Effect of sleep restriction on insulin sensitivity and energy metabolism in post-menopausal women: A randomized, crossover trial

Abstract

Objective.

Investigate the effect of sleep restriction (SR) on insulin sensitivity and energy metabolism in postmenopausal women.

Methods.

In a randomized crossover trial, 14 women underwent 4 nights of habitual sleep (HS, 100% normal sleep) and SR (60% of HS) while following a eucaloric diet. Outcomes included: (1) insulin sensitivity by hyperinsulinemic-euglycemic clamp, defined as the glucose infusion rate (GIR); (2) resting metabolism and substate oxidation by indirect calorimetry; and (3) glucose, insulin, and C-peptide concentrations following a standard meal test.

Results.

Nine postmenopausal women (age 59±4 years, BMI 28.0±2.6 kg/m²) were analyzed. Accelerometer-determined total time in bed was 8.4±0.6 hours during HS vs. 5.0±0.4 hours during SR (38% reduction, p<0.0001). SR reduced low-dose insulin GIR by 20% (HS: 2.55±0.22 vs. SR: 2.03±0.20 mg/kg/min; p=0.01) and high-dose insulin GIR by 12% (HS: 10.48±0.72 vs. SR: 9.19±0.72 mg/kg/min; p<0.001). SR reduced fat oxidation during high-dose insulin infusion (p<0.01) and did not alter resting energy metabolism.

Conclusions.

Four nights of SR reduced insulin sensitivity and fat oxidation in postmenopausal women. These findings underscore the role of insufficient sleep on metabolic dysfunction in post-menopausal status. Larger trials investigating how sleep disturbances cause metabolic dysfunction during menopause are needed across all stages of menopause.

1. Introduction

Menopause is the foremost acceleration point of cardiometabolic disease risk in women. During the menopause transition, dramatic changes in endogenous sex hormones coincide with increases in body weight, fat mass, dyslipidemia, hypertension, and vascular dysfunction (1, 2) . These menopausal changes contribute to a worsened cardiometabolic profile that is independent of chronological aging (2). Sleep disturbances―particularly shortened or fragmented sleep―are a highly prevalent menopausal symptom that are also associated with increased cardiometabolic risk (3, 4). In the Study of Women’s Health Across the Nation (SWAN), the prevalence of sleep disturbances ranged from 16% to 42% in premenopausal women and increased to 39% to 47% and 35% to 60% in perimenopausal and postmenopausal women, respectively (4). Yet, the degree to which sleep disturbances contribute to the unfavorable changes in cardiometabolic risk have not been fully elucidated in menopausal women.

Experimental trials investigating the mechanisms linking sleep disturbances to adverse cardiometabolic risk have predominantly been restricted to men only (5, 6, 7) or in combination with premenopausal women (8, 9, 10, 11, 12). These studies observe impaired glucose tolerance and reduced insulin sensitivity following short-term sleep restriction (SR) or fragmentation. The lack of sleep-related studies in women―particularly among perimenopausal and postmenopausal women―is partly due to the challenges of controlling for menstrual cycle phase (follicular vs. luteal), measuring sex steroids and gonadotropins during the menstrual cycle, as well as the difficulty interpreting the hormonal changes during perimenopause and into the postmenopausal years (2, 13, 14).

To date, only one experimental study has examined the within-subject effect of sleep disturbances during menopause on energy metabolism (15). In this well-controlled study, both normal sleep and sleep fragmentation were administered to premenopausal women as well as a subset of women undergoing experimental menopause (via administration of gonadotropin-releasing hormone (GnRH) agonist leuprolide acetate) (15). Compared to normal sleep, short-term sleep fragmentation reduced fat oxidation (by indirect calorimetry) in premenopausal women; however, sleep fragmentation in the experimental menopause state (hypo-estrogenized) did not further reduce fat oxidation (15). To the best of our knowledge, no other studies have investigated how sleep disturbances (e.g., shortened and fragmented sleep, poor sleep quality, etc.) alter metabolism during perimenopause and into the postmenopausal years.

