1 INTRODUCTION

The annual incidence of anterior cruciate ligament (ACL) injuries among young and middle-aged athletes continues to rise, most of which are surgically reconstructed (ACL reconstruction [ACLR]) with the hopes of restoring knee stability and allowing the athlete to return to their preinjury level of sport.1, 2 Unfortunately, between 15% and 35% of ACLR patients who successfully return to sport will sustain a second ACL injury to either their surgical or contralateral limb.3, 4 Along with the risk for second ACL injuries, ACLR patients have an elevated risk for injuring other knee joint structures (e.g., meniscus),5 developing early signs of osteoarthritis,6 and having self-reported joint pain and impaired quality of life.7 Previous research has demonstrated that ACLR patients have poor movement biomechanics during athletic movements,8-13 and that movement deficits prospectively contribute to their heightened risk for second ACL injuries14 and early signs of knee osteoarthritis.7, 15 Though research has demonstrated that movement deficits are important to address in ACLR patients, movement quality is not typically assessed when determining return to sport readiness.16, 17 Multiple recent clinical reviews have suggested that assessing movement quality is a critical next step in advancing the care for ACLR patients.18, 19

The primary hurdle for incorporating movement quality into return to sport testing is the lack of available methods to assess human movement which are both valid and feasible to implement in clinical settings. Human movement is traditionally measured using three-dimensional (3D) motion analysis and embedded force plates for laboratory-based research. However, as these technologies are expensive and require a large amount of space, they are not practical for most clinical settings. The landing error scoring system (LESS)20 is a clinical screening tool for assessing movement quality which has been recommended for return to sport testing in ACLR patients.21 Clinical screening tools rely on having a trained movement scientist visually assess and score multiple movement qualities using a standardized rubric. While not requiring instrumentation is advantageous for clinical settings, these screening tools are subjective and have limited sensitivity. The subjective nature of the tools leads to intrinsic between-rater and within-rater variability, especially when comparing raters with different experience levels or educational backgrounds.22 Additionally, as aspects of movement (e.g., knee valgus) are scored as either good or bad, clinical screening tools are not well suited for identifying subtle changes or differences in landing mechanics. Likely resulting from these limitations, the LESS was not able to identify specific differences between ACLR patients and healthy controls,23 and was not able to prospectively predict primary ACL injuries in a large cohort of 5047 athletes.24 Assessment tools that are objective and sensitive could improve our ability to identify landing mechanics deficits in clinical settings.

Two-dimensional (2D) video analysis is a common method for measuring movement kinematics, and has been used in clinical settings.25 While only requiring one camera, video analysis has good agreement with traditional 3D motion capture for assessing knee kinematics in the frontal and sagittal plane.26, 27 The knee frontal plane projection angle (FPPA), has been widely assessed with regard to ACL injury prevention, and is computed as the angle formed between the hip, knee, and ankle joints in the frontal plane.27, 28 Knee FPPA is prospectively associated with first ACL injuries in uninjured athletes29 and second ACL injuries in athletes returning to sport following ACLR.14 Despite its clear importance with regard to ACL injuries, there is currently a lack of research assessing knee FPPA in ACLR patients. While it is known that 3D knee abduction is different between ACLR patients and uninjured controls,10, 11 it is unknown whether there is a difference between these groups for the knee FPPA. Video analysis has been used to identify sagittal plane landing mechanics deficits in ACLR patients.25 Welling et al. used video analysis to assess knee flexion during hop testing, and found that ACLR patients landed with significantly less knee flexion on their surgical limb, relative to their non-surgical limb.25 These findings are clinically important, as reduced knee flexion symmetry during a unilateral drop-landing task is prospectively associated with worse scores on the knee injury and osteoarthritis outcome score (KOOS) pain and quality of life sections at 2 years following ACLR.7 However, Welling et al.25 did not compare knee flexion symmetry between ACLR patients and uninjured controls. With a goal of rehabilitation being to restore pathological movement back to a healthy preinjury state, and as uninjured athletes do display some amounts of knee valgus and landing asymmetry, determining whether video analysis can identify differences between ACLR patients and uninjured controls is an important step when considering video analysis as a component of clinical return to sport testing.

