Exogenous Ketosis Improves Sleep Efficiency and Counteracts the Decline in REM Sleep after Strenuous Exercise

ABSTRACT

Introduction

Available evidence indicates that ketone bodies may improve sleep quality. Therefore, we determined whether ketone ester (KE) intake could counteract sleep disruptions induced by strenuous exercise.

Methods

Ten well-trained cyclists with good sleep quality participated in a randomized crossover design consisting of two experimental sessions each involving a morning endurance training and an evening high-intensity interval training ending 1 h before sleep, after which polysomnography was performed overnight. Postexercise and 30 min before sleeping time, subjects received either 25 g of KE (EX_KE) or a placebo drink (EX_CON). A third session without exercise but with placebo supplements (R_CON) was added to evaluate the effect of exercise per se on sleep.

Results

Blood d-β-hydroxybutyrate concentrations transiently increased to ~3 mM postexercise and during the first part of the night in EX_KE but not in EX_CON or R_CON. Exercise significantly reduced rapid eye movement sleep by 26% (P = 0.001 vs R_CON) and increased wakefulness after sleep onset by 95% (P = 0.004 vs R_CON). Interestingly, KE improved sleep efficiency by 3% (P = 0.040 vs EX_CON) and counteracted the exercise-induced decrease in rapid eye movement sleep (P = 0.011 vs EX_CON) and the increase in wakefulness after sleep onset (P = 0.009 vs EX_CON). This was accompanied by a KE-induced increase in dopamine excretion (P = 0.033 vs EX_CON), which plays a pivotal role in sleep regulation. In addition, exercise increased sleep spindle density by 36% (P = 0.005 vs R_CON), suggesting an effect on neural plasticity processes during sleep.

Conclusions

These data indicate that KE ingestion improves sleep efficiency and quality after high-intensity exercise. We provide preliminary evidence that this might result from KE-induced increases in dopamine signaling.

Both acute and regular exercises exert positive effects on sleep quality and quantity, as evidenced by improved sleep onset, slow-wave sleep, and sleep efficiency (1). However, sleep disruptions also frequently occur in athletes, most often resulting in decreased rapid eye movement (REM) sleep (2,3) and increased wakefulness (3) during the subsequent night. This is particularly common before competition (4), during intensive training periods (5,6), after ultraendurance exercise (7), as well as after vigorous exercise ending in close proximity to bedtime (8,9). Sleep impairments do not only harm exercise performance (10) but also hamper postexercise recovery and increase illness and injury rates (11). This results in an overreliance on sleep medication by elite athletes (12), which is accompanied by major side effects (13). As such, there is a need for nonpharmacological interventions that can counteract exercise-induced sleep impairments in athletes.

In this perspective, we recently demonstrated that increasing blood ketone levels postexercise, and before sleep, via ingestion of a ketone ester (KE) drink limited the development of physiological disturbances (e.g., decreased resting, submaximal and maximal heart rate) and improved sustained training load during a 3-wk overload training period (14). Although sleep was not evaluated in this study, KE ingestion attenuated the overload training-induced increase in nocturnal urinary excretion of adrenaline and noradrenaline (14). Elevated nocturnal sympathetic and adrenergic activity are considered to be an important cause of reduced sleep quality in both athletes (15) and other individuals afflicted by chronic insomnia (16). Therefore, it is tempting to suggest that KE may have improved sleep during the training overload period resulting in improved recovery.

Such hypothesis is further supported by evidence that oxidation of the ketone bodies d-β-hydroxybutyrate (βHB) and acetoacetate (AcAc), which are continuously produced from fatty-acid-derived acetyl-CoA in hepatic mitochondria and astrocytes, supports brain metabolism (17). Furthermore, the contribution of ketone bodies to brain metabolism has been shown to depend on the level of neuronal activity. In rats under ketosis, ketone bodies supported ~60% of total neuronal oxidation during wakefulness, which further increases when progressing to a state of anesthesia and providing full support for neuronal oxidation during a state of isoelectricity (18).

Preliminary evidence also indicates that ketone bodies can directly modulate sleep architecture. Intracerebroventricular injection of AcAc, but not βHB, increased slow-wave activity during non-REM (NREM) sleep in a dose-dependent manner and slightly decreased the amount of REM sleep in mice (19). In contrast, another study observed that elevating blood ketone levels by means of a ketogenic diet increased REM sleep in children with epilepsy (20). However, because a ketogenic diet elicits a multitude of metabolic changes, it is unclear whether these effects are a direct result of ketosis per se. Taken together, these data suggest that ketone bodies have the potential to beneficially impact sleep quality and quantity.

Against this background, we aimed to investigate whether ketone bodies can counteract strenuous exercise-induced disruptions in sleep architecture in well-trained cyclists. Therefore, we simulated a strenuous training day, involving a morning endurance training session and a late-evening high-intensity interval training (HIIT) ending 1 h before sleeping time. After each training session and before sleeping time, the subjects received either a KE or a placebo drink, enabling us to investigate the specific effects of ketone bodies in the absence of other metabolic alterations induced by other ketotic interventions such as fasting or a ketogenic diet. We also added an experimental condition without exercise to investigate the effect of exercise per se on sleep architecture. Based on available literature data (2,3,9), we hypothesized that exercise would result in less REM sleep and an increase in wakefulness after sleep onset (WASO) and that these exercise-induced impairments can be counteracted by KE ingestion.

