Gait Mechanics in Women of the ACL-SPORTS Randomized Control Trial: Interlimb Symmetry Improves Over Time Regardless of Treatment Group

Abstract

Women after anterior cruciate ligament (ACL) injury and ACL reconstruction (ACLR) are more likely than men to exhibit asymmetric movement patterns, which are associated with post-traumatic osteoarthritis. We developed the ACL specialized post-operative return-to-sports (ACL-SPORTS) randomized control trial to test the effect of strength, agility, plyometric, and secondary prevention (SAPP) training with and without perturbation training (SAPP+PERT) on gait mechanics in women after ACLR. We hypothesized that movement symmetry would improve over time across both groups but more so among the SAPP+PERT group. Thirty-nine female athletes 3–9 months after primary ACLR were randomized to SAPP or SAPP+PERT training. Biomechanical testing during overground walking occurred before (Pre-training) and after (Post-training) training and one and two years post-operatively. Hip and knee kinematic and kinetic variables were compared using repeated measures ANOVA with Bonferroni corrections for post-hoc comparisons (α=0.05). There was a time by limb interaction effect (p=.028) for peak knee flexion angle (PKFA), the primary outcome which powered the study, characterized by smaller PKFA in the involved compared to uninvolved limbs across treatment groups at Pre-training, Post-training, and one year, but not two years. Similar findings occurred across sagittal plane knee excursions and kinetics and hip extension excursion at midstance. There were no meaningful interactions involving group. Neither SAPP nor SAPP+PERT training improved walking mechanics, which persisted one but not two years after ACLR.

Statement of Clinical Significance:

Asymmetrical movement patterns persisted long after participants achieved symmetrical strength and functional performance, suggesting more time is needed to recover fully after ACLR.

Introduction

Anterior cruciate ligament (ACL) injuries and ACL reconstruction (ACLR) surgeries are occurring at increasing rates, especially among young people who engage in high-level sporting activities.^1–3 Young women continue to sustain ACL injuries at higher rates than young men when participating in the same sport.² Women also are more likely than men to exhibit aberrant and asymmetric lower extremity movement patterns both before and after ACL injury.^4–10 Aberrant and asymmetric movements may not only play a role in primary ACL injury risk,^9,11–13 but also influence the development and progression of post-traumatic osteoarthritis.^14–18 While movement asymmetries may persist years after ACLR,^4,7 little is known about the effect of post-operative rehabilitation programs on movement patterns, particularly in women, after ACLR.

We developed the anterior cruciate ligament specialized post-operative return-to-sports (ACL-SPORTS) randomized control trial to determine the effect of post-operative return-to-sport (RTS) rehabilitation on gait mechanics in both male and female athletes.¹⁹ Specifically, this randomized control trial was designed to test the effect of strength, agility, plyometric, and secondary prevention (SAPP) training with and without perturbation training²⁰ (SAPP+PERT), a specific type of neuromuscular training. The scientific premise for this study was based on evidence that rehabilitation programs incorporating both strength and neuromuscular (perturbation) training have improved movement patterns among individuals after ACL injury,^20–24 and that the effect of post-operative rehabilitation and neuromuscular training on movement patterns was unkown.¹⁹ We recently reported our findings from the men of the ACL-SPORTS trial (as recruitment occurred more quickly among the men).^25,26 Our findings suggest that, among men, there is no difference in response to SAPP versus SAPP+PERT training on walking mechanics; movement asymmetries persisted to some degree at both one and two years after ACLR, but were generally no longer clinically meaningful at two years.^25,26

While the effect of post-operative rehabilitation and RTS training with or without neuromuscular training on the movement patterns of female athletes is unknown, women may respond differently than men for several reasons. Women who are ACL-deficient have shown greater improvements in movement patterns in response to strength and neuromuscular training.^23,27–29 Women are also more likely than men to exhibit aberrant and asymmetric movement patterns after ACL injury;^5–9 women may therefore demonstrate greater baseline movement asymmetry and thus have more room for improvement. Given that women, but not men, walk with greater movement asymmetry 6 months after ACLR compared to pre-operatively,²⁸ differences between men and women may be exacerbated early after ACLR. Finally, our recent findings from the ACL-SPORTS cohort suggest women may take longer than men to regain quadriceps strength after ACLR,³⁰ which could influence their walking mechanics.³¹

