Journal Logo


Caffeine Increases Work Done above Critical Power, but Not Anaerobic Work


Author Information
Medicine & Science in Sports & Exercise: January 2018 - Volume 50 - Issue 1 - p 131-140
doi: 10.1249/MSS.0000000000001408
  • Free


The relationship between power output (PO) and the maximal tolerable duration of high-intensity exercise is characterized by a hyperbolic function, which is defined by two constants: a power-asymptote and a curvature constant (1–3). The power-asymptote, the so-called critical power (CP), was originally defined as the external PO that could be sustained for a very long time and task failure should not occur (2). This definition has been refined to provide a physiological interpretation of the CP and is currently defined as the greatest oxidative metabolic rate that can be sustained with a physiological steady state (1,3–5). It has been assumed therefore that CP represents the boundary between the “heavy” and “severe” exercise intensity domains (1,3–5), where any exercise intensity within the severe domain (>CP) is associated with limited exercise tolerance and non–steady-state responses of blood lactate concentration, pulmonary oxygen uptake (V˙O2), ventilation (E), and heart rate (HR) (3). This non–steady-state response will ultimately lead to attainment of the maximal oxidative metabolic rate (i.e., V˙O2max) if exercise is continued for the maximal tolerable duration (1,3–5).

Although CP is a well-defined parameter, a precise definition of the curvature constant (W′) is more problematic because its physiological determinants have not been fully elucidated (1,3). Originally, W′ was defined as a finite amount of mechanical work able to be performed above the CP, irrespective of the chosen work rate (1,3). The W′ was conceived as representing a “finite anaerobic energy store” or the so-called “anaerobic capacity,” which comprises the energy derived from muscle phosphocreatine (PCr) breakdown, anaerobic glycolysis, and a small amount of aerobic energy linked to O2 stores (1,3,6,7). However, some authors have recently argued that the physiological interpretation of W′ is more complex than simply a marker of anaerobic energy stores (3,4,6,7). For example, Vanhatalo et al. (6) found that hyperoxia (70% O2) reduced W′, a finding that contradicts the definition of W′ as a parameter strictly related to anaerobic metabolism. They also demonstrated that exhaustion during severe-intensity exercise, regardless of the chosen work rate, coincides with the attainment of critical values of fatigue-related intramuscular metabolites such as ADP, Pi, and H+ (6). These same authors reported in another study a positive relationship between the dynamic changes in W′ and the loss of skeletal muscle efficiency (represented by the V˙O2 slow component) during severe-intensity exercise, with exhaustion and depletion of W′ coinciding with attainment of V˙O2max (7). Further evidence suggests that W′ is strongly related to the magnitude of global and peripheral fatigue (8). Together, these findings point to a possible mechanistic link between W′ and the development of fatigue, and argue against W′ exclusively representing the utilization of a fixed anaerobic energy store (8,9). On the other hand, irrespective of the chosen exercise intensity, the amount of anaerobic work completed during supramaximal exercise has been reported to be a constant value when other methods are used to estimate the anaerobic capacity, such as the accumulated oxygen deficit (AOD) or the summation of the fast component of excess postexercise oxygen consumption and the O2 equivalents derived from blood lactate accumulation (10,11).

Given this unresolved controversy, a manipulation able to increase the total work completed above CP could provide novel insights into the relationship between W′ and the anaerobic capacity. Caffeine seems a promising candidate to test this relationship because it increases exercise tolerance during high-intensity exercise, such as that performed at an intensity above the CP (12–15), which should coincide with a greater amount of work that can be performed above the CP (16). However, the mechanism by which caffeine increases exercise tolerance during high-intensity exercise is poorly understood, and there is evidence that caffeine might not have a direct effect on flux through the glycolytic pathway in vivo (15). Rather, caffeine may improve intramuscular milieu by increasing Na+–K+–ATPase activity and increasing calcium release from the sarcoplasmic reticulum during muscle contraction (17–20). Therefore, a potential effect of caffeine on work completed above CP might be a consequence of the delayed fatigue development.

In support of the previously mentioned hypothesis, Simmonds et al. (15) reported that caffeine ingestion increased both time to exhaustion and total O2 uptake (indicative of an increased aerobic contribution) during exercise performed in the severe-intensity domain. Caffeine did not alter V˙O2 kinetics or end-exercise V˙O2, suggesting that the “extra” exercise time under the influence of caffeine might have been supported by a longer time maintained at V˙O2max rather than an increased anaerobic energy contribution. However, time maintained at V˙O2max was not quantified in that study, so it is unknown whether caffeine in fact increases the tolerance to maintain the maximal oxidative metabolic rate. In addition, total work performed above CP was not determined; therefore, a possible link between W′ and anaerobic work was not explored.

Given the discussion previously, we investigated for the first time if caffeine ingestion could increase total work done above CP during a constant-load exercise performed at a severe-intensity exercise, and if this would be accompanied by greater estimated anaerobic energy expenditure. On the basis of previous observations, a further novel aspect of this study was to directly determine if caffeine affects the time to reach, and the time maintained at, V˙O2max. It was hypothesized that caffeine ingestion would enable the completion of more work above the CP but would not be accompanied by greater anaerobic energy expenditure. It was also hypothesized that caffeine would increase the time maintained at V˙O2max, with the “extra” energy to support a longer exercise time derived from aerobic metabolism.


