Alzheimer’s disease, stroke, heart disease, and diabetes all rank among the leading causes of death in the United States. Many behavioral and pharmacological strategies have been used to combat the accelerated incidence of these ailments. Increasing evidence indicates that states of elevated plasma ketones may improve prognosis for each of these conditions (1–3). Although β-hydroxybutyrate (BHB), by definition of its chemical structure, is not a ketone, it is conventionally referred to as a ketone because it is one of the three products of ketogenesis along with acetoacetate and acetone. For the purposes of this study, ketosis has been defined as blood BHB levels of >0.5 mmol·L−1 (4). This degree of ketosis is not extreme as ketoacidosis, which is often observed in uncontrolled type 1 diabetes mellitus when BHB levels reach levels that can cause the blood to become dangerously acidic (5). Establishing a state of ketosis is used as a therapeutic intervention for the treatment of conditions, including epilepsy (5), obesity (3), hypertension (6), and chronic pain (7).
In addition to these benefits, Kashiwaya et al. (2) found that BHB protects against neurodegeneration and toxin-induced nerve damage commonly seen in Parkinson’s and Alzheimer’s diseases. Enhanced synaptic plasticity and neurogenesis, resistance to neuronal degeneration, and enhanced recovery from nerve injury, all function as added neurological benefits of elevated ketone levels (1). Work done by Shimazu et al. (8) found that BHB increases histone acetylation, which induces the expression of certain genes that reduce oxidative stress and protects against cellular damage. In addition, mechanisms have been discovered that describe the role of ketone bodies in decreasing inflammation (9) and improving mitochondrial respiration and ATP production (10). Although ketone bodies have historically been considered a mere byproduct of metabolism (11), research suggests that they play an active role in metabolism and cell signaling (12).
Several studies suggest that many of the benefits described are still achieved even if elevated BHB levels are not maintained for an extended period of time. For example, Anson et al. (13) found that alternate-day fasting resulted in beneficial effects that met or exceeded those of caloric restriction when observing reductions in serum glucose and insulin levels as well as increased resistance of neurons in the brain to excitotoxic stress. The mechanisms explaining how ketone bodies act in benefiting each of these pathways are still being discovered, but research suggests that some of the benefits of maintaining elevated ketones result (at least in part) from the actions of the ketone bodies themselves (12).
Exercise also plays a key role in ketone body production and utilization (14). Courtice and Douglas (15) found that ketone bodies remained elevated for several hours after a 10-mile walk in a phenomenon known as postexercise ketosis. Although ketone bodies are produced from the breakdown of fat in prolonged exercise, these substrates are also readily used by the body during exercise; thus, ketone body concentrations increase moderately during an exercise bout (16). After cessation of exercise, a high level of ketone bodies is observed as acetoacetate and BHB continue production with less utilization (14). Although these and other works help validate the idea that exercise facilitates the need for ketones as metabolic fuel, they do not define the effects of exercise on the time course to achieving ketosis (17).
The state of ketosis is most readily achieved through carbohydrate restriction, complete calorie restriction (fasting), or a combination. Fasting has recently gained popularity as a means for weight loss and other health benefits (18). Although many studies have used fasting to induce a state of ketosis (19), the time course to achieve this state has not been well described (17). An appreciation of this time course in healthy individuals would provide an essential foundation for future research in discovering the physiological pathways and health benefits associated with this state.
To our knowledge, no other study has objectively measured BHB concentration over time in a fasted state compared with that of a fasted state with an initial bout of intense exercise. The aims of this study were to determine the BHB concentration over time during an acute (36-h) fast and to evaluate how an initial bout of exercise influences this concentration. Anton et al. (20) recommend that randomized controlled trials should use biomarkers of the metabolic switch as a measure of fasting compliance and energy source during the fasting period. BHB, insulin, and glucagon are each biomarkers that can be used to observe these effects (21). A secondary aim of the study was to evaluate how mood and hunger changed over the period of the fast, with and without an initial bout of exercise.
We hypothesized that BHB concentrations would be greater during the exercise condition compared with the nonexercise condition over the course of a 36-h fast. We also postulated that the time frame for achieving ketosis (0.5 mmol·L−1 BHB) would be significantly less when the fast was combined with an initial bout of exercise compared with fasting alone. In addition, we expected plasma insulin to decrease and plasma glucagon to increase more dramatically when fasting was combined with exercise. When considering hunger, we expected no difference in perceived hunger between the two fasting protocols (22). Finally, we hypothesized moods of tension, depression, anger, fatigue, and confusion would be higher and vigor would be lower on the exercise day compared with the nonexercise day.
MATERIALS AND METHODS
A randomized crossover design with counterbalanced treatment conditions was used to compare the influence of fasting alone to fasting combined with vigorous exercise on plasma BHB concentration. These two conditions included a 36-h water-only fast beginning at 8:00 pm and ending at 8:00 am, 36 h later. Multiple studies have demonstrated water-only fasting up to and exceeding 36 h to be safe and well tolerated for healthy participants (19). Approval from the university’s institutional review board was obtained before initiating any aspect of this study (IRB2019-319; Brigham Young University, Provo, UT).
Participants completed two treatment conditions on identical days of the week, with a minimum 6-d washout between each session (maximum washout period was 13 d). Using randomizer.org, condition order was randomly assigned to participant numbers before the study (23). The participant numbers were assigned to participants chronologically from the time they joined the study. Before each laboratory session, participants were screened for contraindications to participation as outlined hereinafter. The outcome variables measured were body mass index (BMI), percent body fat, fat mass, hunger, thirst, stomach discomfort, mood, capillary BHB levels, plasma glucagon, and plasma insulin.
