Reduction of risk factors for ACL Re-injuries using an innovative biofeedback approach: A phase I randomized clinical trial

Abstract

Objectives:

Determine the safety and initial efficacy of a novel biofeedback intervention to improve landing mechanics in patients following anterior cruciate ligament reconstruction (ACLR).

Methods:

Forty patients post-ACLR (age: 16.9 ± 2.0 years) were randomly allocated to a biofeedback intervention or an attention control group. Patients in the biofeedback group completed 12 sessions over six-weeks that included bilateral unweighted squats with visual and tactile biofeedback. Patients in the control group completed a six-week educational program. Lower extremity mechanics were collected during a bilateral stop jump at baseline, six-weeks, and 12-weeks post-intervention. Linear mixed-effects models adjusted for sex and graft type determined the main effects of and interactions between group and time.

Results:

No group by time interaction existed for peak knee extension moment symmetry. A group by time interaction existed for peak vertical ground reaction force symmetry (p = 0.012), where patients in the biofeedback group had greater improvements in symmetry between baseline and post-intervention that were not maintained through the retention assessments.

Conclusion:

This novel biofeedback program did not reduce risk factors for second ACL injuries. Future work could develop and test multidisciplinary interventions for reducing second ACL injury risk factors.

Clinicaltrials.gov identifier:

(NCT03273673)

1. Introduction

Athletes who play sports that involve cutting and jumping are at significant risk of sustaining a non-contact anterior cruciate ligament (ACL) injury, with incidence rates continuing to rise among younger athletes (Beynnon et al., 2014; Buller et al., 2015; Longo et al., 2020; Zbrojkiewicz et al., 2018). A majority of ACL injuries are surgically reconstructed (ACLR), with an associated cost of approximately $12,740 in the United States (Lubowitz & Appleby, 2011). Unfortunately, between 15% and 35% of patients who return to sport following ACLR go on to sustain a second ACL injury to either their surgical or contralateral limb (Grindem et al., 2016; Kyritsis et al., 2016; Paterno et al., 2014; Webster & Feller, 2016). A growing body of literature indicates that patients following ACLR and post-operative rehabilitation have significant movement deficits during dynamic tasks such as running (Sigward et al., 2016), jumping (Paterno et al., 2007), landing (Butler et al., 2016; Paterno et al., 2007; Renner et al., 2018; Schmitt et al., 2015), and hopping (Oberläander et al., 2013; Orishimo et al., 2010; Peebles et al., 2019). Deficits in landing mechanics at the time of return to sport have been prospectively identified as a risk factor for second ACL injuries (Paterno et al., 2010). As deficits in landing mechanics do not resolve with time and have been identified as far as two years post-operative (Butler et al., 2016; Paterno et al., 2007), it is imperative that interventional approaches to improve landing mechanics are identified and implemented during rehabilitation.

Greater between-limb differences in knee extension moment (i.e. reduced symmetry) at ground contact and increased surgical limb knee frontal plane range of motion during bilateral landing are associated with an increased incidence of second ACL injuries within the first 12 months following return to sport (Paterno et al., 2010). Since landing from a jump is a quick ballistic movement, modifying landing mechanics can be difficult. Alternatively, the mechanics of squatting are comparable to jumping and landing, with squatting being a slower and more controlled movement that can be repeated in a cyclic pattern, making squatting an ideal movement for motor learning. A recent study found a moderate to strong relationship between squatting and landing biomechanics for both frontal plane kinematic (R = 0.366–0.737) and kinetic symmetry (R = 0.382–0.573) in a cohort of patients following ACLR (Peebles et al., 2021a). Based on the relationship between squatting and landing biomechanics, interventions that focus on improving squatting biomechanics may translate to improved landing biomechanics.

Several studies have used real-time biofeedback to modify movement mechanics during squatting (Bonnette et al., 2020; Chan & Sigward, 2019; Kernozek et al., 2020; White et al., 2013). Chan and Sigward (Chan & Sigward, 2019) recently demonstrated that patients following ACLR can improve their weight distribution symmetry when squatting while visualizing real-time biofeedback by displaying the load under each limb. Bonnette et al. (Bonnette et al., 2020) demonstrated that a single session of visual biofeedback during squatting can change landing mechanics during a drop vertical jump in healthy athletes. Along with visual biofeedback, tactile feedback has been included in movement-based interventions for patients following ACLR. Snyder-Mackler et al. (White et al., 2013) included tactile biofeedback in an intervention for patients recovering from ACLR, where a Thera-band was placed around the knees during bilateral squatting to facilitate hip abduction. It is unknown however if this exercise has an impact on knee alignment during dynamic movements. While biofeedback is a promising approach for improving movement mechanics, it is currently unknown if targeting squatting mechanics during a biofeedback intervention can result in improved landing mechanics.