Assessing kinetic symmetry during bilateral landing is also important with regard to second ACL injuries. A large body of literature has demonstrated that ACLR patients shift weight off of their surgical limb and rely more heavily on their non-surgical limb during a bilateral landing.9, 11-13 Furthermore, reduced knee extension moment symmetry during bilateral landing at the time of return to sport is prospectively associated with an increased incidence of second ACL injuries14 and early signs of knee osteoarthritis.15 While computing knee extension moment requires 3D kinematic and kinetic data, 61% of the variance in knee extension moment symmetry can be explained through plantar force impulse symmetry, using two wireless force sensing insoles.30 This finding suggests that impulse symmetry could be used as a surrogate for knee extension moment symmetry in clinical settings. Additionally, force sensing insoles have been validated against embedded force plates with excellent agreement for impulse symmetry during bilateral landing (intraclass correlation coefficient = 0.967).31 However, using portable devices, such as force sensing insoles to identify kinetic asymmetries during bilateral landing in ACLR patients has not been previously explored.

There has been limited research to date that explores the ability to identify landing mechanics deficits in ACLR patients using clinically feasible instrumentation in non-research settings. The purpose of this study was therefore to compare landing kinematics and kinetics between ACLR patients and uninjured controls in a non-laboratory setting. Based on previous laboratory-based literature comparing landing mechanics between ACLR patients and healthy controls,8-12 we hypothesized that ACLR patients would have (1) reduced impulse symmetry during bilateral landing, (2) increased non-dominant (or surgical limb) knee FPPA range of motion during bilateral landing, and (3) reduced knee flexion range of motion symmetry during unilateral landing, all relative to uninjured controls.

2 METHODS 2.1 Participants

Sixteen patients recovering from a primary unilateral ACLR signed university institutional review board approved informed consent and participated in this case-control study (level of evidence III). All ACLR patients met the following inclusion criteria: (1) between the ages of 14 and 30 years old, (2) released to return to unrestricted physical activity (6–12 months post-ACLR), (3) primary unilateral ACLR, (4) plan to return to a sport that involves jumping and/or cutting, and (5) have no pre-existing condition that limited physical activity. ACLR patients were compared with a sample of 16 gender-matched healthy control participants who participated in a previous jumping/landing study with the same data collection protocol and for which they signed informed consent before participation. An a priori power analysis for the primary outcome measure, impulse symmetry during bilateral landing, was completed using previously published data (ACLR: 82.3% ± 17.3%; Control: 100.6% ± 16%),31, 32 with a significance of 0.05% and 90% power, which resulted in a sample size of 15 participants per group. All healthy control participants met the following inclusion criteria: (1) between the ages of 18 and 30 years old, (2) self-report of being recreationally active, defined as participating in physical activity at least three times per week for at least 30 min, (3) self-report of having prior experience playing a sport that involved landing, (4) self-report of being injury free, defined as no injury in the previous 3 months and no current pain that impacted mobility, (5) self-report of no prior major lower extremity injury or surgery, and (6) no pre-existing condition that limited physical activity.

2.2 Procedure

All healthy control participants completed the following protocol in a university athletic center, and all ACLR patients completed the same protocol in a local rehabilitation clinic (Figure 1). Before testing, all ACLR patients completed the anterior cruciate ligament return to sport after injury scale (ACL-RSI), which quantifies a patients psychological readiness for returning to sport following ACLR,33 and the 12-item short form of the KOOS-12, which quantifies knee pain, function, and quality of life.34 These outcomes were collected to describe the demographics of the ACLR patients. All participants wore their own athletic clothing; however, based on the influence of footwear on movement mechanics,35 all participants wore standardized neutral cushioned running shoes (Air Pegasus; Nike Inc.). Six reflective markers were placed bilaterally on the lateral aspect of the femoral epicondyle, thigh, and shank segments (Figure 1). Thigh and shank markers were used as tracking markers, and therefore were not placed over specific anatomical landmarks.26 Markers were placed so they could be seen in the frontal and sagittal plane, and were at least 15 cm apart to reduce measurement noise.36, 37 Two-dimensional marker trajectories were recorded at 240 Hz in the frontal and sagittal planes using an iPad Pro and an iPhone 6s (Apple Inc.), with an image size of 1280 by 720 pixels. All videos were recorded with the flash on, which illuminated markers and allowed for automated point tracking. The iPad was placed approximately 2.6 m in front of the landing target and 0.75 m off the ground with a horizontally orientation. The iPhone was placed approximately 1.7 m to the left of the landing target and 0.7 m off the ground with a vertical orientation (Figure 1A–C). Plantar impact forces were measured bilaterally using loadsol sensors (Novel Electronics), which were calibrated based on manufacturer guidelines.31 Plantar forces were measured at 100 Hz for the healthy control participants and 200 Hz for ACLR patients due to an advancement in the sensor technology between the collection time points. However, as loadsol measurement accuracy is improved when sampling at 200 versus 100 Hz,31 loadsol data for the ACLR patients was down sampled to 100 Hz for comparison.

