Abstract
Rationale
Walking difficulties are common after a stroke. During rehabilitation, electromechanical and robotic gait‐training devices can help improve walking. As the evidence and certainty of the evidence may have changed since our last update in 2020, we aimed to update the scientific evidence on the benefits and acceptability of these technologies to ensure they remain a viable option for stroke rehabilitation.
Objectives
Primary
• To determine whether electromechanical‐ and robot‐assisted gait training versus physiotherapy (or usual care) improves walking in adults after stroke.
Secondary
• To determine whether electromechanical‐ and robot‐assisted gait training versus physiotherapy (or usual care) after stroke improves walking velocity, walking capacity, acceptability, and death from all causes until the end of the intervention phase.
Search methods
We searched CENTRAL, MEDLINE, Embase, and seven other databases. We handsearched relevant conference proceedings, searched trials and research registers, checked reference lists, and contacted trial authors to identify further published, unpublished, and ongoing trials. The date of the latest search was December 2023.
Eligibility criteria
We included all randomised controlled trials and randomised controlled cross‐over trials in people over the age of 18 years diagnosed with stroke of any severity, at any stage, in any setting, evaluating electromechanical‐ and robot‐assisted gait training versus physiotherapy (or usual care).
Outcomes
Our critical outcome was the ability to walk independently, measured with the Functional Ambulation Category (FAC). An FAC score of 4 or 5 indicated independent walking over a 15‐metre surface, irrespective of aids used, such as a cane. An FAC score less than 4 indicates dependency in walking (supervision or assistance, or both, must be given in performing walking). Important outcomes included walking velocity and capacity, as well as dropouts.
Risk of bias
We used Cochrane's RoB 1 tool.
Synthesis methods
Two review authors independently selected trials for inclusion, assessed methodological quality and risk of bias, and extracted data. We used random‐effects models for the meta‐analysis. We assessed the certainty of evidence using the GRADE approach.
Included studies
We included 101 studies (39 new studies plus 62 studies from previous versions) with a total of 4224 participants after stroke in our review update.
Synthesis of results
Electromechanical‐assisted gait training in combination with physiotherapy probably increases the odds of participants becoming independent in walking (odds ratio (OR) 1.65, 95% confidence interval (CI) 1.21 to 2.25; P = 0.001; I² = 31%; 51 studies, 2148 participants; moderate‐certainty evidence); probably does not increase mean walking velocity (mean difference (MD) 0.05 m/s, 95% CI 0.02 to 0.08; P < 0.001; I² = 58%; 73 studies, 3043 participants; moderate‐certainty evidence); and does not increase mean walking capacity (MD 11 metres walked in 6 minutes, 95% CI 1.8 to 20.3; P = 0.02; I² = 43%; 42 studies, 1966 participants; high‐certainty evidence). Electromechanical‐assisted gait training does not increase or decrease the risk of loss to the study during the intervention or the risk of death from all causes (high‐certainty evidence).
At follow‐up after study end, electromechanical‐assisted gait training in combination with physiotherapy may not increase the odds of participants becoming independent in walking (OR 1.64, 95% CI 0.77 to 3.48; P = 0.20; I² = 69%; 8 studies, 569 participants; low‐certainty evidence), and probably does not increase mean walking velocity (MD 0.05 m/s, 95% CI −0.03 to 0.13; P = 0.22; I² = 66%; 17 studies, 857 participants; moderate‐certainty evidence) or mean walking capacity (MD 9.6 metres walked in 6 minutes, 95% CI −14.6 to 33.7; P = 0.44; I² = 53%; 15 studies, 736 participants; moderate‐certainty evidence).
Our results must be interpreted with caution because (1) some trials investigated people who were independent in walking at the start of the study; and (2) there was variation between trials with respect to the devices used and duration and frequency of treatment.
Authors' conclusions
Moderate‐certainty evidence shows that people who receive electromechanical‐assisted gait training in combination with physiotherapy after stroke are probably more likely to achieve independent walking than people who receive gait training without these devices.We concluded that nine patients need to be treated to prevent one dependency in walking.
Further research should consist of large, definitive pragmatic phase 3 trials undertaken to address specific questions about the most effective frequency and duration of electromechanical‐assisted gait training, as well as how long any benefit may last. Future trials should consider time poststroke in their trial design.
Funding
This Cochrane review had no dedicated funding.
Registration
Protocol (2006): doi:10.1002/14651858.CD006185
Original review (2007): doi:10.1002/14651858.CD006185.pub2
Review update (2013): doi:10.1002/14651858.CD006185.pub3
Review update (2017): doi:10.1002/14651858.CD006185.pub4
Review update (2020): doi:10.1002/14651858.CD006185.pub5
Plain language summary
Do electronic or robotic gait‐training devices help people walk better after a stroke?
Key messages
• Electronic and robotic devices plus physiotherapy probably help people walk independently again after a stroke. They may particularly benefit people in the first three months after a stroke.
• We need more research to find out how often, and for how long, these devices should be used.
What is a stroke?
A stroke happens when the flow of blood to part of the brain is cut off, blocking the supply of oxygen and nutrients to brain cells. This causes a sudden attack of weakness that usually affects one side of the body. If the supply of blood to the brain is stopped, brain cells begin to die. This can lead to brain injury, disability, and even death.
People who survive a stroke are often left with long‐term problems caused by the injury to their brain. They may find physical activities, such as walking, difficult because of weakened leg muscles on one or both sides of their body, stiff joints, or lack of co‐ordination. People may need a long period of rehabilitation, including physiotherapy, before they can recover their former independence. Physiotherapy includes exercise, massage, skills training, and electrical treatment to help people regain movement.
Walking after a stroke
One of the most important goals after a stroke is to help people walk again. Robotic devices (programmed to move and perform certain tasks automatically) and electrically powered mechanical (electromechanical) devices have been developed to help people practise walking. People who have trouble walking after a stroke need a lot of practice to get better. It is unclear if these walking machines are effective.
What did we want to find out?
We wanted to find out if gait‐training devices combined with physiotherapy can help people improve their walking after a stroke compared to not using such devices.
What did we do?
We searched for studies that looked at the use of gait‐training devices to help people learn to walk again after a stroke. We were interested in:
• how many people could walk independently;
• how fast people could walk;
• how far they could walk in 6 minutes;
• how many people dropped out of the study; and
• how many people died.
We looked for studies in which people were assigned to treatment groups at random. This type of study usually gives the most reliable evidence about the effects of a treatment.
What did we find?
We found 101 studies involving 4224 adults (average age 47 to 76 years) who had had a stroke and were learning to walk again. The studies compared the effects of physiotherapy plus electromechanical and robotic devices for gait training with the effects of physiotherapy alone or usual care. In most studies, the training period lasted three to four weeks; the shortest time was 10 days, and the longest was eight weeks.
At the end of training, compared with physiotherapy or usual care, using a gait‑training device plus physiotherapy:
• probably helps more people walk independently (51 studies; 2148 people);
• probably does not increase people's average walking velocity (73 studies; 3043 people);
• does not increase the distance people could walk in 6 minutes (42 studies; 1966 people); and
• does not increase or decrease how many people dropped out of the study, or how many people died (deaths were rare) (101 studies; 4224 people).
For every nine people treated with a device plus physiotherapy, probably one extra person walks independently by the end of treatment.
At follow‐up, using a gait‑training device plus physiotherapy may not help more people walk independently, and probably does not increase people's average walking velocity or the distance people could walk in 6 minutes, compared with physiotherapy or usual care.
What are the limitations of the evidence?
We have low to high confidence in our results. Many studies were of low or poor quality with small sample sizes. Therefore, some studies may have made the benefits of these devices seem greater than they are.
How up‐to‐date is this evidence?
The evidence is current to December 2023.
Summary of findings
Summary of findings 1. Summary of findings table ‐ Electromechanical‐assisted training compared to physiotherapy (or usual care) for people with gait impairments after stroke.
| Electromechanical‐assisted training compared to physiotherapy (or usual care) for people with gait impairments after stroke | ||||||
| Patient or population: people with gait impairments after stroke Setting: Rehabilitation in inpatient our outpatient settings in industrialised countries Intervention: electromechanical‐assisted training Comparison: physiotherapy (or usual care) | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect (95% CI) | № of participants (studies) | Certainty of the evidence (GRADE) | Comments | |
| Risk with physiotherapy (or usual care) | Risk with electromechanical‐assisted training | |||||
| Independent walking at end of intervention phase, all electromechanical devices used(critical outcome) assessed with: (FAC >3) | 417 per 1000 | 542 per 1000 (464 to 617) | OR 1.65 (1.21 to 2.25) | 2148 (51 RCTs) | ⊕⊕⊕⊝ Moderatea,b | Treatment probably increases the odds of participants becoming independent in walking at the end of intervention phase, all electromechanical devices used(critical outcome). For some of the included trials, no effect estimate could be calculated because no events (e.g. no participant reached the ability to walk) or only events (e.g. all participants regained walking) were reported. |
| Independent walking at follow‐up (> 3 months after study end)(critical outcome) assessed with: (FAC > 3) | 569 per 1000 | 684 per 1000 (504 to 821) | OR 1.64 (0.77 to 3.48) | 569 (8 RCTs) | ⊕⊕⊝⊝ Lowc | Treatment may not increase the odds of participants becoming independent in walking at follow‐up after study end, all electromechanical devices used(critical outcome). For some of the included trials, no effect estimate could be calculated because no events (e.g. no participant reached the ability to walk) or only events (e.g. all participants regained walking) were reported. |
| Walking velocity (metres per second) at end of intervention phase | The mean walking velocity (metres per second) at end of intervention phase was 0.54 m/s | MD 0.05 m/s higher (0.02 higher to 0.08 higher) | ‐ | 3043 (73 RCTs) | ⊕⊕⊕⊝ Moderated | Treatment probably does not increase walking velocity (metres per second) at end of intervention phase. |
| Walking velocity (metres per second) at follow‐up (> 3 months after study end) | The mean walking velocity (metres per second) at follow‐up (> 3 months after study end) was 0.6 m/s | MD 0.05 m/s higher (0.03 lower to 0.13 higher) | ‐ | 857 (17 RCTs) | ⊕⊕⊕⊝ Moderated | Treatment probably does not increase walking velocity (metres per second) at follow‐up. |
| Walking capacity (metres walked in 6 minutes) at end of intervention phase | The mean walking capacity (metres walked in 6 minutes) at end of intervention phase was 163.9 m | MD 11.01 m higher (1.76 higher to 20.25 higher) | ‐ | 1966 (42 RCTs) | ⊕⊕⊕⊕ High | Treatment does not increase walking capacity (metres walked in 6 minutes) at end of intervention phase. |
| Walking capacity (metres walked in 6 minutes) at follow‐up (> 3 months after study end) | The mean walking capacity (metres walked in 6 minutes) at follow‐up (> 3 months after study end) was 203.4 m | MD 9.55 m higher (14.58 lower to 33.67 higher) | ‐ | 736 (15 RCTs) | ⊕⊕⊕⊝ Moderated | Treatment probably does not increase walking capacity (metres walked in 6 minutes) at follow‐up. |
| Lost to studyduring intervention phase, dropouts ‐ total | 104 per 1000 | 0 per 1000 (‐2 to 1) | RD 0.00 (‐0.02 to 0.01) | 4224 (101 RCTs) | ⊕⊕⊕⊕ High | Treatment does not increase or decrease the odds of participants dropping out of study during intervention phase, dropouts ‐ total. |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; MD: mean difference; OR: odds ratio | ||||||
| GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect. | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_450209116350583763. | ||||||
a Not downgraded due to the robustness of the results as shown in the sensitivity analysis regarding internal validity. b Downgraded one level for other considerations due to indications of a small‐study effect (asymmetrical funnel plot). c Downgraded two levels for imprecision because 95% CI of summary effect measure crosses the line of no effect (OR = 1). d Downgraded one level for imprecision because 95% CI of summary effect measure crosses the minimal clinically important difference.
Background
Description of the condition
A stroke is a sudden, non‐convulsive loss of neurological function due to an ischaemic or haemorrhagic intracranial vascular event [1]. In general, cerebrovascular accidents are classified by anatomic location in the brain, vascular distribution, aetiology, age of the affected individual, and haemorrhagic versus non‐haemorrhagic nature [2]. Stroke is a leading cause of death and serious long‐term disability in adults [3]. Three months after stroke, 20% of people remain wheelchair bound, and approximately 70% walk at a reduced velocity and capacity [4]. Restoration of walking ability and gait rehabilitation are therefore highly relevant for people who are unable to walk independently after stroke [5], as well as for their relatives. To restore gait, modern concepts of rehabilitation favour a repetitive, task‐specific approach [6, 7]. In recent years it has also been shown that higher intensities of walking practice (resulting in more repetitions trained) resulted in better outcomes for people after stroke [6, 7].
Description of the intervention and how it might work
As an adjunct to overground gait training [8], in recent years treadmill training has been introduced for the rehabilitation of people after stroke [9]. Treadmill training with and without partial body weight support enables the repetitive practice of complex gait cycles for these people. However, one disadvantage of treadmill training might be the effort required by therapists to set the paretic limbs and to control weight shift, thereby possibly limiting the intensity of therapy, especially in more severely disabled people. Automated electromechanical gait machines were developed to reduce dependence on therapists and consist of either a robot‐driven exoskeleton orthosis ([10]; Nam 2019 [11]) or an electromechanical solution with two driven foot plates simulating the phases of gait (Aprile 2019 [12]; [13]).
One example of automated electromechanical gait rehabilitation is the Lokomat [10]. A robot gait orthosis combined with a harness‐supported body weight system is used together with a treadmill. The main difference from treadmill training is that the patient's legs are guided by the robotic device according to a preprogrammed gait pattern. A computer‐controlled robotic gait orthosis guides the patient, and the process of gait training is automated.
A second example is the Gait Trainer GT I, which is based on a double crank‐and‐rocker gear system [13]. In contrast to a treadmill, the electromechanical Gait Trainer GT I consists of two foot plates positioned on two bars, two rockers, and two cranks, which provide the propulsion. The harness‐secured patient is positioned on the foot plates, which symmetrically simulate the stance and swing phases of walking [13]. A servo‐controlled motor guides the patient during walking exercise. Vertical and horizontal movements of the trunk are controlled in a phase‐dependent manner. Again, the main difference from treadmill training is that the process of gait training is automated and is supported by an electromechanical solution.
Other similar electromechanical devices developed in recent years include the Exowalk (Nam 2019), the Haptic Walker [14], the Anklebot [15], and the LOPES (Lower Extremity Powered Exoskeleton) [16]. More recently, new so‐called powered mobile solutions (Buesing 2015 [17]; Calabrò 2018 [18, 19]; Stein 2014 [20]; Watanabe 2014 [21, 22]) and ankle robots (Forrester 2014 [23]; Waldman 2013 [24]) to improve walking have been described in the literature.
Electromechanical devices (such as those previously described) can be used to give non‐ambulatory patients intensive practice (in terms of high repetitions) of complex gait cycles. The advantage of these electromechanical devices compared with treadmill training with partial body weight support may be the reduced effort required of therapists, as they no longer need to set the paretic limbs or assist trunk movements [25].
Why it is important to do this review
Scientific evidence for the benefits of the above‐mentioned technologies may have changed since our Cochrane review was first published in 2007 [26] and last updated in 2020 [27], so an update of the review was required to justify the large equipment and human resource costs needed to implement electromechanical‐assisted gait devices, as well as to confirm the safety and acceptance of this method of training. The aim of this fourth review update was therefore to evaluate the best available evidence about the above‐mentioned approach.
Objectives
Primary
To determine whether electromechanical‐ and robot‐assisted gait training versus physiotherapy (or usual care) improves walking in adults after stroke.
Secondary
To determine whether electromechanical‐ and robot‐assisted gait training versus physiotherapy (or usual care) after stroke improves walking velocity, walking capacity, acceptability, and death from all causes until the end of the intervention phase.
Methods
In our protocol, we stated that we would use the PEDro scale to assess the methodological quality of the included trials. However, Chapter 8 of the latest edition of the Cochrane Handbook for Systematic Reviews of Interventions suggests that scales that yield a summary score should be avoided [28]. In accordance with this suggestion, we no longer wished to use the PEDro scale to assess the methodological quality of included trials, and instead used the Cochrane RoB 1 tool to analyse trial methods.
In our protocol, we planned to quantify heterogeneity with the I² statistic and to use a cutoff of I² = 50% for all comparisons. Additionally, we planned to calculate the overall effects using a random‐effects model instead of a fixed‐effect model when we found substantial heterogeneity. However, for this update, we calculated overall effects using a random‐effects model, irrespective of the level of heterogeneity.
In our 2017 review update, we expanded our post hoc sensitivity analysis of type of device (Analysis 5.1; Analysis 5.2; Analysis 5.3) by introducing a new subgroup of studies using mobile and ankle devices and by adding a new comparison (Analysis 5.3) of different devices for regaining walking capacity [9].
Since the 2020 review update, we have used risk difference for Analysis 1.7 instead of odds ratio. We changed that calculation because such absolute measurements are preferred here, and because we believe that risk differences better reflect the differences in dropout rates between intervention and control groups. We have used risk differences in Analysis 1.7 and Analysis 1.8.
We followed Methodological Expectations of Cochrane Intervention Reviews (MECIR) [29] when conducting the review and PRISMA 2020 [30] for the reporting.
Criteria for considering studies for this review
Types of studies
We included randomised controlled trials and randomised controlled cross‐over trials. In the case of randomised controlled cross‐over trials, we planned to analyse only the first period as a parallel‐group trial.
Types of participants
We included studies with participants of any gender over 18 years of age after stroke, using the World Health Organization (WHO) definition of stroke, or a clinical definition of stroke if the WHO definition was not specifically stated [1].
Types of interventions
We included trials evaluating electromechanical‐ and robot‐assisted gait training plus physiotherapy versus physiotherapy (or usual care) for regaining and improving walking after stroke. We also included automated electromechanical devices that were used in combination with therapies such as functional electrical stimulation applied to the legs during gait training (compared with therapies not using electromechanical devices). We defined an automated electromechanical device as any device with an electromechanical solution designed to assist stepping cycles by supporting body weight and automating the walking therapy process in people after stroke. This category includes any mechanical or computerised device designed to improve walking function. We also searched for electromechanical devices such as robots for gait training after stroke [14, 15, 16].
Electromechanical devices can principally be differentiated into end‐effector and exoskeleton devices. Examples of end‐effector devices are the LokoHelp [31], the Haptic Walker [14], and the Gait Trainer GT I [13]. The definition of an end‐effector principle is that a patient's feet are placed on foot plates, whose trajectories simulate the stance and swing phases during gait training [32]. An example of exoskeleton devices is the Lokomat [10]. Such exoskeletons are outfitted with programmable drives or passive elements, which move the knees and hips during the phases of gait [32].
We did not include non‐weight‐bearing interventions such as non‐interactive devices that deliver continuous passive motion only [33]. To prevent duplication with other Cochrane reviews and protocols (e.g. [34]), we excluded trials testing the effectiveness of treadmill training or other approaches such as repetitive task training in physiotherapy or electrical stimulation alone [35, 36].
There was no criterion for minimum duration or setting of the intervention.
Outcome measures
We used the following outcome measures.
Critical outcomes
Regaining the ability to walk is a very important goal for people after stroke [6, 37, 38]. We therefore defined the critical outcome as the ability to walk independently. We measured the ability to walk with the Functional Ambulation Category (FAC) [39]. An FAC score of 4 or 5 indicated independent walking over a 15‐metre surface, irrespective of aids used, such as a cane. An FAC score less than 4 indicates dependency in walking (supervision or assistance, or both, must be given in performing walking).
If the included studies did not report FAC scores, we used alternative indicators of independent walking, such as:
a score of 3 on the ambulation item of the Barthel Index [40]; or
a score of 6 or 7 for the walking item of the Functional Independence Measure [41]; or
a 'yes' response to the item 'walking inside, with an aid if necessary (but with no standby help)' or 'yes' to 'walking on uneven ground' in the Rivermead Mobility Index [42].
We used the results of the trials regardless of the duration of treatment at the end of the treatment phase and placed no restrictions on interventions and comparators in terms of delivery, dose, duration, or intensity.
For the follow‐up, we prioritised information on the first follow‐up after the end of the intervention phase.
Important outcomes
We defined important outcomes as measures of activity limitations. We used walking velocity (in metres per second; [6]), walking capacity (metres walked in 6 minutes; [6]), and the Rivermead Mobility Index score as relevant measures of activity limitations, if stated by the trialists.
Adverse outcomes
We investigated the safety of electromechanical‐assisted gait‐training devices by examining the incidence of adverse outcomes such as thrombosis, major cardiovascular events, injury, pain, and any other reported adverse event. To measure the acceptance of electromechanical‐assisted gait‐training devices in walking therapies, we used visual analogue scales or withdrawal from the study for any reason (lost to study during intervention phase, dropout rates), or both, during the study period, depending on data provided by the study authors.
Additionally, we used death from all causes as an important outcome.
