Journal Logo


Neuromuscular Fatigue of Cycling Exercise in Hypoxia


Author Information
Medicine & Science in Sports & Exercise: September 2020 - Volume 52 - Issue 9 - p 1888-1899
doi: 10.1249/MSS.0000000000002331


Neuromuscular fatigue can be defined as the progressive change that occurs in the central nervous system and/or muscles from exercise, resulting in a force output that is less than anticipated for a given voluntary contraction or stimulation (1). Although research has largely focused on single-joint isometric exercise tasks (2), dynamic exercise involving large muscle groups, such as cycling or running, have recently been receiving more attention because of their ecological relevance. Nevertheless, moving subjects from the cycle ergometer/treadmill to an isometric ergometer to assess fatigue inevitably results in a delay between exercise cessation and fatigue evaluation, which permits partial recovery of neuromuscular function (3,4) and may lead to fatigue misinterpretation. To circumvent this limitation, we developed an instrumented cycle-ergometer that allows fatigue measurement for the first time with no delay (5). Its special locking system permits the locking of the pedals instantly at a fixed position, allowing the exploration of fatigue throughout exercise and right at exhaustion (EXH). This device significantly reduces the delay typically observed in previous literature and accurately quantifies fatigue magnitude and etiology.

Besides task dependency, neuromuscular fatigue is thought to be affected by environmental conditions such as hypoxia (6). Indeed, there seems to be an earlier development of fatigue under hypoxic conditions, which may result in greater peripheral fatigue in time-matched exercise (7–9). Furthermore, altitude level seems to affect/alter fatigue etiology. In Amann et al. (10), subjects cycling to EXH (81% of normoxic peak power output) in normoxia, moderate hypoxia (FIO2 = 0.15; end-exercise SpO2 = 82%) and severe hypoxia (FIO2 = 0.10; end-exercise SpO2 = 67%). Peripheral fatigue, assessed with peak twitch evoked by peripheral nerve stimulation on relaxed muscles, in normoxia and moderate hypoxia at EXH was significantly greater than in severe hypoxia. Additionally, Millet et al. (11) showed a performance reduction (i.e., time to EXH) in submaximal exercise of the elbow flexors in severe hypoxia (FIO2 = 0.09; SpO2 = 75%) compared with normoxia, whereas the working muscles were held ischemic, that is, were in the same metabolic conditions in both conditions.

This evidence suggests that severe hypoxia impairs performance through a brain-hypoxic effect, independent of peripheral feedback from the locomotor muscles. Thus, performance switches “from a predominantly peripheral origin of fatigue to a hypoxia-sensitive central component of fatigue” (10).

Although the brain-hypoxic effect has been associated with a SpO2 below 75% (10), hypoxia is usually induced via decreased FIO2. As a result, this proposed threshold arises from mean SpO2 group values, whereas the interindividual responses variability for a given FIO2 is known to be particularly high (see, for example, Jubeau et al. [12]). Clamping SpO2 individually at distant levels of hypoxemia (i.e., moderate vs severe hypoxemia) and keeping it constant throughout exercise seems to be an attractive approach to test the validity of an altitude-severity fatigue etiology threshold. Indeed, a reduction in SpO2 as a response to the lowering of FIO2 is observed both at rest and during constant cycling exercise (13,14). Additionally, exercise performance is affected by SpO2—a decline in SpO2 lowers self-selected power output and faster rate of desaturation exacerbates that effect (14).

The previously mentioned studies selected the intensity of the fatigue sessions in hypoxia based on an incremental test performed in normoxia. Due to the known deleterious effect of altitude on aerobic power, the fatigue session in hypoxia is carried out at a higher relative intensity than in normoxia, resulting in shorter exercise duration. This is problematic because exercise duration and intensity influence fatigue etiology (15,16). Consequently, it is not possible to infer the independent effect of hypoxia on central versus peripheral fatigue. To circumvent this issue, Jubeau et al. (12) assessed neuromuscular fatigue after three 80-min bouts of cycling exercise in normoxia and severe hypoxia (FIO2 = 0.12; SpO2 = ~75%) at the same relative intensity (45% peak power output). The authors reported similar central and peripheral fatigue between the two conditions. Despite the unquestionable relevance of this article to understand fatigue after prolonged exercise in hypoxia, some issues are yet to be addressed.

First, SpO2 in hypoxia was approximately 75% (mean group value), which is exactly the threshold thought to trigger a switch between the dominant mechanisms controlling central motor output (10). Thus, further decreasing SpO2 may lead to different fatigue magnitude and etiology. Second, arterial O2 saturation variability might have blunted the effects of severe hypoxia, as some subjects desaturated less than others as discussed above. Third, the effect of hypoxia on prolonged exercise at relative intensities remains to be explored when conducted to EXH since it was not the case in Jubeau et al. (12) Lastly, no study has ever assessed central and peripheral contributions to fatigue throughout exercise, and previous evidence might have misinterpreted exercise-induced neuromuscular fatigue due to the delay to measure fatigue during cycling and at EXH.

Thus, this study aimed to evaluate fatigue development and its etiology during cycling and, for the first time, immediately after EXH in normoxia, moderate, and severe hypoxia at relative and absolute exercise intensities. We hypothesized that: (a) fatigue development would be greater in hypoxic absolute sessions compared with normoxia and dependent on the level of hypoxia; (b) fatigue development and time to EXH (TTE) of the hypoxic relative sessions would not be affected by the level of hypoxia; (c) at EXH, peripheral fatigue would be attenuated in severe (but not moderate) hypoxia, and greater for absolute versus relative exercise intensities; (d) on the contrary, for a given absolute or relative exercise intensity, severe systemic hypoxic stress would per se induce greater central fatigue.


Subjects and Ethical Approval

Thirteen recreationally active men (mean ± SD: age, 27 ± 8 yr; height, 177 ± 6 cm; mass, 75 ± 8 kg; V˙O2peak, 55 ± 9 mL·kg−1·min−1) with no signs or symptoms of neurological, cardiovascular, circulatory, or orthopedic disorders were recruited for the study. Before participation, written informed consent was obtained from all subjects. The study was approved by the ethics board of the University of Calgary (REB 15–2431) and performed in accordance with the 2008 edition of the Declaration of Helsinki. Subjects refrained from any type of physical exercise at least 2 d before testing and did not drink any beverages with caffeine on test days.

