Protocol

Participants performed the RPT, as first described in Fuller et al. [24]. Briefly, participants repetitively moved their dominant arm between a proximal (30% of arm length) and a distal (100% of arm length) target aligned with their body midline at shoulder height while standing. They maintained a rhythm of one movement per second (2 s for a cycle) set by a metronome. The touch-sensitive cylindrical targets (length 6 cm, radius 0.5 cm Quantum Research Group Ltd) provided auditory feedback when activated to help participants maintain the metronome pace. An elliptically shaped mesh barrier was positioned under participants’ elbow motion trajectory, 10 cm below the target’s midpoint, to ensure it remained elevated during the whole task. Participants rated their level of perceived exertion (RPE) using a Borg CR10 scale every minute [31]. They performed the task until they reached a RPE of 8/10. Participants were unaware of this task termination criterion. Previous studies studying the RPT showed objective signs of performance fatigability such as a decrease of maximal isometric shoulder elevation force and pushing-pulling power generation capacity using the same stopping criterion [24, 32]. EMG signs of fatigue were previously reported in several shoulder and elbow muscles [21, 22].

Kinematic data acquisition and analysis

Upper body and trunk kinematic data were acquired at 120 Hz using a six-camera motion capture system (MX3 Vicon, Oxford Metrics Ltd., Oxford, UK). Reflective markers (10 mm diameter) were placed on the trunk (C7, T10, manubrium), upper arm (acromioclavicular joint, deltoid insertion, lateral epicondyle), forearm (middle of the forearm, medial and lateral styloid processes), and hand (second metacarpophalangeal joint, index fingertip) [24, 33]. Before performing the RPT, each participant maintained a static position (standing with the arm in anatomical position or elevated at 90° in abduction) to define the kinematic model. Marker trajectories were low-pass filtered at 15 Hz (zero lag, Butterworth, fourth order).

A generic kinematic model adapted from the Standford VA model available in OpenSim [34, 35] was scaled using static trial data. The model included the following degrees-of-freedom: trunk-global (three translations [XYZ], three rotations [XYZ]), humerothoracic (three translations [XYZ], three rotations [Y₁XY₂]), and elbow [ZY] as well as multiple joints at the wrist and hand (Fig. 1). Trunk lateral flexion (X rotation), axial rotation (Y rotation), and flexion (Z rotation), and humerothoracic plane of elevation (Y₁ rotation), and elevation (X rotation) as well as elbow flexion (Z rotation) waveforms were reported for analysis [36]. As in Wu et al. [36], the humerothoracic elevation angle is the angle between longitudinal axes of the humerus and trunk. The plane of elevation is the plane in which this elevation is observed: 0° is abduction and 90° is flexion. Dominant index (IDX) and lateral epicondyle (ELB) marker positions were also analyzed, as they are directly relevant to the task goals, pointing toward the target for the former and maintaining the elbow above the mesh for the latter. All data were then partitioned into individual forward and backward pointing movements using the target switch signal or peak IDX antero-posterior velocity. For each movement, the following variables were computed:

• Each joint’s average angle (to represent average posture) and range of motion (to represent movement amplitude)

• The Euclidean distance between IDX marker and the distal target’s center at the end of the movement (movement error)

• The average vertical distance between ELB marker and the mesh barrier (elbow height)

• Movement duration timing error (|movement duration – 1 s|).

The mean value and movement-to-movement variability (standard deviation) of each joint angular variable (range of motion and average angle) and of movement errors were calculated across all movements for one condition (Non-Fatigue: first 30 s or Fatigue Terminal: last 30 s). Only mean values were calculated for the elbow height, movement duration, and timing error variables. Only data from forward movements are reported here.

Statistical analysis

Men’s and women’s age, height, weight, and endurance time were compared with independent sample t tests. Two-way ANOVAs (sex [between subjects: men vs women] × Fatigue [repeated measures: Non-Fatigue vs Fatigue Terminal]) were performed on each of the variables. Post hoc paired or independent sample t tests with Holm corrections [37] were performed when interactions were observed for the following planned comparisons: women Non-Fatigue vs women Fatigue Terminal, men Non-Fatigue vs men Fatigue Terminal, women Non-Fatigue vs men Non-Fatigue, and women Fatigue Terminal vs men Fatigue Terminal. The standardized response mean (SRM) was calculated for each variable for men and women with its confidence interval to judge the similarities and differences in motor adaptations to muscle fatigue [38]. The SRM was interpreted qualitatively according to Cohen’s standards as low (0.2 < SRM < 0.5), medium (0.5 < SRM < 0.8), or high (SRM > 0.8) [39].

To assess the association between each Non-Fatigue joint angle variable and endurance time, Pearson correlation coefficients with their 95% confidence interval were computed.

