The effects of cycling with and without functional electrical stimulation on lower limb dysfunction in patients post-stroke: A systematic review with meta-analysis

Abstract

BACKGROUND:

One of the leading causes of disability in the world with enormous economic burden is stroke.

OBJECTIVE:

To quantify the effectiveness of different protocols of cycling with/without functional electrical stimulation on functional mobility after stroke.

METHODS:

Multiple databases were searched till 2018. Data extraction was performed using a pre-determined data collection form. The quality of the evidence was evaluated using the Grading of Recommendations Assessment, Development and Evaluation.

RESULTS:

A total of 14 trials satisfied eligibility criteria and were included. Cycling had a positive effect on the 6-meter walking test performance (SMD, 0.41; 95% CI, 0.11 –0.71; I² = 0% ) compared with no or placebo intervention (control). Compared with control, cycling had a positive effect on 10-meter walking speed (SMD, 0.30; 95% CI, 0.05 –0.55; I² = 0% ), and on balance based on the Berg score (SMD, 0.32; 95% CI, 0.06 –0.57; I² = 49% ). Cycling with functional electrical stimulation had a positive effect on balance (SMD, 1.48; 95% CI, 0.99 –1.97; I² = 91% ) compared with cycling alone.

CONCLUSIONS:

It appears that cycling has a positive effect on walking speed, walking ability and balance. Functional electrical stimulation combined with cycling has positive effects on balance beyond cycling alone.

Keywords

Meta-analysis stroke lower limb cycling disability functional electrical stimulation

1 Introduction

One of the leading causes of disability in the world is stroke (Feigin et al., 2014; Lehmann et al., 2015). People who have sustained a stroke experience activity limitations such as a decreased ability to walk, and decreased participation in everyday life tasks (e.g., property maintenance, managing household chores, self-care, etc) (Billinger et al., 2014; Shariat et al., 2018; Timmermans et al., 2016). These functional limitations are often related to the hemiplegia that occurs as a result of the stroke (Kwakkel, Veerbeek, van Wegen, & Wolf, 2015). The levels of disability associated with stroke result in an enormous economic burden (Wang et al., 2014). Identifying strategies to maximize function and reduce the societal burden associated with stroke is of utmost importance.

In the first few weeks after a stroke, neurological deficits may improve as a result of brain plasticity. This brain plasticity may result from recruitment of different pathways, facilitation of existing but dormant synaptic junctions, aboritzation of dendrites and synaptogenesis (Rossini, Calautti, Pauri, & Baron, 2003). The magnitude of recovery differs greatly among individuals with similar clinical severity in the acute phase. The understanding of the mechanisms which prevent or promote recovery is critical to optimize therapeutic management strategies and maximize function. During treatment, sensory feedback and motor activity is essential (Rossini et al., 2003). Numerous studies have identified a relationship between afferent stimulation and changes in brain activity, such as repetition (Jones, 2000), efficient goal-directed activity (Nudo, Wise, SiFuentes, & Milliken, 1996), and functional electrical stimulation (FES) (Ambrosini, Ferrante, Pedrocchi, Ferrigno, & Molteni, 2011; de Sousa, Harvey, Dorsch, Leung, & Harris, 2016; Popović, Sinkjær, & Popović, 2009).

Outcomes after a stroke can potentially be enhanced through interventions focused on improving impairments, functional limitations and restrictions of participation (Billinger et al., 2014). One such intervention that may address these outcomes is cycling with/without electrical stimulation which has the potential to improve motor capacity and hypothetically result in greater participation and activity performance after a stroke (Barbosa, Santos, & Martins, 2015; Bauer, Krewer, Golaszewski, Koenig, & Müller, 2015). In comparison with conventional training modes, cycling is a low-cost and simple-to-use rehabilitation approach following a stroke (Mazzocchio, Meunier, Ferrante, Molteni, & Cohen, 2008).

Patients with a stroke who participate in daily cycle training have achieved major enhancements in their lower extremity (LE) muscle strength, balance ability, and aerobic capacity (Yang et al., 2014). Furthermore, participants who suffer from chronic stroke and participate in 10–12-weeks of cycling have been shown to experience improvements in cardiorespiratory fitness (Janssen et al., 2008). Even though multiple systems are affected by a stroke, such as motor control, balance, upper-extremity function, endurance, and gait, the majority of studies have only reported data on functional outcomes post stroke (Duncan et al., 2003). Most studies have used a simple form of cycling with/without FES (Bauer et al., 2015; de Sousa et al., 2016; M. Lee et al., 2008; S. Y. Lee et al., 2013; Yang et al., 2014). However, the optimal cycling protocol to maximize outcomes is not known for this population. An increasing number of physical therapists throughout the world are using cycling to treat lower limb dysfunction for patients post stroke (Billinger et al., 2014; Valles et al., 2016). As physical therapists use cycling more frequently, it is essential to continually evaluate the current evidence to refute or support its effectiveness. Therefore, the purpose of this systematic review and meta-analysis was to quantify the effectiveness of different protocols of cycling with/without FES on the lower limbs after stroke and to make recommendations for clinicians currently treating this population. The specific research question is as follow: Is cycling with FES more effective than cycling alone and if so, for which outcome measures?

2 Methodology

2.1 Protocol and registration

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed for this systematic review and meta-analysis (Moher, Liberati, Tetzlaff, Altman, & Group, 2009). This systematic review was registered a priori with PROSPERO on 03.01.2018 (CRD42017077974).

Figure 1 shows the criteria for inclusion of studies.

Fig.1

PRISMA flow chart.

2.2 Eligibility criteria

The following criteria were used for the selection of relevant articles:

Participants: Human subjects post-stroke (Adults ≥18 within 5 years after stroke).

Intervention: Cycling with/without functional electrical stimulation (FES) to the lower limbs

Comparison: cycling alone, cycling with FES, control, placebo, or other interventions

Outcomes: Balance, walking speed, mobility

Study design: Randomized clinical trial

2.3 Information sources

The following 7 electronic databases were searched till 2018; Pubmed, Cochrane Central Register of Controlled Trials, Ovid Medline, EBSCO Cumulative Index of Nursing and Allied Health Literature, Ovid EMBASE, Physiotherapy Evidence Database (PEDro) (www.pedro.org.au), and Occupational Therapy Systematic Evaluation of Effectiveness (www.otseeker.com) for relevant search terms including randomized, stroke, cycling, lower limb dysfunction using Boolean operators.

2.4 Search

In the databases of Cochrane Central Register of Controlled Trials, Medline, ISI Web of Science, EBSCO Cumulative Index to Nursing and Allied Health Literature, and Scopus. We searched our variables in terms of population such as cerebrovascular disorders, basal ganglia, cerebrovascular disease, brain ischemia, post-stroke, stroke, hemiplegia; intervention such as FES, muscle stimulation; study design such as randomized controlled trials, random allocation, controlled clinical trials, control groups, sham intervention, placebo, clinical trials, double-blinding, cross-over studies; outcomes such as gait, lower limb function, motor function, lower extremity, balance, postural control, mobility and walking.

AND was used to combine the search concepts of population, study design and outcome. OR was used with synonyms and similar terms. We used parentheses to help group parts of the search query, especially when we had several parts, and to set the order of the query for the database.

We searched for the subject heading first, then searched text words, combined these synonymous searches with OR using our search history. We repeated for the second, third and subsequent concepts. Finally we combined large search results set with AND.

Concept 1

Search #1

Search #2

Search # 3

Search #4 = #1 OR #2 OR #3

Concept 2

Search#5

Search#6

Search #7

Search#8 = #5 OR #6 OR #7

Search#9 = #4 AND #8

We also used filters, including date of publication (Till 2018), and design of study (only randomized control trials).

2.5 Study selection

Once duplicate studies were removed, 2 reviewers (AS and SN) independently screened the titles and abstracts to determine if studies potentially met eligibility for this systematic review and meta-analysis. If we identified studies that might be eligible or if we couldn’t determine if studies were eligible from the abstract a full text review was performed. Disagreements between reviewers were resolved by consulting a third reviewer (JC) who was blind to other reviewers’ decisions on whether the study should be included. Following the final study selection, we emailed the corresponding author on each of the selected studies and queried them regarding our list of studies and if they thought we might have missed potentially relevant studies for this systematic review. Where it was not possible to determine eligibility according to the information stated in the publication, the trial’s authors were contacted to determine missing information. If multiple studies reported the same data only the article with the longest follow-period was included in the systematic review and meta-analysis.

2.6 Data collection process

Data extraction was performed by the primary investigator (AS), and was compiled into a standardized form (Gattie, Cleland, & Snodgrass, 2017; Howlett, Lannin, Ada, & McKinstry, 2015). Data included sample size, diagnosis, inclusion/exclusion criteria, duration of symptoms, type of cycling intervention (location, technique, and duration), main outcomes, time to outcome, and harm reported. Included studies were analyzed for methodological quality by 2 independent reviewers (AS and SN), based on risk of bias table in the Cochrane handbook. Disagreements between the reviewers were again settled by a 3rd reviewer (JC) who was blind to previous assessment scores.

2.7 Data items

Our variables were based on PICO:

Population: stroke, cerebrovascular, cerebrovascular disorder, hemiparesis, hemiplegia.

