Fatigue is a phenomenon that dysregulates physiological systems, including the autonomic nervous system (ANS) (Tanaka et al., 2011; Yoshiuchi et al., 2004). Heart rate variability (HRV), reflecting ANS activity, serves as a potential biomarker for fatigue (Escorihuela et al., 2020; Matuz et al., 2022; Ni et al., 2022). Compared to individuals with lower levels of physical activity, fatigue may exhibit distinctive HRV characteristics in physically active individuals, attributed to their unique physiological adaptations and response patterns to physical stress (Hegde et al., 2018). Physically active individuals present a prime model for studying these changes as their HRV responses are indicative of their ANS's adaptability to stress and resilience to fatigue (Marcora et al., 2009). Understanding these responses in a physically active population can provide insights into optimizing training regimens and recovery strategies (Seiler et al., 2007).

In addition, Fatigue-induced alterations in ANS activity increase the risk of life-threatening events and decrease performance in daily activities (Marcora et al., 2009; Okutucu et al., 2011). Fatigue affects the balance between the sympathetic nervous system (SNS) and the parasympathetic nervous system (PNS) (Castillo-Aguilar et al., 2021; Hebisz et al., 2020; Matuz et al., 2021; Mizuno et al., 2011) by altering the activation of brain regions that contribute to sympathetic-parasympathetic balance (e.g., the dorsal anterior cingulate cortex) (Critchley et al., 2003; Tanaka et al., 2015). Thus, assessing the effect of different fatigue types on ANS activity is essential to improve autonomic function, reduce fatigue-related morbidity and mortality, as well as optimizing recovery strategies.

Fatigue is classified as physical fatigue (PF), mental fatigue (MF), and physical-mental fatigue (PMF) (Kume et al., 2017), which are induced by different factors and exercise loads, including prolonged physical activity, demanding cognitive tasks, and insufficient sleep (Dutheil et al., 2012; Maynard et al., 2021; Russell et al., 2020; Schmitt et al., 2013). Furthermore, different types of fatigue may affect ANS activity differently (András et al., 2021; Dobbs et al., 2020; Qin et al., 2021). For instance, compared with the control (absence of fatigue), MF changes PNS parameters by increasing the root mean square of successive differences (RMSSD) between normal heartbeats (Matuz et al., 2021; Qin et al., 2021), while PF has the opposite effect (Dobbs et al., 2020; Hebisz et al., 2020). However, previous studies focused primarily on the effects of PF and MF on the ANS (Castillo-Aguilar et al., 2021; Dobbs et al., 2020; Hebisz et al., 2020; Matuz et al., 2021; Mizuno et al., 2011), and little is known about the effects of PMF on ANS function. Although MF and PF affect the ANS by stimulating or inhibiting PNS activity, respectively, the activation of the PNS and SNS is not linearly correlated. Therefore, the effects of PMF on the regulation of ANS activity need to be further explored.

This study evaluated the effects of PMF on ANS function and compared the effects of PF, MF, and PMF on ANS activity. Thus, models of these types of fatigue in healthy younger adults were constructed. ANS activity was assessed by measuring HRV before and after the induction of fatigue. We hypothesized (1) that the induction of PF, MF, and PMF resulted in a corresponding increase in subjectively fatigue levels respectively (e.g., physical fatigue without mental fatigue, mental fatigue without physical fatigue, and physical-mental fatigue) and (2) that different types of fatigue have a different effect on ANS activity. Specifically, PMF and PF leads to high level of sympathetic activity while MF leads to high level of parasympathetic activity.

Forty-five physically active individuals from Beijing Sport University voluntarily participated in the study. Participants were included if they engaged in moderate to vigorous physical activity exceeding six hours weekly (Gianola et al., 2017), and their physical activity habits were assessed using the International Physical Activity Questionnaire (IPAQ) (Craig et al., 2003). The average weekly physical activity was 14.47 ± 3.23 h, spanning 4.27 ± 1.46 days, with a mean MET value of 4919.06 ± 3021.50 per week. Average sedentary and sleep durations were 6.13 ± 1.93 hours and 7.45 ± 0.78 h, respectively. All participants gave written informed consent. The exclusion criteria were subjects with injuries that affected the implementation of fatigue-inducing protocols and subjects with burnout. Nine participants were excluded, and 36 (17 females) were included in the study (Table 1). This study was approved by the Human Subjects Review Committee of Beijing Sport University (2021004H), and all procedures were conducted according to the Declaration of Helsinki.

