Sleep-aligned Extended Overnight Fasting Improves Nighttime and Daytime Cardiometabolic Function

Abstract

Background:

Time-restricted eating (TRE) has gained attention for its potential cardiometabolic health benefits. Existing TRE approaches may have limited adherence and sustainability due to fixed fasting windows with prolonged fasting duration before sleep, or they involve self-selected fasting windows without specifying the duration relative to sleep, a critical period for cardiometabolic regulation. We hypothesized that an individualized approach that extended overnight fasting (EOF) duration by three hours in alignment with habitual sleep time (last meal ≥3 hours before sleep) would enhance nighttime autonomic balance, decrease blood pressure (BP) and heart rate (HR), increase BP/HR dipping, and glucose regulation compared to a control group maintaining habitual eating patterns.

Methods:

In this randomized parallel-arm controlled trial, 39 overweight/obese participants (36 – 75 years) completed either an EOF intervention (13h - 16h fasting) or a control condition (habitual fast of 11h - 13h). Both groups dimmed lights 3 hours before bedtime. The intervention duration was 7.5 weeks.

Results:

Compared to control, EOF intervention significantly improved the co-primary outcome of nighttime dipping of diastolic BP, but not the Matsuda Index of insulin sensitivity. EOF improved secondary measures of nighttime autonomic function and morning oral glucose tolerance, including lower nighttime HR, higher heart rate variability, lower nighttime cortisol and during the oral glucose tolerance test lower glucose level, and higher 30-minute insulinogenic index, indicating improved acute insulin response.

Conclusions:

Extending overnight fasting duration by three hours in alignment with sleep improved cardiometabolic health in middle-aged/older adults by strengthening coordination between circadian- and sleep-regulated autonomic and metabolic activity. This sleep-aligned TRE approach represents a novel, accessible lifestyle intervention with promising potential for improving cardiometabolic function.

Graphical Abstract

INTRODUCTION

There is growing interest in time-restricted eating (TRE), a dietary approach that limits food consumption to a consistent window of 10 hours or less during the daytime ¹. This paradigm is supported by substantial evidence from animal studies, where time-restricted feeding aligned with the dark-light cycle has been shown to improve health and longevity ^2,3. In humans, emerging research suggests that TRE that aligns feeding patterns with the body’s diurnal regulation of autonomic, endocrine, and metabolic processes could represent a promising intervention to support cardiometabolic health ⁴. This structured approach to eating contrasts with modern eating patterns that typically span 14 to 15 hours daily and have been linked to increased risk of cardiovascular and metabolic diseases ⁵.

Research shows that TRE can improve several cardiometabolic markers, including insulin sensitivity, fasting glucose levels, lipid profiles, body weight, and blood pressure (BP) ⁴. Studies have tested the cardiometabolic benefits of TRE using different fasting windows; some implemented long fasting periods at fixed times, beginning in the afternoon (around 3 PM) and extending to the following morning ^6–9. Other studies used self-selected fasting windows that differed between participants, with some beginning close to bedtime and ending several hours after waking ¹⁰. However, the potential benefits of aligning the fasting periods around individual sleep schedules remain largely unexplored. A recent trial where participants began fasting at least three hours before their habitual sleep time showed improved glucose regulation in patients with metabolic syndrome ¹¹, suggesting that fasting should begin several hours prior to bedtime.

During sleep, the body undergoes distinct autonomic and metabolic processes, characterized by reduced metabolic rate, glucose utilization, and sympathetic nervous system activity, processes crucial for cardiometabolic and physiological homeostasis. Since sleep timing varies between individuals due to differences in circadian sleep-wake regulation, anchoring the fasting window around each person’s habitual sleep schedule may enhance these sleep-dependent cardiometabolic processes. This approach is supported by research showing that eating late relative to a fixed bedtime, compared to eating earlier relative to that same fixed bedtime, is associated with weight gain, insulin resistance, and impaired glucose regulation, likely due to disrupted sleep-related autonomic and metabolic activities ¹². Notably, the 2 – 3 hours before bedtime represent a physiologically critical period for metabolism as melatonin levels begin rising, and evidence indicates that eating during elevated melatonin periods negatively impacts metabolic function ¹³.

While previous TRE studies may have achieved their cardiometabolic benefits partly through extended overnight fasting, they did not examine autonomic and metabolic physiology during sleep, as most were conducted in field settings that precluded such mechanistic assessments. Furthermore, previous TRE studies did not account for the potential confounding impact that ambient lighting before usual bedtime has on the circadian system ¹⁴, sleep ¹⁵, and autonomic activity ¹⁶, factors that could significantly influence cardiometabolic outcomes.

Beyond its effects on cardiometabolic function, there is some evidence, mostly from subjective or wearable-derived objective measures, that TRE may improve sleep quality ¹⁷, which could further mediate its effects on cardiometabolic health ¹⁸. However, existing studies show overall mixed results regarding the effects of TRE on sleep, which in part may be due to the reliance on different indirect objective measures of sleep.

To date, TRE research has primarily focused on metabolically vulnerable young and middle-aged populations, particularly those with metabolic syndrome or obesity ⁴. The potential benefits for older adults remain largely unexplored, despite this group facing age-related changes in autonomic and metabolic processes that increase their cardiometabolic disease risk. In addition, older adults often experience age-related sleep disruptions ¹⁹ and might benefit from potential TRE-related sleep improvements.

To address these gaps and to optimize the coordination between circadian and sleep physiology, we designed an extended overnight fasting intervention aligned with individual sleep patterns (EOF intervention). Specifically, we extended the overnight fast by 3 hours, with a minimum 3-hour fast required before each participant’s habitual bedtime. We hypothesized that the EOF intervention would enhance cardiometabolic function in middle-aged and older adults compared to a control group that maintained their habitual eating patterns, with both groups instructed to dim ambient lighting < 100 lux during the 3 hours before bedtime. We tested the hypothesis that EOF intervention would improve nighttime autonomic balance, nighttime BP and HR dipping, daytime glucose regulation, and objective sleep quality.

METHODS

Anonymized data have been made publicly available in Arch, the open access Northwestern University Institutional Repository, and can be accessed at https://doi.org/10.21985/n2-620a-tb23.