Experimental Design

The experimental protocol is shown in Figure 1. In total, subjects completed 8 exercise sessions on a customized electromagnetically braked recumbent cycle ergometer (5): three maximal incremental tests to determine peak V˙O2 (V˙O2peak) and maximal aerobic power output (Wmax), and five fatigue tests to assess neuromuscular fatigue before, during and after cycling. The three maximal incremental tests were separated by at least 48 h, randomly ordered and performed in three different environmental conditions: normoxia (NOR), moderate hypoxia, and severe hypoxia (see below). At least 48 h after completing the last maximal incremental test, subjects performed five fatigue tests in random order interspersed with at least 72 h. Three of the five fatigue tests were performed at an absolute power output in NOR (SpO2 ≥ 98%), moderate hypoxia (MODABS; target SpO2 = 85%), or severe hypoxia (SEVABS; target SpO2 = 70%); that is, the intensity of the fatigue tests was selected based on the Wmax attained in the maximal incremental test performed in normoxia. The two remaining fatigue tests were performed at a relative power output in moderate (MODREL; target SpO2 = 85%) and severe hypoxia (SEVREL; target SpO2 = 70%); that is, the intensity was selected based on the Wmax attained in the maximal incremental test performed in the same hypoxic condition. Subjects were blind to the environmental condition and the power output of the sessions (i.e., ABS or REL).

Panel A illustrates a typical fatigue session. Panel B shows the measurements carried out in each neuromuscular function assessment. NMF, neuromuscular fatigue.


Maximal incremental tests and familiarization

Before the start of the test, subjects were asked to remain relaxed and motionless on the cycle ergometer for 3 min for measurement of resting cardiorespiratory responses. Then, a 2-min period of cycling at very low wattage (20 W) was performed, which was followed by 3 min of warm-up at 50 W. To account for the altitude at which the experiment was carried out (Calgary, 1045 m), FIO2 was held constant at 24% for the normoxic conditions tests. In the hypoxic conditions, the unloaded cycling and warm-up aimed at ensuring that FIO2 was progressively adjusted to the target SpO2 before the actual start of the maximal incremental test. No rest was allowed between unloaded cycling, warm-up, and the start of the fatigue test. The actual test then commenced at 90 W in normoxia, 75 W in moderate hypoxia, and 60 W in severe hypoxia, and comprised 25-W increments every minute until volitional EXH (voluntary stop of pedaling despite verbal encouragement). At the first session, subjects pedaled at a self-selected pace, which was held constant throughout the experiment. In the last two sessions, a familiarization to neuromuscular function testing procedures was carried out after the maximal incremental tests. The familiarization comprised knee extensor isometric voluntary contractions (maximal and submaximal) on the customized cycle ergometer with and without femoral nerve stimulation and transcranial magnetic stimulation (TMS). As soon as subjects performed the neuromuscular function evaluation consistently, the same evaluation was repeated immediately after a cycling sequence; that is, after locking the pedals.

Fatigue tests

After baseline neuromuscular function evaluation, the exposure to normoxia, moderate or severe hypoxia was initiated, and subjects completed 3 min on the cycle ergometer where they were asked to sit relaxed for measurement of resting cardiorespiratory responses and quadriceps muscle oxygenation. This resting period was followed by 2 min of cycling at very low wattage (20 W) and by a 3-min warm-up (30% Wmax attained in the same environmental condition). As for maximal incremental tests, the unloaded cycling and warm-up in the hypoxic conditions aimed at ensuring that FIO2 was progressively adjusted to the target SpO2 before the actual start of the fatigue test. No rest was allowed between unloaded cycling, warm-up, and the start of the fatigue test. The fatigue tests were performed to EXH (voluntary stop of pedaling despite verbal encouragement) with an initial power output set at 55% Wmax determined either in normoxia or in the same environmental condition and held constant for 1 h. In case subjects did not reach EXH in the first hour, the intensity was increased every subsequent 10-min stage by 5% Wmax until EXH.

Neuromuscular function evaluation

Neuromuscular function evaluations were performed on the knee extensors of the right leg before, during, and at EXH of the fatigue tests. Two evaluations were completed before cycling, with 2 min of rest in between. During cycling, evaluations were performed every 5 min for the first 20 min, then every 10 min and finally immediately at EXH. For each evaluation, pedals were locked, and subjects acquired as fast as possible the correct position for testing, which consisted of maintaining a 90° angle at the ankle and knee joint and 100° at the hip joint. This position was acquired to match closely that traditionally held on an isometric chair to assess knee extensor force. Subjects started the evaluation within 1 s after stopping cycling. The evaluation started with a single knee extensor isometric maximal voluntary contraction (MVC) for 5-s, followed by a voluntary reduction in force to match 75% MVC and then 50% MVC with no rest period in-between (3). At each force level, a TMS stimulus was delivered. After each TMS stimulation, subjects were asked to re-contract to the same force level as fast as possible, and a single electrical pulse was delivered to the femoral nerve. After the last single pulse was evoked at 50% MVC, subjects relaxed. After 2–3 s of relaxation, subjects performed a 5-s MVC with one high-frequency doublet being evoked during the force plateau. Three electrical stimuli were then delivered on the relaxed muscle: one high-frequency doublet (Db100), one low-frequency doublet (Db10), and one singlet (peak twitch [Pt]). The stimuli at rest were interspersed with 3 s of rest. After each neuromuscular function evaluation, which was standardized to 40 s, the pedals were unlocked to allow the subject to resume cycling until EXH.