Participants underwent four separate visits, each comprising a specific fatigue-inducing protocol (PF, MF, or PMF) followed by a peak power output assessment. The order of protocols, depicted in Figure 1, was interspersed with a minimum seven-day interval to facilitate complete recovery. This study was conducted from December 2020 through June 2021. To gauge subjective physical and mental fatigue levels, the Rating of Perceived Exertion (RPE) scale and the Visual Analog Scale-Fatigue (VAS-F) were employed. Prior to each session, participants were advised to avoid strenuous physical activities and to refrain from consuming alcohol or caffeine for 24 hours.

PF-inducing protocol: 60-min cycling exercise

The 60-min cycle ergometer exercise included three 20-min blocks with a 10-s interval between blocks, aimed at assessing to the subjective levels of both physical and mental fatigue (Figure 1). The load was 60% of the peak power and remained constant throughout the protocol. Before the experiment, the peak power output of each participant was measured using a cycle ergometer (Ergoselect 100, Ergoline GmbH, Bitz, Germany). The protocol included 1) a warm-up (cycling for 3 min at 30 W and 70 ± 5 rpm) and 2) initial loads of 50 W (males) and 30 W (females) with an increase at a rate of 20 W per 2 min until the test ended or the participant interrupted the test for any reason. The test was terminated when at least two of the following conditions were met: 1) HR ≥ 85% × maximum heart rate (measured using the formula: 208 - [age × 0.7]); (2) RPE ≥ 7; and (3) failure to maintain a speed of 70 ± 5 rpm for more than 10 s. The peak power output and the loads used to induce PF and PMF in both sexes are shown in Table 1.

MF was induced using the 60-min AX-Continuous Performance Task (AX-CPT), which was generated and implemented using E-prime 2.0 (Sharpsburg, USA). The task consisted of three 20-min blocks with a 10-s interval between blocks based on four uppercase letters (A, B, X, Y). The task was presented on a 23-inch monitor (Lenovo L12364A, 1920 × 1080 pixels, 60 Hz refresh rate). Before the experimental test, the participants performed a 3.5-min familiarization task (60 trials) consisting of 42 target pairs (70%, AX), six valid cue-invalid probe pairs (10%, AY), six invalid cue-valid probe pairs (10%, BX), and six invalid cue-invalid probe pairs (10%, BY). Each block of the experimental task (total of 340 trials) consisted of 238 AX (70%), 34 AY (10%), 34 BX (10%), and 34 BY (10%). The letters appeared on the screen randomly. One trial of AX-CPT is presented in Figure 2. Participants were instructed to press the target button (left arrow) as soon as the X probe appeared after observing the A cue; while observing any other letter pair, the participants pressed the right arrow button as soon as possible. Responses to probe stimuli were recorded with a time limit of 1500 ms.

PMF was induced by simultaneously performing AX-CPT and cycling exercises, the same protocols used to induce MF and PF, respectively.

HRV and HR

HRV was assessed before and immediately after each task (Figure 1). Participants were asked to sit still for at least 5 min before the test. Then, HRV data were recorded continuously for 3 min in a sitting position using a cardiac chest band (Movesense OP174, Suunto, Vantaa, Finland) connected to an app (Firstbeat Sports Coach, Jyväskylä, Finland) running on a mobile device (iPad, Apple) for monitoring and recording HR throughout the experiment (Chen et al., 2020; Wu et al., 2020). Recordings were conducted in a quiet, temperature-controlled room. Participants were seated comfortably and asked to breathe naturally with their eyes closed.

This scale was developed by Lian et al. (2005) based on the Maslach Burnout Inventory and consists of 20 items that evaluate college student burnout in three domains: dejection (eight items), improper behavior (six items), and reduced personal accomplishment (six items). Items were rated on a Likert scale ranging from one (strongly disagree) to five (strongly agree). Learning burnout and its dimensions were measured as mean scores with a cut-off point of three.

