XXX

RESEARCH ARTICLES

Use of a Robotic Walking Device for Home and

Downloaded from http://journals.lww.com/jnpt by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW

Community Mobility in Parkinson Disease: A Randomized

Controlled Trial
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdtwnfKZBYtws= on 10/03/2024

Deb A. Kegelmeyer, DPT, MS, GCS, Raquel Minarsch, PT, DPT, NCS, Sandra K. Kostyk, MD, PhD,
David Kline, PhD, Rachel Smith, MPH, and Anne D. Kloos, PT, PhD, NCS

Background/Purpose: Gait impairments in Parkinson disease (PD) Discussion and Conclusions: All RAGT participants could use the
contribute to decreased quality of life. This randomized controlled trial EXOD safely. The RAGT treatment used in this mostly low impair-
examined immediate- and longer-term effects of a single joint robotic ment population of people with PD may be ineffective and/or was
exoskeleton device (EXOD), the Honda Walking Assist device, on gait. insufficiently dosed to see a positive treatment effect. Our findings
Methods: Participants (n = 45) with PD (Hoehn and Yahr stages suggest that RAGT interventions in PD may be more effective in
1-3) were randomized to a robotic-assisted gait training (RAGT) individuals with greater motor impairments.
group (n = 23) or control (CON) group (n = 22). The RAGT group Video Abstract available for more insights from the authors (see
was tested with and without the EXOD at baseline and then received the Video, Supplemental Digital Content 1 available at: http://links.
supervised in-home and community training with the EXOD twice lww.com/JNPT/A459).
weekly for 8 weeks. The CON group received no interventions. Key words: gait, locomotion, Parkinson disease, robotics
Outcome measures included gait speed (primary), gait endurance
(6-minute walk test), perceived ease of walking, and questionnaires (JNPT 2024;48: 102–111)
and logs assessing performance of daily activities, freezing of gait,
and daily activity levels. INTRODUCTION
Results: Forty participants completed the study. No significant
immediate impact of EXOD usage on participants’ gait measures
was found. Differences in gait speed and secondary outcome mea-
I ndividuals with Parkinson disease (PD) develop gait
impairments and balance problems over time that interfere
with their participation in meaningful activities in their homes
sures postintervention were not significantly different between the and communities.1 Interventions that promote motor learning
RAGT and CON groups. Participants with greater disease severity to optimize functional mobility in individuals with neurologic
(worse baseline motor scores) had greater improvements in stride impairments often provide high repetition, task-oriented
length during unassisted walking after the intervention than those movement training.2,3 As examples, repetitive practice of
with lower severity (mean difference: 3.22, 95% confidence interval: stepping through treadmill walking or external cueing tech-
0.05-6.40; P = 0.04). niques improves walking performance in PD populations.4
However, many individuals cannot regularly exercise due to
Division of Physical Therapy (D.A.K., R.M., R.S., A.D.K.) and Departments barriers including lack of access to exercise facilities, physi-
of Neurology and Neurosciences (S.K.K.), College of Medicine, The Ohio
State University, Columbus; Center for Biostatistics (D.K.), The Ohio
cal impairments, and low motivation (apathy).5,6
State University, Columbus; and Department of Biostatistics and Data Robotic devices deliver high repetition task-specific
Science (D.K.), Wake Forest University School of Medicine, Winston- training to promote motor learning, while reducing the need
Salem, North Carolina. for manual assistance or cueing by therapists.7 Clinical trials
This study was funded by a grant from The Michael J. Fox Foundation for of robotic-assisted treadmill training in PD demonstrated
Parkinson’s Research (#16197). Dr Kostyk’s institution has received sal-
ary and research trial funding from Voyager Therapeutics, Neurocrine significant improvements in gait speed, functional mobility,
Biosciences, and Prilenia Therapeutics. Honda R&D Americas, Inc, pro- walking endurance, freezing of gait (FOG), and quality of
vided the Honda Walking Assist exoskeleton devices for this study. The life.8-16 While people with PD (PwPD) accepted and bene-
company did not contribute direct funding for this project, nor did their fited from the training, the robots used in these studies are
staff have any input on study design or in preparation of this manuscript.
ClinicalTrial.gov registration number: NCT03751371.
large, heavy, and stationary (ie, Lokomat, G-ECOSYSTEM)
All remaining authors had no conflicts of interest to report. that are cost prohibitive for most people and require extensive
Supplemental digital content is available for this article. Direct URL citations training before they can be used independently.
appear in the printed text and are provided in the HTML and PDF versions The Honda Walking Assist (HWA) device (Honda
of this article on the journal’s Web site (www.jnpt.org). R&D Co Ltd, Wako, Japan) is a lightweight and portable
Correspondence: Anne D. Kloos, PT, PhD, NCS, Division of Physical
Therapy, College of Medicine, The Ohio State University, 453 W 10th single-joint robotic exoskeleton device (EXOD).17,18 Worn
Ave, Columbus, OH 43210 ([email protected]). around the user’s waist and thighs, internal motors assist hip
Copyright © 2024 Academy of Neurologic Physical Therapy, APTA. flexion during swing initiation and hip extension during
ISSN: 1557-0576/24/482-102 stance walking phases (Figure 1). The device calibrates leg
DOI: 10.1097/NPT.0000000000000467 movements to spatial and temporal cadence targets to

102 JNPT ● Volume 48, April 2024

Copyright © 2024 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
 JNPT ● Volume 48, April 2024 Use of a Robotic Walking Device for Home and Community Mobility in Parkinson Disease
Downloaded from http://journals.lww.com/jnpt by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdtwnfKZBYtws= on 10/03/2024

Figure 1. Honda walking assist device.