Effects of movement repetitions and sex on motor performance

Most of the RPT performance variables did not change with fatigue in men or in women (Fig. 2). Indeed, no Fatigue, Sex, or Sex × Fatigue effects were observed for the movement error, movement error variability, and movement duration variables (all F < 3.195, p > 0.078). However, elbow height decreased with fatigue (main effect of Fatigue: F_1,79 = 45.019, p < 0.001). Moreover, men’s elbow was significantly higher relative to the mesh barrier when compared to women (main effect of sex: F_1,79 = 5.412, p = 0.023). Importantly, at the end of the experiment, elbow height was still at least 10 mm over the mesh barrier (whole sample minimum), with an average elbow height of 96 ± 32 mm for women and 110 ± 32 mm for men. There was also a small but consistent increase in movement timing error with fatigue in both sex subgroups (main effect of Fatigue: F_1,79 = 10.833, p = 0.001).

Effects of movement repetitions and sex on upper body mean kinematic parameters

Non-Fatigue waveforms of the different joint angles (Fig. 3, left panels) show that elbow flexion and humerothoracic plane of elevation contributed importantly to the dynamic component of the reaching task because of their broad ranges of motion. Moreover, the large deviation from the anatomical position and the low range of motion observed for the humerothoracic elevation degrees of freedom (DoF) illustrate the important postural demands of the task at this joint. All trunk angles had a minimal contribution to the task before fatigue (low range of motion and angular position close to neutral). However, trunk contribution to the task increased with fatigue, while elbow and shoulder contributions decreased. This multi-joint movement reorganization with fatigue is highlighted by main effects of Fatigue for all joint angle variables (mean values) except for the average trunk axial rotation angle (Table 2).

		Mean values			Movement-to-movement variability (SD)
		Fatigue	Sex	Sex × Fatigue	Fatigue	Sex	Sex × Fatigue
Elbow flexion	Range of motion	F 1,79  = 8.82	F1,79 = 1.44	F1,79 = 0.17	F 1,79  = 22.77	F 1,79  = 5.75	F1,79 = 0.70
		p = 0.004	p = 0.234	p = 0.678	p < 0.001	p = 0.019	p = 0.407
	Average angle	F 1,79  = 8.27	F1,79 = 0.11	F1,79 = 0.57	F1,79 = 3.06	F 1,79  = 11.39	F1,79 = 0.19
		p = 0.005	p = 0.741	p = 0.451	p = 0.195	p = 0.001	p = 0.662
HT plane	Range of motion	F 1,79  = 6.70	F1,79 = 0.01	F1,79 < 0.01	F 1,79  = 24.65	F1,79 = 2.69	F11,79 = 0.38
		p = 0.011	p = 0.910	p = 0.986	p < 0.001	p = 0.105	p = 0.537
	Average angle	F 1,79  = 8.57	F1,79 = 2.98	F1,79 = 0.64	F 1,79  = 9.42	F1,79 = 1.22	F1,79 = 0.49
		p = 0.004	p = 0.088	p = 0.428	p = 0.003	p = 0.274	p = 0.484
HT elevation	Range of motion	F 1,79  = 25.89	F1,79 = 0.84	F1,79 = 0.96	F 1,79  = 21.39	F1,79 = 0.70	F1,79 = 0.50
		p < 0.001	p = 0.362	p = 0.331	p < 0.001	p = 0.404	p = 0.482
	Average angle	F 1,79  = 191.55	F 1,79  = 7.05	F 1,79  = 4.77	F 1,79  = 5.36	F 1,79  = 7.17	F 1,79  = 4.09
		p < 0.001	p = 0.010	p = 0.032	p = 0.023	p = 0.009	p = 0.047
Trunk lateral flexion	Range of motion	F 1,79  = 55.56	F1,79 = 0.48	F1,79 = 0.10	F 1,79  = 51.01	F1,79 = 1.31	F1,79 = 1.07
		p < 0.001	p = 0.489	p = 0.757	p < 0.001	p = 0.256	p = 0.305
	Average angle	F 1,79  = 108.36	F1,79 = 0.27	F1,79 = 0.81	F 1,79  = 47.25	F1,79 = 0.01	F1,79 = 0.04
		p < 0.001	p = 0.602	p = 0.128	p < 0.001	p = 0.944	p = 0.264
Trunk axial rotation	Range of motion	F 1,79  = 87.18	F1,79 = 0.31	F1,79 = 1.58	F 1,79  = 34.04	F1,79 = 0.13	F1,79 = 0.09
		p < 0.001	p = 0.581	p = 0.213	p < 0.001	p = 0.719	p = 0.765
	Average angle	F1,79 = 0.53	F 1,79  = 6.34	F1,79 = 0.61	F 1,79  = 26.55	F1,79 = 1.58	F1,79 = 0.28
		p = 0.470	p = 0.014	p = 0.437	p < 0.001	p = 0.212	p = 0.602
Trunk flexion	Range of motion	F 1,79  = 13.92	F1,79 = 0.10	F1,79 < 0.01	F 1,78  = 33.97	F1,78 = 0.03	F1,78 = 0.02
		p < 0.001	p = 0.753	p = 0.953	p < 0.001	p = 0.872	p = 0.878
	Average angle	F 1,79  = 53.25	F1,79 = 0.24	F1,79 = 1.61	F 1,78  = 42.78	F1,78 = 1.18	F1,78 = 0.38
		p < 0.001	p = 0.627	p = 0.208	p < 0.001	p = 0.281	p = 0.539