Subjective PF was assessed using the Borg CR-10 on a scale from 0 to 10 (Borg, 2010) and was defined as an RPE score of at least 5 (Ribeiro et al., 2008). Participants were instructed to complete the RPE scale four times: at baseline and immediately after each block (Figure 1).

The level of MF was measured using the VAS-F (Veness et al., 2017), which is a 100-mm horizontal line with the marks ‘none’ (not fatigued at all) and ‘maximum’ (extremely fatigued) on the left and right, respectively. Participants were instructed to mark the line based on their subjective MF. The VAS score was the distance between the marked point and the left endpoint, and a VAS-F value of at least 50 mm was defined as the presence of MF (Dernis-Labous et al., 2003).

AX-CPT performance

Cognitive performance in AX-CPT was recorded automatically by measuring the accuracy and reaction time for four probe pairs (AX, AY, BX, and BY). The average accuracy for these probe pairs every 5 min was measured as cognitive load and mental status during MF- and PMF-inducing tasks.

Cardiac autonomic nervous function was evaluated using the time and frequency domains of HRV. HRV measurements were analyzed utilizing the Firstbeat system (https://sports.firstbeat.com/, Jyväskylä, Finland). The time domain consisted of the standard deviation of all RR intervals (SDRR) and the RMSSD in RR intervals. The frequency domain consisted of low-frequency spectral power (LF; 0.05-0.15 Hz), high-frequency spectral power (HF; 0.15-0.4 Hz), and the LF/HF ratio. The normalized units for LF and HF served as indices of autonomic regulation in the participants, with the following formulas:

To assess the effect of fatigue types on cardiac autonomic activity, RMSSD and nHF were utilized as primary outcomes of parasympathetic activity (Laborde et al., 2017), and SDNN served as a secondary outcome (Yılmaz et al., 2018). The LF is indicative of both sympathetic and parasympathetic activity, with a strong propensity towards sympathetic activity (Hayano and Yuda, 2021). Additionally, the nLF/nHF ratio reflects sympathovagal balance (Laborde et al., 2017).

Physiological responses to fatigue-inducing protocols were evaluated by measuring HR. The effect of PF- and PMF-inducing protocols on internal and external load, respectively, was assessed by measuring training impulse (TRIMP) every 5 min using the formula:

A power analysis was conducted using G*Power 3.1 (parameters: one-way repeated measure ANOVA, within factors, f = 0.25, P = 0.05, power = 0.90, number of groups: 3, number of measurement points: 4) yielding a sample size of 30 participants. For the present study, 36 participants were recruited and included in the final analyses.

Data were expressed as means ± standard deviations (M ± SD). Statistical analyses were conducted using SPSS version 25.0 (SPSS Inc., Chicago, USA). The normality of the distribution of continuous variables was assessed using the Shapiro-Wilk test.

The effects of PMF on ANS activity were examined using a paired-samples t-test. The dependent variables were the outcomes of HRV (RMSSD, SDNN, nLF/nHF, nLF, and nHF). This research undertook a stratified analysis to investigate the differential impacts of fatigue on autonomic nervous system activity across genders.

The effects of PMF and PF on ANS activity were compared using two-way analysis of variance (ANOVA). The dependent variable encompassed the percentage changes in HRV parameters (ΔRMSSD, ΔSDNN, ΔnLF, ΔnHF, and ΔnLF/nHF), with fatigue type and gender serving as independent variables. RPE was included as a covariate because fatigue might affect ANS activity. Similarly, VAS-F (subjective level of mental fatigue) was included as a covariate to analyze differences in the effects of MF- and PMF-inducing protocols.

The effectiveness of the fatigue models (PF, MF, and PMF) was evaluated by one-way repeated-measures ANOVA. The dependent variables of the models were RPE and VAS-F. The effects of cycling exercises alone and AX-CPT combined with cycling on heart rate and TRIMP, and the effects of AX-CPT alone and AX-CPT combined with cycling on cognitive task performance (i.e., accuracy) were assessed using a two-way repeated-measures ANOVA.

