomes are presented in Figure 2 and Tables 2 through 5. Results for all secondary outcomes are available in Supplementary Tables 1 through 4.

Results for the primary outcomes during eyes-open and eyes-closed balance are shown in Figure 2 and Tables 2 and 3. Results for the secondary outcomes appear in Supplementary Table 1.

The visual- and auditory-biofeedback conditions during eyes-open static balance reduced the number of COP data points in the anterolateral foot quadrant (P = .002; ESs = 0.80 and 0.86, respectively) while simultaneously increasing COP data points in the posteromedial quadrant compared with the baseline condition (P = .01; ESs = –0.74 and –0.89, respectively). Furthermore, the auditory-biofeedback condition reduced the number of COP data points in the posterolateral foot quadrant compared with baseline (P = .003; ES = 0.72; Figure 2 and Table 2).

During the eyes-closed trials, we observed a decrease in COP data points in the anterolateral foot quadrant (P < .001, ES = 0.95) and an increase in data points in the posteromedial foot quadrant (P = .006, ES = –0.97) during the auditory-biofeedback condition compared with baseline (Figure 2 and Table 3).

Results for the primary variables during the step-down task are given in Tables 4 and 5 and for the secondary variables in Supplementary Tables 2 through 4. Compared with baseline, the auditory-biofeedback condition increased the lateral heel peak pressure (P = .03, ES = –0.68; Table 4) and pressure-time integral (P = .003, ES = –0.75; Table 5). The auditory-biofeedback condition reduced the pressure-time integral of the lateral forefoot compared with baseline (P = .001, ES = 0.70; Table 5).

Results for the primary variables during the lateral hop are provided in Tables 4 and 5 and for the secondary variables in Supplementary Tables 2 through 4. Visual biofeedback increased the peak pressure (P = .002, ES = –0.73) and pressure-time integral (P = .001, ES = –1.01) in the lateral heel region as well as the lateral midfoot region (P = .001, ES = –0.78; Table 4). Auditory biofeedback did not change plantar pressure from baseline.

Results for the primary variables during the forward lunge are supplied in Tables 4 and 5 and for the secondary variables in Supplementary Tables 2 through 4. The auditory-biofeedback condition decreased the pressure-time integral (P < .001, ES = 0.78) in the lateral forefoot (Table 5). Visual biofeedback did not change plantar pressure during the forward lunge.

The purpose of our study was to determine the real-time effects of 2 novel external focus-of-attention biofeedback devices on static balance and functional-task biomechanics in a cohort of individuals with CAI. Our results partially supported our central hypothesis that both visual and auditory biofeedback would produce changes in static balance and biomechanics during functional tasks. Both modes of external biofeedback contributed to changes in static balance but individually targeted functional activities. Our study provides initial evidence for the utility of various external-biofeedback media in targeting multiple rehabilitation exercises for maximizing motor control and learning.

During eyes-open and eyes-closed static balance, the auditory-biofeedback condition produced a beneficial shift in COP location from the anterolateral to the posteromedial foot quadrant. The visual-biofeedback condition produced similar changes in COP location in the eyes-open trials. According to previous researchers,⁶  healthy individuals had more COP data points in the posteromedial foot quadrant, whereas individuals with CAI had more in the anterolateral foot quadrant. Despite a real-time advantageous shift in COP location, our secondary time-to-boundary outcomes (Supplementary Table 1) indicated an initial worsening of postural control during these trials. The goal of developing balance-training programs through the perspective of dynamic systems theory²⁷  is to manipulate the task in a way that allows patients to explore new avenues for handling a changing environment. We speculate that having less stability is natural, as these individuals are discovering a new COP location, and evidence²⁸  has suggested that postural control continues to improve over time when balance training is coupled with an external focus of attention. Furthermore, individuals with CAI have a heightened reliance on visual information, and traditional balance-training programs are unable to alter that visual reliance.²⁹  In our investigation, the auditory-biofeedback condition caused parallel changes during eyes-open and eyes-closed balance, indicating the potential for improved balance without relying on visual stimulus. As patients continue to use external biofeedback to maintain a posteromedial COP, perhaps their overall stability will also improve, but researchers will need to determine the long-term effects of our external-biofeedback devices on postural control.

