Mobility function and recovery after stroke: preliminary insights from sympathetic nervous system activity

Abstract

Background:

Post-stroke hemiparesis increases the perceived challenge of walking, as indicated by feelings of stress, anxiety, and fear of falling. Perceived challenge is commonly measured by self-report which is susceptible to measurement bias, especially underreporting. There is a need for an objective measure of perceived challenge to overcome limitations of self-report and to advance post-stroke assessment of mobility function. A promising approach is to assess sympathetic nervous system (SNS) activity with skin conductance to detect the physiological stress response associated with challenging walking tasks. The primary purpose of this study was to investigate the feasibility of using skin conductance measurements to detect task-related differences in the challenge posed by complex walking tasks in adults post-stroke. This study also investigated whether rehabilitation of walking after stroke would lead to a reduction in SNS activity measured by skin conductance during walking.

Methods:

Adults post-stroke and healthy young adults performed various lab-based walking tasks including typical walking, walking in dim lighting, walking over obstacles, and dual-task walking (while performing a verbal fluency task). Skin conductance was measured from the palmar surface of each hand and spatiotemporal gait parameters were recorded by an electronic walkway. Continuous decomposition analysis was used to assess changes in skin conductance level (ΔSCL) and skin conductance response (ΔSCR). A subset of the post-stroke participants also underwent a 12-week rehabilitation intervention designed to enhance walking coordination and function.

Results:

SNS activity measured by skin conductance (both ΔSCL and ΔSCR) was significantly greater for the obstacles task and dual-task compared to typical walking in the stroke group (p<0.05). Consistent with this finding, participants also exhibited ‘cautious’ gait behaviors of slower speed, shorter step length, and wider step width during the challenging tasks (p<0.01). Following the rehabilitation intervention, SNS activity decreased significantly for the obstacles task (ΔSCL p=0.04; ΔSCR p=0.008) and dual-task (ΔSCR p=0.01). Additionally, the intervention yielded a reduction in step width for all walking tasks (p≤0.02), and an increase in self-reported balance confidence (p=0.04).

Conclusions:

SNS activity measured by skin conductance is a feasible approach for objectively quantifying task-related differences in the perceived challenge of various walking tasks in people post-stroke. Furthermore, reduced SNS activity during walking following a rehabilitation intervention suggests a beneficial reduction in the physiological stress response evoked by complex walking tasks.

Introduction

Stroke-related motor impairments restrict mobility including reduced community ambulation, independence, and participation in life-role activities.^1,2 An important factor that helps to explain mobility restrictions after stroke is the perceived challenge of walking, as measured by self-reported higher levels of state anxiety³, fear of falling⁴, low mobility self-efficacy⁵, and poor balance confidence.⁶ For instance, common measures of self-reported mobility-related balance confidence include the Activities-specific Balance Confidence (ABC) Scale⁷ and Falls Efficacy Scale.⁸ Both are multi-item questionnaires that assess the perceived level of confidence/concern regarding balance or falls during a variety of common ambulatory activities.

The perceived challenge of walking in adults post-stroke may be especially increased during complex tasks such as walking over irregular terrain or in distracting environments. ^2,9 Impaired performance on complex walking tasks and/or avoidance of environments requiring complex walking contributes significantly to mobility disability and restricted community ambulation.^9,10 Likewise, higher levels of balance confidence and falls self-efficacy are strongly associated with perceived physical function/recovery, less frequent avoidance of challenging environments, more frequent participation in walking-related activities and high satisfaction with community walking.^5,6,11–13

Assessment of the perceived challenge of walking by self-report is a valuable approach⁶, however, this form of assessment can be susceptible to subjective measurement bias and error.^14–16 Examples of common sources of potential bias are over-reporting of positive traits and under-reporting of negative traits (i.e., social desirability bias), choosing extreme scores on a self-report rating scale (i.e., response style effects), and providing responses that the individual thinks will be desirable to the researcher (i.e., demand characteristics).¹⁷ A physiologically-based measure of the perceived challenge of walking would be valuable to reduce measurement error due to bias and enhance objective assessment of this important aspect of mobility function and recovery.

