Individuals Early after ACL Reconstruction Show Intact Motor Learning of Step Length via the Split-Belt Treadmill

Abstract

Background:

Rupturing the anterior cruciate ligament is an orthopedic injury that results in neuromuscular impairments affecting sensory input to the central nervous system. Traditional physical therapy after anterior cruciate ligament reconstruction aims to rehabilitate orthopedic impairments but fails to address asymmetric gait mechanics that are present post-operatively and are linked to the development of post-traumatic osteoarthritis. A first step towards developing gait interventions is understanding if individuals after anterior cruciate ligament reconstruction have the capacity to learn new walking mechanics.

Methods:

The split-belt treadmill offers a task-specific approach to examine neuromuscular adaptations in patients after injury. The potential for changing spatiotemporal gait mechanics via split-belt treadmill adaptation has not been tested early after anterior cruciate ligament reconstruction; nor has the ability to retain and transfer newly learned gait mechanics. Therefore, we used a split-belt treadmill paradigm to compare gait adaptation, retention, and transfer to overground walking between 15 individuals 3–9 months after anterior cruciate ligament reconstruction and 15 matched control individuals.

Findings:

Results suggested individuals after anterior cruciate ligament reconstruction were able to adapt and retain step length symmetry changes as well as controls. There was also evidence of partial transfer to overground walking, similar to controls.

Interpretation:

Despite disruption in afferent feedback from the joint, individuals after anterior cruciate ligament reconstruction can learn a new gait pattern using sensorimotor adaptation, retain, and partially transfer the learned gait pattern early after anterior cruciate ligament reconstruction. This may be a critical time to intervene with gait-specific interventions targeting post-operative gait asymmetries.

1. Introduction

Sustaining an anterior cruciate ligament (ACL) rupture results in reduced afferent input into the central nervous system (CNS) and impaired neuromuscular control.^14,20 An avascular, aneural tendon graft is used to replace the ruptured ACL during anterior cruciate ligament reconstruction (ACLR), which cannot compensate for the sequelae of sensorimotor impairments. Ligamentous injuries also cause pain and inflammation,¹³ peripheral deafferentation,^22,23 and joint laxity.²⁴ Collectively, these impairments induce undesirable neuroplastic changes, contributing to altered sensory input and motor output. Similarly, arthrogenic muscle inhibition (i.e., a sequela of neurologic responses following joint injury)¹⁹ of the quadriceps may further contribute to biomechanical changes seen post injury.

Asymmetric gait mechanics persist for years after ACLR^5,6,37,41 despite full recovery of strength and clinically assessed function.² While asymmetric knee joint mechanics have been associated with the consequent development of post-traumatic osteoarthritis (PTOA),^4,25,40 current clinical interventions have not been successful in restoring mechanics,^7,8 necessitating the development of interventions directly targeting these gait impairments. Prior to developing effective clinical interventions for post-operative gait, it is critical to understand and quantify the ability for individuals who have sustained an ACL injury to learn new walking patterns.

Motor learning can take several forms, each with separate neural mechanisms, and each of which may be impaired or intact in patient populations, based on the sensorimotor impairments associated with that population. For the purpose of this manuscript, learning is operationally defined as a change in motor behavior over time due to practice.¹⁷ One form of learning that has gained considerable attention in rehabilitation is sensorimotor adaptation. Sensorimotor adaptation learning is induced by the presence of a sensory prediction error, or the mismatch between the expected and actual sensory consequences of a movement.³⁹ Sensory prediction errors drive changes in the motor plan and subsequent motor actions over a period of repeated practice until the errors are minimized. Sensorimotor adaptation is reliant upon accurate, intact somatosensation from the limbs to generate the appropriate error signals that drive the change in motor commands. Given the neural changes seen after ACL injury, there may be changes in how individuals who have torn their ACL can adjust their gait pattern to novel circumstances (motor learning) through sensorimotor adaptation. Importantly, if individuals after ACL rupture are not able to adjust their gait in a short-term motor learning process similar to uninjured controls, there may be a reduced capacity for learning new gait patterns in the long-term, which may explain why current clinical interventions have not successfully restored post-operative gait mechanics. Determining the adaptability of the CNS early after ACLR will provide important insight to aid in the design of early rehabilitation strategies directly targeting gait biomechanics.

