Early time-restricted eating improves markers of cardiometabolic health but has no impact on intestinal nutrient absorption in healthy adults

Summary

Early time-restricted eating (eTRE) improves aspects of cardiometabolic health. Although the circadian system appears to regulate nutrient absorption, little is known about the effects of eTRE on intestinal absorption. In this randomized crossover trial, 16 healthy adults follow a controlled, weight maintenance diet for 9 days, consuming all calories between 0800 and 1400 (eTRE schedule) or 0800 and 2000 (control schedule). We measure the energy content of the diet, stool, and urine with bomb calorimetry and calculate intestinal energy absorption. The eTRE schedule is more effective than the control eating schedule for improving markers of cardiometabolic health, including 24-h mean glucose concentrations and glycemic variability, assessed as the mean amplitude of glycemic excursions. However, eTRE has no effect on intestinal energy and macronutrient absorption, gastrointestinal transit time, colonic hydrogen gas production, or stool microbial composition, suggesting eTRE does not impact gastrointestinal function. This trial is registered (ClinicalTrials.gov: NCT04877262).

Graphical abstract

Highlights

• •Early time-restricted eating (eTRE) does not impact intestinal energy absorption
• •eTRE does not affect intestinal macronutrient absorption or transit time
• •eTRE has no effect on colonic hydrogen gas production or stool microbial composition
• •eTRE lowers 24-h glucose, glycemic variability, and fasting insulin and glucose

Dawson et al. found that restricting eating to a 6-h window earlier in the day (0800–1400) does not impact intestinal energy or macronutrient absorption but does improve markers of cardiometabolic health, including 24-h blood glucose, glycemic variability, and insulin sensitivity compared with a 12-h energy-matched eating schedule (0800–2000).

Introduction

Intermittent fasting is a dietary strategy used to facilitate weight loss and reduce cardiovascular disease risk.^1,2 Intermittent fasting is characterized by alternating periods of fasting (≥12 h) and ad libitum eating. Early time-restricted eating (eTRE) is a form of intermittent fasting that confines eating to a short window (4–10 h) in the early part of the day, followed by extended fasting until the next morning.³ Cardiometabolic improvements observed with TRE are often due to weight loss; several studies show that TRE leads to a self-imposed energy deficit and subsequent weight loss.⁴ However, clinical trials that control energy intake to maintain body mass and limit the early feeding window to 10 h or less per day also report improvements in insulin sensitivity, glycemic variability, blood pressure, and appetite.^5,6,7 Synchronizing eating habits with biological circadian rhythms takes advantage of improved glucose tolerance in the morning and may also capitalize on improved nutrient absorption and utilization during this time of day.

The circadian system regulates nutrient absorption and utilization. Following an eTRE schedule (0800–1600) for 2 weeks increased skeletal muscle branched-chain amino acid and glucose uptake following a standardized carbohydrate and protein beverage.⁸ In murine models, circadian system-related genes are ubiquitously expressed throughout the small and large intestine, undergo diurnal oscillations, and are associated with nutrient transporter expression and nutrient absorption.^9,10 In situ models in mice suggest that carbohydrate and lipid absorption are greatest during the dark cycle when mice are more active.⁹ However, these findings are contrasted with a recent study in mice that administered oil via gavage at various time points in the 24-h cycle and found that blood, hepatic, and intestinal lipids were highest during the light cycle when activity is low.¹⁰

Gastrointestinal motility also displays a circadian rhythmicity, where the migrating motility complex is greatest during the daytime and lowest overnight.¹¹ However, the effect of meal volume, versus the effect of meal timing, may have a stronger impact on gastrointestinal motility with eucaloric intermittent fasting (i.e., a greater volume of food is required in a shorter eating window). Intermittent fasting eating patterns may resemble the physiology of overeating, which has been shown to slow whole-gut transit time. Whole-gut transit time was ∼11 h slower when individuals overate (150% weight maintenance energy needs) compared with when the same participants underate (50% weight maintenance energy needs).¹² Furthermore, in the same study, overeating increased the percentage of energy absorbed compared with undereating. More research, particularly in humans, is needed to understand the relationship between meal timing and nutrient absorption. Measuring energy loss in stool can serve as a reliable indicator of intestinal absorption.^12,13,14 Only one study has directly compared stool energy loss between an eTRE and control eating schedule and found that eTRE increased stool energy loss (i.e., decreased absorption) by 32 kilocalories (kcal)/day.¹⁵ However, the stool collection period was short (i.e., 1 day), and the results may not translate to a free-living setting.

Therefore, the objective of the current study was to determine the effects of an eTRE schedule on intestinal energy and macronutrient absorption, components of energy expenditure, gastrointestinal function, and markers of cardiometabolic health, compared with a control eating schedule, in healthy adults consuming a controlled, energy-matched, weight maintenance diet under free-living conditions. We hypothesized that the eTRE schedule would increase energy and macronutrient absorption, slow gastrointestinal transit time, and improve the glycemic profile.

Results and discussion

This randomized, crossover, controlled feeding trial consisted of two 9-day weight maintenance diet periods comparing the effects of eTRE (i.e., all meals consumed in a 6-h window between 0800 and 1400) relative to a control (CON) eating schedule (i.e., all meals consumed in a 12-h window between 0800 and 2000) on intestinal energy and macronutrient absorption, markers of cardiometabolic health, and gastrointestinal function (Figure 1A). Weight maintenance energy requirements were calculated based on each participant’s resting energy expenditure measured by indirect calorimetry and self-reported physical activity level. After a 3-day acclimation period to the study diet and eating schedule, the same 1-day menu was provided to participants each day (study days 4–9) to assess the impact of meal timing on energy and nutrient digestibility independent of food components and diet composition (Figure 1B). Participants were required to consume all foods, beverages, and water provided. The primary endpoints were energy and macronutrient absorption, measured by energy lost in stool relative to energy consumed in the diet. Secondary endpoints included gastrointestinal transit time, hydrogen production, and microbial composition; subjective hunger and appetite; resting energy expenditure; thermic effect of food; glycemic control and variability; postabsorptive and postprandial metabolomics and glucose, insulin, and lipid concentrations; and expression of circadian system-related genes.

Figure 1.

Study design and intervention

(A) Study design and testing procedures. The order of the eTRE (0800–1400) and CON schedule (0800–2000) was randomized. Controlled diets were provided to the participants for each 9-day study period. A continuous glucose monitor (CGM) was applied on day 1, a blue dye marker consumed on days 4 and 7, and a gas-sensing capsule (Atmo Biosciences) ingested on day 5 of each study period. On day 6 of each study period, resting energy expenditure (REE) and thermic effect of food (TEF) measures were performed as illustrated. Blood was drawn while fasted and hourly for 4 h to analyze circulating insulin and metabolites.

(B) Controlled diet. Each menu provided three meals and three snacks every day. Caloric intake was tailored to each participant’s weight maintenance energy requirements, and participants were instructed to finish eating within a 6-h time frame (early time-restricted eating [eTRE]) or 12-h time frame (control [CON]).

Participant characteristics and compliance

Sixty-three individuals expressed interest in the study and were pre-screened to determine eligibility status. Of the 63 respondents, 17 provided written informed consent, were fully screened, and enrolled in the study. Ultimately, 16 participants (31.1 ± 5.2 years; BMI: 23.8 ± 3.4 kg/m²; body fat: 28.9% ± 7.0%) completed the study due to one participant’s withdrawal before the end of the first study period for personal reasons (Figure S1). Baseline participant characteristics are shown in Table S1. Participants consumed 100% of the foods and beverages provided during each study arm (Figure 1B) and were compliant with the prescribed eating windows [eTRE: 6 h (360 min) and CON: 12 h (720 min)]. Eating duration for eTRE was 356 (95% confidence interval [CI]: [352, 359]) minutes and for CON 711 (707, 715) minutes. In addition, body weight remained stable during both eating schedules (eTRE: 0.16 [−0.05, 0.36] kg vs. CON: 0.20 [−0.13, 0.53] kg; p intervention = 0.80). As another measure of compliance, participants maintained similar levels of physical activity during each study period (eTRE: 7,282 [5,514, 9,049] steps/day and CON: 7,324 [6,079, 8,568] steps/day; p intervention = 0.86).

eTRE has no impact on energy or macronutrient absorption

In this healthy cohort, eTRE had no impact on energy digestibility relative to the CON eating schedule (Table 1), although there was marked interindividual variability, with digestibility ranging from 86.3% to 95.3% during the eTRE schedule and 89.7% to 94.3% during the CON schedule, as shown in Figure 2. In addition, there were no differences observed in relative digestibility of fat, protein, or carbohydrate or absolute loss of fat, protein, or carbohydrate in stool when following the eTRE schedule compared with the CON schedule (Table 1).

Table 1.

