Persons with stroke walk slower (0.23–0.78 m/s), as compared to age-matched healthy adults (ages 65–80: 0.81–1.51 m/s), presenting with decreased cadence, stride length, and limb segment movement amplitude.1–3 Hip, knee, and ankle movement is less on the involved side leading to shortened step and stride length, increased double limb support time, decreased stance time on the involved side,2–6 and greater step:stride asymmetry.7 Step time on one side is not equal to half of the stride time on the opposite side.7
Although primarily due to lower extremity impairments, decreases in walking velocity post-stroke may also be related to upper extremity impairments commonly seen in persons with stroke. Reduced arm swing leads to decreased transverse thoracic rotation, transverse pelvic rotation, and stride length as compared with walking with “typical” arm swing.8–10 Previous research in healthy adults has shown that counterrotation between the pelvis and thorax serves to decrease angular momentum within the body during walking11–13; therefore, a reduction in transverse thoracic rotation can lead to a similar reduction in transverse pelvic rotation,14 stride length, and walking velocity.
During gait training, clinicians may be observed instructing patients to move their arms reciprocally back and forth (ie, out-of-phase). The purpose is to use reciprocal arm movement to facilitate lower extremity movement, increasing stride length and walking velocity. This approach is supported by research8,11,14–17 that has shown that there are characteristic coordination patterns between upper and lower extremity movement during walking. For instance, in healthy adults, the arms swing together, forward and backward (in-phase) and are synchronized more with the step frequency (2:1 frequency coordination) at slower walking velocities. However, as walking velocity increases above 0.8 m/s, the arms shift from in-phase to a more out-of-phase pattern and are synchronized with stride frequency (1:1 frequency coordination) versus step frequency (2:1 frequency coordination).11,15–17
These coordination patterns are altered when persons present with upper extremity impairments. The noninvolved arm movement of a person with stroke synchronizes with stride frequency (1:1 coordination) at comfortable walking velocities, while the involved arm movement frequency does not synchronize with stride or step frequencies (asynchronous).8,11,15,18 Currently, there are no data to support specific interventions for upper extremity impairments (eg, stroke) and the coordination between the involved arms and legs during walking. Ford et al18 previously examined the use of an external auditory rhythm during walking in persons with stroke. They reported more of a 1:1 frequency coordination on the involved side when individuals stepped to the beat of an increasing metronome. However, frequency coordination on the involved side became asynchronous when individuals were instructed to move the arms to the beat of an increasing metronome.
Clinicians may altogether avoid incorporating specific arm swing patterns during gait training when persons with stroke have more severe upper extremity deficits. Ford et al18 showed that the there was more of a 1:1 frequency coordination on the involved side when instructions were directed at leg movement versus arm movement. However, that approach may detrimental to upper extremity function and walking function, as previous research8,10,11,14–18 has shown a velocity-dependent coupling between upper body motion and lower body motion during walking.
The purpose of this study was to investigate the effects of instructions to move the arms in- versus out-of-phase on upper and lower extremity movement patterns during walking at increasing treadmill velocities in individuals post-stroke. Based on previous research,8–15 we would expect that instructions to move the arms out-of-phase would lead to greater arm swing, increased transverse trunk rotation, thereby increasing stride length and decreasing stride frequency when walking at a fixed treadmill velocity. We predict that out-of-phase arm movement will coincide with more of a 1:1 frequency coordination pattern compared to in-phase arm movement. We also expect that arm swing will be greater on the noninvolved side versus the involved side.
Eight subjects (ages 14–72 years; four males and four females) who have suffered a stroke (four left side involved, four right side involved) served as a sample of convenience and agreed to participate in this pilot study. Subjects were included if they (1) walked independently without use of an assistive device, (2) had no severe perceptual deficits, (3) had no complicating medical history such as cardiac or pulmonary disorders, and (4) had sufficient motivation to participate. Informed consent and assent was approved by the Institutional Review Board of Saint Francis University and The Pennsylvania State University and obtained before subject participation.
Subjects were instructed to walk on a treadmill according to the following increments 0.22, 0.40, 0.63, 0.85, 1.10, 1.30, 1.52, 1.30, 1.10, 0.85, 0.63, 0.40, 0.22 m/s during three experimental conditions: (1) no instruction, (2) move the arms in-phase, (3) move the arms out-of-phase. The investigators aimed to examine changes in coordination relative to systematic increases and decreases in walking velocity, as research has demonstrated that coordination patterns are reflected in changing stride frequency and walking velocity.8,10,11,14–18 Therefore, systematic increases and decreases in treadmill velocity were chosen over random assignment of speed levels. Each participant had previously walked on a treadmill and before data collection subjects were given 5 to 10 minutes to acclimate to treadmill walking. Before each walking speed level, subjects were given 30 seconds to acclimate to the new speed level, and then data were collected for 30 seconds at each velocity condition. The investigators provided a demonstration of what in-phase and out-of-phase arm swing before data collection. Before data collection at each treadmill speed, the participant was reminded with verbal instructions to move the arms in- or out-of-phase. During data collection, investigators stood on each side of the subject guarding against a loss of balance during walking.
