Acutely Normalizing Walking Speed Does Not Normalize Gait Biomechanics Post-Anterior Cruciate Ligament Reconstruction

Abstract

Aberrant gait biomechanics in individuals with anterior cruciate ligament reconstruction (ACLR) are linked to posttraumatic osteoarthritis (PTOA) development, indicating a need to normalize gait biomechanics to prevent PTOA. ACLR individuals walk slower than uninjured controls and slower speeds are associated with aberrant gait biomechanics. Yet, it is unclear if increasing walking speed normalizes gait biomechanics post-ACLR.

Purpose:

To determine the effect of acutely increasing walking speed on gait biomechanics in ACLR individuals compared to their habitual speed and uninjured matched-controls.

Methods:

Gait biomechanics were collected on 30 ACLR individuals (20 females, age: 22.0±4.2 years, BMI: 24.0±3.0 kg·m⁻²) at their habitual speed and at 1.3 m·s⁻¹, a speed similar to controls, and 30 uninjured matched-controls (age: 21.9±3.8, BMI: 23.6±2.5) at their habitual speed. Functional waveform analyses compared biomechanics between: i) walking at habitual speed vs 1.3 m·s⁻¹ in ACLR individuals; and ii) ACLR individuals at 1.3 m·s⁻¹ vs controls.

Results:

In the ACLR group, there were no statistically significant biomechanical differences between walking at habitual speed (1.18±0.12 m·s⁻¹) and 1.3 m·s⁻¹ (1.29±0.05 m·s⁻¹). Compared with controls (habitual speed: 1.34±0.12 m·s⁻¹), the ACLR group while walking at 1.3 m·s⁻¹ exhibited smaller vertical ground reaction force (vGRF) during early and late stance (13-28, 78-90% stance phase), greater midstance vGRF (47-61%), smaller early-to-midstance knee flexion angle (KFA; 1-44%), greater mid-to-late stance KFA (68-73, 96-101%), greater internal knee abduction moment (69-101%), and smaller internal knee extension moment (4-51, 88-96%).

Conclusions:

Increasing walking speed to a speed similar to uninjured controls did not elicit significant changes to gait biomechanics, and ACLR individuals continued to demonstrate biomechanical profiles that are associated with PTOA development and differ from controls.

INTRODUCTION

Following anterior cruciate ligament reconstruction (ACLR), nearly one third of individuals develop radiographic posttraumatic osteoarthritis (PTOA) within the first decade and approximately half develop PTOA by the second decade (1, 2). It is hypothesized that aberrant gait biomechanics hasten the development of PTOA in individuals with an ACLR (3-5). Aberrant gait biomechanical patterns demonstrated by individuals following ACLR are associated with greater concentrations of biomarkers related to deleterious changes in joint tissue metabolism (3, 6-8) and articular cartilage composition (4, 9-12) which are linked to PTOA development. Specifically, compared to uninjured controls, ACLR patients walk slower and demonstrate less dynamic loading of the ACLR limb, as characterized by smaller first and second peak vertical ground reaction forces (vGRF) and greater vGRF during midstance (4, 13, 14). ACLR also individuals exhibit a stiffened-knee gait pattern, characterized by smaller peak knee flexion angle (KFA) in the first quarter of stance and lesser knee extension during midstance compared to uninjured controls (13). Lastly, ACLR individuals exhibit lesser peak sagittal knee moments compared to uninjured controls (13).

There is a critical need to develop targeted interventions that normalize aberrant gait biomechanics in ACLR patients to reduce the risk and progression of PTOA by optimizing the loads exerted on knee joint tissues. It is hypothesized that modifying aberrant gait biomechanics using advanced rehabilitation techniques early following ACLR may be beneficial for PTOA prevention (15-17). Recent work (15, 16, 18) has demonstrated that cueing ACLR individuals to modify the first vGRF peak using real-time gait biofeedback evokes changes in vGRF throughout stance phase, and simultaneously alters KFA and knee moments (15, 16, 18). Unfortunately, advancing real-time gait biofeedback as a breakthrough therapy must overcome significant technological barriers related to accurately modifying gait biomechanics (15, 16, 18). Thereby, developing methods capable of normalizing aberrant gait biomechanics that do not require laboratory-grade force measuring devices and sophisticated biofeedback technology would be beneficial for advancing the overall goal of mitigating PTOA development following ACLR. The inability to precisely normalize gait biomechanics in a way that promotes joint tissue health remains a critical barrier to reducing the risk of developing PTOA following ACLR.