Nine healthy men (26.6 ± 5.3 yr, 74.0 ± 5.3 kg, 1.73 ± 0.10 m, 15% ± 5% of body fat) participated in this study. Participants were familiar with cycling time-to-exhaustion trials because they had been involved in previous studies from our laboratory. The required sample size was estimated using the reported effect size of caffeine on time to exhaustion during severe-intensity exercise (21). With an alpha of 0.05 and a desired power of 0.95, the total effective sample size necessary to achieve statistical significance was estimated to be six participants. However, assuming that 25% might drop out during the data collection, the sample size was increased to nine participants. Participants were informed about the procedures, risks, and benefits associated with the protocol before signing a consent form agreeing to participate in this study. This study was approved by the Research Ethics Committee of the Federal University of Pernambuco.

Experimental Design

Each participant visited the laboratory on six different occasions, with all tests performed at least 72 h apart. During the first visit, anthropometric measurements (body mass and height, and chest, abdomen and thigh skinfolds) were obtained and body density was converted to a body fat percentage (22,23). Participants then had their resting blood glucose concentration and blood pressure measured and answered a physical activity readiness questionnaire. Immediately after, participants performed an incremental exercise test on an electromagnetic cycle ergometer (Ergo-Fit 167, Pirmasens, Germany) with electrocardiogram monitoring (Micromed; WinCardio, Brasilia, Brazil). This initial screen was performed to exclude individuals with cardiovascular risk. After examination, a cardiologist attested that participants were able to join the subsequent experimental trials.

During the three following visits, participants cycled until exhaustion at different PO values to allow the determination of CP and W′. These trials were performed in a randomized and counterbalanced order using a balanced Latin square design. During the next two visits, participants performed a constant-load exercise in the severe-intensity exercise domain (i.e., >CP) after ingesting caffeine or a placebo in a double-blind manner and in a counterbalanced order. To effectively blind the participants, they were informed that they might ingest only placebo or only caffeine. Participants were instructed to refrain from exhaustive exercise, alcohol, and food or supplements containing caffeine 24 h before each experimental test. They also registered all food and beverages ingested during the 24 h before the first experimental trial and were asked to replicate this in the subsequent visits.

Incremental Test

After a 3-min warm-up at 50 W, PO was increased by 25 W·min−1 until exhaustion. Participants were instructed to maintain a cycling cadence between 70 and 80 rpm. Exhaustion was assumed when the pedal revolutions dropped below 70 rpm for more than 5 s or were reduced to less than 70 rpm on more than three consecutive occasions. Participants were also verbally encouraged to increase their pedal revolutions and to continue cycling maximally if the revolutions dropped below 70 rpm. Carbon dioxide production (V˙CO2), V˙O2, and E were measured breath by breath throughout the test using an automatic analyzer (Metalyzer 3B®; Cortex, Saxony, Germany). Before each test, the analyzer was calibrated using ambient air and a known gas concentration (12% O2 and 5% CO2). The volume was calibrated using a 3-L syringe (Quinton Instruments, Bothel, WA).

The V˙O2max and maximal HR (HRmax) were determined as the average of the O2 and HR values during the last 30 s of the test, respectively. The POmax was determined as the highest PO achieved during the last completed stage. When the participants could not maintain the final PO during an entire stage, the POmax was calculated using the fractional time completed in the last stage multiplied by the increment rate. Two experienced investigators identified the gas exchange threshold (GET) from a cluster of measurements: 1) the first disproportionate increase in V˙CO2 from visual inspection of individual plots of V˙CO2 versus V˙O2, 2) an increase in V˙E/V˙O2 without an increase in V˙E/V˙CO2, and 3) an increase in end-tidal O2 pressure with no fall in end-tidal CO2 pressure. Because of the cumulative delay in V˙O2 kinetics during the incremental test, the PO at GET was corrected by two-thirds of the increment rate (i.e., 25 W × 0.66 = 16.6 W) (24).

Derivation of the Power–Duration Relationship and Estimation of CP and W

Participants performed a 5-min warm-up at 90% of the GET, followed by a 5-min rest. They then cycled until exhaustion (same criteria as those used in the incremental test) at the PO corresponding to 70% of the difference between the GET and V˙O2max (Δ70), or 100% or 120% of V˙O2max. These exercise intensities were selected to yield times to exhaustion between 2 and 10 min, which have been suggested as appropriate to determine the hyperbolic power–time limit relationship (1).

Individual CP and W′ were estimated from the PO and the corresponding time to exhaustion, using least squares fitting of the following regression models (equations 1–3) (1).

where t is the time to exhaustion, W is the total work performed, W′ is the constant curvature of hyperbole power–time, P is the power, and CP is the critical power. The SEE associated with the CP and W′ were expressed as coefficients of variation (CV%, i.e., relative to the parameter estimate). The total error associated with a given model was calculated as the sum of the CV% associated with the CP and W′. The sum of the CV% was optimized for each participant by selecting the model with the smallest total error (equation 1, 2, or 3) to produce the “best-individual-fit” parameter estimates (25). Therefore, the best individual fit was used for further analysis.