Statistical significance was defined as P < 0.05. The effect size was calculated using Cohen’s d index and classified as small (d = 0.2), medium (d = 0.5), or large (d = 0.8) (Jacob, 1992).

The participants completed all study visits. There were no significant differences in subjective fatigue levels (RPE and VAS-F) and HRV parameters (SDNN, RMSSD, nLF/nHF, nLF, and nHF) between the three protocols at baseline. PF- and PMF-inducing protocols significantly increased RPE scores (F = 169.463, P < 0.001, η² = 0.939; F = 222.803, P < 0.001, η² = 0.953) and led to extreme physical exertion (Figure 3A). MF- and PMF-inducing protocols significantly increased VAS-F scores (F = 56.171, P < 0.001, η² = 0.836; F = 140.819, P < 0.001, η² = 0.928) and caused severe mental fatigue (Figure 3B). These results suggested that the PF, MF, and PMF models were effective. The peak power and experimental load (60% of the peak power) were 216.32 ± 37.15 W and 129.74 ± 22.30 W for males and 154.71 ± 30.44 W and 92.82 ± 18.26 W for females (Table 1).

The paired-samples t-test showed that PMF significantly decreased the RMSSD, SDNN, and nHF, while increasing nLF and the nLF/nHF ratio compared to baseline (Table 2). However, there were no significant differences observed in the nLF, nHF, and the nLF/nHF ratio for male participants, nor in the SDNN for female subjects.

An investigation into the influence of fatigue type (PMF vs. PF) and gender (male vs. female) on HRV parameters was performed using a two-way ANOVA. M and SD for ΔRMSSD, ΔSDNN, ΔnLF, ΔnHF, and ΔnLF/nHF, are presented in supplementary Table 3. A significant main effect of gender was found on ΔnHF (F = 4.288, P < 0.05, η² = 0.079), but not others (see supplementary Table 4). For kinematic parameters, two-way repeated-measures ANOVA indicated no significant interactions between fatigue type and exercise block for average HR (F = 1.744, P = 0.060; see Figure 4A) and TRIMP (F = 0.929, P = 0.511; Figure 4B). Heart rate (F = 43.245, P < 0.001, η² = 0.888) and TRIMP (F = 12.048, P < 0.001, η² = 0.692) increased gradually during PF- and PMF-inducing tasks, without significant difference between these tasks (F = 1.600, P = 0.210; F = 1.814, P = 0.182).

In examining the effects of fatigue type (PMF and MF) and gender (male and female) on HRV paraments, a two-way ANOVA was conducted. There was a significant main effect for fatigue type on ΔRMSSD (F = 20.809, P < 0.001, η² = 0.431), ΔSDNN (F = 13.361, P < 0.001, η² = 0.327), but not ΔnLF (F = 3.107, P = 0.083), ΔnHF (F = 1.840, P = 0.180) and ΔnLF/nHF (F = 2.493, P = 0.119). The main effect for gender as well as the interaction effect between fatigue type and gender were not significant. Post-hoc analyses indicated that ΔRMSSD and ΔSDNN were significantly lower in the PMF model than in the MF model (Figure 5). Two-way repeated-measures ANOVA indicated no significant interactions between fatigue types and exercise block for cognitive task performance (F = 1.064, P = 0.388) (Figure 4C). Cognitive performance (i.e., accuracy in the AX-CPT) decreased gradually during MF- and PMF-inducing protocols (F = 5.971, P < 0.001, η² = 0.531) and was lower in the latter (F = 7.643, P < 0.01, η² = 0.101).

This study evaluated the effects of PF, MF, and PMF on HRV, marking the first to assess these specific fatigue types on autonomic nervous system (ANS) activity. Our tasks induced distinct fatigue states, confirmed by subjective, physiological, and cognitive measures. Regarding ANS activity, PMF closely resembles PF by enhancing cardiac sympathetic activity while diminishing parasympathetic activity. PMF decreased RMSSSD, SDNN more strongly than MF, indicating that different fatigue types influenced cardiac ANS activity differently.