We took care to ensure that relevant outcomes were not available because they had not been measured, rather than because they had simply not been reported.
Depending on the above‐stated categories and the availability of variables used in the included trials, we discussed and reached consensus on which outcome measures should be included in the analysis.
Search methods for identification of studies
We searched for trials in all languages and arranged for translation of relevant papers where necessary.
Electronic searches
We searched the following electronic bibliographic databases.
Cochrane Central Register of Controlled Trials (CENTRAL; 2023 Issue 12), in the Cochrane Library (Supplementary material 1).
MEDLINE Ovid (1950 to 13 December 2023) (Supplementary material 1).
Embase Ovid (1980 to 13 December 2023) (Supplementary material 1).
Cumulative Index to Nursing and Allied Health Literature (CINAHL) EBSCO (1982 to 13 December 2023) (Supplementary material 1).
Allied and Complementary Medicine Database (AMED) Ovid (1985 to 13 December 2023) (Supplementary material 1).
Web of Science (Science Citation Index Expanded, Social Sciences Citation Index, Arts and Humanities Citation Index) (1899 to 13 December 2023) (Supplementary material 1).
Physiotherapy Evidence Database (PEDro) (searched 13 December 2023) (Supplementary material 1).
COMPENDEX (1972 to 6 January 2020) (Supplementary material 1).
SPORTDiscus EBSCO (1949 to 13 December 2023) (Supplementary material 1).
Inspec (1969 to 6 January 2020) (Supplementary material 1).
We developed the search strategies with the help of the Cochrane Stroke Group Information Specialist and adapted the MEDLINE search strategy for use with other databases and used the Cochrane Highly Sensitive randomised controlled trial filter [28]. The dates of last searches differ between databases due to changes in our institutions' database licences.
We identified and searched the following ongoing trials and research registers.
World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) (apps.who.int/trialsearch/) (searched 13 December 2023) (Supplementary material 8).
Searching other resources
We also:
-
handsearched the following relevant conference proceedings:
World Congress of NeuroRehabilitation (2002 to 2023);
World Congress of Physical Medicine and Rehabilitation (2001 to 2023);
World Congress of Physical Therapy (2003 to 2023);
Deutsche Gesellschaft für Neurotraumatologie und Klinische Neurorehabilitation (2001 to 2023);
Deutsche Gesellschaft für Neurologie (2000 to 2023);
Deutsche Gesellschaft für Neurorehabilitation (1999 to 2023); and
Asia‐Oceanian Conference of Physical & Rehabilitation Medicine (2008 to 2023);
screened reference lists of all relevant articles; and
contacted trialists, experts, and researchers in our field of study.
Data collection and analysis
Selection of studies
Two review authors (JM, BE) independently read the titles and abstracts of identified references, eliminating any obviously irrelevant studies and ranking the remaining studies as relevant, irrelevant, or possibly relevant based on our inclusion criteria (types of studies, types of participants, aims of interventions, outcome measures). We obtained the full‐text reports of studies deemed relevant or possibly relevant, and the same two review authors independently assessed the full‐text reports for inclusion in the review. We excluded all trials of specific treatment components, such as electrical stimulation as stand‐alone treatment, treadmill training, and continuous passive motion treatment, as these have been addressed in other Cochrane reviews (e.g. [34]). In case of doubt regarding eligibility, we retrieved the full‐text report of the study. Any disagreements between review authors were resolved in consultation with a third review author (JK). We contacted trialists for further information to reach consensus as needed. We recorded the selection process in sufficient detail to complete a PRISMA flow diagram, and listed in Supplementary material 3 all studies that did not match our inclusion criteria regarding types of studies, types of participants, and aims of interventions.
Data extraction and management
Two review authors (JM, BE) independently extracted trial and outcome data from the included studies using a form that had been piloted. We established the characteristics of unpublished trials through correspondence with the trial co‐ordinator or principal investigator. If a review author was involved in any of the selected studies, another review author not involved in the study extracted the study information. We used checklists to independently record the following details.
Methods of generating the randomisation schedule.
Method of concealment of allocation.
Blinding of assessors.
Use of an intention‐to‐treat analysis (all participants initially randomly assigned were included in analyses as allocated to groups).
Adverse events and dropouts for all reasons.
Important imbalance in prognostic factors.
Participants (country, number of participants, age, gender, type of stroke, time from stroke onset to entry into the study, inclusion and exclusion criteria).
Comparison (details of the intervention in treatment and control groups, details of co‐intervention(s) in both groups, duration of treatment, stroke severity, electromechanical device used, duration of study intervention, aetiology (ischaemic/haemorrhage), intensity of treatment per day, description of the control intervention, dropouts).
Outcomes and time points of measures (number of participants in each group and outcome, regardless of compliance).
The two review authors checked all the extracted data for agreement, with a third review author (JK) arbitrating when consensus could not be reached. We contacted trialists to request more information, clarification, and missing data as needed.
If one of the review authors was a co‐author of an included trial, another review author (BE or JK) made eligibility decisions, extracted data, and performed risk of bias and GRADE assessments for that trial.
Risk of bias assessment in included studies
Two review authors (JM, MP) independently evaluated the methodological quality of the included trials using Cochrane's RoB 1 tool, as described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions [43].
We assessed the following risk of bias domains as at low, high, or unclear risk of bias.
Random sequence generation (selection bias)
Allocation concealment (selection bias)
Blinding of participants and personnel (performance bias)
Blinding of outcome assessment (detection bias)
Incomplete outcome data (attrition bias)
Selective reporting (reporting bias)
Other bias
We checked all methodological quality assessments for agreement between review authors, resolving any disagreements by discussion. If one of the review authors was a co‐author of an included trial, another review author (BE or JK) conducted the methodological quality assessment for that trial.
Measures of treatment effect
We used a random‐effects model for all statistical analyses. For dichotomous variables, we calculated and reported Peto odds ratios (ORs) for the critical outcome, and risk differences (RDs) for the important outcomes, with 95% confidence intervals (CIs). For continuous data, we calculated the treatment effect using standardised mean differences (SMDs) and 95% CIs when studies used different scales to measure an outcome, and mean differences (MDs) and 95% CIs when studies used the same method to measure an outcome. We used Cochrane RevMan software for all statistical comparisons [44].
Based on the event rate and on control and event rates, we calculated the number needed to treat for an additional beneficial outcome (NNTB) with 95% CI for the critical outcome of independently walking at the end of the intervention phase [45].
Unit of analysis issues
We anticipated that a majority of trials would have a parallel‐group design. When studies had two or more active intervention groups eligible for inclusion within the same comparison (against a control, placebo, or no‐treatment group), we planned to 'share' control group data between the multiple pair‐wise comparisons to avoid double counting of participants within an analysis. If studies used a randomised controlled cross‐over design, we would analyse data from the first phase only (up to the point of cross‐over). We did not anticipate that any studies would use a cluster‐randomised design.
Dealing with missing data
In the case of missing outcome data, we attempted to analyse data according to the intention‐to‐treat approach. We contacted the trial co‐ordinator or principal investigator if data were missing.
Reporting bias assessment
We inspected funnel plots to assess the risk of publication bias.
Synthesis methods
We pooled the results of all eligible studies to present an overall estimate of the effect of electromechanical‐assisted gait training (meta‐analysis). Clinical diversity and heterogeneity did not contribute to the decision about when to pool trials, but we described clinical diversity and variability in participants, interventions, and outcomes studied in Table 2 and Table 3. If studies had three or more intervention groups, for example two treatment groups and one control group, and the results of these intervention groups did not differ, we combined the results of all intervention groups into one collapsed group and compared this information with the results of the control group.
1. Overview of included studies and synthesis table illustrating key participant characteristics in studies, sorted alphabetically.
| Study ID | Experimental: mean age (SD) |
Control: mean age (SD) |
Experimental: mean time poststroke |
Control: mean time poststroke |
Experimental: sex |
Control: sex |
Experimental: side paresis |
Control: side paresis |
| Akinci 2023 | 61 years | 60 years | > 6 months | > 6 months | Not stated | Not stated | 27 right, 15 left | 9 right, 5 left |
| Alingh 2021 | 61 years | 57 years | 5 weeks | 6 weeks | 10 men, 7 women | 10 men, 5 women | 10 right, 7 left | 8 right, 7 left |
| Aprile 2017 | 58 (20) years | 69 (11) years | > 6 months | > 6 months | 4 men, 2 women | 5 men, 3 women | 0 right, 6 left | 5 right, 3 left |
| Aprile 2019 | 61 (12) years | 56 (9) years | 86 days | 44 days | 9 men, 3 women | 10 men, 4 women | 3 right, 9 left | 9 right, 5 left |
| Aschbacher 2006 | 57 years | 65 years | ≤ 3 months | ≤ 3 months | 2 women | 4 women | Not stated | Not stated |
| Bang 2016 | 54 years | 54 years | 12 months | 13 months | 5 men, 4 women | 4 men, 5 women | 4 right, 5 left | 4 right, 5 left |
| Belas dos Santos 2018 | 44 (12) years | 56 (11) years | 5 years | 11 years | 6 men, 2 women | 5 men, 2 women | 2 right, 2 left, 3 both | 2 right, 2 left, 4 both |
| Bergmann 2018 | 72 (9) years | 71 (10) years | 7.5 weeks | 8 weeks | 10 men, 5 women | 7 men, 8 women | 11 right, 4 left | 12 right, 3 left |
| Brincks 2011 | 61 years | 59 years | 56 (median) days | 21 (median) days | 5 men, 2 women | 4 men, 2 women | 5 right, 2 left | 1 right, 5 left |
| Buesing 2015 | 60 years | 62 years | 7 years | 5 years | 17 men, 8 women | 16 men, 9 women | 13 right, 12 left | 12 right, 13 left |
| Calabrò 2018 | 69 (4) years | 67 (6) years | 10 months | 11 months | 12 men, 8 women | 11 men, 9 women | 12 right, 8 left | 11 right, 9 left |
| Chang 2012 | 56 (12) years | 60 (12) years | 16 (5) days | 18 (5) days | 13 men, 7 women | 10 men, 7 women | 6 right, 14 left | 6 right, 11 left |
| Cho 2015 | 55 (12) years | 55 (15) years | 15 months | 13 months | Not stated | Not stated | 6 right, 4 left, 4 both | 3 right, 1 left, 3 both |
| Choi 2022 | 56 years | 61 years | 18 months | 17 months | 9 men, 9 women | 3 men, 3 women | 9 right, 9 left | 3 right, 3 left |
| Chua 2016 | 62 (10) years | 61 (11) years | 27 (11) days | 30 (14) days | 35 men, 18 women | 40 men, 13 women | 24 right, 29 left | 21 right, 32 left |
| Cotinat 2023 | 59 years | 60 years | 7 years | 7 years | 9 men, 6 women | 11 men, 7 women | 9 right, 7 left | 7 right, 11 left |
| De Luca 2020 | 54 years | 56 years | > 6 months | > 6 months | 11 men, 4 women | 11 men, 4 women | Not stated | Not stated |
| Dias 2006 | 70 (7) years | 68 (11) years | 47 (64) months | 48 (30) months | 16 men, 4 women | 14 men, 6 women | Not stated | Not stated |
| Erbil 2018 | 50 (11) years | 48 (10) years | 39 months | 25 months | 16 men, 13 women | 3 men, 11 women | 11 right, 18 left | 6 right, 8 left |
| Fisher 2008 | Not stated | Not stated | < 12 months | < 12 months | Not stated | Not stated | Not stated | Not stated |
| Forrester 2014 | 63 years | 60 years | 12 days | 11 days | Not stated | Not stated | 9 right, 9 left | 7 right, 9 left |
| Gandolfi 2019 | 63 (11) years | 64 (10) years | 54 (36) months | 53 (41) months | 10 men, 6 women | 13 men, 3 women | 10 right, 6 left | 12 right, 4 left |
| Geroin 2011 | 63 (7) years | 61 (6) years | 26 (6) months | 27 (6) months | 14 men, 6 women | 9 men, 1 woman | Not stated | Not stated |
| Gorsler 2024 | 72 years | 70 years | 37 days | 35 days | 18 men, 7 women | 10 men, 8 women | All left | All left |
| Han 2016 | 68 (15) years | 63 (11) years | 22 (8) days | 18 (10) days | Not stated | Not stated | 20 right, 10 left | 14 right, 12 left |
| Hidler 2009 | 60 (11) years | 55 (9) years | 111 (63) days | 139 (61) days | 21 men, 12 women | 18 men, 12 women | 22 right, 11 left | 13 right, 17 left |
| Hornby 2008 | 57 (10) years | 57 (11) years | 50 (51) months | 73 (87) months | 15 men, 9 women | 15 men, 9 women | 16 right, 8 left | 16 right, 8 left |
| Husemann 2007 | 60 (13) years | 57 (11) years | 79 (56) days | 89 (61) days | 11 men, 5 women | 10 men, 4 women | 12 right, 4 left | 11 right, 3 left |
| Jayaraman 2019 | 59 (9) years | 61 (12) years | 7 (6) years | 5 (3) years | 17 men, 8 women | 16 men, 9 women | 13 right, 12 left | 12 right, 13 left |
| Kang 2021 | 64 years | 63 years | > 6 months | > 6 months | 10 men, 5 women | 8 men, 7 women | 10 right, 5 left | 6 right, 9 left |
| Kayabinar 2019 | 55 (13) years | 63 (10) years | 29 (20) months | 34 (38) months | 20 men, 12 women | 21 men, 11 women | 31 right, 1 left | 29 right, 3 left |
| Kelley 2013 | 67 (9) years | 64 (11) years | 4 years | 1 year | 64% men | 67% men | Not described | Not described |
| Kim 2015 | 54 (13) years | 50 (16) years | 80 (60) days | 120 (84) days | 9 men, 4 women | 10 men, 3 women | 8 right, 5 left | 10 right, 3 left |
| Kim 2019a | 48 (6) years | 46 (14) years | 21 (33) months | 10 (8) months | 9 men, 1 woman | 7 men, 2 women | Not stated | Not stated |
| Kim 2019b | 57 (12) years | 60 (13) years | 2 (2) months | 2 (3) months | 20 men, 5 women | 13 men, 10 women | 14 right, 11 left | 14 left, 9 right |
| Kim 2020 | 54 years both groups | 3 to 12 months both groups | 25 men, 5 women, both groups | Not stated | Not stated | |||
| Kooncumchoo 2021 | 64 years | 64 years | 6.6 years | 5.6 years | 10 men, 5 women | 10 men, 5 women | 7 right, 8 left | 7 right, 8 left |
| Kotov 2021a | 63 years, both groups | 2 months, both groups | Not stated | 29 right, 18 left, both groups | ||||
| Kotov 2021b | 61 years, both groups | Not stated | Not stated | Not stated | Not stated | Not stated | Not stated | |
| Kwon 2018 | Not stated | Not stated | At least 3 months since stroke | At least 3 months since stroke | Not stated | Not stated | Not stated | Not stated |
| Kyung 2008 | 48 (8) years | 55 (16) years | 22 (23) months | 29 (12) months | 9 men, 8 women | 4 men, 4 women | 9 right, 8 left | 4 right, 4 left |
| Lee 2019 | 61 (7) years | 62 (6) years | 1486 (264) days | 1536 (312) days | 7 men, 7 women | 7 men, 5 women | 5 right, 9 left | 7 right, 5 left |
| Lee 2022 | 64 years | 66 years | 15 months | 14 months | 20 men, 13 women | 6 men, 4 women | 15 right, 18 left | 5 right, 5 left |
| Lee 2023a | 63 years | 65 years | 0.9 months | 0.9 months | 15 men, 11 women | 11 men, 12 women | 15 right, 11 left | 11 right, 12 left |
| Lee 2023b | 68 years | 70 years | Not stated | Not stated | 11 men, 6 women | 11 men, 10 women | 10 right, 6 left, 1 bilateral | 10 right, 11 left |
| Li 2021a | 50 years | 50 years | 3.4 months | 2.5 months | 14 men, 1 women | 15 men, 2 women | 6 right, 9 left | 12 right, 5 left |
| Li 2021b | Not stated | Not stated | Not stated | Not stated | Not stated | Not stated | Not stated | Not stated |
| Lin 2022 | 54 years | 57 years | 26 days | 35 days | 14 men, 6 women | 15 men, 5 women | 7 right, 13 left | 12 right, 8 left |
| Louie 2021 | 60 years | 55 years | 37 days | 41 days | 16 men, 3 women | 10 men, 7 women | 8 right, 11 left | 7 right, 10 left |
| Lu 2017 | 50 (13) years | 58 (13) years | 59 months | 28.5 months | 23 men, 7 women | 21 men, 9 women | 13 right, 17 left | 15 right, 15 left |
| Mayr 2008 | Not stated | Not stated | Between 10 days and 6 months | Between 10 days and 6 months | Not stated | Not stated | Not stated | Not stated |
| Mayr 2018 | 68 (12) years | 68 (12) years | 5 (1 to 8) weeks | 4 (2) weeks | 22 men, 15 women | 19 men, 18 women | 20 right, 17 left | 16 right, 21 left |
| Meng 2022 | 59 years | 58 years | not stated | Not stated | 33 men, 29 women | 70 men, 57 women | 34 right, 28 left | 48 right, 56 left |
| Miyagawa 2023 | 65 years | 63 years | 14 to 90 days | 14–90 days | Not stated | Not stated | Not stated | Not stated |
| Molteni 2021 | 62 years | 68 years | 36 days | 34 days | 21 men, 17 women | 21 men, 19 women | 19 right, 19 left | 11 right, 26 left |
| Morone 2011 | 62 (11) years | 62 (14) years | 19 (11) days | 20 (14) days | 15 men, 9 women | 13 men, 11 women | 13 right, 11 left | 15 right, 9 left |
| Mustafaoglu 2020 | 53 years | 55 years | 12 months | 13 months | 21 men, 13 women | 12 men, 5 women | 20 right, 14 left | 12 right, 5 left |
| Nam 2019 | 48 (15) years | 68 (17) years | 530.11 (389) days | 284.81 (309) days | 11 men, 7 women | 6 men, 10 women | Not stated | Not stated |
| Nam 2020 | 60 (11) years | 57 (9) years | 546 (296) days | 600 (506) days | 8 men, 10 women | 14 men, 6 women | Not stated | Not stated |
| Nam 2022 | 62 years | 61 years | 522 days | 767 days | 35 men, 20 women | 34 men, 20 women | 28 right, 27 left | 25 right, 29 left |
| Noser 2012 | 67 (9) years | 64 (11) years | 1354 days | 525 days | 7 men, 4 women | 6 men, 4 women | Not stated | Not stated |
| Ochi 2015 | 62 (8) years | 66 (12) years | 23 (7) days | 26 (8) days | 11 men, 2 women | 9 men, 4 women | 6 right, 7 left | 5 right, 8 left |
| Ogino 2020 | 66 years | 65 years | 8 years | 7 years | 6 men, 2 women | 9 men, 2 women | 4 right, 4 left | 8 right, 3 left |
| Palmcrantz 2021 | 62 years | 61 years | 21 months | 33 months | 11 men, 5 women | 24 men, 8 women | 6 right, 10 left | 10 right, 22 left |
| Park 2018 | 56 (7) years | 57 (9) years | 7 months | 8 months | 16 men, 8 women | 9 men, 7 women | 14 right, 10 left | 8 right, 8 left |
| Park 2020 | 76 years | 70 years | 2 weeks | 2 weeks | 6 men, 2 women | 4 men, 3 women | 3 right, 4 left | 3 right, 4 left |
| Peurala 2005 | 52 (8) years | 52 (7) years | 2 (2) years | 4 (5) years | 26 men, 4 women | 11 men, 4 women | 13 right, 17 left | 10 right, 5 left |
| Peurala 2009 | 67 (9) years | 68 (10) years | 8 (3) days | 8 (3) days | 11 men, 11 women | 18 men, 16 women | 11 right, 11 left | 14 right, 20 left |
| Picelli 2016 | 62 (10) years | 65 (3) years | 6 (4) years | 6 (4) years | 7 men, 4 women | 9 men, 2 women | Not stated | Not stated |
| Pohl 2007 | 62 (12) years | 64 (11) years | 4 (2) weeks | 5 (2) weeks | 50 men, 27 women | 54 men, 24 women | 36 right, 41 left | 33 right, 45 left |
| Saltuari 2004 | 62 (13) years | 60 (19) years | 3 (4) months | 2 (1) months | 4 men, 4 women | 2 men, 6 women | Not stated | Not stated |
| Schroeder 2024 | 60 years | 69 years | 7 to 14 days | 7 to 14 days | 6 men, 4 women | 4 men, 5 women | 4 right, 6 left | 2 right, 7 left |
| Schwartz 2006 | 62 (9) years | 65 (8) years | 22 (9) days | 24 (10) days | 21 men, 16 women | 20 men, 10 women | 17 right, 20 left | 8 right, 22 left |
| Sczesny‐Kaiser 2019 | 63 years | 66 years | 62 months | 102 months | 3 men, 6 women | 2 men, 7 women | 4 right, 5 left | 4 right, 5 left |
| Song 2021 | 61 years | 60 years | 3.4 months | 3.6 months | 12 men, 6 women | 9 men, 9 women | Not stated | Not stated |
| Stein 2014 | 58 (11) years | 57 (15) years | 49 (39) months | 89 (153) months | Not stated | Not stated | Not stated | Not stated |
| Stolz 2019 | 68 (12) years | 63 (16) years | 15 (9) days | 26 (22) days | 8 men, 8 women | 14 men, 6 women | 7 right, 9 left | 11 right, 9 left |
| Talaty 2023 | 63 years | 54 years | 17 days | 17 days | 12 men, 3 women | 10 men, 5 women | 8 right, 7 left | 4 right, 11 left |
| Tanaka 2012 | 63 (10) years | 60 (9) years | 55 (37) months | 65 (67) months | 10 men, 2 women | 9 right, 3 left | ||
| Tanaka 2019 | 64 (12) years | 62 (9) years | 103 (28) days | 92 (38) days | 13 men, 8 women | 14 men, 6 women | 8 right, 13 left | 10 right, 10 left |
| Taveggia 2016 | 71 (5) years | 73 (7) years | 60 (49) days | 39 (31) days | 7 men, 6 women | 10 men, 5 women | Not stated | Not stated |
| Thimabut 2022 | 53 years | 63 years | 56 days | 73 days | 6 men, 7 women | 10 men, 3 women | 7 right, 6 left | 4 right, 9 left |
| Tomida 2019 | 55 (9) years | 61 (12) | 25 (6) days | 31 (10) days | 11 men, 2 women | 5 men, 8 women | 6 right, 7 left | 5 right, 7 left |
| Tong 2006 | 71 (14) years | 64 (10) years | 2 (1) weeks | 2 (1) weeks | 19 men, 11 women | 12 men, 8 women | 13 right, 17 left | 7 right, 13 left |
| Ucar 2014 | 56 years | 62 years | Not stated | Not stated | Not stated | Not stated | Not stated | Not stated |
| Van Nunen 2012 | 53 (10) years | 2 (1) months | 16 men, 14 women | Not stated | Not stated | |||
| Waldman 2013 | 51 (8) years | 53 (7) years | 41 (20) months | 30 (22) months | Not stated | Not stated | Not stated | Not stated |
| Wall 2020 | 52 years | 50 years | 32 days | 42 days | Not stated | Not stated | 2 right, 8 left | 2 right, 5 left |
| Watanabe 2014 | 67 (17) years | 76 (14) years | 59 (47) days | 51 (34) days | 7 men, 4 women | 4 men, 7 women | 6 right, 5 left | 5 right, 6 left |
| Werner 2002 | 60 (9) years | 60 (9) years | 7 (2) weeks | 6 (2) weeks | 8 men, 7 women | 5 men, 10 women | 8 right, 7 left | 8 right, 7 left |
| Westlake 2009 | 59 (17) years | 55 (14) years | 44 (27) months | 37 (20) months | 6 men, 2 women | 7 men, 1 woman | 4 right, 4 left | 3 right, 5 left |
| Wright 2021 | 60 years | 65 years | 31 months | 32 months | 14 men, 2 women | 14 men, 4 women | 5 right, 11 left | 8 right, 10 left |
| Xie 2023 | 59 years | 63 years | 8 weeks | 9 weeks | Not stated | Not stated | 10 right, 5 left | 10 right, 5 left |
| Yeung 2018 | 54 (13) years | 61 (10) years | 4 (3) years | 6 (4) years | 6 men, 3 women | 7 men, 3 women | 4 right, 5 left | 5 right, 5 left |
| Yeung 2021 | 66 years | 64 years | 26 days | 28 days | 16 men, 14 women | 8 men, 9 women | 13 right, 17 left | 11 right, 6 left |
| Yokota 2023 | 65 years | 63 years | < 48 hours | < 48 hours | 7 men, 5 women | 5 men, 5 women | Not stated | Not stated |
| Yoo 2023a | 61 years | 65 years | 19 days | 43 days | 4 men, 5 women | 5 men, 3 women | 3 right, 6 left | 5 right, 6 left |
| Yoo 2023b | 62 years | 66 years | 12 years | 16 years | 10 men, 4 women | 10 men, 5 women | 5 right, 9 left | 8 right, 7 left |
| Yu 2021 | 58 years | 52 years | 7 weeks | 8 weeks | 18 men, 9 women | 12 men, 15 women | 12 right, 15 left | 18 right, 9 left |
| Yun 2018 | 63 (6) years | 64 (3) years | 31 (3) years | 28 (8) years | 10 men, 8 women | 9 men, 9 women | 3 right, 15 left | 4 right, 14 left |
| Zhang 2023 | 57 years | 61 years | 2.5 months | 3.5 months | 14 men, 4 women | 13 men, 3 women | 11 right, 7 left | 8 right, 8 left |
SD: standard deviation
2. Overview of included studies and synthesis table illustrating key demographics of studies including dropouts and adverse events, sorted alphabetically.
