Effect of the Kickstart exoskeleton lower extremity walking system on improving lower extremity walking ability in subacute stroke patients: a randomized controlled trial

Abstract

Background

Walking dysfunction is a primary cause of a reduced ability to perform activities of daily living and decreased quality of life in stroke patients. The Kickstart^® Walk assist system is portable and easy to don and remove. There is a lack of high-quality, randomized controlled trials to validate its effectiveness. The aim of this study was to evaluate the effectiveness of the Kickstart^® Walk Assist system in improving lower limb muscle strength and walking ability in stroke patients.

Methods

Forty-six patients were enrolled and randomly assigned to either a control group (n = 23) or a Kickstart group (n = 23). Both groups received conventional rehabilitation therapy. In addition, patients in the Kickstart group wore the Kickstart^® Walk Assist system for 20 min, and patients in the control group received walking training for 20 min. The outcome measures included the Fugl-Meyer Assessment of Lower Extremity Motor Function (FMA-LE), gait parameters, the 10MWT, the Borg Subjective Fatigue Scale (Borg), and surface electromyography (sEMG).

Results

Compared with the control group, the Kickstart group showed more significant improvements in FMA-LE at 4 and 8 weeks (P = 0.025, P = 0.028), 10-MWT (P = 0.256), Borg at 8 weeks (P = 0.035), sEMG (P < 0.05), and gait parameters (P > 0.05). No adverse events were observed during or after the intervention.

Conclusion

The Kickstart^® Walk assist system can increase stroke patients’ lower limb strength and motor function and improve their walking ability.

Trial registration

This study was registered in the Chinese Clinical Trial Registry (Unique Identifier ChiCTR2300067605) on January 13, 2023.

Introduction

Stroke is characterized by high morbidity, disability, and mortality [1] and high medical resource costs [2]. Some studies have shown that approximately 75% of stroke patients are over 65 years of age, with an increasing prevalence in younger populations [3]. Stroke patients often experience substantial residual functional deficits, with 70–80% facing temporary or permanent disabilities, including reduced muscle strength, balance, and walking ability [4, 5].

Compared with traditional stroke rehabilitation, lower extremity exoskeletons offer patients more consistent, repeatable, and controlled training. These devices reduce the treatment burden on therapists and provide stroke patients with weight-bearing support on the affected side, gait training, and assisted walking, making them effective rehabilitation treatments [6]. However, most exoskeleton robots are expensive, time-consuming to wear, need to be equipped with active power, and have high requirements for the site [7–9], which limits the application and promotion of lower limb rehabilitation robots for stroke patients. Therefore, developing a portable and easy-to-use lower limb rehabilitation exoskeleton would be highly important for clinical practice with stroke patients.

Early studies of lower limb rehabilitation robots aimed to provide reasonable weight support for patients. Researchers developed a pedal-type robot to enhance training functions on this basis [10]. But suffered from inadequate hip constraints, limiting natural gait replication, they lacked monitoring/assessment systems and failed to adapt to individual needs or simulate therapist techniques. However, end-effector robots remain valuable. They offer cost-effective solutions when simplicity is prioritized. For severely paralyzed patients focused on compensatory ambulation—not normalized gait—they provide functional independence. Rehabilitation goals should thus be personalized, with device selection matching patient-specific objectives. Bed-type rehabilitation robots [11] are designed for bedridden rehabilitation training in patients with long-term lower limb motor dysfunction. Utilizing the bed frame as a platform, these systems adjust the bed inclination angle to facilitate patient transition from a supine position to standing during rehabilitation. The primary objective is to mitigate or prevent complications. However, training with bed-type rehabilitation robots (e.g., Erigo) requires patients to maintain a supine position. Consequently, during hip and knee flexion-extension exercises, the lower limbs remain anterior to the body’s midline (sagittal plane). This postural constraint significantly restricts joint range of motion, particularly hip extension. Therefore, conventional bed-type rehabilitation systems often demonstrate constrained capabilities for comprehensive lower limb status evaluation and typically exhibit limited adaptability in dynamically adjusting support levels based on real-time patient biomechanical feedback.

