Treadmill Handrail-use Increases the Anteroposterior Margin of Stability in Individuals’ Post-stroke

Abstract

Treadmills are important rehabilitation tools used with or without handrails. The handrails could be used to attain balance, prevent falls, and improve the walking biomechanics of stroke survivors, but it is yet unclear how the treadmill handrails impact their stability margins. Here, we investigated how 3 treadmill handrail-use conditions (no-hold, self-selected support, and light touch) impact stroke survivors’ margins of stability (MoS). The anteroposterior MoS significantly increased for both legs with self-selected support while the mediolateral MoS of the unaffected leg decreased significantly when the participants walked with self-selected support in comparison to no-hold in both cases. We concluded that the contextual use of the handrail should guide its prescription for fall prevention or balance training in rehabilitation programs.

Introduction

Stroke may lead to long-term disability (Go et al., 2013), and about two-thirds of people with acute stroke lose the ability to independently ambulate (Jørgensen et al., 1995) after the cerebrovascular accident. More than 60% of those who attain independent ambulation walk below community walking optimum speeds due to motor weakness, poor coordination, limited endurance, and gait instability (Chen et al., 2005; Hsiao et al., 2017; Jørgensen et al., 1995). Human movement models have shown that gait stability is attained through the interaction of an individual’s center of mass (CoM) and their base of support (BoS) so that the goal of bipedal movement is to retain the projecting CoM within the BoS using the placement of the foot for every step taken (Bruijn & Van Dieën, 2018; MacKinnon & Winter, 1993; Townsend, 1985). Also, Hof et al. (2008), with an inverted pendulum model, suggested a stability-attaining theory in which a perturbing moment - resulting from a projected center of pressure (CoP) to extrapolated center of mass (xCoM) distance - must be controlled by timely displacing the CoP, to attain balance while walking (Hof, 2008). This theory implies that mediolateral (ML) and/or anteroposterior (AP) stability can be attained through the anterior and lateral placement of the foot in motion to control an individual’s magnitude of the margin of stability (MoS)(Hof, 2008; Tyson et al., 2006), but this balance strategy may be difficult to attain in survivors of stroke due to improper gait adaptations resulting from the paresis of their affected leg following the cerebrovascular accident (Dean & Kautz, 2015; Haarman et al., 2017).

Survivors of stroke often present with altered weight distribution patterns and a greatly increased sway in their posture, as well as a decreased bodyweight excursion towards the paretic side leg resulting in poor gait biomechanics and exposing some to the risks of falls (Dean & Kautz, 2015; Hyndman et al., 2002; Kao et al., 2014; Tyson et al., 2006). Physical therapists look to address these gait asymmetries correlated with balance impairments through therapeutic interventions including the introduction of assistive devices like the walking cane in over-the-ground walking and handrails in treadmill rehabilitation (Hsiao et al., 2017; Ijmker et al., 2015). These devices are recommended in some cases for immediate fall prevention in people who are unable to walk without support, while in other cases are recommended for rehabilitation processes to improve gait biomechanics.

A previous study by Jeka et al. provides evidence for the effectiveness of this intervention by showing that light contact sensory cues at the fingertips are sufficient to reduce postural sway in static or dynamic situations for sighted and blinded individuals (Jeka et al., 1996). Further research has also corroborated this by using light touch on cane handles. Bellicha et al. deduced that the light-grip of an instrumented cane handle reduced AP and ML sway compared to the no-grip of the handle (Bellicha et al., 2021), and Kang et al. also showed that treadmill walking while holding the handrails facilitated somatosensory changes that improved the plantar foot pressure and foot contact area of his stroke participants, thereby improving their gait (Kang et al., 2015). Other studies investigating how assistive device use impacted the gait of survivors of stroke found benefits such as compensation for impaired motor control in the affected limb, normalization of spatial variables, as well as the improvement of functional mobility, and the prevention of falls (Kuan et al., 1999; Tyson & Rogerson, 2009). In their study to investigate how treadmill handrail use impacts the energetics, step parameters, and neuromuscular activity of survivors of stroke, IJmker et al. showed that people with stroke experienced normalization of their step parameters when they walked while holding the treadmill handrails. The study participants took longer steps with decreased step width at a better step length symmetry and experienced a reduction in their energy cost of walking on the treadmill while holding the handrail. In contrast, participants only showed a small decrease in step width when walking with a light touch of the handrail. The investigators suggested that the observed result was due to the balance support available in the treadmill handrail condition (Ijmker et al., 2015). The above studies demonstrate how both a light touch and firm hold of a cane or treadmill handrail could improve posture and step parameters in stroke survivors, yet no study has explicitly shown how different extents of handrail use could influence the biomechanical parameters that define balance to ensure mechanical gait stability in stroke survivors while walking on the treadmill for rehabilitation purposes. An improved understanding of this could guide how handrails are used in treadmill walking for stroke survivors to attain gait stability during rehabilitation.

