Research Design and Subjects
In three trials distributed over two days within a two-week period, neck muscle EMG-force relationships were examined across seven submaximal contraction intensities during sustained isometric constant-force contractions for neck flexion and neck extension. Following approval by the Regional Committee for Medical Research Ethics, a convenience sample of eighteen participants was recruited through personal communication. Inclusion criteria were being healthy and aged 18–65 years; exclusion criteria were neck or shoulder pain within the last three months and known patch allergy. Nine males and nine females with a mean (SD; min-max) age of 29 (7; 20–48) years, height of 1.70 (0.08; 1.53–1.81) m, and weight of 66 (10; 45–84) kg enrolled for participation and signed a written informed consent. Eleven participants reported that they had trained for muscle strength or cardiovascular fitness more than three times per week over the past 12 months. All participants were requested to refrain from alcohol and physical training of the neck and shoulders for 72 h preceding measurements.
Protocol
EMG activity was monitored bilaterally in three areas (Fig. 1): the anterior neck (AN) – centered on the belly of the sternocleidomastoid muscle, approximately one third of the length above the sternal attachment [21], the upper posterior neck (UPN) – centered over the most superficial splenius capitis muscle area, between the sternocleidomastoid and upper trapezius muscles [19–21], and the lower posterior neck (LPN) – over the lower semispinalis capitis, 20 mm lateral to the median line in level with the C7 spinous process [20]. Reference electrodes were placed over sternum and spinous processes. In accordance with current recommendations [23], electrode placement areas were meticulously shaved, abraded, and disinfected with alcohol. Disposable, self-adhesive, pre-gelled Ag/AgCl electrodes with a conductive diameter of 10 mm (Blue Sensor N-00-S, Medicotest A/S, Ølstykke, Denmark) were attached pairwise with an inter-electrode distance of 20 mm, and secured with surgical tape. Skin potential offset was allowed to stabilize for 20 min preceding measurements, after which electrode-skin impedance was consistently below 5 kΩ. To obtain temporally synchronized data, differentially amplified EMG (bandwidth 8–500 Hz, input impedance = 10 GΩ, CMRR = 110 dB at 50 Hz, input noise <1.6 μVRMS, gain = 305) and force signals (TB5 tension load cell, Lahti Precision Ltd, Lahti, Finland; maximal load = 500 kg, error <0.02%, sensitivity <0.1%) were digitized by a 14-bit A/D converter with a sampling frequency of 1000 Hz and stored in an integrated system (Biomonitor ME6000, Mega Electronics Ltd, Kuopio, Finland).
Subjects were seated in a standardized position on a fixed, height-adjustable stool, with a rigid square block stabilizing the torso in a neutral spine posture, and the extremities placed to prevent them from aiding during trials (Fig. 2a). A rigid strap connected to a fixed force transducer was placed over the supraorbital ridge and reference values were acquired through three maximal voluntary contractions for each contraction direction, during which subjects pushed their head against the strap. Subjects were carefully instructed as to the correct technique. To obtain accurate reference values, maximal contractions were rehearsed before trials, verbal encouragement and visual feedback of the force output was provided during trials, and the highest 1-s average of the highest contraction was selected. Reference values were used to calculate submaximal target contraction intensities equivalent to 5, 10, 20, 30, 50, 70, and 90% of maximal force.
Following a 30-min rest, participants met target contraction intensities aided by live visual feedback of produced force. When subjects did not reach target intensities, they were permitted an additional trial. Software malfunction resulted in two cases failing to meet this criterion: one being 18% instead of 20%, and one 26% instead of 30%; both included in the analyses below. Total time for the submaximal protocol was 40 min, followed by a one-hour break with electrodes retained, after which the submaximal contraction procedure was repeated. The same 40-min sampling procedure was repeated again 5–14 days after the first day, with identical preparatory steps. To avoid systematic bias due to muscular fatigue, contraction direction and intensities were conducted in random order, sustained for approximately five seconds, and separated by 2-min rest periods. Data sampling took place in a university motion laboratory, and was carried out by a trained physiotherapist with prior experience of the protocol.
Data Management and Statistics
Data management was conducted in MATLAB (v8.1, The MathWorks Inc., Natick, MA, USA). Unprocessed signals were visually inspected in time and frequency domains. EMG and force signals were then respectively band-passed at 20–400 Hz and low-passed at 25 Hz, with a fourth-order zero phase Butterworth filter. Next, EMG signals were full-wave rectified, and both EMG and force signals were smoothed by a centered 1-s moving average with a 1-ms lag. The signal-to-noise ratio was calculated as the ratio of desired EMG signal power to the noise signal power obtained during rest to provide a measure of EMG signal quality. Automated functions were employed to extract submaximal values from the smoothed data by selecting the most stable second of force data within ±5% of the target intensity, followed by extraction of the coinciding EMG amplitude. All selected values were visually controlled to confirm their adequacy. Torques were calculated for neck flexion and extension from force data, and for gravity on the mass of the head and neck [20], using a constant of 7.9% of body weight [3]. Gravitational torque was then subtracted from neck flexion and added to neck extension. The torque calculation details are provided in Fig. 2b. Because target contraction intensities did not account for gravity, data below 20% of maximal target force in some cases (n = 16 for 5%, n = 6 for 10%, and n = 1 for 20% of target intensity) corresponded to neck flexion torques below zero. These values were discarded in the following analyses, since it was uncertain whether they were related to the examined exertion or to gravity alone. Submaximal force and EMG values were normalized as the percentage of maximal force (% MVC), and maximal voluntary excitation (% MVE), respectively. Finally, to decrease random bias related to measurement error [24], EMG data averaged between both muscle sides and trials were used for EMG-force relationship analyses, whereas EMG data averaged between muscle sides were used for the reliability analysis.
Statistical analyses were conducted in R (v3.1.3, R Core Team 2015, R Foundation for Statistical Computing, Vienna, Austria). EMG-force relationships were visually inspected and modeled with general linear mixed-effects regression via a top-down strategy [25]: following selection of fixed effects, random effects were iterated one by one, and last, fixed effects were reduced until the most parsimonious model was identified. Maximum likelihood estimation was used to fit models in which the number of fixed effects differed, while restricted maximum likelihood estimation was used to fit models in which random effects differed. Akaike’s information criterion and Schwarz’s Bayesian information criterion aided in the model selection [26], and the Pratt-adjusted coefficient of determination [27, 28] and the root mean square error [29] were used as measures of the models’ fit to the data. Statistical assumptions were assessed visually [25]. The influence of antagonistic muscle activity was tested for all sampling areas.
Test-retest reliability of individual subjects’ modeled EMG-force relationships were examined within and between days across sampling areas. The final EMG-force relationship models were refitted for each of the three trials, and the participant-specific regression lines were extracted from the refitted models and compared between trials per percentile across the contraction intensity range. Changes in the mean were presented graphically. An intraclass correlation coefficient (ICC) based on a single measurement two-way random effects model of absolute agreement was utilized to examine test-retest correlation [30, 31], and the standard error of measurement (SEM; i.e., the within-subject standard deviation between trials) [31] was used to investigate within-subject variation. Alpha was set at 0.05 for all analyses.