Journal Logo

SPECIAL COMMUNICATIONS: Methodological Advances

An Innovative Ergometer to Measure Neuromuscular Fatigue Immediately after Cycling


Author Information
Medicine & Science in Sports & Exercise: February 2018 - Volume 50 - Issue 2 - p 375-387
doi: 10.1249/MSS.0000000000001427


Fatigue is a complex phenomenon with multiple possible definitions (e.g., increased rate of perceived exertion or energy cost), but fatigue from exercise is usually defined by exercise physiologists as a decrease in maximal voluntary contraction force (1). To be more accurate, neuromuscular fatigue (NMF) 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 (2). From a physiological perspective, fatigue can originate at several potential sites that are typically classified as being central (the corticospinal pathway proximal to the neuromuscular junction), or peripheral (distal to the neuromuscular junction) (1,3). Central fatigue is an alteration within the central nervous system leading to a derecruitment of motor units and/or a decrease of the discharge frequency below the frequency of tetanic fusion, both factors leading to force decline. Current knowledge on the origins of peripheral fatigue after endurance exercise suggests that such exercise could impair the action potential transmission along the sarcolemma, the excitation–contraction coupling and the actin–myosin interaction. Central and peripheral fatigue are sometimes independent (e.g., peripheral fatigue can occur without central fatigue), but they usually interact, with their relative contributions depending largely on the type of exercise investigated (4,5).

Many studies have attempted to measure NMF in humans using a combination of EMG and force measurements during voluntary contractions as well as electrical nerve stimulation and transcranial magnetic stimulation (TMS) (6). Although the magnitude and etiology of fatigue depend on the exercise under consideration, that is, the so-called task dependency, fatigue is often studied using single-joint isometric, isotonic, and isokinetic exercises (e.g., 7–9). These single-joint experimental setups are convenient for performing an isolated neuromuscular assessment immediately after exercise. Yet many sports, daily living activities, and training or rehabilitation techniques (e.g., walking, cycling, running, weight training) involve dynamic exercise with a large muscle mass. There are well-known differences between single-limb and whole-body exercises in terms of cardiovascular limitations (as shown by larger cardiopulmonary responses when exercising with a large muscle mass) or peripheral versus central fatigue (e.g., 10). Indeed, Rossman et al. (10) suggested that when exercising a small muscle mass, the central nervous system tolerates a greater magnitude of peripheral fatigue and likely a greater intramuscular metabolic disturbance. Thus, investigations of fatigue in whole-body dynamic exercise are directly applicable to real-world settings.

Thus, there is a need for specificity in testing peripheral and central fatigue both during and immediately after such exercises. Many studies, including our own work, have measured knee-extensor fatigue due to cycling by exercising the subjects on a standard stationary bicycle and then transferring the subjects to a specialized chair to isometrically measure central and peripheral fatigue (e.g., 5,10–12). However, time, usually minutes rather than seconds, is required for the transfer. This is problematic because strength (i.e., force during maximal voluntary contractions [MVC]) (e.g., 8,13), voluntary activation (VA) (e.g., 14), and muscle contractile properties (e.g., 13) recover rapidly after short-duration exercise. For instance, significant recovery in skeletal muscle function has been shown to occur within the first 1 to 2 min after exercise (13). Similar to measures of peripheral fatigue, measures of central fatigue are affected if a short time delay exists between task failure and measurement. Changes to EMG parameters elicited by single-pulse TMS, the motor-evoked potential (MEP) and cortical silent period (CSP), occur within a few seconds after the end of exercise (14). Thus, measurements of muscle function must be performed immediately after exercise termination. Consequently, previous studies have likely underestimated the magnitude of NMF development during exercise and have not examined the changes in corticospinal parameters in a timely manner (see Carroll et al. (15) for a recent review). In other words, previous studies have not been able to adequately examine the causes of exercise task failure. In addition, the time course of central and peripheral fatigue responses during cycling is usually not measured with traditional setups, which does not allow determination of the respective contributions of central and peripheral factors at any given point of a fatiguing task. The recognition of the need to reduce the time delay between exercise cessation and fatigue measurement (16) has resulted in the development of several different cycling ergometers (17–19). These ergometers have been used to quantify fatigue in the knee extensors with different measures including isometric force, power, and EMG from 2 to 8 s after exercise. However, none of these ergometers to date have been able to incorporate the current criterion standard isometric force measures for determining the level of central or peripheral fatigue immediately after exercise.

To address these important issues, we developed an innovative cycling ergometer that can switch from cycling (i.e., fatiguing exercise) to isometric mode (i.e., for NMF measurement) within 1 s. The development of this new cycling ergometer was influenced by previously published innovative single-leg studies with experimental setups that allow an instant switch from dynamic exercise in isokinetic (8) and isotonic (9,20) modes, to an isometric mode for neuromuscular assessment without a time delay. Furthermore, the protocol we used was inspired by a recently validated quadriceps intermittent fatigue (QIF) test (7) that allows repetitive muscle fatigue assessments without the limitations of subject motivation or cooperation. By adapting the experimental setups from these previous studies and the QIF protocol to this innovative cycling ergometer, we are proposing to assess for the first time the kinetics of fatigue in cycling without any time delay. Thus, the goal of the present experiment was (i) to assess the feasibility and reliability of the innovative ergometer during an incremental fatigue test and (ii) to compare the fatigue outcomes from this new test with the classic way of measuring fatigue, that is, by moving the subject from a stationary bicycle to an isometric chair and testing them 2-min postexercise.



Twelve healthy young adults (six men and six women; mean ± SD: age, 26 ± 5 yr; body mass, 69 ± 10 kg; height, 175 ± 9 cm) participated in this study. No participants had musculoskeletal or cardiorespiratory conditions. The participants were instructed to refrain from caffeine consumption and strenuous exercise or activity for at least 24 h before the testing sessions. Participants were informed of the experimental protocol, and all associated risks before giving written informed consent. All procedures conformed to the Tri-Council Policy Statement and were approved by the University of Calgary Conjoint Health Research Ethics Board (REB13-1104).

Experimental design

All procedures were conducted in an environmentally controlled laboratory (i.e., temperature, ~21°C; relative humidity, ~35%) at a similar time of the day for each subject. Preexercise and postexercise measurements were taken in the traditional isometric chair setup, and preexercise, during and postexercise measurements were taken on the innovative cycling ergometer (experimental setups described below).