Improving landing mechanics in patients returning to sport following ACLR may be an important treatment to reduce the incidence of ACL re-injury. The purpose of this Phase I randomized clinical trial was two-fold. The first purpose was to determine the safety and initial efficacy of a novel biofeedback intervention to improve landing mechanics associated with risk factors for second ACL injury in patients following ACLR. We hypothesized that patients following ACLR who completed the visual and tactile biofeedback intervention would have greater improvements in bilateral landing mechanics, relative to attention controls who underwent an educational program regarding ACL injury prevention and recovery. The second purpose of this study was to collect data necessary to ensure that future intervention studies are adequately powered to detect statistical improvements in landing mechanics among patients following ACLR.

2. Materials and methods

2.1. Study design

This clinical trial was an assessor blinded, single-center ([REDACTED]), 12-week, parallel design randomized clinical trial. Patients were randomized into one of two groups: biofeedback or attention control. The biofeedback intervention consisted of 12 sessions over a six-week time period (two sessions each week), and the attention control program consisted of six educational sessions on ACL injury prevention and recovery, over a six-week time period. The duration and frequency of the intervention was chosen based on current rehabilitation protocols and previous clinical trials investigating movement-based rehabilitation in patients following ACLR (Chmielewski et al., 2016; White et al., 2013) and injury prevention for uninjured athletes (Taylor et al., 2018; Thompson et al., 2017). Study outcomes were assessed at baseline, prior to the intervention, at six-weeks immediately following the intervention, and at a 12-week retention visit. The study was conducted between February 2018 and March 2020. The study protocol was approved by the Human Research Protection Program (IRB# 17-007) of [REDACTED], and all study patients signed informed consent. Complete detail regarding the design and rationale for this study can be found elsewhere (Queen et al., 2021).

The study sample consisted of 40 patients following ACLR who intended to return to full sport participation after completing rehabilitation. Inclusion criteria were: 1) primary, unilateral ACLR, 2) completed at least 18 weeks of post-operative physical therapy, 3) willing to comply with all study procedures, and 4) aged 14–21 years. Exclusion criteria were: 1) history of more than one ACLR, 2) post-operative complications that required additional surgical intervention, 3) live greater than 60 miles from the research lab, 4) any limitations that would prevent the patient from attending the biofeedback training sessions, 5) participation in another ACL intervention outside of standard post-operative physical therapy, and 6) knee extension moment limb symmetry index greater than or equal to 90% at the baseline study assessment time point (see Measurements and Procedures section).

Participant recruitment was from five local orthopaedic surgeons from a single clinical practice. Eligibility of potential study patients was verified through an initial medical record screening followed by a phone screening and a subsequent in-person screening. Final exclusion criteria were determined during the baseline assessment following the completion of informed consent. The biofeedback intervention was completed by a clinician (physical therapist or athletic trainer) and the biomechanical assessment was completed by a second individual who was blinded to the participant group assignment (research technician). After verifying eligibility, patients were randomized into the biofeedback or control groups using a stratified randomization sequence generated by the project biostatistician. The randomization sequence was created using STATA 15.0 (StataCorp, College Station, TX, USA) statistical software, and was stratified by gender, age (young: 14–17 years old, older: 18–21 years old), and activity level (mild: Tegner score 1–3, moderate: Tegner score 4–7, vigorous: Tegner score 8–11), with a 1:1 allocation using random block size of 2 and 4.