We therefore sought to investigate knee and hip kinematics (angles and excursions) and kinetics as well as medial compartment tibiofemoral loading during walking before and after post-operative training as well as 1 and 2 years after ACLR among the women of the ACL-SPORTS randomized control trial. Our a priori hypothesis was that the women who received SAPP+PERT training, compared to SAPP alone, would exhibit greater improvements in movement symmetry (i.e., more symmetrical sagittal plane knee kinematics, the primary outcome on which the study was powered). We also hypothesized that, regardless of treatment group, movement symmetry would improve over time, especially between 1 and 2 years after ACLR, coincident with return to sport.

Methods

Participants

This study is a prospective, randomized control trial (level of evidence: 1), which was registered at clinicaltrials.gov (NCT01773317) and was approved by the University of Delaware Institutional Review Board. The study was performed at the University of Delaware between December 2011 and August 2018. Written informed consent was obtained from all participants; minors provided assent in addition to consent from their parent/guardian.

Female athletes after primary, unilateral ACLR were eligible for enrollment when they achieved impairment resolution, defined as full and symmetric knee range of motion, minimal to no effusion,³² at least 80% quadriceps strength limb symmetry index, initiation of a running progression, and the ability to hop without pain on each leg. All participants were required to meet the following inclusion/exclusion criteria: at least 12 weeks and less than 10 months after primary, unilateral ACLR; previous participant in level I or II sports (i.e., sports involving jumping, cutting, and pivoting)³³ for at least 50 hours per year; no history of previous ACLR or other severe lower extremity injury or surgery; no concomitant grade III knee ligament (e.g., medial collateral ligament) injury; and no osteochondral defect of 1 cm² or larger.

Participants were randomized using a random number generator and stratified by sex so that 20 women each were allocated to the SAPP and SAPP+PERT treatment groups (Figure 1). Randomization was performed by a research administrator (MC). All physical therapist researchers who performed data collections were blinded to group assignment (single-blinded study). Our a priori power analysis indicated that 36 women were needed to detect differences between groups in sagittal plane knee kinematics (primary outcome) based on the established minimal clinically important difference²⁹ (β = 0.20; α = 0.05, medium effect size = 0.30); we enrolled 40 female athletes to allow for 10% attrition.¹⁹ After training, the researchers were informed that one participant may have sustained a graft rupture prior to enrollment, thus she was excluded from this study (and all analyses).

Figure 1.

The flow chart depicts the enrollment process and testing sessions.

Interventions

We encourage readers to consult the ACL-SPORTS protocol paper, published previously by White et al.,¹⁹ as well as a recent publication by Arundale et al.,³⁴ for additional details about the present study. We will present the general aspects of the study here only briefly. After enrollment, all participants first underwent clinical testing and motion analysis testing (described below) prior to receiving any interventions. Subsequently, participants received 10 sessions (~2x/week) of either SAPP or SAPP+PERT training. All participants in both groups received the same general elements of the SAPP training program, consisting of Nordic hamstrings, standing squats progressing to tuck jumps, drop jumps, triple single leg hopping, and agility drills. All participants in both groups with quadriceps strength indexes below 90% also performed quadriceps strengthening exercises in the physical therapy clinic whereas those with quadriceps strength indexes of 90% or greater continued quadriceps strengthening on their own (outside the clinic). Participants in the SAPP only group of the ACL-SPORTS randomized control trial also received a sham intervention during which the athlete stood on one leg on a stable surface and performed hip flexion against a resistance band with the opposite limb.¹⁹ Participants in the SAPP+PERT treatment group did not receive the sham intervention but instead received the SAPP training described above plus 10 sessions of perturbation training (~30 minutes per session).²⁰ Perturbation training is a specific type of neuromuscular training designed to improve neuromuscular activation patterns and facilitate dynamic knee stability.^{20,22–24,29} During perturbation training,^19,20,22,24 the patient stood on an unstable surface (i.e., rollerboard or rockerboard) while the physical therapist applied movements, or perturbations, to the surface. Perturbations began in a blocked manner with small movements and were progressed by increasing the speed, magnitude, direction (e.g., adding diagonals or rotations), and manner in which the perturbations were delivered (e.g., blocked progressing to random; verbal cues progressing to no verbal cues). Sport-specific distractions, such as throwing, catching, or passing a ball, were added during the later sessions.²⁴ While the interventions, principles, and general progression were standardized, physical therapists were allowed clinical judgment to determine the optimal rate of progression for each individual athlete and appropriate selection of sport-specific distractions. Additional details on perturbation training are available in the ACL-SPORTS protocol paper¹⁹ and from Chmielewski et al.²⁴

