It was desirable to utilize large load magnitudes to elicit the greatest energetic response to loading. It was determined through pilot testing that the maximum load attached to the shank should not exceed 20% of the lower extremity’s mass (i.e., sum of thigh, shank, and foot masses), because larger loads could not be securely and comfortably affixed to the subject’s proximal shank. Therefore, the load used in the Mass5 condition was calculated first for each subject. This established the total load carried on the body for all conditions. Once this load was determined, it was simply redistributed among the torso, proximal shank, and distal shank locations to achieve the desired lower-extremity mass and moment of inertia manipulations. For the Mass5 condition, the total load was concentrated on the proximal shank to minimize the effect of the load on Ihip. The mean proximal shank load of 2.8 kg used in this study was slightly more than the 1.5- to 2.0-kg loads applied to the distal shank or feet in previous research on persons with transtibial amputation (11,14) and on able-bodied subjects (13,23).
The BSLN condition was subsequently established by removing a portion of the proximal shank load for the Mass5, and repositioning that load on the torso. In addition, a portion of the remaining proximal shank load was redistributed distally to match Ihip for the BSLN and Mass5 conditions (Fig. 1 and Table 2). The MOI5 condition was achieved by redistributing most of the proximal shank load of the BSLN condition distally on the shank. This increased Ihip without altering total lower-extremity mass. Table 1 summarizes the average loads applied to the three loading sites to achieve the three experimental conditions, whereas Table 2 summarizes the average values for the lower-extremity mass, center of mass location, and Ihip that resulted from the load manipulations.
A condition in which all load was carried on the torso and no load was affixed to the legs (NL) served as a fourth testing condition (Table 1). The NL condition was included to determine whether adding load to an unloaded leg elicited a change in metabolic cost that was comparable to values associated with limb loading reported in the literature. Moreover, the NL condition also was used for EMG amplitude normalization.
Experimental data collection sessions.
Due to the inability to simultaneously collect metabolic, EMG, and video data, subjects returned for two additional testing visits. The order of the four conditions was randomized within each subject, but balanced across all subjects, similar to a Latin square procedure. For all data collection, subjects performed 6-min walking bouts for each condition at a treadmill speed of 1.5 m·s−1. Subjects were instructed to walk as comfortably as possible.
Aerobic demand was quantified using a TrueMax 2400 Metabolic Measurement System (ParvoMedics Inc., Salt Lake City, UT). The gas analyzers were calibrated using gases of known concentrations of O2 and CO2 before each test. The pneumotach was calibrated using a 3-L reference volume calibration syringe. To measure expired air for metabolic analysis, subjects wore a nose clip and breathed through a metabolic breathing valve attached to the metabolic cart via a hose. Subjects stood quietly for 4 min as resting V̇O2 was collected. The first 4 min of each 6-min trial permitted subjects to reach an aerobic steady state, and served as a load accommodation period. Average aerobic demand (V̇O2) was computed from the last 2 min of the resting trial and each walking trial.
On a separate day, subjects completed their final experimental session, during which EMG and video data were recorded. The order of loading conditions was the same as the order used for metabolic collection. These data were recorded from the left leg only, because of the physical laboratory setup. Testing the same leg for each subject was deemed satisfactory because it has been shown that, on average, gait is symmetrical (9). To synchronize the electromyography data with gait cycle events, footswitch sensors were affixed to the heel and toe of the shoe sole, and simultaneously sampled with EMG data at 1000 Hz.
EMG was used to investigate the relative amplitude of leg muscular activity in an effort to determine potential explanations for the hypothesized metabolic cost outcomes. The rectus femoris (RF), vastus lateralis (VL), biceps femoris (BF), tibialis anterior (TA), soleus (SO), gastrocnemius (GA), and gluteus maximus (GM) were selected for EMG recording because of their predominant role in gait and their accessibility to surface EMG data collection.
Ag-AgCl surface electrodes, integrated with a differential preamplifier circuit (Therapeutics Unlimited Inc., Iowa City, IA; preamplification gain = 35; interelectrode distance = 22 mm; electrode diameter = 8 mm), were placed over the distal half of the muscle belly, such that the contact surfaces were aligned in parallel with the muscle fibers. In an effort to reduce impedance at the skin–electrode interface, the skin of each electrode site was shaved, abraded, and cleaned with isopropyl alcohol. A single reference electrode was placed on the skin proximal to the lateral malleolus.