F and p values are presented for each two-way ANOVA (Sex [between subjects: women vs men] × Fatigue [repeated measures: Non-Fatigue vs Fatigue Terminal]). Itatic fonts present statistically significant effects (p<0.05)

HT humerothoracic

In addition to the various common fatigue-related kinematic changes between men and women, there was a Sex × Fatigue interaction for the average humerothoracic elevation angle. Post hoc analyses showed that while humerothoracic elevation decreased with fatigue for both men (− 9.8° ± 6.0°; t₃₉ = 10.305, p_corrected < 0.001) and women (− 7.2° ± 5.0°; t₄₀ = 9.211, p_corrected < 0.001), this decrease was more important in men. Indeed, men’s average humerothoracic elevation angle was lower than women’s at the end of the experiment (t₇₉ = 3.210, p_corrected = 0.004) while no differences were observed at Non-Fatigue (t₇₉ = 1.456, p_corrected = 0.149). Finally, women’s trunk was more rotated (10.9° ± 5.0°) than men’s (8.9° ± 4.4°) during the whole experiment, so that their reaching shoulder was more advanced.

Effects of muscle fatigue and sex on upper body movement-to-movement variability

Movement-to-movement variability (standard deviation) significantly increased with fatigue for all assessed variables except for the average elbow flexion angle (Table 2, Fig. 4). There was also a Sex × Fatigue interaction for the average humerothoracic elevation angle movement-to-movement variability as women became more variable with fatigue (t₄₀ = − 2.699, p_corrected = 0.030), but not men (t₃₉ = − 0.249, p_corrected = 0.804). In line with this interaction, average humerothoracic elevation angle standard deviation was greater for women than men once fatigued (t_77.4 = 3.241, p_corrected = 0.008), despite similar variability at the beginning of the task (t₇₉ = 1.004, p_corrected = 0.636). Women’s elbow flexion average angle and range of motion were more variable than men’s both before and after fatigue.

Responsiveness of kinematics variables to RPT-induced changes in men and women

While almost all joint angle variables changed with fatigue (main effects of Fatigue on 22/24 variables), the amplitude of those changes varied (Fig. 5). The most consistent change both in men (SRM [95% CI] = − 1.63 [− 2.11 to − 1.15]) and in women (SRM [95% CI] = − 1.44 [− 1.88 to − 1.00]) was a decrease in average humerothoracic elevation angle. Other variables that were highly responsive to RPT-induced changes (SRM > 0.8) were average trunk lateral flexion angle (mean and movement-to-movement variability [men and women]), trunk lateral flexion range of motion (mean [men and women] and movement-to-movement variability [women]), average trunk flexion (mean [men]), and trunk axial rotation range of motion (mean [men and women]).

Association between kinematic variables and endurance time

Significant correlations between Non-Fatigue motor behavior and endurance time were observed for only 4 out of 20 possibilities, and correlations were weak (Additional file 1: Table S1). For men, endurance time correlated significantly and negatively with Non-Fatigue mean (p = 0.036; r [95% CI] = − 0.334 [− 0.025 to − 0.584]) and movement-to-movement variability (p = 0.033; r [95% CI] = − 0.338 [− 0.023 to − 0.587]) of humerothoracic elevation range of motion. Movement-to-movement variability of shoulder average elevation angle (p = 0.032; r [95% CI] = − 0.339 [− 0.031 to − 0.589]) was also weakly and negatively related to endurance time in men. For women, only Non-Fatigue variability of trunk flexion (p = 0.032; r [95% CI] = 0.336 [0.032 to 0.584]) was (positively) associated with the time needed to reach a RPE of 8/10.

