Cueing changes in peak vertical ground reaction force to improve coordination dynamics in walking

Abstract

Biofeedback has been effectively implemented to improve the mediation and distribution of joint loads during gait, however, the inability to effectively coordinate lower limb movement by altering loading patterns may increase pathological stress and risk of injury and deleterious joint changes. This study examined the influence cueing an increase or decrease in lower extremity loading has on inter- and intralimb joint coordination during gait, applied herein for 12 persons following anterior cruciate ligament reconstruction across three loading conditions (control, high, and low). Visual biofeedback was presented on a screen via a force-measuring treadmill with targeted changes prescribed based on stride-to-stride peak vertical ground reaction forces bilaterally. The pattern and stability of coordination dynamics among each of the ankle, hip and knee joint pairs were assessed via discrete relative phase and cross-recurrence quantification analyses for each condition. High and low loading altered the pattern and stability of intralimb coordination; low loading led to decreased coordination stability (20° greater than control condition) and high loading resulted in a more tightly coupled coordination pattern (higher %CDET). With thoughtful consideration for movement control, biofeedback can be used to target mechanisms leading to long-term deleterious joint adaptations.

Introduction

The repeated coordination of multiple joints is fundamental to walking gait (Heiderscheit, 2000) and reflects a system that is adaptable to environmental stimuli and stressors (i.e., a system that can execute the most effective responses and strategies to navigate daily life; Grabiner & Enoka, 1995). To do so involves the complex integration and interaction between different neural and motor aspects for coordinated movement. Therefore, human locomotion blends the stability of rhythmical multi-joint coordinated movements with the flexibility to vary the coordination of these multi-joint movements to accommodate to an ever-changing environment (Riley & Turvey, 2002; Van Emmerik, Hamill, & McDermott, 2005). A reduced capacity to adapt to stressors (e.g., walking on an irregular surface, an unanticipated change in direction) has been hypothesized by Lipsiz and Goldberg (1992) to result from a loss of system complexity (i.e., altered functional capacity of control) due to either a reduction in the integrity of neuromuscular components or an alteration in the coupling function among system components (e.g., the coordination between oscillating limbs). For example, as we age, neurophysiological changes (e.g., slowed neuromotor processing, transmission speed, and neurotransmitter potential) reduce the capacity of neuromuscular control (Lu et al., 2011). As a result, older adults demonstrate a less adaptive gait pattern (i.e., reduced gait speed, more rigid trunk posture, and a stronger coupling of limbs) with a reduced capacity to adapt to external stressors (Kavanagh, Barrett, & Morrison, 2005; Woollacott & Pei-Fang, 1997). Similar changes in gait patterns have also been observed across clinical populations, including individuals with diabetic neuropathy (Wang & Newell, 2012) and those who sustain a musculoskeletal injury (Armitano, Morrison, & Russell, 2018). These changes can impact the complex blend of stability and flexibility that comprise adaptive coordination dynamics, and can reduce the ability of the movement system to appropriately adapt to environmental constraints (e.g., stressors such as an irregular surface). As we described in more detail below, this in turn places the individual at an increased risk of injury (Kavanagh et al., 2005) and may lead to deleterious consequences for overall joint health (Van Emmerik et al., 2005).

Reduced intra-joint coupling within the lower extremity has been associated with increased stress at the joint and surrounding tissues (Van Emmerik et al., 2005) and has been associated with cumulative micro-trauma within the joint (Hamill, van Emmerik, Heiderscheit, & Li, 1999). Specifically, individuals with a reduced coupling lower limb coordination pattern have exhibited altered lower extremity joint loading patterns; e.g., decreased coordination variability and larger impact forces (Noehren, Wilson, Miller, & Lattermann, 2013; Zeni, Rudolph, & Higginson, 2010). Such changes in loading during gait have been associated with changes in joint tissue metabolism (Pietrosimone et al., 2016) and knee cartilage composition (Pfeiffer et al., 2019): both predispose the joint to the onset of osteoarthritis (OA). Therefore, the intricate relationship between the inability to effectively coordinate lower limb movement and changes in loading patterns, may increase pathological stress and risk of injury and/or OA onset in the joint. Thus, a focus on enhancing movement coordination strategies may be beneficial for promoting joint health.

