Hip Joint Contact Loading and Muscle Forces During Running With a Transtibial Amputation

People with a transtibial amputation (TTA) are increasingly interested and participating in physical activities like running [1], which has numerous emotional, social, and physical health benefits [1–3]. However, limited access to appropriate prostheses is a barrier for many people with TTA who wish to run. Running-specific prostheses (RSPs) are designed for running and sprinting tasks with large capacity for energy storage and return. However, RSPs can be difficult to obtain as they are expensive and many insurance providers in the United States do not cover the cost of the devices. Therefore, if a person with TTA desires to run, he or she may do so using their daily-use prosthesis (DUP), which are stiffer devices that are not designed for highly dynamic activities like RSPs.

People with TTA have greater risks and rates of bilateral hip osteoarthritis (OA) compared to the age-matched population without TTA (14% versus 1%) [4–7], but there is conflicting evidence as to whether the intact or amputated side hip of people with TTA is at greater risk [5,7,8]. OA is a degenerative and painful disease characterized by the breakdown of articular cartilage [9]. The causality of OA in people with TTA is not fully understood, but asymmetric, large, and repeated joint contact forces are risk factors for the development of the disease in people with and without TTA [4,10–13]. For people with TTA, asymmetric movement biomechanics and muscle strength further increase OA risk factors [4,14]. Greater device energy storage and return in RSPs reduce hip joint kinetic asymmetry compared to DUPs [15]. Specifically, using compliant RSPs decreases amputated leg positive hip work (i.e., the time integral of positive hip joint power) compared to DUPs during running. Smaller positive hip work from the amputated leg reduces hip joint work asymmetry commonly observed in people with TTA. RSPs are generally more compliant compared to DUPs (RSP: 18–42 kN/m, DUP: 49–72 kN/m) [16,17], although these can vary greatly depending on manufacturer, design, and prescription. Repeated, large joint contact forces are also contributing factors to high rates of hip OA, and are likely affected by the presence of an amputation and device type, but have not yet been quantified in people with TTA during running [4,11,12,18]. Joint contact impulse, which is the time integral of joint contact force, is descriptive of cumulative load. Quantifying impulse over the gait cycle can reveal additional differences in loading than peak force alone [19].

During walking, the gluteus maximus, gluteus medius, vasti, and hamstrings are substantial contributors to hip joint contact forces (HJCFs) [20], in addition to the rectus femoris and iliopsoas providing smaller contributions. People with TTA have reduced amputated leg quadriceps strength and altered activity in the amputated leg gluteus medius and quadriceps during locomotion compared to people without TTA [21–23]. Thus, HJCFs are likely also different between people with and without TTA. Furthermore, characterizing the effect of device type on HJCFs is important for understanding potential effects of prosthesis choice on OA risk factors. Investigating HJCFs becomes especially important for people with TTA during activities with repeated and large magnitude joint loading, such as running. Joint contact forces are nearly impossible to measure directly, and joint intersegmental force computations do not include contributions from muscle forces [24,25]. However, muscle forces substantially contribute to joint contact forces [26]. Alternatively, musculoskeletal modeling and simulation can be used to quantify internal muscle force and HJCF results.

The purpose of this study was to determine how the presence of a unilateral transtibial amputation and the type of prosthesis (DUP versus RSP) affect peak HJCFs and impulses during running. We hypothesized that people with TTA would have greater peak HJCFs during the stance and swing phases and hip joint contact impulses over the running gait cycle compared to people without TTA due to documented hip joint kinetic compensations in this population [15,27]. We also hypothesized that using an RSP would reduce peak HJCFs and impulses in the amputated leg compared with using a DUP, due to prior evidence that use of an RSP reduces the magnitude of hip compensations (i.e., hip joint work) in the amputated leg. Similarly, given hip compensatory strategies for people with TTA, we expected that the amputated leg would have greater peak HJCF compared to the intact leg. To assist in interpretation of HJCF and impulse results, we also examined muscle forces that largely contribute to HJCFs, including the gluteus maximus, gluteus medius, vasti, hamstrings, rectus femoris, and iliopsoas.

Six people with unilateral TTA (5 M/1F, 1.78 m ± 0.057 m, 73.4 kg ± 8.14 kg, 30 ± 7 years) and six people without TTA (5 M/1F, 1.76 m ± 0.062 m, 76.7 kg ± 10.5 kg, 28 ± 5 years) (Table 1) provided written informed consent to participate in the experimental protocol approved by the Institutional Review Board. Participants were free from injury at the time of testing.

