| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Biomotion Laboratory, Department of Orthopaedics, Massachusetts General Hospital Institute of Health Professions, and Harvard Medical School, Boston, Massachusetts 02114
Submitted 29 April 2003 ; accepted in final form 29 August 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
mechanical power; joint angles and moments; hip and ankle compensation; lower extremity arthritis
A major challenge for identifying what locomotor characteristics are linked to disablement is that adaptations in locomotor function also occur without disablement, attributable to common impairments associated with normal aging. Age-related loss of strength, balance control, and cardiorespiratory function may also engender gait adaptations, such as slowed gait, shorter step length, and wider base of support (2, 5, 7, 27). Several studies in recent years suggest that analysis of joint kinematics (angular rotations) and kinetics (moments and powers) can illuminate the mechanistic behaviors responsible for these changes in gait with aging (3, 10, 13, 28, 32). Prior studies report reductions in ankle plantar flexion and plantar flexor moment and power (3, 10, 13, 32), reduced knee flexion and knee extensor moment and power (3, 32), reduced hip extension (12, 13, 28), and increased hip extensor or flexor moment and power (3, 10, 13) in healthy elders compared with young adults. A recent review article discusses these studies in more detail (17).
Of particular interest is how adaptations in the gait of elders with functional limitations in mobility differ from those of healthy elders. Understanding these differences may be important for determining the optimal physical therapy intervention required to prevent disablement in healthy elders and to halt or reverse disablement in those with profound functional limitations. The optimal treatments for these two classifications of older persons may, in fact, be quite different. However, few studies have examined the biomechanics of gait in elderly people with functional limitations (19-21, 23). Reductions in ankle work and increases in hip flexor eccentric (absorption) work have been reported for elders with functional limitations compared with age-matched healthy elders. The difference between these findings and those of studies comparing healthy young and old subjects are primarily the effects occurring with hip musculature. However, the underlying mechanisms of these disparate effects remain elusive.
If one considers the following sequence, young and healthy 
old and healthy old and functionally limited, a question that has not been addressed by prior studies arises: Are kinematic and kinetic changes in gait between the healthy young and healthy elders simply exaggerated for the elderly with lower extremity pathology, or do gait changes from lower extremity pathology reflect neuromuscular adaptations different from those caused by aging? If the former were true, there would be a consistent deterioration in gait kinematic and kinetic parameters among the healthy young, healthy elders, and elders with functional limitations. In other words, the kinematic and kinetic variables that differ between the healthy young and healthy elders should also differentiate healthy elders and functionally limited elders. However, if gait changes in functionally limited elders reflect different neuromuscular adaptations, unrelated to aging, a different set of variables would be expected to distinguish them from healthy elders, compared with those distinguishing young and old healthy people. It is also important that spatiotemporal variables be controlled, e.g., gait speed, to ensure that comparisons are indeed reflective of neuromuscular differences.
In the present study, joint kinematic (peak angular rotations) and kinetic (peak moments and powers) parameters for healthy young and elderly subjects and for elders with functional limitations due to lower extremity pathology are compared to explore the differing biological mechanisms that underlie gait changes, when controlling for spatiotemporal differences. We hypothesize that neuromuscular adaptations responsible for gait changes in functionally limited elders with lower extremity pathology differ from those adaptations that distinguish young and old healthy people. We also hypothesize, based on prior studies, that the primary neuroadaptive mechanisms that discriminate between healthy and functionally limited elders will involve the hip musculature.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
Kinematic and kinetic analysis of gait. Subjects walked from two to five trials unassisted in bare feet along a 10-m walkway. Bilateral body segment kinematics were measured with a four-camera Selspot II optoelectric system (Selective Electronics, Partille, Sweden). Briefly, arrays of three to five light-emitting diodes (LEDs) embedded in rigid plastic disks were securely attached to 11 body segments (both feet, shanks, thighs, and arms, and pelvis, trunk, and head) by using neoprene and leather forms with Velcro straps (Fig. 1). Raw LED data were sampled at 150 Hz and filtered by using a low-pass Butterworth filter (4th order, 6-Hz cutoff, zero lag). LED array trajectories were analyzed by using TRACK software (Massachusetts Institute of Technology, Cambridge, MA) and resolved into three-dimensional (3D), six degree-of-freedom, body segment kinematics. Foot-floor reactions were synchronously sampled (150 Hz, unfiltered) from two adjacent piezoelectric force platforms (Kistler Instruments, Winterthur, Switzerland).
|
A pointing technique was used to establish segment-fixed anatomic coordinate systems relative to the LED arrays (29), as illustrated in Fig. 1. Joint angles were calculated by using Cardan 3-1-2 decomposition of the segment rotational matrices (30). Body segment mass, mass moments of inertia, and centers of mass were computed from regression models for male (16) and female (33) adults. Higher order segmental kinematics were computed from successive numerical differentiation (18) of segmental linear and angular displacements. Force plate data, segment kinematics, and inertial properties were input into a 3D Newtonian inverse dynamic model (18) to compute net joint moments in segment coordinates (about the joint mediolateral axes of the proximal segment). Net joint moments were normalized to body mass and height, resulting in units of Newton meter per kilogram per meter. Net joint powers were computed as the dot product of net joint moments and joint angular velocity in the global reference frame and then transformed into segment coordinates as was done for moments. Normalized moments were used in the above calculations such that power units were in Watts per kilogram per meter.
