Does caffeine supplementation alter energy contribution during a work-based ~30 min cycling time-trial?

Caffeine has been shown to increase anaerobic energy contribution during short-duration cycling time-trials (TT) though no information exists on whether caffeine alters energy contribution during more prolonged, aerobic type TTs. The aim of this study was to determine the effects of caffeine supplementation on longer and predominantly aerobic exercise. Fifteen recreationally-trained male cyclists (age 38±8 y, height 1.76±0.07 m, body mass 72.9±7.7 kg) performed a ~30 min cycling TT following either 6 mg·kg-1BM caffeine (CAF) or placebo (PLA) supplementation, and one control (CON) session without supplementation, in a double--blind, randomised, counterbalance and cross-over design. Mean power output (MPO) was recorded as the outcome measure. Respiratory values were measured throughout exercise for the determination of energy system contribution. Data were analysed using mixed-models. CAF improved mean MPO compared to CON (P=0.01), and a trend towards an improvement compared to PLA (P=0.07); there was no difference in MPO at any timepoint throughout the exercise between conditions. There was a main effect of Condition (P=0.04) and Time (P<0.0001) on blood lactate concentration, which tended to be higher in CAF vs. both PLA and CON (Condition effect, both P=0.07). Ratings of perceived exertion increased over time (P<0.0001), with no effect of Condition or interaction (both P>0.05). Glycolytic energy contribution was increased in CAF compared to CON and PLA (both P<0.05), but not aerobic or ATP-CP (both P>0.05). CAF improved aerobic TT performance compared to CON, which could be explained by increased glycolytic energy contribution.


Introduction
Does caffeine supplementation alter energy contribution during a work-based ~30 min cycling time-trial?
Th e rate of energy supply in a given TT will mainly depend on its intensity and duration. Shorter duration TTs (i.e., 1km; ~1 min) are considered high-intensity and predominantly limited by peripheral fatigue, and anaerobic energy distribution is considered a limiting factor for performance in such tasks 1 . On the other hand, middle-to longerduration TT's (i.e., >4 km) are predominantly supplied by aerobic sources, although an increased anaerobic power output contribution may be expected during specifi c moments, such as uphill climbs and the dispute of positions at the end of the TT 2 . In fact, it has been known for some time that anaerobic capacity moderately correlates to performance during middle-distance cycling TT (r=-0.50) 3 . Th us, any nutritional intervention employed to improve anaerobic energy supply, even during cycling events that are highly aerobic in nature, could potentially improve cycling TT performance.
Caff eine is a popular ergogenic aid that has purported ergogenic properties that can impact upon both peripheral and central fatigue. Indeed, caff eine can directly aff ect the skeletal muscle by increasing the activity of phosphofructokinase, subsequently increasing anaerobic glycolysis 4 . In addition, caff eine administration has also been speculated to promote calcium release from the sarcoplasmic reticulum ryanodine receptor, thereby improving excitation-contraction coupling, as shown by increased force during tetanic stimulation 5 . Caff eine can also promote reduced levels of potassium in the muscle cell interstitium, leading to improved stimulation of the sodiumpotassium pump 6 . Caff eine may also act on the central nervous system as an adenosine A1 and A2A receptor antagonist, leading to an increase in motivational drive and neuromuscular excitability, which may result in a decreased perception of eff ort (i.e., ratings of perceived exertion; RPE) and pain for a given workload 7 .
Th e wide-ranging physiological eff ect of caff eine suggests it can have ergogenic properties across a variety of exercise intensities and durations. Indeed, caff eine has been shown to improve cycling TT performance of 1-8 , 4-9 , 10-10 and 40-11 km. Santos et al. 9 showed an improved 4-km TT performance with caff eine, although overall aerobic and anaerobic Methods Participants contribution was unchanged. This was despite greater anaerobic contribution in the middle section of the test with caff eine. Nonetheless, changes in mean power output (MPO) mirrored changes in anaerobic contribution, supporting the notion that power distribution throughout a TT is regulated primarily by changes in anaerobic contribution 3 . It is unclear whether longer duration exercise tasks with a lower contribution from anaerobic energy sources 12 are improved with caff eine due to modifi cations in anaerobic energy contribution and subsequent power distribution.
The aim of this study was to determine the effect of caffeine supplementation on physiological and performance variables during a prolonged work-based simulated ~30 min cycling TT. We hypothesised that caffeine would improve exercise performance due to an increased anaerobic energy contribution.
Fifteen recreationally-trained 13 male cyclists (TABLE 1) volunteered and gave their written informed consent to participate in this study. All were competitive cyclists with over one year of training experience. Participants were required not to have taken any supplements in the past 6 months, except for carbohydrate and protein, or any previous use of anabolic steroids. Th e study was fi rst approved by the University of São Paulo´s Ethics Review Committee.