Nutrient absorption and gastrointestinal health

	eTRE (n = 16)	CON (n = 16)	p Intervention
Energy digestibility, %	91.7 (90.5, 93.0)	91.7 (90.9, 92.5)	0.95
Fat digestibility, %	95.0 (94.0, 95.9)	94.6 (93.5, 95.6)	0.36
Protein digestibility, %	85.6 (84.1, 87.1)	86.8 (85.3, 88.2)	0.19
Carbohydrate digestibility, %	95.1 (94.1, 96.2)	94.9 (93.5, 96.3)	0.76
Stool energy loss, kcal/daya	211 (177, 252)	216 (180, 259)	0.65
Stool fat, g/daya	3.9 (3.0, 5.1)	4.2 (3.2, 5.6)	0.43
Stool protein, g/day	15.0 (12.2, 17.8)	13.9 (11.2, 16.6)	0.22
Stool carbohydrate, g/day	14.7 (11.3, 18.2)	15.9 (11.2, 20.6)	0.66
Stool fiber, g/daya	9.7 (6.9, 13.6)	10.3 (8.5, 12.5)	0.65
Stool excretion, wet, g/day	166 (128, 204)	178 (143, 214)	0.16
Stool excretion, dry, g/daya	40.7 (32.6, 50.2)	41.4 (34.6, 49.7)	0.58
Stool energy density, wet, kcal/g	1.4 (1.3, 1.5)	1.3 (1.2, 1.5)	0.20
Stool energy density, dry, kcal/g	5.2 (5.1, 5.3)	5.2 (5.1, 5.3)	0.66
WGTT, h	23.3 (18.8, 27.9)	26.1 (21.6, 30.7)	0.31
Bristol stool rating	3.9 (3.4, 4.3)	3.9 (3.5, 4.4)	0.76
Urine excretion, g/day	2,992 (2,482, 3,502)	3,013 (2,434, 3,592)	0.75
Urinary energy loss, kcal/day	106 (91, 121)	100 (88, 112)	0.27
Urinary nitrogen, g/day	11.3 (9.5, 13.2)	10.3 (9.0, 11.5)	0.022
Metabolizable energy, %	87.8 (86.6, 89.1)	88.0 (87.2, 88.8)	0.69

Values are raw means (95% CI). The p values are derived from linear mixed models with intervention (eTRE and CON), study period (1 and 2), and their interaction as fixed effects and subject as a random effect (SAS v.9.4, SAS Institute, Cary, NC, USA). The intervention-by-study-period interaction assesses the presence of carryover effects, sometimes alternatively referred to as sequence effects. There were no study period or intervention-by-study period effects for measures of nutrient absorption and gastrointestinal health, so p values for these effects are not included in the table. eTRE, early time-restricted eating; WGTT, whole-gut transit time.

^aValues are presented as raw geometric means (95% CI), and data are normalized by log transformation in the linear mixed model.

Figure 2.

Energy and macronutrient digestibility

(A–D) Shown are individual energy (A), carbohydrate (B), fat (C), and protein (D) digestibility from the 3-day collection period during the eTRE and CON schedules. Participants who started the eTRE intervention first are represented by gray dots. Participants who started the CON period first are represented by white dots. Energy digestibility percentage is calculated as 100 − (stool energy [kcal] / gross energy of the diet [kcal]) × 100. Macronutrient digestibility percentage is calculated as 100 − (stool macronutrient loss [g] / macronutrient content of the diet [g]) × 100. Compared with the CON schedule, eTRE had no impact on energy (p = 0.95), carbohydrate (p = 0.76), fat (p = 0.36), or protein (p = 0.19) digestibility. All raw data are presented as individual data points and were analyzed using linear mixed models in SAS (v.9.4; SAS Institute, Cary, NC, USA) with intervention (eTRE or CON), period (1 or 2), and their interaction considered fixed effects and participant treated as a random effect.

Our results differ from a carefully controlled metabolic chamber study that reported participants following a 5.5-h eTRE schedule (0800–1330) had increased stool weight (mean Δ ± SEM, 18 ± 5 g/day), stool energy loss (32 ± 9 kcal/day; eTRE: 174 ± 18 kcal/day versus CON: 142 ± 17 kcal/day, p = 0.005), and stool carbohydrate, fat, and protein loss compared with a CON eating schedule (0800–1900).¹⁵ In contrast, our data indicate a nonsignificant decrease in wet stool weight (eTRE: 166 [128, 204] g versus CON: 178 [143, 214] g; p intervention = 0.16) following the eTRE schedule compared with the CON schedule and no change in dry stool weight (eTRE: 40.7 [32.6, 50.2] g versus CON: 41.4 [34.6, 49.7] g; p intervention = 0.58) (Table 1). The energy density of stool (wet) produced during eTRE was numerically (but not statistically) greater than that produced during CON, resulting in equivalent energy digestibility (described above) and absolute stool energy loss between groups. It is difficult to reconcile the contrasting findings between studies, but differences in study population may be responsible, in part, for the discrepant findings. Both studies recruited young, healthy males and females; however, in the current study, the population was primarily White, whereas the Bao et al.¹⁵ study population was primarily Asian. Furthermore, our study population had greater mean BMI (23.8 ± 3.4 versus 21.9 ± 1.7 kg/m²) and consumed ∼900 kcal/day more gross energy, leading to ∼50–60 g/day greater stool weight and ∼50–100 kcal/day greater absolute energy loss. However, mean energy digestibility was similar between studies (i.e., 90%–92%).

A previous controlled-feeding, crossover trial¹⁶ is in agreement with our findings, reporting no significant differences in stool energy loss when the bulk of energy was consumed at lunch (60% energy needs) versus dinner (60% energy needs), with the remainder of energy consumed during lunch or dinner (15%), breakfast (15%), and snacks (10%). Although the study by Papadopoulou et al.¹⁶ did not employ a TRE model, the pattern of consuming most of the diet earlier vs. later in the day is comparable with TRE interventions. In addition, the study population had a similar mean BMI (24.3 ± 5.3 kg/m²) and consumed a similar amount of gross energy (2,606 ± 513 kcal/day) as our study population (Tables S1 and S2). However, stool output was ∼70–80 g/day lower and absolute energy loss ∼120 kcal/day less than in our study population, leading to greater mean energy digestibility (i.e., 94%–95%). In the study by Papadopoulou et al.,¹⁶ participants were allowed to self-select their diet, which was then matched during the second study arm, but self-selected dietary fiber intake may be responsible for mean energy digestibility differences between studies. Dietary fiber is known to increase stool weight and decrease energy digestibility,^17,18 but dietary fiber intake was not reported in the studies by Papadopoulou et al.¹⁶ or Bao et al.¹⁵

Energy and macronutrient digestibility are altered by single-dietary-component substitutions during controlled, isocaloric feeding trials.^19,20,21,22 For example, consuming 42 g/day of cashews decreased energy (2.0%), fat (1.7%), protein (1.1%), and carbohydrate (2.0%) digestibility compared with an isocaloric diet without cashews.²⁰ The same research group found similar reductions in energy and macronutrient digestibility when adding almonds,¹⁹ walnuts,²² and pistachios²¹ to a controlled diet. Importantly, the current study matched food and nutrient intakes between study periods, thereby ensuring that the known effects of whole foods and various nutrients, particularly fiber, on digestibility would not impact the findings.

eTRE has no impact on metabolizable energy

Metabolizable energy, which accounts for stool and urinary energy loss and represents the percentage of gross dietary energy available for metabolism, was not different during eTRE compared with the CON schedule (Table 1).

Daily urine production was not different between the eTRE and CON eating schedule. Similarly, urinary energy loss was not different between the eTRE and CON eating schedule (Table 1). Urinary energy loss is predominantly due to protein metabolism, whereas fat and carbohydrate metabolism contribute little to energy excreted in the urine of healthy adults. Urinary nitrogen loss was significantly greater when following eTRE compared with CON (Table 1). Urinary nitrogen has been used to assess dietary protein intake,²³ but because dietary protein intake was matched between the eating schedules (Table S2), the increase in urinary nitrogen following eTRE may be due to increased protein breakdown. Interestingly, changes in urinary nitrogen in response to the intervention differed by gender (i.e., intervention-by-gender interaction; Table S5). Males had less nitrogen excretion during CON compared with eTRE (p adjusted = 0.009), whereas females had no difference in nitrogen excretion between CON and eTRE (p adjusted = 1.0). Following eTRE led to reduced nitrogen balance (−0.2 [−1.7, 1.3] g/day) compared with the CON schedule (1.2 [−0.1, 2.4] g/day; p intervention = 0.02). Corresponding to urine nitrogen loss patterns, males had greater nitrogen balance during CON compared with eTRE (p adjusted = 0.009), whereas females had no difference between the interventions (p adjusted = 1.0).

Bao et al.¹⁵ observed a trend for greater urinary energy loss in the eTRE group (mean Δ ± SEM, 6.67 ± 3.14 kcal/day, p = 0.058), similar in direction and magnitude to our findings for differences in urinary energy loss between interventions. However, due to greater energy intake in our study population, urinary energy excretion was ∼50 kcal/day higher than in the Bao et al.¹⁵ study. Greater mean urinary energy loss in our study population likely contributed to greater variance in the outcome measure. Furthermore, Bao et al.¹⁵ used a paired t test for comparisons between groups, whereas we used a linear mixed model to adjust for study period (i.e., is an outcome always greater during study period 1 than 2 regardless of diet?) and ensured that there were no carryover effects from the first study period to the second (i.e., intervention-by-study period interaction). Adjustment for study period in crossover design studies accounts for learned effects and/or changes in a person over time. Furthermore, testing for carryover effects when using a crossover study design ensures that diet effects are due to the intervention and not due to the order in which the intervention was received.

eTRE has no impact on gastrointestinal transit time or hydrogen production

Whole-gut transit time (WGTT) was assessed as the average time between participants ingesting a non-absorbable blue dye capsule on days 4 and 7 and the first appearance of the marker in stool on each occasion. WGTT was not significantly different on the eTRE schedule compared with the CON schedule (Table 1). In agreement, several studies^13,15 in obese (n = 9) and lean (n = 12) individuals provide evidence that WGTT is stable, even when meal timing, dietary energy intake, or energy absorption were altered; however, in another study, underfeeding (50% energy needs) adults (n = 27) with obesity reduced WGTT (i.e., faster transit) and decreased relative and absolute energy absorption compared with when those same adults were overfed (150% energy needs).¹² In the current study, females tended to have slower overall WGTT (27.7 [22.9, 32.5] h) relative to males (21.8 [18.0, 25.6] h; p gender = 0.059), but the effects of eating schedule on WGTT did not differ by gender (p interaction = 0.61; Table S5). To control for the possible effects of hormonal fluctuations on gastrointestinal transit time and other outcome measures in the current study, female participants began each study arm on the same respective menstrual cycle day in the mid-follicular phase. Contraceptive use is characterized in Table S1.