Data analysis. Spatiotemporal variables consisted of average anterior center of mass velocity (gait speed), stance duration, step length, and step width. Average gait speed was calculated as the whole body center of mass displacement over time during the gait cycle, where whole body center of mass was calculated from combining the center of mass of individual body segments (29). Stance duration was the elapsed time between heel strike and toe off of stance phase. Step width and step length were computed as the lateral and anterior-posterior distances, respectively, between ankle centers in double support. These data were collected for each walking trial and averaged across trials.
Although body segment data were processed in 3D space, only the sagittal (flexion-extension) plane values were analyzed in this study. Figure 2 shows the kinematic and kinetic peaks for a representative subject. Peak values were collected for each walking trial and averaged over the number of walking trials of each subject. Joint kinematic variables consisted of the following: peak ankle dorsiflexion (AR1) and peak ankle plantar flexion (AR2) in terminal stance/swing initiation; peak knee flexion in early stance loading response (KR1), peak knee extension (or minimum knee flexion) in midstance (KR2), and peak knee flexion angle in swing (KR3); and peak hip extension (HR1) and peak hip flexion (HR2). Joint moments consisted of the following: peak ankle dorsiflexor moment in early stance (AM1) and peak ankle plantar flexor moment in late stance (AM2); peak knee extensor moment in loading response (KM1), peak knee flexor (or minimum knee extensor) moment in midstance (KM2), and peak knee extensor moment in terminal stance (KM3); and peak hip extensor moment in early stance (HM1) and peak hip flexor moment in late stance (HM2). Joint powers consisted of the following: ankle (plantar flexor) eccentric peak power (AP1); ankle (plantar flexor) concentric peak power (AP2); first knee (extensor) eccentric peak power (KP1); first knee (extensor) concentric peak power (KP2); second knee (flexor) concentric peak power (KP3); second knee (extensor) eccentric peak power (KP4); first hip (extensor) concentric peak power (HP1); hip (flexor) eccentric peak power (HP2); and second hip (flexor) concentric peak power (HP3). Eccentric refers to muscle lengthening in absorption of energy, whereas concentric refers to muscle shortening in generating energy.
|
Statistical analyses consisted of multiple analysis of covariance (MANCOVA) of the above kinematic and kinetic variables, controlling for spatiotemporal variances, among the three groups of subjects: healthy young (n = 45), healthy elderly (n = 37), and functionally limited elderly (n = 38). This latter group will be referred to as disabled elders. This was followed by a discriminant analysis to 1) determine the relative importance of the discriminating variables in terms of differentiating age effects (heathy young vs. healthy elders) from disability effects (healthy elders vs. disabled elders) and 2) evaluate the sensitivity and specificity of classification of subjects based on these variables.
The first step was to arrive at an appropriate covariate among potential candidate variables: gait speed, step length, stance duration, and step width. This was accomplished by examining between-group differences in these variables and their correlations. MANCOVA was then used to examine blocks of variables (joint kinematics, joint moments, joint powers) between groups of subjects. If significance on the MANCOVA was achieved for a block, variables yielding significant between-group differences on univariate analysis of covariance (ANCOVA) tests were then examined with Bonferroni post hoc group comparisons. All such variables from each of the three blocks were then designated as "potential discriminating variables," based on whether they discriminated by age or by health status. Thus variables were included if they discriminated one group from the other two. These variables were then individually adjusted for the selected covariate and entered in the discriminant analysis.
A Fisher linear discriminant analysis was conducted by using the potential discriminating variables as independents and grouping variables as dependents, healthy young subjects and healthy elders, and then separately for healthy elders and disabled elders. We then arrived at a set of structure coefficients for each discriminating variable for healthy young healthy elderly, and healthy elderly disabled elderly. The structure coefficients (correlations between the individual discriminating variables and the standardized canonical discriminant functions) were used to gauge the relative importance of each discriminating variable. Classification tables were used to gauge how well the discriminating variables classified individuals. A "leave-oneout analysis" and kappa statistics were computed for the classifications to control for model bias and chance classification, respectively. Classification was also conducted with gait speed as the sole independent variable, to show how much better the kinematic and kinetic variable discriminated groups compared with only using gait speed.
Finally, bivariate correlations among the discriminating variables were examined for the three groups to explain the underlying mechanisms of the observed adaptations. Partial correlations of selected variables (potentially including those not identified as discriminating variables) were conducted as follow-up tests to the bivariate correlation analysis. Because peak kinematic and kinetic variables may have negative or positive sense, depending on sign convention (with the exception of power variables, which are negative or positive, depending on energy flow), all variables were made positive sense for the purpose of the correlation analysis. This was done to make interpretation of the correlation coefficients, direct vs. inverse relationships, more straightforward.
