Secondary Logo

Share this article on:

Is Running Better than Walking for Reducing Hip Joint Loads?

SCHACHE, ANTHONY G.1,2; LIN, YI-CHUNG1; CROSSLEY, KAY M.2; PANDY, MARCUS G.1

Medicine & Science in Sports & Exercise: November 2018 - Volume 50 - Issue 11 - p 2301–2310
doi: 10.1249/MSS.0000000000001689
APPLIED SCIENCES

Purpose Knowledge of hip biomechanics during locomotion is necessary for designing optimal rehabilitation programs for hip-related conditions. The purpose of this study was to: 1) determine how lower-limb muscle contributions to the hip contact force (HCF) differed between walking and running; and 2) compare both absolute and per-unit-distance (PUD) loads at the hip during walking and running.

Methods Kinematic and ground reaction force data were captured from eight healthy participants during overground walking and running at various steady-state speeds (walking: 1.50 ± 0.11 m·s−1 and 1.98 ± 0.03 m·s−1; running: 2.15 ± 0.18 m·s−1 and 3.47 ± 0.11 m·s−1). A three-dimensional musculoskeletal model was used to calculate the HCF as well as lower-limb muscular contributions to the HCF in each direction (posterior–anterior; inferior–superior; lateral–medial). The impulse of the resultant HCF was calculated as well as the PUD impulse (BW·s·m−1) and PUD force (BW·m−1).

Results For both walking and running, HCF magnitude was greater during stance than swing and was largest in the inferior–superior direction and smallest in the posterior–anterior direction. Gluteus medius, iliopsoas, and gluteus maximus generated the largest contributions to the HCF during stance, whereas iliopsoas and hamstrings generated the largest contributions during swing. When comparing all locomotion conditions, the impulse of the resultant HCF was smallest for running at 2.15 m·s−1 with an average magnitude of 2.14 ± 0.31 BW·s, whereas the PUD impulse and force were smallest for running at 3.47 m·s−1 with average magnitudes of 0.95 ± 0.18 BW·s·m−1 and 1.25 ± 0.24 BW·m−1, respectively.

Conclusions Hip PUD loads were lower for running at 3.47 m·s−1 compared with all other locomotion conditions because of a greater distance travelled per stride (PUD impulse) or a shorter stride duration combined with a greater distance travelled per stride (PUD force).

1Department of Mechanical Engineering, University of Melbourne, Melbourne, Victoria, AUSTRALIA; and

2La Trobe Sport and Exercise Medicine Research Centre, La Trobe University, Bundoora, Victoria, AUSTRALIA

Address for correspondence: Anthony G. Schache, B. Physio (Hons), Ph.D., Department of Mechanical Engineering, University of Melbourne, Victoria 3010 Australia; E-mail: anthonys@unimelb.edu.au.

Submitted for publication December 2017.

Accepted for publication June 2018.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.acsm-msse.org).

Knowledge of hip biomechanics during walking and running is necessary for designing optimal rehabilitation programs for hip-related conditions. Differences in hip biomechanics across these two modes of locomotion have been reported in studies using inverse dynamics–based analyses and/or recordings of muscle electromyographic signals (1–3). For example, when switching from fast walking to jogging at the same speed, there is a significant decrease in both the magnitude of the peak hip flexor moment during late stance (2) as well as the mean activation levels of the hip-spanning psoas and iliacus muscles (1). These observations suggest that loads experienced by the hip joint itself are likely to be sensitive to a change in locomotion mode. To improve our understanding of hip biomechanics during walking and running, investigations involving detailed analyses of joint contact forces would be beneficial.

The hip contact force (HCF) during walking and slow running has been recorded in vivo using telemetered joint implants (4–7). Unfortunately, these data are restricted to a small number of older people who have undergone surgical joint replacement procedures. This approach is not practicable for obtaining data on young healthy adults. Computer-based musculoskeletal modeling offers an alternative noninvasive approach for calculating the HCF, which has been demonstrated to be capable of yielding reasonable estimates, at least for walking (8–10). Even though studies have used musculoskeletal models to predict the HCF for various speeds of walking and running in healthy adult participants (11–16), analyses to date have been relatively limited. Most studies have only considered the resultant HCF (11,13,14,16) or have only investigated one mode of locomotion (11,12,14,15). Further research is therefore required to provide a more complete analysis of the HCF during walking and running at a range of speeds.

Muscles are the main contributors to the mechanical loading of joints (17). Despite their influential role, very little is known about the way in which muscles contribute to the HCF during locomotion. To our knowledge, only two studies have systematically investigated muscle contributions to the HCF (18,19). Correa et al. (18) evaluated walking at 1.4 m·s−1, and analyses were based on outputs from a previously published dynamic optimization solution (20), whereas the analyses of Pandy and Andriacchi (19) were based on experimental data collected from five healthy adults for walking at 1.4 m·s−1 and running at 3.4 m·s−1. Both studies found hip-spanning muscles such as the gluteus maximus (GMAX), gluteus medius (GMED), iliopsoas (ILPSO), and hamstrings (HAMS) to be the dominant contributors. Further research is required to more comprehensively understand how individual muscle contributions to the HCF are influenced by a change in both the mode and speed of locomotion.