On the day the fast was initiated, participants were asked to maintain normal eating behaviors and hydration patterns to avoid excess caloric consumption in preparation for the fast. Participants arrived to the laboratory having not eaten for 4 h. Only water was allowed leading up to the visit. After screenings and assessments, a standardized meal was provided, and a 36-h fast initiated. This was a water-only fast, meaning no other food or beverages were allowed during the fasting period. Participants were instructed to stay hydrated throughout the fast. Noncaloric, electrolyte and/or caffeinated beverages/additives were not allowed. Gum chewing was also prohibited.
Based on random assignment, participants either proceeded with a fasting-only regimen or participated in the exercise regimen 30 min after the initiation of the fasting period. During the testing period, participants were required to complete hunger and mood assessments and check and record capillary BHB levels every 2 h, except during sleeping hours. In addition, they returned to the laboratory for venous blood draws every 12 h beginning at 8:00 pm after the standardized meal. Detailed timelines for each condition are visually represented in Figures 1 and 2.
Twenty healthy adults (11 male and 9 female) were recruited through word of mouth, advertisements, and fliers in the local community. Participants were required to be capable of participating in vigorous physical activity without restrictions as measured by a Physical Activity Readiness Questionnaire (PAR-Q). The PAR-Q contained questions regarding contraindications to physical activity. Any “yes” response on the PAR-Q excluded participants. The baseline activity status of each participant was self-reported. Based on these results, 5% of participants exercised less than 30 min·wk−1, 10% of participants exercised between 30 and 90 min·wk−1, 35% of participants exercised between 90 and 150 min·wk−1, 20% of participants exercised between 150 and 225 min·wk−1, and 30% of participants exercised more than 225 min·wk−1.
Potential participants were excluded from the study if they
- were diagnosed with a metabolic disease, orthopedic impairment, eating disorder, or food allergy;
- were taking medications that altered metabolism (24);
- were consuming 60 mg or more of caffeine daily (25);
- were pregnant, lactating, or postmenopausal (26);
- did not have a BMI between 18.5 and <30 kg·m−2 (27); and
- were practicing ketogenic, carbohydrate, or calorie-restricted diets
Body weight and height were measured for all participants at the beginning of each session. Weight was measured using a digital scale (Seca, Hamburg, Germany) accurate to ±0.1 kg with participants dressed in athletic shorts and a T-shirt with shoes removed. Height was measured by a stadiometer accurate to ±0.1 cm (Seca). A GE iDXA (GE, Fairfield, CT) was used to assess fat mass, percent body fat, and visceral adipose tissue (28). Calibration of the DXA scan took place at the beginning of each testing day using a manufacturer-provided calibration block. Scans were analyzed using Encore software version 17.
Mood and hunger ratings
Participants were asked to rate their mood and energy levels using the Brunel Mood Scale (BRUMS), every 2 h while recording capillary BHB levels. The 24-item BRUMS measures six identifiable mood states (tension, depression, anger, vigor, fatigue, and confusion) through a self-report inventory on a 5-point Likert Scale from 0 (not at all) to 4 (extremely) based on current feelings. The BRUMS questionnaire has been validated by Rohlfs et al. (29) as a sensitive and trustworthy replacement for the Profile of Mood States questionnaire in detecting mood states for adolescent and adult populations. Hunger, thirst, and stomach discomfort were also assessed in the same intervals as mood and used a visual analog scale that ranged from 0 (not at all) to 10 (extremely) (30).
Plasma insulin and glucagon
Two 4-mL EDTA tubes of blood were taken from each participant at the median cubital vein within 30 s of tourniquet application. Venipuncture took place every 12 h (times 0, 12, 24, and 36 h) in the human performance laboratory. For processing, each 4-mL tube was inverted to allow for mixture. Both samples were centrifuged for 15 min at 1500g at 3°C within 10 min of collection, after which the plasma was aliquoted and placed in separate vials. Plasma samples were stored in a −80°F freezer until ready for analysis. Insulin and glucagon levels were quantified using standard 96-well microplate enzyme-linked immunosorbent assay (ELISA) kits according to the manufacturer’s instructions (Crystal Chem, Inc., Elk Grove Villiage, IL). The insulin ELISA kit (catalog no. 90095) had an assay range from 0.9–220 mU·L−1 and an analytical sensitivity of 0.25 mU·L−1. For insulin, the intra-assay coefficient of variation was 9.04% and the interassay coefficient of variation ranged between 5.95% and 8.58%. The glucagon ELISA kit (catalog no. 81520) had an assay range from 7.8 to 500 pg·mL−1 and an analytical sensitivity of 1.1 pg·mL−1. For glucagon, the intra-assay coefficient of variation was 6.89% and the inter-assay coefficient of variation ranged from 1.90% to 6.81%.
Capillary BHB assessment
Capillary BHB measurements were assessed on the even hours during the fasting session (with the exception of sleep time) using the Precision Xtra portable BHB meter (Abbott Laboratories, Abington, NC). Before each fasting session, these meters were calibrated using manufacturer control solutions. A 5-μL capillary blood sample was applied to an electrochemical strip inserted into the sensor. The Precision Xtra portable BHB monitor was demonstrated by Byrne et al. (31) to be accurate in measuring real-time whole blood capillary BHB compared with venous whole blood reference samples up to blood levels of 6 mmol·L−1.