Experimental Tests

Participants arrived at the laboratory and ingested caffeine (5 mg·kg−1 body mass) or a placebo (cellulose). This caffeine dose is believed to increase plasma caffeine levels to ~40 μmol·L−1, peaking at 60 to 90 min after ingestion (26). Fifty minutes later, a resting capillary blood sample was collected for determination of plasma lactate concentration ([La]). A 5-min warm-up at 90% of the GET was subsequently performed, followed by a 5-min rest. Participants then cycled until exhaustion in the severe-intensity domain (PO corresponding to 80% of the difference between the GET and V˙O2max, Δ80), followed by a 10-min passive recovery. Capillary blood samples were taken at 1, 3, and 5 min after exercise to measure peak [La]. Gas exchanges and E were measured breath by breath throughout the test using the same gas analyzer described for the incremental test, whereas HR was recorded using an HR monitor (Polar T 31/34®, Kempele, Finland). The end-exercise V˙O2, V˙CO2, REE, E, and HR were determined as the average of the data containing in the last 30 s of the test. The RPE was asked immediately after exhaustion. Participants were also asked before and after each trial to identify which supplement they had ingested.


Plasma lactate concentration

Capillary blood samples were collected from the ear lobe (40 μL). Blood was transferred to microtubes containing 8 μL of EDTA and then centrifuged (4000 rpm) at 4°C for 15 min for plasma separation. The plasma [La] was determined in a spectrophotometer (Q798U2V5; Quimis, Sao Paulo, Brazil) using commercial kits (Labtest Diagnostica, Minas Gerais, Brazil).

Calculation of the total work done above CP and energy system contributions

The amount of work performed above CP was calculated as the power–time integral above CP and expressed both as work (kJ) and a percentage of W′ (%). The aerobic energy expenditure was calculated from the exercise V˙O2 area under the curve using the trapezoidal method, deducting the corresponding V˙O2 rest (27,28). The anaerobic lactic energy expenditure was calculated assuming that an elevation of 1 mM in plasma lactate above basal levels is equivalent to 3 mL O2·kg−1 body mass (28,29). Anaerobic alactic energy expenditure was calculated using the fast component of the V˙O2 recovery (30,31). For this, the 10-min V˙O2 recovery curve was fitted with a biexponential model (equation 4) using the Origin software (version 8.5; Microcal, Northampton, MA):

where V˙O2(t) is the oxygen uptake at a given time, V˙O2 recovery is the mean of V˙O2 during the last 90 s of the recovery, A is the amplitude, δ is the time delay, τ is the time constant, and 1 and 2 denote the fast and slow components, respectively.

The anaerobic alactic energy expenditure was obtained by equation 5:

The contribution of each energy system was measured as equivalent of liters of O2 consumed and then converted to energy equivalent (kJ), assuming that 1 L of O2 is equal to 20.9 kJ. The total anaerobic energy expenditure during exercise was considered as the sum of the alactic and lactic energy expenditure. The total energy expenditure during exercise was considered as the sum of the energy expenditure attributed for each energy system.

Calculation of the time to reach and the time maintained at V˙O2max

First, the breath-by-breath V˙O2 data were interpolated to obtain second-by-second data. Because we were more interested in characterizing the overall V˙O2 response during the exercise and in calculating a corresponding time to achieve V˙O2max, the V˙O2–time plot was fitted using a single-exponential model without time delay (equation 6), with the fitting window commencing at t = 0 s (equivalent to the mean response time (MRT)) (32–34):

where V˙O2(t) is oxygen uptake at a given time, V˙O2 baseline is the mean of V˙O2 during the last 90 s of the rest period, A is the asymptotic amplitude, and MRT is the mean response time (i.e., the time required to attain 63% of the final V˙O2).

Because we were interested in estimating the time taken to reach V˙O2max and the time maintained at V˙O2max (TRV˙O2max and TMV˙O2max, respectively), and the end V˙O2 was not different from V˙O2max achieved during the incremental test (see Results), the parameter A was fixed as the V˙O2max obtained from the incremental test (35). It was assumed that the V˙O2 is projecting to V˙O2max when the value of 1 − e−(t/MRT) from equation 6 is at least 0.95, that is, when t = 3MRT. Thus, TRV˙O2max was defined as 3MRT (36). TMV˙O2max was determined by deducting the TRV˙O2max from the time to exhaustion.

Statistical Analysis

The statistical analysis was performed using the Statistical Package for Social Sciences (SPSS), version 20.0 (SPSS Inc., Chicago, IL). Dependent variables were tested for normality using the Kolmogorov–Smirnov test. The variables were normally distributed and were described as mean and SD. A one-way ANOVA with repeated measures, followed by a Bonferroni post hoc test, was performed to check if end V˙O2 during the constant-load exercise in the placebo and caffeine trials was significantly different from the V˙O2max achieved in the incremental test, and to compare the parameters estimated between the three models. Paired t-tests were used to compare time to exhaustion, estimated mean V˙O2 at 3MRT, TRV˙O2max and TMV˙O2max, energy system contributions, total work done above CP, and end-exercise physiological responses between the caffeine and placebo trials. A potential learning effect was checked by comparing the time to exhaustion between the first and the second experimental trials using a paired t-test. Pearson correlation coefficient was carried out to determine the relationship between W′ and the estimated anaerobic energy expenditure. The significance level was set at 5% (P < 0.05).