Saturation clamping

A pulse oximeter (NPB-295, Nellcor Puritan Bennett Inc., Pleasanton, CA) was placed on the forefinger to continuously monitor SpO2. Subjects inhaled a gas mixture delivered by an Altitrainer 200 (SMTEC, Nyon, Switzerland) through a face-mask to achieve the different environmental conditions. Rather than setting a specific altitude or FIO2 for the hypoxic conditions, SpO2 was clamped at 85% and 70% in moderate and severe hypoxia, respectively. For that purpose, FIO2 was constantly adjusted (range, 0.14–0.18 for moderate hypoxia sessions and 0.09–0.15 for severe hypoxia sessions) to make sure the target SpO2 was attained and stabilized for each subject.

Equipment and Measurements

New cycle ergometer and force assessment

A custom-built recumbent cycling ergometer was utilized for all testing sessions. Details on the cycle ergometer may be found elsewhere (5). Neuromuscular tests were therefore directly conducted on the cycle ergometer, without having to move subjects to an isometric ergometer and thus limiting fatigue misinterpretation.

Electromyographic recordings

EMG activity of the right vastus lateralis (VL), rectus femoris (RF), and biceps femoris (BF) were recorded using bipolar surface electrodes (Meditrace 100; Covidien, Mansfield, MA) of 10-mm diameter with a 30-mm interelectrode distance and the reference electrode placed on the patella. To obtain low impedance (Z < 5 kΩ) at the skin-electrode surface, the skin was gently abraded with fine sandpaper and cleaned with alcohol. EMG signals were recorded and converted from analog to digital at a sampling rate of 2 kHz with a PowerLab system (16/35; ADInstruments). The EMG signal was amplified with octal bio-amplifier (ML138; ADInstruments; common mode rejection ratio = 85 dB, gain = 500) with a bandpass filter (5–500 Hz) and analyzed offline using LabChart 8 software (ADInstruments).

Femoral nerve stimulation

Electrical stimuli were delivered percutaneously to the femoral nerve via a cathode electrode (10-mm stimulating diameter; Meditrace 100, Covidien) secured with tape in the inguinal triangle. The anode, a 50 × 90 mm rectangular electrode (Durastick Plus; DJO Global, Vista, CA), was placed in the gluteal fold. Square-wave stimuli of 1-ms duration were delivered employing a constant current stimulator (DS7A; Digitimer, Welwyn Garden City, Hertfordshire, UK). At the start of each fatigue test session (i.e., still while breathing normoxic air), the optimal stimulus intensity was determined by delivering single pulse stimulation to the femoral nerve until both maximal Pt and maximal M-wave amplitude were obtained. The stimulating intensity (117 ± 76 mA, 94 ± 46 mA, 95 ± 40 mA, 116 ± 63 mA, 106 ± 50 mA in NOR, MODABS, SEVABS, MODREL, and SEVREL, respectively) was supramaximal (i.e., 130% of optimal intensity) and kept constant throughout the fatigue test.

Transcranial magnetic stimulation

The motor cortex area of the right leg was stimulated with a magnetic stimulator (Magstim 2002; The Magstim Company, Dyfed, UK). A concave double-cone coil (diameter, 110 mm; maximum output, 1.4 T) was used to deliver single TMS pulses of 1-ms duration. The coil was positioned over the vertex of the scalp and held tangentially to the skull. At the start of each fatigue test session, optimal TMS site was determined by placing the coil to preferentially activate the left motor cortex (contralateral to the right leg) and to elicit the largest motor-evoked potential (MEP) in the VL with a small MEP in the BF (<10% of VL MEP) during isometric knee extension at 20% MVC with a stimulus intensity of 50% of maximal stimulator power output (17). The optimal stimulus site was marked on a breathable swimming cap, which was worn on the head during the fatigue tests to ensure reproducibility of the TMS conditions. The determination of optimal stimulation intensity comprised delivering TMS during brief (~5 s) submaximal (20% MVC) isometric knee extensions at 20%, 30%, 40%, 50%, 60%, 70%, and 80% of maximal stimulator power output in random order. For each stimulation intensity, subjects performed four contractions interspersed with 10 s of rest. The stimulus intensity (68% ± 11%, 68% ± 9%, 69% ± 9%, 68% ± 9%, 68% ± 10% of maximal stimulator power output in NOR, MODABS, SEVABS, MODREL, and SEVREL, respectively) that elicited the largest right VL MEP and superimposed twitch (SIT) with small MEP of the right BF (<10% of VL MEP) was considered optimal and employed throughout the fatigue test, as previously suggested by our group (e.g., Rupp et al. (18). If the stimulation intensities were not sufficient to observe a plateau in MEP responses, higher intensities (i.e., 90%, 100%) were investigated.

Gas exchanges and HR

Pulmonary gas exchange and ventilation were measured breath-by-breath with a metabolic cart (Quark CPET; COSMED, Rome, Italy) during the maximal incremental tests. Calibration of the gas analyzer was carried out before each test according to manufacturer instructions. HR was continuously recorded during the maximal incremental tests with a HR monitor (Garmin International, Schaffhausen, Switzerland), which was synchronized with the metabolic cart. All data editing, processing, and modeling were performed using OriginLab (version 9.2; OriginLab, Northampton, MA).

Near-infrared spectroscopy measurements

A two-wavelength (780 and 850 nm) portable continuous-wave near-infrared spectroscopy system (Portalite; Artinis Medical Systems, Elst, The Netherlands) was used to assess quadriceps muscle oxygenation during fatigue tests. The signals provided estimates of change in oxyhemoglobin (O2Hb), deoxyhemoglobin (HHb), and total hemoglobin (THb) concentrations. An absolute measure of oxygenated-hemoglobin saturation was represented as the tissue saturation index (TSI = O2Hb/THb, expressed in %). The probe unit was positioned over the muscle belly of the right VL with an interoptode distance of 4 cm and was firmly secured to the skin with double-sided adhesive tape. A black sweatband was placed over the probe to protect the optodes from ambient light. Data were recorded continuously at 10 Hz and filtered with a 2-s wide moving Gaussian smoothing algorithm before analysis.

Blood lactate

Fingertip approximately 5-μL capillary blood samples were withdrawn to measure blood lactate concentration ([La]) with a portable lactate analyzer (Lactate Scout; EKF Diagnostics, Barleben, Germany). This procedure was carried out for each fatigue test session, at rest as well as at 3 and 5 min after EXH.

