Early Time-Restricted Eating Affects Weight, Metabolic Health, Mood, and Sleep in Adherent Completers: A Secondary Analysis

Abstract

Objective.

Data are mixed on whether intermittent fasting improves weight loss and cardiometabolic health. Here, we analyzed the effects of time-restricted eating (TRE) in participants who consistently adhered ≥5 days/week every week.

Methods.

Ninety patients aged 25–75 years old with obesity were randomized to early TRE (eTRE; 8-hour eating window from 07:00–15:00) or a control schedule (≥12-hour window) for 14 weeks. We performed a per-protocol analysis of weight loss, body composition, cardiometabolic health, and other endpoints.

Results.

Participants who adhered to eTRE ≥5 days/week every week had greater improvements in body weight (−3.7 ± 1.2 kg; p=0.003), body fat (−2.8 ± 1.3 kg; p=0.04), heart rate (−7 ± 3 beats/min; p=0.02), insulin resistance (−2.80 ± 1.36; p=0.047), and glucose (−9 ± 5 mg/dl; p=0.047) relative to adherers in the control group. They also experienced greater improvements in mood, including fatigue and anger; however, they self-reported sleeping less and taking longer to fall asleep.

Conclusions.

For those who can consistently adhere at least 5 days/week, eTRE is a valuable approach for improving body weight, body fat, cardiometabolic health, and mood. Further research is needed to determine whether eTRE’s effects of shortening sleep but reducing fatigue are healthful or not.

INTRODUCTION

Intermittent fasting involves cycling periods of eating and fasting for typically 14–48 hours at a time. In animals, intermittent fasting has a wide range of benefits: it decreases body weight and/or body fat, improves insulin sensitivity, lowers glucose levels, decreases blood pressure, improves cardiovascular health, slows cellular aging, slows cancer progression, enhances cognition, slows the progression of neurodegenerative conditions, and even extends lifespan (1, 2, 3, 4).

Among the various types of intermittent fasting, daily intermittent fasting, now known as time-restricted eating (TRE), has become popular. In humans, we define TRE as eating within a consistent daily window of 10 hours or less and fasting for the remaining ≥14 hours of the day. Animal studies suggest that TRE can prevent and reverse diet-induced obesity and improve several cardiometabolic endpoints, including glucose tolerance, cholesterol levels, and blood pressure (5, 6, 7). In humans, a handful of tightly-controlled efficacy studies report that TRE improves cardiometabolic endpoints such as insulin sensitivity, blood pressure, and oxidative stress even when calorie intake is identical to the control group (8, 9, 10, 11, 12). However, studies testing the effectiveness of TRE for losing weight and improving cardiometabolic health in free-living participants are mixed. Several initial studies found that TRE reduces body weight by 1–4% over several weeks (13, 14, 15, 16, 17, 18, 19, 20); however, more recent effectiveness trials are about evenly mixed (21, 22, 23, 24, 25, 26, 27, 28, 29) and include two prominent larger clinical trials with null findings (21, 22).

One consistent finding is that adherence to TRE is moderate to high, usually averaging 5.0–6.4 days/week (71–91% adherence) (13, 14, 15, 17, 18, 21, 30, 31). However, because there is a range of adherence to any intervention, it is important to also determine the effects of TRE in people who adhere. The most common analytical method, intention-to-treat (ITT) analysis, cannot answer this question; it estimates the effects of prescribing a treatment regardless of whether participants adhered. By contrast, per-protocol analyses evaluate the effects in people who do adhere (32). Both analytical methods have advantages and disadvantages (32). ITT analyses better represent the effectiveness of prescribing an intervention but tend to underestimate the true effects of receiving the treatment (33) because they include non-adherent participants and often involve imputation, which can introduce additional statistical noise. By comparison, per-protocol analyses tend to better estimate the efficacy of a treatment, but because the subgroups are not due to randomization, they can be affected by selection and other biases, which can inflate the estimated effects. While researchers and clinicians generally prefer ITT analyses to determine the effectiveness of a public health recommendation, research suggests that the general public usually prefers knowing the per-protocol effects (33). Yet only one prior small study on TRE has performed a per-protocol analysis (34).

