The effects of initiating a 24-hour fast with a low versus a high carbohydrate shake on glycemic control in older adults: a randomized crossover study

Abstract

Purpose

This study measured the impact of macronutrient composition of a pre-fast meal on time-to-ketosis and other metabolic indicators of glycemic control during a 24-hr fast within a population of older, sedentary, overweight adults.

Methods

Twenty-four adults who were over the age of 50, sedentary (< 150 min of weekly exercise), and overweight (BMI ≥ 27) participated in a randomized crossover study. Each of these inclusion criteria have been shown to increase the risk for the development of chronic diseases. Participants began each 24-hr fast with either a high carbohydrate/low fat/moderate protein (HC/LF) or an isocaloric low carbohydrate/high fat/moderate protein (LC/HF) shake. Metabolic indicators included subcutaneous glucose readings every 15 min throughout the study, capillary beta-hydroxybutyrate (BHB), and plasma concentrations of insulin, glucagon, glucagon-like peptide-1 (GLP-1) and glucose-dependent insulinotropic peptide (GIP). Measurements of these hormones took place at 0, 1, 24, and 48 h, and BHB measurements took place at 0, 1, 4, 8, 12 and 24 h.

Results

Glucose levels were higher in the HC/LF group 15 min to 2.25 h after fast initiation (p < 0.05 for all). There was a significant condition by time interaction for BHB (F = 3.84, p < 0.01). Nutritional ketosis (BHB ≥ 0.5 mmol/L) was reached on average by 12 h in the LC/HF condition but was not reached at any point during the fast on average in the HC/LF condition. An hour after consuming the LC/HF shake insulin was 41.9% lower (t = 6.13, p < 0.01), glucagon 23.6% higher (t = -4.72, p < 0.01), GLP-1 26.8% higher (t = -5.16, p < 0.01), and GIP 34.4% higher (t = -3.41, p < 0.01) compared to the HC/LF shake.

Conclusion

A low carbohydrate pre-fast meal can reduce time-to-ketosis in older, sedentary, overweight adults. Those looking to improve glycemic control through fasting or time-restricted eating interventions may consider the macronutrient composition of their pre-fast meal to improve its efficacy.

Introduction

Chronic diseases, such as Alzheimer’s disease, cardiovascular disease, cancer, and diabetes are among the highest causes of death within the United States [1]. In 2018, just over half of the United States adult population had at least one chronic disease [2] while 27% of that same population had more than one chronic disease [2, 3]. Additionally, the treatment costs of chronic conditions in the United States were estimated to total $4.5 trillion dollars in 2022, making these diseases both a significant health and economic burden [3, 4]. Though chronic disease prevalence is already a large issue, the population of adults ages 50 and over is projected to increase as well as the number of those within that population with one or more chronic conditions [5]. Since increased Body Mass Index (BMI) [6, 7], sedentary behavior [8] and metabolic dysfunction [9] are each among the most significant contributing factors in the development of chronic diseases, especially among older populations [10], interventions aimed at reducing these risk factors are of tremendous importance.

In recent years, extensive research has been conducted to gain a deeper understanding of the pathology of chronic diseases and to develop preventative measures. Studies have even shown that Alzheimer’s disease follows a similar pathological development as type 2 diabetes, including mechanisms linked to insulin resistance, neuroinflammation and oxidative stress [11]. While many behavioral and pharmacological strategies have been employed to reduce the incidence of these conditions, increasing evidence indicates that entering a mild to moderate state of ketosis through fasting or a very low carbohydrate diet can also improve the prognosis for these conditions [12–15]. A fast-induced state of ketosis has been shown to enhance glycemic control by improving insulin sensitivity along with one’s ability to metabolically switch between oxidation of carbohydrates and fatty acids [16].

Metabolic switching as defined by Anton et al. is the body’s preferential shift from utilization of glucose to fatty acids and ketones as a fuel source, typically occurring between 12 and 36 h of fasting [16]. During a fast, liver glycogen stores are progressively depleted, initiating the switch to fat metabolism. Free fatty acids can be converted into ketones which can be reliably measured by the amount of beta-hydroxybutyrate (BHB) in circulation [17]. As this metabolic switch takes place, plasma insulin levels are reduced, potentially reducing insulin resistance that occurs in hyper-insulinemic individuals [18]. Frequent incorporation of metabolic switching through intermittent fasting is associated with improved insulin sensitivity and glucose regulation in both animal and human studies [19–21].

Although many studies have evaluated the potential health benefits of fasting, the practice is difficult maintain. Few studies have evaluated potential interventions that could assist in making fasting a more tolerable and/or more beneficial practice. Previously, our laboratory demonstrated that beginning a fast with exercise accelerates the rate of metabolic switching, affectively reducing the time-to-ketosis (BHB 0.5mmol/L) by 3.5 h without impacting subjective hunger or mood ratings by participants [22, 23].

