Insufficient Sleep and Its Effect on Body Weight and Composition: A Pooled Analysis of Randomized Trials

Abstract

Background:

Insufficient sleep is associated with obesity. However, the causal effect on weight status of chronic, mildly insufficient sleep and its potential variability by sex and menopausal status remain unknown.

Objective:

To explore the effect of 6 weeks of sleep restriction (SR) of 1.5 hours per night on energy balance and body weight regulation.

Design:

Pooled analysis of 2 randomized crossover trials. (ClinicalTrials.gov: NCT02960776 and NCT02835261)

Setting:

Outpatient intervention with inpatient and outpatient assessments.

Participants:

Adults (n = 95) aged 20 years or older at elevated cardiometabolic risk with habitual sleep of 7 or more hours per night.

Intervention:

Six weeks of sustained adequate sleep (AS) and SR of 1.5 hours per night separated by a multiweek washout.

Measurements:

Outcome measures included adiposity (assessed using magnetic resonance imaging), body weight, waist circumference, and energy balance behaviors and biomarkers.

Results:

Sleep duration was reduced by 78.4 minutes (95% CI, −83.5 to −73.3 minutes) per night with SR versus AS. Body weight (0.45 kg [CI, 0.33 to 0.57 kg]), waist circumference (0.52 cm [CI, 0.25 to 0.79 cm]), and whole-body volume (0.56 L [CI, 0.19 to 0.93 L]) were increased with SR relative to AS. Leptin levels were elevated with SR versus AS (2.03 ng/mL [CI, 0.38 to 3.68 ng/mL]). Sedentary time was increased by 17.2 minutes (CI, 11.7 to 22.7 minutes) per day with SR versus AS.

Limitations:

The intervention duration may have been too short to identify changes in body composition, power to evaluate individual differences was limited, and effect sizes were modest.

Conclusion:

Prolonged exposure to moderately short sleep may lead to weight gain, suggesting that weight management and cardiometabolic disease prevention programs should consider incorporating sleep strategies to promote AS.

INTRODUCTION

Obesity is a primary risk factor for cardiovascular disease (CVD) (1). Adopting changes in diet and physical activity to achieve a negative energy balance reduces body weight and adiposity; however, achieving and maintaining these behavior changes is difficult, as illustrated by unencumbered increases in the prevalence of obesity (2). Identifying additional modifiable behaviors that may affect weight management could enhance strategies for obesity treatment and cardiometabolic disease prevention.

Sleep is increasingly recognized as an essential factor for cardiovascular health (3). Insufficient sleep is highly prevalent (4, 5) and is consistently associated with elevated risk for obesity (6) and poorer weight loss outcomes (7). Yet existing support for a causal relation between sleep and obesity relies on inpatient trials demonstrating shifts toward a positive energy balance in response to severe sleep restriction (SR) (4 to 5 hours per night) over short-term periods (≤2 weeks) (8, 9), thereby limiting real-world relevance. Thus, causal evidence that typical patterns of short sleep (roughly 6 hours per night) contribute to obesity remains unavailable (10).

The goal of the present study was to assess whether “real-world” chronic short sleep adversely affects body weight regulation in adults. We hypothesized that 6 weeks of mild SR would increase body weight and adiposity compared with adequate sleep (AS) due to adverse changes in energy balance. Furthermore, we aimed to explore whether the effects of SR are more pronounced in women than in men, which would align with higher rates of obesity observed in this group (11) along with previous findings of exacerbated associations between short sleep and adiposity (12) and greater increases in food intake in response to severe SR (13) in women compared with men.

METHODS

Design Overview

This investigation stems from 2 randomized crossover clinical intervention studies of SR in adults conducted between 2016 (start of recruitment) and 2023 (end of follow-up). Both trials included two 6-week interventions performed in outpatient settings: a mild SR intervention, in which participants delayed their bedtime by 1.5 hours per night compared with usual sleep, and maintenance of habitual AS, in which participants maintained their habitual sleep duration of 7 or more hours per night. Sleep intervention periods were separated by a 4- to 6-week washout interval. The sequence of the intervention was randomly allocated in each trial. Due to the nature of the intervention, participants were not blinded to it.