In this study, we looked to answer the question of how different extents of treadmill handrail use impact the stability of people with stroke. We investigated how three handrail-use conditions influence the biomechanical parameters that define balance to impact their gait stability; we investigated how walking without the handrail, with a light touch of the handrail, and with a self-selected hold of the handrail instigated changes to the ML foot placement and the stability margins while people with stroke used a one-sided treadmill handrail on their non-paretic side. We evaluated the anteroposterior and mediolateral margin of stability (AP-MoS and ML-MoS), the step width (SW), and the mediolateral distance of the leading foot to the CoM graded as the lateral foot placement (LFPL) of the paretic and unaffected legs of the participants. We hypothesized that the participants’ paretic and non-paretic leg MoS would increase while the SW would decrease when the participants hold the handrails with Self-selected support compared with a Light Touch or No-Hold of the handrails, as the Ijmker et al. study showed evidence of a more economic gait with improved step parameters when their participants walked with increased support. The offered handrail support may enhance balance control thereby enabling the participants in the study to walk with an optimized step width due to the added base of support by the treadmill handrails; some studies have reported similar findings of increased stability and step width reduction with assistive device use in the stroke population (Boonsinsukh et al., 2009; Ijmker et al., 2015; Kuan et al., 1999).

Methods

Participants

Nineteen persons post-stroke (12F, 7M; Age = 59.21 ± 13.86) were recruited for this study (Table 1). The study took place at the University of Nebraska at Omaha. Participants included were between the ages of 19–80, had a stroke greater than 6 months (mean years since stroke = 2.16 ± 2.07 years), could walk independently with or without an assistive device, had a resting heart rate between 40–100 beats per minute, and had resting blood pressure between 90/60 and 170/90 mmHg. Individuals were excluded if they had pain in their legs or spine that limits their walking, more than one stroke, evidence of a cerebellar stroke on an MRI, any unexplained dizziness in the last 6 months, visual impairments that prevent viewing content on a screen 5 feet away, neurotoxin injection treatments within the past 3 months or an inability to communicate with investigators. Written consent was collected from all participants and the study was approved by the Institutional Review Board at the University of Nebraska Medical Center.

Table 1.

Demographic and clinical characteristics of participants

Participant Demographics / clinical characteristics
All (N)	19
Age (years)	59.21 ± 13.86 (range 39 – 75)
Weight (Kg)	87.54 ± 17.81
Sex	12 F / 7 M
Years since stroke (Avg)	2.16 ± 2.07
Paretic Side	11 L / 8 R
Assistive device use	Yes [n = 7] / No [n = 12]