Intervention: cycling, leg ergometer, functional electrical stimulation, bike training, functional stimulation, lower limb exercise.

Comparison: control, experimental, sham, conventional therapy, standard therapy.

Outcome: balance, walking speed, mobility.

2.8 Risk of bias in individual studies

Random sequence generation (selection bias), Allocation concealment (selection bias), Blinding of participants and personnel (performance bias) all study design outcomes, Blinding of outcome assessment (detection bias all outcomes), Incomplete outcome data (attrition bias) all outcomes, and Selective reporting (reporting bias(were determined using the Cochrane risk of bias that has three level of bias (low risk of bias, high risk of bias, or unclear risk of bias).

2.9 Synthesis of results

Kappa statistics provided an estimate of inter-rater agreement. RevMan (The Nordic Cochrane Centre, Copenhagen, Denmark) was used to perform meta-analyses (Landis & Koch, 1977). Outcome measures used different scales of measurement with 95% CIs were used to investigate differences (Takeshima et al., 2014). Study variability and heterogeneity was tested by a fixed effects model and the I² statistic (Higgins & Green, 2011). The I² value of 25% represents a small, 50% a moderate, and 75% a large degree of heterogeneity (Higgins, Thompson, Deeks, & Altman, 2003). Based on previous research, we described 0.2 as small, 0.5 as moderate, and 0.8 as large effect sizes (Cohn, 1988). As we had fewer than 10 studies for each comparison we did not develop the funnel plot as it has been suggested that power would be artificially inflated (Sterne et al., 2011).

Using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach (Balshem et al., 2011) two reviewers (AS and JC) analyzed each study separately. Once the reviewers appraised the evidence, the quality was categorized on a level of evidence using the following criteria (Guyatt et al., n.d.):

High-quality evidence: Additional research would not be likely to change our confidence in the estimate of effect

Moderate-quality evidence: Additional research is likely to show a significant influence on our confidence in the evaluation of effect and may change the evaluation

Low-quality evidence: Additional research is to be expected to show a significant effect on our confidence in the estimate of effect and can probably change the estimate

Very low–quality evidence: Any effect estimate is very indeterminate.

Randomized clinical trials were scored as high-quality evidence but were downgraded based on the following 5 items: (1) research design and risk of bias (reduced if more than 25% of the participants were from studies used risk of bias table in Cochrane handbook; (2) represented heterogeneity (reduced if significant heterogeneity existed on visual inspection or the I² value was more than 50%); (3) Reduced generalizability(downgraded if more than 50% of the participants were outside the target group); (4) imprecision (reduced if fewer than 40 participants were included in the comparison for continuous data); and (5) other (publication bias) (Saragiotto et al., 2016).

3 Results

3.1 Study selection

Citations for 2084 trials were identified in the original search. After duplicates were removed, 1845 records remained and were screened. Of these, 1718 were excluded after screening based on the title and abstract. 127 Articles were assessed for full text eligibility. Of these, 113 trials were excluded after full-text review, leaving 14 trials for inclusion. See Fig. 1 for a summary of the flow of trials through the review.

3.2 Study characteristics

The 14 trials included in the review included 794 participants who were assigned for randomization and 680 participants who completed the final follow-up. Across the trials, the mean age ranged from 42.5 to 85 years, and 60% of participants were men. All trials included only participants who had suffered a stroke. The mean time after stroke ranged from >1 to 27 months, with 70% of the trials carried out on patients who had a stroke greater than 3 months prior to enrollment. Details about the characteristics of participants have been reported in Table 1.

Table 1
Summary of included trials

Study Design Patients Intervention Outcomes Conclusion Adverse effects

(Duncan et al., 2003) RCT n (randomized = 100, 92 completed trial) Exp: Structured, progressive, physiologically based, therapist-supervised, in-home program Motor recovery of lower and upper extremity This program produced gains in endurance, balance, and mobility None reported

Age (y):>50 90 min3 weeks12–15 weeks Grip strength

Sex: 92 M Con: conventional rehabilitation Isometric strength of dorsi-flexors and knee extensors

Time since stroke: 1–5 mo Gait speed

walk 6 MWD)

Balance

Aerobic capacity

Timing: 0, 12 wk

(Kamps & Schüle, 2005) RCT n (randomized = 40, 37 completed trial) Exp: conventional rehabilitation + cycling Gait velocity and the distance walked. MOTOmed Movement Trainer is a helpful addition to conventional therapy of stroke patients. None reported

Age (y): 55–75 10 min in ever twice a day for 4 months Motor assessments and MOTOmed Data.

Sex: 22 M, 9 F Con: conventional physio- and ergotherapeutic interventions. Timing: 1 to 16 wk

Time since stroke: 3–27 mo

(Katz-Leurer, Sender, Keren, & Dvir, 2006) RCT n (randomized = 24, 23 completed trial) Exp: Conventional Rehabilitation + cycling Balance, Stroke patients can improve their motor and balance abilities after an early short duration of cycling. None reported

Age (y):>50 5 weeks3 weeks The motor function of the lower extremity.

Sex: 13 M, 11 F Con: Conventional Rehabilitation Timing: 0, 3 wk, 6 wk

Time since stroke:>1 mo

(Olney et al., 2006) RCT n (randomized = 74, 54 completed trial) Exp: Supervised exercise sessions primary outcome: walking speed. Supervised programs showed trends to greater improvements in self-reported gains. None reported

Age (y):>20 90 min3 weeks10 weeks secondary outcome Human Activity Profile.

Sex: 29 M, 25 F Con: Home program for 9 (unsupervised). physical and mental health and quality of life,

Time since stroke:>12 mo It was not clear, if they used the conventional/standard rehabilitation as well. lower extremity muscle strength.

The walking heart rate and the resting heart rate.

Timing: 0, 10 wk, 6 mo, 1 yr

(Sullivan et al., 2007) RCT n (randomized = 80, 63 completed trial) 4 groups: Primary outcome: Self-selected walking speed, fast walking speed, and 6-minute walk distance measured before and after intervention and. After chronic stroke, task-specific training during treadmill walking with body-weight support is more effective in improving walking speed and maintaining these gains at 6 months than resisted leg cycling alone. There were 17 adverse events during the intervention period, 8 of which were not study related (4 falls in home, 1 report of low back pain, 1 controlled seizure, 1 participant diagnosed with colon cancer, 1 participant diagnosed with congestive heart failure after randomization but prior to starting intervention).

Age (y):>18 1. Treadmill training and upper-extremity ergometry secondary outcome the LE-FM motor score, Berg Balance Scale, the 16-item Stroke Impact Scale, version 3.0, Medical Outcomes Study 36-Item Short-Form Health Survey, version 2.0, and LE isometric peak torque

Sex: 63 M 2. Cycling and upper-extremity ergometry Timing: 6 we, 6 mo

Time since stroke: 4 mo to 5 yr 3. Treadmill training and cycling

4. Treadmill and LE muscle-specific progressive resistive Exercise

4 weeks6 weeks

(M. Lee et al., 2008) RCT n (randomized = 52, 48 completed trial) Aerobic cycling plus sham progressive resistance training (PRT), sham cycling plus, aerobic cycling plus PRT, or sham cycling plus sham PRT. Primary outcomes walk distance, habitual and fast gait velocities, and stair climbing power. Single-modality exercises targeted at existing impairments do not optimally address the functional deficits of walking No serious adverse events during testing or training

Age (y):45–85 60 min3 weeks12 weeks Secondary outcomes: cardiorespiratory fitness; muscle strength, power, and endurance; and psychosocial attributes.

Sex: 28 M, 20 F Timing: 0, 12 wk

Time since stroke: <3mo

(Ambrosini, Ferrante, Pedrocchi, Ferrigno, & Molteni, 2011) RCT n (randomized = 35, 30 completed trial) Exp: Conventional rehabilitation + Trained cycling and FES Primary outcome motor power of the paretic lower extremity gait speed. FES cycling training significantly improved lower extremity motor functions and accelerated the recovery of overground locomotion in postacute hemiparetic patients. None reported

Age (y):40–70 25 min5 weeks4 weeks Secondary outcome Trunk control, functional abilities of the impaired leg during single-limb standing.

Sex: 18 M, 12 F Con: Placebo Timing: 0, 4 wk, 3–6 mo

Time since stroke: <6 mo

(Lo, Hsu, Hsueh, & Yeh, 2012) RCT n (randomized = 20, 20 completed trial) EXP:CG+FES group Postural control Cycling training, with or without FES may reduce spasticity in stroke patients. None reported

Age (y):>47 20 min1 session Muscle tone

Sex: 10 F,10 M Con:CG Limb dynamic

Time since stroke: 25.54 ± 12.95 mon For M and 29.64 ± 10.36 mon for F 20 min 1 session Timing: 0-immediately after 20 min

Since:25–29 months

(Lee et al., 2013) RCT n (randomized = 16, 16 completed trial) Exp: assistive ergometer training +FES Functional exercise capacity, Berg Balance Scale, and the Korean version of Modified Barthel Index Ergometer training for 4 weeks improved the functional ability of subacute stroke patients. None reported

Age (y):>50 30 min5 weeks4 weeks Peak oxygen consumption, metabolic equivalent, resting and maximal heart rate, resting and maximal blood pressure, maximal rate pressure product, submaximal rate pressure product, submaximal rate of perceived exertion, exercise duration, respiratory exchange ratio, and estimated anaerobic threshold

Sex: 8 M, 8 F Con: assistive ergometer training Timing: 0, 4 wk

Time since stroke: within 6 mo 30 min5 weeks4weeks

Not clear if they have followed the conventional therapies or no.

