Motor adaptation to continuous lateral trunk support force during walking improves trunk postural control and walking in children with cerebral palsy: A pilot study

Abstract

Purpose:

To determine whether the application of continuous lateral trunk support forces during walking would improve trunk postural control and improve gait performance in children with CP.

Materials and Methods:

Nineteen children with spastic CP participated in this study (8 boys; mean age 10.6 ± 3.4 years old). Fourteen of them were tested in the following sessions: 1) walking on a treadmill without force for 1-minute (baseline), 2) with lateral trunk support force for 7-minutes (adaptation), and 3) without force for 1-minute (post-adaptation). Overground walking pre/post treadmill walking. Five of them were tested using a similar protocol but without trunk support force (i.e., control).

Results:

Participants from the experimental group showed enhancement in gait phase dependent muscle activation of rectus abdominis in late adaptation period compared to baseline (P=0.005), which was retained during the post-adaptation period (P=0.036), reduced variability of the peak trunk oblique angle during the late post-adaptation period (P = 0.023), and increased overground walking speed after treadmill walking (P = 0.032). Participants from the control group showed modest changes in kinematics and EMG during treadmill and overground walking performance. These results suggest that applying continuous lateral trunk support during walking is likely to induce learning of improved trunk postural control in children with CP, which may partially transfer to overground walking, although we do not have a firm conclusion due to the small sample size in the control group.

1. INTRODUCTION

Many children with CP exhibit impairments in their walking abilities and manipulation skills (M. H. Woollacott & Shumway-Cook, 2005). One crucial factor that contributes to impairments in gait and manipulation is the dysfunction in trunk postural control (Balzer, et al., 2017; Curtis, et al., 2015). Trunk postural control is a prerequisite for daily activities and for walking. However, many children with CP present with impaired trunk postural control (Heyrman, et al., 2014; Nashner, Shumway-Cook, & Marin, 1983), which derives from the primary brain injury that causes deficits in postural neural networks (Y. M. Kwon, Rose, Kim, & Son, 2017). For instance, children with CP manifest a greater range of motion for pelvis tilt, thorax, head, and kyphosis and lordosis, and greater variability of movement for thorax, spine and shoulder during gait, compared to their typically developed peers (Heyrman, et al., 2013; Romkes, et al., 2007). The abnormal trunk movements during gait have been suggested to reflect compensatory movements due to lower limb motor impairments, but may also be partially due to an underlying trunk control deficit (Heyrman, et al., 2014; Kiernan, 2021). Impairments in trunk postural control significantly limit their walking capacity (Curtis, et al., 2015) and negatively impact their daily activities. Trunk control was the strongest predictor of gait capacity of children with CP (Balzer, et al., 2017). In particular, because balance control in the frontal plane is more challenging during locomotion (Bauby & Kuo, 2000), and children with CP, particularly for those with lower walking function, showed excessive lateral trunk movement, i.e., the trunk tilts to the left/right sides (Attias, et al., 2015), during walking, which may be partially due to impairments in frontal trunk postural control and/or result from the compensation for their lower limb motor deficits. Thus, improvements in trunk postural control, particularly the trunk control in the frontal plane, may potentially facilitate walking in children with CP.

While the impact that trunk postural control dysfunction has on the activity of children with CP has long been recognized, strong evidence of effective interventions for treating this key deficit is lacking (Dewar, Love, & Johnston, 2015). For instance, hippotherapy, an intervention that applies rhythmical perturbation forces to the pelvis during sitting astride using horseback movement, has been used for improving balance and gait in children with CP for decades (Bertoti, 1988). While some studies have shown improvements in postural control and motor function in children with CP after hippotherapy (J. Y. Kwon, et al., 2015), a systemic review indicated a mixed result regarding the effect of hippotherapy on Gross Motor Function Measure (GMFM) scores (Tseng, Chen, & Tam, 2013). In addition, reactive balance training using a moving platform while participants stand on it has been used to improve standing balance in children with CP (El-Shamy & Abd El Kafy, 2014; Shumway-Cook, Hutchinson, Kartin, Price, & Woollacott, 2003). However, this approach may not directly target trunk postural control because this perturbation is applied to the feet (M. Woollacott, et al., 2005), which mostly triggers leg muscle activation (Burtner, Qualls, & Woollacott, 1998). Treadmill training is also a commonly used approach for improving gait and/or postural control (Dodd & Foley, 2007), but it does not target trunk postural control, or the effect of the given perturbation on trunk postural control is too small to induce improvements. Thus, there is a clear need to develop more effective intervention approaches to improve trunk postural control during walking in children with CP.

Results from a previous study indicated that the application of continuous force perturbations to the pelvis during walking might improve dynamic balance control in adults with spinal cord injury (Lin, et al., 2020). It is unclear whether the approach that applying varied perturbation/supporting force to the trunk during walking may induce improvements in trunk postural control during locomotion in children with CP.

Thus, we aimed to determine whether the application of continuous varied trunk supporting force in the frontal plane during walking would induce improvement in trunk postural control in children with CP. We hypothesized that the application of continuous varied lateral trunk supporting force during walking would improve trunk postural control, which might facilitate walking in children with CP. In addition, we hypothesized that the improved trunk postural control might transfer to overground walking given there are some shared neural circuits involved in treadmill and overground walking (Reisman, Wityk, Silver, & Bastian, 2009) in children with CP.

2. METHODS

2.1. Participants

Nineteen children with spastic CP who were recruited from the outpatient clinics of the Shirley Ryan AbilityLab participated in this study. Fourteen of them (5 boys; mean age 12yr 7mo ± 3yr 3mo) (Table 1) were tested in the experimental condition, and additional 5 of them (3 boys, mean age 8yr 3mo ± 3yr 2mo) were tested in the control condition (Table 1). Inclusion criteria were the following: 1) age 4-16 years old; 2) spastic CP with GMFCS levels from I to III; 3) ability to independently walk with or without assistance; 4) ability to signal pain, fear or discomfort; 5) ability to follow instructions. Exclusion criteria were the following: 1) orthotics that cross the knees; 2) lower extremity orthopedic surgery or Botulinum toxin injection within the past 6 months; 3) severe lower extremity contractures, fractures, osseous instabilities; 4) cardiovascular instability. The Northwestern University Medical School Institute Review Board approved this study. Children and their parents signed consent forms before data collection. Children’s GMFCS information was obtained from their medical records.

