The study consisted of two parts. Part 1 was an intra- and inter-rater within- and between-day reliability study and part 2 was a concurrent validity study. The study population was a convenience sample and all participants were recruited from Aalborg University. The inclusion criteria for participants were: no previous or present injury or pain in the lower extremities or back, no medical or neurological conditions, and the ability to walk on a treadmill for a minimum of 20 minutes. Exclusion criteria were: < 18 years of age and BMI > 30. Furthermore participants were not included if they had a highly supinated foot posture and abnormal hypomobility of their midfoot kinematics identified after clinical examination. The study was approved by Aalborg University and conducted in accordance with the Helsinki Declaration and all participants were given written and verbal information about the project and signed an informed consent before participating. The reporting of the study follows the Guidelines for Reporting Reliability and Agreement Studies (GRRAS) [16].
Description of the stretch-sensor
The stretch-sensor is an elastic, flexible, and thin capacitive sensor (Figures 1 and 2). It consists of a stretchable active area that is 15 × 60 mm and a non-stretchable area at both ends that are each 15 × 10 mm, which serve to attach the stretch-sensor to the skin. A change in the stretch of the active area of the sensor causes a linear change in the electrical capacitance. Therefore changes in elongation can be calculated based on the change in the electrical capacitance of the sensor [15]. The thickness of the stretchable area is 0.40–0.60 mm and the thickness is below 1.5 mm in the non-stretchable area, which allows the stretch-sensor to measure in-shoe ND in conventional shoes. The capacitance of the stretch-sensor is measured 200 times per second (200 Hz). The signals from the stretch-sensor are sent to an input box that records the capacitance data on a SD card or transmits the data directly to the computer through a USB cable. Afterwards, the data were analysed using a custom-written script in Matlab [15]. We previously compared the amount of stretch from a calibration slate with the stretch measured from the stretch-sensor and found that the stretch-sensor was valid when compared to a calibration sled with R2 = 0.999 [15].
Attachment of the stretch-sensor to the foot
The stretch-sensor was placed between two points on the medial side of the foot. One attachment point was 20 mm posterior to the malleolus medialis and secured around both malleoli using a Velcro strap which ensured fixation of the sensor. The second attachment point was 20 mm posterior and 20 mm distal to the navicular tuberosity (Figure 1). The choice of attachment points was based on a pilot study in which we investigated different attachment points [15]. The prominence of the malleolus medialis did not allow us to position the distal part of the stretch-sensor directly onto the navicular bone because the stretch-sensor would interfere with the prominence of the malleolus medialis. Therefore, we choose to position it posterior and distal to the navicular tuberosity (Figure 1). Based on previous bone-pin studies by Wolf et al., it appears the entire medial midfoot moves in the same direction during walking [17]. Therefore, the position posterior and distal to the navicular tuberosity is likely a good proxy of ND [15, 18]. The attachment of the stretch-sensor took approximately 2 minutes per participant and each rater practiced the placement of the stretch-sensor on a minimum of 20 subjects before placing the stretch-sensor on the participants included in the study.
Part 1: reliability
Between-day intra-rater, and within- and between-day inter-rater reliability was based on measurements from 27 participants (12 women, 15 men, mean age of 26 years [age 22–39], mean BMI 22.6 [range 19.4–30.0]) recruited from Aalborg University. Rater 1 and rater 2 collected data independently and were blinded to the results from the other rater. In a randomised order, either rater 1 or rater 2 started by positioning the stretch-sensor on the medial side of the foot. Participants then walked without shoes for six minutes on a treadmill [18, 19], which was followed by 1.5 minutes of walking that was recorded using the stretch-sensor. The analysis was made on 10 consecutive steps identified after 30 seconds of recording. After this measurement, the stretch-sensor was repositioned by the other rater, and a second recording was made.
After the barefoot measurements when rater 1 had positioned the stretch-sensor, the procedure was repeated, but this time participants wore shoes while walking on the treadmill (Figure 3). This resulted in two measurements by rater 1 where participants walked with and without shoes while only one measurement was obtained by rater 2, where participants walked without shoes.
The following day, all participants returned and the procedure was repeated. All data were analysed by a third person who was not involved in the data collection.
Part 2: concurrent validity
To investigate the concurrent validity, the static ND was measured with the stretch-sensor and compared with the static ND as performed by Brody measured simultaneous. ND was defined as the change in the height of the naviculare tuberosity from a neutral position to a relaxed stance [12]. Subtalar neutral position was defined as the position where the talus could be palpated equally on the medial and lateral side of the foot [12]. The ND test as performed by Brody was chosen because it is one of the most commonly used clinical measurements of ND. Static ND was measured on 27 new participant (15 women and 12 men, mean age of 25 years [range 18–28], mean BMI 23.5 [range 19.7-30.0]).
The procedure for the measurement of ND was as follows: Rater 1 measured ND as performed by Brody with a ruler and then placed the stretch-sensor on the medial side of the participant’s foot. This was followed by a measurement of ND recorded by the stretch-sensor. Rater 2 repeated the same procedure, and the stretch-sensor was repositioned between the two procedures (Figure 3). Each rater was blinded to the results of the other rater’s assessments, and a third person registered the measurements of ND from the stretch-sensor, preventing raters 1 and 2 from seeing these results. To test the concurrent validity data from rater 1 was used.
Data analysis
To calculate ND for part 1 the data from the stretch-sensor was processed in the custom-written Matlab script called DataAnalyzer. Figure 4 shows the calculation of ND during five consecutive steps. Heel strike and the time point where the stretch-sensor was maximally stretched (which corresponds to the minimal height of the navicular during the stance phase), were marked in DataAnalyzer. ND was then described as the difference between the elongation of the stretch-sensor in the two positions. This approach was based on previous studies investigating dynamic ND which also defined ND as the difference in navicular height from heel strike to the minimal height of the navicular during the stance phase [18]. The calculations were based on 10 consecutive steps, which were identified 30 seconds into the recording. The procedure took approximately 3 minutes per participant. In part 2, the data from the stretch-sensor were collected and visualised directly by the program Datalogger, and ND was calculated as the change in the elongation of the stretch-sensor from subtalar neutral to relaxed stance.
Statistical analysis
In part 1, a two-way random effect model (2.1), single measures, absolute agreement, Intraclass Correlation Coefficients (ICC) were used to express intra- and interrater reliability. ICC > 0.75 was interpreted as acceptable reliability. Limits of Agreements (LoA) were used to express the agreement between the raters [20]. The LoA was calculated as the mean difference between raters ± 1.96 times the standard deviation of the differences between raters. The LoA was presented as a range indicating the maximal potential difference between the two raters in 95% of the ratings. Heteroscedasticity was visually assessed, and there were no trends towards heteroscedasticity. In part 2, a linear regression model was used to investigate the linear association between static ND as measured by the stretch-sensor and static ND as performed by Brody. All the statistical analyses were performed in SPSS 20.0.