One promising approach to gauge the perceived challenge of walking is to measure sympathetic nervous system (SNS) activity. Increased SNS activity is an autonomic stress response (the ‘fight or flight’ response) that contributes to the mobilization of physiological resources to optimize behavioral performance. The SNS is responsive to physical exertion and cognitive load¹⁸ including both current and anticipated demands.¹⁹ SNS activity can be readily and non-invasively measured by recording skin conductance from the palmar surface of the hands.^20,21 The validity and reliability of this approach has been discussed extensively in the 2012 committee report from the Society for Psychological Research Ad Hoc Committee on Electrodermal Measures.²² Several studies have shown that skin conductance is responsive to the increased challenge experienced during complex walking tasks in individuals without neurological deficits. Clark and colleagues²³ reported that SNS activity measured by skin conductance was increased in older adults when they performed obstacle crossing and dual-task walking (with a verbal fluency task) compared to typical walking. Adkin and colleagues²⁴ reported an increase in both skin conductance and self-reported state anxiety (measured by a modified version of The Sport Anxiety Scale) when healthy young adults were asked to rise to their toes while standing at the edge of an elevated platform. Similarly, Hadjistavropoulos and colleagues²⁵ reported increased skin conductance and self-reported state anxiety when older adults performed dual-tasking walking (while holding a tray) on an elevated platform, relative to typical walking. Collectively, these findings suggest that complex walking tasks elicit a physiological stress response from the SNS that can be quantified by skin conductance assessment. However, the feasibility and validity of using this approach for people with post-stroke neurological impairments has not been investigated. Furthermore, it is not known if SNS activity is responsive to rehabilitation-induced recovery of walking function after stroke. Therefore, the purpose of this study was to test the hypotheses that SNS activity measured by skin conductance during walking in adults post-stroke would be increased during lab-based assessment of complex walking tasks relative to typical walking (Hypothesis 1); and reduced in response to a post-stroke gait rehabilitation intervention (Hypothesis 2). We also measured SNS activity during the walking assessments in healthy young adults to provide a reference for how the healthy nervous system responds to the challenge of complex walking tasks. Furthermore, we tested an exploratory hypothesis that greater SNS activity during walking would be associated with worse spatiotemporal gait outcomes (Hypothesis 3). This would support that people who perceive walking to be more challenging also exhibit deterioration in gait outcomes, especially during the more complex tasks.

Methods

Equipment

Skin conductance signals were recorded using a commercially available portable data acquisition unit (Flexcomp Infiniti, Thought Technologies Ltd., QC, Canada). Adhesive electrodes (10mm Ag/AgCl recording surface) with conductive paste (0.5% saline in a neutral base) were placed on the palmar surface of the proximal phalanges of the index and ring fingers, bilaterally (see Figure 1). The palmar measurement site is densely innervated by sudomotor nerves that are highly sensitive to cognitive/emotional stimuli^20,26 including greater attention, and mobility-related fear of falling and/or anxiety.^23–25 Skin conductance signals were sampled at 32 Hz, acquired separately from each hand for each participant, and saved directly to an onboard flash memory card. These data were later downloaded to a computer for offline analysis.

Figure 1.

Equipment and set up for recording skin conductance.

The GAITRite^® system (GAITRite Gold, CIR Systems, PA, USA) was used to measure the spatiotemporal gait data during the walking tasks. GAITRite^® is a portable electronic walkway that is approximately 8.2 m in length. The walkway uses pressure-sensitive sensors to capture the timing and location of foot strikes at a sampling rate of 80 Hz.

Participants

All participants provided written informed consent at the time of enrollment and the study procedures were approved by the local Institutional Review Board. The inclusion criteria for the study included age >21 years; at least 6 months post-stroke; medically stable; able to follow 2-stage commands; 10 meter walking speed <0.8 m/s; Fugl-Meyer lower extremity score <30; and Mini-Mental State Exam Score (MMSE)²⁷ ≥21. All participants were evaluated by a licensed physical therapist and were visually confirmed to exhibit a hemiparetic gait pattern. The inclusion criteria also included the ability to walk independently with or without assistive devices and/or supportive braces. Use of assistive devices was minimized, but some participants did use an ankle/foot orthosis and/or cane when these were needed to safely complete the protocol. The total number of participants needing assistive devices is reported in Table 1. The participants recruited for this study were a sample of convenience obtained from an ongoing clinical trial.

