J Appl Physiol 124: 1491–1501, 2018.

First published February 22, 2018; doi:10.1152/japplphysiol.00930.2017.

RESEARCH ARTICLE

Caffeine increases both total work performed above critical power and
peripheral fatigue during a 4-km cycling time trial
Leandro Camati Felippe,1 X Guilherme Assunção Ferreira,1 Sara Kely Learsi,1 Daniel Boari,2
Romulo Bertuzzi,3 and X Adriano Eduardo Lima-Silva1,4
1
Sport Science Research Group, Federal University of Pernambuco, Pernambuco, Brazil; 2Center of Engineering, Modeling,
and Applied Social Science, Federal University of ABC, São Paulo, Brazil; 3Endurance Performance Research Group
(GEDAE-USP), University of São Paulo, São Paulo, Brazil; and 4Human Performance Research Group, Technological
Federal University of Parana, Parana, Brazil
Submitted 10 October 2017; accepted in final form 18 February 2018

Felippe LC, Ferreira GA, Learsi SK, Boari D, Bertuzzi R, INTRODUCTION

Lima-Silva AE. Caffeine increases both total work performed above
critical power and peripheral fatigue during a 4-km cycling time trial.
J Appl Physiol 124: 1491–1501, 2018. First published February 22,
2018; doi:10.1152/japplphysiol.00930.2017.—The link between total
work performed above critical power (CP) and peripheral muscle
fatigue during self-paced exercise is unknown. We investigated the
influence of caffeine on the total work done above CP during a 4-km
cycling time trial (TT) and the subsequent consequence on the
development of central and peripheral fatigue. Eleven cyclists per-
formed three constant-load exercise trials to determine CP and two
4-km TTs ~75 min after oral caffeine (5 mg/kg) or cellulose (placebo)
ingestion. Neuromuscular functions were assessed before and 50 min
after supplementation and 1 min after TT. Oral supplementation alone
had no effect on neuromuscular function (P ⬎ 0.05). Compared with
placebo, caffeine increased mean power output (~4%, P ⫽ 0.01) and
muscle recruitment (as inferred by EMG, ~17%, P ⫽ 0.01) and
reduced the time to complete the TT (~2%, P ⫽ 0.01). Work
performed above CP during the caffeine trial (16.7 ⫾ 2.1 kJ) was
significantly higher than during the placebo (14.7 ⫾ 2.1 kJ, P ⫽ 0.01).
End-exercise decline in quadriceps twitch force (pre- to postexercise
decrease in twitch force at 1 and 10 Hz) was more pronounced after
caffeine compared with placebo (121 ⫾ 13 and 137 ⫾ 14 N vs.
146 ⫾ 13 and 156 ⫾ 11 N; P ⬍ 0.05). There was no effect of caffeine
on central fatigue. In conclusion, caffeine increases muscle recruit-
ment, which enables greater work performed above CP and higher
end-exercise peripheral locomotor muscle fatigue.
NEW & NOTEWORTHY The link between total work done above
critical power and peripheral fatigue during a self-paced, high-inten-
sity exercise is unclear. This study revealed that caffeine ingestion
increases muscle recruitment, which enables greater work done above
critical power and a greater degree of end-exercise decline in quad-
riceps twitch force during a 4-km cycling time trial. These findings
suggest that caffeine increases performance at the expense of greater
locomotor muscle fatigue.
neuromuscular fatigue; power-time relationship; self-paced exercise;
supplementation; W=
Address for reprint requests and other correspondence: A. E. Lima-Silva
Human Performance Research Group, Academic Dept. of Physical Education
(DAEFI), Technological Federal Univ. of Parana, Pedro Gusso St. 2601,
Neoville, Curitiba, Parana 81310-900, Brazil (e-mail: aesilva@utfpr.edu.br).
http://www.jappl.org 8750-7587/18 Copyright © 2018 the American Physiological Society
Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (150.161.081.015) on June 28, 2018.
Copyright © 2018 American Physiological Society. All rights reserved.
The relationship between cycling power output (PO) and
time to task failure is characterized by a hyperbolic function
(53). The asymptote of the power-time hyperbola has been
termed critical power (CP) and represents the highest oxidative
metabolic rate that can be sustained at physiological steady
state (52, 66). The curvature constant of the hyperbola (W=)
represents a finite amount of energy available for work per-
formed in excess of the CP (44, 52). The physiological deter-
minants of W= have not been fully determined but were first
attributed to “finite anaerobic energy stores,” which comprise
the energy derived from muscle phosphocreatine (PCr) break-
down, anaerobic glycolysis, and a small amount of aerobic
energy linked to O2 stores (34, 52). Recent evidence, however,
suggests that W= may represent more than a simple marker of
anaerobic stores (11, 21, 45, 66). A full depletion of W= and
ultimately task failure during severe-intensity, constant-load
exercise seems to be associated with the attainment of critical
values of metabolites (PCr, H⫹, Pi, and ADP) (21, 39, 68),
which are related to peripheral muscle fatigue (11, 12). This
suggests a possible link between total work done above CP and
the rate of development of peripheral fatigue during constant-
load endurance exercise (11, 15).
The mechanism likely underlying the link between the total
work done above CP and peripheral fatigue is that the loss of
contractile function above CP requires the progressive recruit-
ment of additional, rapidly fatigable type II muscle fibers (18).
The recruitment of these additional motor units is associated
with a faster accumulation of intramuscular metabolites (e.g.,
H⫹ and Pi) known to cause peripheral fatigue (11, 12, 16).
Furthermore, these metabolites also stimulate group III/IV
muscle afferents that limit muscle recruitment and neural
activation of the working muscle (2, 12). This regulatory
feedback loop limits the level of intramuscular metabolic
perturbation and prevents peripheral fatigue from surpassing an
individual “critical fatigue threshold” (2, 4, 12, 41). It therefore
seems plausible to suspect that the same mechanism limiting
peripheral fatigue development may also limit the amount of
work performed above CP (11, 66). Recent studies suggest that
despite different exercise intensities and degrees of neuromus-
cular activation within the severe-intensity domain, both intra-
muscular metabolic changes (PCr, H⫹, Pi, and ADP levels)
(21, 45, 66) and peripheral fatigue reach consistent values at
1491
1492 CAFFEINE INCREASES PERIPHERAL FATIGUE ACCUMULATION
task failure (11, 15) and total work performed above CP (11,
45, 66). These findings suggest that exercise performed within
the severe-intensity domain is terminated when intramuscular
metabolic perturbations reach a critical level and this appar-
ently coincides with the depletion of W= (2, 11).
While the possible link between total work done above CP
and the development of peripheral fatigue has been explored
during constant-load exercise (17), the relationship during
self-paced exercise is largely unknown, with only one study
investigating W= use during TT exercise (20) but without
quantifying the exercise-induced fatigue. In that study, W= was
fully depleted during the first 2.5 min of a 3-min cycling TT
and was not different from the W= use during a severe-intensity
constant-load exercise trial where the participants reached
exhaustion within 3 min (20). This agrees with studies showing
that in middle-distance cycling TTs, which are supposed to be
in the severe-intensity domain (from 500-m to 4-km cycling
TT), the athletes appeared to distribute their energy resources
over the TT (29). This allows athletes to preserve the ability to
provide for anaerobic PO up to the closing stages of the event
(29, 55, 56). Simultaneous monitoring of total work performed
above CP and peripheral fatigue, however, has not been per-
formed; therefore, a further link among pacing control, total
work done above CP, and peripheral fatigue accumulation
cannot be provided.
An experimental increase in the total work completed above
CP could provide novel insights into the relationship between
total work achieved above CP and peripheral fatigue accumu-
lation during self-paced exercise. Caffeine seems a promising
candidate to test this relationship as it has been seen to increase
total work done above CP (5, 57) and to increase PO during a
4-km cycling TT (55). Caffeine acts on the central nervous
system (24, 31) and/or directly on the skeletal muscle (51, 58).
In the central nervous system, caffeine acts as an antagonist
to adenosine receptors, decreasing the rate of perceived
exertion (RPE) for a given PO (26, 31). Caffeine can also
improve calcium release from the sarcoplasmic reticulum
during muscle contraction (62, 63). Together, these central
and peripheral effects of caffeine may attenuate the devel-
opment of fatigue and increase exercise time up to task
failure (49). However, caffeine may also decrease the sen-
sory signals from the muscle to the brain (46). As PO can be
freely altered during a cycling TT, altered sensory signals
with caffeine might allow an increase in muscle recruitment
and therefore a higher PO and more work performed above
CP. It would be expected that a greater amount of work
performed above CP might result in a higher degree of
end-exercise peripheral fatigue.
Therefore, to explore a possible link among pacing strat-
egy, total work done above CP, and fatigue during self-
paced exercise, we designed the present study to test the
following questions: 1) does caffeine increase the total work
done above CP during the TT; and 2) does the expected
caffeine-induced increase in the total work performed above
CP affect end-exercise peripheral fatigue? We hypothesized
that caffeine ingestion would increase muscle recruitment,
enabling a greater amount of work to be done above the CP
and ultimately a higher degree of peripheral fatigue at the
end of the trial.
J Appl Physiol • doi:10.1152/japplphysiol.00930.2017 • www.jappl.org
Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (150.161.081.015) on June 28, 2018.
Copyright © 2018 American Physiological Society. All rights reserved.
MATERIAL AND METHODS
Participants. Eleven moderately trained male cyclists (258 ⫾ 24
km/wk) with regular participation in local competitions (~11 compe-
titions per year) were recruited for this study [age: 34 ⫾ 4 yr; body
mass: 71 ⫾ 11 kg; height: 175 ⫾ 7 cm; body fat: 11 ⫾ 3%; and
V̇O2max: 55 ⫾ 4 ml·kg⫺1·min⫺1]. Participants signed a written in-
formed consent. The study was conducted according to the Declara-
tion of Helsinki and was approved by the Research Ethics Committee
of the Federal University of Pernambuco.
Experimental design. Participants visited the laboratory on six
different occasions. On the first visit, anthropometric measurements
were taken and a maximal incremental test was performed to deter-
mine the gas exchange threshold (GET), peak power output (PPO),
and maximal oxygen uptake (V̇O2max). During the next three fol-
low-up visits, participants were familiarized with the maximum iso-
metric voluntary contraction and electrical stimulation procedures
(MVC ⫹ ES), followed by a constant-workload test performed until
task failure (1 work rate per visit), used later to determine the CP and
W=. Twenty minutes after the constant-workload test, participants
performed a 4-km familiarization cycling TT. Thus participants had
the opportunity to perform three familiarization sessions with MVC ⫹
ES and 4-km cycling TT procedures. In the last two visits, with the