Traditional isometric chair

Subjects sat on a customized chair secured by chest and hip straps with knees and hips flexed at 90°. The right ankle was attached to a force transducer (LC101-2K; Omegadyne, Sunbury, OH) that ran parallel with the knee extensors. Visual feedback of the force produced by the subject was projected on a computer monitor placed in front of the subject. During all measurements, the subjects were instructed to place their hands across their chest. The chair setup was identical for each subject during each test.

Innovative cycling ergometer

Subjects cycled and performed voluntary isometric knee extension on a custom-built recumbent cycling ergometer (Fig. 1A) with electromagnetically braked Velotron system (Racermate Inc, Seattle, WA) and powered by Velotron Coaching Software (Racermate Inc) (see Figure, Supplemental Digital Content 1, Pictures of new cycling ergometer, The cycling ergometer permitted the pedals to be locked instantly in a fixed position. Both the height and fore–aft position of the cycling ergometer seat were adjusted such that subjects could cycle comfortably. When the pedals were locked, hip angle was approximately 100° and right knee and ankle angles were approximately 90°. Subjects were also secured at the hip and chest with noncompliant straps. Isometric force was measured during voluntary and evoked contractions by a previously validated (21) wireless PowerForce pedal force analysis system (Model PF1.0.0; Radlabor GmbH, Freiburg, Germany) located between the pedal and crank. Force was sampled at 500 Hz and recorded using Imago Record (version 8.50; Radlabor GmbH) software. When the pedals were locked in position, the cranks were parallel to the ground. Subjects were instructed to lock their ankle and push their foot forward such that the knee extensors were preferentially activated and force was measured in line with the crank. To provide real-time visual force feedback, the PowerForce signal was transmitted to a PowerLab system (16/35, ADInstruments, Bella Vista, Australia) using a National Instruments 16-bit A/D card (NI PCI-6229; National Instruments, Austin, TX) and connector block (BNC-2111; National Instruments) and displayed on a large computer monitor positioned in front of the subject. The data recorded using Imago Record was analyzed offline using Labchart 8 software (ADInstruments).

Ergometer used for the cycling bouts and to assess NMF isometrically with a lockable pedal position (panel A), overview of the experimental protocol (panel B), overview of the neuromuscular assessment before, during and after the cycling exercise (panel C), and representative traces of individual superimposed Tw and evoked forces on relaxed muscle on the innovative ergometer (panel D).


Electromyographic activity of the right knee extensors, vastus lateralis (VL), rectus femoris (RF), and biceps femoris, was recorded with pairs of self-adhesive surface (10-mm recording diameter) electrodes (Meditrace 100; Covidien, Mansfield, MA) in bipolar configuration with a 30-mm interelectrode distance and the reference on the patella. A low impedance (<10 kΩ) between electrodes was obtained by shaving and gently abrading the skin and then cleaning it with isopropyl alcohol, and the electrodes were taped to the skin to ensure they would stay in place for the duration of the experiment. The EMG signals were analog-to-digitally converted at a sampling rate of 2000 Hz by PowerLab system (16/35; ADInstruments) and octal bio-amplifier (ML138; ADInstruments; common mode rejection ratio = 85 dB, gain = 500) with bandpass filter (5–500 Hz) and analyzed offline using Labchart 8 software (ADInstruments).

Femoral nerve electrical stimulation

To determine stimulation intensity subjects were seated in the isometric chair while single electrical stimuli (1-ms duration) were delivered incrementally to the femoral nerve until plateaus in twitch (Tw) and maximal M-wave amplitudes were reached. The stimuli were delivered via constant-current stimulator (DS7A, Digitimer; Welwyn Garden City, Hertfordshire, UK) to the right femoral nerve via a cathode electrode (10-mm stimulating diameter; Meditrace 100, Covidien) secured with tape and a gauze plug to apply pressure in the inguinal triangle and 50 × 90 mm rectangular anode electrode (Durastick Plus, DJO Global, Vista, CA) in the gluteal fold. The subjects were instructed to relax fully when the stimulation was delivered. Supramaximal stimulation at 130% of the plateau intensity was delivered to ensure maximal Tw and M-wave amplitudes. The mean stimulus intensity used during session 1 was 171 ± 75 mA and session 2 was 151 ± 67 mA (P > 0.05).

Transcranial magnetic stimulation

All stimulations were performed using a 110-mm double-cone coil powered by a Magstim 2002 stimulator (The Magstim Company, Whitland, UK) inducing a posteroanterior current in the brain. To reduce head and neck movements, the subjects wore a cervical collar throughout the experiment. The TMS stimulation site was determined as previously described (22). To determine the TMS stimulation intensity the subject performed brief (2–3 s) voluntary knee-extensor contractions to meet a force target of 20% of their warm-up MVC. The subject was stimulated at the previously determined site during four brief contractions at seven randomly ordered intensities (i.e., 20%, 30%, 40%, 50%, 60%, 70%, and 80% of maximal stimulator output). A response curve was produced from the elicited MEP and the intensity where the MEP size plateaued was chosen (23). The mean stimulus intensities used during sessions 1 and 2 were 62% ± 13% and 61% ± 15%, respectively, of the maximal stimulator output (P > 0.05).


The testing protocol consisted of a familiarization session and two experimental sessions (sessions 1 and 2) that were completed within a 4-wk period. The familiarization session required the subjects to perform MVC on the cycling ergometer and chair with femoral nerve stimulation and TMS as well as cycling ergometer transitions from cycling to “locked pedal” mode (see Fig. 1A). Sessions 1 and 2 were identical and consisted of the experimental protocol outlined in Figure 1B.

At the start of sessions 1 and 2 a standardized warm-up was performed on the isometric chair consisting of five 5-s submaximal contractions of the knee extensors at 10% and 30% of the subject’s familiarization session MVC, followed by three 5-s contractions at 50% MVC, one near-maximal contraction, and finally one MVC.

The first pretest measurements consisted of a 5-s isometric knee extensor MVC with superimposed femoral nerve electrical stimulation (100-Hz paired pulse) delivered when a plateau in maximal force was reached, followed by three more stimuli in the relaxed muscle spaced 3 s apart (100-Hz and 10-Hz paired pulses and single pulse). This was repeated a second time 2 min later. If there was >5% difference in maximal force, a third MVC was performed. The second pretest measurements consisted of two (or three, if >5% difference in MVC) isometric knee extensor MVC with a single TMS pulse delivered at maximal force and once the subject returned to maximal force the femoral nerve was stimulated (single pulse).