Participants details

This study was conducted with approval from Northwestern University’s Institutional Review Board, and all participants provided written informed consent. Participants were enrolled from two studies identical for screening and visit procedures, except for age criterion. One study (NCT03490825) enrolled adults aged 55 – 75 years, while the other one (NCT03490864) enrolled adults aged 35 – 54 years. Both studies were designed as randomized controlled trial where each participant was assigned to one of four intervention types: EOF + placebo, EOF + melatonin, control + melatonin, control + placebo. The main hypothesis of the study is that sleep-aligned EOF would improve cardiometabolic function compared to habitual eating patterns. Due to the COVID emergency, enrollment was prioritized to address the primary hypothesis and only randomization to either EOF + placebo or control + placebo groups continued.

Recruitment occurred between 2018 and 2024 through flyers, social media, and website advertisements. Initial eligibility was determined via an online survey (REDCap) or phone interview. Targeted recruitment strategies were implemented throughout the study to achieve participant demographics representative of the Chicagoland population (approximately 25% racial and ethnic minorities). Candidates who passed initial screening underwent a remote screening visit, during which consent was obtained and questionnaires assessing sleep, mood, physical activity levels, and medical history were administered.

Eligible participants then attended an in-person screening visit for anthropometric measurements, BP and HR, HbA1c, and Mini-Mental State Examination (MMSE). Following the in-person visit, subjects monitored their eating schedules for 7 days using food diaries and a meal logger application on their phones. During this 7-day period, participants also wore a wrist actigraphy device (Actiwatch Spectrum, Philips Respironics) to monitor wake-sleep patterns and completed sleep diaries.

The study aimed to include middle-aged and older adults at various levels of cardiometabolic disease risk, without extreme sleep phenotypes, and who were overall healthy. Eligibility criteria included middle-aged and older adults 35 to 75 years, BMI < 45 kg/m², both sexes. Potential participants needed to have a habitual overnight fasting period of ≤ 13 hours, regular eating (at least 2 meals/day) and sleeping schedules (≤ 2 hours deviation in daily mid-sleep time), self-reported sleep duration ≥ 6.5 hours, habitual mid-sleep time of 1 – 5 am and habitual time in bed < 9 hours. The threshold of ≤13 hours for baseline fasting was chosen based on the median fasting duration from previous data and to ensure that participants had room for meaningful extension of overnight fasting.

Exclusion criteria included history or current sleep disorders, AHI ≥ 30 (based on home sleep recording), anemia, HbA1C ≥ 6.5, diabetes diagnosis or use of any medication for diabetes, endocrine dysfunctions, major psychiatric disorders (DSM-V criteria), moderate depression (Beck depression inventory 2 scale > 16), cognitive impairment (MMSE < 26), unstable medical conditions, gastrointestinal diseases requiring dietary adjustments, recent use of melatonin or psychoactive medications, current hormone replacement therapy, irregular sleep schedules, smoking ≥ 6 cigarettes/day, high caffeine consumption (> 400 mg/day), recent weight loss programs, bariatric surgery, night eating syndrome, and significant visual impairment.

Study design

This randomized parallel study compared a 6-week minimum EOF intervention to a control condition in which participants maintained their usual eating schedules. All participants completed two 4-day and 3-night stays at the Clinical Research Unit (CRU) at Northwestern Memorial Hospital: one before the intervention (baseline) and one after completion of the intervention (post-intervention). During these stays, assessments of sleep, hormones, BP and HR were conducted (Figure 1). Eating schedules at baseline visit were matched to participants’ habitual schedule, while at post-intervention visit they were based on group assignment and self-reported caloric distribution (EOF intervention vs control schedule).

Figure 1.

Study procedures: After screening procedures to determine eligibility, eligible participants enrolled in the study were randomized to either a minimum 6-week intervention in the EOF group (overnight fasting extended by 3 hours and ≥ 3 hours of fasting before bedtime), or to a control group maintaining their habitual eating schedule. * two extra weeks were allowed to match participant’s schedule with CRU availability. At baseline and post-intervention, participants were admitted to the hospital for a 4 days/3 night’s stay to obtain the following autonomic, metabolic, and sleep measures: nighttime heart rate (HR) and heart rate variability (HRV) and cortisol levels, a 3-hour oral glucose tolerance test (OGTT), 15.5-hour ambulatory blood pressure monitoring (ABPM) starting after OGTT completion until the following morning, and nightly polysomnography.

Pre-intervention visit

Prior to the baseline CRU visit, participants underwent one week of monitoring using actigraphy and sleep diaries to determine habitual bedtime. An interview with a trained nutritionist, supplemented by 7-day food diaries and the meal logger application (covering at least 2 workdays and 1 weekend day) obtained at screening, was used to determine the meal schedule and caloric intake during the laboratory stay.

In-laboratory visits

Inpatient visits were 3 nights/4 days long. Participants’ anthropometric measures (height, weight, waist circumference) and vitals (BP and HR) were obtained at admission at each visit. Participants underwent nightly polysomnography (PSG) with sleep opportunity timing and duration tailored to their average habitual sleep times with either a 7-, 8-, or 9-hour time in bed. Based on habitual eating patterns, up to 4 eating occasions were scheduled and an optional snack was offered. Each meal’s caloric and macronutrient distribution was based on habitual eating patterns. Overall caloric intake was calculated based on estimated energy requirements using the Institute of Medicine equations ²⁰ using age, sex, height, weight and sedentary activity level (reflecting the sedentary activity level required by the CRU visit conditions). Ambient light intensity was maintained at approximately 200 – 300 lux during wake periods and < 1 lux during sleep periods. On night 2 lights were dimmed to < 50 lux 3 hours before bedtime for both CRU visits, and in the post-intervention visit lights were dimmed on Night 1 and 3 to < 100 lux to reflect the intervention.

On Night 2, hourly blood draws were taken from an indwelling catheter, from lights off until just after wake time on Day 3. On the morning of Day 3, starting 1 hour after wake, a 3-hour OGTT was performed. After OGTT completion, BP and HR were monitored using ambulatory blood pressure monitoring at 30-minute intervals until wake time on Day 4.