Walking speed is an easily modifiable gait parameter and thus a potentially clinically feasible target to improve gait biomechanics. Previous work has demonstrated that ACLR individuals walk slower than their matched uninjured counterparts in the first 12 months post-ACLR (13) and slower walking speeds are associated with smaller peak vGRF and greater knee moments (13, 19, 20). Furthermore, slower habitual walking speed is associated with idiopathic osteoarthritis development (21, 22), and with deleterious changes to joint tissue metabolism (3, 6, 7) and cartilage composition (4, 9, 10) following ACLR. Increasing walking speed is known to elicit increased 1^st and 2^nd peak vGRF, decreased midstance vGRF, and greater knee flexion excursion in healthy individuals (19, 23-25). Previous literature also suggests that cueing increases in walking speed and step cadence subsequently improves 1^st peak vGRF, knee flexion excursion, and sagittal plane knee moments in individuals 6-12 months post-ACLR (26, 27). There is a link between aberrant gait biomechanics and slower habitual walking speeds following ACLR (3, 10, 13, 14) and cued increases in walking speed influence gait biomechanics (i.e., increased 1^st peak vGRF, knee flexion excursion, and sagittal knee moments) (26, 27). However, it remains unknown if acutely increasing walking speed in individuals with ACLR to match the speed of uninjured controls (13) will normalize gait biomechanics compared to habitual speed and uninjured controls. The capability to normalize gait biomechanics by modifying walking speed would be significant as walking speed is a clinically feasible measure, and cueing changes in walking speed is an accessible approach that is easily understood by patients.

The primary purpose of the current study was to leverage an observational cohort study to compare gait biomechanics between i) ACLR individuals walking at their habitual speed and walking at a speed similar to the habitual speed of uninjured controls (1.3 m·s⁻¹) (13, 28), and ii) ACLR individuals walking at 1.3 m·s⁻¹ compared with matched, uninjured controls walking at their habitual speed. Secondarily, as habitual walking speed variability exists within a cohort, we conducted the same analyses described above for ACLR individuals stratified into slow self-selected speed group as to determine the effect of an increase in speed in a subgroup that would incur the most meaningful increase in speed when cued to achieve the 1.3 m·s⁻¹ speed. The biomechanical variables of interest were vGRF, KFA, and internal knee abduction (KAM) and extension (KEM) moments. We hypothesized that acutely cueing an increase in walking speed to 1.3 m·s⁻¹ in ACLR individuals would result in more dynamic and normalized vGRF, KFA, KAM, and KEM waveforms, similar to uninjured matched controls. Further, we hypothesized the magnitudes of change in gait biomechanics would be greatest in the slow self-selected speed ACLR group.

METHODS

Study Design

We conducted a case-control analysis comparing gait biomechanics of ACLR individuals walking at an evidence-based predetermined speed of 1.3 m·s⁻¹ (13, 28) to: i) the same ACLR individuals walking at their habitual speed; and ii) to gait biomechanics of a matched-control group. The predetermined speed of 1.3 m·s⁻¹ was the habitual walking speed of a similarly matched (i.e., age, sex, and BMI) uninjured control group in a previous study which assessed gait biomechanics between controls and ACLR individuals 6-12 months post-ACLR (13). Additionally, the same predetermined speed was utilized as a normalized control speed in a study assessing gait biomechanics and cartilage composition in individuals 6-18 months post-ACLR, in which participants walked at the predetermined speed until the participants reported feeling comfortable walking at that speed (28). All ACLR individuals in our study were part of a prospective longitudinal cohort study that enrolled ACL injured patients prior to ACLR. ACLR gait biomechanics were assessed during a single session first while walking at their habitual speed (±5%) followed by walking at 1.3 m·s⁻¹ (±5%). The uninjured controls were matched to the ACLR group based on age (±2 years), sex (male, female), and BMI (±3 kg·m⁻²) (29). Gait biomechanics were collected at each control participant’s habitual walking speed during a single session, and dominant limb biomechanics of each control participant were analyzed as part of the current study (13, 29). Limb dominance was defined as the preferred leg to kick a soccer ball (30). All methods and protocols in the present study were approved by the Institutional Review Board at the University of North Carolina at Chapel Hill, and all participants provided written informed consent prior to participation in the study.

Participants

ACLR Participants

Thirty individuals between 18 and 35 years of age with a history of primary unilateral ACLR participated in the study. Participants were recruited from orthopaedic clinics in the university health system, and all underwent a bone-patellar tendon-bone autograft ACLR by one of three participating orthopaedic surgeons. Individuals were excluded if they reported any of the following: ACLR revision surgery; multi-ligament reconstruction at the time of ACLR; removal of >1/3 of their meniscus at the time of ACLR (31-33); contralateral ACLR or injury; history of any other lower extremity orthopaedic surgery; physician diagnosed knee osteoarthritis; history of neuromuscular disorders, balance disorders, or inflammatory arthritis; history of other orthopaedic/musculoskeletal injury within 6 months prior to enrollment and testing; or were pregnant (4, 5, 13, 14). All participants underwent formalized physical rehabilitation following ACLR, yet we did not standardize therapy for this group.

