Effectiveness of Gait Training Using an Electromechanical Gait Trainer, With and Without Functional Electric Stimulation, in Subacute Stroke: A Randomized Controlled Trial

Participants 

All people with a first stroke who were admitted to the inpatient unit of a rehabilitation hospital in Hong Kong were screened from October 2003 to December 2004 as possible participants. Inclusion criteria for this study were (1) a diagnosis of ischemic brain injury or intracerebral hemorrhage shown by magnetic resonance imaging or computed tomography less than 6 weeks after the onset of stroke; (2) sufficient cognition to follow simple instructions and understand the content and purpose of the study (Mini-Mental State Examination score >21)21; (3) the ability to stand upright, supported or unsupported, for 1 minute; (4) significant gait deficit (FAC score <3)22; and (5) no skin allergy to electric stimulation.

Patients were excluded if they had (1) a recurrent stroke or other neurologic deficit that would affect ambulation ability; (2) any additional medical or psychologic condition that would affect their ability to comply with the study protocol—for example, a significant orthopedic or chronic pain condition, major poststroke depression, history of potentially fatal cardiac arrhythmias, implanted cardiac pacemaker, Parkinson’s disease, or clinical signs of newly developed thrombosis of the thigh; (3) aphasia with an inability to follow 2 consecutive step commands or a cognitive deficit; or (4) severe hip, knee, or ankle contracture that would preclude passive range of motion of the leg.

All participants gave informed consent through methods approved by the university and the hospital’s institutional review board. The study protocol and procedures were also approved by the university’s and the hospital’s ethics committees. The study design was a randomized controlled trial of 4 weeks. Randomization was done by computer-generated random numbers. Each participant was assigned to a group according to the list of generated random numbers after the baseline assessment.

Measurements 

Stroke patients who met all of the selection criteria during screening were allowed to participate in the study. All assessments during the study period, including the screening of patients, were performed by a single research physical therapist. Participants were assessed at the PT department of the hospital. Neither participants nor the research physical therapist were blinded to the treatment because it was impractical to do so. The 3 groups were compared in terms of general mobility, gait ability, overground walking speed, and motor impairment.

The EMS has proven to be a valid scale with good interrater reliability (Spearman correlation coefficient ρ=.88) that can be readily applied during daily clinical work23 and hence was used in this study to detect improvement in mobility. This is a 20-point scale assessing 7 elements of mobility, incorporating gait speed and functional reach.

Balance was assessed by the BBS. This has excellent interrater and intrarater reliability for elderly subjects24 and for subjects with acute stroke (interrater intraclass correlation coefficient model 2,1 [ICC2,1]=.98; intrarater ICC2,1=.97).25 It can detect change in balance of people with acute stroke.26 Ambulatory ability was rated by using the FAC scale.22, 27 Participants were rated according to the personnel support needed for gait, regardless of use of an assistive device, according to a 6-point scale. The interrater reliability of the FAC scale has been assessed as .85.27

The leg subscale of the Motricity Index evaluates 3 joint movements (hip flexion, knee extension, ankle dorsiflexion) and was used to analyze motor loss of the affected lower limb after stroke.28 It is sensitive to motor recovery after stroke. This measure has good reliability, with a Cronbach α of .77.29 The Pearson correlations between the Motricity Index and the isometric strength from a dynamometer are good to excellent (r range, .78–.91).29

Overground walking speed was measured by timing a 5-m walk with a stopwatch. The speed was calculated in meters per second. The test-retest variability in the 5-m walk for subjects after stroke (2–6y) was measured by calculating the maximum difference between any 2 readings as a percentage of the lowest time. The results showed that it was reliable on test-retest, with 95% of subjects’ speed within 25% of the lowest time.22 The Pearson correlations between the 5-m and the 10-m walk and between the walks done in the first and second trials were good (r range, .95–.99).22 Each participant was asked to walk as fast as possible on a walkway for 5m using a walking aid if necessary. No other assistance was given, and no orthoses were allowed for the walk. If a walking aid was used for the first timed walk, it was also used for each subsequent timed walk. The walking speed was regarded as 0.0m/s if a participant required manual assistance from a therapist to walk or was unable to finish the whole 5m. Walking speed was measured from 2 trials, with the average of the 2 speeds recorded as the final gait speed of each participant. The 5-m walking test was designed for acute stroke survivors by da Cunha-Filho et al.9 The short distance enables people with relatively poor aerobic fitness, balance, and lower-limb strength to complete the test more readily.