Exoskeleton robots, a major branch of rehabilitation robotics, have attracted significant research interest in recent years, particularly for their potential to improve walking ability in patients with lower limb dysfunction. On the basis of the walking movement of a normal person, the exoskeleton rehabilitation robot can accurately detect the patient’s movement intention to realize the patient’s active and passive combination of training modes, which has been applied in the clinic. An Israeli research team [12] used the Lokomat^® exoskeleton robot for lower limb rehabilitation in 67 patients with subacute stroke and demonstrated that it significantly improved both walking function and National Institute of Health Stroke Scale (NIHSS) scores compared with conventional physical therapy. An Italian research team [13] used the Ekso exoskeleton robot to perform simultaneous lower limb rehabilitation in 12 subacute stroke patients and 11 chronic stroke patients, resulting in improvements in the Trunk Control Test (TCT) and Walking Impairment Questionnaire (WIQ) scores in subacute stroke patients but no significant changes in chronic stroke patients. However, the broader application of exoskeletal robots in rehabilitation therapy has been partially constrained by practical challenges such as time-consuming donning procedures, limited flexibility [14], and difficulties in developing patient-specific training protocols [15]. Louis et al. [16] developed a flexible robotic exoskeleton to address the characteristics of traditional exoskeletal robots, which are bulky and not easy to wear. This robotic exosuit improved the foot-plantar flexion angle of the swing phase while the patient was walking and improved the patient’s gait balance ability by improving the symmetry of the forward propulsive impulse. Compared with ankle‒foot orthoses (AFOs), which limit the range of motion of the ankle, impair forward propulsion during gait [14], and increase energy expenditure during gait [10], robotic exosuits are more energy-conserving and allow patients to walk longer.

Current research on lower limb robotics for stroke patients focuses on affordability [3, 17] and ease of use [15, 18]. There is limited research in China on portable, wearable exoskeleton robots, with most studies focusing on large, stationary models. Therefore, research in this area holds substantial value for clinical applications. This study aims to verify the effectiveness of a portable and easy-to-use lower extremity exoskeleton walking system (Kickstart^® Walk Assist system) in improving stroke patients’ lower extremity movement and walking ability, which has specific clinical research and application value.

Methods

Study design and participants

A total of 46 patients were randomly assigned to the control or Kickstart group, with 23 patients in each group. Both groups received 4 weeks of rehabilitation therapy intervention, and relevant indicators were assessed at baseline and at 2, 4, and 8 weeks after treatment. The treatment of the patients was carried out by two therapists who had been working for more than 5 years. Before the assessment, the senior therapist trained the assessing therapists to standardize the assessment indicators, criteria, and procedures. The flow chart is shown in Fig. 1.

Fig. 1.

Flow chart

A total of 46 stroke patients treated at Wuxi Central Rehabilitation Hospital from March 2022 to December 2023 were recruited. Table 1 shows the baseline characteristics of the patients. The inclusion criteria were as follows: met the diagnostic guidelines for stroke in the 2019 Chinese Classification of Cerebrovascular Diseases [19] and confirmed by cranial CT or MRI; first-ever onset of the disease, aged 35–75 years; duration of the disease between 1 and 6 months, unilateral hemiparesis; modified Ashworth scale (MAS) score ≤ 2 for the major muscle groups of the lower limbs; ability to walk independently or with assistance for > 10 m; range of motion of each lower limb joint on the affected side was basically normal or did not affect walking; no other factors affecting walking; Mini-Mental State Examination (MMSE) [20] score > 17; and informed consent. The exclusion criteria were as follows: unilateral neglect; poor general condition unsuitable for gait training; New York Heart Association (NYHA) functional classification ≥ III [21]; history of traumatic injury with unstable fracture; knee or hip flexion contracture > 15 degrees; multiple epilepsy; and participation in other experiments. The withdrawal criteria were as follows: in the opinion of the investigator, they needed to be withdrawn from the study; they had serious adverse events, developed serious complications, etc.; health might have been compromised by continuing the study; and the subjects were asked to withdraw. There was no statistically significant difference in the baseline demographic data between the two patient groups, as shown in Table 1.

Table 1.

Basic information of patients in the two groups

Variable	Control group(n = 23)	Kickstart group(n = 23)	t/χ2	P
Age(y)	61.83(12.298)	60.43(11.552)	0.395	0.694
Sex			0.107	0.743
Male, n(%)	17(73.9%)	16(69.6%)
Female, n(%)	6(26.1%)	7(28.3%)
Type of stroke			0.674	0.412
Ischemic, n(%)	18(78.3%)	21(91.3%)
Hemorrhagic, n(%)	5(21.7%)	2(8.7%)
Side of stroke			3.136	0.077
Left, n(%)	9(39.1%)	15(65.2%)
Right, n(%)	14(60.9%)	8(34.8%)
Assistance while walking			1.394	0.238
Yes, n(%)	10(43.5%)	14(60.9%)
No, n(%)	13(56.5%)	9(39.1%)
Duration(day)	44.39(29.022)	39.87(5.649)	0.546	0.588

Interventions

Both groups received conventional rehabilitation therapy, including specific physical, exercise, occupational, and traditional therapies. Exercise therapy consisted mainly of strength training of the lower limb and trunk muscle groups, passive stretching, and sit-to-stand transfer training. Each therapy session lasted 40 min and was conducted once daily, 5 days per week, over 4 weeks.