3. Intervention

3.1. Biofeedback

The biofeedback intervention was a six-week program with biweekly sessions on non-consecutive days (12 sessions in total). The goal of the intervention was to heighten awareness of asymmetrical limb loading during squatting. During each intervention session, patients completed 3 sets of 10 squats with visual biofeedback and 3 sets of 10 squats with tactile biofeedback. The visual biofeedback consisted of patients visualizing a bar graph of the weight placed on each limb that was displayed in real-time (Fig. 1A). Weight distribution was quantified using two force plates (AMTI, Watertown, MA) measuring vertical ground reaction force, which was acquired using Qualisys Track Manager software (Qualisys, Goteborg, Sweden) and streamed to Visual 3D software (C-Motion, Bethesda, MD) where the graphs were built and updated in real-time. Along with weight distribution, the bar graph also displayed a green band that spanned from 45% to 55% of the patient’s bodyweight. The goal for each participant was to complete each squat while keeping the bar graphs level and within the green band (i.e. with an limb symmetry index >90%). The tactile biofeedback consisted of a resistance band that was placed around the surgical knee and pulled at a 45-degree angle toward the non-operative limb (Fig. 1B). The purpose of this tactile load was twofold: 1) pulling patients in the direction they typically shift their weight (i.e. towards the non-operative limb) resulted in the participant resisting the load by shifting weight back toward the operative limb to maintain stability, and 2) subjecting the surgical knee to an inwards valgus load resulted in the participant having to activate hip musculature to maintain good frontal plane knee alignment (White et al., 2013). It is important to note that the tactile load was not meant to result in strength gains, rather to train the patients to activate their hip musculature while squatting (White et al., 2013). As the appropriate tactile load for a participant varies with factors such as height, weight, and strength, the clinician who completed the first intervention session determined an appropriate tactile load based on their clinical judgement and measured this load using a handheld dynamometer (Rolyan Smart Handle, Performance Health, Warrenville, IL). The average tactile load was 11.7 ± 2.8 lbs., and the range was 2–15 lbs. The load was then kept consistent throughout the intervention protocol for each participant using the handheld dynamometer.

Fig. 1.

Example study participant completing the visual biofeedback (A) and tactile biofeedback (B) during an intervention session.

All squats were unweighted with the feet shoulder width apart. Squats were completed to the pace of a metronome set at 30 beats per minute, with the goal of completing one squat every two beats. A stool was placed behind each participant at a height such that the patients’ thighs would be parallel with the ground if they were to sit. During each squat, the patients were instructed to go down to the height of the stool without transferring any weight (Fig. 1). The visual and tactile biofeedback exercises were completed in a randomized order, and patients were provided with at least 20 s rest between sets and 10 min rest between exercises (Queen et al., 2021). Each session lasted approximately 30 min.

3.2. Attention control

The six-week attention control program focused on providing patients with education related to expectations upon returning to sport following ACLR and risk factors for second ACL injuries. Patients completed one educational session per week, with weeks one, three, and five completed online and weeks two, four, and six completed in person. Each session lasted approximately 30 min. A short quiz followed each educational session to assess how well the patients understood the content covered in each session.

4. Measurements and Procedures

Patient reported outcomes and biomechanical movement analysis were completed at the baseline assessment, immediate post-intervention, and retention visits. The patient reported outcomes included: 1) the Tegner Physical Activity questionnaire, that quantified the current level of physical activity, 2) the Anterior Cruciate Ligament Return to Sport after Injury (ACL-RSI) scale, that quantified the patients emotions, confidence, and risk-appraisal for returning to sport following ACLR, and 3) the 11-item Tampa Scale for Kinesiophobia (TSK-11) questionnaire, that quantified the patients fear of movement as it relates to the potential for pain and re-injury. Each of these outcomes have good internal consistency and have been used in similar studies (Webster et al., 2008; Woby et al., 2005). Any adverse events or serious adverse events which occurred during the intervention and assessment visits were recorded and investigated by the study team and the Institutional Review Board as required.

Patients wore form-fitting athletic clothing and standardized neutral cushioned running shoes (Nike Pegasus, Nike Inc, Beaverton, OR) during the biomechanical assessment. Retroreflective markers were affixed to the pelvis and lower extremities in a modified Helen-Hayes marker set (Fig. 2A). Markers were placed on bony landmarks by the same tester for every session. Marker coordinate data was collected during movement testing at 240 Hz using a 10-camera motion capture system (Qualisys, Gothenburg, Sweden), and ground reaction forces and moments were measured bilaterally at 1920 Hz using two triaxial force plates (AMTI, Watertown, MA). For the stop-jump trials, patients were asked to run 4–5 steps forward, jump off of one foot, land with two feet on both force plates, immediately change their momentum and jump up to a target height, and land cleanly back on both force plates (Fig. 2B). Approach speed during the stop-jumps was not standardized, but the data collector encouraged participants to complete the task at a fast pace while remaining safe. The patient’s maximum vertical jump height was measured at the baseline assessment (Vertec, Power Systems Inc, Knoxville, TN). The target height for the stop-jump trials was set using a tennis ball hung from the ceiling at 75% of the patient’s maximum vertical jump height. Three practice stop-jumps preceded the data collection trials.

