Purpose: To evaluate the reliability and validity of the CSA (model 7164) accelerometer (MTI) in a wide walking-running speed range in laboratory and field.
Method: Twelve male subjects performed three treadmill walking/running sessions and one field trial with the same continuous protocol involving progressively increasing velocities at 5 min per interval from 3 to 6 km x h(-1) (walking) and 8 to 20 km x h(-1) (running). In the field trial, this protocol was terminated after 35 min (14 km.h(-1)), but the trial then extended with 5-km running at a freely chosen velocity. In both scenarios, two CSAs were mounted on each hip and the step frequency measured at each velocity. Oxygen uptake VO(2) x kg(-1) was measured on the last two treadmill sessions. Correlation analyses were performed for mean CSA output relationship with speed, VO(2) per kilogram, and step frequency.
Results: In all trials, CSA output rose linearly (R2 = 0.92, P < 0.001) with increasing speed until 9 km.h-1 but remained at approximately 10000 counts.min-1 during running, thus underestimating VO(2) per kilogram at speeds > 9 km x h(-1). Estimation errors increased with speed from 11% (P < 0.01) at 10 km x h(-1) to 48% (P < 0.001) at 16 km x h(-1), when assuming a linear relationship. Freely chosen velocities in the field trial ranged from 10.9 to 16.3 km.h-1. No difference in the CSA-speed relationship was observed between the two scenarios. Differences in CSA output between subjects could partially be attributed to differences in step frequency (R = -0.34 (P = 0.02) for walking and R = -0.63 (P < 0.001) for running).
Conclusion: CSA output increases linearly with speed in the walking range but not in running, presumably due to relatively constant vertical acceleration in running. Between-subject reliability was related to step frequency because CSA data are filtered most at higher movement frequencies. Epidemiological CSA data should thus be interpreted with these limitations in mind.