2.3.2. Heart Rate

The electrical activity of the heart was recorded using electrocardiography (ECG) by two pre-gelled Ag/AgCl surface electrodes (35 × 26 mm, KendallTM H93SG ECG Electrodes, Covidien, Zaltbommel, the Netherlands) placed ~5 cm cranial and ~3 cm left-lateral from the distal end of the sternum and over the anterior to midaxillary line at the fifth left rib. The PS11 sampled (1000 Hz) and recorded the ECG signal (PS11-UD, THUMEDI® GmbH & Co. KG, Thum, Germany), from which we calculated the mean heart rate (HRMEAN) for the overall task, for the screwing part (phase 1), and for the fastening part (phase 2).

#### 2.3.3. Forearm Acceleration

Acceleration of the forearm was sampled (4096 Hz) and recorded (PS11-UD, THUMEDI® GmbH & Co. KG, Thum, Germany) using a single-axis accelerometer (resolution 0.005 m/s2) placed ventral over the extensor digitorum, 2 to 3 cm ventral of the styloid process ulnaris, using double-sided adhesive tape. Its orientation, i.e., its sensitive axis, was the orthogonal of the forearm bones ulna and radius. The position and orientation were chosen to avoid mechanical interference with the EMG recording electrodes. The accelerometer measured the flexion-extension and rotational acceleration of the forearm. The RMS of the acceleration data was real-time calculated (250-millissecond moving window with 50% overlap), from which we determined the mean (ACCMEAN) for the overall task, screwing only (phase 1), and fastening only (phase 2). Similar as for muscular activity, we calculated the relative cycle-to-cycle variability of the forearm acceleration (ACCCV), for each experimental day, by dividing the square root of the mean variance across all work cycles for screwing and fastening by the ACCMEAN.

#### *2.4. Statistical Analysis*

We checked whether the parameters were normally distributed by visually inspecting the histograms and skewness and kurtosis values [24,25]. Based on these explorations, RMS10, RMSMEDIAN, RMS90, and RMSCV were not normally distributed and were therefore log-transformed before the statistical analyses. We performed repeated-measures analysis of variance (RM-ANOVA) on the mean parameters (i.e., RMS10, RMSMEDIAN, RMS90, HRMEAN, and ACCMEAN) of the overall one-hour task, with day as the within-subject factor. We also performed RM-ANOVA (day as within-subject factor) on the mean parameters of phases 1 and 2 separately (i.e., RMS10, RMSMEDIAN, RMS90, RMSCV, HRMEAN, ACCMEAN, and ACCCV). In case of a significant main effect of day, we performed post hoc tests using Student's T-Tests with Bonferroni correction and calculated Cohen's *d* effect sizes using the pooled standard deviation of the two respective days as a standardizer [26]. All statistical analyses were performed with JMP (JMP® 13.1.0) and statistical significance was accepted for the main effects when *p* < 0.05 or, for the Bonferroni-corrected post hoc comparisons, when *p* < 0.0167 (i.e., 0.05 divided by three comparisons). Effect sizes were considered small (*d* ≥ 0.2), medium (*d* ≥ 0.5), or large (*d* ≥ 0.8), as suggested by Cohen [27].

#### **3. Results**

All data with statistical outcomes are summarized in Table S1 (Supplementary Material). Data of a different number of subjects was available for each parameter due to failed or unreliable recordings, of which the specific number of subjects is mentioned in Table S1 (Supplementary Material).

#### *3.1. Muscular Fatigue*

None of the muscles showed signs of muscular fatigue. Only the flexor carpi radialis showed a significant decrease in RMSMEDIAN concomitant with a significant increase in MPF, which actually points towards the opposite of muscular fatigue, i.e., recovery of muscular strength [23].