Biomechanical Testing and Variables of Interest

Biomechanical testing occurred during over-ground walking at four post-operative time points: 1) after impairment resolution and before receiving any interventions (Pre-training); 2) following 10 training sessions (Post-training); 3) one year post-operatively (1 year); and 4) two years post-operatively (2 years). Biomechanical testing included bilateral surface electromyography (EMG) collected at 1080 Hz; electrode placement was immediately preceded by shaving and abrading the skin with alcohol-soaked gauze to improve electrode conductance. Electrodes (MA-300 EMG System, Motion Lab Systems, Baton Rouge, LA) were placed on seven lower extremity muscles crossing the knee joint on each limb: the rectus femoris, medial and lateral vasti, medial and lateral hamstrings, and medial and lateral gastrocnemii. Maximal volitional isometric contractions were performed for each muscle group³⁵ and used for EMG normalization. EMG data were high-pass filtered at 30 Hz using a 2^nd order Butterworth filter, rectified, and low-pass filtered at 6 Hz to create a linear envelope. After placing the EMG electrodes and performing maximal volitional isometric contractions, we placed 39 retroreflective markers on the lower extremities and pelvis as previously described.²⁷

Participants walked across our motion analysis laboratory over an embedded force platform (Bertec Corporation, Columbus, OH) at a self-selected gait speed maintained within ± 5% during and across all testing sessions. Kinematic data were captured at 120 Hz using an eight camera motion analysis system (VICON, Oxford, UK) whereas kinetic data were captured at 1080 Hz. We used commercial software (Visual3D, C-Motion, Germantown, MD) to calculate kinematic and kinetic variables via inverse dynamics. All data were normalized to 100% of stance to allow temporal comparison of data across subjects. Kinetic variables were normalized by mass*height (kg*m) to allow comparisons across participants.³⁶ Biomechanical variables of interest included sagittal plane hip and knee kinematics and sagittal and frontal plane hip and knee kinetics. We compared key biomechanical variables at peak knee flexion angle (PKFA). We also calculated sagittal plane hip and knee excursions during the weight acceptance and midstance phases of gait. Weight acceptance was defined as initial contact to PKFA; midstance was defined as PKFA to peak knee extension angle.²⁵

We also used a previously validated³⁷ subject specific, EMG-driven musculoskeletal modeling approach^5,37,38 to estimate medial compartment tibiofemoral joint contact forces. Described previously in detail,^5,37,38 our model uses a hill-type muscle fiber in series with an elastic tendon. The model employs an iterative, simulated annealing process to allow muscle parameters to vary within physiological norms (± 2 standard deviations) to best fit the knee flexion moment curve derived through these EMG-derived estimations to the knee flexion moment curve derived through inverse dynamics. We performed this process for 5 walking trials per limb for each participant, at Pre-training, Post-training, and 2 years. Next, we predicted each trial (for a given limb, at a given timepoint) using the derived muscle parameters and coefficients from the other four trials. We subsequently selected the three best-fitting (highest R² and lowest root mean square) predicted trials, using the same predicted values, for each limb at each timepoint. We used the Winby frontal plane moment algorithm³⁹ to estimate the medial tibiofemoral joint contact force. Peak medial compartment contact force (PMCCF), constrained to the first 50% of stance and normalized by body weight to allow comparison across participants,³⁶ was our modeling-derived variable of interest given the association between PMCCF and early knee osteoarthritis after ACLR.^14,16