Performing a cycling exercise at 60% of the peak power output for 60 min strongly increased RPE and marginally increased VAS-F, suggesting the successful induction of PF without MF (VAS-F < 50), in line with a previous study (Davidow et al., 2020), wherein prolonged high-intensity intermittent running induced PF without MF in rugby union players. In addition, we observed a linear increase in HR and TRIMP, suggesting that an increase in physical workload was consistent with the time-on-protocol.

Performing 60-min AX-CPT induced MF without PF (RPE < 5), evidenced by a strong increase in subjective scores (VAS-F) and a significant decrease in cognitive task performance (accuracy). Previous research has shown that 60-min AX-CPT can effectively induce mental fatigue in healthy college students (VAS > 60) without physical fatigue (VAS < 40) (Loch et al., 2020). Moreover, participants’ responses became gradually faster as accuracy decreased; this effect was also observed during the execution of the PMF-inducing protocol, consistent with previous studies (Delliaux et al., 2019; Thomson et al., 2009). Decreased accuracy may be due to a loss of motivation and low levels of self-efficacy due to MF (Loch et al., 2020) and an increase in wrong responses due to short reaction times (Thomson et al., 2009).

AX-CPT combined with cycling exercise strongly induced PMF, demonstrated by increased subjective PF and MF, increased workload consistent with the time-on-task, and decreased cognitive task performance. The effectiveness of this model in inducing PM and MF has been validated previously (Barzegarpoor et al., 2020; Haji Agha Bozorgi et al., 2021). For instance, Barzegarpoor et al. (Barzegarpoor et al., 2020) found that PF and MF levels increased significantly (RPE > 16, NASA-TLX > 60) in male recreational cyclists after 45-min AX-CPT combined with cycling exercises.

The effect of different types of fatigue on cardiac autonomic activity was assessed by measuring HRV. The results showed that PF remarkably decreased RMSSD, SDNN, nHF, and nLF, suggesting decreased parasympathetic activity. Meanwhile, the increase in the nLF/nHF ratio, an indicator of sympathovagal balance, suggests higher sympathetic activity and lower parasympathetic activity (Michael et al., 2017), in line with a previous study (Dobbs et al., 2020) showing a significant decrease in the time and frequency domains at 30 min after high-volume back squats. In addition, cycling exercise increases HR linearly because of higher sympathetic activity to maintain physical performance (Borresen and Lambert, 2008). Long-term physical exercise-induced PF increases cardiac sympathetic activity and parasympathetic withdrawal.

Regarding MF, the results showed that SDNN, nLF, and nLF/nHF were increased 5 min after 60-min AX-CPT, suggesting that increased cardiac parasympathetic activity is a feature of MF. Moreover, the slight decrease in HR in each exercise block during the MF-inducing task indicates an increase in cardiac parasympathetic modulation (Hallman et al., 2019), in line with previous studies (Matuz et al., 2021; Qin et al., 2021). The previous study reported that healthy college students had a significant increase in parasympathetic activity (lnRMSSD, pNN50, lnHF, lnLF, and SD2) after completing a 1.5-h dual 2-back task (Matuz et al., 2021). However, Delliaux et al. (Delliaux et al., 2019) found that most HRV paraments were remarkedly decreased (a significant decrease in time domain measures and a slight increase in the LF/HF ratio) during a 1-h task-switching task, indicating either decreased parasympathetic activity or increased sympathetic activity. This inconsistency might be because HRV was measured under fatigue conditions at different time points (i.e., during or after cognitive task completion).