The GET was identified at 102.5 ± 24.6 W and POmax at 261.1 ± 26.1 W. The V˙O2max was 2.99 ± 0.39 L·min−1 (40.6 ± 5.8 mL·kg−1·min−1) and HRmax was 180 ± 8 bpm. The mean PO corresponding to Δ70, 100%, and 120% V˙O2max were 213.9 ± 23.6, 261.1 ± 26.1, and 310.5 ± 33.8 W, respectively. There were no differences in CP or W′ estimates between the three models (equations 1–3, P = 0.99; Table 1). The CP and W′ estimate from the best individual fit were 170 ± 22 W and 18.4 ± 2.5 kJ, respectively (Table 1). The mean PO at Δ80 was 229.6 ± 23.6 W and corresponded to 136% ± 7% of CP and 88% ± 2% of POmax. The mean time to exhaustion was not significantly different between the first and second experimental bouts (337 ± 82 and 366 ± 78 s, P = 0.34). Seven of nine participants (83.3%) reported before the trial they were ingesting placebo, when they actually ingested caffeine. The same seven participants continued to believe that they had received the placebo when asked after the trial.

The parameter estimates derived from equations 1–3 and the best individual fit.

Time to exhaustion, total work done above CP, and energy system contributions

Time to exhaustion was higher with caffeine compared with placebo (P = 0.01; Fig. 1A). This was accompanied by a greater total work done above CP in caffeine than in placebo (P = 0.01; Fig. 1B). Total work done above CP corresponded to 130% ± 30% of W′ with caffeine and 95% ± 14% of W′ with placebo (P = 0.01). However, there was no difference between caffeine and placebo for anaerobic lactic, anaerobic alactic, and total anaerobic (lactic + alactic) energy expenditure (P = 0.50, 0.78, and 0.54, respectively; Fig. 2A). There was no significant correlation between W′ and estimated anaerobic energy expenditure for either placebo (r = 0.14, P = 0.71; Fig. 2B) or caffeine (r = 0.09, P = 0.81; Fig. 2C). Aerobic and total energy expenditures were higher in caffeine compared with placebo (P = 0.01; Fig. 2A).

Time to exhaustion (A) and total work done above CP (B) when performing severe-intensity exercise after caffeine or placebo intake. Dotted lines represent curvature constant (W′); straight lines, individual values. *Significantly greater than placebo (P < 0.05).
A, Alactic (AA), anaerobic lactic (AL), total anaerobic (TA), aerobic (A), and total energy expenditure (TEE) in the caffeine and placebo conditions. Relationship between total anaerobic energy expenditure and the curvature constant (W′) of the power–duration curve in the placebo (B) and caffeine (C) conditions.

End-exercise physiological responses and time to reach and time maintained at V˙O2max

The HR and E end values were higher in caffeine compared with placebo (P = 0.01 and 0.04, respectively; Table 2). The HR end corresponded to 99% ± 4% of HRmax for caffeine and 96% ± 4% of HRmax for placebo. There were no effects of caffeine on V˙CO2 end, peak plasma [La], or RPE (P = 0.66, 0.62, and 0.44, respectively; Table 2). The mean V˙O2 profiles for the group and the two participants who had the greatest and lowest improvement in time to exhaustion with caffeine are shown in Figure 3. The V˙O2 end was similar to the V˙O2max reached in the incremental test for both caffeine and placebo (P = 0.99 and 0.37, respectively), and there was no difference between caffeine and placebo (P = 0.13; Table 2). The MRT was similar between caffeine and placebo (66.4 ± 13.2 and 60.3 ± 8.1 s, respectively; P = 0.31). Caffeine did not change TRV˙O2max compared with placebo (P = 0.31; Fig. 4A). However, caffeine increased TMV˙O2max compared with placebo (P = 0.04; Fig. 4B).

The effects of a placebo or caffeine on time to exhaustion and end-exercise values for metabolic, cardiopulmonary, and perceptual responses during exercise in the severe-intensity domain.
A, Oxygen uptake (20-s bins) as a function of time during a severe-intensity exercise bout after caffeine or placebo intake. Note the longer time maintained near the V˙O2max after caffeine ingestion. Oxygen uptake profile after caffeine or placebo intake by the two participants who had the greatest (B) and lowest (C) improvement in time to exhaustion with caffeine.
Time to reach (A) and time maintained (B) at V˙O2max during a severe-intensity exercise in caffeine and placebo conditions. Straight lines represent individual values. *Significantly higher than placebo (P < 0.05).


The principal novel finding of this study was that the amount of work that can be performed above CP during whole-body exercise is not constrained to W′. With caffeine, the longer time to task failure and greater total work performed above CP were accompanied by a higher aerobic rather than anaerobic contribution. Furthermore, the W′ was not related to the anaerobic energy expenditure in either the caffeine or placebo condition. These data are consistent with our hypotheses and indicate that the total work done above CP is not related to an “anaerobic reserve” or an “anaerobic capacity.” This is in accordance with a recent postulation that W′ might be better defined as the “buffer” available to resist exercise intolerance during supra-CP exercise (3), and suggests that the ability to increase the work done above CP might be linked to an enhanced tolerance for metabolic/physiological disturbances and/or a reduced rate of projection of fatigue-related metabolites toward maximal values (6).

