Sustained Limb-Level Loading: A Ground Reaction Force Phenotype Common to Individuals at High-Risk For and Those With Knee Osteoarthritis

Abstract

OBJECTIVE:

To compare the vertical (vGRF), anterior-posterior (apGRF) and medial-lateral (mlGRF) ground reaction force (GRF) profiles throughout the stance phase of gait: 1) between individuals 6-12 months post anterior cruciate ligament reconstruction (ACLR) and uninjured matched controls; and 2) between ACLR and individuals with differing radiographic severities of knee osteoarthritis (KOA) defined as Kellgren and Lawrence (KL) grades KL2, KL3, and KL4.

METHOD:

A total of 196 participants were included in this retrospective cross-sectional analysis. Gait biomechanics were collected from individuals 6-12 months post-ACLR (n=36), uninjured controls matched to the ACLR group (n=36), and individuals with KL2 (n=31), KL3 (n=67), and KL4 OA (n=26). Between-group differences in vGRF, apGRF, and mlGRF were assessed in reference to the ACLR group throughout each % of stance phase using a functional linear model.

RESULTS:

The ACLR group demonstrated lesser vGRF and apGRF in early and late stance compared to the uninjured controls, with large effects (d range: 1.35-1.66). Conversely, the ACLR group exhibited greater vGRF (87-90%; 4.88%BW; d=0.75) and apGRF (84-94%; 2.41%BW; d=0.79) than the KL2 group in a small portion of late stance. No differences in mlGRF profiles were observed between the ACLR and either the uninjured controls or the KL2 group. The magnitude of difference in GRF profiles between the ACLR and OA groups increased with OA disease severity.

CONCLUSION:

Individuals 6-12 months post-ACLR exhibit strikingly similar GRF profiles as individuals with KL2 KOA, suggesting both patient groups may benefit from targeted interventions to address aberrant GRF profiles.

INTRODUCTION

Altered limb loading during walking – measured via the ground reaction force - is exhibited both by individuals at high-risk for developing knee osteoarthritis (KOA) (1) following anterior cruciate ligament injury and reconstruction (ACLR) and in individuals with established radiographic KOA developed via multiple etiologies (2–8). Aberrant limb loading profiles during walking are known contributors to KOA, and are associated with metabolic and compositional changes to joint tissue (9–13). Individuals with an ACLR exhibit lesser peak vertical ground reaction force (vGRF) and vGRF profiles that are less dynamic compared to uninjured controls (i.e., characterized by lesser first and second vGRF peak magnitudes and greater midstance vGRF) (4). A less dynamic limb loading profile is linked to deleterious changes in articular cartilage composition (12,14), altered metabolism within the joint-tissue (11,13), and worse patient-reported outcomes within the first 12 months post-ACLR (15). Further, lesser peak vGRF in the ACLR limb at 6 months post-ACLR, have been linked to worse patient-reported outcomes at 12 months post-ACLR (16). Similarly, individuals with radiographic KOA walk with lesser overall peak limb-loading and a less-dynamic vGRF profile, a profile that worsens with increased radiographic KOA disease severity (2). Therefore, previous literature suggests that a similar GRF phenotype exists between younger individuals at high-risk for KOA follow joint injury and older adults with established symptomatic and radiographic KOA; however, little research exists to compare limb-level loading profiles in individuals at high-risk of KOA to those with established KOA and compared to controls. Identifying a common GRF phenotype between those at high-risk for and individuals with KOA is important for early detection, treatment and monitoring of biomechanical determinants for both KOA onset and progression.

Less dynamic vGRF profiles exhibited by patients at high risk of KOA and those with KOA (3,4) may be harmful, as lesser first and second peak vGRF and greater vGRF at midstance may elicit more sustained loading to the knee (14,17). Articular cartilage is governed by viscoelastic properties and deforms in response to loading in a time-dependent manner (17,18). Consequently, a less dynamic loading profile may cause greater articular cartilage strain and result in accelerated, deleterious biological joint tissue changes as compared to dynamic loading profiles characteristic of uninjured controls. Lesser peak loading has been linked to biological articular cartilage changes in those with ACLR (11,12,19). Similarly, individuals with a Kellgren and Lawrence (KL) grade 2-4 radiographic KOA exhibit lesser 1^st and 2^nd peak vGRF in comparison to uninjured controls (3). Further, greater vGRF at midstance, a key identifying feature of sustained limb-level loading profiles, is strongly linked to deleterious changes in articular cartilage composition following ACLR (14). Overall, previous data indicate a high likelihood that a unifying GRF phenotype exists between those at high risk for KOA onset and in those with established KOA that is indicative of sustained compressive loading of the joint tissue.