Potential participants for the study were sent an e-mail containing a link to an online survey. The online questionnaire was used to ensure participants met the inclusion criteria. As part of the online survey, participants were asked to report any food allergies, and complete a food preference questionnaire and the PAR-Q. Qualifying candidates were invited to participate in the study and were instructed to arrive prepared to exercise. They were also instructed to avoid caffeine consumption and other stimulants on the testing day as well as to refrain from vigorous physical activity for the 24-h period before testing. Adherence to these pretest day protocols was assessed at the beginning of each session through verbal questioning and confirmation. To later reaffirm adherence, baseline insulin and glucagon were assessed and found not to be different between days, suggesting that participants entered the study in a similar metabolic state. If pretest protocols had not been followed, the participant was rescheduled.
Informed consent was given by participants before participation in any aspect of this study. Participants reported to the Human Performance Research Lab at the university for each assessment. Each participant was informed of the main purpose of the study and familiarized with the testing procedures. Training for proper portable BHB meter use took place in accordance with manufacturer guidelines, and participants were given a copy of these testing instructions. Using Qualtrics online survey software (Qualtrics, Provo, UT), participants logged their own capillary BHB blood levels and filled out the questionnaires relating to mood and hunger every 2 h. Participants were reminded to take and record these measurements via automated text messages. Participants were oriented to the Qualtrics software and given login information during the initial orientation. Participants were asked to go about their normal activities of daily living during the testing period and to avoid exercise or strenuous activity, including strength or cardiovascular training, yard work, hiking, or other moderate activity. Participants were also asked to maintain their normal sleeping patterns.
Participants were given exactly the same standardized meal to initiate each fast. The energy needs for each participant were estimated using equations validated by Hall et al. (32). This equation uses height (in centimeters), weight (in kilograms), age (in years), and sex to predict basal metabolic rates and has been validated for accuracy and reliability (32). An activity factor of 1.55 was used to estimate total daily energy requirements (33). Meals were standardized based on macronutrient content (60% CHO, 25% fat, 15% protein) and consisted of combinations of raw almonds, string cheese, crackers, apple slices, beef jerky, and commercially packaged peanut butter and jelly sandwiches. Meals were assembled in the human performance laboratory using both commercially available prepackaged foods and commercially available boxed foods that were weighed to match caloric and macronutrient requirements of the participant. Participants were given 25% (basal metabolic rate × 1.55 × 0.25) of their daily caloric requirements in the standardized meal. The same foods were given on both test days, and participants were instructed to consume all the food provided for each meal. Meal adherence was assessed in each session by direct observation by the researchers. Any noncompliance (not eating all food and/or eating other foods) required participants to redo that particular laboratory session on a different day when they had followed food protocols.
Participants were asked to eat normally leading up to the fast and abstain from food for 4 h before the standardized meal and initiation of the fast to prevent prefast caloric loading and to normalize measured blood markers (34). Consent, orientation, instruction, hunger assessment, anthropometric data, demographic information, and bloodwork were collected during the initiation of each session. Participants returned to the laboratory for anthropometric measurements and bloodwork at 12, 24, and 36 h of fasting where they were reminded to stay hydrated according to testing protocols. Testing procedures and protocols were reviewed with each participant before initiation of each fasting session.
Based on random assignment, participants completed either an exercise condition or a nonexercise condition first. During the nonexercise condition, after all measurements were taken (including a DXA scan) and after consuming a standardized meal, participants immediately proceeded with normal daily activity. During the exercise condition, all measurements were taken (excluding a DXA scan), followed by an exercise regimen 30 min after consuming the standardized meal (35).
Participants exercised on a treadmill at a grade and speed that brought their estimated heart rate reserve (HRR) to 70%. Exercise at this intensity is classified as intense (36) and was used because it has been shown to maximize glucose oxidation compared with lower-intensity training (37). Participants exercised in this manner until an equivalent number of calories was expended as given in the standardized meal. The formula used to calculate the participant’s target heart rate was as follows (38):
Maximal HR estimation was calculated using the following formula (39):
Participants were fitted with a strap-on heart rate monitor (Garmin, Olathe, KS) and instructed to be seated for 5 min to establish resting HR (HRrest). Once HR calculations were complete, subjects began the exercise. The speed and grade were adjusted to meet the target HR within the first 5 min of exercise. Once this speed and grade were set, they were not adjusted for the remainder of the exercise intervention. If a participant was unable to maintain the exercise at this intensity, they were allowed to take a 60-s break, after which the exercise was resumed.
The length of exercise was individualized in order to expend a similar number of calories to that consumed from the standardized meal. This calculation was based on the standard American College of Sports Medicine–established metabolic calculation converting oxygen to calories by multiplying liters of oxygen by 5. The equations that were used are presented hereinafter.
American College of Sports Medicine metabolic equation:
Equations to estimate time on treadmill:
- E = calculated energy based on standardized meal in kilocalories
- kg = weight of the participant in kilograms
- S = speed of the treadmill in meters per minute
- G = grade of the treadmill
- min = time on treadmill
All calculations were performed in a preset, protected spreadsheet to ensure accuracy. Indirect calorimetry was used continuously throughout the exercise to verify energy expenditure (COSMED, Rome, Italy).
