Increased comfortable gait speed is associated with improved gait biomechanics in persons with chronic stroke completing an 8-week forced-rate aerobic cycling intervention: a preliminary study

Abstract

Task-specific gait training is recommended to improve locomotor function post-stroke. Our objective was to determine the effects of a forced-rate aerobic exercise (FE) intervention on gait velocity and biomechanics in the absence of task-specific gait training. Individuals with chronic stroke (N=14) underwent 24 sessions of FE, at a targeted aerobic intensity of 60–80% of their heart rate reserve. Change in comfortable walking speed in addition to spatiotemporal, kinematic, and kinetic variables were measured using 3D motion capture. Overground walking capacity was measured by the six-minute walk test (6MWT). To determine gait biomechanics associated with increased walking speed, spatiotemporal, kinematic, and kinetic variables were analyzed separately for those who met the minimal clinically important difference (MCID) for change in gait velocity compared with those who did not. Participants demonstrated a significant increase in gait velocity from 0.61 to 0.70 m/s (P=0.004) and 6MWT distance from 272.1 to 325.1 meters (P<0.001). Those who met the MCID for change in gait velocity demonstrated significantly greater improvements in spatiotemporal parameters (P=0.041), ground reaction forces (P=0.047), and power generation (P=0.007) compared to those who did not. Improvements in gait velocity were accompanied by normalization of gait biomechanics.

INTRODUCTION

Considerable effort is put toward the recovery of gait post-stroke, with particular emphasis on walking speed, as it is shown to predict disability level.¹ Despite advances in rehabilitation, nearly 75% of individuals do not regain full use of their hemiparetic lower extremities, resulting in deficits in locomotion, increased fall risk, and decreased community reintegration.^2–4 A recently published clinical practice guideline cited strong evidence that high intensity task-specific gait training is effective in improving locomotion.¹ However, we recently reported that two modes of aerobic cycling improved walking capacity in individuals with chronic hemiparesis² in the absence of task-specific gait training, contrasting the guideline.¹

Following stroke, diminished muscle power and abnormal timing and coordination of muscle agonists and antagonists disrupt the modulation of phasic muscle activity, resulting in inefficient movement patterns during gait. Phasic muscle activity comparable to what is observed during gait is induced during cycling training.^3–6 High-rate cycling has also been shown to improve rate-dependent mobility.⁷ Additionally, neuroimaging studies have shown cortical, subcortical, and cerebellar patterns of activation on imaging during cycling tasks, spatially comparable to a lower limb reciprocal tapping task.⁸ Thus, high intensity cycling may elicit a transfer of training to improve locomotor function in persons with stroke.

The current study uses three-dimensional (3D) motion capture to provide a biomechanical analysis of gait to provide insight into whether increases in gait velocity were associated with improvements in motor control or if individuals sacrificed gait mechanics to walk faster. In the absence of training, individuals post-stroke typically increase stride length and cadence, but demonstrate blunted increases in gait kinetics compared to neurologically healthy individuals.⁹ The aims of this project were to determine the effects of an 8-week forced-rate aerobic exercise (FE) intervention on gait velocity and locomotor control using 3D motion capture in individuals with chronic stroke. It was hypothesized that individuals completing FE would demonstrate increases in gait velocity accompanied by an improvement in gait biomechanical characteristics toward normal from baseline to end of treatment (EOT).

METHODS

A randomized clinical trial was conducted to determine the priming effects of FE on upper extremity (UE) recovery in individuals >6 months post-stroke (K01HD092556) clinicaltrials.gov registration number NCT03819764). This project represents an interim analysis of an exploratory aim to determine preliminary effects of an 8-week FE intervention on locomotor control quantified using biomechanical gait data. The study was approved by the Cleveland Clinic IRB (#18–734) and the research was conducted in accordance with the Declaration of the World Medical Association. All participants provided written informed consent.