EMG data were conditioned with an EMG-544 amplifier/processor module (Therapeutics Unlimited Inc.) before sampling. Amplification gains were selectable from 1,000 to 100,000 within a bandwidth of 20–4,000 Hz. Input impedance was greater than 15 MΩ at 100 Hz. Common-mode rejection ratio was 87 dB at 60 Hz. EMG data and footswitch outputs were sampled simultaneously at 1000 Hz with Labview software (National Instruments, Austin, TX). Four data recordings of 5 s, each capturing a minimum of four consecutive stride cycles, were collected during the final minute of each walking bout.
The same marker set that was used to identify segment endpoints and added loads during the orientation visit was used again during the EMG data collection session. Identical video data recording procedures were conducted during all conditions. Added load magnitudes and video records of lower-extremity position were used to confirm that the experimental loading configurations produced the desired moment of inertia (Ihip) changes predicted by the limb inertial model (Table 2).
Mean oxygen consumption values over the last 2 min of the 6-min trial were used as an estimate of each subject’s walking economy. V̇O2 (mL·kg−1·min−1) was converted to watts per kilogram with the conversion factor of 21.1 J·mL−1 O2 (7) and dividing by 60 s. Resting metabolic power was subtracted from each load condition to yield net metabolic power (W·kg−1).
A three-segment (foot, shank, and thigh) sagittal plane model was used to calculate mechanical work, estimated from changes in segmental energies allowing energy transfers within and between segments (17):
where i represents the thigh, shank, or foot segment at time j. Segment masses, centers of mass, and moments of inertia were estimated from de Leva (5). Segment linear and angular velocities were calculated from coordinate data using finite difference equations. The lowest vertical position of the toe marker during stance served as the reference height for calculating potential energy. Mechanical work was computed with the following equation:
Average mechanical power (W·kg−1) for the swing phase was computed by dividing the work by swing time, and normalizing to total subject mass (body mass plus added load).
Temporal gait characteristics were calculated to assess commonly observed differences in the duration of stride cycle components associated with altered limb inertia. Heel strike and toe-off events were identified using voltage data obtained from footswitches and subsequently used to calculate stance, swing, and stride times (ms). Stride frequency was calculated as the inverse of stride time, and converted to units of strides per minute.
A simple pendulum model was used to calculate the period of oscillation (T):
where Iaxis is the pendulum moment of inertia about its axis of rotation (i.e., Ihip), m is its mass (i.e., total lower-extremity mass plus added load), d is its center of mass relative to the axis of rotation (i.e., total lower-extremity mass plus added-load center of mass relative to the hip joint), and g is the acceleration due to gravity. Period of oscillation does not directly correspond to temporal characteristics of gait; however, it is an effective tool in examining relative changes in swing time when lower-extremity inertia is altered (16,22).
Heel strike and toe-off events were also used to normalize EMG data to each stride cycle. Raw EMG data from 12 strides per condition were converted to microvolts, corrected for dc bias, band-pass filtered (20–400 Hz), full-wave rectified, and filtered using a fourth-order, zero-lag, low-pass Butterworth filter with a cutoff frequency of 15 Hz to create a linear envelope. EMG amplitude for each condition was normalized to the peak EMG amplitude of the NL condition, that is, each data point was expressed as a percent (25).
A two-factor, repeated measures ANOVA was used to test for a significant measure (metabolic, mechanical) by load condition (NL, BSLN, MOI5, Mass5) interaction in power. A one-factor (load) repeated measures ANOVA was used to test for significant differences in temporal characteristics. A two-factor (muscle by load condition) ANOVA with repeated measures on both factors was used to test for a significant interaction in EMG. Statistical significance was determined at P < 0.05 for all tests. Tukey’s HSD test was used for all post hoc statistical analyses. Effect sizes (ES) were calculated when differences existed.
The measure by load condition interaction for power was not significant [F(3,39) = 1.31, P = 0.29], indicating that mechanical and metabolic power responded similarly to loading condition (Fig. 2) as hypothesized. The condition main effect was significant [F(3,39) = 61.2, P < 0.001], as power increased with increases in limb mass and/or moment of inertia. The addition of a nominal load to the unloaded leg (NL) to create the BSLN condition resulted in a significant increase in power (P < 0.001; ES = 0.64). There was a significant increase in power from BSLN to MOI5 (P < 0.01; ES = 0.27) and a significant increase from BSLN to Mass5 (P < 0.01; ES = 0.29); however, power for MOI5 and Mass5 was not different (P = 0.99).
It was hypothesized that changes in peak EMG amplitude would reflect changes in energy cost between load conditions; however, changes in EMG were small (Fig. 3). Despite the trend for most muscles to have increased activation for both MOI5 and Mass5, the only significant differences in peak EMG were for VL (Mass5 greater than both BSLN, P = 0.001; ES = 0.59 and MOI5, P = 0.001; ES = 0.46) and for SO (MOI5 greater than Mass5, P < 0.05; ES = 0.74).