Uninjured Control Participants

Thirty uninjured control participants were age- (±2 years), BMI- (±3 kg·m⁻², and within same BMI category), and sex- (male, female) matched to an ACLR participant (29). Control participants were excluded if they were pregnant, diagnosed with inflammatory arthritis, had a history of knee injury or any other lower extremity joint injury (i.e., chronic ankle instability, hip labrum injury), or a history of any lower extremity orthopaedic surgery (13, 29).

Procedures and Measurements

Walking Gait Biomechanics

Each participant was fitted with 26 retroreflective markers placed on the upper and lower extremities, along with a rigid cluster of 3 markers placed on the sacrum, as previously described (4, 5, 13, 14). Following marker placement, overground habitual walking speed was determined from the average speed of 5 walking trials utilizing 2 sets of timing gates (Dashr Timing Systems; Lincoln, NE). Participants were instructed to walk normally at their habitual walking speed, look straight ahead, and maintain a constant speed through the timing gates. After the determination of habitual walking speed, all participants then completed 5 successful data collection trials within 5% of their habitual walking speed (4, 5, 13, 14). Following the habitual walking speed data collection trials, the ACLR group then completed 5 data collection trials walking at 1.3 m·s⁻¹ (±5%). A single trial was deemed successful if: both right and left feet made contact with a single force plate for the entirety of stance phase, a forward gaze was maintained, a gait speed was maintained within ±5% of the designated speed (i.e., habitual speed or 1.3 m·s⁻¹), and gait was not visibly altered or abnormal during the trial (4, 5, 13, 14).

All gait data were collected while participants walked barefoot across an overground walkway. The walkway included two embedded and staggered force plates (Bertec, Corp.; Columbus, OH) for kinetics; a 10-camera, 3-dimensional motion capture system (Vicon Motion Systems; Oxford, UK) for kinematics; and 2 sets of timing gates to ensure each walking trial was within 5% of the designated speed. Kinetic and kinematic data were sampled at 1200 and 120 Hz, respectively, and filtered at 10 Hz using a recursive 4^th order low-pass Butterworth filter (4, 5, 13, 14). All gait biomechanics outcomes (vGRF, KFA, KAM, and KEM) were derived during the stance phase, defined as the interval between heel strike (vGRF > 20 N) and toe-off (vGRF < 20 N) and was time-normalized (1-101%) utilizing custom software (MATLAB, MathWorks, Natick, MA; Visual 3D, C-Motion, Germantown, MD). A static trial for each participant was collected and used to create the segment linkage model and calculate body weight (BW) via force plate measurements. vGRF was normalized to BW; KFA was defined as the angle of the shank relative to the thigh segment using Euler/Cardan angles and positive sagittal plane angles were defined as flexion; and KAM and KEM were calculated using traditional inverse dynamics and normalized to the product of BW and body height. Hip joint centers were calculated via the Bell and Brand hip joint CODA coordinate system (34); knee and ankle joint centers were defined and calculated as the radius of half the distance between the medial and lateral epicondyles and malleoli, respectively. Timing gates were utilized during each testing session to determine habitual walking speed and to ensure that each gait trial was within 5% of habitual speed. For analysis, walking speeds were calculated using an inhouse custom MATLAB program that utilized the trajectory of the sacral plate markers across the 1m walkway directly over the embedded force plates (4, 5, 13, 14). Bilateral biomechanics of the ACLR participants and the dominant limb of the control participants were included in the present analysis.

Statistical Analysis

Primary Analysis

ACLR individuals exhibit aberrant gait biomechanics (vGRF, KFA, KAM, and KEM) throughout multiple portions of stance compared to uninjured controls (i.e., early, mid, and late stance) (13, 14). Therefore, we conducted separate functional waveform analyses to evaluate differences across the stance phase between: i) the ACLR limb at habitual speed vs ACLR limb at 1.3 m·s⁻¹, and ii) the ACLR limb at 1.3 m·s⁻¹ vs limb of uninjured controls at their habitual speed. The functional waveform analyses were performed utilizing the bayesFDA (version 0.3.0) package in R statistical software (version 3.4.3). The average of the 5 walking trials were calculated and utilized for the waveform analyses in which each averaged waveform was assumed to be an independent observation. Bayesian P-splines were fit to estimate the average gait waveforms for each group which were utilized to compute difference curves and corresponding 95% confidence intervals (CI) (5, 13, 14, 29, 35). Comparisons between conditions and groups were considered significantly different at each % of stance phase if the mean difference and corresponding 95% CI did not span zero for greater than 3% of stance phase (35). The largest difference between waveform ensembles and corresponding Cohen’s d effect sizes for each significant comparison are also reported in the present analysis.