Beyond the use of physical activity, altering the time to metabolic switching and the tolerability of a fast may also be accomplished by manipulation of the pre-fast meal. The quantity and the quality of food that is consumed prior to fasting likely impacts the time frame in which metabolic fuel switches from primarily carbohydrate to greater fat utilization [24]. If the pre-fast meal has a lower carbohydrate load, the metabolic switch may occur more quickly, potentially resulting in greater ketone production and utilization earlier during the fast. If true, the metabolic benefits of fasting may be achieved with a shorter fast or result in greater benefits during a fast of similar length.

The purpose of this study was to measure the influence of macronutrient composition in a fast-initiating shake on time-to-ketosis and other markers of glycemic control within a sample of older adults who present with greater risk for chronic diseases [25]. We hypothesized that blood glucose, insulin, glucagon-like peptide-1 (GLP-1) and glucose-dependent insulinotropic peptide (GIP) would be lower while BHB and glucagon would be higher in those who began their fast with a low carbohydrate/high fat/moderate protein (LC/HF) shake compared to those who begin their fast with a high carbohydrate/low fat/moderate protein (HC/LF). We also sought to determine whether there is a lingering impact on glycemic markers the day following the fast after 24 h of free-living conditions.

Methods

Design

A randomized crossover design with counterbalanced treatment conditions was used. One condition provided the participant with a commercially available low-carbohydrate shake to initiate the fast. The other intervention provided the participant with a commercially available high-carbohydrate, yet isocaloric, shake to initiate the fast. The effects of these conditions on markers of glycemic control were assessed. These two conditions involved a 24-hour water-only fast beginning at 8:00 am and ending at 8:00 am the following day. A final assessment 24 h after participants broke their fast took place the next day. Approval from the University’s Institutional Review Board was obtained prior to initiating any aspect of this study and can be found in OpenScience Framework 10.17605/OSF.IO/9N7BU.

Participants completed both fasting conditions outside of the lab with a 7-day washout between each session. Using randomizer.org, condition order was randomly assigned to participant numbers prior to the study. Participants were assigned numbers chronologically based on when they joined the study.

Prior to each laboratory session, participants were screened for contraindications to participation as outlined in the section below. The following outcome variables were measured: Body Mass Index (BMI), fat mass, percent body fat, visceral adipose tissue, capillary BHB, continuous subcutaneous glucose monitoring, and plasma concentrations of insulin, glucagon, GLP-1, and GIP.

Participants

Twenty-four adults (11 female and 13 male) were recruited by word of mouth, advertisements, and fliers in the local community. Participants were 50 years of age or older with a BMI > 27, who did not participate in more than 150 min of physical activity per week [26]. This population was selected because older adults with obesity are at the highest risk of metabolic complications and may benefit from periodic fasting [27]. Participants diagnosed with chronic or metabolic diseases, eating disorders, or food allergies were excluded from the study. Additionally, participants were excluded who were taking medications that alter metabolism, appetite, or neurological function [28]; habitually consuming 60 mg or more of caffeine daily [29]; participating in calorie or carbohydrate-restricted diets; fasting more than once per week; or having irregular sleep patterns. Women were also excluded if they were pregnant or lactating [30].

Potential participants for the study completed an online qualification survey. As part of the online survey, participants were asked to report any food allergies. Qualifying candidates were invited to participate in the study and all interested participants gave written informed consent prior to participation in any aspect of the study. Participants were also instructed to avoid caffeine consumption and other stimulants on testing days as well as to refrain from vigorous physical activity for the 24-hour period prior to testing. Adherence to pre-test day protocols was assessed at the beginning of each session. If pre-test protocols were not followed, the participant was rescheduled.

One hundred and thirty-three individuals were assessed for eligibility for the study as outlined in Fig. 1. Most were excluded due to low BMI, medications, or high amounts of physical activity. Of these individuals, twenty-five qualified and subsequently enrolled in the study. Twenty-four individuals completed all aspects of the study, and one participant chose to no longer participate due to an inability to tolerate the blood draws.

Fig. 1.

Participant flow diagram

Treatment sessions

Participants reported to the 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 ketone meter use took place in accordance with manufacturer guidelines (Abbott Laboratories, Abington, UK), and participants were given a copy of these testing instructions.