Experimental design
Participants attended the laboratory on a total of six separate occasions, separated by a minimum of 72 h, with all trials being performed at the same time of day to ensure results were not aff ected by circadian variation 14 . The first session was for the determination of VO 2max and W max using an incremental cycling test to exhaustion. Th e following two sessions were for familiarisation of the simulated TT protocol. Th e fi nal three sessions comprised of the main trials in which individuals performed the TT protocol following either caff eine (CAF) or placebo (PLA) supplementation, or a control (CON) trial.
Twenty-four hours prior to all trials, participants were required to refrain from alcohol, caff eine and any strenuous exercise. Food intake was monitored during the twenty-four hours prior to the fi rst main trial using a food diary and replicated prior to the remaining main trials. Th e food diaries were analysed by a nutritionist immediately prior to the experimental sessions to ensure that participants had not consumed any caff eine containing foods while energy and macronutrient intake was analysed at a later time by the same nutritionist using specifi c software (Avanutri online, Avanutri, Rio de Janeiro, Brazil).
The main trials followed a double-blind, randomised, counterbalanced and cross-over design. Participants arrived at the laboratory a minimum of 6 h post-prandial. One hour prior to exercise in CAF and PLA, participants ingested a capsule containing either 6 mg•kg -1 BM of caff eine or dextrose alongside 500 mL of water before remaining seated until the commencement of exercise. In CON, participants followed the same procedure although no capsule was ingested. Participants were allowed access to their phones or own reading material throughout this waiting period. Blinding occurred via an outside researcher who prepared each participant's supplements in identical opaque capsules. Participants were randomly assigned to each experimental condition using a Latin Square model 15 .

Incremental cycling capacity test
Participants' performed an incremental cycling capacity test to exhaustion on a cycle ergometer (Lode Excalibur, Germany) to determine individual VO 2max and W max . Individual set up of the cycle ergometer was determined prior to the maximal test, recorded electronically, and maintained for all subsequent trials. Participants performed four submaximal 4-min stages starting at 75 W; this was increased by 50 W each stage until 225 W. Th ereafter, workload increased by 30 W every minute until volitional exhaustion. Ventilatory and gas exchange measurements were recorded using a portable breath-by-breath system (K4 b 2 , Cosmed, Italy); the highest value averaged over a 30-s period during the test was defi ned as VO 2max . Individual W ma was calculated as the last completed stage plus the fraction of time spent in the fi nal non-completed stage multiplied by 30 W.

Cycling Time-Trial
A 5-min cycling warm up was performed at 125 W immediately followed by the individual TT. Individuals were required to complete a predetermined amount of work equivalent to 25 min at 85% of their individual W max in the fastest possible time 16 ; work to be completed was calculated according to the formula, "Total amount of work = 0.85 × W max × 1500 s". Th e cycle ergometer was set in linear mode, meaning work load was cadence dependent according to the formula, "W = α × (rpm) 2 ". Th e α value was based on each individual's W max so that individuals were working at 85% W max when cycling at a cadence of 95 rpm.
Participants were instructed to complete the exercise in the fastest possible time. No motivation or specifi c information was given to the individuals during the test although they were informed when they had completed 25%, 50%, 75% and 90% of the exercise. We have previously shown the test to have a coeffi cient of variation (CV) of 2.9% following one familiarisation trial (17); the CV ((standard deviation/mean) x 100)) between familiarisation trials in the current study was 2.7%.