Table 1.

Characteristics of children with cerebral palsy, walking speeds, and force magnitudes applied to the trunk.

Participant No.	Age yr, mo	Sex	BW kg	BH (m)	Type of CP	GMFCs	More affected side	AFO	Treadmill Walking Speed (m/s)	Support force (N)
1	13, 10	F	49.7	1.62	diplegia	I	R	R	1.1	130
2	11, 7	F	38.2	1.44	diplegia	I	R	No	0.6	79
3	11, 6	F	35.5	1.53	diplegia	I	R	No	0.9	96
4	8, 10	F	43.1	1.38	diplegia	II	R	No	0.4	102
5	10, 1	F	24.6	1.27	triplegia	III	L	Both	0.1	90
6	12, 5	M	35.3	1.46	diplegia	I	L	No	0.7	102
7	5, 2	F	17.1	1.14	diplegia	I	R	N	0.4	90
8	8, 5	F	25.2	1.25	diplegia	II	L	No	0.4	102
9	12, 8	M	64.5	1.69	diplegia	II	R	No	0.8	102
10	15, 8	M	75.8	1.90	diplegia	I	R	No	0.9	113
11	16, 9	M	56.2	1.70	diplegia	II	L	Both	0.4	113
12	14, 3	M	43.1	1.55	triplegia	II	R	Both	0.5	113
13	14, 11	F	48.6	1.62	hemiplegia	I	R	No	0.8	113
14	13, 6	F	45.2	1.58	quadriplegia	I	L	No	0.5	107
15 c	4, 4	M	15.6	1.18	diplegia	II	L	Both	0.1	NA
16 c	6, 2	F	19.3	1.21	diplegia	I	R	none	0.5	NA
17 c	12, 1	F	38.8	1.56	diplegia	II	L	Both	0.8	NA
18 c	9, 1	M	33.6	1.43	triplegia	II	R	none	0.4	NA
19 c	9, 5	M	26.2	1.24	diplegia	II	R	Both	0.4	NA

Note: Diplegia refers that participants showed deficits in both lower limbs. Triplegia refers that participants had deficits in both lower limbs and one upper limb. Hemiplegia refers that participants had deficits in lower limb and upper limb only on one side. Quadriplegia refers that participants showed deficits in both lower limbs and upper limbs. Participant numbers with letter c indicate the participant from the control group. Abbreviations: yr, year; mo, month; BW, body weight; CP, cerebral palsy, GMFC-s, Gross Motor Function Classification System; AFO, ankle-foot orthoses; BH, body height; m, meter; cm/s, centimeters per second; F, female; M, male. L, left; R, right; N, Newton. NA represents not applicable.

2.2. Apparatus

A custom designed trunk-support system, mounted over a treadmill, was used to provide lateral trunk support force while participants walked on a treadmill (Figure 1a). Specifically, two belts (width = 10.5 cm) were attached to the upper trunk (approximately T7-T10) and lower trunk (approximately L1-L4). Each belt was connected to two extension springs (stiffness=1.31 N/mm) through two ropes (0.8 cm in diameter) from the left and right sides of the participant (Figure. 1a). Each rope was affixed to a fixed aluminum frame that was located at the side of a split-belt treadmill through a sleeve bearing. The locations of sleeve bearings were adjusted according to each participant’s height to ensure the pulling direction of the trunk support force was horizontal. A compression/tension load cell (MLP-75, Transducer Techniques, Temecula, CA) was attached to the rope in series and was used to record the pulling force. As shown in Figure 1, the magnitude of the supporting force was linearly related to the displacement of stretch of the spring that was attached to the side of participants, and the magnitude of the supporting force applied to the trunk was the resultant force induced by two stretched springs. In particular, because the trunk obliquity angle varied across a gait cycle, the magnitude of the resultant supporting force also varied within a gait cycle, accordingly. In addition, a GAITRite mat (CIR Systems, Inc., Franklin, NJ, USA) was used for recording participants’ overground walking performance.

Figure 1.

Experimental setup, protocol, the profile of support force applied to the trunk within a gait cycle, and the peak trunk support force during the course of treadmill walking. a) A participant walked on a treadmill with bilateral trunk support force which was applied through a trunk-support system. b) The experimental protocol for both the experimental and the control groups consisted of 3 overground walking sessions (pre, immediately after treadmill walking, and 10 post treadmill walking), and treadmill walking session. c) Profile of the magnitude of the trunk support force induced by the springs on the left (orange line), and right (green line) sides, and the resultant force applied to the trunk (purple) over a gait cycle. The vertical dash line represents the moment of left foot contact. Positive values of force represent the force applied toward the right side of the participant; negative values of force represent the force applied toward the left side of the participant. d) The peaks of the trunk support force applied toward the right side (blue) and the left side (black) of the participant during treadmill walking in the experimental group. As shown in the figure, the magnitudes of the peak force showed prominent variability across different steps during the adaptation period. Each dot represents the peak of resultant force of each stride. Data in Figure c and d were from a representative participant. Abbreviations: OG, overground; TM, treadmill; TS, trunk support; N, newton; RFC, right foot contact; LFC, left foot contact.

Figure 1 Alt Text. Fig. a. A participant walked on a treadmill with the application of bilateral trunk support force which was delivered through two belts, ropes, extension springs, loadcells and a fixed frame. Fig. b. The protocol for the experimental group that was consisted of 3 overground walking sessions (pre, immediately after treadmill walking, and 10 post treadmill walking), and treadmill walking sessions, which included 1-minute walking without trunk support, 7-minute walking with trunk support, followed by a 1-minute walking without trunk support, and a 5-minute walking with trunk support. The protocol for the control group were similar to the experimental group except for no trunk support was provided. Fig. c. Two curves represent the changes of force magnitude of trunk support of the left and right sides of a participant over a gait cycle. The third curve represents the change of resultant force of two sides. Fig. d. Dots with marked variability represent peaks of the trunk support force applied toward both sides of the participant during 7-minute treadmill walking in the experimental group.