Intervention and Compliance

At discharge on Day 4 at baseline visit, participants received specific instructions based on their group assignment. A random number generator was used to create a randomized block design for male and female blocks, by a collaborator not involved in this project. Participants were randomized stratified by sex with a 1:1 allocation ratio to either a control group (maintain habitual eating) or intervention group (+ 3 hours of habitual overnight fast duration with their last meal at least 3 hours before bedtime, minimum of 12 and maximum of 16 hours of overnight fast). Participants were instructed to not consume any food or beverages with calories during the fasting window and water was permitted.

The intervention lasted a minimum of 6 weeks, with an additional 2-week extension permitted to accommodate participants’ schedules with CRU availability.

To minimize the impact from other inputs to the circadian system, both groups were instructed to dim lights to <100 Lux three hours before bedtime. Lighting < 100 lux was chosen as it represents a common level of light in homes’ living rooms ¹⁴ They were also directed to maintain their habitual sleep, physical activity and caloric and macronutrient intake, with the EOF group adjusting meal timing to accommodate the overnight fast assigned.

Compliance was calculated based on self-report from a daily intervention log using the following variables: the clock time that lights were dimmed, time of last meal, time of first meal and bedtime. To monitor compliance, participants submitted daily logs that were reviewed at least every 2 days in the first 1 – 2 weeks of the intervention and once a pattern was established, they were reviewed at least weekly by the study staff. If on 2 consecutive days participants were more than +/− 30 minutes from the assigned fast duration and +/− 15 minutes for the duration of light and meal relative to bedtime, they received individualized coaching calls within about 24 hours. In addition, if the timing of the fast start or end was more than one hour different than scheduled, then participants were also coached.

Participants were considered compliant with EOF if they fasted within 30 minutes of their assigned overnight fast duration (e.g. if assigned 15 hours the fast had to be at least 14.5 hours), similarly they were considered compliant if they dimmed lights and consumed their last meals at least 2.5 hours before bedtime. An identical approach was used to monitor control subjects’ compliance with dimming lights in the evening.

In laboratory measures

Nighttime HR and heart rate variability (HRV)

HR and HRV measurements were obtained from Night 2 PSG and analyzed using a dedicated software (PRANA, PhiTools) ²¹ at a sample frequency of 500 Hz. ECG artifacts were automatically detected, visually inspected and then removed from the analysis. HR was averaged every 5 min across the night in accordance with standard guidelines ²². Only “normal-to-normal” R waves times were included. Spectral power was calculated using fast Fourier transform on time windows with a stable signal of at least 5 minutes for the entire duration of the sleep opportunity period. Spectral power was calculated in the high frequency band (HF: 0.15 to 0.40 Hz) reflecting mostly parasympathetic activity, and in the low-frequency band (LF: 0.04 to 0.14 Hz) reflecting a combination of vagal and sympathetic activities. The LF/HF ratio was calculated as an indicator of sympathovagal balance ²². Data analysis included the first 7 hours of the sleep period, starting from lights off.

Nighttime serum cortisol

Measurements were obtained from hourly blood draws during the sleep opportunity period of Night 2. Samples from the first 7 h of the sleep period, starting from lights off, were used for analysis. Blood samples were immediately centrifuged and stored at – 70°C until processing. Immunoassays were used to measure cortisol (measuring range 3.0 – 1750 nmol/L).

Ambulatory blood pressure monitoring (ABPM) of BP and HR

ABPM of BP and HR was conducted using 30-minute measurements on Day 3 (90217A Spacelabs Healthcare). The analysis included 8.5 hours before lights off (daytime period) and 7 hours after lights off (nighttime period). Time changes in Systolic BP (SBP), diastolic BP (DBP), and HR were analyzed across the entire 15.5-hour period, as well as during daytime and nighttime periods. Average pulse pressure for daytime and nighttime periods was calculated as (average SBP - average DBP). Nighttime dipping of SBP, DBP, and HR was calculated as the difference between the average value during the day and the average value during the night, expressed as a percentage of the daytime mean. The change from non-dipper status (< 10% nighttime dipping) to dipper status (≥ 10% nighttime dipping) was also reported for SBP, DBP and HR.

Homeostatic Model Assessment for insulin resistance (HOMA-IR)

HOMA-IR was calculated using the formula: [fasting glucose (mg/dL) × fasting insulin (μU/mL)] / 405 using fasting glucose and insulin samples taken 50 minutes after wake.

3-hour Oral Glucose Tolerance Test (OGTT)

The Oral Glucose Tolerance Test (OGTT) was conducted on the morning of Day 3. Ten minutes after fasting blood samples were taken, participants ingested a 75-g glucose bolus (Trutol). Blood samples were then collected at 10, 20, 30, 60, 90, 120, 150, and 180 minutes post-ingestion. Time-dependent changes in glucose and insulin were analyzed (mean levels) as well as changes of glucose and insulin levels at each time point of the OGTT as they can inform on the risk to develop type 2 diabetes, specifically 60 minute OGTT glucose levels ²³ and 30-minute insulin levels ²⁴. The insulinogenic index at 30 minutes was calculated as (Insulin30 - Insulin0) / (Glucose30 - Glucose0), a measure of pancreatic β-cells function ²⁴. The Matsuda index was calculated as 10,000/square root of [fasting glucose × fasting insulin] × [mean glucose × mean insulin during OGTT], a measure of insulin sensitivity ²⁵.

PSG

Overnight PSG recordings were conducted each night using a Neurofax EEG-1100 Digital EEG Acquisition System (Nihon-Kohden 8.0) with a sampling frequency of 500 Hz. EEG recordings were obtained from frontal (F3, F4), central (C3, C4), parietal (P3, P4), and occipital (O1, O2) channels referenced to the contralateral mastoid. Additionally, right and left electro-oculograms, three sub-mental electromyograms, and two electrocardiogram (ECG) leads were recorded. The ECG leads were placed above the right collarbone and below the left ribs. PSG recordings were visually scored according to the American Academy of Sleep Medicine scoring criteria ²⁶ by an experienced scorer blinded to the participants’ condition (EOF vs control intervention). PSG measures were calculated for total recording time (duration in minutes from lights off to lights on), total sleep time (duration in minutes of time spent asleep), sleep efficiency (percentage of time spent asleep over the entire recording period), time spent in each sleep stage (N1, N2, N3 [also known as slow-wave sleep], and REM sleep; duration in minutes and percentage of sleep time), sleep onset latency (time from lights off to the first 30-s epoch scored as N1 or N2), wake after sleep onset (time in minutes spent awake after sleep onset and before lights on), and arousal index (number of arousals lasting at least 3 seconds in duration per hour of sleep). Night 1 served as an adaptation night and was not included in the analysis.

