Upper trapezius muscle activity in healthy office workers: Reliability and sensitivity of occupational exposure measures to differences in sex and hand dominance

Abstract

Patterns of cervical muscle activity may contribute to overuse injuries in office workers. The purpose of this investigation was to characterize patterns of upper trapezius muscle activity in pain-free office workers using traditional occupational exposure measures and a modified Active Amplitude Probability Distribution Function (APDF), which considers only periods of active muscle contraction. Bilateral trapezius muscle activity was recorded in 77 pain-free office workers for 1–2 full days in their natural work environment. Mean amplitude, gap frequency, muscular rest, and Traditional and Active APDF amplitudes were calculated. All measures demonstrated fair to substantial reliability. Dominant muscles demonstrated higher amplitudes of activity and less muscular rest compared to non-dominant, and women demonstrated less muscular rest with no significant difference in amplitude assessed by Active APDF compared to men. These findings provide normative data to identify atypical motor patterns that may contribute to persistence or recurrence of neck pain in office workers.

Practitioner Summary

Upper trapezius muscle activity was characterized in a large cohort of pain-free workers using electromyographic recordings from office environments. Dominant muscles demonstrated higher activity and less rest than non-dominant, and women demonstrated less rest than men. Results may be used to identify atypical trapezius muscle activity in office workers.

Introduction

Neck pain, defined as pain in an anatomical location between the superior nuchal line and the spine of the scapula (Guzman et al. 2009), is a significant health problem, ranking as the fourth greatest contributor to global disability (Hoy et al. 2014) with 6–12 month incidence rates ranging from 6% – 17.4% in the general working population (Côté et al. 2009). Compared to other occupational groups, office workers experience particularly high incidence rates ranging from 15.4% - 34.4% annually (Côté et al. 2009). The high incidence of neck pain in office workers, despite the relatively low physical demand of this occupation, has led researchers to investigate the potential contribution of sustained, low amplitude cervical muscle activity to muscle overuse injuries (Sjøgaard, Lundberg, and Kadefors 2000, Hägg 2000) and the development of pain in this population (Blangsted, Hansen, and Jensen 2003, Dennerlein and Johnson 2006, Richter et al. 2009, Szeto, Straker, and O’Sullivan 2009).

Trapezius myalgia is present in one third of office workers with chronic neck pain (Sjøgaard et al. 2006); therefore, the trapezius muscle is a common target of electromyographic (EMG) studies of physical workload (Hansson et al. 2000, Nordander et al. 2000, Østensvik, Veiersted, and Nilsen 2009a). Several methods have been proposed for assessing patterns of muscle activation from prolonged (i.e., whole work day) EMG recordings and their relation to musculoskeletal pain. Calculating the total duration of continuous periods during which the amplitude of the EMG signal remains below a predefined level of muscular rest is often used to measure temporal aspects of muscle rest (Hansson et al. 2000, Veiersted, Westgaard, and Andersen 1993, Veiersted et al. 2013), but does not provide information on the composite magnitude of physical exposures (i.e. EMG amplitude). For this purpose, the Amplitude Probability Distribution Function (APDF) can be used to evaluate static, median, and peak amplitudes of muscle activity that occur over prolonged recording periods (Jonsson 1982, Hansson et al. 2000, Nordander et al. 2000). Static muscle activity is defined as the EMG amplitude exceeded for 90% of the sampled workday (calculated as the 10^th percentile of the APDF). Static muscle activity is thought to represent the level of sustained, low-amplitude muscle activity required for biomechanical functions such as maintaining postural control. Median and peak values are calculated as the 50th and 90th percentiles of the APDF, respectively, and are thought to represent briefer and less frequent bouts of higher amplitude muscle activity required for more dynamic tasks (Jonsson 1982, Hansson et al. 2000).

APDF values are traditionally calculated using the EMG signal from the entire recording period, including periods when EMG amplitude is below the defined level of muscular rest, to estimate the composite load placed on the muscle throughout the workday (referred to hereafter as the Traditional APDF). This measure was originally developed to estimate muscle strain for occupational tasks with limited amounts of muscular rest (Jonsson 1982). According to Jonsson (1982), APDF percentiles provide relevant information about the amplitude of muscle contraction only when muscular rest remains below 10% of the total recording time. With greater periods of rest, the APDF is increasingly influenced by low values of resting EMG that reduce the amplitude of static, median, and peak EMG in a non-linear fashion due to the logarithmic shape of the function.