| Criteria | Electromechanical device used | Duration of study intervention | Aetiology (ischaemic/haemorrhagic) | Intensity of treatment per day | Description of control intervention | Dropouts | Reasons for dropout and adverse events in experimental group | Reasons for dropout and adverse events in the control group | Source of information |
| Akinci 2023 | Lokomat | 6 weeks | Not stated | 40 minutes, 3 times a week | Conventional physiotherapy (40 min) 5 days a week for 6 weeks | 0 of 56 | n.a. | n.a. | Published information |
| Alingh 2021 | LOPES‐II | 6 weeks | 24/8 | 30 minutes, 3 to 5 times a week | Conventional training group, 30 minutes, 3 to 5 times a week | 3 of 34 | Discontinued training or participation | Discontinued training | Published information |
| Aprile 2017 | G‐EO system | 45 days, 20 sessions | Not stated | 45 minutes, 3 times a week | Traditional gait rehabilitation, 1 hour, 3 times a week | 0 of 14 | None | None | Published information |
| Aprile 2019 | G‐EO system | 20 sessions | 8/4 | 45 minutes, 3 times a week | Traditional gait rehabilitation, 3 times a week, 20 sessions | 0 of 26 | None | None | Published information |
| Aschbacher 2006 | Lokomat | 3 weeks | Not stated | 30 minutes, 5 times a week | Task‐oriented physiotherapy, 5 times a week for 3 weeks (2.5 hours a week) | 4 of 23 | Not stated | Not stated | Unpublished information provided by study authors. |
| Bang 2016 | Lokomat | 4 weeks | 13/5 | 60 minutes, 5 times a week (20 sessions) | Treadmill training without body weight support | 0 of 18 | None | None | Published information |
| Belas dos Santos 2018 | Lokomat | 5 months | 4/11 | 60 minutes, 3 times a week | Therapist‐assisted gait training, once a week, 60 minutes, for 5 months | 4 of 19 | Participants not complying with protocol treatment criteria in the intervention group | Not stated | Published information |
| Bergmann 2018 | Lokomat | 2 weeks | 8/7 | 60 minutes, 8 to 10 sessions in 2 weeks | Non‐robotic physiotherapy is described as training of postural control including sensory feedback components in sitting, sit‐to‐stand, standing, and walking, if possible, 60 minutes or 30 minutes with 2 therapists, 8 to 10 sessions in 2 weeks. | 12 of 38 | No pusher behaviour at start of treatment (n = 3), second stroke (n = 2), infection (n = 1), pain in the lower limb (n = 3), transfer to another hospital (n = 1) | No pusher behaviour at start of treatment (n = 1), infection (n = 1) | Published information |
| Calabrò 2018 | Ekso | 8 weeks | Not stated | 45 minutes, 5 days a week | Conventional overground gait training | 0 of 40 | None | None | Published information |
| Brincks 2011 | Lokomat | 3 weeks | Not stated | Not stated | Physiotherapy | 0 of 13 | None | None | Unpublished and published information provided by study authors. |
| Buesing 2015 | Wearable exoskeleton Stride Management Assist (SMA) system | 6 to 8 weeks | Unclear | 3 times per week for a maximum of 18 sessions | Functional task‐specific training (intensive overground training and mobility training) | 0 of 50 | None | None | Published information |
| Chang 2012 | Lokomat | 10 days | Not stated | 30 minutes daily for 10 days | Conventional gait training by physical therapists (with equal therapy time and same number of sessions as experimental group) | 3 of 40 | Not described by group (3 participants dropped out: 1 due to aspiration pneumonia, 2 were unable to co‐operate fully with the experimental procedure) | Unpublished and published information provided by study authors. | |
| Cho 2015 | Lokomat | 8 weeks (2 phases, cross‐over after 4 weeks) | 4/14 (2 both) | 30 minutes, 3 times a week for 4 weeks | Bobath (neurophysiological exercises, inhibition of spasticity and synergy pattern) | 0 of 20 | None | None | Published information |
| Choi 2022 | Lokomat Pro | 6 weeks | Not stated | 30 minutes a day, 5 times a week | Neuro‐Development Treatment | 0 of 24 | None | None | Published information |
| Chua 2016 | Gait Trainer | 8 weeks | Not stated | Not stated | Physiotherapy including 25 minutes of stance/gait, 10 minutes cycling, 10 minutes tilt table standing | 20 of 106 | 2 deaths, 3 refusals, 1 medical problem, 1 transport problem (1 pain as adverse event) | 1 death, 6 refusals, 3 medical problems, 1 administrative problem, 2 inability to contact (no adverse events) | Published information |
| Cotinat 2023 | Lokomat | 2 weeks | 18/15 | Gait training 30 minutes, 5 times a week | Physiotherapist‐aided overground gait training | 1 of 34 | Knee pain during RAGT | None | Published information |
| De Luca 2020 | EksoGT | 8 weeks | Not stated | 60 minutes, 3 times a week | Physiotherapist‐aided gait training | 0 of 30 | None | None | Published information |
| Dias 2006 | Gait Trainer | 4 weeks | Not stated | 40 minutes, 5 times a week | Bobath method, 5 times a week for 5 weeks | 0 of 40 | None | None | Unpublished and published information provided by study authors. |
| Erbil 2018 | RoboGait | 3 weeks | 28/15 | 90 minutes, 5 days/week | Physical therapy including stretching, strengthening exercises, proprioception, weight‐bearing, balance, co‐ordination, and ambulatory training, 90 minutes, 5 days/week | 5 of 48 | Not stated | Not stated | Published information |
| Fisher 2008 | AutoAmbulator | 24 sessions | Not stated | Minimum 3 sessions a week, up to 5 sessions; number of minutes in each session unclear | Standard physical therapy, 3 to 5 times a week for 24 consecutive sessions | 0 of 20 | 14 adverse events; no details provided | 11 adverse events; no details provided | Unpublished and published information provided by study authors. |
| Forrester 2014 | Anklebot | 8 to 10 sessions (with about 200 repetitions) | Not stated | 60 minutes, 8 to 10 sessions | Stretching of the paretic ankle | 5 of 34 | Total of 5 dropouts across both groups (1 medical complication, 1 discharge before end of study, 2 time poststroke > 49 days, 1 non‐compliance) | Published information provided by study authors. | |
| Gandolfi 2019 | G‐EO system | 5 weeks, 10 individual rehabilitation sessions | 13/3 | 45 minutes/2 days per week | Sensory Integration Balance Training including overground gait training, stairs up and down, passive lower limb joint mobilisation and stretching exercises for the same duration as experimental group | 4 of 32 | Not stated for both groups | Published information | |
| Geroin 2011 | Gait Trainer | 2 weeks | Not stated | 50 minutes, 5 times a week | Walking exercises according to the Bobath approach | 0 of 30 | None | None | Unpublished and published information provided by study authors. |
| Gorsler 2024 | Lyra‐THERA gait trainer | 3 weeks | 16/25 | 16/18 | Mobilisation in an upright position with the aid of a standing frame | 7 of 43 | Not related to therapy | Not stated | Unpublished information |
| Han 2016 | Lokomat | 4 weeks | 33/23 | 30 minutes, 5 times a week | Neurodevelopmental techniques for balance and mobility | 4 of 60 | None | 4 unclear reasons | Published information provided by study authors. |
| Hidler 2009 | Lokomat | 8 to 10 weeks (24 sessions) | 47/16 | 45 minutes, 3 days a week | Conventional gait training, 3 times a week for 8 to 10 weeks (24 sessions); each session lasted 1.5 hours | 9 of 72 | Not described by group (9 withdrew or were removed because of poor attendance or a decline in health, including 1 death, which was unrelated to study according to study authors) | Unpublished and published information provided by study authors. | |
| Hornby 2008 | Lokomat | 12 sessions | 22/26 | 30 minutes, 12 sessions | Therapist‐assisted gait training, 12 sessions. Each session lasted 30 minutes. | 14 of 62 | 4 participants dropped out (2 discontinued secondary to leg pain during training, 1 experienced pitting oedema, and 1 had travel limitations) | 10 participants dropped out (4 discontinued secondary to leg pain, 1 experienced injury outside therapy, 1 reported fear of falling during training, 1 presented with significant hypertension, 1 had travel limitations, and 2 experienced subjective exercise intolerance) | Published information provided by study authors. |
| Husemann 2007 | Lokomat | 4 weeks | 22/8 | 30 minutes, 5 times a week | Conventional physiotherapy, 30 minutes per day for 4 weeks. Information as provided by study authors | 2 of 32 | 1 participant enteritis | 1 participant pulmonary embolism | Published information |
| Jayaraman 2019 | Honda Stride Management Assist | 6 to 8 weeks | 33/17 | 45 minutes per session, 3 times per week | Overground gait training, functional task‐specific training | 4 of 54 | 2 transportation problems | 2 transportation problems | Information as provided by study authors |
| Kang 2021 | SUBAR ground walking exoskeletal robot | 3 weeks | 16/14 | 30 minutes, 10 sessions | Traditional neurodevelopmental treatment techniques | 2 of 30 | 2 discontinued intervention | n.a. | Published information |
| Kayabinar 2019 | Robot‐assisted device | 5 weeks | Not stated | 40 minutes, 15 sessions of training, 3 times per week | Conventional gait training 40 minutes, 3 days per week | 2 of 66 | 1 early discharge | 1 early discharge | Information as provided by the study authors |
| Kelley 2013 | Lokomat | 8 weeks | Not stated | 60 minutes, 5 times a week | Overground gait training by physiotherapy on level and uneven surfaces | 1 of 21 | None | 1 withdrew | Information as provided by the study authors |
| Kim 2015 | Walkbot | 4 weeks | 13/13 | 30 minutes, 5 times a week | Conventional physiotherapy (bed mobility, stretching, balance training, strengthening, symmetry training, treadmill training) | 4 of 30 | 1 rib fracture, 3 decline in health condition | Information as provided by the study authors | |
| Kim 2019a | Lokomat | 4 weeks | Not stated | 60 minutes, 20 sessions | Conventional physical therapy (CPT) | 2 of 19 | 1 withdrew | 1 withdrew | Published information |
| Kim 2019b | Morning walk | 3 weeks | 16/32 | 1.5 hours per session, 5 times per week | Conventional physiotherapy | 10 of 58 | 1 medical complication, 1 unstable mood, 1 isolation | 7 early discharge | Published information |
| Kim 2020 | End‐effector robot‐assisted gait training (G‐EO Evolution) | 4 weeks | 17/11 | 30 minutes/day, 5 times a week | Body weight‐supported treadmill training | 2 of 30 | 1 withdrawal | 1 withdrawal | Published information |
| Kooncumchoo 2021 | Invented gait training machine (I‐Walk) | 8 weeks | 24/6 | 30 minutes/day, 3 times a week | Overground gait training | 0 of 30 | None | None | Published information |
| Kotov 2021a | ExoAtlet exoskeleton | 2 weeks | Not stated | 10 to 30 minutes/day, 3 times a week | Ortorent MOTO pedal trainer for active‐passive training for the upper and lower limbs | Not stated | Not stated | Not stated | Published information |
| Kotov 2021b | ExoAtlet exoskeleton | 2 weeks | Not stated | 20 to 40 minutes/day, 5 times a week | Standard rehabilitation | Not stated | Not stated | Not stated | Published information |
| Kwon 2018 | Exowalk | 4 weeks | Not stated | 30 minutes a day, 5 days a week | Physical therapist‐assisted gait training | 0 of 41 | None | None | Published information |
| Kyung 2008 | Lokomat | 4 weeks | 18/7 | 45 minutes, 3 days a week | Conventional physiotherapy, received equal time and sessions of conventional gait training | 10 of 35 | 1 participant dropped out for private reasons (travelling); adverse events not described | 9 participants refused after randomisation (reasons not provided); adverse events not described | Unpublished and published information provided by study authors. |
| Lee 2019 | Gait Enhancing and Motivating System | 4 weeks | 18/8 | 45 minutes, 3 times per week, 10 sessions | Gait training without Gait Enhancing and Motivating System | 2 of 28 | None | 2 withdrew | Published information |
| Lee 2022 | Healbot T | 4 weeks | 30/13 | 30 minutes, 10 sessions | Conventional gait rehabilitation | 4 of 47 | 3 discontinued training | 1 withdrawal | Published information |
| Lee 2023a | Morning Walk (CUREXO) | 4 weeks | 40/9 | 5 times a week | Traditional neurodevelopmental‐treatment techniques | 6 of 55 | 2 withdrawal | 2 switched to other therapy, 1 new disease, 1 early discharge | Published information |
| Lee 2023b | FREE Walk, robotic‐exoskeleton | 4 weeks | 17/11 | 3 days a week and regular rehabilitation programmes 2 days a week | Regular rehabilitation | 0 of 38 | n.a. | n.a. | Published information |
| Li 2021a | BEAR‐H1 exoskeleton robot of lower limb | 4 weeks | 24/8 | Twice a day, 30 minutes, 5 times a week | Conventional training guided by physical therapists | 4 of 36 | 1 personal reasons | 3 personal reasons | Published information |
| Li 2021b | BEAR‐H1 | 4 weeks | Not stated | 30 minutes, 5 days a week | Routine training for walking | 16 of 130 | 8 (not stated) | 8 (not stated) | Published information |
| Lin 2022 | RAGT system (MRG‐P100, HIWIN) | 3 to 4 weeks (15 sessions) | 12/28 | RAGT additional 30 minutes, 5 sessions per week (15 sessions) | Conventional rehabilitation | 0 of 40 | n.a. | n.a. | Published information |
| Louie 2021 | EksoGT powered exoskeleton | Maximum of 8 weeks | 25/11 | RAGT 4 to 5 days a week, for 45 to 60 minutes | Usual care | 2 of 36 | 1 declined assessment | 1 second stroke | Published information |
| Lu 2017 | Lower limb rehabilitation robot | 8 weeks | Not stated | 60 minutes per day conventional rehabilitation training combined with lower limb rehabilitation robot training, 5 days per week | Conventional rehabilitation training combined with body weight support treadmill training | 0 of 60 | n.a. | n.a. | Published information |
| Mayr 2008 | Lokomat | 8 weeks | Not stated | Not stated | Add‐on conventional physiotherapy, received equal time and sessions of conventional gait training | 13 of 74 | 4 participants dropped out (reasons not provided); adverse events not described | 9 participants dropped out (reasons not provided) | Unpublished and published information provided by study authors. |
| Mayr 2018 | Lokomat | 8 weeks | Not stated | 2 hours, 5 times a week | Conventional overground physical therapy | 8 of 74 | 7 change in clinical condition | 5 change in clinical condition, 2 lack of compliance |
Published information |
| Meng 2022 | Walkbot | 4 weeks | Not stated | 45 minutes, 3 times a week | Intensity‐matched enhanced lower limb therapy | 2 of 128 | 2 personal reasons | n.a. | Published information |
| Miyagawa 2023 | Curara | 10 sessions (2 weeks) | Not stated | 30 minutes, 5 times a week | Conventional therapist‐assisted gait training | 5 of 40 | 3 (1 retracted informed consent, 1 mental health issues, and 1 discontinued the trial due to robot‐induced skin problems) | 1 had a second stroke, 1 was excluded from analysis | Published information |
| Molteni 2021 | EksoGT | 3 weeks | 63/12 | 60 minutes, 5 days a week | Conventional gait training | 5 of 80 | 2 medical issues not related to training | 3 medical issues not related to training | Published information |
| Morone 2011 | Gait Trainer | 4 weeks | 41/7 | 40 minutes, 5 times a week | Focused on trunk stabilisation, weight transfer to paretic leg, and walking between parallel bars or on the ground. Participants were helped by 1 or 2 therapists and walking aids if necessary. | 21 of 48 | 12 (hypotension, referred weakness, knee pain, urinary infection, uncontrolled blood pressure, fever, absence of physiotherapist) | 9 (hypotension, referred weakness, knee pain, ankle pain, uncontrolled blood pressure, fever, absence of physiotherapist) | Information as provided by study authors |
| Mustafaoglu 2020 | Lokomat | 6 weeks | 24/27 | 45 minutes, 2 days a week | Conventional gait training | 0 of 51 | n.a. | n.a. | Published information |
| Nam 2019 | Exowalk | 4 weeks | 20/14 | 30 minutes, 5 days a week | Physical therapist‐assisted gait training by conventional method | 6 of 40 | 6 did not complete gait training because of an individual schedule. | Published information | |
| Nam 2020 | Exowalk | 2 weeks | 25/13 | 60 minutes, 5 days a week | Physical therapist‐assisted gait training | 2 of 40 | 2 personal reasons | None | Published information |
| Noser 2012 | Lokomat | Unclear | Not stated | Not stated | Not stated | 1 of 21 | No dropouts; 2 serious adverse events (1 skin breakdown as a result of therapy, 1 second stroke during post‐treatment phase) |
1 dropout due to protocol violation; 2 serious adverse events (1 sudden drop in blood pressure at participant's home leading to brief hospitalisation, 1 sudden chest pain before therapy leading to brief hospitalisation) | Information as provided by study authors |
| Ochi 2015 | Gait‐assistance robot (consisting of 4 robotic arms for thighs and legs, thigh cuffs, leg apparatuses, and a treadmill) | 4 weeks | 10/16 | 20 minutes, 5 times a week for 4 weeks, in addition to rehabilitation treatment | Range‐of‐motion exercises, muscle strengthening, rolling over and sit‐to‐stand and activity and gait exercises | 0 of 26 | None | None | Published information |
| Ogino 2020 | Gait Exercise Assist Robot (GEAR) | 4 weeks | 5/15 | 40 minutes, 5 days a week | Treadmill training | 1 of 20 | 1 withdrew | None | Published information |
| Palmcrantz 2021 | Hybrid Assistive Limb (HAL) | 6 weeks | 21/11 (1 both) | 60 minutes, 3 days a week | Conventional gait and mobility training | 4 of 37 | 2 medical reasons, 2 withdrew | None | Published information |
| Park 2018 | Lokomat Pro | 6 weeks | 20/20 | 45 minutes, 3 times a week | General gait training using a treadmill | 0 of 40 | None | None | Published information |
| Park 2020 | Walkbot | 2 weeks | 14/0 | 30 minutes, 7 days a week | Gait training | 0 of 14 | None | None | Published information |
| Peurala 2005 | Gait Trainer | 3 weeks | 25/20 | 20 minutes, 5 times a week for 3 weeks, in addition to rehabilitation treatment | Walking overground, all participants practised gait for 15 sessions over 3 weeks (each session lasted 20 minutes). | 0 of 45 | None | None | Published information |
| Peurala 2009 | Gait Trainer | 3 weeks | 42/14 | 20 minutes, 5 times a week for 3 weeks, in addition to rehabilitation treatment | Overground walking training; in the control group, 1 or 2 physiotherapy sessions daily | 9 of 56 | 5 dropouts (2 situations worsened after 1 to 2 treatment days; 1 had 2 unsuccessful attempts with the device; 1 had scheduling problems; 1 felt protocol was too demanding) | 4 dropouts (1 felt protocol too demanding; 2 situations worsened after 1 to 2 treatment days; 1 death) | Published information |