Fig. 2.

Visualization of the modified Helen-Hayes marker set used to track pelvic and lower extremity kinematics (A) and example of a study participant completing the stop jump tasks during a biomechanical assessment (B).

Marker coordinate and ground reaction force data were exported to Visual 3D (C-Motion, Bethesda, MD) to compute biomechanical outcomes. Joint center locations and segment orientations were defined using anatomic markers and methods detailed elsewhere (Peebles et al., 2020; Queen et al., 2021). A fourth-order recursive low-pass Butterworth filter with a cutoff frequency of 7 and 100 Hz was used to smooth the kinematic and kinetic data, respectively. Knee joint angles and internal moments were computed during the first landing phase of each trial, with a cardan rotation sequence of sagittal, frontal, and then transverse planes. The landing phase was defined as the time between initial contact, when vertical ground reaction force exceeded 25 N, and when the pelvis center of mass reached its lowest position during ground contact. The primary outcome measure for this study was peak knee extension moment limb asymmetry index (ASI) computed for each stop jump (Equation (1)) and averaged across trials. We assessed symmetry of the peak knee extension moment instead of at initial contact as there can be slight differences as to when the feet contact the ground during a stop jump and knee extension moment at initial contact is quiet variable. Additionally, most ACL injuries occur during the weight acceptance phase of a landing, approximately 50 ms after initial contact (Krosshaug et al., 2007). The secondary outcome measures were peak vertical ground reaction force ASI, ground reaction force impulse ASI, peak knee flexion angle ASI, peak knee abduction angle, knee ab/adduction range of motion, which were also computed for each stop jump and averaged across trials.

$$ASI=ABS\left[\frac{Sx-NSx}{.5(Sx+NSx}\right]$$ Equation 1

where ASI – limb asymmetry index, ABS – absolute value, Sx – surgical limb, and NSx – non-surgical limb.

5. Statistical analysis

This study was a randomized clinical trial, with the goal of determining parameter estimates to use when computing the sample size needed to power a subsequent multi-center clinical trial. As prior clinical trials have not examined knee extension moment symmetry changes across time, we used knee joint position and knee extension moment as proxy variables to determine the sample size for the current study (Benjaminse et al., 2017; Dai et al., 2014; Dowling et al., 2012; Ericksen et al., 2015). Our a-priori power analysis suggested that 14 patients per group would be sufficient for detecting differences between groups with 80% power and a significance of 0.05. A total sample size of 40 patients was chosen to account for a potential dropout rate of 20%. Linear mixed-effects regression models with robust standard errors to account for the small sample size were used to determine the main effects of and interactions between group and time for our primary outcome and all secondary outcome measures. All mixed models adjusted for sex and graft type. A per-protocol analysis design was implemented, where patients were excluded if they: 1) did not attend at least 75% of treatment sessions (i.e., at least 8/12 biofeedback sessions or 4/6 education sessions), or 2) did not complete both the post-intervention and retention assessments (Queen et al., 2021). Significance was set at 0.05, and all statistical analyses were performed using STATA 15 (StataCorp, College Station, TX, USA). Post-hoc effect sizes were calculated for significant interaction effects using Cramer’s V (Cohen, 1988).

6. Results

The number of patients screened, randomized, and who completed the intervention and follow-up assessments are shown in Fig. 3. Forty-one patients met initial inclusion criteria and completed the baseline assessment; however, one participant had a knee extension moment ASI less than 10% (equivalent to an limb symmetry index greater than 90%) and therefore did not meet our final inclusion criteria. Of the 40 patients who were randomized, 22 patients were randomized into the control group and 18 patients were randomize into the biofeedback group. The mean, standard deviation, and group comparison for all baseline demographics are in Tables 1 and 2. There were no significant differences between the control and biofeedback groups at baseline in number of females, age, activity level, body mass index, and time since surgery. The control and biofeedback groups had an unbalanced distribution of graft type, with the control group having more hamstring grafts and the biofeedback group having more patellar tendon grafts (p = 0.01). Four patients in the biofeedback group dropped out of the study due to travel distance and lack of available time. Three patients in the control group had less than 75% adherence and were subsequently removed from the final statistical analyses. No related or unrelated adverse events were reported during the study.

Fig. 3.

CONSORT diagram reflecting all patients who were contacted, recruited, enrolled, and completed the intervention.

