Fatigue and recovery measured with dynamic properties versus isometric force: effects of exercise intensity

One of the most common definitions of neuromuscular (NM) fatigue is an exercise-related decrease in the maximal voluntary force or power output (Bigland-Ritchie et al., 1986; Enoka, 2012). The vast majority of fatigue investigations, including our own studies, have only considered the decrement in isometric maximal voluntary contraction (IMVC) as a fatigue index (e.g. Behm and St-Pierre, 1997; Bigland-Ritchie and Woods, 1984; Doyle-Baker et al., 2018; Froyd et al., 2016; Gandevia et al., 1996; Kennedy et al., 2016; MacIntosh et al., 1994; Morris et al., 2012; Temesi et al., 2014; Thomas et al., 2016; Wüthrich et al., 2014). In fact, in our recently published meta-analysis (Kruger et al., 2018), we were able to identify 29 experimental studies that have examined the effects of age on IMVC loss after NM fatigue versus only 11 studies that have measured power output decrements. One reason for using almost exclusively IMVC as an index of fatigue is probably the simplicity of this measurement, as isometric transducers are more accessible and less expensive than dynamic ergometers. Another reason is that isometric contractions in combination with peripheral nerve/transcranial magnetic stimulation allow investigation of central (voluntary activation) and peripheral (electrically evoked force responses on relaxed muscles) fatigue (Millet et al., 2011). Yet, the assessment of changes in the maximal power output may be equally critical to quantify the impact of a fatiguing task, particularly after dynamic exercise, because of task specificity. Some experimental studies have shown that NM function changes with fatigue can be accompanied by a greater reduction in maximum power output and/or dynamic force production at high velocity than is typically seen for isometric force (Cheng and Rice, 2005, 2009; Westerblad et al., 1998). For instance, Cheng and Rice (2005) reported a ∼46% drop in power output following isotonic single-limb contractions, while isometric force decreased by ∼26%. This apparent discrepancy was related to the fact that power is dependent on both force and velocity, both of which can change with fatigue. Therefore, the assessment of isometric force alone can give an incomplete or misleading interpretation of the overall consequences of fatigue.

Although the assessment of power can give rise to additional insights into NM function that would not be discerned with the measurement of isometric force alone, most of the published studies considering dynamic measures of fatigue have used an isotonic or isokinetic dynamometer to measure power during single-limb tasks (Cheng and Rice, 2005, 2009; Dalton et al., 2015; Wallace et al., 2016). Investigating power responses during dynamic exercises with large muscle mass (e.g. cycling) is relevant because of its functional association with daily activities. For this reason, we recently developed an innovative cycling ergometer that can measure NM fatigue in both dynamic and isometric mode immediately after the cessation of exercise (Doyle-Baker et al., 2018). From evaluation of a single all-out acceleration, torque–velocity (T–V) and power–velocity (P–V) relationships can be obtained. These relationships allow calculation of the maximal theoretical torque (T_max, representing maximal force production capability) and angular velocity (V_max, representing maximal angular velocity until force can be produced), as well as the maximal power output (P_max), which combines both measurements (Samozino et al., 2007). These theoretical values represent the two extremes of the T–V spectrum and well characterize the dynamic force production capabilities at low (T_max) and high (V_max) velocities. They give a better estimation of the maximal dynamic force production capabilities than the instantaneous torques measured during a cycling test as the instantaneous torques largely depend on the resistance set during the sprints, which is not the case for T_max and V_max. Moreover, our cycle ergometer has lockable pedals that allow switching from cycling to isometric contraction modes within seconds (Doyle-Baker et al., 2018).

As stated above, only a few studies have considered both isometric force and power responses after fatiguing tasks (Cheng and Rice, 2005, 2009) and, to the best of our knowledge, no study has examined both IMVC and power output following dynamic exercises with a large muscle mass performed at different intensities/durations. In addition, no study has yet compared the responses of dynamic assessment – P_max, T_max and V_max – with isometric measures of fatigue – IMVC and maximal rate of force development (MRFD) – following different fatiguing cycling exercises. Therefore, the aim of this study was to compare dynamic (T_max, V_max and P_max) with isometric (IMVC and MRFD) measures of fatigue immediately after various intensities of dynamic fatiguing exercise and during the subsequent recovery period.