Consistent with our hypothesis, our results showed that fatigue induced by physical and cognitive load had a markedly different effect on cardiac autonomic function. More specifically, PF increased sympathetic activity and decreased parasympathetic activity, while MF exhibited the opposite effect. These observations were corroborated by variations in HR following and after different types of stimulation (i.e., physical and mental workloads). In comparison to the baseline, HR exhibited a significant increase during physical load and PF scenarios, but a minor decrease during mental load and MF scenarios, suggesting that different types of load-induced fatigue elicit unique patterns of ANS activation. There changes in HR and ANS might be related to the activation of the central autonomic network, which controls sympathetic-parasympathetic balance (Tanaka et al., 2015), such as the dorsal anterior cingulate cortex (ACC) controlling sympathetic output (Critchley et al., 2003). Previous studies have shown that higher parasympathetic activity under MF is closely associated with decreased activity in the medial prefrontal cortex and ACC (Caseras et al., 2008; Lorist et al., 2005). In contrast, Liu et al. (Liu et al., 2007) found that PF increased activity in ipsilateral sensorimotor, prefrontal, orbitofrontal, and anterior cingulate regions. Another explanation for this finding is that moderate-high-intensity physical exercise increases sympathetic activity as a result of arterial baroreflex control and the secretion of epinephrine and norepinephrine, increasing HR to meet the demands of working muscles (maintain physical performance), increase peripheral blood flow (for heat dissipation), and maintain the function of other organs (Hebisz RG et al., 2020). However, acetylcholine release is increased during cognitive task performance (Passetti et al., 2000); this neurotransmitter binds to muscarinic receptors and augments parasympathetic activity, leading to a slight decrease in HR (Shaffer et al., 2014).

Regarding PMF, the 60-min AX-CPT combined with cycling exercises significantly decreased RMSSD, nLF, and nHF and significantly increased the nLF/nHF ratio, suggesting that PMF increased cardiac sympathetic activity and decreased parasympathetic activity. Nonetheless, changes in HRV paraments were similar between PF and PMF, which was unexpected but in line with a previous study (Garde et al., 2002). This result may be related to the regulation of sympathetic and parasympathetic nerves. Parasympathetic nerves on the sinoatrial node, atrial myocardium, and atrioventricular node are activated at rest, while sympathetic nerves are activated during physical activity (Borresen and Lambert, 2008). Sympathetic activation triggers the release of norepinephrine and epinephrine (Shaffer et al., 2014), increasing HR and cardiac contractility to maintain exercise performance. Therefore, sympathetic nerves may play a dominant role during moderate-high intensity physical activity, regardless of whether cognitive tasks are performed. Sympathetic drive persists for some time despite parasympathetic activation to facilitate recovery from medium-high intensity exercises in which sympathetic activity predominates (Borresen and Lambert, 2008).

The effects of MF and PMF on ΔRMSSD, ΔSDNN, ΔnLF, ΔnHF, and ΔnLF/nHF were distinct, suggesting that these protocols influenced cardiac ANS activity differently. PMF increased sympathetic activity and decreased parasympathetic activity, while MF had the opposite effect. One possible explanation for this effect is that sympathetic activity predominates during the dual task to maintain physical performance. In addition, participants tend to maintain exercise performance at the expense of cognitive performance (accuracy), and the latter was significantly lower during the dual task than during AX-CPT.

The study yielded a notable observation that autonomic nervous activity differed among participants based on gender, which was in line with the previous studies (Voss et al., 2015; Woo and Kim, 2015). It was observed that males exhibited a less significant reduction in nHF power in response to fatigue when compared to females. While the precise reasons for this discrepancy are challenging to ascertain, it may be attributed to inherent physiological differences between genders, notably in hormonal responses (Voss et al., 2015). Such differences could potentially influence cardiac autonomic regulation and stress adaptation.

This study analyzed the effects of PF and MF on cardiac autonomic activity under rigorously controlled conditions, including protocol duration and respiratory control during HRV measurements. These findings elucidated how different types of fatigue affected cardiac autonomic activity and showed that HRV was useful for assessing fatigue. For workers with physically and mentally demanding jobs, investigating the association between ANS and PMF is critical to maintaining job performance and reducing fatigue-related morbidity from work-related musculoskeletal disorders (Mehta and Agnew, 2011).

This study has some limitations. First, since the study implemented three distinct fatigue-inducing protocols, measuring HRV during task execution under identical control conditions (i.e., sitting still for at least 5 min) was unfeasible. Second, the effects of long-term fatigue on HRV were not assessed. Third, the effects of fatigue on ANS function may be affected by fatigue-inducing tasks not encountered in real-life situations. Therefore, the association of HRV with ANS function and real-life long-term fatigue needs to be further researched.

Our findings suggest that PF and PMF increased sympathetic activity and decreased parasympathetic function. Further, MF increased parasympathetic activity more strongly than PF and PMF.