Comparison of multi-planar and sagittal-plane stepping machines for walking and balance restoration in chronic stroke: a randomized control trial (RCT)

Abstract

Balance deficits are a common consequence of stroke, increasing the risk of falls. The Pinnacle Trainer (PT), which features a multi-planar exercise trajectory, has been shown to significantly activate hip abductors—key muscles for lateral stability. The elliptical trainer (ET), which simulates gait-like movement, is another commonly used rehabilitation tool. Both may offer viable options for gait training in individuals with chronic stroke. This study investigated the intervention effects of PT and ET on walking and balance abilities in individuals with chronic stroke. Thirty-six individuals with chronic stroke were randomly assigned to one of three groups: Pinnacle Trainer group (n = 12), ET group (n = 12), and control group (n = 12). Each group participated in an 8-week intervention program. The 6-minute walk test, 10-meter walk test, and the center of pressure (COP) displacements during obstacle crossing were measured as outcome measurements. The assessors (one therapist and one biomechanist) were blinded to the participants’ group assignments. All groups demonstrated significant improvements on the walking ability. Compared to the ET and control groups, the PT group showed significant improvements in mediolateral COP displacement, indicating enhanced balance and gait performance. These results support the integration of PT exercises into stroke rehabilitation programs targeting functional balance and mobility.

Introduction

Stroke is a common cause that frequently results in lasting disability^1,2. Individuals who have had a stroke may have hemiplegia and lose their ability to walk, which can cause balance deficits and increase the risk of falls^3–5. Individuals with stroke may have increased postural sway due to a shift in the mean position of the center of pressure (COP) from the affected leg toward the unaffected leg during a quiet stance⁶. Furthermore, stroke survivors who have experienced falls have been found to exhibit greater postural sway than those who have not experienced falls⁷. Thus, improving balance is a crucial rehabilitation goal.

During the swing phase of gait, the activation of the gluteus medius (GM) prevents the hip from dropping; that is, the hip abductor plays an important role during walking to stabilize the pelvis and increase lateral stability to prevent falls. However, impairment of hip abductor activation on the paretic side during weight shifting tends to occur even after walking ability has been restored following a stroke⁸. Hence, an effective and safe training approach for regaining hip abductor muscle strength and re-learning complex balance control is important for individuals to prevent falls after a stroke. Rehabilitation that focuses on task-specific training is essential, as motor skills are best acquired and retained when practicing the target movement itself⁹. The effectiveness of training is further enhanced by integrating variability of practice, which performing the target skill under varying conditions to promote greater adaptability and generalization of the learned motor patterns¹⁰. Beyond simple muscle activation, effective training must adhere to motor learning principles that engage the central nervous system in processing, refining, and automating movement, which typically involves repetition, feedback, and challenging the limits of stability¹¹.

A pinnacle trainer (PT) is a type of stair-climbing-like exercise equipment with bi-planar exercise trajectories where the pedals move upward/downward and medially/laterally during stepping. In other words, users perform joint movement of the lower limbs in the sagittal plane and frontal plane. The bi-planar movement and need for continuous lateral stability inherent to the PT’s motion inherently introduces a higher degree of practice variability and significantly greater demand on frontal plane control compared to typical sagittal-plane exercises. Moreover, exercise with a PT produces significant activation of hip abductors¹², which are essential for maintaining balance during walking or functional balance tasks¹³. PT intervention might thus be a feasible approach for stroke survivors to restore balance ability by combining a strong biomechanical rationale with enhanced motor learning opportunities through variable practice. An elliptical trainer (ET) produces a gait-like exercise trajectory, primarily involving motion in the sagittal plane, which is similar to that in task-orientation training. Exercise with an ET produces significantly lower impact forces on the joints compared to those produced by walking¹⁴. The usability and effectiveness of ETs in restoring walking ability have been reported for individuals with hemiplegia, hip fracture, and osteoarthritis^15,16. Thus, the use of an ET for walking restoration might be a feasible approach for stroke survivors.