(Mayo, MacKay-Lyons, Scott, Moriello, & Brophy, 2013) RCT n (randomized = 87, 65 completed trial) Two dose-equivalent interventions, one involving stationary cycling and the other disability-targeted interventions were tested. Primary outcome walking capacity. Both programs were equally effective in maintaining walking capacity after discharge from stroke rehabilitation. No serious adverse events occurred

Age (y):>50 30 minper day1 year Secondary outcomes physical function included comfortable walking speed with usual assistive device, balance

Sex: 37 M, 28 F EXP: One group exercised on a stationary bicycle, Control group carried out mobility exercises and brisk walking. Stroke Impact Scale – Physical, and Physical Function Index of the Medical Outcomes Study RAND-36 Item Health Survey.

Time since stroke: <12 mo Timing: 0, 1 mo, 6 mo, 12 mo

(Yang et al., 2014) RCT Cross-over study trial n (randomized = 31, 30 completed trial) Exp: first period: conventional rehabilitation (2 hours) +cycling training Activity: lower extremity Additional 4-week biofeedback cycling training could lead to improved LE functional recovery, walking endurance, and speed for patients with chronic stroke. None reported

Age (y): 40–60 Second period: conventional rehabilitation(2 hours) walking

Sex: 22 M, 8 F 30 min5 weeks4 weeks spasticity of the knee extensor

Time since stroke:1–20 mo Con: first period: conventional rehabilitation (2 hours) Timing: 0, 4 wk, 1 month (wash out), 2 month

Second period: conventional rehabilitation (2 hours) +cycling training

30 min5 weeks4 weeks

(Bauer, Krewer, Golaszewski, Koenig, & Müller, 2015) RCT n (randomized = 40, 37 completed trial) Exp: Conventional therapy + active leg cycling with FES Primary outcome: Walking ability, balance and gait. FES-assisted active cycling seems to be a promising intervention during rehabilitation in patients with stroke. Adverse events were not observed

Age (y):>18 20 min* 3 weeks* 4 weeks. secondary outcome: evaluating the patient’s ability to generate volitional muscle contraction in the paretic lower limb gait velocity, muscle tone.

Sex: 19 M, 18 F Con: Conventional Therapy + active leg cycling without FES Timing: 0, 4 wk, 6 wk

Time since stroke: 1 wk-24 mo 20 min* 3 weeks* 4 weeks

(Kim, Cho, Kim, & Lee, 2015) RCT n (randomized = 38, 32 completed trial) EXP: rehabilitation program + cycling exercise Balance stationary cycling training is suitable for stroke rehabilitation and may be used in clinical practice None reported

Age (y):>61 60 min(30 + 30) 5 weeks6 weeks Gait function

Sex: 16 F, 16 M Con: rehabilitation program 30 min5 weeks6 weeks Timing: 0 – 6 weeks

Time since stroke:>6 month

(de Sousa, Harvey, Dorsch, Leung, & Harris, 2016) RCT Eligible = 341 n (randomized = 40, 39 completed trial) Exp: conventional rehabilitation + an incremental, progressive, FES cycling program Primary outcomes: Mobility Functional electrical stimulation cycling does not improve mobility in people with acquired brain injury and its effects on strength are unclear None reported

Age (y): 45–75 17, 23, 32 min in per week5 weeks4 weeks Strength of the knee extensors

Sex: 19M, 20 F Con: conventional rehabilitation 60 min* 4 weeks. Secondary outcomes strength of the knee extensors of the unaffected lower limb, strength of key muscles of the affected lower limb and spasticity of the affected plantar flexors

Time since stroke: <6 mo Timing: 0, 4 wk

Study	Design	Patients	Intervention	Outcomes	Conclusion	Adverse effects
(Duncan et al., 2003)	RCT	n (randomized = 100, 92 completed trial)	Exp: Structured, progressive, physiologically based, therapist-supervised, in-home program	Motor recovery of lower and upper extremity	This program produced gains in endurance, balance, and mobility	None reported
		Age (y):>50	90 min3 weeks12–15 weeks	Grip strength
		Sex: 92 M	Con: conventional rehabilitation	Isometric strength of dorsi-flexors and knee extensors
		Time since stroke: 1–5 mo	Gait speed
				walk 6 MWD)
				Balance
				Aerobic capacity
				Timing: 0, 12 wk
(Kamps & Schüle, 2005)	RCT	n (randomized = 40, 37 completed trial)	Exp: conventional rehabilitation + cycling	Gait velocity and the distance walked.	MOTOmed Movement Trainer is a helpful addition to conventional therapy of stroke patients.	None reported
		Age (y): 55–75	10 min in ever twice a day for 4 months	Motor assessments and MOTOmed Data.
		Sex: 22 M, 9 F	Con: conventional physio- and ergotherapeutic interventions.	Timing: 1 to 16 wk
		Time since stroke: 3–27 mo
(Katz-Leurer, Sender, Keren, & Dvir, 2006)	RCT	n (randomized = 24, 23 completed trial)	Exp: Conventional Rehabilitation + cycling	Balance,	Stroke patients can improve their motor and balance abilities after an early short duration of cycling.	None reported
		Age (y):>50	5 weeks*3 weeks	The motor function of the lower extremity.
		Sex: 13 M, 11 F	Con: Conventional Rehabilitation	Timing: 0, 3 wk, 6 wk
		Time since stroke:>1 mo
(Olney et al., 2006)	RCT	n (randomized = 74, 54 completed trial)	Exp: Supervised exercise sessions	primary outcome: walking speed.	Supervised programs showed trends to greater improvements in self-reported gains.	None reported
		Age (y):>20	90 min3 weeks10 weeks	secondary outcome Human Activity Profile.
		Sex: 29 M, 25 F	Con: Home program for 9 (unsupervised).	physical and mental health and quality of life,
		Time since stroke:>12 mo	It was not clear, if they used the conventional/standard rehabilitation as well.	lower extremity muscle strength.
				The walking heart rate and the resting heart rate.
				Timing: 0, 10 wk, 6 mo, 1 yr
(Sullivan et al., 2007)	RCT	n (randomized = 80, 63 completed trial)	4 groups:	Primary outcome: Self-selected walking speed, fast walking speed, and 6-minute walk distance measured before and after intervention and.	After chronic stroke, task-specific training during treadmill walking with body-weight support is more effective in improving walking speed and maintaining these gains at 6 months than resisted leg cycling alone.	There were 17 adverse events during the intervention period, 8 of which were not study related (4 falls in home, 1 report of low back pain, 1 controlled seizure, 1 participant diagnosed with colon cancer, 1 participant diagnosed with congestive heart failure after randomization but prior to starting intervention).
		Age (y):>18	1. Treadmill training and upper-extremity ergometry	secondary outcome the LE-FM motor score, Berg Balance Scale, the 16-item Stroke Impact Scale, version 3.0, Medical Outcomes Study 36-Item Short-Form Health Survey, version 2.0, and LE isometric peak torque
		Sex: 63 M	2. Cycling and upper-extremity ergometry	Timing: 6 we, 6 mo
		Time since stroke: 4 mo to 5 yr	3. Treadmill training and cycling
			4. Treadmill and LE muscle-specific progressive resistive Exercise
			4 weeks*6 weeks
(M. Lee et al., 2008)	RCT	n (randomized = 52, 48 completed trial)	Aerobic cycling plus sham progressive resistance training (PRT), sham cycling plus, aerobic cycling plus PRT, or sham cycling plus sham PRT.	Primary outcomes walk distance, habitual and fast gait velocities, and stair climbing power.	Single-modality exercises targeted at existing impairments do not optimally address the functional deficits of walking	No serious adverse events during testing or training
		Age (y):45–85	60 min3 weeks12 weeks	Secondary outcomes: cardiorespiratory fitness; muscle strength, power, and endurance; and psychosocial attributes.
			Sex: 28 M, 20 F	Timing: 0, 12 wk
		Time since stroke: <3mo
(Ambrosini, Ferrante, Pedrocchi, Ferrigno, & Molteni, 2011)	RCT	n (randomized = 35, 30 completed trial)	Exp: Conventional rehabilitation + Trained cycling and FES	Primary outcome motor power of the paretic lower extremity gait speed.	FES cycling training significantly improved lower extremity motor functions and accelerated the recovery of overground locomotion in postacute hemiparetic patients.	None reported
		Age (y):40–70	25 min5 weeks4 weeks	Secondary outcome Trunk control, functional abilities of the impaired leg during single-limb standing.
			Sex: 18 M, 12 F	Con: Placebo	Timing: 0, 4 wk, 3–6 mo
		Time since stroke: <6 mo
(Lo, Hsu, Hsueh, & Yeh, 2012)	RCT	n (randomized = 20, 20 completed trial)	EXP:CG+FES group	Postural control	Cycling training, with or without FES may reduce spasticity in stroke patients.	None reported
		Age (y):>47	20 min*1 session	Muscle tone
		Sex: 10 F,10 M	Con:CG	Limb dynamic
		Time since stroke: 25.54 ± 12.95 mon For M and 29.64 ± 10.36 mon for F	20 min *1 session	Timing: 0-immediately after 20 min
		Since:25–29 months
(Lee et al., 2013)	RCT	n (randomized = 16, 16 completed trial)	Exp: assistive ergometer training +FES	Functional exercise capacity, Berg Balance Scale, and the Korean version of Modified Barthel Index	Ergometer training for 4 weeks improved the functional ability of subacute stroke patients.	None reported
		Age (y):>50	30 min5 weeks4 weeks	Peak oxygen consumption, metabolic equivalent, resting and maximal heart rate, resting and maximal blood pressure, maximal rate pressure product, submaximal rate pressure product, submaximal rate of perceived exertion, exercise duration, respiratory exchange ratio, and estimated anaerobic threshold
		Sex: 8 M, 8 F	Con: assistive ergometer training	Timing: 0, 4 wk
		Time since stroke: within 6 mo	30 min5 weeks4weeks
			Not clear if they have followed the conventional therapies or no.
(Mayo, MacKay-Lyons, Scott, Moriello, & Brophy, 2013)	RCT	n (randomized = 87, 65 completed trial)	Two dose-equivalent interventions, one involving stationary cycling and the other disability-targeted interventions were tested.	Primary outcome walking capacity.	Both programs were equally effective in maintaining walking capacity after discharge from stroke rehabilitation.	No serious adverse events occurred
		Age (y):>50	30 minper day1 year	Secondary outcomes physical function included comfortable walking speed with usual assistive device, balance
		Sex: 37 M, 28 F	EXP: One group exercised on a stationary bicycle, Control group carried out mobility exercises and brisk walking.	Stroke Impact Scale – Physical, and Physical Function Index of the Medical Outcomes Study RAND-36 Item Health Survey.
		Time since stroke: <12 mo	Timing: 0, 1 mo, 6 mo, 12 mo
(Yang et al., 2014)	RCT Cross-over study trial	n (randomized = 31, 30 completed trial)	Exp: first period: conventional rehabilitation (2 hours) +cycling training	Activity: lower extremity	Additional 4-week biofeedback cycling training could lead to improved LE functional recovery, walking endurance, and speed for patients with chronic stroke.	None reported
		Age (y): 40–60	Second period: conventional rehabilitation(2 hours)	walking
		Sex: 22 M, 8 F	30 min5 weeks4 weeks	spasticity of the knee extensor
		Time since stroke:1–20 mo	Con: first period: conventional rehabilitation (2 hours)	Timing: 0, 4 wk, 1 month (wash out), 2 month
			Second period: conventional rehabilitation (2 hours) +cycling training
			30 min5 weeks4 weeks
(Bauer, Krewer, Golaszewski, Koenig, & Müller, 2015)	RCT	n (randomized = 40, 37 completed trial)	Exp: Conventional therapy + active leg cycling with FES	Primary outcome: Walking ability, balance and gait.	FES-assisted active cycling seems to be a promising intervention during rehabilitation in patients with stroke.	Adverse events were not observed
		Age (y):>18	20 min* 3 weeks* 4 weeks.	secondary outcome: evaluating the patient’s ability to generate volitional muscle contraction in the paretic lower limb gait velocity, muscle tone.
		Sex: 19 M, 18 F	Con: Conventional Therapy + active leg cycling without FES	Timing: 0, 4 wk, 6 wk
		Time since stroke: 1 wk-24 mo	20 min* 3 weeks* 4 weeks
(Kim, Cho, Kim, & Lee, 2015)	RCT	n (randomized = 38, 32 completed trial)	EXP: rehabilitation program + cycling exercise	Balance	stationary cycling training is suitable for stroke rehabilitation and may be used in clinical practice	None reported
		Age (y):>61	60 min(30 + 30) 5 weeks6 weeks	Gait function
		Sex: 16 F, 16 M	Con: rehabilitation program 30 min5 weeks6 weeks	Timing: 0 – 6 weeks
		Time since stroke:>6 month
(de Sousa, Harvey, Dorsch, Leung, & Harris, 2016)	RCT	Eligible = 341 n (randomized = 40, 39 completed trial)	Exp: conventional rehabilitation + an incremental, progressive, FES cycling program	Primary outcomes: Mobility	Functional electrical stimulation cycling does not improve mobility in people with acquired brain injury and its effects on strength are unclear	None reported
		Age (y): 45–75	17, 23, 32 min in per week5 weeks4 weeks	Strength of the knee extensors
		Sex: 19M, 20 F	Con: conventional rehabilitation 60 min* 4 weeks.	Secondary outcomes strength of the knee extensors of the unaffected lower limb, strength of key muscles of the affected lower limb and spasticity of the affected plantar flexors
		Time since stroke: <6 mo	Timing: 0, 4 wk