In the control group, gait exercise training, including one-legged standing training, step training, toe clearance training, lateral step training, walking training, and stair climbing training, was conducted 5 days a week for 20 min per session for 4 weeks.

In the Kickstart group, following routine physiotherapy and rehabilitation, patients in the Kickstart group underwent gait training via the Kickstart^® Walk Assist system (Real Star Rehabilitation, Shanghai, China). The Kickstart^® Walk assist system is a rehabilitation device consisting of a belt, an external support structure, and an Exotendon (Figs. 2 and 3). The effect of the Exotendon is similar to that of an artificial tendon, which stores energy during the stance phase and releases it during the swing phase of the gait cycle. The exotendon mechanism is inspired by the anatomy of horses’ hind limbs, where long tendons spanning multiple joints store energy during the stance phase. This stored energy is released during the swing phase, reducing the muscle exertion required for movement. The universal hip joint is designed to allow complex movement patterns of the pelvis and hip joints. The knee joint is designed to prevent hyperextension of the knee, and the ankle joint has a built-in ankle‒foot orthosis to help correct foot drop and foot inversion. The patient wore the Kickstart lower extremity exoskeleton for supervised gait training. The Exotendon of the exoskeleton’s tendons is adjusted through a tendon tension adjuster, allowing the patient to fully lift the lower limb off the ground and complete the step motion. Each session lasted 20 min once a day, 5 days a week for 4 weeks (Figs. 2 and 3).

Fig. 2.

Kickstart^® Walk assist system

Fig. 3.

Patient wearing Kickstart

Outcome measures

Participants were assessed at baseline (before randomization), at the 2-week and 4-week time points during the intervention, and at the 8-week follow-up, all by a blinded assessor. During testing, necessary assistance was provided based on the patient’s functional status. Assistance was administered by a single therapist positioned on the affected side of the patient, strictly limited to light touch contact (minimal fingertip contact for balance cueing) and guard contact (hand placement near torso/hips for fall prevention without weight-bearing support). Therapists were explicitly instructed to neither actively push patients nor provide forward propulsion assistance. The assistance method was kept consistent for each patient throughout the study to ensure intervention fidelity. During all gait assessments, walking aids (including canes and orthoses) were strictly prohibited, and critically, the Kickstart system was not worn during post-intervention evaluations.

Primary outcome

The primary outcome was motor function measured with the lower extremity Fugl-Meyer Motor Function Assessment Scale as the main index, as it has good predictability, credibility, and sensitivity for the recovery of lower limb motor function in stroke patients [22]. Higher scores on this 17-item scale, where each item is scored from 0 to 2 for 34 points, indicate better lower limb function.

Secondary outcomes

The secondary outcomes include gait indicators (gait speed, stride length, temporal symmetry ratio (TSR)), the 10-meter walk test, the Borg Subjective Fatigue Scale and surface electromyography (iEMG), and the root mean square (RMS) of the biceps femoris, rectus femoris, iliopsoas and gluteus maximus muscles.

The TSR analyzes the temporal distribution of swing and stance phases during gait, serving as a metric for gait symmetry assessment [23]. The TSR was calculated as follows: TSR = (paretic swing time/stance time)/(nonparetic swing time/stance time).

The 10-meter walk test is reliable for gait testing in stroke patients [24]. A 10-meter track was measured and marked on the floor with 2 m at each end to facilitate acceleration and deceleration. Patients were instructed to walk as quickly and safely as possible, and the time taken to walk 10 m was recorded. The measurements were repeated 3 times, and the results were averaged.

The gait data were collected via the Gait and Balance Function Training and Assessment System (AL-600). Patients walked barefoot and were allowed to practice on the test bed for 2 min to become familiar with it before the test. They were instructed to walk naturally, look straight ahead, with a therapist on each side to prevent falls. The measurements were repeated three times and averaged (Fig. 4).

Fig. 4.

AL-600 gait analysis system

The Borg Rating Scale of Perceived Exertion is a valid measurement tool for determining exercise intensity [25] and correlates well with heart rate and other physiological indicators. The scale ranges from 6 to 20 points, with higher scores indicating greater perceived fatigue. A score above 13 generally correlates with significant respiratory and fatigued symptoms, whereas a score above 17 suggests that exercise cessation may be needed [26].