Participant data are reported as means and SD. Condition and sex were the primary factors in the models. The two conditions being considered were fasting without exercise and fasting with exercise. All models were analyzed using R (40) and JAGS version 4.3.0 statistical software. Results were visualized using GraphPad Prism version 9.0.0 (San Diego, CA). The area under the treatment curve was analyzed using the trapezoidal rule with one observation per subject by treatment with area under the curve as the dependent variable. Area under the curve was computed in attempt to represent a total response with a single number as a measure of intensity of the response. We preferred the Bayesian paradigm for this analysis because it is consistent with the axioms of probability. Thus, we are not constrained to work assuming the null hypothesis is true, nor are our interval estimates only reflective of our confidence in a particular method (assuming the null is true) but are real estimates of the probability the parameter lies in the given interval. In addition, these methods appropriately account for correlations among parameter estimates, so no adjustment in posterior probabilities (PP) is necessary for multiple comparisons (41). PP densities for the parameters of interest were generated using Markov chain Monte Carlo methods. These PP densities could then be used to make appropriate inferential statements. PP values exceeding 0.95 were taken to be statistically significant. Similarly, PP intervals (PPI) not spanning zero were taken to be statistically significant. All chains from the posterior densities were evaluated to determine if convergence was achieved.
Using the standard 0.5 mmol·L−1 (4), the time frame to achieve ketosis for both males and females under the two treatment conditions was also evaluated using the Bayesian paradigm. To evaluate this time frame, an estimate of the curve was needed. After considering a number of possible curves, a cubic curve was found to fit best. The coefficients of the curve were estimated using a Bayesian hierarchical model. The best-fit model had hierarchical parameters (meaning the coefficients for each subject were treated as a draw from the population of possible coefficients) for the intercept, linear, and quadratic coefficients. A hierarchical paradigm was not used on the cubic coefficients.
Responses from the visual analog scales for hunger, thirst, and stomach discomfort were evaluated independently. All responses of the 24-item BRUMS mood questionnaire were divided into six mood factors (anger, depression, tension, vigor, fatigue, and confusion). Once sorted, each mood factor score was summed for a score between 0 and 16. The areas under the curve for each of these factors were analyzed using the Bayesian paradigm as previously described.
The sample size for this study was calculated a priori. Setting significance at 0.05 and power at 80% and conservatively estimating a 10% difference between conditions (effect size of 0.66) resulted in a sample of 20 participants.
We screened 31 individuals for eligibility. Of those screened, 11 were disqualified for reasons outlined in Figure 3. The remaining 20 participants were randomly allocated to condition order. Eleven men and nine women were recruited, and all participants completed all aspects of the study. The demographic characteristics of the participants are outlined in Table 1. The standardized meal fed to participants before each fast was 614.8 (SD, 85.2) kcal. Measured energy expenditure during the exercise bout on the fast and exercise day was 587.6 (SD, 120.1) kcal. The average METs during the prescribed exercise was 9.14 (SD, 1.37). The average respiratory quotient (R) (measured by indirect calorimetry) throughout the prescribed exercise was 0.95 (SD, 0.14), indicating that the major fuel source for the exercise was glucose (37).
TABLE 1 -
Demographic characteristics of participants.
||Male (n = 11)
||Female (n = 9)
||Cumulative (n = 20)
|Visceral adipose, g
|Exercise time, min
| Hawaiian/Pacific Islander
Area under the BHB curve
Of the anticipated 440 BHB samples, 438 were collected and the area under the curve for BHB was calculated for each participant and averaged (Fig. 4). There was no interaction between condition and sex (95% PPI, −6.62 to 18.94 mmol·L−1; PP = 0.47). Similarly, there was no main effect of sex (PPI, −7.53 to 8.27 mmol·L−1; PP = 0.54). Because no differences in sex were observed, men and women were analyzed together.
The mean (SD) area under the curve for the nonexercise intervention was 19.19 (2.59) mmol·L−1, whereas the mean area under the curve for the exercise condition was 27.49 (2.59) mmol·L−1. There was a treatment effect with the exercise condition resulting in a larger area under the curve than the nonexercise treatment (95% PPI, 1.94 to 14.82 mmol·L−1; PP = 0.99).
Time to achieve ketosis
Using 0.5 mmol·L−1 BHB as a minimum threshold for ketosis, a hierarchical model was used to estimate the time to ketosis for each condition. The posterior estimates of the mean (SD) time to ketosis were 21.07 (2.95) h with fasting alone and 17.5 (1.69) h when fasting was combined with exercise. This represents a 3.57-h reduction in time to ketosis when the fast was started with exercise (95% PPI, −2.11 to 10.87; PP = 0.89).
Plasma insulin and glucagon
Areas under the curve for plasma insulin, glucagon, and the insulin/glucagon ratio were calculated for each participant and averaged (Fig. 5). For insulin, there was no interaction between sex and condition (95% PPI, −39.69 to 97.28 μU·mL−1; PP = 0.76). Similarly, there was no main effect of sex (95% PPI, −74.51 to 34.07 μU·mL−1; PP = 0.77) or condition (95% PPI, −21.64 to 36.18 μU·mL−1; PP = 0.67) on insulin area under the curve.
For glucagon, there was no interaction between condition and sex (95% PPI, −365.77 to 361.88 pg·mL−1; PP = 0.62). Similarly, there was no main effect of sex (95% PPI, −289.56 to 95.16 pg·mL−1; PP = 0.85). However, there was a treatment effect with the exercise condition resulting in a 97.13 pg·mL−1 greater glucagon area under the curve than the nonexercise condition (95% PPI, 34.08 to 354.21 pg·mL−1; PP = 0.98).