#### *3.2. Muscular Activity*

#### 3.2.1. M. Triceps Brachii

The factor day had a significant effect on RMSMEDIAN and RMS90 during the overall task (*F* = 8.88, *p* < 0.001), the screwing phase (*F* = 7.64, *p* < 0.001), and the fastening phase (*F* = 5.81, *p* < 0.01). For the overall task, RMSMEDIAN decreased from day 1 (11.80 ± 9.56 %RVE) to day 2 (10.31 ± 8.33 %RVE; *p* < 0.01, *d* = 0.16) and from day 1 (11.80 ± 9.56 %RVE) to day 3 (9.32 ± 6.84 %RVE; *p* < 0.001, *d* = 0.30). For screwing, RMSMEDIAN decreased from day 1 (13.74 ± 11.12 %RVE) to day 2 (12.08 ± 10.17 %RVE; *p* < 0.01, *d* = 0.16) and from day 1 (13.74 ± 11.12 %RVE) to day 3 (10.98 ± 9.18 %RVE; *p* < 0.001, *d* = 0.29). For fastening, RMSMEDIAN decreased from day 1 (10.23 ± 7.97 %RVE) to day 3 (8.23 ± 6.05 %RVE; *p* < 0.01, *d* = 0.29). RMS90 decreased in the overall task (*F* = 5.27, *p* < 0.01) from day 1 (38.03 ± 26.13 %RVE) to day 3 (30.82 ± 17.73 %RVE; *p* < 0.01, *d* = 0.33), in the screwing phase also (*F* = 4.38, *p* < 0.05) from day 1 (38.25 ± 26.20 %RVE) to day 3 (31.74 ± 18.83 %RVE; *p* < 0.01, *d* = 0.29), and in the fastening phase (*F* = 9.09, *p* < 0.001) from day 1 (48.47 ± 36.00 %RVE) to day 2 (31.18 ± 31.80 %RVE; *p* < 0.01, *d* = 0.16) and from day 1 (48.47 ± 36.00 %%RVE) to day 3 (39.26 ± 25.81 %RVE; *p* < 0.001, *d* = 0.30; Figure 2). No changes in RMS10 or RMSCV were found between days.

**Figure 2.** Boxplots of the 10th, 50th, and 90th percentile muscular activity (RMS10, RMSMEDIAN, RMS90) and coefficient of variation (RMSCV) of the triceps brachii over three days for the overall task (O, grey filled boxplots), the screwing phases (S, black filled boxplots), and the fastening phases (F, white filled boxplots).

#### 3.2.2. M. Biceps Brachii

RMSMEDIAN of the overall task was significantly influenced by day (*F* = 5,89, *p* < 0.01), with day 1 holding higher values (28.83 ± 15.01 %RVE) than day 2 (25.51 ± 13.82 %RVE; *p* < 0.01, *d* = 0.23) and day 3 (25.50 ± 13.60 %RVE; *p* < 0.01, *d* = 0.23; Figure 3). RMSMEDIAN for the screwing phase also showed a significant effect of day (*F* = 3.25, *p* < 0.05) but post hoc tests revealed no significant pairwise comparisons. RMS90 was significantly influenced by day during fastening (F = 14.64, p < 0.001) with a decrease from day 1 (148.33 ± 69.01 %RVE) to 2 (122.52 ± 62.60 %RVE; *p* < 0.001, *d* = 0.36) and also from day 1 (148.33 ± 69.01 %RVE) to 3 (114.85 ± 58.87 %RVE; *p* < 0.001, *d* = 0.49). RMS10 and RMSCV did not change between days.

**Figure 3.** Boxplots of the 10th, 50th, and 90th percentile muscular activity (RMS10, RMSMEDIAN, RMS90) and coefficient of variation (RMSCV) of the biceps brachii over three days for the overall task (O, grey filled boxplots), the screwing phases (S, black filled boxplots), and the fastening phases (F, white filled boxplots).

#### 3.2.3. M. Flexor Carpi Radialis

During the fastening phase, RMS90 was significantly influenced by day (*F* = 4.09, *p* < 0.05; Figure 4). Post hoc tests showed a decreased RMS90 from day 1 (65.90 ± 80.80 %RVE) to 3 (52.52 ± 53.75 %RVE; *p* < 0.01, *d* = 0.20). No significant difference between days was found for RMS10, RMSMEDIAN, or RMSCV.

