Abstract
Background A reduced capacity to modify gait to the environment may contribute to the risk of falls in people with poststroke foot drop using an ankle-foot orthosis.
Objective This study aimed to quantify their capacity to restore steady gait after a step modification.
Design This was a cross-sectional, observational study.
Methods Nineteen people in the chronic phase (>6 months) after stroke (mean age=55.0 years, SD=10.1) and 20 people of similar age (mean age=54.6 years, SD=12.0) who were able-bodied were included. Participants were instructed to avoid obstacles that were suddenly released in front of the paretic leg (stroke group) or left leg (control group) while walking on a treadmill. Outcomes were success rates of obstacle avoidance as well as post-crossing step length, step duration, hip flexion angle at foot-strike, and peak hip extension of the steps measured within 10 seconds following obstacle release.
Results Success rates of obstacle avoidance were lower for people poststroke. Moreover, their first post-crossing step length and duration (ie, the nonparetic step) deviated more from steady gait than those of people in the control group (ie, the right step), with lower values for people poststroke. Similar deviations were observed for post-crossing hip flexion and extension excursions.
Limitations People poststroke were relatively mildly impaired and used an ankle-foot orthosis, which may limit the generalizability of the results to other populations poststroke.
Conclusions People with poststroke foot drop using an ankle-foot orthosis had reduced gait adaptability, as evidenced by lower success rates of obstacle avoidance as well as an impaired capacity to restore steady gait after crossing an obstacle. The latter finding unveils their difficulty in incorporating step modifications in ongoing gait.
Walking is a context-specific activity involving instant gait adjustments to a continually changing environment. Gait adjustments are essential when walking on uneven or cluttered terrain to secure adequate foot placement to local environmental features, such as obstacles.1,2 Gait adaptability, including the capacity to modify gait to the environment, is related to fall risk,3 as most falls occur due to trips, slips, or misplaced steps.4
People in the chronic phase after stroke are at elevated risk of falling,5–8 and even in individuals who are well recovered and live in the community, fall rates are high, which may be attributed to impaired gait adaptability.9–15 For example, den Otter and colleagues10 found that people poststroke were less successful in avoiding obstacles on short notice than people in their control group who were able-bodied. Interestingly, the avoidance maneuver induced larger disturbances in the locomotor rhythm in people who were severely affected by stroke than in people who were only mildly affected or who participated in the control group. Although the length and duration of the post-crossing stride were almost instantaneously restored to steady gait values in people who were mildly or not affected by stroke, this was clearly not the case for the people who were severely affected, suggesting that the number of steps required to settle into steady gait after a step modification was considerably larger. Unfortunately, we do not know how much larger because only a single post-crossing stride was analyzed.10 In addition, the group of people who were mildly affected might have performed so well in this study due to the availability of a handrail. Handrail support may well facilitate restoration of steady state walking after crossing the obstacle.
These results expand our insight into gait adaptability in people poststroke: not only is their capacity to avoid obstacles diminished but ostensibly also their capacity to restore steady gait after crossing obstacles. A slower restoration of coordination after a perturbation reflects reduced coordinative stability16,17 and reduced gait adaptability.11–13 So far, however, the number of studies focusing on the restoration of steady gait is limited. Moreover, the studies that have been performed show equivocal results. Some studies showed impaired capacity to restore gait after a step modification,9,11,13 whereas another study demonstrated that there was no impairment in gait restoration.10 This discrepancy between studies might be attributed to differences in the prevalence of a foot drop in the studies' participants. However, none of the abovementioned studies reported on foot drop prevalence. This lack of data is unfortunate because foot drop reduces foot clearance during gait, thereby affecting people's capacity to avoid obstacles. The standard care for foot drop is use of an ankle-foot orthosis (AFO). Although an AFO can improve walking,18 the effect of an AFO on gait adaptability has not yet been investigated. We suggest that both the foot drop itself and the restriction of ankle mobility by the AFO may contribute to reduced gait adaptability.