Both trainers could be feasible options with a relatively low impact force for gait training than level walking¹⁷, besides, they both incorporate motor system and suspension system; thus, these rehabilitation systems could be used for those who are unable to exercise due to balance deficits or muscle strength deficits. However, to our knowledge, there have been no comprehensive comparisons of the intervention effects of these two types of exercise equipment on walking and balance restoration in individuals with chronic stroke. Therefore, the purpose of this study was to compare the effects of PT and ET training on walking and balance abilities in individuals with chronic stroke. The results of current research could provide suggestions for choosing proper exercise machines for specific groups or training purposes. Based on the increased hip abductor activation observed with PT in the previous study¹⁸, we hypothesized that both PT and ET would improve walking ability, but only PT would lead to greater improvements in balance performance.

Methods

Participants

The required sample size was estimated using G*Power version 3.1, based on data from our pilot study. Assuming an effect size of f = 0.4 (large effect), a statistical power of 0.80 and an alpha level of 0.05 for F-tests, the analysis indicated that a minimum of 34 participants would be needed. This study was a parallel group design, with baseline measurement and post-intervention measurement. Simple randomisation using a drawing lots method to allocate participants. Thirty-six individuals with chronic stroke participated in this study. Participants were allocated into the PT group, ET group, or control group using sealed envelope, ensuring an equal and independent chances of assignment. The protocol was approved by the Institutional Review Board of National Cheng Kung University Hospital (IRB no: B-ER-101-059). All methods were performed in accordance with the relevant guidelines and Regulations. All participants signed an informed consent form before the assessments and intervention sessions. The study has been registered in the ISRCTN registry (no. ISRCTN 27645313) on 22/06/2021. All measurements and intervention protocols were conducted in Taiwan.

The CONSORT diagram is shown in Fig. 1. The inclusion criteria were as follows: unilateral hemiplegia occurred after the first stroke, onset time was at least 6 months prior to the study a Brunnstrom stage of above 3, a Mini-Mental Status Examination (MMSE) score of above 24, a functional ambulation category (FAC) score of at least 2, and the ability to stand for at least 10 s with or without assistance.

Fig. 1.

CONSORT diagram for this study.

Interventions

The participants in the PT and ET groups underwent 8-week PT and ET interventions, respectively, in conjunction with conventional therapy. The intervention sessions were conducted three days a week for about 60 min per session. The participants were instructed to exercise on either the PT (S776MA, SportsArt, Taiwan) (Fig. 2A) or ET (E870, SportsArt, Taiwan) (Fig. 2B) for 30 min under the guidance of an experienced therapist and follow by a 30-minute conventional therapy. The purpose of the intervention was to practice correct movement patterns rather than muscle strengthening or cardiopulmonary improvements; hence, the resistance was set at the lowest level for both the PT and ET at the beginning of the intervention and the stepping speed was controlled by the participants. Certified therapist adjusted the stepping number according to the movement correctness. For example, if the participants were breathing heavily and could not maintain their bodies in an upright position during stepping or exhibited irregular movements when stepping, the therapist corrected the movement and decreased the stepping number in a given set. The therapist followed this guideline to train the participants and increased the stepping number gradually when the participants could maintain their posture while stepping. These two machines have incorporated with the suspension system to provide safety while stepping. In addition, both the PT and ET devices were equipped with a motor-assisted mode that helped participants initiate stepping movement during the early phase of training. To ensure safety and allow participants to become familiar with the exercise pattern, the motor-assisted mode was used during the first week of training. From the second week onward, all participants performed active stepping without motor assistance under therapist supervision. Each participant initially completed three sets of 30 stepping repetitions per session using the assigned device (PT or ET). The training volume was progressively increased to five sets of 30 repetitions per session after four weeks of intervention, according to each participant’s tolerance and performance. The stepping speed was self-paced to ensure safety. The therapist supervised exercise posture throughout training session. The number of steps was increased in subsequent sessions when participants were able to complete all sets without compensation movement or imbalance. Typical increments ranged from approximately 10–20% depending on individual tolerance and therapist judgement.

Fig. 2.

Rehabilitation systems. (A) Pinnacle trainer (PT) and (B) Elliptical trainer.

The participants in the control group had an 8-week conventional therapy intervention that included physical therapy and overground gait training three days a week for about 60 min per session.

To ensure comparability, the control and experimental interventions were similar in intervention duration, frequency and therapist involvement.

Outcome measurements

The outcome measurements of this study were divided in primary and secondary measurements. All assessments were performed by an experienced therapist and a biomechanist before and after the 8-week intervention. The assessors (one therapist and one biomechanist) were blinded to the participants’ group assignments.