Experimental Procedure

Sixty-five retroreflective markers were attached to each participant’s torso and upper and lower extremities. The treadmill gait analysis session took place on a split-belt instrumented treadmill with 2 embedded six-degree-of-freedom force platforms and custom-designed instrumented handrails (Bertec Corp, Columbus, OH). Force data were collected at 1000 Hz from each side using the embedded force platforms and handrails. The marker data were collected using a 16-camera motion analysis system at 100 Hz (Vicon Motion Systems, Oxford, UK). All treadmill trials were performed at the participants’ self-selected speed. To determine the participants’ self-selected walking speed, the treadmill was initially set to 0.1 m/s and increased by 0.1 m/s until the participant verbally indicated they were walking at a comfortable speed and the participant proceeded to walk at their comfortable speed for thirty seconds before stopping the treadmill (Hedrick et al., 2021; Liu et al., 2020). The self-selected speed was determined without using the handrails if they could. Each treadmill condition was three minutes long and the three conditions included: no handrails (NHR), light support handrail (5%HR), and self-selected handrail (SSHR) use. Individuals wore a safety harness for all trials with no body weight support. For the SSHR condition, participants could use the handrail however they want, and they were all instructed to hold onto a side handrail with their non-paretic hand alone because not all survivors of stroke fully recover the use of their paretic hand (Barreca et al., 2003). For the 5%HR, real-time feedback of the handrail forces was displayed on a screen that was in front of the participants while they also used the non-paretic side handrail. Participants saw either a red “X” or a green “O” in front of them on the screen. The screen displayed the green “O” while all force (vertical, horizontal, lateral) applied to the treadmill handrail remained below 5% of the participant’s body weight. If the force threshold was exceeded, a red “X” was displayed instantaneously. Before the gait trials, the symbols were explained to the participants, and they were instructed to reduce the amount of force they applied on the handrails if the red “X” appeared. The observed mean ± SD vertical force (normalized to bodyweight in Newtons) applied to the treadmill handrail by all participants was 0.0102 ± 0.006 (N/BW in N) at the 5%HR condition and 0.022 ± 0.011 (N/BW in N) at the SSHR condition, and individual forces for participants ranged from <0.01N to 62.75N and <0.01N to 176.56N for 5%HR and SSHR conditions respectively. The recorded forces varied across a wide range in both conditions for individual participants. For the NHR, participants were instructed not to use the handrails during the three-minute trial. Three out of 19 total participants were unable to complete the NHR trials and were therefore exempted from the statistical analysis, so we analyzed the results for 16 participants. The participants were not allowed any familiarization period for any of the treadmill handrail conditions, and the order of the three conditions was randomized for each participant. If needed, participants were allowed 3–5 minutes or longer of rest between walking trials.

Data Analysis

Kinematic and kinetic data from the treadmill conditions were collected in Nexus (VICON, Oxford, UK). Calculations were performed in Visual 3D software (C-Motion, Inc., Germantown, MD, USA) as well as MATLAB (Mathworks, Natick, MA, USA). A 4^th order low pass Butterworth filter was used in filtering both the kinetic (60Hz) and kinematic (6Hz) data. The MoS, LFPL, and SW were calculated as shown below.

Margin of Stability (MoS):

The MoS is a variable for defining mechanical gait stability in dynamic situations (Fallahtafti et al., 2021), and more studies are beginning to adopt it to quantify mechanical gait stability in post-stroke populations (Watson et al., 2021). While earlier studies had established that the regulation of the position of the center of mass (CoM) relative to an individual’s base of support (BoS - the area bounded by the feet) is the primary condition for stability, the limitation of this condition for dynamic situations warranted further work (Hof, 2008; Hof et al., 2005). In 2005, Hof et al. introduced an extrapolated center of mass concept (XcoM) on the premise that the CoM path is extrapolated in the direction of its velocity (vCoM)(Hof et al., 2005). Using the XcoM in an inverse pendulum model (Figure 1) where the pendulum length ‘Ɩ’ suspends the CoM and ‘g’ is the gravitational force, they calculated the MoS (equation 1) as the perpendicular distance between the position of the XcoM and the BoS, also taken as the tenable range confining the center of pressure (CoP) (Hof, 2008; Hof et al., 2005).

Figure 1.

The inverse pendulum model shows the motion of the black mass (CoM) suspended on the leg length (pendulum height ‘l’) as the leading foot moves in mid-stance. The dotted circles show the swing of the CoM to account for the movement extrapolated CoM (XcoM), which represents the position of the CoM about its velocity (VcoM). ‘g’ represents the gravitational acceleration, and the anterior limit of the BoS in relation to the foot is shown.

Where

$$\text{XcoM}=CoM+\frac{\mathit{\text{vCoM}}+{\text{V}}_{\text{BS}}}{\surd \frac{g}{l}}$$ equation1

