Soleus H-reflex modulation in cerebral palsy and its relationship with neural control complexity: a pilot study

Abstract

Individuals with cerebral palsy (CP) display motor control patterns that suggest decreased supraspinal input, but it remains unknown if they are able to modulate lower-limb reflexes in response to more complex tasks, or whether global motor control patterns relate to reflex modulation capacity in this population. Eight ambulatory individuals with CP (12 – 18 years old) were recruited to complete a task complexity protocol, where soleus H-reflex excitability was compared between bilateral (baseline) and unilateral (complex) standing. We also investigated the relationship between each participant’s ability to modulate soleus H-reflex excitability and the complexity of their walking neural control pattern determined from muscle synergy analysis. Finally, six of the eight participants completed an exoskeleton walking protocol, where soleus H-reflexes were collected during the stance phase of walking with and without stance phase plantar flexor resistance. Participants displayed a significant reduction in soleus H-reflex excitability (-26 ± 25%, p = 0.04) with unilateral standing, and a strong positive relationship was observed between more refined neural control during walking and an increased ability to modulate reflex excitability (R = 0.79, p = 0.04). There was no difference in neuromuscular outcome measures with and without the ankle exoskeleton (p-values all > 0.05), with variable reflex responses to walking with ankle exoskeleton resistance. These findings provide evidence that ambulatory individuals with CP retain some capacity to modulate lower-limb reflexes in response to increased task complexity, and that less refined neural control during walking appears to be related to deficits in reflex modulation.

Introduction

To control task specific movement, humans rely on a dynamic interplay between spinal reflexes and supraspinal reflex modulation driven by sensorimotor input (Rossignol et al. 2006; Dominici et al. 2011). These spinal (i.e., short-latency) reflexes and their modulation have been previously studied using the Hoffman (H-) reflex, which is an electrical analog of the stretch reflex (Palmieri et al. 2004). It has been observed in neurologically intact individuals that reflex excitability, measured as the change in H-reflex amplitude relative to change in background motor activity (Edamura et al. 1991; Funase et al. 1994; Palmieri et al. 2004), will decrease during postures or movements requiring greater supraspinal drive and stability at the ankle (i.e., higher co-contraction); for example, when going from bilateral to unilateral standing (Earles et al. 2000; Huang et al. 2009; Pinar et al. 2010), or when walking on a balance beam compared to walking on a treadmill (Llewellyn et al. 1990). Given the presence of soleus reflex modulation during challenging motor tasks and the significant involvement of central control mechanisms, it would be beneficial to understand how this modulation is affected by neuromuscular disorders affecting the brain, such as cerebral palsy (CP) (Graham et al. 2016). This is of particular interest because there is a growing body of evidence that motor control strategies in CP are marked by decreased or abnormal central drive (Petersen et al. 2013; Willerslev-Olsen et al. 2015; Frisk et al. 2019; Cappellini et al. 2020), likely because of damage and alterations to descending corticospinal tracts (Brouwer and Ashby 1991) and sensorimotor cortices (Kurz et al. 2014a, b). This has been observed in walking, when individuals with CP have decreased neural control complexity of their lower limb muscles, measured as a higher variance in muscle activity explained by one muscle synergy (tVAF₁) (Steele et al. 2015). As supraspinal input is necessary for reflex modulation during challenging motor tasks, an individual with CP may display altered modulation capacity, but this has not yet been evaluated in this population.

Previous studies that have explored soleus H-reflex modulation in individuals with CP versus those without CP have demonstrated deficits in reciprocal inhibition, marked by large soleus H-reflex responses during antagonist activation (Leonard et al. 1990) and atypically high reflex amplitudes during the stance phase of gait (Hodapp et al. 2007). It was also found that treadmill gait training significantly reduced heightened stance phase soleus H-reflex for a cohort of children with CP, which was accompanied by improved measures of mobility such as free walking speed (Hodapp et al. 2009). It was theorized that this was due to improved presynaptic inhibition of the soleus H-reflex from increased corticospinal drive to the lower limb muscles (Hodapp et al. 2009), but this was never fully explored. While neural control complexity does appear to be modifiable in CP with gait training interventions, as was demonstrated in a pilot clinical trial with a resistive ankle exoskeleton (Conner et al. 2021), no study to date has investigated if increased supraspinal drive during walking is related to reflex modulation patterns in CP. A more complete understanding of the relationship between neural control complexity and reflex modulation in CP could provide further information on how this disorder affects neuromuscular control and the mechanisms behind observed improvements with emerging interventions.