METHODS

Ethical Approval and Subjects

Before participation, subjects were examined by a qualified physician using a medical questionnaire and a resting electrocardiogram. People working in late-night shifts and extreme morning and evening chronotypes as determined by the Horne and Östberg questionnaire (21) were excluded from participation. Subjects were also free of psychological and neurological disorders, including depression and anxiety, as assessed using the Beck’s Depression (22) and Anxiety (23) questionnaires. None of the subjects had sleep disorders and reported good sleep quality (Pittsburgh Sleep Quality Index score <5), were nonsmokers, and did not take any medication that could interfere with either sleep or exercise performance. None of the subjects followed a high-fat, low-carbohydrate, ketogenic diet or consumed ketotic supplements during the last 3 months before the study. Ten well-trained male cyclists with good sleep quality (age: 23 ± 4 yr (mean ± SD); body mass: 70.7 ± 4.8 kg; height: 1.79 ± 0.05 m; lactate threshold (LT): 267 ± 38 W; V̇O_2max: 62.9 ± 7.2 mL·kg⁻¹·min⁻¹) and an average cycling volume of 10.8 ± 4.4 h·wk⁻¹ met the inclusion criteria and signed the written informed consent before participation. The study was approved by the Ethics Committee Research UZ/KU Leuven (B3222021000492) and complies with the Declaration of Helsinki (registered at www.clinicaltrials.gov as NCT05439720).

Preliminary Testing

Two weeks before the start of the experimental sessions, subjects completed a maximal graded exercise test on a calibrated cycling ergometer (Cyclus 2; RBM elektronik-automation GmbH, Leipzig, Germany) to determine LT and maximal oxygen uptake rate (V̇O_2max). Initial workload was set at 100 W and increased every 8 min with 40 W until volitional exhaustion to determine LT. Heart rate was measured continuously (Polar H10; Polar, Kempele, Finland), and capillary blood samples were obtained every 4 min to determine blood lactate concentration (Lactate Pro2; Arkray, Amstelveen, the Netherlands). LT was defined as the lowest workload corresponding to a 1 mM blood lactate increase from min 4 to 8 within the same stage. Subsequently, the subjects rested for 15 minute before starting the V̇O_2max test. The V̇O_2max test started at an initial workload of 70 W and was increased by 25 W every 30 s until volitional exhaustion. During the test, V̇O₂ and V̇CO₂ were measured continuously (Cortex Metalyzer 3B; Cortex, Leipzig, Germany), and V̇O_2max was determined as the highest O₂ uptake rate over a 30-s period.

General Study Design

Experiments were conducted according to a double-blinded, placebo-controlled, crossover design (Fig. 1). The study consisted of three experimental sessions, each separated by a 1-wk washout period. The order of the three experimental conditions was randomized by a researcher who was otherwise not involved in the study. Two of the three experimental sessions involved a 120-min cycling endurance training session (ET_120′) starting 2 h after breakfast and a 90-min HIIT (HIIT_90′) ending 1 h before sleeping time. After each training session, and 30 min before sleeping time, the subjects received either 25 g of a KE (EX_KE) or a control drink (EX_CON). To determine the specific effect of strenuous exercise on sleep, we added an additional experimental session without exercise (R_CON). Except for exercise, R_CON followed the exact same protocol as EX_CON and EX_KE. In the R_CON condition, subjects received the control drink at the same time points as during the exercise conditions. Sleep was measured during the subsequent night in each condition using polysomnography (PSG). Subjects were randomly assigned to the experimental conditions in a counterbalanced order by a researcher who was otherwise not involved in the study.

FIGURE 1.

Schematic overview of the experimental design. In a randomized crossover design, subjects (n = 10) participated in three experimental conditions. In two conditions, subjects performed a morning endurance training session (ET_120′) and an evening 90-min high-intensity interval training (HIIT_90′) ending 1 h before sleeping time. Subjects received either 25 g of control (EX_CON) or KE supplement (EX_KE) after each training and 30 min before sleeping time. In addition, one condition without exercise but with control supplements (R_CON) was added to evaluate the effects of exercise on subsequent sleep alone.

Familiarization

Exactly 1 wk before the first experimental session (see Experimental Sessions section), subjects completed the full protocol, but without nutritional intervention, to become accustomed to the exercise protocol, the sleeping facility, and the sleep recordings. From the familiarization session until the end of the study, subjects had to maintain a regular sleep–wake rhythm with at least 7 h of sleep per night. In order not to deviate from their normal sleep–wake rhythm, subjects were allowed to choose their preferred wake-up time (lights on) and sleeping time (lights off). Adherence to this sleep–wake rhythm was monitored throughout the entire study period using a sleep diary and an actigraphic wristband (ActiGraph wGT3X-BT, ActiGraph, Pensacola, FL). Sleep quality and quantity of each night preceding the experimental sessions were assessed using the St. Mary’s sleep questionnaire (24).

Experimental Sessions

The evening before each experimental session, the subjects consumed a carbohydrate-rich dinner (~5600 kJ, 69% carbohydrate, 16% fat, 15% protein) at home and went to bed at their predetermined time. The next morning, they woke-up at their predetermined time, consumed a carbohydrate-rich breakfast at home (∼2600 kJ, 72% carbohydrates, 13% protein, 15% fat), and arrived at the exercise and sleeping facility 2 h after waking up. Each subject stayed in a fully equipped, temperature- and ventilation-controlled private room at the research facility (Bakala Academy—Athletic Performance Center, Leuven) during the entire experimental session to standardize light exposure and environmental temperature during the day (21°C) and night (18°C). The exercise training sessions were performed in a separate room within the facility. ET_120′ started 2.5 h after waking up and consisted of eight consecutive 15-min intervals during which the exercise intensity alternated between 60% and 80% of LT. One hour after ET_120′, subjects received a carbohydrate-rich lunch (~5000 kJ, 69% carbohydrates, 15% protein and 16% fat). In the afternoon, subjects received some free time, but activities (e.g., exercise or naps), food intake, or caffeinated drinks (e.g., coffee) that could interfere with sleep were prohibited. Five hours before sleeping time, subjects received a light evening meal (~1700 kJ, 69% carbohydrates, 15% protein, 16% fat). Two hours later, HIIT_90′ was started, which consisted of a 10-min warm-up (70% of LT), 10× 7-min intervals (3 min at 120% of LT–4 min at 50% of LT), and an all-out sprint at 175% of LT. HIIT_90′ ended 1 h before sleeping time, after which subjects took a lukewarm shower (38°C). Subsequently, the electrodes for the PSG measurement were attached, and subjects went to bed 5 min before their predetermined sleeping time. Next morning, subjects were woken up at their predetermined time. As such, total time in bed was identical between the experimental sessions for a given subject.

