Substrate utilization during exercise is influenced by several factors including the relative intensity and duration of exercise, an individual's training status and the effect of the preceding diet on both the substrate pool and the prevailing hormonal milieu (Hawley et al., 2015). As exercise intensity increases, there is a greater reliance on carbohydrate (CHO) based fuels (i.e., muscle and liver glycogen, blood glucose, lactate) and a reduction in the utilization of fat substrates (Brooks and Mercier, 1994). Ketone bodies provide another potential source of readily oxidized fuel for skeletal muscle, but are predominately associated with conditions of metabolic stress such as starvation, where they are needed to preserve essential function of peripheral tissues including the brain and heart (Robinson and Williamson, 1980; Veech, 2004). However, there has been recent interest in the notion that increasing blood ketone concentrations could contribute to an enhancement of exercise performance by providing a readily available, alternative oxidative substrate for working muscle, sparing the limited stores of muscle glycogen (Pinckaers et al., 2017). As such, models of hyperketonemia from endogenous origin such as chronic exposure to a ketogenic diet (Burke et al., 2017) as well as introduction of exogenous sources of ketone bodies from ketone ester supplements (Cox et al., 2016) have been investigated. Nevertheless, there is some concern that the presence of high circulating concentrations of ketone bodies could inhibit the flux of other muscle substrates, either by impairing (rather than “sparing”) muscle CHO oxidation and/or inhibiting adipose tissue lipolysis (Evans et al., 2017). As such, the situations in which an available ketone supply may benefit exercise capacity or performance may be determined by the duration and/or intensity of exercise and the need for combinations of muscle substrate to meet the metabolic demands.

Ketone bodies, namely D-β-hydroxybutyrate (βHB), acetone and acetoacetate (AcAc), are produced in the liver mitochondria from acetyl-CoA in response to an increased mobilization of free fatty acids (FFA) from adipose tissue in situations of reduced CHO availability (Robinson and Williamson, 1980). As summarized in recent reviews (Egan and D'Agostino, 2016; Pinckaers et al., 2017), under conditions of high CHO availability, circulating concentrations of ketone bodies are low, but are slightly elevated (0.1–0.5 mmol/L) by an overnight fast and further raised by exercising in a fasted state (0.5–1.0 mmol/L). Prolonged fasting/starvation (5 days) causes a maximal increase in rates of ketone body production (1–2 mmol/min or 140–280 g/day), leading to increased plasma concentrations that plateau under normal physiological conditions at ~7–10 mmol/L. Meanwhile, chronic exposure to a ketogenic diet [low CHO (<50 g/d), low-moderate protein (~15% of energy), high fat (75–80% of energy)] raises plasma ketone bodies to 1–2 mmol/L after several days, with concentrations reaching the apparent plateau achieved by prolonged fasting, according to the level of CHO restriction and duration of “keto-adaptation” (Pinckaers et al., 2017). Exogenous forms of ketone bodies include ketone salts, and more recently, ketone esters. Ingestion of the former appears to be less effective in increasing circulating ketone body concentrations and carries a significant salt load (Balasse and Ooms, 1968). Recently, a newly produced ketone monoester, R-3-hydroxybutyl R-3-hydroxybutyrate (Clarke et al., 2012), increases in plasma ketone concentrations (3–6 mmol/L) within the hours following its ingestion (400–600 mg/kg BM), although this may be altered by concomitant intake of food (Evans et al., 2017; Pinckaers et al., 2017).

To investigate the potential benefits to metabolism and sports performance, Cox et al. (2016) studied the effects of ingesting either CHO or CHO plus ketone ester (573 mg·kg⁻¹ BM) on performance in trained cyclists. Their ingestion protocol induced a higher blood D-βHB concentration during submaximal cycling (ranging between ~1.5 and 3 mmol/L) and lead to a subsequent improvement in time-trial (TT) performance by ~2% following ketone ester and CHO ingestion compared to the ingestion of only CHO. However, aspects of the study design are inconsistent with conditions of “real world” cycling competition. Accordingly, we examined the effect of pre-“race” ingestion of a ketone diester on blood ketone body concentrations, substrate metabolism and performance under conditions of elite professional cycling; ingestion of a pre-race CHO-rich meal, inclusion of a warm-up, involvement of world-class cyclists and simulation of a real-life course. We hypothesized that this protocol would result in acute nutritional ketosis but that no performance improvement would be observed due to the high intensity nature of a real-life TT event which is dependent on the high rates of energy production from the oxidation of CHO-based fuels.