Here, we performed the largest per-protocol analysis of TRE to date. We used data from a relatively large randomized controlled weight-loss trial comparing TRE versus eating over a ≥12-hour period. In the parent study, we investigated a form of TRE called early TRE (eTRE), which involves eating early in the day to align with circadian rhythms in metabolism (35). In the ITT analysis, eTRE was more effective than eating over a ≥12-hour window for losing weight, lowering diastolic blood pressure, and improving mood. Here, we performed a per-protocol analysis among completers who consistently adhered to their assigned eating windows at least 5 days/week every week of the intervention. This threshold was inspired by two rodent studies reporting that practicing TRE 5 days per week improved several cardiometabolic endpoints, albeit to lesser degrees than practicing TRE daily (6, 36). We also chose a threshold of ≥5 days/week because it may be more realistic than higher thresholds (e.g., ≥6 days/week), and it is easier for the general public to grasp than percent adherence (e.g., 80%). In the parent study, we hypothesized that eTRE would be more effective for losing weight and body fat and improving cardiometabolic health than eating over a ≥12-hour window. We expected the magnitudes of any benefits and harms to be larger in the per-protocol analysis.

METHODS

Study Overview.

The study protocol has been described in detail elsewhere (28). In brief, the parent study was a 14-week parallel-arm, randomized controlled weight-loss trial. Enrollment was limited to new patients at the Weight Loss Medicine Clinic at the University of Alabama at Birmingham (UAB) Hospital. Patients were eligible if they had a body mass index (BMI) of 30.0–60.0 kg/m², were aged 25–75 years old, were not taking weight loss medication, and did not have diabetes or a severe or unstable medical condition. Patients were excluded if they performed overnight shift work, regularly ate within a less than 10-hour daily period, or regularly ate dinner before 18:00. The study was approved by UAB’s Institutional Review Board (protocol number 300001207) and preregistered on ClinicalTrials.gov (NCT03459703), and all participants provided written informed consent before participating.

Interventions.

Participants were randomized to practice eTRE by eating within an 8-hour window between 07:00–15:00 or to a control (CON) eating schedule, which involved eating over a self-selected ≥12-hour period. The control group was designed to mimic meal-timing patterns of the median American (37). Participants were instructed to follow their assigned eating schedule at least 6 days/week. Participants were randomized to the two groups in a 1:1 allocation ratio, with stratification by biological sex, race (Black vs. not Black), and baseline physical activity level (≤2 days/week vs. ≥3 days/week of exercise), using block sizes of 2. Aside from the eating schedules, we matched all other intervention components across groups.

Both groups received weight-loss counseling involving energy restriction (ER) through the UAB Weight Loss Medicine Clinic. Henceforth, we refer to the two groups as the CON+ER and eTRE+ER groups. Participants met one-on-one with a registered dietitian at baseline and weeks 2, 6, and 10. Participants were counseled to reduce their energy intake by 500 kcal/day below their resting energy expenditure, as measured by indirect calorimetry (ReeVue Indirect Calorimeter, KORR Medical Technologies, Inc.; Salt Lake City, UT). Sedentary participants (defined as ≤2 days/week of structured physical activity) were instructed to exercise 15–25 minutes/day for 5 days/week at 50–60% of their age-predicted maximal heart rate, while active individuals were counseled to exercise 30 minutes/day for 5 days/week at 60–80% of their age-predicted maximal heart rate. Participants were also instructed to attend at least 10 weekly group classes at the clinic on healthy eating, exercise, and behavioral topics.

Outcome Measures.

The co-primary outcomes were weight loss and fat loss and were supplemented by additional measures of body composition. The secondary outcomes were fasting cardiometabolic risk factors, including blood pressure, heart rate, lipids, glucose, insulin, HbA1c, and indices of insulin resistance and β-cell function. We also assessed food intake, appetite, eating behaviors, physical activity, mood, and sleep as tertiary outcomes. All outcomes were assessed during a single baseline testing visit and post-intervention testing visit, which were conducted in the morning following a ≥12-hour water-only fast. The only exception was food record data, which was usually collected in the ≤1.5-week period prior to each testing visit.

Weight Loss and Body Composition.

Metabolic weight was measured in a hospital gown using a Seca mBCA 514 (Seca Corporation; Chino, CA). As interim measurements, we also measured body weight in the non-fasting state in light clothing in the clinic every 2 weeks throughout the intervention. Body composition was measured using dual x-ray absorptiometry (iDXA; GE Lunar Corporation; Madison, WI) and analyzed using enCORE software (version 15).

Cardiometabolic Risk Factors.

Blood pressure, heart rate, and waist circumference were measured using standard procedures. Fasting glucose, total cholesterol, HDL cholesterol, triglycerides, and HbA1c were assayed using a Sirrus Clinical Chemistry Analyzer (Stanbio Laboratory, L.P.; Boerne, TX). LDL cholesterol was calculated using the Friedewald equation. Insulin was measured using a Tosoh AIA-900 Analyzer (Tosoh Corporation; South San Francisco, CA). The inter-assay CVs were 2.5% for glucose, 4.0% for insulin, 4.3% for total cholesterol, 6.6% for HDL cholesterol, and 3.5% for triglycerides. Glucose and insulin were used to calculate the homeostatic model assessments for insulin resistance (HOMA-IR) and β-cell function (HOMA-β).