Walking ability

Primary outcome measurements included a 6-minute walking test, a 10-m walking test, and obstacle crossing, which were used to assess gait and balance. Walking speed is an important indicator of walking capacity¹⁹. These tests were performed to assess walking ability. In the 6-minute walking test, the participants were requested to walk as far as they could in 6 min in each trial. The distance they walked was recorded; a longer distance indicated better walking endurance. The 10-m walking test was conducted three times. For this test, the participants were requested to walk as fast as they could in a safe manner for 10 m. The time required to complete this task was recorded; a shorter time indicated a faster walking speed.

Balance ability

The participants were requested to cross a 5-cm-tall obstacle and to perform at least three successful trials. A 5-cm-tall obstacle crossing test is a reliable test (Intra-Class Correlation Coefficient: 0.916)²⁰, besides, the sensitivity and specificity of this test were (0.929 and 0.933, respectively) for people with stroke²⁰. The forces during obstacle crossing were collected by a force plate (9281b, Kistler, Switzerland) and the COP parameters were then calculated to assess their balance. The sampling rate of the force plate was set at 1000 Hz.

COP parameters, such as COP displacement or COP area, have good reliability under both eyes-open and eyes-closed conditions²¹, and more sensible to evaluate balance capacity compared to Berg Balance Scale for individuals with chronic stroke²². Hence, COP displacement can be used to evaluate balance capacity.

Data analyses

The COP parameters were calculated from the ground reaction forces recorded during the single-stance period on the paretic limb while performing the obstacle-crossing task. During the test, participants led with the paretic limb to step over the obstacle. The COP displacement was computed as the total excursion of the COP trajectory within this stance phase. The maximum COP displacements in the anteroposterior (AP) and mediolateral (ML) directions were determined as the peak-to-peak ranges of COP movement in each direction.

COP displacements indicate balance ability, where a larger COP displacement indicates a greater degree of instability²³.

Statistical analyses

All experimental data are expressed as means ± standard deviation (SD). Shapiro-Wilk tests were performed to test the normality of continuous data. Paired t-tests were performed to test the differences between baseline and post measurements when data were normally distributed; otherwise, the Wilcoxon signed-rank test was used.

One-way analysis of variance (ANOVA) with Tukey post-hoc tests were used to compare the among-group differences at the baseline measurement for all dependent variables. The intervention effects were tested using analysis of covariance (ANCOVA) with Bonferroni correction of post-hoc tests. The pre-measurement and the Brunnstrom stage were used as the co-variate. A mixed-model ANOVA was conducted to examine the effects of different interventions over time among groups. Effect sizes were estimated using partial eta squared (η²ₚ), a common measure for ANCOVA and AOVA analyses. According to conventional benchmarks, an η²ₚ value of 0.01 indicates a small effect, 0.06 a medium effect, and 0.14 a large effect²⁴. The significance level was set at 0.05. SPSS Version 17.0 (IBM SPSS Inc., Chicago, IL, USA) was used to perform the statistical analyses.

Results

The basic demographic data for the participants are shown in Table 1. No significant among-group differences were found in terms of the demographic data (Table 1). Although the inclusion criteria permitted participants with FAC ≥ 2, all enrolled participants had FAC scores of ≥ 3 and were able to ambulate independently during the walking assessments. The Brunnstrom stage in the PT and ET groups was significantly lower than that in the control group (Table 1). This indicates that the participants in the former two groups had significantly worse function than that of those in the control group.

Table 1.

Basic demographics of participants in each group.

Age (years)	PT group	ET group	Control group	p value
	52.67 (13.12)	54.50 (12.99)	54.00 (8.59)	0.952
Gender	6 males, 6 females	8 males, 4 females	11 males, 1 female	0.083
Height (cm)	164.50 (10.53)	166.83 (8.71)	167.92 (7.20)	0.635
Weight (kg)	73.25 (17.05)	71.42 (14.35)	72.69 (14.15)	0.933
Onset time (years)	2.72 (2.03)	2.11 (2.22)	2.55 (2.23)	0.338
Brunnstrom stage	4.67 (0.49)	4.67 (0.89)	5.25 (0.45)	0.045 a, b
Functional ambulation category	4.83 (0.58)	4.33 (0.89)	4.58 (0.51)	0.213

Notes: ^a Significant difference between PT group and control group.