Secondary Analysis

To determine if individuals with slower habitual walking speeds demonstrate greater gait biomechanical changes while walking at 1.3 m·s⁻¹, the ACLR cohort was stratified into slow self-selected speed and fast self-selected speed groups based on the calculated minimal detectable change (MDD) (36). The slow speed group comprised individuals who exhibited average habitual walking speeds that were less than one MDD from 1.3 m·s⁻¹. Next, the same functional waveform analyses described above (5, 13, 14, 29, 35) were completed to detect differences during stance phase between each subgroup’s i) ACLR limb at habitual speed (n = 15 per subgroup) vs ACLR limb at 1.3 m·s⁻¹ (n = 15 per subgroup), and ii) ACLR limb at 1.3 m·s⁻¹ (n = 15 per subgroup) vs limb of uninjured controls at their habitual speed (n = 15 per subgroup).

Supplementary Analysis

Previous literature has also found that the contralateral, uninvolved limb in individuals with unilateral primary ACLR demonstrates aberrant gait biomechanics compared to matched uninjured controls (13). Therefore, to further contextualize the findings in the present study, we conducted a supplementary analysis of the contralateral, uninvolved limb in the ACLR group. We utilized the same approach as proposed for the primary and secondary analyses above between walking speed conditions (i.e., habitual speed and 1.3 m·s⁻¹) and compared with the uninjured controls for the contralateral, uninvolved limb (see Supplemental Table 1 and Supplemental Figures 1-6, Supplemental Digital Content).

RESULTS

Demographics for the ACLR and uninjured control groups are reported in Table 1; peak differences and corresponding effect sizes for each comparison are reported in Table 2.

Table 1.

Participant demographics and outcome measures (Mean ± SD, or n (%), as appropriate)

	ACLR Group (n=30)	ACLR Slow Self-Selected Speed (n=15)	ACLR Fast Self-Selected Speed (n=15)	Controls (n=30)
Sex, n (%)
Female	20 (67)	9 (60)	11 (73)	20 (67)
Male	10 (33)	6 (40)	4 (17)	10 (33)
Age, years	22.0±4.2	22.7±4.8	21.2±3.6	21.9±3.8
BMI, kg·m−2	24.0±3.0	24.2±2.4	23.7±3.6	23.6±2.5
Months Since ACLR	6.2±0.4	6.2±0.3	6.2±0.5	-
Meniscal Injury, n (%)	21 (70)	11 (73)	10 (67)	-
Habitual Speed, m·s−1	1.18±0.12§	1.14±0.08§†¥	1.31±0.04	1.29±0.05
Cued Speed, m·s−1	1.30±0.04	1.31±0.04	1.29±0.04	-

ACLR, anterior cruciate ligament reconstruction; BMI, body mass index; kg·m⁻², kilogram of body weight per square meter height; m·s⁻¹, meters per second

^§Indicates statistically significant difference (p<0.01) compared to controls

^†Indicates statistically significant difference (p<0.01) compared to fast self-selected speed group

^¥Indicates statistically significant difference (p<0.01) compared to 1.3 m·s⁻¹

Table 2.

Stance phase percentiles, peak differences, and effect sizes of biomechanical outcome comparisons for ACLR habitual speed vs ACLR 1.3 m·s⁻¹, and ACLR 1.3 m·s⁻¹ vs controls.

Comparison	Variable	Start (% Stance)	End (% Stance)	Maximum Location (% Stance)	Peak Difference	Effect Size
ACLR Habitual Speed vs 1.3 m·s−1	vGRF (BW)	-	-	-	-	-
	KFA (deg)	-	-	-	-	-
	KAM (BW*h)	-	-	-	-	-
	KEM (BW*h)	-	-	-	-	-
ACLR 1.3 m·s−1 vs Controls	vGRF (BW)	13	28	21	+7.11	0.82
		47	61	54	−7.01	−0.96
		78	90	82	+5.65	0.83
	KFA (deg)	1	44	16	−5.66	−1.00
		69	72	69	+2.14	0.51
		96	101	101	−2.21	−0.47
	KAM (BW*h)	70	100	75	−0.004	−0.57
	KEM (BW*h)	4	51	20	+0.022	1.81
		88	96	91	+0.004	0.89
ACLR Slow Self-Selected Speed vs 1.3 m·s−1	vGRF (BW)	10	13	13	+6.30	0.84
	KFA (deg)	−	−	−	−	−
	KAM (BW*h)	−	−	−	−	−
	KEM (BW*h)	−	−	−	−	−
ACLR Slow Self-Selected Speed 1.3 m·s−1 vs Controls	vGRF (BW)	18	22	20	+7.17	0.80
	KFA (deg)	1	35	17	+4.40	0.84
		92	101	101	+3.30	0.78
	KAM (BW*h)	−	−	−	−	−
	KEM (BW*h)	5	46	21	−0.021	−1.86
		87	94	90	−0.005	−0.99
ACLR Fast Self-Selected Speed vs 1.3 m·s−1	vGRF (BW)	−	−	−	−	−
	KFA (deg)	−	−	−	−	−
	KAM (BW*h)	−	−	−	−	−
	KEM (BW*h)	−	−	−	−	−
ACLR Fast Self-Selected Speed 1.3 m·s−1 vs Controls	vGRF (BW)	18	24	23	+7.27	0.86
		54	56	55	−7.88	−1.00
	KFA (deg)	1	42	16	+6.93	1.16
		72	79	74	−2.99	−0.76
	KAM (BW*h)	87	97	91	+0.004	1.00
	KEM (BW*h)	5	45	20	−0.023	−1.70