2.3. Experimental protocol

In the experimental condition, the protocol included: 1) treadmill walking without support for 1 minute (baseline); 2) walking with lateral trunk support force for 7 minutes (adaptation); 3) walking without support for 1 minute (post-adaptation); 4) standing break for 1 minute; 5) walking with trunk support force for additional 5 minutes (re-adaptation). Treadmill speed was set at their self-selected comfortable speed. Overground walking speed was tested before, immediately after treadmill walking, and 10 minutes after the end of treadmill walking, Figure 1. In the control condition, a protocol that was comparable to that of the experimental condition was used except for no trunk support force was applied, i.e., treadmill walking only (Figure 1b). Before the start of adaptation sessions, participants walked at their fast walking speed on a treadmill for 30 strides. EMG data collected during this time of period were used to normalize the EMG data of other sessions.

During treadmill walking, an overhead harness was used for protection only but no body weight support was provided. For safety, participants were allowed to hold onto the hand railings but were requested to remain to be consistent throughout all the treadmill walking sessions. The springs were pre-extended (pre-stretched 8-11.5 cm, Table 1) by pulling the ropes on both sides simultaneously. The average spring pretension force was 33±12% of body weight, which was determined during standing based on children’s tolerance (i.e., to make sure they felt comfortable).

During overground walking, participants were instructed to walk on an instrumented gait mat (GAITRite^® Platinum Plus Classic, CIR Systems, Inc., Franklin, NJ, USA, dimension: 5.1 × 0.9 m) with self-selected comfortable speed. In addition, participants were also instructed to walk as fast as they could but safely and without running for their fast walking speed. Three trials were conducted for each speed, and the average of 3 trials was used for analysis.

2.4. Data collection

We recorded surface electromyographic (EMG) activity of twelve muscles from the more affected body side: tibialis anterior (TA), soleus (SOL), gastrocnemius medialis (GAS), rectus femoris (RF), vastus medialis (VM), medial hamstring - semitendinosus (MH), hip adductor - magnus (ADD), hip abductor - gluteus medius (ABD), rectus abdominis (RA), external oblique abdominis (OBL), erector spinae - longissimus (ES), and trapezius (TRAP). The more affected side was determined by participants’ self report, and confirmed by a physical therapist based on manual muscle strength test of key leg muscles. The placement sites of EMG electrodes for the RA and OBL were determined according to Ng et al. (Ng, Kippers, & Richardson, 1998). The remaining EMG electrodes were positioned according to SENIAM (Armand, Decoulon, & Bonnefoy-Mazure, 2016; Hermens, Freriks, Disselhorst-Klug, & Rau, 2000). Surface EMG signals were recorded by using the Trigno wireless EMG system (Delsys Inc., Boston, USA) with the signals amplified (×1000) and band-pass filtered (20-450 Hz), and then sampled with an A/D board at 500 Hz using a customized Labview program (National Instruments, Austin, USA).

The kinematic signals of trunk and leg movements were recorded using the Xsens system (Xsens, Enschede, Netherlands) (Zhang, Novak, Brouwer, & Li, 2013), which consisted of 8 wearable sensors, with the sampling frequency at 100 Hz. The placement of these sensors was determined based on the instructions of the Xsens system. These sensors collected 3-dimensional position and angular data of the trunk, hips, knees, and ankles. The kinematic and EMG signals were synchronized using a Labview program and were collected during the treadmill walking sessions.

2.5. Data analysis

All EMG and kinematic data during treadmill walking were analyzed using custom Matlab programs (MathWorks, Natick, USA). The EMG data were first high-pass filtered using a 4^th order Butterworth filter with cutoff frequency at 10Hz (Yan, et al., 2023), then, notch filtered at 55-65 Hz and 115-125 Hz to remove electrical noise (Yan, et al., 2023). All EMG data were rectified and smoothed using a 4^th order Butterworth filter with a lowpass cutoff frequency at 40 Hz (Yan, et al., 2023) (Wu, Kim, Arora, Gaebler-Spira, & Zhang, 2016). The smoothed EMG data were segmented into different gait cycles based on ankle positions starting from the initial contact of the weaker leg. The integral of EMG activity (iEMG) was calculated for the intervals of stance and swing phases of gait cycle. In addition, the integrals of EMG activity of intervals during the preswing (50-60% of gait cycle) and initial swing (60-70% of gait cycle) were also calculated ^{(White & McNair, 2002)}. The EMG data were normalized to the peak iEMG signals of each muscle during participant’s fast-speed walking session.

Kinematic data during treadmill walking were lowpass filtered using a Butterworth 2^nd order filter with cutoff frequency at 10 Hz. Kinematic variables included trunk obliquity, step length and step height. Trunk obliquity was defined as the maximum angle that the trunk tilts to the side ipsilateral to the stance leg (Krautwurst, et al., 2013). The angle was calculated from two segments in the frontal plane over a gait cycle; one segment was the axis through T8 and the center of sacrum and the other was the vertical axis. The variability of trunk obliquity was quantified using the standard deviation of the trunk obliquity angles. The step length was defined as the distance between the ankle locations in the anterior-posterior (AP) direction at the time of foot contact. Step height was defined as the maximum vertical displacement of the ankle over a step. Overground walking data including walking speed, cadence, and step length were derived using the software from GAITRite.

During treadmill walking, spatial-temporal parameters, such as step length, step height, and integrated EMG during stance, swing, preswing and initial swing phases of gait were averaged over the middle 30 steps during baseline, the first 5 steps (early-adapt), the middle 5 steps (mid-adapt), and the last 5 steps during the motor adaptation period (late-adapt), and the first 5 steps (early post-adapt) and the last 5 steps (late post-adapt) during the post adaptation period for each participant (the average of 5 steps was used because motor adaptation was generally short-lived, which was used by investigators in the field (Park, et al., 2021)). To quantify the variability of trunk obliquity across different time periods, we calculated the standard deviation of trunk obliquity angle across the middle 10 steps during baseline, the first and last 10 steps during the adaptation period, and the first and last 10 steps during the post-adaptation period.

2.6. Statistical analysis

We used Shapiro-Wilk W test to assess the normal distribution of variables. Mauchly’s test statistic was used for assessing violations of sphericity. Greenhouse-Geisser correction was used when data violated sphericity assumptions. One-way repeated measures ANOVA was conducted to identify significant differences at different time points. For the sessions of overground walking, we compared walking speed, step length and cadence at 3 time points (OG baseline, immediately after treadmill walking (OG post), 10 minutes after treadmill walking (OG post10). For the sessions of treadmill walking, we compared iEMG of stance phase and swing phase of gait, step length and step height at 3 time points (baseline, early-adapt, late-adapt) to examine the motor adaptation to the trunk support force. In addition, we compared step length, step height, and iEMG of stance phase and swing phase of gait at 3 time points (baseline, early post-adapt, late post-adapt) to examine the motor learning effect induced by the continuous trunk support force. In addition, we also compared the trunk obliquity variability and iEMG of preswing and initial swing phase of gait at baseline, early post-adapt, and late post-adapt.