Testing setup for the (A) university athletic center where all control participants were tested, and the (B,C) two physical therapy/rehabilitation centers where all anterior cruciate ligament reconstruction (ACLR) patients were tested. (D,E) Participants wore their own athletic clothing, six reflective markers, and a pair of standardized running shoes fitted with loadsol sensors [Color figure can be viewed at wileyonlinelibrary.com]

All participants completed seven trials of a bilateral drop vertical jump and seven trials of a unilateral drop landing on each limb. For the bilateral task, participants began standing on a 31 cm high box, were instructed to jump forward towards a target placed half their body height away from the box, land with both feet on the ground, then change direction as quickly as they could and jump vertically as high as possible.23 To improve initial jump height consistency, participants were asked to jump forward off the box, but not up in the air. This horizontal drop vertical jump was chosen based on previous literature which found that most ACL injuries occur when participants contact the ground with their center of mass posterior to their base of support,38 and that landing tasks which are designed to involve both horizontal and vertical momentum at ground contact increase risk factors for ACL injuries.39 For the unilateral drop landing task, participants were instructed to stand unilaterally on top of the 31 cm high box, drop straight off the box, and land on the ground with the same foot.8 For a unilateral landing trial to be counted as successful, the participant needed to maintain balance unilaterally for 2 s without any other extremity contacting the ground. If a trial was unsuccessful, it was redone until seven successful trials were recorded on each limb. Restrictions were not placed on arm movement during either landing task and no additional instructions were provided. Participants were encouraged to practice each task before testing so that they were familiar with each movement. To minimize fatigue, participants were given a 30 s rest between trials and 1-min rest between tasks.

For the healthy control participants, the bilateral landing was performed first, followed by the unilateral landing on the participant's dominant limb followed by the non-dominant limb. Healthy control participants were asked to kick a soccer ball, and the limb with which they kicked the ball was defined as the dominant limb. The dominant limb was tested first for control participants, based on hop testing literature which suggests testing a participants dominant/non-surgical limb first.40 However, when finalizing the testing protocol for the ACLR patients, we decided to vary the testing order to reduce ordering effect. To accomplish this, the data collector subjectively varied which test was performed first (bilateral or unilateral) and which limb performed the unilateral landing first.

2.3 Analysis Impulse symmetry was computed during bilateral landing using the loadsol data. This outcome was chosen as impulse is the strongest ground-based impact symmetry outcome for predicting knee extension moment symmetry,30, 32 which is prospectively associated with second ACL injury risk14 and early signs of knee osteoarthritis.15 Additionally, assessing impulse symmetry using loadsol sensors has excellent concurrent validity relative to using embedded force plates (interclass correlation coefficient = 0.967).31 All loadsol data was analyzed using the Load Analysis Program, which is an open-source MATLAB user interface that can be downloaded here: https://github.com/GranataLab/LAP. Plantar force was normalized to bodyweight, and impulse was computed bilaterally between ground contact and toe off for the first landing during the bilateral task (Figure 2D).30 Ground contact and toe off were identified at the instant when plantar force rose above and fell below 50 N, respectively. Impulse symmetry was computed for each trial using the normalized symmetry index (NSI), which reflects the difference between limbs on a scale from 0% (completely symmetric) to ±100% (completely asymmetric), normalized by the magnitude of outcome variation within each participant (Equation 1).41 We chose this symmetry index as it robust for a wide variety of kinematic and kinetic outcomes, and it is more stable than other symmetry indices and is bound between ±100.41 (1) Visual depiction of outcome measure calculations. (A–C) A line was formed between the thigh and knee markers and between the knee and shank markers in the frontal and sagittal view videos to form the knee frontal plane projection angle and knee flexion angles, respectively. (D) Impulse was computed as the area under the force-time curve for both limbs on the first landing of the bilateral task [Color figure can be viewed at wileyonlinelibrary.com]

