Three-dimensional data-tracking dynamic optimization simulations of human locomotion generated by direct collocation

The aim of this study was to perform full-body three-dimensional (3D) dynamic optimization simulations of human locomotion by driving a neuromusculoskeletal model toward in vivo measurements of body-segmental kinematics and ground reaction forces. Gait data were recorded from 5 healthy participants who walked at their preferred speeds and ran at 2 m/s. Participant-specific data-tracking dynamic optimization solutions were generated for one stride cycle using direct collocation in tandem with an OpenSim-MATLAB interface. The body was represented as a 12-segment, 21-degree-of-freedom skeleton actuated by 66 muscle-tendon units. Foot-ground interaction was simulated using six contact spheres under each foot. The dynamic optimization problem was to find the set of muscle excitations needed to reproduce 3D measurements of body-segmental motions and ground reaction forces while minimizing the time integral of muscle activations squared. Direct collocation took on average 2.7 ± 1.0 h and 2.2 ± 1.6 h of CPU time, respectively, to solve the optimization problems for walking and running. Model-computed kinematics and foot-ground forces were in good agreement with corresponding experimental data while the calculated muscle excitation patterns were consistent with measured EMG activity. The results demonstrate the feasibility of implementing direct collocation on a detailed neuromusculoskeletal model with foot-ground contact to accurately and efficiently generate 3D data-tracking dynamic optimization simulations of human locomotion. The proposed method offers a viable tool for creating feasible initial guesses needed to perform predictive simulations of movement using dynamic optimization theory. The source code for implementing the model and computational algorithm may be downloaded at http://simtk.org/home/datatracking.     

Improved walking function in laboratory does not guarantee increased community walking in stroke survivors: Potential role of gait biomechanics

Reduced daily stepping in stroke survivors may contribute to decreased functional capacity and increased mortality. We investigated the relationships between clinical and biomechanical walking measures that may contribute to changes in daily stepping activity following physical interventions provided to participants with subacute stroke. Following ≤40 rehabilitation sessions, 39 participants were categorized into three groups: responders/retainers increased daily stepping >500 steps/day post-training (POST) without decreases in stepping at 2–6 month follow-up (F/U); responders/non-retainers increased stepping at POST but declined >500 steps/day at F/U; and, non-responders did not change daily stepping from baseline testing (BSL). Gait kinematics and kinetics were evaluated during graded treadmill assessments at BSL and POST. Clinical measures of gait speed, timed walking distance, balance and balance confidence were measured at BSL, POST and F/U. Between-group comparisons and regression analyses were conducted to predict stepping activity from BSL and POST measurements. Baseline and changes in clinical measures of walking demonstrated selective associations with stepping, although kinematic measures appeared to better discriminate responders. Specific measures suggest greater paretic vs non-paretic kinematic changes in responders with training, although greater non-paretic changes predicted greater gains (i.e., smaller declines) in stepping in retainers at F/U. No kinetic variables were primary predictors of changes in stepping activity at POST or F/U. The combined findings indicate specific biomechanical assessments may help differentiate changes in daily stepping activity post-stroke.     

Association between walking ability and trunk and lower-limb muscle atrophy in institutionalized elderly women: a longitudinal pilot study

Background

The aim of this study was to investigate the association between walking ability and muscle atrophy in the trunk and lower limbs.

Methods

Subjects in this longitudinal study were 21 elderly women who resided in nursing homes. The thicknesses of the following trunk and lower-limb muscles were measured using B-mode ultrasound: rectus abdominis, external oblique, internal oblique, transversus abdominis, erector spinae, lumbar multifidus, psoas major, gluteus maximus, gluteus medius, gluteus minimus, rectus femoris, vastus lateralis, vastus intermedius, biceps femoris, gastrocnemius, soleus, and tibialis anterior. Maximum walking speed was used to represent walking ability. Maximum walking speed and muscle thickness were assessed before and after a 12-month period.

Results

Of the 17 measured muscles of the trunk and lower limbs, age-related muscle atrophy in elderly women was greatest in the erector spinae, rectus femoris, vastus lateralis, vastus intermedius, and tibialis anterior muscles. Correlation coefficient analyses showed that only the rate of thinning of the vastus lateralis was significantly associated with the rate of decline in maximum walking speed (r = 0.518, p < 0.05).