Con = control group; Exp = experimental group; RCT = randomised clinical/control trial; F = female; M = male; n = number; mo = month; yr = year.

We included studies that examined FES on the function of the lower limb. Frequency of the interventions ranged from 1 to 7 sessions per week (for studies with cycling from 3–7 sessions) and (for studies with from 1–6 sessions) while the length of each session was from 10 to 90 minutes (for studies with cycling from 6 min–90 min) and (for studies with cycling from 17–30 min). The duration of interventions ranged from 3 to 16 weeks (for cycling alone 3–72 weeks and for cycling with FES about 4 weeks), with the total dose of the stimulation between 4 and 16 hours. Parameters of electrical stimulation varied across studies as the frequency ranged from 20 to 60 Hz and pulse width from 300 to 450 ms. Electrical stimulation was provided by the therapist or the participant, either mechanically (e.g., by aid of a foot switch) or physiologically (e.g., by setting a threshold for muscle activity). Participants performed several movements such as ankle dorsiflexion during the stimulation. Finally, for the control intervention, 10 studies used sham stimulation but in 4 studies, control participants received no stimulation.

Results showed that lower limb function was assessed using 2-, 6-, 10-, or 50-meter walking, Timed “Up&Go”-Test (TUG), Berg Balance Scale (BBS), Postural Assessment Scale for Stroke Patients (PASS), Barthel Index (BI), Functional Mobility Scale (FMS), Tinetti-test, Dynamometer, the Stair Climbing Test, the leg subscale of the Motricity Index, Trunk Control Test (TCT), Upright Motor Control Test (UMCT), Functional Ambulation Classification (FAC), or Performance-oriented Mobility Assessment (POMA) in 13 studies. The Modified Ashworth Scale (MAS), FMS, Dynamometer, Wolf Motor Function, or Hoffmans reflex were used for upper limb function in 6 studies. Furthermore, physiological parameters were assessed in five trials using peak aerobic capacity, physiological cost index, VO₂ max, metabolic equivalent (MET), Exercise tolerance test, or not and Short-Form Health Survey RAND-36. In only one study did no participants experience side effects after an intervention (Sullivan et al., 2007). There were 17 adverse events during the intervention period, 8 of which were not study-related (4 falls in home, 1 report of low back pain, 1 controlled seizure, 1 participant diagnosed with colon cancer, 1 participant diagnosed with congestive heart failure after randomization but prior to starting the intervention) (Sullivan et al., 2007).

Study-related events included: (1) in the BWSTT/UE-EX group, minor hand abrasion and foot pain; (2) in the BWSTT/CYCLE group, foot pain, reduced blood pressure, and increased blood pressure (twice in 1 participant); and (3) in the BWSTT/ LE-EX group, gluteus medius muscle pain and toe pain, with a later diagnosis of toe stress fracture (2 occurrences in 1 participant). Four participants were withdrawn from the study by the administration due to the adverse events. Two adverse events were considered related to the study (foot pain, toe stress fracture), and 2 adverse events were considered not related to the study but due to cardiac conditions (congestive heart failure, high blood pressure not responsive to medication). The participant with colon cancer withdrew from the study.

3.3 Risk of bias within studies

Data on risk of bias of each study are presented in Table 2.