The majority of GRF research has focused on understanding vGRF profiles following ACLR and in KOA populations (3,4,11,20). However, limited research exists to characterize anterioposterior (apGRF) and mediolateral (mlGRF) ground reaction forces, metrics that are linked to the AP and ML shear forces applied to knee tissues (e.g., articular cartilage and bone) (21), in the same ACLR and KOA patient populations. The superficial regions of articular cartilage, which are subject to attenuating shear forces during ambulation, are among the first areas to undergo compositional OA-related changes (17). Lesser peak apGRF and mlGRF, similar to lesser peak vGRF, may reflect a protective tendency to reduce loading magnitude on the knee joint (22). Previous research found that older adults who reported knee pain demonstrated less-dynamic apGRF during walking (e.g., characterized by both lesser braking and propulsive forces) and symptomatic individuals with radiographic KOA and knee pain exhibited greater mlGRF (e.g., greater medially-directed GRF throughout stance phase) in comparison to uninjured controls (20); however, the impact of differing stages of radiographic severity on apGRF and mlGRF profiles was not assessed. Thereby, a comprehensive assessment of vGRF, apGRF, and mlGRF profiles is critical to determine similarities between individuals at high-risk of ACLR and those with KOA and secondarily, if the GRF profiles suggests less dynamic limb-level loading compared to young, uninjured controls. Understanding the GRF profiles exhibited by both the ACLR and KOA groups important for informing the applicability of future interventions that target aberrant GRF profiles. Therefore, the purpose of our hypothesis-generating study was to compare GRF profiles (vGRF, apGRF, mlGRF) throughout the stance phase of gait between individuals 6-12 months post-ACLR and uninjured controls, as well as individuals with KOA stratified by radiographic disease severity (KL2, KL3, or KL4).

MATERIALS AND METHODS

Study Design

Data for the current study incorporated multiple datasets. We included baseline data from a phase I and phase II clinical trial in KOA patients. We also included baseline biomechanical data from ACLR patients enrolled in an experimental crossover study (23) and a convenience sample of uninjured controls matched to the ACLR group. All participants underwent a biomechanical gait analysis at a single timepoint in the same laboratory. If multiple testing sessions existed per participant, data from the baseline session was utilized. Participants were recruited from various settings, including a local orthopedic clinic, intercollegiate athletics, and the broader community. The study design identified the ACLR group as the primary group of interest, with planned comparisons in gait profiles as follows: ACLR vs. uninjured controls, ACLR vs. KL2, ACLR vs. KL3, and ACLR vs. KL4. The Institutional Review Board approved all study methods and all participants provided written, informed consent and assent (if under 18 years of age).

ACLR Participants

Participants were 6-12 months post primary, unilateral ACLR and between 17-35 years of age. Individuals recruited for the ACLR group were excluded if there was a previous history of lower extremity orthopedic surgery prior to ACLR, if a multi-ligament reconstruction was required at ACLR, previous diagnosis of radiographic osteoarthritis, or were pregnant. We also excluded individuals with history of cardiovascular conditions, neurodegenerative conditions, a body mass index (BMI) ≥35 kg/m² or previous history of a lower extremity fracture.

Participants with Radiographic KOA

All participants in the KOA group (40-75 years of age) demonstrated radiographic knee OA (KL grade ≥2) which was assessed by a fellowship-trained musculoskeletal radiologist. OA patients were further separated into three groups based on clinical definitions of radiographic OA severity: KL2 (clear osteophyte formation and joint space narrowing), KL3 (multiple osteophytes, joint space narrowing and potential bone deformation) and KL4 (large osteophytes visible, definite joint space narrowing, sclerosis and severe bone deformation) (24). Specific exclusionary factors for the KOA group included receiving any intra-articular investigational drug, biologic or hyaluronic acid injection within the past 6-months, a corticosteroid injection within the past 3-months, previous history of anti-depressant or stimulant use within the past 4 weeks, or a clinical diagnosis of rheumatoid or psoriatic arthritis. History of a previous joint injury was not considered an inclusion or exclusion criteria for the KOA group and the specific etiology causing KOA in the study participants was not determined.