Adherence.

Participants self-reported when they started and stopped eating all calorie-containing foods or beverages for the day via an electronic survey. eTRE+ER participants were counted as adherent if they ate within a ≤8.5-hour window between 6:30 to 15:30, while CON+ER participants were counted as adherent if they ate over an ≥11.5-hour window. Days with missing data were counted as non-adherent.

Food Intake, Eating Behavior, and Physical Activity.

Participants completed digital food records on 2 weekdays and 1 weekend day using the Remote Food Photography Method (RFPM) (38). In brief, participants took photos of all food and beverages they consumed, and a trained dietitian estimated the portion sizes using standardized references. Food intake was calculated using the Food and Nutrient Database for Dietary Studies (FNDDS, version 6.0) and other sources such as the manufacturer’s information. Records were considered invalid if energy intake was <50% of predicted energy intake, as estimated by the NIDDK body weight planner (www.niddk.nih.gov/bwp) using a physical activity level (PAL) of 1.45. We also used the NIDDK body weight planner to conduct a post hoc analysis of energy intake during weight loss, assuming negligible changes in physical activity and a baseline PAL of 1.45. Participants’ hunger and fullness were measured in the morning (typically 7:30–9:00) in the fasting state during the week 0 and 14 testing visits using retrospective visual analog scales, while eating behavior was measured using the Dutch Eating Behavior Questionnaire (DEBQ). Finally, physical activity was assessed by self-report using the Baecke Physical Activity Questionnaire.

Mood and Sleep.

Mood was measured using the Profile of Mood States-Short Form (POMS-SF). Depression symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9). Sleep was measured by self-report using the Pittsburgh Sleep Quality Index (PSQI) and the Munich Chronotype Questionnaire (MCTQ). For the analyses, we excluded MCTQ data from one individual who reported a sleep duration of fewer than 3 hours.

Statistical Analysis.

All analyses were performed using two-sided tests with α=0.05 in R (version 4.0.3). We limited the per-protocol analysis to completers in both groups who consistently adhered to their assigned eating schedule at least 5 days/week every week. Since body weight was measured at multiple time points, it was analyzed using linear mixed models, with participant as a random intercept and with adjustment for age, sex, and race. All other data were analyzed using independent t-tests, with one exception. For the small number of endpoints that were imbalanced at baseline and whose change scores correlated with baseline values, we instead performed linear regression with adjustment for baseline and/or baseline × group. Values are reported as mean ± SEM, except that baseline characteristics, adherence, and eating windows are reported as mean ± SD.

RESULTS

Attrition.

As previously reported, we screened 656 people and enrolled 90 participants (28). Nine participants withdrew from the CON+ER group, while 11 withdrew from the eTRE+ER group (p=0.80). Only two participants in the eTRE+ER group withdrew due to difficulty following eTRE. We were unable to collect week 14 data from another 11 participants due to the COVID-19 pandemic. In total, 59 participants completed all aspects of the intervention.

Adherence.

Excluding those affected by the COVID-19 pandemic, 53% (21 of 40) of the CON+ER group and 38% (15 of 39) of the eTRE+ER group were consistently adherent and completed the study. Adherent completers in the CON+ER and eTRE+ER groups adhered 6.8 ± 0.3 and 6.3 ± 0.4 days/week, respectively (p<0.001; Figure 1). Adherent completers in both groups ate breakfast around roughly 8:00 (p=0.45). However, adherent completers in the eTRE+ER group ate over a 7.5 ± 0.8-hour period ending at 15:18 ± 0:18, while those in the CON+ER group ate over a 12.8 ± 0.9-hour period ending at 20:45 ± 1:06 (p<0.001). Thus, adherent completers in the eTRE+ER group fasted an additional 5.2 hours per day (p<0.001).

Figure 1. Eating Window.

(A) Adherent completers in the early time-restricted eating plus energy restriction (eTRE+ER) group fasted for an additional 5.2 hours per day relative to those in the control eating schedule plus energy restriction (CON+ER) group. (B) Adherent completers in both groups started eating about the same time of day, but those in the CON+ER group finished eating at 20:45 ± 1:06, whereas those in the eTRE+ER group stopped eating at 15:18 ± 0:18. * p<0.05

Participant Characteristics.