Participants ran on an instrumented treadmill (Bertec, Inc., Columbus, OH) at 3.5 m/s (7:40 min/mile) while GRFs were collected at 2000 Hz. Participants were instrumented with a set of 56 (people without TTA and people with TTA wearing DUPs) or 63 (people with TTA wearing RSPs) active markers. Markers on the DUP were placed similarly to the intact leg. The RSP had seven individual markers on the prosthesis to characterize deformation of the carbon-fiber device (four physical and three digitized) [15]. Kinematics were collected with an optoelectric motion capture system at 100 Hz (3DInvestigator, Northern Digital, Inc., ON, Canada).

Surface electromyography (EMG) signals were collected at 2000 Hz (Bortec, Inc., Canada) on a set of four muscles including the bilateral vastus lateralis, biceps femoris long head, rectus femoris, and medial gastrocnemius (bilateral for people without TTA and intact leg of people with TTA). EMG electrodes were placed parallel to the direction of the muscle fiber (Ag/AgCl, 22 mm interelectrode distance (Blue Sensor N-00-S, Ambu, Inc., Denmark)).

Musculoskeletal models were developed for each participant by scaling generic models in opensim v3.3 to match each participant's size and mass [28–30]. Each model had varying degrees-of-freedom (DOF) and number of Hill-type musculotendon actuators depending on the presence of TTA and the type of prosthesis (Fig. 1). Hill-type musculotendon actuators incorporated activation and deactivation dynamics as a simplified first-order dynamic model and contraction dynamics with force–length–velocity properties [31]. For each model, lumbar motion was modeled with three rotational DOFs relative to the pelvis, the upper arm had three rotational DOFs at the shoulder, and elbow and forearm motion were each modeled with 1 DOF revolute joints. Each DOF of the torso and arms were controlled with idealized torque actuators. Hip motion was modeled with three rotational DOFs, and the knee was modeled with one rotational DOF [28]. RSPs and DUPs were modeled with 4 and 1 sagittal plane DOF, respectively, and actuated with idealized torque actuators that reflected the net joint moments computed from inverse dynamics (Fig. 1). The ankle joint of the intact leg and legs of people without TTA was modeled as 1 DOF. Center of mass locations for the residual leg in models of people with TTA were adjusted using experimentally collected residual leg dimensions and defined as a right frustum [32]. Prosthesis mass was measured and applied to the model for each participant. Inertial properties of the RSP were estimated using previous data [33] according to the shape of RSP. DUP inertial properties were estimated using participant-specific DUP mass in combination with the geometry of a biological foot [34,35]. The pylon was modeled as a hollow tube.

Experimental GRFs and kinematics were low-pass filtered using a fourth order bidirectional Butterworth filter at 6 Hz. An inverse kinematics solution constrained to model DOFs was computed using a least squares optimization approach [36] (C-Motion, Germantown, MD). A residual reduction algorithm was used to ensure dynamic consistency between the inverse kinematics solution and the ground reaction forces [37]. A computed muscle control algorithm (CMC) was then used to determine individual muscle forces by minimizing the sum of squared muscle controls, which reproduced the running kinematics [38] (Fig. 1). Three gait cycles on each leg of each participant were simulated for a total of 108 running simulations.

Individual muscle forces determined from CMC were used to calculate the three-dimensional vector magnitude of the total hip joint contact force (e.g., Ref. [39]) over the stance and swing phases of running. HJCF and muscle force results were low-pass filtered using a bidirectional fourth order Butterworth with a 6 Hz cutoff frequency and normalized by body weight for each participant. Peak stance and swing phase values of HJCF of the intact and amputated legs of people with TTA using RSPs and DUPs and people without TTA were compared in addition to total gait cycle impulse, which was calculated as the time integral of HJCF from heel strike to the following heel strike. Peak muscle forces of the vasti, hamstrings, gluteus maximus, gluteus medius, rectus femoris, and iliopsoas were compared during the stance and swing phases. Vastus medialis, vastus intermedius, and vastus lateralis were combined to represent total vasti muscle force. Biceps femoris long head, semitendinosus, and semimembranosus were combined to represent total hamstrings muscle group force.

Electromyography signals were demeaned, bandpass filtered (20 Hz–500 Hz, bidirectional fourth order Butterworth), full-wave rectified, and low-pass filtered (6 Hz cutoff, bidirectional fourth-order Butterworth) to create a linear envelope [40]. EMG was normalized to the maximum activation of the muscle during the trial for each participant. Muscle activations from the running simulations were similarly, low-pass filtered (6 Hz cutoff, bidirectional fourth order Butterworth) and normalized to peak activation (Fig. 1). EMG and simulation muscle activations were visually compared to ensure similar onset and offset times [41]. Residual forces and moments from the residual reduction algorithm and kinematic tracking errors were also evaluated as a metric of simulation quality.