Nutritional Intervention

Subjects received 32 g of carbohydrates per hour during ET_120′ via an energy cake (650 kJ, 83% carbohydrates, 11% fat, 3% protein; 6D Sports Nutrition, Oudenaarde, Belgium). After each training session or at the same time of the day for the R_CON condition, subjects also received 300 mL of a high-dose protein–carbohydrate recovery shake delivering 64 g of carbohydrates and 30 g of proteins (1585 kJ, 67% carbohydrates, 32% protein, 1% fat; 6D Sports Nutrition). In addition, in EX_KE, subjects received 25 g of a ketone monoester immediately after each training session and 30 min before sleeping time. The pure (R)-3-hydroxybutyl-(R)-3-hydroxybutyrate KE, free of additional ingredients, was purchased from KetoneAid Inc. (Falls Church, VA). The dosing strategy was based on a previous study by our research group (14,25) and aimed to increase blood βHB levels to ~3 mM within 30 min postexercise and just before sleeping time. In the EX_CON and R_CON condition, subjects received a taste-matched placebo drink at the same time points. The placebo drink consisted of collagen Peptan^® (12.5% w/v; 6D Sports Nutrition) and 1 mM bitter sucrose octaacetate (Sigma-Aldrich, Bornem, Belgium) dissolved in water. The same control drink was used as in our previous studies in which subjects were unable to distinguish the control drink from KE (26–30). The total energy content of the three KE supplements was 1468 versus 139 kJ for the three control drinks. An inert low-caloric placebo drink was used to exclude potential effects on sleep from increased carbohydrate intake or slightly increased ketone body production associated with increased fat intake. The supplements were administered in nontransparent tubes to avoid potential visual identification. As subjects received the placebo drink in two experimental conditions, we also provided 2100 mg of maltodextrin in the form of 3× 700-mg gelatin capsules 30 min before sleeping time to further disinform the subjects by saying that the capsules contained different compositions of “slow-release” ketone bodies.

Experimental Measurements

Sleep PSG

Both familiarization and experimental nights were recorded with a digital amplifier (V-amp; Brain Products GmbH, Gilching, Germany) and digitized at a sampling rate of 1000 Hz. Electroencephalographic (EEG) recordings were made from Fz, Cz, Pz, Oz, C3, C4, A1, and A2 according to the international 10–20 system. The A2 electrode served as the reference electrode, and A1 was used as a backup reference electrode. An electrode applied to the middle of the forehead functioned as the ground electrode. Vertical and horizontal eye movements electrooculography (EOG) were recorded with electrodes above and under the right eye and with electrodes attached to the outer cantus of both eyes. EEG and EOG data were recorded with a 0.1-Hz low cut-off filter and a 30-Hz high cut-off filter. Muscle tone and movements were measured with a bipolar submental electromyogram of the chin with a low cut-off filter of 10 Hz and a high cut-off filter of 200 Hz. A 50-Hz notch filter was used to filter out electrical noise. An independent certified sleep technician (Sleep Well PSG, Canada), who was blinded to the treatment conditions, visually scored night recordings according to the Rechtschaffen and Kales guidelines in conjunction with the American Academy of Sleep Medicine guidelines. PSG data were visually scored in 30-s epochs and 0.3- to 35-Hz, 0.3- to 30-Hz, and 10- to 100-Hz bandpass filters were used for EEG, EOG and electromyogram signals, respectively. The following sleep variables were obtained: (i) total sleep time; (ii) WASO; (iii) total NREM and REM sleep; (iv) total N1, N2, and N3 sleep; (v) sleep onset, sleep onset to N2, N3, and REM sleep; and (vi) sleep efficiency (total sleep time/time in bed). A total of two datasets were lost because of technical issues, one from EX_CON and one from EX_KE. In one subject, the computer crashed during the night, and in another, all the electrodes had come off during the night. Therefore, sleep architecture data and sleep event data were analyzed with two missing data points.

Sleep event detection

First, EEG data were preprocessed in BrainVision Analyzer (Brain Products GmbH) by applying a 0.1- to 30-Hz bandpass filter and subsequently transferred to the Python environment (version 3.10.5). Detection of slow waves and sleep spindles was performed as previously described by Nicolas et al. (31). For each channel, we detected the number of slow waves and computed slow-wave density, duration, and peak-to-peak amplitude. Next, slow-oscillation parameters were averaged across the channels for N2 and N3 sleep. Concerning sleep spindles, we first detected the number of spindles and computed spindle density, duration, and amplitude for each channel and subsequently averaged the parameters across the channels for N2 and N3 sleep.