In the current study, we examined the capacity to restore steady gait after a step modification induced by obstacle crossing in people with poststroke foot drop that is corrected by an AFO. To this end, a stroke group of 19 people after stroke and a control group of 20 people of similar age who were able-bodied participated. All participants were instructed to avoid an obstacle that was dropped unexpectedly in front of one foot while walking on a treadmill. We compared success rates of obstacle avoidance and the capacity to re-establish steady gait during the post-crossing steps between stroke and control groups. In line with previous studies,10,15 we expected lower success rates in people poststroke than in people who were able-bodied. Given the reduced propulsion-generating capacity and postural stability when standing on their paretic leg,19–21 we further expected a diminished capacity to restore steady gait in people poststroke, reflected in more persistent deviations in post-crossing step length, step duration, and hip kinematics.
Method
Participants
For the experimental group, 19 community walkers (Functional Ambulation Categories22 rating of 5) in the chronic phase (at least 6 months) after stroke were invited. All participants were recruited from a parallel study23 that investigated the effect of functional electrical stimulation of the peroneal nerve on gait adaptability. As an inclusion criterion for this latter study, the participants had to have stroke-related foot drop for which they used an AFO. Furthermore, participants had to be able to walk comfortably on the treadmill without handrail support at 2 km/h or faster. The following clinical characteristics were scored at inclusion: balance (Berg Balance Scale24), lower extremity muscle strength (Motricity Index25), and lower extremity motor selectivity (Fugl-Meyer Assessment26). For the control group, we invited 20 participants of similar age without gait limitations. The Table lists group characteristics. All participants gave written informed consent before the start of the study.
Characteristics of the Groupsa
Experimental Setup and Procedure
The people poststroke walked on a treadmill at 2 or 3 km/h, whichever was closest to their preferred treadmill walking speed. In the control group, people also were asked to walk at 2 km/h or 3 km/h in order to arrive at similar average walking speeds for both groups. All participants wore their own daily shoes and a safety harness to prevent them from falling. An electromagnet held an obstacle (length, width, and height: 40 × 30 × 1.5 cm, respectively) to a bar at the front side of the treadmill just above the treadmill belt27,28 (Fig. 1A).
Schematic diagram of (A) the experimental setup and (B) the definition of steps and strides in each of the avoiding strategies (short step strategy [SSS], long step strategy [LSS]). A step was defined by the foot contact of one leg and the preceding foot contact of the contralateral leg, whereas a stride was defined by the foot contact of one leg and the preceding foot contact of the same leg.
We synchronously recorded bilateral hip excursions in the sagittal plane with goniometers (Biometrics SG150, Biometrics Ltd, Newport, United Kingdom; sampling rate=1,000 Hz) that were attached vertically to the lateral side of the trunk and leg (crossing the trochanter major) while standing upright. The positions of the reflective markers attached to each foot and to the front edge of the obstacle were recorded with a 6-camera, 3-dimensional motion analysis system (Vicon Motion Systems, Vicon-UK, Oxford, United Kingdom; sampling rate=100 Hz23,29). Marker position recordings were processed online using custom-made software, allowing for obstacle release between mid-stance and mid-swing of the paretic leg of participants poststroke and the left leg of participants in the control group.23,28,29 The timing of release was randomly distributed over trials and uniformly across participants. The obstacle was always released in front of the paretic leg of participants poststroke and the left leg of participants in the control group. Participants were instructed to maintain a distance of about 10 cm from the foot to the (unreleased) obstacle at the moment of foot-strike (Fig. 1A) and to avoid the obstacle when released without stepping aside with the paretic (or left) leg, whereas the other foot could be placed freely. Thus, participants could apply 1 of 2 different strategies30: (1) shortening of the ongoing paretic/left stride to make an additional step in front of the obstacle before crossing it (short step strategy [SSS]) or (2) lengthening of the ongoing paretic/left stride to directly cross the obstacle (long step strategy [LSS]; see Fig. 1B and video, available below). It has been reported that the strategy choice strongly depends on the timing of obstacle release.30,31 By experimentally manipulating this timing, we were able to elicit both SSS and LSS responses.
A video demonstrating the short step strategy (SSS) and long step strategy (LSS) used to avoid the obstacle during treadmill walking.