This study had three goals. First, we evaluated the effect of task complexity on soleus H-reflex modulation in children and young adults with CP by measuring the change in reflex excitability between unilateral and bilateral standing. As individuals with CP demonstrate atypical supraspinal drive (Huang et al. 2009), we hypothesized that that there would be no change in reflex excitability between standing conditions. Next, we investigated the association between each individual’s capacity for reflex modulation and the complexity of their neural control pattern during walking. We hypothesized a significant positive relationship between deficits in neural control complexity during walking and a reduced ability to modulate soleus H-reflex excitability due to the influence of supraspinal input on both of these measures. Finally, to follow up on a previous observation of increased neural control complexity in CP after training with a resistive ankle exoskeleton device (Conner et al. 2021), we explored how walking with the device influences reflex modulation. We hypothesized that individuals with CP would display a decrease in soleus H-reflex excitability when walking with ankle resistance due to the increased supraspinal drive required for this motor task. In testing these hypotheses, we aimed to further elucidate the multifaceted effects of CP on inhibitory and faciliatory supraspinal pathways and global measures of motor control, providing novel information for improving targeted neuromuscular interventions in this patient population.

Materials and methods

This study was approved by the Northern Arizona University Institutional Review Board (#986744), and utilized participants recruited for a clinical trial, which was prospectively registered at ClinicalTrials.gov (NCT04119063). The protocol was completed at the Northern Arizona University – Phoenix Biomedical Campus (Phoenix, AZ), and informed written consent was provided by a parent or legal guardian for each participant.

Participants

The following inclusion criteria was used for recruitment of participants: confirmed diagnosis of CP, age 10 to 18 years old, Gross Motor Function Classification System (GMFCS) levels I – III, the ability to stand on one leg with support, and the capacity to walk for at least ten minutes on a treadmill with support were recruited for this study. Individuals who had undergone a selective dorsal rhizotomy procedure were excluded from participation if a measurable H-reflex was not elicitable due to the effect of this intervention on reflex modulation (Logigian et al. 1994). Given the scheduling demands of a pediatric participant, and the inherent discomfort of electric reflex testing, recruited participants had the option of completing one or both of the following protocols, which were held on separate days: 1) the primary, standing task complexity protocol, and 2) the supplementary, resistive ankle exoskeleton protocol. Eight total participants (7M/1F) were recruited for the primary standing protocol, and six participants completed the walking ankle resistance protocol. Prior to starting either protocol, participants were evaluated by a licensed physical therapist, who collected anthropometrics and measured spasticity level at the ankle using a Modified Ashworth Scale (Table 1).

Table 1.

Participant characteristics and respective protocols completed

	Age (y)	Gender	GMFCS levela	CP typeb	Modified Ashworth Scalec		Protocol(s)
					Ankle Plantar Flexion	Ankle Dorsiflexion
P1	14	F	1	SH	1+	0	Task complexity
P2	16	M	2	SH	2	0
P3	15	M	3	SD	1+	0	Task complexity & Walking ankle resistance
P4	18	M	2	SD	2	4
P5	15	M	2	SD	2	0
P6	13	M	2	SD	3	0
P7	16	M	1	SH	0	0
P8	12	M	I	SH	2	0

^aGMFCS: Gross Motor Function Classification System

^bCP type: spastic hemiplegic (SH), spastic diplegic (SD)

^cModified Ashworth Scale: used to measure spasticity by licensed physical therapist

H-reflex assessment

Wireless surface electromyography (EMG) sensors were placed on the non-dominant leg of each participant for measuring reflex responses and muscle activity (BIOPAC Systems, Goleta, CA, USA; 2000 Hz). Bipolar electrodes (Ag-AgCl) were placed on the soleus and tibialis anterior muscles according to SENIAM recommendations (Hermens et al. 1999).