promote symmetrical walking patterns.19 Advantages of of walking would improve when (1) wearing the EXOD
lightweight, wearable robotics over stationary robotic sys- compared with unassisted walking over 1 session and (2)
tems include easy donning and doffing, greater ease of use, when walking unassisted after the RAGT intervention com-
and free movement in the home and community, thereby pared with before the intervention, and that (3) participants in
yielding greater walking practice.20 the RAGT group would be able to walk safely with the
Laboratory-based studies in healthy adult and neurolo- EXOD.
gical populations to examine the effects of gait training using
EXODs with hip assistance, including the HWA device, have
demonstrated improved gait spatiotemporal parameters and METHODS
endurance.20-25 Healthy elderly and middle-aged adults The study was approved by the Institutional Review
showed temporary improvements in stride length, gait Board of The Ohio State University and was registered at
speed, and energy costs wearing the HWA compared with ClinicalTrials.gov (NCT03751371). A blinded, randomized
unassisted walking.21,22 Long-term HWA usage in indivi- controlled trial design was utilized to assess preliminary
duals with stroke improved gait parameters during unassisted efficacy of the EXOD in individuals with idiopathic PD
walking.20,23 Individuals with moderately advanced PD (n = (Figure 2). Randomization allocation was 1:1 and stratified
12) showed immediate improvements in gait speed, step by Hoehn and Yahr (H&Y) stages 1-3.31 Each stratum rando-
length, and hip flexion and extension excursion while walk- mization scheme was generated in permutated blocks of vary-
ing with the HWA device and improved walking endurance ing size to ensure a balance in the randomization among
during unassisted walking after 10 half-hour sessions of gait stage. Individuals randomized into the RAGT group per-
training with the HWA device in the home setting.25 The formed outcome measures with and without the EXOD
mechanical assistance given by the HWA may provide (1) (hypothesis 1) and then received training with the EXOD in
temporal (rhythmic tactile stimulus on leg) and spatial (atten- their homes or surrounding communities twice weekly for 8
tional stimuli to increase step amplitude) external cues and (2) weeks. The control (CON) group completed baseline out-
force assistance to increase hip mobility and ameliorate lower come measures without the EXOD only and then continued
extremity muscle strength and power deficits26 that improve with their usual care, daily activities, and ongoing exercise
gait in people with mild to moderate PD.27,28 Thus, EXOD regimens. At 8 weeks, both groups underwent postassessment
interventions may help overcome exercise barriers due to lack (hypothesis 2). The randomization scheme was overseen by
of access to exercise facilities by allowing in home gait biostatisticians using the randomizeR package version 1.4.2
training, physical impairments by providing lower extremity in R 3.5.1 (Vienna, Austria).32 Trial assessors were blinded to
step cueing and mechanical assistance, and low motivation by group allocation and administered assessments that were col-
making walking more comfortable and less effortful, thereby lected without the EXOD. Assessors who obtained outcomes
improving home and community mobility for daily life activ- completed with the EXOD were not blinded.
ities and the quality of life in PwPD.29,30
The primary purposes of this study were to determine Participants
the immediate effect of EXOD usage on gait speed and the Adults with idiopathic PD were recruited from central
effect of an 8-week robotic-assisted gait training (RAGT) Ohio. Inclusion criteria were age 50 to 90 years, ambulation
intervention on gait speed while walking unassisted in the without assistance (H&Y stages 1-3), and on stable PD
home and the community, and secondarily to determine the medication doses for 4 weeks prior to the study. Exclusion
safety of EXOD usage in individuals with PD. We hypothe- criteria were presence of other significant cardiac, neurolo-
sized that gait endurance, gait parameters, and perceived ease gic, or orthopedic problems that affect gait; weight

© 2024 Academy of Neurologic Physical Therapy, APTA 103

Copyright © 2024 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
 Kegelmeyer et al JNPT ● Volume 48, April 2024
Downloaded from http://journals.lww.com/jnpt by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdtwnfKZBYtws= on 10/03/2024

Figure 2. Study flow sheet. CON, control; EXOD, exoskeleton device; RAGT, robotic-assisted gait training.

exceeding 220 lb and height exceeding 6′8″; electronic Exoskeleton Device

medical devices embedded in the body; participation in The EXOD, provided by Honda R&D Americas, Inc,
physical therapy; and inability to understand instructions weighs 5.95 lb and has 2 motors that run on a single recharge-
as assessed by inappropriate or lack of responses to simple able battery. The device has Food and Drug Administration
commands. clearance for people with stroke with gait deficiencies and has

104 © 2024 Academy of Neurologic Physical Therapy, APTA

Copyright © 2024 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
 JNPT ● Volume 48, April 2024 Use of a Robotic Walking Device for Home and Community Mobility in Parkinson Disease

established safety and feasibility in healthy older adults and through 7 to compare general activity levels between groups.
people with stroke.20-23 Hip joint angle and torque output are To monitor safety during the study period participants were
monitored using angle and current sensors. Assist torque is asked to record any falls (ie, events that caused the person to
transmitted to the user’s thighs via thigh frames. When the rest inadvertently on the ground or floor) or other adverse
user initiates walking, the EXOD automatically adjusts leg events that occurred. All outcome measures were repeated
Downloaded from http://journals.lww.com/jnpt by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW

movements to reach target walk ratios (step length/cadence) after the 8-week intervention period for both groups.
by increasing hip flexion and/or extension using power sup-
plied by the device. Hypothesis 3: Safety
A trained physical therapist (PT) properly sized and Safety during walking was assessed using counts of
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdtwnfKZBYtws= on 10/03/2024