To understand whether small intestinal transit time, colonic transit time, or fermentation profiles were impacted by eTRE, participants ingested a gas-sensing capsule (Atmo Biosciences, Victoria, Australia) on study day 5. Small bowel transit time and colonic transit time were not significantly different between the eTRE and CON schedule (Table S3). Consistent with these data, the eTRE schedule did not result in a significant difference in hydrogen production. These null findings persisted when hydrogen production was normalized by transit time. The fermentation of carbohydrates and dietary fiber by gastrointestinal bacteria produces several gases, including hydrogen.²⁴ We found that eTRE had no impact on intestinal hydrogen production by directly measuring luminal gas concentrations with an ingestible gas-sensing capsule. Prior work using hydrogen breath testing demonstrated that eating later in the day (2300 vs. 1800) decreased unabsorbed dietary carbohydrates after the breakfast meal the following morning (i.e., decreased hydrogen production), suggesting that night-time eating increases the absorption of carbohydrates.²⁵ Hydrogen breath testing involves collecting breath samples every 20 min for an extended period, which increases participant burden. We reduced the participant burden and increased the number of samples analyzed by using an ingestible capsule capable of collecting continuous gas samples.²⁶ Our results do not necessarily dispute the findings that night-time eating increases carbohydrate absorption because our protocol was designed to reveal differences in eTRE compared with eating over 12 h with the last meal ending at 2000. As evidence of the relationship between night-time eating and weight gain mounts,²⁷ future studies should compare eTRE with late TRE (with and without disrupted sleep) to further explore energy and macronutrient absorption in the context of chronobiology.

eTRE reduces hunger and appetite during the day

Participant ratings of hunger, fullness, capacity to eat, desire to eat, and calculated composite satiety scores are depicted in Figure S2. Participants completed the first survey at 0745 before eating breakfast. There were significant interactions between the intervention and time for each subjective appetite rating (p interaction < 0.0001 for all); however, post hoc analyses revealed that there were no significant differences in any of the 5 hunger and appetite metrics before eating breakfast (all p adjusted = 1). Post hoc analyses indicated that eTRE reduced hunger, capacity to eat, and desire to eat (all p adjusted < 0.0001) while increasing fullness and overall satiety (all p adjusted < 0.0001) at 1200 and 1700 compared with the CON schedule. The last survey completed at 2100 indicated that the hunger and appetite suppression observed in the mid-day had dissipated. Participants reported significantly greater hunger, capacity to eat, and desire to eat and less fullness and overall satiety at 2100 (7 h since the last meal) when following the eTRE schedule compared with the CON schedule (all p adjusted < 0.0001). Previous studies have shown that shifting the eating window to earlier in the day reduces hunger, capacity to eat, and desire to eat during much of the daylight hours.^7,15 Our data corroborate these findings and affirm that eTRE leads to increased satiety during the daytime, which likely contributes to the decreased ad libitum energy consumption observed in other eTRE trials.^7,28

eTRE has no impact on resting energy expenditure or the thermic effect of food

There were no significant differences in resting energy expenditure (REE) during the eTRE schedule (1,483 [1,316, 1,651] kcal/day) compared with the CON eating schedule (1,480 [1,309, 1,651] kcal/day, p intervention = 0.79; Figure 3A). In response to eating a mixed meal (731 [650, 812] kcal) consisting of the controlled breakfast and morning snack consumed on days 1 and 4–9 (Figure 1B), the increase in the thermic effect of food (TEF) over a 4-h postprandial period was not significantly different between the two eating schedules (p interaction = 0.68). The increase in energy expenditure from baseline peaked at 0.26 (0.22, 0.30) kcal/min during the eTRE schedule and 0.26 (0.22, 0.30) kcal/min during the CON schedule (p intervention = 0.76; Figure 3B). As a result, the incremental area under the curve (iAUC) for TEF was not significantly different between the eating schedules (p intervention = 0.59; Figure 3C). REE is regulated by the circadian system independent of energy intake, fasting duration, and activity levels.²⁹ Earlier reports demonstrate that inadequate sleep coupled with circadian disruption significantly reduces REE,³⁰ supporting the notion that decreased energy expenditure could partly explain increased obesity risk in individuals who are active during the biological night.³¹ We found that aligning energy intake to biological circadian rhythms (i.e., eTRE) had no impact on REE, and these results are consistent with the current body of literature.^7,8,15 Without disruption of the circadian system, eTRE is not likely to modulate REE. Similarly, these previous trials also report little impact of eTRE interventions on TEF when analyzing the first meal of the day.^7,15

Figure 3.

Energy expenditure and respiratory quotient

(A–D) Compared with the CON schedule, eTRE had no impact on REE (p intervention = 0.79) (A), peak increase in energy expenditure after a mixed-meal tolerance test (p intervention = 0.76) (B), overall TEF measured over time (intervention-by-time interaction, p = 0.68) and by incremental area under the curve (p intervention = 0.59) (C) but did lower overall respiratory quotient (RQ) values (p intervention < 0.0001) (D). All data are presented as raw means (95% CI) and were analyzed using linear mixed models in SAS (v.9.4, SAS Institute) with intervention (eTRE or CON), period (1 or 2), time point (0, 30, 60, 90, 120, 180, or 240 min), and their interactions considered fixed effects and participant treated as a random effect. The p values for intervention, time, and the intervention-by-time interaction are presented in (C) and (D). There were no significant intervention-by-study period effects (i.e., carryover effects) for energy expenditure or RQ measures.

eTRE lowers mean 24-h glucose and glycemic variability

Temporal glucose patterns were measured by a continuous glucose monitor worn during each study period. Early TRE reduced 24-h glucose AUC (p intervention = 0.0004) and 24-h mean glucose levels (94 [88, 100] mg/dL) compared with the CON eating schedule (101 [94, 107] mg/dL; p intervention = 0.0005; Figures 4A and 4B). Interestingly, eTRE resulted in lower blood glucose during nighttime (defined as 2001–0759) (eTRE: 82.4 [80.9, 83.9] mg/dL and CON: 94.3 [91.1, 97.4] mg/dL; p adjusted ≤ 0.0001) but not during the daytime (0800–2000; eTRE: 105 [102, 107] mg/dL and CON: 107 [104, 111] mg/dL; p adjusted = 1.0; Figure 4B). These results support previous findings that reductions in 24-h blood glucose concentrations are due to lower overnight blood glucose concentrations.^5,8 In addition to evaluating mean glucose concentrations, we also evaluated the fluctuations in glucose levels over 24 h. Higher glycemic variability is associated with hypoglycemia and macrovascular complications in individuals with diabetes,³² but glycemic variability may also have utility in detecting early stages of glucose dysregulation. Glycemic variability, assessed as the mean amplitude of glycemic excursions, was reduced by eTRE in this healthy cohort (eTRE: 37.1 [31.5, 42.6] mg/dL vs. CON: 47.1 [39.1, 55.0] mg/dL; p intervention = 0.007; Figure 4C). This is comparable with previous findings in obese individuals that showed that eTRE decreases glycemic variability.⁵ Early TRE appears to have robust effects on glycemic variability. Interestingly, participants in our study were of normal weight (BMI 23.8 ± 3.4 kg/m²) and consumed 3 meals and 3 snacks within a 6-h eating window during the eTRE schedule compared with 3 meals within a 6-h eating window in the study by Jamshed et al.,⁵ indicating that eating frequency within a shortened eating window does not impact findings related to the effects of eTRE on glycemic variability.

Figure 4.