Before the first treatment condition, weight was measured using a digital scale (Seca, Hamburg, Germany) accurate to ± 0.1 kg, and height was measured by a stadiometer accurate to ± 0.1 cm (Seca, Hamburg, Germany). Body mass index (BMI) was calculated as weight (kg) divided by the square of height (m²). Participants then received a whole-body dual-energy X-ray absorptiometry (DXA) scan to assess body composition, including fat mass, percent body fat, and visceral adipose tissue [31–33]. The scan was performed using a GE iDXA (GE, Fairfield, CT) which was calibrated at the beginning of each day of use using a manufacturer-provided calibration block. Encore software version 17 was used to analyze each scan, and visceral adipose tissue was calculated using the Core Scan application of the machine [34, 35].

Subjects were then provided with a continuous glucose monitor (CGM) (“Freestyle Libre” by Abbott) which was attached to the back of the non-dominant arm by the research team in accordance with manufacturer’s guidelines. Participants were asked to eat normally leading up to the fast but to abstain from food after 8:00 pm the night before the fast to normalize measured blood markers.

Shortly before 8:00 am the next day, participants reported to the lab for a baseline blood draw and finger prick. They also completed a survey on Qualtrics’s online survey software (Qualtrics.com), where they logged their own capillary BHB levels. Based on random assignment, participants then consumed either the HC/LF or LC/HF shake within a 5-minute window. The standardized shake was consumed by 8:00 am, initiating the 24-hour fast. Another blood draw, finger prick, and survey were taken an hour after shake consumption. Participants then proceeded with their normal daily routine with the exception of pricking their finger and logging their own capillary BHB levels at 12:00 pm, 4:00 pm, and 8:00 pm. Participants were reminded via automated text message to take and record these measurements both two hours prior to each appointment and immediately before they needed to take a capillary ketone reading.

Participants were asked to go about their normal activities of daily living during the testing period, though they were asked to avoid strenuous activity including strength or cardiovascular training, yard work, hiking, or other moderate activity during the fasting period and the 24-hours prior. Participants were asked to maintain their normal sleeping patterns. Each fast was a water-only fast, meaning no other food or beverages were consumed during the fasting period besides water. Participants were instructed to stay hydrated throughout the fast. Non-caloric, electrolyte, and caffeinated beverages/additives were not allowed. Gum chewing was also prohibited.

Participants returned to the lab at 8:00 am the next day (24 h) for a blood draw, finger prick, and survey. Following this visit, participants were permitted to break their fast with their own meal and resume their regular activity and diet. Participants reported to the lab the following day at 8:00 am (48 h) for a final blood draw in an overnight fasted state (12 h) before eating their first meal of the day. Between the 24-hr visit and 48-hr visit normal unsupervised eating and activity patterns were practiced. See Fig. 2 for a complete visual of protocols.

Fig. 2.

Protocol timeline of all measurements taken

DXA = dual-energy X-ray absorptiometry. CGM = continuous glucose monitor. LC/HF = low carbohydrate/high fat/moderate protein. HC/LF = high carbohydrate/low fat/moderate protein

Standardized shake

Participants were given a standardized shake to initiate each fast. The energy needs for each participant were estimated using equations validated by Hall et al. used by the National Institutes of Health [36]. This equation uses height (cm), weight (kg), age (years), and sex to predict basal metabolic rates (BMR) and has been validated for accuracy and reliability. An activity factor of 1.4 was used to estimate total daily energy requirements. Participants were given 25% (BMR x 1.40 × 0.25) of their daily caloric requirements in the standardized shake in each fast, though the macronutrient composition differed between shake conditions. The commercially available LC/HF shake– HlthCode^® Chocolate Macadamia flavor– was consisted of the following macronutrient calorie breakdown: carbohydrates = 4.36% (nutrose), Fat = 66.21% (blend of saturated, monounsaturated, and polyunsaturated fats from coconut oil, olive oil, medium chain triglycerides, flaxseed, cocoa butter fat, and ghee), and Protein = 29.43% (whey protein concentrate, collagen, egg whites). The commercially available HC/LF shake– Carnation Breakfast Essentials^® High protein– consisted of the following macronutrient calorie breakdown: carbohydrates = 56.78% (sucrose, maltodextrin, and lactose), Fat = 13.15% (blend of saturated, monounsaturated, and polyunsaturated fats from milk), and Protein = 30.06% (whey protein isolate and whey and casein from milk). Shake volume in each condition was matched within subjects and was computed using a protected spreadsheet calculator that mixed the shakes in volumes of water (LC/HF) or milk (HC/LF) to meet manufacturer recommendations. The calculation for macronutrient and total caloric composition for the HC/LF shake included the nutrition from the 2% milk that the shakes were made with to ensure that the shakes were isovolumetric and isocaloric in both conditions. Consumption adherence was assessed in each session by direct observation by the researchers.

