BackgroundAlthough gait speed is a widely used measure in older people, testing methods are highly variable. We conducted a systematic review to investigate the influence of testing procedures on resulting gait speed. MethodsWe followed the PRISMA checklist for this systematic review. Two independent reviewers screened Pubmed and Embase for publications on pairwise comparisons of testing procedures of usual gait speed. Descriptives were abstracted from the included publications using a predefined extraction tool by two independent reviewers. We defined the cut-off for the minimal clinically imporant diffence in gait speed as 0.1 m/sec. ResultsOf a total of 2109 records identified for screening, 29 reports on 53 pairwise comparisons were analyzed. The median (range) difference in gait speed for dynamic versus static start was 0.06 (-0.02 to 0.35) m/sec (14 reports); for longer versus shorter test distance 0.04 (-0.05 to 0.23) m/sec (14 reports); for automatic versus manual timing 0.00 (-0.05 to 0.07) m/sec (12 reports), for hard versus soft surfaces -0.11 (-0.18 to 0.08) m/sec (six reports), and electronic walkways versus usual walk test 0.04 (-0.08 to 0.14) m/sec (seven reports), respectively. No report compared the effect of finishing procedures. ConclusionsThe type of starting procedure, the length of the test distance, and the surface of the walkway may have a clinically relevant impact on measured gait speed. Manual timing resulted in statistically significant differences of measured gait speed as compared to automatic timing, but was below the level of clinical importance. These results emphasize that it is key to use a strictly standardized method for obtaining a reliable and valid measurement of gait speed.