ACLR, anterior cruciate ligament reconstruction; BW, body weight; h, body height; deg, degrees; vGRF, vertical ground reaction force; KFA, knee flexion angle; KAM, internal knee abduction moment; KEM, internal knee extension moment

Primary Analysis

The ACLR group exhibited an average habitual walking speed of 1.18±0.12 m·s⁻¹ and an average speed of 1.29±0.05 m·s⁻¹ while targeting the cued speed of 1.3 m·s⁻¹. The control group exhibited an average habitual walking speed of 1.34±0.12 m·s⁻¹.

Vertical Ground Reaction Force

The ACLR limb demonstrated no statistically significant differences when walking at 1.3 m·s⁻¹ compared to habitual speed (Fig. 1A, 1B). Yet, the ACLR limb at 1.3 m·s⁻¹ demonstrated significantly smaller vGRF in early (13-28% stance phase) and late (78-90%) stance, and greater midstance vGRF (47-61%) compared with uninjured controls (Fig. 1A, 1C).

Figure 1.

Vertical ground reaction force (vGRF) waveforms normalized to body weight (BW) for the ACLR limb at habitual speed and 1.3 m·s⁻¹, and uninjured controls (1A); and knee flexion angle (KFA; degrees) waveforms for the ACLR limb at habitual speed and 1.3 m·s⁻¹, and uninjured controls (1D). Main effects of speed in individuals with ACLR are shown for the functional analysis of variance (vGRF, 1B; KFA, 1E). Pairwise comparison effects between the ACLR limb at 1.3 m·s⁻¹ and uninjured controls are shown for the functional analysis of variance (vGRF, 1C; KFA, 1F).

Knee Flexion Angle

The ACLR limb demonstrated no statistically significant differences between walking speed conditions (Fig. 1D, 1E). While walking at 1.3 m·s⁻¹, the ACLR limb exhibited smaller KFA during early and midstance (1-44%) and greater KFA in late stance (68-73, 96-101%) compared with uninjured controls (Fig. 1D, 1F).

Internal Knee Abduction Moment and Knee Extension Moment

Compared to habitual walking speed, the ACLR limb demonstrated no statistically significant differences in KAM and KEM while walking at 1.3 m·s⁻¹ (Fig. 2A, 2B, 2D, 2E). Yet, compared with uninjured controls, the ACLR limb at 1.3 m·s⁻¹ exhibited significantly greater KAM during mid-to-late stance (69-100%; Fig. 2A, 2C) and smaller KEM during early, mid-, and late stance (4-51, 88-96%; Fig. 2D, 2F).

Figure 2.

Internal knee abduction (KAM) and extension (KEM) moments normalized to the product of body weight times height (BW x Height) waveforms for the ACLR limb at habitual speed and 1.3 m·s⁻¹, and uninjured controls (KAM, 2A; KEM 2D). Main effects of speed in individuals with ACLR are shown for the functional analysis of variance (KAM, 2B; KEM, 2E). Pairwise comparison effects between the ACLR limb at 1.3 m·s⁻¹ and uninjured controls are shown for the functional analysis of variance (KAM, 2C; KEM, 2F).

Secondary Analysis

The MDD for habitual walking speed was 0.06 m·s⁻¹ in the ACLR group. Therefore, ACLR individuals who walked ≤1.24 m·s⁻¹ (i.e., 1.30 m·s⁻¹ - 0.06 m·s⁻¹ = 1.24 m·s⁻¹) were allocated to the slow self-selected speed group (n=15) and individuals who walked >1.24 m·s⁻¹ were allocated to the fast self-selected speed group (n=15). This ensured that the slow self-selected speed group was cued to walk at an increased speed of >1 MDD from their habitual speed. The average habitual walking speeds for the slow self-selected speed group was 1.14±0.08 m·s⁻¹, and 1.31±0.04 m·s⁻¹ for the fast self-selected speed group. The ≤1.24 m·s⁻¹ group walked slower at their habitual speed compared to the group >1.24 m·s⁻¹ (p<0.01) and compared with controls (p<0.01; Table 1).