Table 1.

Stroke group clinical characteristics

	Full Cohort (n=31)	Intervention Subset (n=9)
Age (in years)	59.2±10.1	63.0±10.9
Stroke chronicity (in months)	15.26±7.93	55.4±52.1
Sex (Female/Male)	12F/19M	4F/5M
Paretic side(Right/Left)	14R/17L	5R/4L
Used AFO and/or cane	8	1
		Pre	Post
ABC scale score (out of 100)	61.46±17.69	70.41±11.01	79.29±9.91
FMA (out of 34)	23.93±4.65	21.0±4.74	20.88±5.94

Abbreviations: Ankle Foot Orthosis (AFO), Activities-specific Balance Confidence (ABC) Scale, Fugl-Meyer Assessment (FMA). Errors are standard deviation.

Participants were excluded if they had any condition that would interfere with the ability to safely and appropriately participate in the gait intervention protocol, such as uncontrolled hypertension, lower extremity pain, severe obesity (body mass index >40), poor cardiopulmonary and/or renal function, severe perceptual or cognitive deficits or active drug/alcohol abuse, significant balance disturbances, lower motor neuron damage or radiculopathy, myocardial infarction or heart surgery in the prior year, bone fracture or joint replacement in the prior six months, or diagnosis of a terminal illness. Inclusion criteria for the young adults were age 18–30 years and self-reported absence of any medical conditions that affected walking ability.

Study assessments

Walking tasks

Participants were instructed to walk at preferred speed around a designated path while performing four separate walking tasks in random order: 1) typical walking (“typical” task), 2) walking in dim lighting (“dim” task), 3) dual-task walking with a verbal fluency task (“verbal” task), and 4) walking over obstacles (“obstacles” task). The distance walked for each lap was 18 meters, of which 8.2 m was performed over an electronic walkway that recorded spatiotemporal gait variables. Two or three laps were performed for each separate walking task, depending on the walking speed of the participant. Slower walkers performed only 2 laps in order to make the total task time equitable across participants and to prevent fatigue. The number of steps acquired from the electronic walkway was similar across walking tasks for each participant, and all recorded steps were included in the data analysis. For typical, the walking path was unobstructed and well-lit. For dim, the lights in the room were turned off, but some exterior light was allowed to come in through an open door. For verbal, the participants were asked to say as many words as possible that began with a randomly assigned letter while walking. The letters for the verbal task were randomly selected from this list: B, D, G, N, P, R, and T. A new letter was assigned at the beginning of each lap. For obstacles, the participants were asked to step over six small objects (shoes) evenly spaced within each lap. Two of the objects were placed on the electronic walkway at approximately one-third and two-thirds of the length of the walkway. This ensured that the entire pass over the walkway included periods of planning for, executing, and/or recovering from the obstacles negotiation task. Task order was randomized for each participant. Each walking task was separated by at least two minutes of rest and was preceded by a quiet standing period of at least 30 seconds to provide reference baseline data for skin conductance measurements.

Clinical assessments for the participants post-stroke

The Activities-specific Balance and Confidence (ABC) Scale was administered to assess the self-reported balance confidence pre-intervention and post-intervention. The lower extremity Fugl-Meyer assessment (FMA) was also conducted to characterize the participants’ motor impairment pre-intervention and post-intervention.

Gait rehabilitation intervention for the participants post-stroke

To test Hypothesis 2, a subset of individuals post-stroke participated in a 12-week rehabilitation intervention focusing on enhancing coordination (60 sessions, by a licensed physical therapist, 3 hours per day, five times per week). Also included were therapeutic exercises to improve balance and strength. The training paradigm included progression from simple to complex movement tasks (i.e., moving from in synergy to out of synergy lower extremity movement patterns) to facilitate motor learning of appropriate coordination patterns. Positions of sidelying, prone, supine, and seated were used to mitigate the effects of gravity and abnormal muscle tone, in order to facilitate the practice of the high-quality movement patterns. The standing position was used in order to practice coordination of movements in the upright position. The newly-acquired coordinated movements were then integrated into the practice of gait. Gait coordination training focused on ankle dorsiflexion, knee flexion, and hip flexion during the swing phase; knee flexion at toe-off and knee extension before heel strike; knee and hip extension during the stance phase; and whole body balance control during weight shifting. The motor learning principles used in the gait training protocol included movement practice as close to normal as possible, high number of repetitions, attention to the motor task, and rapid progression of task difficulty while maintaining the integrity of the task movements. The intervention protocol used in this study has been successfully implemented in prior studies.^28–30