use of a double-blind, random, and counterbalanced design, partici-
pants performed a 4-km cycling TT after ingesting 5 mg/kg body mass
of caffeine (CAF) or cellulose (PLA) 1 h before the trial. The MVC ⫹
ES were applied before and ~50 min after the supplementation and 1
min after the trial. The CAF and PLA were encapsulated in gelatin
capsules of the same color and shape.
All visits were scheduled at the same time of day to avoid any
effect of a circadian variation; the visits were separated by 7 days to
prevent any residual fatigue from the previous test (28). All trials were
performed using the participant’s own bike attached to a Compu-
trainer (RacerMate, Seattle, WA). The tire calibration and its pressure
to the roller were checked before each trial to guarantee no slippage.
The tire was warmed-up for 5 min before each test, following the
manufacturer’s recommendations. The seat position was recorded on
the first visit and replicated during all subsequent test sessions.
Participants refrained from vigorous physical activities, alcohol, and
caffeinated beverages 24 h before each visit. Participants were also
instructed to record their food and drink intake during the 24 h before
the first visit and to repeat this for the subsequent visits. Participants
consumed their last meal 2 h before each test session.
Body composition and maximal incremental test. Body density was
estimated using three skin folds (pectoral, abdomen, and thigh)
equation (43) and then converted to a body fat percentage using the
Siri’s equation (59). The maximal incremental test started with a
5-min warm-up at a PO of 100 W, and then, the PO was increased
30 W every 3 min until participants were unable to maintain a pedal
frequency of between 80 and 90 revolutions per minute (37). The PPO
was determined as the PO maintained during a last complete stage.
When the last stage was incomplete, the PPO was calculated using the
fractional time supported in the last incomplete stage multiplied by the
increment rate. The V̇O2 was measured breath by breath using an
automatic gas-exchange analyzer (Metalyzer 3B; Cortex, Leipzig,
Germany). Before each test, a 3-liter syringe, ambient air, and a
cylinder of known concentrations of O2 (12%) and CO2 (5%) were
used to calibrate the metabolic cart, according to the manufacturer’s
recommendations. The V̇O2max was determined from the average V̇O2
data within the last 30 s of the test. Two experienced investigators
identified visually the GET from a first disproportionate increase in
V̇CO2, an increase in V̇E/V̇O2 without increase in V̇E/V̇CO2, and an
increase in end-tidal O2 pressure with no fall in end-tidal CO2
pressure (40).
Constant workload tests. The constant workload tests were pre-
ceded by a 5-min warm-up at 100 W followed by a 5-min rest.
Participants then cycled until task failure (same criteria used in the
 CAFFEINE INCREASES PERIPHERAL FATIGUE ACCUMULATION
incremental test) at a PO corresponding to 70% of the difference
between the GET and V̇O2max (⌬70), 80% of the difference between
the GET and V̇O2max (⌬80), or 100% V̇O2max. These exercise inten-
sities were performed on separate days and were selected to yield time
to exhaustion between 3 and 12 min, which have been described as
appropriate periods of time to determine the hyperbolic power-time
limit relationship (67). Strong verbal encouragement was provided
throughout the test, and time to exhaustion was recorded to the nearest
second.
Individual CP and W= were estimated from the PO and the corre-
sponding time to exhaustion, using the least squares fitting of the
following regression models (Eqs. 1, 2, and 3) (21):
Nonlinear: t ⫽ W⬘ ⁄ (PO ⫺ CP) (1)
Linear: W ⫽ W ⬘ ⫹CP · t (2)
Linear inverse of time: PO ⫽ 共W ⬘ ⁄ t兲 ⫹ CP (3)
where t is the time to exhaustion, W is the total work performed, W=
is the constant curvature of hyperbole power time, PO is the power
output, and CP is critical power.
The standard error of estimates associated with the CP and W= are
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 (Eqs. 1, 2, or 3) to
produce the “best-individual-fit” parameter estimates (10, 57). This
best individual fit was used for further analysis.
Experimental sessions. Immediately after the participants arrived at
the laboratory, they were attached to the electrodes for ES and EMG
recording, and an intravenous catheter was inserted into the antecu-
bital vein. A blood sample (10 ml) was collected, and then, partici-
pants performed a standardized warm-up for knee extension muscles
(4 ⫻ 5-s isometric contractions at 50, 60, 70, and 80% of the MVC
recorded in the familiarization sessions, interspersed by a 30-s rest).
Following a 5-min rest, six MVC ⫹ ES were measured, with 30-s rest
between them (baseline). Thereafter, participants ingested CAF or
PLA and rested seating comfortably for 40 min, when a second blood
sample was taken (10 mL) and knee extension muscle warm up and
MVC ⫹ ES measurements were repeated (pre-TT). Participants then
rested for 5 min and performed a 5-min warm-up at 100 W on the
bicycle, followed by a 5-min rest. Then, participants were instructed
to perform the 4-km cycling TT in the shortest possible time. Partic-
ipants were oriented to remain seated throughout the trial and were
free to alter gear ratio and/or pedaling frequency during whole trial.
They received visual feedback for distance completed but not for
exercise time, PO, and pedal frequency. The RPE was obtained every
kilometer using the Borg 15-point scale (13). The PO was measured
every second via an interface connecting Computrainer to the com-
puter (RacerMate One software, Seattle, WA). Immediately after the
trial, a blood sample (10 ml) was collected. The postexercise MVC ⫹
ES measurements (post-TT) started 1 min after the end of the trial.
Neuromuscular assessment. Neuromuscular function was assessed
with the participants seated on a custom-made bench with their trunk
and thigh angles of 120° and the knee joint angle set at 90°. A
noncompliant cuff attached to a calibrated linear strain gauge (EMG
System of Brazil, São Jose dos Campos, Brazil) was fixed to the right
ankle just superior to the malleoli for force measurement. A mono-
polar 0.5-cm diameter cathode electrode (Ambu Neuroline 715, Bal-
lerup, Denmark) was positioned at the right femoral nerve. During the
familiarization sessions, the electrode position was moved around the
femoral triangle and percutaneous electrical nerve stimuli were ap-
plied using an electrical stimulator (Neuro-TES; Neurosoft, Ivanovo,
Russia) until the highest values were obtained for muscle membrane
excitability (Mwave) and quadriceps contraction potential (Qtw). The
optimal intensity of stimulation was determined by increasing inten-
sity of a single electrical stimulus (1 Hz and 80-␮s duration) by 30 V
J Appl Physiol • doi:10.1152/japplphysiol.00930.2017 • www.jappl.org
Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (150.161.081.015) on June 28, 2018.
Copyright © 2018 American Physiological Society. All rights reserved.
1493
every 30 s (starting at 100 V) until a plateau occurred in the Mwave and
Qtw (61). The anode was positioned on the gluteal fold opposite the
cathode. Before each session, the position of electrodes was marked
with indelible ink to ensure identical placement and plateau in the
Mwave and Qtw and was checked before each trial.
For the assessment of neuromuscular function, participants were
instructed to produce their maximal torque and maintain it for 5 s
during each MVC. Visual feedback and consistent verbal encourage-
ment were provided throughout the voluntary contractions. Square
wave stimuli (80 ␮s) were delivered to the femoral nerve with the
stimulation intensity set at 120% of plateau M wave and Qtw
(300 ⫾ 50 V). The first stimulation (1 Hz) was applied during the
MVC (superimposed twitch) as soon as the plateau in isometric fo-
rce was reached. Potentiated quadriceps twitch torque evoked by
single (Qtwpot) and paired pulses at 10 Hz (Qtw10) and 100 Hz
(Qtw100) were elicited 2, 4, and 6 s, respectively, after each MVC.
The average of the last four MVC for each moment was used for
further analysis of the neuromuscular function (3, 15). The MVC was
used as a global index of fatigue (65). The maximal voluntary
activation (VA) was used to assess the capacity of the central nervous
system to activate the working muscle to the maximum (i.e., central
fatigue). The VA was estimated by the interpolated-twitch technique,
using the following equation (48): 1 – (superimposed twitch/Qtwpot) ⫻
100. Contraction was excluded from analysis when stimulus was not
applied in the plateau of maximal force (~8% of a total of 288
contractions).
The Qtwpot was used to assess the global peripheral fatigue, while
Qtw10 and Qtw100 were used to assess peripheral low- and high-
frequency fatigue, respectively (50). Furthermore, other parameters of
muscle contractile properties were obtained from the 1-Hz stimulation
twitch: 1) time between the stimulus and peak force (contraction
time); 2) the maximum instantaneous ascending slope of force devel-
opment (maximal rate of twitch torque development); 3) the maxi-
mum instantaneous descending slope of force development (maximal
rate of twitch torque relaxation); and, 4) time required to twitch return
to half of the twitch peak (half relaxation time).
Electromyography recording. During MVC and 4-km cycling TT,
EMG activity from the right vastus lateralis was recorded with a
sample rate of 20,000 Hz (Neuro-MEP-Micro; Neurosoft, Ivanovo,
Russia) using monitoring electrodes with full-surface solid adhesive
hydrogel (Ambu Neuroline 715, Ballerup, Denmark). The skin was
shaved, abraded with emery paper, and cleaned with alcohol to reduce
skin impedance ⬍5 k⍀. Electrodes were placed according to the
Surface Electromyography for the Non-Invasive Assessment of Mus-
cles Standards (36). The position of the electrodes was maintained by
an elastic net bandage wrapped around the thigh to reduce the noise
and marked with indelible ink to ensure identical placement on
subsequent visits.
The raw EMG signal was full-wave rectified and filtered with
second-order Butterworth band-pass filters with cutoff frequencies set
at 20 and 500 Hz to remove external interference noise and movement
artifacts. Root mean square (RMS) was calculated during the 250 ms
before the electrical stimulation in each MVC (RMSMVC). During the
4-km cycling TT, the mean of RMS values for every 400-m segment
was calculated and normalized by RMSMVC at baseline.
M-wave peak-to-peak amplitude (Mwave ampl) was calculated for
each 1 Hz stimulus. The beginning of the M wave was considered as
an increase of 2 SD above the baseline values, while the ending as a
reduction to values ⬍2 SD below the baseline (54).
Systemic responses to exercise. During the 4-km cycling TT, V̇E,
V̇O2, and V̇CO2 were measured breath by breath and averaged for each
400 m (Metalyzer 3B; Cortex). Heart rate was measured using a
transmitter coupled to a gas analyzer. Blood samples (10 ml) were
immediately transferred to a Falcon tube, and the pH was immediately
analyzed by a portable pH meter (HI 8424; Hanna). An additional 4
ml were transferred to tubes with EDTA and centrifuged for 10 min
at 3000 rpm for plasma separation. Plasma ammonia (Randox, Crum-
1494 CAFFEINE INCREASES PERIPHERAL FATIGUE ACCUMULATION