placed the EXOD around the participants’ waist and bilateral observed episodes of loss of balance (LOB), defined as
thighs and adjusted device settings. Flexion or extension a change in posture requiring the examiner walking with the
assistance levels were adjusted to maximize the participants’ participant to use the gait belt and/or manual assistance to
gait endurance and to promote a symmetric gait pattern based prevent a fall. Safe walking was defined as no increase in
on tablet feedback provided by the EXOD. The PT applied incidences of LOB during walking with the EXOD as com-
a gait belt to ensure participant safety during walking. pared with unassisted walking. Longer-term safety of EXOD
use was defined as no increase in incidences of falls through-
Procedures out the intervention timecourse. Safety was also assessed as
The PT investigators obtained demographic informa- 80% of participants and greater in the RAGT group attaining
tion and assessed outcome measures, and a neurologist admi- a safe gait.
nistered the Unified Parkinson Disease Rating Scale
(UPDRS) Part III—Motor Scale1 to determine H&Y stage Intervention
in the investigators’ laboratory. Participants were instructed RAGT group participants received PT supervised,
to take PD medications about 1 hour before assessments. home and community-based (ie, outside home and/or within
surrounding community) walking training wearing the EXOD
Hypothesis 1: Immediate Effect of EXOD twice a week for 45 to 60 minutes for 8 weeks. Participants
Gait endurance was measured with the 6-minute walk were cued to take larger and more symmetrical steps by the
test (6MWT), a reliable and valid measure in the PD device’s assistance. The default device settings were used
population.33,34 Participants were instructed to “walk as far initially (2 Nm of flexion and extension assist bilaterally)
as you can in 6 minutes.” Gait parameter measurements were and were adjusted as necessary (0 for no assist to 4 Nm) to
speed, stride length, swing time, double support time, and promote a symmetrical, safe stepping pattern with primary
coefficients of variation. These were collected during the first goals to increase step length and foot clearance. Hip flexion
straight pass of the 6MWT for 50 ft using portable wearable assist was gradually increased by 0.5 Nm on the side that
sensors on each lower extremity (LEGSys, Biosensics, demonstrated a shorter step length until symmetry was
Newton, Massachusetts).35,36 Participants’ perceived ease of achieved or maximum assist was reached. If maximum hip
walking defined as how easy it felt to walk during the 6MWT flexion assist was reached and a continued step length asym-
was indicated on a visual analog scale (100-mm line with metry was observed, hip extension assist on the contralateral
anchors “Not at all easy” and “Extremely easy”). Gait assess- limb was increased. Participants performed primarily forward
ments were conducted in the RAGT group first during unas- walking during each intervention session (30-45 minutes)
sisted walking, followed by assisted (ie, wearing the EXOD) with an average of approximately 15 minutes of each session
walking, with 15-minute rests between tests. CON group focused on multidirectional stepping, turning, and other chal-
participants never used the device. lenging balance and motor control tasks during gait, with rests
as needed. Because of the heterogeneity of the home/commu-
Hypothesis 2: Effect of RAGT Intervention nity environment and participant severity of symptoms, stan-
Gait assessments (6MWT, gait parameters) were dardization of sessions was limited; however, consistency
obtained first without wearing the EXOD, followed by wear- was attempted when possible. Environment (home vs com-
ing the EXOD in the RAGT group, whereas they were munity), home layout and size, weather, and the participant’s
obtained only without the EXOD in controls. All participants endurance influenced the training activities. Typical sessions
completed the Stanford Self-Efficacy for Measuring Chronic in the community included outdoor forward walking over
Disease 6-Item Scale37 questionnaire to measure participants’ sidewalks and uneven ground including inclines, curbs, or
confidence in performing daily activities, and the Freezing of grass. Emphasis during outdoor training was on endurance,
Gait Questionnaire.38 Both the Stanford Self-Efficacy for speed, step length symmetry, and cadence. Examples of
Measuring Chronic Disease 6-Item Scale and the Freezing indoor community sessions included walking around a track
of Gait Questionnaire have good reliability and validity in at a fitness center, an open gym at a fitness center or church,
individuals with PD.38,39 To determine changes in activity or hallways, ramps, and stairs at a participant’s place of work.
levels, all participants wore activity monitors (PAMSys, Indoor activities inside participants’ homes were focused on
Biosensics, Newton, Massachusetts) to quantitatively mea- multidirectional walking (backward, side stepping), walking
sure physical activity and recorded daily activities in an over various surfaces (hardwood, carpet, etc), obstacle nego-
activity log for 5 consecutive days during weeks 1 and 8. tiation, turns, and stair negotiation. The EXOD settings were
Participants continued to fill out activity logs in weeks 2 modified to facilitate the goal of the activity, such as

© 2024 Academy of Neurologic Physical Therapy, APTA 105

Copyright © 2024 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
 Kegelmeyer et al JNPT ● Volume 48, April 2024

increasing hip extension cues during backward walking. To RESULTS

address freezing, the timing of flexion and extension assist in
the EXOD was adjusted during activities that typically caused Participants
freezing episodes for the participant. Participants wore a gait Between April 2019 and July 2021, 107 individuals were
belt with primarily standby assistance provided by the thera- assessed for eligibility (Figure 3). Reasons for exclusion
Downloaded from http://journals.lww.com/jnpt by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW

pist with occasional contact guard assist during more challen- included not meeting criteria (n = 36), living outside of
ging activities or fatigue, or minimal assistance in the event of 1-hour driving distance from research laboratory (n = 11),
LOB. The PT monitored participants’ heart rates and Ratings and losing interest or not following up (n = 14). Forty-six
of Perceived Exertion (Borg RPE) to maintain a moderate participants were enrolled; 1 individual was withdrawn prior
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdtwnfKZBYtws= on 10/03/2024