^b Significant difference between ET group and control group.

Walking ability

There were no significant among-group differences at the baseline comparisons in the 6-minute walk (p > 0.05) and 10-m walk (p > 0.05) (Table 2). Furthermore, no significant group differences were observed in both 6-minute walk test and 10-minute walk test after an 8-week intervention (P > 0.05) (Table 3).

Table 2.

Comparisons of baseline measurements among groups.

6-min walk (m)	PT group (mean (SD))	Control group (mean (SD))	ET group (mean (SD))	ANOVA p value
	243.58 (117.23)	237.77 (113.99)	225.83 (145.34)	0.940
10-m walk (m/s)	0.71 (0.35)	0.68 (0.32)	0.69 (0.51)	0.986
COP_AP (cm)	2.51 (1.33)	3.11 (1.34)	2.77 (1.40)	0.247
COP_ML (cm)	0.83 (0.48)	0.83 (0.32)	1.06 (0.29)	0.568

Note: AP represents anteroposterior displacement and ML represents mediolateral displacement.

Table 3.

Among-groups comparisons of outcome measurements after interventions.

	PT group			Control group			ET group			ANCOVA p value	η² (Partial)
	Post (mean (SE))	95% CI	Changes (Mean (SD)	Post (mean (SE))	95% CI	Changes (Mean (SD)	Post (mean (SE))	95% CI	Changes (Mean (SD)
6-min walk (m)	274.99 (10.16)	254.27 to 295.71	33.59 (36.36)	249.43 (9.69)	229.67 to 269.19	23.67 (27.82)	269.34 (9.66)	249.64 to 289.04	22.91 (40.83)	0.138	0.116
10-m walk (m/s)	0.76 (0.03)	0.67 to 0.80	0.09 (0.05)	0.74 (0.03)	0.69 to 0.82	0.06 (0.10)	0.80 (0.03)	0.75 to 0.87	0.07 (0.14)	0.242	0.088
COP_AP (cm)	2.44 (0.29)	1.86 to 3.03	−0.29 (0.62)	2.90 (0.29)	2.31 to 3.49	−0.03 (1.47)	2.25 (0.29)	1.67 to 2.83	−0.54 (0.76)	0.281	0.076
COP_ML (cm)	0.51 (0.07)	0.37 to 0.65	−0.34 (0.45)	0.66 (0.07)	0.52 to 0.80	−0.19 (0.36)	0.92 (0.07)	0.78 to 1.07	−0.09 (0.20)	0.001a, b	0.345

Notes: Results are derived from ANCOVA controlling for baseline and Brunnstrom stage. The presented data for each parameter was adjusted results.

a: Significant difference between control group and ET group.

b: Significant difference between PT group and ET group.

CI represents confidence interval.

AP represents anteroposterior displacement.

ML represents mediolateral displacement.

A mixed-model ANOVA was conducted to examine changes in walking distance (6-minute walk test) and walking speed (10-meter walk test) across time (pre-test and post-test) and among the three groups. The main effects of time were significant for both the 6-minute walk test and the 10-meter walk test (F = 26.00, p < 0.001, η²ₚ = 0.441; F = 21.26, p < 0.001, η²ₚ = 0.392, respectively), indicating an overall increase in walking distance and speed following the intervention. The time × group interaction for the 6-minute walk test was not significant (F = 1.93, p = 0.161, η²ₚ = 0.105), suggesting that although walking distance improved over time, the extent of improvement did not differ significantly among the groups (Fig. 3A). For the 10-meter walk test, the time × group interaction was also not significant (F = 2.63, p = 0.087, η²ₚ = 0.137) (Fig. 3B).

Fig. 3.

Comparisons of walking ability between pre-test and post-test for each group. (A) 6-minute walk test, (B) 10-meter walk test.

Balance ability

The COP displacements in the anteroposterior and mediolateral directions during obstacle crossing at the baseline measurement were not significantly different among groups (Table 2). After an 8-week intervention, significant difference of COP displacement in the mediolateral direction among group was observed (P = 0.01, η²ₚ = 0.345). According to the was post-hoc test, the COP displacement in mediolateral direction was significantly smaller in the PT group compared to that in the ET group (Table 3). Nevertheless, no significant difference was observed between the PT group and the control group (Table 3). In contrast, no significant differences in COP displacement in the anteroposterior direction were found among groups after intervention (Table 3).