We compared dynamic properties and isometric force in response to three distinct cycling fatiguing exercises [30 s all-out intensity (AI), 10 min at severe intensity (SI) and 90 min at moderate intensity (MI)] to examine whether the differences and/or similarities between isometric and dynamic properties vary according to the intensity/duration of the protocol. As previously discussed and shown in the literature (Cheng and Rice, 2005, 2009; Dalton et al., 2015; de Haan et al., 1989), we expected that IMVC and P_max would decrease differently after the fatiguing exercises, but the effects of different exercise intensity on the diversity of responses still required investigation. As both IMVC and T_max represent the capacity to produce maximal force and both V_max and MRFD are measures of explosiveness, we would expect a similar magnitude of changes of these properties in response to fatigue. Again, whether this is true following exercise at different intensities is not known. However, studies that have assessed either dynamic or isometric measures of fatigue have shown that these assessments are affected differently, according to the exercise intensity–duration relationship (Black et al., 2017; Hureau et al., 2014; Morris et al., 2012; O'Bryan et al., 2017; Thomas et al., 2016, 2015). Thus, we expected changes in fatigue outcome of different amplitude after the different exercise intensities/durations. For example, fatigue induced by AI exercise may cause a greater decrease in V_max and, by consequence, P_max than the other fatiguing exercises because of an enhancement in [ADP], which has been associated with a reduction in maximal velocity (Westerblad et al., 1998).

Written informed consent was obtained from each participant before beginning the study. This study was approved by the Conjoint Health Research Ethics Board of the University of Calgary (REB #15-2430) and it was conducted according to the guidelines of the Declaration of Helsinki, except for registration in a database.

Ten young adult males volunteered to participate in this study [mean±s.d. age: 27±4 years; height: 181±6 cm; body mass: 79±9 kg; maximal oxygen uptake (V̇_O2,max): 49±7 ml kg⁻¹ min⁻¹]. To be included in this study, participants needed to be healthy, non-smokers, non-obese and without chronic metabolic disease. All participants were healthy according to the Physical Activity Readiness Questionnaire (PAR-Q). Physical activity status was assessed with the Godin Leisure Time Exercise Questionnaire. The participants were classified as physically active (exercise duration >150 min per week of moderate to vigorous intensity exercise) consistent with the Canadian Physical Activity Guidelines (https://csepguidelines.ca). The participants were instructed to refrain from consuming caffeine and alcohol for 12 h prior to testing sessions and from participating in any strenuous exercise the day before testing. The characteristics of the participants are presented in Table 1.

Participants visited the laboratory on seven different occasions. During the first visit, participants performed an incremental exercise test on a recumbent cycle ergometer to volitional exhaustion while their oxygen uptake was measured by open circuit spirometry (Quark CPET, COSMED, Rome, Italy). The initial power output was 90 W with 15 W increments every minute. The Quark CPET metabolic cart was used to determine first ventilatory threshold (VT1), second ventilatory threshold (VT2), and V̇_O2,max. V̇_O2,max was determined to have been achieved when a V̇_O2 plateau was observed (when the V̇_O2 varied by less than 1.5 ml kg⁻¹ min⁻¹ with an increase in workload) (Cunha et al., 2011). If there was no plateau, the V̇_O2 peak was considered as the highest V̇_O2 over 30 s. VT1 was identified as the minimum workload at which the ventilatory equivalent ratio for oxygen (V̇_E/V̇_O2) systematically increased without an increase in the ventilatory equivalent ratio for carbon dioxide (V̇_E/V̇_CO2) and VT2 as the lowest workload where V̇_E/V̇_O2 and V̇_E/V̇_CO2 both increased. VT1, VT2 and V̇_O2,max were obtained by visual inspection of graphs by two independent observers (R.L.K. and L.M.J.). Disagreements were resolved by a third author (G.Y.M.). After the incremental cycling test, the participants were familiarized with the isometric NM function assessment and the T–V cycling sprint test.