lin, UK) and lactate (Labtest, Minas Gerais, Brazil) were measured by
the enzymatic colorimetric method using a spectrophotometer (SP-22;
Biospectro, Curitiba, Brazil).
Data analysis. The amount of work performed above CP during the
4-km cycling TT was calculated as the power-time integral above CP,
recorded at the nearest 1-s interval. Some participants (4 in PLA and
3 in CAF) dropped down below CP (see RESULTS). As exercise
performed below CP reconstitutes partially the W= (60), the W=
reconstitution (W=rec) was taken into account for these participants
and subtracted from the calculated W=use using the model proposed
by Skiba et al. (60):
t
RESULTS
The CP and W= estimates. The mean PO corresponding to
⌬70, 80, and 100% V̇O2max were 250 ⫾ 8, 265 ⫾ 8, and
287 ⫾ 9 W, respectively. The corresponding times to task
failure were 854 ⫾ 116, 540 ⫾ 73, and 307 ⫾ 41 s, respec-
tively. The parameters derived from Eqs. 1, 2, and 3, and from
the best individual fit, are shown in Table 1.
Exercise performance. The mean time to complete the 4-km
cycling TT was significantly faster (1.8 ⫾ 0.6%) in CAF com-
pared with PLA (396 ⫾ 5 vs. 403 ⫾ 6 s, respectively, P ⫽

兰共
W⬘rec ⫽ W⬘exp兲共e⫺共t⫺u兲⁄␶W⬘兲
0 TT are shown in Fig. 1. PO and RMSTT during the 4-km
where W=exp represents the amount of the starting W= that was
currently expended, (t ⫺ u) is equal to the time in seconds when the
participant was recovering below CP, and ␶W= is the time constant of
the reconstitution of the W=.
The ␶W= was calculated using the regression equation previously
reported by Skiba et al. (60):
␶W⬘ ⫽ 546e共⫺0,01Dcp兲 ⫹ 316
where Dcp is equal to the difference between the recovery power and
the participant’s CP.
Reliability measurements. Performance time and neuromuscular
assessments from the second and third familiarization sessions were
used to calculate the reliability of the measurements. Mean between-
day, within-subject coefficients of variation were 1.7 ⫾ 2.3% (range:
1.2–5.0%) for time to cover 4-km cycling TT, 4.8 ⫾ 0.9% (range:
3.7–7.4%) for MVC, 1.8 ⫾ 0.7% (range: 0.3–3.2%) for VA, and
5.1 ⫾ 1.1% (range: 3.4 – 8.3%) for Qtwpot.
Statistical analysis. The distribution of the data was analyzed by
the Shapiro-Wilk test. Once normality was confirmed, the two-way
ANOVA with repeated measures was used to verify the main effects
of supplement, distance, and interaction on PO, root mean square for
time trial (RMSTT), and total amount of work performed above CP
during the 4-km cycling TT. Two-way ANOVA with repeated mea-
sures was also used to test the main effect of supplement, moment
(baseline, pre-TT, and post-TT) and interaction on neuromuscular and
blood parameters. One-way ANOVA with repeated measure was used
to compare the V̇O2 reached during the 4-km cycling TT in CAF and
PLA conditions, with the V̇O2max achieved during the incremental test.
When differences were noted, the Duncan’s new multiple range test
was used to locate these differences. Paired t-test was used to compare
performance time, mean PO, total amount of work performed above
CP, and recovery parameters of W= (W=rec, PO, and time recovery).
Pearson correlation was performed between changes in total amount
of work performed above CP and changes in Qtwpot, Qtw10, and Qtw100
(⌬ between CAF vs. PLA). Statistical significance was reported when
P ⬍ 0.05. Analyses were performed using Statistics Package for
Windows (version 10; StataSoft, Tulsa, OK). Data are reported as
mean ⫾ SE.
0.01).
(4) PO and muscle recruitment. The PO and RMSTT during the
cycling TT were higher in CAF compared with PLA (both P ⫽
0.01, Fig. 1, A and B). There was no interaction between
supplement and distance for PO (P ⫽ 0.78), but there was a
main effect of distance (P ⫽ 0.01). Participants adopted a
fast-start strategy in both conditions, with PO at the first 400 m
being higher than PO from 800 to 4,000 m (for all P ⬍ 0.05,
Fig. 1A). The PO at 800 m was also higher than PO from 1,200
(5) to 3,600 m (for all P ⬍ 0.05). A final sprint was noted under
both conditions, with higher PO at 4,000 m than from 1,200 to
3,600 m (for all P ⬍ 0.05). There was no interaction between
supplement and distance (P ⫽ 0.14) for RMSTT, but there was
a main effect of distance (P ⫽ 0.01). The muscle recruitment
was significantly higher in the first 400 m than from 800 to
3,600 m and at 800 m higher than from 2,800 to 3,600 m under
both conditions (for all P ⬍ 0.05).
Amount of work performed above CP. The distribution and
total amount of work performed above CP during the TTs are
shown in Fig. 2. There was a main effect of supplement (P ⫽
0.01), distance (P ⫽ 0.01), and supplement ⫻ distance inter-
action (P ⫽ 0.02) for work done above CP (Fig. 2A). The work
done above CP at 400 and 4,000 m was significantly higher in
CAF than in PLA condition (P ⫽ 0.04 and 0.01, respectively).
In addition, the work accomplished above CP at 400 m was
significantly higher than between 800 and 4,000 m and at 800
m higher than 1,200 to 3,600 m for both conditions (all P ⬍
0.05). Work done above CP at 4,000 m was also significantly
higher than 1,200 to 3,600 m for both conditions (all P ⬍ 0.05).
The total work done above CP was ~14% higher in the CAF
than under the PLA condition (16.7 ⫾ 2.1 vs. 14.7 ⫾ 2.1 kJ,
respectively, P ⫽ 0.01).
Some participants (4 in PLA and 3 in CAF) dropped below
CP a few times during the trial. The three participants who
dropped below CP under the CAF condition also dropped
below CP under the PLA condition. For these participants,
there was no significant difference for the mean PO performed
below CP between CAF and PLA (13 ⫾ 3 vs. 16 ⫾ 3 W below