At the beginning of the session, the participants had ample time to familiarize themselves with unsupported treadmill walking (at least 5 minutes). Then, a so-called unperturbed walking trial was performed during which the participants' normal gait pattern was registered for 2 minutes. Subsequently, participants performed 5 obstacle-crossing trials to become familiarized with obstacle negotiation, followed by 30 experimental obstacle-crossing trials. Recordings started when the obstacle was attached to the magnet and stopped 10 seconds following obstacle release.
Data Processing
For each trial, we determined whether the obstacle was avoided successfully. Contacting the obstacle by stepping with (a part of) the foot on the obstacle or paretic/left steps beside the obstacle was classified as failure.23 Mean success rates were quantified, stratified for LSS and SSS.
Steps and strides were defined on the basis of foot-strike instances (Fig. 1B), derived from foot marker positions in the vertical and sagittal planes. Step lengths (length) and step durations (duration) were obtained from the sagittal marker positions, after transformation to a reference frame moving with the belt.32 Bilateral hip angles were low-pass filtered (fourth-order, bidirectional Butterworth filter at 10 Hz) and time normalized to the stride. Hip flexion angle at foot-strike (at start of the stride; hipfl) and maximum hip extension (hipext) angle were computed for each leg at every stride. From the unperturbed trial, step duration, step length, hip flexion, and hip extension (durationunper, lengthunper, hipflunper, and hipextunper, respectively) were averaged over 30 strides, starting at stride 50. We included both unsuccessful and successful trials as long as a step modification was involved in the crossing maneuver. Gait was assumed to be substantially modified when crossing stride length deviated significantly from the unperturbed stride length. That is, we excluded trials if the crossing stride length remained within the average unperturbed stride length ±1.96 × SD, totaling 10.4 (±5.3) and 4.9 (±4.8) trials for the stroke and control groups, respectively.
From each trial, length(n), duration(n), hipfl(n), and hipext(n) were computed for all n post-crossing steps (for length and duration) and strides (for hipfl, hipext). Specifically, step (stride) n=1 was defined as the first step (stride) with the nonparetic/right leg after obstacle crossing (Fig. 1B). Subsequent steps (strides) were with alternating legs. Thus, odd step (stride) numbers refer to steps (strides) with the nonparetic/right leg, and even numbers refer to steps (strides) with the paretic/left leg.
Subsequently, all post-crossing variables X(n) were normalized to their unperturbed counterparts Xunper, according to X*(n) = 100 × X(n)/Xunper[%], where X(n) represents each of the variables of interest of the obstacle-crossing trials (ie, length(n), duration(n), hipfl(n), and hipext(n)) and Xunper represents their averaged counterparts of the unperturbed trial (ie, lengthunper, durationunper, hipflunper, and hipextunper). These data then were stratified depending on the obstacle-crossing strategy used (ie, either SSS or LSS) for each trial per participant. All normalized variables X*(n) were averaged across trials for each participant and each strategy. These average values were entered into the statistical analysis.
Data Analysis
For each X*(n) series, a 2-way repeated-measures analyses of variance (ANOVA) with group as a between-subjects factor (2 levels: stroke and control groups) and step (stride) as a within-subjects factor (n=11 levels; at least n=11 post-crossing steps [strides] were available for all participants) was conducted, separately for SSS and LSS. The alpha level in the ANOVAs was set to .025 to correct for dual testing of dependent variables (ie, separately for SSS and LSS). If sphericity was violated, a Greenhouse-Geisser correction was applied (adjusted degrees of freedom are reported). Effect sizes are reported as ηp2. For significant step (stride) × group interactions, 11 post hoc, 2-tailed, independent-samples t tests with Bonferroni correction were conducted. For significant main effects of step (stride), post hoc contrast tests were performed.
Role of the Funding Source
The contribution of Dr Roerdink was supported by Veni grant 451-09-024 from the Netherlands Organization for Scientific Research (NWO). The contribution of Dr van Swigchem was supported by small unrestricted grants from Ness Netherlands BV (now Bioness) and Unu BV in the Netherlands.