Continuous glucose monitoring

(A–C) Compared with the CON schedule, eTRE reduced (A) 24-h continuous glucose total area under the curve (p intervention = 0.0004), (B) 24-h mean glucose levels (p intervention = 0.0005) and nighttime glucose levels (p intervention-by-time interaction = 0.0004; p adjusted ≤ 0.0001), and (C) glycemic variability as measured by the mean amplitude of glycemic excursions (p intervention = 0.007). All data are presented as raw means (95% CI) and were analyzed using linear mixed models in SAS (v.9.4, SAS Institute) with intervention (eTRE or CON), study period (1 or 2), time (daytime or nighttime; B only), and their interaction(s) considered fixed effects and participant considered a random effect. Error bars are excluded from (A) to improve readability. Daytime was defined as 0800–2000 and nighttime as 2001–0759. There were no study period or intervention-by-study period interactions (i.e., carryover effect) for the glycemic variables. ∗p < 0.05.

eTRE improves fasted but not postprandial metabolism in response to a mixed meal

The eTRE schedule also lowered fasting blood glucose (eTRE: 82 [79, 85] mg/dL and CON: 86 [82, 90] mg/dL; p intervention = 0.02), insulin (eTRE: 2.5 [1.9, 3.4] μIU/mL and CON: 3.6 [2.5, 5.3] μIU/mL; p intervention = 0.001), and homeostatic model of insulin resistance (HOMA-IR) (eTRE: 0.51 [0.37, 0.69] and CON: 0.77 [0.52, 1.1]; p intervention = 0.0008) without impacting 24-h urinary c-peptide (eTRE: 23.0 [18.1, 29.1] ng/mL and CON: 23.7 [18.4, 30.4] ng/mL; p intervention = 0.59) compared with CON. Similarly, there were no differences in urinary c-peptide when normalized by 24-h urine volume (eTRE: 69.5 [56.8, 82.2] μg/24 h and CON: 69.5 [58.6, 80.3] μg/24 h; p intervention = 0.99) or energy intake (eTRE: 28.4 [23.2, 34.7] ng/kcal and CON: 29.1 [24.1, 35.2] ng/kcal; p intervention = 0.69). Urinary beta hydroxybutyrate was below the level of detection (0.03 mmol/L) in all but one participant during the CON schedule and one participant during the eTRE schedule. The eTRE schedule had no impact on fasting triglyceride (eTRE: 73.5 [60.1, 89.7] mg/dL and CON: 80.0 [65.9, 97.0] mg/dL; p intervention = 0.08), total cholesterol (eTRE: 187 [168, 209] mg/dL and CON: 181 [164, 201]) mg/dL; p intervention = 0.30), low-density lipoprotein (LDL) cholesterol (eTRE: 103 [88, 121] mg/dL and CON: 99 [87, 114] mg/dL; p intervention = 0.31), or high-density lipoprotein (HDL) cholesterol (eTRE: 65 [56, 75] mg/dL and CON: 63 [54, 73] mg/dL; p intervention = 0.27) concentrations compared with CON. In addition, there was no impact of eTRE on the total cholesterol:HDL cholesterol ratio (eTRE: 2.9 [2.5, 3.3]) compared with CON (2.9 [2.6, 3.3]; p intervention = 0.82).

Following the standardized mixed-meal challenge, there were no differences in glucose AUC (p intervention = 0.08) or insulin AUC (p intervention = 0.17) between eating schedules (Figure 5). Interestingly, in the linear mixed model analysis, eTRE had greater glucose concentrations at 1 h (eTRE: 98.7 [90.8, 106.7] mg/dL vs. CON: 82.7 [69.2, 96.1] mg/dL; p interaction = 0.001, p adjusted = 0.006) following the standardized mixed-meal tolerance test compared with the CON schedule (Figure 5). In the linear mixed model analysis for insulin concentrations, there was an interaction effect between intervention and time (p = 0.02) but no significant differences after post-hoc adjustment (Figure 5). In the current study, participants fasted for different amounts of time during the eTRE (18 h) and CON (12 h) schedules to ensure that all blood draws occurred at the same time of day during each study period. Fasting duration is known to influence the postprandial glucose and insulin response; however, this is more apparent when the fast is extended beyond 24 h. Horton and Hill³³ demonstrated that extended fasting (72 h) results in larger glucose and insulin excursions following a test meal compared with an overnight (13 h) fast. Similar to the study by Horton and Hill,³³ we sampled blood every hour in the postprandial period and observed an increase in blood glucose concentrations at 1 h in the eTRE schedule (i.e., extended fast) compared with the CON schedule, although to a lesser magnitude. The continuous glucose monitor (CGM) data (Figure 4A) suggest that blood glucose is higher during the CON schedule following the morning meal. The reasons for the disparate findings are not entirely clear but may be due to not measuring glucose with enough frequency in the first hour after the meal to detect similar glycemic patterns. A large trial investigating whether postprandial glucose, insulin, and triglycerides could be predicted based on individual characteristics (i.e., age, gender, microbiome composition, body composition, etc.) and meal composition reported large variability in postprandial glycemic and lipemic responses,³⁴ with the mean glucose peak occurring 30 min after consuming a test meal.³⁴ In the current study, it is likely peaks in glucose that occurred from 0–1 h were missed. A limitation of the current study is the lack of blood draws at acute (0–1 h) time points following the mixed-meal tolerance test; therefore, these data (Figures 5A–5C) should be interpreted cautiously.

Figure 5.

Postprandial glycemic and lipid metabolism

(A–C) Values are raw means (95% CI) and were analyzed using a linear mixed model with intervention (eTRE and CON), time (0, 1, 2, 3, and 4 h), study period (1 and 2), and their interactions as fixed effects and subject as a random effect (SAS v.9.4, SAS Institute). Compared with the CON schedule, during the eTRE schedule, participants had lower triglyceride concentrations (p intervention < 0.0001) (A) and greater glucose concentrations 1 h after the mixed meal (intervention-by-time interaction, p = 0.001, p adjusted = 0.006) (B). There was an intervention-by-time interaction for insulin (p interaction = 0.02) but no significant post hoc differences (C). The p values for intervention, time, and the intervention-by-time interaction are presented. Bonferroni corrections were used to adjust for multiple comparisons. There were no study period or intervention-by-study period (i.e., carryover effect) effects for triglyceride or glucose concentrations, but there was a intervention-by-study period effect (p = 0.003) for insulin. Insulin was log transformed prior to analysis to meet model assumptions. ∗p adjusted = 0.006.

For postprandial triglycerides, there were no differences in total AUC (p intervention = 0.68) between eating schedules. However, when analyzing the postprandial time course using a mixed linear model, eTRE had lower total triglyceride concentrations (p intervention < 0.0001) compared with the CON schedule. Similar findings for postprandial triglycerides have been reported in participants with obesity;³⁵ however, other eTRE studies in generally healthy participants with BMI categories ranging from normal weight to obese have found no impact of meal timing on triglycerides.^5,8,36 TRE is proposed to increase hepatic lipid oxidation, resulting in decreased triglyceride synthesis.³⁷ In the current study, there were no measurable differences in fasting respiratory quotient (RQ) values (eTRE: 0.80 [0.79, 0.82] and CON: 0.82 [0.81, 0.84]; p intervention = 0.11); however, eTRE led to lower RQ values over the entire 4-h postprandial measurement period (p intervention < 0.0001), suggesting that whole-body lipid oxidation is increased during eTRE (Figure 3D). This is in contrast to a previous trial in overweight and obese individuals that reported that eTRE reduces fasting but not daytime, non-protein RQ.⁷ Interestingly, Bao et al.¹⁵ reported no eTRE effect on fasting non-protein RQ, which is in line with our observations; however, contrary to our findings, they found that eTRE increased RQ during the daytime (i.e., during the eating window).

eTRE does not impact the plasma metabolome

Targeted metabolomics analyses were performed on plasma samples while fasted, and 2 and 4 h after a mixed meal challenge to explore potential impacts of eTRE on the tricarboxylic acid cycle and glycolysis intermediates. Using metabolomics to identify transient differences in circulating metabolites has the potential to provide insight into mechanisms driving glycemic improvements. In the current study, there were no significant differences between any of the 69 measured metabolites after false discovery rate (FDR) adjustment when data were analyzed using a linear mixed model (p intervention > 0.05, p intervention-by-time interaction > 0.05; Table S8). In a study conducted by Lundell et al.³⁸ comparing the serum metabolome over a 24-h period in participants practicing 8-h TRE for 5 days compared with a 14-h CON schedule, principal-component analysis did not result in a clear separation by time of day or intervention. However, enrichment analysis revealed a time-of-day-specific effect on serum amino acids, including leucine, isoleucine, and valine metabolism. Branched-chain amino acids are not detected in the liquid chromatography-mass spectrometry (LC-MS) method used in the current analysis, thereby precluding a comparison of these specific results with our study.

eTRE has no effect on the fecal microbiome

There were no differences in microbiome alpha (Shannon: p intervention = 0.85; Chao1: p intervention = 0.57) or beta diversity (Bray-Curtis dissimilarity index: p intervention = 1.0) and no differences in the ratio of Firmicutes to Bacteroidota ratio (p intervention = 0.15) between the eTRE and CON schedules (Figure 6). Differential abundance testing was performed using the ANOVA-like differential expression (ALDEx2) method for high throughput sequencing data, which revealed some genus-level differences between the CON and eTRE schedules, but these were not statistically significant after adjusting for FDR. The impact of TRE on the microbiome is equivocal, with trials using late TRE interventions^39,40 and eTRE⁴¹ reporting changes in diversity metrics and other trials^42,43 reporting no impact on microbiome diversity. Few studies have explored the impact of eTRE on the microbiome independent of changes in dietary intake and weight loss, making it difficult to separate the impact of weight loss on the microbiome⁴⁴ from the impact of TRE. Importantly, the current study investigated the effects of eTRE on microbiome diversity and composition while maintaining body mass and matching dietary intake and showed no significant differences between eTRE and the CON eating schedules (Figures 6 and S4).

Figure 6.