After the pretest measurements on the isometric chair, the subject was moved to the cycling ergometer. Two pretest series identical to those performed on the isometric chair were conducted on the cycling ergometer and then the cycling test began. The subjects pedaled at predetermined power outputs scaled to their body mass that increased every 3-min stage. More precisely, the power output was 0.3 W·kg−1 for the first stage and then increased by 0.3 W·kg−1 for the next four stages, by 0.4 W·kg−1 for the following five stages and by 0.5 W·kg−1 for all subsequent stages (Fig. 1B). At the end of every 3-min stage, the pedals were locked instantaneously so that the subject’s right knee was at a 90° angle allowing them to perform an isometric contraction of the right knee extensors. The same measurements performed in the pretest were repeated during a 30-s interval between stages (Fig. 1C). After the measurements were taken, the pedals were immediately unlocked allowing the subject to resume cycling at the predetermined higher power output. For all exercise tests, participants were blinded to elapsed time and power output, receiving visual feedback only for pedal cadence with the instruction to maintain a cadence between 60 and 80 rpm. Power output was increased after every 3-min stage and 30-s measurement interval until the subjects could no longer sustain a cadence above 60 rpm.

One posttest measurement series took place immediately after cycling task failure, before moving the subject from the cycling ergometer back to the isometric chair. Exactly 2 min after task failure on the cycling ergometer, a final posttest measurement was performed on the isometric chair.

Data analysis

Neuromuscular parameters were measured on the isometric chair at preexercise (PRE) and 2 min postcycling (POST-Chair). On the cycling ergometer, data were measured at precycling (PRE), after the completion of each 3-min cycling stage, and then immediately postcycling after task failure (POST-Bike). The last common cycling stage achieved by all participants was stage 7.

The pretest maximal force was defined as the highest prestimulation MVC force from the two MVC with 100-Hz paired electrical pulses before the cycling bout. Maximal voluntary force was always measured from the first MVC in the testing series (i.e., with 100-Hz paired electrical pulses). Voluntary activation was determined by Tw interpolation using the superimposed and potentiated doublet amplitudes elicited by 100-Hz paired-pulse femoral nerve electrical stimulation during and after the MVC and calculated from equation 1.

M-wave peak-to-peak area was calculated from EMG responses to single femoral nerve electrical stimulation in the relaxed muscle state. Motor-evoked potential and M-wave areas were also determined from the EMG responses to the electrical and TMS pulses delivered during the second MVC and MEP are reported normalized to the M wave during MVC (MEP/Msup). Cortical silent period was determined visually by an experienced investigator with its onset defined as the TMS stimulator discharge and the offset as the resumption of continuous voluntary EMG activity (24).

Indices of peripheral fatigue included potentiated Tw amplitude (Tw, N), 100-Hz potentiated doublet amplitude (Db100, N), and 10-Hz potentiated doublet amplitude normalized to Db100 (Db10:100). The presence of low-frequency fatigue, which is the preferential loss of force with stimulation at low frequencies and can occur without or with a smaller reduction in force during stimulation at high frequencies (25), was evaluated from the change in Db10:100 (26).

Statistical analyses

All data are reported as mean ± SD. The assumptions of statistical tests were verified as appropriate, and normal distribution was assessed with Shapiro–Wilk test. When Mauchly test of sphericity was violated, the Greenhouse-Geisser correction was applied since epsilon was always <0.75 (27). To compare variables between the first and second cycling sessions, a two-way repeated-measures ANOVA (session × time) involving all time points from PRE-Bike to stage 7 and POST-Bike was used (i.e., PRE, Stages 1, 2, 3, 4, 5, 6, and 7, and POST-Bike). A Bonferonni pairwise comparison was used for post hoc analysis if an ANOVA indicated a significant main effect or interaction. A three-way repeated-measures ANOVA was performed (session–ergometer–time) to test the difference in the relative changes from PRE to POST between the cycling ergometer versus the traditional isometric chair. The effect size was computed as partial eta squared (ηp2). As recommended by Hopkins (28), relative reliability was assessed with a mixed-model intraclass correlation coefficient (ICC3,1) at the 95% confidence level. An ICC > 0.75 represents excellent reliability, from 0.60 to 0.74 represents good reliability, 0.40 to 0.59 represents fair reliability, and an ICC < 0.40 represents poor reliability (29). Intraclass correlation coefficient was not calculated for VA due to a ceiling effect, which occurs when there is little variability in the scores of all subjects as they approach the maximum possible value (30). Because of the sample size of this study, absolute reliability was determined through typical error expressed as coefficients of variation (CV) ± 95% confidence intervals (CI) (31). Both ICC and CV were calculated using log-transformed data in the online spreadsheet created by Hopkins (31). Statistical significance was set as P < 0.05 for all tests, and all ANOVA were conducted using IBM™ SPSS™ Statistics (version 24, IBM Corp., Somers, New York, NY) software.


Subject data

The mean number of cycling stages completed by the 12 subjects was 9.9 ± 2.0 (range, 7–13) in session 1 and 9.8 ± 1.9 (range, 7–13) in session 2 (P > 0.05). The mean power outputs achieved at task failure were 240 ± 56 W and 238 ± 57 W in sessions 1 and 2, respectively (P > 0.05). All subjects completed the same number of stages for both sessions 1 and 2, except for two subjects who terminated exercise a single stage earlier or later in the second session, respectively. Stage 7 was the last common stage completed among all the subjects in both sessions and as such, was used to examine fatigue for a given workload.

Reliability of force measured before exercise

The results for reliability of absolute PRE force measures are presented in Table 1 including mean values and percent change in mean values for each session. Between-session relative reliability was excellent for all measures on the cycling ergometer, with ICC ranging from 0.90 for 100-Hz doublet force to 0.97 for MVC force. Absolute reliability on the cycling ergometer was excellent, with CV < 5.2% for all measured parameters. Overall, the ICC and CV of the parameters collected on the cycling ergometer were similar to those found on the isometric chair (Table 1).

Reliability of PRE, stage 7 and POST values for voluntary and evoked knee-extensor force on the isometric chair and cycling ergometer.