ACLR Slow Self-Selected Speed

ACLR individuals in the slow self-selected speed group demonstrated increased vGRF during early stance (10-13%) while walking at 1.3 m·s⁻¹ compared to habitual speed (Figs. 3A, 3B). KFA, KAM, and KEM were not significantly different in the ACLR limb while walking at 1.3 m·s⁻¹ compared to habitual speed (Figs. 4-6A, 4-6B). While walking at 1.3 m·s⁻¹ and compared with controls, the slow self-selected speed group demonstrated lesser vGRF in early stance (18-22%; Fig. 3A, 3C); lesser KFA during early stance (1-35%), greater KFA during late stance (92-101%; Fig. 4A, 4C); lesser KEM during early and midstance (5-46%); and greater KEM during late stance (87-94%; Fig. 6A, 6C). There were no statistically significant differences in KAM between the ACLR limb of the slow self-selected speed group and controls (Fig. 5A, 5C).

Figure 3.

Vertical ground reaction force (vGRF) waveforms normalized to body weight (BW) for the ACLR limb for each subgroup at habitual speed and 1.3 m·s⁻¹, and uninjured controls (slow, 3A; fast, 3D). Main effects of speed in individuals with ACLR are shown for the functional analysis of variance (slow, 3B; fast 3E). Pairwise comparison effects between the ACLR limb at 1.3 m·s⁻¹ and uninjured controls are shown for the functional analysis of variance (slow, 3C; fast 3F).

Figure 4.

Knee flexion angle (KFA, degrees) waveforms for the ACLR limb for each subgroup at habitual speed and 1.3 m·s⁻¹, and uninjured controls (slow, 4A; fast, 4D). Main effects of speed in individuals with ACLR are shown for the functional analysis of variance (slow, 4B; fast, 4E). Pairwise comparison effects between the ACLR limb at 1.3 m·s⁻¹ and uninjured controls are shown for the functional analysis of variance (slow, 4C; fast, 4F).

Figure 6.

Internal knee extension moment (KEM) waveforms normalized to body weight times height (BW x height) for the ACLR limb for each subgroup at habitual speed and 1.3 m·s⁻¹, and uninjured controls (slow, 5A; fast, 5D). Main effects of speed in individuals with ACLR are shown for the functional analysis of variance (slow, 5B; fast, 5E). Pairwise comparison effects between the ACLR limb at 1.3 m·s⁻¹ and uninjured controls are shown for the functional analysis of variance (slow, 5C; fast, 5F).

Figure 5.

Internal knee abduction moment (KAM) waveforms normalized to body weight times height (BW x height) for the ACLR limb for each subgroup at habitual speed and 1.3 m·s⁻¹, and uninjured controls (slow 5A; fast 5D). Main effects of speed in individuals with ACLR are shown for the functional analysis of variance (slow, 5B; fast 5E). Pairwise comparison effects between the ACLR limb at 1.3 m·s⁻¹ and uninjured controls are shown for the functional analysis of variance (slow, 5C; fast, 5F).

ACLR Fast Self-Selected Walking Speed

No differences in vGRF, KFA, KAM, and KEM were observed between habitual speed and 1.3 m·s⁻¹ in the fast self-selected speed ACLR group (Figs. 3-6D, 3-6E). Yet, compared with uninjured controls, the ACLR limb at 1.3 m·s⁻¹ demonstrated lesser vGRF in early stance (18-24%) and greater vGRF during midstance (54-56%; Fig. 3D, 3F); lesser KFA in early-to-midstance (1-42%) and greater KFA in late stance (72-79%; Fig. 4D, 4F); greater KAM during late stance (87-97%; Fig. 5D, 5F); and lesser KEM in early-to-midstance (5-45%; Fig. 6D, 6F).

Supplementary Analysis

In the contralateral, uninvolved limb, walking at 1.3 m·s⁻¹ elicited no statistically significant changes in biomechanics compared to habitual speed in the ACLR group. Compared with controls, the contralateral limb demonstrated lesser early-to-midstance KFA (1-47%), greater KFA in late stance (96-101%), greater KAM (59-85%), and lesser KEM (5-45%) while walking at 1.3 m·s⁻¹. No statistically significant differences were observed in vGRF between the contralateral limb and controls. Compared to habitual speed, the slow self-selected speed ACLR group demonstrated increased vGRF and KEM during early stance (vGRF: 10-22; KEM: 16-28%) while walking at 1.3 m·s⁻¹. The slow self-selected speed group exhibited lesser KFA in early stance (1-12%), greater KFA in late stance (82-101%), and lesser KEM (70-83%) and KAM (10-37%) compared with controls. No differences in vGRF were observed in the contralateral limb compared with controls in the slow self-selected speed group while walking at 1.3 m·s⁻¹. The fast self-selected speed group did not demonstrate differences in vGRF, KFA, KAM, and KEM while walking at 1.3 m·s⁻¹ compared to habitual speed. However, the fast self-selected speed group demonstrated lesser vGRF (92-94%), KFA (1-48%), and KEM (7-40%) compared with controls (see Supplemental Table 1 and Supplemental Figures 1-6, Supplemental Digital Content).

