Evaluation of the optimal number of steps to obtain reliable running spatio-temporal parameters and their variability

Spatio-temporal running parameters and their variability help to determine a runner's running style. However, determining whether a change is due to the measurement or to a specific condition such as an injury is a matter of debate, as no recommendation on the number of steps required to obtain reliable assessments exists.

What is the optimal number of steps required to measure different spatio-temporal parameters and study their variability at different running speeds?

Twenty-five runners performed three experimental sessions of three bouts of treadmill running at 8, 10 and 12 km/h separated by 24 h. We measured cadence, stride, step, contact and flight time. We calculated the duty factor and the leg stiffness index (Kleg). Mean spatio-temporal parameters and linear (coefficient of variation, standard deviation) and non-linear (Higuchi fractal index, α1 coefficient of detrended fluctuation analysis) analyses were computed for different numbers of steps. Relative reliability was determined using the intraclass coefficient correlation. The minimal number of steps which present a good reliability level was considered as the optimal number of steps for measurement. Absolute reliability was assessed by calculating minimal detectable change.

To assess the mean values of spatio-temporal running parameters, between 16 and 150 steps were required. We were unable to obtain an optimal number of steps for cadence, stride and step-time variabilities for all speeds. For the linear analyses, we deduced the optimal number of steps for Kleg and the contact time (around 350 steps). Non-linear analyses measurements required between 350 and 540 steps, depending on the parameter.

Researchers and clinicians should optimize experimental conditions (number of steps and running speed) depending on the parameter or the variability analysis targeted. Future studies must use absolute reliability metrics to report changes in response to a specific condition with no bias due to measurement error.