Interestingly, several studies of office workers have reported large (>10%) amounts of muscular rest during prolonged recordings of the trapezius muscle (Aarås et al. 1996, Blangsted, Hansen, and Jensen 2003, Nordander et al. 2000), which might be expected to reduce estimates of static, median, and peak muscle activity. Although such studies provide valuable information on composite loads placed on the muscle throughout the workday, the inclusion of large amounts of muscular rest may decrease the sensitivity of the Traditional APDF to differences in the amplitude of static and dynamic loads placed on the muscle during periods of active contraction. Thus, it may be informative to consider the amplitude of active muscle contraction independently from the amount of muscle rest when assessing physical risk factors for musculoskeletal pain in sedentary occupations such as office work. For example, this approach could help identify whether the magnitude of active muscle contraction required to perform job tasks or the duration of rest breaks throughout the workday should be prioritized by injury prevention programs. The current study proposes an Active APDF methodology in which periods of muscular rest are removed prior to constructing the amplitude probability distribution function to insure that static, median, and peak values reflect the amplitude of muscle contraction when the muscle is active, independent of muscular rest. Importantly, the reliability and sensitivity of this approach to detect distinct patterns of muscle activity throughout the workday is also assessed.

Most previous investigations of muscle activity in the workplace have included populations with diverse occupations (Hansson et al. 2000, Nordander et al. 2000, Østensvik, Veiersted, and Nilsen 2009b). Studies that have focused specifically on office workers often examine standardized tasks performed in a controlled laboratory environment (Szeto, Straker, and O’Sullivan 2009, Dennerlein and Johnson 2006), or analyze isolated office tasks such as keyboarding in the workplace (Nordander et al. 2000). Characterization of more global patterns of muscle activity that occur throughout the workday, including task transitions and extraneous activities that are not performed during typical computer work, may provide additional insight into activity patterns that contribute to overuse injuries in an office setting where physical demands are generally low. Furthermore, differences in cervical muscle activity associated with hand dominance (Nordander et al. 2000, Richter et al. 2009) have rarely been addressed in previous literature despite asymmetrical demands typical of office work, such as using a computer mouse. Similarly, sex differences in work-related cervical muscle activity are not well understood despite a consistently greater incidence of neck pain in women compared to men (Hoy et al. 2014).

The purpose of this investigation was to characterize patterns of upper trapezius muscle activity in a large cohort of pain-free office workers using both Traditional and Active APDF methodologies, in addition to other common summary measures of occupational exposure (e.g. muscular rest, gap frequency, mean EMG). The day-to-day reliability and sensitivity of these measures to intra- and inter-individual differences in hand dominance and sex were examined to determine their utility for future worksite investigations. We expected all occupational exposure measures to demonstrate acceptable reliability for research applications, with the Traditional APDF method providing lower absolute estimates of static, median, and peak muscle activity compared to the Active APDF method. We also expected to observe a greater magnitude of static activity and less muscle rest in the dominant (i.e., more active) compared to non-dominant upper trapezius and in women (i.e., higher incidence of neck pain) compared to men.

Methods

Participants

A convenience sample of 77 healthy office workers was recruited through print and radio advertisements, new employee orientations, and employee bulletins and flyers posted at businesses employing a large number of office workers in the greater Denver area. Eligible participants were within three months of their date of hire and worked > 30 hours per week in an office setting that required the use of a computer for at least 75% of the workday. Participants were screened for the presence of neck pain or associated disorders during the previous year. To avoid the potential for selection bias due to poor recall of pain symptoms, the Neck Disability Index (NDI) (Vernon and Mior 1991) was used to screen for activity limitations caused by pain which are more likely to be remembered than non-interfering neck pain. Participants were included in the study if they reported no neck pain or associated disorders during the previous year, and scored < 5 points on the NDI.

