Associations of Strength and Spatiotemporal Gait Variables With Knee Loading During Gait After Anterior Cruciate Ligament Reconstruction

We recruited 23 participants with primary unilateral ACLR (13 men and 10 women, age = 25.1 ± 6.3 years, height = 174.7 ± 12.0 cm, mass = 76.0 ± 21.9 kg). Based on the data recorded from healthy individuals (unpublished data), we calculated that a sample size of 23 ACLR participants would achieve statistical power of greater than 80% for detecting a significant correlation of 0.5 between knee moments and spatiotemporal gait values. Increased risk of early knee OA development has been well documented in patients after ACLR.³  Thus, to minimize the possible influences of knee-joint degeneration on study outcomes, the participants must have had their unilateral surgery within 5 years of testing (time post-ACLR = 14.4 ± 17.2 months). Volunteers were excluded if they reported a current lower extremity injury or medical condition that would impair performance of the tasks described in the next section. The institutional review board approved the protocol before data collection. All procedures were explained to each participant, and informed consent was obtained.

Maximal concentric isokinetic knee-muscle strength at 60°/s was recorded with a Kin Com III dynamometer (Isokinetic International). The participant was seated in the chair and secured with straps across the trunk and waist. The limb being tested was secured with a thigh strap and the shank pad attached proximal to the malleoli. The axis of rotation of the dynamometer was aligned with the knee-joint axis (ie, femoral epicondyles), and the lever arm of the dynamometer was recorded. Limb weight was controlled using gravity correction before testing for each limb. Oral instructions were given before testing to fold the arms across the chest and extend and flex the lower leg as hard and as fast as possible through the full range of motion (ie, 0° to 90° of knee flexion). Each participant was given 2 practice trials. Four successful trials (ie, going through the full range of motion with arms across the chest) were recorded. The peak extensor moments from the highest 3 trials were then averaged and normalized to body mass (Nm/kg) for subsequent statistical analyses.

We considered clinical assessment of knee-muscle strength and spatial or temporal gait values (eg, step length, stance time) feasible and thus examined these variables as predictors of knee-joint moments. Spatial or temporal gait values were evaluated by instructing the participant to walk at the same self-selected speed across a 16-ft (4.9-m) pressure walkway (model GAITRite-RE16; CIR Systems Inc) 4 times to provide at least 10 footprints or steps for each leg. Specifically, cadence, step length, foot progression angle (ie, the angle between the line of progression and the midline of the footprint), step width (ie, the distance from the heel center of the reference foot to the line of progression of the opposite foot), and stance time (as a percentage of gait cycle) were derived by the GAITRite software. Step length and width were normalized to the body height of each participant.

The kinematics of the lower extremity and ground reaction forces (GRFs) during overground walking were recorded using an 8-camera motion-analysis system (Vicon T series; Oxford Metrics PLC) at 250 Hz and force plates at 1000 Hz (model OPT464508-2K-STT; Advanced Mechanical Technology, Inc), respectively. Reflective markers were placed on anatomical landmarks following the method reported by Tsai et al.⁸  Each participant was instructed to walk at a self-selected speed (1.37 ± 0.17 m/s). Three successful steps (foot completely inside the force plate) were recorded bilaterally for each individual.

Visual3D (version 2021.06.2; C-Motion Inc) was used to process the raw coordinate data in order to compute segmental kinematics and kinetics for both limbs. The trajectory data of the reflective markers were filtered using a fourth-order, zero-lag Butterworth 12-Hz low-pass filter. The local coordinate systems of the pelvis, thigh, shank, and foot were derived from a standing calibration trial. Hip-joint center was estimated using the greater trochanter method,¹⁷  and joint coordinate systems were established based on the International Society of Biomechanics recommendations.¹⁸  Joint kinematics were calculated using Euler angles with the following order of rotations: flexion or extension, abduction or adduction, and internal or external rotation.

Three-dimensional net joint moments were determined using inverse-dynamics equations⁸  and reported as external moments. Peak knee-flexion and -adduction moments during the first half of the stance phase of gait were identified and normalized to body weight and height (% BW·BH) for each walking trial. The beginning and end of the stance phase were determined based on a threshold of 20 N in the vertical GRF.