$$\text{MoS}=\text{BoS}-\left(\mathit{\text{XcoM}}\right)$$ equation2

vCoM is the velocity of the CoM which we calculated using Visual 3D, in relation to the velocity of the anterior boundary of the base of support for each participant $\left({\text{V}}_{\text{BS}}\right)$ referenced as their self-selected walking speed during the treadmill walking trials. Also, ‘Ɩ’ for both legs was derived as the magnitude of the leg length pendulum at the sagittal plane calculated from the leg vectors (distance from the CoM to ankle joint center) in the vertical and anteroposterior direction. The MoS was calculated using Visual 3D, and we derived the AP-MoS and ML-MoS for the participants’ paretic and non-paretic feet. The MoS was calculated at initial contact for each limb with the BoS defined using the toe marker of both legs in the sagittal plane for the AP-MoS, and the 5^th metatarsal marker for both legs in the frontal plane for the ML-MoS. The anteroposterior MoS for the leg was calculated as the distance from the sagittal plane position of the toe markers to the anteroposterior location of the extrapolated CoM position (Fallahtafti et al., 2021; McAndrew Young et al., 2012) while we calculated the mediolateral MoS as the distance from the frontal position of the 5^th metatarsal marker to the mediolateral position of the extrapolated CoM position.

Step width:

The SW was defined as the mediolateral distance between the lateral malleoli markers of the leading and the trailing legs at initial contact of each of the limbs (Hsiao et al., 2017). This analysis was done using Visual 3D.

Lateral foot placement:

The LFPL was calculated as the mediolateral distance between the CoM and lateral malleolus of the leading limb at initial contact (Hsiao et al., 2017). This analysis was done using Visual 3D.

All the variables were averaged across the 3 minutes trials.

Statistical Analysis

We performed a multilevel-model analysis (hierarchical linear models) in R/R-studio (R Core Team, 2021; RStudio Inc., Boston, MA, USA) using the ‘1merTest’ package (Kuznetsova, Brockhoff & Christensen, 2017) to evaluate the effects of the handrail conditions on each variable for the affected and unaffected side at a significance level of 0.05. This multilevel-model analysis is based on the assumptions of linearity, and it was used because it has less strict assumptions than a standard regression model and it enables the incorporation of variables from every level (Raudenbush & Bryk, 2002) while accounting for missing data. With this model, we can evaluate unique intercepts for each participant individually. The model can evaluate variances at its different levels to identify changes within a participant and variables that change across all individuals (Woltman et al., 2012).

The multilevel model was done to analyze the AP-MoS, ML-MoS, SW, and LFPL with a hierarchical approach. The first step in the model tested the effect of the handrail (HR) conditions on the different variables individually, after which we added the leg-side (paretic and non-paretic) as a 2^nd level to the model before testing the interaction between side and handrail condition in a 3^rd level for the model. Where we found fixed effects, we performed an additional analysis with a post-hoc comparison using Tukey’s HSD to adjust for multiple comparisons either at the level of handrail conditions, leg-side, or assessing interactions at the multi-levels.

Results

Relationship of handrail use conditions and leg-side on the MoS

AP-MoS:

The model for AP-MoS that included HR condition as a factor was significant (χ2(2, N=16) = 13.56; p = 0.001), and the other models that included leg-side as a factor (χ2(1, N=16) = 1.15; p = 0.284) and the interaction between HR condition and leg-side (χ2(2, N=16) = 0.29; p = 0.866) were not significant. We selected the significant model and noted a significant main effect of handrail conditions (F (2, 78) = 7.205, p = 0.001) with a large effect (partial η2 = 0.16), and the post-hoc analysis with Tukey’s HSD correction for pairwise comparisons showed a significantly higher AP-MoS at the SSHR condition compared to the NHR condition (t(78) = −3.765, p = 0.001) with a large effect size (Cohen’s d = −0.941), while there was no significant difference between the 5%HR and SSHR (t(78) = −1.459, p = 0.316) condition and between the NHR and 5%HR condition (t(78) = −2.305, p = 0.061.

ML-MoS:

The model for ML-MoS that included both HR condition and leg-side as factors was significant (χ2(1, N=16) = 39.45); p < 0.0001), as well as the model that included the interaction between HR condition and leg-side (χ2(2, N=16) = 9.37; p = 0.023), but the model that had only HR condition as a factor was not significant (χ2(2, N=16) = 0.606; p = 0.739). The model that included the interaction effect was selected due to the improved Akaike’s Information Criteria (AIC = −468.11), and this model showed a significant main effect of side (F (2, 75) = 52.54, p < 0.0001) with a large effect (partial η2 = 0.41) and an interaction effect of the Handrail conditions with the side factor (F (2, 75) = 3.716, p = 0.029) with a medium effect (partial η2 = 0.09). Further analysis on the interaction effect with Tukey’s HSD correction for pairwise comparisons on the handrail conditions at each level of side showed a significantly smaller ML-MoS at the SSHR condition compared to the NHR condition with a large effect (Cohen’s d = 0.88) for the non-paretic leg (t(75) = 2.497, p = 0.039), while there were no other significant differences for comparisons of the handrail conditions of both the paretic and non-paretic legs (p > 0.05). χ2

