Efficacy of Robot-assisted Training on Upper Limb Motor Function After Stroke: A Systematic Review and Network Meta-analysis.

The included trials varied widely in patient baseline, robot type, dose, and duration, and heterogeneity was high for the main outcomes. The funnel plots showed asymmetry for the primary outcomes, so publication bias may have inflated the apparent benefit. GRADE rated the network meta-analysis evidence as intermediate or low, so the findings should not be treated as definitive. Some trials did not report randomization methods clearly, and blinding or dropout reporting was incomplete in several studies. Only one trial reported adverse events, so safety is not well characterized. The network analysis had no indirect comparisons between the three interventions, so the ranking of robot types should be read cautiously. The trials included subacute or chronic stroke patients, so it says little about very early acute stroke. Many outcome measures were subjective scales, and some secondary outcomes had few trials or small samples. The trials compared conventional rehabilitation plus robot-assisted training with conventional rehabilitation alone, so the robot component cannot be separated from the conventional rehabilitation component.

Do not conclude that robot-assisted training is proven better for every stroke patient or that one robot type is clearly superior. The evidence was intermediate to low, heterogeneity was high, publication bias was possible, and the network analysis did not find significant differences between robot types.