Effect of robotic-assisted gait training on objective biomechanical measures of gait in persons post-stroke: a systematic review and meta-analysis

Abstract

Background: Robotic-Assisted Gait Training (RAGT) may enable high-intensive and task-specific gait training post-stroke. The effect of RAGT on gait movement patterns has however not been comprehensively reviewed. The purpose of this review was to summarize the evidence for potentially superior effects of RAGT on biomechanical measures of gait post-stroke when compared with non-robotic gait training alone.

Methods: Nine databases were searched using database-specific search terms from their inception until January 2021. We included randomized controlled trials investigating the effects of RAGT (e.g., using exoskeletons or end-effectors) on spatiotemporal, kinematic and kinetic parameters among adults suffering from any stage of stroke. Screening, data extraction and judgement of risk of bias (using the Cochrane Risk of bias 2 tool) were performed by 2-3 independent reviewers. The Grading of Recommendations Assessment Development and Evaluation (GRADE) criteria were used to evaluate the certainty of evidence for the biomechanical gait measures of interest.

Results: Thirteen studies including a total of 412 individuals (mean age: 52-69 years; 264 males) met eligibility criteria and were included. RAGT was employed either as monotherapy or in combination with other therapies in a subacute or chronic phase post-stroke. The included studies showed a high risk of bias (n = 6), some concerns (n = 6) or a low risk of bias (n = 1). Meta-analyses using a random-effects model for gait speed, cadence, step length (non-affected side) and spatial asymmetry revealed no significant differences between the RAGT and comparator groups, while stride length (mean difference [MD] 2.86 cm), step length (affected side; MD 2.67 cm) and temporal asymmetry calculated in ratio-values (MD 0.09) improved slightly more in the RAGT groups. There were serious weaknesses with almost all GRADE domains (risk of bias, consistency, directness, or precision of the findings) for the included outcome measures (spatiotemporal and kinematic gait parameters). Kinetic parameters were not reported at all.

Conclusion: There were few relevant studies and the review synthesis revealed a very low certainty in current evidence for employing RAGT to improve gait biomechanics post-stroke. Further high-quality, robust clinical trials on RAGT that complement clinical data with biomechanical data are thus warranted to disentangle the potential effects of such interventions on gait biomechanics post-stroke.

Keywords: Cerebrovascular accident; Literature synthesis; Powered exoskeleton; Rehabilitation; Walk.

Fig. 1

PRISMA flowchart for identification and screening of eligible studies for the current review

Fig. 2

Risk of bias summary: review authors’ assessment of each risk of bias item for every included study (generated with the Review Manager Web, The Cochrane Collaboration, 2019, available at revman.cochrane.org)

Fig. 3

Risk of bias graph (generated with the Review Manager Web, The Cochrane Collaboration, 2019, available at revman.cochrane.org)

Fig. 4

A forest plot (generated with the Review Manager Web, The Cochrane Collaboration, 2019, available at revman.cochrane.org) summarizing a pooled effect estimate on change in gait speed (m/s), following robotic-assisted gait training (RAGT) compared with non-robotic gait training (non-RAGT), during walking at a self-selected velocity (SSV) and the fastest velocity possible (FV). CI confidence interval; df degrees of freedom; SD standard deviation

Fig. 5

A forest plot (generated with the Review Manager Web, The Cochrane Collaboration, 2019, available at revman.cochrane.org) summarizing a pooled effect estimate on change in cadence (steps/min), following robotic-assisted gait training (RAGT) compared with non-robotic gait training (non-RAGT), during walking at a self-selected velocity (SSV) and the fastest velocity possible (FV). CI confidence interval; df degrees of freedom; SD standard deviation

Fig. 6

A forest plot (generated with the Review Manager Web, The Cochrane Collaboration, 2019, available at revman.cochrane.org) summarizing a pooled effect estimate on change in step length (cm), following robotic-assisted gait training (RAGT) compared with non-robotic gait training (non-RAGT). *: assessed during walking at a self-selected velocity (SSV); ^: assessed during walking at the fastest velocity possible (FV); CI confidence interval; df degrees of freedom; SD standard deviation

Fig. 7

A forest plot (generated with the Review Manager Web, The Cochrane Collaboration, 2019, available at revman.cochrane.org) summarizing a pooled effect estimate on change in stride length (cm), following  robotic-assisted gait training (RAGT) compared with non-robotic gait training (non-RAGT). *: assessed during walking at a self-selected velocity SSV; ^: assessed during walking at the fastest velocity possible FV; CI confidence interval; df degrees of freedom; SD standard deviation

Fig. 8

A forest plot (generated with the Review Manager Web, The Cochrane Collaboration, 2019, available at revman.cochrane.org) summarizing a pooled effect estimate on change in temporal symmetry (ratio), following robotic-assisted gait training (RAGT) compared with non-robotic gait training (non-RAGT). *: assessed during walking at a self-selected velocity SSV; ^: assessed during walking at the fastest velocity possible FV; CI confidence interval; df degrees of freedom; SD standard deviation

Fig. 9

A forest plot (generated with the Review Manager Web, The Cochrane Collaboration, 2019, available at revman.cochrane.org) summarizing a pooled effect estimate on change in spatial symmetry (ratio), following robotic-assisted gait training (RAGT) compared with non-robotic gait training (non-RAGT). *: assessed during walking at a self-selected velocity SSV; ^: assessed during walking at the fastest velocity possible FV; CI confidence interval; df degrees of freedom; SD standard deviation