RPE and dyspnea scale

To assess subjective perception of effort, RPE was obtained using Borg’s scale (19). The modified Borg dyspnea scale was used to assess the subject’s difficulty of breathing (from 0, no difficulty in breathing to 10, maximal breathing difficulty) (20). During the fatigue tests, RPE and difficulty of breathing were recorded in the last minute before each neuromuscular function evaluation and at EXH.

Data Analysis

In the maximal incremental tests, V˙O2peak was considered as the highest 30-s V˙O2 average and maximal HR was the highest recorded HR during the test.

Concerning the fatigue tests, only the highest score of each variable from the two neuromuscular function evaluations performed before cycling was considered for further analysis. Maximal voluntary contraction was obtained from the first maximal voluntary force of the neuromuscular evaluation (i.e., before the delivery of the TMS), as this contraction did not benefit from recovery. The SIT evoked by TMS at each voluntary contraction (i.e., 100%, 75%, and 50% MVC) were used to calculate cortical voluntary activation (VATMS). For this purpose, the y-intercept of the linear regression between the SIT and voluntary force was used to obtain the estimated resting twitch (ERT) (21). The VATMS was then calculated using the following formula:

This method has been shown to be reliable to determine VATMS in the knee extensors (3,17,22). Cortical silent period (CSP) duration was determined manually as the period between the stimulation and the return of continuous EMG activity (23).

Voluntary activation with peripheral nerve stimulation (VAPNS) was determined with the interpolated twitch technique by quantifying the evoked responses to stimulation of the femoral nerve (in this study 100-Hz doublet, Db100) during an MVC and right after on the relaxed muscle. VAPNS was then calculated using the following formula:

The twitch amplitudes obtained on the relaxed muscle (Db100, Db10, Pt) were measured to describe peripheral fatigue. The ratio between Db10 and Db100 (Db10:100) provided an index of low-frequency fatigue.

Interpolation was used to present the data to account for the intraindividual and interindividual variability in the number of completed stages of fatigue tests (i.e., 5-min stages in the first 20 min and then 10-min stages up to EXH). For that purpose, data were interpolated between stages and EXH so intermediate data points would be attributed to each second of the total TTE. Then, data were expressed as a function of TTE-SEVABS since it was systematically the shortest TTE. In other words, if subject A’s TTE-SEVABS was 15 min, all data of the four remaining fatigue tests would be expressed for 15 min, in a percentage of this duration. For statistical purposes, the duration was then converted to a percentage of TTE-SEVABS so it would be possible to analyze data at fixed time points (i.e., baseline before cycling, 25%, 50%, 75%, 100% TTE-SEVABS). The actual EXH values of the five fatigue tests were also considered for analysis.

Statistical Analysis

Data are presented as mean ± SD. To check if data were normally distributed, Shapiro-Wilk tests were performed. Homogeneity of variances was determined with the Levene’s test. In case sphericity was violated, Greenhouse–Geisser corrections were applied. To determine the effect of graded hypoxia and power output on blood [La], 1 two-way ANOVA (condition × time) with repeated measures was performed on five conditions (NOR, MODABS, SEVABS, MODREL and SEVREL) and three-time points (before exercise, 3 and 5 min after EXH). To determine the effect of graded hypoxia and power output during exercise (i.e., kinetics) on muscle oxygenation, SpO2, RPE, dyspnea, fatigue and neuromuscular function, two-way ANOVA (condition × time) with repeated measures were performed on five conditions (NOR, MODABS, SEVABS, MODREL, and SEVREL) and 5-time points (baseline, 25%, 50%, 75%, and 100% of TTE-SEVABS). Where ANOVA revealed significant main effects or interactions, data were further explored using pair-wise comparisons with a Bonferroni correction. To compare TTE, fatigue and neuromuscular function at EXH in the different conditions, one-way ANOVA with repeated measures were performed on the percentage change from baseline to EXH. Where ANOVA revealed significant interactions, data were further explored using pairwise comparisons with a Bonferroni correction. The statistical analyses were performed with Statistica (version 8, Tulsa, OK).


Power output of incremental tests

Wmax attained in NOR was significantly higher than MOD (247 ± 47 vs 221 ± 41 W, P < 0.001) and SEV (193 ± 39 W, P < 0.001), and Wmax was higher in MOD versus SEV (P < 0.001).

Oxygen saturation

Throughout the fatigue test in NOR, mean SpO2 ranged from 96% to 98%. As expected and presented in Figure 2, no differences were observed in SpO2 between MODREL and MODABS, with values stabilized at 85% ± 1% between 25% and EXH. Likewise, SpO2 was stabilized at 71% ± 2% between 25% and EXH in SEVREL and SEVABS.

Changes in arterial oxygen saturation (Saturation, panel A), TSI (panel B) and THb (panel C) during the five conditions. All panels show data at baseline as well as 25%, 50%, 75%, and 100% of TTE of SEVABS. Shaded area indicates values at EXH for NOR, MODREL, MODABS, and SEVREL. The first x axis represents the actual TTE (min) of the five sessions. Data are mean ± SD. **P < 0.01, ***P < 0.001: significantly different from baseline (main effect of time); †††P < 0.001: significantly different from the same timepoint between NOR and MODABS; ‡‡P < 0.01, ‡‡‡P < 0.001: significantly different from the same timepoint between NOR and SEVABS; $P < 0.05, $$P < 0.01; significantly different from the same timepoint between NOR and MODREL; §§P < 0.01, §§§P < 0.001: significantly different from the same timepoint between NOR and SEVREL; a P < 0.05, aaa P < 0.001: significantly different from the same timepoint between SEVREL and SEVABS.

Power output and time to EXH of fatigue tests

Initial and final power outputs during the fatigue tests are presented in Table 1. Initial power output was significantly higher in NOR than MODREL and SEVREL (P < 0.001). In addition, power output was higher in the absolute compared with the relative sessions (P < 0.001). The power output attained at the end of the fatigue tests was significantly higher in NOR compared with the four other conditions (P < 0.001). Additionally, final power output was significantly higher in SEVABS than SEVREL.