Surface electromyography was performed using a BTS FREEEMG 300 (BTS Bioengineering, Milan, Italy). Before testing, the test site was fully exposed, the hair was removed, and the area was cleaned with 75% alcohol to optimize sensitivity. The two electrodes were placed 2 cm apart and attached to the muscle belly along the path of the muscle fibers. The patient was instructed to perform a maximal isometric contraction and hold it for 7 s. The measurement was repeated three times. The specific locations of the electrodes were as follows (Fig. 5): rectus femoris (a): midpoint between the ilium and the knee joint; biceps femoris (b): lateral to the midline of the posterior thigh, from the gluteal groove to the midpoint of the knee joint; iliopsoas (c): medial to the anterior superior iliac spine; and gluteus maximus (d): midpoint between the greater trochanter of the femur and the sacral vertebrae. The surface EMG data were automatically processed via BTS EMG analysis software. The surface electromyography data were automatically processed via BTS EMG analysis software. The calculation of iEMG involves several steps: First, EMG signals are recorded from the muscles of interest using surface electrodes. These signals are then filtered to remove noise and artifacts, typically using a band-pass filter to isolate the relevant frequency components. Next, the filtered EMG signal is rectified to convert all negative values to positive values, ensuring that the signal represents the total muscle activity. Finally, the rectified EMG signal is integrated over a specific time interval, either by numerical integration or by summing the absolute values of the signal.

Fig. 5.

Electrode position

Sample size calculation

Sample size calculations were conducted via Cohen’s method [27]; this study used G*Power software to estimate the sample size, the effect size was 0.25, the test criterion α was 0.05, the test power (1-β) was 0.95, the number of groups was 2, the number of repetitions was 4, the corr between rep measures was 0.5, and the nonsphericity correction was 1. The sample size was calculated to be 36 cases, considering the 20% attrition rate, and the final sample included 46 cases.

Randomization and blinding

Block randomization was performed via a computer-generated random sequence with random block sizes of 4–6. A total of 46 patients were randomly assigned to the control and Kickstart groups, with 23 patients in each group. Forty-six sealed envelopes, each labeled with a subject number, contained group assignments (Kickstart or control groups). A research team member who was not involved in treatment or evaluation safeguarded the envelopes to prevent premature opening. Once a subject met the inclusion criteria and consented to join the study, their numbered envelope was opened, and they were assigned to the intervention group indicated inside.

This study employed a single-blind design; the patients were assessed by two other therapists who were unaware of the grouping of the patients. During the assessment, the patient’s personal information (such as name and medical record number) was replaced with anonymous codes to reduce the evaluator’s recognition of the patient’s identity, thereby reducing potential bias.

Statistical analysis

All the statistical analyses were performed with SPSS version 21.0 (IBM, Armonk, NY, USA). Data that followed a normal distribution were described via the mean ± standard deviation (x ± SD), and for pre- and postmeasurement data, repeated-measures ANOVA was used with a prior sphericity test. If the sphericity test was not passed, the main effect and interaction effect of group and time were analyzed via Greenhouse correction, and then one-way repeated-measures ANOVA and paired samples tests were used to analyze the main effect and interaction effect of group and simple effects of time were analyzed via one-way repeated-measures ANOVA and paired samples test. Measures that did not follow a normal distribution were described by medians (quartiles); analyses were performed via generalized estimating equations for the main effects of group, time, and interaction effects; and later analyses were performed via Friedman’s test and the signed rank test for simple effects of group and time. The number of cases (proportions) determined via the chi-square test statistically described all the count data. p < 0.05 was considered a statistically significant difference.

Results

Exercise capacity

The FMA-LE group*time interaction was not significant (P = 0.134), but the Kickstart group was superior to the control group at 4 and 8 weeks posttreatment (P = 0.025, P = 0.028), comparative analysis revealed significantly greater improvement in ΔFMA-LE at 4 weeks for the Kickstart group compared to the control group (p = 0.017), and both groups showed improvements over pretreatment at all time points (P < 0.05). The 10-MWT group*time interaction was not significant (P = 0.256), but both groups showed significant improvements over pretreatment at all posttreatment time points (P < 0.05) (Tables 2, 3 and 4).

Table 2.

Changes in FMA-LE scores before and after treatment in the two groups

	Pre(W0)	Post-2w(W2)	Post-4w(W4)	Post-8w(W8)	W0 vs. W2	W0 vs. W4	W0 vs. W8
					P	P	P
Control group(n = 23)	18.48(5.53)	21.35(5.45)	23.61(5.12)	25.65(5.69)	0.001	0.001	0.001
Kickstart group(n = 23)	19.22(7.39)	22.96(6.22)	26.91(4.51)	28.91(3.85)	0.001	0.001	0.001
P	0.703	0.356	0.025	0.028
	F group = 2.316; F time = 74.882; F group*time = 2.230
	P group = 0.135; P time = 0.001; P group*time = 0.134

Table 3.

Differences in FMA-LE scores before and after treatment in the two groups

	Δ2	Δ4	Δ8	Δ2 vs. Δ4	Δ2 vs. Δ8	Δ4 vs. Δ8
				P	P	P
Control group(n = 23)	2(1,4)	5(2,7)	7(3,10)	0.001	< 0.001	0.001
Kickstart group(n = 23)	4(3,5)	8(5,11)	9(5,15)	< 0.001	< 0.001	0.001
Z	-1.552	-2.379	-1.508
P	0.121	0.017	0.131

Notes: Δ2: Change from baseline at Week 2; Δ 4: Change from baseline at Week 4; Δ8: Change from baseline at Week 8

Table 4.

