Effect of Prophylactic Knee Bracing on Anterior Cruciate Ligament Agonist and Antagonist Muscle Forces During Perturbed Walking

Authors: Raneem Haddara, Vahidreza Jafari Harandi, Peter Vee Sin Lee

Abstract

Background:

Anterior cruciate ligament (ACL) injuries most commonly occur after a perturbation. Prophylactic knee braces (PKBs) are off-the-shelf braces designed to prevent and reduce the severity of knee injuries during sports, yet their effectiveness has been debated.

Purpose:

To identify differences in ACL agonist and antagonist muscle forces, during braced and unbraced conditions, while walking with the application of unexpected perturbations.

Study Design:

Controlled laboratory study.

Methods:

A total of 20 recreational athletes were perturbed during walking at a speed of 1.1 m/s, and motion analysis data were used to create patient-specific musculoskeletal models. Static optimization was performed to calculate the lower-limb muscle forces. Statistical parametric mapping was used to compare muscle forces between the braced and unbraced conditions during the stance phase of the perturbed cycle.

Results:

The brace reduced muscle forces in the quadriceps (QUADS), gastrocnemius (GAS), and soleus (SOL) but not in the hamstrings. The peak QUADS muscle force was significantly lower with the brace versus without at 49% to 60% of the stance phase (28.9 ± 12.98 vs 14.8 ± 5.06 N/kg, respectively; P < .001) and again at 99% of the stance phase (1.7 ± 0.4 vs 3.6 ± 0.13 N/kg, respectively; P = .049). The SOL muscle force peak was significantly lower with the brace versus without at 25% of the stance phase (1.9 ± 1.7 vs 4.6 ± 3.4 N/kg, respectively; P = .031) and at 39% of the stance phase (1.9 ± 1.4 vs 5.3 ± 5.6 N/kg, respectively; P = .007). In the GAS, there were no significant differences between conditions throughout the whole stance phase except between 97% and 100%, where the braced condition portrayed a smaller peak force (0.23 ± 0.13 vs 1.4 ± 1.1 N/kg for unbraced condition; P = .024).

Conclusion:

These findings suggested that PKBs that restrict knee hyperextension and knee valgus/varus motion can alter neuromuscular patterns, which result in a reduction of QUADS force.

Clinical Relevance:

Understanding the way PKBs alter muscle function and knee mechanics can provide invaluable information that will help in making decisions about their use. Further studies should investigate different types of braces and perturbations to evaluate the effectiveness of PKBs.

Prophylactic knee braces (PKBs) are off-the-shelf knee braces designed to prevent and reduce the severity of knee injuries, such as anterior cruciate ligament (ACL) tears, sustained during sporting activity.^15^ Rehabilitation from ACL injuries is very expensive, and these injuries often result in pain, predisposition to osteoarthritis and other morbidities, quadriceps (QUADS) weakness and other neuromuscular deficiencies, and other long-term effects that present as a major rehabilitation challenge.^40^ As a result, researchers have focused on injury prevention, such as neuromuscular training programs, that aim to strengthen and alter dangerous neuromuscular coordination patterns to prevent injuries.^26^ The QUADS^11,13,35,39,43^ and gastrocnemius (GAS)^1,13,19,42^ are known as ACL antagonists and have been shown to contribute to ACL tears through anterior tibial translation (ATT) in cadavers, at low flexion angles. On the other hand, the hamstrings (HAMS) and soleus (SOL) are ACL agonists that provide counterbalancing forces against ATT by the QUADS and GAS.^13,19,35,42,43^ Nonetheless, not all programs are effective in reducing ACL injury rates,^48^ with causes including lack of compliance,^46^ specificity toward a particular skill level,^29^ and total exposure time.^57^