Statistical Analysis

We used independent t-tests and chi-square tests of proportions to compare demographic characteristics between groups. We used repeated measures, mixed model analysis of variance (ANOVA) to compare all biomechanical variables of interest. Post-hoc comparisons with Bonferroni corrections for multiple comparisons were made for significant interactions or main effects of time. Differences were also compared to previously established minimum clinically important difference (MCID) values²⁹ to determine whether or not statistically significantly difference values were clinically meaningful. All P values less than 0.05 were established a priori as statistically significant. Statistical analyses were performed using SPSS version 25.0 (IBM Corporation, Armonk, New York, USA).

Results

There were no differences between treatment groups for demographic characteristics, clinical variables, or gait speed among all participants (Table 1a). There were likewise no differences in demographics, clinical variables, or gait speed among only those 35 participants who completed biomechanical testing at all timepoints (Table 1b) and were used in our primary statistical analyses. There were also no statistically significant differences between the 4 participants who missed one (n=3) or two (n=1) follow-up testing sessions compared to the 35 who completed all biomechanical testing sessions (all p > 0.05). Not every subject had data that could be modeled for reasons such as incomplete or suboptimal EMG or poor model tuning. In addition to the 4 subjects lost to follow-up as described above and depicted in Figure 1, there were 6 subjects at Pre-training, 7 subjects at Post-training, and 6 subjects at 2 years that had data that could not be modeled. There were a total of 23 (12 SAPP, 11 SAPP+PERT) women with complete modeling data at all timepoints of interest (i.e., Pre-training, Post-training, and 2 years). There were no differences in demographics of these 23 participants with complete modeling data and the remaining 16 participants with incomplete modeling data.

Table 1.

Demographic characteristics did not differ between the SAPP and SAPP+PERT groups among all participants (Table 1a) or among only those who completed testing at all timepoints (Table 1b).

Table 1a. Variable	SAPP (n=20)	SAPP+PERT (n=19)	P-value
Age at Surgery (years)	18.9 (5.8)	19.0 (8.8)	0.986
Gait Speed (m/s)	1.5 (.1)	1.5 (.1)	0.410
Height (m)	1.65 (.06)	1.65 (.08)	0.986
Weight (kg)	68.8 (10.9)	67.9 (14.3)	0.82
Body Mass Index (kg/m^2)	25.3 (3.3)	24.7 (3.9)	0.631
Pre-Injury Sport Level	19 Level I, 1 Level II	15 Level I, 4 Level II	0.182
Pre-Injury Competition Level	9 School Sponsored, 7 Club Level, 4 Intramural/Recreation	10 School Sponsored, 7 Club Level, 2 Intramural/Recreation	0.707
Mechanism of Injury	6 Contact, 14 Non-contact	5 Contact, 14 Non-contact	0.798
Graft Type	8 Hamstring, 8 BPTB, 4 Allograft	10 Hamstring, 8 BPTB, 1 Allograft	0.368
Table 1b. Variable	SAPP (n=17)	SAPP+PERT (n=18)	P-value
Age at Surgery (years)	19.2 (6.2)	19.0 (9.1)	0.933
Height (m)	1.65 (.07)	1.65 (.08)	0.945
Weight (kg)	69.6 (11.7)	67.4 (14.5)	0.631
Body Mass Index (kg/m^2)	25.4 (3.5)	24.6 (4.0)	0.539
Gait Speed (m/s)	1.5 (.1)	1.5 (.1)	0.310
Pre-Injury Sport Level	16 Level I, 1 Level II	14 Level I, 4 Level II	0.338
Pre-Injury Competition Level	8 School Sponsored, 6 Club Level, 3 Intramural/Recreation	9 School Sponsored, 7 Club Level, 2 Intramural/Recreation	0.858
Mechanism of Injury	5 Contact, 12 Non-contact	4 Contact, 14 Non-contact	0.711
Graft Type	5 Hamstring, 8 BPTB, 4 Allograft	9 Hamstring, 8 BPTB, 1 Allograft	0.233

(Abbreviations: SAPP = strength, agility, plyometric, and secondary prevention; SAPP+PERT = SAPP plus perturbation training; BPTB = bone-patellar tendon-bone autograft.)