Measurements
Minute ventilation (V E ), VO 2 , and respiratory exchange ratio (RER) were measured during all main experimental trials using a breath-bybreath portable gas analyser (K4 b 2 , Cosmed, Italy) that has previously been validated over a range of exercise intensities 18 . Th e gas analyser was calibrated according to the manufacturer's specifi cations prior to every test.
Finger-prick blood samples were taken at baseline; following completion of 25%, 50%, 75%, 90% and 100% of the test; and 3, 5 and 7 min post-exercise. A volume of 20 μL of blood was stored in the same volume of ice-cold 2% NaF solution. Th is was then centrifuged for 5 min at 1 g at 4°C and the resultant plasma was then stored at -80°C until analysis. Plasma lactate was determined spectrophotometrically using an enzymaticcolorimetric method (Biotecnica, MG, Brasil).
Heart rate was monitored consistently throughout exercise at a frequency of 5 Hz using a heart rate monitor with telemetry data transmission (Cosmed, Italy). Ratings of perceived exertion were recorded following 25%, 50%, 75%, 90% and 100% of the test using the 6-20 point Borg scale 19 .

Calculation of the energetic system contribution
Energy contribution was calculated according to the methods of Artioli et al. 20 . Th e net energy generated by aerobic metabolism was calculated by subtracting rest oxygen consumption from exercise oxygen consumption. Oxygen consumption was measured as the participants remained seated quietly on the ergometer for a 5-min period prior to the warm-up; the mean of the fi nal 30 s of this multiplied by the exercise duration was taken as resting oxygen consumption. Th e area under the curve (AUC) of oxygen consumption during the entire TT was calculated using the trapezoidal method. Subsequently, resting oxygen consumption was subtracted from exercise oxygen consumption.
Estimated energy cost of the ATP-CP system was calculated using the fast component of excess post-exercise oxygen consumption by subtracting resting oxygen consumption from the oxygen consumed during the recovery period. For the calculation of the contribution of the glycolytic system, it was assumed that 1 mmol·L -1 of lactate above resting values corresponded to 3 Th ere was no main eff ect of Condition on absolute aerobic energy contribution (P = 0.84) or on absolute ATP-CP contribution (P = 0.34). Th ere was a main eff ect of Condition on absolute glycolytic energy contribution (P = 0.02), with higher contribution shown in CAF vs. CON (TABLE 3). Th ere was a signifi cant main eff ect of Condition on the relative aerobic contribution (P = 0.01), with lower relative contribution in CAF than CON (P < 0.05). Similarly, there was a significant main effect of condition on the relative glycolytic contribution Results mL of oxygen consumed per kilogram of body mass. Delta plasma lactate (i.e., end plasma lactate minus resting plasma lactate), was thus multiplied by 3 and by the athlete's body mass. The obtained value of oxygen in millilitres was converted to litres and subsequently to energy (kJ), assuming that each 1 L of oxygen is equal to 20.92 kJ 20 .
The result obtained from each energy system was summed and the total energy expenditure calculated. Total metabolic work (TMW) was calculated as the sum of the energy systems according to previous methods 21-24 and based upon assumptions described by di Prampero and Ferretti 25 .