A type I error rate (
) of 0.05 was used for the MANCOVA/ANCOVA and kappa statistics. Holm's Bonferroni correction with a family-wise of 0.05 was applied to bivariate correlation analysis. SPSS for Windows (version 10.0, SPSS, Chicago, IL) was used for all statistical analyses.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
Spatiotemporal variables. Gait speed and step length significantly differed among groups (P < 0.001). Post hoc tests revealed that young subjects had significantly higher gait speed and step length compared with healthy elders (P = 0.001 and P < 0.001, respectively) and healthy elders had significantly higher gait speed and step length than disabled elders (P = 0.021 and P = 0.010, respectively). Stance duration and step width were not significantly different between groups (P = 0.059 and P = 0.136, respectively). However, there were strong relationships between gait speed and stance duration (r = -0.809, P < 0.001) and between gait speed and step length (r = 0.857, P < 0.001). Thus only one variable was necessary to represent subject's spatiotemporal differences. Because gait speed was significantly different among all three groups, it was used as a covariate to adjust kinematic and kinetic group means for statistical comparison. These data are summarized in Table 2.
Joint kinematics and kinetics. The MANCOVA tests for each block of variables (with gait speed as a covariate) indicated significant covariate effects and differences among groups for joint rotations (covariate: Wilk's 
= 0.519, P < 0.001; between groups: Wilk's = 0.807, P = 0.041), joint moments (covariate: Wilk's = 0.433, P < 0.001; between groups: Wilk's = 0.785, P = 0.005), and joint powers (covariate: Wilk's = 0.417, P < 0.001; between groups: Wilk's = 0.592, P < 0.001).
For joint rotations (Table 3), ANCOVA tests indicated significant between-group differences for AR2 (P = 0.011) and KR1 (P = 0.013). Bonferroni post hoc tests revealed significantly higher AR2 in the healthy young compared with healthy elders (P = 0.018) and disabled elders (P = 0.034), although no differences were detected between healthy elders and disabled elders (P = 1.00). Significantly higher KR1 was found for healthy elders compared with the healthy young (P = 0.010), whereas no difference was found for disabled elders compared with healthy elders (P = 0.448) or the healthy young (P = 0.503).
|
For joint moments (Table 4), ANCOVA tests indicated significant between-group differences for AM1 (P = 0.003), AM2 (P = 0.003), KM2 (P = 0.016), KM3 (P = 0.029), and HM2 (P = 0.006). Bonferroni post hoc tests revealed significantly higher AM1 for disabled elders compared with the healthy young (P = 0.003) and healthy elders (P = 0.022) but no difference between healthy elders and the healthy young (P = 1.00). Significantly lower AM2 was found for disabled elders compared with the healthy young (P = 0.030) and healthy elders (P = 0.002), with no difference between the healthy young and healthy elders (P = 1.00). Significantly higher KM2 and KM3 were found for the healthy young compared with disabled elders (P = 0.014 and P = 0.024, respectively), whereas no differences were detected for healthy elders compared with the healthy young (P = 0.757 and P = 0.429, respectively) and disabled elders (P = 0.148 and P = 0.433, respectively). HM2 was significantly higher for disabled elders compared with the healthy young (P = 0.008) and healthy elders (P = 0.029), but no difference was detected between the healthy young and healthy elders (P = 1.00).
|
For joint powers (Table 5), ANCOVA tests indicated significant between-group differences for AP2 (P = 0.002), KP2 (P = 0.018), KP4 (P = 0.001), HP1 (P < 0.001), and HP2 (P = 0.007). Bonferroni post hoc tests revealed significantly lower AP2 for disabled elders compared with the healthy young (P = 0.002) and healthy elders (P = 0.035), but no difference between the healthy young and healthy elders (P = 0.585). Significantly lower KP2 was found for disabled elders compared with the healthy young (P = 0.020), but no difference was found for healthy elders compared with disabled elders (P = 1.00) or the healthy young (P = 0.096). KP4 was significantly lower for the healthy young compared with healthy elders (P = 0.016) and disabled elders (P = 0.001), with no differences detected between healthy elders and disabled elders (P = 0.651). HP1 was significantly lower for the healthy young compared with healthy elders (P = 0.003) and disabled elders (P < 0.001), with borderline significance between healthy elders and disabled elders (P = 0.052). HP2 was significantly higher for disabled elders compared with healthy elders (P = 0.026) and the healthy young (P = 0.012), but no differences were detected between healthy elders and the healthy young (P = 1.00).
|
Discriminant analyses. Potentially discriminating variables identified from the MANCOVA/ANCOVA analyses above, as defined by the previously described criteria, consisted of AR2, AM1, AM2, HM2, AP2, KP4, HP1, and HP2. Structure coefficients for this set of variables are shown for the healthy young vs. healthy elders and healthy elders vs. disabled elders in Fig. 3. These data show a high degree of discrimination between the healthy young and healthy elderly groups from HP1, AR2, and KP4 and a relatively low-degree of discrimination from AM1, AM2, HM2, AP2, and HP2. Conversely, there was a high degree of discrimination between healthy elderly and disabled elderly groups from AM1, AM2, HP2, AP2, and HM2, as well as HP1, and a relatively low degree of discrimination from KP4 and AR2.
|
Classification accuracy between the healthy young and healthy elderly subjects, using the above set of variables, was 79.3%. Sensitivity and specificity for identifying an elder >50 yr of age was 0.757 and 0.822, respectively. Cross-validation from the leave-one-out analysis was 68.3%, with sensitivity and specificity of 0.621 and 0.733, respectively, indicating a low degree of bias in the classification of subjects from the study sample. Kappa statistics were significant (
= 0.580, P < 0.001), suggesting that change agreement was low. Based on gait speed alone, classification accuracy was 64.6% (cross-validation was also 64.6%), with sensitivity and specificity of 0.541 and 0.733, respectively. Kappa statistics for classification by gait speed alone were significant, although they indicated a relatively higher degree of change agreement ( = 0.277, P = 0.011).