The efficacy of our visual- and auditory-biofeedback devices differed among our chosen functional tasks. When performing the step down and forward lunge, participants were more responsive to auditory biofeedback, whereas during the lateral hop, they were more responsive to visual biofeedback. During the step down, auditory biofeedback caused participants to adopt a more dorsiflexed, closed-packed position during initial contact and throughout the loading phase. This strategy may benefit patients with CAI, as landing in a more plantar-flexed and inverted ankle position is thought to cause episodes of giving way or recurrent ankle sprain.^9,30  Auditory biofeedback during the forward lunge caused participants to reduce primary and secondary plantar-pressure measures in the lateral midfoot, lateral forefoot, and lesser toes. We observed no increases in plantar pressure in the medial foot column, which indicated that they did not adopt an overly unnatural tactic when performing these tasks. Although the COP trajectory was not measured in this study, previous authors²⁰  who analyzed walking-gait retraining noted that reductions in lateral plantar pressure were accompanied by a medial shift in COP trajectory, which is a beneficial strategy for those with CAI. The visual biofeedback during the lateral hop produced a more closed-packed landing strategy by increasing plantar pressure in the heel and midfoot regions. As with the step down, this appears to be a beneficial strategy for individuals to adopt in order to reduce the inherent risk of sustaining another inversion ankle sprain.

Our results illustrated that auditory biofeedback was beneficial during tasks performed in the sagittal plane, whereas visual biofeedback was advantageous during tasks performed in the frontal plane. Evidence of motor learning supports the use of visual biofeedback during complex tasks (eg, lateral hop) compared with auditory biofeedback during less complex tasks (eg, step down and lunge).²¹  Although our auditory device was designed to provide biofeedback only when the lateral forefoot applied excessive pressure to the sensor and no biofeedback was given during the aerial phases, our participants were able to adjust their feet on initial contact to adhere to the cues given. However, during the lateral-hop tasks, placement of the auditory sensor may have prevented them from altering their landing strategy so as to follow the cue and still perform the task correctly. In contrast, the constant visualization of the crossline laser during the lateral-hop task may have allowed participants to determine a proper biomechanical strategy for adjusting their performance to the biofeedback. Our visual-biofeedback instructions during all tasks were to keep the vertical line of the laser parallel to a piece of tape (ie, transverse-plane motion) and reduce the amount of rotation of the crossline (ie, frontal-plane motion).

Collectively, we chose our tasks to mimic common exercises that are used during ankle rehabilitation and have been the primary focus of earlier authors.^14,15,25  Both devices have been shown to be beneficial in targeting aberrant walking-gait biomechanics,^19,20  whereas our findings did not support one medium over the other. To optimize motor learning, Guadagnoli and Lee³¹  proposed that, when using protocols incorporating feedback, clinicians should be flexible about and cognizant of the demands of the task being performed. Although this framework was primarily built around healthy individuals learning new motor tasks, our results add to this existing framework by including an injured population relearning a skill with ideal biomechanics. More utility may be possible in using these novice external-biofeedback devices congruently with an impairment-based rehabilitation model to improve biomechanics across various tasks.

Our study had limitations. First, our intent was to determine a real-time, single-dose effect; therefore, we cannot draw conclusions about the long-term ability of these visual- and auditory-biofeedback devices to improve biomechanics. Second, our work lacked a neuromuscular perspective, and future researchers should characterize the full range of biomechanical changes that occur while using these novice biofeedback devices. Our results warrant more exploration involving these external-biofeedback devices in a full impairment-based rehabilitation program to determine their overall benefit to patients with CAI.

The visual- and auditory-biofeedback devices improved static balance and functional-task biomechanics differently depending on the exercise. Our findings provide an initial extrapolation of the use of external focus-of-attention biofeedback during rehabilitation after an ankle sprain. Clinicians should consider using low-cost, user-friendly external focus-of-attention biofeedback devices to improve biomechanics and balance during established rehabilitation protocols.