Table 2
Cochrane Table of Risk of Bias

Bias Authors’ judgement Support for judgement

1. (Yang et al., 2014)

Random sequence generation (selection bias) Low risk Computer-generated random numbers sealed

Allocation concealment (selection bias) Low risk Envelopes by an independent individual

Blinding of patients and personnel (performance bias) All outcomes Low risk Participant unblinded

Blinding of outcome assessment (detection bias All outcomes Low risk Assessor, therapist

Incomplete outcome data (attrition bias) All outcomes Low risk One participant dropped out from Group A because of falling at home after agreeing to participate

Selective reporting (reporting bias) Low risk The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way

Other bias Unclear None known

2. (de Sousa, Harvey, Dorsch, Leung, & Harris, 2016)

Random sequence generation (selection bias) Low risk Computer software, a person not involved in the trial created a blocked random allocation schedule for 40 patients

Allocation concealment (selection bias) Low risk Allocations were placed in opaque, sequentially numbered and sealed envelopes

Blinding of patients and personnel (performance bias) All outcomes high risk No blinding

Blinding of outcome assessment (detection bias All outcomes Low risk Assessors, therapist

Incomplete outcome data (attrition bias) All outcomes Low risk A person dropped out (death)

Selective reporting (reporting bias) Low risk The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way

Other bias unclear None known

3. (Kamps & Schüle, 2005)

Random sequence generation (selection bias) High risk Not reported

Allocation concealment (selection bias) High risk Not reported

Blinding of patients and personnel (performance bias) All outcomes High risk Unclear

Blinding of outcome assessment (detection bias All outcomes High risk Unclear

Incomplete outcome data (attrition bias) All outcomes Low risk Because of organizational reasons, 9 patients dropped out of the study so we were only able to report data from 31 patients

Selective reporting (reporting bias) Low risk The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way

Other bias Unclear None known

4. (Bauer, Krewer, Golaszewski, Koenig, & Müller, 2015)

Random sequence generation (selection bias) Low risk Not reported

Allocation concealment (selection bias) Low risk Not reported

Blinding of patients and personnel (performance bias) All outcomes Low risk Patients

Blinding of outcome assessment (detection bias All outcomes Low risk Assessor

Incomplete outcome data (attrition bias) All outcomes Low risk 2 patients dropped out

1 = virus infection

1 = sever pusher

Selective reporting (reporting bias) Low risk The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way

Other bias unclear None known

5. (Ambrosini, Ferrante, Pedrocchi, Ferrigno, & Molteni, 2011)

Random sequence generation (selection bias) Low risk A computer-generated randomization sequence

Allocation concealment (selection bias) Low risk An automated assignment system was used

Blinding of patients and personnel (performance bias) All outcomes Low risk Patients and physiotherapist

Blinding of outcome assessment (detection bias All outcomes Low risk Assessor

Incomplete outcome data (attrition bias) All outcomes Low risk 5 subject dropped out

Selective reporting (reporting bias) Low risk The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way

Other bias Unclear None known

6. (Lee et al., 2008)

Random sequence generation (selection bias) Low risk Concealed coded cards compiled in permuted blocks

Allocation concealment (selection bias) Low risk Telephone screening followed by a history and physical examination and a 12-lead electrocardiogram at rest and during a physician-supervised maximal effort cycling test

Blinding of patients and personnel (performance bias) All outcomes Low risk Patients

Blinding of outcome assessment (detection bias All outcomes Low risk Assessor in primary outcome but did not use a blinded assessor for secondary outcomes

Incomplete outcome data (attrition bias) All outcomes Low risk Four patients dropped out during training and declined to come back for the follow-up assessment because of illness unrelated to exercise training or for personal reasons

Selective reporting (reporting bias) Low risk The study protocol is not available but it is clear that the published reports include all expected outcomes, including those that were pre-specified

Other bias Unclear None known

7. (Duncan et al., 2003)

Random sequence generation (selection bias) Low risk Insufficient information

Allocation concealment (selection bias) Low risk Sealed envelopes

Blinding of patients and personnel (performance bias) All outcomes High risk Participant

Blinding of outcome assessment (detection bias All outcomes Low risk Assessor

Incomplete outcome data (attrition bias) All outcomes Low risk 8 dropped out. Six subjects dropped from the intervention arm: 1 had significant renal insufficiency detected after randomization but before therapy; 1 had subclavian steal syndrome diagnosed the first 2 weeks after randomization; 1 chose to withdraw after 18 visits; and 3 experienced a second stroke. Two subjects dropped out from the usual care group: 1 withdrew from the study immediately after randomization, and 1 did not return for the 3-month assessment.

Selective reporting (reporting bias) Low risk The study protocol is not available but it is clear that the published reports include all expected outcomes, including those that were pre-specified

Other bias Unclear None known

8. (Katz-Leurer, Sender, Keren, & Dvir, 2006)

Random sequence generation (selection bias) Low risk A random number table

Allocation concealment (selection bias) High risk Unclear

Blinding of patients and personnel (performance bias) All outcomes High risk Insufficient information

Blinding of outcome assessment (detection bias All outcomes Low risk Assessor

Incomplete outcome data (attrition bias) All outcomes Low risk No subjects dropped out of the study. However, one woman from the exercise group left the hospital before final measurements (six weeks) were performed.

Selective reporting (reporting bias) Low risk The study protocol is available and all of the study’s pre -specified outcomes that are of interest in the review have been reported in the pre-specified way

Other bias Unclear None known

9. (Olney et al., 2006)

Random sequence generation (selection bias) Low risk Computer-generated randomization list

Allocation concealment (selection bias) Low risk Sealed envelopes

Blinding of patients and personnel (performance bias) All outcomes High risk Patients unblinded

Blinding of outcome assessment (detection bias All outcomes Low risk Therapist blinded

Incomplete outcome data (attrition bias) All outcomes Low risk One subject from Kingston randomized to the supervised program and 1 subject from Ottawa randomized to the unsupervised program refused to complete the program

Selective reporting (reporting bias) Low risk The study protocol is not available but it is clear that the published reports include all expected outcomes, including those that were pre-specified

Other bias Unclear None Known

10. (Lee et al., 2013)

Random sequence generation (selection bias) High risk Insufficient information

Allocation concealment (selection bias) High risk Insufficient information

Blinding of patients and personnel (performance bias) All outcomes High risk Insufficient information

Blinding of outcome assessment (detection bias All outcomes High risk Insufficient information

Incomplete outcome data (attrition bias) All outcomes Low risk All patients completed their therapy sessions

Selective reporting (reporting bias) Low Risk All of outcomes were reported

Other bias Unclear None known

11. (Sullivan et al., 2007)

Random sequence generation (selection bias) High risk Insufficient information

Allocation concealment (selection bias) High risk Insufficient information

Blinding of patients and personnel (performance bias) All outcomes High risk Patients unblinded

Blinding of outcome assessment (detection bias All outcomes Low risk Assessor blinded

Incomplete outcome data (attrition bias) All outcomes High risk 8 patients withdrew and one person refused follow-up assessment

Selective reporting (reporting bias) Low risk The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way

Other bias Unclear None known

12. (Mayo, MacKay-Lyons, Scott, Moriello, & Brophy, 2013)

Random sequence generation (selection bias) Low risk Computer-generated allocation of consecutively recruited patients in randomized blocks of two, four, and six; concealed in opaque envelopes

Allocation concealment (selection bias) Low risk Computer-generated allocation of consecutively recruited patients in randomized blocks of two, four, and six, and concealed in opaque envelopes

Blinding of patients and personnel (performance bias) All outcomes High risk Patients unblinded

Blinding of outcome assessment (detection bias All outcomes Low risk Assessor blinded

Incomplete outcome data (attrition bias) All outcomes High risk 15 patients did not complete their intervention

Selective reporting (reporting bias) Low risk The study protocol is available and all of the study’s pre -specified outcomes that are of interest in the review have been reported in the pre-specified way

Other bias Unclear None known

13. (Lo, Hsu, Hsueh, & Yeh, 2012)

Random sequence generation (selection bias) High risk Insufficient information

Allocation concealment (selection bias) High risk Not reported

Blinding of patients and personnel (performance bias) All outcomes High risk Not reported

Blinding of outcome assessment (detection bias All outcomes High risk Not reported

Incomplete outcome data (attrition bias) All outcomes High risk Insufficient information

Selective reporting (reporting bias) Low risk The study protocol is available and all of the study’s pre -specified outcomes that are of interest in the review have been reported in the pre-specified way

Other bias Unclear None known

14. (Kim, Cho, Kim, & Lee, 2015)

Random sequence generation (selection bias) High risk Insufficient information

Allocation concealment (selection bias) High risk Not reported

Blinding of patients and personnel (performance bias) All outcomes High risk Patients unblinded

Blinding of outcome assessment (detection bias All outcomes Low risk Assessor blinded

Incomplete outcome data (attrition bias) All outcomes Low risk Reported fully

Selective reporting (reporting bias) Low risk The study protocol is available and all of the study’s pre -specified outcomes are reported in the pre-specified way