| Picelli 2016 | G‐EO system evolution | 30 minutes a day for 5 consecutive days | Not stated | 5 days in addition to botulinum toxin injection of calf muscles | None | 0 of 22 | None | None | Published information |
| Pohl 2007 | Gait Trainer | 4 weeks | 124/31 | 20 minutes, 5 times a week | Physiotherapy every weekday for 4 weeks | 11 of 155 | 2 participants refused therapy, 1 increased cranial pressure, 1 relapsing pancreas tumour, 1 cardiovascular unstable |
4 participants refused therapy, 1 participant died, 1 had myocardial infarction | Published information |
| Saltuari 2004 | Lokomat | 2 weeks | 13/3 | A‐B‐A study: in phase A, 30 minutes, 5 days a week | Physiotherapy every weekday for 3 weeks (phase B) | 0 of 16 | None | None | Unpublished and published information provided by study authors. |
| Schroeder 2024 | Ekso GT | 4 weeks | 16/3 | ∼45 minutes, 4 times a week | Typically consisted of daily 60‐minute sessions of physiotherapy and occupational therapy, 5 times/week, besides nursing care | 3 of 26 | 2 corrupted data | 1 corrupted data | Published information |
| Schwartz 2006 | Lokomat | 6 weeks | 49/67 | 30 minutes, 3 times a week | Physiotherapy with additional gait training 3 times a week for 6 weeks | 6 of 46 | 2 participants with leg wounds, 1 with recurrent stroke, 1 refused therapy |
1 participant with recurrent stroke, 1 with pulmonary embolism |
Unpublished and published information provided by study authors. |
| Sczesny‐Kaiser 2019 | Hybrid assistive limb | 6 weeks | 14/4 | 30 minutes, 5 times a week, 30 sessions | Conventional physiotherapy | 0 of 18 | None | None | Published Information |
| Song 2021 | Morning walk | 3 weeks | 25/11 | 5 times a week | Conventional physiotherapy | 24 of 60 | Discontinued training | Discontinued training | Published information |
| Stein 2014 | Bionic leg device (AlterG) | 6 weeks | Not stated | 1 hour, 3 times a week for 6 weeks | Group exercises | 0 of 24 | None | None | Published information |
| Stolz 2019 | Robowalk | 3 months | 29/7 | 30 minutes, 5 days a week | Conventional physiotherapy | 4/40 | None | 2 medically unstable, 2 withdrew | Published information |
| Talaty 2023 | Lokomat | Unclear | 22/8 | Up to four 45‐minute gait training sessions per week | Conventional gait training | 0/30 | None | None | Published information |
| Tanaka 2012 | Gait Master4 | 4 weeks | Not stated | 20 minutes, 2 or 3 times a week (12 sessions) | Non‐intervention (non‐training) | 0 of 12 | None | None | Published information |
| Tanaka 2019 | Stride Management Assist | 10 consecutive days | 29/12 | 1 to 2 hours with 10 minutes or longer including RAGT | Conventional gait training | 5/41 | 1 participant had trouble with leg braces, 2 participants for personal reasons | 2 participants for personal reasons | Published information |
| Taveggia 2016 | Lokomat | 5 weeks | Not stated | 30 minutes, 5 sessions a week | Conventional gait training | 0 of 28 | None | None | Published information |
| Thimabut 2022 | Welwalk system (WW‐1000) | 6 weeks | 15/11 | 40 minutes, 5 sessions a week | Ground ambulation training | 0 of 26 | None | None | Published information |
| Tomida 2019 | GEAR system | 4 weeks | 6/20 | 40 minutes, 7 times a week | Conventional gait training | None | None | None | Published information |
| Tong 2006 | Gait Trainer | 4 weeks | 39/11 | 20 minutes, 5 times a week | Conventional physiotherapy alone, based on Bobath's concept | 4 of 50 | None | 2 participants were discharged before end of study, 1 participant was readmitted to an acute ward, 1 participant had deteriorating condition | Published information |
| Ucar 2014 | Lokomat | 2 weeks | Not stated | 30 minutes, 5 times a week | Conventional physiotherapy at home (focused on gait) | 0 of 22 | None | None | Published information |
| Van Nunen 2012 | Lokomat | 8 weeks | Not stated | 30 minutes, twice a week | Overground walking therapy | 0 of 30 | None | None | Unpublished and published information provided by study authors. |
| Waldman 2013 | Portable rehab robot (ankle device) | 6 weeks | Not stated | 3 times a week, 18 sessions | Stretching plantar flexors and active exercises for ankle mobility and strength | 0 of 24 | None | None | Published information |
| Wall 2020 | HAL | 4 weeks | 19/13 | 60 to 90 minutes, 4 days a week | Conventional therapy | 1 of 33 | 1 did not continue | None | Published information |
| Watanabe 2014 | Single‐leg version of Hybrid Assistive Limb (HAL) | 4 weeks | 11/11 | 20 minutes, 12 sessions | Aimed to improve walking velocity, endurance, balance, postural stability, and symmetry | 10 of 32 | 4 withdrew, 1 epilepsy, 1 technical reasons | 2 pneumonia, 2 discharged | Published information |
| Watanabe 2017 [191, 192, 193] | Robot Suit Hybrid Assistive Limb (HAL) | 4 weeks | 7/5 only intervention group, control group not stated | 3 times a week, minutes not stated | Conventional gait training | 10 of 33 | 4 withdrew, 1 medical problem, 1 technical reasons | 2 medical reasons, 2 early discharge | Published information |
| Werner 2002 | Gait Trainer | 2 weeks | 13/12 | 20 minutes, 5 times a week | Gait therapy including treadmill training with body weight support | 0 of 30 | None | None | Published information |
| Westlake 2009 | Lokomat | 4 weeks (12 sessions) | 8/8 | 30 minutes, 3 times a week | 12 physiotherapy sessions including manually guided gait training (3 times a week over 4 weeks) | 0 of 16 | None | None | Published information |
| Wright 2021 | AlterG Bionic Leg orthosis | 10 weeks | Not described | 30 minutes, 5 to 6 times a week | Usual care, physiotherapy | 3 of 34 | None | 2 declined follow‐up, 1 deep vein thrombosis | Published information |
| Xie 2023 | Soft robotic exoskeleton | 2 weeks | 23/7 | 30 minutes, 5 times a week | Physiotherapy | 0 of 30 | None | None | Published information |
| Yeung 2018 | Exoskeleton ankle robot | 5 weeks | 14/5 | 30 minutes, 20 sessions | Gait training with passive ankle foot orthosis | 0 of 19 | None | None | Published information |
| Yeung 2021 | Power‐assisted and swing‐controlled ankle robot | 20 sessions (10 weeks) | 38/9 | 20 sessions (30 minutes) twice‐weekly | Physiotherapy | 4 of 47 | 2 early discharge, 2 recurrent strokes | None | Published information |
| Yokota 2023 | HAL‐FL05 | 20 sessions | 10/12 | 20 minutes, 20 sessions, 1 to 3 sessions per day, 5 or 6 days a week | Conventional therapy | 2 of 24 | None | 1 refused to participate, 1 medical reason | Published information |
| Yoo 2023a | ExoAtlet Medy | 12 sessions | 12/5 | 30 minutes | Conventional gait training | 8 of 25 | 7 (3 early discharge, 2 general weakness, 2 dissatisfaction) | 1 early discharge | Published information |
| Yoo 2023b | HealbotG | 4 weeks (10 sessions) | 18/11 | 30 minutes, 10 sessions | Treadmill training without RAGT | 1 of 30 | 1 refused | None | Published information |
| Yu 2021 | Gait Training and Evaluation System A3 (NX) | 14 consecutive days | 33/21 | 50 minutes each day for 14 days | Conventional overground gait training is based on traditional neurodevelopmental therapy techniques. | 31 of 85 | 16 training schedule conflict, 10 not meeting inclusion criteria, 5 lost contact | Published information | |
| Yun 2018 | Lokomat | 3 weeks | 11/25 | 30 minutes per day, 5 days a week | Conventional physical therapy based on neurodevelopmental techniques developed by Bobath and the physiotherapy proposed by Karnath | 2 of 19 | 1 recurrent stroke | 1 pneumonia | Published information |
| Zhang 2023 | Lower extremity rehabilitation robot (MANBUZHEKANGFU) | 4 weeks | Not described | 30 minutes, 5 days a week | Conventional overground gait training | 4 of 38 | 1 transferred home | 3 transferred home | Published information |
n.a.: not available RAGT: robotic‐assisted gait training
We used the I² statistic to assess heterogeneity. We considered an I² > 50% as indicating substantial heterogeneity, in which case we explored individual trial characteristics to identify potential sources of the heterogeneity. We used a random‐effects model, regardless of the level of heterogeneity.
Investigation of heterogeneity and subgroup analysis
As planned in our protocol [46], we performed for our critical outcome a formal subgroup analysis following the guidance in the Cochrane Handbook for Systematic Reviews of Interventions [47], comparing participants treated in the acute and subacute phases of their stroke (within three months) with participants treated in the chronic phase (longer than three months).
We performed three further (post hoc) subgroup analyses.
Ambulatory status at the start of study (including only studies that included an independent walker; including only studies that included dependent and independent walkers; and including only studies that included a dependent walker).
Types of devices used in trials (including only studies that used end‐effector devices and including only studies that used exoskeleton devices).
Type of content of the control group (physiotherapy, usual care, or no therapy for walking as comparator).
We assessed subgroups using the formal test for subgroup differences in RevMan.
Equity‐related assessment
We did not investigate equity‐related characteristics in this review.
Sensitivity analysis
As planned in our protocol, we performed a sensitivity analysis of methodological quality for each included study.
We carried out the following sensitivity analyses by including only those studies:
with an adequate sequence generation process;
with adequate concealed allocation;
with blinded assessors for the critical outcome; and
without incomplete outcome data.
We considered it necessary to do a further sensitivity analysis by removing the largest study, Pohl 2007 [48], because some of the review authors (JM and MP) were investigators in this large trial. We carried out this sensitivity analysis by including all studies without the largest study (Pohl 2007).
Certainty of the evidence assessment
We created one summary of findings table using the following outcomes.
-
Critical outcome measure: independent walking:
at the end of intervention phase, all electromechanical devices used. Scale from 0 to infinity;
at follow‐up after study end. Scale from 0 to infinity.
-
Important outcome measure: mean walking velocity (metres per second):
at the end of intervention phase. Scale from 0 to infinity;
at follow‐up. Scale from 0 to infinity.
-
Important outcome measure: mean walking capacity (metres walked in 6 minutes):
at the end of intervention phase. Scale from 0 to infinity;
at follow‐up. Scale from 0 to infinity.
Important outcome measure: lost to study during intervention phase: number of dropouts.
Two review authors (BE, JM) used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness, and publication bias) to assess the certainty of evidence as it relates to the studies that contribute data to the meta‐analyses for the prespecified outcomes [49]. We used the methods and recommendations described in Section 8.5 and Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions [50], employing GRADEpro GDT software [51]. We justified all decisions to downgrade the certainty of evidence using footnotes and made comments to aid the reader's understanding of the review where necessary.
If there were multiple points at which follow‐up data were collected in the trials, we selected outcome data from the first follow‐up after the end of the intervention in the trials.
Consumer involvement
Consumers were not involved in this version of the review due to limited resources, although the review authors did use core outcome sets for the review’s outcomes, which were developed with consumer involvement.
Results
Description of studies
See Included studies and Supplementary material 2, Supplementary material 3, and Supplementary material 5.
We contacted 10 trial authors to request or clarify data, with two (20%) responding.
We provide a full description of the included trials in Characteristics of included studies and a summary of trial and participant characteristics in Table 2 and Table 3.
Results of the search
A flow diagram of the study selection process for this update is shown in Figure 1.
1.
Study flow diagram.
Our searches of the electronic databases and trials registers generated 6674 new unique references for screening. After excluding non‐relevant citations, we obtained the full text of 120 new papers, of which 39 new trials were included in the review.
Included studies
We included 101 studies involving a total of 4224 participants (see Supplementary material 2; Figure 1; Table 2; Table 3). All included studies investigated the effects of electromechanical‐ or robot‐assisted gait‐training devices in improving walking after stroke (Akinci 2023 [52, 53]; Alingh 2021 [54]; Aprile 2017 [55]; Aprile 2019; Aschbacher 2006 [56]; Bang 2016 [57]; Belas dos Santos 2018 [58]; Bergmann 2018 [59, 60]; Brincks 2011 [61]; Buesing 2015; Calabrò 2018; Chang 2012 [62, 63]; Cho 2015 [64]; Choi 2022 [65]; Chua 2016 [66, 67]; Cotinat 2023 [68, 69]; De Luca 2020 [70]; Dias 2006 [71]; Erbil 2018 [72]; Fisher 2008 [73, 74]; Forrester 2014; Gandolfi 2019 [75, 76]; Geroin 2011 [77]; Gorsler 2024 [78]; Han 2016 [79]; Hidler 2009 [80]; Hornby 2008 [81]; Husemann 2007 [82, 83]; Jayaraman 2019 [84, 85]; Kang 2021 [86]; Kayabinar 2019 [87]; Kelley 2013 [88]; Kim 2015 [89]; Kim 2019a [90, 91]; Kim 2019b [92, 93]; Kim 2020 [94, 95]; Kooncumchoo 2021 [96]; Kotov 2021a [97]; Kotov 2021b [98]; Kwon 2018 [99]; Kyung 2008 [100, 101]; Lee 2019 [102]; Lee 2022 [103]; Lee 2023a [104]; Lee 2023b [105]; Li 2021a [106]; Li 2021b [107, 108]; Lin 2022 [109, 110]; Louie 2021 [111]; Lu 2017 [112]; Mayr 2008 [113, 114]; Mayr 2018 [115]; Meng 2022 [116]; Miyagawa 2023 [117, 118]; Molteni 2021 [119, 120]; Morone 2011 [121]; Mustafaoglu 2020 [122]; Nam 2019; Nam 2020 [123]; Nam 2022 [124]; Noser 2012 [125]; Ochi 2015 [126]; Ogino 2020 [127]; Palmcrantz 2021 [128, 129]; Park 2018 [130]; Park 2020 [131]; Park 2021 [132]; Peurala 2005 [133, 134, 135]; Peurala 2009 [136]; Picelli 2016 [137]; Pohl 2007; Pournajaf 2023 [138]; Rojek 2020 [139]; Saltuari 2004 [140]; Schroeder 2024 [141, 142]; Schwartz 2006 [143]; Sczesny‐Kaiser 2019 [144]; Song 2021 [145, 146]; Stein 2014; Stolz 2019 [147]; Talaty 2023 [148]; Tanaka 2012 [149]; Tanaka 2019 [150]; Taveggia 2016 [151, 152]; Thimabut 2022 [153, 154]; Tomida 2019 [155]; Tong 2006 [156, 157, 158, 159]; Ucar 2014 [160]; Van Nunen 2012 [161, 162]; Waldman 2013; Wall 2020 [163]; Watanabe 2014; Werner 2002 [164]; Westlake 2009 [165]; Wright 2021 [166, 167, 168, 169]; Xie 2023 [170]; Yeung 2018 [171, 172]; Yeung 2021 [173]; Yokota 2023 [174]; Yoo 2023a [175]; Yoo 2023b [176]; Yu 2021 [177]; Yun 2018 [178]; Zhang 2023 [179]).
Sample size, number of participants
The included studies had different sample sizes, ranging from 155 participants in Pohl 2007 to 12 participants in Tanaka 2012.
Population
The included studies recruited adult patients between seven and 14 days (Schroeder 2024) and seven to eight years (Ogino 2020) after stroke.
Age
The mean age was 61 years across all studies (Table 2).
Sex
More men than women (42%) were included in the studies (Table 2).
Ambulatory status
One in four participants were unable to walk at the start of treatment. One in two participants in the treatment group who were initially unable to walk were able to walk independently by the end of the intervention phase.
Setting
Most studies randomised participants in the inpatient rehabilitation setting, but in some studies participants were treated in the outpatient setting.
More detailed information about participant characteristics, including inclusion/exclusion criteria and baseline demographics, is provided in Supplementary material 2.
Interventions and comparators
A total of 38 different electromechanical‐ or robot‐assisted gait‐training devices were examined in the intervention group. Most trials compared a device with physiotherapy, walking overground, usual care, or a combination of these comparators. Sometimes usual care included physiotherapy, but this was not always clearly described by the study authors. Only two trials did not describe any intervention as a comparator. In most trials, the training period lasted three to four weeks, the shortest being 10 days and the longest being eight weeks.
Outcomes
Most studies described the ability to walk independently (and used appropriate scales such as the FAC); 73 studies measured walking velocity and 42 studies measured walking capacity.
All trials provided information about dropouts. Deaths were rare and were only reported in three trials.
Excluded studies
We excluded 239 full‐text articles, for the following reasons (Figure 1):
irrelevant participant characteristics (11 articles);
irrelevant interventions (61 articles);
irrelevant comparison (16 articles)
irrelevant outcome (11 articles)
irrelevant study types (92 articles);
multiple publications (48 articles).
We provide the reasons for exclusion of 38 studies in Supplementary material 3, among which two studies used an irrelevant intervention, 14 used irrelevant comparisons, five used an irrelevant outcome, two used an irrelevant study design, and 15 studies were excluded for other reasons, such as duplication of an already included trial.
Ongoing studies
We identified 80 ongoing studies. Fourteen studies formerly categorised as awaiting classification have since been excluded given that we were unable to contact study authors for years. For further details on the ongoing studies, see Supplementary material 5. One study for which we were unable to make contact with the trialists is still awaiting classification; see Supplementary material 4.
Risk of bias in included studies
Risk of bias in the included studies is described in detail in Supplementary material 2; Supplementary material 9; Figure 2.
2.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
In general, there was low or unclear risk of bias arising from the randomisation and allocation process across the included studies. However, there was a low or high risk of bias for blinding and selective reporting across studies.
We did not find important other potential sources of bias. There were no important differences in risk of bias ratings between the two raters' assessments.