Changes in the 10MWT before and after treatment in the two groups

	Pre(W0)	Post-2w(W2)	Post-4w(W4)	Post-8w(W8)	W0 vs. W2	W0 vs. W4	W0 vs. W8
					P	P	P
Control group(n = 23)	20.32(11.82,48.33)	15.28(10.22,46.58)	12.62(8.60,45.92)	10.24(7.57,34.12)	0.001	0.001	0.001
Kickstart group(n = 23)	30.94(15.38,49.30)	28.96(13.65,40.29)	24.54(13.35,33.65)	20.65(11.48,36.07)	0.003	0.001	0.001
P	0.210	0.203	0.227	0.073
	Waldχ2 group = 1.289; Waldχ2 time =  47.754; Waldχ2 group*time = 2.727
	P group = 0.256; P time = 0.001; P group*time = 0.256

Gait parameters

Gait speed, stride length, and the TSR group*time interaction were not significant (P = 0.858, P = 0.378, and P = 0.061, respectively). However, both groups’ gait speed improved significantly over pretreatment at 4 and 8 weeks posttreatment (P < 0.05). Stride length was greater than pretreatment length in the control group at 8 weeks posttreatment (P = 0.042) and in the Kickstart group at 2, 4, and 8 weeks posttreatment (P = 0.014, P = 0.003, P = 0.017). TSR was better than pretreatment in the Kickstart group at 4 and 8 weeks posttreatment, and there was no significant difference in the control group. The gait parameter changes did not significantly differ between the two groups (P > 0.05). See Table 5.

Table 5.

Changes in gait parameters before and after treatment in the two groups

	Pre(W0)	Post-2w(W2)	Post-4w(W4)	Post-8w(W8)	W0 vs. W2	W0 vs. W4	W0 vs. W8
					P	P	P
Speed(m/min)
Control group(n = 23)	28.93(17.25)	34.70(18.69)	39.35(19.77)	41.47(19.86)	0.057	0.002	0.001	F group = 1.590; F time = 24.233; F group*time = 0.153
Kickstart group(n = 23)	22.19(15.59)	29.21(19.31)	32.47(19.83)	33.92(21.48)	0.120	0.003	0.001	P group = 0.214; P time = 0.001; P group*time = 0.858
P	0.171	0.332	0.245	0.223
Stride length(cm)
Control group(n = 23)	53.53(20.36)	57.75(19.77)	59.19(17.88)	63.48(18.83)	0.568	0.337	0.042	F group = 0.4829; F time = 11.353; F group*time = 1.002
Kickstart group(n = 23)	46.69(18.85)	55.51(20.56)	58.09(23.29)	57.86(26.34)	0.014	0.003	0.017	P group = 0.491; P time = 0.001; P group*time = 0.378
P	0.243	0.708	0.858	0.409
TSR(%)
Control group(n = 23)	1.84(1.13,2.38)	1.32(0.80,2.06)	1.56(1.01,1.91)	1.34(1.11,1.83)	0.162	0.465	0.248	Waldχ2 group = 0.787; Waldχ2 time = 3.685; Waldχ2 group*time = 5.579
Kickstart group(n = 23)	1.51(1.21,3.93)	1.63(1.34,2.67)	1.18(1.03,1.99)	1.21(1.06,2.07)	0.648	0.019	0.004	P group = 0.375; P time = 0.158; P group*time = 0.061
P	0.538	0.033	0.583	0.767

The Borg rating scale of perceived exertion

There was a group*time interaction effect for Borg (P = 0.028), with a more significant reduction in perceived exertion in the Kickstart group than in the control group at 8 weeks (P = 0.035). Compared with the pretreatment group, both groups presented significant reductions (P < 0.05) (Table 6).

Table 6.

Changes in Borg scale scores before and after treatment in the two groups

	Pre(W0)	Post-2w(W2)	Post-4w(W4)	Post-8w(W8)	W0 vs. W2	W0 vs. W4	W0 vs. W8
					P	P	P
Control group(n = 23)	14.43(1.93)	13.35(1.87)	12.13(2.60)	11.52(2.13)	0.001	0.001	0.001
Kickstart group(n = 23)	14.48(3.03)	12.91(2.83)	11.87(2.85)	10.04(2.46)	0.001	0.001	0.001
P	0.954	0.542	0.747	0.035
	F group = 0.635; F time = 77.822; F group*time = 3.447
	P group = 0.430; P time = 0.001; P group*time = 0.028

Surface electromyography

Rectus femoris muscle

There was a group*time interaction effect for RMS (P = 0.001), which was greater in the Kickstart group than in the control group at 4 and 8 weeks after treatment (P = 0.025, P = 0.016). While the iEMG group*time interaction was not significant (P = 0.116, P = 0.408), both groups showed improvements over pretreatment at each posttreatment time point (P < 0.05). See Fig. 6.

