Synchronized Dual-Task Training During Robot-Assisted Gait Training After Stroke: A Three-Arm Randomized Controlled Trial

Background Robot-assisted gait training (RAGT) and dual-task training (DTT) each improve walking after stroke, but whether delivering dual tasks synchronously during RAGT confers additive benefit is unknown. We compared synchronized robot–dual-task training (RDT) against each component alone. Methods In this single-center, three-arm, assessor- and statistician-blinded randomized controlled trial, 99 patients with post-stroke walking dysfunction were allocated to RDT (n = 33), RAGT (n = 33), or DTT (n = 33) (after removal of one duplicate data-entry record), each added to conventional rehabilitation (20 min/day, 5 days/week, 8 weeks). The primary outcome was the Fugl-Meyer Assessment of the Lower Extremity (FMA-LE); secondary outcomes were the Berg Balance Scale (BBS), activities of daily living (ADL; Modified Barthel Index), and six-minute walk test (6MWT) categories. Between-group change scores were compared by ANCOVA adjusted for baseline values, with pairwise post hoc comparisons. Results All groups improved significantly on all outcomes (all within-group P < 0.001). FMA-LE gains were greatest with RDT (Δ14.27 ± 4.99) versus RAGT (Δ8.55 ± 2.72; adjusted mean difference 4.87, 95% CI 3.39–6.35; P = 1.1×10⁻⁸; d = 1.43) and DTT (Δ6.39 ± 2.84; adjusted mean difference 5.99, 95% CI 4.38–7.61; P = 3.8×10⁻¹⁰; d = 1.94); RAGT also exceeded DTT (adjusted mean difference 1.54, 95% CI 0.40–2.69; P = 0.009). BBS and ADL improved most with RDT (Δ21.67 and Δ31.52, respectively); all nine pairwise comparisons across outcomes were statistically significant, including ADL RDT versus RAGT (adjusted mean difference 5.33, 95% CI 1.03–9.62; P = 0.016, not surviving Bonferroni correction). Week-8 6MWT categories favored RDT (P = 0.006 vs RAGT; P < 0.001 vs DTT); 66.7% of RDT participants walked > 300 m, and none were unable to complete the test. Findings were robust in baseline-stratified analyses, in a sensitivity analysis excluding 13 over-age participants, and in an ordinal logistic regression of 6MWT categories adjusted for baseline grade. Conclusions In this completed three-arm randomized trial, synchronized dual-task training during RAGT was associated with greater improvements than dual-task training alone across all evaluated outcomes—lower-limb motor function, balance, activities of daily living, and walking endurance—and with greater improvements in FMA-LE, BBS, and ADL than robot training alone, although the ADL contrast versus RAGT was modest and did not survive Bonferroni correction. Diffusion tensor imaging correlates of these clinical gains and longer-term follow-up will be reported separately. Trial registration Chinese Clinical Trial Registry (ChiCTR), http//www.chictr.org.cn/ registration number ChiCTR2500115695 date of registration 29 December 2025. National Medical Research Registration and Filing Information System record MR5025090484.