intensity (RPE 13-16) for better intervention compliance and to randomization due to meeting exclusion criteria that was not
comparison across groups. Participants were progressed by disclosed during screening. Forty-five participants were rando-
increasing distance or time spent walking while maintaining mized to RAGT (n = 23) or CON (n = 22) groups. Five
EXOD settings as long as step symmetry and foot clearance participants (3 RAGT, 2 CON) did not complete the study
was achieved. Participants were instructed to continue usual due to a laboratory shutdown caused by COVID-19 pandemic
activities and to inform investigators of changes (eg, medica- (n = 3), an unrelated hospitalization (n = 1), and dropping out to
tions, starting new exercise programs, or physical therapy) receive physical therapy (n = 1).
made during the intervention period. The participants never Baseline participant characteristics are summarized in
wore or used the EXOD without therapist supervision. Table 1. Group characteristics were comparable, except that
more CON group participants were male. There were 22 fallers
Analysis and 20 nonfallers, (RAGT fallers = 11, nonfallers = 9; CON
A priori power analyses predicted that 23 participants fallers = 11, nonfallers = 11).
randomized to the RAGT group would have at least 80%
power to detect a difference of 0.61 standard deviations in Table 1. Participant Characteristics
gait speed using a 2-sided paired t test with type I error rate of RAGT, Median CON, Median
0.05 for hypothesis 1, and 18 participants per group would Characteristic Levels (Q1, Q3) (Q1, Q3)
detect a difference of 0.96 standard deviations in gait speed Age, y 72 (65.21, 78.56) 72 (65.51, 77.03)
after week 8 (hypothesis 2) using a 2-sided t test with type Assistive device use No 15 (65%) 14 (64%)
I error rate of 0.05 (PASS 15, Kaysville, Utah). Yes 8 (35%) 8 (36%)
For hypothesis 1, the primary analysis compared the Number of falls in past 1 (0.00, 3.00) 1 (0.00, 2.00)
6 mo
difference in gait speed with/without the EXOD in the Hoehn & Yahr stage Stage 1 1 (4%) 0 (0%)
RAGT group only using a 2-sided paired t test at the 0.05 Stage 2 18 (78%) 18 (82%)
significance level. Secondary outcomes gait endurance, gait Stage 3 4 (17%) 4 (18%)
parameters (stride length, swing time, double support time, Sex Female 10 (43%) 2 (9%)
Male 13 (57%) 20 (91%)
and coefficients of variation), and perceived ease of walking Years since diagnosis 5 (3.00, 7.00) 5 (3.50, 10.00)
for hypothesis 1 were analyzed similarly using appropriate Years since symptom 8 (6.00, 11.00) 8 (6.00, 16.00)
methods for paired data. For hypothesis 2, the primary onset
analysis assessed the effect of treatment on gait speed at 8 Abbreviations: CON, control group; Q1, quartile 1; Q3, quartile 3; RAGT, robotic-
weeks using a linear regression model that included an assisted gait training group.
indicator of treatment, stage, and baseline gait speed. This xxxx
analysis followed intention to treat principles to preserve the
balancing properties of randomization.40,41 For analysis, Hypothesis 1: Immediate Effect of EXOD Analysis
H&Y stages 1 and 2 were combined because there was At baseline, there were no significant differences in gait
only 1 participant enrolled in stage 1. The test of treatment speed, or any secondary gait measures, when participants
effect was 2-sided and at the 0.05 significance level. Post assigned to the RAGT group (n = 23) walked with EXOD
hoc subanalyses were performed to determine whether par- assist compared with without the device (Table 2). Gait-
ticipant gait parameter responses to the intervention differed related measures did not change significantly on first trial
according to the participants’ baseline fall history status or wearing the EXOD, though some measures worsened slightly
to their baseline disease severity. The effect was estimated in some participants.
per a 5 unit increase in their baseline UPDRS motor scores,
which is the minimal clinically important difference in the Hypothesis 2: Effect of RAGT Intervention Analysis
UPDRS motor scale.42 Participants who reported 1 fall and No significant differences were noted in gait speed or
greater during the previous 6 months at baseline were desig- the secondary outcome measures between RAGT and CON
nated as “fallers.” Secondary outcomes for hypothesis 2 groups after the intervention period (Table 3, see Table,
were assessed similarly to the primary outcome using appro- Supplemental Digital Content 2, available at: http://links.
priate generalized linear models. This resulted in 11 separate lww.com/JNPT/A460, Outcome Measure Unadjusted
models—1 for each secondary outcome. Data (incidences of Means), except for instances of LOB during the 6MWT for
LOB and falls) were analyzed using descriptive statistics for the RAGT group compared with the CON group. In general,
hypothesis 3. All analyses were done using SAS 9.4 (Cary, changes in outcomes were in the positive direction, including
North Carolina). increases in speed, stride length, and the 6MWT distance.

106 © 2024 Academy of Neurologic Physical Therapy, APTA

Copyright © 2024 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
 JNPT ● Volume 48, April 2024 Use of a Robotic Walking Device for Home and Community Mobility in Parkinson Disease
Downloaded from http://journals.lww.com/jnpt by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdtwnfKZBYtws= on 10/03/2024

Figure 3. CONSORT diagram. PT, physical therapist; RAGT, robotic-assisted gait training.

Baseline fall history status was not related to the parti- Hypothesis 3: Safety Analysis
cipants’ responses to the intervention for any outcomes. All RAGT group participants were able to use the
Participants with higher baseline UPDRS motor scores had EXOD safely. There were no adverse events reported
significantly greater improvement in stride length during related to the study or EXOD usage. There were no differ-
unassisted walking following the intervention than those ences in incidences of LOB between groups and while
with lower scores (mean difference: 3.22, 95% confidence wearing the EXOD during gait testing (see Supplemental
interval: 0.05-6.40; P = 0.04, Figure 4). Digital Content 2, available at: http://links.lww.com/JNPT/
Because of a combination of activity monitor technical A460).
problems and participants’ lack of compliance wearing activ-
ity monitors during weeks 1 and 8 of the study, there were
insufficient data for valid statistical analysis of electronically DISCUSSION
monitored activity data. Data gathered from activity diaries This randomized controlled trial examined the short-
indicated that the majority of participants in both groups and long-term impact of using a single-joint EXOD on
engaged primarily in light activity (eg, taking a walk or slow improving unassisted and assisted mobility in individuals
bicycle ride) across the day and very few engaged in moder- with mild to moderate PD. In contrast to robotic treadmill
ate- (eg, doing laundry, yard work) or heavy- (eg, running, systems, the lightweight and portable EXOD used in this
chopping wood) level activity. Reported activity levels did not study offered the unique opportunity to study the effects
change in either group across the 8 weeks. of RAGT delivered in the home and community

© 2024 Academy of Neurologic Physical Therapy, APTA 107

Copyright © 2024 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
 Kegelmeyer et al JNPT ● Volume 48, April 2024