Classification accuracy between healthy elders and disabled elders, using the above set of variables, was 73.3%. Sensitivity and specificity for identifying an elder with functional limitations was 0.737 and 0.730, respectively. Cross-validation from the leave-one-out analysis was 68.0%, with sensitivity and specificity of 0.711 and 0.649, respectively, again indicating a low degree of bias in the classification of subjects from the study sample. Kappa statistics were significant ( = 0.467, P < 0.001), suggesting that change agreement was low. Based on gait speed alone, classification accuracy was 61.3% (cross-validation was also 61.3%), with sensitivity and specificity of 0.605 and 0.622, respectively. Kappa statistics for classification by gait speed alone were, however, barely significant, indicating a high degree of change agreement ( = 0.227, P = 0.049).
Correlation analysis. Bivariate correlations are summarized in Table 6; all significant correlations had r > 0.45 and P 
0.005. Four relationships among the discriminating variable set existed, regardless of grouping: AM2 with AP2, HM2 with KP4, HM2 with HP2, and KP4 with HP2. One relationship existed for the healthy young group that was not present for the other two groups: AM1 with AM2. One relationship existed for the healthy elder group that was not present for the other two groups: AM1 with AP2. Two relationships existed for healthy elders and disabled elders that were not present for the healthy young: AR2 with AP2 and AM1 with KP4. Finally, a total of seven relationships existed for the disabled elders that were not present for the healthy young or healthy elders: AR2 with AM2, AR2 with HM2, AM1 with HM2, AM1 with HP2, HM2 with AP2, AP2 with KP4, and AP2 with HP2.
|
A post hoc follow-up test was conducted to examine the relationships among HM2, KP4, and HP2. The reasoning behind this post hoc comparison was that 1) all three variables were significantly related to one another and 2) the relationships were consistent for all three groups of subjects. A partial correlation analysis was conducted in which the relationship between two of the variables was explored while partialing out the variance of the third. Significant correlations remained for HM2 vs. HP2 when partialing out KP4 and for HM2 vs. KP4 when partialing out HP2; however, no significant relationship remained between HP2 and KP4 when partialing out HM2. These findings were consistent across groups of subjects. Although hip extension angle (HR1) was not identified as one of the discriminating variables, because HR1 occurs approximately at HM2 and HP2, we computed the correlation between HR1 and the three kinetic variables: HM2, KP4, and HP2. HR1 was directly related to HM2 and HP2 but was not related to KP4. Furthermore, the relationship between HR1 and HP2 was eliminated when partialing out HM2.
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Prior studies provide relatively strong, although not conclusive, evidence that changes in gait with aging are more a result of neuromuscular adaptations than self-selected gait alterations (e.g., reductions in walking speed and step length) to increase stability (17). Judge et al. (10) reported that elders increase hip flexor power to compensate for reduced ankle plantar flexor power, presumably to assist in advancing the leg into swing phase. Kerrigan et al. (13) reported that hip flexor contracture, restricting hip extension in the late-stance phase, may cause gait changes. DeVita and Hortobagyi (3) reported increases in hip extensor moment and power in healthy elderly subjects who walked at the same speed as healthy young subjects, with simultaneous reductions in ankle and knee power. Although these studies do not agree in all respects, they all suggest that neuromuscular changes occur with aging that cause changes in gait characteristics. The present study largely supports the findings of these previous investigations and gives further insight into how functional limitations influence gait characteristics, beyond the effects of aging. Below we discuss the kinematic and kinetic variables that discriminated by age and by health status and how these variables may be useful for identifying those at risk for disablement, and we also discuss the interrelationships among these variables to elucidate the neuroadaptive mechanisms of disablement.
Separating the effects of age and disability on gait. Data from the present study suggest that gait changes with lower extremity impairment manifest differently than do gait changes with aging alone. As hypothesized, kinematic and kinetic variables that discriminated between healthy elders and young subjects differed from those variables that discriminated between healthy elders and elders with functional limitations, suggesting that gait changes due to lower extremity pathology and impairment involve different neuromuscular adaptations than those for aging. As shown in Fig. 3, the most prominent age-related effects, for healthy elderly compared with young subjects, were reduced peak ankle plantar-flexion angle in late stance (AR2), increased peak eccentric knee extensor power in the late-stance phase (KP4), and increased peak concentric hip extensor power in the early-stance phase (HP1). However, the effects that discriminated best by health status, for disabled elders compared with healthy elders, were increased peak ankle dorsiflexor moment in early stance (AM1), decreased peak ankle plantar flexor moment in late stance (AM2), increased peak hip flexor moment in late stance (HM2), decreased peak concentric ankle plantar-flexor power in late stance (AP2), and increased peak eccentric hip flexor power in mid-to-late stance (HP2). Interestingly, increased HP1 also registered as potentially important for discriminating disabled elders from healthy elders.