Baseline characteristics are shown for adherent completers in Table 1 and for all participants in Table S1. Demographically, adherent completers were aged 44 ± 12 years, had a mean BMI of 38.4 ± 6.4 kg/m², and were 72% female, 69% White, and 22% Black. There were baseline imbalances between groups for blood pressure and the Baecke total physical activity score (p≤0.04), so we adjusted for baseline values when analyzing these endpoints.

Table 1. Baseline Characteristics.

Mean ± SD.

Characteristic or Risk Factor	All Participants (n=36)	CON+ER (n=21)	eTRE+ER (n=15)	p
Demographics
Age, years	44 ± 12	42 ± 12	46 ± 11	0.33
Sex				1.00
Female	26 (72%)	15 (71%)	11 (73%)
Male	10 (28%)	6 (29%)	4 (27%)
Race				1.00
Asian	1 (3%)	1 (5%)	0 (0%)
Black	8 (22%)	5 (24%)	3 (20%)
White	25 (69%)	14 (67%)	11 (73%)
More Than One Race	2 (6%)	1 (5%)	1 (7%)
Ethnicity				0.50
Hispanic	0 (0%)	0 (0%)	0 (0%)
Not Hispanic	34 (94%)	19 (90%)	15 (100%)
Unknown or Not Reported	2 (6%)	2 (10%)	0 (0%)
Body Composition
BMI, kg/m2	38.4 ± 6.4	38.3 ± 6.0	38.5 ± 7.1	0.93
Weight, kg	107.2 ± 21.9	104.4 ± 21.7	111.1 ± 22.4	0.38
Fat Mass, kg	49.5 ± 15.6	47.6 ± 14.9	52.2 ± 16.7	0.39
Fat-Free Mass, kg	57.7 ± 10.8	56.9 ± 11.7	59.0 ± 9.8	0.58
Cardiometabolic Risk Factors
Glucose, mg/dl	109 ± 17	106 ± 14	114 ± 20	0.12
Insulin, mIU/l	20.7 ± 14.3	19.0 ± 11.7	23.1 ± 17.5	0.40
HOMA-IR	5.65 ± 4.26	4.94 ± 3.31	6.65 ± 5.28	0.24
HOMA-β	178 ± 125	179 ± 127	176 ± 128	0.95
HbA1c, %	5.6 ± 0.3	5.6 ± 0.4	5.6 ± 0.3	0.97
Systolic BP, mm Hg	126 ± 12	123 ± 9	131 ± 14	0.04
Diastolic BP, mm Hg	83 ± 8	81 ± 7	87 ± 9	0.02
Heart Rate, beats/mina	74 ± 9	72 ± 9	77 ± 6	0.09
Total Cholesterol, mg/dl	203 ± 40	198 ± 36	210 ± 46	0.36
Triglycerides, mg/dl	125 ± 65	122 ± 71	128 ± 56	0.80
LDL Cholesterol, mg/dl	118 ± 28	115 ± 25	122 ± 32	0.49
HDL Cholesterol, mg/dl	60 ± 16	58 ± 17	63 ± 16	0.38
Eating Habits
Eating Duration, h/day	13.2 ± 1.6	13.2 ± 1.7	13.0 ± 1.4	0.71
Eating Start Time, h:m	7:17 ± 1:01	7:26 ± 0:59	7:05 ± 1:04	0.32
Eating End Time, h:m	20:26 ± 1:38	20:40 ± 1:52	20:07 ± 1:12	0.33
Mood
Total Mood Disturbance Score	2.6 ± 2.6	1.9 ± 2.1	3.5 ± 3.0	0.07
Vigor-Activity	1.9 ± 0.9	2.1 ± 1.0	1.6 ± 0.6	0.09
Fatigue-Inertia	1.4 ± 0.8	1.3 ± 0.7	1.6 ± 1.0	0.22
Depression-Dejection	0.6 ± 0.5	0.5 ± 0.4	0.7 ± 0.5	0.34
Anger-Hostility	0.8 ± 0.5	0.7 ± 0.4	0.9 ± 0.7	0.26
Anxiety-Tension	1.0 ± 0.7	0.9 ± 0.6	1.1 ± 0.8	0.35
Confusion-Bewilderment	0.7 ± 0.4	0.6 ± 0.4	0.7 ± 0.5	0.37
PHQ-9 Score	5.1 ± 4.4	4.7 ± 4.4	5.6 ± 4.6	0.56
Sleep
Sleep Duration, h/day	7.4 ± 1.0	7.4 ± 1.0	7.4 ± 0.9	0.97
Sleep Onset, h:m	23:02 ± 0:58	23:08 ± 0:53	22:53 ± 1:05	0.46
Sleep Offset, h:m	6:28 ± 0:56	6:34 ± 1:06	6:18 ± 0:35	0.37
Chronotype, h:m	3:10 ± 0:55	3:16 ± 1:04	3:00 ± 0:39	0.43
Sleep Quality	6.1 ± 3.2	5.7 ± 2.7	6.7 ± 3.7	0.36
Sleep Latency, min	19 ± 14	21 ± 15	17 ± 14	0.44
Sleep Efficiency, %	93 ± 5	93 ± 5	93 ± 5	0.98
Sleep Inertia, min	15 ± 19	13 ± 21	17 ± 15	0.59