Table 1.

Participant demographics and patient reported outcomes compared between the biofeedback and control group at the baseline assessment time point.

	Biofeedback (n = 14)	Control (n = 19)	Total (n = 33)	Group Comparison
Gender
Male	8	10	18	p = 0.797
Female	6	9	15
Age at Surgery (years)
Mean ± SD	16.8 ± 1.8	17.0 ± 2.3	16.9 ± 2.0	p = 0.821
Median [range]	17 [14–21]	17 [14–21]	17 [14–21]
BMI (kg/m2)
Mean ± SD	24.5 ± 4.3	24.4 ± 3.8	24.4 ± 3.9	p = 0.937
Median [range]	23.9 [17.6–35.6]	23.8 [19.5–34.4]	23.8 [17.9–35.6]
Time Since Surgery at Baseline (months)
Mean ± SD	6.2 ± 1.5	5.5 ± 1.4	5.8 ± 1.4	p = 0.171
Median [range]	6.3 [3.2–8.8]	5.3 [3.5–9.7]	5.7 [3.2–9.7]
Tegner Before Injury
Mean ± SD	9.2 ± 0.8	8.9 ± 1.2	9.0 ± 1.0	p = 0.380
Median [range]	9 [7–10]	9 [6–10]	9 [6–10]
Tegner at Baseline
Mean ± SD	6.4 ± 3.4	4.9 ± 2.5	5.5 ± 3.0	p = 0.167
Median [range]	6 [0–10]	5 [0–10]	5 [0–10]
ACL-RSI
Mean ± SD	66.5 ± 22.9	57.6 ± 20.4	61.4 ± 21.6	p = 0.250
Median [range]	71 [15–93]	60 [24–100]	64 [5–100]
TSK-11
Mean ± SD	21.9 ± 2.8	24.1 ± 2.7	23.2 ± 2.9	p = 0.026 *
Median [range]	23 [16–25]	25 [17–28]	24 [16–28]

Note: groups were compared using an independent samples t-test; ACL-RSI – Anterior Cruciate Ligament Return to Sport after Injury; TSK-11 – 11-item Tampa Scale for Kinesiophobia; BMI - Body Mass Index;

^*indicates a significant between-group difference (p < 0.05).

Table 2.

Participant demographics compared between the biofeedback and control groups.

	Biofeedback (n = 14)	Control (n = 19)	Total (n = 33)	Group Comparison
Surgical Procedure
Patellar Tendon	10	7	17	p = 0.049 *
Hamstring	4	12	16
Presence of Meniscal Pathology
Yes	9	12	21	p = 0.947
No	5	7	12
Mechanism of Injury
Contact	3	4	7	p = 0.662
Incidental Contact	2	1	3
Non-Contact	9	14	23
Sport at Time of Injury
Football	5	6	11	p = 0.649
Basketball	3	5	8
Soccer	4	3	7
Lacrosse	–	2	2
Rugby	1	1	2
Volleyball	–	1	1
Gymnastics	1	–	1
Cadet Training	–	1	1
Race
Black/African American	–	1	1	p = 0.083
White	10	18	28
Unknown or Not Reported	1	–	1
More than one race	3	–	3
Ethnicity
Hispanic or Latino	1	1	2	p = 0.452
Not Hispanic or Latino	13	13	29
Unknown/Not Reported	–	2	2

Note: groups were compared using a Chi-Squared test;

^*indicates a significant between-group difference (p < 0.05).

Changes in the primary and secondary outcomes over the study period are reported in Table 3. An interaction between group and time was not observed for our primary outcome measure, peak knee extension moment ASI. There was, however, a group by time interaction for one of the secondary outcomes, namely, peak vertical ground reaction force ASI (p = 0.012), where patients in the intervention group had greater improvements in symmetry between the baseline and post-intervention assessments. However, this improvement in symmetry was not maintained between the post-intervention and retention assessment time points, during which the biofeedback intervention program was not performed. Fig. 4 depict the significant main effects of time and interaction between group and time for the symmetry outcomes. A main effect of time was observed for peak knee extension moment ASI (p = 0.002), peak vertical ground reaction force ASI (p = 0.030), vertical ground reaction force impulse ASI (p = 0.004), and peak knee flexion ASI (p = 0.013). For each outcome, symmetry improved across time. Fig. 5 depict the significant main effects of time for knee kinematic outcomes. There was a main effect of time for peak surgical and non-surgical limb knee flexion (p < 0.001), where knee flexion increased across time (see Fig. 5). There was also a main effect of time for peak non-surgical limb knee abduction (p = 0.043), where peak knee abduction increased across time. No effects of time or interactions between time and group were observed for knee frontal plane range of motion on the surgical or non-surgical limb (all p > 0.250).