In gait, sensorimotor adaptation is commonly assessed using a split-belt treadmill.^12,16,27,28 Split-belt treadmills have a separate belt under each leg, each with its own motor, so a participant can walk either with both legs at the same speed (“tied-belt”), or with the legs going at different speeds (“split-belt”).²⁷ Walking in the split-belt configuration disrupts the normal interlimb symmetry of specific spatiotemporal gait parameters (e.g., step length), which induces the essential sensory prediction errors that drive the CNS to update its feedforward motor commands to restore the previous symmetry pattern despite the fact that the legs continue to move at different speeds.¹⁸ Split-belt treadmill adaptation of interlimb step length symmetry has been shown in individuals post-stroke²⁹ and in uninjured individuals.¹²

Changing symmetry of spatiotemporal parameters during split-belt treadmill walking demonstrates the capacity of the CNS to adjust gait in response to a shift in sensory input, and provides insight into both the neural control of gait and the capacity for motor learning in these populations.³ Step length symmetry is a very frequently used gait parameter to quantify sensorimotor adaptation capacity and magnitude, largely because of its highly reliable robust and reliable responses during split-belt adaptation.^12,16,29 Therefore, despite the fact that step lengths are not typically asymmetric in individuals after ACLR, it is an ideal variable to initially assess in this population because its responses to the perturbation induced by the split belts is so well characterized such that any deviations from the norm should be relatively easy to detect.^{1,3,12,18,21,27–30} If individuals after ACLR cannot adapt this gait parameter, it may be representative of the inability to address the aberrant joint-specific mechanics seen in this patient population.

One published study to date has found that some spatiotemporal parameters adapt in individuals 36 ± 24 months after ACLR in a single session of split-belt treadmill adaptation.³⁴ While this study measured immediate aftereffects, it is unknown if this type of motor learning is possible during the initial post-operative rehabilitation time point (i.e., prior to return to sport), when gait is initially addressed clinically, and when the afferent signaling from the knee joint is likely to be more impaired. Similarly, there was no second exposure to the split-belt with the same split condition, which would provide information on the ability of individuals after ACLR to retain new walking mechanics from one exposure to the next. Lastly, there was no test of transfer to a more salient overground environment.

Therefore, the purpose of this study was to determine the ability of individuals 3–9 months after ACLR to learn, retain, and transfer a new gait pattern by adapting and storing new step length symmetry patterns, and to compare the extent of adaptation, retention, and transfer of step length in these individuals to uninjured, age-, sex- and activity level-matched¹⁰ control individuals. We hypothesized that individuals after ACLR would be able to adapt and retain a novel step length symmetry pattern, but to a lesser extent than controls.

2. Methods

2.1. Participants

Fifteen individuals between the ages of 13 and 30 years old who underwent ACLR were recruited from the University of Delaware Physical Therapy Clinic and the surrounding area (Table 1). Participants were included if this was their first ACL rupture and they had no previous history of serious lower extremity injury, were 3 to 9 months post ACLR, and were verbally cleared for return to run by a health care provider. Participants were excluded if they had any cardiovascular or neurological conditions which contraindicated exercise. Fifteen control subjects were matched by age (±2 years), sex, and level of sport participation, and were recruited from the University of Delaware community and surrounding area. Controls were excluded if they had a previous history of ACL rupture, or major lower extremity injury requiring surgery. This study was approved by the University of Delaware Institutional Review Board and informed consent was obtained from all participants.

Table 1:

Demographics for each group (mean ± SD).