Participants

Individuals with a single, unilateral stroke with residual UE hemiparesis were recruited. Inclusion criteria: 1) ≥ 6 months post-stroke, 2) UE Fugl-Meyer motor score 19–55, 3) Ambulatory ≥20 meters (with or without assistive device), and 4) 18–85 years of age. Exclusion criteria: 1) hospitalization for myocardial infarction, heart failure or heart surgery within 3 months, 2) cardiac arrhythmia, 3) hypertrophic cardiomyopathy, 4) severe aortic stenosis, 5) pulmonary embolus, 6) contractures, and 7) other contraindication to exercise. All participants underwent a metabolic exercise stress test to ensure safe cardiopulmonary response to maximal exercise.

Forced Exercise Intervention

The FE protocol was based on methodology from our previous studies^2,10–13 and involved supervised exercise on a custom-engineered stationary semi-recumbent cycle ergometer for 45-minute sessions. The FE cycle motor supplemented pedaling rate 30% greater than the participant’s voluntary rate achieved during their exercise stress test. The target heart rate zone (60–80%) was determined for each participant using the Karvonen formula. Heart rate was continuously monitored and displayed using a Wahoo chest strap (Wahoo, Atlanta, GA). Clip-in cycling shoes were used, and individuals completed FE sessions without the use of ankle foot orthoses in order to facilitate normal pedaling mechanics, allowing for contribution from the plantarflexors during the downstroke phase of pedaling.

Walking Capacity

The six-minute walk test (6MWT) was administered to assess overground walking capacity.¹⁴

Gait Analysis

Biomechanical gait data were collected using the Computer Assisted Rehabilitation ENvironment (CAREN) (Motekforce Link, Netherlands) system at baseline and following the 8-week FE intervention by an engineer.^15,16 The primary outcome was change in comfortable self-selected gait velocity obtained during two 2-minute trials following an acclimatization trial.

Secondary outcomes included spatiotemporal, kinematic, and kinetic gait parameters. Kinematics included sagittal plane range of motion of the hip, knee, and ankle. Kinetic data included peak vertical ground reaction forces (vGRF), peak anterior-posterior (AP) braking and propulsion forces, and peak lateral GRF. Total power generation for the hip, knee, and ankle were computed as the positive area under the curve. Peak hip and knee extension and ankle plantarflexion moments during stance phase were obtained. Methods to compute each gait variable are outlined in the supplemental material.

Statistical Analysis

Descriptive statistics were computed to describe demographic and exercise variables for the overall sample and for the dichotomized groups separating participants who met the minimal clinical important difference (Group_MCID) value for change in gait velocity of 0.1 m/s compared to those who did not. Groups were compared on demographics and exercise variables. Normality of data was determined using visual inspection of histograms and normal Q-Q plots, along with the Shapiro-Wilk test and homogeneity of variance was assessed with Levene’s test. The change in comfortable gait velocity and 6MWT from baseline to EOT were analyzed using separate paired t-tests with an alpha of 0.05. The remaining spatiotemporal, kinematic and kinetic variables were analyzed using separate 2-way multivariate analysis of variance (MANOVA) models, with time as the within factor and group (Group_MCID or MCID not achieved) as the between group factor. If the MANOVA was significant at an alpha of 0.05, post-hoc comparisons were conducted using separate linear models with Bonferroni corrections to adjust for multiple comparisons. Partial eta squared values were used to determine effect size of the univariate comparisons within each MANOVA, defined in Table 2.

Table 2: Spatiotemporal and Kinematic Gait Variables.

	Baseline	EOT	Difference
Six-Minute Walk Test (m)*a	272.1 ± 128.4	325.1 ± 147.6	53.0 ± 36.4
Gait Velocity (m/sec)*a	0.61 ± 0.34	0.70 ± 0.32	0.09 ± 0.09