To evoke a soleus H-reflex, stimulating electrodes (Ag-AgCl) were placed in the popliteal fossa of the participant’s non-dominant leg for stimulation of the posterior tibial nerve, with the cathode electrode just superior to the popliteal fossa and the anode electrode just inferior to the popliteal fossa (Palmieri et al. 2004). Stimulations were 1 ms, square-wave pulses ranging from 0.1 – 50 mA. Standing, H-reflex amplitudes were evoked at a low level of stimulation, and the position of the stimulating electrodes were adjusted to find an ideal stimulating location, as indicated by a large H-reflex amplitude without the presence of an M-wave. H-reflexes were then evoked at progressively higher stimulation magnitudes until the magnitude required to elicit H_max was reached. This setup was completed prior to H-reflex testing for both protocols. Full soleus H-reflex recruitment curves, including maximal muscle responses (M_max), were not measured for these pediatric participants due to the associated discomfort of electrically stimulating a muscle for maximal contraction. Instead, this study utilized within-subjects, paired sample comparisons that were normalized to baseline condition values (i.e., bilateral standing values for the standing protocol, and unresisted walking values for the walking protocol) so that normalization to M_max was not required. In other words, M_max would not change between conditions. Additionally, by determining the magnitude required to elicit H_max, a stimulation magnitude lower than this could be used to ensure that evoked H-reflexes were on the ascending limb of the H-reflex recruitment curve (Hodapp et al. 2007).

Task complexity protocol

Soleus H-reflex responses were recorded for participants under two conditions: 1) bilateral standing (i.e., baseline condition), and 2) unilateral standing on the non-dominant leg (i.e., complex condition (Huang et al. 2009; Pinar et al. 2010)) (Fig. 1A). Given the variation in functional level associated with CP, and the possibility that not all participants could stand on one leg unsupported, each participant was given the option to support themselves during the complex condition by placing one hand on a stabilizing surface that was approximately waist height. If the participant chose to use support for the complex condition, they were instructed to also use support for the baseline condition.

Figure 1. Study design.

(A) Task complexity protocol, where soleus H-reflexes were elicited with stimulation of the posterior tibial nerve under two conditions: a baseline, bilateral standing condition and a complex, unilateral standing condition (B) Walking ankle resistance protocol, where soleus H-reflexes were elicited during mid-stance under a baseline walking and when walking with an ankle exoskeleton device delivering resistance to plantar flexion proportional to a user’s real-time estimated ankle moment.

Starting with the baseline condition and a stimulation level of 0.1 mA, the stimulation magnitude was increased until a measurable M-wave was present, but the magnitude was below that needed to elicit H_max. The size of this M-wave was held constant (within ± 10%) across conditions to ensure a consistent, effective stimulus strength. For both conditions, 15 reflexes were evoked, and background EMG (bEMG) activity of the soleus and tibialis anterior was measured 50 ms before each respective stimulation (Schneider et al. 2000).

Baseline neural control complexity was assessed by collecting lower limb muscle activity while participants walked barefoot overground at a self-selected speed. Surface EMG electrodes (Noraxon, Scottsdale, AZ, USA; 1000 Hz) were placed on the soleus, tibialis anterior, vastus lateralis, and semitendinosus muscles of the non-dominant limb according to SENIAM recommendations (Hermens et al. 1999), and six gait cycles on the non-dominant side were recorded. Walking EMG data is not available for P1, who did not complete the overground walking trials after opting out of further testing. Walking EMG data for P2 was collected on a treadmill at the participant’s self-selected speed.

Walking ankle resistance protocol

Participants were outfitted with a lightweight untethered lower limb robotic device, the details of which can be found in (Lerner et al. 2018; Conner et al. 2020a) (Fig. 1B). Briefly, this device consisted of an actuation & control assembly worn at the waist, and ankle assemblies worn bilaterally on the legs. Motors actuated the carbon fiber ankle assemblies via Bowden cables to generate ankle torque in the sagittal plane. Using custom force sensors placed under the forefoot, the ankle moment was estimated and a proportional level of resistance to plantar flexion could be applied in real-time during each stance phase (Bishe et al. 2021). The device was controlled via a custom MATLAB graphical user interface (v2019b, Natick, MA, USA), and weighed approximately 1.75 kg.

Prior to the H-reflex testing, participants completed five “acclimation” visits across two weeks to learn how to properly engage with the resistive exoskeleton control. During these acclimation visits, participants walked with adaptive plantar flexion resistance and real-time soleus biofeedback on the non-dominant limb for at least 20 minutes, totaling approximately 120 minutes over the five visits. Resistance level was progressively increased from 0.1 – 0.2 Nm/kg across visits, while treadmill speed was held constant. Biofeedback was not used during H-reflex testing to avoid the confounding effect of a dual-task on reflex modulation (Weaver et al. 2012). A sixth visit was then completed for walking H-reflex testing.