Table 3.

Study outcomes compared between groups and time points.

	Baseline	Follow-up 1	Follow-up 2	Group × Time
				X2	Effect size Cramer’s V
Peak KEM ASI
Biofeedback	47.72 ± 7.36 (33.29–62.15)	41.71 ± 5.34 (31.25–52.17)	35.94 ± 5.64 (24.89–46.99)	0.46, NS	0.10
Control	44.75 ± 3.58 (37.73–51.76)	41.43 ± 3.75 (34.08–48.79)	34.17 ± 3.34 (27.62–40.73)
Peak vGRF ASI
Biofeedback	35.77 ± 5.62 (24.76–46.78)	24.40 ± 3.35 (17.83–30.97)	31.83 ± 3.91 (24.17–39.48)	8.83, p = 0.012	0.42
Control	28.38 ± 2.18 (24.12–32.65)	26.36 ± 2.91 (20.66–32.06)	23.10 ± 2.11 (18.95–27.24)
vGRF Impulse ASI
Biofeedback	44.84 ± 7.16 (30.81–58.86)	30.53 ± 5.97 (18.82–42.23)	34.69 ± 4.40 (26.08–43.31)	4.21, NS	0.29
Control	40.88 ± 5.58 (29.95–51.81)	36.99 ± 6.09 (25.05–48.93)	34.76 ± 4.61 (25.73–43.79)
Sx Frontal Plane ROM
Biofeedback	6.97 ± 0.89 (5.22–8.71)	7.85 ± 1.04 (5.82–9.88)	7.83 ± 1.09 (5.69–9.97)	2.64, NS	0.23
Control	9.90 ± 1.05 (7.85–11.95)	9.13 ± 0.98 (7.22–11.04)	10.06 ± 0.98 (8.15–11.97)
NSx Frontal Plane ROM
Biofeedback	8.07 ± 0.91 (6.28–9.86)	8.91 ± 1.21 (6.54–11.28)	8.44 ± 0.98 (6.51–10.36)	2.23, NS	0.21
Control	7.42 ± 0.88 (5.70–9.14)	7.53 ± 0.56 (6.43–8.62)	8.73 ± 0.88 (7.01–10.45)
Peak Knee Flexion ASI
Biofeedback	5.93 ± 0.80 (4.37–7.49)	5.43 ± 0.57 (4.32–6.54)	4.42 ± 0.61 (3.23–5.62)	0.78, NS	0.13
Control	7.53 ± 0.95 (5.66–9.40)	7.27 ± 0.99 (5.32–9.21)	5.45 ± 0.75 (3.98–6.91)
Peak Sx Knee Flexion
Biofeedback	78.73 ± 3.10 (72.66–84.80)	82.20 ± 3.04 (76.23–88.17)	85.11 ± 3.59 (78.07–92.16)	0.74, NS	0.12
Control	78.49 ± 2.61 (73.38–83.60)	81.36 ± 2.66 (76.13–86.58)	85.98 ± 2.80 (80.49–91.48)
Peak NSx Knee Flexion
Biofeedback	82.63 ± 3.22 (76.32–88.94)	85.28 ± 3.27 (78.87–91.70)	87.33 ± 3.67 (80.14–94.52)	0.61, NS	0.11
Control	83.02 ± 2.61 (77.91–88.13)	85.66 ± 2.69 (80.40–90.93)	89.01 ± 2.87 (83.38–94.64)
Peak Sx Knee Abduction
Biofeedback	3.57 ± 1.42 (0.79–6.35)	5.66 ± 1.34 (3.03–8.28)	4.06 ± 1.46 (1.19–6.93)	1.34, NS	0.17
Control	6.95 ± 1.77 (3.48–10.42)	7.31 ± 1.75 (3.88–10.75)	7.60 ± 1.78 (4.10–11.10)
Peak NSx Knee Abduction
Biofeedback	4.01 ± 1.10 (1.86–6.16)	5.66 ± 1.48 (2.76–8.55)	5.78 ± 1.31 (3.21–8.35)	0.02, NS	0.02
Control	5.19 ± 1.63 (2.00–8.38)	6.90 ± 1.36 (4.23–9.58)	6.84 ± 1.62 (3.68–10.01)

Note. Data are expressed as mean ± robust standard error (95% confidence interval). Models controlled for sex and surgery graft procedure. KEM – knee extension moment; vGRF – vertical ground reaction force; ROM – range of motion; Sx – surgical limb; NSx – non-surgical limb. Effect size for Cramer’s V for 2 degrees of freedom is interpreted as: 0.07 (small) 0.21 (medium) 0.35 (large).