There were small but statistically significant differences in swing time [F(3,39) = 3.44, P < 0.05] and stride time [F(3,39) = 3.52, P < 0.05] among the loading conditions (Table 3). Specifically, swing time was 2.2% greater (P < 0.05; ES = 0.42) and stride time was 1.1% greater (P < 0.05; ES = 0.26) for MOI5 compared with NL. However, there were no differences in temporal gait characteristics among the three experimental conditions (BSLN, MOI5, Mass5). There were no differences in stance time among conditions [F(3,39) = 0.91, P = 0.44]. There was a significant difference in pendulum period of oscillation [F(3,39) = 171.36, P < 0.001] among the loading conditions (Table 3). Post hoc analyses indicated that pendulum periods of oscillation for all conditions were different from each other; most notably, MOI5 was 1.7% greater than BSLN (P < 0.001; ES = 1.00), whereas Mass5 was 2.1% smaller than BSLN (P < 0.001; ES = 1.27).
Results of published studies examining the relationship between lower-extremity inertia and metabolic cost of walking in unilateral, transtibial amputees (14) and nonpathological subjects (2,13,23) suggest that metabolic cost of walking increases as lower-extremity inertia increases. However, these experimental manipulations did not reveal whether the higher metabolic demands were associated with increases in lower-extremity mass or moment of inertia about the hip joint (Ihip), because both inertial parameters were altered with the addition of load to a limb. The primary focus of our project was to extend our understanding of the effect of systematic manipulations of limb inertial characteristics on metabolic energy cost of walking, with potential implications for design considerations of lower-extremity prostheses. Our results indicated metabolic cost increased similarly (3.4% for MOI5; 4.0% for Mass5) when Ihip and lower-extremity mass were independently increased by 5% from a baseline load condition.
Martin (12) reported that lower-extremity mechanical work increased with increases in the metabolic cost of running with lower-extremity loading. Furthermore, mass applied only to the foot elicited a greater mechanical and metabolic response than when the same mass was applied only to the thigh. Likewise, mechanical power of the lower extremity during the swing phase in our study increased in parallel with metabolic power, suggesting the increased metabolic demand was associated with higher mechanical demands under increased inertia conditions. Interestingly, two recent investigations of walking by lower-extremity amputees observed little or no relationship between metabolic cost and mechanical work demands when prosthesis inertia was manipulated. Scherer and colleagues (20) compared the effect of a lighter titanium prosthesis versus a heavier stainless steel prosthesis on the mechanical work and metabolic cost of walking in transtibial and transfemoral amputees. Although the mechanical work demands of walking were lower for the titanium prosthesis than for the steel prosthesis, there was no difference in metabolic cost between these conditions. Similarly, Foerster et al. (6) reported a poor correlation between lower-extremity mechanical work and metabolic cost when mass was added to the prosthetic limb.
Contrasts of proximal and distal lower-extremity loading effects during walking have not been investigated in able-bodied individuals. Lehmann and colleagues (11), however, systematically investigated the effects of load magnitude and position on walking energy demand of transtibial amputees. Metabolic cost increased 4.2 and 7.9%, respectively, when mean loads of 1.0 and 1.5 kg were shifted from a location near the prosthetic limb’s center of mass to a distal location near the ankle. We observed a comparable effect. Metabolic cost was 3.4% higher for the MOI5 condition relative to the BSLN condition. This increase in energy cost was directly associated with an average shift of 0.52 kg distally on the leg. Collectively, these results have important implications for prosthesis mass distribution. For example, when the pylon of a prosthetic limb is enhanced with shock absorbing mechanisms, prosthesis designers should always consider proximal placement of these components, because a more distal distribution of mass will contribute to increased energy cost of walking.
We are aware of no published studies of walking or running that have attempted to independently manipulate total limb mass while maintaining a limb’s moment of inertia. Lehmann and colleagues (11) reported metabolic cost was not affected by the addition of 1.0 and 1.5 kg near the prosthetic limb center of mass. Similarly, Czerniecki and colleagues (4) reported no change in walking metabolic cost when 0.68- and 1.34-kg loads were applied near the prosthesis center of mass for persons with transfemoral amputation. Mattes and colleagues (14) reported no change in the cost of walking when 0.85 kg was added to the distal aspect of the prosthetic limb shank of unilateral transtibial amputees, but a 6% increase in cost when 1.70 kg were added. Collectively, results from these three studies suggest that metabolic demand of walking for lower-extremity amputees is less sensitive to mass increases if the load is positioned more proximally on a limb. In contrast, we observed a 4.0% higher energy cost for our Mass5 condition relative to the BSLN condition. The Mass5 condition was created by adding an average load of 0.81 kg bilaterally to the proximal aspect of the shank while also shifting an average of 0.68 kg from the distal to the proximal load site. To be consistent with trends observed by Lehman et al. and Czerniecki et al., we should have seen little or no effect of adding mass proximally to the shank.