On the sixth visit, participants began with H-reflex testing of the baseline condition, followed by approximately 20 – 30 minutes of resistance re-acclimation, and ending with H-reflex testing of the resisted condition (Fig. 1B). For both conditions, stimulations were delivered during midstance on the non-dominant side, identified by force sensors embedded in the footplates of the device. Starting with the baseline condition, stimulation level was increased until a measurable M-wave was present, but the level was below that needed to elicit H_max. H-reflexes were then evoked and recorded if the associated M-wave amplitudes were within ± 10% of this initial M-wave to ensure a consistent, effective stimulus strength (Fig. 1B). The first twelve reflexes that met this criteria were recorded for each condition (Hodapp et al. 2009), in addition to bEMG activity of the soleus and tibialis anterior 50 ms prior to each stimulation (Schneider et al. 2000).

Three-dimensional kinematics of the non-dominant ankle joint and lower limb muscle activation levels were collected while participants walked under baseline (i.e., un-resisted) and resisted conditions. Participants were outfitted with reflective markers according to Vicon’s lower body Plug-In Gait model (Vicon, Denver, CO, USA; 100 Hz) and with surface EMG electrodes (Noraxon, Scottsdale, AZ, USA; 1000 Hz) on the soleus, tibialis anterior, vastus lateralis, and semitendinosus muscles of the non-dominant limb according to SENIAM recommendations (Hermens et al. 1999). Ten gait cycle were recorded for each condition at matched speeds during non-stimulation steps.

Data processing

H-reflex amplitudes were measured as the peak-to-peak magnitude of the raw, non-rectified EMG signal (unadjusted H-reflex) where the H-reflex occurred (typically 40 – 50 ms post-stimulation) (Palmieri et al. 2004). Background EMG signals were band-passed filtered (4^th order Butterworth, 20 – 400 Hz) and rectified, and the mean activity 50 ms before stimulation was calculated. To account for any difference in background muscle activity between conditions and the known facilitation effect that increased voluntary activation of the soleus has on H-reflex amplitude (Capaday 1997; Stein et al. 2007) , we also calculated an adjusted H-reflex measure, where H-reflex amplitudes were normalized by dividing by their respective bEMG mean activity (Kao et al. 2010).

Co-contraction at the ankle was calculated to determine the stabilization effort by this joint for each condition. Specifically, activation levels of each muscle were normalized to the mean activity of that muscle during the baseline condition of each protocol. Next, a co-contraction index (CCI) was calculated over the 50 ms prior to stimulation using the following equation (1) (Knarr et al. 2012):

$$\text{CCI}={\sum }_{i=1}^{101}\frac{\mathit{\text{LEMG}}\left(i\right)}{\mathit{\text{MEMG}}\left(i\right)}\left(\mathit{\text{LEMG}}\left(i\right)+\mathit{\text{MEMG}}\left(i\right)\right)$$ (1)

where i represents the individual time points of stance phase (0 – 100%, or 101 total data points), LEMG represents the normalized magnitude of the less active muscle at time point i, and MEMG represents the normalized magnitude of the more active muscle at time point i. Mean H-reflex amplitude, soleus bEMG, adjusted H-reflex amplitude, and ankle co-contraction level was calculated for each participant and condition.

Inverse kinematics analysis was performed using the collected marker data to calculate ankle joint angles of the non-dominant limb during the baseline and resisted conditions. Joint angle data were low-pass filtered (4^th order Butterworth, 6 Hz low-pass cutoff), averaged across gait cycles, and time-normalized to the gait cycle (0 – 100%). Average ankle angle at midstance, when H-reflexes would have been evoked, was then calculated for reach participant across walking conditions.

For the barefoot overground walking trials and treadmill walking trials (baseline and resisted), EMG signals of the lower limb muscles were de-meaned, bandpass filtered (4^th order Butterworth, 40 – 400 Hz), rectified, and low pass filtered (4^th order Butterworth, 10 Hz). The filtered signals were then down-sampled to align with marker data, and time-normalized to the gait cycle (0 – 100%). Activation levels were then normalized to the average maximum activation of each respective muscle within a gait cycle, with the resisted trials using the baseline treadmill values. The variance in muscle activity accounted for by one muscle synergy (tVAF₁) was then calculated using non-negative matrix factorization as a measure of neural control complexity (Steele et al. 2015) for three conditions: 1) barefoot overground walking, 2) unresisted treadmill walking, and 3) resisted treadmill walking. Specifically, the following equation (2) was used (Lee and Seung 1999; Steele et al. 2015):