PKBs are meant to combat the factors that contribute to the failure of training programs as well as targeting less modifiable risk factors such as the person’s anatomy. PKBs were initially designed to protect the knee from contact injuries, such as a lateral impact to the knee joints, without compromising normal function. However, researchers have also sought to study how such an externally applied device can affect and alter muscle function and knee mechanics; they have found differences in kinematics,^7,59^ kinetics,^7,59^ and even muscle forces^16,23^ during various movements. Despite the popularity and widespread use of these braces, research in this area has shown conflicting results.^14,23,36,50,54^ The majority of the motion studies^7,59^ has examined changes in kinematics and kinetics with PKBs. Two studies^16,23^ have examined the effect of bracing on muscle forces using in vivo motion analysis as well as a combined in vivo/in vitro method. Moreover, two other studies^7,50^ have focused on using surface electromyography (EMG) to find differences in muscle activity due to bracing during a variety of dynamic activities. Ewing et al^16^ found that the HAMS and vasti (vastus lateralis, vastus medialis, vastus intermedius) muscles produced significantly greater flexion and extension torques, respectively, and greater peak muscle forces with bracing during an anticipated double-leg drop landing. Similarly, Hangalur et al^23^ also examined the effects of a PKB during drop landing and found a reduction in muscle forces and ACL strain. Osternig and Robertson^50^ reported significant reductions during running in the EMG activity of 6 lower-limb muscles crossing the knee and ankle with prophylactic bracing compared with control condition. Likewise, Branch et al^6^ evaluated 2 types of custom-fitted braces on ACL-deficient and healthy patients and reported a decrease in QUADS and HAMS activity during the stance phase of side-step cutting. However, much is left unknown about the effect of PKB on muscle forces during other movements, specifically during unplanned, unanticipated movements.

Current literature in motion analysis evaluates knee braces during anticipated and planned motions, which may not provide us with a full understanding of their effectiveness. Over 96% of ACL injuries occur during sudden, unexpected, dynamic movements. These injuries most commonly occur during awkward dynamic body movements, landing, and perturbations.^5,21,31,33,49^ Perturbations make the athlete unbalanced or lose control, which can ultimately lead to injury. As a result, examining the effect of a prophylactic brace on muscle forces in relation to ACL injury during unplanned perturbation may provide a better understanding of the effectiveness of the braces in a more realistic scenario. In turn, this will provide further insight into the efficacy of knee bracing and potential brace designs that can prevent injury. Currently, PKBs are not widely used as preventive interventions due to an absence of direct evidence of their effectiveness. The dynamic evaluation of the effectiveness of PKBs is a complex problem influenced by many variables; the braces can mechanically alter and restrict joint kinematics, or they can modify the neuromuscular activity of the lower limb.

As the effect of prophylactic bracing on neuromuscular coordination patterns is still not fully understood, we opted to examine how a PKB can perform during unanticipated movements. The purpose of this study was to examine the effect of PKBs during an unanticipated perturbation during which the wearer cannot plan for the specific movement. This will provide insight on how braces that restrict knee hyperextension and knee valgus/varus affect neuromuscular coordination and muscle forces during unanticipated perturbations. Although ACL injuries do not usually happen during walking, potentially injurious movement patterns during a disturbance to natural balance while walking could provide insight on what may be reproduced, on a higher scale, during high-impact and high-speed athletic tasks that can effectively tear the ACL.^28^ Furthermore, perturbation training, at low speeds, is used for nonoperative rehabilitation after ACL injury and postoperatively after reconstruction, on the assumption that control at low speeds will transfer to higher speed activities.^24^ Moreover, perturbed walking has been proven to reveal dynamic stabilization strategies; perturbations have commonly been used as models of injury mechanism to understand the neuromuscular and biomechanical responses to potentially injurious actions.^17,18,38,45,47,52,53,55^ It was hypothesized that bracing would decrease ACL antagonist muscle forces (QUADS and GAS) and increase ACL agonist muscle forces (HAMS and SOL).