The walking patterns of the subjects were studied in the Motion Analysis Laboratory at Saint Francis University on a treadmill (Biodex, RTM 400l). The belt of the treadmill is 2 meters long and 0.5 meters wide, is driven by a 2.35-horsepower electric motor, and has a continuous speed control (range, 0.1–8.0 m/s). To minimize metallic interference during data collection, the metal side rails on the treadmill were replaced by oak and refastened to the treadmill with nylon screws. The speed of the belt is influenced minimally when subjects are walking on it (tolerance is ± 0.05 m/s).
Three-dimensional (3D) kinematic data were collected through a Skill Technologies 6D Research System (Skill Technologies Inc.). The 6D Skill Technologies System uses electromagnetic tracking to reproduce human movement in real time. The 6D Skill Technologies Motion Analysis System uses a motion capture unit (MCU) manufactured by Polhemus Inc., using eight magnetic sensors (60 Hz per sensor), which detect rotation and translation around the x-axis (sagittal plane axis), y-axis (frontal plane axis), and z-axis (transverse plane axis). Motion is captured through a 2-inch cube transmitter attached to the UltraTrak System (Polhemus Inc.) and located 12 inches from the edge of the treadmill belt. Each sensor was fastened to a limb segment with double-sided tape (location of the sensors). Sensors were placed on the following bony landmarks: proximal one third of the anterior surface of each tibia, lower one third of each femur, immediately distal to the insertion of the middle deltoid on each humerus, the sacrum, and the sternum. Each sensor is tethered (30-foot lead) to the capture board of the MCU. Each sensor wire ran toward the base of the thoracic spine, and then all eight sensors were wrapped together running directly posterior to the subject. Once sensors were properly fastened, they were aligned, using the 6D Motion Capture Software (Skill Technologies Inc.).
Before data collection, an electromagnetic map of the walking area was developed. Using algorithms from the skMapper software (Skill Technologies Inc.) a 1 × 1 × 2.45-inch 3D grid was created. The map was created by moving a mapping rod, which contained eight sensors (2.45 inches apart), from point to point in a positive y direction (anterior), and once that row was complete, the mapping rod was moved over 1 inch in a positive x direction (right/lateral), and the process was repeated in the positive y direction again. The map contained 32 points in the positive y direction and 18 points in the positive x direction. The map for this experiment reduced possible error to maximal 0.17 cm in the x, y, and z directions.
Raw data from each sensor were collected at 60 Hz and reconstructed by Skill Technologies 6D Research System software. Data were filtered at 6 Hz (low-pass Butterworth Filter) and Cardan angles with XY’Z”sequence were calculated, using Skill Technologies 6D Research System software. Angular displacement in the sagittal plane for the upper extremities was calculated by measuring the angular changes around the x-axis between the humerus sensors and the sternum sensor. Angular displacement for the lower extremities was calculated by measuring angular displacement around the x-axis between the thigh sensors and the sacrum sensor. Angular displacement in the transverse plane for both the thorax and pelvis was calculated by measuring the angular displacement around the z-axis between sternum sensor and the stationary transmitter and the sacrum sensor and the stationary transmitter, respectively (see Table 1 for description of dependent measures).
Data Analysis: Relative Power Index (RPI)
The first maximum of the shoulder angular data in stride cycle (i) was used for the 2:1 frequency coupling between arm and leg movement. Stride cycles were identified by two consecutive positive maxima from the hip angular position data for each trial. Next, we determined whether the power in arm movement frequency was higher at stride or step frequency during walking (see Wagenaar and van Emmerik17). Movement frequencies and corresponding power in shoulder and hip angle were estimated by calculating the power spectral density (PSD) function of the shoulder and hip angle time series (Welch’s averaged, modified periodogram method, 1024 samples, no overlap, Hanning window17). The PSD was normalized by dividing it by the mean power calculated over the 0.1- to 2.5-Hz frequency range for each trial separately. In the PSD of the hip angle time series, the frequency with the largest power represented the stride frequency. The step frequency was the second peak at twice the stride frequency. The power in the PSD of the shoulder angle time series at the stride and step frequencies were identified, using the stride and step frequencies from lower extremity sagittal plane motion.