Primary and secondary outcomes of the study

Primary outcomes of the study are BP nighttime dipping and the Matsuda Index from the OGTT. Secondary outcomes include additional measures from the OGTT including fasting glucose and insulin, HOMA-IR, OGTT 1-hour glucose, 30-minute insulin and insulinogenic index at 30 minutes, as well as nighttime cortisol levels, 15.5-h BP and HR, nighttime sleep from PSG. Outcome assessors were blinded to treatment allocation.

Statistical analysis

Linear mixed models (LMM) using restricted maximum likelihood (REML) estimation with unbounded variance components and participant as a random effect were used to assess differences between EOF and control groups. The group × visit interaction term tested whether changes from baseline to post-intervention visits differed between groups for nighttime HR and HRV, nighttime plasma cortisol, glucose and insulin, OGTT measures, BP and HR measurements from ABPM, and PSG variables. For the OGTT, differences between the EOF and control groups in changes from baseline to post-intervention were also assessed for glucose at 60 minutes and insulin at 30 minutes. Normality assumptions were checked using the “Shapiro-Wilk test”. Variables not normally distributed, such as OGTT insulin levels, were analyzed after natural log transformation to meet the distributional assumptions of LMM framework. Comparisons between EOF and the control group for demographics, baseline cardiometabolic and sleep characteristics, and eating habits (Table 1) were performed using unpaired 2-sided t-tests for continuous variables or Fisher exact test for proportions. The Fisher exact test for proportions was used to test between groups differences in changes of BP and HR dipping status from baseline to post-intervention visit.

Table 1.

Baseline demographics, health metrics and eating habits.

	EOF (n = 21)	CN (n = 18)	p
Demographic data
Age (years)	52.1 ± 9.7	52.6 ± 12.5	0.89
Sex (females, n)	16	15	1
Race (White, Black/African American/Asian/Other, n)	17/1/1/2	14/1/1/2	1
Participants employed (n)	16	15	1
Cardiometabolic parameters
BMI (kg/m2)	29.4 ± 4.1	33.7 ± 4.3	0.003
Participants with obesity (BMI ≥30 kg/m2, n)	6	13	0.010
Waist circumference (inches)	97.6 ± 10.5	106.9 ± 10.5	0.009
Hemoglobin A1c %	5.6 ± 0.36	5.6 ± 0.3	0.72
Participants with A1c 5.7 %–6.5 % (n)	8	10	1
Fasting glucose levels (mg/dL)	102.2 ± 9.9	102 ± 9.1	0.99
SBP (mmHg)	128.8 ± 12.2	122.8 ± 12.2	0.23
DBP (mmHg)	76.2 ± 8.8	71.8 ± 8.6	0.77
HR (beat per minute)	72.7 ± 10.3	71.7 ± 9.3	0.77
Nighttime non-dipper for SBP (n)	8	5	0.76
Nighttime non-dipper for DBP (n)	6	3	0.71
Nighttime non-dipper for HR (n)	11	9	1
Participant taking BP medications	2	2	1
Post-menopausal (n)	8/21	6/18	1
Sleep and Activity
Apnea Hypopnea Index (number/hour)	4.6 ± 5.9	3.9 ± 3.8	0.68
PROMIS sleep disturbance questionnaire (T score)	51.3 ± 6.8	54.7 ± 6.6	0.12
Totals sleep time (hour:min)	6:46 ± 00:52	7:09 ± 00:34	0.07
Wake after sleep onset (minutes)	38.8 ± 14	40.6 ± 12.7	0.68
Activity count (a.u.)	238872 ± 65459	228258 ± 92678	0.69
Eating Habits
Overnight fasting duration (hour:min)	11:25 ± 1:00	11:43 ± 0:49	0.37
Time between last meal and bedtime (hour:min)	2:29 ± 1:12	2:34 ± 0:49	0.80

a.u. = arbitrary units; BMI = body mass index; DBP = diastolic blood pressure; HR = heart rate; SBP = systolic blood pressure.

Individual contributions to each analysis along with participant demographics are reported in Table S1. Participants with > 20% missing values (ranging up to 100%) for any complete assessment period (daytime, nighttime, or OGTT) at BL or post-intervention visit were excluded from analysis, as multiple imputation was deemed inadequate for such extensive gaps in time-dependent cardiometabolic measurements. For missing values < 20%, single missing values were imputed based on the nearest values, while multiple missing values for HR/HRV, BP and cortisol profiles were handled using LMM model for between-group analyses. Eleven participants were excluded from nighttime HRV analysis due to ECG artifacts, 15 from nighttime cortisol analysis, 9 from fasting glucose and insulin analysis, 13 from 3-hour OGTT analysis, and 12 from 30-minute insulinogenic index analysis due to blood draw issues or hemolyzed samples, 4 from BP and HR ABPM analysis due to device recording failure; 9 from sleep analysis on night 2 and 2 from night 3 due to PSG recording issues.

The pre-specified primary and secondary outcomes represent interconnected components of autonomic, metabolic, and sleep systems rather than strictly independent endpoints. Several variables are calculated from the same physiological measures, making multiple testing corrections less informative and potentially overly conservative. Therefore, we reported unadjusted p-values for all the outcomes tested and analyzed them as an integrated physiological response ²⁷. Nevertheless, following multiple comparison correction, physiologically relevant secondary findings remain significant, supporting our interpretation that results arise from biological mechanisms rather than chance (Table S1).

All statistical calculations were performed using JMP 16 software (SAS Institute Inc.) and statistical significance was set at P < 0.05. Data are presented as mean ± SDs unless otherwise specified.

RESULTS

Participants

A total of 379 participants were screened and assessed for eligibility. Of these, 55 participants were enrolled in the study and randomly assigned to one of two groups: the EOF intervention or control group. Of these, 39 completed the study and were included in the final analysis (EOF intervention n = 21, control group n = 18, study CONSORT diagram in Figure S1). Baseline demographic information, health metrics and eating habits are shown in Table 1.