	ACLR (n=15)	Control (n=15)	P-value
Age (y)	20.83 ± 3.54	20.84 ± 3.07	0.67
Sex (f)	9 (60%)	9 (60%)	1.00
BMI (kg/m2) *	24.00 (22.10–26.10)	21.90 (19.80–23.60)	0.04
Tied Treadmill Gait Speed (m/s)	1.05 ± 0.17	1.06 ± 0.08	0.78
Treadmill Baseline SAI (%)	0.01 ± 2.30	−0.01 ± 1.15	0.98

^*Median and IQR; ACLR, anterior cruciate ligament reconstruction; BMI, body mass index; SD, standard deviation; SAI, Step length asymmetry index

2.2. Clinical Testing and Demographics

For the control group, limb dominance was determined with the question, “Which side do you use to kick a ball?” and the nondominant leg was considered the ‘involved’ limb. Patient characteristics, mechanism of injury, surgical details (e.g., meniscus status, graft type), and time since surgery were collected at the time of data collection for the ACL group. All participants underwent assessments of ³⁸knee joint proprioception and quadriceps strength.

Knee joint proprioception was assessed using active joint position sense.^9,15 Participants were seated on an electromechanical dynamometer (System 3; Biodex, Shirley, NY, USA) with their hips positioned to 90° of flexion and their knee positioned at 45° of flexion. The uninvolved knee was tested first, followed by the involved. Participants were instructed to close their eyes and straighten their knee from the starting 45° angle towards extension (0°). The tester locked the Biodex once the participants knee was anywhere between 18° and 22° and the participants were asked to actively control their knee (so it was not supported only by the dynamometer) and remember that knee joint position. Participants then actively returned to 45° and were asked to actively match the previous position at the speed of their choice and hit a button to lock the dynamometer in place once they believed they were in the same position. Participants underwent 2 practice trials followed by 5 real trials, each with a new target point between 18° and 22°. There was no feedback provided to participants after any of the trials performed. Error was calculated as the absolute difference from the initial angle to the participant-selected angle, and data were averaged across all 5 trials for statistical analysis.

Quadriceps strength and activation deficits were assessed using a burst superimposition technique during a maximal voluntary isometric contraction (MVIC).^35,36 Participants were seated on an electromechanical dynamometer (System 3; Biodex, Shirley, NY, USA) with their hips and knees flexed to 90 degrees and the axis of the dynamometer in line with the axis of the knee joint. The pelvis, thigh, and lower leg were stabilized with straps. Two self-adhesive electrodes (3 × 5 inches) were secured to the thigh, one proximally over the rectus femoris and one distally over the vastus medialis. Participants were familiarized to the testing procedures and performed 2 warmup sub-maximal trials (50% and 75% effort) and 1 warmup maximal (100% effort) trial. Participants were familiarized to the muscle stimulation with brief bouts of 10, 30 and 60 V stimulations. Participants then performed an MVIC with a supramaximal burst of electrical stimulation (600 microseconds, 135 V, 100-pulse-per-second train). Central activation ratio (CAR) measures a participant’s volitional activation of the quadriceps, representative of muscle inhibition, determined by comparing strength of the electric elicited force and the volitional force ((MVIC/electrically elicited contraction) *100%). Involved and uninvolved MVICs were normalized to body weight (kg) when reported individually (Table 2). A quadriceps strength index (QI) was used to measure strength of the involved limb’s quadriceps compared to the uninvolved limb’s quadriceps. A participant’s QI was calculated as a limb symmetry index (LSI=(involved limb MVIC/uninvolved limb MVIC) * 100%). An LSI of 100% indicates full symmetry, and an LSI <100% indicates the involved limb’s strength is lower compared to the uninvolved limb.

Table 2.

Clinical outcome measures for each group (mean ± SD).