Assessment of perceived exertion, sleep quality, and blinding of supplementation

Rating of perceived exertion (RPE) was assessed via a visual 6–20 Borg rating scale immediately after ET_120′ and HIIT_90′. Immediately after waking up, subjects filled in the St. Mary’s Hospital Questionnaire (24) to evaluate perceived sleep quality. At the end of the study, subjects were asked to identify which supplement they received during each experimental session, as well as to indicate their confidence in their choice on a scale from 0 (no idea at all) to 10 (completely certain). They were also asked to indicate whether they believed the maltodextrin capsules they received just before bedtime contained “slow-release” ketone bodies and to indicate their certainty on a scale ranging from 0 to 10.

Capillary and venous blood measurements and analyses

βHB and glucose levels were measured (GlucoMen areo 2K-meter with β-ketone and glucose sensor strips; A. Menarini Diagnostics, Firenze, Italy) in capillary blood samples that were collected from a hyperemic earlobe (i) just before ET_120′ (baseline), (ii) 30 min after ET_120′ and (iii) HIIT_90′, (iv) just before sleeping time, and (v) at wake-up time. To ensure double blinding, a researcher who was not involved in any of the other experimental measurements measured blood βHB concentrations. Furthermore, blood lactate concentrations were measured at the end of ET_120′ and HIIT_90′ (Lactate Pro2, Arkray). In addition, 15 min before sleeping time, a venous blood sample was taken from an antecubital vein in vacuum tubes containing either EDTA or silica clot activator (Becton Dickinson (BD) Vacutainer, Erembodegem, Belgium). Next, blood samples were centrifuged at 1500 rpm for 10 min at 4°C, and plasma and serum samples were stored at −80°C for later analyses. Commercially available high-sensitive enzyme-linked immuno sorbent assay kits were used to determine plasma adrenaline, noradrenaline, dopamine (BA E-6600; LDN, Nordhorn, Germany), and serum serotonin concentration (BA E-5900R, LDN).

Urine collection and analyses

After ET_120′, subjects had to empty their bladder completely. Diurnal urine was collected from then until the start of HIIT_90′. Before sleeping time, subjects emptied their bladder again, and subsequently, nocturnal urine was collected until wake-up time. Diurnal and nocturnal urine was collected in urine flasks prepared with 10 mL of hydrochloric acid. A small volume (~10 mL) of urine was frozen (−80°C) for subsequent analyses of adrenaline, noradrenaline, and dopamine using a commercially available enzyme-linked immuno sorbent assay kit (BA E-6600, LDN) within 4 wk after the final experimental session.

Determination of nocturnal blood ketone levels

Nocturnal βHB concentrations could not be determined during the experimental sessions, as this would obviously disturb sleep. Therefore, we performed an additional experiment to have an indication about the degree of ketosis that was achieved during the night of the EX_KE condition. Three subjects (age: 25 ± 4 yr; body mass: 76.5 ± 3.3 kg; height: 1.75 ± 0.05 m; body mass index: 23.8 ± 2.1 kg·m⁻²; LT: 241 ± 12 W) who had not participated in the main study completed the evening protocol as described above. First, they consumed the standardized evening meal (~1700 kJ, 69% carbohydrates, 15% protein, 16% fat) and started HIIT_90′ 2 h later. Next, they consumed 25 g of KE followed by the recovery shake. Thirty minutes before sleeping time, the second dose of 25 g of KE was consumed. Starting from 1 h before sleeping time, a capillary blood sample from a hyperemic earlobe was obtained every hour until wake-up time (8 h of sleep) to determine nocturnal blood βHB levels (GlucoMen LX plus meter with LX β-ketone, A. Menarini Diagnostics).

Study Outcomes and Statistical Analyses

The primary outcomes were sleep efficiency and the nocturnal excretion of catecholamines. Secondary outcomes were total REM and NREM sleep, WASO, and blood catecholamine concentrations. All statistical analyses were performed using GraphPad Prism version 9.4.0 (GraphPad Software Inc., La Jolla, CA). A one-way analysis of variance was performed to examine differences between the experimental conditions for measurements obtained at a single time point per experimental session. A two-way repeated measures analysis of variances was used to analyze differences over multiple time points. Data containing missing values (i.e., sleep architecture and sleep event data) were analyzed by a mixed model. This mixed model used a compound symmetry covariance matrix and was fit using restricted maximum likelihood. The Geisser–Greenhouse correction was used when the assumption of sphericity was violated (Mauchly’s test). In case of a significant main effect on the one-way ANOVA test, or an interaction effect on the two-way ANOVA test (P < 0.05), this was followed by the Šidák post hoc test. Multiple comparisons were performed between R_CON and EX_CON to investigate the effects of exercise and between EX_CON and EX_KE to investigate the effects of KE. All data are presented as means ± SD. Effect sizes are presented as partial eta-squared (η_p²) for main and interaction effects, and in case of post hoc pairwise comparisons as Cohen’s d. Confidence intervals (95%) were included for significant pairwise comparisons for the primary and secondary outcomes. An a priori power analysis in G*Power (version 3.1.9.7) indicated that to obtain a significant one-way ANOVA effect (P < 0.05) on sleep efficiency (primary outcome) with a statistical power of 0.80, nine subjects were required. The effect size (Cohen’s F = 1.30) used for power calculation was based on a previous study in which it was found that presleep vigorous exercise disrupts sleep efficiency (8).