Biofeedback is an approach that has been regularly utilized to promote corrective changes to movement patterns. Through the implementation of biofeedback during gait, Hausdorff and colleagues (1996) demonstrated that the intrinsic temporal patterns had the capacity to be augmented. That is, the underlying control processes governing gait have the capacity to be manipulated in a prescribed manner through strategic biofeedback. These findings have since been expounded upon as a means of understanding the underlying temporal structure of stride intervals (Hausdorff, Zemany, Peng, & Goldberger, 1999; Kaipust, McGrath, Mukherjee, & Stergiou, 2013) as well as to understand the flexibility of the underlying control processes of healthy movement control (Rhea, Kiefer, D’Andrea, Warren, & Aaron, 2014a; Rhea et al., 2014b). Similarly, this approach has also been applied in rehabilitative and clinical settings to target deviations in movement patterns; patterns that are known to arise from coordination dynamics among various components within the system (Diniz et al., 2011); that may place an individual at risk of injury through applying a mechanistic focus to the biofeedback implemented. For example, individuals with a lower extremity injury (e.g., patellofemoral pain, anterior cruciate ligament injury) have a tendency to alter the loading rate of the affected joint, which has been suggested to increase the individual’s risk of later developing OA (Noehren et al., 2013; Pietrosimone et al., 2019; Utting, Davies, & Newman, 2005). Vertical ground reaction force (vGRF) is a fundamental measure of lower-extremity loading during gait and is associated with changes in joint tissue metabolism (Pietrosimone et al., 2016), knee cartilage composition (Pfeiffer et al., 2019), and patient-reported outcomes (Pietrosimone et al., 2018); predisposing such individuals to OA development. Recent research has employed visual biofeedback using peak vGRF to cue a change in loading mechanics to optimize gait in individuals with an anterior cruciate ligament reconstruction (ACLR, Luc-Harkey et al., 2018). Luc-Harkey and colleagues (2018) demonstrated the successful implementation of visual loading biofeedback that corresponded to a 5% increase or decrease in vGRF (i.e., high and low loading, respectively) to influence gait kinetics and kinematics in ACLR individuals. The findings demonstrate several joint-specific implications of cueing changes in loading that may successfully improve walking mechanics in an effort to reduce the risk of OA development. However, despite the known fundamental relationship between coordination and joint loading, it remains unknown how facilitating a change in knee joint load would impact lower limb coordination during gait. This is important to understand given that when an effector asymmetry has been applied during gait, research has shown deviations in limb coordination, and reduced stability (Russell, Kalbach, Massimini, & Martinez-Garza, 2010). A missing component of both motor control and biomechanics research is the understanding of how coordination is influenced by the prescription of visual biofeedback to modulate gait dynamics. Given the disruption in loading rate following ACLR, individuals with an ACLR present as the ideal population to assess a prescribed visual biofeedback approach to ultimately improve gait stability and coordination. This is important because cueing changes without understanding the implications for coordination could lead to the implementation of biofeedback interventions to target mechanisms that induce a positive behavioral change, but inadvertently reduce adaptability. To understand the functional implications of loading on movement strategies, a necessary step is to determine how the control and coordination across multiple joints is affected by cueing changes in loading during gait.

Nonlinear measures, such as relative phase (ϕ) and cross recurrence quantification analyses (CRQA), have been used to quantify patterns and variability of coordination (Abernethy, Burgess-Limerick, Engstrom, Hanna, & Neal, 1995; Schöner, Jiang, & Kelso, 1990; Shockley, Butwill, Zbilut, & Webber Jr, 2002; Webber Jr & Zbilut, 2005), enhancing our understanding of movement strategies in clinical populations. ϕ analyses have been used extensively to quantify patterns of coordination between limbs or joints during gait (Armitano et al., 2018; Drewes et al., 2009; Kiefer et al., 2013; Kurz, Stergiou, Buzzi, & Georgoulis, 2005; Miller, Meardon, Derrick, & Gillette, 2008). The measure of mean ϕ (Mϕ) describes the oscillatory phase relation between two joints with in-phase (Mϕ=0°) typically observed in lower leg intralimb coordination and anti-phase (Mϕ=180°) observed in lower leg interlimb coordination during gait (Abernethy et al., 1995; Schöner et al., 1990). In addition, the standard deviation of ϕ (SDϕ) describes the coordination stability between two joints. Deviations from in-phase or anti-phase coordination reveals changes in the pattern of coordination between limbs or joints such that greater asymmetry leads to reduced coordination stability while greater symmetry suggest a more rigid coordination pattern. While ϕ and SDϕ provide information about changes in coordination patterns and coordination stability, respectively, they cannot easily discern the underlying mechanisms driving such changes. CRQA is a nonlinear analysis tool used to further distinguish such mechanisms and, specifically, whether such changes are due to noise or coupling strength. CRQA assesses the dynamic nature of two time series (e.g., the ankle and knee sagittal plane joint angles) evolving together over time (see Methods for more detailed technical description).