**Figure 4.** Boxplots of the 10th, 50th, and 90th percentile muscular activity (RMS10, RMSMEDIAN, RMS90) and coefficient of variation (RMSCV) of the flexor carpi radialis over three days for the overall task (O, grey filled boxplots), the screwing phases (S, black filled boxplots), and the fastening phases (F, white filled boxplots).

#### 3.2.4. M. Extensor Digitorum

We found a significant main effect of day for RMSMEDIAN in the overall task (*F* = 8.56, *p* < 0.001), during screwing (*F* = 6.32, *p* < 0.01), and during fastening (*F* = 3.31, *p* < 0.05). During the overall task, RMSMEDIAN decreased from day 1 (45.29 ± 22.23 %RVE) to 2 (42.51 ± 23.57 %RVE; *p* < 0.05, *d* = 0.12) and from day 1 (45.29 ± 22.23 %RVE) to 3 (40.46 ± 22.00 %RVE; *p* < 0.001, *d* = 0.22). RMSMEDIAN decreased from day 1 (50.34 ± 25.56 %RVE) to 3 (46.00 ± 26.25 %RVE) during screwing (*p* < 0.001, *d* = 0.17) and decreased from day 1 (34.76 ± 18.44 %RVE) to 3 (31.76 ± 19.19 %RVE) during fastening (*p* < 0.05, *d* = 0.16). RMS10 differed significantly between days for the overall task (*F* = 18.35, *p* < 0.001), screwing (*F* = 13.72, *p* < 0.001), and fastening (*F* = 3.50, *p* < 0.05; Figure 5). In the overall task, RMS10 decreased from day 1 (12.70 ± 7.06 %RVE) to 2 (10.13 ± 6.59 %RVE; *p* < 0.001, *d* = 0.38) and from day 1 (12.70 ± 7.06 %RVE) to 3 (9.12 ± 5.84 %RVE; *p* < 0.001, *d* = 0.55). In the screwing phase, RMS10 decreased from day 1 (17.45 ± 9.45 %RVE) to 2 (15.03 ± 8.73 %RVE; *p* < 0.01, *d* = 0.27) and from day 1 (17.45 ± 9.45 %RVE) to 3 (14.04 ± 8.07 %RVE; *p* < 0.001, *d* = 0.39). In the fastening phase, RMS10 decreased from day 1 (16.43 ± 9.30 %RVE) to 3 (15.18 ± 10.42 %RVE; *p* < 0.05, *d* = 0.13). The factor day also significantly influenced RMS90 for the overall task (*F* = 5.96, *p* < 0.01), screwing (*F* = 5.24, *p* < 0.01), and fastening (*F* = 4.27, *p* < 0.05). RMS90 decreased from day 1 (83.00 ± 45.24 %RVE) to 3 (75.60 ± 43.64 %RVE) in the overall task (*p* < 0.001, *d* = 0.17), decreased from day 1 (84.22 ± 47.10 %RVE) to 3 (77.29 ± 45.61 %RVE) in the screwing phase (*p* < 0.01, *d* = 0.15) and decreased from day 1 (84.14 ± 38.57 %RVE) to 3

(76.82 ± 38.52 %RVE) in the fastening phase (*p* < 0.01, *d* = 0.19). No significant effect of day on RMSCV was found.

**Figure 5.** Boxplots of the 10th, 50th,and 90th percentile muscular activity (RMS10, RMSMEDIAN, RMS90) and coefficient of variation (RMSV) of the extensor digitorum over three days for the overall task (O, grey filled boxplots), the screwing phases (S, black filled boxplots), and the fastening phases (F, white filled boxplots).