Classifications by age and by health status, as predicted from the discriminate analyses, were moderately high in accuracy: 79% by age and 73% by health status. The sensitivity of the discriminating variable set to identify healthy elders from the healthy young was 76%, whereas the specificity in falsely identifying the healthy young as healthy elders was 82%. Whereas there is probably little clinical value in being able to identify healthy elders from healthy young subjects, the above analysis suggests that joint kinematic and kinetic measures are more accurate in discriminating between different groups than is gait speed alone, which had a lower sensitivity, 54%, but similar specificity, 73%. It probably makes sense that specificity would be higher than sensitivity, as there are likely fewer young people who walk slower than expected compared with healthy elders who walk faster than expected.
The sensitivity of the discriminating variable set to identify disabled elders from healthy elders was almost 74%, whereas the specificity in falsely identifying healthy elders as disabled elders was slightly lower at 70%. The kinematic and kinetic variables offered a more sensitive and specific mechanism for identifying disabled elders than did gait speed alone, which had sensitivity and specificity of 60 and 62%, respectively. The low kappa statistic for discriminating by gait speed also suggested a high degree of chance agreement, thus further emphasizing the importance of examining muscle-joint interactions during gait in elderly groups rather than relying on time-distance measures.
Although there is a clear need to better understand the differential effects of aging and disablement on function, and the above analysis suggests that differences in age-related and disability-related neuromuscular adaptation exist, the variables identified as age-related and disability-related discriminators do not entirely agree with results of past studies. To be confident that age and disability effects can truly be separated, enabling the use of gait analysis to reliably identify those at risk for disablement, these discrepancies need to be better understood.
With respect to age-related discriminators, a reduction in peak ankle-plantar flexion angle, shown in the present study to discriminate groups by age and not by health status, appears consistent with past studies (3, 10, 13). Agreement regarding increased late-stance knee power absorption in elderly subjects is varied. Winter et al. (32) reported an increase in late-stance knee absorption (our KP4, Winter's K3 power) in fit elders compared with healthy young subjects, and Kerrigan et al. (13) reported increased late-stance knee power absorption in elders when they were walking faster than healthy subjects. Whereas this finding is consistent with the findings of the present study, other studies (3, 10) have not shown this effect. We also found hip extensor power (concentric, early stance) to be significantly increased for elders compared with healthy young subjects, consistent with the findings of DeVita and Hortobagyi (3), who reported increased positive hip work for healthy elders who walked at the same speed as young adults. However, studies by Judge et al. (10) and Kerrigan et al. (13) and a prior report of ours (19) did not show a significant difference in hip extensor power between young and old subjects.
The study by Kerrigan et al. (13) also highlighted the effect of significantly reduced hip extension in elderly subjects that persisted when elders walked faster than young subjects, whereas we and others (3, 10) found no significant difference in hip extension by age group. In addition, DeVita and Hotobagyi (3) reported significantly increased hip extensor angular impulse (area under the joint moment curve) and significantly decreased hip flexor angular impulse for older adults compared with young adults; however, in our study, there were no significant differences between young and elderly healthy subjects in either hip extensor or hip flexor moment, which is in agreement with Kerrigan et al. (13). DeVita and Hortobagyi (3), however, did not report peak moment values, so our joint moment results and those of Kerrigan et al. (13) may not be comparable to theirs.
With respect to disability-related discriminators, some of the identified variables in this report have been shown in prior studies to discriminate by age. In contrast to the studies by others (3, 10, 13, 32), but in agreement with our past work (19), we found ankle plantar flexor power to be a strong discriminator of health status and not of age-related changes in gait. The study by Winter et al. (32) showed a significant reduction in late-stance ankle plantar flexion power for elderly subjects (our AP2, Winter's A4 power); however, it is important to recognize that the study by Winter et al. did not control for gait speed differences, as we did here and in our previous work (19), in explaining the reduction of ankle plantar flexor power. Furthermore, the study by Judge et al. (10) also found no significant difference between young and elderly groups' ankle plantar flexor power when correcting for step length differences. However, the study by Kerrigan et al. (13) found that significantly reduced ankle plantar flexor power persisted when healthy elders walked faster than young subjects, and DeVita and Hortobagyi (3) reported significantly reduced ankle plantar flexor work in healthy elderly adults who walked at the same speed as young adults.
We also found reduced peak ankle plantar flexor moment to discriminate by health status, and not by age, which is in contrast to the age-related decrease in ankle plantar flexor angular impulse reported by DeVita and Hortobagyi (3). Other disability-related discriminatory variables, i.e., increased hip flexor moment in late stance, increased ankle dorsiflexor moment in early stance, and increased hip flexor eccentric (absorption) power in mid-to-late stance, have not been reported in past studies to be age dependent, and thus our findings are at least indirectly in agreement with prior studies in that respect. Importantly, the finding of increased hip power absorption in the disabled group compared with the healthy groups is consistent with our previous investigations of mechanical energy expenditures during gait in elders with lower extremity impairments and pathology (20, 21).
