Low Accelerometer Sampling Rates Attenuate Tibial Impact Acceleration Peaks During Running

Abstract

High tibial acceleration peaks have been associated with tibial stress fracture in runners. Field-testing with wearable wireless accelerometers in now commonplace, but some devices have a lower sampling frequency than in laboratory testing. PURPOSE: To determine the influence of sampling rate on peak axial tibial acceleration and peak resultant tibial acceleration magnitude during running. METHODS: As part of a larger study, 19 healthy adults were recruited (10 women; 31±6 years; 1.70±0.08 m; 68.6±11.6 kg) and provided informed consent to participate. A precision accelerometer sampling at 1000Hz was attached to the distal anteromedial aspect of the right tibia. Participants ran at 3.0m/s in the laboratory for five good trials making contact with a force plate sampling at 1000Hz. Raw data were down-sampled to 500Hz and 100Hz, common sampling rates for wearable wireless accelerometers. All data were low-pass filtered at 70Hz. Stance phase was identified by foot contact on the force plate. Peak positive axial acceleration and peak resultant acceleration were determined for each trial and averaged. One factor repeated measures analysis of variance with least significant difference post-hoc tests determined whether peaks differed among sampling rates. Effect sizes were calculated to aid interpretation of the data. RESULTS: Both peak axial and peak resultant acceleration were significantly reduced when tibial acceleration was sampled at 100Hz compared to 500Hz or 1000Hz (p = 0.041). Reductions were 0.7g or about 10% of the peak magnitude at 1000Hz, a small effect. Values were stable between the 500Hz and 1000Hz sampling rates. CONCLUSION: We recommend that tibial acceleration data are sampled at rates of 500Hz or greater to avoid attenuation of peaks producing erroneously low values for both peak axial and peak resultant acceleration.Table: Mean and standard deviation of peak axial and peak resultant tibial acceleration during running when sampling data at different rates