Other bias Unclear None known

Bias	Authors’ judgement	Support for judgement
1. (Yang et al., 2014)
Random sequence generation (selection bias)	Low risk	Computer-generated random numbers sealed
Allocation concealment (selection bias)	Low risk	Envelopes by an independent individual
Blinding of patients and personnel (performance bias) All outcomes	Low risk	Participant unblinded
Blinding of outcome assessment (detection bias All outcomes	Low risk	Assessor, therapist
Incomplete outcome data (attrition bias) All outcomes	Low risk	One participant dropped out from Group A because of falling at home after agreeing to participate
Selective reporting (reporting bias)	Low risk	The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way
Other bias	Unclear	None known
2. (de Sousa, Harvey, Dorsch, Leung, & Harris, 2016)
Random sequence generation (selection bias)	Low risk	Computer software, a person not involved in the trial created a blocked random allocation schedule for 40 patients
Allocation concealment (selection bias)	Low risk	Allocations were placed in opaque, sequentially numbered and sealed envelopes
Blinding of patients and personnel (performance bias) All outcomes	high risk	No blinding
Blinding of outcome assessment (detection bias All outcomes	Low risk	Assessors, therapist
Incomplete outcome data (attrition bias) All outcomes	Low risk	A person dropped out (death)
Selective reporting (reporting bias)	Low risk	The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way
Other bias	unclear	None known
3. (Kamps & Schüle, 2005)
Random sequence generation (selection bias)	High risk	Not reported
Allocation concealment (selection bias)	High risk	Not reported
Blinding of patients and personnel (performance bias) All outcomes	High risk	Unclear
Blinding of outcome assessment (detection bias All outcomes	High risk	Unclear
Incomplete outcome data (attrition bias) All outcomes	Low risk	Because of organizational reasons, 9 patients dropped out of the study so we were only able to report data from 31 patients
Selective reporting (reporting bias)	Low risk	The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way
Other bias	Unclear	None known
4. (Bauer, Krewer, Golaszewski, Koenig, & Müller, 2015)
Random sequence generation (selection bias)	Low risk	Not reported
Allocation concealment (selection bias)	Low risk	Not reported
Blinding of patients and personnel (performance bias) All outcomes	Low risk	Patients
Blinding of outcome assessment (detection bias All outcomes	Low risk	Assessor
Incomplete outcome data (attrition bias) All outcomes	Low risk	2 patients dropped out
1 = virus infection
1 = sever pusher
Selective reporting (reporting bias)	Low risk	The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way
Other bias	unclear	None known
5. (Ambrosini, Ferrante, Pedrocchi, Ferrigno, & Molteni, 2011)
Random sequence generation (selection bias)	Low risk	A computer-generated randomization sequence
Allocation concealment (selection bias)	Low risk	An automated assignment system was used
Blinding of patients and personnel (performance bias) All outcomes	Low risk	Patients and physiotherapist
Blinding of outcome assessment (detection bias All outcomes	Low risk	Assessor
Incomplete outcome data (attrition bias) All outcomes	Low risk	5 subject dropped out
Selective reporting (reporting bias)	Low risk	The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way
Other bias	Unclear	None known
6. (Lee et al., 2008)
Random sequence generation (selection bias)	Low risk	Concealed coded cards compiled in permuted blocks
Allocation concealment (selection bias)	Low risk	Telephone screening followed by a history and physical examination and a 12-lead electrocardiogram at rest and during a physician-supervised maximal effort cycling test
Blinding of patients and personnel (performance bias) All outcomes	Low risk	Patients
Blinding of outcome assessment (detection bias All outcomes	Low risk	Assessor in primary outcome but did not use a blinded assessor for secondary outcomes
Incomplete outcome data (attrition bias) All outcomes	Low risk	Four patients dropped out during training and declined to come back for the follow-up assessment because of illness unrelated to exercise training or for personal reasons
Selective reporting (reporting bias)	Low risk	The study protocol is not available but it is clear that the published reports include all expected outcomes, including those that were pre-specified
Other bias	Unclear	None known
7. (Duncan et al., 2003)
Random sequence generation (selection bias)	Low risk	Insufficient information
Allocation concealment (selection bias)	Low risk	Sealed envelopes
Blinding of patients and personnel (performance bias) All outcomes	High risk	Participant
Blinding of outcome assessment (detection bias All outcomes	Low risk	Assessor
Incomplete outcome data (attrition bias) All outcomes	Low risk	8 dropped out. Six subjects dropped from the intervention arm: 1 had significant renal insufficiency detected after randomization but before therapy; 1 had subclavian steal syndrome diagnosed the first 2 weeks after randomization; 1 chose to withdraw after 18 visits; and 3 experienced a second stroke. Two subjects dropped out from the usual care group: 1 withdrew from the study immediately after randomization, and 1 did not return for the 3-month assessment.
Selective reporting (reporting bias)	Low risk	The study protocol is not available but it is clear that the published reports include all expected outcomes, including those that were pre-specified
Other bias	Unclear	None known
8. (Katz-Leurer, Sender, Keren, & Dvir, 2006)
Random sequence generation (selection bias)	Low risk	A random number table
Allocation concealment (selection bias)	High risk	Unclear
Blinding of patients and personnel (performance bias) All outcomes	High risk	Insufficient information
Blinding of outcome assessment (detection bias All outcomes	Low risk	Assessor
Incomplete outcome data (attrition bias) All outcomes	Low risk	No subjects dropped out of the study. However, one woman from the exercise group left the hospital before final measurements (six weeks) were performed.
Selective reporting (reporting bias)	Low risk	The study protocol is available and all of the study’s pre -specified outcomes that are of interest in the review have been reported in the pre-specified way
Other bias	Unclear	None known
9. (Olney et al., 2006)
Random sequence generation (selection bias)	Low risk	Computer-generated randomization list
Allocation concealment (selection bias)	Low risk	Sealed envelopes
Blinding of patients and personnel (performance bias) All outcomes	High risk	Patients unblinded
Blinding of outcome assessment (detection bias All outcomes	Low risk	Therapist blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	One subject from Kingston randomized to the supervised program and 1 subject from Ottawa randomized to the unsupervised program refused to complete the program
Selective reporting (reporting bias)	Low risk	The study protocol is not available but it is clear that the published reports include all expected outcomes, including those that were pre-specified
Other bias	Unclear	None Known
10. (Lee et al., 2013)
Random sequence generation (selection bias)	High risk	Insufficient information
Allocation concealment (selection bias)	High risk	Insufficient information
Blinding of patients and personnel (performance bias) All outcomes	High risk	Insufficient information
Blinding of outcome assessment (detection bias All outcomes	High risk	Insufficient information
Incomplete outcome data (attrition bias) All outcomes	Low risk	All patients completed their therapy sessions
Selective reporting (reporting bias)	Low Risk	All of outcomes were reported
Other bias	Unclear	None known
11. (Sullivan et al., 2007)
Random sequence generation (selection bias)	High risk	Insufficient information
Allocation concealment (selection bias)	High risk	Insufficient information
Blinding of patients and personnel (performance bias) All outcomes	High risk	Patients unblinded
Blinding of outcome assessment (detection bias All outcomes	Low risk	Assessor blinded
Incomplete outcome data (attrition bias) All outcomes	High risk	8 patients withdrew and one person refused follow-up assessment
Selective reporting (reporting bias)	Low risk	The study protocol is available and all of the study’s pre -specified (primary and secondary) outcomes that are of interest in the review have been reported in the pre-specified way
Other bias	Unclear	None known
12. (Mayo, MacKay-Lyons, Scott, Moriello, & Brophy, 2013)
Random sequence generation (selection bias)	Low risk	Computer-generated allocation of consecutively recruited patients in randomized blocks of two, four, and six; concealed in opaque envelopes
Allocation concealment (selection bias)	Low risk	Computer-generated allocation of consecutively recruited patients in randomized blocks of two, four, and six, and concealed in opaque envelopes
Blinding of patients and personnel (performance bias) All outcomes	High risk	Patients unblinded
Blinding of outcome assessment (detection bias All outcomes	Low risk	Assessor blinded
Incomplete outcome data (attrition bias) All outcomes	High risk	15 patients did not complete their intervention
Selective reporting (reporting bias)	Low risk	The study protocol is available and all of the study’s pre -specified outcomes that are of interest in the review have been reported in the pre-specified way
Other bias	Unclear	None known
13. (Lo, Hsu, Hsueh, & Yeh, 2012)
Random sequence generation (selection bias)	High risk	Insufficient information
Allocation concealment (selection bias)	High risk	Not reported
Blinding of patients and personnel (performance bias) All outcomes	High risk	Not reported
Blinding of outcome assessment (detection bias All outcomes	High risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	Insufficient information
Selective reporting (reporting bias)	Low risk	The study protocol is available and all of the study’s pre -specified outcomes that are of interest in the review have been reported in the pre-specified way
Other bias	Unclear	None known
14. (Kim, Cho, Kim, & Lee, 2015)
Random sequence generation (selection bias)	High risk	Insufficient information
Allocation concealment (selection bias)	High risk	Not reported
Blinding of patients and personnel (performance bias) All outcomes	High risk	Patients unblinded
Blinding of outcome assessment (detection bias All outcomes	Low risk	Assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	Reported fully
Selective reporting (reporting bias)	Low risk	The study protocol is available and all of the study’s pre -specified outcomes are reported in the pre-specified way
Other bias	Unclear	None known

3.4 Synthesis of results

3.4.1 Cycling vs. control on walking speed

The effect of cycling on walking speed was examined by pooling data after the intervention from 6 trials consisting of 249 participants. Cycling improved walking speed compared to control (SMD, 0.30; 95% CI, 0.05 –0.55; I² = 0% ). Five trials used 10 MWK test and one trial (Olney et al., 2006) used 6 MWK test as an outcome measurement. The overall results showed that cycling had a positive and significant effect on walking speed in comparison with control (P < 0.02) (Fig. 2).

Fig.2

Standardized mean difference (95% CI) of walking speed after cycling as compared with a control from five studies. Sensitivity analysis showed that walking speed was significant except by excluding one trial at a time from pooled effects to determine whether any one study was particularly influential. Susceptibility analysis showed there is stability results for this meta-analysis.