Methods

Participant Recruitment and Preparation

Between July 2018 and November 2019, twenty healthy participants (10 men and 10 women) were recruited for this study. Participation criteria included individuals between the ages of 18 and 30 years who are regular participants in sports that included jumping, cutting, and/or pivoting >50 hours per year. Exclusion criteria included a history of or current chronic injury such as a fracture, tendon tear, or ligament tear to either lower extremity; no history of vestibular dysfunction; and no recent history (past 6 months) of acute lower-limb injury that necessitated medical attention. All participants signed an informed consent form approved by the relevant human research ethics committees. Each participant was fitted with a commercial PKB (K8 2.0; POD Active Pty) on both legs. The knee brace consisted of upper and lower forged carbon composite frames connected by medial and lateral hinges that included synthetic ligaments. These ligaments were inspired by the properties of native knee ligaments, providing progressive and multidirectional motion control. Proper fit was determined by measuring the width of the participant’s knees using the calipers and instructions provided from the manufacturer for a comfortably tight fit. A total of 45 retroreflective markers (0.014-m diameter) were mounted on each participant’s trunk, pelvis, thigh, shank, and feet using a custom marker set based on the study by Dorn et al^12^ (with minor variations). Instead of a sacral cluster, 2 posterior superior iliac spine markers were placed. Calibration markers were affixed to the participant’s medial and lateral femoral condyles and medial and lateral malleoli to define joint centers. The medial and lateral epicondyle markers were removed during the trials due to placement of the braces. Surface EMG was placed on the skin over the muscle bellies of the rectus femoris, vastus medialis, vastus lateralis, biceps femoris, semitendinosus, medial GAS, SOL, and tibialis anterior.

Instrumentation

The trials were completed using a computer-assisted rehabilitation environment (Motekforce Link). This system features a semi-immersive virtual reality screen, 10-camera (MX T20 S) motion capture system (Vicon Inc), and 1 × 2–m dual-belt treadmill instrumented with 2 force plates (Motekforce Link). EMG data were sampled at 2400 Hz using a wireless EMG system (Avanti Trigno; Delsys).

Perturbation Protocol

The patients walked at a speed of 1.1 m/s for 5 minutes before any perturbations were administered. The perturbation was a deceleration perturbation in which the belt under the dominant leg suddenly decelerated to a speed of 0 m/s with a maximum acceleration of 2.9 m/s^2^ about 85 milliseconds after toe off.^22^ This ultimately applies the perturbation at heel strike, causing the patient to experience sudden deceleration and temporarily lose balance. Three successful perturbations were recorded with a minimum of a 15-second washout period between them without the PKB (control) and another 3 with the PKB.

Data Preprocessing

The raw (labeled) kinematic marker, ground-reaction force (GRF), and muscle EMG data from each trial were exported from VICON Nexus (Version 2.8.2; VICON, Oxford Metrics). Marker trajectories and GRF data were smoothed with 10-Hz fourth-order Butterworth filters.^32^ Raw EMG data were also high-pass filtered using a fourth-order, zero-lag recursive Butterworth filter with a cutoff frequency of 20 Hz, full-wave rectified, and then low-pass filtered with a cutoff frequency of 15 Hz.^16^ The resulting linear envelopes were normalized to each muscle’s maximum EMG amplitude reached during stance phase.

Rigid-Body Modeling

Rigid-body modeling was performed using OpenSim, an open-source musculoskeletal modeling software, by following a standard pipeline.^10^ A 10-segment, 25 degrees-of-freedom (DOF) generic model (Gait 2392) was scaled to each patient’s mass and anthropometry using the static trial data. The hip was modeled as a ball-and-socket joint with 3 DOF, while the knee was modeled as a translating hinge joint with 3 DOF (flexion-extension, rotations, and valgus-varus). The lower-limb joints were actuated by 92 musculotendon units, where each unit was represented by a Hill-type muscle in series with an elastic tendon.^56^ Inverse kinematics computed the joint angles at each time instant by matching the model markers with the experimental marker trajectories such that the sum of the squared error distances between the 2 marker sets was minimized.^37^ Residual reduction analysis was used to create simulations that were dynamically consistent with the experimentally recorded GRFs.^10,44^ Next, static optimization solved for the individual musculotendinous forces by decomposing the net joint moments in an optimization problem that minimized the sum of the squares of all muscle activations.^9^ The muscle force results from static optimization were validated against EMG; data were qualitatively consistent with the computationally estimated muscle activations (Figure 1). Each walking cycle was normalized to 100% starting from heel strike to toe-off.

