Effects of Physical Exercise Interventions on Dual-Task Gait Speed Following Stroke: A Systematic Review and Meta-Analysis.

OBJECTIVES: To estimate the treatment effects of exercise and/or gait training interventions on dual-task walking in people with stroke. The secondary objective was to conduct subgroup analyses to compare the treatment effects of interventions involving dual-task training to those without any dual-task training.

DATA SOURCES: A systematic search of the literature was conducted in 6 databases (PubMed, CINAHL, EMBASE, Web of Science, Cochrane Central Register of Controlled Trials, and Physiotherapy Evidence Database) up to July 18, 2017.

STUDY SELECTION: Randomized controlled trials (RCTs), nonrandomized controlled trials, or uncontrolled studies involving individuals with stroke and examining the effects of exercise and/or gait training interventions on dual-task gait speed.

DATA EXTRACTION: We extracted data on participant characteristics, intervention duration, frequency, and type; pre and post gait speed and secondary nongait task performance for single and dual-task conditions, types of tasks used for dual-task assessment and dual-task prioritization instructions.

DATA SYNTHESIS: Of 313 articles identified, 7 studies involving 12 independent treatment arms (n=124) met the inclusion criteria. There was a significant pre-post intervention increase in dual-task gait speed (MD: 0.03m/s, 95% CI: 0.01, 0.06) and single-task gait speed (MD: 0.06m/s, 95% CI: 0.03, 0.09). Dual-task training tended to have a larger effect on dual-task gait speed than interventions without dual-task training. Between-group analysis of three RCTs found evidence of superiority of dual-task gait training over single-task gait training for improving dual-task gait speed (MD: 0.08m/s, 95% CI: 0.02, 0.14).

CONCLUSIONS: Exercise and gait training interventions, especially those involving dual-task practice, may improve dual-task gait speed after stroke, but the clinical significance is unclear. Current effect size estimates lack precision due to small sample sizes of existing studies.