Plasma processing

Within 5 min of phlebotomy, whole blood samples were centrifuged at 1500 x g for 10 min to separate the plasma. 40µL of a protease inhibitor cocktail (Halt™ Protease Inhibitor Cocktail 100x, Thermo Fisher Scientific, Inc.) was added to the plasma sample which was then vortexed and separated into three cryovials for subsequent storage at -80 °F. Fluorescence values for insulin, glucagon, GLP-1, and GIP were obtained using the Human Metabolic Hormone Magnetic Bead Panel multiplex kit (Millipore Sigma, Cat. #HMHEMAG-34 K) and a MAGPIX™ Multiplex Reader (Luminex Corporation, Austin, Texas, USA). Standard curves were created from the diluted standard fluorescence values and concentrations for each plasma sample value were interpolated.

Assay precision and variability

To assess assay precision, intra-assay and inter-assay coefficient of variation (CV) values were calculated for insulin, glucagon, GIP, and GLP-1. Intra-assay CV was determined by calculating the mean and standard deviation of duplicate wells for each sample, while inter-assay CV, reflecting variability across different assay runs, was calculated from quality control samples measured across six plates. The intra-assay CV for all analytes ranged from 6.0 to 8.4%, and the inter-assay CV ranged from 4.5 to 13.4%, consistent with the manufacturer’s reported assay performance.

Statistical analysis

The a priori sample size for this study was calculated to detect a 25% difference in the area under the BHB concentration curve between conditions. Based on an alpha of 0.05, a power of 85%, and a moderate effect size of 0.66, we determined that 24 participants were needed for adequate statistical power. The statistical software package R was used to calculate the sample size for the study based on the above parameters. An effect size of 0.66 is considered moderate, which is typical for human randomized crossover trials in this field [37]. Participants were recruited until 24 individuals completed the entire study.

Data are presented as means with standard deviations. Prior to analysis, each variable was assessed for normality using the Proc Univariate procedure in SAS to determine both skewedness and kurtosis. All variables were normally distributed except for GIP and glucagon, which exhibited positive skewness. To address this skewness, we conducted analyses both with and without log transformation and compared the results. Log transformation diminished the skewness of the data but did not alter the interpretation of the results compared to the non-log transformed data. Therefore, for ease of interpretation, we present the results from the non-log transformed data in this paper.

To assess differences in the area under the glucose and BHB curves, a linear mixed model for repeated measures using Proc Mixed in SAS was employed. The trapezoidal rule was applied to calculate the area under the curve (AUC) for each treatment, with AUC serving as the dependent variable to represent the overall intensity of the response. The AUC for glucose was calculated over two separate intervals: the first three hours following the administration of the pre-fast snack, and then for the entire 24-hour fasting period.

A linear mixed model for repeated measures using Proc Mixed in SAS was also utilized to evaluate both the main and interactive effects of condition and time across the 24-hour fast. This analysis was repeated to assess the main and interactive effects of condition and time from baseline to 24 hours post-fasting (t = 48) for both fasting conditions. The latter analysis aimed to investigate the residual effects of the two different fasting protocols on all variables of interest under similar conditions (ad libitum feeding followed by an overnight fast). In the model, ‘participant’ was included as random effects, allowing us to examine mean differences between conditions (as fixed effects) while also accounting for the variability or correlation structure among repeated measures within subjects (as random effects). A compound symmetry variance structure was used, and we applied the Kenward-Rogers degrees of freedom adjustment to correct for heterogeneity in variances and correlations. In this analysis, glucose levels were measured every 15 min, resulting in 96 total observations over the 24-hour period. In contrast, beta-hydroxybutyrate (BHB) was measured at six time points (baseline, 1, 4, 8, 12, and 24 h), and GLP-1, GIP, insulin, and glucagon were assessed at three time points (0, 1, and 24 h) and at two time points in an additional analysis (0, and 48 h).

Significant effects (p ≤ 0.05) were further evaluated using the least squared means method to obtain post-hoc pairwise comparisons using a Tukey adjustment to control for familywise error rate. F-values represent the overall main and interactive effects, while t-values reflect post-hoc pairwise comparisons. All statistical analyses were conducted using SAS software version 9.4 for Windows.

Results

The demographic characteristics of the participants are outlined in Table 1. The isocaloric and isovolumetric LC/HF and HC/LF shakes provided to participants at the beginning of each fast contained 591.11 ± 106.58 kcals.

Table 1.