Data analysis

Data analysis was conducted with Matlab version R2015a (The Mathworks, Natick MA) using Ledalab v3.4.7 and custom analysis programs. The raw skin conductance data were down-sampled to 8 Hz and visually examined for the presence of motion artifact, as indicated by high frequency fluctuations in the signal amplitude. Relatively few artifacts were identified, and the ones that were found were removed and replaced by linear interpolation. Continuous decomposition analysis was performed in Ledalab v3.4.7 to separate the tonic component (skin conductance level, or SCL) and the phasic component (skin conductance response, or SCR) of the signal. A minimum amplitude criterion of 0.04 μS was applied to achieve a balance between sensitive detection of SCRs and minimizing the effects of movement artifacts.^22,31

For each walking task, two values were extracted from the tonic skin conductance level (SCL) data. The first value was the minimum SCL during the baseline period of quiet standing that preceded each walking task. The minimum SCL was selected in order to capture the most relaxed state of the SNS. The second value was the mean SCL during the duration of the walking task. The primary outcome variable for SCL is the percent change in SCL (denoted by ΔSCL) between the resting and walking phases of each task using the following formula:

$$\Delta \text{SCL }=\left[\left(\text{Walking Average }–\text{ Resting Minimum}\right)/\text{ Resting Minimum}\right]*100$$

A similar approach was used for analysis of SCR. The rate of SCRs was calculated during the duration of the resting phase that preceded the walking task, as well as during the duration of the walking task. The rate of SCRs refers to the number of SCRs detected, divided by the duration of the recording period. The primary outcome for SCR is the change in the rate of SCR (denoted by ΔSCR) from the resting to the walking phase of each task using the following formula:

$$\Delta \text{SCR }=\left(\text{Rate of SCR during walking }–\text{ Rate of SCR during rest}\right)$$

Statistics

Statistical analysis was conducted using JMP software (JMP^® 11. Cary, NC: SAS Institute Inc.). Pearson’s correlation coefficient was used to examine the association between the skin conductance data acquired from the paretic and non-paretic hands (stroke group), and the left and right hands (young group). The association was examined for the data from each task, as well as the pooled data from all tasks. The false discovery rate procedure was applied to account for multiple comparisons. ³²

Hypothesis 1 tested that SNS activity measured by skin conductance during walking would be increased during lab-based tasks of complex walking relative to typical walking in adults post-stroke. To test Hypothesis 1, task-dependent differences in SNS activity were assessed with separate two-way repeated measures ANOVA models (Task × recording site) for each skin conductance variable (ΔSCL and ΔSCR) and for each group (stroke and young). Recording site refers to the side that the skin conductance signal was recorded from (i.e., paretic and non-paretic hand for the stroke group, and left and right hand for the young group). Additionally, task-dependent differences in spatiotemporal gait outcomes in the adults post-stroke were assessed by one-way repeated measures ANOVA for walking speed, and by two-way repeated measures ANOVA (Task × limb) for step width, step length and step length variability.

Hypothesis 2 tested that SNS activity measured by skin conductance during walking would be reduced in response to a post-stroke gait rehabilitation intervention. To test Hypothesis 2, separate three-way repeated measures ANOVA (Time × Task × recording site) was used to compare intervention-induced changes in SNS activity for each skin conductance variable (ΔSCL and ΔSCR). Post hoc analysis using Tukey’s HSD was used to further investigate significant main effects. The assumption of sphericity for the ANOVA models were tested by Mauchly’s test, and any violations were corrected using the Huynh-Feldt correction (if ϵ > 0.75) or the Greenhouse-Geisser correction (if ϵ < 0.75). Intervention-induced changes in the spatiotemporal gait parameters measured pre-intervention and post-intervention were statistically compared by paired t-tests for each walking task.