The FIM instrument30 and Barthel Index31 were used to provide a comprehensive view of overall function. The FIM measures not only self-care activities and mobility but also communication and cognitive functions. It is well validated, the interrater reliability is high, and the Cronbach α was .88 to .91 in people with stroke.32, 33 The Barthel Index is a validated and widely used instrument to measure dependency in activities of daily living, and the Cronbach α for reliability has been shown to be .84 to .85 in people with stroke.32, 33 The FIM and Barthel Index were performed before and after the 4 weeks of intervention by nurses who were blinded in this study.

Interventions 

All participants received their regular weekday 40-minute PT sessions and 1.5-hour multidisciplinary treatments, which consisted of occupational therapy, speech therapy, and psychology throughout the study period. All participants participated in the 4 weeks of training during the study period in the same hospital. The study consisted of 1 training session per weekday during the 4 weeks. To prevent contamination FES was not allowed to be used during the regular PT sessions. Each participant of the experimental groups (EGT, EGT-FES) underwent gait training with BWS by the electromechanical gait trainer, and each participant of the EGT-FES group also received FES to the paretic lower limb during the gait training. Participants in the CGT group underwent overground walking for their gait training. Figure 1 shows the training program details for each group.

View Large Image Download to PowerPoint

Fig 1. CONSORT flowchart of the training programs.

The CGT group received conventional PT gait training based on the principles of proprioceptive neuromuscular facilitation and Bobath concepts.34, 35 It was conducted by each subject’s own hospital physical therapist, who administered his/her regular rehabilitation. The aim of the Bobath treatment is to improve a person’s posture and movement.34 Through specialized handling techniques, muscle stiffness may be reduced, muscle control against gravity may be increased, and fluctuating muscle activity may be stabilized. Participants could undergo the overground walking gait training with or without a walking aid or orthoses and with or without manual assistance depending on their abilities. The conventional rehabilitation provided by the hospital was tailored to each patient’s needs, focusing on his/her particular impairments and disabilities, and therefore each patient received varying degrees of specific therapies. Each session of conventional treatment was documented, with each type of activity and its duration recorded.

The EGT group used the electromechanical gait trainer. Each gait-training session was to last 20 minutes with an optional rest break (of 1–3min) after the first 10 minutes. The electromechanical gait trainer simulated a normal gait cycle in a symmetric manner with a ratio of 60% to 40% between the stance and swing phases. The gait trainer also supported part or all of each person’s body weight via a harness attached by ropes to a gear system, according to the participant’s ability to lift his/her foot during the swing phase.13 If partial or full body weight needed to be supported, the participant’s physical therapist manually tugged on the system’s pulley. Each harnessed participant was positioned over the gait trainer’s footplates, to which his/her feet were secured. The propulsion of the footplates helped with the movement of the feet and legs during both the stance and swing phases. Step length and walking speed could be adjusted from 34 to 48cm and from 0 to .70m/s, respectively. Other training variables included the percentage of body-weight support provided and each participant’s use of the gait trainer’s front horizontal bar for hand support to increase stability. The target training velocity was relatively slow (.20–.60m/s) to avoid overexertion of participants.15 Body weight was partially supported to compensate for the paresis of the affected lower limb, and this support was reduced as soon as a participant could take his/her full body weight. The physical therapist would give assistance during gait training to help with each participant’s knee extension and would give verbal cueing to facilitate each participant’s head and trunk extension and midline awareness when the participant leaned to the paretic side or increased trunk kyphosis. Participants who achieved adequate balance on the gait trainer were encouraged not to use the horizontal bar for hand support to further train their balance and postural control of the lower limbs.