Results
All participants in the stroke group used an AFO for stroke-related foot drop. The AFO was a customized polypropylene brace of various designs that restricted ankle plantar flexion while leaving free at least 15 degrees of dorsiflexion mobility. In 2 participants, the AFO was a simple spring-type brace providing a dorsiflexion moment about the ankle.23 Furthermore, participants poststroke had relatively mild balance problems and demonstrated mild to moderate paresis of the affected leg (Table). In 1 patient, a trip occurred, and this trial was excluded from further analysis. On average, 25.1 (SD=4.8) and 19.6 (SD=5.3) trials were included for participants in the control and stroke groups, respectively. Three participants (2 in the stroke group and 1 in the control group) showed a strong preference for LSS, regardless of variations in the timing of obstacle release. Because these participants did not apply SSS in any of the trials, statistical analyses of SSS were performed for 17 people in the stroke group and 19 people in the control group. Mean success rates were 62.9% (SD=24.6) and 29.1% (SD=34.4) in the stroke group and 98.2% (SD=4.0) and 97.7% (SD=6.3) in the control group for LSS and SSS, respectively (P<.001 for group differences).
Normalized Step Length (Length*) and Step Duration (Duration*)
In general, the first post-crossing step length and step duration deviated more from steady gait in the stroke group (ie, the nonparetic step) than in the control group (ie, the right step; Fig. 2). For LSS, this observation was confirmed by significant group × step interactions for duration* and length* (F3.05,112.72=24.85, P<.001, ηp2=0.40 and F1.96,72.48=11.00, P<.001, ηp2=0.23, respectively). Post hoc t tests for n=1 revealed 14.2% and 22.4% lower values in the stroke group compared with the control group for duration* and length* (t37=5.88, P<.001 and t37=3.85, P<.001, respectively). For duration*, the same held for n=3 (t37=3.80, P=.001). For SSS, significant group × step interactions were observed only for duration* (F3.57,121.35=29.28, P<.001, ηp2=0.46), with 8.3% and 4.4% lower values for n=1 and n=3, respectively, for the stroke group compared with the control group (t34=6.60, P<.001 and t34=3.51, P=.001). For length* in SSS, the group × step interaction tended toward significance (F2.43,82.55=3.07, P=.04, ηp2=0.08).
Mean (95% confidence interval) normalized step lengths (upper panels) and step durations (lower panels) for the crossing step (ie, step n=0) and subsequent post-crossing steps (ie, steps n=1–5) for the long step strategy (LSS) and the short step strategy (SSS). Steps n=6–11 are not visualized because steady gait was fully restored from step n=5 for both groups. Values were normalized to unperturbed counterparts. Significant differences between groups for a given step n are indicated with asterisks. Note that even steps were with the paretic (left) leg, whereas odd steps were with the nonparetic (right) leg (see Fig. 1). Post-crossing steps (ie, n≥1) were included in the statistical analysis.
Normalized Hip Flexion (Hipfl*) and Hip Extension (Hipext*) Excursions
Average kinematic profiles for the groups are provided in Figure 3. Figure 4 shows that hipfl* and hipext* deviated from unperturbed gait, in particular for n=2 (ie, the first stride with the paretic/left leg; Fig. 1B). For hipfl*, there was no difference between groups; for hipext*, the deviation was greater in the stroke group than in the control group. Specifically, for LSS, a significant main effect of stride was observed for hipfl* (F2.53,93.63=43.64, P<.001, ηp2=0.54) and a significant group × stride interaction was observed for hipext* (F3.22,119.00=12.57, P<.001, ηp2=0.25). Post hoc tests revealed that hipfl* was significantly larger for n=2 than for all other strides (F1,37=84.43, P<.001, ηp2=0.69). Hipext* for n=2 was 43.0% smaller in the stroke group than in the control group (t37=4.29, P<.001). For SSS, a similar pattern of results was found for hipfl* (main effect of stride: F3.31,112.37=9.65, P<.001, ηp2=0.22), with overall larger values for n=2 (F1,34=19.55, P<.001, ηp2=0.37). The group × stride interaction for hipext* was absent (F2.73,92.83=1.08, P=.36, ηp2=0.03).