Hypothesis 3 tested associations between skin conductance and spatiotemporal parameters of gait. The purpose was to conduct a preliminary assessment of whether greater task-related SNS activity is associated with deterioration of gait performance. Each participant’s ΔSCL value from typical walking was subtracted from the ΔSCL value from each complex walking task. An analogous procedure was used to calculate task-related differences for ΔSCR and for gait parameters (speed, step width, step length, and step length variability). Pearson’s correlation coefficient was used to test the association between the task-related changes in each skin conductance measure (i.e., ΔSCL and ΔSCR) and gait parameters.

Results

Thirty-one adults with chronic post-stroke hemiparesis participated in the cross-sectional walking assessments to test Hypotheses 1 and 3. A subset of nine participants underwent the intervention protocol for testing Hypothesis 2. Demographic and clinical information for all stroke participants are shown in Table 1. A group of eight healthy young adults (age = 22.4 ± 3.7 years) was also tested on each walking assessment to provide reference data on task-related changes in SNS activity for a healthy unimpaired nervous system.

Pearson’s correlation coefficient was used to assess the consistency of skin conductance measurements from each recording site for the stroke and young groups (i.e., paretic versus non-paretic ΔSCL, and paretic versus non-paretic ΔSCR; left versus right ΔSCL, and left versus right ΔSCR). For both groups, the recording sites were significantly associated for most walking tasks (p<0.05), as reported in Table 2.

Table 2.

Stroke and young group bilateral skin conductance associations

Walking Task	Stroke Group		Young Group
	ΔSCL	ΔSCR	ΔSCL	ΔSCR
Typical	r=0.65***	r=0.56**	r=0.96***	r=0.79*
Dim	r=0.61***	r=0.28	r=0.97***	r=0.67
Verbal	r=0.66***	r=0.51**	r=0.97***	r=0.53
Obstacles	r=0.48*	r=0.62**	r=0.96***	r=0.97***
Pooled tasks*	r=0.73***	r=0.50***	r=0.96***	r=0.70***

^***p<0.001

^**p<0.01

^*p<0.05

Pooled tasks* - ΔSCL and ΔSCR were pooled across all walking tasks within each group.

Significance levels corrected for multiple comparisons with False Discovery Rate Procedure.

SNS activity measured by skin conductance for each walking task

Within the stroke group, there was a significant main effect of Task on ΔSCL (p<0.0001, Figure 2A). Post hoc analysis revealed that ΔSCL was significantly greater for obstacles (12.4%) compared to verbal (9.0%, p=0.04), dim (5.4%, p=0.0005), and typical (4.3%, p<0.0001). ΔSCL for verbal was significantly greater than typical (p=0.02). The main effect of recording site was not significant (p=0.82).

Figure 2. Change in skin conductance level (ΔSCL) and skin conductance response (ΔSCR) for the stroke and young groups.

Panels A, B, C and D show the change in skin conductance (i.e., ΔSCL and ΔSCR) from the resting to the walking period of each task. The stroke group is shown in black and the young group is shown in light grey. The error bars denote the standard error.

Within the young group, there was a significant main effect of Task on ΔSCL (p=0.009, Figure 2B). Post hoc analysis revealed that ΔSCL was significantly greater for obstacles (21.0%) compared to verbal (13.1%, p=0.03), dim (3.8 %, p=0.01), and typical (3.4 %, p=0.04). ΔSCL for verbal was significantly greater than dim (p=0.04). A trend towards greater ΔSCL was observed for verbal compared to typical (p=0.07). The main effect of recording site was not significant (p=0.97). Resting and walking SCL for each group and for each walking task has been shown in Table 3.1.

Table 3.

Stroke and young group skin conductance (SCL and SCR) at rest and during walking tasks

Table 3.1
Walking Task	Stroke Group		Young Group
	Resting SCL	Walking SCL	Resting SCL	Walking SCL
Obstacles	5.95±4.71	6.73±4.18	8.72±2.75	10.19±2.58
Verbal	5.98±4.59	6.46±4.39	9.53±2.43	10.60±2.22
Dim	6.13±4.76	6.44±4.53	8.55±2.57	8.82±2.60
Typical	5.99±4.72	6.29±4.35	8.21±2.24	8.51±2.60

Table 3.2
Walking Task	Stroke Group		Young Group
	Resting SCR	Walking SCR	Resting SCR	Walking SCR
Obstacles	0.18±0.15	0.30±0.19	0.21±0.11	0.46±0.29
Verbal	0.14±0.09	0.27±0.18	0.29±0.17	0.46±0.14
Dim	0.17±0.14	0.24±0.21	0.18±0.12	0.29±0.17
Typical	0.20±0.20	0.24±0.21	0.15±0.07	0.29±0.13

Values are group mean ± standard deviation.