Abbreviations: CON+ER, control eating schedule plus energy restriction; eTRE+ER, early time-restricted eating plus energy restriction; BMI, body mass index; BP, blood pressure; HOMA-IR, homeostasis model assessment for insulin resistance; HOMA-β, homeostasis model assessment for β-cell function.

^aOne baseline value for heart rate in the eTRE+ER group was excluded.

Weight Loss and Body Composition.

Figure 2 shows the results for weight loss and body composition. At the end of the 14-week intervention, adherent completers in the CON+ER group lost 3.9 ± 0.9 kg (4.1 ± 0.8%) of body weight (p<0.001), whereas those in the eTRE+ER group lost 7.6 ± 1.0 kg (7.0 ± 0.9%) (p<0.001). Thus, adherent completers in the eTRE+ER group lost an additional 3.7 ± 1.2 kg (p=0.003) or 2.9 ± 1.1% (p=0.009) of weight. eTRE+ER was also more effective for losing body fat. eTRE+ER decreased body fat by an additional 2.8 ± 1.3 kg (−3.2 ± 0.8 vs. −6.0 ± 1.1 kg; p=0.04) and trunk fat by an additional 1.6 ± 0.7 kg (p=0.04) relative to CON+ER. In contrast, there were no differences in fat-free mass (−1.2 ± 0.3 vs. −1.8 ± 0.6 kg; −0.7 ± 0.6 kg; p=0.25), appendicular lean mass (−0.3 ± 0.4 kg; p=0.42), visceral fat (n=30; −0.1 ± 0.1; p=0.37), and waist circumference (−1.8 ± 1.5 cm; p=0.25).

Figure 2. Weight and Body Composition.

Among adherent completers, early time-restricted eating plus energy restriction (eTRE+ER) was more effective for losing (A) weight, (B) body fat, and (D) trunk fat than the control eating schedule plus energy restriction (CON+ER). There were no differences in (C) fat-free mass, (E) visceral fat (n=30), (F) waist circumference, and (G) appendicular lean mass between groups. * p<0.05

Cardiometabolic Risk Factors.

Cardiometabolic risk factors are shown in Figure 3. Among adherent completers, eTRE+ER decreased heart rate by an additional 7 ± 3 beats/min (p=0.02), fasting glucose by 9 ± 5 mg/dl (p=0.047), and insulin resistance as measured by HOMA-IR by 2.80 ± 1.36 (p=0.047) relative to CON+ER. There were no between-group differences in systolic blood pressure (−6 ± 3 mm Hg; p=0.10), diastolic blood pressure (−4 ± 2 mm Hg; p=0.09), fasting insulin (−9.8 ± 5.1 mIU/l; p=0.07), HbA1c (−0.1 ± 0.1%; p=0.46), HOMA-β (−78 ± 53; p=0.15), total cholesterol (−14 ± 10 mg/dl; p=0.18), triglycerides (−8 ± 17 mg/dl; p=0.64), LDL cholesterol (−9 ± 8 mg/dl; p=0.30), and HDL cholesterol (−3 ± 3 mg/dl; p=0.20).

Figure 3. Cardiometabolic Risk Factors.

Among adherent completers, early time-restricted eating plus energy restriction (eTRE+ER) was more effective at decreasing (A) fasting glucose, (C) insulin resistance as measured by HOMA-IR, and (H) heart rate (n=35) than the control eating schedule plus energy restriction (CON+ER). There were no differences in (B) fasting insulin, (D) HOMA-β, (E) HbA1c, (F) systolic blood pressure, (G) diastolic blood pressure, (I) total cholesterol, (J) triglycerides, (K) LDL cholesterol, and (L) HDL cholesterol between groups. * p<0.05

Food Intake, Eating Behavior, and Physical Activity.