Statistical Analysis

The mean of three successful trials per condition for each patient was calculated, and the dominant leg was used for analysis. Muscles were combined into functional groups by calculating the sum of the contributions from each muscle within the QUADS (rectus femoris, vastus medialis, vastus intermedius, and vastus lateralis), HAMS (biceps femoris long head, biceps femoris short head, semimembranosus, and semitendinosus), and GAS (medial and lateral compartments of GAS). Statistical parametric mapping (SPM) was used to statistically compare muscle forces between men and women.^20^ Specifically, an SPM paired t test was used to compare the QUADS, HAMS, GAS, and SOL muscle forces during braced and unbraced conditions (α = .05). All SPM analyses were implemented using the open-source spm1d code (Version M0.1; www.spm1d.org) in Matlab (R2019a, 9.6.0.3427; The Mathworks).

Results

The study participants consisted of 10 healthy men (age, 22.2 ± 3.50 years; height, 1.76 ± 0.07 m; mass, 71.3 ± 11.6 kg) and 10 healthy women (age: 22.5 ± 3.24 years; height, 1.64 ± 0.08 m; mass, 57.1 ± 7.02 kg). Compared with the unbraced condition, the brace reduced muscle forces in the QUADS, GAS, and SOL. The peak QUADS muscle force was significantly lower with versus without the brace at 49% to 60% of the stance phase (28.9 ± 12.98 vs 14.8 ± 5.06 N/kg, respectively; P < .001) and at 99% of the stance phase (1.7 ± 0.4 vs 3.6 ± 0.13 N/kg; P = .049) (Figure 2). There were no significant differences between conditions in the HAMS (Figure 3). The SOL muscle forces were significantly lower for the braced condition at 25% of the stance phase (1.9 ± 1.7 N/kg [braced] vs 4.6 ± 3.4 N/kg [unbraced]; P = .031) and at 39% (1.9 ± 1.4 N/kg [braced] vs 5.3 ± 5.6 N/kg [unbraced]; P = .007) (Figure 4). In the GAS, the only significant difference between conditions was at 97% to 100% of the stance phase, during which the braced condition portrayed a smaller GAS peak force (0.23 ± 0.13 vs 1.4 ± 1.1 N/kg for unbraced; P = .024) (Figure 5). There were no significant differences in knee joint angles between braced and unbraced conditions (Figure 6).

Discussion

The brace significantly reduced QUADS peak forces from 28.9 to 14.8 N/kg at P < .001, which is in line with our hypothesis. Moreover, we hypothesized an increase in HAMS force with the PKB. However, our study showed no significant differences between braced and unbraced conditions during walking with perturbations. Looking at the lower leg muscles, the GAS’s force—which is an ACL agonist—was reduced in the last 3% of the gait cycle from 1.4 to 0.23 N/kg with bracing. This does not fulfill our hypothesis, as a reduction occurred in the terminal stance phase, whereas a reduction in the peak force was desired. Conversely, we hypothesized a force increase in the SOL—an ACL agonist; however, the brace had no effect on the highest peak force. Instead, the brace slightly delayed the SOL force production, causing lower SOL forces with the brace to be observed at 25% and 39% of the stance phase.

The reduction in QUADS peak forces from 28.9 to 14.8 N/kg at P < .001 agreed with our hypothesis. This is a beneficial change, as the QUADS can strain the ACL when they contract at low flexion angles and have been shown to induce ACL tears in cadavers.^11,13,35,39,43^ Cadaveric and in vivo studies^11,13,30,35,39^ have shown that QUADS contraction has an effect on ACL strain between 0° to 45° of flexion. The smaller the knee flexion, the greater the ACL strain with the same QUADS force. This was evident in a study by Beynnon et al^4^, in which they performed a static in vivo study where they applied 100 N of anterior shear load at 30° of knee flexion; 2 of 7 braces provided strain-shielding effects. However, when this was compounded with isometric QUADS contraction, none of the braces were able to provide protection to the ACL. Similarly, Branch et al^6^ also conducted an in vivo study that applied anterior shear load with active QUADS force and showed similar results. These studies^4,6,23^ show that PKBs do not reduce ACL injury by solely providing mechanical stability; changes in neuromuscular coordination and native firing patters are also required to reduce ACL strain.