Subsequently, participants randomly completed the six experimental sessions separated by at least 3 days. The responses to three durations of cycling exercises (30 s all-out intensity, 10 min at severe intensity and 90 min at moderate intensity) were assessed by either isometric or dynamic measures of fatigue. The measurements were performed before (Pre), after (Post) and at +1, +2, +4 and +8 min after each cycling task.

Participants performed three different durations of exercise on a recumbent cycling ergometer adapted from a Velotron ergometer (Racer Mate, Seattle, WA, USA). For the AI exercise trial, participants were instructed to cycle as fast and as hard as possible against the same resistance previously determined for the T–V sprint test for 30 s. The SI cycling task was performed at a power output corresponding to 5% above VT2. Participants were instructed to keep the cadence constant between 60 and 80 rpm. If they were not able to keep the cadence above 60 rpm (n=3), power output was slightly reduced (by 10 W) so that they could finish the task. When this happened, the same power output reduction was applied during the other SI session. For the MI session, intensity was determined as the power output corresponding to 20% below VT1. Participants were also instructed to keep the cadence constant at their preferred rate (between 60 and 80 rpm). In the SI and MI conditions, participants were instructed to choose a cadence between 60 and 80 rpm and to keep it constant during the exercise. The average cadence sustained during the SI and MI tasks was 76±5 and 76±6 rpm, respectively. Rating of perceived exertion (RPE) was again assessed within the first revolutions and within the last 30 s of exercise. For the AI exercise, RPE was assessed before and immediately after cessation of the task.

The NM function assessment was performed on the same recently validated recumbent cycling ergometer (Doyle-Baker et al., 2018). This ergometer includes instrumented pedals (Model PF1.0.0, Radlabor GmbH, Freiburg, Germany) that measure bilateral force applied to the crank in two directions (radial and tangential). The pedals can be locked at the 90 deg knee position to allow the subject to perform a unilateral isometric contraction similar to that which can be performed in an isometric chair (Doyle-Baker et al., 2018). In this study, the NM function test was conducted with the right leg only. To perform IMVCs with the right knee extensors, the left pedal needs to be locked at a fixed leg position, allowing the right knee and ankle to be at the 90 deg position. During the IMVC, the individuals were secured by non-compliant straps at the hip and chest and were instructed to place their hands across their chest. The participants performed a 5 min standard warm-up, which consisted of cycling for 3 min at 30%, 1 min at 45% and 1 min at 60% of the power output corresponding to the beginning of the V̇_O2,max plateau. Following the warm-up, the first NM evaluation was carried out (Fig. 1A), without moving from the ergometer. First, an explosive IMVC was performed as fast and as hard as possible (<1 s) to measure MRFD. Participants were instructed to contract as fast as possible and to quickly relax right after reaching the maximal force. One second after full relaxation, participants performed a 5 s IMVC during which they attempted to contract as hard as possible and to maintain this level for ∼5 s while receiving strong verbal encouragement. MRFD and IMVC were not measured during the same contraction because many subjects have difficulty maintaining a force plateau when contracting as fast as possible. Therefore, the assessment of both outcomes in the same contraction could jeopardize the voluntary activation measurement (data not shown) by reducing the possibility of reaching a plateau during the IMVC. The pre-assessment was performed twice with 2 min of rest between repeats. If there was a noticeable difference (∼5–10%) in the IMVC, we asked the participant to repeat the sequence. The participants then performed one of the three fatiguing cycling exercises. At 10 s (Post) and 1, 2, 4 and 8 min after the end of the exercise, the participants repeated the same NM function assessments as prior to the trial. The 10 s delay between exercise cessation and the beginning of the post-exercise measurement was intentionally chosen to match the timeline of the ‘Post’ time point for the T–V sprint test. Force data were collected from the crank sensors at 500 Hz during the experiment using Imago Record (v.8.50; Radlabor GmbH). These data were saved and exported to Labchart 8 software (ADInstruments, Bella Vista, Australia) for later analysis.