Whereas there are many consistencies among different studies that have examined the age-related changes in lower extremity neuromuscular function in gait, the above discussion, in concert with a recent review of some of the above studies (17), suggests that there are many critical differences among investigations that preclude one from equivocally identifying the key neuromuscular adaptations that undergird age-related and disability-related changes in gait. However, a better understanding of the interrelationships among these kinematic and kinetic variables may help explain some of the differences among studies and work toward developing a clinically informative understanding of the underlying mechanisms of neuromuscular adaptations.
Neuromuscular adaptive mechanisms during gait with age and disablement. The second hypothesis of our study was that the primary neuroadaptive mechanisms that discriminate between healthy and functionally limited elders will involve the hip musculature. Although this hypothesis was partially supported, hip flexor absorption power and hip flexor moment discriminated only between healthy and disabled groups, we also found ankle dorsiflexor moment, ankle plantar flexor moment, and ankle plantar flexor power to be strong discriminators between healthy elders and elders with functional limitations, with the largest discriminating effects coming from ankle plantar flexor moment (Fig. 3). In addition, we found hip extensor power generation to discriminate by age and, to a lesser extent, by health status. These results suggest that factors that limit gait in elders with lower extremity dysfunction are not simply due to changes in motor control of single muscle-joint systems, but appear to be tightly coupled interactions among multiple muscle-joint systems, as previously suggested by DeVita and Hortobagyi (3), albeit to explain age-related differences.
The correlation matrices in Table 6 show the multifaceted and varied relationships that exist among the three groups of subjects. The first observation is that the strength and number of significant correlations increase with age and even more so with disability: of the five significant correlations found for the young adults, only one was unique to the young subjects, with the other four existing for all three groups; of the seven significant correlations for healthy elders, one was unique to healthy elders and two were common to the disabled elderly group; and of the 13 significant correlations for the disabled elders, seven were unique to that group. This finding alone suggests that motor control of gait in elders, particularly those with lower extremity dysfunction, becomes exceedingly coupled and potentially less able to adapt to shifting demands of the environment. This finding may also reveal the existence of a neurophysiological adaptation of central control pathways that tightly regulate the firing patterns of lower extremity muscles in disabled populations. To understand these adaptations, one must examine closely the relationships among lower extremity biomechanical variables for all three groups.
Relationships independent of age or health status. Four relationships were common to all three groups of subjects studied: peak ankle plantar flexor moment was directly related to peak ankle plantar flexor power; peak hip flexor moment was directly related to peak knee extensor power absorption; peak knee extensor power absorption was directly related to peak hip flexor power absorption; and peak hip flexor power absorption was directly related to peak hip flexor moment. The relationship between ankle plantar flexor moment and power for the three groups illustrates that ankle plantar flexor power-generating capacity is closely coupled to plantar flexor moment generation, which probably becomes more important with limited dynamic range of movement of the ankle, as observed for the elderly groups.
The post hoc partial correlations among the peak hip flexor moment, peak hip power absorption, and peak knee extensor power absorption showed that the relationship between hip flexor power absorption and knee extensor power absorption was moderated by hip flexor moment. This suggests that increased hip flexor moment in the latter portion of the stance phase increases the amount of thigh segment energy that needs to be dissipated by hip flexors and knee extensors, hence the proportional increase in hip flexor and knee extensor power absorption. The fact that peak hip extension angle was directly related to peak hip flexor power absorption and peak hip flexor moment but only related to hip flexor power absorption through hip flexor moment (as the relationship between HR1 and HP2 was nonsignificant when partialing out HM2) is interesting, considering that peak hip extension was not different between groups (Table 3), but hip flexor moment was significantly higher for the disabled subjects (Table 4). This suggests that hip flexor moment is equally sensitive to hip extension for the three groups, but offset higher for the disabled subjects. As suggested by Kerrigan et al. (13) and Judge et al. (10), reduced hip extension in elders may be due to increased anterior tilt of the pelvis, which potentially allows a larger moment arm for the hip flexors to act on.
Relationships unique to healthy subjects. Peak ankle dorsiflexion moment was significantly and inversely related to peak ankle plantar flexion moment for the healthy young subjects only. This finding suggests that young healthy subjects control the amount of "push-off" ankle torque by altering their center of mass position, hence center of pressure position relative to the ankle (ankle moment arm), during the stance phase; a posterior shift would tend to increase dorsiflexor moment while decreasing plantar flexor moment, whereas an anterior shift would tend to have the opposite effect. The ability to shift the locus of the center of pressure excursion relative to the ankle may represent a feedback control mechanism, perhaps somatosensory in nature, for regulating propulsion. Neither healthy elders nor disabled elders exhibited this relationship, suggesting an age-related reduction in somatosensory function of the ankle and/or sole of the foot.