MANOVA Models:		GroupMCID (N=5)		Did not meet MCID (N=9)
Spatiotemporal Variables*b	Limb	Baseline	EOT	Baseline	EOT
Cadence (steps/min)*c†	N/A	78.9 ± 28.7	93.4 ± 29.7	88.1 ± 13.2	87.8 ± 11.8
Step Length (cm)	Paretic††	32.9 ± 15.0	40.4 ± 13.7	45.2 ± 16.8	48.5 ± 14.9
	Non-paretic†	36.1 ± 11.06	44.2 ± 9.76	45.7 ± 19.9	48.0 ± 17.2
Stance Time (sec)	Paretic	1.33 ± 0.80	1.05 ± 0.55	0.95 ± 0.20	0.94 ± 0.16
	Non-paretic	1.34 ± 0.78	1.06 ± 0.55	1.02 ± 0.23	1.01 ± 0.19
Swing Time (sec)	Paretic	0.40 ± 0.06	0.38 ± 0.09	0.46 ± 0.06	0.45 ±0.06
	Non-paretic	0.40 ± 0.08	0.38 ± 0.10	0.38 ± 0.04	0.39 ± 0.04
Single Support Time (%)	Paretic†	25.1 ± 6.0	27.6 ± 4.0	27.6 ± 4.17	28.5 ± 3.41
	Non-paretic†††	25.8 ± 5.8	28.2 ± 4.1	31.9 ± 4.70	32.0 ± 4.94
Kinematic Variables
Hip ROM, flex/ext, degrees	Paretic†	30.5 ± 6.07	33.1 ± 7.03	33.7 ± 12.7	35.7 ±12.7
	Non-paretic†	36.2 ± 7.21	41.1 ± 9.07	38.7 ± 11.6	42.1 ± 12.3
Knee ROM, flex/ext, degrees	Paretic††	44.2 ± 13.2	48.5 ± 11.7	37.3 ± 16.7	39.2 ±17.3
	Non-paretic††	45.3 ± 9.16	47.7 ± 6.61	54.4 ± 16.1	54.8 ± 15.3
Ankle ROM, dorsi-/plantarflexion, degrees	Paretic†	17.6 ± 4.4	20.2 ± 4.11	17.8 ± 6.44	19.0 ± 6.05
	Non-paretic†	19.6 ± 2.78	23.7 ± 5.08	22.8 ± 9.47	23.8 ± 9.37

ROM: range of motion; flex/ext: flexion/extension

^*Statistically significant analyses in bold print; P ≤ 0.05 for paired t-test and MANOVA; after Bonferroni correction, post-hoc univariate comparisons for spatio-temporal variables significant at P ≤ 0.01.

^a:Paired t-test significant;

^b:MANOVA significant;

^c:Variable significant in post-hoc analysis

^†:Small effect size (partial eta squared 0.01–0.05);

^††:Medium effect size (partial eta squared 0.06–0.13);

^†††:Large effect size (partial eta squared ≥0.14)

Italicized values denote improvement in variable toward normal

RESULTS

Fourteen participants were included in this study. Demographics, baseline characteristics, and exercise variables are summarized in Table 1. Changes in spatiotemporal, kinematic, and kinetic parameters are shown in Table 2.

Table 1: Participant Demographics and Exercise Characteristics.

	Overall (n=14)	≥ MCID (n=5)	< MCID (n=9)	P-value
Age (years)	63.6 ± 13.4	65.2 ± 19.3	62.7 ± 10.3	0.36
Male sex (versus female), n	11 (78%)	5 (100%)	6 (67%)	0.05*
Dominant Side Affected, n	6 (43%)	2 (40%)	6 (67%)	0.69
Time Since Stroke (months)	39 [14, 80]	38 [14, 104]	40 [12, 83]	1.00
Modified Rankin Score, n
2		4 (80%)	5 (55.6%)	0.285
3		1 (20%)	4 (44.4%)
Stroke Type, n
Ischemic	12 (86%)	4 (80%)	8 (89%)
Hemorrhagic	2 (14%)	1 (20%)	1 (11%)
Height (cm)	174.2 ± 9.3	175.9 ± 8.7	173.2 ± 10.0	.625
Mass (kg)	84.0 ± 19.5	98.7 ± 19.3	75.8 ± 14.9	.029*
Exercise characteristics
Cadence (RPM)	75.3 ± 7.1	78.9 ± 7.6	73.2 ± 6.4	0.36
Percentage of HRR	59 ± 10%	60 ± 7%	59 ± 11%	0.80
Power (watts)	70.3[9.5, 113.9]	100.3[−0.5, 117.6]	65.5 [13.2, 101.7]	0.90
Session Duration, min	44.7 ± 0.5	44.8 ± 0.3	44.7 ± 0.6	1.00