For the insulin/glucagon ratio, there was no interaction between condition and sex (95% PPI, −3.07 to 44.79; PP = 0.79). Similarly, there was no main effect of sex (95% PPI, −7.29 to 28.11; PP = 0.88). However, there was a treatment effect with the nonexercise condition resulting in a 20.83 greater area under the curve than the exercise condition (95% PPI, 4.70 to 24.22; PP = 0.99).
Hunger, thirst, and stomach discomfort
Hunger, thirst, and stomach discomfort were measured using a visual analog scale (Fig. 6). There was no interaction between condition and sex for hunger (95% PPI, −8.45 to 69.90 cm; PP = 0.78), thirst (95% PPI, −22.93 to 40.91 cm; PP = 0.49), or stomach discomfort (95% PPI, −12.14 to 58.42 cm; PP = 0.39). Similarly, there was no main effect of sex on hunger (95% PPI, −25.93 to 59.96 cm; PP = 0.78), thirst (95% PPI, −37.59 to 35.85 cm; PP = 0.48), or stomach discomfort (95% PPI, −45.37 to 35.28 cm; PP = 0.40). There was no main effect of condition on hunger (95% PPI, −13.29 to 26.77 cm; PP = 0.76), thirst (95% PPI, −12.80 to 19.22 cm; PP = 0.65), or stomach discomfort (95% PPI, −8.37 to 27.35 cm; PP = 0.85).
Areas under the curve for the six moods assessed in the BRUMS were calculated and averaged for each participant. Mean responses for each time point can be visualize in Figure 7. There was no interaction between condition and sex for depression (95% PPI, −15.96 to 53.17; PP = 0.38), tension (95% PPI, −18.73 to 54.25; PP = 0.55), vigor (95% PPI, −13.46 to 99.21; PP = 0.52), fatigue (95% PPI, −23.37 to 79.34; PP = 0.27), and confusion (95% PPI, −8.27 to 52.42; PP = 0.63). We concluded no main effect of sex for depression (95% PPI, −43.87 to 33.19, PP = 0.37), tension (95% PPI, −38.42 to 43.23; PP = 0.52), vigor (95% PPI, −59.84 to 56.80; PP = 0.48), fatigue (95% PPI, −79.09 to 16.49; PP = 0.10), and confusion (95% PPI, −38.63 to 36.04; PP = 0.54). Similarly, there was no treatment effect for tension (95% PPI, −4.16 to 32.02; PP = 0.93), vigor (95% PPI, −19.36 to 37.45; PP = 0.74), fatigue (95% PPI, −34.37 to 16.70; PP = 0.25), and confusion (95% PPI, −3.18 to 26.77; PP = 0.94). A treatment effect for depression indicated a greater area under the curve on the nonexercise day compared with the exercise day (95% PPI, 0.01 to 28.94; PP = 0.95).
For anger, there was an interaction between condition and sex (95% PPI, 3.63 to 80.36; PP = 0.95). Therefore, it would not be appropriate to interpret a main effect for anger. Females reported more anger (mean (SD), 39.66 (17.71)) on the nonexercise day than males (15.81 (14.25)), and males reported more anger on the exercise day (mean (SD), 22.21 (14.35)) than females (4.63 (17.50)).
The main purpose of the study was to evaluate how an initial bout of exercise affects BHB concentration over a 36-h fast. We found that completing a bout of vigorous aerobic exercise at the beginning of the 36-h fast increased the concentration of BHB by 43%, compared with fasting alone, an effect that was independent of sex. We also observed a 3.5-h average reduction in time to ketosis (BHB of 0.5 mmol·L−1) in the exercise condition. These findings help to address the paucity of research in the understanding of BHB concentrations during the first 36 h of a water-only fast.
Several studies have evaluated BHB concentrations during an acute fast. These studies have a number of limitations that make it difficult to generalize their results. For example, Haymond et al. (42) measured BHB every 4 h over the course of an 86-h fast in 10 healthy men and 10 healthy women. Although mean BHB levels were not reported at each measurement time, the graph provided suggests that adults reach BHB levels of 0.5 mmol·L−1 between 16 and 18 h. Our fasting-only group achieved similar BHB concentrations after 21.07 (SD, 2.95) h, which is several hours slower. Browning et al. (19) measured BHB every 4 h over the course of a 48-h fast in nine healthy men and nine healthy women. Although mean BHB levels were not reported at each measurement time, the graphs provided suggest that adults reach BHB levels of 0.5 mmol·L−1 within 24–26 h, which is several hours longer than observed in the fasting-only group in our study (19). Neither of these studies controlled any aspect of diet coming into the fast, which may directly alter the time course to ketosis, especially if the participants prepared for the fast by consuming more food. The difference of time course in the Browning study may also be explained by the fact that the sample included adults with average BMIs of 25 kg·m−2 for men and 27 kg·m−2 for women (19).
One factor thought to alter the BHB concentration over time is sex. Haymond et al. (42) suggested that women build BHB more quickly than men. However, we did not observe this sex difference in our study under either condition. Our results are reflective of work by Browning et al. (19), who also found no such difference between sex in fasted conditions. Although sex did not play a role in most of the mood states, there was a difference in anger with women reporting more anger on the nonexercise day and men reporting more anger on the exercise day. Although overall anger scores were very low in both conditions, exercise may influence hormone production differently between males and females, and we speculate that this may play a role in the observed differences.