Force and VA outcome measures during the cycling test

The two-way repeated-measures ANOVA revealed no statistically significant interaction between session and time for any force outcome measure (all P > 0.350), or significant main effects between sessions (all P > 0.244). The two-way repeated-measures ANOVA for MVC during the cycling test showed a statistically significant main effect for time (F[2.080, 22.975] = 49.688, P < 0.001) with a large effect size (ηp2 = 0.819). Maximal voluntary contractions decreased continuously throughout the test (Fig. 2A) and pairwise comparisons indicate that compared with MVC at PRE, there was a significant difference from stage 4 onward (P = 0.007).

Volitional and evoked knee-extensor force and VA during the cycling ergometer tests in sessions 1 and 2. MVC (panel A), VA (panel B), Db100 (panel C), Tw force (panel D), 10 to 100 Hz doublet ratio (Db 10:100, panel E). All values that are significantly different from PRE are denoted with *(P < 0.05).

Figure 2B shows VA measurements throughout the cycling test. There was a statistically significant main effect for time (F[8, 88] = 3.376, P = 0.002, ηp2 = 0.235). However, the post hoc pairwise comparisons revealed no significant differences between time points (all P > 0.195). Reliability results of VA for both the cycling ergometer and isometric chair are displayed in Table 2.

Reliability of M-waves in the relaxed muscle and MEP (normalized to M-wave areas during MVC, Msup) and CSP at MVC from the VL and RF at PRE, stage 7 and POST on the isometric chair and cycling ergometer.

Main effects for time were also significant across all other force measures (Db100: F[2.846, 31.308] = 61.056, P < 0.001, ηp2 = 0.847; Tw: F[2.726, 29.983] = 46.322, P < 0.001, ηp2 = 0.808; Db10:100: F[2.975, 32.728] = 13.993, P < 0.001, ηp2 = 0.560). Pairwise comparisons revealed a significant decrease from PRE in Tw force (P = 0.005) after only one stage of cycling, in Db100 force after stage 4 (P = 0.027), and in Db10:100 (P = 0.018) after stage 7.

The reliability of force measures at stage 7, the last common cycling stage among all subjects, and at POST for the cycling ergometer and isometric chair are presented in Table 2. Between-session reliability was excellent for all cycling ergometer measurements made at stage 7 with ICC ranging from 0.83 for MVC to 0.91 for Db10:100. At POST-Bike, all ICC values were within the good to excellent range on the cycling ergometer (0.62–0.91), as well as on the isometric chair (0.70–0.80). The CV values for all parameters at stage 7 and POST-Bike were higher than at PRE and ranged from 4.0% to 11.9%.

Comparison of force outcome measures on the cycling ergometer and isometric chair

The three-way repeated-measures ANOVA revealed no statistically significant interaction between session, ergometer, and time (all P > 0.057) for all force measures. There was a statistically significant two-way ergometer–time interaction for MVC (F [1, 11] = 18.726, P = 0.001, ηp2 = 0.630), Db100 (F [1, 11] = 17.505, P = 0.002, ηp2 = 0.636), and Db10:100 (F [1, 11] = 16.055, P = 0.002, ηp2 = 0.616). No other simple two-way interaction was statistically significant (all P > 0.057). The main effect for the difference between the cycling ergometer and isometric chair was statistically significant for all force measures (all P < 0.042), except for Tw (P = 0.839). The main effect for time from PRE to POST was significant for all force measures: MVC (F [1, 11] = 121.332, P < 0.001, ηp2 = 0.917), Db100 (F [1, 11] = 136.343, P < 0.001, ηp2 = 0.932), Db10:100 (F [1, 11] = 146.047, P < 0.001, ηp2 = 0.936), and Tw (F [1, 11] = 203.457, P < 0.001, ηp2 = 0.953). There was a greater reduction of mean MVC measured on the cycling ergometer at POST-Bike (63% ± 12% PRE) in comparison to that measured on the isometric chair 2 min later (72% ± 12% PRE). The reduction in mean Db100 force from PRE- to POST-Bike was also greater on the cycling ergometer (62% ± 9% of PRE) than at POST-Chair (75% ± 12% of PRE). Surprisingly, Db10:100 decreased from the time when it was measured at POST-Bike (0.77 ± 0.10) to when it was measured at POST-Chair (0.65 ± 0.10). There was no change in mean Tw measured at POST-Bike (53% ± 14% of PRE) on the cycling ergometer and POST-Chair (54% ± 12% of PRE) on the isometric chair.

EMG outcome measures

Figure 3 shows the changes in M-waves, MEP, and CSP over the course of the cycling test. Mean results and percent change in mean VL and RF MEP areas, CSP, and M-wave areas measured in resting muscle, along with their respective reliability outcomes, are displayed in Table 2 (for MEP and M-wave amplitudes, see Table, Supplemental Digital Content 2, Reliability of MEP and M-wave amplitudes, There were no statistically significant interaction effects (all P > 0.457) between session and time for any EMG measures as determined from the two-way repeated-measures ANOVA. No significant differences were found between sessions for M-waves or MEP in either muscle (all P > 0.155). There was a significant time effect for M-wave area in the VL, which increased from PRE- to POST-Bike (F [2.957, 32.531] = 9.055, P < 0.001, ηp2 = 0.452). Conversely, there was no significant time effect for RF M-wave area, or VL or RF MEP (all P > 0.072).

EMG responses in VL and RF during the cycling ergometer tests for sessions 1 and 2. M-wave area in the relaxed muscle (panels A and B), MEP areas normalized to M-wave areas during MVC (panels C and D) and CSP durations (panels E and F). All values that are significantly different from PRE are denoted with *(P < 0.05).

The absolute and relative reliabilities of M-wave area in the VL at PRE were comparable between the isometric chair and the cycling ergometer (CV of 14.3% and 14.8% and ICC of 0.61 and 0.50, respectively). All MEP measured at PRE had good to excellent relative reliability, with cycling ergometer ICC of 0.88 and 0.79, and isometric chair ICC of 0.66 and 0.88, in the VL and RF, respectively. Motor-evoked potential and M-wave areas from stage 7 on the cycling ergometer showed fair to excellent relative reliability with ICC ranging from 0.51 to 0.77 as well as low CV for M-wave areas (4.7% and 5.2% for VL and RF, respectively). At POST-Bike and POST-Chair, there was comparable absolute reliability in M-wave areas, with all CV ranging between 5.1% and 7.3%. All ICC were fair to excellent (0.45–0.76).