Men’s and women’s kinematic changes with fatigue during the repetitive pointing task

Many studies have previously shown that women are more fatigue-resistant when sustaining an isometric contraction or performing series of high-intensity intermittent contractions (reviewed in [12]). However, during dynamic tasks, sex differences in fatigability are less consistent [14, 22]. In the present study, endurance times, based on perceived fatigue, were similar between men and women. Nevertheless, mean humerothoracic elevation angle decreased more in men than in women. This may indicate that humerothoracic elevators reached a higher intensity of fatigue in male participants. It is known that shoulder elevation maximal force decreases by about 5% by the end of the RPT [24]. However, significant inter-individual differences (standard deviation = 170% of the mean reported changes) exist in the amount of muscle force decrease following the RPT. Those inter-individual differences may be critical when studying the effects of individual factors, such as participants’ sex. Hunter et al. [13] showed that women rated greater levels of perceived exertion than men for a similar time point relative to time to task failure during intermittent isometric elbow flexion (e.g., 75% of the time to task failure). Moreover, following fatiguing knee dynamic contractions, men’s maximal force decreased more than women’s despite similar subjective fatigue ratings [43]. In line with those results, it could be argued that men were closer to their actual time to task failure and had a greater force decrease than women at a RPE of 8/10 in our study. This could explain why men’s humerothoracic elevation decreased more than women’s. However, if sex differences in the RPE response curve were the sole factor explaining these results, we would expect Sex × Fatigue interactions on most of the assessed variables as evidence that men would indeed have greater fatigue-related changes than women. However, since sex differences in kinematic adaptations were observed only in a single DoF, it is unlikely that the choice of stoppage criterion explained all our findings.

Another specificity of the present report is the multi-joint dynamic nature of the motor task when compared to the single-joint isometric contractions reported in most of the previous studies comparing men’s and women’s fatigability [12]. Although a significant decrease in maximal force after the RPT was observed in shoulder elevators [24], it is likely that fatigue developed at different sites of the body simultaneously for some participants. For instance, previously observed increases in EMG activity of the trapezius, anterior deltoid, biceps brachii, and triceps brachii suggest some muscle fatigue for both the shoulder and elbow muscles during the RPT [22]. However, men and women differed on some variables such as elbow flexion variability as well as mean trunk rotation from the beginning of the RPT. This indicates that they were not doing exactly the same task, which may influence the structures most at risk of developing muscle fatigue. Moreover, the RPT may lead to a preferential development of muscle fatigue in different structures between sexes. Based on the literature on sex differences in fatigability, we may expect men being more challenged by the postural (isometric) component of the task [12]. Indeed, we observed a greater decrease in humerothoracic elevation, a DoF mostly involved in the postural component of the task, in male participants. Moreover, three variables describing baseline behavior at this DoF out of four were associated with endurance time, but only in men. Men’s greater sensibility to fatigue induced by the postural component of the RPT may be related to their lower concentration of type 1 muscle fibers and lower muscle perfusion [12]. The greater level of muscle fatigue experienced by men in the muscles involved in humerothoracic elevation may also be related to anthropometric differences. Men being heavier and having heavier upper limbs, the absolute force they must produce during the RPT is larger.

Men’s and women’s variability during the repetitive pointing task

Most kinematic adaptations during the RPT were similar between men and women. However, in addition to the greater decrease in humerothoracic elevation with fatigue for men when compared to women, changes in humerothoracic elevation variability also differed between sexes. It increased with fatigue, but only in women. This suggests that men and women use slightly different kinematic strategies to adapt to muscle fatigue, with men modifying their mean movements to a greater instance and women increasing movement-to-movement variability. By decreasing mean humerothoracic elevation angle, men may reduce the amount of force produced by humerothoracic elevators. On the other hand, women’s strategies may be to increase movement-to-movement variability to spread the load necessary to perform the RPT across redundant structures from one repetition to the other. In their recent study on EMG data, Srinivasan et al. [22] also showed sex differences in motor adaptations during the RPT. Men increased more their upper trapezius muscle EMG variability than women. Moreover, they showed that with fatigue, biceps EMG movement-to-movement variability increased in women but decreased in men. Based on those results, the authors suggested that women used a more elbow-based strategy to maintain their task performance despite fatigue while men’s strategy targeted more the shoulder. We did not find any evidence of the more elbow-based strategy in women from our kinematic data. The relationships between EMG data and movement kinematics are not straightforward, and many methodological [44], physiological [45], and biomechanical [46] factors can alter them. For instance, the biceps brachii, for which different fatigue-related adaptations have been shown between men and women [22], is a bi-articular muscle. Changes in its activation can have an impact at the elbow and/or at the shoulder joint. Nevertheless, the results of both the current study and that of Srinivasan et al. point to the same conclusion that men and women’s fatigue adaptations, although similar, are not exactly identical [21, 22].