Synthesis of results
1. Electromechanical‐ and robot‐assisted gait training plus physiotherapy versus physiotherapy (or usual care)
Independent walking (critical outcome)
1.1. Independent walking at the end of the intervention phase, with all electromechanical devices used
Fifty‐one trials with a total of 2148 participants measured our critical outcome independent walking at study end. However, for some of the included trials, no effect estimate (OR) was feasible because no events (e.g. no participant reached the ability to walk) or only events (e.g. all participants regained walking) were reported (Analysis 1.1) (Supplementary material 6, Supplementary material 7) [47].
The use of electromechanical devices in gait rehabilitation for people after stroke probably increased the chance of walking independently (OR 1.65, 95% CI 1.21 to 2.25; P = 0.001; I² = 31%; moderate‐certainty evidence; Table 1). However, several studies investigated at least some participants who were already independent in walking at the start of the study, while other studies included only non‐ambulatory participants (Analysis 4.1). Of the total population of 4224 participants, 49.7% were independent and 34.7% were dependent walkers (or non‐ambulatory) at the start of the study.
1.2. Independent walking at follow‐up after study end
Eight trials with a total of 569 participants measured our critical outcome independent walking at follow‐up after study end. However, for two of these trials (125 participants), no effect estimate (OR) was feasible because no events (e.g. no participant reached ability to walk) or only events (e.g. all participants regained walking) were reported (Analysis 1.2). The use of electromechanical devices for gait rehabilitation of people after stroke may not increase the chance of walking independently at follow‐up after study end (OR 1.64, 95% CI 0.77 to 3.48; P = 0.20; I² = 69%; low‐certainty evidence). However, some included trials investigated participants who were already independent in walking at the start of the study. We could draw no definitive conclusion regarding a longer‐lasting effect of the use of electromechanical devices.
Mean walking velocity (important outcome)
1.3. Walking velocity (metres per second) at the end of the intervention phase
Seventy‐three trials with a total of 3043 participants provided data for walking velocity (m/s) at study end (Analysis 1.3). The use of electromechanical devices for gait rehabilitation probably did not increase walking velocity. The pooled MD (random‐effects model) for walking velocity was 0.05 m/s (95% CI 0.02 to 0.08; P < 0.001; I² = 58%; moderate‐certainty evidence). In some studies, participants who were unable to walk were regarded as having a walking velocity of zero metres per second.
1.4. Walking velocity (metres per second) at follow‐up
Seventeen trials with a total of 857 participants provided data for walking velocity (m/s) at follow‐up after study end (Analysis 1.4). The use of electromechanical devices for gait rehabilitation probably did not increase walking velocity at follow‐up after study end. The pooled MD (random‐effects model) for walking velocity was 0.05 m/s (95% CI −0.03 to 0.13; P = 0.22; I² = 66%; moderate‐certainty evidence; Analysis 1.4). In some studies, participants who were unable to walk were regarded as having a walking velocity of zero metres per second.
Mean walking capacity (important outcome)
1.5. Walking capacity (metres walked in 6 minutes) at the end of the intervention phase
Forty‐two trials with a total of 1966 participants provided data for walking capacity (metres walked in 6 minutes) at study end (Analysis 1.5). The use of electromechanical devices in gait rehabilitation did not increase the walking capacity of people after stroke. The pooled MD (random‐effects model) for walking capacity was 11 metres walked in 6 minutes (95% CI 1.76 to 20.25; P = 0.02; I² = 43%; high‐certainty evidence; Analysis 1.5).
1.6. Walking capacity (metres walked in 6 minutes) at follow‐up
Fifteen trials with a total of 736 participants provided data for walking capacity (metres walked in 6 minutes) at follow‐up after study end (Analysis 1.6). The use of electromechanical devices for gait rehabilitation probably did not increase walking capacity at follow‐up after study end. The pooled MD (random‐effects model) for walking capacity was 9.6 metres walked in 6 minutes (95% CI −14.6 to 33.7; P = 0.44; I² = 53%; moderate‐certainty evidence; Analysis 1.6).
Lost to study during intervention phase, dropouts (important outcome)
1.7. Adverse outcomes: acceptability of electromechanical‐assisted gait‐training devices during the intervention phase in terms of dropouts
All trialists provided information about participants who dropped out from all causes during the trial period, but some trials reported no events/dropouts (Analysis 1.7). Data were not available to focus specifically on dropouts due to acceptability. We have therefore analysed data related to 'lost to study', which include reasons that are unlikely to have anything to do with acceptability.
The use of electromechanical devices for gait rehabilitation of people after stroke did not increase or decrease the risk of participants dropping out from the study (RD (random‐effects model) 0.00, 95% CI −0.02 to 0.01; P = 0.82; I² = 0%; high‐certainty evidence). The reasons for dropouts and all adverse events are described in detail for each trial in Table 3.
Death from all causes until the end of the intervention phase (important outcome)
1.8. Death from all causes until the end of the intervention phase
Only three larger trials reported any deaths during the intervention period (Chua 2016; Hidler 2009; Pohl 2007). In Pohl 2007, one participant in the control group died as the result of aspiration pneumonia, and one participant in the treatment group died due to recurrent stroke. One death occurred in Hidler 2009, but the group in which it occurred was not stated. We therefore used a worst‐case (conservative) scenario and counted the one death for the experimental group. In Chua 2016, the deaths occurred after the treatment period. The use of electromechanical devices for gait rehabilitation of people after stroke did not increase or decrease the risk of participants dying during the intervention period (RD (random‐effects model) 0.00, 95% CI −0.01 to 0.01; P = 0.87; I² = 0%; high‐certainty evidence; Analysis 1.8).
Subgroup and sensitivity analyses
Using the results from the critical outcome, it is possible to explore the apparent effectiveness of electromechanical‐assisted devices for regaining walking ability. Of 1113 initially non‐ambulatory participants in the treatment group, 546 (49%) were walking independently at the end of the intervention phase. We used the critical outcome of independently walking at the end of the intervention phase for all included participants (OR 1.65) to calculate the NNTB. Together with our control event rate of 42% (432 out of 1035 initially non‐ambulatory control participants were independently walking at the end of treatment), we calculated an NNTB of 9 (95% CI 7 to 11) [45]. This means that every ninth dependency in walking ability after stroke could be avoided with the use of electromechanical‐assisted devices.
Because people in different stages of their stroke were recruited (some in the first three weeks and some years after their stroke), our effect estimate could have been influenced by the duration of illness. We therefore performed a subgroup analysis comparing participants in the first three months and chronic phase of stroke. However, the results suggest that these categories of stroke duration probably do not affect the results of the treatment (Supplementary material 9).
We argue that more than half of the 4224 participants included were able to walk independently at baseline (see Description of studies and Supplementary material 2). We therefore performed three further subgroup analyses by ambulatory status at the start of the study (Analysis 4.1; Analysis 4.2; Analysis 4.3). We did not find subgroup differences between ambulatory and non‐ambulatory participants in these analyses. The results suggest that people's walking ability at the start of the study probably does not affect the results of the treatment.
About 38 different devices used in 101 trials were assessed in this review. We put these into different subgroups: end‐effector devices, exoskeleton devices, mobile devices, and ankle devices. In these analyses, we found subgroup differences between the device subgroups in ability to walk at the end of the intervention (Analysis 5.1). Furthermore, walking velocity and walking capacity at the end of the intervention phase were higher when specific device subtypes were used (such as end‐effector devices) compared with training (e.g. by an exoskeleton device) (Analysis 5.2; Analysis 5.3).
We performed a subgroup analysis comparing trials that offered physiotherapy with trials that did not offer physiotherapy (or only described it as 'usual care' or no therapy was given) to the control group (Analysis 6.1). Our formal subgroup analysis showed no differences in regaining independent walking (P = 0.06) between participants who were dependent or independent walkers at baseline. However, there were only a few trials where physiotherapy was not part of the control group, or was inadequately prescribed, or was not prescribed at all.
We conducted a sensitivity analysis to examine the effect of methodological quality on the odds of walking independence. We also analysed the results after removing the largest study. We found that the benefit was relatively robust when we removed trials with an inadequate sequence generation process, inadequate allocation concealment, no blinded assessors for the critical outcome, and incomplete outcome data, and when we removed Pohl 2007 (Analysis 2.1).
All sensitivity and subgroup analyses can be found in Supplementary material 9; Analysis 2.1; Analysis 3.1; Analysis 4.1; Analysis 4.2; Analysis 5.1; Analysis 5.2; Analysis 5.3; Analysis 6.1.
Equity assessment
We did not investigate equity‐related characteristics in this review.
Reporting biases
We created three funnel plots to explore possible small‐study and publication bias; see Figure 3; Figure 4; Figure 5.
3.
Funnel plot of comparison: 1 Electromechanical‐ and robot‐assisted gait training plus physiotherapy versus physiotherapy (or usual care), outcome: 1.1 Independent walking at end of intervention phase, all electromechanical devices used.
4.
Funnel plot of comparison: 1 Electromechanical‐ and robot‐assisted gait training plus physiotherapy versus physiotherapy (or usual care), outcome: 1.3 Walking velocity (metres per second) at end of intervention phase.
5.
Funnel plot of comparison: 1 Electromechanical‐ and robot‐assisted gait training plus physiotherapy versus physiotherapy (or usual care), outcome: 1.5 Walking capacity (metres walked in 6 minutes) at end of intervention phase.
In one of our three funnel plots (Figure 3), we could see a slight asymmetry in favour of positive studies, which could indicate a bias due to non‐publication or other factors such as methodological or clinical diversity of the studies.
Discussion
The aim of this review was to evaluate the effects of electromechanical‐ and robot‐assisted gait‐training devices for improving walking after stroke. We sought to estimate the likelihood or chance of becoming independent in walking as a result of these interventions, which is a main rehabilitation goal for people who have had a stroke ([5, 37]; Hornby 2008; [38]).
Summary of main results
We included 101 trials with a total of 4224 participants after stroke (with an average age of 47 to 76 years). The effects of physiotherapy plus electromechanical and robotic devices for gait training were compared with the effects of physiotherapy alone or usual care. In most studies, the training period lasted three to four weeks; the shortest time was 10 days, and the longest eight weeks. The main results are presented in Table 1.
At the end of training, compared with physiotherapy or usual care, using a gait‑training device plus physiotherapy:
probably helps more people walk independently (moderate‐certainty evidence);
probably does not increase people's average walking velocity (moderate‐certainty evidence);
does not increase the distance people could walk in 6 minutes (high‐certainty evidence); and
does not increase or decrease how many people dropped out of the study or how many people died (high‐certainty evidence); deaths were rare.
For every nine people treated with a device plus physiotherapy, probably one extra person was able to walk independently by the end of treatment.
At follow‐up after the intervention, compared with physiotherapy or usual care, using a gait‑training device plus physiotherapy may not help more people to walk independently (low‐certainty evidence), and probably does not increase people's average walking velocity (moderate‐certainty evidence) or the distance people could walk in 6 minutes (moderate‐certainty evidence).
Overall completeness and applicability of the evidence
The clinical importance of improvements in walking velocity and capacity has not been established. In addition, the long‐term sustainability of these benefits (beyond three months) is questionable, with only eight trials examining the extended effects on regaining walking ability. However, it is noteworthy that the use of electromechanical‐ or robot‐assisted gait‐training devices appears to be safe and well‐tolerated, as evidenced by the absence of increased adverse events, dropouts, or deaths among participants receiving this intervention compared to physiotherapy or usual care.
Despite these findings, several important aspects of electromechanical‐assisted gait training remain unclear, as follows:
optimal frequency of training sessions;
ideal duration for using assistive technologies;
most effective timing for application of the intervention.
These points highlight the need for further research to determine the most efficient and effective protocols for electromechanical‐assisted gait training in rehabilitation settings.
Limitations of the evidence included in the review
The applicability and external validity of the evidence are limited due to several factors. First, the exclusion of specific patient groups, such as individuals over 80 years of age, those with unstable cardiovascular conditions, cognitive or communication deficits, and limited range of motion in lower limb joints at baseline, restricts the generalisability of the findings. Additionally, heterogeneity among trials further impacts external validity. Differences in trial design (e.g. two‐arm, three‐arm, parallel‐group, or cross‐over trials), duration of follow‐up, participant selection criteria, therapy interventions (particularly intervention duration), and participant characteristics (e.g. time since stroke onset and stroke severity at baseline) contribute to variability in the description of interventions and comparators, thereby limiting broader applicability of the results.
There was heterogeneity between the trials in terms of how many people were lost to follow‐up. Some trials reported no dropouts, while others reported almost 50% lost to follow‐up. In addition, many of the trials had no follow‐up at all. This may have biased our results about walking recovery in the long term.
It is not clear whether observed differences between experimental and control groups depend on the intensity of therapy, in terms of repetition of gait practice. Time devoted to therapy is a crude measure of intensity. A 30‐minute therapy session could include no walking practice or high‐intensity walking practice with many steps taken. Reviews of the effectiveness of arm robotic therapy suggest that the positive benefit of robotic therapy may be lost when the intensity of practice is matched between experimental and control groups [180]. However, the numbers of repetitions in experimental and control groups were not exactly counted in any of the included studies. Further studies should therefore ascertain whether the benefits described here are still apparent when the intensity of gait practice (e.g. step repetitions) is exactly matched between groups.
In the description of the original trials, it was sometimes unclear what the comparator was, for example usual care could mean that physiotherapy was provided or not, or that other care was provided. As the control intervention often improves walking, there could be a difference between trials where no physiotherapy was offered and trials where physiotherapy was not offered or was given to the control group.
Subgroup analyses
The results of our three subgroup analyses by ambulatory status at the start of the study suggest that the people's walking ability at the start of the study probably does not affect the results of the treatment.
The results of our subgroup analyses between the device subgroups in ability to walk at the end of the intervention suggest that the device used during the intervention probably affects the outcome of the treatment. This could be interpreted to mean that the type of device used probably plays a role in improving walking function after stroke. This is in line with former versions of this review from 2013 [181], from 2017 [9], and from 2020 [27], and is consistent with other reviews that compared the effects of different types of devices on walking ability after stroke [38, 182]. However, in the absence of a direct empirical comparison between electromechanical‐assisted gait‐training devices, this point warrants further investigation. Taken together, we have low to high overall confidence in our results.
In our subgroup analysis comparing trials that included physiotherapy with trials that did not include physiotherapy in the control group (or only described it as 'usual care' or provided no treatment), there were no differences in regaining independent walking. There were few trials where physiotherapy was not part of the control group, or was not described adequately or at all. Finally, the results of the second subgroup were based on only one trial. The conclusions of this subgroup analysis are therefore limited.
It would be interesting for clinicians to know the effect of different treatment intensities and time after stroke on walking ability. Unfortunately, the information in the trial descriptions about the intensity and time after stroke in the trials was rarely precise enough or only given in ranges, so we have not yet done a subgroup analysis. We believe that such subanalyses would be more reliable with more detailed individual patient data.
Risk of bias
We noted methodological differences in the mechanisms of randomisation and in the allocation concealment methods used, as well as in blinding of our critical outcome and the presence or use of intention‐to‐treat analysis and reporting that could have affected the reliability of the results. Some studies were of low or poor quality, had a high risk of bias or a small sample size, or both. Consequently, some trials may have made the benefits of these devices seem greater than they are. We did not find important other potential sources of bias. There were no important differences in risk of bias ratings across raters' assessments.
The certainty of the evidence in the review was high, moderate, or low, downgraded primarily due to imprecision.
Trials investigating electromechanical‐ and robot‐assisted gait‐training devices are subject to potential methodological limitations. These include the inability to blind the therapist and participants, so‐called contamination (provision of the intervention to the control group), and co‐intervention (when the same therapist unintentionally provides additional care to either the treatment or comparison group). All these potential methodological limitations introduce the possibility of performance bias. However, this was not supported in our sensitivity analyses by methodological quality. After examining the effects of methodological quality on the odds of independence in walking, we found that the benefits were relatively robust when we removed trials with an inadequate sequence generation process, inadequate concealed allocation, no blinded assessors for the critical outcome, and incomplete outcome data (Analysis 2.1). We therefore did not downgrade the certainty of evidence for this outcome. However, we found that the odds of independence in walking were slightly lower after the largest included study (Pohl 2007, N = 155) was removed, but a clinically relevant benefit for participants was still observed (Analysis 2.1).
We did use funnel plots to investigate the potential for publication bias. By visual inspection of these plots, we found some graphical evidence of publication bias for our critical outcome (Analysis 1.1), but not for Analysis 1.4 or Analysis 1.3 (see Figure 3, Figure 4, and Figure 5). This could be explained by various reasons, such as publication bias and small‐trial effects. However, as the asymmetry was present in only one of the three outcomes, it is not clear whether this represents true publication bias or is due to other factors. In light of the graphical evidence of publication bias for our critical outcome, we have downgraded the certainty of the evidence from high to moderate.
Ongoing studies
Many studies on this topic have not yet been completed, or their results have not yet been published. Given that we found several ongoing studies of substantial size, these ongoing studies could potentially impact our overall conclusion when they are included in the review (see Supplementary material 5).
Although we found that electromechanical‐assisted gait training in combination with physiotherapy probably helped more people walk independently, it should be mentioned that we do not know yet whether these devices provide any cost benefit. Further studies should investigate, under the premise that gait practice is matched in terms of objective measures of intensity, the long‐term costs of regaining walking ability and the cost‐effectiveness of these devices.
Limitations of the review processes
The risk of publication bias is present in all systematic reviews. However, we performed an extensive search for relevant literature in electronic databases and handsearched conference abstracts and searched for trials in all languages, arranging for translation of relevant papers where necessary. Additionally, we asked study authors, trialists, and experts in the field for information on other unpublished and ongoing trials. Although we conducted an extensive literature search, we acknowledge the possibility that we may not have identified all relevant studies.
We used standard Cochrane review methodology to minimise bias and followed MECIR standards when conducting this review and reporting our findings. None of the authors has any financial conflicts of interest related to the topic of this review, and none had any direct involvement in the analysis of the clinical trials included in this review.
We experienced a limitation in the review process when attempting to obtain additional information from trial investigators. Despite our efforts to contact them for data clarification, we did not receive responses to all questions.
Agreements and disagreements with other studies or reviews
A recent and relevant review describes the effects of new so‐called powered mobile solutions [183]. We included 23 studies of mobile devices in this update. When pooling these results, we did not find improvements in walking velocity and walking capacity; this result is in agreement with the recent review of [183]. Additionally, we included four studies describing the effects of ankle robots in improving walking (Forrester 2014; Waldman 2013; Yeung 2018; Yeung 2021). When pooling these studies, we did not find evidence for improvements in walking velocity and walking capacity.
However, currently only a few clinical comparisons of two or more interventions to improve walking ability have been reported, although in practice, it is crucial to know which device performs more effectively than others in a given situation. The rehabilitation team also encounters difficulties in deciding which specific form of treatment to prescribe for a stroke patient. A methodological approach to solving this problem might be the network meta‐analysis. Such a network meta‐analysis recently investigated conventional gait training, training on a treadmill (with or without body weight support and/or in combination with or without a walking velocity paradigm), and electromechanical‐assisted gait training with end‐effectors or exoskeletons [38]. Overall, 95 randomised controlled trials involving a total of 4458 post‐stroke patients were included in the analysis. The review authors found that gait velocity (the primary endpoint) and gait training assisted by end‐effectors led to a mean improvement of 0.16 m/s [38]. This result of the network meta‐analysis is in accordance with the results of our review update, especially for our sensitivity analysis about the type of device used for regaining walking velocity (Analysis 5.2).
It is possible that some devices might be better for very severely affected patients who are not able to walk. However, no clear evidence about this is currently available. A recent large cohort study with broad inclusion criteria and not many exclusion criteria showed that in patients who are not able to walk (those with FAC equal to 0 and FAC equal to 1), gait training with end‐effector devices improved walking ability [184]. Our main finding is that for improving walking ability, the type of device is not important (Analysis 5.1).
Our review in a broader context
Electromechanical devices for gait rehabilitation are not only used for stroke patients, but also for other neurological diseases such as spinal cord injury or multiple sclerosis.
People with spinal cord injury
Several studies have demonstrated effects of robot‐assisted gait training in people with spinal cord injury [185, 186, 187, 188]. For instance, a systematic review and meta‐analysis showed improved lower extremity function [186], while another concluded that robot‐assisted gait training seems to be helpful for people with spinal cord to improve the time in the timed‐up‐and‐go‐test (but not gait distance, walking velocity, leg strength, 10‐metre walking test (10 MWT), and 6‐minute walking test (6 MWT)) [185].
In their meta‐analysis, Park and colleagues found that robot‐assisted gait training significantly improved walking abilities [188]. In this line, Moriarty and colleagues showed that robotic therapy could improve ambulation in people with spinal cord injuries compared to standard treatment only [187].
People with multiple sclerosis
Recently, the effects of robot‐assisted gait training were also summarised in people with multiple sclerosis [189, 190].
One systematic review stated that the evidence is uncertain on the effectiveness of robot‐assisted gait training on balance and gait in people with multiple sclerosis, but that a multimodal rehabilitation approach, including robot‐assisted gait training, should be encouraged [189]. Another systematic review included 16 studies enrolling 536 participants. They found (small) improvements at the end of the intervention with regard to walking velocity, walking endurance, mobility, and balance [190].