Demographic characteristics of participants

	Male (n = 13)	Female (n = 11)	Cumulative (n = 24)
Age (years)	56.9 ± 6.8	59.8 ± 8.0	58.3 ± 7.3
Weight (kg)	102.8 ± 13.8	89.3 ± 23.0	96.6 ± 19.4
Height (cm)	183.6 ± 6.8	164.4 ± 4.3	174.8 ± 11.3
BMI (kg/m2)	30.9 ± 3.7	33.6 ± 8.1	32.2 ± 6.3
	Male (n = 10)	Female (n = 10)	Cumulative (n = 20)
% Body Fat*	35.1 ± 5.3	47.2 ± 5.7	40.9 ± 8.2
Visceral adipose tissue (g)*	2178.5 ± 848.6	1465.1 ± 992.9	1840.6 ± 965.5

Mean ± standard deviation

*Body fat percentage and visceral adipose tissue measurements were only obtained for 20 of the 24 participants

Glucose

There was no difference in glucose concentrations between shake conditions at baseline (t = 1.71, p = 0.0871). However, there was a significant condition by time interaction over the course of the fast (F = 4.68 and p < 0.0001). Continuous glucose measurements were obtained every 15 min by the CGM throughout the duration of the study, and our follow-up analysis showed significantly higher glucose concentrations 15 min after consumption of the HC/LF shake compared to the LC/HF shake (t = 3.36, p = 0.0008), continuing until 2.25 h after shake consumption (t = 2.82, p = 0.0048). From 2.5 h until the conclusion of the 24-hour fast there were no longer differences between conditions at any time point (p > 0.05 for all).

The mean glucose concentrations throughout the duration of the 24-hour fast are shown in Fig. 3. AUC for the first 3-hrs after consuming the shake in the HC/LF condition was 322.95 ± 51.83 mg/dL*hour. This AUC was 16.0% greater than the LC/HF condition which had an AUC of 275.17 ± 45.83 mg/dL*hour in the same timeframe (F = 21.89, p = 0.0001). Total AUC for the full 24-hours of the fast in the HC/LF condition was 2111.57 ± 261.60 mg/dL*hour and 2034.65 ± 247.40 mg/dL*hour for the LC/HF condition, which is only 3.7% higher (F = 2.62, p = 0.1201).

Fig. 3.

Mean glucose concentrations over time

* Indicates a significant difference between conditions from 0.25 to 2.25 h after shake consumption

Concentration curves were not smoothed but directly connected between data points which are plotted every 15 min

Beta-Hydroxybutyrate

There was a significant condition by time interaction for BHB (F = 3.84, p = 0.0023; Table 2). There were no differences in BHB concentrations between shake conditions at baseline, 1 h, and 4 h after beginning the fast. At 8 h, BHB concentrations were higher in the LC/HF condition (t = 2.26, p = 0.0245). BHB concentrations remained higher in the LC/HF condition at 12 h (t = 4.55, p < 0.0001) and 24 h (t = 3.74, p = 0.0002). The average AUC for the LC/HF fast was 10.23 ± 4.30 mmol/L*hour, while the average AUC for HC/LF fast was 6.54 ± 2.51 mmol/L*hour, which was significantly higher by 44.0% (F = 13.30, p = 0.0007; Fig. 4).

Table 2.

BHB concentrations (mmol/L) at various timepoints for each condition

BHB (mmol/L)
	0	1	4	8	12	24	F value	p-value
LC/HF	0.15 ± 0.05a	0.22 ± 0.09a	0.23 ± 0.11a	0.40 ± 0.27b*	0.54 ± 0.27c*	0.50 ± 0.28c*	3.84	0.0023
HC/LF	0.13 ± 0.07a	0.15 ± 0.05a	0.18 ± 0.07a	0.29 ± 0.15b	0.32 ± 0.16b	0.31 ± 0.21b

Mean ± standard deviation. F and p-values refer to the condition by time interaction for each outcome

^a−c Indicates a significant difference within condition. * Indicates a significant difference between conditions at a given timepoint. LC/HF = low carbohydrate/high fat. HC/LF = high carbohydrate/low fat

Fig. 4.

Mean BHB concentrations over time

*Indicates a significant difference between conditions from 8 to 24 h after shake consumption

Symbols represent the mean concentration at each timepoint with the bars representing the standard deviation

The average AUC was 44.0% more in the LC/HF condition compared to the HC/LF condition (F = 13.30, p = 0.0007) as indicated by the greyed area between the curves

BHB concentrations increased over the 24-hour fasting period, regardless of condition (F = 34.09, p < 0.0001) (Table 2). Follow-up analysis indicated that BHB concentrations in the HC/LF and LC/HF conditions did not change between 0 and 1 h (t = -0.35, p = 0.7289; t = -1.73, p = 0.0841, respectively) or 1 to 4 h of fasting (t = -0.84, p = 0.4003; t = -0.20, p = 0.8430, respectively). In the LC/HF condition, BHB concentrations increased between 4 and 8 h (t = -4.14, p < 0.0001) and between 8 and 12 h (t = -3.27, p = 0.0012) but remained similar between 12 and 24 h of fasting (t = 1.07, p = 0.2841). In the HC/LF condition, BHB concentrations increased between 4 and 8 h (t = -2.56, p = 0.0112) but remained similar from 8 to 12 h (t = -0.57, p = 0.5726) and 12 to 24 h of fasting (t = 0.12, p = 0.9080).