Within the stroke group, there was a significant main effect of Task on ΔSCR (p=0.002, Figure 2C). Post hoc analysis revealed that ΔSCR was significantly greater for obstacles (0.13 responses/sec, p=0.007) and verbal (0.13 responses/sec, p=0.003) compared to typical (0.04 responses/sec). ΔSCR for verbal was significantly greater than dim (0.06 responses/sec, p=0.02). A trend towards greater ΔSCR was observed for obstacles compared to dim (p=0.09). The main effect of recording site was not significant (p=0.79). Within the young group, the main effect of Task on ΔSCR was not significant (p=0.39, Figure 2D). Resting and walking SCR for each group and for each walking task has been shown in Table 3.2.

Task-related differences in spatiotemporal measurements of gait

Gait data for stroke participants are shown in Table 4. There was a significant main effect of Task on gait speed (p<0.001), step length (p<0.0001), step length variability (p<0.0001), and step width (p=0.008). Post hoc analyses were conducted for each gait variable. Gait speed was significantly slower for obstacles compared to verbal, dim, and typical (p<0.001). Gait speed was significantly slower for verbal compared to dim (p=0.0003) and typical (p=0.0001). Step length was significantly shorter for obstacles compared to verbal, dim, and typical (p<0.001). Step length for verbal was significantly shorter than dim (p<0.0002) and typical (p<0.0001). Step length variability was significantly greater for obstacles compared to verbal, dim, and typical (p<0.01). Step width was significantly greater for verbal (p=0.02) and dim (p=0.016) compared to typical. The main effect of Limb was not significant for any comparisons (p>0.05).

Table 4.

Gait measures in post-stroke participants

	Full Cohort (n=31)		Intervention Subset (n=9)
			Pre		Post
Walking speed (cm/sec)
Typical	53.25±20.40		53.72±20.63		54.73±17.01
Dim	52.39±18.87		52.83±18.95		54.22±17.33
Verbal	47.29±18.05		49.97±21.32		45.62±17.99
Obstacles	39.86±18.12		41.30±20.93		46.37±10.49
Step width (cm)
Typical	19.38±4.71		20.28±4.98		17.20±4.57
Dim	20.22±4.71		21.32±4.62		17.26±4.48
Verbal	20.0±4.53		20.56±4.46		18.59±4.90
Obstacles	19.62±4.68		19.23±4.11		15.58±3.10
Step length (cm)	Paretic	Non-paretic	Paretic	Non-paretic	Paretic	Non-paretic
Typical	43.29±10.44	39.99±10.90	39.39±11.50	39.88±9.70	42.38±7.25	41.30±8.29
Dim	42.78±12.82	39.44±12.37	41.68±18.20	41.10±15.41	42.89±6.04	39.47±8.95
Verbal	40.10±11.30	35.42±11.66	37.49±13.53	36.56±13.55	35.23±7.31	32.65±11.28
Obstacles	38.22±19.40	31.54±12.60	29.23±16.49	38.06±13.06	44.06±9.57	35.06±11.03
Step length variability	Paretic	Non-paretic	Paretic	Non-paretic	Paretic	Non-paretic
Typical	2.66±1.36	3.04±1.56	2.97±1.91	3.45±1.45	3.27±2.32	2.68±0.47
Dim	3.63±5.80	4.47±7.08	6.45±10.25	7.79±12.59	2.78±1.32	3.29±0.86
Verbal	2.96±2.94	2.89±1.26	3.75±5.09	2.52±0.75	2.38±1.03	2.92±0.71
Obstacles	12.21±12.37	10.74±13.77	11.28±8.42	16.19±23.99	10.97±3.28	6.80±3.55

Values are group mean ± standard deviation.