Conclusions

This longitudinal study suggests that reduced walking ability may be associated with muscle atrophy in the trunk and lower limbs, especially in the vastus lateralis muscle, among frail elderly women.     

Influence of locomotion speed on biomechanical subtask and muscle synergy

This paper investigates the relationship of biomechanical subtasks, and muscle synergies with various locomotion speeds. Ground reaction force (GRF) of eight healthy subjects is measured synchronously by force plates of treadmill at five different speeds ranging from 0.5 m/s to 1.5 m/s. Four basic biomechanical subtasks, body support, propulsion, swing, and heel strike preparation, are identified according to GRF. Meanwhile, electromyography (EMG) data, used to extract muscle synergies, are collected from lower limb muscles. EMG signals are segmented periodically based on GRF with the heel strike as the split points. Variability accounted for (VAF) is applied to determine the number of muscle synergies. We find that four muscle synergies can be extracted in all five situations by non-negative matrix factorization (NMF). Furthermore, the four muscle synergies and biomechanical subtasks keep invariant as the walking speed changes.     

Compliant bipedal model with the center of pressure excursion associated with oscillatory behavior of the center of mass reproduces the human gait dynamics

Although the compliant bipedal model could reproduce qualitative ground reaction force (GRF) of human walking, the model with a fixed pivot showed overestimations in stance leg rotation and the ratio of horizontal to vertical GRF. The human walking data showed a continuous forward progression of the center of pressure (CoP) during the stance phase and the suspension of the CoP near the forefoot before the onset of step transition. To better describe human gait dynamics with a minimal expense of model complexity, we proposed a compliant bipedal model with the accelerated pivot which associated the CoP excursion with the oscillatory behavior of the center of mass (CoM) with the existing simulation parameter and leg stiffness. Owing to the pivot acceleration defined to emulate human CoP profile, the arrival of the CoP at the limit of the stance foot over the single stance duration initiated the step-to-step transition. The proposed model showed an improved match of walking data. As the forward motion of CoM during single stance was partly accounted by forward pivot translation, the previously overestimated rotation of the stance leg was reduced and the corresponding horizontal GRF became closer to human data. The walking solutions of the model ranged over higher speed ranges (~1.7 m/s) than those of the fixed pivoted compliant bipedal model (~1.5 m/s) and exhibited other gait parameters, such as touchdown angle, step length and step frequency, comparable to the experimental observations. The good matches between the model and experimental GRF data imply that the continuous pivot acceleration associated with CoM oscillatory behavior could serve as a useful framework of bipedal model.     

Estimates of muscle function in human gait depend on how foot-ground contact is modelled

Computational analyses of leg-muscle function in human locomotion commonly assume that contact between the foot and the ground occurs at discrete points on the sole of the foot. Kinematic constraints acting at these contact points restrict the motion of the foot and, therefore, alter model calculations of muscle function. The aim of this study was to evaluate how predictions of muscle function obtained from musculoskeletal models are influenced by the model used to simulate ground contact. Both single- and multiple-point contact models were evaluated. Muscle function during walking and running was determined by quantifying the contributions of individual muscles to the vertical, fore-aft and mediolateral components of the ground reaction force (GRF). The results showed that two factors – the number of foot-ground contact points assumed in the model and the type of kinematic constraint enforced at each point – affect the model predictions of muscle coordination. Whereas single- and multiple-point contact models produced similar predictions of muscle function in the sagittal plane, inconsistent results were obtained in the mediolateral direction. Kinematic constraints applied in the sagittal plane altered the model predictions of muscle contributions to the vertical and fore-aft GRFs, while constraints applied in the frontal plane altered the calculations of muscle contributions to the mediolateral GRF. The results illustrate the sensitivity of calculations of muscle coordination to the model used to simulate foot-ground contact.     