The 2D video data was used to compute non-dominant (or surgical) knee FPPA range of motion during bilateral landing and knee flexion range of motion symmetry during unilateral landing. These outcomes were chosen based on their prospective association with second ACL injury incidence14 and self-reported knee pain and quality of life.7 All 2D video data was analyzed using automated video analysis for dynamic systems (AVADS), which is an open-source MATLAB user interface that can be downloaded here: https://github.com/GranataLab/AVADS.37 Recent work has demonstrated that using AVADS to compute knee FPPA and knee flexion range of motion is valid relative to 3D motion capture (interclass correlation coefficient = 0.907–0.986) and repeatable between-days (interclass correlation coefficient = 0.751–0.944).26 The reflective markers were automatically tracked throughout each video, and all marker kinematics were low-pass filtered using a fourth order recursive Butterworth filter, with a cutoff frequency of 13 Hz, which was determined based on the sampling frequency and movement dynamics.42 Knee FPPA and flexion angle were computed between a line connecting the thigh and knee markers and a line connecting the knee and shank markers.26, 43 Only range of motion outcomes were of interest, therefore, joint center locations were not estimated, and thigh and shank segment angles were not aligned anatomically.26 Initial contact was determined visually in each video. During the bilateral landing, knee FPPA range of motion during the first landing phase was computed for the surgical limb of ACLR patients and non-dominant limb of control participants. The landing phase was defined between initial contact and peak knee flexion, which was determined in the sagittal plane.14 Knee flexion range of motion was computed during the unilateral landing for each limb. Knee flexion range of motion symmetry was then computed using the NSI and averaged across trials for each participant.

The Shapiro–Wilk test and Levene's test were used to ensure that all landing outcomes were normally distributed and had equal variance between groups. Outcomes were then compared between ACLR patients and control participants using an independent samples t test. Cohen's d was computed to quantify effect size, with a value between 0.2 and 0.5 indicating a small effect, 0.5 and 0.8 indicating a medium effect, and greater than 0.8 indicating a large effect.44 Significance was set at an alpha of 0.05. All data are reported as mean ± standard deviation, and all statistical comparisons were completed using SPSS Version 24 (IBM).

3 RESULTS

Participant demographics for each group can be found in Table 1. For the ACLR patients, the average time following surgery at the testing visit was 9.7 ± 2.0 months. Twelve participants had a non-contact injury mechanism, two were incidental contact, and two were contact. Nine patients had bone-patellar tendon-bone autografts and seven had hamstring autografts. Eight patients had concomitant meniscal injuries. The average score on the ACL-RSI was 66.3 ± 27, and the average score on the KOOS-12 was 12.8 ± 9.8.

Table 1. Participant demographics ACLR patients Control participants Gender 7 Male, 9 female 7 Male, 9 female Race 14 White, 2 black, 0 Asian 13 White, 0 black, 3 Asian Age (years) 17.3 ± 1.4 23.6 ± 2.8 Height (cm) 172.1 ± 11.0 177.3 ± 7.9 Weight (kg) 73.8 ± 15.0 71.9 ± 16.5 ACL-RSI 66.3 ± 27 - KOOS-12 12.8 ± 9.8 - Abbreviations: ACLR, anterior cruciate ligament reconstruction; ACL-RSI, anterior cruciate ligament return to sport after injury scale; KOOS, knee injury and osteoarthritis outcomes score.

There were significant differences between groups with large effect sizes for impulse NSI during bilateral landing and knee flexion range of motion NSI during unilateral landing (both p < 0.005, d > 1.0; Table 2), where the ACLR patients had reduced symmetry relative to the control participants (Figure 3). No significant between-group difference was observed for knee FPPA range of motion during bilateral landing (p = 0.111, d = 0.58).