The trials included largely the same assessments, but one recruited women only and primarily focused on cardiometabolic risk factors (Trial 1 [ClinicalTrials.gov: NCT02835261]), whereas the other recruited women and men and primarily focused on measures of energy balance regulation (Trial 2 [ClinicalTrials.gov: NCT02960776]). The 2 trials had largely concordant eligibility criteria, with minor differences (Supplement Table 1, available at Annals.org). Although they were not prespecified, we performed pooled analyses due to the similarities in the 2 study interventions to allow us greater power to detect effects, particularly among subgroups.

Settings and Participants

Persons aged 20 years or older were recruited at Columbia University Irving Medical Center from the New York City metropolitan area via advertisements on fliers and websites and were screened using a multistep process. Prospective participants first completed a brief phone screening followed by an in-person screening visit. During the screening visit, individuals had their height and weight measured and completed questionnaires on health history and lifestyle behaviors. In Trial 1, participants were enrolled over a range of body mass index (BMI) values spanning normal weight to obesity (18.5 to 34.9 kg/m²); in Trial 2, participants with overweight or class I obesity (BMI of 25 to 33.5 kg/m²) or with normal weight (BMI of 20 to 24.9 kg/m²) and 1 parent with type 2 diabetes, hyperlipidemia, or CVD were eligible. Both trials excluded persons who had cardiovascular, metabolic, or neurologic disease; reported sleep or psychiatric disorder symptoms, including eating disorders; were at high risk for obstructive sleep apnea; had weight loss greater than 5 kg in the previous 3 months or reported efforts to achieve weight change; were taking medications known to affect sleep or appetite; consumed large amounts of caffeine (>300 mg/d); had an extreme morning or evening chronotype; had done shift work or traveled across time zones in the previous month; operated heavy machinery; or were commercial long-distance drivers. Persons with a history of drug or alcohol misuse were excluded. In both trials, women were excluded if they were pregnant, were less than 1 year postpartum, were planning to become pregnant, or were using hormonal contraceptives or menopausal hormone therapy. Supplement Table 1 provides information on similarities and differences between Trials 1 and 2. All procedures in each trial were approved by the university’s institutional review board. Participants provided written informed consent before participation and agreed to refrain from operating motor vehicles during the study.

Participants who remained eligible after initial screening then completed a 2-week home-based screening to evaluate their usual sleep duration using wrist actigraphy (ActiGraph GT3X+ [ActiGraph LLC]) and sleep diaries. Participants who slept at least 7 hours on at least 70% of the screening nights were invited to participate and were scheduled for their pre-randomization baseline visit (Supplement Figure 1, available at Annals.org). Randomization occurred upon completion of those assessments. A total of 95 men and women were enrolled across the 2 randomized crossover trials (Trial 1: n = 50; Trial 2: n = 45).

Randomization and Interventions

Sleep intervention conditions were identical in both trials. The order of the sleep conditions was randomly assigned separately across participants from each trial and was stratified by menopausal status in Trial 1 and sex in Trial 2. A study statistician who was blinded to the study conditions provided the order assignments, determined using an electronic random sequence generator. The order of sleep conditions was revealed to the participant by the research assistant upon completion of baseline measures preceding the first intervention period. The research assistant did not have a priori knowledge of the allocated sequence.

During AS, participants were instructed to maintain their usual bedtimes and wake times to achieve total sleep time (TST) of at least 7 hours per night. In the SR intervention, participants were instructed to delay their bedtime (based on their average time from screening) to achieve a reduction in TST of about 1.5 hours per night. Sleep schedules were personalized based on habitual sleep patterns from the actigraphy sleep screening. Adherence was monitored using wrist actigraphy, worn on the nondominant wrist at all times throughout each intervention period, and nightly sleep diaries, which were reviewed at least biweekly by the research assistant. Adjustments were made to the sleep schedule to ensure that target TSTs were achieved in each sleep condition.

Outcomes and Follow-up

Anthropometry, and particularly adiposity, was the main focus of the pooled analysis. Additional outcomes of interest for the pooled analyses included physical activity and energy balance–regulating hormones. Energy expenditure was not part of the pooled analyses as it was only measured in Trial 2. Information on all outcomes measured in each trial is provided in Supplement Table 1.