Uninjured, Matched-Control Participants

Individuals between 18-35 years of age with no previous history of lower extremity injury or surgery were recruited to serve as uninjured controls for the ACLR group. We excluded individuals with cardiovascular restrictions, neurodegenerative conditions, or a BMI≥35 kg/m². The uninjured control participants were recruited from the local community and matched to participants in the ACLR group based on sex, BMI (±3 kg/m²) and age (±2 yrs.).

Gait Biomechanical Collection

All participants performed a gait analysis in personal athletic footwear at the individual’s habitual, comfortable walking speed. Participants were asked to walk at a comfortable pace “as if walking over a sidewalk” according to previously published protocols (25). Once acclimated to the laboratory, participants performed 5 successful walking trials across a 6m walkway with 2 embedded force plates (FP406010, Bertec Corporation, Columbus, Ohio), with gait speed determined from timing gates (TF100, TracTronix). Habitual walking speed was determined from the average of 5 walking trials. A successful trial was identified as a trial with the participant contacting each force plate with a single foot, not visibly altering stride characteristics, and maintaining a gait speed between ±5% of the habitual walking speed. Participants did not use an assistive device during walking trials. Force data was sampled at either 960Hz (n=34) or 1200Hz (n=162) based on the methods of the original studies in which data were collected. All force data were low-pass filtered at 10Hz (4^th order Butterworth) and down sampled to a standardized 101 data points (1-100% of stance) of stance phase. Stance phase was defined as the time period between heel strike (vGRF>20 N) and toe off (vGRF<20 N; Visual 3D, C-Motion, Germantown, MD; 2020 x64). vGRF, apGRF and mlGRF were all normalized to body weight (%BW) and the average of 5 walking trials per person, for each biomechanical variable of interest, was utilized for statistical analysis.

Primary Statistical Analysis

Demographic information was reported in Table 1. Group differences in age, BMI and self-selected walking speed were separately assessed using one-way analysis of variance (ANOVA) where the five-level group was considered the main factor and post hoc Tukey pairwise comparisons evaluated any differences in baseline demographics. Between-group differences in GRF (vGRF, apGRF, mlGRF) throughout each % of the stance phase were assessed using functional linear model with group waveform effects (26,27). Specifically, the analysis modeled group waveforms using Bayesian penalized splines (28). All analyses were performed in the R programming language (R Core Team, 2022, version 4.1.2) using open-source (bayesFDA) available on GitHub (29). Ninety-five percent confidence intervals (CI) were constructed and visualized around the mean differences, and regions where the CI of the difference did not include zero were considered statistically significant differences. Maximum between-group differences and associated Cohen’s d effect sizes in regions of stance phase where the 95% CI did not include zero were reported.

TABLE 1.

Demographics (Mean ± SD)

	ACLR (n=36)	Uninjured Controls (n=36)	KL2 (n=31)	KL3 (n=67)	KL4 (n=26)	p value
Age	21.23 ± 3.77c,d,e	21.56 ± 3.74c,d,e	60.23 ± 7.71a,b	62.46 ± 7.58a,b	64.38 ± 6.36a,b	<0.0001
Sex (% Female)	47.22%	47.22%	58.06%	55.22%	50%	-
BMI	24.99 ± 3.47c,d,e	25.04 ± 3.59c,d,e	27.65 ± 4.06a,b	28.80 ±3.50a,b	30.09 ±3.99a,b	<0.0001
Walking Speed	1.23 ±0.09b,e	1.42 ± 0.12a,c,d,e	1.22 ± 0.19b,e	1.16 ± 0.21b	1.06 ± 0.25a,b,c	<0.0001

Denotes a statistically significant difference (p<0.05) from ACLR^a, Uninjured Controls^b, KL2^c, KL3^d or KL4^e

Post Hoc Statistical Analysis

Walking speed is known to impact the magnitude of GRF profiles (30,31); therefore, we conducted additional post hoc supplementary analyses to assess the impact of walking speed on vGRF, apGRF and mlGRF gait patterns. For univariate analyses, ANCOVA models can be used to adjust group effects to control for covariates. Similarly, the primary analyses were reimplemented using covariate-adjusted functional linear models with group waveform effects (i.e., walking speed; Supplemental Table 1 and Supplemental Figures 1–3).