The MEP areas of biceps femoris were minimal at PRE, that is, 0.0065 ± 0.0064 mV·s−1 (i.e., <6% and ~10% of VL and RF MEP areas, respectively) and did not increase during the protocol (i.e., 0.0058 ± 0.0074 mV·s−1 at POST-Bike).

Cortical silent period measured in the VL and RF did not change over time (both P > 0.060), and there was no main effect for session in either muscle (both P > 0.602). Measures of CSP on the cycling ergometer had excellent relative reliability at PRE-Bike (ICC: VL, 0.86; RF, 0.86) and poor relative reliability at stage 7 (ICC: VL, 0.34; RF, 0.34) and POST-Bike (ICC: VL, 0.36; RF, 0.42). Absolute reliability ranged in CV from 9.0% to 10.2% in the VL, and 8.9% to 10.4% in the RF through PRE, stage 7, and POST-Bike. In comparison, the isometric chair showed good relative reliability at PRE-Chair (ICC: VL, 0.73; RF, 0.60), and poor to good relative reliability at POST-Chair (ICC: VL, 0.18; RF, 0.56). The absolute reliability in the isometric chair had CV values ranging from 8.9% to 13.2% from PRE- to POST-Chair. Note that the lower bound of the ICC 95% CI was greater than 0.60 for half of the EMG measures on both the cycling ergometer and isometric chair.


The main findings of this study are that (i) the innovative cycling ergometer allowed for the first time neuromuscular assessments of fatigue without any recovery period after cycling, that is, a dynamic exercise with a large muscle mass; (ii) measures of voluntary and evoked knee extensor force production performed during the cycling test showed excellent relative reliability and low CV; (iii) EMG parameters were more variable but still showed good to excellent relative reliability and low CV. Overall, the new cycling ergometer represents a major methodological advance and was found to be as reliable as traditional measurements performed on an isometric chair.

The need for an innovative ergometer

To the best of our knowledge, this is the first time NMF has been assessed isometrically and intermittently during and immediately after an intermittent cycling bout (17). Many studies have investigated NMF in various muscle groups in purely isometric or isolated dynamic conditions (6). However, two main issues are that single-joint isometric (or even dynamic) conditions, although relevant in an experimental context, fail to adequately model the NMF resulting from most real-world exercises. The other experimental design, that is, combining cycling with a transfer for isometric neuromuscular assessments, results in significant time delays (sometimes up to 3 min) between task failure and measurement of neuromuscular function (5,11,32–34). This is problematic since recovery of neuromuscular function begins the moment exercise stops (13). As a result, most studies have underestimated the extent to which muscle fatigue develops during dynamic exercise with a large muscle mass. For example, Thomas et al. (5) transferred subjects from a stationary bicycle to an isometric chair to perform neuromuscular assessments on the knee extensors. These transfers are reported to have taken up to 2.5 min, meaning valuable data on the level of fatigue was not collected in that period. This same problem applies to many studies from our group, for instance (e.g., 11,12), and other research groups (e.g., 32–34). With our cycling ergometer, the pedals can be instantaneously locked to allow for neuromuscular assessment. Previous studies have attempted to examine force (or power) and/or EMG changes during and after cycling with no delay (e.g., 17–19), but no ergometer allows for the instantaneous measurement of changes in either VA or evoked forces on relaxed muscle as the present one does.

Here, the innovative cycling ergometer was used in combination with an adapted version of the QIF test (7) and subjects experienced a progressive decline in MVC from preexercise levels by 19% at stage 7 and 37% at POST-Bike. The MVC changes were mainly attributed to peripheral mechanisms of fatigue, as evidenced by reductions in Db100 (−23% at stage 7 and −38% at POST-Bike) and Tw force (−28% at stage 7 and −47% at POST-Bike). This was accompanied by low-frequency fatigue because Db10:100 decreased from 0.96 at PRE-Bike to 0.86 at stage 7 and 0.77 at POST-Bike, which suggests a failure in excitation–contraction coupling. Physiologically, low-frequency fatigue is likely due to reduced Ca2+ release from the sarcoplasmic reticulum (35). The time effect for VA was statistically significant, showing that the new cycling ergometer can detect central changes despite the lack of significant pairwise comparisons. This suggests that the central contributions to induced fatigue over the course of this cycling protocol were minimal. The lack of any statistically significant change in MEP areas and CSP durations would also support the conclusion that the fatigue induced by this cycling test was mainly peripheral. However, one needs to keep in mind that the subjects were young and healthy, that is, the current test may still be able to detect central fatigue abnormalities in elderly or clinical populations or during much longer cycling protocols, which are more likely to induce central fatigue (5).

The issue of recovery before neuromuscular assessment described above does not mean that previous experiments studying fatigue have produced irrelevant results. Yet, the innovative cycling ergometer will allow the etiology of fatigue to be revisited in cycling, that is, dynamic exercise with a large muscle mass similar to most locomotor/rehabilitation activities, and in various environmental conditions, types of exercise and different populations. A good example is the effects of hypoxia on fatigue. It has been proposed that there is a shift from peripheral to central limitations as hypoxia severity increases from moderate to severe, but to date this has not been measured immediately at task failure (36,37). Furthermore, recovery of peripheral and central fatigue has different kinetics so recovery matters not only in terms of fatigue magnitude but also in terms of fatigue etiology (15).

Reliability of force outcome measures during cycling

The main purpose of the present experiment was to test the reliability of the different measures used in this innovative testing design. Comparable reliability studies have previously been conducted in isometric conditions for both upper and lower limbs (7,38–43). In comparison to both Place et al. (42) and Bachasson et al. (7) the present innovative cycling ergometer had similar excellent reliability scores at PRE for MVC, Tw, and Db100. The absolute reliability of VA results at PRE from the present study is also in agreement with the rest of the literature (7,40,42).

In the present study, stage 7 was chosen as a point of comparison because it was the last common stage completed by all subjects, that is, it allows investigation of fatigue after a given workload. Overall, all measures at stage 7 showed excellent relative reliability and low CV. The reliabilities of the other force parameters at stage 7 in the present study are similar to those found in the QIF study halfway through the protocol (18) for Tw, Db100, and Db10:100.