Results for Hypothesis 1: SAPP vs. SAPP+PERT Training Effect

There were no interaction or main effects involving treatment group for the primary outcome variable, peak knee flexion angle (PKFA, Figure 2). There were no significant 3-way interactions and no group by limb interaction effects for any variable of interest; there were, however, two significant group by time interaction effects and two main effects of group (Supplemental Table). There was a group by time interaction effect for hip extension excursion during weight acceptance (Figure 3), however the only statistically significant post-hoc comparison was between Post-training and 2 years in the SAPP+PERT group and this difference of 1.4° was well below the MCID of 3.0°²⁹. There were no statistically different (p ≥ .311) or clinically meaningful differences between groups at any timepoint for hip extension excursion during weight acceptance (group differences were all less than 1.1°). There was also a significant group by time interaction effect for hip extension moment at PKFA (Figure 4). Within the SAPP group only, there were smaller hip extension moments (collapsed across limb) at 2 years compared to any of the earlier timepoints (post-hoc Bonferroni p < .05), but no differences among any of the earlier timepoints (i.e., Pre-training, Post-training, and 1 year). There were no differences over time in the SAPP+PERT group and no differences between groups at any timepoint. While there were no interaction effects for hip abduction moment at PKFA, there were main effects of group and limb (Figure 5), characterized by larger values in the SAPP+PERT versus SAPP group as well as in the involved compared to uninvolved limb, regardless of time. There was also a main effect of group for internal knee extensor moment at PKFA, characterized by larger moments in the SAPP+PERT versus SAPP group collapsed across limb and time (Figure 6); the interaction effect of time by limb for internal knee extensor moment is described below.

Figure 2. Peak Knee Flexion Angle.

There were no interaction or main effects of group for peak knee flexion angle (PKFA), the primary study outcome. There was an interaction effect of time by limb for PKFA as well as a main effect of limb for PKFA. Clinically meaningful differences were present across both groups at both Pre-training and Post-training, but not at one or two years after ACLR (bar = exceeds MCID). (Note: due to rounding, the interlimb difference in the SAPP+PERT group is 3.1° at pre-training, which exceeds the MCID of 3.0°²⁹, whereas the interlimb difference at one year is only 2.9°.)

Figure 3. Hip Extension Excursion During Weight Acceptance.

There was a significant group by limb interaction effect of hip extension excursion during weight acceptance; however, no differences were clinically meaningful. (Brackets indicate the statistically significant difference between the Post-training and 2 year timepoint within the SAPP+PERT group, collapsed across limbs.)

Figure 4. Hip Internal Extension Moment at PKFA.

There was a significant time by group interaction effect (p = .022) for hip internal extension moment at peak knee flexion angle (PKFA), characterized by smaller values at 2 years compared to any earlier timepoints in the SAPP group only (bracket denotes difference).

Figure 5. Hip Internal Abduction Moment at PKFA.

While there were no interaction effects for hip internal abduction moment at peak knee flexion angle (PKFA), there were main effects of group and limb, characterized by larger values in the involved compared to uninvolved limb as well as in the SAPP+PERT versus SAPP group (indicated by large bracket), regardless of time.

Figure 6. Knee Internal Extension Moment at PKFA.

There was an interaction effect of time by limb for knee internal extension moment at peak knee flexion angle (PKFA), characterized by large early interlimb asymmetry that resolved across groups by two years after ACLR. Interlimb differences exceeded MCID values across both groups at Pre- and Post-training, but only in the SAPP+PERT group at one year (bar = exceeds MCID²⁹). There was also a main effect of group with higher knee extension moments across time and limb in the SAPP+PERT group compared to the SAPP group (indicated by the large bracket).