Ground reaction forces and lower-limb joint kinetics of turning gait in typically developing children

Turning is a common locomotor task essential to daily activity; however, very little is known about the forces and moments responsible for the kinematic adaptations occurring relative to straight-line gait in typically developing children. Thus, the aims of this study were to analyse ground reaction forces (GRFs), ground reaction free vertical torque (T_Z), and the lower-limb joint kinetics of 90° outside (step) and inside (spin) limb turns. Step, spin, and straight walking trials from fifty-four typically developing children were analysed. All children were fit with the Plug-in Gait and Oxford Foot Model marker sets while walking over force plates embedded in the walkway. Net internal joint moments and power were computed via a standard inverse dynamics approach. All dependent variables were statistically analysed over the entire curves using the mean difference 95% bootstrap confidence band approach. GRFs were directed medially for step turns and laterally for spin turns during the turning phase. Directions were reversed and magnitudes decreased during the approach phase. Step turns showed reduced ankle power generation, while spin turns showed large T_Z. Both strategies required large knee and hip coronal and transverse plane moments during swing. These kinetic differences highlight adaptations required to maintain stability and reorient the body towards the new walking direction during turning. From a clinical perspective, turning gait may better reveal weaknesses and motor control deficits than straight walking in pathological populations, such as children with cerebral palsy, and could potentially be implemented in standard gait analysis sessions.     

Estimates of muscle function in human gait depend on how foot-ground contact is modelled

Computational analyses of leg-muscle function in human locomotion commonly assume that contact between the foot and the ground occurs at discrete points on the sole of the foot. Kinematic constraints acting at these contact points restrict the motion of the foot and, therefore, alter model calculations of muscle function. The aim of this study was to evaluate how predictions of muscle function obtained from musculoskeletal models are influenced by the model used to simulate ground contact. Both single- and multiple-point contact models were evaluated. Muscle function during walking and running was determined by quantifying the contributions of individual muscles to the vertical, fore-aft and mediolateral components of the ground reaction force (GRF). The results showed that two factors--the number of foot-ground contact points assumed in the model and the type of kinematic constraint enforced at each point--affect the model predictions of muscle coordination. Whereas single- and multiple-point contact models produced similar predictions of muscle function in the sagittal plane, inconsistent results were obtained in the mediolateral direction. Kinematic constraints applied in the sagittal plane altered the model predictions of muscle contributions to the vertical and fore-aft GRFs, while constraints applied in the frontal plane altered the calculations of muscle contributions to the mediolateral GRF. The results illustrate the sensitivity of calculations of muscle coordination to the model used to simulate foot-ground contact.     

Modulation of shoulder muscle and joint function using a powered upper-limb exoskeleton

Robotic-assistive exoskeletons can enable frequent repetitive movements without the presence of a full-time therapist; however, human-machine interaction and the capacity of powered exoskeletons to attenuate shoulder muscle and joint loading is poorly understood. This study aimed to quantify shoulder muscle and joint force during assisted activities of daily living using a powered robotic upper limb exoskeleton (ArmeoPower, Hocoma). Six healthy male subjects performed abduction, flexion, horizontal flexion, reaching and nose touching activities. These tasks were repeated under two conditions: (i) the exoskeleton compensating only for its own weight, and (ii) the exoskeleton providing full upper limb gravity compensation (i.e., weightlessness). Muscle EMG, joint kinematics and joint torques were simultaneously recorded, and shoulder muscle and joint forces calculated using personalized musculoskeletal models of each subject’s upper limb. The exoskeleton reduced peak joint torques, muscle forces and joint loading by up to 74.8% (0.113 Nm/kg), 88.8% (5.8%BW) and 68.4% (75.6%BW), respectively, with the degree of load attenuation strongly task dependent. The peak compressive, anterior and superior glenohumeral joint force during assisted nose touching was 36.4% (24.6%BW), 72.4% (13.1%BW) and 85.0% (17.2%BW) lower than that during unassisted nose touching, respectively. The present study showed that upper limb weight compensation using an assistive exoskeleton may increase glenohumeral joint stability, since deltoid muscle force, which is the primary contributor to superior glenohumeral joint shear, is attenuated; however, prominent exoskeleton interaction moments are required to position and control the upper limb in space, even under full gravity compensation conditions. The modeling framework and results may be useful in planning targeted upper limb robotic rehabilitation tasks.     

Model of loadings acting on the femoral bone during gait

An evaluation of the model of loadings acting on the femoral bone during the whole gait cycle was the main aim of the paper. A computer simulation of the musculoskeletal system based on the gait data collected during gait was used to determine the muscle forces as well as the hip joint reaction. Kinematic parameters as well as the ground reaction force for ninety-nine healthy persons of both sexes (18–36 years old) who had no history of musculoskeletal disease were registered during normal gait with preferred speed and used as inputs for musculoskeletal modelling and numerical simulation with the use of the AnyBody software. Time waveforms of the values of force generated by 21 muscles having attachments on the femoral bone as well as the hip joint reaction force were obtained. Directions of particular forces were presented using a femoral coordinate system. Attachment points for all muscle forces were obtained on the basis of the unscaled standard model with the length of the femur equal to 0.41 m. The presented model of loadings acting on the femoral bone element can be useful for the biomechanical analysis of bone development and remodelling as well as for the optimisation of implant or bone stabilizer design and pre-clinical testing.     