The purpose of this study was to assess the effect of cueing an increase or decrease in peak vGRF on intralimb and interlimb coordination of the lower extremity during gait. We hypothesized that both high and low loading (i.e., increasing and decreasing vGRF) would alter interlimb and intralimb coordination. Specifically, we anticipated that high and low loading conditions that modified vGRF would, in turn, result in a less flexible, more rigid pattern of coordination.

Materials and Methods

Participants

Individuals were recruited from the university orthopedic practice and the University of North Carolina health system affiliated physical therapy clinics as part of a larger randomized trial (Luc-Harkey et al., 2018). Twelve individuals (age: 20±4 years; 8.8±1 month post-surgery; 4F) with a history of unilateral ACLR were assessed for this study (see Table 1). A priori power analysis was performed using G*Power software to ensure that the number of subjects included is adequate to reach statistical power (Faul, Erdfelder, Buchner, & Lang, 2009). Based on the statistical design (see Statistical Analysis section below), to detect a large effect (>0.6; Cohen, 2013) with an α set at 0.05, the analysis revealed 9 participants would be adequate for achieving power of .95. All participants were within 6–12 months post-ACLR, had a vGRF (bilaterally) greater than 1.05 of their body weight (in order to perform the low loading condition), had been cleared to return to full activity and engaged in at least 30 minutes unrestricted physical activity three times per week. Exclusion criteria included: 1) a lower-extremity musculoskeletal injury in the past six months, 2) additional lower-extremity surgeries other than the ACLR, 3) clinically diagnosed OA or symptoms relating to OA, 4) cardiovascular restrictions for physical activity, and 5) women who were pregnant. All procedures were in compliance with the University of North Carolina at Chapel Hill’s Institutional Review Board (IRB number: 13–2385) and guidelines thus followed regulations on ethical treatment of participants.

Table 1.

This table illustrates a summary of participant demographics.

Sex	4 Female, 8 Male
Age (years)	20±4
Height (cm)	171.5±6.04
Mass (kg)	72.5±14.3
BMI	24.5±4
Time since surgery (months)	8.8±1.3
Graft type	QD=2; PT=10
Tegner	6.5±1.2
Self-selected gait speed	1.25±0.11

Abbreviations: The data is presented as the mean ± standard deviation. BMI= body mass index, QD= quadriceps tendon autograft, PT= patellar tendon autograft.

Procedures

Participants completed a total of four visits. During the initial visit, written informed consent was obtained followed by collection of demographic information (see Table 1). Participants were then instructed to walk overground at a “comfortable” pace through two sets of infrared timing gates (TF100, Trac Tronix, Lenexa, Kansas, United States) for a total of five trials. The average gait speed of these trials was used as their preferred speed for treadmill walking. Participants then walked on a dual-belt, instrumented treadmill (Bertec Corp, Columbus, OH, USA) to capture baseline measures of vGRF. Participants completed a 5-minute acquisition period before data were recorded. A custom MATLAB (Mathworks R14, version 7.0 USA) algorithm was used during the baseline collection to extract the peak vGRF from the first 50% of the stance phase (see Figure 1). Baseline peak vGRF was determined for each limb by extracting and calculating the average of the left vGRF peaks and right vGRF peaks from 90 seconds of treadmill walking.

Figure 1.

This is a representative time series of vGRF during gait. The baseline peak vGRF was determined by extracting the peak vGRF from the first 50% of the stance phase from each limb. The left vGRF peaks and right vGRF peaks were then averaged from 90 seconds of treadmill walking (see Methods for more detailed description).

During three subsequent visits, participants completed three loading conditions (control, high, low) randomly allocated over three separate sessions (8.5±2.3 days between each session). Prior to data collection, retroreflective markers were placed on the lower-extremity of each participant by a single biomechanist across all testing sessions using previous methodology (Bell, Pedersen, & Brand, 1990; Luc-Harkey et al., 2018). Kinematic data were collected with an 8-camera three-dimensional motion capture system (Qualisys Medical AB, Gothenburg, Sweden) and Qualisys Track software. Joint-specific kinematic outcome variables (i.e., frontal and sagittal ankle, knee, and hip joint angular positions) were calculated for 3000 consecutive steps (roughly 30 minutes).

