BackgroundThis study was a prospective, randomized, open, blinded-endpoint trial with the aim of examining whether gait training with a gait-assistance robot (GAR) improves gait disturbances in subacute non-ambulatory hemiplegic stroke patients more than overground conventional gait training. The GAR adopts a robot arm control system with full weight bearing and foot pressure visual biofeedback. MethodsTwenty-six hemiplegic patients were randomly assigned to either the GAR-assisted gait training (GAGT) group or the overground conventional gait training (OCGT) group. Both groups underwent 60min of standard physical therapy and 20min of GAGT or OCGT 5days a week for 4weeks. The primary outcome measure was the Functional Ambulation Classification (FAC). The secondary outcome measures were the peak torque of the extensor muscles in the lower extremities and a 10-m walking test. The lower extremity function was evaluated using the Fugl-Meyer Assessment, and activities of daily living were assessed using the Functional Independence Measure. ResultsThe GAGT group demonstrated significantly greater improvements in FAC and peak torque on the unaffected side (p=0.02) than the OCGT group. Additionally, gait speed tended to be faster (p=0.07) in the GAGT group. ConclusionsGAGT combined with standard physical therapy in subacute non-ambulatory hemiplegic patients led to significant improvements in gait and peak torque on the unaffected side compared to OCGT.