This is the first study to report the effect of pre-exercise supplementation with a ketone diester on the performance of a cycling TT under conditions simulating real-life competition: laboratory simulation of a World Championship TT course in world-class male road cyclists who undertook nutritional strategies mimicking competition practices with respect to CHO and caffeine supplementation. Although, our protocol achieved hyperketonemia, as evidenced by increases in serum βHB and AcAc concentrations, there was an impairment of TT cycling performance in these elite cyclists. This outcome appears to be linked to the general observation of gut discomfort and intolerance among the study participants, with symptoms ranging from mild to severe. Our investigation adds important information to sports nutrition, by adding a real-world element.

The primary aim of our study was to address recent reports of enhanced sports performance associated with an acute increase in blood ketone concentrations following the intake of a ketone ester drink (Cox et al., 2016), by re-examining this concept in a more ecologically valid protocol. The investigation was undertaken as a collaborative project with a World Tour professional team of the International Cycling Union (UCI), offering an opportunity for them to make an evidence-based decision regarding the potential use of a highly discussed performance aid (Abraham, 2015). A number of features were included in the study design to optimize the reliability and validity of data, including opportunities to mimic the conditions under which the performance aid (ketone ester supplement) would be used. Team sports scientists were able to provide a simulation of the profile of the 2017 World Championships TT course (Bergen, Norway) on a cycle ergometer. Furthermore, the world-class cyclists from the team who participated in the study were personally motivated to receive individual and group results, and were highly experienced in TT cycling as well as familiarized to the specific laboratory-based course simulation. Race nutrition strategies (24 h prior, pre-race and during race) were standardized and made realistic by involving meal preparation by the team chef to suit both sports nutrition guidelines (Thomas et al., 2016) and the cultural practices of the riders. This included attention to achieve adequate CHO availability in pre-race meals (as opposed to the overnight fasted state involved in other studies such as Balasse, 1979; Fery and Balasse, 1986; Cox et al., 2016), as well as the real-life intake of caffeine and CHO supplements during the pre-race and within-race practices.

Our primary finding of a 2.0 ± 1% (58 s) longer time to completion in the TT following ketone ingestion supports our initial hypothesis that ketone ingestion would not enhance TT performance (Figure 2). Although, the gut disturbances were the likely cause of the performance impairment, we note that none of the cyclists achieved a faster time in the ketone trial, even when they reported very minor symptoms. Furthermore, the RPE in the ketone trial were similar to those on the placebo trial despite a lower power output and heart rate. This suggests that the gastrointestinal discomfort and/or some direct effect of ketones on the brain increased the perception of effort, and in accordance with the psychobiological model of pacing (Pageaux et al., 2014), our highly experienced cyclists reduced their work output to enable the TT to be completed without premature exhaustion. This finding warrants further investigation to confirm and explore the mechanisms. In the meantime, we note that the outcome of impaired performance with ketone ingestion is in contrast to previous studies (Clarke and Cox, 2015; Cox et al., 2016). Indeed, Clarke and Cox (2015) and Cox et al. (2016) reported 1–2% improvements in 30 min rowing performance and 30 min TT performance, respectively, following ingestion of a similar ketone dose to the current study (573 vs. 500 mg/kg BM, respectively) in combination with CHO, compared to CHO alone.

Due to the lack of a commercial supply, we were unable to obtain the ketone monoester supplement used in the study of Cox et al. (2016). However, we were able to source a diester that is currently being investigated as a potential treatment for seizures resulting from central nervous system oxygen toxicity (D'Agostino et al., 2013) and used in a similar dose to Cox et al. (2016). This diester contains a racemic mixture of βHB (i.e., contains both D- and L- enantiomers of the βHB) and has the ability to elevate both βHB and AcAc in a 1:1 ratio. The use of enzymatic analysis in the current study measures only the D- enantiomer which is the main circulating form of βHB and the most likely to have a direct effect on substrate metabolism and skeletal muscle responses (Yamada et al., 2010).

We provided the ketone drink in two doses, with the first bolus ingested 70 min prior to the TT. Based on previous research, our aim was to reach peak βHB concentrations at ~1 h following ingestion (i.e., prior to the TT; Pinckaers et al., 2017). We measured a modest increase in serum βHB concentrations in the ketone trial, reaching >0.3 mmol/L following the warm-up, but to our surprise serum βHB concentrations peaked immediately following the TT (>0.4 mmol/L; Figure 4B). The capillary whole blood samples analyzed for βHB concentrations measured values 2- to 3-fold greater than the serum samples (Figure 5B). This variation in D-βHB concentrations via enzymatic analysis (serum) and whole blood is consistent with previous literature reporting a –0.5 to 0.6 mmol/L higher concentration with handheld monitors (Pineda and Cardoso, 2015). This variation in measuring blood ketones in a controlled laboratory setting highlights the challenges athletes face in the field when aiming to reach and stay within the “optimal” range for a performance benefit (Egan and D'Agostino, 2016). We also measured a peak in serum AcAc concentrations following the warm-up, reaching ~0.5 mmol/L (Figure 4C). Therefore, when we consider total circulating ketones measured (i.e., βHB and AcAc) and the L- enantiomer that has not been measured, it is likely that athletes would be in the “optimal” range of 1–3 mmol/L for a proposed performance benefit (Egan and D'Agostino, 2016). Although “nutritional ketosis” was achieved, βHB concentrations reported in the current study are much lower than those reported previously (Cox et al., 2016). Cox et al. (2016) reported an increased in βHB concentrations to ~2 mmol/L within 20 min of ketone ester ingestion when co-ingested with CHO or ~4 mmol/L when ingested alone. This variation in serum βHB is likely explained by a range of factors including the different ketone esters used, the elite training status of cyclists in the current study and the different pre-ingestion nutritional strategies were the current study focused on appropriate race preparation practices.