Summary statistics presented as mean ± standard deviation, median [Q1, Q3], or n (%) for categorical data; MCID – minimal clinically important difference for change in gait velocity; RPM- revolutions per minute; HRR- heart rate reserve

Walking Capacity and Spatiotemporal Characteristics of Gait

Walking capacity measured by the 6MWT (Fig 1a) improved significantly from 272.1±128.4 meters to 325.1±247.6 meters (P < 0.001).¹⁷ Change in gait velocity for each participant is shown in Figure 2. Gait velocity improved from 0.61±0.34 m/s to 0.70±0.32 m/s, P = 0.004 (Fig 1b). The MANOVA revealed a significant effect for change in spatio-temporal gait variables favoring Group_MCID, V = 0.84, F(7, 6) = 4.72, P = 0.041.

Figure 1:

Spaghetti plots depicting Six-Minute Walk Test (1a) and gait velocity performance (1b) for each participant and the cohort (black bold-faced line) at baseline and EOT. Responders (dashed lines) and non-responders (solid lines) who met the MCID value for the 6MWT of 34.4 meters and the MCID value for change in gait velocity of 0.1 m/sec are delineated. Variability in baseline gait velocity is observed across participants, whether or not improvements in gait velocity exceeding the MCID value of 0.1 m/s was obtained. Significant improvements were observed in both metrics from baseline to EOT, denoted with an asterisk.

Figure 2:

Change in gait velocity from baseline to post-intervention for each forced exercise participant revealing all but two demonstrating improvements and five exceeding the minimal clinically important difference of 0.1 m/sec. Females participants are indicated by red filled bars.

Kinematic parameters of gait

Average sagittal plane range of motion (ROM) for the hip, knee, and ankle showed a modest increase at EOT. The MANOVA did not reveal a significant between group effect for change in kinematic gait variables, V = 0.47, F(6, 7) = 1.02, P = 0.482.

Kinetic Parameters of Gait

A significant effect favoring Group_MCID was observed for change in GRF (V = 0.624, F(4, 9) = 3.73, P = 0.047) and for change in joint power (V = 0.686, F(3, 10) = 7.27, P = 0.007). Increased joint moments were primarily observed with hip extension and ankle plantarflexion. The MANOVA did not reveal a significant between group effect for change in joint moment (V = 0.433, F(3, 10) = 2.55, P = 0.115).

DISCUSSION

Our results support our working hypothesis that an 8-week FE cycling intervention contributed to a significant improvement in gait velocity, accompanied by improvements in select spatiotemporal, kinematic, and kinetic gait characteristics. Improvements in select kinetic variables indicate that participants’ compensatory strategies did not worsen in order to walk faster. While gait biomechanics were measured during treadmill walking, significant improvements in overground walking were also observed during the 6MWT, exceeding the MCID value of 34.4 meters.¹⁷ Improvements on the 6MWT if converted to velocity were greater in magnitude than the improvements measured during treadmill walking. This discrepancy has been observed in prior studies and it is plausible that individuals were more comfortable with overground walking compared to treadmill.¹⁸ Nonetheless, our novel findings may have considerable clinical relevance, as moderate- to high-intensity cycling may be a viable and efficacious option to improve locomotor function in individuals with chronic stroke.¹

While task-specificity has been considered important in motor learning, a transfer of training has been shown to occur between distinct motor tasks, particularly when the kinematic and spatiotemporal requirements of the tasks are similar.¹⁹ Although cycling and walking are different tasks, both require the rapid reciprocal activation and relaxation of lower extremity muscles synergistically.^4,5,20–24 Thus, high-rate cycling associated with FE, combined with the consistent rhythmic motion induced by the motorized bike, may have elicited a transfer of training resulting in improved locomotor control.

Improvements in Spatiotemporal Gait Parameters Following FE

Overall, FE participants demonstrated improved spatiotemporal gait characteristics across all variables. However, Group_MCID had significantly greater increase in gait cadence, approaching normative values for healthy adults reported at 111.6±8.3 steps/minute.²⁵ Increases in step length were also observed, accompanied by improvements in step symmetry approaching 0.5 at EOT and an increase in percentage of time spent in single limb stance for both limbs, indicative of improved balance and single limb stance stability.

