Subjects
Between June and July 2013, healthy adult (> 18 years) subjects were recruited through advertisements placed in the offices of a private-sector physiotherapy practice in Sydney, Australia. Exclusion criteria included current arm, shoulder or hand pathology precluding pain-free active upper limb ROM. Subjects were screened using the QuickDASH, an 11-item questionnaire of arm, shoulder or hand pain/dysfunction, scored between 0 (no pain/dysfunction) and 100 (most severe pain/dysfunction) [11]. If prospective subjects reported a QuickDASH score > 0, the researcher questioned the subject further to ascertain current pain status with unloaded, active upper limb ROM. The study was approved by Western Sydney Local Health District’s Human Research Ethics Committee, and all subjects provided written informed consent before participation.
General demographic data, e.g. date of birth, were obtained directly from subjects. Height and weight were measured using a combined scale/stadiometer (Tanita WB-3000, Wedderburn, Sydney).
Study protocol
While standing erect, subjects had their arms positioned in a series of 16 positions for each of right then left arms, in the following order:
-
Flexion (anteversion): arm-by-side, 30°, 45°, 60°, 90°, 120°, 150°, maximal active ROM
-
Abduction: arm-by-side, 30°, 45°, 60°, 90°, 120°, 150°, maximal active ROM.
The scapula was not stabilised, i.e. the set angles of flexion/abduction were achieved using a subject-determined combination of glenohumeral and scapulothoracic motion.
Conventional goniometry (using a 360°, 20 cm clear plastic goniometer) was used to measure flexion/abduction angle in arm-by-side and maximal positions, and to position the arm in the requisite, or ‘set’ angle of flexion/abduction for set positions (“goniometer-set angle”, gΘ). Body landmarks used for goniometry were:
-
Flexion: Lateral aspect of acromion and lateral border of the humerus (forearm held in neutral pronation/supination).
-
Abduction: Anterior aspect of acromion and midline of humerus (forearm held in neutral pronation/supination).
A plumb line was used to provide a vertical reference line.
Once positioned, subjects maintained each position for 30 s by grasping an adjustable suction-cup/handle affixed to an adjacent wall. The handle was orientated vertically for flexion positions and horizontally for abduction positions to minimise subjects’ muscle activity. Subjects were able to rest between positions as required.
Subjects wore the SMA throughout positioning (on the right arm for right arm flexion/abduction, on the left arm for left arm flexion/abduction). The SMA has a ‘timestamp’ button, which marks its data output electronically, thereby allowing for delineation of data between positions. Once a subject’s arm position had been measured/set, the researcher pressed the timestamp button to commence data collection; 30 s later the researcher again pressed the timestamp button to end data collection.
The SenseWear Mini Armband
The SMA (BodyMedia Inc., Pittsburgh, USA) is a lightweight (50 g), wireless physical activity monitor worn on the upper arm. The SMA incorporates a manufacturer-calibrated, tri-axial accelerometer and physiological sensors, data from which are used to calculate step count and energy expenditure according to proprietary ‘black box’ algorithms. Data from the tri-axial accelerometer can be retrospectively downloaded and obtained independently of physiological data, using SenseWear Professional Software (version 7.0). Raw accelerometer data are initially obtained in the form ‘x, y, z’ , where −1 ≤ x, y, z ≤ 1, representing acceleration of the SMA in transverse (pitch), forward (roll), and longitudinal axes, relative to gravity.
When the SMA is stationary (e.g. is in a set position), accelerometer data represent the component of gravity acting on the SMA along each axis. Hence when the SMA is stationary and vertically upright, the longitudinal accelerometer (z) should theoretically read at 1 (the full component of gravity acting along the longitudinal axis), when stationary and horizontal, 0 (gravity acting perpendicular to the longitudinal axis), and when stationary and vertically upside down, −1 (the full component of gravity again acting along the longitudinal axis (see Figure 1). The manufacturer-provided algorithm for determining angular displacement of the SMA relative to gravity, i.e. from the vertical plane (elevation), is: elevation (“SMA-derived elevation angle”, a1Θ) = cos−1(z) [12].
In this study we obtained longitudinal accelerometer data (z) at a frequency of 1 Hz (i.e. we obtained 30 data points for each position for each subject). The 30 data points for each position were averaged to provide a single time-averaged data point. The time-averaged z data points were then transformed to angular elevation angle (a1Θ) using the manufacturer-provided algorithm.
The position of the SMA on the upper arm was standardised as follows: the subject was asked to perform resisted shoulder abduction in the functional position to allow for palpation of the deltoid tuberosity; marks were then placed on the subject’s arm at the location of the deltoid tuberosity and on the posterior aspect of the arm directly in the line of the deltoid tuberosity. The SMA was then centred vertically over the posterior marker using the manufacturer-supplied velcro armband.
Statistical analysis
The statistics package IBM SPSS Statistics Version 20 was used to analyse data. Two–tailed tests with a 5% significance level were used throughout. Simple descriptive statistics were used to summarise subject data and SMA-derived elevation angle (a1Θ) for goniometer-set shoulder positions. Differences between goniometer-set (gΘ) and SMA-derived elevation angle (a1Θ) were calculated for each of flexion and abduction, and ANOVAs performed to investigate if differences were associated with subject and study condition (left vs right arm, flexion vs abduction).
Bland-Altman plots were used to assess agreement between gΘ and a1Θ for each of arm flexion and abduction [13]. The Pearson product–moment correlation statistic was used to measure correlations between gΘ and a1Θ.
A priori sample size calculations suggested that 200 data points for each study condition, i.e. 25 subjects by 8 positions in each of left arm flexion, left arm abduction, right arm flexion and right arm abduction, would allow us to calculate confidence intervals ± 0.24 sd for the 95% limits of agreement between gΘ and a1Θ (where sd is the standard deviation of the differences between gΘ and a1Θ).