Anthropometric Measurements and Body Composition

Anthropometric measurements were taken at each study visit. Height, weight (Tanita WB-3000 scale), and waist circumference (WC) were measured in duplicate by trained staff using standard procedures (https://wwwn.cdc.gov/nchs/nhanes/ContinuousNhanes/Manuals.aspx?BeginYear=2015). The scale was calibrated to the nearest 0.1 kg and the stadiometer was calibrated to the nearest 0.1 cm; BMI was calculated as the ratio of weight in kilograms to the square of height in meters. Waist circumference was measured at the level of the umbilicus. Baseline and end point measurements were taken in the morning after an overnight fast lasting more than 12 hours.

Whole-body magnetic resonance imaging (MRI) (1.5T 6X Horizon system [General Electric]) was performed to measure adiposity at baseline and the end point of each intervention period. Participants were positioned in the scanner in a supine position with their arms stretched overhead. The scanner was landmarked at the L4-L5 intervertebral disc, with 1-cm slice images taken from L4-L5 to the top of the fingers and from L4-L5 to the end of the toes. Image slices were spaced 4 cm apart for the upper and lower body and 2 to 4 cm apart for the visceral area (10 cm below to 20 cm above L4-L5). Imaging analysis was performed to quantify whole-body and tissue volumes (in liters) using specialized software (sliceOmatic [TomoVision]) (14). The estimated volume of each tissue was calculated based on the slice thickness and interslice interval, assuming a cylindrical shape.

Biomarkers of Energy Balance Regulation

Whole blood was collected at baseline, week 3, week 4 (subset), and the end point of each intervention period after an overnight fast lasting more than 12 hours. Before processing, some samples were treated to inhibit the cleavage of peptides. Processed samples were aliquoted and banked for batch analyses of leptin, glucagon-like peptide-1 (GLP-1), and ghrelin (radioimmunoassays [MilliporeSigma]).

Sleep and Activity

Rest and activity levels were assessed during screening and every day of each intervention period using wrist actigraphy (ActiGraph GT3X+). Sleep was scored using the Cole-Kripke algorithm (ActiLife, version 6); sleep diaries were used to enhance the accuracy of bedtime and wake time estimates (15).

Raw accelerometer data were also processed using the R package GGIR (version 2.6-0) (https://wadpac.github.io/GGIR) (16) to obtain information on physical activity. The program calibrated accelerometer data using local gravity as a reference and identified periods of sustained abnormally high values (17). Data processing included detection of non-wear time (18); days with wear time less than 16 hours were excluded from analysis. For valid wear days, detected non-wear periods were imputed using the participant-specific average at similar time points on other days. Physical activity outcomes were based on calculation of the average magnitude of dynamic acceleration corrected for gravity, averaged over 5-second epochs; thresholds to distinguish sedentary behavior, light-intensity physical activity, and moderate-to-vigorous physical activity (MVPA) were 0 to 49, 50 to 99, and 100 or more mG, respectively. Times spent in each category were summed across each valid wear day with sleep times excluded. The current investigation focused on daily time spent sedentary and engaging in MVPA.

Total Daily Energy Expenditure

Free-living total daily energy expenditure (TDEE) was assessed at the end point of each intervention period (n = 34 [Trial 1]) using doubly labeled water. At the week 4 visit, participants provided a urine sample after an overnight fast lasting more than 12 hours. They were given a dose of doubly labeled water providing 0.23 g of ¹⁸O per kilogram of total body water and 0.15 g of ²H per kilogram of total body water, which was estimated from body weight using standard equations. To ensure that the labeled isotope was consumed entirely, the cup was twice refilled with tap water, and the participant consumed the full amount each time. The time of dosing was recorded, and urine was collected 3 and 4 hours later. Participants collected postdosing urine samples at 24 hours, 7 days, and 14 days. All urine specimens were processed and frozen for batch analyses using established methods (19–21).

Statistical Analysis

Descriptive characteristics of the full sample and subgroups (men, premenopausal women, and postmenopausal women) assessed at screening are presented as means and SDs for continuous variables and as numbers and percentages for categorical variables.