Exclusion criteria included: 1) objective signs of structural pathology upon physical examination by a licensed physical therapist, including but not limited to shoulder bursitis, impingement, tendonitis, fracture, and cervical nerve or disc impairment with radiculopathy or loss of sensory or motor function, 2) self-reported fibromyalgia diagnosis or musculoskeletal pain present in more than four body regions concurrently, 3) self-reported systemic illness including cancer, rheumatic, cardiovascular, or neurological disease, 4) prior surgery involving the cervical spine or shoulders, 5) acute (< 12 weeks prior to study) injury of the neck or shoulders, 6) untreated psychiatric condition, 7) uncontrolled hypertension, 8) pregnancy, and 9) an inability to type or comprehend written and oral instructions in English. All participants provided written informed consent according to study procedures approved by the Colorado Multiple Institutional Review Board.

Data collection protocol

Participants first completed a demographics questionnaire at a familiarization session. Bilateral upper trapezius muscle activity and electrocardiography (ECG) were recorded with a portable data monitor on a representative workday selected by the participant based on convenience. A subset of participants (N = 63) completed a second day of worksite monitoring within two weeks of the first to examine the reliability of occupational exposure summary measures. Participation in the second day of recording was based on participants’ willingness and availability to wear the monitor on a second workday. An investigator met the participant at a private location near his or her workplace 15 minutes prior to the start of the workday for equipment setup. After visual verification of signal quality, two 10 s submaximal reference voluntary efforts (RVE) and resting measurements were collected. RVEs were measured with both arms held at 90 degrees of flexion in the sagittal plane and 45 degrees of abduction in the horizontal plane, while wearing 1 kg wrist weights with the forearms in full supination. This position was slightly modified from Hansson et al., (2000) to reduce the setup time required in the workplace, and for comparison with ongoing laboratory studies using the same reference position. Resting measurements were collected during quiet sitting with the shoulders relaxed and hands resting in the lap for 10 s. Trained laboratory personnel monitored the EMG signal online to confirm the absence of muscle activity during resting measurements, and provided verbal feedback as necessary to insure complete muscle relaxation.

Participants wore a portable data monitor continuously in a minimally obtrusive waist pack positioned on the anterolateral aspect of the waist throughout the workday as they performed their usual activities. Participants were instructed to ignore the device and not make any changes to their usual work routine. No observation of work activities was performed by study investigators to minimize observation bias. The portable data monitoring system was removed at the end of the workday by study investigators after repeating resting EMG measurements at the workplace.

Data recording and processing

The portable data monitor (Delsys Myomonitor IV; Delsys, Boston, MA, USA) recorded surface EMG bilaterally from the upper trapezius. Surface electrodes (DE-2.3 Single Differential Surface EMG Sensor; Delsys) with two parallel 1 × 10mm silver surface contacts with an inter-contact distance of 10 mm were used. Signals were amplified (x1000 V/V) and band-pass filtered using hardwired settings (20–450 Hz) prior to sampling (1000 Hz). Electrodes were placed on the upper trapezius muscle belly 2 cm lateral to the midline between the seventh cervical vertebra and the posterior acromion process (Farina et al. 2002). Electrode positions were marked and covered with a water-proof protective sealant to improve the reliability of sensor placement when multiple days were recorded. ECG was recorded with a bi-lead sensor (SP-X14 EKG Sensor for Myomonitor System; Delsys) which amplified (x1000 V/V) and band-pass filtered (0.5 – 30 Hz) the signal prior to sampling (1000 Hz). ECG leads were attached to 1.75 inch diameter foam and solid gel electrodes with 1 cm Ag/AgCl conducting snaps (Red Dot 9640; 3M, St. Paul, MN, USA) which were placed vertically on the flat portion of the sternum (Marker and Maluf 2014). EMG and ECG signals were referenced to a bony surface on the right clavicle with a self-adhesive electrode pad. EMG and ECG data were stored on an internal 1GB memory card in consecutive one-hour data sets for offline processing. Each data set was 10 s short of an hour (3590 s) to allow for automated internal data storage.