Results for Hypothesis 2: Effect of Time

There were time by limb interaction effects characterized by early interlimb asymmetry that resolved by 2 years for six variables of interest: peak knee flexion angle, knee flexion excursion during weight acceptance, knee extension excursion during midstance, hip extension excursion during midstance, knee internal extension moment at PKFA, and peak medial compartment contact force (Supplemental Table). For peak knee flexion angle (Figure 2), the primary study outcome, post-hoc comparisons with Bonferroni adjustment revealed that there were statistically significant differences between limbs (collapsing across groups) at Pre-training, Post-training, and 1 year (p ≤ 0.002), but not at 2 years (p = 0.161). Regardless of group, participants walked with clinically meaningfully²⁹ smaller knee flexion angles in the involved compared to uninvolved limb at Pre- and Post-training, but not at 1 or 2 years after ACLR. There were supportive findings for sagittal plane knee moments (Figure 6) and excursions during weight acceptance (Figure 7) and midstance (Figure 8). Likewise, there was an interaction of time by limb for hip extension excursion during midstance (Figure 9); post-hoc comparisons for hip extension excursion during midstance revealed statistically significant and clinically meaningful differences between limbs at Pre-training and Post-training (p < .001) but not at 1 year (p = .097) or 2 years (p = .304). There was a time by limb interaction effect for peak medial compartment contact force (Figure 10); while only the difference between limbs at Post-training was statistically significant (p = .026), the minimal detectable change of 0.30 body weight (BW)⁴⁰ and meaningful inter-limb difference of 0.4 BW⁴¹ were not exceeded at Post-training (interlimb difference: - 0.27) or any other timepoint.

Figure 7. Knee Flexion Excursion During Weight Acceptance.

There was an interaction effect of time by limb for knee flexion excursion during weight acceptance, characterized by early interlimb asymmetry that resolved by one year after ACLR. Only the interlimb differences at Pre-training for the SAPP group and at Post-training for the SAPP+PERT group were clinically important (bar = exceeds MCID²⁹).

Figure 8. Knee Extension Excursion During Midstance.

There was an interaction effect of time by limb for knee extension excursion during midstance, characterized by early interlimb asymmetry that resolved across groups by two years after ACLR. Interlimb differences exceeded MCID values across both groups at Pre- and Post-training, but only in the SAPP+PERT group at one year (bar = exceeds MCID²⁹).

Figure 9. Hip Extension Excursion During Midstance.

There was an interaction effect of time by limb for hip extension excursion during midstance; clinically meaningful differences between limbs occurred across groups at both Pre-training and Post-training, but not at 1 or 2 years (bar = exceeds MCID²⁹).

Figure 10. Peak Medial Compartment Contact Force.

There was a time by limb interaction for peak medial compartment contact force. While the difference between limbs at Post-training was statistically significant (p = .026), the difference (− 0.27) did not exceed the meaningful inter-limb difference threshold of 0.4 body weight (BW)⁴¹ at Post-training or any other timepoint.

Discussion

The purpose of this study was to compare knee and hip kinematics and kinetics as well as medial compartment tibiofemoral loading during walking before and after post-operative return-to-sport training with and without perturbation as well as 1 and 2 years after ACLR among the women of the ACL-SPORTS randomized control trial. Our a priori hypothesis, that the women who received SAPP+PERT training, compared to SAPP alone, would exhibit greater improvements in movement symmetry, was not supported. Our second hypothesis, that movement symmetry would improve over time regardless of treatment group, was largely supported, particularly among knee sagittal plane kinematics and kinetics. Participants walked with generally symmetric gait patterns by 2 years after ACLR, but not at earlier timepoints. Our findings suggest that meaningful movement asymmetries persist during gait even after participants achieve symmetrical strength, have high functional performance, and return to sports,³⁰ supporting the notion that more time is needed to recover after ACLR^42,43 and/or task-specific training is needed to restore movement symmetry.^44,45