Our overall objective was to investigate the effect of SR on whole-body insulin sensitivity and energy metabolism in postmenopausal women. We hypothesized that SR would reduce insulin sensitivity (by 2-step hyperinsulinemic-euglycemic clamp), as well as reduce fat oxidation under basal and insulin-stimulated conditions (by indirect calorimetry). We also hypothesized that SR would increase insulin secretion to maintain glucose homeostasis (in response to a standard meal test).

2. Methods

The trial was conducted at the Pennington Biomedical Research Center (PBRC; Baton Rouge, Louisiana, USA), approved by the PBRC Institutional Review Board, and conducted in accordance with the Helsinki Declaration of 1975. The study was preregistered at clinicaltrials.gov prior to enrolling participants [NCT04286451]. Participants were recruited from the greater Baton Rouge area between July 2020 and April 2022 via listserv, flyers, social media, and website advertisements.

2.1. Participants

Postmenopausal women of all races and ethnicities with a BMI between 25 and 35 kg/m² (inclusive) and self-reported ≥6.5 hours of habitual nightly sleep were eligible. Postmenopausal status was defined by self-reported absence of menses for at least 1-year (3, 16). Women had to be within 10 years of their last menstrual cycle; have nocturnal time in bed of ≥6.5 hours per night (for at least 5 out of 7 days each week); and agree to abstain from napping or moderate-to-vigorous physical activity during both sleep conditions. Habitual time in bed was self-reported and further confirmed by 7-day wrist accelerometry combined with a sleep diary (during the run-in period). Key exclusion criteria included history of diabetes (or fasting blood glucose >126 mg/dL), unstable weight in the last 3 months (i.e., gain or loss ≥5% of body weight), concomitant illnesses (e.g., polycystic ovary syndrome, sleep apnea), use of medications that could affect sleep or metabolism (e.g., antidepressants, antipsychotics, hypnotics, sedatives), or inability to curtail caffeine consumption. A full list of inclusion and exclusion criteria are presented in Supplementary Table 1. All women provided written informed consent prior to completing any trial-related activities, including determination of eligibility.

2.2. Trial Design

This randomized, single-blinded, crossover pilot clinical trial investigated the effect of SR on insulin sensitivity and energy metabolism in postmenopausal women. SR was achieved by limiting time in bed. Women were randomized to follow either 4 nights of habitual sleep (100% of normal sleep duration confirmed by 7-day accelerometry and a sleep diary) or 4 nights of SR (40% reduction in habitual sleep, including 20% delayed sleep onset and 20% early waking) and then crossover to the other arm after a minimum 3-week washout period. The study was conducted primarily in an outpatient setting with women sleeping in the inpatient unit only during SR. To minimize confounding from diet, the study provided eucaloric meals (50% carbohydrate, 30% fat, and 20% protein) throughout each sleep condition to meet weight maintenance energy requirements and ensure comparable macronutrient meal composition. An energy requirement equation generated from doubly-labeled water and based on age, sex, fat mass, and fat-free mass was used (17). While participants food preferences were considered to ensure consumption of meals /snacks provided, information related to habitual food intake was not collected. It is likely that the macronutrient composition of habitual diet differed from the diet provided during the habitual and restricted sleep period. All food was prepared by the PBRC Metabolic Kitchen and matched across arms. Women were instructed to consume meals at times that mimicked their normal eating habits (as documented during 7-day run-in period). No other food or beverage (except for water and non-caffeinated, noncaloric drinks) were allowed. Participants were also instructed to abstain from any moderate-to-vigorous physical activity during both the sleep conditions. Physical activity was monitored during both run-in and sleep period by accelerometry. Following each sleep condition, inpatient testing included assessment of: (1) insulin sensitivity by 2-step hyperinsulinemic-euglycemic clamp; (2) resting metabolism and substrate oxidation by indirect calorimetry; and (3) serum glucose, insulin, and C-peptide response to a standard meal test. A graphical illustration of the study protocol is provided in Figure 1.

Figure 1. Study Protocol.