Real-Time Biofeedback

Participants completed a 5-minute acquisition period for each condition before data were recorded. During the control condition, no biofeedback was displayed. For the high and low visual biofeedback conditions a custom MATLAB algorithm was used to continuously compute the average left and right vGRF and display the visual cueing on a 72-inch screen directly in front of the treadmill. Kinematic data were sampled at 100Hz. The high and low loading conditions corresponded to a 5% body weight increase or decrease for each leg (i.e., high or low conditions, respectively) from the baseline vGRF. Bilateral target lines were displayed on the screen corresponding to the loading condition (see Figure 2). Prior to beginning the high and low loading conditions, participants were provided a demonstration of how the bars on the graph of the visual biofeedback would continuously display the vGRF for each limb. Participants were instructed to match the height of each vertical bar (peak vGRF from the first 50% of stance) to the red horizontal line during each step (see Figure 2). All participants were instructed to find a strategy that would maximize their likelihood of consistently reaching the target red horizontal line.

Figure 2.

Feedback was displayed as a vertical bar graph on a on a 72-inch screen while participants walked on a treadmill (A). (B) Provides an image of how the feedback was displayed on the screen. The bars on the bar graph represented each limb and displayed the magnitude of the average peak vGRF while participants performed the walking trials (sample data in C). A red horizontal line was displayed across the screen and represented the target vGRF requested for each condition. The peaks circled in graph (C) correspond to the two vGRF impact peaks for the left and right legs. The peak vGRF was determined as the peak vGRF from the first 50% of stance phase (see Methods for more detail). Participants were instructed to alter their gait to match each vertical bar (i.e., peak vGRF) to the red horizontal line (vGRF target) during each step.

Analyses

Relative Phase Analysis.

Joint angular position data were calculated using marker kinematics. Marker position data were filtered in Visual 3D using a low pass 4^th order Butterworth filter (10 Hz cutoff frequency) and used to calculate joint angular positions. To approximate the interlimb and intralimb phase relation, discrete relative phase was used to quantify coordination (Fuchs, Jirsa, Haken, & Scott Kelso, 1995). These analyses were performed on the sagittal plane joint angular positions using custom algorithms developed in MATLAB for Mϕ analyses. The Mϕ of interlimb (affected vs. unaffected: ankle, knee, hip) and bilateral intralimb (ankle-knee, knee-hip, ankle-hip) joints pairs were computed as a ratio of the time between successive peaks of peak flexion for respective joints over time. The specific joint pairs were chosen to provide an in-depth evaluation of the influence of loading biofeedback on lower limb coordination. For each stride (i) sagittal plane peak flexion for the ankle, knee, and hip joints was detected bilaterally and used to calculate intralimb and interlimb phase plane motion for each of the six joint pairs (intralimb: ankle-knee, ankle-hip, knee-hip, interlimb: ankle-ankle, knee-knee and hip-hip pairs). To quantify the Mϕ of the coordination for each joint pair in degrees based on a unit cycle, the ratio of the time between successive peaks ($\mathrm{\Delta }t$) to the stride time (T) of the involved limb (interlimb) or most distal joint of the joint pairs (intralimb) being assessed:

$${\varphi }_i=\left(\frac{\mathrm{\Delta }{t}_i}{T_i}\right)\times 360°$$

The Mϕ (i.e., coordination pattern) for all intralimb and interlimb joint pairs was then computed. SDϕ was calculated to quantify the coordination stability between joints.

Cross Recurrence Quantification Analysis.

The raw, unfiltered sagittal plane joint angle data for each joint (ankle, knee, hip) on each limb was also processed and submitted to CRQA (Shockley et al., 2002; Webber Jr & Zbilut, 2005). CRQA calculates the time-correlated activity between two signals (i.e., one joint pair of two joint angle time series in the current implementation) by embedding the two separate time series in a single reconstructed phase space via time-delayed copies of the original measured signal. See Marwan et al. for complete method (Marwan, Romano, Thiel, & Kurths, 2007). More than 25% of the trials were randomly selected and assessed, individually, for parameter selection to ensure proper resolution with minimal saturation given the analysis of the matrix. The following parameters were used in the final analysis: delay (t) equal to 33 data points (approximately ¼ of the gait cycle duration), an embedding dimension (m) of 5 based on a false nearest neighbors analysis, and the radius set to 0.8% of the maximum distance (Marwan et al., 2007; Shockley et al., 2002; Webber Jr & Zbilut, 2005).