Effect of stable and unstable load carriage on walking gait variability,dynamic stability and muscle activity of older adults

Load carriage perturbs the neuromuscular system, which can be impaired due to ageing. The ability to counteract perturbations is an indicator of neuromuscular function but if the response is insufficient the risk of falls will increase. However, it is unknown how load carriage affects older adults. Fourteen older adults (65 ± 6 years) attended a single visit during which they performed 4 min of walking in 3 conditions, unloaded, stable backpack load and unstable backpack load. During each walking trial, 3-dimensional kinematics of the lower limb and trunk movements and electromyographic activity of 6 lower limb muscles were recorded. The local dynamic stability (local divergence exponents), joint angle variability and spatio-temporal variability were determined along with muscle activation magnitudes. Medio-lateral dynamic stability was lower (p = 0.018) and step width (p = 0.019) and step width variability (p = 0.015) were greater in unstable load walking and step width variability was greater in stable load walking (p = 0.009) compared to unloaded walking. However, there was no effect on joint angle variability. Unstable load carriage increased activity of the Rectus Femoris (p = 0.001) and Soleus (p = 0.043) and stable load carriage increased Rectus Femoris activity (p = 0.006). These results suggest that loaded walking alters the gait of older adults and that unstable load carriage reduces dynamic stability compared to unloaded walking. This can potentially increase the risk of falls, but also offers the potential to use unstable loads as part of fall prevention programmes.     

The evolution of human running: effects of changes in lower-limb length on locomotor economy

Previous studies have differed in expectations about whether long limbs should increase or decrease the energetic cost of locomotion. It has recently been shown that relatively longer lower limbs (relative to body mass) reduce the energetic cost of human walking. Here we report on whether a relationship exists between limb length and cost of human running. Subjects whose measured lower-limb lengths were relatively long or short for their mass (as judged by deviations from predicted values based on a regression of lower-limb length on body mass) were selected. Eighteen human subjects rested in a seated position and ran on a treadmill at 2.68 ms(-1) while their expired gases were collected and analyzed; stride length was determined from videotapes. We found significant negative relationships between relative lower-limb length and two measures of cost. The partial correlation between net cost of transport and lower-limb length controlling for body mass was r=-0.69 (p=0.002). The partial correlation between the gross cost of locomotion at 2.68 ms(-1) and lower-limb length controlling for body mass was r=-0.61 (p=0.009). Thus, subjects with relatively longer lower limbs tend to have lower locomotor costs than those with relatively shorter lower limbs, similar to the results found for human walking. Contrary to general expectation, a linear relationship between stride length and lower-limb length was not found.     

Impact of ankle foot orthosis stiffness on Achilles tendon and gastrocnemius function during unimpaired gait

Ankle foot orthoses (AFOs) are designed to improve gait for individuals with neuromuscular conditions and have also been used to reduce energy costs of walking for unimpaired individuals. AFOs influence joint motion and metabolic cost, but how they impact muscle function remains unclear. This study investigated the impact of different stiffness AFOs on medial gastrocnemius muscle (MG) and Achilles tendon (AT) function during two walking speeds. We performed gait analyses for eight unimpaired individuals. Each individual walked at slow and very slow speeds with a 3D printed AFO with no resistance (free hinge condition) and four levels of ankle dorsiflexion stiffness: 0.25 Nm/°, 1 Nm/°, 2 Nm/°, and 3.7 Nm/°. Motion capture, ultrasound, and musculoskeletal modeling were used to quantify MG and AT lengths with each AFO condition. Increasing AFO stiffness increased peak AFO dorsiflexion moment with decreased peak knee extension and peak ankle dorsiflexion angles. Overall musculotendon length and peak AT length decreased, while peak MG length increased with increasing AFO stiffness. Peak MG activity, length, and velocity significantly decreased with slower walking speed. This study provides experimental evidence of the impact of AFO stiffness and walking speed on joint kinematics and musculotendon function. These methods can provide insight to improve AFO designs and optimize musculotendon function for rehabilitation, performance, or other goals.     