Given that walking and running are popular forms of exercise in healthy adults, it is also important to consider how hip joint loading varies across these two locomotion modes. It is possible that the load accumulated at the hip for a given distance travelled may be less for running compared to walking. Even though the peak resultant HCF has been shown to be approximately 30% higher for running at 12 km·h−1 (3.3 m·s−1) compared with running and walking at 6 km·h−1 (1.7 m·s−1) (13), the accumulated load could be lower for running at 3.3 m·s−1 because of a shorter stride duration and a smaller number of strides required to travel a given distance. Per-unit-distance (PUD) loads at the knee for walking and running have been investigated previously (21); however, to our knowledge, no previous studies have investigated PUD loads at the hip across a range of common locomotion conditions.

The aim of the current study was twofold: first, to determine how a change in mode and speed of locomotion influences the contributions of individual lower-limb muscles to the HCF; and second, to compare both absolute and PUD loads at the hip joint during walking and running at a range of speeds. We hypothesised that: (i) the ipsilateral hip muscles (specifically, GMED, GMAX, ILPSO, and HAMS) would be the dominant contributors to the HCF independent of locomotion mode and speed; and (ii) PUD loads at the hip would be less for running compared to walking.

Back to Top | Article Outline

METHODS

Experimental gait data

Eight participants (four males and four females; age, 28.1 ± 8.4 yr; height, 1.75 ± 0.06 m; and body mass, 71.3 ± 7.0 kg) gave informed consent to participate in the study. All participants were experienced runners free of any recent or pre-existing musculoskeletal injury that was likely to affect their ability to walk and run. Ethical approval for the study was obtained from the relevant institutional human research ethics committee (University of Melbourne ref number: 0830526).

Participants were asked to walk and run over ground at a range of designated steady-state speeds (walking, 1.5 and 2.0 m·s−1; running, 2.0 and 3.5 m·s−1). Locomotion speed for each trial was monitored using timing gates (Speedlight Telemetry Timing; Swift Performance Equipment, Walcol, Australia) positioned 20 m at each end of the calibrated measurement volume. Participants were provided with verbal feedback after each trial to achieve the prescribed locomotion speeds. For each locomotion condition, participants performed repeated trials until a single trial was obtained whereby the locomotion speed was within 10% of the desired speed. Adequate rest was given between locomotion conditions to prevent fatigue. Participants used their preferred foot strike pattern and self-selected their stride frequency and stride length.

Three-dimensional kinematic data were collected using a 22-camera, video-based, motion analysis system (VICON Motion Systems Ltd., UK) sampling at a rate of 250 Hz. Small reflective markers (14 mm in diameter) were placed at specific locations on the upper and lower limbs as described previously by Dorn et al. (22). Ground reaction force data were collected from eight force plates (Kistler Instrument Corp., Amherst, NY) embedded in the laboratory floor and sampling at a frequency of 1500 Hz. Ground reaction force data were filtered using a fourth-order, low-pass, Butterworth filter with a cut-off frequency of 40 Hz.

Back to Top | Article Outline

Musculoskeletal model

We used a three-dimensional musculoskeletal model to calculate joint kinematics, joint kinetics and muscle forces based on the experimental data. The skeleton was represented as a generic 10-segment, 23-degree-of-freedom (DOF) mechanical linkage, the full details of which have been previously described by Anderson and Pandy (20,23). The pelvis was modeled as a single rigid body with six DOF; the hip was modeled as a three DOF ball-and-socket joint, and the knee as a single DOF hinge joint. Two segments were used to model each foot: a hindfoot segment and a toes segment. The lower leg was connected to the hindfoot (i.e., the ankle-subtalar complex) using a two DOF universal joint that comprised of two nonintersecting hinge joints. The toes articulated with the hindfoot via a single DOF hinge joint. The upper body (head, arms and torso) was represented as a single segment that articulated with the pelvis via a ball-and-socket back joint. The lower limbs and trunk were actuated by 54 muscle–tendon units. Each lower-limb was actuated by 24 muscle–tendon units, whereas six abdominal and back muscle–tendon units actuated relative movements of the pelvis and upper body. Each muscle–tendon unit was modeled as a Hill-type muscle with passive and active muscle fascicles connected in series with a compliant tendon (24). The force generating properties and attachment sites of the muscle–tendon units were based on data from Anderson and Pandy (23).

The generic musculoskeletal model was scaled to each participant’s anthropometric measurements. A static standing calibration trial with all markers in situ was captured for each participant prior to the commencement of the walking and running trials. This static trial was used to determine a scaling factor for each body segment by calculating the relative distance between a pair of markers measured on the segment and the corresponding virtual pair located in the model. An inverse kinematics analysis was used to calculate a set of joint angles for each time instant, whereby the sum of the squares of the differences between the experimentally recorded markers and the corresponding virtual markers defined in the model was minimised (25). A fourth-order, low-pass, Butterworth filter with a cutoff frequency of 7 Hz was used to smooth the joint angles before further processing. Net joint torques generated about the back, hip, knee and ankle joints were computed using a standard inverse dynamics approach. The joint torques were then decomposed into individual muscle forces using a static optimization algorithm that minimized the sum of all muscle activations squared subject to each muscle’s force–length–velocity properties (26). A nonlinear performance criterion was selected because it has the potential to allow more synergistic and antagonistic muscle activity compared to linear criteria (27). Furthermore, it has been shown to produce muscle forces with time histories similar to experimentally recorded electromyographic signals during locomotion conditions consistent with those considered in the present study (22,26,28). A pseudoinverse force decomposition method (29) was used to compute the contributions of all lower-limb muscle forces to the vertical, fore-aft, and mediolateral ground reaction forces. An individual muscle force and its ground reaction force contribution were applied to the model to determine both the resultant HCF as well as each muscle’s contribution to the three components of the HCF (18). All force data were normalized to each participant’s body weight (BW). Our convention defined the HCF as the force directly applied to the hip joint (femoral head) by the pelvis. It was expressed in the femoral reference frame whereby positive represented a force directed in the inferior, posterior and lateral directions (Fig. 1).