EMG signals were pre-processed with custom scripts implemented in Spike2 (Cambridge Electronic Design, Cambridge, UK) and Matlab (MathWorks Inc., Natick, MA, USA) prior to occupational exposure analyses. Each hour long data file was first visually inspected for artifacts caused by wire movement, which were removed and replaced with the mean of the preceding 0.5 s of the EMG signal. The ECG signal was used to identify periods of ECG contamination in the EMG signal, which were subsequently removed using a validated filtered template subtraction technique to minimize the loss of EMG signal in the process of ECG contamination removal (Marker and Maluf 2014). EMG was then root-mean-square (RMS) processed with a 100 ms window, updating every 10 samples. System noise was removed by subtraction of the average RMS value recorded in the baseline resting period from the remaining RMS EMG signal in a power sense (Jackson, Mathiassen, and Dempsey 2009, Hansson 2011). RVE values were computed as the mean of the RMS processed signal for a stable 8 s period within each RVE trial and then averaged across the two trials (Jackson, Mathiassen, and Dempsey 2009). EMG was normalized to RVE prior to computation of occupational exposure summary measures. Figure 1 shows representative EMG data before and after processing and normalization.

Figure 1.

Electromyographic (EMG) data for a representative 100 s sample prior to processing (A) and after root mean square (RMS) processing and normalization to a reference voluntary effort (RVE) (B). The dashed line represents the 3% RVE muscular rest threshold. The Traditional Amplitude Probability Distribution Function (APDF) derived from normalized RMS data is shown in (C) and the Active APDF is shown in (D) for visualization of the 10^th (static), 50^th (median), and 90^th (peak) percentiles (dotted lines). No data below the 3% RVE muscular rest threshold in (B) is included in the calculation of the Active APDF or percentile values in D. The Traditional APDF (C) is shifted leftward and has a steeper initial slope than the Active APDF (D). Note that both APDF calculations were performed using data from a complete 1-hour EMG recording, not just the 100 s period shown in A and B.

Data analysis and occupational exposure summary measures

All occupational exposure analyses were performed using custom software written in LabVIEW (National Instruments, Austin, TX, USA). To facilitate the assessment of within-day variability of EMG recordings, occupational exposure summary measures were calculated for each complete hour of data collected. Following reliability analyses, summary measures were averaged across the workday for each participant and then averaged across days (Fethke et al. 2012) for participants who completed two days of data collection. It should be noted that calculation of ADPF percentiles is a non-linear process, so averaged values reflect composite muscle activity for a one-hour recording period rather than the full workday. All summary measures were calculated separately for the dominant and non-dominant upper trapezius muscles.

Mean amplitude across the entire recording period was calculated as a global index of muscular load. Gaps in muscular activity were defined as any periods in which muscle activity fell below 3% RVE for at least 0.125 s (Hansson et al. 2000). Gap frequency was expressed as the number of gaps/min and muscular rest was defined as the summed duration of all gaps expressed as a percentage of total recording time. A threshold value of 3% RVE has previously been used to approximate 0.5% of maximum voluntary contraction (MVC) (Hansson et al. 2000), which is the recommended threshold for identifying muscular rest (Veiersted et al. 2013).

Static, median, and peak amplitudes of muscle activity were calculated using two procedures. First, they were identified from a Traditional APDF as the 10^th, 50^th, and 90^th percentiles of the entire RMS processed signal, respectively. These values are equivalent to the static, median, and peak percentile rankings of the APDF (Jonsson 1982), illustrated in Fig 1C. A second APDF was then calculated by removing all values below the 3% RVE muscular rest threshold prior to calculating percentiles for the remaining RMS processed signal to assess the amplitude of muscle activity only during periods of active muscle contraction. The results of this modified analysis are referred to as Active APDF values (Figure 1D). Comparing Figures 1C and 1D illustrates how large periods of muscle rest cause a leftward shift and steep initial increase in the Traditional APDF that reduce estimates of static, median, and peak muscle activity.

Statistical analysis

All statistical analyses were performed using R (R Development Core Team, Vienna) and SAS (SAS Institute, Cary, N.C.). Between day reliability was calculated for all occupational exposure summary measures among participants who completed two days of data collection, using intraclass correlations (ICC(2,1)). Variance components associated with participant ($S_P^2$), day ($S_D^2$, within participant), and hour ($S_H^2$, within day) were calculated with a mixed model Analysis of Variance (ANOVA), entering participant, day, and hour as random effects. Variance components were computed separately for each muscle (dominant and non-dominant), and the first-order autoregressive covariance structure was applied to account for correlation between consecutive hours of data collection. Given the large imbalance between men and women in the sample, sex differences were calculated using non-parametric comparisons (Wilcoxon Rank-Sum Test) for age, body mass index (BMI), mean amplitude, gap frequency, muscular rest, and Traditional and Active APDF variables. Occupational exposure summary measures for the dominant and non-dominant muscles were averaged when investigating sex differences. Summary measures were also compared between the dominant and non-dominant upper trapezius using paired t-tests.