In the present study, caffeine ingestion enabled greater time to exhaustion during severe-intensity exercise. This is in accordance with previous studies reporting increased exercise tolerance with caffeine during high-intensity trials to exhaustion (13–15,21). However, the effect of caffeine on time to exhaustion in the present study was larger (~33%; increasing from ~300 to ~400 s) compared with the more modest effect (~8%–15%) typically reported in the literature for time-to-exhaustion trials (13–15,21). This large improvement in performance cannot be associated with any conscious perception of which supplement was being ingested, because seven of nine participants (83%) were unable to correctly identify which supplement they ingested. It is interesting to note that the largest improvement in time to exhaustion reported previously (~15%) occurred during a supramaximal exercise lasting ~180 s (15,21). On the other hand, the magnitude of improvement was considerably lower (~8%) when the duration of supramaximal exercise was shorter (~100 s) (13). Together, these results suggest that the ergogenic effect of caffeine on exercise tolerance during high-intensity exercises may be more pronounced when exercise lasts more than 3 min. This matches with the average time necessary to reach V˙O2max (Fig. 3), which reinforces our hypothesis that the positive effect of caffeine on time to exhaustion during severe-intensity exercise might be linked to a maintenance of maximal oxidative metabolic rate for longer (i.e., longer time maintained V˙O2max).

It is controversial which energy system(s) might provide the energy to support this longer duration of exercise until exhaustion (14,15). In the present study, we found no influence of caffeine on the energy provided by the alactic and lactic energy systems (and consequently on the total anaerobic energy expenditure). It has been reported that caffeine intake increases the AOD during supramaximal exercise, indicating that caffeine may enhance the total energy provided by anaerobic sources (13,15,21). However, the mechanism(s) by which caffeine might increase the total energy produced via anaerobic energy systems is(are) speculative, and there is no strong evidence to support an effect of caffeine on increasing glycolytic anaerobic energy capacity. In fact, Simmonds et al. (15) demonstrated that although caffeine increased the AOD when measured over the entire exercise bout, indicating increased total amount of anaerobic energy production, no changes were observed when AOD was calculated over the same duration of exercise for placebo and caffeine. These authors suggested that the caffeine-induced increase in AOD may have been a consequence of, rather than a cause of, prolonged exercise (15) and does not necessarily indicates a direct effect of caffeine on anaerobic metabolism. This is corroborated by Poli et al. (14), who found a higher time to exhaustion at 115% of V˙O2max with caffeine, but without significant changes in alactic and lactic contributions and ultimately total anaerobic energy expenditure. Together, these results challenge the hypothesis that caffeine increases the anaerobic capacity.

Although the anaerobic energy expenditure during severe-intensity exercise in the present study was not changed with caffeine, the total work done above CP was 35% larger when compared with placebo. The amount of work performed above CP during the placebo trial was ~96% of the total W′, but it was ~127% of the total W′ during the caffeine trial. If the amount of energy that can be provided by anaerobic metabolism could be represented by W′, it should be expected that any caffeine-induced increase in the W′ would be accompanied by a proportional increase in the estimated anaerobic energy expenditure. However, the correlation between W′ and anaerobic energy expenditure was not significant for either the caffeine or the placebo condition. This is consistent with a recent interpretation that the W′ seems not to represent a fixed anaerobic capacity (1,3,6). For instance, Vanhatalo et al. (6) found that hyperoxia reduced W′, when no changes would be expected if W′ represented a fixed anaerobic energy reserve. Similarly, sodium bicarbonate ingestion, which may improve high-intensity exercise performance by facilitating enhanced energy supply through anaerobic glycolysis (37), should increase the total amount of work done above CP. However, Vanhatalo et al. (38) found that despite notably enhanced blood-buffering capacity, bicarbonate ingestion had no effect on the W′. Our results, showing an increased total work done above CP without any increase in the anaerobic energy expenditure after caffeine ingestion, add to evidence suggesting that the anaerobic energy that can be expenditure above CP might be a fixed amount, but this might not be well represented by W′, especially when an ergogenic agent is used to postpone exhaustion.

In the present study, the MRT and time to reach the V˙O2max were not altered with caffeine, which suggests that caffeine does not speed the V˙O2 response and consequently the rate of aerobic energy provision. However, the longer time to exhaustion in the caffeine condition was linked with a longer time at the V˙O2max. In addition, the ability to exercise longer and to do more work above CP with caffeine was accompanied by a larger cardiopulmonary disturbance (i.e., higher end V˙E and HR). This suggests that caffeine enabled participants to maintain a maximal oxidative metabolic rate for longer, which in turn resulted in a more pronounced cardiopulmonary disturbance. This might be linked to an effect of caffeine on the central nervous system and/or the skeletal muscle. In the central nervous system, caffeine might reduce perceived effort via a direct blockade of the adenosine A2a receptors in the brain (39,40), permitting more external work to be performed for a given conscious perception (15,40,41). Caffeine might also have a peripheral effect on increasing muscle function (42), perhaps by increasing calcium release from the sarcoplasmic reticulum during muscle contractions (42) and by delaying potassium accumulation (43), which could reduce muscle sensory signals to the brain and consequently increase the effort tolerance. Our findings of larger E and HR values at exhaustion suggest that caffeine enables uncomfortable feelings related to the exercise (e.g., dyspnea and leg muscle pain) to be sustained for longer during severe-intensity exercise (44). We also note that participants reached the same RPE at the moment of exhaustion (18 ± 1 units), although they exercised longer, suggesting that caffeine might attenuate the conscious perception of discomfort caused by the large cardiopulmonary disturbance during a severe-intensity exercise.