A mixed-model ANOVA was conducted to examine changes in balance ability (COP displacements in anteroposterior and mediolateral directions) across time (pre-test and post-test) and among the three groups. For balance ability in anteroposterior direction, the main effects of time and the time × group interaction were not significant (F = 2.43, p = 0.129, η²ₚ = 0.069, F = 0.921, p = 0.408, η²ₚ = 0.053, respectively) (Fig. 4A). The main effects of time were significant for the COP displacement in mediolateral directions (F = 12.42, p < 0.001, η²ₚ = 0.273), indicating an overall improved balance in mediolateral direction following the intervention. However, the time × group interaction for the COP displacement in mediolateral direction was not significant (F = 1.48, p = 0.243, η²ₚ = 0.082), suggesting that although balance ability improved over time, the extent of improvement did not differ significantly among the groups (Fig. 4B).

Fig. 4.

Comparisons of COP displacement between pre-test and post-test for each group in (A) Anteroposterior direction, (B) Mediolateral direction.

Discussion

This study investigated the effects of stepping machine interventions on walking and balance abilities in individuals with chronic stroke. The results of this study revealed that the PT group demonstrated significant improvement on the walking and balance ability, especially a significant reduction of COP displacement in the ML direction. These results supported our hypothesis the multi-planar exercise trajectory of the PT would enhance balance more effectively than the ET. The multi-planar motion in PT may have facilitated lateral stability control by requiring active weight shifting in both frontal and sagittal planes, while the ET’s sagittal-only trajectory may provide less challenge to lateral postural control.

As the Brunnstrom stage reflects the degree of motor recovery following stroke, it may directly affect an individual’s capacity for balance control and walking performance. In this study, significant baseline differences in Brunnstrom stage were found between the control and training groups. Therefore, the Brunnstrom stage was included as a covariate in the ANCOVA to control for its potential confounding influence on the outcomes. This statistical adjustment allowed for a more accurate comparison of intervention effects by accounting for baseline differences in motor function.

Walking ability

In the present study, all groups showed significant within-group improvements in walking speed after 8 weeks; however, neither intervention group demonstrated significant differences compared with the control group. Although previous research reported significantly greater improvements in walking velocity following robot-assisted gait and stair-climbing training compared with conventional physiotherapy²⁵. Participants in Hesse et al., study were subacute stroke, whereas the present study recruited chronic stroke. Recovery potential and responsiveness to intervention are known to be time-dependent, that is, earlier post-stroke periods are characterized by heightened neural plasticity and greater capacity for functional gains, while improvements tend to plateau as patients transition to the chronic stage. Consequently, between-group differences in walking outcomes may be smaller in chronic cohorts, even when within-group improvements are observed. Therefore, all groups demonstrated significant improvement on walking speed, yet, there were no significant differences among group at post-measurement. Walking speed is an important indicator of walking function in individuals with stroke²⁶, and, reduced walking speed increased fall risks²⁷. Although the progression of recovery typically slows during the chronic stage, participants in all groups showed improvement over the 8-week period, suggesting that structured repetitive training (including stepping exercise and overground walking) may help maintain walking ability.

On the other hand, all groups showed significant improvement on 6-minute walk test after an 8-week intervention in the present study. The average improvements in walking endurance in the present study were 40.23, 32.58, and 13.75 m for the PT, ET, and control groups, respectively. The improvements in walking endurance were greater than those in a previous study that recruited participants with chronic stroke for overground walking intervention or partial BWS treadmill walking intervention²⁸. Possible reasons for the greater improvements in the present study are the duration and type of intervention. In Middleton et al.’s study²⁸, the intervention of treadmill walking or overground walking was only 10 days; in contrast, the intervention in the present study was 8 weeks. In the present study, the average improvements in walking distance in the PT group (40.23 m) were greater than the minimal detectable changes (MDC) (34.37 m)²⁹. The proportion of participants who achieved a clinically meaningful improvement (post–pre ≥ 34.37 m) was 16.7% (2/12) in the control group, 41.7% (5/12) in the ET group, and 58.3% (7/12) in the PT group. These findings indicate that a higher proportion of participants in the PT group demonstrated meaningful improvements in walking endurance compared with the control group. The smaller improvement observed in the ET group may be due to its predominant activation of the quadriceps and soleus, which mainly contribute to forward propulsion in the sagittal plane. This could explain why ET improved walking endurance but not lateral stability to the same extent as PT. Although both interventions led to greater average improvements than the control group, only the PT group showed a higher proportion of participants exceeding the MDC, suggesting that PT training may be more effective in promoting clinically meaningful gains in walking endurance compared to ET group and the control group.