Fig. 6.

Changes in sEMG signals in the biceps femoris between the two groups before and after treatment, *p < 0.05

Biceps femoris

The iEMG and RMS group*time interaction effects were not significant (P = 0.151, P = 0.128). The iEMG and RMS were better in the Kickstart group than they were before treatment at every time point after treatment (P < 0.05), and there was no significant change in the control group (P > 0.05). See Fig. 7.

Fig. 7.

Changes in the sEMG signals of the rectus femoris between the two groups before and after treatment, *p < 0.05

Iliopsoas muscle

There was a group*time interaction effect for iEMG and RMS (P = 0.005, P = 0.002), with both groups showing improvement over pretreatment at all posttreatment time points (P < 0.05). The iEMG and RMS were significantly better in the Kickstart group than in the control group at 4 and 8 weeks posttreatment (P = 0.014, P = 0.020 for iEMG; P = 0.016, P = 0.019 for RMS). See Fig. 8.

Fig. 8.

Changes in sEMG signals in the iliopsoas between the two groups before and after treatment, *p < 0.05

Gluteus maximus muscle

The iEMG and RMS group*time interaction effects were not significant (P = 0.069, P = 0.186) but were better than those before treatment at each posttreatment time point in both groups (P < 0.05). After 4 weeks of treatment, the iEMG was better in the Kickstart group than in the control group (P = 0.030). See Fig. 9.

Fig. 9.

Changes in sEMG signals in the gluteus maximus between the two groups before and after treatment, *p < 0.05

Discussion

By comparing the use of the Kickstart^® Walk Assist system with conventional gait training in stroke patients, this study aimed to investigate the beneficial effects of the Kickstart system in improving patients’ walking ability and lowering limb strength. Compared with the control group, following a 4-week intervention and an 8-week follow-up, the Kickstart group showed significant improvements in lower limb motor function (FMA-LE) and subjective fatigue (Borg scale), particularly in muscle strength of the iliopsoas and rectus femoris. However, no significant between-group differences were observed in gait parameters (speed, stride length, TSR), as both groups exhibited similar improvements over time. These findings suggest that the Kickstart system primarily enhances neuromuscular control and reduces fatigue, while gait improvements may derive from conventional therapy shared by both groups.

Consistent with our findings, the Guidelines for Stroke Rehabilitation and Recovery in Adults, published by the American Heart Association and the American Stroke Association [4], state that robot-assisted rehabilitation can be combined with conventional rehabilitation to improve motor function and mobility in stroke patients. When walking with the Kickstart^® Walk Assist system, the affected side passively stretches the extensor tendon during the support phase by extending the hip and knee, thereby storing energy. This stored energy is then released in the swing phase, assisting hip flexion and lower limb elevation, allowing a more effective forward swing of the affected side. As a result, the propulsive power of the limb on the hemiplegic side is improved during walking, and the demand for muscle strength on the hemiplegic side is greatly reduced [17], improving the patient’s ability to walk. These findings align with our study results, in which the FMA-LE and 10-MWT scores improved following training with the Kickstart^® Walk Assist system, indicating enhanced lower limb motor function and walking ability. The minimal clinically important difference (MCID) denotes the slightest change in treatment outcomes perceived as beneficial and meaningful by patients or clinicians, serving as a quantifiable metric to assess the significance of clinical improvement. A research investigation into the MCID for the Fugl-Meyer assessment of lower extremities in individuals who have experienced a stroke revealed that the MCID for the FMA-LE is 6 points [28]. Both groups achieved clinically meaningful improvements in FMA-LE (> 6 points) at 8 weeks post-treatment (Control: Δ7.17; Kickstart: Δ9.69), indicating that intensive rehabilitation contributed substantially to functional recovery. However, the Kickstart group reached this threshold earlier (Δ6.69 at 4 weeks vs. Control Δ5.13) and maintained greater improvement magnitude (Δ9.69 vs. 7.17 at 8 weeks, P = 0.028). Comparative analysis revealed significantly greater improvement in ΔFMA-LE at 4 weeks for the Kickstart group compared to the control group, whereas no statistically significant between-group difference was observed at 8 weeks. This suggests that the Kickstart system may accelerate recovery within the context of conventional therapy, rather than replace its effects. Research has established that an increase of 0.2 m per second in walking speed constitutes a more rigorous MCID standard for stroke patients [29]. Notably, both groups achieved a clinically meaningful MCID improvement in the 10-MWT following 8 weeks of treatment. In robotic gait training for lower limb rehabilitation, the exoskeleton provides additional support to reduce the weight load on the lower limb. The induced plastic changes in gait patterns could be unconsciously enhanced by highly repetitive movements and obvious proprioceptive and somatosensory feedback under the correct movement pattern [30, 31] to improve the patient’s walking ability. At the 8-week posttreatment follow-up, patients still showed good improvement in motor function, which is consistent with the idea that repetitive [32, 33], high-intensity, intensive therapy in the early stages of stroke leads to sustained recovery of patient function on the basis of functional relearning and may also be related to ongoing structural and functional reorganization in the ipsilateral or contralateral neural network [34]. Lin et al. [35] conducted a 4-week intervention and 3-month follow-up in stroke patients and reported that robot-assisted gait training significantly improved patients’ lower limb motor function in the long term. Consistent with recent systematic review [36], rehabilitation robots can significantly improve lower limb motor function, balance capacity, and postural stability, thereby enhancing patients’ independent ambulation ability. Although not directly measured, our observed improvements in FMA-LE (Δ > 6 points, exceeding the MCID), gait speed (> 0.2 m/s MCID) represent established predictors of independent community ambulation.