Differences in stance and swing phase peak HJCFs and muscle forces, as well as total gait cycle hip joint contact impulse, were each assessed in rstatisticalcomputingsoftware v 1.1.153 using three separate, two-factor analyses of variance (ANOVAs) [42] with factors of (1) side and (2) prosthesis condition. Separate ANOVAs were used to include within- and between-subjects comparisons, which has been previously used in similar research designs [43]. The side factor had two levels (amputated/left versus intact/right) and the prosthesis condition factor had two levels (RSP versus DUP, RSP versus No TTA, DUP versus No TTA) (Table 2). Tukey's correction for multiple comparisons was used for post hoc pairwise comparisons when significant main or interaction effects were found (ɑ = 0.05).

Simulations reproduced the inverse kinematics solution, with the largest root-mean-squared (RMS) rotational errors in arm pronation/supination (i.e., intact side arm pronation = 7.04 ± 13.32 deg), and the largest translational error in vertical translation of the pelvis for people wearing RSPs (1.52 ± 3.02 cm) (Appendix Table 6). Residual forces and moments at the pelvis were less than 1% overall bodyweight/bodyweight-height (Appendix Table 7). Experimental EMG and simulated muscle activations were similar, although simulated gastrocnemius and vastus lateralis peak activations were delayed compared to EMG signals (Fig. 2).

During the stance phase, there was a significant main effect of side, and a significant interaction of side × prosthesis condition when comparing people wearing RSPs and people without TTA, in addition to significant main effects of side and prosthesis condition when comparing people wearing RSPs to people wearing DUPs (p ≤ 0.04). The intact leg of people wearing RSPs had a greater peak HJCF compared to the amputated leg (p = 0.02) (Fig. 3, Table 3). Similarly, the intact leg of people wearing DUPs had a greater stance phase peak HJCF compared to the amputated leg (p = 0.05) (Fig. 3, Table 3). The amputated leg of people wearing DUPs had greater stance phase peak HJCF compared to the amputated leg of people wearing RSPs (p < 0.01) (Table 3). Overall stance phase peak HJCF was greater for people wearing DUPs compared to people wearing RSPs (p < 0.01) and intact legs had greater peak HJCFs compared to amputated legs, regardless of prosthesis type (p = 0.03).

During the swing phase, there were significant main effects of group for all ANOVAs (p ≤ 0.01) (Table 3). People with TTA wearing DUPs and RSPs had greater peak HJCF compared to people without TTA (p ≤ 0.01) (Fig. 3, Table 3) and people wearing DUPs had greater peak HJCFs compared to people wearing RSPs (p < 0.01). In addition, the intact leg of people wearing DUPs had greater swing phase peak HJCF compared to wearing RSPs (p < 0.01), and the amputated leg of people with TTA had greater peak HJCFs compared to the leg of people without TTA regardless of prosthesis type (p ≤ 0.03) (Fig. 3, Table 3). Finally, the intact leg of people wearing DUPs had greater swing phase peak HJCFs compared to the leg of people without TTA (p = 0.02) (Fig. 3, Table 3).

There was a significant main effect of side and an interaction effect of side × prosthesis condition when comparing people without TTA to people wearing RSPs (p < 0.04). There were also significant main effects of side and prosthesis condition when comparing people with TTA wearing DUPs and RSPs (p < 0.03). The intact leg of people wearing RSPs and DUPs had a greater hip joint contact impulse compared to the amputated leg (p = 0.02, p < 0.01, respectively) (Table 3) over the total gait cycle. In addition, people wearing DUPs had greater overall hip joint contact impulse compared to when they were wearing RSPs (group effect, p < 0.01) and the intact leg had greater hip joint contact impulse compared to the amputated leg, regardless of prosthesis type (p = 0.03).

There was a significant main effect of side for people without TTA compared to people using a DUP (p = 0.04) and a significant main effect of prosthesis condition when comparing people using a DUP to people without TTA and to people using an RSP (p ≤ 0.02) over stance. People using DUPs had greater stance phase leg gluteus maximus force compared to people without TTA and compared to people using RSPs (main effect, p ≤ 0.02) (Table 4). The amputated leg of people wearing DUPs had greater peak gluteus maximus force compared to people without TTA (p = 0.03) (Fig. 4, Table 4).