The reliability of the force parameters at POST-Bike are more variable than at PRE or stage 7, but are still good to excellent and comparable to the QIF reliability (7). The relative reliability of the MVC and Tw forces of the cycling ergometer at POST-Bike was excellent (ICC of 0.91 and 0.88, respectively), and good for Db100 and Db10:100 (ICC of 0.62 and 0.67, respectively), similar to data at task failure in the QIF study. At POST-Bike, the present study and the QIF test were similar in terms of CV for MVC (6.5% vs 5.3%), Db10:100 (8.0% vs 6.4%), but slightly higher for Db100 (9.0% vs 5.1%), and Tw (11.9% vs 5.1%). Our results were collected immediately after cycling task failure and as such represent the most accurate reduction in force production capacity measured to date after a dynamic exercise with a large muscle mass. Subjects were visibly fatigued at task failure, as exhibited by uncontrollable and involuntary tremors in the legs, which could contribute to the variability in the measures at POST-Bike. Regardless, the good reliability at POST-Bike and the excellent reliability of the cycling ergometer at stage 7 demonstrate its applicability as a tool for measuring NMF induced by cycling exercise.

Comparison of force outcome measures on the cycling ergometer and isometric chair

Froyd et al. (13) showed that significant recovery in skeletal muscle function occurs within the first 1–2 min after exercise. Even quicker, recovery of EMG parameters elicited by single-pulse TMS (MEP and CSP) occurs within a few seconds after the end of exercise (44). In the present study, we performed neuromuscular assessments (assessment duration = 30 s) on the cycling ergometer immediately after the subject stopped cycling, and then moved the subject during a 1.5-min delay to the classical isometric chair setup. Exactly 2 min after cycling task failure, another neuromuscular assessment was performed in the isometric chair because this is what is usually done in classical setups. The mean MVC measured at POST-Bike on the cycling ergometer was 63% of PRE and then 2 min later on the isometric chair it was 72% of PRE, that is, a recovery of nearly 25% of the total reduction in MVC force over the course of the test. Db100 had an even greater recovery in force from 62% of PRE- at POST-Bike to 75% of PRE measured on the isometric chair. These results confirm previous findings that recovery occurs quickly and further emphasize the need to assess NMF immediately at task failure.

Surprisingly, there was no significant difference in Tw measured at POST-Bike and POST-Chair and Db10:100 was further decreased at POST-Chair compared with POST-Bike. One possible explanation is the differences in pretension in the knee extensors when the leg is attached to the force transducer on the isometric chair versus the cycling ergometer (45). Indeed, all the studies performed on the knee extensors have used some variation of a custom-built chair adjusted so that subjects’ knees and hips were at a specified degree of flexion (typically 90°), and with a force transducer attached above the malleolus of one leg (5,7,26,39,40,42,46). This isometric chair experimental setup is identical to that used in our lab and we have attempted to replicate this position when the cycling ergometer is locked in the isometric mode. Our cycling ergometer has instrumented force pedals that have been previously validated by Stapelfeldt et al. (21), with their reliability confirmed by the excellent inter-session reproducibility measured at rest in the present study. Nevertheless, pretension was probably different in the isometric chair and the changes in muscle–tendon stiffness with fatigue (47) may have compromised a valid comparison between the innovative bike and the isometric chair for the single Tw and low-frequency doublet.

Furthermore, depending on the force variables considered, values measured on the cycling ergometer were ~55 to 65% of the values measured on the isometric chair. The main reason is likely the fact that the moment arm is about 30% longer on the cycling ergometer which means that torque was in reality approximately 10% to 20% lower on the cycling ergometer than on the chair. The remaining difference can probably be attributed to the reduced stability on the cycling ergometer versus the chair.

Reliability of EMG outcome measures

Motor-evoked potential and CSP found in the present studies are in line with those previously found in the VL and RF muscles in our own experiments, for instance (22,48), as well as others (19,49). Overall, the EMG outcome measures were less reliable than the force parameters but were still good to excellent. The absolute reliability at PRE in the present study was better than that in Thomas et al. (5) for VL M-wave and MEP areas. The ICC for M-wave area at PRE reported by Thomas et al. (5) was excellent, whereas on the cycling ergometer, it was only good. This may be due to the fact that Thomas et al. (5) marked electrode placement with ink to ensure consistent placement while we decided not to do that to be more consistent with future experiments using the innovative cycling ergometer. For instance, when assessing the effects of a training intervention lasting several weeks, it is impractical to mark the electrode position on the skin. The ICC was excellent (0.88) in the present study for MEP area, whereas only good (0.74) in Thomas et al. (5). These reproducibility results are similar to those found in other reliability studies of EMG measures on the knee extensors, such as M-wave amplitude assessed by Place et al. (42) (ICC, 0.71; CV, 12.2%) and O’Leary et al. (41) (ICC, 0.62; CV, 11.5%).

The relative reliability of the M-wave and MEP areas was good to excellent at stage 7. Absolute reliability for the VL M-wave area was low with a CV of 4.7%, whereas the VL MEP area CV was 21.7%. Place et al. (42) studied the reliability of M-wave amplitude changes for the VL and RF muscles after isometric exercise and reported, at POST, that ICC were 0.70 and 0.87 and CV were 17.0% and 13.0%, respectively. Although the present study looked at M-wave area rather than amplitude, we showed better absolute reliability at POST-Bike but worse relative reliability. Reliability of MEP areas at POST-Bike for both VL and RF had CV of 13.3% and 15.8%, and good to excellent ICC of 0.76 and 0.59. This is the first time that such measures have been performed immediately after cycling task failure.

The poor reliability for CSP could be explained by the experimental design. Indeed, the determination of TMS intensity was based on MEP response curves, not CSP responses (50), probably explaining why this parameter, known to be very dependent on stimulus intensity, shows a range in reliability (CV from 9.0% to 10.4%; ICC from 0.36 to 0.86). Although mean TMS intensity was not different between the two sessions, it is not known if this would have been the same if CSP responses had been used to determine TMS intensity. Thus, if TMS intensity had been based on CSP response curves rather than MEP, it is likely both absolute and relative reliabilities would have been better for this measure. This needs to be confirmed in future studies, but if this is correct, it means that researchers may have to choose between reliable MEP and reliable CSP.