DISCUSSION

Contrary to our hypothesis, walking at 1.3 m·s⁻¹ elicited no significant changes towards more normalized and dynamic waveforms across all biomechanical variables of interest (vGRF, KFA, KAM, and KEM) compared to habitual speed for our entire ACLR cohort. Moreover, despite walking at a similar speed of 1.3 m·s⁻¹, the ACLR group still demonstrated different gait biomechanics compared with the control group, indicating that acutely increasing and normalizing walking speed post-ACLR may not sufficiently modify gait biomechanics to match uninjured controls. Consistent with our hypothesis, the slow self-selected speed (≤1.24 m·s⁻¹) subgroup demonstrated improvements towards more normalized vGRF waveforms during early stance phase when cued to walk at a faster speed of 1.3 m·s⁻¹. However, biomechanical outcomes from both the slow (vGRF, KFA, and KEM) and fast (vGRF, KFA, KAM, and KEM) self-selected speed groups collected at 1.3 m·s⁻¹ remained significantly different and less dynamic compared with uninjured controls. Interestingly, our supplementary analyses found that the uninvolved, contralateral limb demonstrated vGRF biomechanics that were similar to uninjured controls while walking at 1.3 m·s⁻¹, but KFA, KAM, and KEM remained different from controls. Therefore, while slower walking speed may be a contributing factor to PTOA-related outcomes following ACLR (8, 12) and linked to aberrant gait biomechanics (13), our results suggest that acutely increasing walking speed alone is not sufficient in normalizing gait biomechanics post-ACLR. Our findings do, however, suggest that individuals with a slow self-selected speed post-ACLR could benefit from increasing walking speed in addition to participating in a targeted, precision-based intervention which fully normalizes biomechanics during gait. Hence, there remains a significant clinical need for a breakthrough intervention, beyond acutely normalizing walking speed, to successfully restore normal gait biomechanics post-ACLR.

Walking speed is known to influence the magnitudes of gait biomechanics (19, 37-39) and faster walking speeds are associated with greater first and second peak vGRF, smaller midstance vGRF, and greater peak KFA compared to slower speeds in uninjured controls and individuals with ACLR (19, 37, 39). ACLR individuals walk slower than uninjured matched controls (13), and ACLR patients who walk slower demonstrate more deleterious cartilage health outcomes compared to their counterparts who walk at faster speeds (8, 12). Yet, while previous research has examined the effects of walking speed on gait asymmetries following ACLR (27), this is the one of the first studies to assess the effect of acutely increasing walking speed to match the speed of uninjured controls on the magnitudes of gait biomechanics linked to PTOA development in ACLR individuals. A recent study (27) examined the effects of increasing walking speed, based on a percentage of habitual speed, on discrete gait biomechanical outcomes and gait asymmetry; however, this study investigated differences when both groups (i.e., ACLR and controls) were cued to walk at a faster speed. Therefore, we sought to determine if gait biomechanics could be normalized in ACLR individuals by increasing walking speed in ACLR individuals with slow habitual speeds to match the average habitual walking speed of uninjured controls (13). Our findings show that increasing habitual walking speed does not elicit increases in vGRF and KFA during early stance, which is conflicting with previous literature (19, 37-39). This may be attributable to a potential lack of meaningful change in the increase in walking speed from habitual speed in some of the ACLR group. The slow self-selected speed group demonstrated improvements in vGRF during early stance when walking at a faster speed and this finding is consistent with previous literature (19, 37-39). Despite walking at a similar speed, gait biomechanics of the ACLR individuals remained different from controls which suggests that determinants (e.g., neuromuscular deficiencies, altered motor control, changes to joint tissue, fear or pain) contribute to the development and perpetuation of aberrant gait biomechanics following ACLR. While walking at both their habitual speed and 1.3 m·s⁻¹, ACLR individuals demonstrated biomechanical profiles suggestive of more sustained loading across the limb (i.e., lower peak vGRF and greater vGRF during midstance) and stiffened knee pattern (i.e., lesser KFA in early stance and greater KFA in late stance), which is a biomechanical profile linked to PTOA development (3, 40-43). Furthermore, altered KAM is associated with deleterious changes to medial tibiofemoral cartilage (5) and joint contact forces that are hypothesized to exacerbate PTOA development (43). KAM remained significantly greater in ACLR participants compared with controls thus indicating that increasing walking speed does not normalize frontal plane kinetics which are linked to poor cartilage outcomes (5). Altogether, when cued to walk at 1.3 m·s⁻¹, ACLR individuals still demonstrated persistent aberrant gait biomechanical loading patterns (i.e., less dynamic vGRF waveforms), stiffened knee strategy (i.e., lesser KFA), and altered frontal and sagittal plane loading (i.e., KAM and KEM) compared with controls. It may be that other neuromuscular and compensatory mechanisms are driving aberrant gait biomechanics development following ACLR beyond what can be explained by adaptations in walking speed. For example, neural activation to the quadriceps following ACLR is known to be altered which may contribute to aberrant biomechanics (44, 45), and interventions such as muscle vibration have shown to elicit some improvements in gait biomechanics (46). It may be that walking speed serves as a clinical screening tool to identify individuals who may demonstrate aberrant biomechanics rather than a primary target for biomechanical interventions as increasing beyond a MDD elicited little to no biomechanical changes in the slow self-selected speed group. Thus, the pertinent clinical need remains for a breakthrough intervention to effectively target and improve these aberrant gait biomechanics in ACLR patients to reduce PTOA risk.