Table 2. Baseline Gait Measures With and Without Table 3. Mean Adjusted Outcome Measures at the End of
Exoskeleton Devicea Intervention for the RAGT Compared With the CON Group
Comparison (With to Without Mean Adjusted
Device) Difference 95% CI P Adjusted Mean 95%
Outcome Mean CON RAGT Difference CI P
Downloaded from http://journals.lww.com/jnpt by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW

Gait speed, m/s −0.08 −0.16 to 0.0726

0.01 Gait speed, m/s 1.27 1.29 0.02 −0.10 to 0.7869
Stride length, cm −3.55 −8.90 to 0.1834 0.14
1.81 Stride 132.65 133.35 0.70 −9.03 to 0.8849
Stride length CV 0.00 −0.02 to 0.7064 length, cm 10.43
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdtwnfKZBYtws= on 10/03/2024

0.02 Stride length 0.07 0.07 −0.01 −0.05 to 0.7247

Swing time, s −0.72 −1.87 to 0.0856 CV 0.03
0.43 Swing time, s 41.77 41.42 −0.35 −1.34 to 0.4784
Swing time CV 0.00 −0.01 to 0.7026 0.64
0.02 Swing time 0.06 0.05 −0.01 −0.03 to 0.5124
Double support time, s 1.39 −0.89 to 0.2199 CV 0.02
3.67 Double 16.57 17.18 0.60 −1.34 to 0.5328
Double support time CV −0.01 −0.05 to 0.6821 support 2.55
0.03 time, s
6MWT, m −19.32 −41.4 to 0.0837 Double 0.23 0.19 −0.05 −0.14 to 0.3150
2.80 support 0.05
6MWT ease −0.04 −0.87 to 0.9144 time CV
0.79 6MWT, m 419.61 428.79 9.18 −26.1 to 0.6007
6MWT loss of balance, log mean −0.26 −0.70 to 0.2284 44.45
0.18 6MWT ease 7.45 6.30 −1.15 −2.49 to 0.0894
0.19
Abbreviations: CI, confidence interval; CV, coefficient of variation; 6MWT, 6MWT LOB −2.00 −3.83 −1.83 −2.94 to 0.0019
6-Minute Walk Test.
a (log rate) −0.72
n = 23.
FOGQ 9.24 8.89 −0.36 −2.21 to 0.7001
xxxx
1.50
SSE 7.98 7.91 −0.06 −0.85 to 0.8705
environment. While we demonstrated the safety of EXOD 0.72
usage during home physical therapy, there was no signifi- Abbreviations: CI, confidence interval; CON, control group; CV, coefficient of
cant immediate effect of EXOD usage or an 8-week variation; FOGQ, Freezing of Gait Questionnaire; LOB, loss of balance; RAGT,
RAGT intervention on the primary outcome gait speed. robotic-assisted gait training group; 6MWT, 6-Minute Walk Test; SSE, Stanford Self-
Efficacy for Measuring Chronic Disease 6-Item Scale.
These findings suggest that the RAGT intervention may xxxx
not be an effective intervention for gait training in the
early-stage PD population. However, participants with participants who had particularly high numbers of LOB.
greater disease severity demonstrated greater improve- The lack of significant findings may be due to the high
ments in stride length after RAGT, suggesting that the variability in the participants’ performances on outcome
intervention may be better suited for individuals with measures, as indicated by wide confidence intervals.
greater mobility impairments. Although most participants were classified as being in
EXOD use did not immediately change the RAGT H&Y stage 2, we observed a wide range of performances
group participants’ gait measures or perceived ease of walk- among this group, possibly because we used the original
ing compared with unassisted walking. The novelty of the H&Y stages, which lacks the 2.5 rating included in
device may have resulted in cautious mobility. While partici- a modified version.43 Future studies with larger sample
pants briefly practiced walking with the EXOD prior to test- sizes and randomization using the modified H&Y stages
ing, longer practice sessions wearing EXODs, supervised by could address this problem. Alternatively, the lack of effect
PTs, may be needed to improve self-efficacy and walking may be that the intervention was not long or frequent
performances prior to testing and independent use. It is pos- enough.44-46 Although the 8-week intervention reflects
sible that the greater differences in immediate gait measure a typical duration for home health physical therapy, some
findings compared with ours reported by Kawashima et al25 studies reported improvements with exoskeletons using 20
may be because their gait measures were an average of the to 25 sessions.44-46 Another possibility is that the RAGT
initial 6 trials and participants were tested after wearing the intervention is not effective in the PD population.
device with assistance turned off before testing them with the Disease severity, but not fall history status, was found
assistance turned on, resulting in increased exposure and to influence participants’ responses to the intervention.
practice using the EXOD. All RAGT participants could Participants with greater disease severity had greater increase
walk safely wearing the device, which supports the safety of in stride length during unassisted walking postintervention
device usage in early PD disease stages (H&Y 1-3). than those with less disease severity. Some participants
Following the 8-week RAGT intervention, there were demonstrated minimal gait deficits and therefore had little
no significant changes in the participants’ perceived ease room for improvements. Although disease severity did not
and ability to walk unassisted. Greater reduction in LOB impact gait speed, improved stride length is a major goal of
occurrences during the 6MWT for the RAGT group com- gait training in PwPD that has important implications for gait
pared with the CON group is likely explained by a few endurance and fall risk.47,48 In support of our finding of

108 © 2024 Academy of Neurologic Physical Therapy, APTA

Copyright © 2024 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
 JNPT ● Volume 48, April 2024 Use of a Robotic Walking Device for Home and Community Mobility in Parkinson Disease
Downloaded from http://journals.lww.com/jnpt by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdtwnfKZBYtws= on 10/03/2024

Figure 4. Mean difference in stride length per 5 unit increase in UPDRS-motor score among RAGT group. Participants with higher
symptom severity at baseline showed a greater increase in stride length during unassisted walking following the intervention than
those with lower scores. In contrast, participants in the control (CON) group with higher symptom severity at baseline showed
a decline in stride length after the 8-week period. CON, control; RAGT, robotic-assisted gait training; UPDRS, Unified Parkinson
Disease Rating Scale.