All-out 7 s sprints (Fig. 1B) were performed on the same electronically braked recumbent cycling ergometer (Velotron) equipped with an eddy current braking system, which allows control of the resistance directly from the software. The participants performed two 7 s sprints against a resistance which was equivalent to 5% or 7% of body weight. The individual resistance was chosen based on the goodness of fit of the T–V and P–V relationships: highest coefficient of determination (R²), a greater range of velocities covered by experimental points (i.e. pedal downstrokes), and points distributed as equally as possible to both sides of the P–V relationship curve. This individual resistance would allow the subjects to achieve maximal power at about 50% of the sprint duration. Sprints were separated by 2 min rest. During the 7 s sprints, the participants were vigorously encouraged to pedal as fast as possible, remaining seated on the saddle and secured by chest and hip straps. The sprints were recorded with a video camera (Fujifilm FinePix XP80 Waterproof Digital Camera with 2.7 inch LCD, Tokyo, Japan) with a 240 Hz sampling frequency. Reflective markers were placed on the crank and on the pedal to measure angular displacement during the sprint. The camera was synchronized with the Labchart 8 software with a light signal. After 3 min of rest, one of the three cycling exercises was performed. At 10 s (Post) and +1, +2, +4 and +8 min after the end of the exercise, the 7 s sprint was repeated. This 10 s delay was necessary to bring the flywheel to a complete stop from which the subsequent acceleration was initiated. During the 7 s sprint cycling tests, the instantaneous torque at the pedal crank was obtained from the instrumented pedals (sum of the tangential force of both pedals multiplied by the crank length) and recorded at 240 Hz in the Imago software, and the instantaneous angular velocity was calculated from the camera and Kinovea software after tracking the crank angular displacement at each 4.17 ms (240 Hz). Power was then calculated as the product of instantaneous torque and angular velocity.

After the MI condition, one individual could not achieve optimal velocity during the T–V relationship tests. The lower ability to rapidly overcome the flywheel resistance after NM fatigue compromised the dynamic properties assessment. Consequently, V_max was highly extrapolated from the T–V relationship curve, which increased the inaccuracy of not only V_max but also P_max computation. Because of that, this subject was excluded from this analysis.

The highest IMVC and MRFD were recorded before the cycling task (Pre). MRFD was measured as the maximal slope of the force–time curve (Aagaard et al., 2002) (Fig. 1A). Instantaneous resultant torque, angular velocity and power output recorded at 240 Hz during cycling were averaged from each pedal downstroke of the acceleration phase (Samozino et al., 2007). Individual T–V relationships were determined from pedal downstroke averaged values using least-square linear regression (Samozino et al., 2007). Therefore, each point in the T–V and P–V relationship curves represents a pedal downstroke (Fig. 1B). T–V relationships were extrapolated to obtain T_max (the theoretical maximal torque that the lower limbs could produce) and V_max (the theoretical maximal velocity at which the lower limbs could still produce torque) as the intercepts of the T–V relationships with the force and velocity axis, respectively. P_max was determined as the product of T_max and V_max divided by 4 (Vandewalle et al., 1987). The dataset is given in Tables S1–S5.

All data were analyzed using the Statistical Package for Social Sciences (SPSS) version 24 (SPSS Inc., Chicago, IL, USA). Descriptive statistics (mean±s.d.) were used to describe the data. Levene's test was used to test the homogeneity of variance. Mauchly's sphericity test was used to assess the sphericity assumption. If sphericity was not confirmed, the Greenhouse–Geisser correction factor was applied. A one-way analysis of variance (ANOVA) was used to test reproducibility of the NM function and T–V sprint tests at baseline across the three testing sessions. Relative reliability was assessed through coefficient of variation (CV) and intraclass correlation coefficients (ICC) and the absolute reliability was tested through the calculation of typical error of measurement (TEM) (Hopkins, 2000).