Results

A total of 60 women and 17 men participated. Poor signal quality identified during visual inspection of the data led to the exclusion of 12 days of data (unstable system noise periodically exceeding 3% RVE), completely removing all days of data collection for three participants. Seventy-four participants remained in the final analyses (58 women, 15 men). Mean age (SD) was 31 (7.6) years (women: 31 (7.8), men: 31 (7.1)) and mean BMI (SD) was 24 (4.7) kg/m² (women: 24 (5.0), men: 25 (3.1)). Two days of EMG recordings remained for 55 participants in the final analysis, who did not differ from the full sample of participants with regard to demographic characteristics (p ≥ 0.05). On average, 6.65 hours of data were analyzed per workday (range: 4 – 8 hours). One participant was left-hand dominant. There were no significant differences between men and women in age and BMI.

Table 1 reports ICC values and 95% confidence intervals for between-day reliability of all occupational exposure summary measures in the dominant and non-dominant upper trapezius muscles. Reliability was qualitatively described as recommended by Landis and Koch (Landis and Koch 1977), with ICC values interpreted as: 0–0.20 – slight agreement, 0.21–0.40 – fair agreement, 0.41–0.60 – moderate agreement, 0.61–0.80 – substantial agreement, and 0.81–1.00 – almost perfect agreement. Values ranged from 0.27 to 0.77, demonstrating fair to substantial agreement for all occupational exposure summary measures. Reliability was higher for the non-dominant compared to the dominant trapezius muscle for the majority of summary measures, and reliability of the Active APDF static value was substantially higher than reliability of the Traditional APDF static value.

Table 1.

Between-day reliability of occupational exposure outcomes

	Dominant Trapezius		Non-Dominant Trapezius
	ICC (2,1)	95% Confidence Interval	ICC (2,1)	95% Confidence Interval
Mean Amplitude	0.71	0.54 – 0.82	0.76	0.61 – 0.85
Muscular Rest	0.40	0.15 – 0.60	0.52	0.30 – 0.69
Gap frequency	0.50	0.27 – 0.67	0.50	0.27 – 0.67
Traditional APDF:
Static	0.27	0.01 – 0.50	0.23	0.00 – 0.46
Median	0.77	0.64 – 0.86	0.77	0.63 – 0.86
Peak	0.71	0.55 – 0.82	0.74	0.59 – 0.84
Active APDF:
Static	0.73	0.58 – 0.83	0.54	0.33 – 0.71
Median	0.75	0.60 – 0.84	0.75	0.60 – 0.84
Peak	0.70	0.54 – 0.82	0.64	0.46 – 0.77

ICC – Intraclass Correlation CoefficientAPDF – Amplitude probability distribution function

Mean amplitude, muscular rest, gap frequency, and Traditional and Active APDF values for the dominant and non-dominant upper trapezius, along with variance components due to participant, day, and hour, are presented in Table 2. All summary measures were significantly different between the two muscles. The largest difference was observed in the amount of muscular rest, with the non-dominant muscle being inactive for approximately 7% more of the total recording time than the dominant muscle. All APDF values (static, median, and peak) were greater in the dominant compared to the non-dominant upper trapezius. Relative variance components were similar to previous studies (Fethke et al. 2012), and variance due to day of recording was consistently the smallest component (except in dominant muscular rest and dominant and non-dominant Traditional APDF static values). It should be noted that variance due to hour (within day) is dependent on the length of the sampling window, and would be expected to differ for sampling windows other than 1 hour in length. Results of these comparisons did not meaningfully change when men were removed from the analysis.

Table 2.

Comparison of occupational exposure outcomes for the dominant and non-dominant trapezius muscles.