greater disease severity having a positive influence on inter- compensatory strategies (eg, assistive device use) to prevent
vention response, the study by Kawashima et al25 had a higher falls and therefore were classified as “nonfallers.” An
percentage of participants with moderate disease (H&Y 3) in inverted U curve relationship found between fall frequency
the EXOD group compared with the CON group and reported and disease severity that was mediated by activity levels in
a significant increase in distance and calculated gait speed for PwPD supports this idea.49
the 3-minute walk test (3MWT) in the EXOD group com-
pared with no change in controls.25 Other possible factors that Limitations
may have contributed to the improved gait endurance A limitation of our study is the participant heterogene-
reported in the study by Kawashima et al compared with ity given the relatively low sample size, which likely con-
this trial were the ability to implement continuous overground tributed to the negative findings regarding primary measure
walking outdoors (without constraints of indoor space within gait speed improvements. Disruptions in recruitment and
homes), over a longer intervention period (3 months), and follow-up of participants due to COVID-19 pandemic restric-
better control of exercise/activity levels in the CON group.25 tions were partially responsible for lower participant num-
However Kawashima et al reported no significant changes in bers. In addition, the high proportion of early-stage
the 10-m walk test, while the 3MWT distance significantly participants with minimal gait impairments likely lessened
improved, thereby causing the calculated gait speed on the the room for improvements. Based on our findings, we cal-
3MWT to be increased.25 Thus, the use of the 3 MWT by culate that for 80% power with type I error rate of 0.05 and
Kawashima et al versus the 6 MWT and the use of calculated with a standard deviation of 0.31, there would need to be 91
gait speed using the 3MWT versus gait speed measurement evaluable participants per group to detect a 0.13 m/s change
over shorter distance using LEGSys may underlie the in gait speed at week 8, an amount that was reported as
differences in our findings. Based on our findings, PD a substantially meaningful change for gait speed in older
researchers designing future EXOD studies should consider adults.50
enrolling participants with moderate to severe disease severity Gait training activities were individualized to accom-
(H&Y 2.5-4). The lack of influence for fall history status on the modate each participant’s physical abilities and his or her
intervention response may be that some participants with specific home and community environments (eg, size of
greater gait impairments and/or who experienced previous space, inclines, obstacles, access to outdoors). These accom-
falls reduced their walking and/or implemented modations introduced some variation in the dosages and

© 2024 Academy of Neurologic Physical Therapy, APTA 109

Copyright © 2024 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
 Kegelmeyer et al JNPT ● Volume 48, April 2024

complexities of the interventions delivered. However, these collection and analysis. They also thank The Michael J. Fox
accommodations were necessary for greater transfer of Foundation, Honda Research and Development, and people
training effects into their daily lives. A HWA device–spe-
with PD who supported this project.
cific limitation was torque restriction. The device stops
providing assist when the motors reach a heat threshold,
Downloaded from http://journals.lww.com/jnpt by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW

which frequently occurred with longer times walking, higher

environmental temperatures, and participants’ higher gait REFERENCES
speed or size. More consistent assist provided throughout 1. Goetz CG, Fahn S, Martinez-Martin P, et al. Movement Disorder
each session may have produced a larger effect. While many Society-sponsored revision of the Unified Parkinson’s Disease Rating
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdtwnfKZBYtws= on 10/03/2024