Time-normalized hip angle courses of the paretic leg (stroke group [gray lines]) or left leg (control group [black lines]), averaged over all participants within the groups. The crossing stride and 2 subsequent strides are depicted for the short step strategy (SSS; upper panel) and long step strategy (LSS; lower panel). Unperturbed reference joint angle courses are given with dotted lines. Note that the crossing stride and strides 2 and 4 are subsequent strides that started with the paretic (left) leg. The nonparetic (right) leg is not depicted. Time instants of paretic leg foot-strike (fs) are indicated with vertical dashed lines. Instants of hip flexion (hipfl*) at the start of stride 2 and peak hip extension (hipext*) during stride 2 are indicated with arrows.
Mean (95% confidence interval) normalized hip flexion angle at foot-strike (hipfl*) and peak hip extension excursion (hipext*) for the crossing stride (ie, stride n=0) and 11 subsequent strides (ie, strides n=1–5) for LSS and SSS. Strides n=6–11 are not visualized because steady gait was fully restored from stride (n=5) for both groups. Values were normalized to unperturbed counterparts. Significant differences between groups for a given stride n are indicated with asterisks. Note that even strides were started with the paretic (left) leg, whereas odd strides were started with the nonparetic (right) leg (see Fig. 1). Post-crossing strides (ie, n≥1) were included in the statistical analysis.
An additional analysis for the strategy that yielded most pronounced group differences (ie, LSS) demonstrated similar results when only successful avoidance trials were included.
Discussion
The goal of this study was to investigate gait adaptability of people who had poststroke foot drop and used an AFO, using a treadmill-based obstacle-avoidance paradigm with time-critical obstacle presentations to elicit step modifications. Notwithstanding the observation that people poststroke were able to adjust gait to a certain extent, their gait adaptability was strongly reduced compared with that of people who were able-bodied.
First, people poststroke demonstrated lower obstacle avoidance success rates (Table), particularly in SSS, which is considered the favorable strategy for the more difficult obstacle presentations (ie, late in the gait cycle). In these time-critical situations, people poststroke were more often unable to shorten the ongoing stride, resulting in obstacle contacts with the toe. Failures in LSS were due mostly to insufficient lengthening of the stride, resulting in obstacle contacts with the heel.33
Second, people poststroke required more post-crossing steps to return to steady gait. Post-crossing group differences were most pronounced for the LSS obstacle crossing strategy because of the larger crossing step involved (length*(0); Fig. 2). In contrast, the first post-crossing step in the control group already largely matched unperturbed gait (length*(1) and duration*(1) close to 100%; Fig. 2), the first post-crossing step in people poststroke was shorter in length and duration, even in individuals with relatively preserved balance capacities according to clinical evaluation. These findings largely corroborate earlier work of den Otter and colleagues,10 who reported an impaired restoration of walking cadence in the first post-crossing stride. However, this finding was reported only for the group of people who were severely affected by stroke, whereas the group of people who were only mildly affected did not display restoration difficulties after a step modification, even though their clinical scores may be comparable to those of the participants in the present study (ie, their Brunnström stages 5–6 match fairly well with our average Fugl-Meyer Assessment score of 64%). This discrepancy between studies may be explained by the use of handrail support during treadmill walking, which was prohibited in the current study but allowed by den Otter and colleagues.10 Handrail support may have facilitated the return to steady gait after a step modification. In addition, the severity of foot drop in the participants in the study by den Otter and colleagues was not reported. In the current study, all participants with stroke had foot drop and used an AFO to correct this condition. Reduced ankle motor control may contribute to the shorter post-crossing stance phase on the affected leg, albeit that the AFO may have improved this impairment.18,34
For the first post-crossing step, the nonparetic step length of people poststroke was shortened to a greater extent than the right step of participants in the control group (LSS, length*(1); Fig. 2). Moreover, the nonparetic step durations were shorter for people poststroke up to the third post-crossing step (LSS and SSS, duration*(1) and duration*(3); Fig. 2). We propose 2 explanations for these group differences in post-crossing step parameters: (1) an asymmetry in propulsion generating capacity between body sides in people poststroke and (2) a reduction of postural stability during single-limb support on the paretic leg in people poststroke with foot drop.