Fig. 4.

Unadjusted symmetry outcomes across the 12-week study. Note: ASI – limb asymmetry index, vGRF – vertical ground reaction force. The black horizontal bars outline the second and third quartile, whiskers represent the first and fourth quartile, X represents the mean, the middle line in the plot box represents the median, and dots indicate outliers.

Fig. 5.

Unadjusted joint kinematic outcomes across the 12-week study. Note: NSx – non-surgical limb, Sx – surgical limb, ROM – range of motion. The black horizontal bars outline the second and third quartile, whiskers represent the first and fourth quartile, X represents the mean, the middle line in the plot box represents the median, and dots indicate outliers.

7. Discussion

Developing effective methods to improve lower extremity biomechanics during landing is an important step towards reducing the high incidence of second ACL injuries in athletes returning to sport following ACLR. This phase I clinical trial explored the safety and initial efficacy of using a novel visual and tactile biofeedback intervention to reduce risk factors for second ACL injuries. Patients in the intervention group performed slow cyclic squatting while receiving real-time weight distribution visual biofeedback and while resisting a tactile band pulling their surgical knee medially into a valgus position. The present study suggests that the intervention was safe for patients following ACLR preparing to return to sport. However, as there was not a significant group by time interaction for our primary outcome measure, peak knee extension moment symmetry, the results of the present study suggest that this simple intervention does not reduce risk factors for second ACL injuries. ACL injury prevention programs typically involve having athletes complete plyometric exercises while a trained movement scientist gives verbal feedback of key injury-risk metrics (e.g. knee flexion or valgus angles) (Taylor et al., 2018; Thompson et al., 2017). These ACL injury prevention programs have had mixed results for improving movement biomechanics and reducing the incidence of ACL injuries (Grimm et al., 2015; Taylor et al., 2015, 2018; Thompson et al., 2017; Zebis et al., 2016). Snyder-Mackler et al. conducted a randomized control trial focused on improving movement mechanics in patients following ACLR, with a control group who completed plyometric training only and an intervention group who completed plyometric training and a novel perturbation training program (Arundale et al., 2017; Capin et al., 2017; White et al., 2013). The authors found similar improvements for gait symmetry and hop distance symmetry across time for both the intervention and control groups (Arundale et al., 2017; Capin et al., 2017). Chan and Sigward found that patients following ACLR had acute improvements in vertical ground reaction force impulse symmetry while completing bilateral squatting with a visual biofeedback stimulus displaying the load under each limb (Chan & Sigward, 2019). The present study utilized a similar load distribution visual biofeedback stimulus to improve weight distribution symmetry in conjunction with tactile feedback for improving frontal plane knee alignment during squatting, and explored whether this intervention transferred to improved landing biomechanics in patients following ACLR. We did identify a group by time interaction for vertical ground reaction force symmetry, where patients in the intervention group had greater improvements in symmetry between the baseline and post-intervention assessment time points relative to patients in the control group. However, improvements were not retained at 12-week follow-up. There was also no evidence suggesting that the intervention improved peak knee extension moment symmetry or frontal plane knee alignment during landing. While introducing real-time visual biofeedback is a promising approach for augmenting injury prevention programs, the intervention program explored in the present phase I clinical trial did not improve second ACL injury risk factors.

Other authors have also explored whether using visual biofeedback to improve squatting biomechanics transfers to improved landing biomechanics (Bonnette et al., 2020; Kernozek et al., 2020; Marshall et al., 2020). There is evidence that visual biofeedback of knee valgus angles and moments during squatting can translate to improved knee valgus during unilateral and bilateral landing (Bonnette et al., 2020; Kernozek et al., 2020). Bonnette et al. (Bonnette et al., 2020) developed a squatting biofeedback program which combined lateral trunk lean, the ratio between knee and hip flexion, knee abduction moment, and vertical ground reaction force symmetry into one non-specific visual stimulus. To test whether the squatting biofeedback had any acute transfer effects on landing mechanics, 11 healthy female athletes completed three drop vertical jumps before and immediately after four sets of ten squats with the visual biofeedback stimulus (Bonnette et al., 2020). The authors found increases in knee flexion angle and decreases in knee extension moment following the biofeedback training (Bonnette et al., 2020). However, there were no changes in knee abduction angles or moments, and changes in kinetic symmetry during landing were not quantified. While the results of the present study suggest that biofeedback intervention program was not an effective method to reduce second ACL injury risk, biofeedback may still be a promising tool for motor learning and future work should continue to develop novel interventional approaches for improving movement biomechanics and reducing injury risk.

