**Background:** People with Parkinson's disease (PD) experience significant difficulties with dual-tasking due to executive dysfunction and loss of automaticity. Gait impairments during dual-task walking include decreased speed and step length, increased variability, and more freezing episodes. Previous reviews focused on single-task gait and balance effects, not dual-task performance itself. This systematic review aimed to establish evidence on whether motor–cognitive training improves dual-task performance in PD.
**Methods:** The review adhered to PRISMA guidelines and was preregistered (PROSPERO CRD42021278518). Searches were conducted up to September 28, 2021, in CINAHL, Web of Science, MEDLINE Ovid, and CENTRAL. Eligible studies were RCTs or quasi-RCTs in adults (≥18 years) with idiopathic PD, involving motor–cognitive training (≥2 sessions) with measurable dual-task outcomes. Two authors independently screened, extracted data, and assessed risk of bias using Cochrane RoB2. Meta-analyses used random-effects models with mean differences (MD) and 95% confidence intervals (CI). Certainty was assessed using GRADE.
**Key Results:** Eleven studies (17 reports, 597 randomized, 564 analyzed) were included. Pooled mean age was 68.9 years (SD 7.4), mean H&Y 2.0 (SD 0.4), mean disease duration 6.8 years (SD 4.8). Meta-analyses showed significant improvements favoring motor–cognitive training for: dual-task gait speed (8 studies, MD 0.12 m/s, 95% CI 0.08 to 0.17), dual-task cadence (5 studies, MD 2.91 steps/min, 95% CI 0.08 to 5.73), dual-task stride length (6 studies, MD 10.12 cm, 95% CI 4.86 to 15.38), and dual-task cost on gait speed (2 studies, MD −8.75%, 95% CI −14.57 to −2.92). No significant effects were found for stride length SD (p=0.42), stride time SD (p=0.26), double support (p=0.13), reaction time (p=0.47), or Timed Up and Go cognitive (p=0.17). Sensitivity analysis using fixed-effects models showed a significant effect on TUG cog (MD −2.16 s, 95% CI −4.05 to −0.27). Certainty of evidence was high for gait speed, cadence, stride length, and dual-task cost; moderate for double support; and low for reaction time and TUG cog.
**Clinical Implications:** This is the first systematic review demonstrating that motor–cognitive training improves dual-task gait performance in PD. The 0.12 m/s improvement in dual-task gait speed approaches clinically meaningful thresholds for single-task gait speed (0.06–0.22 m/s). Improvements in stride length (10.12 cm) and reduced dual-task cost (−8.75%) suggest meaningful functional gains. However, lack of effect on gait variability is concerning given its association with fall risk. Most participants had mild-to-moderate disease (mean H&Y 2.0), limiting generalizability to early or advanced stages. Responder analyses from included studies suggest those with higher baseline cognition and lower dual-task gait speed may benefit most. The review highlights critical gaps: poor reporting of cognitive task performance, lack of preregistered analysis plans, and insufficient description of cognitive status. Future research should identify optimal patient subgroups, training doses, and approaches.