Prior to each intervention arm, women wore an accelerometer for 7 days to measure habitual sleep habits for determining prescribed sleep bedtime, waketime, and ultimately total in-bed duration. Each intervention arm consisted of 4 nights of habitual sleep (100% of normal sleep duration) or 4 nights of sleep restriction (40% reduction in habitual sleep, including 20% delayed sleep onset and 20% early waking) and then to crossover to the other arm after a minimum 3-week washout period. On day 4, women consumed a standard meal test (Ensure^® Original Nutritional Shakes, 440-kcal) while serum blood was collected (indicated by ▲) over a 2-hour period from approximately 6:00 to 8:00 PM. On day 5, women awoke at their prescribed waketime for inpatient assessment of insulin sensitivity measured by a 2-step hyperinsulinemic-euglycemic clamp, as well as REE and RER measured by indirect calorimetry (IC) during both low-dose and high-dose insulin infusion. Serum blood was collected during each steady-state period (indicated by ▲). Urine was also collected throughout the duration of the clamp for calculation of urinary nitrogen excretion. * Denotes the 40% reduction in habitual sleep included 20% delayed sleep onset and 20% early waking.

2.3. Screening Visit

Eligibility criteria was verified and baseline study measures were obtained during this visit. Metabolic weight, height, vitals (blood pressure, resting heart rate), and body composition were measured under fasted conditions. A fasting blood sample was collected for measurement of glucose, insulin, free fatty acids (FFA), and follicle-stimulating hormone (FSH). A screening health questionnaire to assess general health (e.g., medical conditions, diet and exercise habits, and reproductive health including self-reported absence of menstrual cycles for 1-year) was collected. Sleep-related questionnaires including the Epworth Sleepiness Scale (ESS) (18), Insomnia Severity Index (ISI) (19), Pittsburgh Sleep Quality Index (PSQI) (20), and Berlin Questionnaire (21) were used to confirm absence of major sleep disturbances and sleep apnea. Women with both an ESS >10 (indicating excessive daytime sleepiness, sleep apnea, or narcolepsy) and ISI >14 (indicating moderate or severe insomnia) were excluded.

2.4. Sleep Prescription and Compliance

SR was achieved by reducing nocturnal time-in bed and not allowing any daytime naps. Personalized sleep prescription for habitual and restricted sleep periods were provided based on regular sleep habits. Women completed a 7-day run-in period prior to each sleep condition. During this time, women were asked to maintain their usual sleep and meal schedules. Detailed activity log, sleep diary (including waketimes, showering, mealtimes, bedtimes, etc.), and wrist accelerometry (ActiGraph™ wGT3X-BT) were used to determine eligibility, sleep prescription, and verify compliance. The first run-in period was used to objectively confirm that women achieved ≥6.5 hours of habitual nightly sleep, as well as to determine individual sleep prescriptions. Women were randomized after the first run-in period. The second run-in period was used to ensure similar sleep patterns were followed upon entry into the second sleep condition.

Individual habitual sleep schedules were derived using average bedtime (“in-bed time”) and waketime (“out-of-bed time”) to determine total time in bed. SR schedules were determined by reducing the total time in bed from habitual sleep by 40% while maintaining sleep midpoint (i.e., 20% delayed sleep onset and 20% early waking). Our trial was designed based on common sleep disturbances experienced by menopausal women (i.e., increased sleep latency and early morning awakening) (4). To ensure compliance during SR, women slept all 4 nights at the PBRC inpatient unit where bedtime and waketime were closely monitored. During habitual sleep, women were given detailed bedtime and waketime instructions and allowed to sleep the first 3 nights at home while the fourth (final) night was spent at the PBRC inpatient unit. Room lights were kept off during their prescribed sleep time. Daily compliance to each sleep schedule and absence of daytime naps was monitored by accelerometry and a questionnaire related to time in bed and sleep disturbances was administered following each night of sleep. Participants were asked to refrain from moderate-to-vigorous physical activity during both sleep periods, but no specific instructions were provided for pre- and post-sleep activities. If desired, participants were allowed walks of low intensity (light activity).

2.5. Study Procedures

2.5.1. Weight and body composition

Daily body weight was measured (Scale Tronix 5200, Welch Allyn, Inc.; Skaneateles Falls, NY) in the fasted state while wearing a surgical gown, which was subtracted from total weight. Fat mass (FM), fat-free mass (FFM), percent body fat, visceral adipose tissue (VAT) mass, and bone mineral density (BMD) were also measured in the fasted state using dual X-ray absorptiometry whole-body scanner (Lunar iDXA; General Electric, Milwaukee, WI, USA) at screening. iDXA scans were analyzed with the enCORE software version 13.60.033. DXA scan obtained at screening was not repeated during the study.