Intervention duration and compliance

All participants completed the required 6-week minimum intervention, with an average duration of 7.5 ± 0.7 weeks. The additional time beyond the 6 weeks was provided to coordinate participants’ individual schedules with hospital stay availability. For the participants in the EOF group, compliance to adhering to the assigned overnight fasting window was 88 ± 10 %, to the minimum 3-hour fasting before bedtime was 95 ± 6 %, and to dimming light 3 hours before bedtime was 87 ± 12 %. For the participants in the control group, compliance to dimming lights 3 hours before bedtime was 87 ± 11 %.

Eating patterns

The self-reported overnight fasting duration post-intervention was 14:51 ± 00:51 hour:min in the EOF group and 11:50 ± 00:42 hour:min in the control group (p < 0.001). The time between the last meal and bedtime post-intervention was 4:24 ± 00:33 hour:min in the EOF group and 2:41 ± 00:47 hour:min in the control group (group × visit p < 0.001). There was a significant increase in last meal to bedtime interval in the EOF group (1:55 ± 1:08 hour:min, p < 0.001) but not in the control group (0:07 ± 0:17 hour:min, p = 0.096).

Caloric intake, BMI and waist circumference

Change in total daily calories from baseline to post-intervention was not different between EOF (baseline = 1954 ± 381 Kcal, post-intervention = 1995 ± 356 Kcal) and the control group (baseline = 1934 ± 376 Kcal, post-intervention = 1945 ± 339 Kcal, group × visit p = 0.73). Individual distribution of meals and calories at baseline and post-intervention during the hospital stay for the EOF group is shown in Figure 2, and for the control group in Figure S2.

Figure 2.

Individual distribution of meals and calories at baseline and post-intervention during the hospital stay for participants in the EOF group (n = 21).

Change in BMI from baseline to post-intervention was not significantly different between EOF (baseline= 29.4 ± 4.1 kg/m², post-intervention= 28.9 ± 3.8 kg/m²) and the control group (baseline = 33.7 ± 4.3 kg/m², post-intervention = 33.7 ± 4.5 kg/m², group × visit p = 0.052).

Change in waist circumference from baseline to post-intervention was also similar between EOF (baseline = 97.6 ± 10.5 inches, post-intervention = 97.9 ± 9.9 inches) and the control group (baseline = 106.9 ± 10.5 inches, post-intervention = 107.2 ± 12.3 inches, group × visit p = 0.59).

Nighttime autonomic and metabolic measures

Nighttime HR and HRV

The EOF group exhibited a reduction in average nighttime HR from baseline to post-intervention (Δ = −2.26 ± 3.9 bpm) compared to the control group (Δ = − 0.03 ± 3.4 bpm, group × visit p < 0.001, Figures 3 A, B). Similarly, the EOF group showed a reduction in average nighttime LF/HF from baseline to post-intervention (Δ = −0.39 ± 0.71) compared to the control group (Δ = 0.19 ± 0.97, group × visit p < 0.001, Figures 3 C, D).

Figure 3.

Nighttime autonomic balance and cortisol levels. 5 minutes bins of heart rate (HR, A and B) and LF/HF (ratio between low and high frequency component of heart rate variability, C and D), and hourly serum cortisol levels (E and F) during the first 7 hours of the sleep period starting from the time of lights off, at baseline and post intervention visits in EOF and control groups. EOF group (n = 15) had lower HR, LF/HF from baseline to post intervention compared to the control group (n = 13). EOF group (n = 12) had lower cortisol levels from baseline to post intervention compared to the control group (n = 12). Error bars represent standard errors.

Nighttime serum cortisol

The EOF group had a significant decrease in nighttime cortisol levels from baseline to post-intervention (Δ = − 1.04 ± 1.6 mcg/dL) compared to the control group (Δ = 1.01 ± 1.3 mcg/dL, p = 0.007, Figures 3 E, F).

BP and HR from ABPM

Figure 4 shows baseline and post-intervention 15.5-hour ABPM profiles of systolic BP (SBP), diastolic BP (DBP) and HR obtained at 30-minute intervals from afternoon through the first 7 hours of the sleep period starting from the time of lights off. Average data for the 15.5-hour, daytime and nighttime periods and pulse pressure, and BP and HR nighttime dipping are reported in Table S3.

Figure 4:

15.5-hour blood pressure (BP) and heart rate (HR) profile. Systolic BP (SBP, A and B), diastolic BP (DBP, C and D) and HR (E and F) at baseline and post intervention visits, in EOF (n = 19) and control (n= 16) groups. Data are aligned with the time of lights off. Nighttime DBP was significantly lower in EOF vs control group from baseline to post-intervention (LMM: group × visit p = 0.033). Daytime HR was significantly higher and nighttime HR significantly lower from baseline to post-intervention in EOF vs the control group (LMM: group × visit p < 0.001 and p = 0.029, respectively). Error bars represent standard errors.*p < 0.05; **p < 0.01.

SBP and DBP

Analysis of 15.5-hour SBP and DBP changes from baseline to post-intervention revealed no significant differences between EOF (SBP Δ = 0.3 ± 5.7 mmHg, DBP Δ = − 0.3 ± 3.8 mmHg) and the control group (SBP Δ = 0.7 ± 5.9 mmHg group × visit p = 0.95, DBP Δ = 1.1 ± 4.0 mmHg, group × visit p = 0.58).

During the daytime period, changes in SBP and DBP from baseline to post-intervention were similar between EOF (SBP Δ = 1.1 ± 8.0 mmHg, DBP Δ = 0.8 ± 4.5 mmHg) and the control group (SBP Δ = − 0.3 ± 6.1 mmHg, group × visit p = 0.19, DBP Δ = 0.0 ± 4.1 mmHg, group × visit p = 0.14).

During the nighttime period, there was a significant reduction from baseline to post-intervention in DBP in EOF (Δ = − 1.8 ± 4.7 mmHg) compared to the control group (Δ = 1.6 ± 5.3 mmHg, group × visit p = 0.033). Changes in nighttime SBP from baseline to post-intervention were similar between EOF (Δ = − 0.3 ± 5.7 mmHg) and the control group (Δ = 1.5 ± 8.0 mmHg, group × visit p = 0.071).

