Effects of Robot-Assisted Therapy for Upper Limb Rehabilitation After Stroke: An Umbrella Review of Systematic Reviews

BACKGROUND:Robotic rehabilitation, which provides a high-intensity, high-frequency therapy to improve neuroplasticity, is gaining traction. However, its effectiveness for upper extremity stroke rehabilitation remains uncertain. This study comprehensively reviewed meta-analyses on the effectiveness of upper extremity robot-assisted therapy in patients with stroke.METHODS:We combined results from 396 randomized controlled trials (RCTs) in 16 meta-analyses and conducted a new meta-analysis using nonoverlapping RCTs and 6 additional RCTs published after 2024. Duplicate studies were removed, all data were from RCTs, and a random-effects model resolved heterogeneity. Effects were analyzed by comparing robot-assisted therapy with conventional therapy at the same dose and as an add-on to conventional therapy.RESULTS:Compared with conventional therapy, the effect of robot-assisted therapy on the Fugl-Meyer assessment was summarized as a significant standardized mean difference (SMD) of 0.29 (95% CI, 0.14-0.44; number of individual RCTs reanalyzed, 100 RCTs), and the additional effect of robot-assisted therapy was an SMD of 0.42 (95% CI, 0.23-0.61; 16 RCTs). However, these Fugl-Meyer assessment improvements did not meet the minimum clinically important difference thresholds identified in previous studies: 12.4 for subacute and 3.5 for chronic stroke. For activities of daily living, only the additional effect was significant by SMD of 0.35 (95% CI, 0.17-0.54; 26 RCTs), muscle strength was significant by SMD of 0.46 (95% CI, 0.22-0.70; 31 RCTs), and spasticity was not significant by SMD of -0.25 (95% CI, -0.55 to 0.06; 25 RCTs).CONCLUSIONS:Robot-assisted therapy shows statistically significant improvements in motor recovery as measured by the Fugl-Meyer assessment in patients with stroke, both at the same dose and as an add-on to conventional therapy; however, these improvements do not meet the minimum clinically important difference. These benefits are consistent across different stages of stroke recovery, different types of robotic devices, duration of intervention, and training sites. However, the heterogeneity of included studies in patient population, stroke severity, intervention protocol, and robot type limits generalizability. High-quality trials are needed to better define the value of robot-assisted therapy across various devices and strategies.