Table 1. The parameter estimates derived from Eqs. 1, 2, and 3 and the best individual fit
R2 CP, W SEE, W CV, % W=, kJ SEE, kJ CV, % TE, %

Nonlinear 0.997–1.000 225 ⫾ 7 3.9 ⫾ 1.4 3.3 ⫾ 1.7 15.9 ⫾ 1.6 1.1 ⫾ 1.5 9.8 ⫾ 7.4 14.1 ⫾ 7.8
Linear 0.975–1.000 223 ⫾ 7 6.5 ⫾ 2.9 3.3 ⫾ 1.9 15.6 ⫾ 1.5 3.0 ⫾ 1.0 9.5 ⫾ 3.2 12.8 ⫾ 5.8
Linear inverse of time 0.914–1.000 226 ⫾ 8 6.9 ⫾ 2.8 3.4 ⫾ 2.2 15.5 ⫾ 1.5 2.5 ⫾ 1.0 10.7 ⫾ 4.1 14.1 ⫾ 5.8
Best individual fit 0.998–1.000 223 ⫾ 8 3.5 ⫾ 0.8 2.2 ⫾ 0.9 16.0 ⫾ 2.0 1.3 ⫾ 0.4 8.7 ⫾ 3.0 10.9 ⫾ 4.4
Data are expressed as means ⫾ SE (except R2 that are minimum and maximum); n ⫽ 11. CP, critical power; SEE, standard error of estimate; CV, coefficient
of variation; W=, constant curvature of hyperbole power time; TE, total error, which indicates the sum of CV% associated with CP and W=.

J Appl Physiol • doi:10.1152/japplphysiol.00930.2017 • www.jappl.org

Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (150.161.081.015) on June 28, 2018.
Copyright © 2018 American Physiological Society. All rights reserved.
 CAFFEINE INCREASES PERIPHERAL FATIGUE ACCUMULATION
CP, respectively, P ⫽ 0.93). The amount of time spent below CP
during the TT was also not significantly different between CAF
and PLA (86 ⫾ 20 vs. 87 ⫾ 23 s, respectively, P ⫽ 0.46).
Consequently, the W= reconstitution during the trial was min-
imal and not different between CAF and PLA (1.2 ⫾ 0.1 vs.
1.6 ⫾ 0.1 kJ, respectively, P ⫽ 0.86). A detailed description of
the distance points when PO was reduced below CP for these
participants is provided in supplementary Table S1 (Supple-
mental Material for this article is available online at the Journal
website).
Neuromuscular function. The absolute values for neuromus-
cular parameters at the baseline, pre-TT, and post-TT are
reported in Table 2. As there was no significant difference
between baseline and pre-TT for any variable (all P ⬎ 0.05,
Table 2), the percentage of changes in some key neuromuscu-
lar parameters of central and peripheral fatigue after the 4-km
TT was also expressed relative to the baseline values (Fig. 3).
J Appl Physiol • doi:10.1152/japplphysiol.00930.2017 • www.jappl.org
Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (150.161.081.015) on June 28, 2018.
Copyright © 2018 American Physiological Society. All rights reserved.
1495
Fig. 1. Power output (PO; A) and EMG activity (B) for each
400-m segment (lines) and mean values (columns) during
the time trial (TT) after placebo (PLA) and caffeine (CAF)
ingestion. Data are reported as means ⫾ SE (n ⫽ 11).
Horizontal dotted line represents the group average for
critical power (CP). MVC, maximum isometric voluntary
contraction. *Mean power output (PO) and root mean
square of time trial (RMSTT) are significantly higher in
CAF than in PLA. †PO and RMS at 400 m were signifi-
cantly higher than those from 800 to 4,000 m for both
conditions. ‡PO at 800 m is significantly higher than 1,200
to 3,600 m for both conditions. §PO at 4,000 m is signifi-
cantly higher than PO from 1,200 to 3,600 m for both
conditions. #RMS at 800 m is significantly higher than
RMS from 2,800 to 3,600 m for both conditions. Statistical
analysis was performed using two-way ANOVA with re-
peated measures followed by a Duncan’s new multiple
range test for comparison at each 400 m and a paired t-test
for mean values.
There was no main effect of supplement (P ⫽ 0.18) or
interaction supplement and time (P ⫽ 0.82) for MVC. How-
ever, there was a main time effect, with a reduction in MVC
post-TT, in comparison to baseline and pre-TT (P ⫽ 0.01).
There was no main effect of condition, time, or interaction for
RMSMVC and VA (all P ⬎ 0.05). However, the Qtwpot and
Qtw10 decreased post-TT compared with baseline and pre-TT
(P ⫽ 0.01), but there was a significant supplement ⫻ time
interaction (P ⫽ 0.01), with the reduction being more pro-
nounced in CAF than in PLA (P ⫽ 0.03 and 0.02, respec-
tively). There was a significant main time effect (all P ⬍ 0.05),
but without main effect of supplement or supplement ⫻ time
interaction (all P ⬎ 0.05) for Qtw100, maximal rate of twitch
torque development, half-relaxation time, and maximal rate of
twitch torque relaxation, with all reducing at post-TT com-
pared with the baseline and pre-TT. There was no time,
supplement, or interaction for the contraction time (all P ⬎
1496 CAFFEINE INCREASES PERIPHERAL FATIGUE ACCUMULATION

Fig. 2. Distribution (A) and total work performed above critical power (CP; B) during the time trial (TT) under placebo (PLA) and caffeine (CAF) conditions.
Data are reported as means ⫾ SE (n ⫽ 11). *Work done above critical power (CP) in CAF is significantly higher than corresponding distance point for PLA
in A and total work done above CP (W=) use in CAF is significantly higher than PLA in B. †Significantly higher than 800 to 4,000 m for both conditions.
‡Significantly higher than 1,200 to 3,600 m under both conditions. Statistical analysis was performed using two-way ANOVA with repeated measures followed
by a Duncan’s new multiple range test for comparison at each 400 m (A) and a paired t-test for total work performed above CP (B).

0.05). The Mwave ampl increased post-TT compared with base-
line and pre-TT, under both conditions (P ⫽ 0.01), but there
was a supplement ⫻ time interaction, with greater increases in
CAF compared with PLA (P ⫽ 0.01).
Relation between the amount of work performed above CP
and peripheral fatigue. There was no correlation between the
changes (⌬) in the amount of work performed above CP with
the changes in the Qtwpot (r ⫽ 0.60, P ⫽ 0.11), Qtw10 (r ⫽
⫺0.19, P ⫽ 0.63), and Qtw100 (r ⫽ 0.15, P ⫽ 0.69).
Systemic responses to exercise and blood parameters. The
systemic responses during TTs are shown in Fig. 4. The V̇O2
reached during the trial was similar (P ⫽ 0.12) between CAF
and PLA (3.46 ⫾ 0.14 and 3.41 ⫾ 0.15 l/min, respectively),
and both were similar (P ⫽ 0.45) to the values achieved during
the maximal incremental test (3.51 ⫾ 0.19 l/min). The V̇O2 and
V̇CO2 increased from rest to the first 800 m (P ⫽ 0.01) and then
remained stable throughout the trial under both conditions (all
P ⬎ 0.05). Heart rate increased progressively from rest to
1,600 m (P ⫽ 0.01) and then remained stable until the end of
the trial under both conditions (all P ⬎ 0.05). The RPE
increased progressively throughout the trial (P ⫽ 0.01), with-
out differences between conditions (P ⫽ 0.10).
There was no difference for blood pH, plasma lactate, and
ammonia between baseline and pre-TT under both conditions.
In post-TT, blood pH dropped, and plasma lactate and ammo-
nia increased significantly (all P ⬍ 0.05), without differences
between CAF and PLA (P ⬎ 0.05, Table 3).
DISCUSSION
The main novel finding of the present study was that caffeine
ingestion increased muscle recruitment during the TT leading