Table 2 shows changes in food intake, eating behavior, and physical activity. Relative to CON+ER, eTRE+ER did not affect energy intake (−5 ± 177 kcal/day; p=0.98), macronutrient intake (p≥0.80) or micronutrients, food groups, or food components (p≥0.06; data not shown), as measured by digital food records. Post hoc weight-loss modeling produced a higher but still non-significant estimate for the between-group difference in energy intake (−350 ± 174 kcal/day; p=0.053). eTRE+ER also did not affect appetite (p≥0.13), eating restraint (0 ± 3; p=0.99), emotional eating (0 ± 3; p=0.91), external eating (0 ± 2; p=0.98), or self-reported physical activity (p≥0.31).

Table 2. Food Intake, Eating Behavior, and Physical Activity.

Among adherent completers, there were no between-group differences in food intake, appetite, eating behaviors, or physical activity.

	CON+ER Within-Group Change		eTRE+ER Within-Group Change		Between-Group Difference
Endpoint	Mean ± SEM	p	Mean ± SEM	p	Mean ± SEM	p
Food Intake by RFPM (n=26)
Energy Intake, kcal/day	−625 ± 134	<0.001	−630 ± 96	<0.001	−5 ± 177	0.98
Carbohydrate Intake, g/day	−84 ± 18	<0.001	−85 ± 22	0.003	−1 ± 29	0.97
Fat Intake, g/day	−29 ± 8	0.003	−26 ± 7	0.004	3 ± 11	0.80
Protein Intake, g/day	−9 ± 7	0.23	−9 ± 7	0.24	0 ± 11	0.98
Weight Loss Modeling
Energy Intake, kcal/day	−466 ± 101	<0.001	−816 ± 150	<0.001	−350 ± 174	0.053
Appetite
Average Hunger, mm	−8 ± 4	0.046	−16 ± 6	0.02	−8 ± 7	0.25
Hunger When Fasting, mm	−18 ± 7	0.02	−3 ± 10	0.74	15 ± 12	0.22
Fullness When Eating, mm	−20 ± 7	0.007	−6 ± 6	0.32	14 ± 9	0.13
Eating Behaviors
Restrained Eating	9 ± 2	<0.001	9 ± 2	<0.001	0 ± 3	0.99
Emotional Eating	−3 ± 2	0.13	−3 ± 2	0.26	0 ± 3	0.91
External Eating	−4 ± 1	0.006	−3 ± 2	0.04	0 ± 2	0.98
Physical Activity
Global Physical Activity Indexa	0.3 ± 0.1	0.06	0.3 ± 0.2	0.10	0.0 ± 0.2	0.96
Work Index	−0.1 ± 0.1	0.22	0.0 ± 0.1	0.66	0.1 ± 0.1	0.44
Sport Index	0.0 ± 0.1	0.91	0.0 ± 0.1	0.65	0.0 ± 0.1	0.87
Leisure Index	0.3 ± 0.1	0.08	0.5 ± 0.2	0.006	0.2 ± 0.2	0.31

Abbreviations: CON+ER, control eating schedule plus energy restriction; eTRE+ER, early time-restricted eating plus energy restriction; RFPM, Remote Food Photography Method, which is equivalent to collecting a photographic food record.

Mood.

Figure 4 shows mood data from the POMS-SF and PHQ-9. Among adherent completers, eTRE+ER increased vigor-activity (0.5 ± 0.2; p=0.04) and decreased fatigue-inertia (−0.6 ± 0.3; p=0.045), depression-dejection (−0.4 ± 0.1; p=0.01), and anger-hostility (−0.4 ± 0.2; p=0.03) relative to CON+ER. eTRE+ER was also more effective for improving overall mood as measured by the total mood disturbance score (−2.4 ± 0.9; p=0.009). There were no between-group differences in tension-anxiety (−0.4 ± 0.2; p=0.14), confusion-bewilderment (−0.2 ± 0.2; p=0.42), or the PHQ-9 score (−1.4 ± 1.3; p=0.31), which indicates the severity of clinical depression.

Figure 4. Mood.

Among adherent completers, early time-restricted eating plus energy restriction (eTRE+ER) was more effective at improving (A) the total mood disturbance score and subscores for (B) vigor-activity, (C) fatigue-inertia, (D) depression-dejection, and (E) anger-hostility than the control eating schedule plus energy restriction (CON+ER). There were no differences between groups in (F) tension-anxiety, (G) confusion-bewilderment, or (H) the severity of clinical depression as measured by the Patient Health Questionnaire-9 (PHQ-9) score. * p<0.05

Sleep.

