Davis et al. 2024, In-lab vs real-world running gait

Experimental wearable study in Sensors comparing a single in-lab run (treadmill, or a measured outdoor course) with five self-selected real-world runs, in two cohorts (49 and 19 runners), across speed, step length, vertical oscillation, stance time and leg stiffness. A person’s own in-lab data represented only 32.5% of their real-world data points multivariately; univariate overlap ranged from 65.7% (leg stiffness, the worst) to 95.2% (speed, the best). Pooling in-lab data across many runners raised representativeness to about 89–90%, and restricting to flat, straight terrain changed overlap by less than 1%. The lesson: one lab or treadmill assessment is a weak proxy for how an individual actually runs in the wild, especially for leg stiffness. Open access.