RESULTS

Blood βHB concentration and side effects

A condition × time interaction effect was detected for blood βHB concentration (P < 0.001, η_p² = 0.80, Fig. 2A). Blood βHB concentrations were similar between all conditions at baseline and remained around ~0.1 ± 0.1 mM during the entire session in R_CON and EX_CON. Conversely, within 30 min upon KE ingestion, blood βHB levels increased respectively to 2.8 ± 1.2, 2.0 ± 0.5, and 3.1 ± 0.3 mM after ET_120′ and HIIT_90′, and at sleeping time (all P < 0.001 for EX_KE vs EX_CON). At wake-up time, blood βHB concentrations had returned to baseline values in all conditions (~0.1 mM, P = 0.998). In the additional experiment performed to determine nocturnal βHB concentrations, blood βHB concentrations were slightly higher (4.0 ± 0.4 mM) at sleeping time than in EX_KE (Fig. 2B). During the night, blood βHB concentration gradually decreased to 1.5 ± 0.2 mM 3 h after sleeping time. During the second part of the night, βHB levels remained around 0.5 mM in all subjects until 6 h after sleeping time. After 8 h of sleep, βHB concentrations had returned to baseline values in all subjects (0.2 ± 0.1 mM). None of the subjects reported anecdotical gastrointestinal distress or any other side effect from the supplements or meals provided during the study.

FIGURE 2.

Effect of KE intake on diurnal (A) and nocturnal (B) blood d-βHB concentration. Individual data points together with mean ± SD. Immediately after a morning intermittent endurance training (ET_120′) and an evening 90-min high-intensity interval training (HIIT_90′) ending 1 h before sleeping time, subjects (n = 10) received either control (EX_CON) or KE supplements (EX_KE) in a randomized crossover design. In addition, one condition without exercise but with control supplements (R_CON) was added to evaluate the effects of exercise on subsequent sleep alone. Panel A shows the blood D-βHB concentrations just before the start of ET_120′ (baseline), 30 min after ET_120′, and HIIT_90′, just before sleeping time and just after waking up. Panel B shows the blood d-βHB concentrations from an additional set of subjects (n = 3) who followed the same evening protocol and in which D-βHB concentrations were measured at each hour of the night. *P < 0.05 between indicated conditions. Gray area indicates the time during which the subjects slept.

Sleep architecture and sleep arousals

Time in bed was on average 528 ± 46 min for all experimental conditions. A main effect was detected for total sleep time (P = 0.049), sleep efficiency (P = 0.048, Fig. 3A), and WASO (P = 0.004, Fig. 3B). Exercise tended to reduce total sleep time by ~12 min (R_CON: 495 ± 46 min vs EX_CON: 483 ± 42 min, P = 0.107, d = 0.03, 95% CI of difference = −2 to 26 min), thereby decreasing sleep efficiency by ~2.5% (R_CON: 93.9% ± 3.4% vs EX_CON: 91.5% ± 4.5%, P = 0.108, d = 0.04, 95% CI of difference = −0.4 to 4.8%). The trend for a decrease in sleep efficiency after exercise resulted from a doubling in WASO (R_CON: 19 ± 10 min vs EX_CON: 37 ± 20 min, P = 0.004, d = −0.94, 95%, CI of difference = −29 to – 6 min). Interestingly, compared with EX_CON, KE ingestion increased total sleep time by ~17 min (EX_KE: 499 ± 48 min, P = 0.041, d = −0.05, 95% CI of difference = −30 to −1 min), resulting in a 3% increase in sleep efficiency (EX_KE: 94.5% ± 3.5%, P = 0.040, d = −0.05, 95% CI of difference = −5.5 to −0.1%). This was accompanied by a ~90% decrease in WASO compared with EX_CON (EX_KE: 20 ± 15 min, P = 0.009 vs EX_CON, d = 0.62, 95% CI of difference = 4 to 29 min).

FIGURE 3.

Effect of exercise and KE intake on sleep architecture and sleep onset. Individual data points together with mean ± SD. Immediately after a morning intermittent endurance training (ET_120′) and an evening 90-min high-intensity interval training (HIIT_90′) ending 1 h before sleeping time, subjects (n = 10) received either control (EX_CON) or KE supplements (EX_KE) in a randomized crossover design. In addition, one condition without exercise but with control supplements (R_CON) was added to evaluate the effects of exercise on subsequent sleep alone. Panels A, B, and C show sleep efficiency, WASO, and sleep onset, respectively. Panels D and E show REM sleep and NREM sleep, respectively. *P < 0.05 between indicated conditions.

Finer examination of sleep architecture revealed that neither exercise nor KE affected sleep onset (11 ± 9 min, P = 0.088, Fig. 3C) or sleep onset to N2 (6 ± 6 min, P = 0.334), N3 (13 ± 8 min, P = 0.903), and REM (106 ± 44 min, P = 0.947). However, a significant main effect was detected for the duration of REM (P = 0.002, Fig. 3D) and NREM (P = 0.044, Fig. 3E) sleep. Exercise resulted in a reduction of REM sleep by 26% (R_CON: 125 ± 19 min vs EX_CON: 93 ± 14 min, P = 0.001, d = 0.36, 95% CI of difference = 13 to 49 min) and an increase in total NREM sleep of 3% (R_CON: 370 ± 49 min vs EX_CON: 390 ± 48 min, P = 0.027, d = −0.08, 95% CI of difference = −34 to −2 min). The latter effect was driven by small, but nonsignificant increases in N1 (R_CON: 16 ± 9 min vs EX_CON: 21 ± 9 min, P = 0.122), N2 (R_CON: 205 ± 47 min vs EX_CON: 216 ± 52 min, P = 0.238), and N3 (R_CON: 148 ± 68 min vs EX_CON: 153 ± 65 min, P = 0.960) sleep. Interestingly, KE ingestion counteracted the exercise-induced decline in REM sleep duration (EX_KE: 117 ± 26 min, P = 0.011 vs EX_CON, d = −0.35, 95% CI of difference = −43 to −6 min) but did not impact total NREM sleep duration (EX_KE: 382 ± 44 min, P = 0.358 vs EX_CON, d = 0.03, 95% CI of difference = −7 to 26 min). Arousal density data during NREM (0.22 ± 0.11 min⁻¹, P = 0.516) and REM (0.194 ± 0.11 min⁻¹, P = 0.519) sleep were similar between all conditions.