The results of the present study indicate that more complex and/or holistic intervention approaches need to be developed and tested with the goal of reducing second ACL injury risk factors. While the intervention group had acute improvements in peak vertical ground reaction force symmetry, there were no improvements for knee extension moment symmetry. As the visual biofeedback was based on vertical ground reaction force symmetry, it is possible that targeting peak knee extension moment symmetry during squatting could have been a more effective approach. Recent work developed a method to estimate knee extension moment in real-time through taking the cross-product between the three-dimensional ground reaction force and the position of a marker placed on the lateral aspect of the knee (Munsch et al., 2020). This real-time estimate had excellent agreement with traditional methods of computing knee extension moment (Munsch et al., 2020), and could be implemented during squatting. Along with targeting movement biomechanics, recent literature suggests that reduced surgical limb muscle strength (Schmitt et al., 2015) and reduced psychological readiness for return to sport (Peebles et al., 2021b) are significantly associated with peak knee extension moment symmetry during landing. Targeting these factors could therefore aid in improving knee extension moment symmetry. Finally, it is possible that the best way to improve landing biomechanics is to design an intervention which incorporates jumping and landing. While landing from a jump likely occurs too quickly for real-time biofeedback to be useful, it is possible that providing patients following ACLR with immediate post-trial feedback following each landing trial could be an effective interventional approach.

While patients in the intervention group had improvements immediately following the intervention for a secondary outcome measure, peak vertical ground reaction force symmetry, these improvements were not maintained at the follow-up assessment. To address the issue of retention, future work should implement a graded intervention program where treatment frequency tapers during the return to sport transition. Along with changing treatment frequency, future programs could explore increasing the difficulty of the intervention through changing squatting speed, introducing weighted squats, or including more challenging movement tasks.

The present study had limitations which should be considered when interpreting the results. The primary limitation was that the present study did not account for graft type when randomizing patients into the intervention and control groups. Therefore, graft type distribution was unbalanced between study groups – the biofeedback group had 10 patients with a patellar tendon graft and four with a hamstring graft, whereas the control group had seven patients with a patellar tendon graft and 12 with a hamstring graft. As graft type does affect kinetic symmetry in patients following ACLR (Mueske et al., 2018), this important between-group difference could have impacted the study findings. Therefore, future intervention studies targeting landing mechanics in patients following ACLR should include graft type as a factor in the randomization scheme. Additionally, the control group completed only six educational sessions, three which were completed online. As compliance for online sessions is difficult to assess, this is another source of potential bias. The results of our study could also have been affected because of the differential attrition in the biofeedback group, which was most frequently due to the travel time required to attend each session. Another limitation is the 90% limb symmetry index cutoff which was included as an inclusion criterion to ensure that only patients following ACLR which had asymmetrical landing mechanics were included in the present study. This 90% cutoff was chosen as it is commonly used for determining return to sport readiness based on hop distance symmetry. A better understanding of what normal and pathological levels of symmetry are for landing kinetics could have improved this inclusion criteria. Finally, the present study had a heterogeneous study sample, with a mix of sexes, varying stages of skeletal maturity, different graft types performed different surgeons, different rehabilitation protocols, and a wide range in time since surgery. As the present study was a phase 1 clinical trial with a low sample size, the wide heterogeneity in our study likely impacted the study outcomes and resulting conclusions.

In conclusion, the results of the present study indicate that the novel visual and tactile biofeedback program did not alter risk factors for second ACL injuries in patients recovering from ACLR. While there were improvements in peak vertical ground reaction force symmetry for the intervention group between the baseline and post-intervention assessment time points, these improvements were not retained at the retention assessment. While the present intervention program prioritized using clinic-friendly methods, future work could develop and test more multidisciplinary interventions that incorporate movement mechanics, muscle strength, and psychological readiness to return to sport.