FIGURE 1

FIGURE 1

Back to Top | Article Outline

Data analysis

Data were analyzed for a single “test” leg for each participant (right leg for 4 participants; left leg for four participants). For each participant, a single stride cycle (ipsilateral initial foot contact to the next ipsilateral initial foot contact) for the designated “test” leg for each locomotion condition was extracted for analysis. Stride time was measured as the duration in time from ipsilateral initial foot contact to the next ipsilateral initial foot contact for the stride cycle of interest. Locomotion speed was then calculated by determining the horizontal distance that the pelvis segment travelled during the stride cycle of interest, and dividing this distance by the measured stride time.

All three components of the HCF were time normalized (0%–100% of the stride cycle) and averaged across participants to generate group mean curves. Similarly, muscle contributions to each component of the HCF were time normalized (0%–100% of the stride cycle) and averaged across participants to generate group mean curves (see Figures, Supplemental Digital Content 1–4, Contributions of individual lower-limb muscles to each component of the HCF during walking at 1.50 ± 0.11 m·s−1 (Supplemental Digital Content 1, http://links.lww.com/MSS/B314), walking at 1.98 ± 0.03 m·s−1 (Supplemental Digital Content 2, http://links.lww.com/MSS/B315), running at 2.15 ± 0.18 m·s−1 (Supplemental Digital Content 3, http://links.lww.com/MSS/B316) and running at 3.47 ± 0.11 m·s−1 (Supplemental Digital Content 4, http://links.lww.com/MSS/B317)). For each component of the HCF as well as the resultant force, data were integrated over the duration of the stride cycle to calculate the impulse (BW·s). Similarly, the contributions of individual lower-limb muscles to each component of the HCF were integrated over time to calculate the magnitude of their impulse (BW·s), but in this instance, the impulse was calculated for the stance and swing phases separately given the anticipated differences in hip muscle activations between these phases. The contributions of individual lower-limb muscles to each component of the HCF were analyzed descriptively.

For each locomotion condition, the impulse of the resultant HCF was divided by the horizontal distance that the pelvis segment travelled for the stride cycle of interest to calculate the PUD impulse expressed in body weight-seconds per meter. The PUD impulse was then divided by the measured stride time to calculate the PUD force expressed in bodyweight per meter as per Miller et al. (21). The PUD impulse allows locomotion conditions to be compared where distance has been normalized, but these data are still influenced by differences in stride time among the locomotion conditions. The PUD force allows locomotion conditions to be compared where both distance and time have been normalized; thus, these data essentially represent the average HCF per stride taking into account the distance travelled per stride, which is usually longer for running than walking (21).

One way repeat-measures ANOVA tests were used to determine whether locomotion condition significantly influenced the following parameters for the resultant HCF: (a) impulse; (b) PUD impulse; and (c) PUD force. Effect size was estimated using partial η 2, where >0.25 was considered a large effect. If a significant main effect for locomotion condition was found (P < 0.05), post hoc pairwise comparisons were used to determine differences between the various locomotion conditions. The level of significance for the post hoc comparisons was adjusted to P < 0.008.

Back to Top | Article Outline

RESULTS

The average speeds for the two walking conditions were 1.50 ± 0.11 m·s−1 and 1.98 ± 0.03 m·s−1, whereas the average speeds for the two running conditions were 2.15 ± 0.18 m·s−1 and 3.47 ± 0.11 m·s−1. When expressed in the femoral reference frame, the contact force on the femoral head was primarily directed inferiorly, laterally and posteriorly throughout the stride cycle (Fig. 1). Irrespective of locomotion condition, the magnitude of the HCF was greater during stance than swing, and it was largest in the inferior–superior direction and smallest in the posterior–anterior direction.

The profile of the HCF during stance for walking was characterized by a biphasic pattern, with both peaks having a similar magnitude (Fig. 2). There was little change in the magnitude of the HCF with faster walking. The average magnitude (and timing as a percentage of stride cycle) of the first peak in the resultant HCF profile was 5.53 ± 1.18 BW (16.48% ± 2.84%) and 5.65 ± 1.42 BW (13.31% ± 3.41%) for walking at 1.50 and 1.98 m·s−1, respectively, whereas the average magnitude (and timing as a percentage of stride cycle) of the second peak in the resultant HCF profile was 5.23 ± 0.86 BW (46.89% ± 2.55%) and 5.51 ± 1.34 BW (48.14% ± 1.86%) for walking at 1.50 and 1.98 m·s−1, respectively. The profile of the HCF during stance for running increased immediately following foot-strike, remained elevated throughout stance, and then declined prior to toe-off (Fig. 2). The magnitude of the HCF increased with faster running. The average magnitude (and timing as a percentage of stride cycle) of the peak in the resultant HCF profile that occurred around mid-stance was 9.53 ± 1.90 BW (15.61% ± 2.24%) for running at 2.15 m·s−1 and 11.03 ± 3.53 BW (12.32% ± 3.10%) for running at 3.47 m·s−1.