Time to EXH and W during the fatigue sessions.

TTE-NOR was significantly longer than TTE-MODABS, TTE-SEVABS, and TTE-SEVREL (all P < 0.001, Table 1). In addition, TTE-MODREL was significantly longer than TTE-MODABS (P < 0.001) and TTE-SEVREL was significantly longer than TTE-SEVABS (P < 0.001).

MVC changes

Before the fatigue sessions, there was no MVC difference between any of the conditions (Table 2). Maximal voluntary contraction was significantly reduced from 25% TTE-SEVABS whatever the condition (P < 0.001, Fig. 3A). Although MVC displayed a similar reduction in NOR and MODABS, it was less depressed in NOR than in SEVABS from 25%. The MVC decrease in NOR was similar to that of MODREL and SEVREL at isotime. Additionally, MVC in MODREL was less reduced than in MODABS at 100% (−9% ± 9% vs –18% ± 12%, respectively, P = 0.019) and also less reduced in SEVREL compared to SEVABS from 25% to 100% TTE. At EXH, the decrease ranged from −26% to −31%, with no differences among sessions.

MVC and evoked responses before the fatigue sessions.
Changes in MVC (panel A), peak twitch (panel B), Db100 (panel C), Db10:100 (panel D), VAPNS (panel E), and VATMS (panel F) during the five conditions. All panels show data at baseline as well as 25%, 50%, 75%, and 100% of TTE of SEVABS. Shaded area indicates values at EXH for NOR, MODREL, MODABS, and SEVREL. The first x axis represents the actual TTE (min) of the five sessions. Data are mean ± SD. **P < 0.01, ***P < 0.001: significantly different from baseline (main effect of time); †P < 0.05: significantly different from the same timepoint between NOR and MODABS; ‡P < 0.05; ‡‡‡P < 0.001: significantly different from the same timepoint between NOR and SEVABS; $P < 0.05: significantly different from the same timepoint between NOR and MODREL; §P < 0.05: significantly different from the same timepoint between NOR and SEVREL; ¥P < 0.05, ¥¥P < 0.01: significantly different from the same timepoint between MODREL and MODABS; a P < 0.05, aa P < 0.01, aaa P < 0.001: significantly different from the same timepoint between SEVREL and SEVABS.

Peripheral fatigue

Before the fatigue sessions, there was no electrically evoked force difference between any of the conditions for any of the parameters (Table 2). Pt, Db100 and Db10:100 decreased throughout all conditions from 25% to 100% (P < 0.001). Pt was significantly lower in MODABS versus NOR at 50% (P = 0.025), and a greater reduction was observed in SEVABS versus NOR from 25% (P < 0.001). NOR, MODREL and SEVREL presented a similar Pt reduction at isotime (Fig. 3B). The MODREL and MODABS were similarly reduced, whereas the decrease in SEVABS was significantly higher than in SEVREL from 25%.

Concerning Db100 at isotime (Fig. 3C), MODABS and NOR were similarly depressed, whereas SEVABS was significantly lower than NOR from 75% (P < 0.001). Also, NOR displayed a similar decrease as MODREL and SEVREL. Although MODREL and MODABS were similarly reduced, there was a lower reduction in SEVREL versus SEVABS from 50%.

Regarding Db10:100 at isotime (Fig. 3D), NOR and MODABS displayed the same decrease, whereas SEVABS was significantly lower than NOR from 50%. NOR was reduced to the same extent than MODREL and SEVREL at isotime. A lower reduction was found in MODREL versus MODABS at 75% (−17% ± 11% vs –25% ± 13%, respectively, P = 0.002), and in SEVREL versus SEVABS from 50%.

At EXH, Pt was less depressed in SEVREL (−33% ± 17%) than NOR (−46% ± 16%, P = 0.048). No differences between conditions were observed for Db100 or Db10:100.

Central fatigue, corticospinal excitability, and intracortical inhibition

Both VAPNS and VATMS decreased to a similar extent across all sessions, and the change was significant from 50% (Figs. 3E and 3F). No difference among conditions and no interaction effect was observed throughout exercise for both VAPNS and VATMS. At EXH, the 4% to 10% decrease for VAPNS and the 9% to 13% decrease for VATMS did not differ between conditions.

The MEP responses are presented in Table 3 (Supplemental Digital Content 1, which illustrates MEP responses before, during and immediately after the fatigue sessions, The VL MEP50 was smaller than baseline at 25%, 75%, and 100% and RF MEP50 significantly decreased at 25% in all conditions. Throughout exercise, MEP75 and MEP100 remained unchanged in both VL and RF. At EXH, MEP was similar in the five conditions at all contraction intensities in both VL and RF.

The CSP responses are presented in Table 4 (Supplemental Digital Content 2, which illustrates CSP responses before, during, and immediately after the fatigue sessions, The VL CSP50 did not change, whereas RF CSP50 was significantly shorter in all conditions from 50%. VL CSP75 was significantly shorter from 50% in all conditions, whereas RF CSP75 remained unchanged. Although VL CSP100 did not change, RF CSP100 was significantly shorter in SEVABS than in NOR at 100% (P = 0.016). At EXH, no differences were observed between conditions at any of the contraction intensities, whatever the muscle considered.

Perceptual responses

RPE increased significantly in all conditions from 25% (Fig. 4A; P < 0.001). MODABS and SEVABS were significantly higher than NOR from 50% and 25%, respectively. The MODREL and NOR were similarly increased, whereas SEVREL exhibited higher RPE than NOR at 75% (14 ± 2 vs 12 ± 2, respectively, P = 0.008) and 100% (14 ± 2 vs 13 ± 2, respectively, P < 0.001). At EXH, RPE did not differ between conditions.