Values obtained after exercise (Post) and during recovery were normalized to the highest baseline measures collected at the beginning of each session (Pre). To compare the changes of dynamic measures versus isometric force production properties after fatigue and during recovery (IMVC versus P_max, IMVC versus T_max, and V_max versus MRFD), two-way repeated measures ANOVA performed on the relative decrease (Post, and +1, +2, +4 and +8 min) were used for each type of fatiguing exercise. Spearman (correlations with IMVC) and Pearson (all other correlations) correlation coefficients were calculated to explore the associations between dynamic measures versus isometric force production properties (i.e. IMVC versus P_max, IMVC versus T_max and V_max versus MRFD). A one-way ANOVA followed by Bonferroni post hoc test was used to compare the relative changes for each dependent variable (IMVC, MRFD, P_max, V_max and T_max) at the end of the exercise (Post) between three types of fatiguing exercise. Then, a two-way repeated measures ANOVA was conducted to compare the influence of the three cycling exercises on each dependent variable (IMVC, MRFD, P_max, V_max and T_max) during the recovery (Post, +1 min, +2 min, +4 min, and +8 min). For the two-way ANOVA, when a main effect of time or fatiguing task was observed, Bonferroni post hoc tests were applied. If there was an interaction, multiple paired t-tests were performed followed by Holm–Bonferroni correction for multiple comparisons. Effect sizes for the ANOVA were reported as partial eta-squared (partial η²). For hypothesis testing, the 95% confidence level was predetermined as the minimum criterion to denote a statistical difference (P≤0.05).

Reliability of NM function and T–V tests at baseline (Pre) is presented in Table 2. There was no difference at baseline between the three exercise conditions for any variable. Mean power outputs for the three fatiguing exercises are presented in Fig. 1C.

There was a time×variable interaction after the AI exercise (F_4.36=13.035, P<0.001, partial η²=0.592; Fig. 2A), showing that P_max decreased more than IMVC after exercise (Post) (P=0.005) and at +1 min of recovery (P=0.048), but the difference was no longer significant from +2 min. Conversely, IMVC was more reduced than P_max at the ‘Post’ time point and during recovery from the SI exercise (variable, F_1,9=15.700, P=0.003, partial η²=0.636; Fig. 2A) and the MI exercise (variable, F_1,8=77.809, P<0.001, partial η²=0.907; Fig. 2A) without any time×variable interaction. There was no correlation between the decrease in IMVC and the decrease in P_max (r=0.079, P=0.683; Fig. 3A).

There was no difference between the decrease in IMVC and the decrease in T_max after the AI exercise (Fig. 2B). After the SI exercise, there was a time×variable (IMVC and T_max) interaction (F_4,36=3.738, P=0.012, partial η²=0.293; Fig. 2B). IMVC was significantly more reduced than T_max after exercise (Post) and at +1 min and +2 min of recovery (P<0.042), i.e. the differences between the two variables tended to shrink during recovery. IMVC was also more reduced than T_max after exercise (Post) and during recovery after the MI task (variable, F_1,8=36.138, P<0.001, partial η²=0.819; Fig. 2B) without any time×variable interaction. There was no correlation between the decrease in IMVC and the decrease in T_max (r=−0.070, P=0.720; Fig. 3B).

There was a time×variable interaction for the AI task (F_4,36=9.748, P<0.001, partial η²=0.520; Fig. 2C), SI task (F_4,36=7.953, P<0.001, partial η²=0.469; Fig. 2C) and MI task (F_4,36=2.859, P=0.039, partial η²=0.263; Fig. 2C). MRFD demonstrated a much greater depression than V_max after exercise (Post) and during recovery for all the fatiguing exercises (all P<0.05). There was a moderate but significant correlation between the reduction in V_max and the reduction in MRFD (r=0.426, P=0.021; Fig. 3C).

There was a significant difference in the relative change in the measured variables (Pre to Post) between the exercise conditions for IMVC (F_2,29=8.810, P=0.001; Fig. 4A), P_max (F_2,28=32.546, P<0.001; Fig. 4B) and V_max (F_2,28=24.572, P<0.01; Fig. 4C) but not for T_max and MRFD (Fig. 4D,E). IMVC was lower after the SI trial (62±12% of baseline) than after both the AI (81±8% of baseline) and MI trials (79±13% of baseline) (P<0.006). The reduction of P_max was greater after the AI trial (to 58±12% of baseline), followed by the SI trial (to 74±8% of baseline) and then by the MI trial (to 94±10% of baseline) (all P<0.01). V_max decreased more after the AI trial (to 71±10% of baseline), followed by the SI trial (to 87±11% of baseline) and was not affected by the MI trial (to 103±9% of baseline) (P<0.005).