FIGURE 2

FIGURE 2

The net impulse of the HCF over the stride cycle was influenced by a change in locomotion speed and mode (Fig. 3). Average values ranged from 1.90 ± 0.27 BW·s (running at 2.15 m·s−1) to 2.51 ± 0.26 BW·s (walking at 1.50 m·s−1) in the inferior–superior direction, 0.68 ± 0.14 BW·s (running at 2.15 m·s−1) to 0.89 ± 0.29 BW·s (running at 3.47 m·s−1) in the lateral–medial direction, and 0.53 ± 0.15 BW·s (running at 2.15 m·s−1) to 0.76 ± 0.28 BW·s (running at 3.47 m·s−1) in the posterior–anterior direction. The average net impulse of the HCF decreased by 10.8%, 19.3%, and 21.6% in the inferior–superior, lateral–medial and posterior–anterior directions, respectively, when increasing walking speed from 1.50 to 1.98 m·s−1. When switching from walking at 1.98 m·s−1 to running at 2.15 m·s−1, the average net impulse of the HCF decreased by 15.2% in the inferior–superior direction, with minimal change evident in the lateral–medial and posterior–anterior directions; it then increased by 11.1%, 30.9%, and 43.4% in the inferior–superior, lateral–medial, and posterior–anterior directions, respectively, with faster running.

FIGURE 3

FIGURE 3

The largest contributions to the HCF during the stance phase of both walking and running were generated by GMED, ILPSO, and GMAX (Table 1). During walking, GMED contributed to the first peak of the HCF in all three directions and to the second peak of the HCF in the inferior–superior and lateral–medial directions (see Figures, Supplemental Digital Content 1 and 2, Contributions of individual lower-limb muscles to each component of the HCF during walking at 1.50 ± 0.11 m·s−1 (Supplemental Digital Content 1, http://links.lww.com/MSS/B314) and walking at 1.98 ± 0.03 m·s−1 (Supplemental Digital Content 2, http://links.lww.com/MSS/B315)). ILPSO contributed to the second peak of the HCF in the inferior–superior and posterior–anterior directions. Finally, GMAX contributed to the first peak of the HCF in the inferior–superior and lateral–medial directions. During running, GMED contributed to the HCF in all three directions, ILPSO mainly contributed to the HCF in the inferior–superior and posterior–anterior directions, and GMAX contributed to the HCF in the inferior–superior and lateral–medial directions (see Figures, Supplemental Digital Content 3 and 4, Contributions of individual lower-limb muscles to each component of the HCF during running at 2.15 ± 0.18 m·s−1 (Supplemental Digital Content 3, http://links.lww.com/MSS/B316) and running at 3.47 ± 0.11 m·s−1 (Supplemental Digital Content 4, http://links.lww.com/MSS/B317)). Locomotion condition was found to influence the magnitude of the impulse delivered by the various muscles to the HCF during stance (Table 1). For example, when switching from walking at 1.98 m·s−1 to running at 2.15 m·s−1, the impulse delivered by ILPSO in the inferior–superior and posterior–anterior directions decreased by 60% and 54%, respectively.

TABLE 1

TABLE 1

The largest contributions to the HCF during swing were generated by ILPSO and HAMS (Table 2). ILPSO contributed to the HCF in the posterior–anterior and inferior–superior directions during terminal swing for walking and during early swing for running (see Figures, Supplemental Digital Content 1–4, Contributions of individual lower-limb muscles to each component of the HCF during walking at 1.50 ± 0.11 m·s−1 (Supplemental Digital Content 1, http://links.lww.com/MSS/B314), walking at 1.98 ± 0.03 m·s−1 (Supplemental Digital Content 2, http://links.lww.com/MSS/B315), running at 2.15 ± 0.18 m·s−1 (Supplemental Digital Content 3, http://links.lww.com/MSS/B316) and running at 3.47 ± 0.11 m·s−1 (Supplemental Digital Content 4, http://links.lww.com/MSS/B317)). HAMS contributed to the HCF in the inferior–superior direction during terminal swing for both walking and running. The magnitude of the impulse delivered by the ipsilateral hip muscles to the HCF during swing increased with each incremental locomotion condition (Table 2).

TABLE 2

TABLE 2

A significant main effect for locomotion condition was found for the impulse of the resultant HCF (P < 0.001; Partial η 2 = 0.640), the PUD impulse (P < 0.001; Partial η 2 = 0.944) and the PUD force (P < 0.001; Partial η 2 = 0.787) (Fig. 4). The impulse of the resultant HCF (Fig. 4; left panel) for walking at 1.50 m·s−1 was significantly greater than that for walking at 1.98 m·s−1 (2.79 ± 0.32 BW·s vs 2.45 ± 0.36 BW·s; P = 0.001) and running at 2.15 m·s−1 (2.79 ± 0.32 BW·s vs 2.14 ± 0.31 BW·s; P < 0.001). Also, the impulse of the resultant HCF for running at 2.15 m·s−1 was significantly less than that for running at 3.47 m·s−1 (2.14 ± 0.31 BW·s vs 2.51 ± 0.51 BW·s; P = 0.003). The PUD impulse (Fig. 4; middle panel) significantly differed for all comparisons (P < 0.001), except between walking at 1.98 m·s−1 and running at 2.15 m·s−1 (1.34 ± 0.21 BW·s·m−1 vs 1.27 ± 0.19 BW·s·m−1; P = 0.275). The PUD force (Fig. 4; right panel) for walking at 1.50 m·s−1 was significantly greater than that for walking at 1.98 m·s−1 (1.65 ± 0.22 BW·m−1 vs 1.44 ± 0.26 BW·m−1; P = 0.001) and running at 3.47 m·s−1 (1.65 ± 0.22 BW·m−1 vs 1.25 ± 0.24 BW·m−1; P < 0.001). Also, the PUD force for running at 2.15 m·s−1 was significantly greater than that for running at 3.47 m·s−1 (1.62 ± 0.26 BW·m−1 vs 1.25 ± 0.24 BW·m−1; P < 0.001).

