Reliability of electromyography during 2000 m rowing ergometry

Twelve, well-trained, male competitive club rowers volunteered to take part in the study (mean ± standard deviation; age 23.6 ± 5.1 years, stature 1.86 ± 0.05 m, body mass 86.5 ± 8.4 kg, 2000 m rowing ergometer time 6:33 ± 0:09 min:s, and rowing experience of 7.2 ± 5.4 years). The cohorts of participants were featured in a previous study (same data collection occasions) which detailed consistency of pacing of the repeated 2000 m sessions (S1, S2, S3) in relation to power output and metabolic responses [19]. All participants had competed at national level events, such as the ‘Head of the River Race’, the ‘Henley Royal Regatta’, the ‘National Rowing Championships of Great Britain’, or the ‘British Universities and Colleges Sports Rowing Championships’. Additionally, participants had extensive prior experience of performing 2000 m rowing ergometer tests prior to their involvement in the study. Participants were informed of the experimental procedures and any potential risks involved and provided written informed consent to participate in the study. The study was approved by the local ethics committee in-line with the Helsinki Declarations for research with human volunteers.

The study followed a repeated measures design to determine the reliability of surface EMG during 2000 m rowing ergometry in trained rowers. Each participant performed three repeat laboratory testing sessions (S1, S2, S3) interspersed by 3–7 days between each session. Participants were asked to arrive at the laboratory in a hydrated state, which was confirmed upon arrival. In addition, participants were required to abstain from exercise on the day of testing and strength training in the 72 h before testing. Electrodes were attached to seven muscle sites for rowing specific EMG analysis [11‐14] in accordance with procedures as described below within Table 1. Following this, participants completed a five-minute self-paced warm-up on a rowing ergometer, followed by a two-minute rest, before performing five maximal rowing power strokes on the rowing ergometer. After a subsequent five-minute rest period, the participants were then instructed to warm-up for a further five minutes on the rowing ergometer after, which they performed the 2000 m time-trial test.

Maximal stroke power, as well as the 2000 m test, was assessed using an air-braked rowing ergometer (Concept 2 Model C, Concept 2 Ltd, Wilford, Notts, UK) with a drag factor set at 140 (in accordance with the British International Rowing guidelines for ergometer testing) [16]. Following the five-minute self-paced warm-up, participants were provided a short break of two minutes and then returned to the ergometer to initially row sub-maximally for one minute, at which point they were instructed to perform two ‘build up’ strokes, followed by the first of five consecutive maximal effort power strokes. All participants were required to hold a rate of 30 s.min⁻¹ during the power strokes, as described previously [17]. For each testing session, the mean rectified EMG recorded during the power strokes was used to normalize the mean rectified EMG recorded during the 2000 m test as subsequently detailed. This method is in accordance with the ‘sprint method’ of EMG normalization, which has seen to produce lower coefficient of variation values during subsequent repeated sessions in running and cycling compared to traditional MVC normalization [3, 4].

Following the power stroke test, participants were provided with a five-minute break and then returned to the ergometer. Participants performed a five-minute self-paced warm-up on the rowing ergometer before the initiation of each 2000 m test. During each session, the only feedback available to participants was the stroke rate and distance remaining [20]. Surface EMG was recorded from seven muscle sites; gastrocnemius, biceps femoris, gluteus maximus, erector spinae, vastus medialis, rectus abdominis and latissimus dorsi respectively (Table 1), and collected during 2000 m test at each visit. These muscles have been shown to contribute importantly to rowing performance [2, 11, 13, 14]. Surface EMG was collected at a sampling frequency of 1000 Hz and amplified (1000x) using a 16-channel wireless telemetric system (Myon RFTD-E16, Myon AG, Baar, Switzerland) interfaced with a multifunction data acquisition module (USB-6210, National instruments, Austin, Texas, USA). Site preparation and electrode placement were performed in accordance with the SENIAM guidelines [23] with exception of the rectus abdominis and latissimus dorsi for which procedures described by Ng et al. [24] and Horsley et al. [25] respectively were adopted. For each site, reduction in skin impedance was achieved prior to the attachment of the electrodes by shaving and cleaning the skin surface superior to the muscle belly with alcohol, followed by skin abrasion with a paper towel and then application of a standard electrode gel constituent [26]. Data were recorded within commercially available software (MyoResearch XP, Noraxon, Scottsdale, Arizona, USA) prior to being exported for analysis in LabChart (LabChart 7, AD Instruments, Oxford, UK). Once exported, the raw EMG data were high-pass-filtered with a cut-off frequency of 15 Hz, and the filtered data were fully rectified. On each testing occasion, mean rectified EMG recorded during each 500 m segment of the 2000 m test was normalized using the mean rectified EMG recorded during the power strokes. This was subsequently expressed as a percentage factor in relation to 1.0 (i.e., 1.0 = 100% of power strokes mean rectified EMG). EMG data were successfully collected from between 10 and 12 participants per muscle site per 2000 m session. Where 10 or 11 participants are listed in Tables 2‐4, errors in data collection or data outliers were detected for 1 to 2 involved participants not included. This was due to electrodes being displaced during the 2000 m rowing test. Participants were identified as outliers using the ‘mean ± three standard deviations’ method [27]. Specifically, if a participant’s mean EMG value over a 2000 m session was lesser or greater than the cohort mean EMG by three standard deviations, then the participant’s data were excluded from the sample for that particular muscle.