During recovery, there was a time×fatiguing exercise interaction for IMVC (F_8,72=2.973, P=0.006, partial η²=0.248; Fig. 4A), P_max (F_8,64=10.631, P<0.001, partial η²=0.571; Fig. 4B) and V_max (F_8,64=3.899, P<0.001, partial η²=0.328; Fig. 4C) but there was neither an interaction nor an effect of time or fatiguing exercise for T_max and MRFD (Fig. 4D,E).

There was a time×condition interaction for RPE (F_2.59=9.726; P<0.001). Although at the end of the exercises RPE was slightly increased, it was significantly greater after the AI (18±3) and the SI (19±1) tasks than following the MI exercise (15±4) (P=0.009 and P<0.001, respectively).

This is the first investigation to compare the magnitude of NM fatigue quantified by dynamic and isometric assessment immediately after dynamic fatiguing exercise of different intensity and duration as well as during recovery for a large muscle mass. The main findings of this study are that dynamic properties and isometric force present divergent responses to fatigue induced by cycling, suggesting distinct mechanisms for decline in maximal velocity, power, force and rate of force development. Interestingly, the relative comparisons between the different indices vary between exercise durations. Therefore, and despite the advantage that isometric measurements allow the use of evoked stimulation to understand fatigue etiology, the assessment of isometric force only after fatiguing exercise may result in an underestimation of the functional impairment after fatigue, particularly after very severe-intensity exercises (e.g. AI), in which P_max is much more reduced than IMVC. The different patterns of decline and recovery of the dynamic and isometric variables following distinct types of exercise further suggest that these measures of fatigue do not entirely share the same physiological mechanisms. Thus, dynamic and isometric assessments are not interchangeable but instead complement each other to better understand the consequences of fatiguing exercise on NM properties.

IMVC decreased the most after the SI exercise and showed some recovery in the following 8 min but appeared to recover only modestly after the AI and not at all after MI exercise (Fig. 2A). This is a complex pattern of change that can only be explained by different physiological mechanisms of fatigue being associated with the different exercises. In contrast, P_max decreased the most after the AI test and least after the MI trial. Peak power output is the product of torque and velocity at the optimal torque and corresponding optimal velocity. For this reason, changes in maximal torque and maximal velocity will influence the decline in peak power. Considering that the relationship between torque and angular velocity for cycling is linear (MacIntosh et al., 2004), P_max occurs at 0.5 of apparent maximum torque and 0.5 of apparent maximum velocity. For this reason, the decrease in P_max should be proportional to the product of the relative values for T_max and V_max. In confirmation, this was the case with our data, e.g. the impact of the AI test on V_max (decreased 27% to 0.73 of original), T_max (decreased 18% to 0.82 of original) and P_max (was 58% of original; compare with 0.73×0.82=0.599 or 60%) after exercise (Post) (Fig. 4). For this exercise, recovery of P_max (−18% after 8 min) relies on recovery of V_max (to +4%) while T_max (−21%) does not appear to change much during the recovery period. Therefore, the greater decrease in P_max after the AI exercise is mainly explained by the fact that T_max decreased similarly across the exercises whereas V_max decreased more after the AI exercise (Fig. 4). This may be related to the greater fatigue in fast-twitch motor units which would be expected to be most pronounced during the AI trial (Gregor et al., 1979; Mannion et al., 1995), because this trial required the greatest power output and most likely included recruitment of fast-twitch motor units, right from the beginning of the trial. P_max recovery after the AI trial was faster than IMVC (and T_max) recovery because of the fast recovery of V_max (Fig. 2C), as cross-bridge kinetics recovers quickly as it may be related to ATP concentration (see below), which recovers quickly post-exercise.