FIGURE 4

FIGURE 4

Back to Top | Article Outline

DISCUSSION

This study used a musculoskeletal modeling approach to determine the HCF during walking at average steady-state speeds of 1.50 ± 0.11 m·s−1 and 1.98 ± 0.03 m·s−1 as well as running at average steady-state speeds of 2.15 ± 0.18 m·s−1 and 3.47 ± 0.11 m·s−1. Our overall goal was to understand how these locomotion conditions affected the contributions of individual lower-limb muscles to the HCF as well as the PUD loads experienced by the hip joint. We found that: 1) the major hip-spanning muscles (GMED, ILPSO, and GMAX) provided the largest contributions to the HCF during walking and running; 2) a change in locomotion mode affected the contribution from ILPSO most of all, with the magnitude of the impulse generated by ILPSO decreasing by more than 50% when switching from walking at 1.98 m·s−1 to running at 2.15 m·s−1; and 3) both PUD impulse and PUD force were lower for running at 3.47 m·s−1 compared to all other locomotion conditions.

Previous studies have recorded the HCF during locomotion in vivo using telemetered joint implants (4–7). Overall, when comparing the calculated HCF from the present study with published in vivo data, curve profiles share a similar shape/timing, but our peak magnitudes are generally higher, especially for running. Our peak magnitudes for the resultant HCF averaged 5.7 BW for walking at 1.50 m·s−1 and 9.5 BW for running at 2.15 m·s−1. In comparison, Rydell (7) recorded in vivo data from a 56-yr-old female patient 26 wk after surgery and reported peak resultant HCF magnitudes of 3.3 BW for walking at 1.4 m·s−1 and 4.3 BW for running at 2.5 m·s−1. Kotzar et al. (6) recorded in vivo data from a 72-yr-old male patient 8 wk after surgery and reported peak resultant HCF magnitudes ranging from 3.2 to 3.9 BW for walking at 1.8 m·s−1. Bergmann et al. (5) recorded in vivo data from an 82-yr-old male patient 30 months after surgery and reported peak resultant HCF magnitudes of 4.7 BW for walking at 1.4 m·s−1 and 5.5 BW for running at 1.9 m·s−1. Finally, Bergmann et al. (4) recorded in vivo data from three male patients (between 51 and 61 yr of age) at time points ranging from 11 to 14 months after surgery and reported an average peak resultant HCF magnitude of 2.5 BW for walking at 1.5 m·s−1. Based on these results, we acknowledge that our model-based predictions are likely to overestimate the true magnitude of the HCF to some extent. However, when comparing results from the present study with published in vivo data numerous factors must be kept in mind. First, in vivo data have been recorded from an artificial hip joint and it is unknown how well such data resemble the contact mechanics of a normal joint (17). Second, in vivo data have only been obtained from older participants (≥51 yr of age), some of whom had coexisting morbidities. Third, in many instances in vivo data have been recorded from participants less than 12 months after major surgery. Thus, muscle activation patterns and the mechanics of moment production by the muscles are likely to be altered (17). Fourth, in vivo data display some variability in profile and magnitude, including within-participant stride-to-stride variability (7,8) as well as between-participant variability (4). Finally, to our knowledge, in vivo data for running at speeds beyond 2.5 m·s−1 have not been published. For these reasons, we believe that differences between our model-based predictions and published in vivo data are to be expected and do not invalidate the main conclusions from this study.

The way in which muscles contribute to knee joint loading during locomotion has been well investigated (19,30–32); however, the same cannot be said for the hip joint. Correa et al. (18) calculated the HCF for walking at 1.35 m·s−1 and found GMED contributed most to the HCF in the inferior–superior and lateral–medial directions, while ILPSO contributed most in the posterior–anterior direction. This result is consistent with our data for walking at 1.50 m·s−1 (Table 1 and Figure, Supplemental Digital Content 1, Contributions of individual lower-limb muscles to each component of the HCF during walking at 1.50 ± 0.11 m·s−1, http://links.lww.com/MSS/B314). Pandy and Andriacchi (19) calculated the HCF for running at 3.4 m·s−1 and found GMED and GMAX to be the dominant contributors during stance, consistent with findings from this study (Table 1 and Figure, Supplemental Digital Content 4, Contributions of individual lower-limb muscles to each component of the HCF during running at 3.47 ± 0.11 m·s−1, http://links.lww.com/MSS/B317). However, they found the contribution from rectus femoris during stance to be greater than that from ILPSO, whereas the opposite was the case in the present study. The most notable effect when switching from walking at 1.98 m·s−1 to running at 2.15 m·s−1 was a decrease of approximately 55% in the stance phase contribution from ILPSO to the HCF in the posterior–anterior and inferior–superior directions (Table 1 and Figures, Supplemental Digital Content 2 and 3, Contributions of individual lower-limb muscles to each component of the HCF during walking at 1.98 ± 0.03 m·s−1 (Supplemental Digital Content 2, http://links.lww.com/MSS/B315) and running at 2.15 ± 0.18 m·s−1 (Supplemental Digital Content 3, http://links.lww.com/MSS/B316)). This result is consistent with results of Andersson et al. (1), who recorded psoas and iliacus electromyographic signals via intramuscular fine wire electrodes and found muscle activation intensity to be about twice as high for walking compared to running at a speed of 3 m·s−1. It therefore seems that switching from walking to running reduces the mechanical load on the major hip flexors during stance, which could be an initiating factor for the transition in locomotion mode (2).