3.4.2 Cycling vs control on walking ability

The effect of cycline vs. no training was examined by pooling data after intervention from 5 trials consisting of 177 participants. Cycling improved walking ability compared with control (SMD, 0.41; 95% CI, 0.11 –0.71; I² = 0%). The analysis included 4 trials of walking ability measured by the 6 MW test and one trial (Kim, Cho, Kim, & Lee, 2015) used 10 MWK test as outcome measurements. The analysis included 4 RCTs and one cross-over study (measure of T₂ after the first period intervention and control) of walking speed that measured 6 MW test. The overall results showed that cycling had a positive effect on walking ability in comparison to a control group, and this effect was significant (P < 0.007) (Fig. 3).

Fig.3

Standardized mean difference (95% CI) of walking ability after cycling as compared with a control from four studies. Sensitivity analysis showed that walking ability was significant by excluding one trial at a time from pooled effects to determine whether any study was particularly influential. Susceptibility analysis showed by excluding studies with <20 subjects there was stability of results for this meta-analysis.

3.4.3 Cycling vs control on balance

The effect of cycling vs. a control group was examined by pooling data after intervention from 5 trials consisting of 251 participants. Cycling improved balance compared with a control (SMD, 0.32; 95% CI, 0.06 –0.57; I² = 49%). The analysis included 4 trials of balance that were measured using the Berg Balance Scale but one trial (Katz-Leurer, Sender, Keren, & Dvir, 2006) measured balance using the postural assessment scale for patients post-stroke (PASS). The overall results showed that cycling had a significant effect on balance in comparison with control (P < 0.01) (Fig. 4).

Fig.4

Standardized mean difference (95% CI) of balance after cycling as compared with a control from five studies. Sensitivity analysis showed that Balance was significant except by excluding one trial at a time from pooled effects to determine whether any one study was particularly influential. Susceptibility analysis showed by excluding studies with <20 subjects there was stability of results for this meta-analysis.

3.4.4 Cycling with FES vs. control on balance

The effect of cycling with FES vs. a control group was examined by pooling data from 2 trials consisting of 96 participants (Bauer et al., 2015; S. Y. Lee et al., 2013). Cycling with FES improved balance compared with a control group (SMD, 1.48; 95% CI, 0.99 –1.97; I² = 91%). The analysis included one trial of balance that was measured using the Berg Balance Scale but one trial (Bauer et al., 2015) used the Functional ambulation classification (FAC) and performance-oriented mobility assessment (POMA). The overall results indicated that cycling with FES had a significant and positive effect on balance (P < 0.00001) (Fig. 5).

Fig.5

Standardized mean (95% CI) of balance after cycling with FES as compared with a control from two studies. Sensitivity analysis showed that balance was significant by excluding one trial at a time from pooled effects to determine whether any one study was particularly influential. Suceptability analysis showed by excluding studies with <20 subjects there was stability of results for this meta-analysis.

4 Discussion

4.1 Summary of main results

We completed a systematic review and meta-analysis examining the effects of cycling with/without FES on the lower limbs in survivors post stroke compared to control groups. We identified 14 randomized controlled trials (479 participants) investigating this clinical question. In this review participants’ age range was from 42.5 to 85 years. The mean time after stroke ranged from > 1 to 27 months. The duration of interventions ranged from 3 to 16 weeks (for cycling alone 3–72 weeks and for cycling with FES about 4 weeks), with the total dose of the stimulation between 4 and 16 hours. Parameters of electrical stimulation varied across studies as the frequency ranged from 20 to 60 Hz and pulse width from 300 to 450 ms.

However, because of a lack of sufficient data, we could not examine the effects of this on lower limb function and balance in the long-term. We cannot suggest the most effective protocol of cycling with/without FES for patients post stroke with lower limb disability because of the inconsistency of previous trial methods.

Our results showed that cycling had a positive effect on the walking ability (SMD, 0.41; 95% CI, 0.11 –0.71; I² = 0%; P < 0.007) compared with no or placebo intervention (control). Compared with control, cycling had a positive effect on walking speed (SMD, 0.30; 95% CI, 0.05 –0.55; I² = 0%; P < 0.02), and on balance based on the Berg score (SMD, 0.32; 95% CI, 0.06 –0.57; I² = 49%; P < 0.01). Concurrent use of cycling with functional electrical stimulation had a positive effect on balance compared with control? (SMD, 1.48; 95% CI, 0.99 –1.97; I² = 91%; P < 0.00001) and compared with cycling alone, the effect was higher.

4.2 Limitations

There were some limitations to the literature upon which this systematic review was based. First, we included all applicable studies, however many of them included small sample sizes which may limit their generalizability. Additionally, we only included English language literature. Third, studies used different measures of function and disability which required us to report combined outcomes as the standardized mean difference which is difficult for clinicians to translate to actual practice. Fourth, no systematic searching was done for unpublished literature. Fifth, primary and secondary outcomes were not defined clearly in all the studies.

4.3 Agreements and disagreements with other studies or reviews

In a previous systematic review with meta-analysis (Howlett et al., 2015). Howlett et al. have examined the effects of FES on physical activity of patients post stroke generally focusing on lower-limb or upper-limb function. The results showed that FES had a small to moderate positive effect on activity compared with no or placebo intervention. The results suggested that FES was more beneficial than training alone with a moderate effect size. These authors used lower limb outcomes mostly focused on walking speed. Balance, which is important in functional ability of the lower limb, was not assessed. Additionally, they did not examine the effects of cycling combined with FES, so it was not possible to predict if cycling with FES is more effective for balance compared with cycling or FES alone.

In a meta-analysis, Robbins et al. reported that FES resulted in significant positive effects on increasing walking speed compared to walking training alone or no intervention in patients post stroke, based on the results of three controlled trials in patients chronic post stroke (Robbins, Houghton, Woodbury, & Brown, 2006). However the effects on balance, other determinants of functional ability of lower extremity were not evaluated. Pereira et al. in 2012 reported that FES improved walking distance in the 6-min walk test with a small but significant treatment effect, based on six controlled trials but walking ability and balance were not assessed (Pereira, Mehta, McIntyre, Lobo, & Teasell, 2012).

5 Conclusion

5.1 Implications for practice

Rehabilitation may offer additional benefit. However, the very limited literature suggests that more studies are needed comparing FES cycling directly with other modalities of exercise such as balance training, strength training, power training or combinations, to determine its relative efficacy. In addition, the additional cost and lack of generalizability of FES cycling to home or non-specialized rehabilitation services is an important issue that needs to be considered.

5.2 Implications for research

Cycling is superior to control for improving walking speed, walking ability, and balance. Cycling with FES has a significant and positive effect on balance compared to cycling without FES. Although more research is needed, patients post stroke with lower limb disability could use cycling with FES as part of their rehabilitation program. In order to determine the ideal protocol of prescribing cycling with FES as a part of post stroke more research is needed and future research should including clinically meaningful outcomes related to functional mobility such as falls and fall-related injuries and other co-morbidities related to higher physical activity levels and community ambulation.

Conflict of interest

The author(s) declared no potential conflicts of interests with respect to the research, authorship, and/or publication of this article.

Funding

This study was supported by Sports Medicine Research Center, Neuroscience Institute, Tehran University of Medical Sciences, Tehran, Iran (IR.TUMS.VCR.REC.1396.4226).

Footnotes

Acknowledgments

Our special thanks go to Prof. Ramin Kordi, Sports Medicine Research Center, Neuroscience Institute, Tehran University of Medical Sciences, Tehran, Iran, for his support.

References

Ambrosini,

, Ferrante,

, Pedrocchi,

, Ferrigno,

, & Molteni,

(2011). Cycling induced by electrical stimulation improves motor recovery in postacute hemiparetic patients, Stroke, 42(4), 1068–1073.

Balshem,

, Helfand,

, Schünemann,

H. J

, Oxman,

A. D

, Kunz,

, Brozek,

, ... & Norris,

(2011). GRADE guidelines: Rating the quality of evidence. Journal of Clinical Epidemiology, 64(4), 401–406.

Barbosa,

, Santos,

C. P

, & Martins,

(2015). The application of cycling and cycling combined with feedback in the rehabilitation of stroke patients: A review. Journal of Stroke and Cerebrovascular Diseases, 24(2), 253–273.

Bauer,

, Krewer,

, Golaszewski,

, Koenig,

, & Müller,

(2015). Functional Electrical Stimulation–Assisted Active Cycling— Therapeutic Effects in Patients With Hemiparesis From Days to Months After Stroke: A Randomized Controlled Pilot Study. Archives of Physical Medicine and Rehabilitation, 96(2), 188–196.

Billinger,

S. A

, Arena,

, Bernhardt,

, Eng,

J. J

, Franklin,

B. A

, Johnson,

C. M

, ... Roth,

E. J.

(2014). Physical activity and exercise recommendations for stroke survivors. Stroke, 45(8), 2532–2553.

Cohn,

(1988). Statistical power analysis for the behavioral sciences. Lawrence Earlbam Associates, Hillsdale, NJ.

de Sousa,

D. G

, Harvey,

L. A

, Dorsch,

, Leung,

, & Harris,

(2016). Functional electrical stimulation cycling does not improve mobility in people with acquired brain injury and its effects on strength are unclear: A randomised trial. Journal of Physiotherapy, 62(4), 203–208.