For swing phase, there was a significant prosthesis condition effect when comparing people wearing RSPs to people wearing DUPs (p = 0.02). People wearing DUPs had greater peak gluteus maximus force compared to people wearing RSPs (Table 5).

During stance phase, there were significant main effects of side and prosthesis condition when comparing people wearing RSPs to people without TTA (p = 0.02, p = 0.05) and a significant main effect of side when comparing people wearing RSPs and DUPs (p = 0.03). There was significant side × prosthesis condition interaction when comparing people wearing RSPs to people without TTA (p < 0.01) and to people wearing DUPs (p = 0.04). People wearing RSPs had greater stance phase peak hamstrings force compared to people without TTA (p = 0.048), and the intact leg had greater peak hamstrings force compared to the amputated leg (p < 0.01) and compared to people without TTA (p = 0.04). Finally, the intact leg had greater peak hamstrings force compared to the amputated leg regardless of prosthesis type (p = 0.03) (Fig. 4, Table 4).

During swing phase, there was a significant main effect of side and interaction of side × prosthesis condition when comparing people wearing RSPs to people without TTA and people wearing DUPs (p ≤ 0.04). The intact leg of people wearing RSPs had greater swing phase peak hamstrings force compared to the amputated leg (p < 0.01) (Table 5).

In stance, there was a significant effect of side when comparing people without TTA to people wearing DUPs (p = 0.02) (Table 3). The amputated leg of people wearing DUPs had greater stance phase peak iliopsoas muscle force compared to the intact leg (p = 0.04) (Fig. 4, Table 4).

There were no significant differences in iliopsoas peak force during swing phase.

There were no significant differences in peak gluteus medius or rectus femoris force for prosthesis condition or side during the stance phase.

During swing phase, there was a significant effect of prosthesis condition when comparing gluteus medius force of people without TTA and people with TTA. People with TTA had greater peak swing phase gluteus medius force when wearing DUPs (group effect, p < 0.01) and RSPs (p < 0.01). The amputated and intact leg of people with TTA had greater peak force compared to people without TTA (p ≤ 0.05).

For the rectus femoris during swing phase, there was a significant main effect of side when comparing people wearing RSPs to people without TTA and to people wearing DUPs (p ≤ 0.03) and a significant interaction of side × prosthesis condition when comparing people without TTA to people wearing RSPs. The intact leg had greater peak rectus femoris force during swing phase compared to the amputated leg (p < 0.01) (Table 5).

There were significant main effects of side (p < 0.01) when comparing all groups and a significant side × prosthesis condition interaction for people without TTA compared to people with TTA (p < 0.01) during stance. The intact leg of people wearing DUPs and RSPs had greater stance phase peak vasti force compared to the amputated leg (p < 0.01) and compared to people without TTA (p ≤ 0.04). The intact leg had greater overall force compared to the amputated leg regardless of prosthesis type (p < 0.01) (Fig. 4, Table 4).

During swing phase, there was a significant effect of prosthesis condition when comparing people without TTA to people wearing DUPs (p = 0.04) and a significant effect of side when comparing people wearing RSPs and DUPs (p = 0.03). There was a significant side × prosthesis condition effect when comparing people without TTA to people wearing RSPs (p = 0.05). People wearing DUPs had greater peak vasti force during swing phase compared to people without TTA and the intact leg of people with TTA had greater peak vasti force compared to the amputated leg (p = 0.03) (Table 5).

There were no differences in stance phase peak HJCF and total gait cycle impulse between people with and without TTA, contrary to our expectations. Using an RSP significantly reduced bilateral HJCF and impulse and resulted in smaller amputated leg peak HJCF compared to the use of a DUP during stance phase, which supported our hypothesis. However, people with TTA had greater intact leg stance phase peak HJCF and total gait cycle impulse compared to the amputated leg regardless of prosthesis type, which did not support our hypothesis. The intact leg of people wearing RSPs had a smaller peak HJCF during swing phase compared to the use of a DUP. Smaller amputated side vasti muscle force for people with TTA regardless of prosthesis type and greater amputated side gluteus maximus force when using a DUP both likely contributed to the resulting asymmetric HJCFs during stance phase. During swing phase, larger peak gluteus medius force for people with TTA likely contributed to greater peak HJCF for people with TTA compared to people without TTA. Muscular weakness, overloading of a joint, and asymmetric loading are all risk factors for the onset of OA [4,9]. Our results suggest that these risk factors are greater in people with TTA compared to people without TTA, especially given asymmetric running biomechanics [15].