Post-hoc test with Least Significant Difference (LSD) adjustment was used (Hayter, 1986). When data failed to pass the normality test, e.g., iEMG of RA during the stance phase, log-transformed data were used to perform ANOVA.

To compute the number of steps with enhanced muscle activity of trunk muscles during the adaptation period, we counted the number of steps which the iEMG exceeded 95% confidence interval values of iEMG during baseline (Park, et al., 2022), which was calculated using the iEMG of the middle 30 steps during baseline. Pearson correlation coefficient was calculated to determine the association between the number of steps with enhanced muscle activity of RA and OBL during the adaptation period and the changes in iEMG of RA and OBL from baseline to the late post-adaptation period, which was used to determine whether the motor performance during the adaptation period had an impact on the motor learning during the late post-adaptation period.

Spearman correlation was used to determine the association between baseline walking speed and the change in walking speed from OG baseline to OG post10, which was used to determine whether the baseline walking speed had an impact on the improvement in overground walking speed after treadmill walking. An Independent t test was used to compare the change in walking speed between children with GMFCS level I vs. level II. All statistical analyses were performed using IBM SPSS Statistics 26 (IBM Corp., Armonk, USA) with the significant level at P < 0.05.

3. RESULTS

3.1. The trunk obliquity during treadmill walking

For the experimental group, the application of trunk support force had an impact on the peak trunk obliquity on the more affected side during treadmill walking. For instance, the peak trunk obliquity on the more affected side increased during the adaptation period, as in Figure 2a. Following the release of the trunk support force, the peak trunk obliquity of the post adaptation period returned to a level that was comparable to baseline.

Figure 2.

Group average of trunk obliquity towards the more affected side and less affected side, and the variability of trunk obliquity during treadmill walking. Group average curves and shaded standard errors of trunk obliquity towards both sides during the course of treadmill walking for the experimental group (a) and the control group (b). The trunk obliquity was defined as the maximum angle between the axis that goes through T8 and the center of sacrum, and the vertical axis in the frontal plane within a gait cycle. The shaded area in green represents the ± 1 standard error. The horizontal dash lines represent the average of trunk obliquity at baseline of treadmill walking. Group average of the variability (standard deviation) of the trunk obliquity of the more affected side (upper panel) and less affected side (lower panel) during baseline, adaptation period, and post-adaptation periods in the experimental group (c) and the control group (d). Error bars indicate means ± standard error for each condition. The Asterisk (*) indicates a significant difference.

Figure 2 Alt Text: Fig. a. Two curves represent group average of the peak trunk obliquity angles of the more affected side and less affected side of participants from the experimental group. The trunk support induced an increase in peak trunk obliquity angle of the more affected side, but a modest change in the less affected side. Fig. b. Bar figures represent the variability of trunk obliquity angle of participants from the experimental group and the control group. For the experimental group, the variability of trunk obliquity angle were significantly decreased from baseline to late-adaptation, from baseline to late post-adaptation, and from early post-adapt to late post-adapt. For the control group, there was no significant change in the variability of trunk obliquity angle. Fig. c. Two curves represent group average of the peak trunk obliquity angles of both sides of participants from the control group, which showed no marked change during the course of treadmill walking. Fig. d. Bar figures represent the variability of trunk obliquity angle of participants from the control group. There was no significant difference in trunk obliquity variability was observed for the control group.

The application of trunk support had an impact on the variability of trunk obliquity on the more affected side during treadmill walking. Specifically, the variability of trunk obliquity on the more affected side significantly differed across 3 time points (i.e., baseline, early-adapt, and late-adapt periods), F(2,22) = 3.723, P = 0.040, Figure 2c. Post-hoc analysis indicated that the variability of trunk obliquity on the more affected side significantly decreased from baseline to the late-adapt (P = 0.019). In addition, the variability of trunk obliquity on the more affected side significantly differed from baseline to post-adapt (i.e., the early post-adapt and late post-adapt periods), F(2,22) = 4.811, P = 0.018), Figure 2c. Post-hoc analysis indicated that the variability of trunk obliquity on the more affected side significantly decreased from baseline to the late post-adapt (P = 0.023), and it also decreased from the early post-adapt to late post-adapt period (P = 0.014). There was no significant change in the variability of trunk obliquity on the less affected side (P > 0.05).

For the control group, the peak trunk obliquity had a modest change during the course of treadmill walking (Figure 2c). In addition, the variability of trunk obliquity also had a modest change (Figure 2d).

3.2. EMG activity of the trunk and leg muscles of the more affected side during treadmill walking

For the experimental group, the application of continuous bilateral trunk support force induced enhancement in muscle activity in trunk muscles (i.e., RA and OBL) of the more affected side during walking. For instance, group averages of iEMG of RA and OBL during stance phase of gait enhanced during the middle and the late-adapt periods although had a modest change during the early-adapt period compared to baseline, Figure 3a and 3b. In addition, the enhanced muscle activity of RA and OBL was retained after the removal of the trunk support during the early and the late post-adapt periods.

Figure 3.

Group average of muscle activity of trunk muscles from the more affected side during treadmill walking. Group average curves and shaded standard errors of integrated muscle activity of RA and OBL for the experimental (a and b) and control (c and d) groups during the course of treadmill walking. Data shown were the integral of EMG during the stance phase of gait, and were smoothed using a span of three points for the sake of clarity. The shaded areas represent ± 1 standard error for each stride. The horizontal dash lines in Figures a and b represent the average of muscle EMG during baseline of treadmill walking. Abbreviations: RA, rectus abdominis; OBL, external oblique.

Figure 3 Alt Text: Four curves represent the group average of integrated muscles activities of RA and OBL for two groups. Muscle activity of RA (a) and OBL (b) increased from baseline to the late adaptation period and was remained higher after the release of trunk support in the experimental group. Muscle activities of RA (c) and OBL (d) did not show marked changes during the course of treadmill walking for the control group.