RESULTS

A total of 196 individuals were included in this cross-sectional analysis, comprised of ACLR patients (n=36), an uninjured control group matched to ACLR patients (n=36) and individuals with KL2 (n=31), KL3 (n=67) and KL4 KOA (n=26). Age, walking speed, and BMI were statistically different between groups (all p<0.0001; Table 1). No statistically significant differences between the ACLR group and the uninjured control group were observed for age and BMI (p>0.05). No differences in walking speed were observed between the ACLR and KL2 or KL3 groups (p>0.05). The KL4 group had slower habitual walking speeds in comparison to all groups (p<0.05), except for the KL3 group (Table 1).

Primary Analyses

Vertical Ground Reaction Forces

ACLR individuals exhibited lesser vGRF in early (8-31% of stance, maximum between-group difference: −12.87%BW, maximum effect size: d=−1.35) and late stance (76-94%; −9.13%BW; d=−1.66) and greater vGRF in midstance (43-64%, 11.95%BW, d=1.75) compared to the uninjured control group (Figures 1; 2A–2B). The ACLR group demonstrated greater vGRF in a small portion of late stance (87-90%, 4.88%BW, d=0.75) compared to the KL2 group (Figures 2C–2D). Similarly, the ACLR group exhibited greater vGRF compared to the KL3 group in late stance (78-95%, 9.17%BW, d=1.19, Figures 2E–2F). The ACLR group demonstrated greater vGRF compared to the KL4 group in early (14-29%, 8.44%BW, d=0.94) and late stance (75-96%, 15.3%BW, d=2.27, Figures 2G–2H).

Figure 1.

Group mean vertical (A), anterioposterior (B), mediolateral (C) ground reaction forces (vGRF, apGRF, and mlGRF) normalized by body weight (BW) are illustrated across the entirety of stance phase (1-100%) for the ACLR (black), uninjured controls (red), KL2 (blue), KL3 (green) and KL4 (grey) groups.

Figure 2.

Group mean vertical ground reaction forces (vGRF), normalized by body weight (BW) are illustrated. vGRF is reported by group between the ACLR (black) and uninjured control groups (red; A), ACLR and KL2 (blue; C), ACLR and KL3 groups (green; E), and the ACLR and KL4 groups (grey; G) across the entirety of stance phase (1-100%). Gray boxes indicate areas where the 95% confidence intervals did not include zero. Corresponding vGRF mean differences with associated 95% confidence intervals (grey bands) between groups are shown at right (B,D,F, and H).

Anterior-Posterior Ground Reaction Forces

ACLR individuals exhibited lesser apGRF in early (9-31%, −4.42%BW, d=−1.44) and late stance (70-99%, −4.43%BW, d=−1.58, Figures 3A–3B) than the uninjured control group. Minimal differences in apGRF profiles were observed between ACLR individuals and the KL2 group, with the ACLR group demonstrating greater apGRF in late stance (84-94%; 2.41%BW; d=0.79) compared to the KL2 KOA group (Figure 3C–3D). Similarly, the ACLR group walked with greater apGRF in late stance than the KL3 group (82-97%, 3.32%BW; d=0.95, Figures 3E–3F). In comparison to the KL4 KOA group, ACLR individuals exhibited more dynamic apGRF profiles, with lesser apGRF at heel strike (2-3%; −1.26%BW, d=−1.09), greater apGRF in early stance (10-30%, 2.82%BW, d=0.99), and greater apGRF throughout late stance (78-99%, 6.40%BW, d=1.96, Figures 3G–3H).

Figure 3.

Group mean anterioposterior ground reaction forces (apGRF), normalized by body weight (BW) are illustrated. apGRF is reported by group between the ACLR (black) and uninjured control groups (red; A), ACLR and KL2 (blue; C), ACLR and KL3 groups (green; E), and the ACLR and KL4 groups (grey; G) across the entirety of stance phase (1-100%). Gray boxes indicate areas where the 95% confidence intervals did not include zero. Corresponding apGRF mean differences with associated 95% confidence intervals (grey bands) between groups are shown at right (B,D,F, and H).

Medial-Lateral Ground Reaction Forces

No statistically significant differences in mlGRF profiles were observed between the ACLR and uninjured controls or between the ACLR and KL2 group. ACLR individuals walked with greater mlGRF in early stance (5-9%, 0.86%BW, d=0.48; 17-24%, 1.24%BW, d=0.63) than the KL3 group (Figures 4E–4F). The ACLR group demonstrated greater mlGRF in early (5-8%, 1.15%BW, d=0.71; 19-24%, 1.40%BW, d=0.74) and late stance (84-87%, 1.18%BW, d=0.56) than the KL4 group (Figures 4G–4H).