Computational stability of human knee joint at early stance in Gait: Effects of muscle coactivity and anterior cruciate ligament deficiency

As one of the most complex and vulnerable structures of body, the human knee joint should maintain dynamic equilibrium and stability in occupational and recreational activities. The evaluation of its stability and factors affecting it is vital in performance evaluation/enhancement, injury prevention and treatment managements. Knee stability often manifests itself by pain, hypermobility and giving-way sensations and is usually assessed by the passive joint laxity tests. Mechanical stability of both the human knee joint and the lower extremity at early stance periods of gait (0% and 5%) were quantified here for the first time using a hybrid musculoskeletal model of the lower extremity. The roles of muscle coactivity, simulated by setting minimum muscle activation at 0–10% levels and ACL deficiency, simulated by reducing ACL resistance by up to 85%, on the stability margin as well as joint biomechanics (contact/muscle/ligament forces) were investigated. Dynamic stability was analyzed using both linear buckling and perturbation approaches at the final deformed configurations in gait. The knee joint was much more stable at 0% stance than at 5% due to smaller ground reaction and contact forces. Muscle coactivity, when at lower intensities (<3% of its maximum active force), increased dynamic stability margin. Greater minimum activation levels, however, acted as an ineffective strategy to enhance stability. Coactivation also substantially increased muscle forces, joint loads and ACL force and hence the risk of further injury and degeneration. A deficiency in ACL decreases total ACL force (by 31% at 85% reduced stiffness) and the stability margin of the knee joint at the heel strike. It also markedly diminishes forces in lateral hamstrings (by up to 39%) and contact forces on the lateral plateau (by up to 17%). Current work emphasizes the need for quantification of the lower extremity stability margin in gait.     

Statistical shape modelling versus linear scaling: Effects on predictions of hip joint centre location and muscle moment arms in people with hip osteoarthritis

Marker-based dynamic functional or regression methods are used to compute joint centre locations that can be used to improve linear scaling of the pelvis in musculoskeletal models, although large errors have been reported using these methods. This study aimed to investigate if statistical shape models could improve prediction of the hip joint centre (HJC) location. The inclusion of complete pelvis imaging data from computed tomography (CT) was also explored to determine if free-form deformation techniques could further improve HJC estimates. Mean Euclidean distance errors were calculated between HJC from CT and estimates from shape modelling methods, and functional- and regression-based linear scaling approaches. The HJC of a generic musculoskeletal model was also perturbed to compute the root-mean squared error (RMSE) of the hip muscle moment arms between the reference HJC obtained from CT and the different scaling methods. Shape modelling without medical imaging data significantly reduced HJC location error estimates (11.4 ± 3.3 mm) compared to functional (36.9 ± 17.5 mm, p = <0.001) and regression (31.2 ± 15 mm, p = <0.001) methods. The addition of complete pelvis imaging data to the shape modelling workflow further reduced HJC error estimates compared to no imaging (6.6 ± 3.1 mm, p = 0.002). Average RMSE were greatest for the hip flexor and extensor muscle groups using the functional (16.71 mm and 8.87 mm respectively) and regression methods (16.15 mm and 9.97 mm respectively). The effects on moment-arms were less substantial for the shape modelling methods, ranging from 0.05 to 3.2 mm. Shape modelling methods improved HJC location and muscle moment-arm estimates compared to linear scaling of musculoskeletal models in patients with hip osteoarthritis.     

Effects of activation frequency and force on low-frequency fatigue in human skeletal muscle.

No comparison of the amount of low-frequency fatigue (LFF) produced by different activation frequencies exists, although frequencies ranging from 10 to 100 Hz have been used to induce LFF. The quadriceps femoris of 11 healthy subjects were tested in 5 separate sessions. In each session, the force-generating ability of the muscle was tested before and after fatigue and at 2, approximately 13, and approximately 38 min of recovery. Brief (6-pulse), constant-frequency trains of 9.1, 14.3, 33.3, and 100 Hz and a 6-pulse, variable-frequency train with a mean frequency of 14.3 Hz were delivered at 1 train/s to induce fatigue. Immediately postfatigue, there was a significant effect of fatiguing protocol frequency. Muscles exhibited greater LFF after stimulation with the 9.1-, 14.3-, and variable-frequency trains. These three trains also produced the greatest mean force-time integrals during the fatigue test. At 2, approximately 13, and approximately 38 min of recovery, however, the LFF produced was independent of the fatiguing protocol frequency. The findings are consistent with theories suggesting two independent mechanisms behind LFF and may help identify the optimal activation pattern when functional electrical stimulation is used.     