Nutritional ketosis, defined as a BHB concentration of ≥ 0.5 mmol/L [17, 38], was achieved on average after 12 h of fasting in the LC/HF condition. In contrast, this level of BHB production was not attained on average by the end of the 24-hour fast during the HC/LF condition (0.31 ± 0.21).

Hormones

There was a significant condition by time interaction during the 24-hour fast for all hormones reported (see Table 3). There were no significant differences in insulin, glucagon, GLP-1, or GIP concentrations between the LC/HF and HC/LF shake conditions at baseline (t = 0.02, p = 0.9859; t = 0.00, p = 0.9968; t = 0.01, p = 0.9931; t = -0.06, p = 0.9548, respectively). An hour post-shake consumption, there was a difference between conditions, with insulin 41.9% lower (t = 6.13, p < 0.0001; Fig. 5a), glucagon 23.6% higher (t = -4.72, p < 0.0001; Fig. 5b), GLP-1 26.8% higher (t = -5.16, p < 0.0001; Fig. 5c), and GIP 34.4% higher (t = -3.41, p = 0.0009; Fig. 5d) after consuming a LC/HF shake compared to the HC/LF shake. By 24 h, there were no longer any significant differences between conditions for any of the hormones measured (t = 0.06, p = 0.9538; t = -1.09, p = 0.2771; t = 0.44, p = 0.6576; t = -0.11, p = 0.9105, respectively).

Table 3.

Hormone concentrations (pg/mL) at various timepoints for each condition during and 24 h after concluding the fast. A direct comparison between baseline and 48 h was taken to compare two overnight fasted conditions preceded by uncontrolled food intake

		0 h	1 h	24 h	F- value	p-value	48 h	F-value	p-value
Insulin pg/mL	LC/HF	3680.1 ± 3557.5a	5028.9 ± 3233.9b*	3639.9 ± 3587.7a	9.25	0.0002	3924.4 ± 3455.6	0.01	0.9079
	HC/LF	3809.6 ± 3550.8a	7696.1 ± 4556.5b	3687.5 ± 3622.8a			3915.9 ± 3365.8
Glucagon pg/mL	LC/HF	99.10 ± 71.78a	174.8 ± 84.4b*	102.5 ± 62.2a	9.64	0.0002	109.2 ± 72.3	1.41	0.2411
	HC/LF	100.1 ± 76.8a	137.9 ± 86.2b	110.2 ± 72.8a			101.5 ± 68.3
GLP-1 pg/mL	LC/HF	218.9 ± 86.4a	445.5 ± 131.8b*	231.2 ± 92.7a	10.62	< 0.0001	256.2 ± 123.4†	1.55	0.2191
	HC/LF	223.8 ± 96.3a	354.3 ± 128.0b	243.8 ± 96.7a			239.7 ± 76.6†
GIP pg/mL	LC/HF	160.9 ± 105.9a	842.4 ± 400.0b*	103.0 ± 50.0a	6.22	0.0030	331.5 ± 394.8†	0.25	0.6221
	HC/LF	152.0 ± 62.8a	595.2 ± 249.7b	88.4 ± 55.6a			275.1 ± 313.1†

Mean ± standard deviation. F and p-values refer to the condition by time interaction for each outcome. ^a−b Indicates a significant difference within condition. * Indicates a significant difference between conditions at a given time-point. ^†Indicates a significant main effect of time from baseline to 48-hrs. LC/HF = low carbohydrate/high fat. HC/LF = high carbohydrate/low fat

Fig. 5.

Concentrations of insulin (a), glucagon (b), GLP-1 (c), and GIP (d) over time. *Indicates a significant difference between conditions at 1-hr after shake consumption

Symbols represent the mean concentration at each timepoint with the bars representing the standard deviation

Symbols and connecting lines are offset to prevent overlapping at each timepoint

Hormone concentrations were measured again 24 h after breaking the fast. Between breaking their fast and arriving in the lab 24 h later, participant food intake was not controlled. Differences between baseline and 48 h were determined to see whether there was a lingering impact of fasting as both measurements were taken after an overnight fast after 24 h of uncontrolled food intake. While there was no condition by time interaction between 0 and 48 h for any of the hormones analyzed (see Table 3), there was a significant main effect for time for GLP-1 and GIP (F = 7.81, p = 0.0076; F = 7.51, p = 0.0088, respectively). Both incretin concentrations were elevated 24 h after breaking the fast compared to baseline (see Table 3).