2.5.2. Standard meal test

Prior to testing on day 4 (evening), participants were instructed to have lunch before 12:00 pm and refrain from eating or drinking anything (except water) for at least 6 hours. On arrival at the PBRC inpatient unit, an intravenous catheter was inserted into an antecubital vein to obtain venous blood samples around 6 PM. After a brief rest period, a 440-kcal test meal (two 8 oz. Ensure^® Original Nutritional Shakes) consisting of 58.7% of energy from carbohydrate (64–66 grams total), 24.8% of energy from fat (12 grams total), and 16.5% of energy from protein (18 grams total) was consumed within 5 minutes under supervision. Serum blood was drawn for measurement of glucose, insulin, and C-peptide at time points −5 (baseline), 15, 30, 60, and 120 minutes. Area-under-the-curve (AUC) was calculated from baseline concentrations.

2.5.3. Two-step hyperinsulinemic-euglycemic clamp

On the morning of day 5, insulin sensitivity was measured by a 2-step hyperinsulinemic-euglycemic clamp. In the fasted state (pre-insulin, basal condition), resting energy expenditure (REE) and respiratory exchange ratio (RER) were measured for 30 minutes by indirect calorimeter (DeltaTrac; Sensor Medics, Yorba, CA, USA). Serum blood was also collected for measurement of basal glucose, insulin, FFA, and C-peptide concentrations. Next, insulin was infused for 180 minutes at 10 mIU/min/m² (low-insulin dose) followed by 120 minutes at 80 mIU/min/m² (high-insulin dose) with variable glucose (20% dextrose solution) infused at a rate to maintain plasma glucose concentrations at 90 mg/dL. Insulin sensitivity was defined as the glucose infusion rate (GIR) during the final 30-min steady-state period at both low- and high-insulin infusion (22) and expressed per kg of body weight. Similarly, REE and RER was repeated during these steady-state periods, as well as serum blood collected for measurement of glucose, insulin, FFA, and C-peptide concentrations. Urine was collected throughout the clamp for measurement of urinary nitrogen excretion to calculate protein oxidation (23). Basal carbohydrate and fat oxidation rates, as well as non-protein RER, were calculated using previously derived equations (24). At both low- and high-insulin doses, the oxidative component of the GIR was also calculated (24), while the non-oxidative component was calculated as the difference between the total GIR and its oxidative component (25). The metabolic clearance rate of insulin (MCR_I) was calculated as the insulin infusion rate (mIU/min/m²) divided by the increase in plasma insulin concentration above the basal concentration (μU/mL), as previously described (22, 26).

2.5.4. Serum and urine chemistry

Serum blood was collected for measurement of glucose (DXC 600 Pro; Beckman Coulter), insulin (Immulite 2000 XPi; Siemens), FFA (DXC 600 Pro; Beckman Coulter; reagent kit from Wako), C-peptide (Immulite 2000 XPi; Siemens), FSH (Immulite 2000 XPi; Siemens), and lipids (DXC 600 Pro; Beckman Coulter). Urine collected during clamp testing (day 5) was assayed for nitrogen by pyrochemiluminescence on an Antek 9000 Series Nitrogen & Sulfur Analyzer (Antek Instruments, Inc.; Houston, TX).

2.6. Quantification and Statistical Analysis

2.6.1. Statistical power

Our goal was to collect preliminary data to determine the treatment effect and variability estimates to inform the design of a larger clinical trial. To date, no clinical studies in postmenopausal women have been conducted that utilize the hyperinsulinemic-euglycemic clamp to assess insulin sensitivity (our primary endpoint). Although a formal power analysis was not feasible due to the lack of prior data, we planned to enroll up to 14 women to achieve a goal of 10 completers in 1 year with an anticipated dropout rate of 20%. The dropout rate was conservatively estimated based on our past pilot studies with similar length, procedures, and participation burden when hyperinsulinemic-euglycemic clamps were utilized. We expected detectable differences in insulin sensitivity between conditions given the high precision of the hyperinsulinemic-euglycemic clamp technique and the within-subject design.