The Effects of FE on Kinematic Gait Parameters

Overall, participants demonstrated modest improvements in sagittal plane kinematics of the hip, knee and ankle. Numerous mechanisms are responsible for diminished ROM post-stroke, including the inability to rapidly activate and relax the limb extensors in a rhythmic manner.²⁶ This impaired motor control interferes with smooth and timely transitions between stance and swing phases of gait and is characterized by abnormal muscle co-contractions.²⁷ The downstroke phase in cycling is driven by extension and forces passive flexion (relaxation of the extensors) of the contralateral limb during its upstroke phase. While we did not obtain EMG data to determine the effect on activation patterns, high cadence cycling may train muscles to work synergistically, facilitating smoother intra- and interlimb reciprocal activation, similar to activation patterns used to coordinate joint angle accelerations and decelerations during gait.^4–6

Changes in Kinetic Variables Following FE

While increased total power generation was evident at the hip, knee, and ankle overall, the Group_MCID demonstrated the greatest increases. Hip extension and ankle plantarflexion power provide the main propulsive forces in human gait.^28,29 Although increased values of total power were also evident with ankle plantarflexion, gait testing was conducted with some participants wearing prescribed ankle-foot orthoses (AFOs), limiting their ability to generate propulsive power with plantarflexion through terminal stance. Albeit a limitation of our ability to investigate a transfer of training, participants were not safe ambulating without AFOs, and a reflection of change in comfortable walking speed was sought. Increased hip extension and ankle plantarflexion moments were also observed, complementary to the changes seen in hip and ankle power generation but may have been blunted by AFOs.

Increased magnitude of AP braking forces and propulsion forces were observed with loading response and terminal stance, respectively, indicative of normalization of deceleration which results in posterior shear forces that occur as the limb initiates weight acceptance and anterior shear which occurs with forward propulsion as the individual’s center of mass progresses anterior to the foot.³⁰ Increased lateral GRF values were also observed, indicating greater weight shift onto the paretic limb.³⁰ These data align with changes observed in the spatiotemporal variables, as improvements in single limb support percentage and increased step symmetry are noted with normalization of lateral paretic limb weight shifting.

Modest improvements in peak vGRF data were also observed for the paretic limb approaching non-paretic limb values, indicating increased load symmetry. In the absence of training, at increased walking speed, individuals post-stroke increase spatiotemporal gait parameters but have reduced and disproportionate increases in ground reaction forces.⁹ Collectively, the changes in GRF data observed are indicative of normalization of the kinetic forces that are responsible for the momentum-driven action in human gait.

Using FE to Overcome Limitations of High-Intensity and High-Repetition Training

Stationary cycling, and in particular FE cycling, has numerous advantages for individuals post-stroke including safety, the ability to complete thousands of repetitions in a single session, replicating high cadence associated with normal gait, and the global benefits to aerobic exercise. Cycling on a semi-recumbent stationary ergometer requires less postural control than walking, providing a safe modality to train in an intense manner without considerable focus on balance. Cycling at an average cadence of 75.3 RPM, our participants completed an average of 2635 revolutions per session during the main 35-minute exercise set, which, if calculated to strides, would equate to walking 896 meters. From a training perspective, cycling and walking do not directly equate; however, it is apparent that the FE approach provided a highly repetitious, intensive, and consistent exercise rate, which cannot be easily replicated during overground or treadmill-based gait training.

Study Limitations

We report the effects of FE cycling on a small sample of 14 individuals without control group data for comparison, potentially limiting the precision of the estimates of effect. The multiple variables obtained through 3-D motion capture included in the models increases the risk of a Type I error. The primary inclusion criteria for this trial were related to UE function which enhanced heterogeneity of baseline ambulatory status. Participants were tested using AFOs they wore during community ambulation and used one or both handrails as needed, potentially impacting biomechanical gait data. However, these conditions were kept consistent for each participant during baseline and post-intervention trials. Therefore, these results may still be used as a basis of comparison between the pre- and post-intervention time points and support the use of moderate- to high-intensity cycling to improve gait velocity, biomechanics, and walking capacity in individuals with chronic stroke.

Data availability:

Data will be made available from the corresponding author upon request.