Participants in the EGT-FES group underwent the same interventions as the EGT group, except for the additional feature of FES during their gait-trainer sessions. Each participant received electric stimulation modalities, including waveform and pulse width with fixed values. The stimulation intensity was adjusted by the supervising physical therapist according to how successful the correct limb movement was elicited and to each participant’s comfort threshold (table 1). A pair of self-adhesive electrodes (Platinum Blue 901220; size, 5×5cm square electrode)b was attached to each participant’s quadriceps on the paretic side and stimulated in the stance phase to facilitate weight acceptance. Another pair of electrodes (Ultraflex 881150; size, 3.8mm round electrode)b was attached to each participant’s common peroneal nerve on the paretic side and stimulated during the swing phase to generate ankle dorsiflexion. The stimulation sites were determined beforehand, with each participant in a seated position, and when a correct functional response was obtained—that is, the knee extended when the quadriceps were stimulated and the ankle dorsiflexed when the peroneal nerve was stimulated. Stimulation intensity was then raised until the functional movement over the desired range of motion (knee angle <20° from full extension, ankle angle in neutral position or dorsiflexed position) was achieved with comfort for each participant; then the sites were marked on the skin with nonconductive, semipermanent ink. Electrodes were then attached to the same marked sites throughout the 4-week intervention period. Before each participant’s first training session, intermittent stimulation was tested continuously on the participant for at least 10 minutes to rule out skin allergy. Two connection wires linked the gait-trainer control box and the 2 single-channel FES stimulatorsc that were developed to synchronize the gait phase and the stimulation timing for the quadriceps and the common peroneal nerve.

Table 1. FES Parameters Set During EGT-FES Training Sessions

	Parameters	Quadriceps Muscle	Common Peroneal Nerve
	Stimulation phase	Stance phase	Swing phase
	Pulse width (μs)	400	400
	Rising edge ramp (s)	0.3	0.3
	Falling edge ramp (s)	0.3	0.3
	Waveform	Rectangular pulse	Rectangular pulse
	Extension (s)	0.1	0.1
	Current (mA)	50–85	50–70

Blood pressure was monitored continuously during each gait-training session to safeguard participants’ health. Set criteria for cessation of intervention included if a participant reported having a headache, confusion, or onset of angina and if an excessive change in blood pressure was detected. A rest break was given if a participant requested a rest because of fatigue. A daily log sheet was kept to record gait speed generated by the gait trainer, percentage of body-weight support from the harness, total walking distance, and number of rest breaks during training. If a participant missed more than 3 scheduled sessions because of medical reasons or an inability to participate, he/she was withdrawn from the study.

These interventions have not been investigated in the subacute stroke population; therefore, parameters for sample size calculation were not available. We chose a sample size of 50, because this number of participants could be reasonably recruited with the available resources.

Statistical Analysis 

Difference scores of all outcome measures were calculated from posttreatment scores minus pretreatment scores to evaluate the effect of gait training. Intention-to-treat analysis was used for all of the recruited participants. Data that were missing because participants had withdrawn from the study were replaced by the last scores obtained. SPSSd was used to compare the differences in outcome measurements. For nonparametric variables, the Friedman test for repeated measurements was used to compare the 3 assessments within a group, and the Wilcoxon signed-rank test was used to perform pairwise multiple comparisons. The Kruskal-Wallis test was used for comparison between the 3 groups (CGT, EGT, EGT-FES). If a significant difference was found between any of the 3 groups, the Mann-Whitney U test was used to perform multiple comparisons between the 2 groups. For continuous variables, 1-way analysis of variance was used to determine differences in the clinical outcome measures across the 3 groups. An α level of P less than .05 was assumed to be significant, and the Bonferroni adjustment was then used for post hoc comparison. No correction for multiple statistical testing was performed.