Alternatively, the greater decrease in IMVC after the severe-intensity but longer-duration exercise (SI) may be associated with an increase in concentration of metabolites [e.g. inorganic phosphate (Pi), protons (H⁺), and plasma potassium (K⁺)] for a prolonged period (Black et al., 2017). Higher concentrations of Pi can reduce free Ca²⁺ availability by binding Ca²⁺ inside the sarcoplasmic reticulum (Fryer et al., 1995). Nevertheless, it is worth considering that decreased myoplasmic [Ca²⁺] can result from a variety of possible mechanisms, e.g. decreased Ca²⁺ release due to metabolic inhibition of ryanodine receptors (MacIntosh et al., 2012), and decreased availability of Ca²⁺ due to redistribution of Ca²⁺ to mitochondria and across the sarcolemma as well as sequestration in the sarcoplasmic reticulum (Allen et al., 1995). Although changes within the muscle (i.e. peripheral fatigue) are possibly the main factors affecting IMVC after fatigue in this study, it is important to note that fatigue can also be related to a failure in the central nervous system to maximally drive motoneurons voluntarily, i.e. central fatigue (Gandevia, 2001). In other words, a decrease in [Ca²⁺] can also be the result of a decreased frequency of action potentials (Glass et al., 2018), particularly high threshold motor units (Gandevia, 2001).

We conclude from these observations that (i) the decrease in IMVC probably relates to decreased myoplasmic [Ca²⁺] and/or Ca²⁺ sensitivity, (ii) compromised P_max results from a combination of decreases in both maximal velocity and force and (iii) metabolic disturbance affecting V_max recovers quickly, allowing partial recovery of P_max in the absence of T_max recovery. For these reasons, the assessment of P_max and IMVC, both of which are used in the common definition of fatigue, are not similarly affected after a given exercise and, therefore, they are not interchangeable.

The difference between trials for T_max was small and recovery was not apparent (Fig. 4D). One could expect the mechanisms responsible for depression and recovery of IMVC and T_max to be the same, as both represent the maximal force produced by a muscle. Force depression in response to fatigue is closely related to lower myoplasmic concentrations of Ca²⁺ and/or decreased Ca²⁺ sensitivity due to fatigue (MacIntosh et al., 2012), which should similarly affect these two indices. We found low-frequency fatigue (assessed by the low-frequency doublet to high-frequency doublet ratio Db10:100) after the AI and SI exercise (Krüger et al., 2019). In addition, lower levels of ATP and high concentrations of Pi during fatigue can affect force production by reducing the energy charge, decreasing the specific force per cross-bridge and the rate of cross-bridge dissociation (Edwards et al., 1975; Jones et al., 2009), which should also affect both IMVC and T_max.

Nevertheless, our results show that T_max and IMVC are affected differently by the physiological mechanisms of fatigue, as although T_max and IMVC were similarly affected after the AI trial, IMVC was decreased much more than T_max after the SI and MI trials (Fig. 2B). There are a few methodological aspects that need to be considered. First, T_max was extrapolated from averaged values from one pedal downstroke, which includes non-maximal force levels applied to the pedal at the beginning and at the end of the downstroke. In contrast, IMVC represents the peak voluntary force. Thus, while T_max is measured from an effort that lasts only a fraction of a second, IMVC represents the maximal force sustained for ∼5 s. Second, different muscles contribute distinctly to the development of force during cycling sprints and IMVC. Indeed, T_max was extrapolated from dynamic measurements during cycling sprints. Studies have shown that there is a substantial contribution in the hip transfer power across the pelvis to the leg during maximal cycling (Driss and Vandewalle, 2013; Elmer et al., 2011). In other words, the co-activation of knee extensors and hip extensor–knee flexor muscles increases the energy transfer between hip and knee joints to maximize cycling efficiency and total power production during maximal cycling (Driss and Vandewalle, 2013; Elmer et al., 2011). This is different from what happens during an IMVC, in which the quadriceps muscles are predominantly recruited during the contractions while the knee flexors play a very small role in force production (Bampouras et al., 2017). Accordingly, the contribution of the knee flexors during the cycling sprint may reduce the total force and/or power applied by the knee extensors to the pedal by the quadriceps muscles (Bobbert et al., 2016). Thus, it is possible that the contrast between the different contraction modes and muscle groups involved in the assessments may explain the discrepancy between the fatigue-induced reduction in T_max and IMVC. Still, it is interesting to note that this is not the case for the AI exercise.