Figure 4.

Group mean mediolateral ground reaction forces (mlGRF), normalized by body weight (BW) are illustrated. mlGRF is reported by group between the ACLR (black) and uninjured control groups (red; A), ACLR and KL2 (blue; C), ACLR and KL3 groups (green; E), and the ACLR and KL4 groups (grey; G) across the entirety of stance phase (1-100%). Gray boxes indicate areas where the 95% confidence intervals did not include zero. Corresponding mlGRF mean differences with associated 95% confidence intervals (grey bands) between groups are shown at right (B,D,F, and H).

Post Hoc Analysis: Walking Speed-Adjusted Model

Vertical Ground Reaction Forces

After controlling for speed, ACLR individuals exhibited lesser vGRF in early stance (18-27%) and greater vGRF during midstance (51-61%) in comparison to the uninjured control group. The ACLR group demonstrated greater vGRF in late stance in comparison to the KL2 group (79-94%), KL3 group (75-96%), and KL4 group (74-96%). Additionally, the ACLR group demonstrated greater vGRF in comparison to the KL4 group in early stance (7-11%).

Anterior-Posterior Ground Reaction Forces

ACLR individuals still demonstrated lesser apGRF in early (15-22%) and late stance (75-86%) in comparison to the uninjured controls. The ACLR group walked with greater apGRF in late stance in comparison to the KL2 (83-95%), KL3 (85-96%) and KL4 groups (83-99%). The ACLR group also exhibited lesser apGRF in a small portion of early stance compared to the KL3 (15-19%) and greater apGRF at heel strike compared to the KL4 groups (2-4%).

Medial-Lateral Ground Reaction Forces

No differences in mlGRF profiles between the ACLR group and the uninjured controls or the KL2 group were found after controlling for gait speed. The ACLR group demonstrated greater mlGRF in comparison to the KL3 group in early (4-13%; 16-23%) and late stance (83-87%). Similarly, the ACLR group walked with greater mlGRF in early (4-13%) and late stance when compared to the KL4 group (81-89%; Supplementary Figure 3).

DISCUSSION

The ACLR group exhibited similar GRF profiles compared to the KL2 group throughout early and midstance, with differences in vGRF and apGRF occurring in a small portion of late stance (ranging between 84-94%) and no differences in mlGRF profiles. The ACLR group also demonstrated similar GRF profiles in early and midstance in comparison to the KL3 group, with greater differences in vGRF apparent in a larger portion of late stance (ranging between 78-97%). The magnitude of GRF differences between the ACLR and KOA groups increased with disease severity, with GRF profiles becoming less dynamic in advanced-stage KOA. As expected, the uninjured control group demonstrated the most dynamic vGRF and apGRF profiles of the groups that were assessed as part of the current study. GRF profile comparisons did not change substantially after controlling for walking speed in our post hoc supplementary analysis. Overall, these data suggest a unifying GRF phenotypic profile exists between ACLR patients at high-risk for KOA onset and individuals with mild (KL2) and moderate (KL3) radiographic KOA disease severity. ACLR individuals and those with KOA demonstrate more sustained or less dynamic limb loading profiles (i.e., lesser GRF peaks and greater midstance GRF) compared to uninjured controls. Further, individuals with advanced radiographic KOA disease (i.e., KL4 group) clearly demonstrated the least dynamic limb-level loading profile. Data from our study indicate that less dynamic limb-level loading profiles should be further studied as a unifying mechanistic link between high-risk KOA populations and those who have already developed KOA. Modifying less dynamic limb-level loading profiles may be a viable intervention option for preventing and slowing KOA onset and progression.