$$tVA{F}_1=1-\frac{\Vert EMG-W^{\text{*}}C\Vert }{{\Vert \mathit{\text{EMG}}\Vert }^2}$$ (2)

where EMG is a matrix containing the normalized EMG data recorded for each muscle across the ten gait cycles; W is a 4 × 1 matrix that represents the relative contribution of individual muscles in a synergy; and C represents the activation level of a synergy over the gait cycle. A lower tVAF₁ value is indicative of increased neural complexity (Steele et al. 2015).

Statistical analysis

Our primary outcome measure for the task complexity protocol was adjusted soleus H-reflex amplitude. Our secondary outcome measures included unadjusted soleus H-reflex amplitude, soleus bEMG and ankle co-contraction. For the walking ankle resistance protocol, our primary outcome measure was also adjusted soleus H-reflex amplitude, and our secondary outcome measures included unadjusted soleus H-reflex amplitude, soleus bEMG, ankle co-contraction, ankle angle at midstance and tVAF₁.

To account for the small sample sizes used in this study, we opted for non-parametric analyses (Ghasemi and Zahediasl 2012). For comparing our outcome measures between conditions for both protocols, we used Wilcoxon signed-rank tests. We also assessed the relationship between neural control complexity (tVAF₁) during barefoot overground walking and reflex modulation with unilateral standing using a Spearman’s Rank correlation coefficient, whereby 0.3 was considered a weak relationship, 0.5 a moderate relationship, and 0.7 a strong relationship (Mukaka 2012). Significance level was set at α = 0.05.

Results

Task complexity protocol

There was a significant decrease in adjusted H-reflex amplitude for unilateral standing versus bilateral standing (p = 0.04, mean ± SD: −26 ± 25%). No statistical difference in unadjusted soleus H-reflex amplitude (p = 0.48) was observed between conditions, but a significant increase in soleus bEMG (p = 0.04; 63 ± 65%) and ankle co-contraction level (p = 0.02; 78 ± 60%) was found with unilateral standing (Fig. 2; see Supplementary Fig. 1 for individual reflexes). A representative response is presented in Fig. 3, whereby P6 had a relatively unchanged reflex response despite a large increase in background soleus activation, resulting in a decreased adjusted reflex response (i.e., reflex response relative to background motor activity), accompanied by a large increase in co-contraction at the ankle.

Figure 2. Task complexity results.

Box and whisker plots of soleus H-reflex amplitude, soleus background muscle activity (bEMG), adjusted soleus H-reflex (H-reflex normalized to bEMG), and co-contraction at the ankle during unilateral standing relative to bilateral standing; + indicates an outlier; *p < 0.05.

Figure 3. Representative standing response.

Representative reflex response (P6) from the baseline (blue) to complex (red) condition, with the average reflex tracing across 15 trials for both conditions (shading indicates ± 1 standard deviation), and the associated soleus background muscle activity (bEMG), adjusted soleus H-reflex (H-reflex normalized to bEMG), and ankle co-contraction level.

There was a strong positive (Mukaka 2012) and statistically significant relationship (R = 0.79, p < 0.05) between participants’ overground walking tVAF₁ and adjusted H-reflex response relative to baseline for the task complexity protocol (Fig. 4).

Figure 4. Neural control complexity and reflex modulation.

Relationship between neural control complexity (tVAF1) during walking at a self-selected speed and the adjusted soleus H-reflex response during unliteral standing relative to bilateral standing for P2 – P8. R indicates the Spearman’s Rank correlation coefficient; *p < 0.05.

Walking ankle resistance protocol

There was a significant decrease in ankle angle at midstance for the resistance condition (-2.3 ± 1.5 degrees, p = 0.03), indicating slightly more plantarflexion relative to baseline (1 – 4 degrees, Fig. 5). Overall, ankle kinematics appeared similar between conditions, and not likely to be a factor influencing reflex responses. No significant difference in tVAF₁ was found between conditions (p = 0.92). For H-reflex testing, there were no significant differences for any measures during the resisted condition versus baseline (p > 0.05 for all measures) (Fig. 6; see Supplementary Fig. 2 for individual reflexes). However, there was variability in the responses to the resisted condition, as shown in Fig. 7. For example, P6 displayed a substantial decrease in H-reflex amplitude with the resisted condition, and this remained even when adjusting for changes in soleus bEMG (Fig. 7A). This decreased reflex amplitude was accompanied by a 16% decrease in tVAF₁, representing a more complex neural control strategy. Conversely, P4 displayed a substantial increase in H-reflex amplitude, which remained above baseline even after correcting for changes in soleus bEMG, and had a 1% decrease in tVAF₁ (Fig. 7B).