Evidence is available indicating that the impulse of a given load may exhibit a stronger relationship with pathology than purely instantaneous parameters such as the peak magnitude (33,34). The advantage of impulse is that it takes into account both the time-varying magnitude and the duration of application of the load. Studies have also calculated the product of the impulse and the number of loading cycles completed per day to obtain an estimate for the total (or cumulative) daily exposure to joint load (35,36). For example, Tateuchi et al. (36) recorded baseline gait analysis data from 50 female patients with secondary hip osteoarthritis (OA) walking at a self-selected speed. They found a higher cumulative hip joint load in the frontal plane (specifically, the impulse of the hip abductor moment multiplied by the average number of steps per day as recorded by a pedometer) to be a significant independent predictor of radiographic progression of hip OA over a 12 month period. No association was found for the peak magnitude of the hip abductor moment. We therefore believe that our decision to use the impulse as the primary metric for quantifying biomechanical load in the present study was well justified from a clinical perspective.

Is running more likely to initiate joint articular cartilage disease than walking? This question has recently been explored in a study by Miller et al. (21) focusing on the knee joint. They found the knee PUD force to be similar for walking at 1.45 m·s−1 (0.75 ± 0.08 BW·m−1) and running at 3.17 m·s−1 (0.80 ± 0.14 BW·m−1). It was proposed that this result offered a biomechanical explanation for why most runners do not develop knee OA (21,37). Interestingly, current evidence would suggest that recreational running is not a risk factor for developing hip OA either (38,39), and it may even provide a protective effect relative to a more sedentary lifestyle (38). Despite being higher than equivalent loads for the knee, we found the hip PUD force for running at 2.15 m·s−1 (1.62 ± 0.26 BW·m−1) to be similar to that for walking at 1.50 m·s−1 (1.65 ± 0.22 BW·m−1), whereas the hip PUD force for running at 3.47 m·s−1 (1.25 ± 0.24 BW·m−1) was approximately 25% less than that for walking at 1.50 m·s−1 (1.65 ± 0.22 BW·m−1). Thus, our data further substantiate the proposition of Miller et al. (21,37) that PUD loads may be relevant in understanding why runners without comorbidities that place them at risk of bone and joint pathology do not appear to have an elevated risk of developing OA. Even though the peak magnitude of the HCF is greater for running compared to walking (Fig. 2), in healthy populations recreational running on its own might not be more harmful for the hip than walking. We do appreciate that this situation may only apply for a certain amount of running. Because accumulated loads likely do not have a linear effect on tissue damage, there will probably be a point with extreme running volumes where, regardless of PUD loads, the load accumulation exceeds the threshold for tissue tolerance.

This study was associated with some limitations that warrant discussion. First, we used a musculoskeletal modeling approach to predict the HCF during walking and running. This approach is inherently reliant on various input parameters and assumptions that have been previously outlined (20). Nevertheless, our HCF predictions were mostly consistent in magnitude with equivalent data from previous model-based studies despite differences in modeling methods. For example, previous studies have reported a peak HCF of approximately 5 BW for walking at 1.5 m·s−1 (11,13,14), which approximates our prediction of 5.68 ± 1.06 BW for walking at this speed. We predicted a peak magnitude for the resultant HCF of 12.62 ± 2.75 BW for running at 3.47 m·s−1. This result is consistent with equivalent data from Edwards et al. (12) and Giarmatzis et al. (13), but greater than data reported by Rooney and Derrick (15) (7.9 BW for running at 4.3 m·s−1) and van den Bogert et al. (16) (5.2 BW for running at 3.5 m·s−1). As previously discussed, our predicted HCF data are greater in magnitude when compared with published in vivo data, especially for running. Although we respect these discrepancies, we note that our main findings were based on within-subject comparisons, and thus will be relatively insensitive to any potential inaccuracies in our modeling approach. Furthermore, we used a static optimization algorithm in the present study based on the findings of Wesseling et al. (40), who compared a range of muscle optimization techniques for calculating the HCF and found static optimization to yield predictions that best matched in vivo experimental data from Bergmann et al. (4). Second, we acknowledge that the present study only involved a relatively discrete sample (N = 8) of healthy adult participants. However, our main outcome measures (the impulse of the resultant HCF, the PUD impulse, and the PUD force) were all found to be significantly influenced by locomotion speed with an effect size of at least 0.64. We therefore believe that our study was sufficiently powered, and a larger sample would not have altered the conclusions. Third, we predicted the HCF for walking and running, which are linear tasks where hip joint motion is primarily restricted to the sagittal plane. However, the hip is a ball-and-socket joint, and thus it has the capacity to rotate appreciably in all three anatomical planes. More complex multidirectional tasks (e.g. pivoting/cutting) may have greater potential for generating adverse joint loading conditions. Further research is required to compare the HCF across a complete spectrum of common functional activities. Fourth, sex-specific differences in hip joint biomechanics during locomotion have previously been documented (41). We therefore acknowledge the potential for the HCF to display sex-specific differences. Although the cohort used in the present study involved an equal distribution of male and female participants, the small sample size precluded us from investigating sex-specific effects.