The results from these patient groups demonstrate the therapeutic potential of electromechanical‐assisted gait training for individuals affected by spinal cord injury and multiple sclerosis. The findings of the current review, when considered in conjunction with the results of other systematic reviews, suggest that this type of technology probably enhances the gait rehabilitation of people with diverse neurological impairments. It could therefore be argued that there is no good reason to believe that the benefits of electromechanical devices for gait rehabilitation will be significantly different for people with neurological conditions other than stroke. In addition, there are conditions, such as hemiparesis following removal of a brain tumour, where trials will never be conducted and the best available evidence will be considered.
Authors' conclusions
Implications for practice
This Cochrane review provides moderate‐certainty evidence that the use of electromechanical‐assisted gait‐training devices in combination with physiotherapy probably increases the chance of regaining independent walking ability among people after stroke. These results could be interpreted as preventing one participant from remaining dependent in walking after stroke for every nine treated. However, this benefit to patients is not supported by the follow‐up results after study end.
Gait‐training devices probably do not increase walking velocity and do not increase walking capacity (moderate‐ and high‐certainty evidence).
This review, together with other systematic reviews, increases the probability that electromechanical‐assisted gait‐training devices in combination with physiotherapy will assist other patients with similar loss due to other causes of neurological damage such as partial spinal cord injury who are unable to walk, or people with multiple sclerosis who have recently lost the ability to walk.
Equity‐related implications for practice
We did not investigate equity‐related characteristics in this review.
Implications for research
Given the large numbers of completed and ongoing studies, there is no need for additional small studies; however, well‐designed, large‐scale, multicentre studies are needed to evaluate the benefits and harms of electromechanical‐assisted gait training for walking after stroke, including only non‐ambulatory people in the very early stages after stroke. Comparisons between different devices are also currently lacking. Future research should include estimates of the costs (or savings) associated with electromechanical gait training. Further analyses should investigate whether non‐ambulatory or ambulatory people benefit most, and trials should include outcome measures in the activities of daily living and quality of life domains. In future updates of this review, we will consider investigating the effects of different control interventions using subgroup analysis. Additionally, in the next update, we will compare the effects of different durations and intensities of treatment (e.g. less than versus more than four weeks, five days per week versus less than five days).
Future studies should investigate the potential benefits of electromechanical gait training for people with other neurological conditions, such as partial spinal cord injury, multiple sclerosis, and traumatic brain injury, who find walking difficult or impossible, and compare whether there are disease‐specific subtypes.
Equity‐related implications for research
We did not investigate equity‐related characteristics in this review.
Supporting Information
Supplementary materials are available with the online version of this article: 10.1002/14651858.CD006185.pub5.
Supplementary materials are published alongside the article and contain additional data and information that support or enhance the article. Supplementary materials may not be subject to the same editorial scrutiny as the content of the article and Cochrane has not copyedited, typeset or proofread these materials. The material in these sections has been supplied by the author(s) for publication under a Licence for Publication and the author(s) are solely responsible for the material. Cochrane accordingly gives no representations or warranties of any kind in relation to, and accepts no liability for any reliance on or use of, such material.
Supplementary material 1 Search strategies
Supplementary material 2 Characteristics of included studies
Supplementary material 3 Characteristics of excluded studies
Supplementary material 4 Characteristics of studies awaiting classification
Supplementary material 5 Characteristics of ongoing studies
Supplementary material 6 Analyses
Supplementary material 7 Data package
Supplementary material 8 World Health Organization International Clinical Trials Registry Platform (WHO ICTRP)
Supplementary material 9 Supplementary material: sensitivity and subgroup analysis
New search for studies and content updated (no change to conclusions)
Additional information
Acknowledgements
We thank all participants and supporters.
Editorial and peer‐reviewer contributions
Cochrane Stroke supported the authors in the development of this review update. The following people conducted the editorial process for this article:
Sign‐off Editor (final editorial decision): Derick T Wade, Professor of Neurological Rehabilitation, Centre for Movement, Occupation and Rehabilitation Sciences (MOReS), Oxford Brookes University, Headington Campus, Oxford;
Managing Editor (selected peer reviewers, provided support and editorial guidance to authors, edited the article): Luisa M Fernandez Mauleffinch, Cochrane Central Editorial Service;
Editorial Assistant (conducted editorial policy checks, collated peer‐reviewer comments, and supported the editorial team): Addie‐Ann Smyth, Cochrane Central Editorial Service;
Copy Editor (copy editing and production): Lisa Winer, Cochrane Central Production Service;
Peer reviewers (provided comments and recommended an editorial decision): Dr Maxence Compagnat, MD, PhD, HAVAE UR20217, University of Limoges, PMR Department, University Hospital Center of Limoges, Limoges, France (clinical/content review); Edozie Iweka, Advanced Practice Radiographer, University Hospitals of Derby and Burton NHS Foundation Trust, Derby, UK (consumer review); Jo‐Ana Chase, Cochrane Evidence Production and Methods Directorate (methods review); Nuala Livingstone, Cochrane Evidence Production and Methods Directorate (methods review); and Jo Platt, Central Editorial Information Specialist (search review).
Contributions of authors
Jan Mehrholz (JM) contributed to the conception and design of the protocol and drafted the protocol. He searched electronic databases and conference proceedings, screened titles and abstracts of references identified by the search, selected and assessed trials, extracted trial and outcome data, guided the analysis and interpretation of data, and contributed to and approved the final manuscript of the review.
Joachim Kugler (JK) evaluated and extracted trial and outcome data, assessed the methodological quality of selected trials, contributed to the interpretation of data, and contributed to and approved the final manuscript of the review.
Marcus Pohl (MP) contributed to the conception and design of the review, drafted the protocol, and assessed the methodological quality of selected trials. Together with JM, he contacted trialists about unpublished data and entered the data, carried out statistical analysis, helped with the interpretation of data, drafted the review, and approved the final manuscript of the review.
Bernhard Elsner (BE) searched electronic databases and conference proceedings, screened titles and abstracts of references identified by the search, selected and assessed trials, extracted trial and outcome data, guided analysis and interpretation of the data, and contributed to and approved the final manuscript of the review.
Some authors involved in previous published versions of this review in 2006 [46], 2007 [26], 2013 [181], 2017 [9], and 2020 [27] are no longer included on the author byline: Cordula Werner and Simone Thomas. Some of the content retained in this review reflects their contributions.
Declarations of interest
Jan Mehrholz: was co‐author of two included trials (Gorsler 2024; Pohl 2007).
Joachim Kugler: none known.
Marcus Pohl: was a co‐author of one included trial (Pohl 2007).
Bernhard Elsner: none known.
These review authors (MP, JM) did not participate in quality assessment or data extraction for these studies.
Jan Mehrholz was a former co‐editor of the Cochrane Stroke Review Group. The author had no role in the editorial process of this review.
Sources of support
Internal sources
-
Technical University Dresden, Lehrstuhl Public Health, Germany
Internal
-
Klinik Bavaria Kreischa, Wissenschaftliches Institut, Germany
Internal
-
Universität zu Lübeck, Institut für Gesundheitswissenschaften, Department of Physiotherapy, Germany
Internal
External sources
-
No external support, Other
None
Registration and protocol
Protocol (2006)[46]
Original review (2007)[26]
Review update (2013)[181]
Review update (2017)[9]
Review update (2020)[27]
Data, code and other materials
As part of the published Cochrane review, the following are made available for download for users of the Cochrane Library: full search strategies for each database; full citations of each unique Cochrane Database of Systematic Reviews report for all studies included, ongoing or awaiting classification, or excluded at the full‐text screen, in the final review; study data, including study information, study arms, and study results or test data; consensus risk of bias assessments; and analysis data, including overall estimates and settings, subgroup estimates, and individual data rows. Appropriate permissions have been obtained for such use. Analyses and data management were conducted within Cochrane’s authoring tool, Review Manager, using the inbuilt computation methods.
What's new
| Date | Event | Description |
|---|---|---|
| 10 June 2024 | New citation required but conclusions have not changed | The conclusions of the review have not changed. |
| 10 June 2024 | New search has been performed | We have updated the searches to January 2024 and revised the text as appropriate. We have included 101 studies with 4224 participants in this update, compared with 62 trials with 2440 participants in the previous version of this review (from 2020). |
History
Protocol first published: Issue 4, 2006 Review first published: Issue 3, 2007
| Date | Event | Description |
|---|---|---|
| 6 January 2020 | New search has been performed | We have updated the searches to January 2020 and revised the text as appropriate. We have included 62 studies with 2440 participants in this update, compared with 36 trials with 1472 participants in the previous version of this review (from 2017). |
| 6 January 2020 | New citation required but conclusions have not changed | The conclusions of the review have not changed. |
| 10 May 2017 | New citation required and conclusions have changed | The conclusions of the review have changed. The previous version of this review concluded that, for the primary outcome (walking), the number needed to treat was 5 patients to prevent 1 dependency; this updated version of our review concludes that 7 patients need to be treated to prevent 1 dependency in walking. |
| 10 May 2017 | New search has been performed | We have updated the searches to September 2016 and revised the text as appropriate. We have included 36 studies with 1472 participants in this update, compared with 23 trials with 999 participants in the previous version of this review (from 2013). |
| 23 January 2013 | New citation required and conclusions have changed | The conclusions of the review have changed. The previous version of this review concluded that, for the primary outcome (walking), the number needed to treat was 6 patients to prevent 1 dependency; this updated version of our review concludes that 5 patients need to be treated to prevent 1 dependency in walking. |
| 14 January 2013 | New search has been performed | We have updated the searches to December 2012 and revised the text as appropriate. We have included 23 trials with 999 participants in this update, compared with 17 trials with 837 participants in the previous version of this review (from 2009). |
| 28 July 2010 | Feedback has been incorporated | Feedback and author response are included in the Feedback section, and an error in the Abstract has been corrected. |
| 16 October 2009 | New search has been performed | We have updated the searches to April 2009 and revised the text as appropriate. The conclusions of the review have not changed. We have included 17 trials with 837 participants in this update, compared with 8 trials with 414 participants in the previous version of this review (from 2007). The previous version of this review concluded that, for the primary outcome (walking), the number needed to treat was 4 patients to prevent 1 dependency; this updated version of our review concludes that 6 patients need to be treated to prevent 1 dependency in walking. |
| 6 August 2008 | Amended | We have converted the review to the new review format. |
References
- 1.World Health Organization. Cerebrovascular accident, stroke. www.who.int/topics/cerebrovascular_accident/en/ (accessed 1 February 2006).
- 2.Adams HP Jr, Bendixen BH, Kappelle LJ, Biller J, Love BB, Gordon DL, et al. Classification of subtype of acute ischemic stroke: definitions for use in a multicenter clinical trial (TOAST: Trial of Org 10172 in Acute Stroke Treatment). Stroke 1993;24(1):35-41. [DOI] [PubMed] [Google Scholar]
- 3.GBD 2016 Stroke Collaborators. Global, regional, and national burden of stroke, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet Neurology 2019;18(5):439-58. [DOI] [PMC free article] [PubMed]
- 4.Jorgensen H, Nakayama H, Raaschou H, Olsen T. Recovery of walking function in stroke patients: the Copenhagen stroke study. Archives of Physical Medicine and Rehabilitation 1995;76:27-32. [DOI] [PubMed] [Google Scholar]
- 5.Bohannon RW, Horton MG, Wikholm JB. Importance of four variables of walking to patients with stroke. International Journal of Rehabilitation Research 1991;14:246-50. [DOI] [PubMed] [Google Scholar]
- 6.Hornby TG, Reisman DS, Ward IG, Scheets PL, Miller A, Haddad D, et al. Clinical practice guideline to improve locomotor function following chronic stroke, incomplete spinal cord injury, and brain injury. Journal of Neurologic Physical Therapy 2020;44(1):49-100. [DOI] [PubMed] [Google Scholar]
- 7.Winstein CJ, Stein J, Arena R, Bates B, Cherney LR, Cramer SC, et al. Guidelines for adult stroke rehabilitation and recovery. Stroke 2016;47(6):e98-e169. [DOI] [PubMed] [Google Scholar]
- 8.States RA, Pappas E, Salem Y. Overground physical therapy gait training for chronic stroke patients with mobility deficits. Cochrane Database of Systematic Reviews 2009, Issue 3. Art. No: CD006075. [DOI: 10.1002/14651858.CD006075.pub2] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Mehrholz J, Thomas S, Werner C, Kugler J, Pohl M, Elsner B. Electromechanical-assisted training for walking after stroke. Cochrane Database of Systematic Reviews 2017, Issue 5. Art. No: CD006185. [DOI: 10.1002/14651858.CD006185.pub4] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Colombo G, Joerg M, Schreier R, Dietz V. Treadmill training of paraplegic patients using a robotic orthosis. Journal of Rehabilitation Research and Development 2000;37(6):693-700. [PubMed] [Google Scholar]
- 11.Nam YG, Lee JW, Park JW, Lee HJ, Nam KY, Park JH, et al. Effects of electromechanical exoskeleton-assisted gait training on walking ability of stroke patients: a randomized controlled trial. Archives of Physical Medicine and Rehabilitation 2019;100(1):26-31. [DOI] [PubMed]
- 12.Aprile I, Iacovelli C, Goffredo M, Cruciani A, Galli M, Simbolotti C, et al. Efficacy of end-effector robot-assisted gait training in subacute stroke patients: clinical and gait outcomes from a pilot bi-centre study. NeuroRehabilitation 2019;45(2):201-12. [DOI] [PubMed] [Google Scholar]
- 13.Hesse S, Sarkodie-Gyan T, Uhlenbrock D. Development of an advanced mechanised gait trainer, controlling movement of the centre of mass, for restoring gait in non-ambulant subjects. Biomedizinische Technik [Biomedical Engineering] 1999;44(7-8):194-201. [DOI] [PubMed] [Google Scholar]
- 14.Schmidt H, Hesse S, Bernhardt R, Krüger J. HapticWalker – a novel haptic foot device. ACM Transactions on Applied Perception 2005;2(2):166-80. [Google Scholar]
- 15.Massachusetts Institute of Technology. MIT develops Anklebot for stroke patients. web.mit.edu/newsoffice/2005/stroke-robot.html (accessed 20 December 2005).
- 16.Veneman J, Kruidhof R, Helm FCT, Kooy H. Design of a series elastic- and Bowdencable-based actuation system for use as torque-actuator in exoskeleton-type training robots. In: International Conference on Rehabilitation Robotics; 2005 June 28-July 1; Chicago (IL). 2005.
- 17.Buesing C, Fisch G, O'Donnell M, Shahidi I, Thomas L, Mummidisetty CK, et al. Effects of a wearable exoskeleton stride management assist system (SMA) on spatiotemporal gait characteristics in individuals after stroke: a randomized controlled trial. Journal of Neuroengineering and Rehabilitation 2015;12:69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Calabrò RS, Naro A, Russo M, Bramanti P, Carioti L, Balletta T, et al. Shaping neuroplasticity by using powered exoskeletons in patients with stroke: a randomized clinical trial. Journal of Neuroengineering and Rehabilitation 2018;15(1):35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.NCT03162263. Improving functional outcomes and neuroplasticity by using Ekso. https://clinicaltrials.gov/show/NCT03162263 (first posted 22 May 2017).
- 20.Stein J, Bishop L, Stein DJ, Wong CK. Gait training with a robotic leg brace after stroke. American Journal of Physical Medicine & Rehabilitation 2014;93(11):987-94. [DOI] [PubMed] [Google Scholar]
- 21.Watanabe H, Tanaka N, Inuta T, Saitou H, Yanagi H. Locomotion improvement using a hybrid assistive limb in recovery phase stroke patients: a randomized controlled pilot study. Archives of Physical Medicine and Rehabilitation 2014;95(11):2006-12. [0003: 9993] [DOI] [PubMed] [Google Scholar]
- 22.JPRN-UMIN000022335. Effects of rehabilitation using the hybrid assistive limb in stroke patients. www.who.int/trialsearch/Trial2.aspx?TrialID=JPRN-UMIN000022335; (first posted not known).
- 23.Forrester LW, Roy A, Krywonis A, Kehs G, Krebs HI, Macko RF. Modular ankle robotics training in early subacute stroke: a randomized controlled pilot study. Neurorehabilitation and Neural Repair 2014;28(7):678-87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Waldman G, Yang C-Y, Ren Y, Liu L, Guo X, Harvey RL, et al. Effects of robot-guided passive stretching and active movement training of ankle and mobility impairments in stroke. NeuroRehabilitation 2013;32(3):625-34. [DOI] [PubMed] [Google Scholar]
- 25.Hesse S, Schmidt H, Werner C, Bardeleben A. Upper and lower extremity robotic devices for rehabilitation and for studying motor control. Current Opinion in Neurology 2003;16(6):705-10. [DOI] [PubMed] [Google Scholar]
- 26.Mehrholz J, Werner C, Kugler J, Pohl M. Electromechanical-assisted training for walking after stroke. Cochrane Database of Systematic Reviews 2007, Issue 4. Art. No: CD006185. [DOI: 10.1002/14651858.CD006185.pub2] [DOI] [PubMed] [Google Scholar]
- 27.Mehrholz J, Thomas S, Kugler J, Pohl M, Elsner B. Electromechanical-assisted training for walking after stroke. Cochrane Database of Systematic Reviews 2020, Issue 10. Art. No: CD006185. [DOI: 10.1002/14651858.CD006185.pub5] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Higgins JP, Savović J, Page MJ, Elbers RG, Sterne JA. Chapter 8. Assessing risk of bias in a randomized trial. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, Welch VA, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.0 (updated July 2019). Cochrane, 2019. Available from https://training.cochrane.org/handbook/archive/v6.
- 29.Higgins JP, Lasserson T, Thomas J, Flemyng E, Churchill R. Methodological Expectations of Cochrane Intervention Reviews. London: Cochrane, 2023. [Google Scholar]
- 30.Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021;372:n71. [DOI: 10.1136/bmj.n71] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Freivogel S, Schmalohr D, Mehrholz J. Improved walking ability and reduced therapeutic stress with an electromechanical gait device. Journal of Rehabilitation Medicine 2009;41:734-9. [DOI] [PubMed] [Google Scholar]
- 32.Hesse S, Waldner A, Tomelleri C. Innovative gait robot for the repetitive practice of floor walking and stair climbing up and down in stroke patients. Journal of NeuroEngineering and Rehabilitation 2010;7:30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Nuyens GE, De Weerdt WJ, Spaepen AJ Jr, Kiekens C, Feys HM. Reduction of spastic hypertonia during repeated passive knee movements in stroke patients. Archives of Physical Medicine and Rehabilitation 2002;83(7):930-5. [DOI] [PubMed] [Google Scholar]
- 34.Mehrholz J, Thomas S, Elsner B. Treadmill training and body weight support for walking after stroke. Cochrane Database of Systematic Reviews 2017, Issue 8. Art. No: CD002840. [DOI: 10.1002/14651858.CD002840.pub4] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.French B, Thomas LH, Coupe J, McMahon NE, Connell L, Harrison J, et al. Repetitive task training for improving functional ability after stroke. Cochrane Database of Systematic Reviews 2016, Issue 11. Art. No: CD006073. [DOI: 10.1002/14651858.CD006073.pub3] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Pollock A, Farmer SE, Brady MC, Langhorne P, Mead GE, Mehrholz J. Interventions for improving upper limb function after stroke. Cochrane Database of Systematic Reviews 2014, Issue 11. Art. No: CD010820. [DOI: 10.1002/14651858.CD010820.pub2] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Bohannon R. Rehabilitation goals of patients with hemiplegia. International Journal of Rehabilitation Research 1988;11(2):181-3. [Google Scholar]
- 38.Mehrholz J, Pohl M, Kugler J, Elsner B. The improvement of walking ability following stroke: a systematic review and network meta-analysis of randomized controlled trials. Deutsches Arzteblatt International 2018;115:639-45. [DOI: 10.3238/arztebl.2018.0639] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Holden MK, Gill KM, Magliozzi MR, Nathan J, Piehl-Baker L. Clinical gait assessment in the neurologically impaired: reliability and meaningfulness. Physical Therapy 1984;64(1):35-40. [DOI] [PubMed] [Google Scholar]
- 40.Wade DT, Collin C. The Barthel ADL Index: a standard measure of physical disability? International Disability Studies 1988;10(2):64-7. [DOI] [PubMed] [Google Scholar]
- 41.Hamilton BB, Laughlin JA, Fiedler RC, Granger CV. Interrater reliability of the 7-level functional independence measure (FIM). Scandinavian Journal of Rehabilitation Medicine 1994;26(3):115-9. [PubMed] [Google Scholar]
- 42.Collen FM, Wade DT, Robb GF, Bradshaw CM. The Rivermead Mobility Index: a further development of the Rivermead Motor Assessment. International Disability Studies 1991;13(2):50-4. [DOI] [PubMed] [Google Scholar]
- 43.Higgins JP, Altman DG, Sterne JA. Chapter 8. Assessing risk of bias in included studies. In: Higgins JP, Green S, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 (updated March 2011). The Cochrane Collaboration, 2011. Available from https://training.cochrane.org/handbook/archive/v5.1/.