Duncan,

, Studenski,

, Richards,

, Gollub,

, Lai,

S. M

, Reker,

, ... Rigler,

(2003). Randomized clinical trial of therapeutic exercise in subacute stroke. Stroke, 34(9), 2173–2180.

Feigin,

V. L

, Forouzanfar,

M. H

, Krishnamurthi,

, Mensah,

G. A

, Connor,

, Bennett,

D. A

, ... Truelsen,

(2014). Global and regional burden of stroke during: Findings from the Global Burden of Disease Study. The Lancet, 383(9913), 245–255.

10.

Gattie,

, Cleland,

J. A

, & Snodgrass,

(2017). The Effectiveness of Trigger Point Dry Needling for Musculoskeletal Conditions by Physical Therapists: A Systematic Review and Meta-analysis. Journal of Orthopaedic & Sports Physical Therapy, 47(3), 133–149.

11.

Guyatt,

G. H

, Oxman,

A. D

, Vist,

G. E

, Kunz,

, Falck-Ytter,

, & Alonso-Coello,

P. (n.d.)

. GRADE: An emerging consensus on rating quality of evidence and strength of recommendations. BMJ [Internet]. 2008 Apr 26 [cited 2011 Aug 8]; 336(7650), 924–926.

12.

Higgins,

J. P. T

, & Green,

(2011). Cochrane handbook for systematic reviews of interventions (Vol. 4). John Wiley & Sons.

13.

Higgins,

J. P. T

, Thompson,

S. G

, Deeks,

J. J

, & Altman,

D. G.

(2003). Measuring inconsistency in meta-analyses. BMJ: British Medical Journal, 327(7414), 557.

14.

Howlett,

O. A

, Lannin,

N. A

, Ada,

, & McKinstry,

(2015). Functional electrical stimulation improves activity after stroke: A systematic review with meta-analysis. Archives of Physical Medicine and Rehabilitation, 96(5), 934–943.

15.

Janssen,

T. W

, Beltman,

J. M

, Elich,

, Koppe,

P. A

, Konijnenbelt,

, de

Haan, A

, & Gerrits,

K. H.

(2008). Effects of Electric Stimulation–Assisted Cycling Training in People With Chronic Stroke. Archives of Physical Medicine and Rehabilitation, 89(3), 463–469.

16.

Jones,

E. G.

(2000). Cortical and subcortical contributions to activity-dependent plasticity in primate somatosensory cortex. Annual Review of Neuroscience, 23(1), 1–37.

17.

Katz-Leurer,

, Sender,

, Keren,

, & Dvir,

(2006). The influence of early cycling training on balance in stroke patients at the subacute stage. Results of a preliminary trial. Clinical Rehabilitation, 20(5), 398–405.

18.

Kim,

, Cho,

, Kim,

Y. L

, & Lee,

(2015). Effects of stationary cycling exercise on the balance and gait abilities of chronic stroke patients. Journal of Physical Therapy Science, 27(11), 3529–3531.

19.

Kwakkel,

, Veerbeek,

J. M

, van

Wegen, E. E. H

, Wolf,

S. L.

(2015). Constraint-induced movement therapy after stroke. The Lancet Neurology, 14(2), 224–234.

20.

Landis,

J. R

, Koch,

G. G.

(1977). The measurement of observer agreement for categorical data. Biometrics 159–174.

21.

Lee,

, Kilbreath,

S. L

, Singh,

M. F

, Zeman,

, Lord,

S. R

, Raymond,

, & Davis,

G. M.

(2008). Comon of effect of aerobic cycle training and progressive resistance training on walking ability after stroke: A randomized Sham exercise–controlled study. Journal of the American Geriatrics Society, 56(6), 976–985 paris.

22.

Lee,

S. Y

, Kang,

S.-Y

, Im,

S. H

, Kim,

B. R

, Kim,

S. M

, Yoon,

H. M

, & Han,

E. Y.

(2013). The effects of assisted ergometer training with a functional electrical stimulation on exercise capacity and functional ability in subacute stroke patients. Annals of Rehabilitation Medicine, 37(5), 619–627.

23.

Lehmann,

M. F

, Alfieri,

D. F

, Meloni,

M. Z

, Flauzino,

, Scavuzzi,

B. M

, de Oliveira,

S. R

, ... Reiche,

E. M. V.

(2015). Metabolic syndrome increases oxidative stress but does not influence disability and short-time outcome in acute ischemic stroke patients. Metabolic Brain Disease, 30(6), 1409–1416.

24.

Mazzocchio,

, Meunier,

, Ferrante,

, Molteni,

, & Cohen,

L. G.

(2008). Cycling, a tool for locomotor recovery after motor lesions? Neuro Rehabilitation, 23(1), 67–80.

25.

Moher,

, Liberati,

, Tetzlaff,

, Altman,

D. G

, & Group,

(2009). Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. e. PLoS Medicine, 6(7), 1000097.

26.

Nudo,

R. J

, Wise,

B. M

, SiFuentes,

, Milliken,

G. W.

(1996). Neural substrates for the effects of rehabilitative training on motor recovery after ischemic infarct. Science, 272(5269), 1791.

27.

Olney,

S. J

, Nymark,

, Brouwer,

, Culham,

, Day,

, Heard,

, ... Parvataneni,

(2006). A randomized controlled trial of supervised versus unsupervised exercise programs for ambulatory stroke survivors. Stroke, 37(2), 476–481.

28.

Pereira,

, Mehta,

, McIntyre,

, Lobo,

, & Teasell,

R. W.

(2012). Functional electrical stimulation for improving gait in persons with chronic stroke. Topics in Stroke Rehabilitation, 19(6), 491–498.

29.

Popović,

D. B

, Sinkjær,

, & Popović,

M. B.

(2009). Electrical stimulation as a means for achieving recovery of function in stroke patients. NeuroRehabilitation, 25(1), 45–58.

30.

Robbins,

S. M

, Houghton,

P. E

, Woodbury,

M. G

, & Brown,

J. L.

(2006). The therapeutic effect of functional and transcutaneous electric stimulation on improving gait speed in stroke patients: A meta-analysis. Archives of Physical Medicine and Rehabilitation, 87(6), 853–859.

31.

Rossini,

P. M

, Calautti,

, Pauri,

, & Baron,

J.-C.

(2003). Post-stroke plastic reorganisation in the adult brain. The Lancet Neurology, 2(8), 493–502.

32.

Saragiotto,

B. T

, Maher,

C. G

, Yamato,

T. P

, Costa,

L. O. P

, Costa,

L. C. M

, Ostelo,

R. W. J. G

, & Macedo,

L. G.

(2016). Motor control exercise for nonspecific low back pain: A Cochrane Review. Spine, 41(16), 1284–1295.

33.

Shariat,

, Ansari,

N. N

, Shaw,

B. S

, Kordi,

, Kargarfard,

, & Shaw,

(2018). Cycling training and functional electrical stimulation for post-stroke patients. Revista Brasileira de Medicina Do Esporte, 24(4), 300–302.

34.

Sterne,

J. A. C

, Sutton,

A. J

, Ioannidis,

J. P. A

, Terrin,

, Jones,

D. R

, Lau,

, ... Schmid,

C. H.

(2011). Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. Bmj, 343, d4002.

35.

Sullivan,

K. J

, Brown,

D. A

, Klassen,

, Mulroy,

, Ge,

, Azen,

S. P

, & Winstein,

C. J.

(2007). Effects of task-specific locomotor and strength training in adults who were ambulatory after stroke: Results of the STEPS randomized clinical trial. Physical Therapy, 87(12), 1580–1602.

36.

Takeshima,

, Sozu,

, Tajika,

, Ogawa,

, Hayasaka,

, & Furukawa,

T. A.

(2014). Which is more generalizable, powerful and interpretable in meta-analyses, mean difference or standardized mean difference? BMC Medical Research Methodology, 14(1), 30.

37.

Timmermans,

, Roerdink,

, van Ooijen,

M. W

, Meskers,

C. G

, Janssen,

T. W

, & Beek,

P. J.

(2016). Walking adaptability therapy after stroke: Study protocol for a randomized controlled trial. Trials, 17(1), 425.

38.

Valles,

K. B

, Montes,

, de

Jesus Madrigal, M

, Burciaga,

, Martínez,

M. E

, & Johnson,

M. J.

(2016). Technology-assisted stroke rehabilitation in Mexico: A pilot randomized trial comparing traditional therapy to circuit training in a Robot/technology-assisted therapy gym. Journal of Neuroengineering and Rehabilitation, 13(1), 83.

39.

Wang,

, Zhang,

, Ayala,

, Dunet,

D. O

, Fang,

, & George,

M. G.

(2014). Costs of hospitalization for stroke patients aged years in the United States. Journal of Stroke and Cerebrovascular Diseases, 23(5), 861–868.

40.

Yang,

H.-C

, Lee,

C.-L

, Lin,

, Hsu,

M.-J

, Chen,

C.-H

, Lin,

J.-H

, & Lo,

S. K.

(2014). Effect of biofeedback cycling training on functional recovery and walking ability of lower extremity in patients with stroke. The Kaohsiung Journal of Medical Sciences, 30(1), 35–42.