A final methodological issue that warrants some discussion relates to the approach used to filter the experimental data. In the present study, we used a fourth-order, low-pass Butterworth filter with a cutoff frequency of 7 Hz to smooth the joint angles calculated from the inverse kinematics analysis, whereas we used the same filter with a cutoff frequency of 40 Hz to smooth the recorded ground reaction force data. This particular filtering approach was chosen primarily on the basis that it allowed our computational modeling process to operate most efficiently and effectively. We acknowledge that a mismatch in cutoff frequencies has the potential to create filtering-induced artifacts in our predicted HCF data during the initial stance period, especially for the high-impact locomotion condition of running at 3.47 m·s−1 (42). To overcome this issue, we focussed on the impulse of the HCF as a key outcome measure, the magnitude of which we expect to remain relatively invariant to alternative filtering approaches. Furthermore, we note that for some participants when running at 3.47 m·s−1 the magnitude of the “impact” spike in the resultant HCF exceeded (by a small amount) the peak that occurred around mid-stance; however, for both running conditions we specifically ignored the magnitude of any “impact” spike and only calculated and reported the magnitude of the peak in the resultant HCF that occurred around mid-stance. For these reasons, we are confident that any potential filtering-induced artifacts in our data did not in any way influence the main outcomes from our study.

In conclusion, we performed a comprehensive analysis of the HCF during walking and submaximal running in a cohort of healthy adult participants. Although the peak magnitude of the resultant HCF increased substantially with a change in locomotion mode from walking at 1.98 m·s−1 to running at 2.15 m·s−1, the impulse of the resultant HCF decreased. The GMAX, GMED, and ILPSO muscles were found to be the main contributors to the HCF during walking and running. Interestingly, it was ILPSO that was most affected by a change in locomotion mode, with its HCF contribution decreasing by approximately 50% when switching from walking to running. Finally, PUD loads at the hip were less for running at 3.47 m·s−1 compared with walking at either speed. The lower PUD impulse was due to the greater distance traveled per stride, whereas the lower PUD force was due to the shorter stride duration in addition to the greater distance traveled per stride. Such findings may explain (at least in part) why recreational running does not appear to be associated with an elevated risk of developing hip OA in healthy people.

Financial support for this project was provided by the Australian Research Council Linkage Project grant (LP110100262).

The authors have no conflicts of interest to declare. The results of the present study do not constitute endorsement by the American College of Sports Medicine. The results of this study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation.