Table 2. Descriptive statistics and group comparisons for landing kinematics and kinetics Outcome Group Sx/ND NSx/D NSI p Value, effect size BL impulse ACLR 0.48 ± 0.09 0.62 ± 0.09 19.86 ± 8.66 p < 0.001 a, d = 1.95 Control 0.53 ± 0.09 0.54 ± 0.08 1.78 ± 9.79 BL knee FPPA ROM ACLR 16.2 ± 6.2 12.3 ± 6.4 - p = 0.111 b, d = 0.58 Control 13.0 ± 4.8 10.9 ± 5.8 - UL knee flexion ROM ACLR 34.9 ± 9.6 44.1 ± 10.7 17.72 ± 14.54 p = 0.004 a, d = 1.09 Control 35.8 ± 7.5 37.7 ± 8.4 3.99 ± 10.17 Abbreviations: BL, bilateral landing; D, dominant limb; FPPA, frontal plane projection angle; ND, non-dominant limb; NSI, normalized symmetry index; ROM, range of motion; UL, unilateral landing.

Plantar force impulse normalized symmetry index (NSI) and frontal plane projection angle (FPPA) range of motion (ROM) during bilateral landing and knee flexion (KF) range of motion NSI during unilateral landing compared between ACLR patients and control participants. * Indicates statistical differences between groups (p < 0.05). ACLR, anterior cruciate ligament reconstruction

4 DISCUSSION

This study identified important kinematic and kinetic deficits in landing mechanics among ACLR patients using data which was collected in a non-laboratory setting using validated and clinically feasible methods. Consistent with two of our hypotheses, we observed greater between-limb asymmetries in ACLR patients during both bilateral and unilateral landing, relative to heathy control participants. These between-group differences are consistent with previous laboratory-based research which used 3D motion capture and embedded force plates to assess landing mechanics.8, 9, 12 The present study demonstrates that deficits in landing mechanics can be identified in non-laboratory settings, which provides a foundation for clinical translation of movement analysis and could aid in return to sport decision making.

Our first hypothesis was supported, as the present study found that ACLR patients continue to offload their surgical limb and rely more on their non-surgical limb during bilateral landing, despite having been cleared to return to full sport participation. This between-group difference is clinically important based on the large effect size (d = 1.95), the large number of previous case-control studies with similar findings,9, 12 and the two prospective studies that reported the association between reduced kinetic symmetry and an increased risk for second ACL injuries14 and early signs of knee osteoarthritis.15 While identifying this landing deficit in a non-research setting is an important step, future work needs to determine if plantar force impulse symmetry is prospectively associated with second ACL injury risk, and begin to establish cutoffs for using this outcome to determine whether patients are ready to safely return to sport.

Our second hypothesis was not supported, as there was not a significant difference between groups for knee FPPA range of motion during bilateral landing (p = 0.111, d = 0.58). While this finding was unexpected, it does not contradict previous literature. Delahunt et al.10 and Goerger et al.11 found that ACLR patients had increased hip adduction, hip internal rotation, and knee abduction angles relative to uninjured controls during bilateral landing. These outcomes reflect the peak knee joint angle based on 3D analysis, whereas the present study assessed the knee FPPA range of motion, which is a 2D outcome. While knee FPPA range of motion may not be different between ACLR patients and healthy controls, knee FPPA is prospectively associated with both second ACL injuries in patients returning to sport following ACLR14 and first ACL injuries in healthy uninjured athletes.29 Therefore, we believe that knee FPPA is important to assess and should be included in future research focusing on predicting and preventing second ACL injuries.

Our third hypothesis was supported, as ACLR patients had reduced knee flexion range of motion symmetry during unilateral landing, relative to uninjured controls (p = 0.004, d = 1.09). This finding is consistent with a recent systematic review, which concluded that ACLR patients have low knee flexion during unilateral landings on their surgical limb, relative to their non-surgical limb.45 This landing deficit reflects a decreased ability to appropriately absorb and dissipate energy through the knee joint and quadriceps musculature when landing unilaterally on the surgical limb.46, 47 Low knee flexion has been associated with both larger surface-based impact forces47 and cadaveric modeling-derived peak ACL strain.48 Additionally, low knee flexion symmetry during unilateral landing in ACLR patients has been prospectively associated with worse knee pain and knee-related quality of life.7 Similar to kinetic symmetry during bilateral landing, knee flexion symmetry during unilateral landing does not appear to resolve with time following surgery.49 A recent cross-sectional study compared hop testing symmetry between patients who were 9 months post-ACLR, 12 months post-ACLR, 18 months post-ACLR, and 24 months post-ACLR.49 While the authors found that hop distance symmetry improved with time, there were no differences between groups for knee flexion symmetry during the hop tests.49 Therefore, knee flexion symmetry during unilateral landing should be improved before releasing patients to return to sport. Future work should determine if knee flexion symmetry during unilateral landing is associated with second ACL injury risk and work towards including this outcome in return to sport testing.