Table 2. Absolute values for neuromuscular function at baseline and pre and post a 4-km cycling TT after PLA and
CAF ingestion
Baseline Pre-TT Post-TT

PLA CAF PLA CAF PLA CAF

MVC, N 405 ⫾ 35 422 ⫾ 38 409 ⫾ 33 436 ⫾ 35 375 ⫾ 29† 396 ⫾ 36†

VA, % 90 ⫾ 2 90 ⫾ 1 89 ⫾ 2 89 ⫾ 2 86 ⫾ 2 88 ⫾ 1
RMSMVC, mV 0.22 ⫾ 0.04 0.22 ⫾ 0.04 0.25 ⫾ 0.03 0.27 ⫾ 0.05 0.25 ⫾ 0.04 0.28 ⫾ 0.05
Qtwpot, N 159 ⫾ 22 163 ⫾ 16 158 ⫾ 11 165 ⫾ 15 137 ⫾ 14† 121 ⫾ 13*†
Qtw10, N 208 ⫾ 17 214 ⫾ 16 203 ⫾ 17 207 ⫾ 17 156 ⫾ 11† 146 ⫾ 13*†
Qtw100, N 235 ⫾ 14 253 ⫾ 14 230 ⫾ 8 251 ⫾ 13 216 ⫾ 8† 212 ⫾ 9†
CT, ms 107 ⫾ 10 102 ⫾ 11 110 ⫾ 13 108 ⫾ 15 104 ⫾ 18 106 ⫾ 17
MRFD, N/ms 1,568 ⫾ 9 1,541 ⫾ 10 1,567 ⫾ 9 1,580 ⫾ 10 1,266 ⫾ 8† 1,227 ⫾ 8†
RT0.5, ms 89 ⫾ 12 90 ⫾ 14 86 ⫾ 12 85 ⫾ 8 72 ⫾ 12† 63 ⫾ 18†
MRR, N/ms 1,494 ⫾ 201 1,562 ⫾ 250 1,440 ⫾ 208 1,571 ⫾ 240 1,410 ⫾ 199† 1,451 ⫾ 215†
Mwave ampl, mV 19 ⫾ 4 20 ⫾ 3 20 ⫾ 5 21 ⫾ 4 22 ⫾ 4† 24 ⫾ 3†*
Data are reported as means ⫾ SE; n ⫽ 11. PLA, placebo; CAF, caffiene; MVC, maximal voluntary contraction; VA, percent of voluntary activation; RMSMVC,
root mean square at MVC; Qtwpot, potentiated quadriceps twitch torque evoked by single stimulus, Qtw10 and Qtw100, potentiated quadriceps twitch torque evoked
by paired electrical stimulation at a frequency of 10 Hz and 100 Hz; CT, contraction time; MRFD, maximal rate of force development; RT0.5, half-relaxation
time; MRR, maximal rate of relaxation; Mwave ampl, maximal M-wave amplitude; TT, time trial. Statistical analysis was performed using two-way ANOVA with
repeated measures followed by a Duncan’s new multiple range test. *Significantly different from the corresponding PLA. †Significantly different from baseline
and pre-TT.

J Appl Physiol • doi:10.1152/japplphysiol.00930.2017 • www.jappl.org

Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (150.161.081.015) on June 28, 2018.
Copyright © 2018 American Physiological Society. All rights reserved.
 CAFFEINE INCREASES PERIPHERAL FATIGUE ACCUMULATION
to a greater amount of work performed above CP and improved
performance. A higher end-exercise peripheral fatigue, how-
ever, was noted when TT was performed under caffeine influ-
ence, but this increased level of peripheral fatigue was not
associated with the increased amount of work performed above
CP. This is the first study showing that caffeine improves
exercise capacity but at the expense of an exaggerated level of
end-exercise locomotor muscle fatigue.
Caffeine ingestion reduced time to complete the 4-km cy-
cling TT by ~2% confirming previous reports (55, 56). Pacing
adopted by all subjects in the present study was a U-shaped PO
profile (Fig. 1A), which is also in accordance with results from
other studies with 4-km cycling TT (29, 55, 56). Caffeine
supplementation increased mean PO during the trial without
changing the pacing strategy, but it is interesting to note that
the points where work above CP was higher with caffeine
when compared with the placebo were in the first and last 400
m (Fig. 2A). This might indicate that an optimal work output
during these two critical points increases overall performance
during the TT. This is in agreement with recent studies report-
ing that augmenting deceptive feedback (i.e., following an
avatar set at a PO 2% higher than baseline) improves overall
performance for allowing increase PO either at the first (1) or
last (27) part of a middle-distance TT. This is also in accor-
dance with studies showing that “enforcing” a faster start
improves overall performance during middle-distance TT (1,
9). However, it is not known whether an “enforced” final sprint
would improve overall performance. Together, these findings
suggest that during a complexly regulated, middle-distance
cycling TT, maximum performance in the first and last parts of
J Appl Physiol • doi:10.1152/japplphysiol.00930.2017 • www.jappl.org
Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (150.161.081.015) on June 28, 2018.
Copyright © 2018 American Physiological Society. All rights reserved.
1497
Fig. 3. Percentage of change from baseline to post time trial
for maximum isometric voluntary contraction (MVC; A);
voluntary activation (VA; B); potentiated quadriceps twitch
torque evoked by single (Qtwpot; C) and paired pulses at 10
Hz (Qtw10; D) and 100 Hz (Qtw100; E); and maximal
M-wave amplitude (Mwave ampl; F) in the placebo (PLA) and
caffeine (CAF) conditions. *Significantly different from
PLA. Statistical analysis was performed using two-way
ANOVA with repeated measures followed by a Duncan’s
new multiple range test.
a middle-distance TT might be obtained by internal and/or
external environmental changes.
The higher performance and the consequent greater total
amount of work performed above CP under the experimental
condition may be explained by the central effects of caffeine.
Although RPE was similar between conditions (Fig. 4D), the
higher average PO with a similar RPE during the caffeine trial
indicates a lower PO/RPE ratio (6, 42, 55). Moreover, the
higher muscle recruitment, as suggested by the greater mean
RMSTT throughout the trial (Fig. 1B), further supports a
beneficial central effect of caffeine in terms of exercise per-
formance. The most recognized central effect of caffeine is its
antagonistic action on adenosine receptors in the central ner-
vous system (23), which results in a reduced effort perception
(25, 26). Caffeine may also have hypoalgesic effects, which
could increase tolerance for discomfort during exercise (25). In
addition, a greater motor-evoked potential, self-sustained firing
of motor units, and increased spinal excitability with caffeine
consumption have been reported (46, 69, 70). Caffeine, when
consumed at physiological doses, increases neural excitability
and the release of excitatory neurotransmitters, reducing the
threshold of neural activation (30, 31). Caffeine also increases
the serotonin concentration in the raphe nucleus, which plays
an important role in transmitting the signal coming from the
motor cortex to the exercising muscles (8). Together, these
effects of caffeine on motor cortex and/or spinal excitability
seem to contribute to the increase in muscle recruitment and
the subsequent greater amount of work performed above CP.
It is tempting to suggest that these caffeine-induced effects
on the muscle recruitment resulting in increased total work
1498 CAFFEINE INCREASES PERIPHERAL FATIGUE ACCUMULATION

Fig. 4. Response of V̇O2 (A), V̇CO2 (B), heart rate (HR; C), and rate of perceived exertion (RPE; D) during the time trial (TT) in the placebo (PLA) and caffeine
(CAF) conditions. Data are reported as means ⫾ SE (n ⫽ 11). Horizontal dashed line in A represents the V̇O2max achieved at the end of the maximal incremental
test. †Significantly higher than 400 m for both conditions. ‡Significantly higher than 400, 800, and 1,200 m for both conditions. §Higher than preceding point
for both conditions. Statistical analysis was performed using two-way ANOVA with repeated measures followed by a Duncan’s new multiple range test for
comparison between the distance points.