Cross training effects of non-paralytic dorsiflexion muscle strengthening exercise on paralytic dorsiflexor muscle activity,gait ability,and balancing ability in patients with chronic stroke: A randomized,controlled, pilot trial

Objective:To investigate the effects of non-paralytic dorsiflexion muscle strengthening exercise on functional abilities in chronic hemiplegic patients after stroke.Methods:A total of 21 patients with chronic stroke underwent dorsiflexion muscle strengthening exercise (MST) 5 times a week for 6 weeks (the experimental group, MST to non-paralytic dorsiflexion muscles, n=11; the control group, MST to paralytic dorsiflexion muscles; n=10). Paralytic dorsiflexor muscle activities (DFA) and 10 m walking tests (10MWT) and timed up and go tests (TUG) were measured before and after intervention.Results:A significant increase in DFA was observed after intervention in the experimental and control groups (p<0.05) (experimental 886.6% for reference voluntary contraction (RVC), control 931.6% for RVC). TUG and 10MWT results showed significant reductions post-intervention in the experimental and control groups (experimental group -5.6 sec, control -4.8 sec; experimental group -3.1 sec, control, -3.9 sec; respectively). No significant intergroup difference was observed between changes in DFA or between changes in TUG and 10MWT results after intervention (p>.05).Conclusion:Strengthening exercise performed on non-paralytic dorsiflexion muscles had positive cross-training effects on paralytic dorsiflexor muscle activities, balance abilities, and walking abilities in patients with chronic stroke.     

Calf muscle work and segment energy changes in human treadmill walking

The relation between changes in potential and kinetic energy in a seven-segment model of the human body and the work of m. triceps surae was investigated in four subjects walking on a treadmill at speeds between 0.5 and 2.0 m/s. Segment energy levels were determined by means of tachometers attached with strings to various points on the subject's body. Muscle work was assessed by electromyogram to force processing. M. triceps surae is active during stance, first doing negative (eccentric) work and ending with a short period of positive (concentric) work at “push-off”. It turned out that in normal walking these muscles provide the major part of positive work for the initiation of swing at push-off. Only at large step lengths, when push-off starts well before contralateral heel contact, is there a minor pushing forward of the trunk. In the negative work phase, m. triceps surae seem to check the forward speed of the trunk. A related decrease of trunk kinetic energy is not present, however, but this may be obscured by the simultaneous action of m. quadriceps femoris and, in a later stage, by a transfer of energy from the decelerating contralateral (swing) leg to the trunk. Energy of the trunk segment shows a sharp decline in double stance and a more gradual increase in the first half of single stance. Evidence is given that this effect is due to quadriceps action in the knee flexion-extension movement during stance. The presented results are incorporated in a general picture of energy flows in human walking.     

Effects of bilateral medial knee osteoarthritis on intra- and inter-limb contributions to body support during gait

Patients with knee OA show altered gait patterns, affecting their quality of living. The current study aimed to quantify the effects of bilateral knee OA on the intra-limb and inter-limb sharing of the support of the body during gait. Fifteen patients with mild, 15 with severe bilateral knee OA, and 15 healthy controls walked along a walkway while the kinematic and kinetic data were measured. Compared with the controls, the patients significantly reduced their knee extensor moments and the corresponding contributions to the total support moment in the sagittal plane (p<0.05). For compensation, the mild OA group significantly increased the hip extensor moments (p<0.05) to maintain close-to-normal support and a more symmetrical inter-limb load-sharing during double-limb support. The severe OA group involved compensatory actions of both the ankle and hip, but did not succeed in maintaining a normal sagittal total support moment during late stance, nor a symmetrical inter-limb load-sharing during double-limb support. In the frontal plane, the knee abductor moments and the corresponding contributions to the total support moment were not affected by the changes in the other joints, regardless of the severity of the disease. The observed compensatory changes suggest that strengthening of weak hip muscles is essential for body support during gait in patients with knee OA, but that training of weak ankle muscles may also be needed for patients with severe knee OA.