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

Only 13 randomized trials were found among 2857 screened citations, so many outcomes could not be pooled. Sample sizes were small: individual trials ranged from 12 to 63 participants, and over half (62%) involved 30 participants or less. Most participants were in the chronic phase after stroke, with mean times up to 104 months; only three trials were in the subacute phase. No trial included non-ambulatory people, so the results do not apply to patients who cannot walk. GRADE certainty was very low for walking speed, step rate, step length, stride length and temporal symmetry, partly because of small samples and inconsistent results. Risk of bias was a concern: randomization was unclear or based on hospital record numbers in two studies, dropouts were often excluded, and selective reporting was likely in most studies. Dropout rates ranged from 0 to 37%, and some withdrawals were related to training problems such as fear of falling, pain, skin lesions or exercise intolerance. Training intensity and individual adjustments were poorly reported, especially in comparator groups, and conventional gait training was not consistently defined. Robotic devices, body-weight support, overground versus treadmill training, walking aids, assessment speed and gait-analysis systems varied widely. Kinetic data were absent: no eligible trial measured forces after training, and kinematic data were available from only three trials. Only English-language full-text articles were searched.

Do not read the step length, stride length or temporal symmetry results as proof that the robot alone improves walking after stroke. The pooled walking speed and step rate results showed no significant difference, and the certainty for these biomechanical outcomes was very low.