	ACLR (n=15)	Control (n=15)	P-value
Quad Strength (MVIC)
Involved (Nm/kg)	2.54 ± 0.75	3.22 ± 1.02	0.023
Uninvolved (Nm/kg)	3.40 ± 0.75	3.36 ± 1.00	0.45
LSI (%)	74.32 ± 14.15	95.93 ± 8.15	<0.001
Quad Inhibition (CAR)
Involved (%)	95.90 ± 8.39	97.23 ± 5.82	0.48
Uninvolved (%)	96.36 ± 8.12	97.23 ± 6.19	0.49
Uninvolved Knee Proprioception Error (degrees)	2.21 ± 0.81	1.98 ± 1.00	0.50
Involved Knee Proprioception Error (degrees)	2.23 ± 1.79	2.08 ± 0.88	0.78

ACLR, anterior cruciate ligament reconstruction; MVIC, maximum voluntary isometric contraction; LSI, limb symmetry index; CAR, central activation ratio; SD, standard deviation

2.3. Split-Belt Treadmill Walking Paradigm

Participants underwent a split-belt treadmill locomotor adaptation paradigm based on prior literature.^1,28,30 The session consisted of 6 walking periods: overground baseline (5 trials), treadmill baseline (3 minutes), adaptation (adaptation I, 12 minutes), washout (12 minutes), a second adaptation period (adaptation II, 12 minutes), and overground post-adaptation (5 trials, Fig 1). During the overground walking trials, participants walked at their self-selected speed, with speed maintained within ±5% of the overground baseline trials using timing gates. During the treadmill baseline and washout periods, the treadmill belts were tied. Participants walked at their self-selected comfortable treadmill walking speed, determined prior to testing by asking participants to choose a speed they could maintain for up to 40 minutes of walking. During the first and second adaptation periods, the treadmill belts were split, in a 2:1 belt speed ratio,^1,27 to assess adaptation to a novel gait perturbation. When the belts were split, the fast belt remained at the self-selected comfortable walking speed and the slow belt was set to half that speed. The uninvolved limb was always placed on the fast belt and involved limb was always placed on the slow belt. Participants were instructed to keep just their fingertips on the treadmill handle in front of them during all walking periods, to reduce translation of the whole body on the treadmill belts.

Fig 1.

Split-belt treadmill adaptation paradigm. Participants walked at a self-selected speed while the belts were ‘tied’ or ‘split’ (2:1 speed ratio) with the uninvolved limb moving twice the speed of the involved limb.

The purpose of adaptation I was to assess for the presence and magnitude of any sensorimotor adaptation (Fig 1). The washout period assessed for the presence of any aftereffects, which would be characterized by the walking symmetry being perturbed a similar amount but in the opposite direction compared to the initial split-belt period. Aftereffects are taken as an indicator of storage of the newly learned gait pattern by the CNS. The adaptation II period assessed the degree to which there was any retention of the newly learned pattern following the behavior being washed out. Retention was characterized by a reduced initial perturbation caused by the split-belt configuration at the beginning of adaptation II compared to adaptation I. Finally, the post-adaptation overground period tested for any transfer of the newly adapted gait pattern from the treadmill to an overground environment.

2.4. Gait Data Collection

Thirty-nine retroreflective markers were placed on bony landmarks on both lower extremities and on rigid, multi-marker grouping tracking shells located on both thighs, shanks, and the pelvis.¹¹ Marker trajectories were captured using an 8-camera motion analysis system (Vicon MX, Los Angeles, CA) at a sampling rate of 100 Hz. A dual-belt treadmill with two embedded 6-component force plates (Bertec Corporation, Columbus, OH) collected forces and moments at 1000 Hz, time-synchronized with the kinematics in Vicon Nexus software.

2.5. Gait Data Analysis

Kinematic and force plate data were calculated using commercial software (MATLAB, TheMathWorks, Natick, MA, USA and Visual3D, C-Motion, Germantown, MD, USA). All kinematic data were filtered with a low-pass fourth-order Butterworth filter at 6 Hz. Moments were normalized by mass*height. Initial contact was defined as the time when the force reading first exceeded 20 N and toe off as the time when the force reading first decreased below 20 N for each step. Step lengths were calculated as the distance between the leading and trailing calcaneus markers at the time of initial contact. The primary outcome measure was step length asymmetry index (SAI), measured in the manner previously established in several motor learning studies:²¹

$$SAI=\frac{{StepLength}_{INVOLVED}-{StepLength}_{UNINVOLVED}}{{StepLength}_{INVOLVED}+{StepLength}_{UNINVOLVED}}\times 100\mathrm{\%}$$

This asymmetry index allows the difference between step lengths to be expressed as a percentage of the total stride length. An SAI of ‘0’ indicates full symmetry and a positive SAI indicates the involved limb has a greater step length.

