Are impact accelerations during treadmill running representative of those produced overground?

Although many runners train overground, measuring impact accelerations on a treadmill may be advantageous for researchers and clinicians. Previous investigations of peak and rate of acceleration (peak_accel, rate_accel) during treadmill running compared to overground running have not examined both the relative consistency and absolute agreement of these measures, or the effect of treadmill stiffness.

(1) Are peak_accel and rate_accel produced during running on a stiff and less stiff treadmill 'representative' of those produced during overground running? (2) Are peak_accel and rate_accel measured on treadmills of different stiffness 'representative' of each other?

Eighteen participants ran at a self-selected pace on three surfaces: Treadmill 1 (reduced stiffness), Treadmill 2 (increased stiffness) and overground on asphalt, whilst peak_accel and rate_accel were recorded at the shank and lower back. Relative consistency (ICC (3,1)), absolute agreement (Bland-Altman analysis) and systematic differences (ANOVA/Friedman's Tests) were assessed.

ICCs revealed moderate to excellent relative consistency in peak_accel and rate_accel between surfaces, with higher consistency for measures at the lower back. Absolute agreement was low, with the Bland Altman limits of agreement exceeding the clinical acceptable range for all comparisons. For systematic differences in means, peak_accel and rate_accel at the shank were significantly higher overground than on either treadmill; with no difference evident at the lower back. No differences were found for surface with respect to shank or lower back peak_accel and rate_accel between treadmills.

Moderate to excellent relative consistency of peak_accel and rate_accel between the surfaces suggests that using different surfaces in research involving rank ordering of participants by acceleration magnitude may be acceptable (e.g. prospective studies examining if impact accelerations are related to injury). However, low absolute agreement indicates that data collected on treadmills of different stiffness and overground should not be used interchangeably (e.g. running-retraining studies).