Sleep event-related analyses

A main effect was detected for sleep spindle density (P = 0.005, Fig. 4A) during N2 sleep. Post hoc analyses revealed that exercise significantly increased sleep spindle density by ~36% during N2 sleep (R_CON: 1.81 ± 1.15 spindles per minute vs EX_CON: 2.48 ± 1.09 spindles per minute, P = 0.005, d = −0.45, 95% CI of difference = −0.95 to −0.18 spindles per minute), without modulating sleep spindle duration (P = 0.612, Fig. 4B) or amplitude (P = 0.816, Fig. 4C) during N2 sleep. KE intake had no effect on sleep spindle density, duration, or amplitude during N2 sleep. Similarly, neither exercise nor KE intake affected the density (P = 0.366, Fig. 4D), duration (P = 0.065, Fig. 4E), or amplitude (P = 0.092, Fig. 4F) of sleep spindles during N3 sleep. Also, slow-oscillation density (N2: 1.48 ± 0.77 min⁻¹; N3: 8.74 ± 3.68 min⁻¹), duration (N2: 1.30 ± 0.08 s; N3: 1.26 ± 0.08 s), and peak-to-peak amplitude (N2: 119.60 ± 6.04 μV; N3: 124.19 ± 10.57 μV) were identical between conditions (all P values >0.234).

FIGURE 4.

Effect of exercise and KE intake on sleep spindle characteristics during N2 and N3 sleep. Individual data points together with mean ± SD. Immediately after a morning intermittent endurance training (ET_120′) and an evening 90-min high-intensity interval training (HIIT_90′) ending 1 h before sleeping time, subjects (n = 10) received either control (EX_CON) or KE supplements (EX_KE) in a randomized crossover design. In addition, one condition without exercise but with control supplements (R_CON) was added to evaluate the effects of exercise on subsequent sleep alone. Panels A, B, and C show sleep spindle density, duration, and amplitude during N2 sleep, respectively. Panels D, E, and F show sleep spindle density, duration, and amplitude during N3 sleep, respectively. *P < 0.05 between indicated conditions.

Subjective sleep quality

A main effect was detected for the question “How clear-headed did you feel after getting up in the morning?” on the St. Mary’s sleep questionnaire (P = 0.005, η_p² = 0.20). Post hoc analyses indicated that the subjects felt sleepier just after waking up the morning after exercise compared with nonexercise (R_CON: 3 ± 1 vs EX_CON: 2 ± 1, P = 0.008). Scores for EX_KE (2 ± 1) were similar as in EX_CON (P = 0.181). None of the other questions of the St. Mary’s Hospital Questionnaire differed between conditions (all P values >0.250). Furthermore, overall perceived sleep quality as evaluated by the question “How well did you sleep last night?” was similarly scored at 3 (R_CON: 3 ± 1, range: 2–5; EX_CON: 3 ± 1, range: 2–4; EX_KE: 3 ± 1, range 2–5), indicating that subjects were satisfied with their sleep in all conditions (P = 0.442).

Urinary volume and catecholamine excretion

A main effect was found for urine production during the night (P = 0.031, η_p² = 0.30), but not during the day (P = 0.215, η_p² = 0.12). Post hoc testing indicated that exercise lowered urine production during the night by ~30% (R_CON: 995 ± 442 mL vs EX_CON: 729 ± 248 mL, P = 0.042), whereas it was unaffected by KE intake (EX_KE: 729 ± 342 mL, P > 0.999). A main effect was also detected for the nocturnal excretion of adrenaline (P < 0.001, η_p² = 0.57), noradrenaline (P = 0.012, η_p² = 0.39), and dopamine (P = 0.005, η_p² = 0.44). Post hoc testing revealed that exercise increased the total nocturnal excretion of adrenaline (Fig. 5A) and noradrenaline (Fig. 5B) by 130% (R_CON: 4.15 ± 2.60 nmol vs EX_CON: 9.48 ± 5.15 nmol, P < 0.001, d = −1.14, 95% CI of difference = −8.08 to −2.59 nmol) and 40% (R_CON: 40.74 ± 14.67 nmol vs EX_CON: 56.60 ± 20.09 nmol, P = 0.009, d = −0.49, 95% CI of difference = −27.91 to −3.80 nmol), respectively, without impacting total nocturnal dopamine excretion (P = 0.146, Fig. 5C). However, KE ingestion increased total nocturnal dopamine excretion by 20% (EX_KE: 1345 ± 200 nmol, P = 0.033 vs EX_CON, d = −0.22, 95% CI of difference = −347 to −15 nmol), without affecting total nocturnal adrenaline (P = 0.414 vs EX_CON) and noradrenaline (P = 0.753) excretion. Neither exercise nor KE ingestion affected total diurnal urinary excretion of adrenaline (P = 0.732, η_p² = 0.03), noradrenaline (P = 0.560, η_p² = 0.06), and dopamine (P = 0.718, η_p² = 0.03).

FIGURE 5.

Effect of exercise and KE intake on nocturnal catecholamine excretion. Individual data points together with mean ± SD. Immediately after a morning intermittent endurance training (ET_120′) and an evening 90-min high-intensity interval training (HIIT_90′) ending 1 h before sleeping time, subjects (n = 10) received either control (EX_CON) or KE supplements (EX_KE) in a randomized crossover design. In addition, one condition without exercise but with control supplements (R_CON) was added to evaluate the effects of exercise on subsequent sleep alone. Panels A, B, and C show the total excretion of adrenaline, noradrenaline, and dopamine in nocturnal urine, respectively. *P < 0.05 between indicated conditions.