Statistical analyses were completed with SPSS (version 25; IBM Corp) using a significance level of P ≤ .05. Between-limbs differences in peak knee-flexion and -adduction moments, peak knee-extensor strength, and spatiotemporal variables were examined with paired t tests. We calculated the Pearson correlation coefficient (r) to evaluate the association between each of the variables (ie, maximal isokinetic knee-muscle strength and spatial or temporal gait values) with the peak knee-flexion and -adduction moment in the ACLR limb during walking. To identify the variables associated with between-limbs differences in knee-joint moments, the between-limbs differences (ie, surgical leg minus uninjured leg) were computed for the peak knee-flexion and -adduction moments and each predictor variable as appropriate. The Pearson correlation coefficient (r) was then used to explore the association of the between-limbs differences in each of the predictor variables with the between-limbs differences in the peak knee-flexion and -adduction moments. Absolute values of the correlation coefficients of 0.75, 0.5, and 0.25 was considered excellent, moderate, and fair, respectively. Variables that were significantly correlated with knee-joint moments were then included in a stepwise regression (ie, the stepwise method provided in SPSS; variable entry: P < .05; variable removal: P > .10) to identify the variables with the strongest associations with the peak knee-flexion and -adduction moments.

Decreased peak knee-extensor strength, peak knee-flexion and -adduction moment, and stance time were observed for the ACLR leg compared with the uninjured leg (Table 1). The peak knee-flexion moment for the ACLR leg during gait was positively correlated with knee-extensor strength and step length and negatively correlated with step width of the ACLR leg (Table 2). Therefore, knee-extensor strength, step length, and step width were included in the subsequent stepwise regression analysis. Among these 3 variables that were significantly associated with the peak knee-flexion moment, the stepwise regression indicated that knee-extensor strength and step width were the most dominant predictors (Figure 1), together accounting for 43% of the variance in the peak knee-flexion moment (Table 3).

The peak knee-adduction moment for the ACLR leg during gait was positively correlated with walking speed, cadence, and step length of the ACLR leg (Table 2). The peak knee-adduction moment was also negatively correlated with step width and stance-phase time (Table 2). Thus, walking speed, cadence, step length, step width, and stance-phase time were included in the subsequent stepwise regression analysis. Among these 5 variables that were significantly associated with the peak knee-adduction moment, the stepwise regression demonstrated that stance-phase time and step width were the most dominant predictors (Figure 2), together accounting for 61% of the variance in peak knee-adduction moment (Table 4).

The between-limbs differences in the peak knee-flexion and -adduction moments were not associated with knee-muscle strength or any spatiotemporal gait variable (Table 5).

In this study, we aimed to determine whether variables that are feasible to evaluate and modify in a clinical setting informed knee-joint loading during gait after unilateral ACLR. Several variables were significantly associated with peak knee-flexion and -adduction moments. A greater knee-flexion moment during gait had a fair to moderate association with greater knee-extensor strength, longer step length, and narrower step width. When compared with the knee-flexion moment, associations of the examined measures with the knee-adduction moment were stronger. Specifically, greater knee-adduction moment during gait was moderately to strongly correlated with a faster walking speed, increased cadence, longer step length, narrower step width, and shorter stance-phase time. Among those variables, knee-extensor strength and step width were the dominant predictors, accounting for a fair proportion of the variance (44%) in the peak knee-flexion moment, whereas step width and stance-phase time together accounted for 62% of the variance in the peak knee-adduction moment.

The associations between the predictor variables evaluated and knee-flexion and -adduction moments during gait can be explained, in part, by biomechanical principles that influence net joint moment calculations. The net joint moment is primarily determined by (1) the GRF and the subsequent JRF, (2) the moment arm of the GRF/JRF to the joint, and (3) the angular acceleration of the segment or joint of interest during a dynamic movement. Thus, factors that change the moment arm of the GRF/JRF to the joint (eg, step length, step width) or accelerate or decelerate the angular motion of the knee (eg, knee-muscle strength, stance-phase time) are expected to affect knee-joint moments during walking.

Consistent with our findings, decreased knee-flexion or -adduction moments or both have been commonly observed in surgical limbs after ACLR when compared with the contralateral uninjured limbs^13,19,20  or the limbs of healthy control participants.^9,10,21  However, increased knee-joint moments have also been reported in participants with ACLR compared with healthy control individuals.²²  Because too little or too much loading is believed to contribute to early knee OA,^10,11,13,14  correcting altered knee-loading profiles during rehabilitation after ACLR is a much-needed addition to current clinical care.

Several spatiotemporal variables correlated with knee moments and therefore may serve as objective targets to assist in normalizing altered knee-loading profiles after ACLR (eg, gait training to modify step length, step width, or stance-phase time during gait). For example, our results suggest that patients with reduced knee-flexion and -adduction moments could be provided feedback to increase step length and decrease step width during walking in order to increase their knee moments. These findings are in general agreement with those of prior studies^15,16  in which healthy participants modified their step length and step width during a single session.