Recently, real-time gait biofeedback has been used to modify gait biomechanics via cuing an increase in first peak vGRF in ACLR patients (15, 16). Increasing first peak vGRF during real-time gait biofeedback training subsequently modifies gait biomechanics throughout stance phase including more dynamic vGRF waveforms (i.e., increased first and second peak vGRF, and decreased midstance vGRF), increased knee flexion excursion (i.e., greater KFA in early stance and greater knee extension in late stance), and improved internal KAM and KEM in ACLR individuals (15). These previous real-time gait biofeedback studies (15-18) demonstrate the ability to successfully modify gait biomechanics with acute cueing during a single session which is promising for future interventions to prevent PTOA following ACLR. The slow self-selected speed subgroup in the present study demonstrated some improvements towards more dynamic vGRF waveforms and less sustained limb-level loading patterns when cued to walk at 1.3 m·s⁻¹ compared to habitual speed. Therefore, individuals who walk slower than 1.24 m·s⁻¹ may benefit most from a combination therapy targeting biomechanical variable magnitudes via real-time biofeedback along with increasing walking speed to maximize the ability to normalize gait biomechanics post-ACLR. Future research should also seek to establish specific threshold values for walking speed to identify individuals who are likely to demonstrate significantly altered gait biomechanics or PTOA-related outcomes, or as an additional screening tool for clinical trial gait retraining interventions aimed at improving aberrant biomechanics in ACLR patients. Our results suggest that walking speed modifications are likely not a stand-alone replacement for more sophisticated biomechanical cueing to restore gait biomechanics post-ACLR. It is possible that increasing walking speed in uninjured individuals with slow habitual walking speeds may decrease the risk of idiopathic knee osteoarthritis development, yet this was outside the scope of the current study.

The present study introduces novel findings on the influence of walking speed on gait biomechanics 6 months post-ACLR, but the results should be interpreted with considerations. First, we assessed acute changes in biomechanics between walking speeds at a single timepoint post-ACLR and it remains unclear whether gait biomechanics would further improve if walking speed was incorporated into targeted intervention programs over multiple weeks. Secondly, participants walked at their 6-month post-ACLR habitual speed and a single cued speed of 1.3 m·s⁻¹, but it may be that gait biomechanics improve when ACLR individuals walk at their pre-ACLR habitual. Similarly, we utilized an evidence-based control speed (13, 28) as the cued increased walking speed, but it is not clear if walking at a speed faster than 1.3 m·s⁻¹ , or at a speed relative to habitual speed (e.g., 15-20% increase/decrease from habitual), could elicit improvements in gait biomechanics. The matched control group, on average, did walk slightly faster (1.34 m·s⁻¹) than our predetermined speed (1.3 m·s⁻¹), yet the difference between those speeds was not statistically different (Table 1), and the difference did not exceed the reported minimal clinically significant walking speed difference (~0.1 m·s⁻¹) (47), or our calculated MDD (0.06 m·s⁻¹). Lastly, we did not investigate the specific individual biomechanical profiles used by each participant to achieve the 1.3 m·s⁻¹ walking speed; therefore, we cannot reach definitive conclusions as to why increasing walking speed did not normalize biomechanics as well as why ACLR individuals’ biomechanics remained different than uninjured controls despite walking at a similar speed.

CONCLUSIONS

Overall, we found that acutely cueing a speed of 1.3 m·s⁻¹ did not normalize gait biomechanics in the ACLR limb compared to matched uninjured controls. Individuals post-ACLR who walk with a slower self-selected speed tended to demonstrate improvements towards more normalized vGRF during early stance when cued to walk at a faster speed of 1.3 m·s⁻¹ compared to their habitual speed. Our data suggest that acutely increasing walking speed is not a sufficient intervention target to fully normalize aberrant gait biomechanics in ACLR patients and there is still a need for a breakthrough gait retraining intervention that specifically cues gait biomechanics following ACLR to decrease PTOA risk.