Fecal microbiome composition

(A–E) Compared with the CON schedule, eTRE had no impact on alpha diversity, measured by the Shannon and Chao1 indices with pairwise comparisons for intervention made using the Kruskal-Wallis test (A), or beta diversity, measured by principal-coordinates analysis (PCoA) plot of the bacterial community using Bray-Curtis distance and permutational multivariate ANOVA (PERMANOVA) (B). Bacterial composition at the major phylum and genus level during eTRE and CON schedules is shown (C). Differential abundance testing was performed using the ANOVA-like differential expression (ALDEx2) method, and 20 taxa with high effect value are shown (D). The ratio of Firmicutes to Bacteroidota did not differ (p intervention = 0.15) between the eTRE and CON schedules (E). See also Figure S4.

eTRE reduces PER1 and PER2

Circadian system-related genes were measured in cDNA isolated from mononuclear cells following an overnight fast. Fasting activates 5′ AMP-activated protein kinase (AMPK), leading to increased nicotinamide phosphoribosyltransferase (NAMPT) activity and subsequent activation of sirtuin 1 (SIRT1). When activated, SIRT1 binds with the CLOCK:BMAL1 complex and inhibits the transcription of Per2.⁴⁵ In the current study, eTRE reduced the absolute number of copies of PER1 (eTRE: 35.6 [16.7, 54.6] copies/μL compared with CON: 60.7 [38.1, 83.2] copies/μL; p intervention = 0.02) and PER2 (eTRE: 35.7 [19.8, 51.7] copies/μL compared with CON: 51.6 [35.2, 67.9] copies/μL; p intervention = 0.04; Figure S3). The expression of CLOCK (p intervention = 0.26), BMAL1 (ARNTL) (p intervention = 0.70), CRY1 (p intervention = 0.12), CRY2 (p intervention = 0.20), REV-ERBα (NR1D1) (p intervention = 0.41), and RORA (p intervention = 0.22) were not impacted by eTRE. These data conflict with another study reporting that eTRE has little impact on PER1 and PER2 when assessed in fasted mononuclear cells.⁵ Another study reported that eTRE increases the amplitude of BMAL1 (ARNTL) and PER2.⁴¹ The reasons for the disparate results are unclear but may be due to differences in PCR methods (digital vs. qPCR), study duration, and timing of blood draws. In addition, these findings should be interpreted with caution as only one time point was evaluated.

eTRE has no impact on sleep

At baseline, participants reported a mean sleep wake-up time of 0629 ± 70 min and bedtime of 2227 ± 61 min. Average sleep duration during the study was not different between the eTRE (474 [448, 501] min) and CON (471 [446, 495] min; p intervention = 0.31) schedules (Table S4). This is consistent with other TRE trials lasting up to 12 weeks.^28,42 However, in a 5-week weight loss trial, individuals following an energy-restricted eTRE schedule reported getting 30 min less sleep compared with their 7.4 h at baseline but having improved mood compared with the CON group.⁴⁶ Following an eTRE schedule while in energy balance does not seem to meaningfully impact sleep.

Limitations of the study

This study employed a well-controlled outpatient diet intervention and crossover study design that allowed investigation of the effects of eTRE on energy and macronutrient digestibility; metabolizable energy; whole-gut, small bowel, and colonic transit time; in vivo hydrogen production; subjective ratings of hunger and appetite; and cardiometabolic health in a free-living population. However, we acknowledge several study limitations that may warrant caution when interpreting the results. The length of the 3-day collection period may not have been adequate to detect intervention-related changes in gastrointestinal measures. However, in a study by Basolo et al.,¹² collecting stool for 3 days was long enough to detect changes in energy loss in stool following short-term periods of overfeeding and underfeeding.¹² Additional trials are needed to determine whether following an eTRE schedule for a longer period would alter intestinal nutrient absorption. Moreover, we did not control the order of foods consumed by the participants during the mixed-meal challenge, which may have affected postprandial metabolism observations. Small-scale studies suggest that consuming carbohydrates first in the meal period results in higher glucose peaks and greater glycemic variability.^47,48 While we assume that variability is reduced by participants eating foods and beverages in a similar order during each study arm in this crossover trial, we did not measure or instruct the order in which foods were consumed.

Insufficient power may have contributed to the null findings in the current study. However, for relative energy digestibility (primary outcome), actual variability (standard deviation: 1.5% in the CON group and 2.3% in the eTRE group) was similar to that used for the a priori sample size calculation (i.e., 2%). Furthermore, the current study had 92% power to detect a 1.3% difference between groups, but no difference between groups (0.02 ± 1.43%) was detected, resulting in a post hoc effect size of 0.01 and power of 5%. Furthermore, an effect size of 0.75 (for 80% power) to 0.97 (for 95% power) was necessary to detect differences between groups. Based on our post hoc analysis, we were adequately powered to detect the following differences between groups: 1.3% for metabolizable energy, 1.8% for fat digestibility, 1.9% for carbohydrate digestibility, 33 kcal/day for absolute energy loss in stool, and 15 kcal/day for absolute energy loss in urine. Protein digestibility, WGTT, and fasting triglycerides were less well powered than expected due to high variability for each measure.

Last, we only assessed gene expression while participants were fasted. Future studies should analyze circadian gene expression at time points across the 24-h cycle. The current study was conducted in young, healthy adults; therefore, results may not be generalizable to older adults with chronic diseases like obesity, type 2 diabetes, or cardiovascular disease.

Conclusion

In summary, following an eTRE schedule for 9 days had no impact on intestinal energy or macronutrient absorption, gastrointestinal transit time, or hydrogen gas production; however, we showed that eTRE lowers mean 24-h glucose concentrations, glycemic variability, and fasted blood glucose and insulin concentrations and improves insulin sensitivity in healthy, normal-weight adults. Importantly, we cannot identify whether the cardiometabolic improvements were related to the fasting duration or the meal timing. The combination of these two factors is likely complementary; however, future studies should examine the impact of various TRE schedules and fasting durations to optimize health outcomes and minimize disruptions of daily life. Importantly, we observed that interindividual intestinal absorption was highly variable, ranging from 89.7% to 94.3% (coefficient of variation: 1.7%) during the CON schedule. Therefore, future research should focus on identifying factors that predict individual energy digestibility to facilitate more precise energy intake recommendations.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
Human blood, urine, and stool samples	This paper	N/A
Critical commercial assays
Direct-zol RNA purification kit	Zymo Research, Irvine, CA, USA	R2050
High-Capacity cDNA Reverse Transcription Kit	Applied Biosystems; Waltham, MA, USA	4374966
QIAcuity Probe PCR Kit	Qiagen; Germantown, MD, USA	250102
TaqMan™ Gene Expression Assay (FAM), Circadian gene primer, Hs00154147_m1	ThermoFischer Scientific; Waltham, MA, USA	4453320
TaqMan™ Gene Expression Assay (FAM), Circadian gene primer, Hs00231857_m1	ThermoFischer Scientific; Waltham, MA, USA	4453320
TaqMan™ Gene Expression Assay (FAM), Circadian gene primer, Hs00242988_m1	ThermoFischer Scientific; Waltham, MA, USA	4453320
TaqMan™ Gene Expression Assay (FAM), Circadian gene primer, Hs01007553_m1	ThermoFischer Scientific; Waltham, MA, USA	4453320
TaqMan™ Gene Expression Assay (FAM), Circadian gene primer, Hs00172734_m1	ThermoFischer Scientific; Waltham, MA, USA	4453320
TaqMan™ Gene Expression Assay (FAM), Circadian gene primer, Hs00901393_m1	ThermoFischer Scientific; Waltham, MA, USA	4448892
TaqMan™ Gene Expression Assay (FAM), Circadian gene primer, Hs00253876_m1	ThermoFischer Scientific; Waltham, MA, USA	4453320
TaqMan™ Gene Expression Assay (FAM), Circadian gene primer, Hs00536545_m1	ThermoFischer Scientific; Waltham, MA, USA	4453320
QIAamp Powerfecal Pro DNA Kit	Qiagen, Valencia, CA, USA	51804
AMPure® XP magnetic beads	Beckman Coulter, Brea, CA, USA	A63881
Deposited data
All data used in figures and supplemental figures are available to download as the source data file	Mendeley Data, V1	Mendeley Data: https://doi.org/10.17632/wkhmz338h2.1
Metabolomics data	National Metabolomics Data Repository	Metabolomics Workbench: http://dx.doi.org/10.21228/M8672X
Earth Microbiome Project benchmarked protocol	http://www.earthmicrobiome.org	N/A
Software and algorithms
SAS	SAS Institute Inc., Cary, NC	Version 9.4
GraphPad Prism	San Diego, CA, USA	Version 9.1.2
Dual energy X-ray absorptiometry software	Hologic Inc., Bedford, MA	Apex
Digital PCR	QIAcuity Software Suite	Version 2.1
Miseq Control Software and Miseq Reporter	Illumina Inc	Version 2.5
QIIME2 (Quantitative Insights into Microbial Ecology)	www.qiime2.org	Version 2–2022.8
DADA2	Callahan et al., 201649	Version 1.20.0
LabFront	https://www.labfront.com/	N/A
Other
Stadiometer	Seca, Chino, CA	217
Digital scale	A&D Medical, San Jose, CA	352BLE
Blood pressure monitor	Omron, Kyoto, Japan	HEM-705CPN
Dual energy X-ray absorptiometry	Hologic Inc., Bedford, MA	Horizon W
Garmin vivosmart 4 watches	Garmin, Olathe, KS	010-01995-12
Continuous glucose monitor/reader	Abbott, Alameda, CA, USA	Freestyle Libre Pro/Freestyle Libre Pro Reader
Resting energy expenditure/thermic effect of food	Parvo Medics, Salt Lake City, UT	2400 TrueOne
Brilliant blue FCF	Millapore Sigma, Burlington, MA	80717
Commercial blender	Waring, McConnellsburg, PA	CB15
Freeze dryer	Labconco, Kansas City, MO	701211250
Bomb calorimeter	Parr Instrument Co., Moline, Illinois	6200
Oxygen bomb	Parr Instrument Co., Moline, Illinois	1108
Pellet press	Parr Instrument Co., Moline, Illinois	2811
Gas sensing capsule	Atmo Biosciences; Victoria, Australia	N/A
QIAcuity One Digital PCR System	Qiagen; Germantown, MD, USA	911001
Nanodrop spectrophotometer	Thermo Scientific	ND-ONEC-W
Qubit-4 fluorimeter	InVitrogen, Carlsbad, CA, USA	Q33238
Illumina MiSeq	Illumina Inc., San Diego, CA, USA	SY-410-1003
DxC 600	Beckman Coulter; Brea, CA	A27318
Immulite	Siemens Corporation; Washington, DC	2000