We start with the former explanation by addressing concomitant changes in post-crossing hip kinematics (Fig. 4). The increased hip flexion angles in both groups at the start of the first post-crossing stride with the paretic (left) leg (n=2) correspond with increased hip flexion angles at the end of the crossing strides and thus roughly reflect how far the foot is placed in front of the trunk at foot-strike after crossing the obstacle, most compellingly so for LSS (Fig. 4). This increased forward foot placement is in line with the increased crossing step length with the paretic (left) leg (length*(0); Fig. 2). The increased forward foot placement relative to the trunk strongly reduces the forward momentum over the leading paretic leg (ie, in the subsequent stance phase of the paretic leg). Both legs, therefore, must deliver more mechanical work to progress the trunk over the paretic stance limb.10,35 Given the asymmetry in propulsion-generating capacity between body sides in people poststroke,19 the weaker paretic leg in particular may be unable to generate the relatively increased propulsion required to progress the trunk forward to maintain speed.35 Thus, the trunk does not displace far enough forward over the supporting foot during single-limb support of the paretic leg. In the current study, this inadequate trunk displacement resulted in smaller peak hip extension excursions of the paretic leg (LSS, hipext*(2); Fig. 4) and smaller first post-crossing nonparetic leg step lengths (LSS, length*(1); Fig. 2) for people poststroke than for people in the control group.
Besides asymmetric propulsion-generating capacity, people poststroke also show postural stability–related asymmetries in stance and swing durations,20 with shorter nonparetic leg than paretic leg step durations in our study (Table). Post-crossing nonparetic step durations were shortened to an even greater extent, as evidenced by significantly lower duration* values for n=1 and n=3 for both LSS and SSS (Fig. 2), suggesting that people poststroke shorten single-limb support stance duration on the paretic leg during the first post-crossing steps. We note that for both groups and for both crossing strategies, the crossing step was larger than the step lengths observed for unperturbed gait (ie, length*(0)>100%; Fig. 2), thereby challenging balance after landing. This balance challenge may be particularly harmful for people poststroke, and in particular those with foot drop, as the weaker paretic leg manifests post-obstacle landing. People poststroke functionally adjusted their gait to this challenging situation by adopting shorter nonparetic step duration, thereby reducing single-support stance duration on the paretic leg.
Finally, we may preclude that obstacle contacts in the failure trials contributed to the results reported for the people poststroke, as an additional analysis of step modifications only for successful trials in LSS demonstrated similar results. We conclude, therefore, that people with poststroke foot drop using an AFO have an impaired gait adaptability compared with people who are able-bodied walking at the same gait speed. This difference is evidenced by a reduced capacity to cross obstacles (Table) as well as by a reduced capacity to restore steady gait after a step modification. We attributed the latter finding to a reduced propulsion-generating capacity with the paretic leg relative to the nonparetic leg as well as to postural imbalance during single-limb support on the paretic leg. These findings of reduced gait adaptability may have implications for safe community ambulation of people in the chronic phase after stroke. Reduction of this capacity in people poststroke might putatively relate to the increased fall risk in this population. However, prospective measurement of falls in daily life is needed to confirm such a relationship. A limitation of this study was that all participants in the stroke group were using an AFO, which may have impaired their adaptability of gait. Hence, one should be cautious in generalizing our results to the population of people with stroke who do not use an AFO for foot drop or to those who do not have foot drop.
Footnotes
All authors provided concept/idea/research design and writing. Dr van Swigchem provided data collection. Dr van Swigchem and Dr Daffertshofer provided data analysis. Dr Weerdesteyn and Dr Geurts provided project management. Dr Geurts provided study participants. Dr Roerdink, Dr Weerdesteyn, Dr Geurts, and Dr Daffertshofer provided consultation (including review of manuscript before submission).
The study protocol was approved by the Medical Ethics Committee of the Arnhem-Nijmegen region.
An abstract of this research was presented at Gait & Mental Function, 1st Joint World Congress of the International Society for Posture & Gait Research; June 24–28, 2012; Trondheim, Norway.
The contribution of Dr Roerdink was supported by Veni grant 451-09-024 from the Netherlands Organization for Scientific Research (NWO). The contribution of Dr van Swigchem was supported by small unrestricted grants from Ness Netherlands BV (now Bioness) and Unu BV in the Netherlands.
- Received March 20, 2013.
- Accepted February 14, 2014.
- © 2014 American Physical Therapy Association