Compared with pretreatment patients, patients showed significant improvements in step speed, stride length, and the TSR after 4 weeks of training with the Kickstart^® Walk Assist system. These results are consistent with the findings of Li et al. [37], who reported that 4 weeks of training with a lower limb rehabilitation robot improved lower limb motor function, step speed, and stride length in stroke patients. Previous studies have shown that when patients use an exoskeleton robot to assist in walking, their gait is closer to normal, reducing balance dysfunction due to lower limb strength deficits [38]. Slower walking speeds can lead to gait asymmetry in stroke patients [39]. Stroke patients usually suffer from insufficient forward propulsion, reduced muscle strength, and bilateral asymmetry in stride length, resulting in slow gait speed, asymmetry, and reduced gait stability [40, 41]. Patients’ gait is characterized by a reduced range of motion of the lower limb joints in the swing phase and a reduced ability of the affected limb to swing forward, resulting in prolonged support time on the unaffected side and a prolonged swing phase on the affected side [42]. Allen et al. [43] suggested that increasing the hip flexion moment on the hemiplegic side prior to swing may improve gait symmetry and increase a patient’s stride speed and stride length, which is one of the design principles of the kickstart exoskeletal lower limb walking system. However, there was no statistically significant difference in the improvement of walking speed between the two groups.The absence of significant between-group differences in gait speed improvement may be attributed to the following factors. The Kickstart system primarily assists hip flexion during swing phase via exotendon energy release, whereas walking speed depends more critically on ankle plantarflexion propulsion during late stance. Since the device’s ankle-foot orthosis (AFO) component immobilizes the ankle to correct foot drop, it may limit push-off force generation—a biomechanical trade-off for stability. The control group had higher baseline gait speed, potentially allowing more ‘room for improvement’ in speed metrics. Notably, both groups achieved comparable clinically meaningful gains (MCID > 0.2 m/s), indicating that conventional training sufficiently addressed speed deficits.

RMS denotes the root mean square value of the instantaneous electromyogram amplitudes of all muscles engaged in an activity over a specified duration, indicating the intensity and quality of muscle activity. Conversely, iEMG represents the cumulative discharge from all muscles involved in the activity over the same period, facilitating the evaluation of muscle fatigue and the intensity of muscle activity. The surface EMG results revealed significant group*time interactions, with iEMG and RMS values of the hip and knee joint flexor and extensor muscle groups showing improvements compared with pretreatment values. The rectus femoris and iliopsoas muscles demonstrated the most significant improvements, reflecting enhanced lower limb strength. The reduction in the Borg score in the Kickstart group was also greater than that in the control group, indicating continuous relief of muscle fatigue, reduced subjective fatigue during walking, and improved walking ability. Assistive exoskeletons are more efficient at saving walking metabolic expenditure in hemiplegic patients because of the lack of ankle plantarflexion strength and the need for hip flexion strength to complete the lower limb swing during walking [44]. Therefore, exoskeleton-assisted rehabilitation training is more appropriate for people with muscle weakness and stroke patients with better underlying conditions. Compared with cable-operated hip exoskeleton robots, which can provide only unidirectional, low traction forces [45], the external tendons of the Kickstart lower limb exoskeleton walking system can provide stronger assistance forces for hip flexion. A lower limb rehabilitation robot significantly improved hip and knee-related muscle strength in stroke patients in the subacute phase [18]. In stroke patients, a reduction in the hip flexion moment on the affected side during exercise can lead to fatigue, and fatigue can also affect muscle strength in the affected limb [46].