Pulse pressure changes from baseline to post-intervention were similar between EOF and control groups during daytime (EOF Δ = 0.3 ± 7.3 mmHg, controls Δ = − 0.4 ± 5.2 mmHg, group × visit p = 0.75) and nighttime (EOF Δ = 1.3 ± 5.1 mmHg, controls Δ= 0.0 ± 5.5 mmHg, group × visit p = 0.49) periods.

HR

Analysis of 15.5-hour HR changes from baseline to post-intervention showed no significant differences between EOF (Δ = 0.1 ± 3.1 bpm) and the control group (Δ = − 0.5 ± 2.9 bpm, group × visit p = 0.11). However, the EOF group had a significant increase in daytime HR (Δ = 1.5 ± 4.1 bpm) and a significant reduction in nighttime HR (Δ = − 1.8 ± 2.7 bpm) from baseline to post-intervention compared to the control group (daytime Δ = − 0.8 ± 3.6 bpm, group × visit p = 0.003, nighttime Δ = − 0.1 ± 3.1 bpm, group × visit p = 0.029).

Nighttime dipping of BP and HR

Change in SBP nighttime dipping was similar between EOF (Δ = − 1.3 ± 6.2 %) and the control group (Δ = 1.5 ± 7.7 %, group × visit p = 0.25). However, significant between-group differences emerged for DBP and HR dipping, with large effect sizes (Cohen’s d = 0.9). The EOF group showed significantly greater nighttime dipping in both DBP (Δ = − 3.5 ± 6.4%) and HR (Δ = − 4.7 ± 5.3%) compared to controls (DBP: Δ = 2.3 ± 7.0%, group × visit p = 0.019; HR: Δ = 1.2 ± 6.4%, group × visit p = 0.003).

For SBP, of 8 participants in the EOF group (one taking BP medication) who were non-dippers (< 10 % nighttime dipping) at baseline, 5 became dippers at post-intervention, while of 5 participants in the control group who were non-dippers at baseline (none taking BP medication) 1 became non-dipper at post-intervention (p = 0.60).

For DBP, of the 6 participants in the EOF group who were non-dippers at baseline (1 taking BP medication) 4 became dippers at post-intervention, 3 participants in the control group who were non-dippers at baseline (none taking BP medications) remained non-dippers at post-intervention (p = 0.49).

For HR, of the 11 participants in the EOF group who were non-dippers at baseline (none taking BP medications) 7 became dippers at post-intervention, of 8 participants in the control group who were non-dippers at baseline (none taking BP medications) 2 became dipper at post-intervention (p = 0.42).

Fasting glucose and insulin, and HOMA-IR

There were no differences between groups in changes of fasting glucose levels (EOF Δ = − 3.5 ± 5.9 mg/dL, controls: Δ = − 3.5 ± 6.6 mg/dL, group × visit p = 0.99), insulin levels (EOF Δ = − 1.6 ± 2.8 μU/mL, controls: Δ = 0 ± 2.7 μU/mL, group × visit p = 0.14), and HOMA-IR (EOF Δ = − 0.5 ± 0.9, controls Δ = − 0.2 ± 1.0, group × visit p = 0.34), from baseline to post-intervention (Table 2).

Table 2.

Fasting glucose and insulin and metabolic measures from a 3-hour oral glucose tolerance test (OGTT).

Variables	EOF (n = 15)		Control (n = 15)
	Baseline	Post-intervention	Baseline	Post-intervention	p
Fasting glucose (mg/dL)	102 ± 10	99 ± 9	102 ± 9	99 ± 9	0.99
Fasting insulin (μU/mL)	12 ± 7	10 ± 6	12 ± 5	12 ± 5	0.14
HOMA-IR [(mg/dL × μU/mL)/405]	3.0 ± 1.8	2.5 ± 1.5	3.3 ± 1.2	3.1 ± 1.1	0.34
30-minute Insulinogenic Index from OGTT* (μU/mL/mg/dL)	0.9 ± 0.4	1.2 ± 1.1	1.9 ± 1.7	1.1 ± 0.5	0.018
Matsuda Index&	28.3 ± 15.8	31.4 ± 15.1	21.5 ± 13.9	23.31 ± 18.9	0.68

HOMA-IR= Homeostatic Model Assessment for Insulin Resistance.

^*= 13 participants in the EOF group and 14 in the control group contribute to 30-minute Insulinogenic Index analysis.

^&= 13 participants in the EOF group and 13 in the control group contribute to Matsuda Index analysis. P values pertain to the interaction term (Group × Visit) of the LMM.

3-hour OGTT

Mean glucose levels during the OGTT after the glucose bolus, were significantly lower from baseline to post-intervention in EOF (Δ = − 5.3 ± 14.7 mg/dL) compared to the control group (Δ = 4.9 ± 13.6 mg/dL, group × visit p = 0.028, Figures 5 A, B). Analysis also showed that glucose levels at 60 minutes after the glucose bolus were significantly lower in EOF compared to controls (p = 0.028).

Figure 5.

3-hour oral glucose tolerance test (OGTT) in EOF (n = 13) and control (n = 13) groups at baseline and post intervention visits. EOF group had glucose levels after the glucose bolus significantly lower from baseline to post intervention compared to the control group, particularly at 60 minutes after the glucose bolus (p = 0.028). Change in insulin levels was similar between groups however, insulin levels at 30 minutes after the glucose bolus were significantly lower in EOF vs controls (p = 0.034). Insulin levels were analyzed using natural log-transformed data due to their non-normal distribution to meet the distributional assumptions of the LMM framework. Error bars represent standard errors. *p < 0.05. 

Mean insulin levels after the glucose bolus, did not differ between EOF (Δ = − 8.4 ± 11.3 μU/mL) and the control group (Δ = − 4.6 ± 13.6 μU/mL, group × visit p = 0.36, Figures 5 C, D). However, insulin levels at 30 minutes after the glucose bolus were lower in EOF vs the control group (p = 0.034).

The insulinogenic index at 30 minutes of the OGTT (Table 2) was significantly higher from baseline to post-intervention in EOF (Δ = 0.3 ± 0.8) compared to the control group (Δ = − 0.8 ± 1.4, group × visit p = 0.018).