V_max decreased much more after the AI test than after the other trials (Fig. 4C). In fact, our study shows that V_max does not always decrease after dynamic fatiguing tasks, as it was decreased by only 13% after the SI exercise and not reduced after the MI trial. The characteristic feature of changes in V_max after the AI and SI exercise was that recovery appears to occur within the very first minutes after the end of the exercise. This implies that V_max is decreased by a factor(s) associated with high metabolic rate that recovers quickly. Intriguingly, changes in MRFD after the three trials were very similar and there was no apparent recovery during the 8 min following these trials (Fig. 4E). For this reason, changes in MRFD cannot be associated with any factor that recovers during this time, such as metabolic disturbance. Thus, we believe the physiological mechanism of fatigue detected by V_max, but not by MRFD, is related to a quickly recovering property of muscle metabolism such as increased [ADP]. Increased [ADP] and/or decreased [ATP] would reduce the energy charge, resulting in less force per cross-bridge as discussed above, and possibly a decreased rate of cross-bridge dissociation. It is known that increased [ADP] slows the maximal velocity of isolated single muscle fibers (Westerblad et al., 1998; Westerblad and Lannergren, 1995), which is dependent on the rate of dissociation of cross-bridges. These metabolic disturbances should resolve within a few minutes of stopping the exercise, allowing recovery of processes that are affected by these mechanisms. It is not clear whether or not a decrease in maximal velocity results from decreased Ca²⁺ release as dantrolene, a drug that inhibits Ca²⁺ release, has been reported to impair V_max estimated from the fit of the Hill equation to measured force and velocity of shortening in whole muscle (Kristensen et al., 2018), while maximal rate of unloaded shortening is not affected by dantrolene in single fibers (Allen et al., 1995). Thus, it is probable that the high energy demand during the AI exercise contributed to increased ADP levels after exercise (Post), which compromised the ability of the muscle to produce V_max and, as a consequence, P_max (Bogdanis et al., 1995). Alternatively, the decrease in V_max after the AI exercise may also be associated with the greater fatigue in fast-twitch motor units, which would be expected to be most active during the 30 s sprint all-out trial (Gregor et al., 1979; Mannion et al., 1995), as previously discussed.

Previous studies have proposed that the decrease in both MRFD and V_max is correlated with a reduction in the maximal motor unit firing rate in fatigue (Harwood and Rice, 2012; Morel et al., 2015; Thorlund et al., 2008). One of the main reasons for this positive association could be that the capacity to produce both a rapid rise in force (MRFD) and force at a high velocity (V_max) relies on a high discharge rate at the onset of the contraction (i.e. motor unit firing rate) (Maffiuletti et al., 2016). However, the pattern of fatigue and recovery in MRFD and V_max was quite different across the exercise trials, so the same physiological changes associated with a given exercise must affect these measures differently. It is important to recognize that lack of recovery of MRFD, but not V_max, is consistent with persistent low-frequency fatigue during recovery (data not shown). Low-frequency fatigue is known to be a consequence of decreased peak [Ca²⁺].

It should be noted that the SI and MI exercises were not performed to exhaustion. Nevertheless, there was clear evidence of NM fatigue following all the exercise trials and the RPE was significantly increased from baseline (Pre) at the end of the exercises (AI: from 8±2 to 18±3; SI: from 9±2 to 19±1; and MI: from 7±1 to 15±4). Moreover, the workload was exactly the same for dynamic and isometric measurements so the comparison of variables is still valid.

This study showed that isometric versus dynamic measurements identify distinct fatigue responses following different durations of dynamic exercise with large muscle mass. Most of the differences in the responses between isometric force and dynamic properties are explained by different physiological mechanisms contributing to the given measure of fatigue. Our findings indicate that isometric and dynamic assessment of fatigue are not interchangeable following dynamic exercise with a large muscle mass and that the results from isometric function cannot be extrapolated to dynamic function and vice versa. The assessment of dynamic properties and isometric force together can bring greater insight into the overall magnitude and etiology of fatigue. Therefore, future researches should consider the task specificities (e.g. exercise contraction mode, intensity and duration) when using different fatigue measures (dynamic versus isometric) to assess NM fatigue. Whether the differences between isometric and dynamic indices of fatigue for a given exercise depend on age, training status, disease or sex remains to be investigated to better understand the effects of these parameters on fatigability.