Linear mixed models were used to assess effects of sleep condition (SR vs. AS) on outcomes. The outcome of TST was assessed for fidelity to the sleep intervention protocol. Primary outcomes were body weight, WC, and MRI-derived tissue distribution; secondary outcomes were biomarkers of energy balance regulation, time spent sedentary and engaging in MVPA, and TDEE. Except for TDEE, which was assessed only once per intervention period, dependent variables were postbaseline values, and models were adjusted for baseline values from the corresponding intervention period (screening values for actigraphy measures).

In addition to baseline (or screening) values of the outcome measure assessed, the final models included, at a minimum, sleep intervention condition (SR vs. AS), study week, trial (Trial 1 vs. Trial 2), and phase order (SR followed by AS vs. AS followed by SR) as fixed effects and participant as a random effect. For actigraphy-measured outcomes, day was also included as a fixed effect. For times spent sedentary and engaging in MVPA, time in bed was also included as a fixed effect in all models to account for the difference in wake hours between sleep conditions. In addition to the aforementioned fixed effects, the initial models also included participant age, gender, and race or ethnicity; non-significant terms were removed from the final models. For all outcomes, we also tested separately for trial-by-sleep-condition interaction to confirm that treatment effects did not differ by trial before proceeding to the main effect analyses. Of note, the trial-by-sleep-condition interaction was not significant for any outcome at a P value of 0.05 and is therefore not described in the results. For exploratory purposes, analyses were repeated after stratification by gender and menopausal status (men vs. premenopausal women vs. postmenopausal women).

All analyses were conducted as intention-to-treat. Results of mixed models are presented as baseline-adjusted mean differences between sleep conditions (SR vs. AS) and 95% CIs or least-squared means and 95% CIs. Results were considered significant at a P value less than 0.05 (P values are not reported).

Role of the Funding Source

The study funders, the National Institutes of Health and the American Heart Association, were not involved in the design or conduct of the study and did not participate in statistical analyses, manuscript preparation, or data interpretation. They played no role in the decision to publish these findings.

RESULTS

Study Sample

A total of 4147 persons were screened for eligibility, of whom 10.4% underwent actigraphy screening (Supplement Figure 1). Of those who completed the full screening process (n = 433), 21.9% underwent baseline assessments and were randomly assigned to an intervention order (n = 95). The study included 72 (76%) women, 15 of whom were postmenopausal (12 in Trial 1 and 3 in Trial 2) (Table 1). Most participants identified as a member of a racial or ethnic minority group.

Table 1.

Baseline Characteristics of Enrolled Participants

Characteristic	Full Sample (n = 95)	Gender/Menopausal Status			Randomization Order
		Men (n = 23)	Premenopausal Women (n = 57)*	Postmenopausal Women (n = 15)†	AS First (n = 46)	SR First (n = 49)
Mean age (SD), y	34.2 (12.6)	32.2 (10.5)	29.2 (6.6)	56.7 (8.6)	32.9 (12.2)	35.5 (13.0)
Mean body mass index (SD), kg/m2	25.6 (3.3)	26.2 (3.7)	24.9 (2.8)	27.4 (3.9)	25.5 (3.2)	25.7 (3.5)
Racial minority, n (%)‡	50 (52.6)	12 (52.2)	30 (52.6)	8 (53.3)	21 (45.7)	29 (59.2)
Ethnic minority, n (%)§	24 (25.3)	8 (34.8)	11 (19.3)	5 (33.3)	10 (21.7)	14 (28.6)
Highest education below college degree, n (%)	16 (16.8)	6 (26.1)	6 (10.5)	4 (26.7)	9 (19.6)	7 (14.3)
Not working full-time, n (%)	47 (49.5)	9 (39.1)	30 (52.6)	8 (53.3)	24 (52.2)	23 (46.9)
Mean Pittsburgh Sleep Quality Index score (SD)	2.8 (1.7)	2.4 (1.2)	2.8 (1.8)	3.5 (2.1)	2.7 (1.7)	2.9 (1.7)
Mean objective sleep duration (SD), min	455.2 (22.8)	457.7 (25.6)	456.0 (23.1)	448.3 (16.1)	450.9 (18.9)	459.2 (25.5)

AS = adequate sleep; SR = sleep restriction.