Statistical Analysis
All data were analysed using the SAS statistical package, (SAS 9.2, SAS Institute Inc., USA) and are presented as mean ± 1SD. Statistical signifi cance was accepted at P ≤ 0.05. Mixed-model analyses with repeated measures were used to compare the overall eff ect of caff eine on TT performance and energy contribution (aerobic, glycolytic, ATP-CP) with Condition (CON, PLA and CAF) used as a fixed factor and participants as a random factor. A mixed-model was used to analyse the effect of caffeine on RPE, where Condition (CON, PLA and CAF) and Time (0, 25, 50, 75, 90 and 100%) were considered fixed factors. Another mixed-model was used to determine the effect on lactate, with Trial (CON, PLA and CAF) and Time (0, 25, 50, 75, 90 and 100%, 3, 5 and 7 min post-exercise) considered fixed factors. Magnitude based inferences 26 were used to determine the practical significance of caffeine on TT performance (overall and throughout the exercise) using a spreadsheet to establish the likelihood of a meaningful effect on exercise capacity. The smallest worthwhile improvement in MPO was calculated using half the CV of the test 27, 28 from Oliveira et al. 17 . Qualitative descriptors were assigned to the positive percentile scores as follows: <1%, almost certainly not; 1-5%, very unlikely; 5-25%, unlikely; 25-75%, possibly; 75-95%, likely; 95-99%, very likely; >99%, almost certainly.
Th ere was a main eff ect of Time (P < 0.0001) on MPO throughout the TT, but no main eff ect of Condition or a Condition x Time interaction (P = 0.93); MPO was not diff erent between trials at any time point throughout exercise (FIGURE 1). Magnitude based inferences showed that MPO was possibly to likely improved in CAF throughout the TT, except during the fi nal 10% of the test (TABLE 2).
Th ere was a main eff ect of Condition (P = 0.04) and Time (P < 0.0001) on blood lactate concentration, although only a tendency towards a Condition x Time interaction (P = 0.07). Blood lactate tended to be higher in CAF vs. both PLA and CON (Condition eff ect, both P = 0.07), with no diff erences between trials at specifi c time points (FIGURE 2, PANEL A). RPE increased over time (P < 0.0001), although there was no eff ect of Condition (P = 0.80) or a Condition x Time interaction (P = 0.97) (FIGURE 2, PANEL B). (P = 0.01); relative glycolytic contribution was higher in CAF compared to both PLA and CON (both P < 0.05; TABLE 3). There were no differences in ATP-CP relative contribution (P > 0.05). There was no effect of Condition on TMW (P = 0.92).    Th e main fi ndings of this study showed that caff eine increased overall MPO during a simulated ~30 min TT. Th e improvement with caff eine could be partly explained by a parallel increase in glycolytic energy contribution.