performed above CP led to greater peripheral fatigue. The
increases in total work performed above CP, however, were not
related to the increases in peripheral fatigue indexes after
caffeine ingestion (⌬ of changes). These suggest that a higher
peripheral fatigue after caffeine ingestion is not directly related
to the prior higher work performed above CP. However, these
data indicate that a given critical peripheral threshold can be
surpassed with oral caffeine. It should also be noted that
caffeine increases end-exercise decline in quadriceps twitch
force when stimulated at low (10 Hz) but not at high frequency
(100 Hz) (Fig. 3, D and E, respectively). Caffeine increases
low-frequency force (63), perhaps by increasing calcium re-
lease from the sarcoplasmic reticulum during muscle contrac-
tions (62). These improved muscle functions may reduce
muscle sensory signals to the brain, enabling greater muscle
recruitment with a consequent higher degree of end-exercise
peripheral fatigue.
A similar reduction in MVC between caffeine and the
placebo after the TT indicates that overall fatigue accumulation
was similar between conditions (Fig. 3A). Some alteration in
muscle activation and/or excitation-contraction coupling may
have occurred to compensate the higher peripheral fatigue with
caffeine to preserve a similar reduction in MVC. No significant
changes in VA after the trial for both placebo and caffeine

Table 3. Blood pH, and plasma lactate and ammonia at baseline, pre-TT, and post-TT for CAF and PLA conditions
Baseline Pre-TT Post-TT

PLA CAF PLA CAF PLA CAF

pH 7.36 ⫾ 0.1 7.35 ⫾ 0.1 7.36 ⫾ 0.1 7.36 ⫾ 0.1 7.18 ⫾ 0.1† 7.19 ⫾ 0.1†
Lactate, mmol/l 1.3 ⫾ 0.1 1.2 ⫾ 0.1 1.2 ⫾ 0.1 1.3 ⫾ 0.1 12.6 ⫾ 0.9† 13.8 ⫾ 1.1†
Ammonia, ␮mol/l 41.9 ⫾ 7.5 34.3 ⫾ 5.6 41.2 ⫾ 10.5 34.1 ⫾ 7.4 78.4 ⫾ 12.0† 91.5 ⫾ 13.7†
Data are expressed as means ⫾ SE; n ⫽ 11. PLA, placebo; CAF, caffiene; TT, time trial. †Post-TT significantly different from baseline and pre-TT in both
conditions (P ⬍ 0.05). Statistical analysis was performed using two-way ANOVA with repeated-measures followed by a Duncan’s new multiple range test.