Hip joint contact impulse is important to consider as cumulative loading is associated with the development of OA [44]. Greater impulsive joint loading indicates higher cumulative loading, as a high impulse over many gait cycles results in greater cumulative load. The intact leg of people with TTA had a greater total gait cycle impulse compared to the amputated leg (Table 3), which may be detrimental for long-term joint health [44]. Reduced overall hip joint contact impulse with the use of an RSP may be beneficial for long-term joint health during running. Changes in peak HJCF are driven by differences in muscle forces, which we also observed across conditions and legs.

During stance phase, the use of a DUP resulted in greater peak gluteus maximus force compared to people without TTA or people using RSPs (p ≤ 0.03). The gluteus maximus is important for providing body support and propulsion during the stance phase of running [45], and also contributes to positive hip joint work. Smaller propulsion from DUPs compared to RSPs may result in greater required gluteus maximus force as a result of limited energy storage and return capabilities of DUPs. Greater gluteus maximus force may also contribute to greater amputated side positive hip work when using DUPs compared to RSPs, which has been previously observed in these participants [15]. Smaller gluteus maximus peak activation in the amputated leg of people wearing RSPs compared to DUPs likely contributes to smaller peak amputated HJCF (Fig. 3), especially as gluteus maximus is a large contributor to HJCF during walking [20].

Gluteus medius is important for providing body support during the stance phase of running for people without TTA [45], is important for medial/lateral balance control, and is the largest contributor to HJCF during walking [20] for people without TTA. People with TTA characteristically favor weight-bearing on the intact leg, which results in associated poor amputated side hip abductor strength [46]. While no significant differences in gluteus medius peak force were found during stance phase, there was greater overall bilateral peak gluteus medius force during swing phase shortly after toe-off in people with TTA compared to people without TTA, regardless of prosthesis type. The role of the gluteus medius during the swing phase of running is not as well-understood as it generally has greater levels of activation during stance. However, different components of the gluteus medius control both hip internal and external rotation, in addition to hip abduction. Altered swing limb trajectories in people with TTA may be indicative of greater force requirements from this muscle.

The iliopsoas contributes to leg swing initiation by transferring power from the trunk to the amputated leg during walking [47] and may produce greater force in people using DUPs due to the lack of biarticular gastrocnemius function, and associated smaller propulsion compared to RSPs and biological legs (Fig. 4, Table 4). Observed greater peak force from the amputated side iliopsoas for people wearing DUPs (Fig. 4) may contribute to greater amputated side HJCF near the end of stance phase (Fig. 3).

The vasti are also important to consider in interpreting HJCF, even though they do not span the hip joint. The vasti muscles, which are uni-articular knee extensors, substantially contribute to HJCF through dynamic coupling by accelerating the knee and hip joints into extension [20,48]. Smaller vasti force output during stance phase likely contributes to the observed smaller stance phase peak HJCF and impulse in the amputated leg compared to the intact leg for people with TTA, particularly as the hip is extending during this period (Figs. 3 and 4). Supporting this notion, people with TTA have atrophied quadriceps resulting in smaller force output [49,50]. While we did not account for asymmetric muscle strength in the musculoskeletal model, the optimization predicted reduced force requirements from the amputated leg vasti muscles. Thus, reduced thigh muscle strength is consistent with our observed reduction in simulated amputated leg vasti force (Fig. 4, Table 4), and the large difference between amputated and intact legs, regardless of prosthesis type. This asymmetry in vasti force likely contributes to peak HJCF and impulse asymmetry (Figs. 3 and 4).

People with TTA included in our study used their own DUPs and RSPs, which was intended to capture the habituated running mechanics of people with TTA. However, variation due to prosthesis design is important to consider and is a potential limitation to this study. For example, participant 4 used a DUP (Table 1) that was much more responsive than a traditional DUP design as it allowed for greater deflection, thus storing and returning more energy. The use of this device decreased amputated leg peak HJCF relative to the intact leg (Fig. 5). In contrast, participant 2 used a DUP with a geometry that had a smaller deflection range (Ottobock Advantage DP2) and resulted in amputated leg peak HJCF that was much greater than the intact leg. Further, individual experience may play a significant role in running mechanics as well as device choice. For example, in addition to a more responsive DUP, Participant 4 was a consistent runner, with frequent access to prosthetic adjustments. Individual differences in running biomechanics and prosthesis choice likely resulted in variable results in magnitude and timing of hip joint loading and muscle forces across participants. The effect of specific device characteristics and running experience on running biomechanics should be further investigated in future work.