Relationship of handrail use conditions and leg-side on the SW

SW:

The model for SW that included HR condition as a factor was significant (χ2(2, N=16) = 69.32; p < 0.0001), and the other models that included only leg-side as a factor (χ2(1, N=16) = 0.2797; p = 0.597) and the interaction between HR condition and leg-side (χ2(2, N=16) = 0.120; p = 0.942) were not significant. The significant model showed a main effect of the handrail conditions (F(2, 78) = 53.77, p < 0.0001) with a large effect (partial η2 = 0.58), and post-hoc analysis using Tukey’s HSD correction for pairwise comparisons showed a significantly larger SW at the NHR condition compared to the 5%HR condition (t(78) = 8.287, p < 0.0001) with a large effect (Cohen’s d = 2.07), and at the NHR condition compared to the SSHR condition (t(78) = 9.542, p < 0.0001) with a large effect (Cohen’s d = 2.39), while there was no significant difference between the 5%HR and SSHR (t(78) = 1.255, p = 0.425) condition.

Relationship of handrail condition and leg-side on the LFPL

LFPL:

The models for LFPL that included only HR condition as a factor (χ2(2, N=16) = 6.788; p = 0.034), the leg-side as a factor with HR condition (χ2(1, N=16) = 67.87; p < 0.0001), and the interaction between HR condition and leg-side (χ2(2, N=16) = 9.787; p = 0.008) were all significant. The model that included the interaction between HR condition and leg side was selected due to improved Akaike’s Information Criteria (AIC = −497.49). This model showed a main effect of HR condition (F(2, 75) = 8.766, p < 0.001) with a large effect (partial η2 = 0.19), a main effect of leg-side (F(1, 75) = 113.23, p < 0.0001) with a large effect (partial η2 = 0.60), and an interaction effect of HR condition and leg-side (F(2, 75) = 4.88, p = 0.010) with a large effect (partial η2 = 0.12). Further analysis was done on the interaction effect with Tukey’s HSD correction for pairwise comparisons on the handrail conditions at each level of the leg side, and the non-paretic side showed a significantly smaller LFPL at the 5%HR condition compared to the NHR condition (t(75) = 2.732, p = 0.021) with a large effect (Cohen’s d = 0.97) and at the SSHR condition compared to the NHR condition (t(75) = 5.118, p < 0.0001) with a large effect (Cohen’s d = 1.81). There were no other significant differences in comparisons of the HR conditions of both the non-paretic and paretic legs (p > 0.05).

Discussion

In this study, we investigated how three different treadmill handrail-use conditions impacted the margin of stability in the stance phase of the gait of persons with stroke. Survivors of stroke may experience an unstable gait due to a decreased ability to control the placement of their foot which may result from poor coordination, decreased lower limb excursion, increased nonparetic leg compensation, and the imbalance of their paretic or weak leg (Dean & Kautz, 2015; Kao et al., 2014; Lamontagne et al., 2007; Tyson et al., 2006). Assistive devices could aid in ensuring their stability while walking, so in this study, we investigated how the use of treadmill handrails impacted the stability of people with stroke by evaluating the mediolateral (ML-MoS) and anteroposterior margin of stability (AP-MoS), the step width (SW), and the lateral foot placement (LFPL) of both their paretic and nonparetic legs while they walked over the treadmill.

Our results showed that the treadmill handrail conditions significantly impacted the AP-MoS of the participants without taking the side of the leg into account, and the AP-MoS was significantly larger only at the SSHR condition compared to the NHR condition, a finding consistent with our hypothesis. On the contrary, the impact of the handrail condition on the ML-MoS was dependent on the side of the leg (paretic or nonparetic) as was noted through the significance of the model representing the interaction between the handrail conditions and the leg-side, which resulted in a significantly smaller ML-MoS at the SSHR condition compared to the NHR condition only for the nonparetic leg. The observed interaction did not result in any significant differences in the ML-MoS between the NHR, 5%HR, and the SSHR conditions of the paretic leg. Also, consistent with the study hypothesis, the handrail conditions impacted the step width independent of the leg side so that the step width of the participants decreased significantly when they had increased support at either the 5%HR or the SSHR conditions independent of the sides of their legs.