	Dominant Trapezius				Non-Dominant Trapezius				p-value
	Mean (SD)	${\mathit{S}}_{\mathit{P}}^2$ (%)	${\mathit{S}}_{\mathit{D}}^2$ (%)	${\mathit{S}}_{\mathit{H}}^2$ (%)	Mean (SD)	${\mathit{S}}_{\mathit{P}}^2$ (%)	${\mathit{S}}_{\mathit{D}}^2$ (%)	${\mathit{S}}_{\mathit{H}}^2$ (%)
Mean Amplitude (%RVE)	26.1 (10)	80.6 (45)	28.5 (16)	68.5 (39)	21.3 (9.4)	73.8 (52)	18.2 (13)	51.1 (36)	< 0.001
Muscular Rest (% recording time)	27.0 (12)	70.1 (23)	91.3 (30)	139.5 (46)	34.0 (14)	118.7 (32)	89.7 (25)	157.6 (43)	< 0.001
Gap Frequency (gaps/min)	12.4 (4.6)	12.5 (27)	7.3 (16)	26.0 (57)	15.1 (6.4)	25.3 (31)	22.1 (27)	33.7 (42)	< 0.001
Traditional APDF:
Static (%RVE)	1.2 (1.6)	1.1 (13)	2.1 (24)	5.5 (63)	0.71 (1.2)	0.5 (12)	1.0 (24)	2.7 (64)	0.006
Median (%RVE)	17.1 (9.8)	82.3 (48)	18.4 (11)	70.3 (41)	12.6 (9.2)	72.6 (48)	15.0 (10)	62.1 (41)	< 0.001
Peak (%RVE)	61.9 (21)	350.0 (45)	123.5 (16)	295.9 (38)	52.5 (20)	319.3 (50)	84.6 (13)	236.1 (37)	< 0.001
Active APDF:
Static (%RVE)	7.6 (2.8)	6.5 (47)	1.7 (12)	5.6 (41)	6.8 (2.9)	5.5 (38)	4.5 (31)	4.3 (30)	0.006
Median (%RVE)	26.6 (8.9)	65.4 (49)	19.9 (15)	48.6 (36)	23.8 (10)	84.2 (55)	24.0 (16)	43.9 (29)	0.002
Peak (%RVE)	73.0 (10)	324.5 (45)	116.3 (16)	274.0 (38)	66.7 (19)	268.6 (44)	124.7 (20)	218.0 (36)	0.001

RVE – Reference Voluntary Effort, APDF – Amplitude probability distribution function, $S_P^2$ – Variance due to participant, $S_D^2$ – Variance due to day (within participant), $S_H^2$ – Variance due to hour (within day)

Variance components also expressed as percent of total variance.

Comparisons between women and men for mean amplitude, gap frequency, muscular rest, and Traditional and Active APDF values (averaged across dominant and non-dominant muscles) are reported in Figure 2. Women showed a significantly greater mean amplitude of muscle activity and less muscular rest compared to men, with no significant difference observed in gap frequency. All Traditional APDF values were significantly higher in women compared to men, whereas no significant sex differences were observed for Active APDF values.

Figure 2.

Comparison of women (N = 58) and men (N = 15) for occupational exposure summary measures averaged across the dominant and non-dominant trapezius muscles. Values are mean (SD). Note that Traditional APDF values are significantly different between women and men, whereas Active APDF values are not.

* p < 0.05

To insure that the results were not biased by combining data from participants with either one or two days of data collection, all analyses were repeated using only mean values from participants with two days of recording. No significant changes in the results were observed.

Discussion

This study characterized patterns of muscle activity in the dominant and non-dominant upper trapezius muscles in the largest studied cohort of pain-free office workers to date. All occupational exposure measures demonstrated fair to substantial reliability between days and low relative variance due to day of recording, supporting their use in clinical and research applications. Both muscles demonstrated large amounts of muscular rest (greater than 25% of the workday), with the non-dominant muscle spending significantly more time at rest than the dominant. Traditional and Active APDF amplitude measures were all lower in the non-dominant muscle compared to the dominant. Women demonstrated significantly less muscular rest compared to men. The amplitude of trapezius muscle activity was found to be significantly higher for women when assessed with the Traditional APDF, whereas no significant sex difference in EMG amplitude was observed after removing the influence of muscular rest with the Active APDF method.