Figure 5. Ankle kinematics and muscle activity.

Individual participants’ average ankle angles (first row) and soleus muscle activation (second row) for baseline (blue) and resisted (red) walking over the gait cycle, with the non-dimensional (nd) walking speed used for both conditions; shading indicates ± 1 standard deviation.

Figure 6. Walking ankle resistance results.

Box and whisker plots of soleus H-reflex amplitude, soleus background muscle activity (bEMG), adjusted soleus H-reflex (H-reflex normalized to bEMG), and co-contraction at the ankle during resisted walking versus baseline walking; + indicates an outlier.

Figure 7. Individual walking responses from two participants.

Average reflex tracings (shading indicates ± 1 standard deviation) for baseline (blue) and resisted (red) treadmill walking for (A) a participant (P6) who displayed decreased reflex excitability with resistance and (B) a participant (P4) who displayed increased reflex excitability with resistance, with each participant’s associated soleus background muscle activity (bEMG), adjusted soleus H-reflex (H-reflex normalized to bEMG), and tVAF1, or the muscle activity accounted for by one muscle synergy, where a lower value indicates increased neural control complexity.

Discussion

The purpose of this study was to examine soleus H-reflex modulation in children and young adults with CP, and how reflex modulation related to measures of motor control. We observed a significant decrease in soleus H-reflex amplitudes relative to background motor activity, or the excitability of the reflex response during unilateral standing compared to bilateral standing. This ambulatory cohort of individuals with CP were able to modulate their reflexes with this more challenging motor task, likely as a result of preserved supraspinal input. We also observed a very strong positive relationship between neural control complexity during walking and reflex modulation between bilateral vs unilateral stance. The individuals who had greater decreases in reflex excitability also had higher levels of neural control complexity during walking (i.e., lower tVAF₁). We did not observe this same decrease in reflex excitability at the group level when participants walked with a resistive ankle exoskeleton device that had previously demonstrated a training effect of increased neural control complexity (Table 2). These findings contribute to our understanding of how ambulatory individuals with CP modulate their ankle plantar flexor reflexes in response to more demanding motor tasks, and how this relates to their motor control strategies while walking.

Table 2.

A priori hypotheses and associated findings

Hypothesis	Supported by findings?
Individuals with CP would display no change in reflex excitability from bilateral to unilateral standing	Not supported Participants did have a significant reduction in reflex excitability from bilateral to unilateral standing (-26 ± 25%, p = 0.04)
There would be a significant positive relationship between more refined neural control during walking and an increased ability to modulate reflex excitability	Supported A strong positive relationship was observed (R = 0.79, p = 0.04)
Individuals with CP would display a decrease in soleus H-reflex excitability when walking with ankle exoskeleton resistance	Not supported No difference in reflex excitability was observed between walking with and without ankle exoskeleton resistance (p > 0.05), which variable responses to this paradigm

The increase in activation of the soleus during unilateral vs bilateral standing was expected given the increased demand that this task places on the ankle musculature, particularly for individuals with CP who often lack stability and coordination at this joint (Conner et al. 2022). Increased voluntary activation of the soleus will increase the reflex response at a matched stimulation level (Stein et al. 2007), meaning the increased voluntary activation seen with unilateral standing would result in a greater reflex response if modulation did not take place. However, we observed no change in the adjusted H-reflex, suggesting that modulation did take place. A direct comparison to previously published results for soleus H-reflex modulation in nondisabled individuals for bilateral and unilateral standing is difficult given slight methodological differences. For example, a previous study assessing reflex modulation with unilateral standing did not normalize reflex amplitudes to background soleus activation. However, we can use the reported increase in background soleus activation of approximately 11% from bilateral to unilateral standing to calculate an adjusted H-reflex response, which comes out to an approximately 23% decrease in reflex excitability for individuals without CP during unilateral standing (Huang et al. 2009). This is very similar to the 26% decrease seen for individuals with CP in the present study, and the slight difference may be due to the fact that unilateral standing was a more difficult and/or less stable condition relative to bilateral standing for individuals with CP, necessitating a greater decrease in reflex excitability.