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Claire Berryman (claire.berryman@pbrc.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •Data used in figures and supplemental figures are available to download at Mendeley Data and are publicly available as of the date of publication. The DOI is listed in the key resources table.
• •Metabolomics data are available to download at the Metabolomics Workbench and are publicly available as of the date of publication. The DOI is listed in the key resources table.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and study participant details

Human participants

This study was approved by the Florida State University Institutional Review Board and registered on clinicaltrials.gov (NCT04877262) and reporting followed the CONSORT checklist (Table S7). Healthy adults who were normal-weight (BMI:18.5–24.9 kg/m²) or overweight (BMI: 25–29.9 kg/m²) and aged 20-45 years, were recruited from the greater Tallahassee, Florida area between December 2021 and May 2022 and provided informed consent prior to participating. Potential participants were excluded if they had any disease or condition known to interfere with metabolism or normal gastrointestinal function (diabetes, cardiovascular disease, kidney disease, prior bariatric surgery, suspected or known fistulas, gastrointestinal obstruction, gastrointestinal disease, colonoscopy within 3 months of the study, alcoholism, substance abuse disorders); irregular menstrual cycle in the past 6 months; weight fluctuations greater that 5% during the previous 6 months; antibiotic use within 3 months of participation; allergies or intolerances to foods included in the controlled diet; habitual use of laxatives, stool softeners, or anti-diarrheal medications (≥1x/week); whole-gut transit time >72 h; or were pregnant or lactating.

Method details

Study design

This randomized, crossover, controlled feeding trial consisted of two 9-day weight maintenance diet periods separated by at least 2 weeks for male participants and 3-weeks for female participants, as illustrated in Figure 1A. Resting energy expenditure was measured during the screening visit after a 12-h fast with no exercise or caffeine consumption for 12 and 24 h, respectively, using open-circuit, indirect calorimetry (2400 TrueOne, Parvo Medics, Salt Lake City, UT, USA). Participants recorded their physical activity for the 7 days preceding the screening visit and agreed to maintain similar exercise habits during the study. Based on self-reported physical activity, physical activity coefficients were estimated for each participant (1.2 = sedentary, little or no exercise and desk job; 1.38 = lightly active, light exercise or sports 1–3 days/week; 1.55 = moderately active, moderate exercise/sports 3–5 days/week; 1.73 = very active, hard exercise or sports 6–7 days/week; 1.9 = extremely active, hard daily exercise/sports and physical job) and multiplied by their resting energy expenditure to determine weight maintenance energy requirements. Female participants began each study arm in the mid-follicular phase of their menstrual cycle to control for the effects of reproductive hormones on gastrointestinal function and other outcome measures. Participants were randomized to follow the early time-restricted eating (eTRE) schedule (i.e., all meals consumed in a 6-h window between 0800 and 1400) and a control eating schedule (i.e., all meals consumed in a 12-h window between 0800 and 2000) for 9-days. Following a 3-day acclimation period (days 1–3) to the study diet and eating schedule, participants collected all stool and urine for a 3-day collection period to determine stool and urine energy loss. A non-absorbable dye marker (Brilliant Blue FCF, Millipore Sigma, Burlington, MA, USA) was used to mark the beginning (provided the morning of day 4) and end (provided the morning of day 7) of each collection period. The energy content of the controlled diet, stool, and urine samples was measured by bomb calorimetry. Participants maintained a physical activity log during the first study arm and replicated the activities during the second study arm. In addition, physical activity was monitored using Garmin vivosmart watches (Garmin, Olathe, KS, USA), which were worn continuously during each study period. Participant data were synced to the Labfront and Garmin applications.

Participants spent 1 h at the laboratory on study days 4, 5, 7, 8, and 9 to consume breakfast under supervision, review meal intake, meal timing, and physical activity logs, and return biospecimens. Participants were allowed to eat remaining snacks and meals at home. On study day 6, participants arrived in the morning to the laboratory for a mixed-meal tolerance test (MMTT). Participants spent 6 h in the laboratory to measure REE, fasting and postprandial blood metabolite, lipid, and insulin concentrations, and thermic effect of food (TEF) in response to the MMTT.

Controlled diet

Participants were provided a controlled diet that matched their weight maintenance energy needs (55% carbohydrate, 15% protein, 30% fat). Energy needs were determined by applying a physical activity factor to REE measured during the screening visit. After a 3-day acclimation period, participants consumed the same 1-day menu during each study period, only varying in the times of day the food was consumed (Figure 1B). Meals and snacks were consumed at pre-specified times during the day for each study arm, but participants were free to consume foods in their preferred order (e.g., for the afternoon snack, participants could eat all the carrots, followed by the hummus or consume the carrots and hummus together). All foods and beverages were measured to 0.1 g and preparation and packaging was supervised by a Registered Dietitian. Participants were required to eat all meals and snacks provided and refrain from consuming non-study foods, except for a calorie-free beverage of choice each day (up to 12 oz/day). Participants recorded the type and amount of calorie-free beverage they consumed each day during study period 1 and then matched consumption of the type and amount of calorie-free beverage during study period 2. Water was provided (32.6 mL/kg) and matched for each study period. Participants ate breakfast under supervision, documented the start and end time of each meal and snack, and returned all food containers to the lab.

Mixed meal tolerance test

The MMTT was completed on day 6 during each study period (Figure 1A). Participants arrived at the laboratory on day 6 at 0630. Participants had weight and REE measured and fasting blood drawn before consuming the MMTT. The MMTT consisted of the day 6 breakfast and morning snack (Figure 1B) and was provided based on individualized weight maintenance energy needs [731 kcal (650, 812); 60% carbohydrate, 16% fat, 28% protein]. Blood was drawn at 1, 2, 3, and 4 h from the start of the MMTT. Similarly, TEF was measured for 15 min starting at 15, 45, 75, 105, 165, and 225 min from the start of the MMTT.

Gross, digestible, and metabolizable energy determination

Meals and snacks were prepared in duplicate during the collection phase. One diet was provided to the participant to consume and the other was used to measure the gross energy of the diet via bomb calorimetry. The energy of the diet, stool, and urine was determined as previously described.⁵⁰ The diet was weighed and then blended in a commercial blender (Waring, McConnellsburg, PA, USA). Twenty-five-gram aliquots were frozen at −80°C, freeze-dried (Freezone 12L, Labconco, Kansas City, MO, USA), and pelleted into 1-g pellets (Parr Instrument Co., Moline, Illinois, USA) before the energy density was determined using a 6200 isoperibol calorimeter with an 1108 oxygen bomb (Parr Instrument Co., Moline, Illinois, USA).