Changes in RPE (panel A), dyspnea (panel B) during the five conditions. Panel C shows changes in blood lactatemia 3 and 5 min after EXH. All panels show data at baseline as well as 25%, 50%, 75%, and 100% of TTE of SEVABS. Shaded area indicates values at EXH for NOR, MODREL, MODABS, and SEVREL. The first x axis represents the actual TTE (min) of the five sessions. Data are mean ± SD. ***P < 0.001: significantly different from baseline (main effect of time); †P < 0.05, ††P < 0.01, †††P < 0.001: significantly different from the same timepoint between NOR and MODABS; ‡‡‡P < 0.001: significantly different from the same timepoint between NOR and SEVABS; $P < 0.05: significantly different from the same timepoint between NOR and MODREL; §§P < 0.01, §§§P < 0.001: significantly different from the same timepoint between NOR and SEVREL; ¥¥P < 0.01, ¥¥¥P < 0.001: significantly different from the same timepoint between MODREL and MODABS; a P < 0.05, aa P < 0.01, aaa P < 0.001: significantly different from the same timepoint between SEVREL and SEVABS.

Dyspnea increased significantly in all sessions from 25% (Fig. 4B; P < 0.001). The MODABS and SEVABS were significantly higher than NOR from 50% and 25%, respectively. Although NOR and MODREL displayed a similar increase, dyspnea was higher in SEVREL versus NOR from 50%.

Muscle oxygenation

There was a significant decrease in TSI in all conditions from 25% TTE (Fig. 2B; P < 0.001). Tissue saturation index was significantly lower in MODABS and SEVABS compared with NOR from 25%. Tissue saturation index was significantly higher in NOR versus MODREL from 50% and NOR versus SEVREL from 25%. Also, TSI in MODREL and SEVREL were similar, whereas it was significantly lower in SEVABS versus MODABS from 50%. Although TSI in MODREL and MODABS were similar, this parameter was significantly lower in SEVABS versus SEVREL from 25%. At EXH, TSI was higher in NOR versus SEVABS (P = 0.015). Total hemoglobin significantly increased in all conditions from 50%. No interaction effect was observed in this variable, and no conditions effects were found at EXH.

Blood lactate concentration

Lactate values were significantly higher post EXH (i.e., 3 and 5 min into recovery) than at rest in all conditions (Fig. 4C; P < 0.001). Overall, the longer the TTE, the lower the lactatemia.


The purpose of this study was to evaluate the effects of various levels of arterial saturation on fatigue development and etiology during cycling exercise and at EXH. The fundamental originalities of this work were that fatigue was measured for the first time with no delay in hypoxia during and at EXH of a dynamic exercise involving a large muscle mass (providing the first accurate description of the time course for this type of ecological exercise) and that exercise was performed both at relative and absolute intensities. The main findings of the present study are: (i) TTE of not only the absolute but also the relative sessions were reduced with hypoxia level; (ii) at EXH, MVC reduction was similar among conditions despite different arterial and tissue saturation; (iii) at EXH, there was less peripheral fatigue (as shown by Pt) in SEVREL versus NOR but central fatigue was similar between all sessions. Performance and peripheral fatigue at EXH in hypoxic relative sessions were decreased with altitude compared with normoxia, suggesting the brain-hypoxic effect on performance may happen in severe and moderate hypoxia.

Neuromuscular fatigue during the fatigue sessions

For a given absolute workload, hypoxia induced greater force and TSI declines, as well as greater RPE and dyspnea ratings compared with normoxia. The severity of hypoxia accentuated the differences in these parameters. Similar to our findings, previous literature showed greater MVC decline and peripheral fatigue in hypoxia compared with normoxia (7–10,24,25).

The intensity of the absolute sessions was based on the maximal incremental test in normoxia. Indeed, initial power output in MODABS and SEVABS corresponded to 61% Wmax and 71% Wmax in their respective environment. Since these intensities were higher than the 55% Wmax in NOR, metabolic and ventilatory adjustments were triggered, and ATP supply could not be exclusively attained by oxidative phosphorylation early in the fatigue tests. The contribution to ATP resynthesis from anaerobic glycolysis and creatine phosphate increased, which resulted in the accumulation of H+ and Pi, known to impair muscle function (26) and cause peripheral fatigue. Concomitantly, lactate concentration increased considerably and could no longer be stabilized, which is supported by the higher blood [La] observed after absolute sessions compared with NOR. The higher blood [La] likely accompanied by progressive accumulation of H+ observed in the blood at high exercise intensities (27) might have increased the perception of effort and dyspnea ratings. Since III/IV afferent fibers are sensitive to the mechanical and metabolic state of the working muscles (28), high H+ likely activated these afferents and informed the central nervous system about the disturbances occurring at the periphery. Furthermore, the high exercise intensities might have led to the recruitment of additional motor units (including type II). These feedforward and feedback mechanisms likely resulted in the earlier attainment of maximal tolerable RPE and thus shorter TTE.

Although in Kayser et al. (29), the fatigue tests in normoxia and severe hypoxia were performed at relative intensities, subjects cycled at maximal power output, which is substantially higher than the intensity used in our study (55% of maximal power output). Thus, this is the first study with prolonged submaximal exercise to EXH that accounts for the altitude-induced decrease in V˙O2peak, that is, subjects exercised in hypoxia at relative intensities. This design allows us to determine how far neuromuscular perturbations observed in hypoxic absolute sessions are due to the associated higher relative intensity and/or to the hypoxemic stress per se. Interestingly, MVC reduction and peripheral fatigue were similarly reduced in hypoxic relative sessions and NOR. Although RPE and dyspnea were similar between MODREL versus NOR, these parameters were higher in SEVREL compared with NOR.

In the quadriceps muscle, MODREL and SEVREL displayed lower values of muscle oxygenation throughout the fatigue sessions when compared with NOR. The similar muscle TSI after repeated sprints in normoxia and hypoxia observed by Billaut et al. (30) is in contrast with our findings and may be explained by the discrepancies in the protocols used. Our results suggest that hypoxia per se lowers muscle oxygenation despite lower mechanical muscle constraints than when considering absolute exercise intensities. Yet, peripheral fatigue was similar among hypoxic relative sessions and NOR. This means that various levels of muscle deoxygenation may be associated with the same amount of peripheral fatigue. It is also possible that the lower O2 availability in severe hypoxia also acted directly on the brain and increased perceptual responses (i.e., RPE and dyspnea ratings) despite similar peripheral fatigue throughout SEVREL.