The present study used load sensing insoles to quantify kinetic symmetry and AVADS to automate marker digitization for extracting kinematic outcomes from video data. AVADS has the potential to be a useful video analysis tool, as it can simultaneously track multiple markers, post-process marker kinematics (e.g., low-pass filter and spline-fill gaps in data), and quickly compute a variety of outcomes.37 Additionally, the ability to extract continuous joint angle time series allows for plotting joint ensemble curves, assessing human movement variability, and employing additional statistical analyses, such as statistical parametric mapping. However, AVADS does require an active MATLAB license, which is often not available in clinical settings. Additionally, as videos need to be transferred to a computer for analysis, this method is not practical for providing patients with immediate feedback during rehabilitation. There are mobile applications available which allow users to compute 2D joint angles using manual digitization on a mobile device directly after recording a video (e.g., see the Angles Video Goniometer app50). Mobile application based manual video analysis could be a quicker and easier method for computing the kinematic outcomes assessed in the present study. We strongly recommend that clinicians place physical markers (e.g., tape) over anatomic landmarks while recording videos to increase the reliability of using manual digitization for computing joint angles.51 Additionally, while the present study used load sensing insoles to measure bilateral landing kinetics, there are other options available for assessing landing kinetics in clinical settings. Previous work has shown that portable force plates are also valid relative to embedded force plates,52 and can be used to identify between-limb asymmetries in single-leg vertical hopping in ACLR patients.53 Based on the evidence that landing mechanics are important to quantify in ACLR patients, and the fact that there are now multiple assessment tools available which are accurate, relatively inexpensive, and portable, clinicians should begin to incorporate quantitative measures of landing quality into their return to sport decision making.

The present study has limitations which should be taken into consideration when interpreting the study findings. The primary limitation is that the ACLR patients and healthy control participants differed in age and were tested at different locations. These differences likely impacted the outcomes of the present study. For example, the control participants were tested in an athletic center with a hard-wood floor surface and the ACLR patients were tested in rehabilitation facility with a carpeted or rubber floor. The hard wood floor surface was stiffer, which likely increases impact kinetics and could have also altered kinematics. If these assessment methods are translated into clinical settings for widespread return to sport assessment, fully standardizing testing conditions will be difficult. Therefore, the fact that we observed between group differences which aligned with our hypotheses and previous literature despite differences in testing location speaks to the external validity of the assessment methods employed in the present study. An additional and important limitation is that the testing protocol was not the same for the ACLR patients and uninjured controls. The loadsol data was sampled at 100 Hz when testing the control participants and 200 Hz when testing the ACLR patients, however, the later was down-sampled to improve group comparisons. Additionally, while the control participants completed the unilateral landing on their dominant limb and then non-dominant limb, the order of testing was varied for ACLR patients in attempts to reduce an ordering affect. While we do not believe these limitations would have changed our findings, they are important to consider when interpreting the study results.

In conclusion, ACLR patients have reduced plantar force impulse symmetry during bilateral landing and reduced knee flexion range of motion symmetry during unilateral landing, relative to healthy control participants. These differences are consistent with previous literature,8, 9, 12 and were found in a non-laboratory setting using data collection and assessment methods which are inexpensive, portable, and user-friendly. This study is an important step in translating biomechanical research into clinical practice.

ACKNOWLEDGMENT

Financial support for this project was provided by the Virginia Tech Graduate Student Assembly (Graduate Research Development Program).

AUTHOR CONTRIBUTIONS

All authors were fully involved in the study design, data analysis, and manuscript preparation. Additionally, all authors have read and approved the final submitted manuscript.