Unilateral standing resulted in significantly higher levels of co-contraction at the ankle, demonstrating the increased stabilization that was necessary at this joint during this condition (Darainy and Ostry 2008). Increased activation of the tibialis anterior muscle relative to soleus activation can result in reciprocal inhibition, where activation of the tibialis anterior inhibits the stretch reflex arc of the soleus by inhibiting Ia afferents. It was previously found that children with CP have deficits in reciprocal inhibition, indicated by increased soleus H-reflexes with the onset of tibialis anterior activation (i.e., increased ankle co-contraction) as opposed to decreased reflex amplitudes (Leonard et al. 1990), as seen in nondisabled children and adults. Our finding of decreased reflex amplitude with increased co-contraction at the ankle in this cohort of individuals with CP could be due to the increased task complexity of unilateral standing, which may have resulted in greater supraspinal input (Llewellyn et al. 1990). This task-dependent input from supraspinal centers via presynaptic inhibition of the Ia afferents (Taube et al. 2008) has been proposed to be important during less stable conditions (i.e., unilateral standing) when a large reflex could detrimentally impact motor performance (Llewellyn et al. 1990), and a central control mechanism to preserve the finite motor pool for the task at hand (Chen and Zhou 2011).

The complexity of neuromuscular control during walking appears to be strongly related to the ability to modulate plantar flexor reflexes. The merging of muscle synergies (i.e., higher tVAF₁) has been thought to reflect decreased supraspinal input due to damages to corticospinal pathways in individuals post-stroke (Clark et al. 2010). Individuals with CP similarly demonstrate higher tVAF₁ relative to nondisabled peers (Steele et al. 2015) which is likely caused in-part by damages to corticospinal tracts (Farmer et al. 1991) leading to decreased supraspinal drive. As supraspinal input is a necessary factor in decreasing reflex excitability with increased task complexity, the strong positive relationship observed in this study between tVAF₁ and reflex excitability appears to support the idea that higher tVAF₁ is indicative of decreased supraspinal drive. This relationship was previously proposed by Ting (Ting 2007), who suggested that the descending pathways responsible for context-dependent control of afferent input and reflex arcs work in concert with muscle synergies to create a simple and predictable hierarchal control architecture.

Hodapp and colleagues (Hodapp et al. 2009) proposed that the reduction in stance-phase soleus H-reflex amplitude was a result of improved supraspinal drive for increased presynaptic inhibition after treadmill training in CP. The findings from the present study appear to support this theory, indicating that individuals with CP retain the ability to reduce their reflex excitability with increased supraspinal drive. This is also supported by the finding that during a specific motor task, “experts” at that motor task will utilize a larger set of muscle synergies (i.e., a lower tVAF₁) relative to “novices” (Sawers et al. 2015), possibly representing greater descending input by the experts. With repeated training, individuals with CP may become more adept at treadmill walking and employ a more complex control strategy defined by greater supraspinal input, which could then decrease reflex excitability.

Training with the exoskeleton device used in the walking ankle resistance protocol has demonstrated benefits on neuromuscular control, such as increased neural control complexity and walking ability for individuals with CP (Conner et al. 2020b, 2021), findings that align with other positive results of resistive robotic gait training for individuals with CP (Kang et al. 2017; Wu et al. 2017). The present study evaluated the effect that walking with this device had on soleus H-reflex modulation during the stance phase of gait, which has been observed to be significantly higher relative to nondisabled peers (Hodapp et al. 2007). The measured reflex response to resistance was highly variable, with no statistical group difference in any of the primary outcome measures. There was no group-level increase in neural control complexity (i.e., decrease in tVAF₁) when walking with resistance, and this may have been due to the fact that minimal instructions were given to participants when walking with resistance, which was previously used to facilitate resistance engagement (Conner et al. 2020a, b, 2021), in addition to the removal of biofeedback when H-reflex testing took place. While we felt this was necessary to avoid confounding factors on reflex modulation, future studies may consider reflex testing with these components incorporated. On an individual basis, however, increased neural control complexity does appear to influence reflex modulation when walking with resistance, as indicated by P6 (Fig. 7A), who displayed a marked decrease in reflex response alongside a marked increase in neural control complexity, while P4 (Fig. 7B) displayed a marked increase in reflex response and virtually no change in neural control complexity. Still, it is difficult to draw broad conclusions without further testing, especially given the multifaceted nature of reflex responses and the inherent heterogeneity of motor control strategies with CP.