Integrated EMG of RA (F(2, 20) = 6.848, P = 0.005, ANOVA) and OBL (F(2, 24) = 3.591, P = 0.043) during the stance phase of gait differed significantly across 3 time points (i.e., baseline, early post-adapt, and late post-adapt periods), Figure 4a. Post-hoc analysis indicated that iEMG of RA significantly increased from 52±8% (mean±SE), at baseline, to 85±16% (P = 0.007) during the early post-adapt period, and retained at 83±17% (P = 0.036) during the late post-adapt period. Similarly, integrated EMG of OBL significantly increased from 51±20% at baseline to 70±31% (P = 0.039) during the early post-adapt period, and retained at 69±32% (P = 0.036) during the late post-adapt period. No significant difference in iEMG of ES was observed from baseline to the early and late-adapt periods (P = 0.227). Also, there were no significant differences in iEMG of leg muscles (P > 0.05).

Figure 4.

Group average of muscle activity of trunk and leg muscles from the more affected side during treadmill walking. a) Group average and standard errors of integral of EMG (iEMG) of RA and OBL muscles during the stance phase of gait. b) Group average and standard errors of iEMG of GAS, SOL, VM, and OBL muscles during the swing phase of gait. EMG data in a and b were compared across three 3 time points (i.e., baseline, the early post-adaptation period, and late post-adaptation periods). c) Group average and standard errors of iEMG of RA and ABD muscles during the stance phase of gait. The data were compared across baseline, early adapt period and late adapt periods. Group average of muscle activation pattern of RA (d) and OBL (e) within a gait cycle during baseline, early post-adapt period, and late post-adapt periods (left panel). Means and standard errors of iEMG of RA and OBL during the preswing (middle panel) and initial swing (right panel). Integral of EMG were compared across baseline, early post-adapt, and late post-adapt periods. Asterisks indicate significant differences. Abbreviations: RA, rectus abdominis; OBL, external oblique; iEMG, integrated EMG.

Figure 4 Alt Text: Fig. a. Bar figures represent integrated EMG of RA and OBL during stance phase. There was a significant increase in RA and OBL muscle activity during stance phase from baseline to early post-adapt and late post-adapt. Fig. b. Bar figures represent integrated EMG of GAS, SOL, VM and OBL during swing phase. There was a significant increase in GAS, SOL, VM, OBL muscle activity during swing phase from baseline to early post-adapt. Fig. c. Bar figures represent integrated EMG of RA and ABD during stance phase. There was a significant increase in RA muscle activity from baseline to late adapt; a significant decrease in OBL muscle activity from baseline to early adapt, and a significant increase in OBL muscle activity from early adapt to late adapt. Fig. d, e. Three curves represent averaged EMG of RA (d) and OBL (e) during baseline, early post-adaptation and late post-adaptation within one gait cycle, and bar figures represent the integreated EMG of RA (d) and OBL (e) during pre-swing and initial swing. There were significant increases in iEMG of RA and OBL during pre-swing and during initial swing from baseline to early post-adapt and late post-adapt.

Additionally, integrated EMG of RA (F(2, 20)=7.31, P=0.004, ANOVA) and ABD (F(2, 24) = 5.377, P = 0.012) during the stance phase of gait differed significantly across 3 time points (i.e., baseline, early-adapt, late-adapt), Figure 4c. Post-hoc test indicated that iEMG of RA significantly increased from baseline to the late-adapt period (P = 0.005). Moreover, the iEMG of ABD during the early-adapt period was significantly smaller than baseline (P = 0.014). No significant differences in iEMG of ES (P = 0.243) and OBL (P = 0.256) were observed across the 3 time points. Also, there were no significant differences in iEMG of other leg muscles (P > 0.05).

Integrated EMG of GAS (F(2, 24) = 3.765, P = 0.038), SOL (F(2, 22) = 3.984, P = 0.033), VM (F(2, 24) = 3.826, P = 0.036), and OBL (F(2, 24) = 4.087, P = 0.03) during the swing phase of gait significantly differed across 3 time points (i.e., baseline, early post-adapt, and late post-adapt periods) (Figure 4b). Post-hoc analysis indicated that iEMG of GAS (16% increase, P = 0.04), SOL (27% increase, P = 0.03), VM (15% increase, P = 0.037), and OBL (17% increase, P = 0.043) significantly increased from baseline to the early post-adapt period, but showed no significant difference between baseline and the late post-adapt period (P > 0.05). No significant differences existed in iEMG of other trunk and leg muscles across the 3 time points (P > 0.05).

The application of continuous trunk support force also had an impact on the muscle activity of both RA and OBL during the preswing and initial swing phase of gait. For instance, as shown in Figures 4d and e (left panels), the integrated muscle activation of RA and OBL during preswing and initial swing significantly differed across the 3 time points, i.e., baseline, early and late post-adapt periods (middle and right panels in Figures 4d and e, preswing: RA, F(1.262, 12.623) = 9.645, P = 0.006; OBL, F(2,22) = 7.265, P = 0.004; initial swing: RA, F(2,22) = 14.556, P < 0.001; OBL, F(2,22) = 6.916, P = 0.005). Post-hoc analyses indicated that the integrated muscle activity of RA during preswing and initial swing increased significantly from baseline to the early post-adapt (preswing: P = 0.008; initial swing: P = 0.002) and the late post-adapt periods (preswing: P = 0.011; initial swing: P = 0.002). Additionally, the integrated muscle activity of OBL increased significantly from baseline to the early post-adapt (preswing: P = 0.011; initial swing: P = 0.046) and the late post-adapt periods (preswing: P = 0.006; initial swing: P = 0.004). Furthermore, participants who had a greater number of steps with increased iEMG of RA and OBL during the adaptation period demonstrated greater enhancements in iEMG of RA and OBL from baseline to the late post-adapt period (RA: R² = 0.512, P = 0.013; OBL: R² = 0.478, P = 0.013; Figure 5).

Figure 5.

Correlation between the number of steps with iEMG exceeding the baseline level and the change in iEMG of RA and OBL from baseline to the late post-adaptation period. Abbreviations: RA, rectus abdominis; OBL, external oblique; iEMG, integrated EMG.

Figure 5 Alt Text: Two lines show that the changes in iEMG of RA and OBL of stance phase from baseline to late post adaptation period were highly correlated with the number of steps with iEMG of stance phase exceeding baseline.