- 44.Review Manager (RevMan). Version 8.18.0. The Cochrane Collaboration, 2025. Available at https://revman.cochrane.org.
- 45.Sackett DL, Deeks JJ, Altman DG. Down with odds ratios! Evidence-Based Medicine 1996;1:164-6. [Google Scholar]
- 46.Mehrholz J, Werner C, Kugler J, Pohl M. Electromechanical-assisted training for walking after stroke. Cochrane Database of Systematic Reviews 2006, Issue 4. Art. No: CD006185. [DOI: 10.1002/14651858.CD006185] [DOI] [PubMed] [Google Scholar]
- 47.Deeks JJ, Higgins JP, Altman DG. Chapter 10. Analysing data and undertaking meta-analyses. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, Welch VA, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.0 (updated July 2019). Cochrane, 2019. Available from https://training.cochrane.org/handbook/archive/v6.
- 48.Pohl M, Werner C, Holzgraefe M, Kroczek G, Mehrholz J, Wingendorf I, et al. Repetitive locomotor training and physiotherapy improve walking and basic activities of daily living in subacute, non-ambulatory stroke patients: a single-blind, randomised multi-centre trial (DEutsche GAngtrainerStudie, DEGAS). Clinical Rehabilitation 2007;21(1):17-27. [DOI] [PubMed] [Google Scholar]
- 49.Atkins D, Best D, Briss PA, Eccles M, Falck-Ytter Y, Flottorp S, et al. Grading quality of evidence and strength of recommendations. BMJ 2004;328:1490. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Schünemann HJ, Vist GE, Higgins JP, Santesso N, Deeks JJ, Glasziou P, et al. Chapter 15. Interpreting results and drawing conclusions. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, Welch VA, editor(s). Cochrane Handbook for Systematic Reviews of Interventions Version 6.0 (updated July 2019). Cochrane, 2019. Available from https://training.cochrane.org/handbook/archive/v6.
- 51.GRADEpro GDT. Version accessed 5 March 2025. Hamilton (ON): McMaster University (developed by Evidence Prime), 2025. Available at https://www.gradepro.org.
- 52.Akıncı M, Burak M, Yaşar E, Kılıç RT. The effects of robot-assisted gait training and virtual reality on balance andgait in stroke survivors: a randomized controlled trial. Gait & Posture 2023;103:215-22. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 53.NCT05419791. The effects of robot assisted gait training with virtual reality. https://clinicaltrials.gov/show/NCT05419791 (first posted 11 June 2022).
- 54.Alingh JF, Fleerkotte BM, Groen BE, Rietman JS, Weerdesteyn V, Asseldonk EHF, et al. Effect of assist-as-needed robotic gait training on the gait pattern post stroke: a randomized controlled trial. Journal of Neuroengineering and Rehabilitation 2021;18(1):26. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Aprilea I, Iacovelli C, Padua L, Galafate D, Criscuolo S, Gabbani D, et al. Efficacy of robotic-assisted gait training in chronic stroke patients: preliminary results of an Italian bi-centre study. NeuroRehabilitation 2017;41(4):775-82. [DOI] [PubMed] [Google Scholar]
- 56.Aschbacher B. Comparing gait training in patients after stroke with task oriented physiotherapy or robot-assisted treadmill training: a feasibility study. Unpublished 2006.
- 57.Bang DH, Shin WS. Effects of robot-assisted gait training on spatiotemporal gait parameters and balance in patients with chronic stroke: a randomized controlled pilot trial. NeuroRehabilitation 2016;38:343-9. [DOI] [PubMed] [Google Scholar]
- 58.Belas dos Santos M, Barros de Oliveira C, dos Santos A, Garabello Pires C, Dylewski V, Arida RM. A comparative study of conventional physiotherapy versus robot-assisted gait training associated to physiotherapy in individuals with ataxia after stroke. Behavioural Neurology 2018;2018. [DOI] [PMC free article] [PubMed]
- 59.Bergmann J, Krewer C, Jahn K, Muller F. Robot-assisted gait training to reduce pusher behavior: a randomized controlled trial. Neurology 2018;91(14):e1319-27. [DOI] [PubMed] [Google Scholar]
- 60.DRKS00003444. Treatment study with locomotor training vs conventional physiotherapy in patients with pusher behaviour. www.drks.de/drks_web/navigate.do?navigationId=trial.HTML&TRIAL_ID=DRKS00003444 (first registered 28 November 2013).
- 61.Brincks J. The order of gait training, including Lokomat® and Physiotherapy, does not influence gait symmetry in subacute ambulatory persons with stroke. In: World Physical Therapy , editor(s). Physiotherapy, Supplement S1, 16th World Congress of Physical Therapy, 2011 June 20-23, Amsterdam. Vol. 97. World Confederation of Physical Therapy, 2011.
- 62.Chang WH, Kim MS, Huh JP, Lee PK, Kim YH. Effects of robot-assisted gait training on cardiopulmonary fitness in subacute stroke patients: a randomized controlled study. Neurorehabilitation and Neural Repair 2012;26(4):318–24. [DOI] [PubMed] [Google Scholar]
- 63.Kim M, Kim YH, Lee PKW, Hyong MK, Jung PH. Effect of robot-assisted gait therapy on cardiopulmonary fitness in subacute stroke patients. Neurorehabilitation and Neural Repair 2008;22:594. [DOI] [PubMed] [Google Scholar]
- 64.Cho DY, Park SW, Lee MJ, Park DS, Kim EJ. Effects of robot-assisted gait training on the balance and gait of chronic stroke patients: focus on dependent ambulators. Journal of Physical Therapy Science 2015;27(10):3053-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 65.Choi W. Effects of robot-assisted gait training with body weight support on gait and balance in stroke patients. International Journal of Environmental Research and Public Health 2022;19:5814. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66.Chua J, Culpan J, Menon E. Efficacy of an electromechanical gait trainer poststroke in Singapore: a randomized controlled trial. Archives of Physical Medicine and Rehabilitation 2016;97:683-90. [DOI] [PubMed] [Google Scholar]
- 67.ACTRN12616000844448. An electromechanical gait trainer for ambulation training after stroke. http://www.who.int/trialsearch/Trial2.aspx?TrialID=ACTRN12616000844448 2016.
- 68.Cotinat M, Celerier M, Arquillière C, Flipo M, Prieur-Blanc N, Viton JM, et al. Robotic gait training and botulinum toxin injection improve gait in the chronic post-stroke phase: a randomized controlled trial. Annals of Physical and Rehabilitation Medicine 2024;67(1):101785. [DOI] [PubMed] [Google Scholar]
- 69.NCT03659266. Effects of combined treatment by botulinum toxin and Lokomat® on walking ability in chronic stroke. www.clinicaltrials.gov/ct2/show/NCT03659266 (first posted 15 January 2020).
- 70.De Luca R, Maresca G, Balletta T, Cannavò A, Leonardi S, Latella D, et al. Does overground robotic gait training improve non-motor outcomes in patients with chronic stroke? Findings from a pilot study. Journal of Clinical Neuroscience 2020;81:240-5. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 71.Dias D, Laíns J, Pereira A, Nunes R, Caldas J, Amaral C, et al. Partial body weight support in chronic hemiplegics: a randomized control trial. Europa Medicophysica 2007;43(4):499-504. [PubMed] [Google Scholar]
- 72.Erbil D, Tugba G, Murat TH, Melike A, Merve A, Cagla K, et al. Effects of robot-assisted gait training in chronic stroke patients treated by botulinum toxin-a: a pivotal study. Physiotherapy Research International 2018;23(3):e1718. [DOI] [PubMed] [Google Scholar]
- 73.Fisher S. Use of autoambulator for mobility improvement in patients with central nervous system (CNS) injury or disease. Neurorehabilitation and Neural Repair 2008;22:556. [Google Scholar]
- 74.Fisher S, Lucas L, Thrasher TA. Robot-assisted gait training for patients with hemiparesis due to stroke. Topics in Stroke Rehabilitation 2011;18(3):269-76. [DOI] [PubMed] [Google Scholar]
- 75.Gandolfi M, Vale N, Dimitrova E, Zanolin ME, Mattiuz N, Battistuzzi E, et al. Robot-assisted stair climbing training on postural control and sensory integration processes in chronic post-stroke patients: a randomized controlled clinical trial. Frontiers in Neuroscience 2019;13:1143. [DOI: 10.3389/fnins.2019.01143] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 76.NCT03566901. Robot-assisted stair climbing training. https://clinicaltrials.gov/ct2/show/record/NCT03566901; (first posted 25 June 2018).
- 77.Geroin C, Picelli A, Munari D, Waldner A, Tomelleri C, Smania N. Combined transcranial direct current stimulation and robot-assisted gait training in patients with chronic stroke: a preliminary comparison. Clinical Rehabilitation 2011;25(6):537-48. [DOI] [PubMed] [Google Scholar]
- 78.Gorsler A, Ernst D, Grittner U, Harnack D, Koßmehl P, Mehrholz J, et al. Effects of end-effector controlled gait training on walking ability in non-ambulatory patients with left-sided neglect. Clinical Rehabilitation (manuscript under review).
- 79.Han EY, Im SH, Kim BR, Seo MJ, Kim MO. Robot-assisted gait training improves brachial-ankle pulse wave velocity and peak aerobic capacity in subacute stroke patients with totally dependent ambulation: randomized controlled trial. Medicine 2016;95:e5078. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 80.Hidler J, Nichols D, Pelliccio M, Brady K, Campbell DD, Kahn JH, et al. Multicenter randomized clinical trial evaluating the effectiveness of the Lokomat in subacute stroke. Neurorehabilitation and Neural Repair 2009;23(1):5-13. [DOI] [PubMed] [Google Scholar]
- 81.Hornby TG, Campbell DD, Kahn JH, Demott T, Moore JL, Roth HR. Enhanced gait-related improvements after therapist- versus robotic-assisted locomotor training in subjects with chronic stroke: a randomized controlled study. Stroke 2008;39(6):1786-92. [DOI] [PubMed] [Google Scholar]
- 82.Husemann B, Müller F, Krewer C, Laß A, Gille C, Heller S, et al. Effects of locomotion training with assistance of a robot-driven gait orthosis in hemiparetic patients after stroke: a randomized controlled pilot study. Stroke 2007;38(2):349-54. [DOI] [PubMed] [Google Scholar]
- 83.Husemann B, Müller F, Krewer C, Laß A, Gille C, Heller S, et al. Effects of locomotion training with assistance of a driven gait orthosis in hemiparetic patients after stroke. Neurologie & Rehabilitation 2004;10(4):25. [Google Scholar]
- 84.Jayaraman A, O'Brien MK, Madhavan S, Mummidisetty CK, Roth HR, Hohl K, et al. Stride management assist exoskeleton vs functional gait training in stroke: a randomized trial. Neurology 2019;92(3):132. [DOI] [PubMed] [Google Scholar]
- 85.NCT03222505. Walk assist device to improve community ambulation. https://clinicaltrials.gov/ct2/show/NCT03222505; (first posted 19 July 2017).
- 86.Kang CJ, Chun MH, Lee J, Lee JY. Effects of robot (SUBAR)-assisted gait training in patients with chronic stroke: randomized controlled trial. Medicine (Baltimore) 2021;100(48):e27974. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 87.Kayabinar E, Ozalp M, Kocyigit MF, As I, Elbasan B. The effects of robotic and conventional gait training in addition to neurodevelopmental treatment on balance, mobility, and health-related quality of life in patients with stroke. Neurological Sciences and Neurophysiology 2019;36(2):112-9. [Google Scholar]
- 88.Kelley CP, Childress J, Boake C, Noser EA. Over-ground and robotic-assisted locomotor training in adults with chronic stroke: a blinded randomized clinical trial. Disability & Rehabilitation: Assistive Technology 2013;8(2):161-8. [DOI] [PubMed] [Google Scholar]
- 89.Kim S-Y, Yang L, Park IJ, Kim EJ, Park MS, You SH, et al. Effects of innovative WALKBOT robotic-assisted locomotor training on balance and gait recovery in hemiparetic stroke: a prospective, randomized, experimenter blinded case control study with a four-week follow-up. IEEE Transactions on Neural Systems & Rehabilitation Engineering 2015;23(4):636-42. [DOI] [PubMed] [Google Scholar]
- 90.Kim HY, Shin J-H, Yang SP, Shin MA, Lee SH. Robot-assisted gait training for balance and lower extremity function in patients with infratentorial stroke: a single-blinded randomized controlled trial. Journal of Neuroengineering and Rehabilitation 2019;16(1):99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 91.NCT02680691. Robot assisted gait training in patients with infratentorial stroke. https://clinicaltrials.gov/show/NCT02680691 (first posted 11 February 2016).
- 92.Kim J, Kim DY, Chun M, Kim SW, Jeon H, Hwang C, et al. Effects of robot-(Morning Walk®) assisted gait training for patients after stroke: a randomized controlled trial. Clinical Rehabilitation 2019;33(3):516‐23. [DOI] [PubMed] [Google Scholar]
- 93.Lee JK, Dae Yul K. The effect of morning walk (end-effector type robot) training for patients with hemiparesis due to stroke. European Stroke Journal 2018;3(1):124. [Google Scholar]
- 94.Kim H, Park G, Shin JH, You JH. Neuroplastic effects of end-effector robotic gait training for hemiparetic stroke: a randomised controlled trial. Scientific Reports 2020;10(1):12461. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 95.NCT04054739. Cortical activity and gait function for robotic gait training in hemiparetic stroke. www.clinicaltrials.gov/ct2/show/NCT04054739 (first posted 28 January 2020).
- 96.Kooncumchoo P, Namdaeng P, Hanmanop S, Rungroungdouyboon B, Klarod K, Kiatkulanusorn S, et al. Gait improvement in chronic stroke survivors by using an innovative gait training machine: a randomized controlled trial. International Journal of Environmental Research and Public Health 2021;19(1):224. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 97.Kotov SV, Isakova EV, Lijdvoy VY, Petrushanskaya KA, Pismennaya EV, Romanova MV, et al. Robotic restoration of gait function in patients in the early recovery period of stroke. Neuroscience and Behavioral Physiology 2021;51(5):583-9. [Google Scholar]
- 98.Kotov SV, Romanov AI, Silina EV, Stupin VA, Isakova EV, Lijdvoy VY, et al. Efficiency of leg exoskeleton use in rehabilitation of cerebral stroke patients. Experimental and Applied Biomedical Research 2021;22(3):257-64. [Google Scholar]
- 99.Kwon BS, Nam YG, Lee HJ, Jo EH, Lee JW. Effects of electromechanical assisted gait training with Exowalk on walking ability of chronic stroke patients: a randomized controlled trial. Annals of Physical and Rehabilitation Medicine 2018;Volume 61(Supplement):e35. [DOI: 10.1016/j.rehab.2018.05.077] [DOI] [PubMed] [Google Scholar]
- 100.Kyung HJ, Kim YH. Effects of robot-assisted gait therapy on locomotor recovery in stroke patients. In: Asia-Oceanian Conference of Physical & Rehabilitation Medicine; 2008 May 16-19; Nanjing, China. 2008.
- 101.Jung KH, Ha HG, Shin HJ, Ohn SH, Sung DH, Lee PKW, et al. Effects of robot-assisted gait therapy on locomotor recovery in stroke patients. Journal of Korean Academy of Rehabilitation Medicine 2008;32:258-66. [Google Scholar]
- 102.Lee H-J, Lee S-H, Seo K, Lee M, Chang WH, Choi B-O, et al. Training for walking efficiency with a wearable hip-assist robot in patients with stroke: a pilot randomized controlled trial. Stroke 2019;50(12):3545-52. [DOI] [PubMed] [Google Scholar]
- 103.Lee J, Chun MH, Seo YJ, Lee A, Choi J, Son C. Effects of a lower limb rehabilitation robot with various training modes in patients with stroke: a randomized controlled trial. Medicine (Baltimore) 2022;101(344):e31590. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 104.Lee J, Kim DY, Lee SH, Kim JH, Kim DY, Lim KB, et al. End-effector lower limb robot-assisted gait training effects in subacute stroke patients: a randomized controlled pilot trial. Medicine (Baltimore) 2023;102(42):e35568. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 105.Lee YH, Ko LW, Hsu CY, Cheng YY. Therapeutic effects of robotic-exoskeleton-assisted gait rehabilitation and predictive factors of significant improvements in stroke patients: a randomized controlled trial. Bioengineering (Basel) 2023;10(5):585. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 106.Li DX, Zha FB, Long JJ, Liu F, Cao J, Wang YL. Effect of robot assisted gait training on motor and walking function in patients with subacute stroke: a random controlled study. Journal of Stroke and Cerebrovascular Diseases 2021;30(7):105807. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 107.Li Y, Fan T, Qi Q, Wang J, Qiu H, Zhang L, et al. Efficacy of a novel exoskeletal robot for locomotor rehabilitation in stroke patients: a multi-center, non-inferiority, randomized controlled trial. Frontiers in Aging Neuroscience 2021;13:706569. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 108.ChiCTR2100044475. A prospective, multicenter, randomized controlled, non inferiority trial to evaluate the effectiveness and safety of the lower extremity exoskeleton. https://trialsearch.who.int/Trial2.aspx?TrialID=ChiCTR2100044475 (first registered 19 March 2021).
- 109.Lin YN, Huang SW, Kuan YC, Chen HC, Jian WS, Lin LF. Hybrid robot-assisted gait training for motor function in subacute stroke: a single-blind randomized controlled trial. Journal of Neuroengineering and Rehabilitation 2022;19(1):99. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 110.ISRCTN15088682. Using the robotic gait training system (RGTS) to improve mobility functions for stroke patients. www.who.int/trialsearch/Trial2.aspx?TrialID=ISRCTN15088682 (first received 27 January 2020).
- 111.Louie DR, Mortenson WB, Durocher M, Schneeberg A, Teasell R, Yao J, et al. Efficacy of an exoskeleton-based physical therapy program for non-ambulatory patients during subacute stroke rehabilitation: a randomized controlled trial. Journal of Neuroengineering and Rehabilitation 2021;18(1):149. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 112.Lu JL, Chen ZM, Wu H, Yang W, Chen HH. Effect of lower limb rehabilitation robot on lower limb motor function of hemiplegic patients after stroke. Chinese Journal of Contemporary Neurology and Neurosurgery 2017;17(5):334-9. [Google Scholar]
- 113.Mayr A, Saltuari L, Quirbach E. Impact of Lokomat training on gait rehabilitation. Neurorehabilitation and Neural Repair 2008;22(5):596. [DOI] [PubMed] [Google Scholar]
- 114.Mayr A, Quirbach E, Picelli A, Kofler M, Smania N, Saltuari L. Effect of early robot-assisted gait retraining in nonambulatory patients with stroke: a single-blind randomized controlled trial. Unpublished 2016. [DOI] [PubMed]
- 115.Mayr A, Quirbach E, Picelli A, Kofler M, Smania N, Saltuari L. Early robot-assisted gait retraining in non-ambulatory patients with stroke: a single blind randomized controlled trial. European Journal of Physical and Rehabilitation Medicine 2018;54(6):819‐26. [DOI] [PubMed] [Google Scholar]
- 116.Meng G, Ma X, Chen P, Xu S, Li M, Zhao Y, et al. Effect of early integrated robot-assisted gait training on motor and balance in patients with acute ischemic stroke: a single-blinded randomized controlled trial. Therapeutic Advances in Neurological Disorders 2022;15:17562864221123195. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 117.Miyagawa D, Matsushima A, Maruyama Y, Mizukami N, Tetsuya M, Hashimoto M, et al. Gait training with a wearable powered robot during stroke rehabilitation: a randomized parallel-group trial. Journal of Neuroengineering and Rehabilitation 2023;20(1):54. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 118.JPRN-UMIN000034237. Rehabilitation trial using robotic wear curara(R) for patients with cerebrovascular and neurodegenerative diseases. Part 1: cerebrovascular disease. www.who.int/trialsearch/Trial2.aspx?TrialID=JPRN-UMIN000034237 (first received 28 January 2020).
- 119.Molteni F, Guanziroli E, Goffredo M, Calabrò RS, Pournajaf S, Gaffuri M, et al. Gait recovery with an overground powered exoskeleton: a randomized controlled trial on subacute stroke subjects. Brain Sciences 2021;11(1):104. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 120.NCT03395717. Stroke rehabilitation with exoskeleton-assisted gait. www.clinicaltrials.gov/show/NCT03395717 (first posted 28 January 2020).