Change in the Matsuda index from baseline to post-intervention (Table 2) was not different between EOF (Δ = 3.1 ± 7.5) and the control group (Δ = 1.8 ± 6.9, group × visit p = 0.68).

Sleep macro-architecture

Sleep parameters did not differ between baseline and post-intervention in either the EOF or control groups during Night 2 and Night 3 (Tables S4 and S5). In particular, there were no statistically significant differences between EOF and control groups in total sleep time, sleep efficiency, wake after sleep onset, percentage of non-rapid eye movement (NREM) and REM sleep stages, and arousal index (all p > 0.05).

DISCUSSION

Extended overnight fasting aligned with sleep, in which participants increased their overnight fasting by 3 hours with a minimum three-hour fast before bedtime, resulted in improvements compared to controls in key measures of cardiometabolic health including nighttime autonomic balance, BP and HR nighttime dipping, and glucose regulation, in middle-aged and older adults.

EOF intervention did not significantly change the Matsuda Index of insulin sensitivity, which was a co-primary outcome of the study. This null finding might be explained by several factors. First, half of our population consisted of subjects with normal glucose tolerance, who are primarily sensitive to changes in acute insulin response rather than insulin sensitivity ^28,29. In addition, the relatively small sample size may have provided insufficient statistical power to detect changes in insulin sensitivity. Furthermore, evidence from one study suggests that Matsuda Index might have limited sensitivity to capture changes in insulin sensitivity in longitudinal studies ³⁰. However, EOF had beneficial effects on other secondary measures of glucose regulation. The EOF group showed improved glucose tolerance, including lower one-hour OGTT glucose levels and a higher 30-minute insulinogenic index, both indicators of improved pancreatic β-cell function ^23,31. The acute insulin response is critical to maintain postprandial glucose homeostasis, and its reduction can predict diabetes onset in individuals with normal fasting and OGTT glucose levels ^28,29. These findings collectively suggest that EOF enhanced glucose uptake and utilization, potentially through increased insulin responsiveness. Although some previous TRE studies showed reduction in glucose levels, the EOF intervention in our study did not modify fasting glucose levels. This difference may be due to the inclusion of participants with more severe metabolic disorders and baseline glucose levels exceeding 99 mg/dL ³² in the previous studies.

A notable strength of this study is the continuous BP and HR monitoring at 30-minute intervals across a 15.5-hour period, from afternoon through morning awakening. Most previous studies have evaluated cardiovascular measures at a single timepoint during the day ³³. Consistent with our hypothesis EOF improved nighttime BP and HR regulation and their day-to-night dipping (a primary outcome of the study), which are established indices of cardiovascular disease risk, when compared to the control condition ^34–36. Specifically, the EOF group showed a significant reduction in nighttime DBP (3%) and greater DBP nighttime dipping (average increase of 3.5% in the day-to-night DBP ratio). These changes occurred alongside greater nighttime HR dipping of 5%, suggesting an overall healthier profile of cardiovascular function during sleep. This is further supported by the finding that approximately 60% of participants in the EOF group transitioned from non-dipper to dipper status for SBP, DBP, or HR, compared to approximately 25% in the control group. Given the association between blunted nighttime BP and HR dipping and adverse cardiovascular outcomes, these findings have clinical significance. Non-dipping BP patterns, defined as less than 10% reduction in BP from day to night, are linked to an increased risk of cardiovascular events and target organ damage, independent of BP levels ^35,37. The clinical relevance of our findings is underscored by the evidence that each 5% increase in nighttime BP dipping has been associated with a 17% reduction in cardiovascular events ³⁸. The observed improvement of 3.5% in DBP nighttime dipping under EOF intervention represents a meaningful change with a large effect size that may confer cardiovascular benefits. Furthermore, blunted nighttime HR dip has also been linked to all-cause mortality, further supporting EOF’s cardiovascular benefits ³⁹. Specifically, based on the prospective study from Ben-Dov et al. ³⁹ where each standard deviation improvement in HR dipping was associated with a 20% reduction in all-cause mortality risk, our intervention’s effect on HR dipping could potentially translate to an estimated 14% reduction in long-term mortality risk.

Interestingly, the enhanced day-to-night HR dipping observed with EOF was driven by both significantly lowered nighttime HR and significantly higher daytime HR compared to the control group. This finding is particularly relevant in our population of middle-aged and older adults, who typically experience age-related reductions in the amplitude of circadian and diurnal physiological rhythms, including autonomic activity ¹⁹. Notably, a dampened day-to-night variation in autonomic measures including HR can contribute to incident cardiovascular events and/or progression of cardiovascular disorders ⁴⁰. Our findings suggest that EOF might promote a more robust day/night rhythm of HR, consistent with healthy profile of circadian regulation of autonomic activity ⁴¹.

Enhanced nighttime BP dipping during EOF is consistent with findings from an animal study showing that time-restricted feeding in the active period increases BP dipping in diabetic mice, but interestingly not in wild type mice ⁴². Our study is the first to show this effect in middle-aged and older adults who are more likely to exhibit impaired nocturnal BP dipping ³⁴ due to autonomic (i.e. increased sympathetic and decreased parasympathetic activity) ³⁶, or sleep and circadian factors (i.e. sleep fragmentation and reduced amplitude of autonomic rhythms) ¹⁹. The stronger effect of EOF on DBP compared to SBP dipping may reflect the intervention’s effect on reducing sympathetic activation, and consequently peripheral vascular resistance and cardiac output ⁴³. However, the differential response might also relate to age-related changes in BP regulation, as DBP becomes more sensitive to autonomic modulation with aging ⁴⁴.

The improvement in the nighttime autonomic balance with EOF is further corroborated by its effect on secondary measures of nighttime autonomic function: nighttime HRV and plasma cortisol. Compared to baseline, the EOF group showed lower HR and LF/HF ratio from HRV, changes not observed in the control group, consistent with a decreased nighttime dominance of sympathetic over parasympathetic activity under EOF conditions. The observed 2.3 bpm reduction in average nighttime HR and 0.43 reduction in LF/HF ratio, while seemingly modest, aligns with previous research showing that changes in average nighttime HR and HRV of comparable magnitude in response to intervention were associated with significant improvements in glucose metabolism and insulin sensitivity on the following morning in healthy young adults ⁴⁵. This suggest that improvement in nighttime autonomic balance under EOF might have contributed to enhanced insulin response during the OGTT. Notably, improved nighttime sympathovagal balance has also been shown to correlate with improved insulin sensitivity during intravenous glucose tolerance testing the following morning ⁴⁶.