^*Trial 1: n = 38; Trial 2: n = 19.

^†Trial 1: n = 12; Trial 2: n = 3.

^‡Any race other than White, based on self-report.

^§Hispanic/Latino/Latina ethnicity, based on self-report.

Sleep Duration

Nightly sleep duration was reduced relative to screening by 78.4 minutes (95% CI, −83.5 to −73.3 minutes) with SR versus AS (367.5 [CI, 363.1 to 371.9] vs. 445.8 [CI, 441.6 to 450.0] minutes per night). Moreover, among participants who completed the SR intervention, 67 (80.7%) achieved a nightly reduction in TST of at least 75 minutes; the majority (39 [58%]) reduced their TST by 90 or more minutes per night (Figure 1). Differences in TST between SR and AS did not differ between trials.

Figure 1.

Means and 95% CIs (raw data) for nightly TST at screening (represented as week 0) and each week of the intervention periods (top panel), individual-level changes in TST from screening during the AS intervention period (middle panel), and individual-level changes in TST from screening during the SR intervention period (bottom panel). Dashed lines indicate 75- and 90-minute reductions from screening in TST. The figure includes participants from intention-to-treat analyses. AS = adequate sleep; SR = sleep restriction; TST = total sleep time.

Body Weight, Adiposity, and Biomarkers of Energy Regulation

Body weight increased by 0.45 kg (CI, 0.33 to 0.57 kg) with SR versus AS in the full sample (Figure 2). In addition, SR led to elevated WC compared with AS (Table 2). For MRI-derived outcomes, an effect of sleep condition was also observed for whole-body volume, which increased by 0.56 L (CI, 0.19 to 0.93 L) with SR compared with AS. However, there was no effect of sleep condition on percentage of whole-body volume as adipose or skeletal muscle tissue (Table 2).

Figure 2.

Means and 95% CIs (raw data) for body weight at baseline, week 3, and the end point of each intervention period (left panel) and means and 95% CIs for changes (raw data) from baseline to the end point of the SR and AS intervention periods (right panel). The figure includes participants from intention-to-treat analyses. AS = adequate sleep; SR = sleep restriction.

Table 2.

Measures of Anthropometry, Body Composition, and Energy Balance Biomarkers and Behaviors After 6 Weeks of AS and Mild Sleep Curtailment

Measure	SR*	AS*	SR vs. AS†
Body weight, kg	71.6 (71.4 to 71.7)	71.1 (71.0 to 71.3)	0.45 (0.33 to 0.57)
Waist circumference, cm	90.9 (90.4 to 91.5)	90.4 (89.8 to 91.0)	0.52 (0.25 to 0.79)
WBV, L	68.4 (68.1 to 68.7)	67.9 (67.6 to 68.1)	0.56 (0.19 to 0.93)
Total adipose tissue, percentage of WBV	35.6 (35.4 to 35.8)	35.6 (35.4 to 35.8)	−0.07 (−0.36 to 0.22)
Subcutaneous adipose tissue, percentage of WBV	32.5 (32.3 to 32.6)	32.5 (32.3 to 32.6)	0.01 (−0.24 to 0.26)
Visceral adipose tissue, percentage of WBV	2.12 (2.08 to 2.17)	2.18 (2.14 to 2.22)	−0.06 (−0.12 to 0.00)
Skeletal muscle, percentage of WBV	34.1 (34.0 to 34.3)	34.0 (33.9 to 34.1)	0.13 (−0.07 to 0.33)
Fasting leptin level, ng/mL	26.5 (24.9 to 28.1)	24.4 (22.8 to 26.0)	2.03 (0.38 to 3.68)
Fasting ghrelin level, pg/mL	811.8 (783.9 to 839.7)	839.1 (811.9 to 866.2)	−27.3 (−56.7 to 2.1)
Fasting glucagon-like peptide-1 level, pmol/L	21.4 (20.1 to 22.7)	21.6 (20.3 to 22.9)	−0.26 (−1.38 to 0.86)
Sedentary behavior, min/d	805.9 (798.4 to 813.3)	788.6 (781.3 to 795.9)	17.2 (11.7 to 22.7)
Moderate-to-vigorous physical activity, min/d	88.1 (84.7 to 91.6)	88.2 (84.8 to 91.6)	−0.1 (−2.8 to 2.7)
Total daily energy expenditure, kcal	2439 (2283 to 2595)	2415 (2262 to 2567)	24 (−113 to 161)

AS = adequate sleep; SR = sleep restriction; WBV = whole-body volume.