Muscular rest, mean muscle activity, and gap frequency

Mean muscle activity is often used as a global measure of muscular load across the full workday (Richter et al. 2009), and therefore includes periods of rest during which there is no muscular load. We observed that muscular rest was significantly less and mean muscle activity was significantly greater in the dominant trapezius muscle. The same pattern was observed when comparing sexes, with women showing greater muscle activity and less muscular rest. This pattern is to be expected, given that periods of muscular rest are included in the calculation of mean muscle activity thereby decreasing its mean amplitude. Differences in muscular rest and mean activity between the dominant and non-dominant muscles indicate that the non-dominant trapezius is utilized less by healthy office workers, likely decreasing global loads placed on this muscle throughout the workday. These measures also suggest a higher global workload for women compared to men; the origin of this difference is further investigated with the discussion of Active APDF measures, below.

Gap frequency was also greater in the non-dominant muscle, but did not differ between women and men. The inter-limb difference in gap frequency is not solely due to the increased amount of muscular rest, as muscular rest and gap frequency reflect distinct motor patterns (Hansson et al. 2000), in which the maximum amount of muscular rest possible would be obtained with fewer rather than more periods of muscle activity separating periods of rest. Therefore, the difference in gap frequency between muscles likely reflects differences in work demands and neuromuscular control of the dominant and non-dominant upper extremities (Hansson et al. 2000).

Traditional and Active APDF

The office workers included in the study sample were similar to those of previous studies (Aarås et al. 1996, Nordander et al. 2000, Blangsted, Hansen, and Jensen 2003), showing large periods of muscular rest throughout the workday. For comparison to previous studies, these periods of muscular rest were included in the calculation of Traditional APDF values. Muscular rest periods were subsequently removed prior to calculating percentile values for the remaining signal, thereby creating the Active APDF to estimate the amplitude of muscle activity only during periods of active muscle contraction. Traditional APDF values were similar to previous studies using the same method, showing static values as low as 0.1% MVC (Holte and Westgaard 2002) or 2.3% RVE (Nordander et al. 2000). While providing important information about composite loads on the muscle throughout the day, these estimates are influenced by large amounts of rest and likely underestimate the amplitude of muscle contraction while the muscle is active (as shown in Figure 1C–D). The low values of static muscle activity may also be insensitive to subtle differences in the magnitude of muscle activity required to maintain postural control of the dominant (more active) and non-dominant (less active) limbs, resulting in few and conflicting reports of hand dominance effects on muscle activity (Nordander et al. 2000). In the current study, however, both Traditional and Active APDF values were significantly lower in the non-dominant limb.

In contrast, findings from the Traditional and Active APDF analyses differed in the comparison of men and women. The Traditional method resulted in significantly higher values for static, median, and peak levels of muscle activity in women, whereas sex differences were less apparent after removing periods of muscular rest with the Active APDF method which showed no significant differences between men and women. One interpretation of this discrepancy is that Traditional APDF values were heavily influenced by significantly reduced amounts of muscular rest in women, indicating that higher APDF values reflect differences in muscular rest between the sexes rather than differences in the amplitude of static and dynamic contractions required to perform job tasks. Contrary to our initial hypothesis, this interpretation suggests that although the upper trapezius muscle in women may be at rest for a smaller proportion of the workday compared to men, its pattern of activation when not at rest is similar to men. These findings highlight the potential utility of the Active APDF method, as we show for the first time that the higher global workloads (i.e., mean muscle activity) observed in women is due to reduced muscular rest but not higher levels of EMG activity while the muscle is active. Future studies should determine whether reduced amounts of muscular rest contribute to the higher prevalence of neck pain in female compared to male office workers (Paksaichol et al. 2012) and whether scheduled rest breaks (as opposed to task modifications to reduce physical workloads) can reduce the prevalence of neck among women.