Improved nighttime autonomic balance in the EOF group was accompanied by a significant reduction in average nighttime plasma cortisol of 1 mcg/dL (12% reduction), providing further evidence of reduced sympathetic activation ⁴⁷. Additionally, one experimental study indicates that even a single night change in cortisol of this magnitude can significantly alter insulin sensitivity and glucose regulation the following morning ⁴⁸.

Despite the significant improvement in nighttime autonomic balance, we did not observe changes in sleep macro-architecture. Prior data examining the effects of TRE on sleep quality show conflicting results, with most studies assessing sleep quality using wearables or diaries ¹⁷. Interestingly, two studies reported a reduction in sleep efficiency with TRE ^49,50, raising the concerns that extended fasting might disrupt sleep through hunger-related arousal mechanisms. This might be particularly relevant in one study where the last meal was served at 3 pm, far from sleep time ⁴⁹. In our study, starting fasting 3 hours prior to sleep did not compromise sleep continuity or architecture. This finding may be particularly relevant for older adults, who often experience age-related sleep disruptions ¹⁹. It is important to note that improvements in cardiometabolic function under EOF conditions were not accompanied by changes in BMI and waist circumference, or caloric intake, which affect cardiometabolic function ⁵¹. Collectively, these findings suggest that EOF can potentially exert beneficial cardiometabolic effects without significant effects on sleep macrostructure or caloric intake.

A novel design element of our study is that we controlled light exposure by reducing ambient lighting before bedtime (< 100 lux) in both groups, thereby reducing the potential impact of the known effects of evening light exposure on circadian rhythms, sleep, and cardiometabolic function ^14–16. The improvements in cardiometabolic measures observed only in the EOF group, despite the similar evening light conditions, indicate that the benefits were driven by the EOF intervention. However, structuring the eating window can impact the regularity of other behaviors, which in turn may have potential additional effects on cardiometabolic function.

Several limitations of our study should be acknowledged. First, the higher proportion of women compared to men may limit the generalizability of the findings to men, as sex is known to affect autonomic, metabolic, and sleep functions. Second, baseline imbalances between groups may have influenced our results: more participants with non-dipper BP status were randomized to the EOF intervention, and the control group had higher baseline BMI and waist circumference, with approximately double the prevalence of obesity compared to the EOF group. However, the observed concomitant improvements in nighttime HR and HRV, which were not different at baseline between the two groups, support the overall beneficial effects of EOF on nighttime cardiovascular function. Additionally, when BMI was introduced into the analyses of primary and secondary outcome measures as a covariate (data not shown), it did not affect the results. Third, the 7.5-week intervention duration, while sufficient to demonstrate significant effects on physiology, may not be sufficient to change weight or waist circumference as seen in longer-term TRE ⁷. Fourth, while physical activity level and light exposure were controlled in the study design, we cannot rule out that subtle changes in activity patterns or light exposure might have contributed to our findings. Lastly, our findings represent the combined effect of extending total overnight fasting duration by 3 hours and implementing a minimum 3-hour pre-sleep fasting period. Their relative contributions or potential synergistic effects cannot be determined from our current study design.

In summary, our results demonstrate that extending the overnight fasting duration by three hours, with at least a 3-hour fasting period before habitual bedtime, enhances the diurnal rhythm of autonomic activity, characterized by reduced sympathetic and increased parasympathetic activity during sleep, promote nocturnal BP and HR dipping, as well as glucose uptake and utilization through more efficient insulin response. These findings have important clinical implications, particularly for middle-aged and older adults at risk for cardiometabolic disease. With an adherence rate nearing 90%, sleep-aligned fasting (EOF) represents a feasible, non-pharmacological intervention for cardiometabolic disease risk reduction. To determine the generalizability and long-term results of this approach, randomized clinical trials powered to assess age and sex differences, and in populations with hypertension and/or diabetes are needed. This study also highlights the potential strength of multimodal interventions such as sleep aligned EOF with other behavioral and environmental factors to target complementary mechanisms to enhance cardiometabolic outcomes.

Clinical Implications.

Existing time restricted eating (TRE) approaches involve self-selected fasting windows without specific alignment with sleep, which is known to play an important role in cardiometabolic regulation. This study demonstrates that extending overnight fasting (EOF) by 3 hours, with meals ending at least 3 hours before bedtime, improves cardiometabolic function during both sleep and daytime in overweight and obese middle-aged and older adults with a habitual fasting duration of 13 hours or less. Specifically, in this 7.5-week randomized controlled trial, the EOF group showed improved day-to-night diastolic blood pressure dipping by 3.5% and heart rate dipping by 5% compared to controls, who maintained habitual eating patterns. EOF also improved the acute insulin response to an oral glucose challenge, indicating better pancreatic β-cell and metabolic function.

This novel approach personalizes time-restricted eating by aligning it with individuals’ circadian sleep-wake rhythm, without changing habitual dietary preferences. Anchoring TRE windows and duration to sleep timing can improve the alignment of the dynamic interactions between metabolic, cardiovascular and sleep physiology. Given its high adherence rate in this study, this may be a more accessible non-pharmacological strategy for improving cardiometabolic health, particularly in middle-aged and older adults who are at higher risk for cardiometabolic disease.

Nonstandard Abbreviations and Acronyms

ABPM

ambulatory blood pressure monitoring

BL

baseline

CRU

clinical research unit

DBP

diastolic blood pressure

LMM

linear mixed model

HRV

heart rate variability

HOMA-IR

homeostatic model of insulin resistance

LF/HF

low frequency to high frequency ratio

MMSE

mini mental state examination

NREM

non rapid eye movement

OGTT

oral glucose tolerance test

PSG

polysomnography

REM

rapid eye movement

SBP

systolic blood pressure

TRE

time restricted eating