Energy system contribution
Caffeine supplementation improved cycling performance compared to the control condition in which no supplement was ingested, but failed to reach signifi cance compared to the placebo trial. Th is was, perhaps, somewhat unexpected, although this could be partially explained by expectancy. When an individual receives a supplement, whether it contains the active ingredient or an inert substance, this can increase their expectancy of obtaining a positive improvement. A positive outcome stemming from the belief that a positive intervention has been received is called the placebo eff ect 29 . We have previously shown that expectancy associated with caff eine supplementation may modify the ergogenic eff ect of both the active intervention and the placebo session 30 . Other studies have also demonstrated the ergogenic potential of expectancy with caff eine supplementation 31 . Th us, the current data suggest that part of the ergogenic eff ect shown here with caff eine supplementation may be due to the placebo eff ect, although the placebo trial was not improved compared to the control. Although the current results showed that caff eine could improve ~30 min cycling TT performance via an increased overall MPO, there was no modifi cation in pacing strategy Discussion throughout the exercise. Th is is in contrast to previous research showing an altered pacing strategy with caff eine during a 4 km TT 9 ; individuals maintained a higher MPO in the middle of the test with caff eine. In the current study, no statistical diff erences were shown at any time point between trials, though it could be argued that routine analyses may not have been sensitive enough to detect small but potentially worthwhile diff erences. Closer inspection of pacing suggests that caff eine supplementation resulted in moderate increases in MPO throughout the test, except for the fi nal 10%. Indeed, magnitude based inferences suggest that the diff erences during these sections were possibly to likely meaningful, and it is possible that the cumulation of these increased power outputs over the diff erent time points led to the overall improvement with caff eine compared to the control session. Th us, it appears that caff eine resulted in small, but potentially worthwhile, increases in MPO throughout the TT.
As expected, the majority of energy provision came from aerobic sources during the ~30 min TT. Despite this, we showed that overall glycolytic energy contribution was higher with caff eine compared to the control trial; the relative contribution from glycolytic energy sources (to total energy expenditure) was higher with caff eine compared to both the control and placebo trial. Th ese results contrast with those of Santos et al. 9 , who showed no overall changes in anaerobic contribution with caff eine during a 4-km TT, although they did not separate glycolytic from ATP-CP contribution. Despite this, their results did show greater anaerobic contribution at several isolated time points throughout the TT. Unfortunately, in the current study we were unable to calculate energy contribution at diff erent time points throughout the TT, which could have provided more information regarding the exact moments at which individuals were able to maintain an increased glycolytic energy contribution. Nonetheless, our results suggest that the improved performance shown here with caff eine may be partly due to an increased glycolytic energy contribution meaning they could maintain a higher mean power output throughout exercise. Th is was mirrored by a trend towards increased blood lactate values with caff eine, also indicative of an increased glycolytic energy contribution. However, the absolute increases in glycolytic energy contribution were low, and it is known that the eff ect of caff eine is pleiotropic. Th erefore, it cannot be ruled out that the mechanism by which performance was improved was also due to caff eine's other mechanisms of action.
Our results reaffirm the sports-performance enhancing properties of caff eine, and yet, only 9 and 8 out of 15 individuals improved their performance (above the test CV) compared to control and placebo. Th e reasons for this inter-individual variation is not entirely clear but could be due to genetic variation in caff eine metabolism or diff erences in habitual caff eine consumption. Single nucleotide polymorphism of the CYP1A2 gene responsible for caff eine metabolism leads to fast (TT allele homozygotes and slow (C allele carriers) metabolizers 32 , although the infl uence of these polymorphisms on the response to acute caff eine supplementation is inconsistent [33][34][35] and beyond the scope of this study. Habitual consumption of caff eine has been suggested to modify the acute response to caff eine supplementation 36 , although more recent data from our lab suggests otherwise 37 . Diff erences may have been due to the relative doses given as Beaumont et al. 36 gave an acute dose equal to that of habitual consumption (0.3 g·kg -1 BM) whereas Gonçalves et al. 37 gave an acute dose in excess of average intake. Here our acute dose was well in excess of the habitual intake of our volunteers (6 vs. 2.5 mg·kg -1 BM) and thus cannot explain the lack of an eff ect in some individuals.
Th ere are two main limitations with the current investigation. Firstly, we were unable to measure caffeine concentration, although we are certain that the relatively high dose used here resulted in suffi cient increases to elicit an ergogenic eff ect 38 . A further limitation of this study is that we were unable to calculate energy systems contribution at various time points throughout the cycling TT. Th is is due to the method employed 20 and the variable lactate concentrations throughout the test. The method requires subsequent lactate values to be higher than the previous measure, to accurately determine energy contribution during that period. However, since pacing during cycling exhibits a "U"-pacing profi le 39 with an unsustainable high power output at the beginning, this meant that lactate concentration following the initial section of the TT was higher than the subsequent sections, hampering our calculations. Further research should use methods that allow quantifi cation of the contribution of all energy systems at time points throughout more prolonged cycling TT performance.
In summary, our results showed that caffeine can improve MPO during a ~30min TT, and an increased glycolytic energy contribution may have contributed to this improved performance. Despite an overall improved performance, pacing was not changed with caffeine. Our results highlight the importance of anaerobic metabolism even during predominantly aerobic based activity and demonstrate that the ergogenic eff ect of caff eine is via an increased contribution from glycolytic energy sources. Th ese data suggest that cyclists can expect to experience moderate but potentially worthwhile gains in performance with caff eine during aerobic cycling TTs.