We hypothesized that exercise would result in more rapid rise of BHB concentration and a shorter time to ketosis. The basis for this hypothesis was related to reducing liver glycogen, which would facilitate the switching of metabolic fuels from glucose to fat. Schranner et al. (43) concluded that an acute bout of endurance exercise raises BHB levels significantly a few hours after the exercise but has little effect immediately after the exercise. The exercise prescribed in the study was calculated to closely match the calories of the standardized meal given to the participants at the beginning of the fast, and the intensity was chosen to target glucose utilization specifically. Affectively, the exercise bout in our study allowed for the earlier consumption of stored glycogen and subsequent use of fatty acids and ketone bodies for metabolic fuel. However, given the design of this study, we are not able to determine if the exercise itself is an independent driver of BHB production or if the increased production was a result of an energy deficit. Although we speculate that at least a portion of the outcome is derivative of the lower insulin (21), an energy deficit may be a major driver of BHB production as liver glycogen stores deplete.
Although understanding BHB concentration trends was the main focus of this study, we also assessed insulin and glucagon levels, as these two hormones both exert strong influences on lipolysis and ketone body production (21). As seen in Figure 5, insulin and the insulin/glucagon ratio most dramatically changed within the first 12 h of fasting, after which they seemed to level off. The reduction in plasma insulin seen in this study is supported by the work of Sutton et al. (44) and supports the notion that fasting does not have to extend beyond 12 h to appreciate the majority of the reduction in plasma insulin. The marked reduction in the insulin/glucagon ratio indicates that hydrolysis of adipose tissue triacylglycerol was taking place (45). This ratio provides insight for individuals looking to lose weight through the combination of fasting and exercise. Interestingly, glucagon inclined steadily in both conditions over the duration of both fasts, with the exercise condition remaining markedly higher. Although these trends are interesting and follow the established counterregulatory roles of insulin and glucagon on each other to protect against hypoglycemia, the antiketogenic effects of insulin seem to play a larger role than the proketogenic effects of glucagon (46). In fact, Capozzi et al. (45) recently demonstrated that glucagon is not necessary to facilitate an increase in BHB concentration in response to a fast.
A secondary question that was addressed in this study was how fasting altered feelings of hunger, thirst, and stomach discomfort and how these perceptions differed with or without exercise. This purpose was added to the study in order to examine the utility of adding exercise at the beginning of a fast. Regardless of the potential benefits of combining fasting with exercise, compliance would be difficult to maintain if it presented unpleasant side effects. Wegman et al. (47) found a significant increase in hunger and stomach discomfort over the duration of intermittent fasting. Our results confirm these findings and further show that, although both hunger and stomach discomfort increased throughout each fast, ratings did not change with the addition of exercise. Thus, combining an initial bout of exercise with fasting increases the BHB and glucagon concentrations while decreasing the insulin/glucon ratio over the course of a 36-h fast. These metabolic changes take place without increasing subjective ratings of hunger or fasting discomfort.
To our knowledge, this is the first study to observe how negative affect is influenced by the addition of exercise to a fast. Work by Appleton and Baker (48) found that fasting was associated with negative affect, and work by Bodnar (49) recently reiterated physical activity as a stimulant for the release of endogenous opioids such as β-endorphin and β-lipoprotein, which can affect perceptions of pain, mood, and discomfort. In the nonexercise condition of this study, moods of depression, confusion, fatigue, tension, and vigor exhibited a positive association with BHB concentrations. In the exercise condition, these moods also increased over the course of the fast but not more than in the nonexercised condition, and depression was actually reduced in the exercise condition. Thus, a greater degree of ketosis was achieved without a greater degree of negative affect. We speculate that the attenuation of negative moods is due, at least in part, to the well-documented effects of exercise on endogenous opioid production (49). Although anger was relatively low overall, because men and women responded differently, this relationship may be worth investigating in future studies to explore sex differences.
The enhanced concentration of BHB observed when combining exercise with fasting can benefit individuals participating in a variety of dietary practices. Although various fasting protocols such as time-restricted eating, alternate-day fasting, and prolonged fasting are currently popular, some fasting styles may be more beneficial for elevating the concentration of BHB than others. For example, someone practicing a 16-h fast would likely not reach BHB concentrations of 0.5 mmol·L−1, even when beginning the fast with exercise. Even an 18- or 20-h fast would likely be inadequate without exercise. The exercise-induced 3.5-h reduction in time to ketosis observed in this study may provide a feasible means to accelerate the metabolic changes observed during fasting. However, we also recognize that there may be benefits from achieving BHB levels less than those of 0.5 mmol·L−1 and that a shorter fasting period may be more tolerable and improve compliance. However, literature is lacking in evidence to support specific claims relating to benefits of lower BHB levels.
A few limitations should be considered when interpreting the results of this study. First, although each participant abstained from food for 4 h before presenting to the laboratory, we did not control the food intake of the participants that took place early in the day. The amount and type of food and drink consumed before the fast and standardized meal may have affected the metabolic state, especially when the participant was anticipating a subsequent 36-h fast. However, we asked participants to follow normal dietary patterns and to not overconsume at these meals. In addition, neither plasma insulin nor glucagon differed between conditions at baseline, suggesting that the participant entered the conditions in a similar metabolic state on both days.
Second, hydration was encouraged but not directly monitored in this study, and hydrations seem to have an effect on ketone body production (50). However, we did measure thirst, and the only difference between conditions was immediately after exercise. In addition, overall thirst scores were very low.
Third, we recognize that the ratio of BHB to acetoacetate grows over time during the initial days of fasting. However, acetoacetate was not measured in this study. Without this biomarker and its ratio to BHB, we cannot make conclusions on total ketone body production, and our discussion and conclusions are limited to BHB concentrations only.