For the control group, group averages of iEMG of RA and OBL during the stance phase of gait showed modest changes during treadmill walking (Figures 3c and 3d). Integrated EMGs of other muscles also showed modest change during the course of treadmill walking (P > 0.05).

3.3. Step length and step height during treadmill walking

For the experimental group, the application of continuous trunk support force had an impact on step length during walking. Specifically, the group average of step length of the less affected side showed a prominent increase approximately 15 steps after the application of the support force during the early adaptation period, Figure 6a. The increased step length of the less affected side was retained during the middle and late adaptation periods. Following the release of the support force, the step length of the less affected side was reduced during the early post-adaptation period, and it gradually returned to a level that was comparable to baseline within approximately 10 steps, and remained at this level for the rest of the post-adaptation period.

Figure 6.

Group average of gait parameters during treadmill walking. Group average curve and shaded standard errors of step length of the less affected leg during the course of treadmill walking for the experimental (a) and control (b) groups. Step length data were averaged across participants and smoothed using a span of three steps. The shaded area represent ± 1 standard error for each stride. The horizontal dash line represents the average of step length during baseline of treadmill walking. c) Group average and standard errors of step length of both limbs at baseline, early and late adapt, and early and late post-adapt periods during treadmill walking for the experimental group. The asterisk (*) indicates a significant difference.

Figure 6 Alt Text: Fig. a. A curve represents the average of step length of the less affected side during the course of treadmill walking for the experimental group. Step length of the less affected leg increased with the trunk support and decreased when the trunk support was removed. Fig. b. A curve represents the average of step length of the less affected side for the control group, which showed modest change during the course of treadmill walking. Fig. c. Bar figures represent the average of step length of both legs for the experimental group. Step length of the less affected leg significantly increased from baseline to late adaptation period. Step length of the more affected leg showed no significant change.

Statistically, the application of trunk support force had a significant effect on the step length of the less affected side during the adaptation period (F(2, 22)=4.927, P = 0.017, Figure 6c). Post-hoc analysis indicated that the step length of the less affected leg significantly increased from baseline to the late-adapt (P = 0.009), and from the early-adapt to the late-adapt period (P = 0.043). In contrast, the application of the trunk support force had no significant impact on the step length of the more affected side (P > 0.05 and step height of both legs (P > 0.05, Figure 4b). In addition, step length of the less affected side differed from baseline to the post adaptation period (F(2,22) = 4.678, P = 0.020). Specifically, the step length significantly increased from baseline to the late post-adapt period (P = 0.045), although it significantly decreased from baseline to the early post-adapt (P = 0.025). The application of trunk support had no significant impact on the stance time on both legs (P > 0.05). For the control group, step length showed a modest change during the course of treadmill walking (Figure 6b).

3.4. Overground walking

For the experimental group, one session of treadmill walking with the application of continuous trunk support force had a significant impact on overground walking speed (F(2, 26) = 11.86, P < 0.001, ANOVA), Figure 7a. Post-hoc analysis indicated that overground walking speed significantly increased from OG baseline (0.84±0.35 m/s) to OG post10 (0.99±0.34m/s, P = 0.032), although the speed showed no significant change immediately after treadmill walking (OG post, 0.89±0.31m/s, P = 0.256). Fast walking speed tended to increase from 1.35±0.46 m/s at OG baseline to 1.42±0.46m/s at OG post10, although this was not significant (P = 0.087). Furthermore, the change in normal walking speed from OG baseline to OG post10 was dependent on the GMFCS functional level of participants (Figure 7e). Specifically, children with GMFCS level II exhibited a greater improvement in walking speed than children with level I (t (8.234) = −5.768, P < 0.001) (Figure 7f). Additionally, children who walked with a slower speed during OG baseline also showed greater improvement in walking speed at OG post10 (R² = 0.36, P = 0.023) (Figure 7g).

Figure 7.

Group average of gait parameters during overground walking. a) Group average and standard errors of walking speed (a), cadence (b), and step length of both limbs (c and d) at 3 time points (i.e., baseline, immediately post treadmill walking, and 10-minute post treadmill walking. Black bars represent experimental group; gray bars represent control group. e) changes in walking speed from OG baseline to OG post10 for children with GMFCS levels I, II, and III in experimental group. Each dot represents the walking speed of each participant. f) Group average and standard errors of the changes in walking speed from OG baseline to OG post10 for children with GMFCS level I and II. The asterisk (*) indicates a significant difference. g) Correlation between the change in walking speed from OG baseline to OG post10 and walking speed at OG baseline. Abbreviations: OG, overground; OG post, overground walking immediately after treadmill walking; OG post10, overground walking 10 minutes after treadmill walking; GMFCS, Gross Motor Function Classification System.

Figure 7 Alt Text: Bar figures represent overground walking speed (a), cadence (b), step length (c,d). The walking speed, cadence, step length of both limbs during overground walking significantly increased from baseline to immediately after treadmill walking, as well as 10-minute after treadmill walking for the experimental group, and showed no significant change for the control group. Bar figures represent the change in walking speed from baseline to 10-minute after the end of treadmill walking for different GMFCS levels (e, f). A line represent the correlation between the change in overground walking speed from baseline to 10-minute after the end of treadmill walking and baseline walking speed (g).

The spatial-temporal parameters also showed significant changes after one session of treadmill walking with trunk support (Figures 7b, 7c and 7d). Specifically, for normal speed, cadence (F(2, 26) = 5.612, P = 0.009), and step length (F(1.1, 14.304) = 6.118, P = 0.007, less affected leg; F(2, 26) = 9.454, P = 0.001, more affected leg, ANOVA) significantly changed after treadmill walking with trunk support force. Post-hoc analysis indicated that cadence (105±17 steps/min) and step length (55±13 cm, less affected; 56±13 cm, more affected) at 10-minute post treadmill walking significantly increased in comparison to OG baseline (cadence = 97±22 steps/min, P = 0.021; step length = 49±16 cm, less affected limb, P = 0.004; step length = 50±14 cm, more affected limb, P = 0.001), although had no significant change at immediate post-treadmill walking (P > 0.05; cadence = 100±17 steps/min, step length = 52±14 cm, less affected; step length = 51±14 cm, more affected). Double-support time significantly changed after one session of treadmill walking with trunk support (P = 0.03), although post-hoc test showed no significant difference between OG baseline and 10-minute post treadmill walking. For fast speed, no significant changes in cadence (P > 0.05), step length (P > 0.05), and double-support time (P > 0.05) were observed after treadmill walking with trunk support. No significant effects were observed for other variables (P > 0.05).