SAIs were calculated over every walking stride and several key time periods, or ‘epochs’, of interest were identified. For treadmill walking, the epochs were: late baseline, early and late adaptation I, early and late washout, and early and late adaptation II. The ‘early’ epochs consisted of the average SAIs from the first 5 strides of the respective period; the ‘late’ epochs were the SAIs averaged over the last 50 strides. For overground walking, since only 5 trials were collected during each period, we averaged all 5 strides to measure the treadmill epochs, overground baseline and overground post-adaptation.

Sensorimotor adaptation was quantified as a reduction in the SAI from early to late adaptation I, following the initial perturbation from baseline expected in the early adaptation I epoch. The presence of a negative aftereffect, or a decrease in the SAI during early washout compared to late baseline, was taken as an indicator of CNS storage of the newly adapted pattern. Retention was quantified as a reduction in the SAI from early adaptation I to early adaptation II, showing that the same perturbation in belt speeds caused a less pronounced perturbation of the SAI during the second exposure (i.e., some of the learned response was retained). To assess transfer, we used a previously established measure from prior split-belt adaptation overground transfer studies:^1,30

$$Transferindex=\frac{{SAI}_{{Overground}_{POST}}-{SAI}_{{Overground}_{BASELINE}}}{{SAI}_{{Treadmill}_{EALYWASHOUT}}-{SAI}_{{Treadmill}_{LATEBASELINE}}}$$

2.6. Statistical Analysis

SPSS software was used for statistical comparisons (IBM Corporation, Chicago, IL, USA, Version 26). Normality was assessed in all clinical variables, and medians with interquartile ranges were used when normality was violated. Independent t-tests were performed to test differences between groups in all clinical outcomes and baseline gait speed and baseline SAIs between groups (Table 1). To test for adaptation, storage of aftereffects, and retention within the ACLR group and between groups, we used a 2×5 analysis of variance (ANOVA) model with a between-subjects factor of group (ACLR, control) and a within-subjects repeated measures factor of epoch (walking period). Separate main effects of group and epoch would be examined barring a significant interaction effect. Post-hoc Bonferroni corrected pairwise comparisons were used when the ANOVA was significant. To test transfer, an independent t-test was used to compare transfer indexes between groups.

3. Results

Individuals who underwent ACLR were an average of 6.15 ± 1.81 months after surgery. Twelve individuals after ACLR had bone-patellar-tendon-bone autografts, two had hamstring autografts, and one had a quadriceps tendon autograft. Six of the 15 had a concomitant meniscal repair at the time of ACLR. Fourteen suffered non-contact injuries, and one had a contact injury. Two ruptured their ACL playing volleyball, three playing soccer, two playing basketball, two playing football, one playing lacrosse, two skiing, one rock climbing, one doing martial arts, and one cheerleading. Demographics between groups were not statistically different except for BMI, which was not matched (Table 1). There were no differences in gait speed or SAI between groups at baseline.

3.1. Clinical Outcomes

Individuals after ACLR had a lower quadriceps strength on the involved limb (P=0.023; Table 2) and quadriceps strength LSI (P<0.001, Cohen’s d=1.9) compared to the control group but no difference in quadriceps inhibition (P>0.48). Interestingly, there were no significant differences between groups in uninvolved or involved limb proprioception (P>0.50).