### *3.3. Heart Rate*

HRMEAN differed significantly between days (Figure 6) for the overall task (*F* = 5.91, *p* < 0.01), screwing (*F* = 9.38, *p* < 0.001), and fastening (*F* = 6.35, *p* < 0.01). For the overall task, HRMEAN decreased from day 1 (91 ± 15 bpm) to 2 (88 ± 12 bpm; *p* < 0.01, *d* = 0.27) and from day 1 (91 ± 15 bpm) to 3 (88 ± 11 bpm; *p* < 0.01, *d* = 0.28). For screwing, HRMEAN decreased also from day 1 (92 ± 15 bpm) to 2 (88 ± 12 bpm; *p* < 0.001, *d* = 0.27) and from day 1 (92 ± 15 bpm) to 3 (88 ± 11 bpm; *p* < 0.001, *d* = 0.29). For fastening it decreased from day 1 (89 ± 15 bpm) to 2 (85 ± 12 bpm; *p* < 0.01, *d* = 0.28) and from day 1 (89 ± 15 bpm) to 3 (85 ± 12 bpm; *p* < 0.01, *d* = 0.29).

#### *Int. J. Environ. Res. Public Health* **2019**, *16*, 1231

**Figure 6.** Average heart rate (HRMEAN) over the three days for the overall task (O, grey filled plots), the screwing phases (S, black filled plots), and the fastening phases (F, white filled plots). Error bars represent the SD across subjects.

#### *3.4. Forearm Acceleration*

ACCMEAN significantly decreased over days in the overall task (*F* = 13.66, *p* < 0.001) and over the screwing phases (*F* = 9.10, *p* < 0.001; Figure 7). For the overall task, ACCMEAN decreased from day 1 (237.13 ± 56.34 mm/s2) to 2 (218.65 ± 51.36 mm/s2; *<sup>p</sup>* < 0.001, *<sup>d</sup>* = 0.34) and from day 1 (237.13 ± 56.34 mm/s2) to 3 (209.45 ± 49.66 mm/s2; *<sup>p</sup>* < 0.001, *<sup>d</sup>* = 0.52). For screwing, ACCMEAN decreased also from day 1 (261.84 ± 65.51 mm/s2) to 2 (244.35 ± 60.95 mm/s2; *<sup>p</sup>* < 0.01, *<sup>d</sup>* = 0.28) and from day 1 (261.84 ± 65.51 mm/s2) to 3 (235.34 ± 58.49 mm/s2; p < 0.001, d = 0.43). We found a main effect of day for ACCCV during fastening (*F* = 3.59, *p* < 0.05). Post hoc tests revealed that ACCCV significantly increased from day 1 (1.52 ± 0.29) to day 3 (1.61 ± 0.36; *p* < 0.01, *d* = −0.25).

**Figure 7.** Average acceleration (ACCMEAN) and coefficient of variation (ACCCV) of the forearm over three days for the overall task (O, grey filled plots), the screwing phases (S, black filled plots), and the fastening phases (F, white filled plots). Error bars represent the SD across subjects.

#### **4. Discussion**

The objective of this study was to investigate whether physical requirements and motor variability decreased over three days of repetitive screwing among novices. The results largely support our hypothesis, showing a decrease over days in the static, median, and peak EMG activity levels for the extensor digitorum and biceps brachii, but to a lesser extent for the flexor carpi radialis and triceps brachii. No significant differences in physical requirements between days 2 and 3 were detected. Similarly, acceleration of the forearm and heart rate decreased over days. We found no support for the hypothesis that relative cycle-to-cycle variability of the muscles decreased over days, whereas the relative variability of forearm acceleration significantly increased over days.

#### *4.1. Biomechanical and Cardiovascular Control Strategies*

The simulated screwing task activated several upper and lower arm muscles, of which we have measured only a selection. It hereby appeared that the biceps brachii produced the highest activity levels as can be indicated from the normalized EMG values, which relates to its main function of lower arm supination [28] and the T -handle torx screwdriver not requiring much grip force.

The extensor digitorum and triceps brachii clearly decreased their median and peak muscle activity. Similar findings were reported by previous studies that found reductions in muscular activity as a result of motor learning during training (e.g., [29–31]). The overall lowered muscular activity production for the same occupational task may indicate that the early stage of motor learning is a process at the level of the central nervous system, which is supported by the decreased heart rate of ~4 bmp (similar to ~4%) over days, since a decreased heart rate reflects less inhibited parasympathetic activation of the autonomic nervous system [32].