Skin conductance before and after the gait rehabilitation intervention

For ΔSCL, there was a significant main effect of Time (p=0.02) and Task (p=0.03), such that ΔSCL was significantly lower post-intervention compared to pre-intervention during walking assessments (Figure 3A). Post hoc analysis revealed that ΔSCL was significantly lower for obstacles at post-intervention compared to pre-intervention (p=0.04). ΔSCL did not change significantly for verbal (p=0.35), dim (p=0.23) and typical (p=0.23). The main effect of recording site was not significant (p=0.21). The interaction effect of Time × Task was not significant (p=0.33).

Figure 3. Change in skin conductance level (ΔSCL) and skin conductance response (ΔSCR) for the stroke group pre-intervention and post-intervention.

Panels A and B show the change in skin conductance (i.e., ΔSCL and ΔSCR, respectively) from the resting to the walking period of each task pre-intervention and post-intervention. Skin conductance at pre-intervention is shown in light grey and post-intervention is shown in black. The error bars denote the standard error.

For ΔSCR, there was a significant main effect of Time (p=0.02), such that ΔSCR was significantly lower post-intervention compared to pre-intervention during walking assessments (Figure 3B). Post hoc analysis revealed that ΔSCR was significantly lower for obstacles (p=0.008) and verbal (p=0.01) at post-intervention compared to pre-intervention. ΔSCR did not change significantly for dim (p=0.59) and typical (p=0.29). The main effect of recording site was not significant (p= 0.44). The interaction effect of Time × Task was not significant (p=0.47).

Gait and clinical outcomes before and after the rehabilitation intervention

The pre-intervention and post-intervention spatiotemporal gait data are reported in Table 4. Statistical comparisons of gait parameters measured pre-intervention and post-intervention revealed that step width decreased significantly post-intervention for obstacles (p=0.004), verbal (p=0.03), dim (p=0.02), and typical (p=0.006) tasks. However, gait speed, step length, and step length variability did not change significantly post-intervention (p>0.05 for all measures).

Self-reported balance confidence measured by the Activities-specific Balance Confidence (ABC) Scale improved significantly following the post-stroke gait rehabilitation intervention (70.41% ± 11.01 pre-intervention versus 79.29% ± 9.91 post-intervention, p=0.04). The Fugl-Meyer lower extremity score did not change significantly (21.0 ± 4.74 pre-intervention versus 20.88 ± 5.94 post-intervention).

Association between SNS activity and spatiotemporal measurements of gait

The test of association between skin conductance measurements of SNS activity (ΔSCL and ΔSCR) and gait measures of speed, step width, step length and step length variability did not yield statistically significant findings. However, several trends were observed. For the dim task, greater ΔSCL showed a trend towards an association with three out of the four measured gait metrics including slower gait speed (p=0.13), shorter step length (p=0.08), and higher step length variability (p=0.09). Likewise, for the obstacles task, greater ΔSCL showed a trend towards an association with slower gait speed (p=0.05) and shorter step length (p=0.08).

Discussion

The results of this study support the feasibility of measuring sympathetic nervous system activity with skin conductance to gauge the perceived challenge of walking tasks in people post-stroke. First, SNS activity was found to be acutely increased during complex walking tasks relative to typical walking. The increase in SNS activity is consistent with a heightened physiological stress response due to the greater perceived challenge experienced during the performance of complex walking tasks. Second, SNS activity during complex walking tasks was attenuated following a gait rehabilitation intervention. This finding suggests that the tasks were perceived as less challenging after the intervention. The study also evaluated potential associations between SNS activity and walking performance outcomes, but the findings were generally weak and not statistically significant.

Skin conductance components SCL and SCR are well-established measurements of SNS activity.²² SCL represents slow changes in skin conductivity over several seconds. SCR represents fast changes in skin conductance^22,33 that may be more closely time locked to underlying activity of the sudomotor nerves.³⁴ Representative data depicting SCL and SCR during the walking tasks are shown in Figures Figures 4A and 4B, respectively. An interesting finding is that skin conductance acquired from the paretic and non-paretic hands were strongly correlated. This finding supports that the present results are driven by central SNS activity, and are relatively robust to potential deterioration of measurement conditions after stroke, such as due to spastic clenched fist or altered skin health.

Figure 4. Example raw skin conductance data during the walking tasks.