Subject data were grouped according to involved side versus noninvolved side. For example, movement amplitude was calculated for involved side (ØAi) versus noninvolved side (ØAni), while mean point estimate of relative phase between the arms (MParms), and the frequency coordination (RPI) between the arms and legs was determined in relation to the involved side (RPIi) and noninvolved side (RPIni). All eight subjects were able to walk at five velocity levels, (0.22, 0.40, 0.63, 0.40, 0.22 m/s). As a result, the effects of systematically varying walking velocity on MParms ØAi, ØAni, RPIi, RPIni, and ØPT were evaluated with a within-group analysis of variance with repeated measures, including two within factors1: velocity manipulation (0.22, 0.40, 0.63, 0.40, 0.22; total five levels) and instruction (no instruction, move the arms in-phase, move the arms out-of-phase; total three levels). Differences in ØA and RPI between the involved and noninvolved arm were also examined. Comparisons were made using a within-groups analysis of variance with repeated measures1: arm side (involved and noninvolved arm; two levels), velocity (0.22, 0.40, 0.63, 0.40, 0.22; total five levels), and instruction (no instruction, move the arms in-phase, move the arms out-of-phase; total three levels). Post hoc comparisons revealed no significant differences in dependent measures between same ascending and descending velocity levels. Therefore, data from the decreasing velocity levels were combined with the increasing velocity levels (0.22. 0.40, and 0.63 m/s) and an additional within-group analysis of variance with repeated measures, including two within factors1: velocity manipulation (0.22, 0.40, 0.63; total three levels) and instruction (no instruction, move the arms in-phase, move the arms out-of-phase; total three levels) was done.8,18 In the case of statistically significant interaction effects, a Tukey analysis was applied to locate differences between conditions. A two-tailed significance level of 0.05 was chosen. Statistical analysis was done using Statistica (Statsoft Inc.).
The results revealed a significant main effect (p < 0.01) for instruction as there was greater out-of-phase motion between the arms when walking with no instructions (mean, 177.8 degrees; SD = 13.4) and walking with instructions to move the arms out-of-phase (mean, 164 degrees; SD = 2.7) compared to walking with instructions to move the arms in-phase (mean, 112.3 degrees; SD = 7.4). There were no statistically significant (p < 0.06) interaction effects between velocity and instruction. Although unlike the no instruction and out-of-phase instruction condition, as walking velocity increased during the in-phase condition, there was a gradual increase in MParms (0.22 m/s = 106.2 degrees; 0.40 m/s = 110.1 degrees; 0.63 m/s = 120.6 degrees).
Arm Swing Amplitude
There were significant main effects (p < 0.01) for both instruction and arm side. Instructions to move the arms in- or out-of-phase led to greater ØA compared to no instruction. The ØAni was also significantly greater than ØAi (Fig. 1). There was a significant interaction effect between instruction and arm side. The ØAni was significantly greater than the ØAi (p < 0.01) during each instruction condition; however, the difference was greatest when individuals were asked to move the arms out-of-phase (Fig. 1).
There was also a significant interaction (p < 0.01) effect between instruction and velocity for both the ØAni and ØAi. At 0.22 m/s, the ØAni was significantly (p < 0.01) greater with instructions to move the arms out-of-phase versus in-phase (Fig. 1). As velocity increased to 0.40 and 0.63 m/s, there was no longer a significant difference (p < 0.99) in ØAni between the two conditions. For the ØAi, there was no significant difference (p < 0.34) between in-phase and out-of-phase conditions at 0.22 m/s. However, as walking velocity increased, the ØAi was significantly greater (p < 0.01) when instructed to move the arms in-phase versus out-of-phase (Fig. 1).
Frequency Coordination: RPI
There were no significant main effects for arm side or interaction effects between arm side and phase or arm side and velocity for RPI (Fig. 2). There was a significant (p < 0.01) main effect for velocity for both RPIni and RPIi as the frequency coordination moved closer to 1:1 as treadmill velocity increased (Fig. 2). There was also a significant (p < 0.01) main effect for instruction showing that there was more 1:1 frequency coordination when instructions were given to move either in-phase or out-of-phase versus no instructions. There were significant interaction effects between instruction and velocity for both RPIn and RPIi. Post hoc analysis showed that when subjects walked at 0.63 m/s and were given instructions to move either in- or out-of-phase, there was more (p < 0.01) 1:1 frequency coordination compared to lower walking velocities during each of the three instruction conditions (Fig. 2).
There were significant main effects for both velocity and instruction for ØPT. ØPT significantly (p < 0.01) increased with walking velocity and was greatest when instructions were given to move the arms out-of-phase (mean, 12 degrees; SD = 5) versus in-phase or no instruction (mean, 8 degrees; SD = 3). There were no significant interaction effects between velocity and instruction for ØPT.
There was a significant main effect for velocity as stride frequency increased with velocity (p < 0.01). There was not a significant main effect for instruction or significant effects between velocity and instruction for stride frequency.