The similar VATMS and VAPNS declines across fatigue sessions suggest that, unlike hypothesized, power output and hypoxia did not influence independently the ability to drive all motoneurons maximally. In our design, the greater MVC drop in absolute sessions, especially SEVABS, was explained by higher peripheral fatigue. Furthermore, VA decline does not seem to be sensitive to the so-called 75% SpO2 threshold. In fact, we showed that exercising in hypoxia below or above 75% SpO2 does not affect the magnitude of VA decline. Rather, severe hypoxia per se may trigger higher RPE and dyspnea. It is important to emphasize that MEP/Mmax and CSP responses were not affected by power output or hypoxia, which suggests that the 75% SpO2 threshold did not induce changes in corticospinal excitability and inhibition.

Neuromuscular fatigue at EXH

The MVC loss, peripheral fatigue, RPE were similar between hypoxic absolute sessions and NOR at EXH. However, there was a greater TSI drop and higher blood [La] in hypoxic absolute sessions versus NOR. The similar decline in MVC is in agreement with previous research (7,10,31,32). However, because these studies were unable to test neuromuscular function immediately at EXH, our study provided new insight into fatigue magnitude and etiology right at EXH in the context of prolonged exercise. Contrary to our hypothesis and the findings of Amann et al. (10), SEVABS resulted in the same magnitude of peripheral fatigue compared to NOR. Since, at EXH, the intensity of MODABS and SEVABS corresponded to 66% Wmax and 71% Wmax (NS) of the same environmental condition, similar peripheral fatigue between hypoxic absolute sessions and NOR (70% Wmax) is not so surprising, as it is known that in addition to the environmental conditions, exercise intensity can also influence fatigue etiology (15,16,27). Indeed, short, high-intensity exercise, such as in SEVABS, exacerbates peripheral fatigue. Thus, the expected lower peripheral fatigue due to premature exercise stop related to brain hypoxic effects in SEVABS might have been counterbalanced by exercise intensity, which ultimately resulted in comparable peripheral fatigue in absolute sessions and NOR. The reason why peripheral fatigue was lower in SEVABS versus NOR in Amann et al. (10) might be related to the delay to assess fatigue at EXH. This is particularly true in high-intensity exercise like in SEVABS, where fatigue is thought to be reversed very quickly (4). Furthermore, exercise intensities used in our study, that is, 55% to 70% of relative Wmax, differed from Amann et al. (10), which were set at 80% of absolute Wpeak.

The shorter TTE in hypoxic relative sessions together with the attenuated peripheral fatigue at EXH supports, although for the first time with a prolonged exercise and no delay between the fatiguing exercise and the fatigue assessment, the hypothesis proposed by Amann et al. (10) and Millet et al. (11), that a brain-hypoxic effect plays a significant role in the decision to stop exercise earlier in hypoxia. The causes of this brain-hypoxic effect are not fully known, but lower O2 availability in the brain due to the direct effect of hypoxia is a strong candidate (7). Despite having accounted for the hypoxia-induced decrease in V˙O2peak, TTE was still shorter in relative sessions versus NOR. Thus, we believe that, at least for the exercise intensities used in this study, the brain-hypoxic effect may also occur at lower altitudes so that the 75% SpO2 threshold where “performance switches from a predominantly peripheral origin of fatigue to a hypoxia-sensitive central component of fatigue” (10) may actually not exist.

We acknowledge that an alternative point of view may underline the fact that, due to the protocol characteristics, that is, increase in power output from 1 h of exercise in some sessions (see Table 1), the final power outputs in SEVREL (~60%) did not coincide anymore to the same relative exercise intensity achieved at EXH in NOR (~70%). At EXH, lower peripheral fatigue in SEVREL compared with NOR may then be attributed, at least partly, to the lower relative exercise intensity finally achieved rather than to the hypoxic effect of severe hypoxia per se.

Contrary to our hypothesis, and despite the reduction in TTE, severe systemic hypoxic stress per se did not induce greater inability of the motor cortex to maximally drive the motoneurons (i.e., central fatigue as assessed by TMS). The performance reduction when exercising in hypoxia at relative intensities may be mediated by factors other than the ability of the motor cortex to drive locomotor muscles, such as an exacerbated perceived effort.


This study confirms, but assessed for the first time without time delay in cycling, that at absolute power output (i.e., higher relative intensity), hypoxia exacerbates the magnitude of peripheral fatigue. Conversely, we showed that, when workload intensities are defined relative to the environmentally triggered maximal capabilities, hypoxia per se does not accentuate peripheral fatigue, yet exercise tolerance is reduced. This is likely due to a direct brain-hypoxic effect resulting in increased perception of effort, as evidenced by the attenuated peak twitch at EXH, and/or subcortical areas prominent for their potential to disrupt performance. Our findings question the existence of a threshold at 75% SpO2 proposed before. Future research should match closely exercise intensity in normoxia and hypoxia to describe fatigue etiology for the same exercise duration.

The authors would like to thank subjects for their time and effort. This study was supported by the University Savoie Mont Blanc as part of the doctoral work of José Mira. The results of the present study do not constitute endorsement by the American College of Sports Medicine.

Conflict of Interest: The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. The authors report no conflict of interest.