The present study provides important insights into the mechanisms controlling task failure during exercise performed at a severe intensity. Our results indicate that the amount of work that can be performed above CP during whole-body exercise may be a constant value for a given condition but can be increased after ingestion of a supplement able to postpone exhaustion (e.g., caffeine), potentially through increased tolerance to accrual of peripheral fatigue (9). On the other hand, total anaerobic energy expenditure was constant, which suggests that total anaerobic work rather than the amount of work that can be performed above CP is a constant value (10). Therefore, W′ should not be used as an indicator of anaerobic capacity. We also found that caffeine increases time maintained at V˙O2max. If it is assumed that V˙O2 is a proxy of muscle PCr utilization (45), it could be assumed that the severe-exercise intensity performed in the present study was tolerated long after muscle PCr had reached its nadir. It has previously been reported that the exercise could be maintained for ~150 s after the nadir of muscle PCr concentration had been reached during the lowest work rate within severe-intensity exercise (see Fig. 2 in Vanhantalo et al. [6]). These results suggest that in some conditions, exercise in the severe-intensity domain can be maintained even after intramuscular metabolites have attained their minimal/maximal nadir values. Together, these findings provide interesting insights into the nature of W′ and the mechanism(s) determining exercise tolerance during severe-intensity exercise. Further studies, however, should investigate the effect of caffeine on other exercise intensities within the severe-intensity domain to determine the effect of caffeine on the shape of the power–duration relationship. In addition, intramuscular metabolites should also be measured to provide evidence that caffeine may increase time to task failure during severe-intensity exercise by increasing the time maintained at the nadir of intramuscular metabolites such as PCr, ADP, and Pi.

It is important to recognize some limitations of the present study. First, we used only a single test to characterize the V˙O2 response during a severe-intensity exercise task, which could affect our calculations of TRV˙O2max and TMV˙O2max. However, it should be mentioned that signal-to-noise ratio is naturally increased for severe-intensity exercise and a reduced number of trials is necessary (46). In addition, that the V˙O2 curves of caffeine and placebo were superimposed suggests that more trials would not have improved the estimation. It should be also highlighted that although the V˙O2 response during a severe-intensity exercise might be better characterized by a double-exponential model, a monoexponential model fitting from the onset of exercise through the entire data has been recommended when the interest is to provide information on the overall response kinetics (i.e., MRT) (47), which was essential to make estimates of TRV˙O2max and TMV˙O2max. Because V˙O2max obtained in the incremental test was not different from end V˙O2 in both the caffeine and placebo trials, this monoexponential model also enabled us to insert the V˙O2max as an amplitude-fixed term in equation 4, which increased the confidence in our estimate of the MRT parameter and both TRV˙O2max and TMV˙O2max. Another observation is that because plasma [La] during exercise is a balance between lactate production and clearance, and the O2 equivalent from plasma [La] may be variable, the estimate of anaerobic lactic metabolism from plasma [La] must be interpreted with caution. Similarly, components other than PCr resynthesis may influence V˙O2 after muscular contraction and therefore could affect the estimate of anaerobic alactic metabolism. In addition, it should be recognized that recreationally active men were recruited for the present study, and any effect of learning or training adaptation with the experimental time to exhaustion bout cannot be disregarded. However, we believe that this has had a minimal effect on our outcomes because time to exhaustion was not different between the first and second experimental sessions, and differences between trials (~29 s) were lower than differences between placebo and caffeine (~100 s). Finally, although the mechanisms by which caffeine increased exercise tolerance could be central and/or peripheral in origin, we were not able to precisely determine the site(s) where caffeine acted in the present study. Future studies should consider monitoring brain activity using functional magnetic resonance imaging and peripheral metabolites using 31P magnetic resonance spectroscopy or via a muscle biopsy, which could directly measure the key components of glycolytic flux and PCr use during exercise, to provide insights into the mechanism(s) governing the increased exercise tolerance and work done above CP with caffeine intake.

Although the present study provides evidence that caffeine increases time to exhaustion and total work performed in excess of CP by increasing the time maintained at V˙O2max, the present results may also have practical implications for further use of caffeine. Because of the lack of free time in modern life, high-intensity training has been gradually incorporated into training routines to promote health benefits and to reduce body mass (48). However, it may be hard for nonathletes to tolerate this kind of training. Our results indicate that caffeine not only increases the tolerance for high-intensity exercise but also allows for greater energy expenditure. Thus, caffeine might have the potential to be used in training routines that have the goal of reducing body mass. In addition, training at or near V˙O2max might place maximal stress on the physiological processes and structures that limit V˙O2max, providing the optimal stimulus for training adaptation (49). However, continuous work at such high intensities cannot be sustained for a long time, which limits total training time expended at this intensity in each training session. Several models of interval training programs to overcome this limitation have been proposed (49). However, if combined with interval training, caffeine might add benefit by increasing the time maintained at V˙O2max. It would be interesting to investigate in further studies the accumulative effect of combining caffeine and interval training on metabolic adaptations and weight loss.


Caffeine increased time to exhaustion and total work done above CP during exercise performed in the severe-intensity domain (i.e., >CP), but this was not related to a greater anaerobic energy expenditure. Rather, the ability to perform more work above CP was linked to the ability to maintain exercise longer at the maximal oxidative metabolic rate (i.e., at V˙O2max) and to tolerate larger cardiopulmonary disturbances, with the energy to support this extra work being provided by aerobic rather than anaerobic metabolism. These results provided additional evidence suggesting that W′ may not represent a strict anaerobic energy reserve or a fixed anaerobic capacity.

The authors thank all participants who took part in this study. They declare no conflict of interest. No financial support was received. The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. The results of this study do not constitute endorsement by the American College of Sports Medicine. The page charge for this paper has been paid by PROEX-CAPES agreement.