3.2. Split-belt Treadmill Walking: SAI Adaptation, Storage and Retention Effects

Figure 2A shows the primary outcome measure, SAI, plotted for all treadmill walking strides (averaged in bins of 3 strides for visualization purposes), averaged over all participants in each group. During the baseline period, SAIs for both groups centered around 0%, indicating near perfect symmetry of step lengths. During adaptation I, both groups were initially driven to walk with a large positive SAI (longer involved /slow belt step length), the expected step length response to the perturbation induced by the split belts. Over the course of many strides, the asymmetry gradually reduced back down to near-baseline levels in both groups. This demonstrates the adaptation, or short-term learning, of a new step symmetry pattern that restores normal gait symmetry despite the treadmill belts continuing to be split. In the washout period, both groups demonstrated the characteristic negative aftereffect. This initial reversal of the SAI to a large negative SAI is a hallmark of sensorimotor adaptation and demonstrates that the CNS has stored the newly learned gait pattern such that when the belts are restored to the tied configuration, the gait pattern is once again incorrect, this time in the opposite direction of asymmetry. Over the course of washout, the learned pattern was gradually “de-adapted”, and the walking pattern returned to near normal levels. Finally, in adaptation II, the initial positive perturbation in SAI was much smaller compared to adaptation I. This reduced perturbation shows that, even though the adaptation was washed out, some aspects of the task were retained such that the same size belt split doesn’t perturb step symmetry as much during a second exposure.

Fig 2.

A. Stride by stride data of step length asymmetry indices (SAI). For visualization purposes only, the data are shown in bins of 3 strides and the length of each treadmill walking period is trimmed to the participant with the fewest strides. Solid lines represent group averaged data for individuals after ACLR (purple) and controls (blue). Shading around the lines represents ± 1SEM. Gray shaded regions indicate when belts are tied. B. Group average SAIs for each epoch of interest. Error bars represent ± 1SEM. Asterisks indicate significant post hoc differences for the main effect of epoch. L BL, late baseline; EA I, early adaptation I; LA I, late adaptation I; EWO, early washout; LWO, late washout; EA II, early adaptation II; LA II, late adaptation II.

Figure 2B shows the group average SAIs at each of the key epochs that were statistically compared. The ANOVA found no interaction of epoch*group (F=0.20, P=0.84, η_p²=0.007), nor any main effect of group (F=0.34, P=0.56, η_p²=0.012), but there was a main effect of epoch (F=150.92, P<0.001, η_p²=0.84). SAIs were significantly greater during early adaptation I compared to late baseline (post hoc, P<0.001); SAIs during late adaptation I were significantly lower than during early adaptation I (P<0.001), indicating participants in both groups were able to adapt their step lengths. SAIs during early washout were significantly greater (in the opposite direction) than late baseline (P<0.001), suggesting storage of the learned pattern. Finally, SAIs during early adaptation II were significantly lower than early adaptation I (P<0.001), and early adaptation II was not statistically different than late baseline (P=0.31), suggesting very strong retention of the adapted gait pattern in both groups.

3.3. Overground Transfer Effects

Both groups transferred the newly learned SAI from the split-belt treadmill to overground walking, although the amount of transfer was relatively low (approximately 20%; see Fig 3). There was no difference in transfer indices between groups (P=0.50).

Fig 3.

Transfer index for individuals after ACLR (purple) and control subjects (blue). Error bars indicate SEM.

4. Discussion

We found that individuals after ACLR were capable of adapting step lengths in response to the split-belt perturbation as we hypothesized. However, the amount of adaptation, retention, and transfer did not differ between groups as we had expected. In short, this work indicates that the ability to learn and store a new gait pattern, assessed using step length symmetry, in individuals after ACLR remains intact.

Our results support that early (within 9 months) after ACLR, individuals can adapt their step length, and is similar to what has been reported in the previous study by Roper et al. in individuals 36± 24 months post ACLR.³⁴ Both Roper et al. and we found no significant differences in step length asymmetry between individuals after ACLR and controls in either adaptation or storage (washout) magnitudes,³³ although Roper et al. found slightly reduced perturbation values in early adaptation and early washout, probably due to minor methodological differences. The ability for individuals after ACLR to adapt their step length during the early to mid-post-operative phase of rehabilitation is encouraging as it indicates gait may be malleable during rehabilitation. While the results are specific to spatiotemporal parameters, they are promising first steps to assessing the adaptability of gait after ACLR. Importantly, step length can be modified in a variety of ways and does not rely solely on the knee joint to achieve symmetry. Future research should explore how the knee joint specific gait mechanics (e.g., knee flexion and extension angles) respond given the disruption in knee joint afferent feedback. Our study also included a re-adaptation period, which showed that both individuals after ACLR and controls show a similar (and very robust) ability to retain the learned gait pattern after a washout period.³²