^*Postbaseline value (95% CI) from linear mixed models adjusted for baseline value of the corresponding measure as well as (at minimum) assessment time point, trial, and phase order (intention-to-treat analysis).

^†Mean difference (95% CI) of postbaseline values between SR and AS from linear mixed models adjusted for baseline value of the corresponding measure as well as (at minimum) assessment time point, trial, and phase order (i.e., main effect of sleep condition) (intention-to-treat analysis).

Concordant with changes in weight status, fasting leptin levels were elevated with SR versus AS (Table 2). Circulating ghrelin tended to be lower with SR than AS; however, these differences did not reach statistical significance. There was no effect of sleep condition on fasting levels of GLP-1.

Physical Activity, Sedentary Time, and TDEE

In the full sample, sedentary time increased by 17.2 minutes (CI, 11.7 to 22.7 minutes) per day with SR compared with AS; there was no effect of SR on time spent engaging in MVPA (Table 2). Among the 34 participants for whom TDEE data were available (15 women; mean age, 28.8 years [CI, 27.1 to 30.5 years]; mean BMI, 26.1 kg/m² [CI, 24.9 to 27.2 kg/m²]), no effect of sleep condition was observed.

Subgroup Analysis

Results of exploratory analyses stratified by gender and menopausal status are presented in Supplement Table 2 (available at Annals.org).

DISCUSSION

The current pooled analyses show that, compared with AS, 6 weeks of mild SR increased body weight with no detectable difference in relative adiposity. Accordingly, increases in body weight with SR corresponded to numerically higher concentrations of leptin, a marker of energy stores (22). Finally, in all groups, sedentary time, adjusted for time in bed, was increased with SR compared with AS.

Before the current study, a 3-week parallel-group study in which healthy young men either curtailed their sleep by 1.5 hours per night or maintained habitual sleep of more than 7 hours per night for 3 weeks showed no difference in weight change between conditions (23). However, closer visual inspection of the temporal patterns of weight fluctuations showed an initial decrease followed by a marked increase after 2 weeks of SR, suggesting a potential threshold of accumulated sleep debt that is needed to affect body weight. Support for this is provided through comparison of sleep debt and weight accumulations between the current study and a previous inpatient trial by Covassin and colleagues (8). In that trial, participants maintained a 5-hour sleep opportunity for 2 weeks, resulting in a sleep debt of 2955 minutes and yielding a weight gain of about 0.5 kg relative to the adequate sleep duration control. In the present study, the average sleep debt accumulated was 3280 minutes over 6 weeks (roughly 10% higher than in Covassin and colleagues' study [8]), with a similar between-condition difference in weight change of about 0.5 kg. These findings suggest that cumulative sleep debt likely affects body weight.

Although 6 weeks of mild SR increased whole-body volume determined by MRI, the gold standard tool for body composition assessment, it did not disproportionately affect adipose versus lean tissue distribution; that is, there was no preferential deposition of adipose tissue due to SR. Although our study may be underpowered to detect small effects in tissue distribution, the finding aligns with the previous observation by Covassin and colleagues; compared with healthy sleep, 2 weeks of severe SR does not differentially affect body fat percentage (8). The authors did, however, observe an absolute increase in abdominal adiposity, measured using dual-energy x-ray absorptiometry, in response to SR. In our study, we found that 6 weeks of SR increased WC relative to AS. Evaluation of changes in body composition over longer periods of adequate and mild short sleep will be needed to confirm this effect.