Balance ability

Stroke survivors frequently have sensorimotor system impairments³⁰ that may result in poor balance and increase the risk of falling during dynamic movement. Thus, balance impairments following a stroke are common. Postural sway has been reported to be greater in individuals after a stroke compared to that in age-matched healthy controls³¹. Although no significant between-group differences were found in COP displacement in the AP direction based on the ANCOVA results, the ET group showed the largest numerical reduction among the three groups. This trend may indicate that ET training, which emphasizes sagittal-plane movement, could contribute to improvements in anterior–posterior postural control, even though the changes did not reach statistical significance. In contrast, the PT and the control group demonstrated significantly smaller COP displacement in the ML direction than that in the ET group after an 8-week intervention. Lateral stability control, which is defined as control of the center of the body mass in the frontal plane during movement, has been shown to be important for weight shifting movements, such as sit-to-stand activities³². Poor lateral stability has been found to be related to balance deficits and COP displacement in the ML direction has been shown to be associated with the risk of falls⁵. Therefore, the improvement in COP displacement in the ML direction may be more crucial for balance compared to that in the AP direction. Furthermore, balance has been shown to be improved significantly after a resistance intervention on the lower extremities with additional GM strengthening³², whereas no such improvement was found without GM strengthening. Hence, hip abductor muscles are essential for controlling lateral postural stability of the trunk during walking or obstacle crossing^13,33. Although both PT and ET are stepping-based training devices, they differ in training characteristics and targeted movement control. The PT exercise trajectory involves not only upward–downward but also outward movements (frontal plane movement) that require hip abduction during stepping³⁴. Such multi-planar movements could plausibly promote greater engagement of the hip abductor muscles and enhance lateral stability control during training. In an electromyographyic study, significant hip abductor activation was observed during stepping on a PT^12,34, however, quadriceps and soleus activation was greater than that of the gluteus medius during ET exercise³⁵, reflecting its sagittal-plane movement characteristics. These distinct training characteristics may lead to different neuromuscular adaptations and balance outcomes³⁶. Consistent with these mechanistic distinctions, only the PT group demonstrated a significant reduction in ML COP displacement, indicating that its additional frontal-plane control demands may have translated into measurable improvements in balance performance.

Additionally, participants in the control group in the present study received conventional physical therapy, including overground walking. Although both walking and ET exercise occur primarily in the sagittal plane, hip abductor activation is required during the single leg stance phase of walking to maintain pelvis and center of mass stability³⁷. This interpretation is supported by the biomechanical study showing that joint moments in the sagittal plane at the knee and hip were greater during ET exercise than during level walking, whereas, the joint moments in the frontal plane of the lower extremity were greater during level walking¹⁴. These findings may explain why the COP displacements in ML directions decreased significantly during obstacle crossing after an 8-week PT and conventional therapy intervention compared with the ET intervention. Taken together, these results suggest that PT training may enhance ML balance through increased frontal-plane control demands and potential involvement of the hip abductors, although this proposed mechanism remains speculative because electromyography or kinematic data were not collected in the present study.