Although measuring fatigue from an ecological exercise such as cycling immediately at task failure is a major methodological advance, the cycling ergometer is still only assessing neuromuscular parameters when it is in isometric mode. Further research must investigate the reliability of the force–velocity relationship and neuromuscular parameters (e.g., Tw response) in dynamic mode on the cycling ergometer. Also, although the subjects were systematically instructed to keep the ankle fixed during the isometric MVC, this was not ensured mechanically. Using a brace is a possibility, but because it would take several seconds to install, this solution was not retained. The excellent reliability of force responses at rest and during fatigue suggest that with proper familiarization and practice, maintaining a fixed ankle is not a major issue (see document, Supplemental Digital Content 3, Complementary experiment comparing the force produced on the cycling ergometer from immobilized versus free ankles, The innovative cycling ergometer uses recumbent cycling (Fig. 1) which may be problematic to assess fatigue in competitive cyclists due to the lack of movement specificity. Yet it was designed to assess NMF during and immediately after dynamic exercise with a large muscle mass, not during cycling per se, which is a major asset for testing many populations, especially in a clinical setting. The recumbent bike is likely more suitable for subjects from clinical populations because of increased stability and accessibility, and because it provides better access to the head for investigators delivering TMS. Another potential limitation of this reliability study is that it was conducted on a young and healthy population whose tolerance to electrical stimulation and TMS may not be generalizable to elderly and clinical populations. Further research is required to determine the feasibility of using the cycling ergometer with elderly and clinical subjects.


Previous studies using cycling as the fatiguing exercise may have (i) underestimated the extent to which muscle fatigue develops, (ii) not been able to adequately examine the causes of exercise termination, and (iii) not been able to determine the time course of central and peripheral fatigue. The results of the present study suggest that an innovative cycling ergometer, which aims to address these issues, is a reliable tool for assessing neuromuscular function and fatigue in the knee extensors before, during, and immediately at task failure. The force measures used to assess neuromuscular function were highly reliable and showed comparable, if not superior, reproducibility to the current criterion standard assessments performed in single-joint isometric conditions. Although, the EMG parameters were more variable than force measures, they were still good in comparison to similar measures in the literature and to those recorded on the isometric chair from the present study. This new cycling ergometer will allow for the reliable investigation of neuromuscular function and fatigue in cycling, during and immediately at task failure, and permit the characterization of the time course of recovery from dynamic exercise. Thus, it has real-world applications by allowing the etiology of fatigue in cycling to be revisited in various environmental conditions (e.g., hypoxia, heat), types of exercise (e.g., prolonged cycling, repeated sprints), and different athletic and clinical populations.

Special thanks are given to Andrzej Stano and the Science Workshop of the University of Calgary for their help in designing the new cycling ergometer and to Dr. Saied Jalal Aboodarda for his assistance with data analysis.

This project was funded by The University of Calgary (URGC research grant 1029914).

The authors have no conflict of interest in relation to the present scientific paper. The results are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation, and do not constitute endorsement by the American College of Sports Medicine.

G. M., R. H., and J. T. developed the study methodology, D. B., M. M., and J. T. collected the data. D. B. analyzed the results. D. B., J. T., and G. M. drafted the article. All authors reviewed and revised the work. All authors reviewed the final article and approved it for submission.