These results suggest that there can be improved stability margins for the paretic limb of survivors of stroke when they have increased support from the treadmill handrails with a self-selected hold (SSHR), but this may only be significant for the AP-MoS. The handrail-use conditions also impacted the stepping strategy of the participants in a manner reflective of findings in studies that have shown that a commonly adopted motor control mechanism for attaining gait stability and increasing the AP and ML MoS is by taking wider and/or longer steps, which invariably results in increased step parameters like the step width (Bruijn & Van Dieën, 2018; Dean & Kautz, 2015; McAndrew Young & Dingwell, 2012). Here we observed that the handrail condition significantly decreased the SW for the paretic and nonparetic side at the 5%HR and the SSHR conditions, and this result was similar to findings in the study of Ijmker et al. where participants following stroke walked with smaller step width when they had increased support from a treadmill handrail at a handrail-hold condition compared to a light touch (Ijmker et al., 2015); these findings are attributed to increased balance control at the handrail-hold conditions. While we may also try attributing the decreased SW to a sign of improved stability, the MoS – which was the balance control measure in our study - was significantly larger for both legs only in the AP direction while there was a significantly smaller nonparetic leg ML-MoS at the SSHR condition compared to the NHR condition. The observed smaller nonparetic leg ML-MoS was in tandem with the decreased lateral placement of the nonparetic leg as the handrail support increased; our results showed that this lateral foot placement was significantly impacted by the handrail conditions while interacting with the leg-sides as an additional factor, and this resulted in statistically significant decreases in the LFPL of only the nonparetic leg at the 5%HR and SSHR conditions compared to the NHR condition. The handrail support on the nonparetic side may have imposed an adaptation restricted only to the nonparetic foot in response to the side of handrail placement. This adaptation may have resulted in the observed decrease in nonparetic LFPL resulting in the smaller nonparetic leg ML-MoS. Also, this LFPL (Barreca et al., 2003; Richards & Pohl, 1999) measures tell us about the distance of the stance leg relative to the CoM position and is distinct from the ML-MoS in that the stability margin measures the mediolateral base of support in relation to the extrapolated center of mass while the LFPL does not (Bruijn & Van Dieën, 2018; Hof et al., 2005; Hsiao et al., 2017; Townsend, 1985). The observed similarities in their results are therefore expected and the decreased non-paretic leg ML-MoS may be attributed to the decreased lateral foot placement of the same leg at the 5%HR and SSHR conditions.

Previous studies showed that the light touch of the fingertip on a stable object surface sends sensory information to the brain, and this information could aid in mediating postural sway in people standing on one leg, both legs and those with balance issues (Bellicha et al., 2021; Jeka, 1997; Lackner et al., 1999). This current study investigated how different treadmill handrail conditions impact the margin of stability in stance for stroke survivors. The high level of importance of treadmills in the rehabilitation of stroke patients for high-intensity training and walking re-education makes this study clinically relevant (Kuys et al., 2011). Studies have reported improved symmetry, longer paretic leg single stance period, improved walking distance and speed, and a more normalized paretic leg muscle activation pattern following treadmill rehabilitation (Gelaw et al., 2019; Hesse et al., 1999; Kuys et al., 2011). This study is the first to report how different handrail use conditions impact the stability of survivors of stroke in treadmill walking using the margin of stability measures. We showed that the use of treadmill handrails with a self-selected hold could better improve the anteroposterior margin of stability of the paretic limb of stroke patients, while the treadmill handrail-use condition had a statistically significant interaction effect with the side of the participant’s legs relative to the mediolateral stability margins, and this resulted in a smaller non-paretic leg ML-MoS when the participants used the handrail with a self-selected force, a finding referenced to the placement of the non-paretic foot.