participants in the RAGT group reported improved ease and Scale (MDS-UPDRS): process, format, and clinimetric testing plan.
Mov Disord. 2007;22(1):41-47.
ability to walk, and/or family members noticed improve- 2. Johansson H, Hagströmer M, Grooten WJA, Franzén E. Exercise-
ments in their walking, we did not systematically collect induced neuroplasticity in Parkinson’s disease: a metasynthesis of the
qualitative data about participants’ and caregivers’ percep- literature. Neural Plast. 2020;2020:8961493.
tions of the intervention’s effects on their walking. It is 3. Cramer SC, Sur M, Dobkin BH, et al. Harnessing neuroplasticity for
clinical applications. Brain. 2011;134(pt 6):1591-1609.
possible that ease of walking could change over a training 4. Harro CC, Shoemaker MJ, Frey OJ, et al. The effects of speed-depen-
period because of better fitness and not because of the device dent treadmill training and rhythmic auditory-cued overground walking
use. Because there was no active treatment control (eg, on gait function and fall risk in individuals with idiopathic Parkinson’s
equally dosed gait training without EXOD but with a 6-lb disease: a randomized controlled trial. Neurorehabilitation. 2014;34
weighted belt), we cannot distinguish the EXOD effect from (3):557-572.
5. Ellis T, Motl RW. Physical activity behavior change in persons with
the RAGT treatment effect. However, hypothesis 1 results neurologic disorders: overview and examples from Parkinson disease
suggest that the immediate effect of EXOD use on gait and multiple sclerosis. J Neurol Phys Ther. 2013;37(2):85-90.
parameters is minimal. The small improvements in gait 6. Schootemeijer S, van der Kolk NM, Ellis T, et al. Barriers and motiva-
parameters observed after RAGT treatment may be due to tors to engage in exercise for persons with Parkinson’s disease.
J Parkinsons Dis. 2020;10(4):1293-1299.
increased levels of walking as they correspond with effects 7. Poli P, Morone G, Rosati G, Masiero S. Robotic technologies and
seen after comparable amounts of nonrobotic conventional rehabilitation: new tools for stroke patients’ therapy. Biomed Res Int.
gait training in PwPD.51,52 Future studies may wish to con- 2013;2013:153872.
sider including the Functional Gait Assessment and/or the 8. Barbe MT, Cepuran F, Amarell M, Schoenau E, Timmermann L. Long-
Timed Up and Go as recommended by the PD EDGE for term effect of robot-assisted treadmill walking reduces freezing of gait in
Parkinson’s disease patients: a pilot study. J Neurol. 2013;260(1):296-298.
their ability to measure functional mobility in this 9. Paker N, Bugdayci D, Goksenoglu G, Sen A, Kesiktas N. Effects of
population. robotic treadmill training on functional mobility, walking capacity,
motor symptoms and quality of life in ambulatory patients with
Parkinson’s disease: a preliminary prospective longitudinal study.
Neurorehabilitation. 2013;33(2):323-328.
CONCLUSIONS 10. Sale P, De Pandis MF, Le Pera D, et al. Robot-assisted walking training
This study demonstrates that it is feasible and safe for for individuals with Parkinson’s disease: a pilot randomized controlled
PTs to perform gait training using a robotic EXOD in the trial. BMC Neurol. 2013;13:50.
homes and the communities of PwPD. When initiating 11. Furnari A, Calabro RS, De Cola MC, et al. Robotic-assisted gait training
in Parkinson’s disease: a three-month follow-up randomized clinical
EXOD device usage with PwPD, PTs should provide gait trial. Int J Neurosci. 2017;127(11):996-1004.
training to allow them to become comfortable and confi- 12. Fundaro C, Giardini A, Maestri R, Traversoni S, Bartolo M, Casale R.
dent walking with the device before testing or to ensure Motor and psychosocial impact of robot-assisted gait training in a
safe walking with the device prior to independent use. The real-world rehabilitation setting: a pilot study. PLoS One. 2018;13(2):
RAGT program used in this mostly low impairment popu- e0191894.
13. Capecci M, Pournajaf S, Galafate D, et al. Clinical effects of robot-
lation of PwPD may not be effective enough and/or may assisted gait training and treadmill training for Parkinson’s disease.
not have been sufficiently dosed to see a positive treatment A randomized controlled trial. Ann Phys Rehabil Med. 2019;62
effect. RAGT may be more beneficial in individuals with (5):303-312.
PD in H&Y stages 2.5 and greater who have gait and 14. Rodriguez-Fernandez A, Lobo-Prat J, Font-Llagunes JM. Systematic
review on wearable lower-limb exoskeletons for gait training in neuro-
balance deficits. Thus, researchers designing studies exam- muscular impairments. J Neuroeng Rehabil. 2021;18(1):22.
ining the effects of robotic EXODs on gait and mobility in 15. Yun SJ, Lee HH, Lee WH, Lee SH, Oh BM, Seo HG. Effect of
PwPD should consider disease severity when selecting robot-assisted gait training on gait automaticity in Parkinson disease:
participants. Future studies should also explore the effects a prospective, open-label, single-arm, pilot study. Medicine (Baltimore).
of single-joint versus multijoint EXOD usage on gait in 2021;100(5):e24348.
16. Picelli A, Capecci M, Filippetti M, et al. Effects of robot-assisted gait
PwPD. training on postural instability in Parkinson’s disease: a systematic
review. Eur J Phys Rehabil Med. 2021;57(3):472-477.
17. Young AJ, Ferris DP. State of the art and future directions for lower
ACKNOWLEDGMENTS limb robotic exoskeletons. IEEE Trans Neural Syst Rehabil Eng.
2017;25(2):171-182.
The authors thank Jessica Donovan, Dori Walterhouse, 18. Satoh Y, Yamada T, Arai Y, et al. The immediate effect of the Honda
Dr Barbara Changhizi, Alex Janis, Steven Ciolek, Madison Walking Assist Device on foot and ankle function in hemiplegic stroke
patients. J Phys Ther Sci. 2020;32(6):405-409.
Blake, Marielle Lynch, Sara McKeeman, Anna Weibel, Julia 19. Shimada H, Kimura Y, Suzuki T, et al. The use of positron emission
Goodwin, and Nathan O’Malley for their assistance on data tomography and [18F]fluorodeoxyglucose for functional imaging of

110 © 2024 Academy of Neurologic Physical Therapy, APTA

Copyright © 2024 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
 JNPT ● Volume 48, April 2024 Use of a Robotic Walking Device for Home and Community Mobility in Parkinson Disease

muscular activity during exercise with a stride assistance system. IEEE 36. Aminian K, Najafi B, Bula C, Leyvraz PF, Robert P. Spatio-temporal
Trans Neural Syst Rehabil Eng. 2007;15(3):442-448. parameters of gait measured by an ambulatory system using miniature
20. Buesing C, Fisch G, O’Donnell M, et al. Effects of a wearable exoske- gyroscopes. J Biomech. 2002;35(5):689-699.
leton stride management assist system (SMA(R)) on spatiotemporal gait 37. Ritter PL, Lorig K. The English and Spanish Self-Efficacy to Manage
characteristics in individuals after stroke: a randomized controlled trial. Chronic Disease Scale measures were validated using multiple studies.
J Neuroeng Rehabil. 2015;12:69. J Clin Epidemiol. 2014;67(11):1265-1273.
Downloaded from http://journals.lww.com/jnpt by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW

21. Shimada H, Suzuki T, Kimura Y, et al. Effects of an automated stride 38. Giladi N, Tal J, Azulay T, et al. Validation of the freezing of gait
assistance system on walking parameters and muscular glucose meta- questionnaire in patients with Parkinson’s disease. Mov Disord.
bolism in elderly adults. Br J Sports Med. 2008;42(11):922-929. 2009;24(5):655-661.
22. Otsuki R, Matsumoto H, Ueki M, et al. Automated stride assistance 39. Dal Bello-Haas V, Klassen L, Sheppard MS, Metcalfe A. Psychometric
device improved the gait parameters and energy cost during walking of properties of activity, self-efficacy, and quality-of-life measures in
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC1y0abggQZXdtwnfKZBYtws= on 10/03/2024