Finally, the exercise prescribed in this study was intense and for a relatively long duration. It may be difficult for many people to complete this type of exercise protocol. However, the results do suggest that exercise does have an effect on BHB concentration, and it is likely that other exercise protocols that are less intense or of shorter duration may still influence this relationship but to a lesser extent.
The results of this project speak to the potential use of combining fasting with exercise as a clinical intervention for cardiometabolic disease. Future research should explore the effects of these interventions on specific biomarkers of cardiometabolic disease such as the inflammatory profile. In addition, future research might explore the effects of incorporating exercise at various times throughout a fast as well as the effects of lower-intensity exercise or high-intensity interval training on BHB concentrations during a fast.
Maintaining elevated BHB levels, even irregularly, has been linked to a number of positive health outcomes. The findings of this study suggest that combining a fast with an initial bout of exercise generates ketones earlier in the fast and yields higher BHB concentrations compared with fasting alone without meaningfully affecting hunger, thirst, stomach discomfort, or mood between conditions. In addition, insulin and the insulin/glucagon ratio experience a marked reduction in each condition within the first 12 h of fasting, whereas glucagon rose steadily in both conditions. Thus, exercising at the beginning of a fast may improve the metabolic outcomes of fasting.
We would like to thank the research assistants who helped collect data and analyze plasma samples for this study. The results of this study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. The results of the present study do not constitute endorsement by the American College of Sports Medicine.
This research was funded by a mentoring environment grant by Brigham Young University.
The authors declare no conflict of interest.
Author Contributions: Conceptualization, L. S. D., B. W. B., and B. T. B.; project administration, L. S. D. and C. L. B.; methodology, L. S. D., B. W. B., B. T. B., L. E. D., and L. A. T.; software, L. S. D. and B. W. B.; validation, B. W. B., B. T. B., L. E. D., and L. A. T.; formal analysis, G. F.; investigation, L. S. D., C. L. B., and H. L. Y.; resources, B. W. B.; data curation, G. F., C. L. B., and H. L. Y.; writing—original draft preparation, L. S. D.; writing—review and editing, B. W. B, B. T. B., L. E. D., and L. A. T.; supervision, B. W. B. All authors have read and agreed to the published version of the manuscript.
1. Mattson MP, Moehl K, Ghena N, Schmaedick M, Cheng A. Intermittent metabolic switching, neuroplasticity and brain health. Nat Rev Neurosci
2. Kashiwaya Y, Takeshima T, Mori N, Nakashima K, Clarke K, Veech RL. D-beta-hydroxybutyrate protects neurons in models of Alzheimer’s and Parkinson’s disease. Proc Natl Acad Sci U S A
3. Abbasi J. Interest in the ketogenic diet grows for weight loss and type 2 diabetes. J Am Med Assoc
4. Volek JS, Noakes T, Phinney SD. Rethinking fat as a fuel for endurance exercise. Eur J Sport Sci
5. Kang HC, Lee YM, Kim HD, Lee JS, Slama A. Safe and effective use of the ketogenic diet in children with epilepsy and mitochondrial respiratory chain complex defects. Epilepsia
6. Goldhamer AC, Lisle DJ, Sultana P, et al. Medically supervised water-only fasting in the treatment of borderline hypertension. J Altern Complement Med
7. Michalsen A, Kuhlmann MK, Lüdtke R, Bäcker M, Langhorst J, Dobos GJ. Prolonged fasting in patients with chronic pain syndromes leads to late mood-enhancement not related to weight loss and fasting-induced leptin depletion. Nutr Neurosci
8. Shimazu T, Hirschey MD, Newman J, et al. Suppression of oxidative stress by beta-hydroxybutyrate, an endogenous histone deacetylase inhibitor. Science
9. Youm YH, Nguyen KY, Grant RW, et al. The ketone metabolite beta-hydroxybutyrate blocks NLRP3 inflammasome-mediated inflammatory disease. Nat Med
10. Maalouf M, Sullivan PG, Davis L, Kim DY, Rho JM. Ketones inhibit mitochondrial production of reactive oxygen species production following glutamate excitotoxicity by increasing NADH oxidation. Neuroscience
11. Harvey KL, Holcomb LE, Kolwicz SC Jr. Ketogenic diets and exercise performance. Nutrients
12. Puchalska P, Crawford PA. Multi-dimensional roles of ketone bodies in fuel metabolism, signaling, and therapeutics. Cell Metab
13. Anson RM, Guo Z, de Cabo R, et al. Intermittent fasting dissociates beneficial effects of dietary restriction on glucose metabolism and neuronal resistance to injury from calorie intake. Proc Natl Acad Sci U S A
14. Koeslag JH. Post-exercise ketosis and the hormone response to exercise: a review. Med Sci Sports Exerc
15. Courtice FC, Douglas CG. The effects of prolonged muscular exercise on the metabolism. Proc R Soc Lond B Biol Sci
16. Koeslag JH, Levinrad LI, Lochner JD, Sive AA. Post-exercise ketosis in post-prandial exercise: effect of glucose and alanine ingestion in humans. J Physiol
17. Harvey CJDC, Schofield GM, Williden M. The use of nutritional supplements to induce ketosis and reduce symptoms associated with keto-induction: a narrative review. PeerJ
18. Mattson MP, Longo VD, Harvie M. Impact of intermittent fasting on health and disease processes. Ageing Res Rev
19. Browning JD, Baxter J, Satapati S, Burgess SC. The effect of short-term fasting on liver and skeletal muscle lipid, glucose, and energy metabolism in healthy women and men. J Lipid Res
20. Anton SD, Moehl K, Donahoo WT, et al. Flipping the metabolic switch: understanding and applying the health benefits of fasting. Obesity (Silver Spring)
21. Muoio DM, Newgard CB. Mechanisms of disease: molecular and metabolic mechanisms of insulin resistance and beta-cell failure in type 2 diabetes. Nat Rev Mol Cell Biol
22. Sumithran P, Prendergast LA, Delbridge E, et al. Ketosis and appetite-mediating nutrients and hormones after weight loss. Eur J Clin Nutr
23. Suresh K. An overview of randomization techniques: an unbiased assessment of outcome in clinical research. J Hum Reprod Sci
24. Verhaegen AA, Van Gaal LF. Drug-induced obesity and its metabolic consequences: a review with a focus on mechanisms and possible therapeutic options. J Endocrinol Invest
25. Johnston KL, Clifford MN, Morgan LM. Coffee acutely modifies gastrointestinal hormone secretion and glucose tolerance in humans: glycemic effects of chlorogenic acid and caffeine. Am J Clin Nutr
26. Morrow PG, Marshall WP, Kim HJ, Kalkhoff RK. Metabolic response to starvation. II. Effects of sex steroid administration to pre- and postmenopausal women. Metabolism
27. Healthy Weight Tools [Internet]. Available from: https://www.nhlbi.nih.gov/health/educational/lose_wt/BMI/bmicalc.htm
. Accessed January 7, 2020.