Despite the positives in this study, it is limited first in that we are not certain if the observed adaptations in the participants during treadmill walking would be retained following rehabilitation, or if these results are transferrable to overground walking with the use of a cane or other walking aids. Furthermore, participants who could not perform the NHR conditions were excluded from the statistical analysis in this study, and we believe this insinuates a need for caution in the translation of the results for individual patients. Also, the restriction of the handrail use to the non-paretic side may have influenced the observed results considering that the impacts of the handrail conditions on the foot placement was isolated to the non-paretic leg. Future studies could explore these limitations to further inform handrail use conditions in treadmill walking and rehabilitation. The results from this study may not be translatable to overground walking with the use of a cane since the treadmill environment differs from walking overground (Puh & Baer, 2009) as we also noted that only two participants walked at a faster preferred walking speed on the treadmill compared to their overground speed which we calculated as the average speed in a 6 minutes’ walk test; a paired t-test showed that the means of the participants overground speed (0.99 ± 0.27 m/s) significantly differed (t(15) = −5.441, p < 0.0001) from their treadmill walking speed (0.68 ± 0.38 m/s). Future studies can investigate how assistive device use impacts stability margins overground to tackle this limitation.

In making treadmill rehabilitation decisions, factors such as the phase of stroke (acute or chronic), the initial level of stability and tendency of the patient to fall, the goal of the intervention, and the observed changes in progression-to-recovery will have to be considered by the physical therapist in making decisions on the use of support devices (handrails and/or harness) during interventions (Gelaw et al., 2019). While we understand that support devices are necessary for individuals unable to walk independently and those at immediate risk of falls, it is necessary to evaluate how these devices impact stability and gait in individuals in the chronic phase of stroke as some may benefit more from walking without assistive device use. Also, physical therapists could construct programs in which they initiate perturbations or introduce conditions that challenge a patient’s stability as an adopted means of training balance; smaller MoS in such cases may be positively aimed even with the use of treadmill handrails, whereas large stability margins could aid fall prevention. Our findings inform the use of treadmill handrails in the rehabilitation of survivors of stroke, as one of the major goals of rehabilitation is to improve the stability of the paretic leg of these individuals to encourage increased weight transfer to this leg while walking. An increase in an individual’s MoS may imply an improvement in their stability (Hof et al., 2005) which could result in the reduction of fall risks in such individuals. Survivors of stroke may gain greater benefits from the self-selected use of treadmill handrails during rehabilitation, but there is also a need to exercise caution in making such therapeutic choices due to likely adaptation mechanisms that could result in a possible decreased non-paretic leg mediolateral margin of stability despite the observed improved anteroposterior margin of stability. Factors such as the patient’s level of independence, and their ability to walk with or without the use of the treadmill handrails could be helpful information in guiding the clinician or physical therapist’s decision, knowing that the end goal of rehabilitation is to wean off assistive devices to ensure eventual peak independence.

Conclusion

This study demonstrated that self-selected treadmill handrail use can increase the anteroposterior margin of stability of the paretic or affected leg of survivors of stroke. Also, handrail use significantly impacted the mediolateral stability margins relative to the paretic and nonparetic leg sides with a significantly smaller nonparetic leg mediolateral margin of stability with a self-selected handrail force, likely resulting from a more medial placement of the nonparetic foot with handrail use. These findings inform the use of handrails in treadmill walking for survivors of stroke, indicating that fall prevention in unstable stroke patients may be achieved through self-selected handrail use, although a wider contextual use of treadmill handrails should be considered in cases where the goal of treadmill walking is to train stability. In such cases, decreasing the individual’s stability margins may be necessary to challenge stability for balance training. Further research will give a more vivid structure on how handrail use could impact paretic leg use in treadmill rehabilitation.

Figure 2.

Violin plots of the anteroposterior (a) and mediolateral (b) margins of stability in meters (m) for the paretic and nonparetic sides with individual data points shown for each participant. Crossbar indicates sample means, and ‘*’ entries indicate a significant relationship between any of the 3 different trials of No handrail (NHR), Light handrail (5%HR), and Self-selected handrail support (SSHR).

Figure 3.

Violin plots of the Step width (a) and lateral foot placement (b) [in meters (m)] for the paretic and non-paretic sides showing individual data points for participants. Crossbars show sample means and ‘*’ entries indicate where a significant relationship exists between the 3 different trials of No handrail (NHR), Light handrail (5%HR), and Self-selected handrail support (SSHR).