Plasma catecholamine concentration

A main effect was found for plasma adrenaline (P = 0.003, η_p² = 0.39), noradrenaline (P = 0.042, η_p² = 0.30), and dopamine (P = 0.043, η_p² = 0.32) concentrations. Exercise raised presleep plasma adrenaline and dopamine levels by ~110% (R_CON: 0.10 ± 0.08 nM vs EX_CON: 0.22 ± 0.08 nM, P < 0.001, d = −1.4, 95% CI of difference = −0.23 to −0.01 nM) and ~45% (R_CON: 0.23 ± 0.12 nM vs EX_CON: 0.33 ± 0.18 nM, P = −0.49, d = −0.49, 95% CI of difference = −0.19 to 0.00 nM), respectively. KE ingestion had no effect on plasma adrenaline (EX_KE: 0.24 ± 0.18 nM, P = 0.711 vs EX_CON) or dopamine (EX_KE: 0.34 ± 0.15 nM, P = 0.969 vs EX_CON). Despite a main effect for plasma noradrenaline, post hoc analyses revealed no differences between the conditions (R_CON: 2.42 ± 0.75 nM, EX_CON: 2.85 ± 0.58 nM, EX_KE: 3.20 ± 0.62 nM). Neither exercise nor KE affected serum serotonin levels (R_CON: 10.93 ± 3.76 nM, EX_CON: 10.28 ± 3.43 nM, EX_KE: 9.59 ± 3.34 nM, P = 0.371).

Blood glucose, blood lactate, and RPE

An interaction effect was found for blood glucose (P = 0.002, η_p² = 0.23), but not for blood lactate levels (P = 0.103, η_p² = 0.24) or RPE (P = 0.430, η_p² = 0.07). Blood glucose levels were similar between R_CON and EX_CON at most time points. However, immediately after (P = 0.020) and 30 min (P = 0.003) after ET_120′, as well as at the start of HIIT_90′ (P = 0.025), blood glucose levels were ~0.85 mM lower in EX_CON compared with R_CON. KE ingestion did not alter blood glucose levels at any time point (P values >0.104 at all time points). Blood lactate levels were ~3 and ~4.5 mM after ET_120′ and HIIT_90′, respectively, but were not affected by KE. RPE values after ET_120′ (16 ± 2) and HIIT_90′ (18 ± 1) were similar between EX_CON and EX_KE.

Identification of supplementation

For both the EX_CON and EX_KE conditions, 60% of the subjects correctly identified the supplement, with a degree of confidence of 4 ± 2 (range: 2–8). For the R_CON condition, 50% identified their supplementation correctly with a degree of confidence of 3 ± 2 (range: 0–6). In addition, 70% of the participants believed that the maltodextrin capsules indeed contained “slow-release” ketone bodies, with a degree of confidence of 5 ± 2 (range: 2–9).

DISCUSSION

Preliminary evidence obtained from rodents indicated that ketone bodies have the potential to beneficially impact sleep (19,20). Against this background, we investigated whether KE intake can counteract the negative effect of strenuous exercise on sleep (19). Therefore, we simulated an exercise training day comprising a morning endurance training session (ET_120′) and a late-evening HIIT session (HIIT_90′). Immediately after each training session as well as presleep, subjects received either a KE supplement or a placebo. We also included a nonexercise condition in which subjects received the placebo supplement. This allowed us to investigate (i) the effect of strenuous exercise on sleep architecture and (ii) the ability of ketone bodies to alter exercise-induced alterations in sleep architecture. Strenuous morning and late-evening exercise disturbed sleep architecture as evidenced by reduced REM sleep, increased NREM sleep, and increased WASO. Interestingly, postexercise and presleep intake of KE counteracted the exercise-induced decrease in REM sleep and WASO, as well as increased total sleep time and sleep efficiency, but did not impact subjective sleep quality. In addition, KE also increased urinary dopamine excretion during sleep.

In general, both acute and chronic exercise exert positive effects on sleep quality (1,32). However, during strenuous exercise periods such as grand tours in cycling (5), ultraendurance exercise (7), or when vigorous exercise ends in close proximity to bedtime, sleep quality is generally impaired (3,9,33). In line with these earlier observations, the exercise protocol used in the current study was effective to impair sleep quality without affecting sleep latencies. Specifically, exercise resulted in an increase in WASO and a trend for a decrease in total sleep time and sleep efficiency. In addition, exercise also substantially reduced REM sleep duration, which is consistent with other studies using late-evening exercise (3,9,33). Similar to previous studies, such decline in REM sleep was accompanied by an increase in NREM sleep (2,9). This increase in NREM sleep at the expense of REM sleep is believed to allow subjects to maximally benefit from the restorative effects of NREM sleep (33).

The main finding of the study was that postexercise and presleep KE ingestion counteracted the exercise-induced decrease in sleep efficiency. Our data suggest that this effect was driven by a KE-induced decrease in WASO and an increase in REM sleep duration. The latter results are in line with a previous study showing that a ketogenic diet increases REM sleep in children with epilepsy (20). This suggests that the increase in REM sleep after a ketogenic diet is likely mediated by ketosis per se and not by other metabolic alterations induced by a ketogenic diet. In contrast with such hypothesis, a study using intracerebroventricular injection of ketone bodies in mice observed no impact of βHB on REM sleep and even a decrease in REM sleep after lithium AcAc injection (19). But sleep stages are differently regulated in rodents compared with humans, suggesting that the effects of ketone bodies on sleep are species specific (34). Furthermore, the decrease in REM sleep upon lithium AcAc injection may simply result from the lithium ions as previous research showed a reduction in REM sleep in rats upon intraperitoneal injection of 150 mL·kg⁻¹ lithium carbonate (35). This is also further supported by the fact that transgenic mice overexpressing Glycogen synthase kinase 3β, which is inhibited by lithium in the millimolar range (36), display severe fragmentation of the sleep–wake cycle (37).

