How robot-assisted gait training affects gait ability, balance and kinematic parameters after stroke: a systematic review and meta-analysis.

High heterogeneity in several outcomes: I² was 94% for walking speed, 87% for FAC and stride length, and 81% for cadence, meaning the included studies varied widely in their results and the pooled estimates are less reliable. Eight of the 27 studies did not state whether blinding was used, one did not use it, and nine did not mention allocation concealment, which weakens the internal validity of the pooled results. The Berg Balance Scale result is internally inconsistent: the Results section reports a significant difference (P=0.01), but the Discussion describes it as no significant advantage for the robot group. Only two of the 27 studies combined robot training with conventional gait training, so the findings apply to robot training used as a standalone intervention, not as an add-on. Several outcomes were based on very few studies: non-affected step length (3 studies), spatial symmetry (4 studies), and each joint-angle measure (4 studies each), making those pooled estimates fragile. The authors note that differences in robot systems (exoskeleton vs end-effector, HAL vs other platforms) and treatment protocols across studies make direct comparison difficult and contribute to heterogeneity.

The authors state they have no conflict of interest with any financial organization. One included study (Nam, reference 39) was flagged as high risk of bias due to research and development funding.

Do not read the positive 6-minute walk, step-length, and functional ambulation results as proof that robot-assisted gait training is broadly superior to conventional training. The primary outcome of walking speed showed no significant difference (P=0.14), several other gait measures were null, and the subgroup finding that overground robots outperform treadmill-based ones rests on very few studies with high heterogeneity.