Another aim of this investigation was to gain insight into mechanisms underlying weight change due to chronic SR. We (13) and others (9, 24) previously showed that acute, extreme curtailment of sleep increases energy intake by about 300 kcal/d. Although a topic of debate (25), one potential explanation for this effect is dysregulation of hormonal control of appetite (26). In our study, we observed a minimal effect of 6 weeks of SR of 1.5 hours per night on fasting levels of the orexigenic hormone ghrelin and the anorexigenic hormone GLP-1. Interestingly, ghrelin levels tended to decrease in response to mild SR, which may suggest suppressed hunger despite weight gain; however, the wide degree of imprecision around estimates precludes meaningful interpretation of the effect of mild SR on physiologic determinants of hunger. Leptin was increased, corresponding with the increase in body weight observed with SR. It is well established that decreases in leptin drive eating to replenish energy stores while increases in leptin have negligible effects (22, 27). Possible explanations for inconsistent findings in the field could be that fasting levels of energy balance hormones do not reflect their dynamic profiles (as levels oscillate during the 24-hour day) or do not translate well to actual eating behaviors.

Daily sedentary time, assessed from actigraphy, increased during SR compared with AS in the absence of changes in MVPA. That prolonged SR did not affect MVPA is unsurprising given that participants were generally sedentary at study onset. The finding that SR increased sedentary time aligns with results from observational studies (28). If replicated, this finding can have important clinical relevance, as greater time spent sedentary may contribute to adiposity gains (29) and predicts elevated risk for cardiometabolic diseases and mortality (30–33).

As in previous studies of extreme, acute SR (13, 34), we observed no difference in TDEE from doubly labeled water between sleep intervention conditions. When using whole-room indirect calorimetry, higher 23-hour energy expenditure has been observed in response to severe SR (35, 36) owing to the added energetic cost of maintaining wakefulness. That body weight is increased with mild SR versus AS in the absence of changes in TDEE in the free-living context of the current study further supports the notion that participants are compensating for the added costs of wakefulness through increased sedentary time and energy intake (9, 24, 37). Although inferred increases in daily energy intake in response to mild SR are likely smaller than those of short-term, severe SR (24), their cumulative effects over time still lead to weight gain.

The limitations of our study reflect the need to balance intervention duration with feasibility and adherence to an intensive protocol. Given the small sample size, generalizability was limited. The study also was not powered to formally test whether treatment effects differed by sex and menopausal status. In addition, although this study reflects a long intervention period for the field, the 6-week duration of SR may have been too short to observe substantial changes in body tissue distribution. Metabolic assessments were limited to morning fasted state to reduce participant burden and attrition. The lack of additional measures across the 24-hour day precludes any observations of intervention effects on metabolites known to fluctuate with circadian rhythmicity, such as ghrelin, and limits our interpretation of those findings.

Despite these limitations, the randomized crossover design, outpatient intervention setting, and detailed objective outcome measures allowed us to establish causality while maintaining a high degree of ecological validity, closely modeling insufficient sleep experienced regularly by U.S. adults. Adherence to the sleep intervention conditions was excellent, and SR reflected “real-world” short sleep, where 78% of adults reporting less than 7 hours of sleep per night report sleeping 6 hours (38). The sample size, although modest in absolute terms, is considered large in the field of experimental sleep curtailment.

The increase in body weight of about 0.5 kg attributable to AS is equivalent to the average annual weight change among young to middle-aged adults (39) and provides causal evidence for a role of mildly insufficient sleep in risk for weight gain. Given the positive relationship between weight status and CVD incidence (1), these findings highlight a potential pathway by which chronic short sleep could accelerate the development of CVD among those with existing risk factors. Future studies are needed to reliably determine the role of sex hormones in influencing the effect of short sleep on body composition.

In conclusion, prolonged mildly short sleep leads to weight gain in adults at elevated risk for cardiometabolic disease. The magnitude of change mirrors population trends in annual weight gain during early and middle adulthood (39) and may contribute to increased risk for obesity and cardiometabolic disorders over time. These findings highlight the importance of discussing sleep duration at health care encounters and support guidance to maintain adequate sleep duration to improve weight management and obesity prevention across the lifespan.

Data Sharing Statement:

Available with the article online.