Before the first meal on day 4 and day 7, a blue dye marker (Brilliant Blue FCF, Millipore Sigma) was ingested by participants to mark the start and end of fecal collections, respectively. The stool was collected, aggregated, and weighed from the appearance of the first marker through the appearance of the second dye marker. The stool was blended with an equivalent amount of double-distilled water before 50 g aliquots were freeze-dried (Freezone 12L, Labconco, Kansas City, MO, USA) for approximately 72 h. The freeze-dried stool was weighed (dry weight), pelleted into 1-g pellets, and combusted via bomb calorimetry. Digestible energy was defined as:

$$\text{Digestible}\text{energy}\left(\text{kcal}\right)=\text{gross}\text{energy}\text{of}\text{the}\text{diet}\left(\text{kcal}\right)-\text{stool}\text{energy}\text{loss}\left(\text{kcal}\right)\text{.}$$

Digestibility was calculated as:

$$\text{Digestibility}\left(\%\right)=100-\left[\text{stool}\text{energy}\left(\text{kcal}\right)/\text{gross}\text{energy}\text{of}\text{the}\text{diet}\left(\text{kcal}\right)\right]\mathrm{x}100\text{.}$$

Urine was collected for 72 h (days 4, 5, and 6) using 3.5 L collection containers (Simport, Quebec, Canada). The total urine collection was mixed, weighed, aliquoted (600 mL), and stored in a −80°C freezer (Fisher Scientific, Hanover Park, IL, USA) until analyzed. The urine was freeze-dried (Freezone 12L, Labconco, Kansas City, MO, USA) for approximately 1 week. Freeze-dried urine was weighed, pelleted into 1-g pellets, and combusted via bomb calorimetry. Metabolizable energy (ME) was defined as:

$$\text{ME}\left(\text{kcal}\right)=\text{gross}\text{energy}\text{of}\text{the}\text{diet}\left(\text{kcal}\right)–\text{stool}\text{and}\text{urine}\text{energy}\text{loss}\left(\text{kcal}\right)\text{.}$$

$$\text{ME}\left((,\%,)\right)=100-\left[\text{stool}\text{and}\text{urine}\text{energy}\right(\text{kcal})/\text{gross}\text{energy}\text{of}\text{the}\text{diet}(\text{kcal}\left)\right]\times 100.$$

One-gram freeze-dried pellets from the controlled diet, stool, and urine were run in quadruplicate. CVs ranged from 0.13 to 1.5% for the diet samples, 0.10 to 3.3% for the stool samples, and 0.23 to 5.5% for the urine samples. The total kcals for the controlled diet, stool, and urine were calculated as the number of kcals measured via bomb calorimetry multiplied by the total dry weight of the sample. Nitrogen in the diet, stool, and urine was measured by the Dumas method (Eurofins, Madison, WI, USA). Fat in the diet and stool were extracted, saponified with 0.5N methanolic sodium hydroxide, and methylated with 14% BF3-methanol. The resulting methyl esters of the fatty acids were extracted with heptane. The methyl esters of the fatty acids were analyzed by gas chromatography (Eurofins, Madison, WI, USA; Table S6). Total fiber was determined by the enzymatic-gravimetric method. Ash was determined by electric furnace and moisture was measured by vacuum drying (Eurofins, Madison, WI, USA). Total carbohydrates of the diet and stool were calculated as:

$$\text{Carbohydrate}\left(\mathrm{g}\right)=100-\left[\text{protein}\left(\mathrm{g}\right)+\text{fat}\left(\mathrm{g}\right)+\text{ash}\left(\mathrm{g}\right)+\text{moisture}\left(\mathrm{g}\right)\right]$$

Macronutrient digestibility was calculated as:

$$\text{Macronutrient}\text{digestibility}\%=100-\left[\text{macronutrient}\text{in}\text{stool}\left(\mathrm{g}\right)/\text{macronutrient}\text{in}\text{diet}\left(\mathrm{g}\right)\right]\mathrm{x}100\text{.}$$

Nitrogen balance was calculated as:

$$\text{Nitrogen}\text{balance}\left(\mathrm{g}/\mathrm{d}\right)=\left(\text{protein}\text{intake}\left(\mathrm{g}/\mathrm{d}\right)/6.25\right)–\left(\text{urinary}\text{nitrogen}\left(\mathrm{g}/\mathrm{d}\right)/0.8+2.5\mathrm{g}\right)\text{,}$$

where 2.5 is the assumed sum of fecal and integumental nitrogen loss.

Whole-gut transit time and stool ratings

Whole-gut gastrointestinal transit time was measured once during screening and twice (starting on day 4 and day 7) during each study period and calculated as the difference in time between the ingestion of the blue dye marker and the first appearance of the dye in stool. The two measurements during each study period were averaged to get whole-gut transit time. Participants rated each bowel movement 1–7 according to the Bristol Stool scale.⁵¹

Small bowel transit time, colonic transit time, and hydrogen production

Participants ingested a gas-sensing capsule (Atmo Biosciences; Victoria, Australia) after breakfast on study day 5. The Atmo Gas Capsule measures luminal gas concentrations (hydrogen and oxygen) through a gas-permeable membrane and relays the data to a hand-held receiver. Segmental transit times are determined using oxygen as a marker of intestinal location.²⁶ While the Atmo Gas Capsule can measure gastric emptying time, the protocol involves extended fasting (6h) after participants ingest the capsule with a meal, which would have prevented participants from adhering to the eTRE and CON meal schedules; therefore, gastric emptying time is not reported. The hydrogen data are presented as the percentage relative to all other gas constituents comprising the intestinal tract. The hydrogen data is presented as quartiles based on the total time from capsule ingestion to excretion.

Hunger and appetite assessment

Participants completed 100 mm visual analog scales assessing subjective hunger, fullness, capacity to eat, and desire to eat at four time points during study days 4 through 7 (0745, 1200, 1700, and 2100). A composite satiety score (CSS)^52,53 was calculated as:

$$\text{CSS}=(\text{fullness}+\left(100–\text{desire}\text{to}\text{eat}\right)+\left(100–\text{hunger}\right)+\left(100–\text{capacity}\text{to}\text{eat}\right)/4$$

Anthropometric measurements

Height was measured to the nearest 0.1 cm using a stadiometer (Seca 217, Chino, CA, USA) during the screening visit. Body mass was measured to the nearest 0.1 kg on the morning of screening, study day 1 and days 4–9 using a calibrated digital scale (A&D Medical wireless weight scale UC-352BLE, San Jose, CA, USA). Body weight change is reported as the difference between the final body weight and day 4 body weight (beginning of collection period). Body composition (fat-free mass, fat mass, and body fat %) was measured at baseline using dual-energy X-ray absorptiometry (DEXA, Discovery W, Hologic Inc., Bedford, MA, USA) by a certified technician. Blood pressure was measured at baseline using an automated blood-pressure monitor (Omron HEM-705CPN, Kyoto, Japan).

Resting energy expenditure and thermic effect of food

Resting energy expenditure (REE) was measured after a 12-h fast with no exercise or caffeine consumption for 12 and 24 h, respectively, using open-circuit, indirect calorimetry (2400 TrueOne, Parvo Medics, Salt Lake City, UT, USA) at baseline and on the morning of day 6. Participants rested in the supine position in a quiet, dimly lit, thermoneutral room for 30 min before REE measurements. REE was assessed using a clear, ventilated hood placed over the participant’s head. Respiratory gas exchange (oxygen consumption and carbon dioxide production) was measured for 25 min. The last 15 min of the test were analyzed after the participant had reached a steady state. REE was calculated using the Weir equation. The thermic effect of food (TEF) was measured using the same open-circuit, indirect calorimetry methods described above. Gas exchange and TEF were measured for 15 min starting at 15, 45, 75, 105, 165, and 225 min after the start of the MMTT in a semi-recumbent position (Figure 1A). TEF is reported at each timepoint, as the incremental area under the curve, and as peak increase in energy expenditure. Respiratory quotient (RQ) was measured as the volume of carbon dioxide exhaled divided by the volume of oxygen inhaled. The mean RQ for 15 min during the REE testing period and each 15-min increment during the postprandial testing is reported.

Continuous glucose monitoring

Participants wore a continuous glucose monitor (Freestyle Libre Pro, Abbott, Alameda, CA, USA) from study day 1 through study day 9. The CGM sensor was applied to the subcutaneous adipose tissue of the tricep area, per the manufacturer’s instructions. Glucose measures collected from midnight (00:00) on study day 4 through 23:59 on study day 5 (48 h total) were used to analyze average 24-h glucose levels. Data from study day 6 were excluded because the participants adhered to the eTRE or CON eating window (i.e., 6-h or 12-h), but were not able to adhere to the prescribed mealtimes due to the MMTT. Total area under the curve (AUC) for glucose and mean 24-h, daytime, and nighttime glucose were calculated from glucose measures collected every 15 min from study day 4 through study day 5 (as described above). Daytime was defined as 0800–2000 and nighttime as 2001-0759. Glucose variability is reported as mean amplitude of glycemic excursions (MAGE).⁵⁴

Blood collection

Trained clinical staff inserted a venous catheter (Becton Dickinson, Franklin Lakes, NJ, USA) into the antecubital vein following REE measurements on day 6. Venous blood samples were collected while fasted, and then every hour for 4 h following a MMTT. Whole blood was collected in EDTA, cell preparation, and serum separating vacutainer tubes (BD Vacutainer, Franklin Lakes, New Jersey, USA). The serum was allowed to clot at room temperature. Blood samples were centrifuged at 3000 RPM for 15 min at 4°C before aliquots were frozen at −80°C until analysis.

Serum and urinary measures

Serum glucose, cholesterol, and triglycerides were measured on a DXC600 instrument (Beckman Coulter; Brea, CA) using standard reagents (WAKO Chemicals USA; Richmond, CA). Low-density lipoprotein cholesterol (LDL) was calculated using the Friedewald equation.⁵⁵ Serum insulin was measured using chemiluminescent immunoassays on an Immulite 2000 instrument (Siemens Corporation; Washington, DC). Insulin resistance was estimated using the homeostatic model of insulin resistance (HOMA-IR), where HOMA-IR = fasting glucose (mmol $·$ L⁻¹) x insulin (μIU $·$ mL⁻¹)/22.5.⁵⁶ Urinary C-peptide concentrations were measured using the Immulite 2000 C-peptide assay (Siemens Healthineers, Erlangen, Germany). Urinary beta-hydroxybutyrate concentrations were measured using procedures validated by Pennington Biomedical Research Center. Briefly, β-Thionicotinamide adenine dinucleotide, a phosphate buffer, and 3-hydroxybutyrate dehydrogenase were added to urine samples sequentially and beta-hydroxybutyrate was assayed by measuring the rate of Thio-NADH production spectrophotometrically.