J Appl Physiol • doi:10.1152/japplphysiol.00930.2017 • www.jappl.org

Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (150.161.081.015) on June 28, 2018.
Copyright © 2018 American Physiological Society. All rights reserved.
 CAFFEINE INCREASES PERIPHERAL FATIGUE ACCUMULATION
suggest that there was no central fatigue after the trial and
caffeine did not increase maximal muscle activation (Fig. 3B).
Membrane excitability, as represented by the M wave, in-
creased after the trial under both conditions (Fig. 3F), corrob-
orating previous findings for high-intensity exercise (32, 64).
The reason for this increased M wave after exercise is not
clear, but it has been demonstrated that an epinephrine-induced
increase in Na⫹-K⫹ pump activity (22), as probably happens
during a high-intensity exercise, that might increase the mem-
brane potential (38), would explain the increased M wave
found here and elsewhere (32, 64). M-wave increases were
higher for caffeine, however, which suggests that a potential
effect of caffeine on membrane excitability may have pre-
vented a more pronounced reduction in MVC and Qtw. Caf-
feine has been reported to decrease extracellular K⫹ accumu-
lation after high-intensity exercise, which may preserve muscle
fiber membrane excitability (51, 58).
One aspect that should be taken into account in our study
is the possible influence of the protocol identifying CP. The
CP was identified during severe-intensity, constant-load
exercises, while the amount of work performed above CP
was measured during a self-paced exercise. This implies that
CP was constant during whole TT and no different from the
constant-load exercise. While we are not aware of any study
verifying CP changes during a TT, a recent study has
suggested that the CP derived from the self-paced protocol
was greater than CP derived from the constant-load protocol
(11). This suggests that CP may have been underestimated,
and consequently, in the present study, the amount of work
output above CP was overestimated. However, any effect of
protocol estimating CP should affect the amount of work
above CP similarly under both conditions. Another aspect
that should also be noted is the possible influence of caffeine
itself on the CP (5, 19). While one study noted that caffeine
did not change CP but increased W= identified during a
3-min all-out trial (19), another study reported a higher CP
with a concomitant reduction in W= with caffeine when
identified during constant workload tests (5). Therefore, a
potential influence of caffeine on CP itself cannot be fully
disregarded.
A limitation of the present study is that we were unable to
confirm the effects of caffeine abstinence and of an increase in
plasma caffeine concentration. An analysis of food records,
however, confirmed that the participants did not ingest foods or
beverages containing caffeine 24 h before the trials. In addi-
tion, previous studies have supported the use of 5 mg/kg of
caffeine to raise plasma caffeine to ~40 ␮mol/l, a concentration
able to cause physiological alterations and improvement in
exercise performance, similar to what is reported in the present
study (7, 47). Although care was taken to ensure that post-TT
neuromuscular testing was completed immediately after exer-
cise, there was usually a delay between the dismounting from
the cycle ergometer and move to the knee extension chair.
However, the transition time (1 min) was kept constant across
the conditions. Since recovery began immediately after the end
of exercise (33), the observed exercise-induced changes might
have been enhanced if no recovery time had been required. In
fact, both peripheral (33) and central (35) fatigue parameters
can recovery rapidly within 1 min after exercise. Another
important aspect is that we recruited recreationally trained
individuals; therefore if these results would be replicable in
J Appl Physiol • doi:10.1152/japplphysiol.00930.2017 • www.jappl.org
Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (150.161.081.015) on June 28, 2018.
Copyright © 2018 American Physiological Society. All rights reserved.
1499
well-trained and professional athletes remains unknown, as the
caffeine effect might be reduced with an increase in the
performance level, particularly during evening trials (14).
In conclusion, caffeine increases muscle recruitment and
consequently the amount of work performed above CP and PO,
improving exercise performance. This increased performance,
however, might result in a higher accumulation of peripheral
fatigue.
ACKNOWLEDGMENTS
L. C. Felippe, G. A. Ferreira, and S. K. Learsi are grateful to Coordination
for the Improvement of Higher Education Personnel (CAPES) for scholarships.
We all thank Markus Amann and Dr. Marcos Silva-Cavalcante for advice and
critical feedback on the paper. The English text of this paper has been revised
by Sidney Pratt (Recife, Brazil), Master of Arts in Teaching Program at The
Johns Hopkins University, and Royal Society of Arts Diploma-Teachers of
English as a Foreign Language Program at Cambridge University.
GRANTS
This work was supported by the Brazilian National Council for Scientific
and Technological Development (Special Visitor Researcher Of Science With-
out Borders Program Process No. 406201/2013-7).
DISCLOSURES
No conflicts of interest, financial or otherwise, are declared by the authors.
AUTHOR CONTRIBUTIONS
L.C.F., R.C.B., and A.E.L.-S. conceived and designed research; L.C.F.,
G.A.F., S.K.L., and A.E.L.-S. performed experiments; L.C.F., D.B., and
A.E.L.-S. analyzed data; L.C.F., D.B., R.C.B., and A.E.L.-S. interpreted results
of experiments; L.C.F. and A.E.L.-S. prepared figures; L.C.F., R.C.B., and
A.E.L.-S. drafted manuscript; L.C.F., G.A.F., S.K.L., and A.E.L.-S. edited and
revised manuscript; L.C.F., G.A.F., S.K.L., D.B., R.C.B., and A.E.L.-S.
approved final version of manuscript.
REFERENCES
1. Aisbett B, Le Rossignol P, McConell GK, Abbiss CR, Snow R. Effects
of starting strategy on 5-min cycling time-trial performance. J Sports Sci
27: 1201–1209, 2009. doi:10.1080/02640410903114372.
2. Amann M, Blain GM, Proctor LT, Sebranek JJ, Pegelow DF, Demp-
sey JA. Group III and IV muscle afferents contribute to ventilatory and
cardiovascular response to rhythmic exercise in humans. J Appl Physiol
(1985) 109: 966 –976, 2010. doi:10.1152/japplphysiol.00462.2010.
3. Amann M, Dempsey JA. Locomotor muscle fatigue modifies central
motor drive in healthy humans and imposes a limitation to exercise
performance. J Physiol 586: 161–173, 2008. doi:10.1113/jphysiol.2007.
141838.
4. Amann M, Proctor LT, Sebranek JJ, Pegelow DF, Dempsey JA.
Opioid-mediated muscle afferents inhibit central motor drive and limit
peripheral muscle fatigue development in humans. J Physiol 587: 271–
283, 2009. doi:10.1113/jphysiol.2008.163303.
5. Arcoverde L, Silveira R, Tomazini F, Sansonio A, Bertuzzi R, Lima-
Silva AE. Effect of caffeine ingestionon anaerobic capacity quantified by
different methods. PLoS One 12: e0179457, 2017. doi:10.1371/journal.
pone.0179457.
6. Astorino TA, Cottrell T, Talhami Lozano A, Aburto-Pratt K, Duhon
J. Effect of caffeine on RPE and perceptions of pain, arousal, and
pleasure/displeasure during a cycling time trial in endurance trained and
active men. Physiol Behav 106: 211–217, 2012. doi:10.1016/j.physbeh.
2012.02.006.
7. Battram DS, Graham TE, Richter EA, Dela F. The effect of caffeine on
glucose kinetics in humans–influence of adrenaline. J Physiol 569: 347–
355, 2005. doi:10.1113/jphysiol.2005.097444.
8. Berkowitz BA, Spector S. The effect of caffeine and theophylline on the
disposition of brain serotonin in the rat. Eur J Pharmacol 16: 322–325,
1971. doi:10.1016/0014-2999(71)90034-3.
9. Bishop D, Bonetti D, Dawson B. The influence of pacing strategy on
VO2 and supramaximal kayak performance. Med Sci Sports Exerc 34:
1041–1047, 2002. doi:10.1097/00005768-200206000-00022.
1500 CAFFEINE INCREASES PERIPHERAL FATIGUE ACCUMULATION
10. Black MI, Jones AM, Bailey SJ, Vanhatalo A. Self-pacing increases
critical power and improves performance during severe-intensity exercise.
Appl Physiol Nutr Metab 40: 662–670, 2015. doi:10.1139/apnm-2014-
0442.
11. Black MI, Jones AM, Blackwell JR, Bailey SJ, Wylie LJ, McDonagh
ST, Thompson C, Kelly J, Sumners P, Mileva KN, Bowtell JL,
Vanhatalo A. Muscle metabolic and neuromuscular determinants of
fatigue during cycling in different exercise intensity domains. J Appl
Physiol (1985) 122: 446 –459, 2017. doi:10.1152/japplphysiol.00942.
2016.
12. Blain GM, Mangum TS, Sidhu SK, Weavil JC, Hureau TJ, Jessop JE,
Bledsoe AD, Richardson RS, Amann M. Group III/IV muscle afferents
limit the intramuscular metabolic perturbation during whole body exercise
in humans. J Physiol 594: 5303–5315, 2016. doi:10.1113/JP272283.
13. Borg GA. Psychophysical bases of perceived exertion. Med Sci Sports
Exerc 14: 377–381, 1982. doi:10.1249/00005768-198205000-00012.
14. Boyett JC, Giersch GE, Womack CJ, Saunders MJ, Hughey CA,
Daley HM, Luden ND. Time of day and training status both impact the
efficacy of caffeine for short duration cycling performance. Nutrients 8:
E639, 2016. doi:10.3390/nu8100639.
15. Broxterman RM, Craig JC, Smith JR, Wilcox SL, Jia C, Warren S,
Barstow TJ. Influence of blood flow occlusion on the development of
peripheral and central fatigue during small muscle mass handgrip exercise.
J Physiol 593: 4043–4054, 2015. doi:10.1113/JP270424.
16. Broxterman RM, Layec G, Hureau TJ, Amann M, Richardson RS.
Skeletal muscle bioenergetics during all-out exercise: mechanistic insight
into the oxygen uptake slow component and neuromuscular fatigue. J Appl
Physiol (1985) 122: 1208 –1217, 2017. doi:10.1152/japplphysiol.01093.
2016.
17. Broxterman RM, Skiba PF, Craig JC, Wilcox SL, Ade CJ, Barstow
TJ. W= expenditure and reconstitution during severe intensity constant
power exercise: mechanistic insight into the determinants of W=. Physiol
Rep 4: e12856, 2016. doi:10.14814/phy2.12856.
18. Cannon DT, Bimson WE, Hampson SA, Bowen TS, Murgatroyd SR,
Marwood S, Kemp GJ, Rossiter HB. Skeletal muscle ATP turnover by
31P magnetic resonance spectroscopy during moderate and heavy bilateral
knee extension. J Physiol 592: 5287–5300, 2014. doi:10.1113/jphysiol.
2014.279174.
19. 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 116: 1693–1702, 2016. doi:10.1007/s00421-016-3423-x.
20. Chidnok W, Dimenna FJ, Bailey SJ, Wilkerson DP, Vanhatalo A,
Jones AM. Effects of pacing strategy on work done above critical power
during high-intensity exercise. Med Sci Sports Exerc 45: 1377–1385,
2013. doi:10.1249/MSS.0b013e3182860325.
21. Chidnok W, Fulford J, Bailey SJ, Dimenna FJ, Skiba PF, Vanhatalo
A, Jones AM. Muscle metabolic determinants of exercise tolerance