Although we have not measured the same increase in circulating ketone concentrations as (Cox et al., 2016), ketones increased appropriately to alter metabolic responses compared to when a placebo was ingested. Blood glucose concentration was lowered in the ketone trial by ~1 mmol/L within 30 min following ingestion of the first ketone dose, and following the TT blood glucose was ~2 mmol/L lower than in the placebo trial. Additionally, we reported a 4.5 mmol/L (35%) reduction in blood lactate concentration following the TT in the ketone trial compared to the placebo trial. These findings of reduced blood glucose and blood lactate concentrations are consistent with the data of Cox et al. (2016) during 60 min of exercise at 75% Wmax and following a 30 min TT, respectively. We also measured lower circulating FFA during the TT following ketone ingestion compared to a placebo. Participants were cycling at 340–350 W during the TT and thus estimated contribution of FFA oxidation to total energy expenditure would likely be low as at this intensity, the muscle relies predominately on CHO-based fuels (Hawley and Leckey, 2015). However, the difference in FFA concentration between the ketone and placebo trials could be related to circulating ketone bodies having the ability to suppress lipolysis via inhibition of catecholamine's (Bjorntorp and Schersten, 1967).

As ketone bodies can be readily oxidized by skeletal muscle, expired gas was collected during the incremental warm-up. No differences in RER were measured between the ketone and placebo trials, although this could be related to the absolute exercise intensities attained by our elite subjects (155, 310, 355 W). Alternatively the high ketone concentrations in the urine suggest that the ketones are not being oxidized at the skeletal muscle. We have not estimated rates of substrate oxidation rates due to βHB and AcAc yielding respiratory exchange quotient values of 0.89 and 1.00, respectively and thus without appropriate correction factors for CO₂ displacement and urine volume this would lead to an inaccurate representation of substrate utilization (Frayn, 1983; Pinckaers et al., 2017). We also reported a modest reduction in heart rate during the TT in the ketone trial compared to the placebo trial (5 bpm) which may be associated with a slightly reduced average power output and in the ketone trial.

Of the 10 participants who completed the trials, all reported gastrointestinal discomfort associated with the intake of the ketone ester including dry retching, mild to moderate nausea, reflux and minor discomfort. Furthermore, one participant was unable to start the TT due to prolonged vomiting and dizziness. This participant also experienced the highest concentrations of serum AcAc concentration when compared to the other 10 participants (Figure 4C, participant X), suggesting bioavailability of the ketone diester may impact individual responses following ketone ingestion. Although the side effects of ketone esters are not frequently discussed in the literature, Clarke et al. (2012) has provided evidence that participants have experienced a range of adverse effects including vomiting, nausea, diarrhea, and abdominal pain. These side effects have been associated with high dose of ketone ester and the consumption of the ester with a milk-based drink (Clarke and Cox, 2015). While it is possible that different dosing strategies, or the use of a different ketone ester product might eliminate or greatly reduce the gut problems seen in the current study, it is unclear whether a performance enhancement could be expected with exogenous ketone use in sporting events undertaken under the conditions employed in our study.

In conclusion, the results of the current study show that ingestion of a 1,3-butanediol acetoacetate diester under conditions of optimal race nutrition (i.e., CHO fed) results in increases in βHB and AcAc concentrations. The diester was associated with gut discomfort and intolerance among the cyclists with symptoms ranging from mild to severe. Despite optimal nutritional support (i.e., CHO breakfast, feeding during the TT and caffeine ingestion) for performance, ketone ingestion was associated with an increase in perception of effort, leading to an impairment of TT performance in elite professional cyclists.

JL, MR, MQ, and JH: Conception and design, Collection and assembly of data, Data analysis and interpretation, Manuscript writing, Final approval of manuscript (required). LB: Conception and design, Financial support, Collection and assembly of data, Data analysis and interpretation, Manuscript writing, Final approval of manuscript (required).