During gait training, physical therapists may instruct patients to move the arms back and forth while they walk. The purpose is to use the reciprocal arm movement and the coordination between upper and lower body movement to increase stride length and over-ground walking velocity. The results from this study demonstrate that during treadmill walking at slower velocities, individuals can produce an out-of-phase arm motion leading to greater trunk rotation. However, there was not a significant decrease in stride frequency (increased stride length) during the treadmill walking. The lack of significant changes in stride parameters is likely due to data being collected (1) while walking on a treadmill and (2) across a small range (0.22–0.63 m/s) of slower walking velocities. Data were collected during treadmill walking to control for walking velocity and analyze the changes in frequency coordination secondary to systematic changes in walking (treadmill) velocity. Although with a small sample and a narrow range of slower walking velocities, it is not surprising that no significant changes in stride parameters were observed. In line with previous studies,8,9,12,18 these findings demonstrate the connection between reciprocal arm movement and transverse trunk rotation. Reciprocal arm swing may be effective in facilitating trunk motion during walking. However, further study is necessary to examine whether an instructed out-of-phase arm pattern also facilitates longer strides and increased velocity during over-ground walking in individuals who walk slow (<0.8 m/s) as well as fast (>0.8 m/s).
Currently, there is no evidence supporting instructions to swing the arms in-phase during gait training. Previous findings8,12–14 would suggest that this type of pattern can lead to decreased transverse trunk rotation, stride length, and over-ground walking velocity. In-phase upper extremity movement is characteristic of slower walking velocities and coincides with more of a 2:1 versus 1:1 frequency coordination; therefore, providing these instructions would seem counter to the purpose of increasing walking velocity.15–17 In this current study, the arms did show a trend to moving more out-of-phase, despite instructions to move in-phase, as velocity increased up to 0.63 m/s. We would expect that as treadmill or over-ground walking velocity increased above 0.8 m/s, the in-phase arm pattern would be more difficult to maintain and would shift to a more out-of-phase pattern.15–17
In contrast to previous studies,15–17 both in- and out-of-phase patterns led to more of a 1:1 coordination pattern (Fig. 2), with the different upper extremity phasing patterns having no statistically significant effect on frequency coordination. These results further support the hypothesis that 1:1 frequency coordination is a strong attractor pattern for human locomotion. The instructions given did not specify the frequency or amplitude of arm movement. Both instructions led to greater arm swing with no significant difference in movement amplitude between them (Fig. 1). Changes in arm swing (Fig. 1) across different treadmill velocities was related to the instructions to move in- or out-of-phase, the involved side versus noninvolved side, and continuous attraction to more of a 1:1 frequency coordination pattern. The involved arm required less amplitude to maintain 1:1 frequency coordination when it was moving out-of-phase with the noninvolved arm versus in-phase with the noninvolved arm at 0.40 and 0.63 m/s (Fig. 1). These results are in line with those of Donker et al19 who examined the changes in the arm frequency/amplitude relationship during over-ground walking when a mass was added to one arm. They reported decreased arm movement amplitude with added mass, but arm muscle activity increased in response to the added mass, maintaining arm movement frequency and its coordination with leg movement (Fig. 3).
These data show the flexibility and adaptability of coordination patterns during treadmill walking and also further supports the previous proposal8,11,15–19 that 1:1 frequency coordination is a strong attractor pattern for human walking. The question still remains as to what type of instruction and treadmill velocity should be used for a given patient. A limitation of this current study is that there are no data characterizing the upper extremity impairments of the participants. In this small sample, the involved arm and the noninvolved arm behaved differently in response to instructions (phasing) and treadmill velocity. These data reveal that the involved arm can increase its amplitude in response to instructions (ie, in-phase instruction, increasing treadmill velocity) to achieve a 1:1 coordination pattern. In-phase arm movement during gait training may negatively affect stride parameters; however, it may facilitate increased arm swing on the involved side if individuals train at slower treadmill speeds.
During gait training, physical therapists aim to improve walking velocity and therefore general walking function. In doing so, they often provide instructions directed at the movement patterns that contribute to pathological gait. However, at present, there are few studies that detail what and how information addresses and improves upper and lower body coordination during walking in patients with stroke. The findings from this study demonstrate the adaptability of coordination patterns of persons with stroke during treadmill walking. These individuals altered arm swing patterns in response to instructions showing that using arm patterns during gait training is possible. However, future study is needed to examine how upper extremity movement dysfunction as a result of a stroke affects coordination patterns during walking. Research is needed to assess how interventions addressing upper extremity function affect walking and also how gait training at higher velocities and imposed upper extremity phasing (out-of-phase versus in-phase) affects arm function.