following exhaustion: relationship to the “critical power”. J Appl Physiol
(1985) 115: 243–250, 2013. doi:10.1152/japplphysiol.00334.2013.
22. Clausen T. Na⫹-K⫹ pump regulation and skeletal muscle contractility.
Physiol Rev 83: 1269 –1324, 2003. doi:10.1152/physrev.00011.2003.
23. Davis JK, Green JM. Caffeine and anaerobic performance: ergogenic
value and mechanisms of action. Sports Med 39: 813–832, 2009. doi:10.
2165/11317770-000000000-00000.
24. de Morree HM, Klein C, Marcora SM. Cortical substrates of the effects
of caffeine and time-on-task on perception of effort. J Appl Physiol (1985)
117: 1514 –1523, 2014. doi:10.1152/japplphysiol.00898.2013.
25. Doherty M, Smith P, Hughes M, Davison R. Caffeine lowers perceptual
response and increases power output during high-intensity cycling. J
Sports Sci 22: 637–643, 2004. doi:10.1080/02640410310001655741.
26. Doherty M, Smith PM. Effects of caffeine ingestion on rating of
perceived exertion during and after exercise: a meta-analysis. Scand J Med
Sci Sports 15: 69 –78, 2005. doi:10.1111/j.1600-0838.2005.00445.x.
27. Ducrocq GP, Hureau TJ, Meste O, Blain GM. Increased fatigue
response to augmented deceptive feedback during cycling time trial.
Med Sci Sports Exerc 49: 1541–1551, 2017. doi:10.1249/MSS.
0000000000001272.
28. Fernandes AL, Lopes-Silva JP, Bertuzzi R, Casarini DE, Arita DY,
Bishop DJ, Lima-Silva AE. Effect of time of day on performance,
hormonal and metabolic response during a 1000-M cycling time trial.
PLoS One 9: e109954, 2014. doi:10.1371/journal.pone.0109954.
29. Foster C, Snyder AC, Thompson NN, Green MA, Foley M, Schrager
M. Effect of pacing strategy on cycle time trial performance. Med Sci
J Appl Physiol • doi:10.1152/japplphysiol.00930.2017 • www.jappl.org
Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (150.161.081.015) on June 28, 2018.
Copyright © 2018 American Physiological Society. All rights reserved.
Sports Exerc 25: 383–388, 1993. doi:10.1249/00005768-199303000-
00014.
30. Fredholm BB. Astra Award Lecture. Adenosine, adenosine receptors and
the actions of caffeine. Pharmacol Toxicol 76: 93–101, 1995. doi:10.1111/
j.1600-0773.1995.tb00111.x.
31. Fredholm BB, Bättig K, Holmén J, Nehlig A, Zvartau EE. Actions of
caffeine in the brain with special reference to factors that contribute to its
widespread use. Pharmacol Rev 51: 83–133, 1999.
32. Froyd C, Beltrami FG, Millet GY, Noakes TD. Central regulation and
neuromuscular fatigue during exercise of different durations. Med Sci Sports
Exerc 48: 1024 –1032, 2016. doi:10.1249/MSS.0000000000000867.
33. Froyd C, Millet GY, Noakes TD. The development of peripheral fatigue
and short-term recovery during self-paced high-intensity exercise. J
Physiol 591: 1339 –1346, 2013. doi:10.1113/jphysiol.2012.245316.
34. Gaesser GA, Wilson LA. Effects of continuous and interval training on
the parameters of the power-endurance time relationship for high-intensity
exercise. Int J Sports Med 9: 417–421, 1988. doi:10.1055/s-2007-
1025043.
35. Gruet M, Temesi J, Rupp T, Levy P, Verges S, Millet GY. Dynamics
of corticospinal changes during and after high-intensity quadriceps exer-
cise. Exp Physiol 99: 1053–1064, 2014. doi:10.1113/expphysiol.2014.
078840.
36. Hermens HJ, Freriks B, Disselhorst-Klug C, Rau G. Development
of recommendations for SEMG sensors and sensor placement proce-
dures. J Electromyogr Kinesiol 10: 361–374, 2000. doi:10.1016/
S1050-6411(00)00027-4.
37. Hettinga FJ, De Koning JJ, Broersen FT, Van Geffen P, Foster C.
Pacing strategy and the occurrence of fatigue in 4000-m cycling time
trials. Med Sci Sports Exerc 38: 1484 –1491, 2006. doi:10.1249/01.mss.
0000228956.75344.91.
38. Hicks A, McComas AJ. Increased sodium pump activity following
repetitive stimulation of rat soleus muscles. J Physiol 414: 337–349, 1989.
doi:10.1113/jphysiol.1989.sp017691.
39. Hogan MC, Richardson RS, Haseler LJ. Human muscle performance
and PCr hydrolysis with varied inspired oxygen fractions: a 31P-MRS
study. J Appl Physiol (1985) 86: 1367–1373, 1999. doi:10.1152/jappl.
1999.86.4.1367.
40. Howley ET, Bassett DR Jr, Welch HG. Criteria for maximal oxygen
uptake: review and commentary. Med Sci Sports Exerc 27: 1292–1301,
1995. doi:10.1249/00005768-199509000-00009.
41. Hureau TJ, Romer LM, Amann M. The ‘sensory tolerance limit’: a
hypothetical construct determining exercise performance? Eur J Sport Sci
18: 1–12, 2016. doi:10.1080/17461391.2016.1252428.
42. Irwin C, Desbrow B, Ellis A, O’Keeffe B, Grant G, Leveritt M.
Caffeine withdrawal and high-intensity endurance cycling performance. J
Sports Sci 29: 509 –515, 2011. doi:10.1080/02640414.2010.541480.
43. Jackson AS, Pollock ML. Generalized equations for predicting body
density of men. Br J Nutr 40: 497–504, 1978. doi:10.1079/BJN19780152.
44. Jones AM, Wilkerson DP, Burnley M, Koppo K. Prior heavy exercise
enhances performance during subsequent perimaximal exercise. Med Sci
Sports Exerc 35: 2085–2092, 2003. doi:10.1249/01.MSS.0000099108.
55944.C4.
45. 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 294:
R585–R593, 2008. doi:10.1152/ajpregu.00731.2007.
46. Kalmar JM, Cafarelli E. Central fatigue and transcranial magnetic
stimulation: effect of caffeine and the confound of peripheral transmission
failure. J Neurosci Methods 138: 15–26, 2004. doi:10.1016/j.jneumeth.
2004.03.006.
47. Magkos F, Kavouras SA. Caffeine use in sports, pharmacokinetics in
man, and cellular mechanisms of action. Crit Rev Food Sci Nutr 45:
535–562, 2005. doi:10.1080/1040-830491379245.
48. Merton PA. Voluntary strength and fatigue. J Physiol 123: 553–564,
1954. doi:10.1113/jphysiol.1954.sp005070.
49. Meyers BM, Cafarelli E. Caffeine increases time to fatigue by maintain-
ing force and not by altering firing rates during submaximal isometric
contractions. J Appl Physiol (1985) 99: 1056 –1063, 2005. doi:10.1152/
japplphysiol.00937.2004.
50. Millet GY, Bachasson D, Temesi J, Wuyam B, Féasson L, Vergès S,
Lévy P. Potential interests and limits of magnetic and electrical stimula-
tion techniques to assess neuromuscular fatigue. Neuromuscul Disord 22,
Suppl 3: S181–S186, 2012. doi:10.1016/j.nmd.2012.10.007.
 CAFFEINE INCREASES PERIPHERAL FATIGUE ACCUMULATION
51. Mohr M, Nielsen JJ, Bangsbo J. Caffeine intake improves intense
intermittent exercise performance and reduces muscle interstitial potas-
sium accumulation. J Appl Physiol (1985) 111: 1372–1379, 2011. doi:10.
1152/japplphysiol.01028.2010.
52. Moritani T, Nagata A, deVries HA, Muro M. Critical power as a
measure of physical work capacity and anaerobic threshold. Ergonomics
24: 339 –350, 1981. doi:10.1080/00140138108924856.
53. Morton RH. The critical power and related whole-body bioenergetic
models. Eur J Appl Physiol 96: 339 –354, 2006. doi:10.1007/s00421-005-
0088-2.
54. Rodriguez-Falces J, Place N. Differences in the recruitment curves
obtained with monopolar and bipolar electrode configurations in the
quadriceps femoris. Muscle Nerve 54: 118 –131, 2016. doi:10.1002/mus.
25006.
55. Santos RA, Kiss MA, Silva-Cavalcante MD, Correia-Oliveira CR,
Bertuzzi R, Bishop DJ, Lima-Silva AE. Caffeine alters anaerobic dis-
tribution and pacing during a 4000-m cycling time trial. PLoS One 8:
e75399, 2013. doi:10.1371/journal.pone.0075399.
56. Silva-Cavalcante MD, Correia-Oliveira CR, Santos RA, Lopes-Silva
JP, Lima HM, Bertuzzi R, Duarte M, Bishop DJ, Lima-Silva AE.
Caffeine increases anaerobic work and restores cycling performance
following a protocol designed to lower endogenous carbohydrate avail-
ability. PLoS One 8: e72025, 2013. doi:10.1371/journal.pone.0072025.
57. Silveira R, Andrade-Souza VA, Arcoverde L, Tomazini F, Sansonio A,
Bishop DJ, Bertuzzi R, Lima-Silva AE. Caffeine increases work done
above critical power, but not anaerobic work. Med Sci Sports Exerc 50:
131–140, 2018. doi:10.1249/MSS.0000000000001408.
58. Simmonds MJ, Minahan CL, Sabapathy S. Caffeine improves supra-
maximal cycling but not the rate of anaerobic energy release. Eur J Appl
Physiol 109: 287–295, 2010. doi:10.1007/s00421-009-1351-8.
59. Siri WE. Body composition from fluid spaces and density: analysis of
methods. 1961. Nutrition 9: 480 –491, 1961.
60. Skiba PF, Jackman S, Clarke D, Vanhatalo A, Jones AM. Effect of
work and recovery durations on W= reconstitution during intermittent
exercise. Med Sci Sports Exerc 46: 1433–1440, 2014. doi:10.1249/MSS.
0000000000000226.
J Appl Physiol • doi:10.1152/japplphysiol.00930.2017 • www.jappl.org
Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (150.161.081.015) on June 28, 2018.
Copyright © 2018 American Physiological Society. All rights reserved.
1501
61. Skurvydas A, Brazaitis M, Kamandulis S, Sipaviciene S. Peripheral and
central fatigue after muscle-damaging exercise is muscle length dependent
and inversely related. J Electromyogr Kinesiol 20: 655–660, 2010. doi:
10.1016/j.jelekin.2010.02.009.
62. Tallis J, James RS, Cox VM, Duncan MJ. The effect of a physiological
concentration of caffeine on the endurance of maximally and submaxi-
mally stimulated mouse soleus muscle. J Physiol Sci 63: 125–132, 2013.
doi:10.1007/s12576-012-0247-2.
63. Tarnopolsky M, Cupido C. Caffeine potentiates low frequency skeletal
muscle force in habitual and nonhabitual caffeine consumers. J Appl
Physiol (1985) 89: 1719 –1724, 2000. doi:10.1152/jappl.2000.89.5.1719.
64. Temesi J, Mattioni Maturana F, Peyrard A, Piucco T, Murias JM,
Millet GY. The relationship between oxygen uptake kinetics and neuro-
muscular fatigue in high-intensity cycling exercise. Eur J Appl Physiol
117: 969 –978, 2017. doi:10.1007/s00421-017-3585-1.
65. 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 47: 537–546, 2015. doi:10.1249/
MSS.0000000000000448.
66. Vanhatalo A, Black MI, DiMenna FJ, Blackwell JR, Schmidt JF,
Thompson C, Wylie LJ, Mohr M, Bangsbo J, Krustrup P, Jones AM.
The mechanistic bases of the power-time relationship: muscle metabolic
responses and relationships to muscle fibre type. J Physiol 594: 4407–
4423, 2016. doi:10.1113/JP271879.
67. Vanhatalo A, Doust JH, Burnley M. A 3-min all-out cycling test is
sensitive to a change in critical power. Med Sci Sports Exerc 40: 1693–
1699, 2008. doi:10.1249/MSS.0b013e318177871a.
68. Vanhatalo A, Fulford J, DiMenna FJ, Jones AM. Influence of hyper-
oxia on muscle metabolic responses and the power-duration relationship
during severe-intensity exercise in humans: a 31P magnetic resonance
spectroscopy study. Exp Physiol 95: 528 –540, 2010. doi:10.1113/
expphysiol.2009.050500.
69. Walton C, Kalmar J, Cafarelli E. Caffeine increases spinal excitability
in humans. Muscle Nerve 28: 359 –364, 2003. doi:10.1002/mus.10457.
70. Walton C, Kalmar JM, Cafarelli E. Effect of caffeine on self-sustained
firing in human motor units. J Physiol 545: 671–679, 2002. doi:10.1113/
jphysiol.2002.025064.