For the control group, walking speed showed modest changes after one session of treadmill walking only. Specifically, the normal and fast walking speeds were 0.86±0.26 m/s and 1.42±0.3 m/s at OG baseline, 0.88±0.24 m/s and 1.37±0.33 m/s at immediately post treadmill walking (OG post), and 0.87±0.28 m/s and 1.37±0.43 m/s at 10 minutes after the end of treadmill walking (OG post10) (Figure 7a, gray bars). In addition, participants did not show an increase in cadence during normal walking speed, i.e., the cadence was 112±14 steps/min at OG baseline, 112±15 steps/min at OG post, and 111±20 steps/min at OG post10 (Figure 7b, gray bars). Moreover, step length of both legs during overground walking showed modest increases (i.e., less affected leg: 46±15 cm at OG baseline, 47±13 cm at OG post, and 47±15 cm at OG post10; more affected leg: 46±13 cm at OG baseline, 47±13 cm at OG post, and 48±12 cm at OG post10) (Figure 7c and 7d, gray bars).

4. DISCUSSION

In children with CP, we examined the motor adaptation to the application of continuous lateral trunk support force during treadmill walking, and the transfer of the motor adaptation to overground walking. We found that the application of continuous lateral trunk support force during walking might induce an improvement in trunk postural control in children with CP. We also found an increase in overground walking speed after one session of treadmill walking with the application of trunk support, suggesting a transfer of motor adaptation to overground walking. Additionally, the baseline functional level and the children’s GMFCS level had an impact on the gain in overground walking speed. The gain in overground walking speed was negatively correlated to their baseline walking speed, and children with GMFCS II showed a greater improvement in overground walking speed than those children with GMFCS I. These results suggest that repeated application of lateral support force is likely to induce improvements in trunk postural control, which may transfer to overground walking and result in an increase in overground walking speed in children with CP, although we do not have a firm conclusion due to the lack of randomization of testing condition. Targeting improvement in trunk postural control may induce an increase in walking in children with CP, particularly for these children with lower GMFCS level.

The application of continuous trunk lateral support during walking might induce an increase in trunk postural control in children with CP. Many children with CP show impairments in trunk postural control during walking (Attias, et al., 2015), which may contribute to their impairments in walking function (Seyyar, Aras, & Aras, 2019). For instance, many children with CP show greater variability in trunk motor performance (Romkes, et al., 2007), which may be due to the impairments in sensory, motor, and integration of the sensory and motor information (Churchland, Afshar, & Shenoy, 2006; Osborne, Lisberger, & Bialek, 2005; Stein, Gossen, & Jones, 2005). In this study, we found that children with CP showed an improvement in trunk postural control during treadmill walking, indicated as reduced variability of the trunk obliquity angle, after the application of continuous trunk support force during walking. One possible mechanism is that the trunk support force applied might not be a pure “supporting” force, but rather, at least partially, a perturbation force, which induces motor adaptation. For instance, the magnitude of the resultant supporting force, as shown in Figure 1, was dependent on the displacement of the trunk position in the frontal plane during walking, which varied within one gait cycle (i.e., the magnitude of force was not constant within a gait cycle). In addition, given that children with CP often show greater variability in trunk displacement across different steps (Romkes, et al., 2007), the peak magnitude of the resultant support force might also vary across different steps. The varied trunk supporting force might force the CNS to recruit additional trunk muscle activation to stabilize the trunk movement. Over time, the CNS of children with CP might adapt to the varied trunk supporting force by forming a model of the environmental uncertainty and re-optimize the motor commands to reduce the variance in trunk motor performance (Izawa, Rane, Donchin, & Shadmehr, 2008), resulting in reduced variability in the trunk obliquity angle during the late adaptation period.

Further, repeated walking practice with improved trunk postural control might also induce motor learning of trunk control, which was supported by the retention of improved trunk postural control during the late post-adaptation period. For instance, participants showed reduced variability in trunk obliquity during the late post-adaptation period when the trunk supporting force was removed, suggesting that a short-term motor memory of improved trunk postural control might be formed after one session of practice. This is consistent with previous studies in healthy controls (Classen, Liepert, Wise, Hallett, & Cohen, 1998; Jax & Rosenbaum, 2007) and in individuals post-stroke (Park, et al., 2022) showing that repetition of a specific movement or muscle activation pattern induced by the application of external perturbation or the presence of obstacles might affect movement pattern or muscle activation after the removal of the perturbation/obstacle. It is possible that the neuronal network mediating trunk movement might be optimized with the repetitive movement of improved trunk motor performance during the adaptation period, and use-dependent motor learning mechanisms may be involved in this process (Diedrichsen, White, Newman, & Lally, 2010).

The application of continuous trunk support force may induce learning of gait phase dependent muscle activity of RA and OBL in children with CP, which may be responsible for the improvement in trunk postural control. In this study, we observed increased muscle activity of both RA and OBL during the post-adaptation period. In particular, as shown in Figure 3, we found that participants showed an improved phasic EMG activation pattern in RA and OBL across gait cycle during the post-adaptation period. One potential mechanism is that the application of trunk support force might elicit enhanced activation of trunk muscles, such as RA and OBL, to counteract the supporting force. In particular, because the magnitude of the supporting force was varied, depending on the displacement of the trunk in the frontal plane, across a gait cycle, the magnitude of the counteracting force might also change, accordingly, resulting in a phase dependent enhancement in the magnitude of the trunk muscle activity of RA and OBL. Over time, the continuous activation of enhanced phase dependent muscle activity of the RA and OBL might induce a retention of enhanced phasic muscle activity of RA and OBL during the post-adaptation period, which may be achieved through use-dependent motor learning mechanisms (Diedrichsen, et al., 2010). In particular, a phasic increase in activity of trunk muscles, which may be a more effective and efficient pattern of postural muscle control, rather than a constant excessive increase, which may create a functionally rigid trunk (Prosser, Lee, Barbe, VanSant, & Lauer, 2010), might induce an improvement in trunk postural control in children with CP.