healthy middle-aged females but not those of young controls. J Phys individuals with Parkinson disease. Physiother Can. 2011;63(1):47-57.
Ther Sci. 2016;28(12):3361-3366. 40. Lee YJ, Ellenberg JH, Hirtz DG, Nelson KB. Analysis of clinical trials
23. Shimada H, Hirata T, Kimura Y, et al. Effects of a robotic walking by treatment actually received: is it really an option? Stat Med. 1991;10
exercise on walking performance in community-dwelling elderly adults. (10):1595-1605.
Geriatr Gerontol Int. 2009;9(4):372-381. 41. Ruiz-Canela M, Martinez-Gonzalez MA, de Irala-Estevez J. Intention to
24. Panizzolo FA, Cimino S, Pettenello E, Belfiore A, Petrone N, treat analysis is related to methodological quality. BMJ. 2000;320
Marcolin G. Effect of a passive hip exoskeleton on walking distance (7240):1007-1008.
in neurological patients [published online ahead of print March 5, 2021]. 42. Sanchez-Ferro A, Matarazzo M, Martinez-Martin P, et al. Minimal
Assist Technol. 2022;34(5):527-532. clinically important difference for UPDRS-III in daily practice. Mov
25. Kawashima N, Hasegawa K, Iijima M, et al. Efficacy of wearable device Disord Clin Pract. 2018;5(4):448-450.
gait training on Parkinson’s disease: a randomized controlled open-label 43. Jankovic J, McDermott M, Carter J, et al. Variable expression of
pilot study [published online ahead of print February 8, 2022]. Intern Parkinson’s disease: a base-line analysis of the DATATOP cohort.
Med. 2022;61(17):2573-2580. The Parkinson Study Group. Neurology. 1990;40(10):1529-1534.
26. Gamborg M, Hvid LG, Thrue C, et al. Muscle strength and power in 44. Berriozabalgoitia R, Bidaurrazaga-Letona I, Otxoa E, Urquiza M,
people with Parkinson disease: a systematic review and meta-analysis. Irazusta J, Rodriguez-Larrad A. Overground robotic program preserves
J Neurol Phys Ther. 2023;47(1):3-15. gait in individuals with multiple sclerosis and moderate to severe
27. Nieuwboer A, Kwakkel G, Rochester L, et al. Cueing training in the impairments: a randomized controlled trial. Arch Phys Med Rehabil.
home improves gait-related mobility in Parkinson’s disease: the 2021;102(5):932-939.
RESCUE trial. J Neurol Neurosurg Psychiatry. 2007;78(2):134-140. 45. McIntosh K, Charbonneau R, Bensaada Y, Bhatiya U, Ho C. The safety
28. Osborne JA, Botkin R, Colon-Semenza C, et al. Physical therapist manage- and feasibility of exoskeletal-assisted walking in acute rehabilitation
ment of Parkinson disease: a clinical practice guideline from the American after spinal cord injury. Arch Phys Med Rehabil. 2020;101(1):113-120.
Physical Therapy Association. Phys Ther. 2022;102(4):pzab302. 46. Nam YG, Lee JW, Park JW, et al. Effects of electromechanical
29. Tu XJ, Hwang WJ, Ma HI, Chang LH, Hsu SP. Determinants of generic exoskeleton-assisted gait training on walking ability of stroke patients:
and specific health-related quality of life in patients with Parkinson’s a randomized controlled trial. Arch Phys Med Rehabil. 2019;100
disease. PLoS One. 2017;12(6):e0178896. (1):26-31.
30. Soh SE, McGinley JL, Watts JJ, et al. Determinants of health-related 47. Ambrus M, Sanchez JA, Sanchez Miguel JA, Del-Olmo F. Test-retest
quality of life in people with Parkinson’s disease: a path analysis. Qual reliability of stride length-cadence gait relationship in Parkinson’s dis-
Life Res. 2013;22(7):1543-1553. ease. Gait Posture. 2019;71:177-180.
31. Hoehn M, Yahr M. Parkinsonism: onset, progression and mortality. 48. Lai YR, Lien CY, Huang CC, et al. Clinical disease severity mediates
Neurology. 1967;17:427-442. the relationship between stride length and speed and the risk of falling in
32. Uschner D, Schindler D, Hilgers R-D, Heussen N. randomizeR: an Parkinson’s disease. J Pers Med. 2022;12(2):192.
R package for the assessment and implementation of randomization in 49. Mactier K, Lord S, Godfrey A, Burn D, Rochester L. The relationship
clinical trials. J Stat Softw. 2018;85(8):1-22. between real world ambulatory activity and falls in incident Parkinson’s
33. Steffen T, Seney M. Test-retest reliability and minimal detectable disease: influence of classification scheme. Parkinsonism Relat Disord.
change on balance and ambulation tests, the 36-item short-form health 2015;21(3):236-242.
survey, and the Unified Parkinson Disease Rating Scale in people with 50. Perera S, Studenski S, Chandler JM, Guralnik JM. Magnitude and
parkinsonism. Phys Ther. 2008;88(6):733-746. patterns of decline in health and function in 1 year affect subsequent
34. Kobayashi E, Himuro N, Takahashi M. Clinical utility of the 6-min walk 5-year survival. J Gerontol A Biol Sci Med Sci. 2005;60(7):894-900.
test for patients with moderate Parkinson’s disease. Int J Rehabil Res. 51. Miyai I, Fujimoto Y, Yamamoto H, et al. Long-term effect of body weight-
2017;40(1):66-70. supported treadmill training in Parkinson’s disease: a randomized con-
35. Tasvuran Horata E, Cetin SY, Erel S. Effects of individual progressive trolled trial. Arch Phys Med Rehabil. 2002;83(10):1370-1373.
single- and dual-task training on gait and cognition among older healthy 52. Tomlinson CL, Patel S, Meek C, et al. Physiotherapy versus placebo or
adults: a randomized-controlled comparison study. Eur Geriatr Med. no intervention in Parkinson’s disease. Cochrane Database Syst Rev.
2021;12(2):363-370. 2013;2013(9):CD002817.

© 2024 Academy of Neurologic Physical Therapy, APTA 111

Copyright © 2024 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.