We conducted an additional analysis specifically evaluating the three-way interaction between sex, condition, and time for each outcome. However, this analysis showed no significant three-way interaction for any outcome (all p-values < 0.05), suggesting no different response between the sexes.

Discussion

Initiating a fast with a LC/HF shake resulted in significantly greater BHB production compared to the HC/LF shake, supporting our research hypothesis. Although glucose concentrations remained comparable between conditions for over half of the 24-hour fasting period, the initial rise in blood glucose following the HC/LF shake influenced time-to-ketosis. Nutritional ketosis was achieved on average approximately 12 h into the fast following the LC/HF shake, whereas nutritional ketosis was not attained on average by the conclusion of the 24-hour fast in the HC/LF condition.

The depletion of liver glycogen is the primary trigger for the shift toward ketone production [39]. Based on this, we hypothesized that the time-to-ketosis would be shorter in the LC/HF condition due to the low carbohydrate content of the shake, which closely mirrors a very-low-carbohydrate diet. This likely led to less liver glycogen at the onset of fasting, facilitating more rapid depletion of glycogen stores. Since liver glycogen typically provides energy for approximately 10–14 h [39], starting a fast with less liver glycogen accelerates the onset of ketosis as observed in the LC/HF condition compared to the HC/LF shake.

The time course of ketosis onset has been examined in various fasting studies. While Browning et al. reported ketosis between 24 and 28 h of fasting [40], other studies have observed ketosis within 20–24 h [22, 37, 41]. In each of these studies, the carbohydrate content of pre-fasting meals reflected the recommended macronutrient composition according to general dietary guidelines [42]. Thus, it was unexpected that ketosis was not achieved on average in the HC/LF condition in our study, which featured a comparable carbohydrate load to these previous studies.

Glycemic elevations during the first two hours of fasting in the HC/LF condition along with the demographic characteristics of our participants likely contributed to this delayed metabolic transition. Aging is associated with glucose intolerance [43], indicating these participants may not have responded well to the glucose load at the beginning of the fast. Additionally, animal models suggest that aging delays metabolic switching [44, 45]. Human studies indicate that reduced muscle mass and increased adiposity in older individuals may impair mitochondrial function, which may contribute to delayed metabolic transitions [46, 47]. This is particularly relevant as our participants were older, sedentary adults with overweight or obesity.

Insulin is secreted in response to glycemic elevations. One hour after shake consumption, both conditions showed elevated insulin levels; however, the HC/LF shake induced an insulin spike that was 1.5 times greater than that observed following the LC/HF shake, supporting our research hypothesis. By the end of the 24-hour fast, insulin levels between the two conditions converged, as expected given the short half-life of insulin [48]. No persistent effects on insulin or glucose levels were observed 24 h after breaking the fast.

Insulin secretion is partially regulated by incretins, including GLP-1 and GIP. These hormones, released from the L-cells and K-cells of the small intestine, account for 50–70% of insulin secretion [49]. Interestingly, our study found higher GLP-1 and GIP concentrations after the LC/HF shake, despite its lower carbohydrate content, contrary to our research hypothesis. Current literature suggests that while the role of incretins is closely tied to glycemic control [50], fat consumption seems to have a greater influence on incretin release than carbohydrate consumption [37, 49]. Previous research has reported elevated GLP-1 concentrations in response to high-fat diets [51, 52], as well as a more sustained rise in GIP after fat consumption [53].

Incretins also regulate glucagon secretion, with GLP-1 inhibiting and GIP stimulating glucagon release [50, 54, 55]. The elevated GIP levels following the LC/HF shake likely contributed to the higher glucagon concentrations observed in this condition one hour after consumption. It was surprising that glucagon levels also increased in the HC/LF condition; however, glucagon can stimulate insulin secretion when blood glucose is elevated after a meal [56].

By 24 h of fasting, glucagon levels returned to baseline with no differences between conditions. This pattern is similar to previous studies that indicate stable glucagon levels in rats [57] and stable or slightly reduced glucagon levels in humans after 22 to 24 h of fasting [37, 58]. There is evidence that humans can maintain blood glucose levels fairly well in a fasted state, with glucagon mainly increasing when blood glucose falls below approximately 85 mg/dL [59]. Although blood glucose fell below this level during the fast, they had returned to baseline levels by the time glucagon was measured at the conclusion of the fast. Additionally, though decreased glucagon has been reported in other studies [37, 58], these studies began the fasting period in the evening when glucagon levels are highest after a meal [55], whereas the fasting periods in our study began in the morning after an overnight fast where glucagon levels are lower [55].