- 121.Morone G, Bragoni M, Iosa M, De Angelis D, Venturiero V, Coiro P, et al. Who may benefit from robotic-assisted gait training? A randomized clinical trial in patients with subacute stroke. Neurorehabilitation and Neural Repair 2011;25(7):636-44. [DOI] [PubMed] [Google Scholar]
- 122.Mustafaoglu R, Erhan B, Yeldan I, Gunduz B, Tarakci E. Does robot-assisted gait training improve mobility, activities of daily living and quality of life in stroke? A single-blinded, randomized controlled trial. Acta Neurologica Belgica 2020;120(2):335-44. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 123.Nam YG, Park JW, Lee HJ, Nam KY, Choi MR, Yu CS, et al. Further effects of electromechanically assisted gait trainer (Exowalk®) in patients with chronic stroke: a randomized controlled trial. Journal of Rehabilitation Medicine 2020;52(9):jrm00097. [DOI: 10.2340/16501977-2723] [DOI] [PubMed] [Google Scholar]
- 124.Nam YG, Ko MJ, Bok SK, Paik NJ, Lim CY, Lee JW, et al. Efficacy of electromechanical-assisted gait training on clinical walking function and gait symmetry after brain injury of stroke: a randomized controlled trial. Scientific Reports 2022;12(1):6880. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 125.NCT00975156. NCT00975156 [Improving ambulation post stroke with robotic training]. clinicaltrials.gov/show/NCT00975156 (first posted 10 September 2009).
- 126.Ochi M, Wada F, Saeki S, Hachisuka K. Gait training in subacute non-ambulatory stroke patients using a full weight-bearing gait-assistance robot: a prospective, randomized, open, blinded-endpoint trial. Journal of the Neurological Sciences 2015;353(1-2):130-6. [DOI] [PubMed] [Google Scholar]
- 127.Ogino T, Kanata Y, Uegaki R, Yamaguchi T, Morisaki K, Nakano S, et al. Effects of gait exercise assist robot (GEAR) on subjects with chronic stroke: a randomized controlled pilot trial. Journal of Stroke and Cerebrovascular Diseases 2020;29(8):104886. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 128.Palmcrantz S, Wall A, Vreede KS, Lindberg P, Danielsson A, Sunnerhagen KS, et al. Impact of intensive gait training with and without electromechanical assistance in the chronic phase after stroke - A multi-arm randomized controlled trial with a 6 and 12 months follow up. Frontiers in Neuroscience 2021;15:660726. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 129.NCT02545088. New technology for individualised, intensive training of gait after stroke - study II (HAL-RCT-II). www.clinicaltrials.gov/ct2/show/NCT02545088 (first posted 2 September 2016).
- 130.Park J, Chung Y. The effects of robot-assisted gait training using virtual reality and auditory stimulation on balance and gait abilities in persons with stroke. NeuroRehabilitation 2018;43(2):227-35. [DOI] [PubMed] [Google Scholar]
- 131.Park C, Oh-Park M, Dohle C, Bialek A, Friel K, Edwards D, et al. Effects of innovative hip-knee-ankle interlimb coordinated robot training on ambulation, cardiopulmonary function, depression, and fall confidence in acute hemiplegia. NeuroRehabilitation 2020;46(4):577-87. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 132.Park C, Oh-Park M, Bialek A, Friel K, Edwards D, You JSH. Abnormal synergistic gait mitigation in acute stroke using an innovative ankle-knee-hip interlimb humanoid robot: a preliminary randomized controlled trial. Scientific Reports 2021;11(1):22823. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 133.Peurala S, Tarkka I, Pitkänen K, Sivenius J. The effectiveness of body weight-supported gait training and floor walking in patients with chronic stroke. Archives of Physical Medicine and Rehabilitation 2005;86:1557-64. [DOI] [PubMed] [Google Scholar]
- 134.Peurala SH, Pitkanen K, Sivenius J, Tarkka IM. Body-weight supported gait trainer exercises with or without functional electrical stimulation improves gait in patients with chronic stroke. Neurorehabilitation and Neural Repair 2006;20(1):98. [Google Scholar]
- 135.Peurala SH, Pitkanen K, Sivenius J, Tarkka I. Body-weight supported gait exercise compared with floor walking in chronic stroke patients. Archives of Physical Medicine and Rehabilitation 2004;85:E7. [DOI] [PubMed] [Google Scholar]
- 136.Peurala SH, Airaksinen O, Huuskonen P, Jäkälä P, Juhakoski M, Sandell K, et al. Effects of intensive therapy using gait trainer or floorwalking exercises early after stroke. Journal of Rehabilitation Medicine 2009;41(3):166-73. [DOI] [PubMed] [Google Scholar]
- 137.Picelli A, Bacciga M, Melotti C, Marchina E, Verzini E, Ferrari F, et al. Combined effects of robot-assisted gait training and botulinum toxin type A effect on spastic equinus foot in patients with chronic stroke: a pilot, single blind, randomized controlled trial. European Journal of Physical and Rehabilitation Medicine 2016;52(6):759-66. [PMID: ] [PubMed] [Google Scholar]
- 138.Pournajaf S, Calabrò RS, Naro A, Goffredo M, Aprile I, Tamburella F, et al. Robotic versus conventional overground gait training in subacute stroke survivors: a multicenter controlled clinical trial. Journal of Clinical Medicine 2023;12(2):439. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 139.Rojek A, Mika A, Oleksy Ł, Stolarczyk A, Kielnar R. Effects of exoskeleton gait training on balance, load distribution, and functional status in stroke: a randomized controlled trial. Frontiers in Neurology 2020;10:1344. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 140.Saltuari L. Efficiency of Lokomat training in stroke patients. Neurologie & Rehabilitation 2004;10(4):S4. [Google Scholar]
- 141.Schroeder J, Yperzeele L, Embrechts E, Loureiro-Chaves R, Hallemans A, Lafosse C, et al. Exoskeleton-assisted training to enhancelower limb motor recovery in subacute stroke: does timing matter? A pilot randomized trial. Frontiers in Stroke 2024;3:1379083. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 142.NCT03727919. Exoskeleton-assisted training to accelerate walking recovery early after stroke. www.clinicaltrials.gov/ct2/show/NCT03727919 (first posted 16 January 2020).
- 143.Schwartz I, Katz-Leurer M, Fisher I, Sajin A, Shochina M, Meiner Z. The effectiveness of early locomotor therapy in patients with first CVA. In: Collaborative Evaluation of Rehabilitation in Stroke Across Europe (CERISE) Congress; 2006 February 10-11; Leuven, Belgium. 2006.
- 144.Sczesny-Kaiser M, Trost R, Aach M, Schildhauer TA, Schwenkreis P, Tegenthoff M. A randomized and controlled crossover study investigating the improvement of walking and posture functions in chronic stroke patients using HAL Exoskeleton - the HALESTRO Study (HAL-Exoskeleton STROke Study). Frontiers in Neuroscience 2019;13:259. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 145.Song KJ, Chun MH, Lee J, Lee C. The effect of robot-assisted gait training on cortical activation in stroke patients: a functional near-infrared spectroscopy study. NeuroRehabilitation 2021;49(1):65-73. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 146.KCT0003090. The effect of end-effector type rehabilitation robot (Morning-Walk) training for stroke patients. cris.nih.go.kr/cris/en/search/search_result_st01.jsp?seq=11942 (first received 28 January 2020).
- 147.Stolz R, Nayyar R, Louie J, Bower KJ, Paul SK, Ng L. The effectiveness of a novel cable-driven gait trainer (Robowalk) combined with conventional physiotherapy compared to conventional physiotherapy alone following stroke: a randomised controlled trial. International Journal of Rehabilitation Research 2019;42(4):377-84. [DOI] [PubMed] [Google Scholar]
- 148.Talaty M, Esquenazi A. Feasibility and outcomes of supplemental gait training by robotic and conventional means in acute stroke rehabilitation. Journal of Neuroengineering and Rehabilitation 2023;20(1):134. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 149.Tanaka N, Saitou H, Takao T, Iizuka N, Okuno J, Yano H, et al. Effects of gait rehabilitation with a footpad-type locomotion interface in patients with chronic post-stroke hemiparesis: a pilot study. Clinical Rehabiilation 2012;26(8):686-95. [DOI] [PubMed] [Google Scholar]
- 150.Tanaka N, Matsushita S, Sonoda Y, Maruta Y, Fujitaka Y, Sato M, et al. Effect of stride management assist gait training for poststroke hemiplegia: a single center, open-label, randomized controlled trial. Journal of Stroke and Cerebrovascular Diseases 2019;28(2):477-86. [DOI] [PubMed] [Google Scholar]
- 151.Taveggia G, Borboni A, Mulé C, Villafañe JH, Negrini S. Conflicting results of robot-assisted versus usual gait training during postacute rehabilitation of stroke patients: a randomized clinical trial. International Journal of Rehabilitation Research 2016;39(1):29-35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 152.NCT02095795. Effectiveness of an exoskeleton gait training versus manual therapy in subacute post stroke patients. clinicaltrials.gov/ct2/show/NCT02095795; (first posted 26 March 2014).
- 153.Thimabut N, Yotnuengnit P, Charoenlimprasert J, Sillapachai T, Hirano S, Saitoh E, et al. Effects of the robot-assisted gait training device plus physiotherapy in improving ambulatory functions in patients with subacute stroke with hemiplegia: an assessor-blinded, randomized controlled trial. Archives of Physical Medicine and Rehabilitation 2022;103(5):843-50. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 154.TCTR20180419004. Effect of the robotic-assisted gait training device (Welwalk®) plus physiotherapy in improving the ambulatory function in sub-acute hemiplegic stroke patients: investigator- blinded, randomized controlled trial. apps.who.int/trialsearch/Trial2.aspx?TrialID=TCTR20180419004 (first received 16 January 2020).
- 155.Tomida K, Sonoda S, Hirano S, Suzuki A, Tanino G, Kawakami K, et al. Randomized controlled trial of gait training using Gait Exercise Assist Robot (GEAR) in stroke patients with hemiplegia. Journal of Stroke and Cerebrovascular Diseases 2019;28(9):2421-8. [DOI] [PubMed] [Google Scholar]
- 156.Tong RKY, Ng MF, Li LS. Effectiveness of gait training using an electromechanical gait trainer, with and without functional electric stimulation, in subacute stroke: a randomized controlled trial. Archives of Physical Medicine and Rehabilitation 2006;87(10):1298-304. [DOI] [PubMed] [Google Scholar]
- 157.Ng MFW, Tong RKY, Li LSW. A pilot study of randomized clinical controlled trial of gait training in subacute stroke patients with partial body-weight support electromechanical gait trainer and functional electrical stimulation: six-month follow-up. Stroke 2008;39:154-60. [DOI] [PubMed] [Google Scholar]
- 158.Ng MFW, Tong RKY, So EFM, Li LSW. The therapeutic effect of electromechanical gait trainer and functional electrical stimulation for patients with acute stroke. Neurorehabilitation and Neural Repair 2006;20(1):97 (Abstract F1D-7). [Google Scholar]
- 159.Li LSW, Tong RYU, Ng MFW, So EFM. Effectiveness of gait trainer in stroke rehabilitation. Journal of the Neurological Sciences 2005;238 Suppl 1:S81. [Google Scholar]
- 160.Uçar DE, Paker N, Buğdaycı D. Lokomat: a therapeutic chance for patients with chronic hemiplegia. NeuroRehabilitation 2014;34(3):447-53. [1053: 8135] [DOI] [PubMed] [Google Scholar]
- 161.Van Nunen M. RCT evaluating the effectiveness of robot-assisted treadmill training in restoring walking ability of stroke patients. In: Olver, J , editor(s). 7th World Congress of Neurological Rehabilitation; 2012 May 16-19; Melbourne, Australia. Melbourne: World Federation for Neurorehabilitation, 2012:210.
- 162.Van Nunen MPM, Gerrits KHL, Konijnenbelt M, Janssen TWJ, Haan A. Recovery of walking ability using a robotic device in subacute stroke patients: a randomized controlled study. Disability & Rehabilitation: Assistive Technology 2015;10(2):141-8. [DOI] [PubMed] [Google Scholar]
- 163.Wall A, Borg J, Vreede K, Palmcrantz S. A randomized controlled study incorporating an electromechanical gait machine, the Hybrid Assistive Limb, in gait training of patients with severe limitations in walking in the subacute phase after stroke. PLOS One 2020;15(2):e0229707. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 164.Werner C, Von Frankenberg S, Treig T, Konrad M, Hesse S. Treadmill training with partial body weight support and an electromechanical gait trainer for restoration of gait in subacute stroke patients: a randomized crossover study. Stroke 2002;33(12):2895-901. [DOI] [PubMed] [Google Scholar]
- 165.Westlake K, Patten C. Pilot study of Lokomat versus manual-assisted treadmill training for locomotor recovery post-stroke. Journal of NeuroEngineering and Rehabilitation 2009;6:18. [DOI: 10.1186/1743-0003-6-18] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 166.Wright A, Stone K, Martinelli L, Fryer S, Smith G, Lambrick D, et al. Effect of combined home-based, overground robotic-assisted gait training and usual physiotherapy on clinical functional outcomes in people with chronic stroke: a randomized controlled trial. 2021 Clinical Rehabilitation;6:882-93. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 167.NCT03104127. Effect of using a lower limb robotic device for patients with chronic stroke. www.clinicaltrials.gov/show/NCT03104127 (first posted 27 January 2020).
- 168.Wright A, Stone K, Lambrick D, Fryer S, Stoner L, Tasker E, et al. A community-based, bionic leg rehabilitation program for patients with chronic stroke: clinical trial protocol. Journal of Stroke and Cerebrovascular Diseases 2018;27(2):372-80. [DOI] [PubMed] [Google Scholar]
- 169.Wright A, Jobson S, Smith G, Faulkner J. Effect of robotic-assisted gait training on the stance and swing phase of overground walking in patients with stroke. Cerebrovascular Diseases 2018;45:367. [Google Scholar]
- 170.Xie R, Zhang Y, Jin H, Yang F, Feng Y, Pan Y. Effects of soft robotic exoskeleton for gait training on clinical and biomechanical gait outcomes in patients with sub-acute stroke: a randomized controlled pilot study. Frontiers in Neurology 2023;14:1296102. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 171.Yeung L-F, Ockenfeld C, Pang M-K, Wai H-W, Soo O-Y, Li S-W, et al. Randomized controlled trial of robot-assisted gait training with dorsiflexion assistance on chronic stroke patients wearing ankle-foot-orthosis. Journal of NeuroEngineering and Rehabilitation 2018;15(1):51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 172.NCT03184259. Interactive exoskeleton robot for walking. https://clinicaltrials.gov/ct2/show/record/NCT03184259; (first posted 12 June 2017).
- 173.Yeung LF, Lau CC, Lai CW, Soo YO, Chan ML, Tong RK. Effects of wearable ankle robotics for stair and over-ground training on sub-acute stroke: a randomized controlled trial. Journal of Neuroengineering and Rehabilitation 2021;18(1):19. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 174.Yokota C, Tanaka K, Omae K, Kamada M, Nishikawa H, Koga M, et al. Effect of cyborg-type robot Hybrid Assistive Limb on patients with severe walking disability in acute stroke: a randomized controlled study. Journal of Stroke and Cerebrovascular Diseases 2023;32(4):107020. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 175.Yoo HJ, Bae CR, Jeong H, Ko MH, Kang YK, Pyun SB. Clinical efficacy of overground powered exoskeleton for gait training in patients with subacute stroke: a randomized controlled pilot trial. Medicine (Baltimore) 2023;102(4):e32761. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 176.Yoo M, Chun MH, Hong GR, Lee C, Lee JK, Lee A. Effects of training with a powered exoskeleton on cortical activity modulation in hemiparetic chronic stroke patients: a randomized controlled pilot trial. Archives of Physical Medicine and Rehabilitation 2023;104(10):1620-9. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 177.Yu D, Yang Z, Lei L, Chaoming N, Ming W. Robot-assisted gait training plan for patients in poststroke recovery period: a single blind randomized controlled trial. BioMed Research International 2021;2021:5820304. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 178.Yun N, Joo MC, Kim SC, Kim MS. Robot-assisted gait training effectively improved lateropulsion in subacute stroke patients: a single-blinded randomized controlled trial. European Journal of Physical and Rehabilitation Medicine 2018;54(6):827-36. [DOI] [PubMed] [Google Scholar]
- 179.Zhang H, Li X, Gong Y, Wu J, Chen J, Chen W, et al. Three-dimensional gait analysis and sEMG measures for robotic-assisted gait training in subacute stroke: a randomized controlled trial. BioMed Research International 2023;11:7563802. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 180.Mehrholz J, Hädrich A, Platz T, Kugler J, Pohl M. Electromechanical and robot-assisted arm training for improving generic activities of daily living, arm function, and arm muscle strength after stroke. Cochrane Database of Systematic Reviews 2012, Issue 6. Art. No: CD006876. [DOI: 10.1002/14651858.CD006876.pub3] [DOI] [PubMed] [Google Scholar]
- 181.Mehrholz J, Elsner B, Werner C, Kugler J, Pohl M. Electromechanical-assisted training for walking after stroke. Cochrane Database of Systematic Reviews 2013, Issue 7. Art. No: CD006185. [DOI: 10.1002/14651858.CD006185.pub3] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 182.Mehrholz J, Pohl M. Electromechanical-assisted gait training after stroke: a systematic review comparing end-effector and exoskeleton devices. Journal of Rehabilitation Medicine 2012;44(3):193-9. [DOI] [PubMed] [Google Scholar]
- 183.Louie DR, Eng JJ. Powered robotic exoskeletons in post-stroke rehabilitation of gait: a scoping review. Journal of NeuroEngineering and Rehabilitation 2016;13:article no. 53. [MEDLINE: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 184.Reichl S, Weilbach F, Mehrholz J. Implementation of a gait center training to improve walking ability and vital parameters in inpatient neurological rehabilitation – a cohort study. Journal of NeuroEngineering and Rehabilitation 2020;17(1):38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 185.Bin L, Wang X, Jiatong H, Donghua F, Qiang W, Yingchao S, et al. The effect of robot-assisted gait training for patients with spinal cord injury: a systematic review and meta-analysis. Frontiers in Neuroscience 2023;17:1252651. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 186.Duan R, Qu M, Yuan Y, Lin M, Liu T, Huang W, et al. Clinical benefit of rehabilitation training in spinal cord injury: a systematic review and meta-analysis. Spine (Phila Pa 1976) 2021;46(6):E398-E410. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 187.Moriarty B, Jacob T, Sadlowski M, Fowler M, Rowan C, Chavarria J, et al. The use of exoskeleton robotic training on lower extremity function in spinal cord injuries: a systematic review. Journal of Orthopaedics 2024;65:1-7. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 188.Park JM, Kim YW, Lee SJ, Shin JC. Robot-assisted gait training in individuals with spinal cord injury: a systematic review and meta-analysis of randomized controlled trials. Annals of Rehabilitation Medicine 2024;48(3):171-91. [PMID: ] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 189.Facciorusso S, Malfitano C, Giordano M, Del Furia MJ, Mosconi B, Arienti C, et al. Effectiveness of robotic rehabilitation for gait and balance in people with multiple sclerosis: a systematic review. Journal of Neurology 2024;271(11):7141-55. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 190.Yang FA, Lin CL, Huang WC, Wang HY, Peng CW, Chen HC. Effect of robot-assisted gait training on multiple sclerosis: a systematic review and meta-analysis of randomized controlled trials. Neurorehabilitation and Neural Repair 2023;37(4):228-39. [PMID: ] [DOI] [PubMed] [Google Scholar]
- 191.Watanabe H, Goto R, Tanaka N, Matsumura A, Yanagi H. Effects of gait training using the Hybrid Assistive Limb® in recovery-phase stroke patients: a 2-month follow-up, randomized, controlled study. NeuroRehabilitation 2017;40(3):363‐7. [DOI] [PubMed] [Google Scholar]
- 192.JPRN-UMIN000022335. Effects of rehabilitation using the hybrid assistive limb in stroke patients. http://www.who.int/trialsearch/Trial2.aspx?TrialID=JPRN-UMIN000022335; (first posted 2016).
- 193.Watanabe H, Tanaka N, Goto R, Yanagi H. Effects of gait training with a Hybrid Assistive Limb® in stroke patients: a randomized controlled study with a 2-month follow-up. Archives of Physical Medicine & Rehabilitation 2016;97(10):e97. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary material 1 Search strategies
Supplementary material 2 Characteristics of included studies
Supplementary material 3 Characteristics of excluded studies
Supplementary material 4 Characteristics of studies awaiting classification
Supplementary material 5 Characteristics of ongoing studies
Supplementary material 6 Analyses
Supplementary material 7 Data package
Supplementary material 8 World Health Organization International Clinical Trials Registry Platform (WHO ICTRP)
Supplementary material 9 Supplementary material: sensitivity and subgroup analysis
Data Availability Statement
As part of the published Cochrane review, the following are made available for download for users of the Cochrane Library: full search strategies for each database; full citations of each unique Cochrane Database of Systematic Reviews report for all studies included, ongoing or awaiting classification, or excluded at the full‐text screen, in the final review; study data, including study information, study arms, and study results or test data; consensus risk of bias assessments; and analysis data, including overall estimates and settings, subgroup estimates, and individual data rows. Appropriate permissions have been obtained for such use. Analyses and data management were conducted within Cochrane’s authoring tool, Review Manager, using the inbuilt computation methods.