Overall, the results of both the Traditional and Active APDF analyses indicate that the amplitude of muscle activity in office workers typically remains low. In the current study, muscle activity was normalized to a submaximal reference instead of a maximal contraction, and peak activity (90^th percentile) was still below 100%. Although submaximal reference contractions are reliable (Bao, Mathiassen, and Winkel 1995, Delisle et al. 2009), normalized EMG values are more difficult to interpret than those normalized to maximal contractions. However, 100% RVE recorded while holding a 1 kg weight with the arm extended would likely not represent a very demanding load on the muscle. Previous research utilizing a similar RVE method reported RVE to be approximately 20% of a maximum contraction (Balogh et al. 1999). This indicates low levels of muscle activity in the current population, as peak values were consistently less than 100% RVE.

Limitations

It should be emphasized that Active APDF values provide information on the amplitude of muscle activity during known periods of active muscle contraction. As these values are independent of rest periods, they must be interpreted alongside other measures such as muscular rest and gap frequency to provide a comprehensive assessment of work-rest patterns which have been examined by previous studies with conflicting results (Hoe et al. 2012). It is particularly important to interpret Active APDF values in combination with muscular rest because the proportion of resting values that are removed differs between participants and percentiles are calculated from variable amounts of data. The present study did not consider all available occupational exposure summary measures as each provides unique information, and no single or standard set of measures is currently accepted as the gold standard (Van Eerd et al. 2012). Other methods such as Exposure Variation Analysis (EVA)(Mathiassen and Winkel 1991) and SUstained Low-level Muscle Activity (SULMA) (Østensvik, Veiersted, and Nilsen 2009b) may provide additional insight regarding temporal aspects (e.g. repetitiveness) and individual variability of physical workload (Srinivasan and Mathiassen 2012) which may also contribute to overuse injuries in the workplace. Thus, the Active APDF is proposed as an independent measure of EMG amplitude during periods of active muscle contraction that can complement, rather than replace, existing methods of exposure analysis.

In contrast to the Traditional APDF method, one disadvantage of the Active APDF is that it requires a threshold for distinguishing active muscle contraction from muscular rest, which can be difficult in the presence of low-amplitude muscular activity (Veiersted et al. 2013). The use of a 3% RVE threshold for muscular rest in this investigation may be considered conservative, potentially resulting in the misinterpretation of low-amplitude muscle activity as muscular rest for some participants. This was a large field study, in which participants performed a multitude of tasks in many different worksites without direct surveillance of factors that can affect signal quality, and factors such as environmental electrical noise and movement artifacts varied widely between participants. Therefore, a conservative 3% RVE threshold was selected to insure that the rest threshold would only be breached by true increases in muscle activity. This RVE threshold has been used previously (Hansson et al. 2000) to approximate the recommended muscular rest threshold of 0.5% MVC (Hansson et al. 2000, Veiersted et al. 2013).

Normalization to a submaximal reference contraction does not permit direct comparisons of the amplitude of muscle activity to a physiologic maximum or to previous studies using different normalization methods. However, RVE normalization methods are reliable (Delisle et al. 2009, Bao, Mathiassen, and Winkel 1995) and have been recommended in previous studies (Hansson et al. 2000). This method was chosen based on the results of a pilot study demonstrating greater reliability of submaximal compared to maximal voluntary exertions when performed within the constraints of the participants’ worksites. RVE normalization was also selected to facilitate comparison of the present results to future studies of patients with chronic neck pain, in which maximal effort is inhibited by pain and not considered a valid normalization method (Burden 2010).

Another limitation is that participants were not observed throughout the work day to synchronize periods of muscle activity with specific tasks, as has been done previously (Hansson et al. 2000, Nordander et al. 2000). This was done to minimize the influence of study observers on the behavior of participants to insure externally valid measurements. It is possible, however, that significant associations with self-reported activity in the workplace may exist for specific office tasks that are masked by more global measures of muscle activity across the full workday. Finally, the proportion of men in the present sample was low, although representative of the lower proportion of male office workers in the general population (Gjerdingen et al. 2001).

Conclusions

This is the largest study to date describing upper trapezius motor patterns for pain-free office workers in their natural work environment. A modification to the Traditional APDF calculation is proposed in which periods of muscular rest are removed prior to calculating amplitude distribution percentiles for the active muscle. These analyses are reliable and sensitive to differences in motor patterns associated with hand dominance and sex. Findings from the present study provide normative data that may be useful in identifying atypical motor patterns that contribute to the persistence or recurrence of chronic neck pain in office workers.