The catecholaminergic system also plays a critical role in the regulation of sleep. In this perspective, we observed that KE supplementation increased nocturnal dopamine excretion. Interestingly, increased central dopamine signaling has previously been shown to initiate the transition from NREM to REM sleep in mice (38). In addition, REM sleep is usually impaired in patients afflicted with Parkinson’s disease, which further highlights the important role of dopamine in REM sleep regulation (39). The observed increase in nocturnal dopamine excretion is in line with recent work from our research group, showing that KE intake during a 100-km ultrarun increased postexercise plasma dopamine levels (40). Considering that dopamine cannot cross the blood–brain barrier, the increased dopamine levels in plasma and urine suggest that exogenous ketosis stimulates dopamine release by the kidneys. But interestingly, others showed that a ketogenic diet also increased dopamine activity in the cerebral cortex of mice (41). Furthermore, βHB can increase intracellular NADPH levels, thereby supporting the conversion of dihydrobiopterin to tetrahydrobiopterin, a critical coenzyme in dopamine synthesis (42). Our data, taken together with available literature, suggest that the observed increase in REM sleep after exogenous ketosis is probably mediated by an increase in cerebral dopamine signaling.

We previously showed postexercise and presleep KE to blunt the increase in nocturnal adrenaline and noradrenaline secretion during a 3-wk overload training period (14). Given that elevated nocturnal adrenergic activity is a primary cause of poor sleep quality in athletes, or other individuals with insomnia (15), a KE-induced suppression in adrenergic activity may also beneficially impact sleep. Nonetheless, in the conditions of the current study, adrenaline and noradrenaline levels were not affected by KE intake. It is worth noting that other studies produced equivocal results related to the effect of ketone bodies on sympathetic activity as βHB has been shown to act as both an agonist (43) and antagonist (44) for G protein-coupled receptor 41 (GPR41) located in sympathetic ganglia. Furthermore, we recently showed that exogenous ketosis during exercise raised plasma noradrenaline concentrations (28). Therefore, we suggest that acute exogenous ketosis does not affect nocturnal adrenaline or noradrenaline activity after acute exercise and thus does not impact sleep through inhibition of sympathetic and adrenergic activity. As such, the observed decrease in catecholamine secretion during overload training likely does not represent an acute effect but rather an adaptive response to prolonged KE supplementation in combination with overload training.

Besides the effect of KE, we observed that exercise increased the number of sleep spindles during N2 sleep. These results are similar to previous data from Aritake-Okada et al. (45) in which four bouts of 40-min cycling at 40% of V̇O_2max during the day increased fast sigma power in the 13- to 16-Hz range during the subsequent night. In contrast, the study by Mograss et al. (46) found no effect on sleep spindle density when a 40-min moderate cycling exercise (at 60% of age-predicted HR_max) was followed by a 60-min nap (46). Discrepancies between studies could be explained by differences in the amount of NREM sleep, timing of exercise, and exercise intensity and duration. Our study and the study by Aritake-Okada et al. (45) used long bouts of exercise (>1 h) and nocturnal sleep recordings as compared with the study by Mograss et al., showing no exercise-induced spindle modulation with nap paradigms. Interestingly, compelling evidence indicates that sleep spindles have a major function in neural plasticity and in memory consolidation (47). In conjunction with our data, this suggests that exercise induces an increase in sleep spindle activity, which may aid the consolidation of recently learned motor tasks.

In the present study, subjects consumed the KE or placebo drink postexercise and before sleeping time. The lower energy content of the placebo versus KE drink resulted in a ~10% energy discrepancy throughout the day and a ~26% energy discrepancy from the evening meal onward between the control conditions and EX_KE. However, earlier data showed that a one- or twofold difference in presleep (48) or prenap (49) energy intake, as well as 2 d of caloric restriction (50), did not impact subsequent sleep outcomes. Therefore, it is unlikely that the higher caloric content of the KE drink compared with placebo impacted the sleep outcomes in the conditions of the current study. In addition, we cannot exclude that just a single 25-g KE dose, or even a lower dose before sleeping time, rather than the multiple postexercise dosing regimen used here, may suffice to counteract exercise-induced sleep dysregulations. Therefore, future studies must go in search of optimal dosing strategies. Furthermore, studies have shown that evening exercise only disrupts sleep when it ends closely before sleeping time (9). Because this situation rarely occurs in real-life conditions, it would be of interest to investigate the effects of KE intake under other circumstances in which sleep of athletes may become disrupted such as ultraendurance exercise (7), intensive training/competition periods (5), and high-altitude training camps (51). The current findings must not be simply extrapolated to other conditions associated with sleep disturbances. It remains to be established whether the beneficial effects of KE on exercise-induced sleep disturbances also appear under nonchallenged conditions or in other healthy or diseased populations afflicted by impaired sleep quality and quantity. In conclusion, here we demonstrate that exogenous ketosis counteracts an exercise-induced suppression of REM sleep and increase in WASO and therefore improves sleep duration and sleep efficiency after a strenuous exercise day. This indicates that KE ingestion is a potent nonpharmacological intervention to improve sleep quality in athletes. This study did not aim to unravel the molecular mechanisms supporting these effects, but we provide preliminary evidence that a KE-induced upregulation of dopamine signaling is potentially involved.