CRQA provides a suite of output measures that are informative of the underlying coordination dynamics as computed from the structures of darkened pixels visible on a given recurrence plot (Figure 3). Seven recurrence variables are typically derived from the matrix parameters stated above, and the current study focused on three of these variables. The first, percent cross recurrence (%CREC), indicates the percentage of shared angular positions found within the embedding space based on the number of times both time series were in similar locations. The percentage of darkened pixels relative to all possible pixels, as illustrated in figure 3, equates the %CREC between the two time series (the percentage of darkened pixels expresses the degree of overlap of the two trajectories; the greater the percentage of darkened pixels, the higher the degree of overlap). %CREC has been shown to be inversely proportional to the amount of noise found within a given signal (Shockley et al., 2002). Second, percent cross determinism (%CDET) was examined, and is indicative of the percentage of time recurrent points fall on a diagonal line, or how often the two joints evolved together over time. The longer period of time two time series repeat together reflects a more rigid pattern. Third, to determine the extent the joint pairs were coordinated, relative entropy (rENTR) was computed. rENTR is calculated by normalizing the entropy value against the total number of lines found in the recurrence plot, providing a measure of the degree of regularity between two oscillators (Davis, Pinto, & Kiefer, 2017). A more rigid pattern would reflect rENTR values closer to zero.

Figure 3.

Sample of intralimb cross-recurrence plots of the hip and knee joint for the control (A), and high loading (B) walking conditions. The diagonal lines depict how often the trajectories of, for example, the hip and knee joint co-evolve together. The occurrence of recurrent points on diagonal lines in the high load condition (B) reflects the %CDET difference in load when compared to the control condition (A). B demonstrate a sequence of recurrent points circled in red, the longer period of time the two time series repeat together, the greater the frequency of recurrent points along the diagonal and, thus, the higher the %CDET.

Statistical Analysis

Interlimb Coordination.

To assess interlimb changes in coordination between the affected and unaffected limbs, we conducted separate repeated measures ANOVAs to compared the interlimb coordination at three lower limb joints (affected vs. unaffected ankle, affected vs. unaffected knee, affected vs. unaffected hip) across loading conditions (control vs. high vs. low) to analyze each of the dependent coordination measures (Mϕ, SDϕ, %CREC, %CDET, and rENTR), with α=.05. If a significant main effect or interaction among load conditions were observed, follow-up tests were conducted using Bonferroni-corrected post-hoc analyses were completed, when necessary.

Intralimb Coordination.

For the analysis of intralimb coordination, we conducted separate repeated measures ANOVAs for each joint pair (ankle-knee, knee-hip, ankle-hip) across loading conditions (control vs. high vs. low) for each of the dependent coordination measures (Mϕ, SDϕ, %CREC, %CDET, and rENTR), with α=0.05. Significant effects were explored using post-hoc analyses within the statistical design by applying Bonferroni-corrected post-hoc analyses when necessary. Effect sizes were computed based on Cohen’s d. All statistical analyses were performed using SAS statistical software (SAS Institute Inc., Release 8.0).

Results

Interlimb Coordination.

The resulting Mϕ for all loading conditions revealed the right and left limbs oscillated on average at 179° (range: 161° to 198°) as expected (Schöner et al., 1990) and did not significantly vary among loading conditions (p-values>.05; affected-unaffected ankle: d=0.14, knee: d=0.26, hip: d=0.11). Further, there were no differences in SDϕ (Affected-unaffected ankle: d=0.68, knee: d=0.87, hip: d=0.54; p-values>.05), nor were effects found in the CRQA (%CREC: affected-unaffected ankle: d=0.21, knee: d=0.01, hip: d=0.05; %CDET: affected-unaffected ankle: d=0.5, knee: d=0.29, hip: d=0.19; rENTR: affected-unaffected ankle: d=0.01, knee: d=0.15, hip: d=0.21; p-values>.05).

Intralimb Coordination.