There are two potential neural pathways that may be involved for the improved trunk postural control in children with CP observed in this study. On one hand, the application of varied trunk support force might induce greater head movement (both the magnitude and its variability) in the frontal plane (i.e., sideway head tilt or horizontal head displacement), which was inferred by the observation of increased trunk obliquity angle during the adaptation period given the close connection between the trunk and the head (Figure 2a). The vestibular system may rapidly sense linear and angular changes in the movement and orientation of the head, e.g., the otolith organs sense linear acceleration and semicircular canals sense angular motion of the head (MacKinnon, 2018). Vestibular afferents from the otolith organs and semicircular canals may project to the vestibular nuclei and further to the thalamus and cerebellum for interpretations in the higher brain center, and also further to the spinal cord via the vestibulospinal tracts, which modulate postural adjustments to the increased trunk obliquity (MacKinnon, 2018). On the other hand, the enlarged movement of the trunk and head, induced by the application of trunk support force, may be sensed by the proprioceptive and cutaneous receptors. For instance, muscle spindles may detect lengthened extrafusal muscle fibers of the trunk and neck which was induced by enlarged trunk and head movement (Ergen & Ulkar, 2007). Golgi tendon organs in trunk muscles may detect the load related information applied to the trunk muscles through group II and Ib afferents. In addition, the enlarged trunk movement might also induce changes in the distribution of pressure beneath the foot, which may be detected by cutaneous receptors on the sole of the feet (Dietz & Duysens, 2000). Afferents from the proprioceptive and cutaneous receptors may project to the spinal cord and give off three collateral branches. One of them terminates in the spinal cord to facilitate postural reflexes (MacKinnon, 2018).

The improvement in trunk postural control may facilitate walking in children with CP. In this study, the application of trunk support force might help participants to stabilize their trunk movement during the adaptation period. Thus, the improvement in trunk motor performance might facilitate establishing a more stable pelvis movement, particularly for the more affected side, which may allow the participants to take a longer step on the less affected leg, i.e., the contralateral leg. In contrast, a deterioration in trunk postural control may compromise leg motor performance. For instance, during the time period of early post-adaptation, the release of trunk supporting force might temporarily compromise the trunk motor performance because participants might rely on the support force to stabilize their trunk movement, resulting in a reduced step length on the less affected side.

In addition, the improvement in trunk motor control might transfer to overground walking, resulting in an improvement in overground walking speed. This is in line with findings from previous studies in individuals post stroke (Park, et al., 2022) and people with incomplete spinal cord injury (Lin, et al., 2019; Yen, Schmit, Landry, Roth, & Wu, 2012), which suggested that adapted motor control that was developed during treadmill walking practice might partially transfer to overground walking. One potential mechanism is that some neural circuits that control trunk movement during treadmill and overground walking contexts may be partially overlapped. Therefore, motor skills obtained from treadmill walking may partially transfer to overground walking, although there are some differences in environmental contexts between treadmill and overground walking (Stolze, et al., 1997). The increases in step length and cadence after 10 minutes post treadmill walking might lead to the improvement in overground walking speed (i.e., 0.15 m/s ± 0.34 m/s). The increase in walking speed at 10 minutes post treadmill walking rather than immediately after treadmill walking may be due to the fatigue caused by treadmill walking, which was consistent with a previous study in individuals post-stroke (Park, et al., 2020). In particular, participant’s GMFCS level had an impact on the gains in walking speed with these less impaired participants (i.e., GMFCS level I) also showed less functional gains. One possible reason may be due to the ceiling effect of improvements in trunk postural control in these children with GMFCS level I who show fewer deficits in trunk postural control, i.e., larger deficits in trunk postural control were found in more severely impaired children (Heyrman, et al., 2013).

Clinical applications and limitations

Results from this study may have potential clinical applications for improving trunk postural control and walking function in children with CP. For instance, our results indicate that providing lateral trunk support force using a passive elastic device during treadmill walking may induce improvement in trunk postural control and overground speed (i.e., 0.15m/s) in children with CP. In addition, this approach may be more beneficial for these children with GMFCS II than those with GMFCS I. These findings may be used by researchers/physical therapists to develop new approaches or new studies to determine whether functional improvement in trunk postural control will facilitate walking in children with CP.

This study has the following limitations. First, the sample size was small and the distribution of the GMFCS levels was not even. Future work should include more subjects, especially those at GMFCS level III, to test this hypothesis. Second, the group assignment for the experimental condition and the control condition was not randomized but in a sequence order (i.e., we recruited the first 14 participants in the experimental group and then recruited additional 5 participants in the control group, although the sample size was small). However, the goal of this study was to determine whether children with CP would adapt to the continuous lateral trunk support force. In addition, the increase in overground walking speed after treadmill walking only, i.e., the control group, was 0.02m/s, which was comparable to results from a previous study indicated that the increase in overground walking speed after 6 weeks of treadmill training was 0.04m/s (Wu, Kim, Arora, Gaebler-Spira, & Zhang, 2017), suggesting that the increase in walking speed in the experimental group (0.15m/s) might not be only due to the treadmill walking only. Thus, the improvement in trunk postural control might have partial contributions to the increase in overground walking speed, although we do not have a firm conclusion due to the lack of randomization of testing order, particularly due to the small number of sample size in the control group. Thirdly, we did not record muscle activity during overground walking. We do not know whether the enhanced activation of trunk muscles might transfer treadmill to overground walking. Fourth, we were not able to assess the trunk postural control using a clinical scale, such as trunk control measurement scale, due to time constraints of the protocol. Lastly, for safety, participants were allowed to hold onto handrails. However, if one participant preferred to hold onto the handrails, the participant was required to hold onto the handrails for all testing sessions.

In conclusion, the application of continuous lateral trunk support force during treadmill walking is likely to induce improvement in trunk postural control in children with CP, although we do not have a firm conclusion due to the small sample size of participants from the control group. The improvement in trunk postural control might transfer from treadmill to overground walking, resulting in improvement in overground walking speed, although we do not have a firm conclusion due to the lack of randomization of the order of condition. Results from this pilot study may provide insights for the development of new methods for improving trunk postural control and walking function in children with CP, although further studies are needed.

Highlights.

Continuous lateral trunk support force during walking is likely to induce improved trunk postural control in children with CP.

Enhanced trunk postural control may transfer to overground walking, resulting in an improvement in overground walking speed.

Targeting improvement in trunk postural control may potentially improve walking function in children with CP.