Observing the behaviour of all four muscles separately made us conclude that not all behaved similarly over the three days. The flexor carpi radialis and biceps brachii mainly contributed to the fastening phase of the experimental task, since their peak muscular activity levels decreased between day 1 and days 2 and 3, especially during fastening. On the other hand, the triceps brachii and extensor digitorum showed the most prominent changes during the overall task, as indicated by their decreased median and peak activity levels from day 1 to days 2 and 3. These results indicate that the flexor muscles became more efficient, especially during fastening, and may benefit more from specific training than the extensor muscles that showed less specific changes between the three days.

The strong decreased forearm acceleration after the first day, especially during screwing, could point to a more efficient motor program that has significantly developed already after one day. The resulting movement patterns of screwing being smoother on days 2 and 3 may have resulted in more muscle relaxation, as reflected by the decreased static muscular activity level of the extensor digitorum. Note that all changes in physical requirements were detected at day 2 or 3, compared to day 1. This may indicate that the first day is highly important for developing motor control strategies, whereas no significant changes or improvements happen between days 2 and 3.

#### *4.2. Motor Variability*

The amount of motor variability of muscular activity levels remained equal over days in this study, which could imply that the screwing task was not new enough or too simple to provoke developments in motor variability, or that the follow-up was not long enough to be able to find significant developments. The relative variability of forearm acceleration, on the other hand, showed a slight but significant increase regarding the first two days (rate of ~7% increase). This strongly implies that movement strategies and muscular activation strategies develop in a different way. As suggested by previous studies, motor control strategies are optimized with increasing work experience, which could evoke an increased motor variability as a strategy to adapt to performance constraints such as muscular fatigue [33,34] and acute or chronic pain [35].

#### *4.3. Practical Implications*

#### 4.3.1. Importance of Familiarization and Randomization

This study with a within-subject design showed decreased muscular activity and acceleration patterns among novices who gain experience in performing a simple screwing task repeated over three days. Decreases were most prevalent between day one and the other two days but were absent between days 2 and 3. This finding indirectly highlights the importance of both sufficient familiarization to experimental tasks and randomization of experimental conditions in studies. Familiarization is especially important when designing a study including repeated measures, to decrease the influence of early stages in the motor learning process. However, it is not clear how much familiarization time is needed, as this probably highly depends on the task complexity. For example, in a study among novice and expert butchers, Madeleine, et al. [36] showed that motor strategies are in development during a six-month follow-up. This does not mean that follow-up times or familiarization periods should be equal to or longer than six months, because this is unrealistic to strive for in studies. However, familiarization should not be too short and include at least some physical practice. Based on the current study results, a two-minute familiarization period, as used by Qin et al. [37], might be too short to reliably measure and interpret physical requirements of occupational tasks. Not reporting task familiarization at all, like Wang et al. [38] did, makes the reader assume that familiarization was not part of the experimental protocol at all and could mean that outcomes were influenced by early stages of the motor learning process. Since the first day of this repetitive task seemed to be most prominent in developing motor control strategies, researchers should provide enough familiarization when investigating occupational tasks.

For comparing different experimental conditions, one can use a between-subject design or a within-subject design. In a between-subject design, classically, subjects are randomly assigned to a group (control vs. intervention). In a within-subject design, subjects will perform all experimental conditions (cross-over design) and are randomly assigned to a sequence of experimental conditions. In most cases, the randomization is performed counterbalanced, meaning that first the number of sequences or the number of subjects per group is determined, after which the assignment to one of the conditions or sequences is based on randomization. The well-known reasons to apply randomization in studies are twofold as follows: (1) Decreasing systematic errors created by a specific experimental manipulation [39] and (2) increasing generalizability or external validity [40]. In addition, we emphasize that randomization may also decrease effects of lack of or insufficient familiarization. This means that an over- or underestimation of the original research question is equally distributed across conditions or across intervention groups. When having a close look at the recent conference proceedings of the International Ergonomics Association 2018 [41], only six of the 223 short papers published, mention randomization in their study. This emphasizes that few studies take randomization into consideration and, as a result, the quality of the study and, therefore, also the relevance of the outcomes may decrease. We therefore recommend applying randomization when possible and, especially when recruiting novices (e.g., students), including a familiarization phase to decrease most of potential motor learning effects.