Also, patient cues to reduce stance-phase time could be used to increase the peak knee-adduction moment in those exhibiting a reduced knee-adduction moment. Simple strategies to modify step length and width could include placing targets on the floor^15,16  and using a metronome to assist in pacing. An important consideration, highlighted by Favre et al in 2016,¹⁶  is that cues to change a single gait variable, such as step width, can result in other involuntary changes to gait (eg, foot progression angle). Furthermore, a single gait-retraining target may influence both sagittal- and frontal-plane knee moments. Thus, research is needed to determine whether modifications of the variables identified in this study can normalize altered knee loading in patients after ACLR and thereby delay the onset or reduce the severity (or both) of knee OA.

Knee-extensor weakness is a common persistent impairment^23,24  that was also observed in this study. The knee extensors function to eccentrically control knee flexion during weight acceptance, so a positive correlation between knee strength and peak knee-flexion moment was not surprising. This positive correlation suggests that strength improvements could result in increased knee-flexion moments.²⁵  However, whether weakness is responsible for decreased knee loading or, alternatively, the altered movement pattern perpetuates the knee weakness remains an open question.

Asymmetry in knee net joint moments between limbs during walking is a typical finding after ACLR and a potential marker of early knee OA risk.^12,13  As such, we also evaluated between-limbs differences of the clinical predictor variables as predictors of between-limbs differences in knee-flexion and -adduction moments. However, none of the variables we assessed were associated with the between-limbs difference in knee-flexion or -adduction moment. These data indicate that the measured variables were less sensitive as predictors of between-limbs asymmetry in knee joint moments.

In the present study, an isokinetic dynamometer and a pressure walkway were used to quantify knee-muscle strength and the spatiotemporal variables of gait, respectively. Our findings serve as a first step in emphasizing the potential value of assessing patients' muscle strength and spatiotemporal gait variables (eg, cadence, step length, step width) to modulate knee-joint moments after ACLR. The pressure walkway was selected because it provides accurate spatiotemporal gait values^26,27  that can be collected quickly in a clinical setting and has been used to assess functional outcomes in patients after ACLR.²⁸  Therefore, the pressure walkway is one objective low-technology solution for addressing a persistent clinical problem after ACLR. Moreover, prior authors^15,16  indicated that simple strategies, such as placing marks on the floor, could be implemented to change spatiotemporal gait values that influence knee-joint loading. Importantly, some spatiotemporal characteristics, such as step length, may be altered by changing cadence while walking on a treadmill, whereas manipulating step length overground is likely more condition specific. However, walking speed and stride or step lengths must be monitored and manipulated during overground walking, whereas the constant speed during treadmill walking may allow participants to focus on 1 factor.

We acknowledge that many current and emerging technologies, such as inertial sensors, could also yield valid spatiotemporal gait information plus lower extremity joint angles when compared with motion-capture systems.^29,30  Future investigators should evaluate whether using alternative tools to measure the same spatiotemporal variables, incorporating more measurements, or both can supply similar or improved knee-joint loading information. For example, including peak knee-flexion angle would likely improve the regression model for knee-flexion moment and including frontal-plane trunk-lean angle could be a helpful variable for influencing the knee-adduction moment.¹⁶ 

Several limitations in our study must be acknowledged. First, given the cross-sectional experimental design, causal relationships between the variables evaluated and knee-joint moments cannot be established based on the correlation or regression analyses. Moreover, participants' knee moments during gait were quantified using the motion-capture system, and spatiotemporal gait variables were quantified using the pressure walkway during separate walking trials. Although we expected comparable gait mechanics during all walking trials, given that the participants walked at their self-selected speeds, the inability to quantify knee moments and spatiotemporal gait variables concurrently due to the limitations of the technology could have influenced the relationships among the variables. Lastly, our patients had different characteristics with respect to factors such as sex and time post-ACLR. Future studies with a larger sample would allow patient characteristics to be included in the regression analysis.

Our work is a first step in bridging the research-to-clinic gap. We identified several variables that predicted knee-joint moments and that we propose can feasibly be evaluated in a clinical setting. Among the identified variables, knee-extensor strength, step width, step length, and stance-phase duration were the dominant variables associated with peak knee-flexion and -adduction moments. As such, these variables could be used to monitor and guide interventions to modify knee-joint loading during ACLR rehabilitation.

This study was supported, in part, by the Georgia Foundation for Physical Therapy.