Variability of L is greater in individuals with amputation on uneven terrain (hypothesis 2)
The uneven terrain surface evoked increases in the variability of angular momentum profiles in comparison to flat treadmill walking, as anticipated. In comparison to the average results, the effect of terrain on L variability was more apparent.
The variability of
LC and
LR was greater on UT in individuals both with and without amputation, however it was higher on average in the former group, in support of our second hypothesis. The higher variability of
LC at sound foot contact is consistent with the proposition that a reduced ability to control movement over the prosthetic side stance phase results in greater fluctuations in whole body angular momentum at the foot contact of the swing limb. This would emphasize a greater demand on the sound side for correcting movement to maintain balance whilst walking on the uneven surface. The increase in v
LC only reached significance unilaterally in the control group; a finding that may be related to limb dominance. Although it was not measured explicitly, there were more right than left ‘prosthetic sides’ which may have introduced a bias given the tendency for similarities in lateral preference amongst the general population [
32]. Such a finding would corroborate previous work that exposes a difference in functional roles taken on by the two limbs during gait tasks, with respect to propulsion and control (see [
33] for a review).
It was anticipated that
LR on the prosthetic side would be directly related to the contours encountered, given a reduced capacity to brake during prosthetic single limb stance, and the potential for the passive component to hinder or exaggerate shank progression. Thus, the variability would be increased further on UT on the prosthetic side in comparison to the sound side. The variability of
LR was greater on UT, however values were similar on both sides, and therefore our second hypothesis was only partially supported. As the first 50% of the stride incorporates both the single limb stance of the ipsilateral limb and an initial double support period it is possible that corrections are made by the sound side during push off to redirect the ground reaction force and attenuate potential fluctuations in
L during prosthetic single leg stance [
9]. It is also possible that this is related to differences in prosthetic step length or time. The analysis of the single and double support phases of gait in isolation may provide further insight into the relative contributions of the sound and prosthetic limbs.
Given the variability observed, it is plausible that the lack of a large change in average
L is a result of averaging across several steps, during which the ground reaction force due to the surface profile might either act to increase or reduce angular momentum. For example, it may be only the steps that contacted a descending contour or lower level that lead to an increase in angular momentum. Pearson’s correlations, performed post-hoc to explore this further, revealed only a minority of participants with a significant relationship between the step-to-step change in surface height and
LC (further detail provided in Additional file
2). However, inconsistencies in the extent and direction of the relationship across participants indicated the lack of a uniform, definitive response.
Overall, our findings point to a greater potential for destabilization in the individuals with amputation, and a greater demand on the sound side for controlling movement. However, it is unclear to the extent to which these effects of UT may be ameliorated by appropriate intervention, either via rehabilitation or prosthetic technology. Unimpaired individuals have been shown to refine their movement on uneven terrain over a short period of time [
22]. Participants in this study were only examined over a very short time period and were given limited time to familiarize with the terrain prior to testing; necessary in order to avoid excessive fatigue. It is possible that more efficient movement strategies may have been adopted with further exposure to the terrain surface. This would suggest that problems experienced on uneven ground by individuals with an amputation may be due to a lack of practice and familiarity rather than solely a deficit of the prosthesis.
This study was restricted to the assessment of individuals using passive prostheses. The provision of positive net work during stance to address the propulsion deficit of passive devices has been shown to reduce, although not completely normalize, the range of
L on slopes [
34]. Microprocessor control of powered and non-powered devices may produce more appropriate lower extremity behavior in different loading contexts and aid the user in negotiating uneven terrain (see [
16] for a relevant review). It is likely, however, that the extent to which such a device will facilitate walking on uneven terrain will depend on the effectiveness of the control algorithm employed. For example, should the push-off of a powered device be inappropriately timed due to changes in contact patterns of the foot with the ground, increases in v
LC and v
LR might be observed on UT. Such findings may be of high utility for the identification of the deficits of control algorithms and potential solutions for their refinement. However, regardless of the extent to which the foot replicates biological action, there will be unpredictability introduced by the component itself if the user is not directing its motion. Electromyographic control (see [
35] for an example) that increases the influence of the user on device behavior may lead to an improvement in whole-body coordination on non-level surfaces, permitting safer and more efficient walking.
There were a number of limitations to this study. It is of note that foot contact timings were based on kinematic features, specifically the relative velocities of the feet with respect to the pelvis [
36]. They therefore may not have captured the true moment of interaction of the foot with the ground, affecting the
L values extracted at heel contact.
LR and v
LR, in contrast, would likely be unaffected. The potential effect of incorrectly identifying the correct contact event was explored in the 60-stride time series of two unimpaired participants. Shifts of 0, 1 and 2% of the gait cycle, corresponding to up to 2.5 frames, were introduced to the foot contact event timings with a uniform random distribution across strides.
LC and v
LC were re-calculated and compared to the original values. Differences in average
LC between the manipulated and original time series were less than 10% of the grand mean
LC and comparable across terrains, increasing confidence in our comparisons. The differences in v
LC were considerably higher; between 20 and 100% of the grand mean. When the walking surface fluctuates, however, it seems likely that the variability in foot contact timings would be under- rather than overestimated when based on kinematic patterns alone. In this case, the differences in v
LC between FT and UT would actually be greater. Nevertheless, more precise identification of contact events through the use of foot switches or accelerometers could lead to our results for v
LC being refuted.
The exclusion of the arms, hands and head from the model may have led to inaccuracies in estimation of
L, however, the contributions of these segments to sagittal plane
L have been shown to be negligible in comparison to those of the other segments [
6]. The calculation of the inertial parameters of the lower limb of the affected side might be more influential. In the absence of geometrical information and precise center of mass values for the residual limb and prosthesis we chose to modify the inertial properties of the shank segment according to the generic correction presented by Ferris et al. [
28]. A brief examination into the effect of computationally manipulating inertial values revealed that considerable differences in the range of
L could be attributed simply to a difference in the mass and center of mass position values of the lower leg input into the model. In fact, applying no correction for the prosthesis resulted in some cases in no observable difference in
L between the sound and prosthetic sides. It is possible that the use of more precise estimates of inertial properties through reaction board and oscillation techniques [
37] could refute the results of our study, however within-limb findings would likely hold.
Along similar lines, height, mass and preferred walking speed were not matched across groups, all of which have a bearing on absolute values of angular momentum. Although normalization to these factors improves the appropriateness of comparisons across individuals and studies, the results should be interpreted with caution.
Participants were asked to maintain the hand posture they had adopted during familiarization on the uneven terrain, i.e. either no use of handrails, or light touch on handrails. More participants in the amputation group used the handrails, and three participants changed their hand position when they returned to the UT after the FT trial due to a lack of confidence. With the inclusion of hand position as a covariate, no effect of hand position was found (p > 0.05 for all main effects and interactions), however, suggesting that this did not confound our results. Further, the greater use of rails in the amputation group would more likely induce a reduction rather than an increase in variability, and without rails the differences between groups would be larger.
The use of handrails by the majority of participants precluded the assessment of coronal and transverse plane
L, and medial-lateral dynamics, both of which may provide complementary insight into the deficits of the prosthetic limb and the control strategies employed in light of them. The lack of a subtalar joint, for example, may limit inversion and eversion, with implications for lateral stability [
38,
39]. That most participants were unable to perform the task without handrail use is in itself of interest. Future work focused on whether the source of this inability is due to a mechanical or a perception deficit is warranted.