2.6.2. Randomization

The randomization code was generated using block randomization in SAS version 9.4 (SAS Institute, Inc.). Since we continued to enroll until the planned 10 women completed the intervention, randomization was performed in a 1:1 allocation and using block sizes of 4 and 2, ensuring an equal number of women completed each of the two possible sequences. Allocations were concealed from women until after they enrolled in the trial. Furthermore, the principal investigator remained blinded throughout the intervention.

2.6.3. Data collection and Statistical analyses

Study data were collected and managed using REDCap electronic data capture tools hosted at Pennington Biomedical Research Center.(27, 28) All analyses are performed using SAS version 9.4 with a significance level of α=0.05. Of the 14 women randomized, 9 women were included in the final analysis; all 9 women had data from both the habitual sleep and restricted sleep conditions. Of the 5 women that were not included in the final analysis, one woman was deemed non-compliant to her prescribed sleep condition and another woman had a 2.4 kg weight difference between sleep conditions. Another 3 women completed their first sleep condition (all habitual) but did not complete their respective SR condition or undergo the study procedures (Figure 2). All baseline characteristics, sleep prescription, actual sleep duration, and other sleep characteristics are expressed as raw mean ± SD. A linear mixed effect model is used to assess metabolic differences in the response to habitual and SR conditions. Model covariates included treatment and crossover order effects (period and sequence). The model used an unstructured covariance matrix to allow correlation within subjects to vary. Means and differences in treatment outcomes are defined by least squares means ± SEM from the linear mixed model with p-values based on model t-tests.

Figure 2. Participant flow diagram.

Individuals were initially assessed via an online questionnaire followed by an in person visit. If an individual failed to meet a criterion, that person was excluded, and no further criteria were measured. Furthermore, individuals who did not have confirmed ≥6.5 hours of habitual nightly sleep by accelerometry were excluded from the trial prior to randomization.

3. RESULTS

3.1. Participants

These 9 postmenopausal women (7 White, 1 Black, 1 Asian) had a mean age of 59±4 years (range 54 to 65 years), BMI of 28.0±2.6 kg/m², and a fasting glucose of 97±9 mg/dL (Table 1). Women were an average of 7.2±3.1 years past their last menstrual period and had postmenopausal FSH concentrations of 77±21 IU/L (range of 52 to 111 IU/L). On average, women were prescribed a habitual time in bed (mean±SD) of 8.4±0.6 hours and a 40% restricted time in bed of 5.0±0.4 hours per night. Sleep and physical activity were comparable during the 7-day run-in period prior to habitual and SR condition (Table 2). During the intervention, compared to habitual sleep, women had an average (mean±SD) reduction in total time in bed of 38% per night with SR (8.4±0.6 vs. 5.0±0.4 hours; p<0.0001). The decreased time in bed was associated with reduced sleep duration and increased sleep quality (Table 3). Body weight was similar between habitual and SR conditions, respectively (74.3±3.2 vs. 74.3±3.2 kg; p=0.94) and did not change throughout each sleep condition (p=0.61). Physical activity did not change during the two sleep conditions (Table 3)

Table 1.

Participant characteristics at screening (n=9)

Screening characteristics are presented as mean ± SD, unless otherwise noted.

Characteristics	Baseline Value
Anthropometry
Age (years)	59 ± 4
Race (White/Black/Asian, n)	7 / 1 / 1
Height (cm)	162.3 ± 8.2
Weight (kg)	74.2 ± 11.2
BMI (kg/m2)	28.0 ± 2.6
Systolic BP (mmHg)	116 ± 19
Diastolic BP (mmHg)	73 ± 9
Heart rate (bpm)	63 ± 4
Time since last menstrual period (years)	7.2 ± 3.1
Blood chemistry
Glucose (mg/dL)	97 ± 9
Insulin (mU/L)	9.8 ± 3.8
HOMA-IR	2.37 ± 1.05
FFA (mmol/L)	0.614 ± 0.171
Follicle-stimulating hormone (IU/L)	77 ± 21
Total cholesterol (mg/dL)	256 ± 63
LDL cholesterol (mg/dL)	159 ± 44
HDL cholesterol (mg/dL)	71 ± 17
Triglycerides (mg/dL)	130 ± 67
Body composition
Percent body fat (%)	43.8 ± 4.6
Fat mass (kg)	32.7 ± 7.3
Fat-free mass (kg)	41.4 ± 5.6
VAT mass (kg)	1.19 ± 0.67
Bone mineral density (g/cm2)	1.07 ± 0.12