1. MacIntosh BR, Rassier DE. What is fatigue? Can J Appl Physiol. 2002;27(1):42–55.
2. Place N, Millet GY. Quantification of neuromuscular fatigue: what do we do wrong and why? Sports Med. 2020;50:439–47.
3. Mira J, Lapole T, Souron R, Messonnier L, Millet GY, Rupp T. Cortical voluntary activation testing methodology impacts central fatigue. Eur J Appl Physiol. 2017;117(9):1845–57.
4. Froyd C, Millet GY, Noakes TD. The development of peripheral fatigue and short-term recovery during self-paced high-intensity exercise. J Physiol. 2013;591(Pt 5):1339–46.
5. Doyle-Baker D, Temesi J, Medysky ME, Holash RJ, Millet GY. An innovative ergometer to measure neuromuscular fatigue immediately after cycling. Med Sci Sports Exerc. 2018;50(2):375–87.
6. Verges S, Rupp T, Jubeau M, et al. Cerebral perturbations during exercise in hypoxia. Am J Physiol Regul Integr Comp Physiol. 2012;302(8):R903–16.
7. Goodall S, González-Alonso J, Ali L, Ross EZ, Romer LM. Supraspinal fatigue after normoxic and hypoxic exercise in humans. J Physiol. 2012;590(Pt 11):2767–82.
8. Amann M, Romer LM, Pegelow DF, Jacques AJ, Hess CJ, Dempsey JA. Effects of arterial oxygen content on peripheral locomotor muscle fatigue. J Appl Physiol. 2006;101(1):119–27.
9. Amann M, Pegelow DF, Jacques AJ, Dempsey JA. Inspiratory muscle work in acute hypoxia influences locomotor muscle fatigue and exercise performance of healthy humans. Am J Physiol Regul Integr Comp Physiol. 2007;293:R2036–45.
10. Amann M, Romer LM, Subudhi AW, Pegelow DF, Dempsey JA. Severity of arterial hypoxaemia affects the relative contributions of peripheral muscle fatigue to exercise performance in healthy humans. J Physiol. 2007;581(Pt 1):389–403.
11. Millet GY, Muthalib M, Jubeau M, Laursen PB, Nosaka K. Severe hypoxia affects exercise performance independently of afferent feedback and peripheral fatigue. J Appl Physiol. 2012;112(8):1335–44.
12. Jubeau M, Rupp T, Temesi J, et al. Neuromuscular fatigue during prolonged exercise in hypoxia. Med Sci Sports Exerc. 2017;49:430–9.
13. Twomey R, Wrightson J, Fletcher H, Avraam S, Ross E, Dekerle J. Exercise-induced fatigue in severe hypoxia after an intermittent hypoxic protocol. Med Sci Sports Exerc. 2017;49:2422–32.
14. Farra SD, Cheung SS, Thomas SG, Jacobs I. Rate dependent influence of arterial desaturation on self-selected exercise intensity during cycling. PLoS One. 2017;12(3):e0171119.
15. Froyd C, Beltrami FG, Millet GY, Noakes TD. Central regulation and neuromuscular fatigue during exercise of different durations. Med Sci Sports Exerc. 2016;48(6):1024–32.
16. Thomas K, Goodall S, Stone M, et al. Central and peripheral fatigue in male cyclists after 4-, 20-, and 40-km time trials. Med Sci Sports Exerc. 2015;47(3):537–46.
17. Sidhu SK, Bentley DJ, Carroll TJ. Cortical voluntary activation of the human knee extensors can be reliably estimated using transcranial magnetic stimulation. Muscle Nerve. 2009;39(2):186–96.
18. Rupp T, Jubeau M, Wuyam B, et al. Time-dependent effect of acute hypoxia on corticospinal excitability in healthy humans. J Neurophysiol. 2012;108(5):1270–7.
19. Borg G. Borg’s Perceived Exertion and Pain Scales. Champaign: Hum Kinet; 1998.
20. Burdon JG, Juniper EF, Killian KJ, Hargreave FE, Campbell EJ. The perception of breathlessness in asthma. Am Rev Respir Dis. 1982;126(5):825–8.
21. Todd G, Taylor JL, Gandevia SC. Measurement of voluntary activation based on transcranial magnetic stimulation over the motor cortex. J Appl Physiol. 2016;121(3):678–86.
22. Goodall S, Romer LM, Ross EZ. Voluntary activation of human knee extensors measured using transcranial magnetic stimulation. Exp Physiol. 2009;94(9):995–1004.
23. Sidhu SK, Bentley DJ, Carroll TJ. Locomotor exercise induces long-lasting impairments in the capacity of the human motor cortex to voluntarily activate knee extensor muscles. J Appl Physiol. 2009;106(2):556–65.
24. Amann M, Venturelli M, Ives SJ, et al. Peripheral fatigue limits endurance exercise via a sensory feedback-mediated reduction in spinal motoneuronal output. J Appl Physiol. 2013;115(3):355–64.
25. Goodall S, Twomey R, Amann M, et al. AltitudeOmics: exercise-induced supraspinal fatigue is attenuated in healthy humans after acclimatization to high altitude. Acta Physiol (Oxf). 2014;210(4):875–88.
26. Westerblad H, Allen DG, Lännergren J. Muscle fatigue: lactic acid or inorganic phosphate the major cause? News Physiol Sci. 2002;17:17–21.
27. Thomas K, Elmeua M, Howatson G, Goodall S. Intensity-dependent contribution of neuromuscular fatigue after constant-load cycling. Med Sci Sports Exerc. 2016;48:1751–60.
28. Amann M, Sidhu SK, Weavil JC, Mangum TS, Venturelli M. Autonomic responses to exercise: group III/IV muscle afferents and fatigue. Auton Neurosci. 2015;188:19–23.
29. Kayser B, Narici M, Binzoni T, Grassi B, Cerretelli P. Fatigue and exhaustion in chronic hypobaric hypoxia: influence of exercising muscle mass. J Appl Physiol. 1994;76(2):634–40.
30. Billaut F, Kerris JP, Rodriguez RF, Martin DT, Gore CJ, Bishop DJ. Interaction of central and peripheral factors during repeated sprints at different levels of arterial O2 saturation. PLoS One. 2013;8(10):e77297.
31. Amann M, Eldridge MW, Lovering AT, Stickland MK, Pegelow DF, Dempsey JA. Arterial oxygenation influences central motor output and exercise performance via effects on peripheral locomotor muscle fatigue in humans. J Physiol. 2006;575(Pt 3):937–52.
32. Romer LM, Haverkamp HC, Amann M, Lovering AT, Pegelow DF, Dempsey JA. Effect of acute severe hypoxia on peripheral fatigue and endurance capacity in healthy humans. Am J Physiol Regul Integr Comp Physiol. 2007;292(1):R598–606.


Supplemental Digital Content

Copyright © 2020 by the American College of Sports Medicine