Self-reported sleep quality, duration, and timing are shown in Figure 5. Sleep quality did not differ between groups (0.2 ± 0.7; p=0.79). However, eTRE+ER decreased sleep duration by 30 ± 13 min (p=0.03), increased sleep latency by 7 ± 3 min (p=0.04), and reduced sleep efficiency by 2 ± 1% (p=0.04) relative to CON+ER. There were no differences in the timing of sleep, including sleep onset (8 ± 13 min; p=0.51), sleep offset (−22 ± 14 min; p=0.16), sleep inertia (1 ± 3 min; p=0.78), or chronotype (1 ± 16 min; p=0.94).

Figure 5. Sleep.

Among adherent completers, early time-restricted eating plus energy restriction (eTRE+ER) decreased (A) sleep duration and (G) sleep efficiency and increased (F) sleep latency relative to the control eating schedule plus energy restriction (CON+ER). There were no differences in (B) sleep onset, (C) sleep offset, (D) chronotype, (E) sleep quality, and (H) sleep inertia between groups. * p<0.05

DISCUSSION

We recently conducted a randomized controlled weight-loss trial comparing eTRE versus eating over a ≥12-hour period for losing weight and improving cardiometabolic health. In the main ITT analysis, eTRE was superior for losing weight, lowering blood pressure, and improving mood relative to eating over a ≥12-hour period (Table S2). Here, we performed a secondary per-protocol analysis, which is also the first relatively large per-protocol analysis of TRE in humans. We used an adherence threshold of ≥5 days per week every week, which was inspired by two rodent studies reporting benefits from practicing TRE only on weekdays (6, 36) and by data suggesting that participants typically adhere to TRE 5.0–6.4 days/week (13, 14, 15, 17, 18, 21, 30, 31).

Participants who consistently adhered to eTRE+ER at least 5 days/week lost more body weight, body fat, and trunk fat than their counterparts in the CON+ER group. The weight-loss effect was larger in the per-protocol analysis than the ITT analysis (−3.7 kg vs. −2.3 kg) and corresponded to a relative energy deficit of 350 ± 174 kcal/day (p=0.053), as estimated by weight-loss modeling. Although previous studies on weight loss are mixed, our results are in line with recent meta-analyses suggesting that TRE has modest to moderate effects on body weight (26, 29). We could not trace this weight loss benefit to changes in appetite, eating restraint, physical activity, or food intake—a shortcoming that is likely due to the well-known limitations of food records, as most studies report that TRE decreases food intake and does not affect physical activity (13, 16, 17, 19, 39, 40, 41). Our discrepant results may be explained by the fact that we excluded food records that severely underreported food intake and/or we were not sufficiently well-powered to detect differences in food intake with only 26 valid food records. In contrast, eTRE’s favorable effects on fat loss in the per-protocol analysis were not present in the ITT analysis. Our results for fat loss mirror findings in a per-protocol analysis by Tinsley et al. (2019), who found that an 8-hour TRE intervention was superior to eating across the day (~13.6-hour window) for losing body fat—an effect that was also not present in the ITT analysis (34). This suggests that people who consistently adhere to eTRE lose more body fat and trunk fat without any negative effects on fat-free mass.

Among consistently adherent participants, eTRE+ER also improved fasting glucose, insulin resistance, and heart rate. None of these effects were observed in the ITT analysis. Our per-protocol findings are consistent with most efficacy studies, which report that eTRE improves glucose levels, insulin levels, insulin sensitivity, and/or β-cell responsiveness (9, 10, 12, 23, 42). Studies on other versions of TRE report more mixed results. For heart rate, only one prior study on TRE reported an improvement (43). In our per-protocol analysis, eTRE+ER did not improve lipids, systolic blood pressure, or diastolic blood pressure. However, the size of the effect for diastolic blood pressure (p=0.09) was identical to the ITT analyses (−4 vs. −4 mm Hg), which did find a statistically significant effect. One previous controlled feeding study found that eTRE reduces blood pressure (9), while the results for other versions of TRE are mixed but lean null.

Participants who consistently adhered to eTRE+ER at least 5 days/week also had greater improvements in overall mood. eTRE+ER modestly increased energy levels and reduced feelings of fatigue, dejection, and anger relative to CON+ER. The positive findings for anger are novel and were not present in the ITT analysis. It is unclear what mechanisms may mediate these effects on mood, so further investigation is needed. However, eTRE may potentially have some undesirable effects on sleep in the context of an energy-restricted diet. Adherent completers in the eTRE+ER group self-reported taking a few minutes longer to fall asleep and sleeping half an hour less relative to adherent completers in the CON+ER group. One plausible explanation for these findings is that practicing eTRE+ER made participants hungrier at night, disrupting sleep. However, in a post hoc analysis, we found no correlations between sleep duration or sleep latency and either mean hunger or hunger while fasting (p≥0.26). Another possibility is that participants may have stayed up later and/or woken up earlier to prepare meals for the day. Alternatively, it is possible that these changes in sleep may instead be healthful. Despite sleeping less, adherent participants in the eTRE+ER group reported less fatigue and similar sleep quality to their counterparts in the CON+ER group. The decrease in sleep duration and fatigue with no change in sleep quality could instead reflect better sleep consolidation. Thus, future studies should use objective measures of sleep, such as actigraphy or polysomnography, to determine whether the changes we observed in sleep are salubrious or not.