Descriptive statistics are presented as mean (± standard deviation) unless otherwise stated. Statistical analyses were conducted using SPSS 25.0 (Chicago, IL, USA) with an alpha level set at p < 0.05. A one-way ANOVA was used to investigate between-sessions differences in 2000 m mean EMG for each of the seven muscle sites. Between-sessions (S1 to S2, S2 to S3) reliability and repeatability of EMG for the 2000 m test was measured by calculating coefficient of variation (CV) and intra-class correlation coefficient (ICC), both expressed with ± 90% confidence intervals [28]. CV descriptor categories for interpreting EMG reliability were adopted as defined by Kordi et al. [5] (‘good’ < 10%, ‘acceptable’ 10.0–19.9%, ‘weak’ 20.0–29.9% and ‘very weak’ ≥ 30.0%). Previously implicated correlation interpretations were adopted for ICC as follows; ‘trivial’ < 0.1, ‘small’ 0.1–0.29, ‘moderate’ 0.3–0.49, ‘large’ 0.5–0.69, ‘very large’ 0.7–0.89, ‘almost perfect’ 0.9–1.0 [5, 7]. The EMG data during the repeated sessions were further divided into four × 500 m segments (0–500 m, 500–1000 m, 1000–1500 m, 1500–2000 m). Subsequently, therefore, a three x four (session x segment) repeated measures ANOVA was used to investigate differences in each respective EMG output. Assumptions of sphericity were assessed using Mauchly’s test of sphericity, with any violations adjusted by use of the Greenhouse–Geisser (GG) correction. If a significant main effect across time was shown, then post hoc differences across sessions were analyzed with use of the Bonferroni’s correction.

There were no significant differences identified in maximal power output for the power stroke test across sessions (S1: 529 ± 44 W, S2: 525 ± 45 W, S3: 524 ± 42 W) (F_2,11 = 0.049, p = 0.953). Reliability across power stroke sessions was high as indicated by mean CV (3.1%) and ‘very large’ mean ICC (r = 0.87). Reliability of EMG varied considerably across muscles with rectus abdominis, biceps femoris and erector spinae having ‘acceptable’ mean CV (11.3–18.8%), whereas for gluteus maximus, vastus medialis, gastrocnemius and latissimus dorsi mean CV ranged from 28.2 to 41.6% (Table 2).

As we previously published elsewhere, there were no significant differences between-sessions for 2000 m performance time mean power or stroke rate [19]. There were no significant differences in mean rectified EMG muscle activation between-sessions for any of the seven muscles; gastrocnemius (F_2,9 = 0.460, p = 0.636), biceps femoris (F_2,11 = 0.039, p = 0.962), gluteus maximus (F_2,9 = 0.476, p = 0.627), erector spinae (F_2,9 = 0.022, p = 0.979), vastus medialis (F_2,9 = 2.068, p = 0.146), rectus abdominis (F_2,11 = 1.762, p = 0.188) and latissimus dorsi (F_2,10 = 1.116, p = 0.341) (Table 3). Mean CV for six out of the seven EMG muscle sites was ‘acceptable’ (12.3–18.6%); however, for gastrocnemius, CV was classed as ‘weak’ (28.6%). Mean ICC values across EMG muscle sites were ‘very large’ (r = 0.75–0.89) for erector spinae, gluteus maximus and rectus abdominis, ‘large’ (r = 0.57–0.63) for vastus medialis and biceps femoris, and moderate (r = 0.31–0.35) for gastrocnemius and latissimus dorsi.

The CV for mean rectified EMG across the seven muscles for each 500 m segment of the 2000 m test is shown in Table 4. Latissimus dorsi displayed the lowest mean CV for rectified EMG across each 500 m segment of the 2000 m time session (12.4–15.4%) and gastrocnemius showed the highest (27.7–31.4%). There were significant within-session differences across session segments for biceps femoris EMG (F(GG)_2,11 = 4.921, p = 0.029) whereby activation was significantly greater during 1500–2000 m (0.72 ± 0.14) in comparison to 1000–1500 m (0.65 ± 0.11) (p = 0.009). For gluteus maximus (F(GG)_2,9 = 13.276, p = 0.001), activation was significantly greater during 1500–2000 m (0.59 ± 0.16) in comparison to 500–1000 m (0.53 ± 0.16) (p = 0.002), and 1000–1500 m (0.53 ± 0.16) (p = 0.001). For vastus medialis (F(GG)_2,9 = 11.985, p = 0.002), activation was significantly greater during 0–500 m (0.57 ± 0.09) than 500–1000 m (0.49 ± 0.09) (p = 0.002) and 1000–1500 m (0.49 ± 0.10) (p = 0.013). Vastus medialis activation was also significantly greater during 1500–2000 m (0.57 ± 0.14) compared to 500–1000 m (p = 0.023) and 1000–1500 m (p = 0.006). In relation to mean values for all seven muscles across segments of the 2000 m test, the pattern was for a decrease in EMG amplitude during 500–1000 m and 1000–1500 m compared to 0–500 m and 1500–2000 m (Fig. 1). There were no between-sessions 500 m segment differences in EMG at any of the muscle sites (p > 0.05).