Mounting evidence suggests recovery from ACLR does not take four to six months, as once thought, but may require one to two years or longer.^42,43 Our findings support and supplement the belief that recovery takes longer, even after participants meet stringent objective criteria.^19,46 The participants in the present study had resolved all impairments prior to enrollment and had a mean of 91% quadriceps strength index before training,³⁰ yet still walked with clinically meaningful²⁹ interlimb sagittal knee asymmetry before and after training. Movement asymmetries were not resolved globally until 2 years after ACLR, long after most participants had achieved symmetrical strength and functional performance and returned to sports.³⁰ These findings indicate that time and the physical challenge accompanying return to sports may be important considerations for restoring movement asymmetry, although targeted interventions may expedite this process.^44,45 Future research studies examining real-time visual, auditory, and/or tactile feedback to correct aberrant movement patterns are warranted.

A few recent studies have linked early post-traumatic osteoarthritis after ACLR to aberrant walking mechanics,^14,16,41,47 notably underloading of the involved limb⁴⁷ and especially the medial compartment of the tibiofemoral joint.^14,16 In the present study, there was a shift toward medial tibiofemoral compartment underloading of the involved versus uninvolved limb at Post-training (7 ± 2 months after ACLR) that was not present at Pre-training (6 ± 2 months after ACLR) or 2 years after ACLR. The shift in walking mechanics was accomplished through both decreased peak medial compartment contact force (PMCCF) in the involved limb and increased PMCCF in the uninvolved limb at Post-training relative to both Pre-training and 2 years. Similar findings of medial compartment underloading at Post-training also occurred among the men of the ACL-SPORTS trial²⁶ and correspond to a potentially important change in activity patterns that occur as athletes begin to return to their sports. The early phases of rehabilitation are focused almost exclusively on the involved, surgical limb,⁴⁶ but during the 10 post-operative return-to-sport training sessions, participants begin doing higher level activities on both limbs for the first time after surgery. During this return-to-sport training, participants may develop movement strategies that rely more heavily on the uninvolved limb,^14,26,48,49 favoring the involved limb due to lingering deficits or lack of confidence.⁵⁰

To the study’s merit, we executed a randomized control trial with blinded assessors, minimal loss to follow-up, and comprehensive and rigorous testing at multiple timepoints using a repeated measures design. We included patients with ACLR surgeries using different graft types and performed by different surgeons, which makes our sample more heterogeneous and reflective of a clinical population but also possibly introduced variability. We did not control rehabilitation prior to enrollment in the study, but all participants met the same, rigorous enrollment criteria.^19,46 Limitations include lack of pre-injury and pre-operative data. We included all female participants with complete biomechanical data in our analyses, including those women who sustained a second ACL injury. Secondary analyses excluding those with second injuries resulted in similar findings as in our primary, planned analyses. The interventions were not task-specific, thus may not be optimally designed to change movement patterns during gait,^44,45 but had been shown to improve gait and neuromuscular control strategies when delivered pre-operatively.^22–24 While the present study included only women and thus the findings may not apply to men given biomechanical differences according to sex,^4–9 we previously reported similar findings in the men of the ACL-SPORTS trial,^25,26 indicating that neither SAPP nor SAPP+PERT training improved gait mechanics in men or women after ACLR.

In conclusion, our findings suggest that strength, agility, plyometric, and secondary prevention (SAPP) training with and without perturbation training do not meaningfully improve walking mechanics among young female athletes. Asymmetrical gait mechanics persist to a large degree until two years after ACLR, long after patients have achieved symmetrical strength and functional performance³⁰ and have returned to sports. Our findings contribute to growing evidence suggesting full recovery after ACLR may take longer than previously thought, up to two years after ACLR.

Supplementary Material

Supplemental Table

Supplemental Table. This table presents all p values for main effects, two-way interaction effects, and 3-way interaction effects for all biomechanical variables of interest. Values less than p = 0.05 are statistically significant, as denoted by an asterisk (*). (Abbreviation: PKFA = peak knee flexion angle.)