Following ethical approval (Sheffield Hallam University Ethics Board), 13 male rearfoot strikers were recruited for the study (30 ± 7 years; height 1.78 ± 0.08 m; mass 77 ± 10 kg). Only rearfoot strikers were recruited to ensure participants displayed similar loading characteristics [4]. All participants were running at least 16 km per week and free from any lower limb musculoskeletal injury at the time of testing. Two accelerometers were used to measure tibial acceleration: (1) A consumer wireless accelerometer (RunScribe version 1, Scribe Labs, California, USA) containing an MPU-9150 inertial measurement unit (Invensense, California comprising a tri-axial accelerometer, magnetometer and rate gyroscope) encased in a housing (total mass, 9.55 g) (Fig. 1a) and (2) a research grade uniaxial piezoresistive accelerometer (model 352C22, PCB Piezotronics, Stevenage, UK), considered a gold standard (reference sensor), and used in previous studies to measure tibial acceleration during running—with a between days reliability of 0.87 (ICC2,1) [5] (Fig. 1b). The reference sensor was attached to a small piece of thermoplastic (total mass, 1.65 g) (Fig. 1b); connected via a wire to a PCB signal conditioner mounted remotely (model 480E09; gain = 10). Although the sensor could not be attached to the bone and was still skin mounted, acceptable accuracy has been reported for skin mounted sensors with a mass of less than 3 g [7]. This supports the choice of the reference sensor in offering the most accurate means of measuring tibial acceleration in vitro. Both sensors were sampled at 1000 Hz.
The reference accelerometer was mounted using double-sided tape to the largest surface of the consumer accelerometer with the sensitive axes of both aligned visually (Fig. 2). The two adjoined sensors were attached to the distal antero-medial aspect of the tibia, 5 cm above the medial malleolus [4] using double-sided tape, ensuring that the sensitive axes of the sensors were aligned with the long axis of the tibia. This site was chosen due to the thin layer of soft tissue overlying the bone, thus reducing the effect of soft tissue oscillations generated during impact [8]. Tension was applied to the skin at the attachment site to help minimise soft tissue motion [6], and the sensors tightly over wrapped with elastic bandage about the circumference of the shank. To ensure consistency across participants, the same investigator applied the sensors on each occasion.
Following a warm up, participants ran at three different speeds on a treadmill in a randomised order categorised as low: (2.5 m s− 1), medium: (3.5 m s− 1), and high: (4.5 m s− 1). Participants were allowed to rest between trials to avoid fatigue. Each trial involved the participant running at the target speed for a total of 40 s, with 10 s to regulate running gait and then a 30 s data collection period. In a further set of trials, the reference sensor was mounted independently at the same location to assess the agreement of the sensor when attached directly to the skin and when mounted on the consumer accelerometer. All trials were completed in a single session and participants wore their own running shoes.
All data from both sensors were bandpass filtered between 2 and 75 Hz with a 2nd order Butterworth filter and converted to units of g using custom Matlab software (Mathworks, R2014a). Residual analysis of the data of ten participants across velocities determined the filter cut off choice [6]. The sensors were synchronised by way of the participant stamping their foot before starting a run so that the same series of foot strikes were analysed for the data from each sensor. Peak positive tibial accelerations—defined as the maximum value during stance—were calculated for each foot strike using both sensors to allow direct peak to peak comparisons across the 30 s run.
Intraclass correlation coefficients (ICC2,1) and 95% limits of agreement (LOA) were used to assess sensor agreement. An ICC value > 0.75 was considered good, whilst 0.4–0.75 was considered moderate [9]. Narrower confidence intervals (CI) of the LOA were considered an indication of the agreement between sensors with an acceptable level of agreement determined specifically for the application of developing real time feedback systems for runners. As the data were normally distributed, a paired samples t test was used to assess agreement of peak acceleration using the gold standard accelerometer when attached to the consumer accelerometer and when mounted independently (P < 0.05).