1. Gandevia SC. Spinal and supraspinal factors in human muscle fatigue. Physiol Rev. 2001;81(4):1725–89.
2. MacIntosh BR, Rassier DE. What is fatigue? Can J Appl Physiol. 2002;27(1):42–55.
3. Place N, Yamada T, Bruton JD, Westerblad H. Muscle fatigue: from observations in humans to underlying mechanisms studied in intact single muscle fibres. Eur J Appl Physiol. 2010;110(1):1–15.
4. 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.
5. Thomas K, Goodall S, Stone M, Howatson G, St Clair Gibson A, Ansley L. Central and peripheral fatigue in male cyclists after 4-, 20-, and 40-km time trials. Med Sci Sports Exerc. 2015;47(3):537–46.
6. Millet GY, Martin V, Martin A, Vergès S. Electrical stimulation for testing neuromuscular function: From sport to pathology. Eur J Appl Physiol. 2011;111(10):2489–500.
7. Bachasson D, Millet GY, Decorte N, Wuyam B, Levy P, Verges S. Quadriceps function assessment using an incremental test and magnetic neurostimulation: a reliability study. J Electromyogr Kinesiol. 2013;23(3):649–58.
8. Pageaux B, Lepers R, Marcora SM. Reliability of a novel high intensity one leg dynamic exercise protocol to measure muscle endurance. PLoS One. 2016;11(10):e0163979.
9. Cheng AJ, Rice CL. Fatigue and recovery of power and isometric torque following isotonic knee extensions. J Appl Physiol. 2005;99:1446–52.
10. Rossman MJ, Venturelli M, McDaniel J, Amann M, Richardson RS. Muscle mass and peripheral fatigue: a potential role for afferent feedback? Acta Physiol (Oxf). 2012;206(4):242–50.
11. Jubeau M, Rupp T, Perrey S, et al. Changes in voluntary activation assessed by transcranial magnetic stimulation during prolonged cycling exercise. PLoS One. 2014;9(2):e89157.
12. Lepers R, Maffiuletti NA, Rochette L, Brugniaux J, Millet GY. Neuromuscular fatigue during a long-duration cycling exercise. J Appl Physiol (1985). 2002;92(4):1487–93.
    13. 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(5):1339–46.
    14. Pageaux B, Angius L, Hopker JG, Lepers R, Marcora SM. Central alterations of neuromuscular function and feedback from group III–IV muscle afferents following exhaustive high-intensity one-leg dynamic exercise. Am J Physiol Regul Integr Comp Physiol. 2015;308(12):R1008–20.
    15. Carroll TJ, Taylor JL, Gandevia SC. Recovery of central and peripheral neuromuscular fatigue after exercise. J Appl Physiol. 2016;122(5):1068–76.
    16. Allen D, Westerblad H. What limits exercise during high-intensity aerobic exercise? Eur J Appl Physiol. 2010;110(3):661–2.
    17. Coelho AC, Cannon DT, Cao R, et al. Instantaneous quantification of skeletal muscle activation, power production, and fatigue during cycle ergometry. J Appl Physiol (1985). 2015;118(5):646–54.
    18. Siegler JC, Marshall PW, Raftry S, et al. The differential effect of metabolic alkalosis on maximum force and rate of force development during repeated, high-intensity cycling. J Appl Physiol (1985). 2013;115(11):1634–40.
    19. Sidhu SK, Cresswell AG, Carroll TJ. Motor cortex excitability does not increase during sustained cycling exercise to volitional exhaustion. J Appl Physiol. 2012;113(3):401–9.
    20. Andersen P, Adams RP, Sjogaard G, Thorboe A, Saltin B. Dynamic knee extension as model for study of isolated exercising muscle in humans. J Appl Physiol (1985). 1985;59(5):1647–53.
    21. Stapelfeldt B, Mornieux G, Oberheim R, Belli A, Gollhofer A. Development and evaluation of a new bicycle instrument for measurements of pedal forces and power output in cycling. Int J Sports Med. 2007;28(4):326–32.
    22. Temesi J, Rupp T, Martin V, et al. Central fatigue assessed by transcranial magnetic stimulation in ultratrail running. Med Sci Sports Exerc. 2014;46(6):1166–75.
    23. Temesi J, Gruet M, Rupp T, Verges S, Millet GY. Resting and active motor thresholds versus stimulus -response curves to determine transcranial magnetic stimulation intensity in quadriceps femoris. J Neuroeng Rehabil. 2014;11(40):1–13.
    24. Taylor JL, Butler JE, Allen GM, Gandevia SC. Changes in motor cortical excitability during human muscle fatigue. J Physiol. 1996;490(2):519–28.
    25. Edwards RH, Hill DK, Jones DA, Merton PA. Fatigue of long duration in human skeletal muscle after exercise. J Physiol. 1977;272(3):769–78.
    26. Verges S, Maffiuletti NA, Kerherve H, Decorte N, Wuyam B, Millet GY. Comparison of electrical and magnetic stimulations to assess quadriceps muscle function. J Appl Physiol. 2009;106(2):701–10.
    27. Field AP (ed). 4th ed. London: SAGE Publications, Limited; 2015.
    28. Hopkins WG. Measures of reliability in sports medicine and science. Sports Med. 2000;30(1):1–15.
    29. Cicchetti DV, Sparrow SA. Developing criteria for establishing interrater reliability of specific items: applications to assessment of adaptive behavior. Am J Ment Defic. 1981;6(4):284–90.
    30. Todd G, Gorman RB, Gandevia SC. Measurement and reproducibility of strength and voluntary activation of lower-limb muscles. Muscle Nerve. 2004;29(6):834–42.
    31. Hopkins WG. Reliability from consecutive pairs of trials (Excel spreadsheet). 2002 June 1, 2016 [cited 2016, June 1]. Available from:
    32. Pearcey GE, Murphy JR, Behm DG, Hay DC, Power KE, Button DC. Neuromuscular fatigue of the knee extensors during repeated maximal intensity intermittent-sprints on a cycle ergometer. Muscle Nerve. 2015;51(4):569–79.
    33. Weavil JC, Sidhu SK, Mangum TS, Richardson RS, Amann M. Fatigue diminishes motoneuronal excitability during cycling exercise. J Neurophysiol. 2016;116(4):1743–51.
    34. Sidhu SK, Weavil JC, Mangum TS, et al. Group III/IV locomotor muscle afferents alter motor cortical and corticospinal excitability and promote central fatigue during cycling exercise. Clin Neurophysiol. 2017;128(1):44–55.
    35. Westerblad H, Duty S, Allen DG. Intracellular calcium concentration during low-frequency fatigue in isolated single fibers of mouse skeletal muscle. J Appl Physiol (1985). 1993;75(1):382–8.
    36. 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 (1985). 2012;112:1335–44.
    37. 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(1):389–403.
    38. Souron R, Farabet A, Millet GY, Lapole T. Reliability of the functional measures of the corticospinal pathways to dorsiflexor muscles during maximal voluntary contractions. J Neurol Sci. 2016;369:368–74.
    39. Goodall S, Romer L, Ross E. Voluntary activation of human knee extensors measured using transcranial magnetic stimulation. Exp Physiol. 2009;94(9):995–1004.
    40. 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.
    41. O’Leary TJ, Morris MG, Collett J, Howells K. Reliability of single and paired-pulse transcranial magnetic stimulation in the vastus lateralis muscle. Muscle Nerve. 2015;52(4):605–15.
    42. Place N, Maffiuletti NA, Martin A, Lepers R. Assessment of the reliability of central and peripheral fatigue after sustained maximal voluntary contraction of the quadriceps muscle. Muscle Nerve. 2007;35(4):486–95.
    43. Allen GM, Gandevia SC, McKenzie DK. Reliability of measurements of muscle strength and voluntary activation using twitch interpolation. Muscle Nerve. 1995;18(6):593–600.
    44. Taylor JL, Allen GM, Butler JE, Gandevia SC. Supraspinal fatigue during intermittent maximal voluntary contractions of the human elbow flexors. J Appl Physiol (1985). 2000;89(1):305–13.
    45. Viitasalo JT. Effects of pretension on isometric force production. Int J Sports Med. 1982;3(3):149–52.
    46. Bachasson D, Guinot M, Wuyam B, et al. Neuromuscular fatigue and exercise capacity in fibromyalgia syndrome. Arthritis Care Res. 2013;65(3):432–40.
    47. Morales-Artacho AJ, Lacourpaille L, Guilhem G. Effects of warm-up on hamstring muscles stiffness: cycling vs foam rolling. Scand J Med Sci Sports [Internet]. 2017 01/2017:[1–11 pp.].
    48. Temesi J, Arnal PJ, Davranche K, et al. Does central fatigue explain reduced cycling after complete sleep deprivation? Med Sci Sports Exerc. 2013;45(12):2243–53.
    49. Kennedy DS, McNeil CJ, Gandevia SC, Taylor JL. Effects of fatigue on corticospinal excitability of the human knee extensors. Exp Physiol. 2016;101(12):1552–64.
    50. Kimiskidis VK, Papagiannopoulos S, Sotirakoglou K, Kazis DA, Kazis A, Mills KR. Silent period to transcranial magnetic stimulation: construction and properties of stimulus–response curves in healthy volunteers. Exp Brain Res. 2005;163(1):21–31.


    Supplemental Digital Content

    © 2018 American College of Sports Medicine