Looking at the QUADS peak in our study, the peak occurred between 30° and 40° of knee flexion angle (Figure 6) with the brace, which falls within the “low flexion” angle range of 0° to 45°. This decrease in QUADS force from our results is in agreement with the study by Hangalur et al,^23^ where the brace reduced peak QUADS force from 5653 to 4666 N during a drop-landing motion. Similarly, our findings are in line with Osternig and Robertson^50^ and Branch et al,^6^ who both reported significant reductions in QUADS EMG while wearing a knee brace compared with an unbraced condition during running and side-step cutting, respectively. On the other hand, Ewing et al^16^ illustrated the opposite findings during drop landing. In that study, the QUADS showed an increase in force in the braced condition compared with the unbraced condition. Although Ewing et al^16^ studied a similar brace made by the same company, opposite results were observed in relation to QUADS forces. This is likely the result of the difference in movement or the absence of the anticipation factor in our study compared with their planned movement study.

The HAMS are ACL agonists that can decrease ACL strain.^13,19,35,43^ Therefore, we hypothesized an increase in HAMS force with the PKB. However, our study showed no significant differences between braced and unbraced conditions during walking with perturbations. On the other hand, a few studies examining muscle forces^23^ and EMG^6,50^ reported a reduction in HAMS muscle force and activity. Ewing et al,^16^ however, illustrated an increase in HAMS forces during the braced condition in landing. This disparity in results could be attributed to the variations in movements (ie, running,^50^ side-step cutting,^6^ drop-landing,^16,23^ and walking with perturbation), which may evoke different responses. Nonetheless, the studies conducted by Hangalur et al^23^ and Ewing et al^16^ were drop-landing and showed opposite results. This suggests that all PKBs may not work equally, instead altering the muscle activations and forces in a similar pattern.

The GAS force, which is an ACL agonist, was reduced from 1.4 to 0.23 N/kg with bracing in the final 3% of the gait cycle. Although a reduction in the GAS force is desired as it increases ACL strain by inducing ATT,^1,13,19,42^ the reduction happened at the end of the stance phase and not during the large peak GAS force between 60% and 80% of the stance phase. This does not fulfill our hypothesis, as a reduction in the peak force was desired. Correspondingly, the study by Ewing et al^16^ did not find any GAS force differences during the drop landings. Yet, the study by Osternig and Robertson^50^ depicted a significant reduction in GAS EMG activity during running. Again, this might be credited to the differences in the PKB brace design.

Conversely, the SOL is an ACL agonist; it induces posterior tibial translation at all flexion angles, with the greatest translations happening at 50° of flexion.^13^ However, the brace had no effect on the highest peak force. Instead, the brace slightly delayed the SOL force production, causing lower SOL forces with the brace to be observed at 25% and 39% of the stance phase. Similarly, neither Ewing et al^16^ nor Osternig and Robertson^50^ found any differences in SOL muscle forces and EMG during landing and running, respectively.

Our results do not fully align with the literature describing examined muscle forces during planned landing.^16,23^ This is likely the result of differences in movement (that is, drop landing versus perturbation) or because of differences in the biomechanics between planned and unplanned movements, as shown by several studies.^3,8,34,51,58^ These differences suggest that bracing might not consistently provide favorable muscle force patterns during a wide range of movements. Currently, PKBs are not widely used as preventive interventions due to an absence of direct evidence of their effectiveness. The dynamic evaluation of the effectiveness of PKBs is a multifaceted challenge affected by many variables; the braces can mechanically change and restrict joint angles, or they can modify the neuromuscular activity of the lower limb.^14^ The braces used in this study are meant to restrict hyperextension of the knee using mechanical stoppers, reduce knee valgus/varus, and add overall joint stability using the synthetic ligaments. Nonetheless, because of the complexity of the braces’ mechanisms of action, it is difficult to precisely compare our results with other prophylactic brace studies. Other studies have used different braces, conducted different motions, and had patients perform planned and anticipated movements.^3, 8,34,51,58^