Our overall loading magnitude results (peak HJCF: 8-11 times body weight (BW)) were in the range of multiple previous experimental and modeling studies, which have examined HJCF. Specifically, in vivo HJCF measurements from instrumented hip implants during running have ranged between 5 and 6 times bodyweight [51,52]. However, results from instrumented implants were collected at a lower running velocity (2.2 m/s) and may be influenced by surgical approaches that alter gait biomechanics [53]. Musculoskeletal modeling has also been used to examine HJCF during running at 3.33 m/s using a static optimization approach with peak loads estimated at approximately 9 times bodyweight during the stance phase and approximately 5 times bodyweight during swing phase [54], similar to our results.

While model muscle activations and EMG were similar, we observed timing shifts in the vastus lateralis and gastrocnemius (Fig. 2), and this result should be acknowledged as a limitation. Changes in timing between model muscle activations and EMG may be due to limitations both in experimental muscle activity measurement, modeling of muscle activation and deactivation dynamics, and the movement optimization algorithm. EMG signals are sensitive to sensor placement and movement artifact. We followed recommended SENIAM placement guidelines [55]; however, cross-talk between muscles may have occurred. Further, EMG signals reflect activity at the muscle belly, rather than the muscle activation occurring along the length of the muscle. Timing shifts may also be a result of differences between the simulations and the participants in electromechanical delay. Electromechanical delay in vivo has been observed at levels exceeding 100 ms [56]. In addition, the muscle recruitment problem that is solved within CMC may result in different muscle activity results across agonist muscles. For example, semitendinosus, semimembranosus, and the biceps femoris long head are biarticular hip extensors, and thus may be activated to varying levels in a group, which may be different than the singular EMG signal that is experimentally collected. Even with some shifts in timing, overall shape of the simulated muscle activation and EMG signals was comparable. Further, the simulations tracked the experimental running kinematics well with small residual forces and kinematic tracking errors, giving us confidence in simulation results. In addition, peak HJCFs from our simulations were similar to previously collected experimental and simulated values.

Musculoskeletal modeling and simulation are useful to estimate in vivo loading and its implications for injury risk. Incorporating muscle forces as a part of estimating hip joint contact forces is important given their substantial contributions to joint loading. In a previous study of knee joint contact loading, muscular forces added approximately two more bodyweight multiples of force during walking compared to simulations only accounting for intersegmental loading [57]. These results suggest that joint contact load differences at the hip for people with TTA are largely affected by changes in muscular contraction, rather than changes in prosthesis mass. Muscles adapt to different device and skeletal characteristics, and it is challenging to separate the effects of device alone from the effects of muscles and their associated joint kinetics that have adapted to using a specific device. However, specifying relative contributions to HJCF from muscular load, prostheses, and skeletal characteristics may be an important area of future work to inform device design and movement training.

Runners with TTA using RSPs had smaller overall peak HJCFs during the stance phase and smaller intact leg peak HJCFs during the swing phase compared to when using DUPs. People with TTA also had greater peak HJCFs during swing compared to people without TTA. The presence of an amputation resulted in greater stance phase intact leg peak HJCF and impulse over the gait cycle compared to the amputated leg and greater overall peak HJCFs during swing compared to people without TTA, which may be detrimental to long-term hip joint health. However, the use of an RSP decreased overall hip joint force impulse compared to the use of a DUP during the gait cycle, which may be especially important for repeated, high impact activities such as running. The use of an RSP also resulted in smaller peak HJCFs just after toe-off in the amputated leg compared to the use of a DUP, which may be a result of the lighter RSP. HJCFs are dependent on muscle force contributions from leg muscles. The use of a DUP resulted in greater amputated leg gluteus maximus peak force compared to people without TTA and compared to people using RSPs during stance, which may reflect the lack of energy storage and return of the DUP compared to the RSP. Swing phase peak gluteus medius forces from both legs were greater for people with TTA compared to people without TTA. Vasti atrophy in the amputated leg likely also plays an important role in the asymmetric HJCF observed for people with TTA regardless of the prosthesis type they use.

This work was supported by the Office of the Assistant Secretary of Defense for Health Affairs through the Orthotics and Prosthetics Outcomes Research Program under Award No. W81XWH-15-1-0518. Opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the Department of Defense. Additional support was provided by the 2018 American Society of Biomechanics Graduate Student Grant-In-Aid.