Peak ankle dorsiflexor moment in healthy elders was significantly and directly related to peak ankle plantar flexor power in late stance for healthy elderly subjects only. The direct (positive) relationship, coupled with the fact that peak ankle dorsiflexor moment occurs in the early-stance phase and peak ankle plantar flexor power occurs in the late-stance phase, suggests a mechanism whereby ankle dorsiflexors stabilize the trunk in the early-stance phase when the contralateral limb is executing push-off. No such relationship existed for the healthy young or disabled elders, although the relationship for the disabled elders was close to reaching significance (Table 6).
Relationships unique to elderly subjects. Peak ankle plantar flexion angle was significantly and directly related to peak ankle plantar flexor power, and early-stance ankle dorsiflexor moment was significantly and directly related to late-stance knee extensor absorption power, for both the healthy elders and disabled elders, but not healthy young subjects. The relationship between peak ankle plantar flexor angle and peak concentric power of the plantar flexors, occurring only in elders, confirms that restricted movement of the ankle, which may occur for a number of reasons (17), is associated with reduced ankle power at terminal stance. The fact that healthy elders did not have a similarly reduced ankle power output as disabled elders (Table 5), while exhibiting the same reduction in ankle plantar flexion (Table 3), suggests that healthy elders are able to generate enough ankle plantar flexor moment (Table 4) to compensate for reductions in ankle range of motion. This assertion is supported by the moderately strong positive correlation between ankle plantar flexor moment and power observed for only the healthy elders, as discussed above.
Both healthy and disabled elders also demonstrated a significant correlation between peak ankle dorsiflexor moment and late-stance knee absorption power that was not observed in young subjects. As we argued above for the healthy elders' relationship between peak dorsiflexor moment and plantar flexor power, it is plausible that the increased dorsiflexor moment, which was significant for disabled elders (Table 4), functions to stabilize the trunk to permit the contralateral leg to generate a greater push-off, thus increasing the energy needed to be dissipated by the knee extensors of the advancing preswing leg. It is likely, however, that healthy and disabled elders in advancing the preswing leg use different mechanisms, as disabled elders did not increase ankle plantar flexor power generation proportional to increases in ankle dorsiflexor moment. Rather, the mechanism used by disabled elders appears to involve the hip muscles.
Relationships unique to disabled elders. Peak ankle dorsiflexor moment in disabled elders was significantly and directly related to peak hip flexor moment and peak hip flexor power absorption for only the disabled elderly subjects. As noted above, peak ankle dorsiflexor moment was also related to late-stance peak knee absorption power for disabled elders, as well as healthy elders. As argued above, the relationship between early-stance peak ankle dorsiflexor moment and latestance peak hip moment and knee and hip powers suggests that ankle dorsiflexors act to increase trunk stabilization in response to the increased effort of the contralateral limb to provide forward propulsion of the body and advancement of the swing leg.
However, peak ankle plantar flexor power was significantly and directly related to peak hip flexor moment, peak knee extensor absorption power, and peak hip flexor power absorption for only the disabled elders. The positive relationship finding is somewhat unexpected, as one might expect an inverse relationship between ankle plantar flexor power and hip flexor moment in this group that would indicate a compensatory mechanism for reduced plantar flexor output, assisting with swing leg advancement as we previously suggested (17). Recalling that ankle plantar flexor power was significantly lower in the disabled group compared with the healthy groups, but that hip flexor moment (as well as hip absorption power) was significantly higher for the disabled group compared with the healthy group, suggests that any further increase in ankle plantar flexor output would require the hip flexors to further increase their moment and, consequently, increase power absorption of the hip flexors. However, because hip flexor moment peaks before ankle plantar flexor power peak, it is unlikely that the hip flexors are responding to increased muscular output of the ankle plantar flexors.
Given the reduced peak ankle plantar flexion angle in elders, also directly related to hip flexor moment (for only disabled elders), and its link to reduced ankle plantar flexor moment and power in disabled elders (Table 6), the role played by hip muscles in the late stance for the disabled group must be to provide the primary source of propulsion of the body. Because the hip flexors (iliacus, psoas) and biarticular hip flexors/knee extensors (sartorius, rectus femoris, and tensor fascia latae) are acting eccentrically when hip flexor moment peaks, the increased tension of these muscles could drive the pelvis and trunk forward in a manner to increase whole body momentum. It, therefore, makes sense that ankle dorsiflexors, which peak in moment for the contralateral limb at approximately the same time, would be increased to help stabilize the trunk, as predicted in prior dynamic simulations (26, 34).
The only biomechanical variable examined that did not correlate significantly with any other variables for any of the groups was peak hip extensor power, although it did approach significance for the disabled elders with peak ankle dorsiflexor moment (Table 6). As discussed above, peak hip extensor power was also the only variable to discriminate among all three groups of subjects; its value increased with age and more so with disablement (although age discrimination was stronger than discrimination by health status, as shown in Fig. 3). As with ankle dorsiflexors, the hip extensors primarily function to stabilize the trunk in the early-stance phase. The relative independence of hip extensor peak power to magnitudes of other lower extremity biomechanical variables, however, suggests that the hip extensors respond to the state of the upper body after heel strike. Not surprisingly, increases in the peak power of hip extensors with aging and with disablement suggest an increasing need to stabilize the trunk in gait after the transition from single-limb to double-limb support.