The ability to transfer a new learned movement to a new environment is a critical feature to rehabilitation, as it allows patients to transfer skills learned in the clinic to their preferred environment. In the test of transfer, results suggested there was partial transfer from treadmill to overground walking in both groups. The numbers seen in both the control group and in individuals after ACLR (ACLR, 22%; control, 16%) were similar to the magnitude of transfer seen in control groups from other studies (i.e., approximately 15–28%^30,1). A higher transfer index in the pathologic population (individuals after ACLR) was similarly seen in stroke survivors compared to controls³⁰, though in the current study the difference was not significant. It is important to note that we assessed transfer after a second exposure to the split-belt treadmill, which may have affected the amount of transfer overground. This is because subsequent exposures after an initial exposure are known to result in smaller size initial perturbations at the beginning of the adaptation and washout periods^18,27,32 (this was seen in our SAI data, comparing early adaptation II to early adaptation I). Although this is an indicator that the newly learned gait pattern has been retained, it also results in smaller aftereffects (whether on the treadmill or overground), as individuals learn to quickly “switch” between two separate walking patterns, one for the split-belt treadmill and one for “tied” treadmill belts and for overground walking. Thus, the level of overground transfer we observed here is likely an under-representation of what would have been exhibited if we had tested transfer after just one split-belt exposure. Further research may be necessary to completely understand the roles of environmental context and motor pattern switching in individuals after ACLR.

Interestingly, we did not observe any differences between groups in knee joint proprioception, measured with active joint repositioning. Chaput et al.⁹ tested active joint position sense in individuals after ACLR and a control group with the same methodology and also showed no difference between groups, with a similar amount of error in each group (3.8 ± 1.63 degrees error after ACLR, 4.3 ± 1.17 degrees error for controls). While some systematic reviews have noted statistically significant differences between individuals after ACLR and controls in proprioception, the differences noted are small, on average 0.35 degrees,³¹ and may not be clinically meaningful differences. In fact, one group found individuals after ACLR were able to reproduce knee angles better on their involved limb compared to their uninvolved limb.²⁶ These data suggest that although there may be proprioceptive changes after ACL rupture, furthered by surgery, they may not fully explain gait changes after injury. It is also important to note that active joint position sense testing does not reflect kinesthesia, or dynamic limb proprioception, and does not reflect activity of the Golgi tendon organs that are specific for load detection, both of which probably play a more significant role in proprioceptive function during gait. Further, it has been suggested that individuals after ACLR may use unique information from uninjured individuals to detect gait symmetries.³³ For these reasons, the fact that none of our participants after ACLR had evidence of significantly impaired active joint position sense does not fully answer the question as to whether knee joint proprioception impairments affect short-term motor learning of a new step length symmetry pattern via sensorimotor adaptation.

There are limitations to consider when interpreting this study. Graft type was not controlled in this study, and individuals with 3 different graft types were included. The active joint position sense as a measure of proprioception was not validated nor was it task specific and may not capture the proprioception required for gait. Gait speed was controlled tightly in the sample; therefore, it was not able to be used as an outcome variable. Finally, overground transfer was assessed after a second exposure to the split-belt treadmill, rather than a single exposure as is typically performed.

5. Conclusion

Individuals 3–9 months after ACLR can learn a new gait symmetry pattern in the short-term using sensorimotor adaptation-based motor learning. Specifically, they show normal adaptation, retention when exposed a second time, and overground transfer of a new step length symmetry pattern acquired during split-belt treadmill walking. Results from this study suggest that despite disruptions in sensory feedback after ACL rupture, individuals may be able to modify their gait mechanics using sensory error-based motor learning approaches early post-operatively.