Our results using able-bodied individuals clearly suggest that metabolic demand is equally sensitive to similar relative increases in limb mass and moment of inertia. Extrapolating our results to lower-extremity amputees suggests that prosthetic designers should be equally sensitive to any increases in limb mass and mass distribution when considering different design configurations. Mechanical advances in prosthetic ankle joint design may improve gait mechanics (3,18); however, these components may also have a greater mass than standard components, resulting in an increase in Ihip and, most likely, energy cost. Ankle-joint components are located distal to the loading positions examined in this study; therefore, Ihip would mostly likely increase a greater magnitude than investigated here, which would concomitantly affect energy cost. Lightweight components and proximal positioning of these components obviously should be considered when feasible for prosthetic limb design. However, because our results conflict with outcomes observed by others in their investigations of the effects of prosthetic limb mass and mass distribution changes on walking energy demand in lower-extremity amputees, the independent effects of mass and moment of inertia manipulations needs to be assessed directly for both transtibial and transfemoral amputees.
It was hypothesized that peak EMG amplitude for several prominent lower-extremity muscles would increase as both mass and moment of inertia increased. Despite increases in energy cost of approximately 4.0% from BSLN to both MOI5 and Mass5, and a trend for muscle excitation to be greater with increased mass and moment of inertia (Fig. 3), there were few statistically significant differences in EMG between load conditions. Thus, we have no objective evidence from our EMG assessment from which to conclude that an increase in the local demand on leg musculature explains the increased demand of walking under increased mass and moment of inertia conditions produced by our loading schemes. Nevertheless, it is worth noting that EMG levels were approximately 10% higher for BSLN, Mass5, and MOI5 conditions relative to the NL condition in which all load was carried on the torso (Fig. 3). Clearly, though, our EMG assessment was not sufficiently sensitive to distinguish our Mass5 and MOI5 conditions from the BSLN condition.
It has been suggested that the swing phase of gait involves minimal activation of muscles that cross the hip and knee (15,16), and the dynamics of the swinging limb are often compared with that of a simple pendulum (21). There is a consistent relationship between walking swing time and the period of oscillation of a simple pendulum; however, the values of each parameter are not equal. Changing the inertial characteristics of the limb would alter the period of oscillation described by a simple pendulum model, which has been shown to reflect increases or decreases in walking swing time (14,19). Our data show that swing time increased from BSLN for the MOI5 condition, and decreased from BSLN for the Mass5 condition, which is consistent with the passive pendulum model’s period of oscillation (Table 3). The trend for swing time to change in parallel with the pendulum period of oscillation suggests that the limb was swinging at a resonant frequency, which results in a minimization of metabolic cost (10). If the changes in swing time did not correspond to the changes in period predicted by the pendulum model, metabolic cost would have increased to a greater extent, reflecting the increased muscle activation required to alter the natural period of the swing limb.
In conclusion, our results showed similar increases in metabolic cost when both lower-extremity mass and moment of inertia about the hip joint were each increased independently from a nominal baseline condition. Mechanical power paralleled the changes in metabolic cost. Thus, independent alterations in each inertial characteristic influenced metabolic and mechanical power similarly. In addition, there were few differences in EMG amplitude across load conditions that support the observed changes in energy cost. Improving the comfort, stability, and performance of a prosthesis typically requires the addition of materials to the limb, thus increasing the limb’s mass and moment of inertia. Therefore, multiple combinations of mass magnitude and mass distribution should be investigated to more accurately identify the interplay between segment inertial characteristics and lower-extremity mechanical work that affect energy cost.
1. Andriacchi, T. P., J. A. Ogle, and J. O. Galante. Walking speed as a basis for normal and abnormal gait measurements. J. Biomech
. 10:261–268, 1977.
2. Barnett, S. L. Ankle weight effect on gait: orthotic implications. Orthopedics
3. Barth, D. G., L. Schumacher, and S. Sienko Thomas. Gait analysis and energy cost of below-knee amputees wearing six different prosthetic feet. J. Prosthet. Orthot
. 4:63–75, 1992.