Our study has several strengths, including its novelty, rigor, and assessment of multiple health endpoints. Our study also has some limitations, including our sample size was moderate, we enrolled mostly females, and we measured sleep and physical activity by self-report. We also recognize that per-protocol analyses can be affected by potential bias. First, per-protocol analyses may have sampling bias and may not be representative of the study population as a whole. For example, participants with greater improvements in cardiometabolic and psychological health may be motivated to be more adherent to eTRE, which could inflate the estimated per-protocol effects. Second, our adherence metric reflects both self-reported adherence to the eating schedule and the number of missing surveys, so our per-protocol analysis may have also selected for conscientiousness and/or adherence to energy restriction rather than adherence to the eating schedule alone. Thus, we cannot rule out bias. Though, in our trial, we found no evidence of differential bias in weight loss among completers versus dropouts and no evidence of bias by chronotype. Adherers to eTRE+ER did not have an earlier chronotype and did not eat dinner any earlier than non-adherent participants, though they had higher fasting glucose and diastolic blood pressure, ate breakfast about 40 min earlier, and ate over a longer eating window at baseline. Other preliminary research suggests that people who successfully adhere to TRE report having regular schedules, improvements in energy levels and physical health, positive psychological impacts, good social support, and better alignment with their daily routines, whereas though who struggle to adhere report feelings of hunger and sluggishness, irregular daily routines, inadequate diet quality during the eating window, difficulties with self-monitoring, and social situations that discourage TRE (44). Interestingly, key drivers of adherence appear to differ greatly across individuals and their unique circumstances (44). Further research is needed to determine the key characteristics of people who can adhere to eTRE, as well as who benefits most. Finally, we cannot say whether the effects generalize to other versions of TRE. We have previously hypothesized that eTRE may be more efficacious than practicing TRE by skipping breakfast—a conjecture now supported by two moderately-sized randomized trials (23, 45). We speculate that the glycemic improvements and changes in sleep are mostly due to eating earlier in the day because the time of day influences glucose tolerance more than the fasting duration (35) and studies on other versions of TRE report no effects on sleep duration. Whereas, we speculate that the weight loss and mood benefits are due to both the longer fasting duration and eating earlier since studies on other versions of TRE also report weight loss (25, 26). However, more research is needed.

In conclusion, adults with obesity who practiced eTRE at least 5 days/week every week lost more weight and body fat and had greater improvements in cardiometabolic health and mood than those who consistently ate over a ≥12-hour period. Therefore, consistently following eTRE had additional benefits for losing body fat and improving fasting glucose, insulin resistance, heart rate, and anger-hostility that were not present in the ITT analysis. Participants who consistently adhered to eTRE also slept less and took longer to fall asleep, yet reported no negative effects on sleep quality or fatigue, which merits further investigation. Future studies should continue to investigate the per-protocol effects of TRE and determine who benefits most.

Study Importance Questions.

What is already known about this subject?

• Effectiveness trials are mixed as to whether a form of intermittent fasting called time-restricted eating (TRE) improves body weight and cardiometabolic health.

• Only one prior study investigated the effects of TRE in adherent participants by performing a per-protocol analysis, and it found that TRE was better for losing body fat than the control eating schedule.

What are the new findings in your manuscript?

• Participants who adhered to early TRE (eTRE) at least 5 days/week every week lost more weight and body fat and experienced greater improvements in heart rate, insulin resistance, and fasting glucose than participants who consistently ate over a ≥12-hour window.

• In addition, several aspects of their mood improved—including fatigue-inertia and anger-hostility—although they also had increased sleep latency, shorter sleep duration, and reduced sleep efficiency.

How might your results change the direction of research or the focus of clinical practice?

• Consistently adhering to eTRE at least 5 days/week is an effective approach for losing weight and improving cardiometabolic health and mood.

• eTRE may reduce fatigue but causes adherers to sleep less and take longer to fall asleep, which merits further investigation.