The findings from this study may have promising implications for the field of neurorehabilitation. This is the first study, to our knowledge, to assess soleus H-reflex modulation with increased task complexity in CP. The findings suggest that during challenging and less stable static tasks, individuals with CP are able to reduce their soleus H-reflex excitability. The magnitude of decrease seems to be related to neural control complexity during walking (i.e., tVAF₁), and may be associated with differences in supraspinal input during a motor task. As tVAF₁ is predictive of treatment outcomes for individuals with CP (Schwartz et al. 2016), this connection with reflex modulation supports the notion that supraspinal drive is a major contributing factors to neuromuscular deficits and therapeutic outcomes in CP. For this reason, those interventions that specifically promote increased supraspinal drive may be the best treatment option for comprehensively addressing this multifaceted movement disorder.

There are notable limitations of the present study that should be considered. First, we tested a relatively small sample size of a classically heterogenous patient population. Still, we observed significant changes during our standing task complexity protocol, and gathered novel information on a rather understudied aspect of CP. Second, while we were able to use previously published results on reflex modulation with unilateral stance in healthy adults for comparison with our participants with CP, a more robust comparison would have been to an age-matched control group of adolescents without CP. Third, we only analyzed soleus H-reflex amplitudes during mid-stance of the gait cycle, and it is possible that there were differences in phase dependent modulation between baseline and resisted walking. Limiting our analysis to one part of the gait cycle was deliberate in order to reduce the number of stimulations necessary for this pediatric population, and it has been previously found that individuals with CP already demonstrate the phase dependent modulation seen in nondisabled peers (Hodapp et al. 2007). Fourth, it has been demonstrated previously that modulation of the soleus H-reflex can differ from modulation of a mechanically-induced stretch reflex (Morita et al. 1998), so we are not able to draw conclusions on the latter with relation to task complexity for individuals with CP. It is possible, however, that a future study could utilize an ankle exoskeleton device, such as the one used in this study, to induce stretch reflexes and explore this further. Finally, we did not assess the training effect of adaptive plantar flexion resistance on reflex modulation, as was done previously with assistive robotic gait training (Kao et al. 2010), or the effect of adding audiovisual biofeedback to this modality, which should be explored in future investigations.

Conclusion

A small cohort of children and young adults with CP demonstrated the ability to decrease their soleus H-reflex excitability during a more challenging motor task, likely due to a combination of reciprocal inhibition and supraspinal input. Reflex modulation of the ankle plantar flexors during unilateral vs bilateral standing was significantly related to neural control complexity while walking. This highlights the importance of effective central control mechanisms for both volitional control of lower limb muscles during dynamic tasks like walking and effective communication with spinal reflex loops in this patient population. Future work should investigate how these reflex responses change over time with motor skill acquisition, and the contribution of variations in reflex modulation patterns to differences in mobility-related measures and responses to interventions. Importantly, this study adds to the limited understanding of lower limb reflex modulation in adolescents and young adults with CP, which has been relatively understudied to date. As these findings seem to indicate, reflex modulation may be an important component to consider when addressing neuromuscular deficits associated with CP, and should continue to be explored.

Supplementary Material

Supplementary Figure 1

Supplementary Figure 1. Task complexity individual reflex responses. Individual soleus H-reflex tracings (15/condition) with bilateral (blue) and unilateral (red) standing. The spike at ~20 ms on each plot is the electrical stimulation picked up by the soleus EMG sensors, with the M-wave occurring ~10 ms after and the H-reflex ~30–40 ms after.

Supplementary Figure 2

Supplementary Figure 2. Walking ankle resistance individual reflex responses. Individual soleus H-reflex tracings (12/condition) with unresisted (blue) and resisted (red) walking. The spike at ~20 ms on each plot is the electrical stimulation picked up by the soleus EMG sensors, with the M-wave occurring ~10 ms after and the H-reflex ~30–40 ms after.

Abbreviations:

bEMG

background electromyography

CP

cerebral palsy

CCI

co-contraction index

H-reflex

Hoffman reflex

M_max

maximal muscle response

tVAF1

the variance in muscle activity explained by one muscle synergy

Availability of data and materials:

The data that support the findings of this study are available from the corresponding author, upon reasonable request.