1. Jones AM, Vanhatalo A, Burnley M, Morton RH, Poole DC. Critical power: implications for determination of V˙O2max and exercise tolerance. Med Sci Sports Exerc. 2010;42(10):1876–90.
2. Monod H, Scherrer J. The work capacity of a synergic muscular group. Ergonomics. 1965;8:329–38.
3. Poole DC, Burnley M, Vanhatalo A, Rossiter HB, Jones AM. Critical power: an important fatigue threshold in exercise physiology. Med Sci Sports Exerc. 2016;48(11):2320–34.
4. Jones AM, Wilkerson DP, DiMenna F, Fulford J, Poole DC. Muscle metabolic responses to exercise above and below the “critical power” assessed using 31P-MRS. Am J Physiol Regul Integr Comp Physiol. 2008;294(2):R585–93.
5. Vanhatalo A, Black MI, DiMenna FJ, et al. The mechanistic bases of the power–time relationship: muscle metabolic responses and relationships to muscle fibre type. J Physiol. 2016;594(15): 4407–23.
6. Vanhatalo A, Fulford J, DiMenna FJ, Jones AM. Influence of hyperoxia on muscle metabolic responses and the power–duration relationship during severe-intensity exercise in humans: a 31P magnetic resonance spectroscopy study. Exp Physiol. 2010;95(4): 528–40.
7. Vanhatalo A, Poole DC, DiMenna FJ, Bailey SJ, Jones AM. Muscle fiber recruitment and the slow component of O2 uptake: constant work rate vs. all-out sprint exercise. Am J Physiol Regul Integr Comp Physiol. 2011;300(3):R700–7.
8. Broxterman RM, Craig JC, Smith JR, et al. Influence of blood flow occlusion on the development of peripheral and central fatigue during small muscle mass handgrip exercise. J Physiol. 2015;593(17): 4043–54.
9. Broxterman RM, Ade CJ, Craig JC, Wilcox SL, Schlup SJ, Barstow TJ. Influence of blood flow occlusion on muscle oxygenation characteristics and the parameters of the power–duration relationship. J Appl Physiol (1985). 2015;118(7):880–9.
10. Medbo JI, Mohn AC, Tabata I, Bahr R, Vaage O, Sejersted OM. Anaerobic capacity determined by maximal accumulated O2 deficit. J Appl Physiol (1985). 1988;64(1):50–60.
11. Zagatto AM, Bertuzzi R, Miyagi WE, Padulo J, Papoti M. MAOD determined in a single supramaximal test: a study on the reliability and effects of supramaximal intensities. Int J Sports Med. 2016;37(9): 700–7.
12. Azevedo R, Silva-Cavalcante MD, Gualano B, Lima-Silva AE, Bertuzzi R. Effects of caffeine ingestion on endurance performance in mentally fatigued individuals. Eur J Appl Physiol. 2016;116(11–12): 2293–303.
13. Bell DG, Jacobs I, Ellerington K. Effect of caffeine and ephedrine ingestion on anaerobic exercise performance. Med Sci Sports Exerc. 2001;33(8):1399–403.
14. Poli RA, Miyagi WE, Nakamura FY, Zagatto AM. Caffeine improved time to exhaustion, but did not change alternative maximal accumulated oxygen deficit estimated during a single supramaximal running bout. Int J Sport Nutr Exerc Metab. 2016;26:549–57.
15. Simmonds MJ, Minahan CL, Sabapathy S. Caffeine improves supramaximal cycling but not the rate of anaerobic energy release. Eur J Appl Physiol. 2010;109(2):287–95.
16. Cheng CF, Hsu WC, Kuo YH, Shih MT, Lee CL. Caffeine ingestion improves power output decrement during 3-min all-out exercise. Eur J Appl Physiol. 2016;116(9):1693–702.
17. Lindinger MI, Graham TE, Spriet LL. Caffeine attenuates the exercise-induced increase in plasma [K+] in humans. J Appl Physiol (1985). 1993;74(3):1149–55.
18. Lindinger MI, Willmets RG, Hawke TJ. Stimulation of Na+, K(+)-pump activity in skeletal muscle by methylxanthines: evidence and proposed mechanisms. Acta Physiol Scand. 1996;156(3): 347–53.
19. Tallis J, James RS, Cox VM, Duncan MJ. The effect of a physiological concentration of caffeine on the endurance of maximally and submaximally stimulated mouse soleus muscle. J Physiol Sci. 2013;63(2):125–32.
20. Tarnopolsky M, Cupido C. Caffeine potentiates low frequency skeletal muscle force in habitual and nonhabitual caffeine consumers. J Appl Physiol (1985). 2000;89(5):1719–24.
21. Doherty M. The effects of caffeine on the maximal accumulated oxygen deficit and short-term running performance. Int J Sport Nutr. 1998;8(2):95–104.
22. Jackson AS, Pollock ML. Generalized equations for predicting body density of men. Br J Nutr. 1978;40(3):497–504.
23. Siri WE. Body composition from fluid spaces and density: analysis of methods. 1961. Nutrition. 1993;9(5):480–91.
24. Whipp BJ, Davis JA, Torres F, Wasserman K. A test to determine parameters of aerobic function during exercise. J Appl Physiol Respir Environ Exerc Physiol. 1981;50(1):217–21.