Circadian-related gene expression

Whole blood was collected while participants were fasted in Vacutainer Mononuclear Cell Preparation Tubes (BD Vacutainer, New Jersey, USA). Mononuclear cells were isolated per the manufacturer’s instructions and stored in Trizol at −80°C until analysis. RNA was isolated using the Direct-zol RNA purification kit (Zymo Research; Irvine, CA, USA). Total RNA was quantified by NanoDrop spectrophotometer (NanoDrop; Thermofisher; Waltham, MA, USA) and reverse transcribed to cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems; Waltham, MA, USA). Gene expression was assessed by digital PCR using the QIAcuity Probe PCR Kit and QIAcuity One Digital PCR System (Qiagen; Germantown, MD, USA). Commercially available TaqMan primers were used to analyze circadian genes: CLOCK (Hs00231857_m1), BMAL1 (ARNTL) (Hs00154147_m1), PER1 (Hs00242988_m1), PER2 (Hs01007553_m1), CRY1 (Hs00172734_m1), CRY2 (Hs00901393_m1), REV-ERBα (NR1D1) (Hs00253876_m1), RORA (Hs00536545_m1) (ThermoFischer Scientific; Waltham, MA, USA). Results are presented as copies/μL.

Plasma metabolomics

Whole blood was collected while participants were fasted, and 2-and 4-h after a mixed meal challenge in EDTA vacutainer tubes (BD Vacutainer, New Jersey, USA). Blood samples were centrifuged at 3000 RPM for 15 min at 4°C and plasma separated before aliquots were placed in Eppendorf Tubes pre-rinsed with 100 μL methanol and frozen at −80°C until analysis. Targeted metabolomics was run on plasma samples using protein precipitation extraction with ultra-performance liquid chromatography tandem quadruple mass spectrometry (UPLC-MS), as described previously.⁵⁷ Plasma (25 μL) was added to 1.5 mL Eppendorf tubes prior to the addition of 10 μL of 1 μM internal standard solution. Next, 750 μL chilled methanol was added and samples were vortexed 30 s before centrifugation at 15,000 × G for 10 min. Subsequently, the supernatant was collected and added to 1.5 mL high-performance liquid chromatography (HPLC) amber glass vials, dried, and reconstituted in 100 μL 3:1 acetonitrile: methanol containing 1-cyclohexyl-ureido, 3-dodecanoic acid (CUDA; Sigma-Aldrich, St. Louis, MO, USA) solution. The solution was vortexed 30 s, transferred to microfilter tubes, and centrifuged at 10,000 × G for 3 min prior to transfer to an HPLC vial. Using a Waters Acquity I-Class UPLC (Waters, Milford, MA, USA), metabolites were separated using a 150 × 2.0 mm Luna NH2 column (Phenomenex, Torrance, CA). Analysis was performed on an API 4000 QTRAP (Sciex, Framingham, MA) using negative ion mode electrospray ionization and multiple reaction monitoring. Quantification was performed using AB Sciex MultiQuant version 3.0.

Fecal microbiome analysis

Fecal microbiome profiles were determined as per our previously described method.^{58,59,60,61,62} Briefly, the subjects were provided with a fecal sample collection kit, along with detailed instructions for sample collection on study day 6. Stool samples were maintained at <40°F until the participant returned to the lab within 24 h of producing the sample. Upon receipt at the lab, the specimens were stored immediately at −80°C until further processing. High-quality genomic DNA was extracted using the QIAamp Powerfecal Pro DNA Kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions and was quantified using a Nanodrop spectrophotometer (Thermo Scientific). Microbiome composition was analyzed in accordance with the Earth Microbiome Project benchmarked protocol (http://www.earthmicrobiome.org) by using a barcoded high-throughput sequencing approach. The V4 hypervariable region of the microbial 16S ribosomal RNA gene was PCR-amplified; the resulting amplicons were cleaned up using AMPure XP magnetic beads (Beckman Coulter, Brea, CA, USA); quantified using Qubit-4 fluorimeter (InVitrogen, Carlsbad, CA, USA) and dsDNA HS assay kit (Life Technologies, Carlsbad, CA, USA); and the final amplicon libraries were made as described previously.⁶³ The purified PCR products were pooled in equimolar concentrations and sequenced on 2x300-bp Illumina MiSeq run (Illumina Inc., San Diego, CA, USA) for paired-end sequencing. To avoid the bias of variation in DNA extraction, or PCR reaction conditions and primers on microbial community composition recovered by amplicon sequencing, all samples were batch-processed and batch-analyzed. The sequencing quality control was executed with on-board Miseq Control Software and Miseq Reporter (Illumina Inc.) and the obtained sequences were de-multiplexed, quality-filtered, clustered, and annotated using the QIIME2 (Quantitative Insights into Microbial Ecology) software package (www.qiime2.org; ver. 2–2022.8).⁶⁴ Raw sequences were trimmed and denoised with DADA2⁴⁹ using the q2-dada2 plugin.^64,65 All produced amplicon sequence variants (ASVs) were aligned with the multiple sequence alignment program MAFFT.⁶⁶ The classification of ASVs was accomplished using a naive Bayes taxonomy classifier,⁶⁵ which was developed for the sklearn classifier against pre-built from the 99% SILVA 138 database.^65,67 Shannon and Chao1 index were used as alpha-diversity metrics, and Bray-Curtis dissimilarity index was used for the PCoA analysis of beta-diversity. To identify differentially abundant taxa between CON and eTRE, the linear discriminant analysis (LDA) effect size (LEfSe),⁶⁸ ANOVA-like differential expression (ALDEx2) method,⁶⁹ and multivariate analysis by linear models (MaAsLin2)⁷⁰ were used. All p values were corrected using the Benjamini–Hochberg procedure. LEfSe cladogram was generated using differential features with LDA>2.0. Spearman’s correlation was used to identify the correlation between CON and eTRE.

Quantification and statistical analysis

Sample size calculation

We assumed 2% (∼40–60 kcal/day) was the minimal difference in energy digestibility that would impact metabolic health (i.e., clinically relevant). We based the standard deviation estimate of 2% on the study by Jumpertz and colleagues who reported that mean relative stool energy loss was 4.9 ± 1.8% when lean participants consumed a 2400-kcal diet and 3.8 ± 1.1% when those same participants consumed a 3400-kcal diet.¹³ The sample size calculation indicated that 16 participants provided 96% power to detect a 2% difference in energy digestibility (primary outcome) between eTRE and the control eating schedule, with a standard deviation of 2% and an α of 0.05.

Randomization and blinding

The randomization scheme was conducted using balanced permutations to stratify by gender and generated from an online platform (randomization.com). Research staff and participants were not blinded to the intervention due to the need for research staff to instruct participants on meal timing and participants to follow prescribed eating windows. In addition, none of the analyses were blinded due to staffing constraints.

Statistical analyses

Statistical analyses were performed in SAS (version 9.4; SAS Institute Inc., Cary, NC) and GraphPad Prism (version 9.1.2; San Diego, CA, USA) with significance set at α = 0.05. Normality for each variable was assessed using the univariate procedure to evaluate skewness and kurtosis and visually inspect the distribution. Normally distributed data are presented as mean (95% CIs) and non-normally distributed data as geometric mean (95% CIs). Segmental transit time and hydrogen production data are presented as median (interquartile range) and were analyzed with the Wilcoxon test. Differences in dietary intake between intervention groups were analyzed using a paired t-test. Incremental AUC (iAUC) was calculated for TEF over the postprandial period using the trapezoidal rule. All areas below the baseline were excluded from iAUC calculations. Total AUC was calculated for 24-h CGM data and postprandial blood measures using the trapezoidal rule. Data that was not normally distributed was log-transformed before analysis. Differences in dependent variables between the meal schedules were evaluated using linear mixed models with compound symmetry. Intervention (eTRE or control), study period (1 or 2), and their interaction were considered fixed effects and participant was treated as a random effect.

Further, time series data (subjective ratings of hunger and appetite, postprandial metabolites, and TEF) were analyzed with a mixed effect linear model with intervention (eTRE and control), time point, study period (1 and 2), and their interactions as fixed effects and subject as a random effect and using the autoregressive covariance structure. The Bonferroni-adjustment for multiple comparisons was used when a significant interaction effect was present. For metabolomics analyses, multiple comparisons were adjusted using the Benjamini-Hochberg procedure⁷¹ with significance accepted at p < 0.05 to control the false discovery rate (FDR).

Additional resources

This study was registered on clinicaltrials.gov (NCT04877262).

Published: January 16, 2024

Supplemental information

Table S8. Fasting and postprandial metabolomics in response to a mixed-meal tolerance test during the CON or eTRE schedule, related to Figure 5

Values are raw means ± SD and were analyzed using linear mixed models in SAS (version 9.4; SAS Institute, Cary, NC) with intervention (eTRE or control), study period (1 or 2), time (0, 2, or 4 h), and their interactions considered fixed effects and participant treated as a random effect. All metabolites were log-transformed for statistical analysis. Multiple comparisons were adjusted by the Benjamini-Hochberg procedure with significance accepted at P < 0.05 to control the false discovery rate. Abbreviations: CON, control; eTRE, early time-restricted eating; Int; intervention.