#### 4.3.2. Motor Variability in An Occupational Context

This study assessed the functional development of motor variability in a manual materials handling task over three days. Although we could not find changes in motor variability over three days, previous studies showed that experienced workers show more variable movement patterns [6], which may indicate that experienced workers are already in the final improvement stage of motor learning where they have developed movement behaviour enabling them to adapt to environmental constraints [42,43]. Furthermore, motor variability has also been suggested to play a beneficial role with respect to internal constraints, such as muscular fatigue [33], pain [44], and external constraints, such as task precision and pace [45]. Confounding factors are also suggested to play a role in the association between motor variability and internal/external constraints, such as sex [46], age [47], and health status. Our current knowledge about the role of motor variability in relation to the aforementioned concepts is still limited. However, when we continue studying motor variability in occupational settings, we might be able to (1) identify workers who are more prone to develop musculoskeletal disorders than others and (2) design (individually adjusted) work tasks and workstations in such a way that motor variability can be promoted [48].

### *4.4. Methodological Considerations*

Our study supports the hypothesis that muscular activity decreased over three days of repetitive screwing and fastening among novices, although it should be noted that our sample size was rather large due to the exploratory design, which is usually not suitable to examine a hypothesis. In general, effect sizes of the reported results were rather small and only few results were accompanied by moderate effect sizes (*d* = 0.50), including the high muscular activity level of the biceps brachii, the static activity level of the extensor digitorum, and forearm acceleration. Muscular co-contraction has been related to aspects of motor learning in a previous study, [16], by calculating a complex index including the muscular activity and torque levels around a joint, which was beyond the scope of this study.

#### **5. Conclusions**

This study showed that physical requirements start developing at task onset, as reflected by the measures of muscular activity, acceleration, and heart rate. However, not all of these measures showed a similar pattern over days, i.e., the extensor digitorum and biceps brachii already showed changes between days 1 and 2, whereas the flexor carpi radialis and triceps brachii showed changes between days 1 and 3. This may indicate the more prominent role of both arm extensor muscles in the screwing and fastening task, regarding the earlier changes. However, changes between days 2 and 3 in physical requirements were absent. Motor variability of the selected muscles did not change between days, but variability of forearm acceleration increased from day 1 to 3, which may reflect that movement strategies develop differently than muscular activation strategies. We emphasize that these developments are part of daily life and should be considered when designing and interpreting studies in terms of task familiarization and randomization of experimental conditions.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1660-4601/16/7/1231/ s1, Figure S1: Example of a recording of the electrical activity of the triceps brachii of subject 823. The graphs show the raw RMS signal for day 1 (upper), day 2 (middle), and day 3 (lower); Table S1: Results of the one-way repeated measures analysis of variance with the within-subject factor 'day'.

**Author Contributions:** The specific contributions of each author were as follows: conceptualization, T.L., R.S., M.A.R., and B.S.; methodology, T.L.; software, R.S.; validation, T.L.; formal analysis, T.L.; investigation, R.S. and B.S.; resources, B.S.; data curation, T.L., R.S., and B.S.; writing—original draft preparation, T.L.; writing—review and editing, T.L., R.S., M.A.R., and B.S.; visualization, T.L.; supervision, T.L., M.A.R., and B.S.; project administration, R.S. and B.S.; funding acquisition, M.A.R.

**Funding:** The work of the Institute of Occupational and Social Medicine and Health Services Research Tübingen is supported by an unrestricted grant of the employers' association of the metal and electric industry Baden-Württemberg (Südwestmetall).

**Acknowledgments:** We would like to thank Sarah Arnold, Anja Meyer, Felix Epple, and Theresa Kreidler for their assistance in the data analysis.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

International Journal of *Environmental Research and Public Health*