28. Bailey BW, LeCheminant G, Hope T, Bell M, Tucker LA. A comparison of the agreement, internal consistency, and 2-day test stability of the InBody 720, GE iDXA, and BOD POD® gold standard for assessing body composition. Meas Phys Educ Exerc Sci
29. Rohlfs I, Rotta T, Andrade A, Terry P, Krebs R, Carvalho T. The Brunel of Mood Scale (BRUMS): instrument for detection of modified mood states in adolescents and adults athletes and non athletes. FIEP Bull
30. Stubbs RJ, Hughes DA, Johnstone AM, et al. The use of visual analogue scales to assess motivation to eat in human subjects: a review of their reliability and validity with an evaluation of new hand-held computerized systems for temporal tracking of appetite ratings. Br J Nutr
31. Byrne HA, Tieszen KL, Hollis S, Dornan TL, New JP. Evaluation of an electrochemical sensor for measuring blood ketones. Diabetes Care
32. Hall KD, Sacks G, Chandramohan D, et al. Quantification of the effect of energy imbalance on bodyweight. Lancet
33. Energy and Protein Requirements [Internet]. Available from: https://pubmed.ncbi.nlm.nih.gov/3937340/
. Accessed December 19, 2019.
34. Plumelle D, Lombard E, Nicolay A, Portugal H. Influence of diet and sample collection time on 77 laboratory tests on healthy adults. Clin Biochem
35. Ball JR. Effect of eating at various times on subsequent performances in swimming. N Y State J Med
36. Jette M, Sidney K, Blumchen G. Metabolic equivalents (METS) in exercise testing, exercise prescription, and evaluation of functional capacity. Clin Cardiol
37. Purdom T, Kravitz L, Dokladny K, Mermier C. Understanding the factors that effect maximal fat oxidation. J Int Soc Sports Nutr
38. Solheim TJ, Keller BG, Fountaine CJ. VO2
reserve vs. heart rate reserve during moderate intensity treadmill exercise. Int J Exerc Sci
39. Roy S, McCrory J. Validation of maximal heart rate prediction equations based on sex and physical activity status. Int J Exerc Sci
40. Team RC. R: A Language and Environment for Statistical Computing
. Vienna, Austria: R Foundation for Statistical Computing; 2020.
41. Kruschke JK. Bayesian data analysis. Wiley Interdiscip Rev Cogn Sci
42. Haymond MW, Karl IE, Clarke WL, Pagliara AS, Santiago JV. Differences in circulating gluconeogenic substrates during short-term fasting in men, women, and children. Metabolism
43. Schranner D, Kastenmuller G, Schonfelder M, Romisch-Margl W, Wackerhage H. Metabolite concentration changes in humans after a bout of exercise: a systematic review of exercise metabolomics studies. Sports Med Open
44. Sutton EF, Beyl R, Early KS, Cefalu WT, Ravussin E, Peterson CM. Early time-restricted feeding improves insulin sensitivity, blood pressure, and oxidative stress even without weight loss in men with prediabetes. Cell Metab
. 2018;27(6):1212–21 e3.
45. Capozzi ME, Coch RW, Koech J, et al. The limited role of glucagon for ketogenesis during fasting or in response to SGLT2 inhibition. Diabetes
46. McGarry JD, Foster DW. Hormonal control of ketogenesis. Biochemical considerations. Arch Intern Med
47. Wegman MP, Guo MH, Bennion DM, et al. Practicality of intermittent fasting in humans and its effect on oxidative stress and genes related to aging and metabolism. Rejuvenation Res
48. Appleton KM, Baker S. Distraction, not hunger, is associated with lower mood and lower perceived work performance on fast compared to non-fast days during intermittent fasting. J Health Psychol
49. Bodnar RJ. Endogenous opiates and behavior: 2018. Peptides
50. Johnson RE, Passmore R. Interrelations among post-exercise ketosis (Courtice–Douglas effect), hydration and metabolic state. Metabolism