Figure 4 illustrated the differences in intralimb Mϕ across the three joint segments for each condition. Main effects of loading condition for the ankle-knee (F_1,10=11.67, p=.007) and ankle-hip (F_1,10=18.85, p=.001) joint pairs were observed for intralimb Mϕ. For the ankle-knee joint pair, the low loading condition exhibited significantly greater Mϕ compared to the control (p=.02, d=2.65) and high (p=.026, d=1.82) loading conditions (control: 81.78°±4.57; high: 86.07°±6.69°; low: 100.17°±8.67°). For the ankle-hip joint pair, the low loading condition exhibited significantly greater Mϕ compared to the control (p=.004, d=2.14) and high (p=.006, d=2.13) loading conditions intralimb Mϕ observed for each joint pair under the three loading conditions (control: 23.59°±5.41°; high: 23.11°±6.16°; low: 40.90°±10.10°). Both high and low loading conditions demonstrated lower SDϕ values compared to the control condition however, these differences were not statistically significant (p’s > .05).

Figure 4.

This figure depicts the differences in mean relative phase values between the three different intralimb segments that were evaluated: A. the ankle-knee, B. the knee-hip, and C. the ankle-hip. * indicates a significant difference from the control condition. ¥ indicates a significant difference from the high loading condition.

Figure 3 shows representative cross-recurrence plots of an example participant. A significant loading condition effect was observed for %CDET for both the ankle-knee (F_1,10= 3.67 p=.031) and knee-hip segments (F_1,10= 7.80 p=.019); the high loading condition led to a more predictable coupling (i.e., higher %CDET) of the ankle-knee (p=.017, d= 0.78) and knee-hip (p=.001, d= 0.90) joint pairs than the control condition. No other effects were significant (%CDET: ankle-hip: d=0.17; %CREC: ankle-knee d=.14, ankle-knee d=.13, : ankle-hip: d=0.09; rENTR: ankle-knee d=.57, ankle-knee d=0.41, : ankle-hip: d=.42; p-values>.05).

Discussion

The current study was designed to assess the implementation of biofeedback to cue changes in loading and the influence this has on lower extremity interlimb and intralimb joint coordination dynamics. While there were similarities in the general patterns of intralimb coordination (i.e., Mϕ, SDϕ, and CRQA) across loading conditions, the loading conditions did change the pattern and predictability of intralimb coordination compared to the control condition. These findings support our primary hypothesis that cueing a change in peak vGRF alters movement coordination strategies. Specifically, the low loading condition resulted in significantly higher Mϕ, with increased intralimb joint coupling while the high loading condition led to a more predictable (higher %CDET) coupling pattern.

Typically, Mϕ is approximately 0° for intralimb coordination between lower limb joints, and is approximately equal to 180° for interlimb coordination between the matched lower-limb joints (Haddad, van Emmerik, Whittlesey, & Hamill, 2006). During the low loading condition, the values between the ankle-knee, knee-hip, and ankle-knee joint pairs stabilized farther away from 0° relative to the control and high loading conditions. This suggests participants may have varied their kinematic strategy during the low condition, which also resulted in alterations in joint coupling. A possible explanation for these findings could be that the low loading condition was more difficult to perform than the other conditions, and this was a sentiment expressed by all but one participant. Constrained interactions across joint pairs would produce more predictable coupling patterns, consistent with our %CDET findings.

While the results trended toward lower SDϕ during the low loading condition, %CDET was more sensitive to load-based changes, and revealed that the high loading condition had a greater influence on coordination dynamics. Specifically, the %CDET results reflected an increase in the predictability of coordination between the intralimb joint pairs over time. These results suggest ACLR persons exhibit a more predictable, possibly less flexible, coupling pattern when walking with higher magnitudes of vGRF. While our study did not utilize a comparison-control design, the findings are of significant importance in light of previous literature assessing individuals with ACLR, revealing ACLR gait demonstrates a more deterministic ankle-hip movement strategy (Kurz et al., 2005), decreased movement regularity at the knee (Moraiti, Stergiou, Vasiliadis, Motsis, & Georgoulis, 2010), and reduced intralimb coupling strength (Armitano et al., 2018). Armitano and colleagues (2018) also demonstrated that those individuals with greater deviations from antiphase coordination demonstrated decreased coordination stability within an ACLR population. To provide biofeedback in a more prescribed manner, it is of interest to assess the influence of biofeedback on coordination at the individual level to provide more targeted biofeedback interventions. Changes in coordination following injury have been expressed as the system relearning the degrees of freedom that define coordinative movements (Newell, 2003). The goal, then, is to provide biofeedback regarding lower extremity loading that both positively modifies mechanistic factors underlying positive health outcomes and establishes adaptable movement strategies (i.e., a combination of stability and flexibility).