Limitations and conclusions. An important aspect of studies on age-related changes in gait is experimental control of gait speed. In the studies by Judge et al. (10) and Kerrigan et al. (13), elderly subjects were studied both at their preferred gait speed and at their fastest gait speed. Fast gait trials were used to determine differences from young subjects that are not related to gait speed. The disadvantage of this approach is that subjects are required to walk at a nonpreferred, nonnatural pace, which is probably nonrepresentative of subjects' optimal neuromuscular function. In the report by DeVita and Hortobagyi (3), the elderly subjects studied walked at the same self-selected preferred velocity as young subjects. Subjects in the present study, however, walked at significantly different gait velocities. Although the design selected by DeVita and Hortobagyi is perhaps the preferred way to control for confounding effects of gait speed among healthy subjects, functionally limited subjects typically walk at a considerably lower self-selected gait speed. Presumably, self-selected gait speed optimizes energy and neuromotor control demands. Instead of requiring subjects to walk at a nonnatural pace for comparison purposes [such as in the designs of Judge et al. (10) and Kerrigan et al. (13)], we chose instead to statistically control for gait speed.
There are other limitations of the present study. We did not examine kinematic and kinetic variables in the frontal and transverse planes, which might also be important for discriminating between age and disability changes in gait. We also did not include pelvic or upper body kinematics, which might also have been useful for describing the effects of age, gender, and disability. We would, however, likely have needed a much larger sample to produce reliable results, considering the already large number of variables that we report on in this paper. Furthermore, we only analyzed the net joint moments and powers and use these, as do many other studies, to infer what is happening with the musculature surrounding the joints. More concrete solutions to uncovering the nature of neuromuscular adaptations can probably only be obtained by examining the actual roles of individual muscles, as opposed to the "net" effects to which inverse dynamic analysis is limited. Forward dynamic simulations and induced acceleration analyses, which are gaining popularity (11, 24, 26, 28, 34), may also be suited to answer some of these questions.
Another potential limitation is the selected 50-yr-old age cut point for stratifying the healthy sample into young and elderly groups. The age distribution of the healthy group was bimodal, having modes at 30 and 70 yr of age, with only 9 of the 82 subjects between the ages of 40 and 60 yr. As it was not feasible to create three groups, the reasonable cut point of 50 yr was selected to create two groups of healthy subjects, with each having a close-to-normal age distribution. Finally, another potential confounder in the study was the varied underlying pathologies of the sample of functionally limited elders. This reflects a reality in disability research: it is highly improbable that one could obtain a sample of elders with identical pathologies and comorbidities. Moreover, we purposely chose not to limit the generalizability of our results by presuming that a single diagnosis yields singular impairments or functional limitations. All subjects in the disabled group shared a common underlying impairment: lower extremity muscle weakness. Furthermore, all subjects reported functional limitations, and all underlying pathologies for this sample were classified as musculoskeletal pathologies.
Despite these limitations, our data lead us to several important conclusions.
First, we conclude that analysis of the sagittal plane lower extremity biomechanics (kinematics, moments, and powers) during gait can be used to separate aging effects from disability effects, specifically those due to lower extremity pathology, such as arthritis. We conclude further that decreased ankle plantar flexion angle, increased knee extensor power absorption, and increased hip extensor power are age-related changes in gait, whereas increased ankle dorsiflexor moment, decreased ankle plantar flexor moment, increased hip flexor moment, decreased ankle plantar flexor power, and increased hip flexor absorption power are changes related to lower extremity dysfunction. We found that this latter set of variables enabled a 73% correct classification for disabled and nondisabled subjects over 50 yr of age.
Second, we conclude that the above between-group differences and differences in the interrelationships among variables point to different mechanisms of neuroadaptive patterning with age and disablement. Relationships among some biomechanical variables were persistent across age and disability groups, suggesting a nonadaptive aspect to motor control that is probably reflex driven as it involved eccentric contractions of hip flexors and knee extensors. Biomechanics of the lower extremity joints appear to become increasingly coupled with aging and disablement, with an increased reliance on the hip flexors and extensor to provide both propulsion and stability. Our data suggest the hip flexors are central to late-stance propulsion for elders with lower extremity dysfunction, and the contralateral ankle dorsiflexors play an increasingly vital role in providing trunk stability during late stance.
In summary, the findings of this study suggest that biomechanical analysis of gait provides the necessary information to reveal the underlying determinants of movement dysfunction, relevant to a motor control framework of physical therapy intervention. Indeed, more detailed biomechanical studies are needed to better understand these neuromuscular adaptations with disability and whether they can be reduced or reversed with appropriate rehabilitative intervention.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
This work was supported in part by National Institutes of Health Grants R01-AG-12561 and R01-AR-45278 and by the National Arthritis Research Foundation.
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  H. H. C. M. Savelberg and K. Meijer The Effect of Age and Joint Angle on the Proportionality of Extensor and Flexor Strength at the Knee Joint J. Gerontol. A Biol. Sci. Med. Sci., November 1, 2004; 59(11): 1120 - 1128. [Abstract] [Full Text] [PDF]
|
|
| ||||||
63% cycle for this subject).