Back to Top | Article Outline

REFERENCES

1. Andersson EA, Nilsson J, Thorstensson A. Intramuscular EMG from the hip flexor muscles during human locomotion. Acta Physiol Scand. 1997;161(3):361–70.
2. Pires NJ, Lay BS, Rubenson J. Joint-level mechanics of the walk-to-run transition in humans. J Exp Biol. 2014;217(Pt 19):3519–27.
3. Prilutsky BI, Gregor RJ. Swing- and support-related muscle actions differentially trigger human walk-run and run-walk transitions. J Exp Biol. 2001;204(Pt 13):2277–87.
4. Bergmann G, Deuretzbacher G, Heller M, et al. Hip contact forces and gait patterns from routine activities. J Biomech. 2001;34(7):859–71.
5. Bergmann G, Graichen F, Rohlmann A. Hip joint loading during walking and running measured in two patients. J Biomech. 1993;26(8):969–90.
6. Kotzar GM, Davy DT, Goldberg VM, et al. Telemeterized in vivo hip joint force data: a report on two patients after total hip surgery. J Orthop Res. 1991;9(5):621–33.
7. Rydell NW. Forces acting on the femoral head-prosthesis. A study on strain gauge supplied prostheses in living persons. Acta Orthop Scand. 1966;37(88 Suppl):1–132.
8. Brand RA, Pedersen DR, Davy DT, Kotzar GM, Heiple KG, Goldberg VM. Comparison of hip force calculations and measurements in the same patient. J Arthroplasty. 1994;9(1):45–51.
9. Heller MO, Bergmann G, Deuretzbacher G, et al. Musculo-skeletal loading conditions at the hip during walking and stair climbing. J Biomech 2001;34:883–93.
10. Stansfield BW, Nicol AC, Paul JP, Kelly IG, Graichen F, Bergmann G. Direct comparison of calculated hip joint contact forces with those measured using instrumented implants. An evaluation of a three-dimensional mathematical model of the lower limb. J Biomech. 2003;36:929–36.
11. Crowinshield RD, Brand RA, Johnston RC. The effects of walking velocity and age on hip kinematics and kinetics. Clin Orthop Relat Res. 1978;(132):140–4.
12. Edwards WB, Gillette JC, Thomas JM, Derrick TR. Internal femoral forces and moments during running: implications for stress fracture development. Clin Biomech. 2008;23(10):1269–78.
13. Giarmatzis G, Jonkers I, Wesseling M, Van Rossum S, Verschueren S. Loading of hip measured by hip contact forces at different speeds of walking and running. J Bone Miner Res. 2015;30(8):1431–40.
14. Röhrle H, Scholten R, Sigolotto C, Sollbach W. Joint forces in the human pelvis-leg skeleton during walking. J Biomech. 1984;17(6):409–24.
15. Rooney BD, Derrick TR. Joint contact loading in forefoot and rearfoot strike patterns during running. J Biomech. 2013;46(13):2201–6.
16. van den Bogert AJ, Read L, Nigg BM. An analysis of hip joint loading during walking, running, and skiing. Med Sci Sports Exerc. 1999;31(1):131–42.
17. Herzog W, Longino D, Clark A. The role of muscles in joint adaptation and degeneration. Langenbecks Arch Surg. 2003;388(5):305–15.
18. Correa TA, Crossley KM, Kim HJ, Pandy MG. Contributions of individual muscles to hip joint contact force in normal walking. J Biomech. 2010;43(8):1618–22.
19. Pandy MG, Andriacchi TP. Muscle and joint function in human locomotion. Annu Rev Biomed Eng. 2010;12:401–33.
20. Anderson FC, Pandy MG. Dynamic optimization of human walking. J Biomech Eng. 2001;123:381–90.
21. Miller RH, Edwards WB, Brandon SC, Morton AM, Deluzio KJ. Why don’t most runners get knee osteoarthritis? A case for per-unit-distance loads. Med Sci Sports Exerc. 2014;46(3):572–9.
22. Dorn TW, Schache AG, Pandy MG. Muscular strategy shift in human running: dependence of running speed on hip and ankle muscle performance. J Exp Biol. 2012;215(Pt 11):1944–56.
23. Anderson FC, Pandy MG. A dynamic optimization solution for vertical jumping in three dimensions. Comput Methods Biomech Biomed Engin. 1999;2(3):201–31.
24. Zajac FE. Muscle and tendon: Properties, models, scaling, and application to biomechanics and motor control. In: Bourne JR, editor. CRC Critical Reviews in Biomedical Engineering. 19. Baca Raton: CRC Press; 1989. p. 359–411.
25. Lu TW, O’Connor JJ. Bone position estimation from skin marker co-ordinates using global optimisation with joint constraints. J Biomech. 1999;32(2):129–34.
26. Anderson FC, Pandy MG. Static and dynamic optimization solutions for gait are practically equivalent. J Biomech. 2001;34(2):153–61.
27. Pedersen DR, Brand RA, Cheng C, Arora JS. Direct comparison of muscle force predictions using linear and nonlinear programming. J Biomech Eng. 1987;109(3):192–9.
28. Glitsch U, Baumann W. The three-dimensional determination of internal loads in the lower extremity. J Biomech. 1997;30(11–12):1123–31.
29. Lin YC, Kim HJ, Pandy MG. A computationally efficient method for assessing muscle function during human locomotion. Int J Numer Meth Biomed Engng. 2011;27(3):436–49.
30. Sasaki K, Neptune RR. Individual muscle contributions to the axial knee joint contact force during normal walking. J Biomech. 2010;43(14):2780–4.
31. Sritharan P, Lin YC, Pandy MG. Muscles that do not cross the knee contribute to the knee adduction moment and tibiofemoral compartment loading during gait. J Orthop Res. 2012;30(10):1586–95.
32. Winby CR, Lloyd DG, Besier TF, Kirk TB. Muscle and external load contribution to knee joint contact loads during normal gait. J Biomech. 2009;42(14):2294–300.
33. Bennell KL, Bowles KA, Wang Y, Cicuttini F, Davies-Tuck M, Hinman RS. Higher dynamic medial knee load predicts greater cartilage loss over 12 months in medial knee osteoarthritis. Ann Rheum Dis. 2011;70(10):1770–4.
34. Chang AH, Moisio KC, Chmiel JS, et al. External knee adduction and flexion moments during gait and medial tibiofemoral disease progression in knee osteoarthritis. Osteoarthritis Cartilage. 2015;23(7):1099–106.
35. Maly MR, Robbins SH, Stratford PW, Birmingham TB, Callaghan JP. Cumulative knee adductor load distinguishes between healthy and osteoarthritic knees—a proof of principle study. Gait Posture. 2013;37(3):397–401.
36. Tateuchi H, Koyama Y, Akiyama H, et al. Daily cumulative hip moment is associated with radiographic progression of secondary hip osteoarthritis. Osteoarthritis Cartilage. 2017;25(8):1291–8.
37. Miller RH. Joint loading in runners does not initiate knee osteoarthritis. Exerc Sport Sci Rev. 2017;45(2):87–95.
38. Alentorn-Geli E, Samuelsson K, Musahl V, Green CL, Bhandari M, Karlsson J. The association of recreational and competitive running with hip and knee osteoarthritis: a systematic review and meta-analysis. J Orthop Sports Phys Ther. 2017;47(6):373–90.
39. Williams PT. Effects of running and walking on osteoarthritis and hip replacement risk. Med Sci Sports Exerc. 2013;45(7):1292–7.
40. Wesseling M, Derickx LC, de Groote F, et al. Muscle optimization techniques impact the magnitude of calculated hip joint contact forces. J Orthop Res. 2015;33(3):430–8.
41. Boyer KA, Beaupre GS, Andriacchi TP. Gender differences exist in the hip joint moments of healthy older walkers. J Biomech. 2008;41(16):3360–5.
42. Bisseling RW, Hof AL. Handling of impact forces in inverse dynamics. J Biomech. 2006;39(13):2438–44.
Keywords:

BIOMECHANICS; MUSCULOSKELETAL MODEL; CONTACT FORCE; IMPULSE; PER-UNIT-DISTANCE LOAD

Supplemental Digital Content

Back to Top | Article Outline
© 2018 American College of Sports Medicine