Balance is associated with muscle strength, the proprioceptive system, and the vestibular system. Proprioception and the ability to control balance are impaired after a stroke³⁸. Previous studies have stated that mechanoreceptors, efferent pathways, and the central nervous system influence neuromuscular function³⁸. Stepping with pedal-type exercise equipment, such as an elliptical trainer or pinnacle trainer, is a form of closed kinetic chain (CKC) exercise. It induces greater stretching of the calf muscles compared to overground walking, which is a kind of open kinetic chain (OKC) exercise. The increased stretch of the calf muscles may stimulate mechanoreceptors, facilitating the central nervous system’s ability to control the affected side and producing greater afferent information from muscle spindles³⁹. Previous studies indicated that muscle activation on the paretic side of the lower extremity increased, and balance improved significantly in individuals with chronic stroke who underwent a 6-week CKC exercise intervention compared to stroke participants in the OKC group and the control group³⁹. These findings have important clinical implications. The PT may serve as a safe and effective rehabilitation modality for improving both gait and balance in individuals with chronic stroke. Its design could incorporate with suspension system, which offers a feasible alternative to conventional gait training, particularly for patients with limited balance who may not tolerate overground or treadmill-based interventions. Furthermore, the PT’s ability to activate hip abductors and support lateral stability makes it especially valuable for fall prevention. These findings collectively suggest that the superior balance improvement in the PT group aligns with its greater frontal-plane control demands and potential activation of the hip abductors, while the ET’s primarily sagittal trajectory may limit transfer to mediolateral stability tasks.

Study limitations

This study had some limitations. Firstly, although previous study stated that healthy participants showed a significant hip abductor moment during exercise on a PT, motor control might be influenced by pain and stability problems in stroke survivors, which might interfere with proper muscular activation¹². This study did not measure the biomechanical characteristics of the lower extremities (including kinetics and muscle activation) during exercise on the equipment. Hence, the underlying mechanism of improved balance after PT intervention should thus be investigated in the future.

Second, this study assessed the walking ability by measuring 6-minute walk test and 10-meter walk test, however we did not evaluate the gait quality, for instance, the symmetry of the stride length, length of the stance/swing phases and so on. The gait quality evaluations were suggested in future study.

Third, although the power of ANCOVA in the present study for walking speed and COP results ranged from 0.82 to 0.96, the sample size was relatively small (12 participants per group). Future studies with larger cohorts are warranted to confirm these findings and improve generalizability.

Fourth, the Brunnstrom stage in the control group was higher at baseline compared with the other groups. The Brunnstrom stage represents an important indicator of motor recovery, thus, this imbalance could potentially influence walking and balance outcomes. Although we included the Brunnstrom stage as a covariate in the ANCOVA to statistically control for its potential confounding effect, residual confounding cannot be completely ruled out.

Finally, characteristics of participants were chronic and relatively high baseline functional levels (such as mean FAC scores close to 5). Therefore, the findings of this study are most applicable to individuals with chronic stroke who retain basic mobility, whereas generalizability may be limited for those in the acute or subacute stage, with cognitive impairment, or requiring full assistance for movement. Besides, since the intervention was delivered in a single clinical setting, applicability to other healthcare systems or cultural contexts should be interpreted with caution.

Conclusion

The pinnacle trainer was associated with greater improvements in mediolateral balance control during obstacle crossing, compared with the elliptical trainer. Although no significant between-group differences were observed in walking ability after the 8-week intervention, all groups demonstrated significant improvements over time. Notably, a higher proportion of participants in the PT group exceeded the minimal detectable change for walking endurance, suggesting greater clinical relevance of PT training for walking recovery. Overall, both PT and ET remain feasible training options for maintaining walking function over time in individuals with chronic stroke.

Author contributions

Conceptualization: YL You, TS Kuan, LC Kuo, and FC Su; Methodology: YL You, CF Lin, TS Kuan, CJ Lin, HF Chieh, SY Lee, HY Tseng, LC Kuo, and FC Su; Formal Analysis: YL You and LC Kuo; Investigation: YL You, CJ Lin, HF Chieh, SY Lee, HY Tseng, and LC Kuo; Data Curation: YL You, CF Lin, TS Kuan, LC Kuo, and FC Su; Formal Analysis: YL You, CF Lin; Project administration: LC Kuo and FC Su; Supervision: LC Kuo and FC Su; Funding acquisition: YL You and FC Su; Resources: FC Su; Writing – Original Draft Preparation: YL You, LC Kuo, and FC Su; Writing – Review & Editing: YL You, CF Lin, TS Kuan, CJ Lin, HF Chieh, SY Lee, HY Tseng, LC Kuo, and FC Su.

Funding

This study was supported by the Ministry of Economic Affairs, Taiwan, under grant no. 104-EC-17-A-17-S3-015, China Medical University, Taiwan, under grant no. CMU111-N-20, and National Science and Technology Council, Taiwan, under grant no. NSTC 112-2222-E-039-002-.

Data availability

The datasets analysed during the current study are available from the corresponding authors on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.