The current CRQA findings indicate that cueing higher joint loading induces altered intralimb control strategies on both the affected and unaffected limb, which is believed to be a contributing factor to the initiation and progression of OA (Hurwitz, Sumner, & Block, 2001). However, there is evidence to suggest that both insufficient (Pietrosimone et al., 2018; Pietrosimone et al., 2017) and excessive (Blackburn, Pietrosimone, Harkey, Luc, & Pamukoff, 2016) lower limb loading can have deleterious consequences on joint health. Therefore, one explanation is that the cueing task is novel and, as a result, the system reduces its degrees of freedom to adapt to these changes in much the same way that an individual learns a new technique to a previously learned motor task (e.g., movement skills; Newell, 2003). Alternatively, the high loading condition may be revealing implications for joint health that would be made clearer through a long-term evaluation of this biofeedback on the coordination dynamics during gait. It is therefore of interest to determine the longitudinal training effects of cueing changes in mechanical load and its implications for coordinated movement strategies and long-term joint health (Pietrosimone et al., 2016; Pietrosimone et al., 2019). Specifically, while biofeedback can elicit positive changes to gait dynamics, the lack of a clear understanding of the effect(s) of a specific biofeedback type or regimen on coordination could result in motor behavior changes that are less functionally adaptive. In addition, it is unknown if the results of the current study reflect how cueing changes in peak vGRF will ultimately influence coordination, or if the changes are due to reorganization of the normal control processes that govern gait dynamics.

We showed that the addition of biofeedback influenced intralimb coordination; however, no differences were observed for the interlimb coordination of any of the joint pairs. Several studies indicate that intralimb and interlimb coordination provide different scales for quantifying the behavioral patterns generated during locomotion (Barela, Whitall, Black, & Clark, 2000; Haddad et al., 2006); e.g., assessment of intralimb coordination provides insight into the limb-specific coordinative patterns while interlimb coordination assessment provides higher-order information of the coupling behavior. While there is no known research that has assessed the implications of loading biofeedback on coordination dynamics, the results of the current study may suggest that the coordination of the degrees of freedom within each limb adjust and reorganize based on imposed constraints (i.e., the use of loading biofeedback) while preserving the interlimb coordination. It is also important to note that the biofeedback was prescribed separately for each limb; right side modified to prescribed right limb targets and left side modified to prescribed left limb targets. If, rather, the biofeedback had been prescribed so that both limbs were to target a bilateral average, we might have anticipated intra and interlimb changes. Therefore, depending on the design of the biofeedback paradigm utilized, it is still important for future research to consider the influence imposed constraints (such as biofeedback) have on interlimb coordination (and limb coordination symmetry) to get a more complete picture of the response by the locomotor system (Sadeghi, Allard, Prince, & Labelle, 2000).

The results of the study highlight important differences in gait coordination with cueing an increase or decrease in vGRF. However, there are limitations that should be considered when interpreting the findings. First, the cohort of participants recruited were a population with a propensity to walk with altered lower limb loading rates and, coincidentally, reduced intralimb coupling. This was purposeful to assess whether cueing an increase or decrease in loading would lead to improved gait dynamics, however, by not being a comparison-control design, the generalizability of the findings may be specific to this particular gait pattern. Further, the use of a treadmill has been shown to stabilize movements, obscuring subtle differences in coordination that might be illuminated when assessing over-ground walking (Dingwell, Cusumano, Cavanagh, & Sternad, 2000). It is of interest to investigate whether coordination dynamics are influenced in a similar manner while performing the biofeedback task during over-ground walking as well.

Conclusions

The results of this study illustrate that a biofeedback paradigm designed to influence a change in peak vGRF also modifies the intralimb coordination dynamics during gait. Alterations in coordination and stability under loading conditions were reflected by changes in Mϕ and %CDET. This can be interpreted as the underlying control processes governing gait have the propensity to be manipulated in a prescribed manner to target changes for healthy coordination dynamics. Overall, the current study shows the potential of applying biofeedback interventions to target mechanisms that induce a positive behavioral change and to promote movement adaptability.