Our cross-sectional study is unable to discern if limb-level loading contributed to a direct mechanistic pathway linking KOA risk to disease progression. However, we found that individuals only 6-12 months post-ACLR exhibit similar limb-level loading to individuals with established, mild-KOA (i.e., KL2). Importantly, young individuals (i.e., 21±4 yrs) at high-risk of KOA following ACLR have limb-loading profiles that resemble older individuals with radiographic KOA (i.e., 60±8 yrs). Differences between ACLR individuals and those with moderate to severe KOA (KL3 and 4) suggests that peak loading magnitude decreases and loading profiles become more sustained (i.e., less dynamic) as KOA disease severity worsens. Previous literature has identified that individuals within 12 months post-ACLR, who were classified as symptomatic based on the Englund definition on the Knee Osteoarthritis Outcome Score (32), demonstrated less-dynamic vGRF loading profiles in comparison to asymptomatic ACLR patients within the same post-operative timeframe (15). Altered GRF loading profiles are linked to both knee-related symptoms and biological changes in the ACLR and KOA groups (9,11–16,20), highlighting the need to target aberrant GRF profiles to improve patient outcomes. Data from the current study indicate that progressive changes in limb-level loading, measured primarily in vGRF and apGRF profiles, are associated with changes across the continuum of KOA disease severity, identifying a striking linkage between individuals at high-risk for KOA and those with mild, moderate and severe KOA.

Our study identifies a GRF phenotype that links individuals at risk for KOA and individuals with mild radiographic KOA (i.e., KL2). Our findings are important in the context of previous research, as less dynamic limb-level loading is linked to deleterious joint tissue changes indicative of KOA development (11,12,19,33). Less dynamic vGRF profiles have been reported in ACLR individuals compared to uninjured controls in the first 12-months post ACLR (4). Specifically, discrete components of a sustained vGRF loading profiles (e.g., lower first peak and higher vGRF at midstance) are associated with deleterious biological tissue changes indicative of KOA development (11,12,14,19) and worse patient-reported outcomes (15). Additionally, individuals who went on to develop radiographic knee OA at 5 years post-ACLR walked with lower medial compartment joint contact forces at 6 months post-ACLR compared to individuals who did not develop KOA in that timeframe (19). Interestingly, the majority of previous research has focused on understanding the link between the first vGRF peak and KOA onset and progression (12,15,16,33–35); yet the features of the vGRF waveform that differentiate between the ACLR group and KL2 and KL3 KOA subgroups groups were near the second vGRF peak. The present study identified the magnitudes of between-group differences in vGRF in late stance increased between the ACLR group and with KOA severity [KL2 (4.88%), KL3 (9.17%), and KL4 groups (15.3%)]. Although a minimally important difference for GRF magnitude has yet to be developed, previous literature has found a 6% body weight difference in vGRF at the 1^st bimodal peak between patients at 6 and 12 months post-ACLR and uninjured controls (4), suggesting that the vGRF profiles between ACLR and KL2 patients were more closely aligned than between the ACLR patients and uninjured controls in the present study. Future research should assess the potential link between the magnitude of the second vGRF peak and outcomes of KOA onset and disease severity. Overall, a less dynamic vGRF profile, exhibited by individuals 6-12 months following ACLR and in individuals with KL2 radiographic KOA, may be an important biomechanical phenotype and potential mechanism linking sustained vGRF loading and KOA onset and development.

ACLR individuals demonstrated lesser braking (i.e., apGRF at early stance) and propulsive forces (i.e., apGRF at late stance; −4.5% BW) compared to the uninjured control group but exhibited similar propulsive forces (−2.5% BW) compared to the KL2 group. No differences in mlGRF were observed between the ACLR group and the uninjured controls or the KL2 group. Previous research found no apparent differences in apGRF but found higher laterally-directed mlGRF in early and late stance and greater medially-directed mlGRF throughout midstance between individuals with radiographic KOA and uninjured controls after controlling for gait speed (20); however, the study did not stratify by radiographic KOA severity. Although little is known about apGRF and mlGRF profiles in the ACLR population, our study suggests that altered apGRF profiles are apparent 6-12 months post-ACLR and apGRF profiles are similar between ACLR individuals and individuals with KL2 KOA. apGRF and mlGRF provide an estimate of the shear forces applied to the articular cartilage (21). Altered articular cartilage loading (i.e., aberrant GRF profiles) may result in breakdown of the superficial zone of cartilage, caused by a reduction in proteoglycan density and increasing disorganization of the collagen network (17,36). Disruption of the superficial zone may contribute to a reduced capacity to withstand compressive loads (17) (i.e., estimated by vGRF), highlighting the need for future research assessing the link between apGRF and mlGRF on KOA outcomes. It is also possible that a shorter step length may contribute to lesser braking and propulsive apGRF in the ACLR and KL2 KOA groups. Future studies should assess the impact of spatiotemporal aspects of gait on differentiating KOA disease severity. Additionally, we found that individuals with KL4 KOA exhibited lesser medially-directed mlGRF profiles in early and late stance compared to ACLR individuals. The current impact of apGRF and mlGRF is unclear and future research should determine the link between apGRF and mlGRF and biological outcomes consistent with KOA development.