Panels A and B show example skin conductance level (SCL) and skin conductance response (SCR) recorded during the walking tasks. Each black dot on the raw data plot indicates a point in time at which a SCR was detected by the analysis algorithm.

The first study hypothesis was that SNS activity would be increased during lab-based assessment of complex walking tasks relative to typical walking in adults post-stroke. This hypothesis was confirmed, as both ΔSCL and ΔSCR increased during lab-based complex walking tasks compared to the typical task (i.e., obstacles and verbal >typical). The young group also demonstrated increased ΔSCL during complex walking compared to the typical task, including a significant increase for the obstacles and a trend for verbal. Cumulatively, these results demonstrate the responsiveness of skin conductance to both physical and cognitive sources of physiological stress, consistent with prior studies.^23,25,35

The rationale for including a healthy young group was to provide additional context for interpreting the stroke group data by also observing how an unimpaired nervous system responds to the same walking tasks. We did not make direct statistical comparisons between the young and stroke groups because we did not hypothesize that there would be a group difference. Rather, our primary objective was to assess task-dependent changes in SNS activity within each group. In general, visual observation of data indicates that the SNS of stroke-injured and healthy adults behaves similarly in response to complex walking tasks. A notable difference was that the stroke group exhibited a significant increase of both ΔSCL and ΔSCR for complex walking tasks, while in the young group only ΔSCL was significantly increased. The reason for a lack of increase in ΔSCR for the young group is unclear because both ΔSCL and ΔSCR are measures of SNS activity. It may be that SCRs are more indicative of specific anxiety-provoking events. For example, a moment of unsteadiness might trigger an acute SNS response (i.e., SCR) that would be expected to occur more often for individuals with hemiparetic gait. ³⁶

The second study hypothesis was that SNS activity would be reduced in response to a gait rehabilitation intervention. This hypothesis was supported by the findings that ΔSCL and ΔSCR were reduced significantly in response to the gait rehabilitation intervention. In agreement with reduced SNS activity, step width was reduced during the walking assessments post-intervention. A narrower step width suggests a less cautious gait pattern and is consistent with improved balance confidence and less fear of falling.³⁷ Indeed, there was an increase in the self-reported balance confidence after the intervention, measured by the ABC Scale. A future randomized controlled trial will be needed to confirm and expand upon these preliminary findings of rehabilitation-induced changes in the perceived challenge of walking after stroke.

Study considerations and limitations

A consideration for this study is that we were not able to control for walking speed across participants due to the overground nature of the walking tasks. This is beneficial in terms of allowing participants to engage naturally in each task, but might also allow them to reduce the challenge level of some tasks by walking more slowly. This could lead to underestimation of SNS activity for a given task. The effect of walking speed on SNS activity can be addressed in future studies such as by using a treadmill to control speed. If anything, we expect that standardizing speed across tasks would further strengthen task-related differences in SNS activity.

The present study investigated if a gait intervention designed to enhance walking coordination and function would also decrease the perceived challenge of walking. The post-intervention findings are based on a secondary analysis performed with a convenience sample from an ongoing clinical trial of neurorehabilitation of walking. Therefore, a limitation of the study design is that the intervention was not specifically designed to reduce the perceived challenge of walking after stroke. Although the findings of this study support robust reductions in SNS activity especially for the most difficult walking tasks, we cannot be certain about the aspects of the intervention that contributed to the decrease in SNS activity. Furthermore, the lack of a non-intervention control group and the participants’ exposure to the walking assessment tasks pre-intervention and post-intervention limits the ability to attribute the reduction in SNS activity solely to the intervention. Future studies would be needed to address the limitations of this design.

This study did not include a concurrent second measure of self-reported perceived challenge upon completion of the experimental walking task. A concurrent measure of self-report could provide additional context by which to interpret task-related changes in skin conductance. Future study designs should include a second measure of self-reported challenge, and also compare the self-report and skin conductance measurement approaches. Future studies should also more fully investigate the potential relationship between SNS activity and walking performance outcomes.

Finally, this study presents promising findings that support the feasibility of measuring skin conductance to assess the perceived challenge of walking. However, it will be important for future studies to establish the test-retest reliability of the skin conductance measure for people post-stroke and for the specific walking tasks performed here (and/or for other walking tasks that could be included as part of future clinical trials).