While insulin and glucagon levels were indistinguishable at the end of the fast, the differences an hour after shake consumption may have played a role when considering the time-to-ketosis. Insulin inhibits ketosis by reducing lipolysis, while glucagon increases lipolysis and thereby oxidation of ketones [60]. FOXA2, a key regulator of ketone production, is inhibited by insulin and activated by glucagon [61]. In our study, insulin was lower and glucagon was higher after the LC/HF shake. These changes in hormone concentrations may be partly responsible for the increased BHB production following the LC/HF shake.

The observed reduction in time-to-ketosis after beginning a fast with a LC/HF shake, as opposed to a HC/LF shake, may have broader implications for those practicing intermittent fasting or time-restricted eating. Given that intermittent fasting is characterized by alternating periods of fasting and eating, the accelerated metabolic transition after a very low carbohydrate meal in this study suggests that such benefits may be replicated during each fasting interval. If practiced consistently, this approach could potentially yield sustained improvements in glycemic control. While prolonged fasting periods typically result in extended durations of ketosis, the findings of this study suggest that daily attainment of nutritional ketosis is achievable within shorter eating windows if the final meal of the day is characterized by a very low carbohydrate content.

Additionally, fasting may have implications for insulin sensitivity. Incretin concentrations were higher 24 h after concluding either fast than at baseline. This difference may be a lingering effect of fasting since the measurement was taken when participants were in an overnight fasted state and insulin and glucagon concentrations were similar to baseline. However, these results should be interpreted with some caution as our study did not include a non-fasting intervention. GLP-1/GIP agonists have recently gained popularity as drugs for weight loss and improvement of glycemic control. These medications have an insulin-sensitizing effect by reducing insulin secretion as a result of improved glucose tolerance [62]. Though these medications provide much higher incretin dosages than were potentially induced by fasting in our study, higher incretin concentrations a day after concluding the fast may contribute to improvements in insulin sensitivity.

Limitations and strengths

It is important to acknowledge the limitations of this study to accurately interpret the findings and enhance the quality of future research. First, plasma samples were collected at 0, 1, 24 and 48 h without intermediate sampling. More frequent sampling would have provided deeper insight into the time course of hormonal changes throughout the fasts; however, the selected time points aimed to capture the immediate impact of the pre-fast meal as well as the lingering impact at the conclusion of the fast and a day later while minimizing participant burden and risk.

Second, this study did not standardized activities of daily living, water intake, or menstrual cycle phase. While this behavioral variability introduced some degree of inconsistency, it also provided a more realistic experience, thereby enhancing the generalizability of the results. Participants also served as their own control, reducing variability between conditions. Additionally, the women in this study were all above the average age for beginning menopause in the United States [63], and though we did not measure menopausal status, we did not observe any differences between men and women in this study.

Finally, we used shakes rather than mixed meals to begin each fast. Although shakes are digested more quickly than solid foods [64], shakes are commercially available and consumed frequently as a meal replacement. In addition, we were able to precisely control the calorie and macronutrient content of these shakes, customizing them for each participant with high fidelity.

Despite these limitations, there were several beneficial aspects of this protocol. We continuously monitored blood glucose levels throughout the entire study and collected BHB measurements at more frequent intervals than the glycemic hormones, giving us greater insight to the time course to ketosis. Our focus on an unmedicated, at-risk population, allowed us to isolate the impact of macronutrient composition on glycemic variables while fasting in individuals who are more likely to have an abnormal metabolic profile.

Conclusion

Older, sedentary, and overweight adults seeking to use fasting to improve their glycemic control can utilize their pre-fast meal to enhance the fast’s effectiveness. Acute glycemic elevations at the beginning of a fast push the time course of metabolic switching further into the fast. Reducing the glycemic impact of the pre-fast meal appears to reduce the time-to-ketosis, thereby extending the metabolic benefits of fasting. These results can assist individuals who are at greater risk for chronic disease improve their glycemic control by optimizing the metabolic benefits of fasting. Future work should seek to understand whether these acute benefits can have longer-lasting impacts when coupled with forms of intermittent fasting or time-restricted eating.

Author contributions

LSD, NEP and BWB designed the study. EZG, LSD, PGG, CGJ and BWB obtained the data, coordinated the study, and drafted the initial paper. EZG, LSD and BWB performed statistical analysis. All authors contributed to the final draft and all authors read and approved the final paper.

Funding

Research reported in this publication was supported by a gerontology research grant from Brigham Young University. Research reported in this publication was also supported by the National Center for Advancing Translational Sciences of the National Institutes of Health under the Award Number TL1TR002368. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Data availability

The data obtained in this study is openly available in OpenScience Framework at DOI: 10.17605/OSF.IO/9N7BU.

Declarations

Ethics approval and consent to participate

Approval from the University’s Institutional Review Board was obtained prior to initiating any aspect of this study, and informed written consent was obtained from all participants prior to participation in any aspect of the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.