Since the effect of prophylactic bracing on neuromuscular coordination patterns is still not fully understood, we opted to examine how a PKB can perform during unanticipated movements. The purpose of this study was to examine the effect of a specific kind of PKB during an unanticipated perturbation, where the wearer cannot anticipate the specific movement. This provided us with insight on how braces that restrict knee hyperextension and knee valgus/varus affect neuromuscular coordination and muscle forces during unanticipated perturbations. Considering the disparity of epidemiological research on PKBs, determining the extent to which different kinds of PKBs alter lower-limb function during a variety of movements will help us in understanding the efficacy of preventive bracing with similar properties.

Limitations of the Study and Future Work

Musculoskeletal modeling is a great tool to predict in vivo muscle forces without invasive methods; nonetheless, it has some limitations. First, the musculoskeletal models were scaled based on a generic model. Using real imaging data from the subjects, such as magnetic resonance imaging, would create a more accurate model. Second, we assumed that the inverse dynamics joint moments were satisfied by muscle forces alone and therefore did not include the effect of the brace, which may also exert a moment about the knee joint, directly into the model.^16^ Unlike functional knee braces, which are specifically designed to off-load certain compartments of the tibiofemoral joint, the exact mechanism of prophylactic knee bracing is unknown. Therefore, nonmuscular contributions were assumed to be negligible compared with the moments generated by the muscles. Third, static optimization methods do not include activation dynamics and have been critiqued for their failure to predict cocontraction. Despite this, static optimization in musculoskeletal modeling is a powerful technique to quantify how muscles function and has been used and validated by many researchers.^1,2,16,25,41^ Additionally, the perturbations were introduced during walking, which is unlikely to cause injury. Nevertheless, perturbed walking has been proven to reveal dynamic stabilization strategies; perturbations have commonly been used as models of injury mechanism to understand the neuromuscular and biomechanical responses to potentially injurious actions.^17,18,45,47,52,53,55^ Many studies^17,18,22,45,47,52,53,55^ have performed perturbations during walking and were able to observe differences. For example, Hurd et al^28^ studied the kinematic differences in normal and perturbed walking between male and female athletes. They concluded that female athletes exhibited characteristics during normal and perturbed walking, which may increase the risk of an ACL injury, and the repetition of these harmful patterns during provocative athletic movements may lead to injury. Furthermore, potentially injurious movement patterns during a disturbance to natural balance while walking could provide insight on what may be reproduced, on a higher scale, during high-impact and high-speed athletic tasks that can effectively tear the ACL.^28^ Moreover, the ACL ligament strain was not directly measured; therefore, it is difficult to accurately discern the impact of bracing on the ACL. Instead, we can infer what might happen to the ACL based on surrounding muscle forces. Finally, the perturbations may have introduced an inertial artifact in the sagittal plane moment component of the GRF. Although the compensatory mechanism developed by Hnat and van den Bogert^27^ has shown great success in in high speeds (96.75%), the model was able to accurately predict only the pitch moment at the speed of 1.2 m/s by 72.06%. Given that our study was conducted at 1.1 m/s, which is at a slower speed than in the study by Hnat and van den Bogert, we assumed the artifact would be smaller and even more inaccurately compensated for using an inertial compensation method. Therefore, we opted to not compensate for the error, as the predicted output underestimates the pitch moment. Despite these limitations of musculoskeletal modeling, it remains a powerful tool in understanding biomechanics without invasive procedures. Future work can build upon these findings, such as examining the effects of other forms of perturbations and unplanned movements using PKBs.

Conclusion

This study showed that PKBs that restrict knee hyperextension and knee valgus/varus altered neuromuscular patterns, thereby resulting in a reduction of QUADS force. Further research may include examining the effects of other types of perturbations and unplanned movements using PKBs. Understanding the way PKBs alter muscle function and knee mechanics can provide invaluable information that will help in decision-making regarding their use.