4. Czerniecki, J. M., A. Gitter, and K. Weaver. Effect of alterations in prosthetic shank mass on the metabolic costs of ambulation in above-knee amputees. Am. J. Phys. Med. Rehabil
. 73:348–352, 1994.
5. de Leva, P. Adjustments to Zatsiorsky-Seluyanov‘s segment inertia parameters. J. Biomech
. 29:1223–1230, 1996.
6. Foerster, S. A., A. M. Bagley, C. D. Mote, and H. B. Skinner. The prediction of metabolic energy expenditure during gait from mechanical energy of the limb: a preliminary study. J. Rehabil. Res. Dev
. 32:128–134, 1995.
7. Foss, M. L., and S. J. Keteyian. Fox’s Physiological Basis for Exercise and Sport
. Boston: WCB/McGraw-Hill, 1998.
8. Gailey, R. S., M. A. Wenger, M. Raya, et al. Energy expenditure of trans-tibial amputees during ambulation at self-selected pace. Prosthet. Orthot. Int
. 18:84–91, 1994.
9. Hannah, R. E. Kinematic symmetry of the lower limbs. Arch. Phys. Med. Rehabil
. 65:155–158, 1984.
10. Holt, K. G., J. Hamill, and R. O. Andres. The force-driven harmonic oscillator as a model for human locomotion. Hum. Mov. Sci
. 9:55–68, 1990.
11. Lehmann, J. F., R. Price, R. Okumura, K. Questad, B. J. de Lateur, and A. Negretot. Mass and mass distribution of below-knee prostheses: effect on gait efficacy and self-selected walking speed. Arch. Phys. Med. Rehabil
. 79:162–168, 1998.
12. Martin, P. E. Mechanical and physiological responses to lower extremity loading during running. Med. Sci. Sports Exerc
. 17:427–433, 1985.
13. Martin, P. E., T. D. Royer, and S. J. Mattes. Effect of symmetrical and asymmetrical lower extremity inertia changes on walking economy. Med. Sci. Sports Exerc
. 29:S86, 1997.
14. Mattes, S. J., P. E. Martin, and T. D. Royer. Walking symmetry and energy cost in persons with unilateral transtibial amputations: matching prosthetic and intact limb inertial properties. Arch. Phys. Med. Rehabil
. 81:561–568, 2000.
15. Mena, D. Analysis and synthesis of human swing leg motion during gait and its clinical applications. J. Biomech
. 14:823–832, 1981.
16. Mochon, S., and T. A. McMahon. Ballistic walking. J. Biomech
. 13:49–57, 1980.
17. Pierrynowsi, M. R., R. W. Norman, and D. A. Winter. Mechanical energy analyses of the human during local carriage on a treadmill. Ergonomics
18. Powers, C. M., L. Torburn, J. Perry, and E. Ayyappa. Influence of prosthetic foot design on sound limb loading in adults with unilateral below-knee amputations. Arch. Phys. Med. Rehabil
. 75:825–829, 1994.
19. Royer, T. D., P. E. Martin, and S. J. Mattes. Perturbability of temporal symmetry of gait by symmetrical and asymmetrical lower extremity inertia changes. Med. Sci. Sports Exerc
. 29:S113, 1997.
20. Scherer, R. F., J. J. Dowling, G. Frost, M. Robinson, and K. McLean. Mechanical and metabolic work of persons with lower-extremity amputations walking with titanium and stainless steel prostheses: A preliminary study. J. Prosthet. Orthot
. 11:38, 1999.
21. Selles, R. W., J. B. Bussmann, R. C. Wagenaar, and H. J. Stam. Effects of prosthetic mass and mass distribution on kinematics and energetics of prosthetic gait: a systematic review. Arch. Phys. Med. Rehabil
. 80:1593–1599, 1999.
22. Selles, R. W., J. B. Bussmann, R. C. Wagenaar, and H. J. Stam. Comparing predictive validity of four ballistic swing phase models of human walking. J. Biomech
. 34:1171–1177, 2001.
23. Skinner, H. B., and R. L. Barrack. Ankle weighting effect on gait in able-bodied adults. Arch. Phys. Med. Rehabil
. 71:112–115, 1990.
24. Waters, R. L., and S. Mulroy. The energy expenditure of normal and pathologic gait. Gait Posture
25. Yang, J. F. Electromyographic amplitude normalization methods: improving their sensitivity as diagnostic tools in gait analysis. Arch. Phys. Med. Rehabil
. 65:517–521, 1984.
Keywords:©2005The American College of Sports Medicine
BIOMECHANICS; METABOLIC COST; GAIT; LOAD CARRIAGE; ELECTROMYOGRAPHY