The propulsive force of the lower limb during gait comes mainly from the reaction force of the ground when the ankle is plantarflexed. Owing to weak ankle plantarflexion and increased calf triceps tension in some stroke patients, the ankle joint was fixed in the design of the Kickstart^® Walk Assist system. Under these conditions, propulsion primarily relies on knee extension and hip extension. It has been concluded that the knee extension moment has a facilitating and complementary relationship with the propulsive force [47], and the hip extension angle caused by the hip extensor muscle groups is one of the primary sources of propulsive force increase [48]. With repetitive and continuous robot-assisted exercise training, improving patients’ lower limb strength increases propulsive force, allowing positive changes in walking ability, gait parameters, and lower limb strength.

Mental health problems after stroke are among the most common problems associated with stroke [49]. Psychological studies have shown that rehabilitation can improve the quality of life of stroke patients and thus improve their mental health [15], and walking has a positive effect on alleviating mental health problems such as anxiety and depression [50]; thus, achieving the ability to walk independently is key to the mental health of stroke patients and their return to life and society. The Kickstart^® Walk assist system assists stroke patients in achieving more extraordinary walking ability, which fully mobilizes the patients’ rehabilitation initiative, increases their confidence in rehabilitation, and contributes to the patient’s rehabilitation process.

The main limitation of this study is its limited sample size, necessitating more extensive randomized controlled trials to further investigate the effects of the Kickstart^® Walk Assist system on functional improvement in stroke patients. Second, the present study included only 4 weeks of follow-up. Although it illustrates, to some extent, the long-term effects of the Kickstart^® Walk Assist system on gait, lower limb function, and muscle strength in stroke patients, a more extended follow-up period is needed to confirm the durability of the treatment effects. Third, while the Kickstart group showed earlier and greater FMA-LE improvement, the control group’s comparable 8-week outcome (Δ > 6 points) underscores that high-frequency conventional training remains foundational. Future studies should isolate device effects through dose-matched designs. In addition, the outcome measure of this study was peripheral motor function, which does not include the assessment of neurophysiological changes in patients, and the effect of the Kickstart^® Walk Assist system on neural remodeling in stroke patients through the periphery should be further investigated to elucidate its mechanism of action. We recognize that patient compliance and intervention intensity differences can affect the study results. Therefore, future research will evaluate compliance via both quantitative and qualitative methods, such as therapy frequency, duration, and patient feedback, to better assess the practical impact of the Kickstart exoskeleton for lower limb walking assistance.

Conclusion

In this study, stroke patients received the Kickstart^® Walk Assist system and conventional walking training, and lower limb motor function, surface electromyography, subjective fatigue, and gait parameters were analyzed through 4-week interventions and 8-week posttreatment follow-up. We found that the Kickstart^® Walk Assist system can enhance forward propulsion by improving lower limb strength in stroke patients, improve gait symmetry and balance, improve patients’ lower limb motor function and walking ability, and have long-term therapeutic effects on patients recovering from stroke. While our study did not directly assess psychological health, enhancing patients’ walking ability with the Kickstart exoskeleton increased their confidence and psychological well-being, which is crucial for holistic stroke recovery.

Abbreviations

FMA-LE

Fugl-Meyer assessment of lower extremity motor function score

sEMG

Surface electromyography

RMS

Root mean square

iEMG

Integrated electromyography

10-MWT

10-MWT

Borg

Borg Subjective Fatigue Scale

NIHSS

National Institute of Health Stroke Scale

TCT

Trunk Control Test

WIQ

Walking Impairment Questionnaire

AFOs

Ankle‒foot Orthoses

MAS

Modified Ashworth scale

MMSE

Mini-Mental State Examination

NYHA

New York Heart Association

TSR

temporal symmetry ratio

MCID

Minimal clinically important difference

Author contributions

CPL and YQL conceived and designed the study; YS and PC carried out the investigation and assessment; JYY performed the study and the formal analysis; CLW and YTZ contributed to the original draft; XWS and CY reviewed and revised the manuscript. All the authors read and approved the final manuscript. CPL, CLW and JYY contributed equally to this work and should be considered co-first authors.

Funding

This study was supported by the National Clinical Medical Research Centre Cultivation Program of Nanjing (No. 303103136AA22), major sports research projects of Jiangsu Sports Bureau (No. ST242102) and Jiangsu Province Hospital clinical diagnosis and treatment of technological innovation “Open bidding for selecting the best candidates” project (No. JBGS202414).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

All participants provided written, informed consent for study participation. The study was performed according to the Declaration of Helsinki and was approved by the Research Ethics Committee, Wuxi Mental Health Center/Wuxi Central Rehabilitation Hospital (NO. WXMHCIRB 2022LLky037).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.