In the present study, there were no statistically significant differences in walking speed between the ACLR group and the KL2 or KL3 KOA groups. Expected between-group differences in BMI and age were observed between the ACLR and KOA groups, as all KOA groups had a greater BMI and were older than the ACLR group (Table 1). Due to the strong link between GRF and gait speed (30), we decided to control for between-group differences in walking speed post hoc. After covarying by walking speed, minimal differences in GRF profiles were observed in comparison to the primary model, suggesting that factors contributing to aberrant GRF profiles are multi-factorial. However, the ACLR group walked at the same speed as the KL2 and KL3 group, increasing the confidence in the interpretation of our results of a unifying GRF profile observed between individuals 6-12 months post-ACLR and early KOA individuals. Although there are inherent demographic differences between our ACLR and KOA groups, our study was focused on the ACLR group and was not designed to determine if GRF profiles differed between individuals with KOA in comparison to uninjured controls. Previous literature has already established differences in gait biomechanics between individuals with KOA and uninjured older adults without KOA (3,37,38). Future studies should assess the influence of other demographics on GRF profiles in KOA patients.

Due to the cross-sectional design of our study, it is unknown if aberrant gait biomechanical profiles existed before individuals sustained an ACL injury or developed structural KOA. Further, a history of knee joint injury was not part of the inclusion criteria for the KOA patients in our study. Therefore, our KOA participants may be more generalizable to the population of people with established KOA and not specific to a particular knee OA etiology (i.e., post-traumatic OA) or phenotype. Post-traumatic KOA may represent distinct disease processes compared to OA-related processes that occur during disease development from other etiologies. Previous research identified differences in neuromuscular activation patterns and gait kinetics between individuals with post-traumatic and idiopathic KOA; however, the group with post-traumatic KOA (56 ± 9 years old) was much older than the present ACLR group (21.2 ± 3.8 years old), who did not have a KOA diagnosis (39). While our data demonstrate similarities between GRF profiles in ACLR patients following ACLR and a cohort of individuals with knee-OA from unknown causes, further research is needed to determine if gait biomechanics differ between different knee OA phenotypes. Due to inter-study variability in project design and differences in marker set-ups, we did not assess kinematic profiles and differences in knee joint moments. Future studies should examine knee kinematic and kinetic profiles in the same ACLR group within 6-12 months post-ACLR and individuals with KOA, as stratified by disease severity. Additionally, our analysis focused on the ACLR limb as previous research identified that vGRF was relatively similar between limbs of individuals at 12 months post-ACLR (4), while both limbs differed compared to uninjured controls. Although it is well-established that altered gait biomechanics persist long-term following ACLR (4,8,8,19,40), longitudinal cohort studies are limited, and future research should assess the time-dependent progression of aberrant GRF profiles in both limbs following post-ACLR relative to KOA development and progression.

Previous research has demonstrated the capability of modifying vGRF profiles in ACLR individuals by cuing an increase in peak vGRF with real-time gait biofeedback in a laboratory setting (23,41). Increasing the dynamic nature of the vGRF profile with real-time gait biofeedback has been found to acutely alter biological outcomes related to joint tissue breakdown (42), suggesting a mechanobiological link between limb-level loading magnitude and KOA-related joint tissue changes. Data from the current study suggests that cueing a more dynamic vGRF waveform in patients with KOA as previously conducted in those with an ACLR (23,41) may be similarly beneficial for reestablishing healthy limb-level loading that may slow KOA progression. Recent advances in technology have made it possible to estimate GRF in real-world settings using load sensing insoles and accelerometers (43–45); therefore, it is possible that GRF can be used as a therapeutic target for gait retraining outside of the laboratory environment in the near future.

ACLR individuals demonstrate less dynamic limb-level loading profiles than uninjured controls. Limb-level loading profiles worsen with progressed radiographic KOA disease stage. Overall, these data suggest a unifying GRF profile exists between ACLR patients at high-risk for KOA onset and individuals with mild radiographic KOA disease severity (KL2). Therefore, developing interventions to modify aberrant ground reaction force profiles in both people at high-risk of KOA following ACLR and those with established KOA may be appropriate for reestablishing healthy dynamic limb-level loading that may prevent and slow KOA progression.