25. Black MI, Jones AM, Bailey SJ, Vanhatalo A. Self-pacing increases critical power and improves performance during severe-intensity exercise. Appl Physiol Nutr Metab. 2015;40(7):662–70.
26. Battram DS, Graham TE, Richter EA, Dela F. The effect of caffeine on glucose kinetics in humans—influence of adrenaline. J Physiol. 2005;569(Pt 1):347–55.
27. Bertuzzi RC, Franchini E, Kokubun E, Kiss MA. Energy system contributions in indoor rock climbing. Eur J Appl Physiol. 2007;101(3):293–300.
28. Bertuzzi RC, Franchini E, Ugrinowitsch C, et al. Predicting MAOD using only a supramaximal exhaustive test. Int J Sports Med. 2010;31(7):477–81.
29. di Prampero PE, Ferretti G. The energetics of anaerobic muscle metabolism: a reappraisal of older and recent concepts. Respir Physiol. 1999;118(2–3):103–15.
30. Beneke R, Pollmann C, Bleif I, Leithäuser RM, Hütler M. How anaerobic is the Wingate Anaerobic Test for humans? Eur J Appl Physiol. 2002;87(4–5):388–92.
31. Margaria R, Edwards HT, Dill DB. The possible mechanisms of contracting and paying the oxygen debt and the role of lactic acid in muscular contraction. Am J Physiol. 1933;106:689–715.
32. Bailey SJ, Vanhatalo A, DiMenna FJ, Wilkerson DP, Jones AM. Fast-start strategy improves V˙O2 kinetics and high-intensity exercise performance. Med Sci Sports Exerc. 2011;43(3):457–67.
33. De Andrade Nemezio KM, Bertuzzi R, Correia-Oliveira CR, Gualano B, Bishop DJ, Lima-Silva AE. Effect of creatine loading on oxygen uptake during a 1-km cycling time trial. Med Sci Sports Exerc. 2015;47(12):2660–8.
34. Whipp BJ, Ozyener F. The kinetics of exertional oxygen uptake: assumptions and inferences. Med Sport. 1998;51(2):139–49.
35. Caputo F, Denadai BS. Exercise mode affects the time to achieve V˙O2max without influencing maximal exercise time at the intensity associated with V˙O2max in triathletes. Int J Sports Med. 2006;27(10): 798–803.
36. Whipp BJ, Rossiter HB. The kinetics of oxygen uptake: physiological inferences from the parameters. In: Poole DC, Jones AM, editors. Oxygen Uptake Kinetics in Sport, Exercise and Medicine. London: Routledge; 2005. pp. 62–94.
37. Bishop D, Edge J, Davis C, Goodman C. Induced metabolic alkalosis affects muscle metabolism and repeated-sprint ability. Med Sci Sports Exerc. 2004;36(5):807–13.
38. Vanhatalo A, McNaughton LR, Siegler J, Jones AM. Effect of induced alkalosis on the power–duration relationship of “all-out” exercise. Med Sci Sports Exerc. 2010;42(3):563–70.
39. Fredholm BB, Battig K, Holmen J, Nehlig A, Zvartau EE. Actions of caffeine in the brain with special reference to factors that contribute to its widespread use. Pharmacol Rev. 1999;51(1): 83–133.
40. Graham TE, Rush JW, van Soeren MH. Caffeine and exercise: metabolism and performance. Can J Appl Physiol. 1994;19(2):111–38.
41. Santos Rde A, Kiss MA, Silva-Cavalcante MD, et al. Caffeine alters anaerobic distribution and pacing during a 4000-m cycling time trial. PLoS One. 2013;8(9):e75399.
42. Tallis J, Duncan MJ, James RS. What can isolated skeletal muscle experiments tell us about the effects of caffeine on exercise performance? Brit J Pharmacol. 2015;172(15):3703–13.
43. Mohr M, Nielsen JJ, Bangsbo J. Caffeine intake improves intense intermittent exercise performance and reduces muscle interstitial potassium accumulation. J Appl Physiol (1985). 2011;111(5):1372–9.
44. Gliottoni RC, Motl RW. Effect of caffeine on leg-muscle pain during intense cycling exercise: possible role of anxiety sensitivity. Int J Sport Nutr Exerc Metab. 2008;18(2):103–15.
45. Whipp BJ, Rossiter HB, Ward SA, et al. Simultaneous determination of muscle 31P and O2 uptake kinetics during whole body NMR spectroscopy. J Appl Physiol (1985). 1999;86(2):742–7.
46. Ozyener F, Rossiter HB, Ward SA, Whipp BJ. Influence of exercise intensity on the on- and off-transient kinetics of pulmonary oxygen uptake in humans. J Physiol. 2001;533(Pt 3): 891–902.
47. Whipp BJ, Ward SA, Rossiter HB. Pulmonary O2 uptake during exercise: conflating muscular and cardiovascular responses. Med Sci Sports Exerc. 2005;37(9):1574–85.
48. Whyte LJ, Ferguson C, Wilson J, Scott RA, Gill JM. Effects of single bout of very high-intensity exercise on metabolic health biomarkers in overweight/obese sedentary men. Metabolism. 2013;62(2):212–9.
49. Midgley AW, McNaughton LR, Jones AM. Training to enhance the physiological determinants of long-distance running performance: can valid recommendations be given to runners and coaches based on current scientific knowledge? Sports Med. 2007;37(10):857–80.


© 2018 American College of Sports Medicine