Abbreviations: BMI, body mass index; BP, blood pressure; HOMA-IR, homeostatic model assessment for insulin resistance; FFA, free fatty acids; LDL, low-density lipoprotein; HDL, high-density lipoprotein; VAT, visceral adipose tissue.

Table 2: Participant Sleep and physical activity levels leading to study periods.

Washout effects for sleep restriction (reported as least squares mean ± SEM) and associated p-values were derived from linear mixed modelling. Data are from 7-day wrist accelerometry. All data are presented as 9 paired sleep conditions.

Endpoint	Run-In		Washout Effect
	Habitual	Restricted	∆ ± SEM	p
Sleep
Time in bed (min)	497 ± 11	498 ± 9	−0 ± 11	0.98
In-bed time	10:08 PM ± 17 min	10:12 PM ± 18 min	5 ± 7 min	0.43
Out-bed time	06:24 AM ± 22 min	06:29 AM ± 23 min	5 ± 12 min	0.32
Total sleep time (min)	444 ± 13	455 ± 14	11 ± 13	0.43
Sleep latency (min)	8 ± 3	5 ± 4	−3 ± 5	0.55
Sleep efficiency (%)	89 ± 2	91 ± 2	2 ± 2	0.3
Wake after sleep onset (min)	45 ± 7	38 ± 8	−7 ± 4	0.12
Sleep fragmentation index (%)	24.1 ± 2.8	18.8 ± 2.9	−5.3 ± 2.4	0.06
Physical activity
Total (min/day)	207 ± 28	208 ± 27	1 ± 8	0.95
Sedentary (min/day)	653 ± 21	670 ± 21	17 ± 18	0.37
Light (min/day)	403 ± 17	384 ± 17	−19 ±14	0.22
Moderate (min/day)	194 ± 23	196 ± 23	2 ± 8	0.84
Vigorous (min)	12 ± 7	11 ± 7	−1 ± 1	0.35
Steps (count/day)	8535 ± 825	9192 ± 836	657 ± 727	0.39

Table 3: Participant sleep and physical activity during study periods.

Treatment effects for sleep restriction (reported as least squares mean ± SEM) and associated p-values were derived from linear mixed modelling. Data are from 4-day wrist accelerometry. All data are presented as 9 paired sleep conditions. Endpoints significantly affected by sleep restriction (p≤0.05) are highlighted in grey.

Endpoint	Sleep condition		Treatment Effect
	Habitual	Restricted	∆ ± SEM	p
Sleep
Time in bed (min)	506 ± 11	305 ± 12	−201 ± 8	<0.001
In-bed time	09:43 PM ± 18 min	11:46 PM ± 17 min	123 ± 9 min	<0.001
Out-bed time	06:09 AM ± 21 min	04:51 AM ± 21 min	99 ± 9 min	<0.001
Total sleep time (min)	458 ± 10	284 ± 11	−174 ± 9	<0.001
Sleep latency (min)	5 ± 1	3 ± 1	−1 ± 1	0.27
Sleep efficiency (%)	91 ± 2	93 ± 1	3 ± 1	0.03
Wake after sleep onset (min)	43 ± 7	18 ± 6	−25 ± 6	0.003
Sleep fragmentation index (%)	21.9 ± 3.1	15.1 ± 2.8	−6.8 ± 1.9	0.01
Physical activity
Total (min/day)	123 ± 17	113 ± 17	−10 ± 10	0.36
Sedentary (min/day)	460 ± 25	491 ± 20	31 ± 19	0.15
Light (min/day)	258 ± 20	269 ± 15	12 ± 20	0.58
Moderate (min/day)	115 ± 14	105 ± 14	−10 ± 10	0.36
Vigorous (min)	8 ± 4	8 ± 5	−0 ± 1	0.77
Steps (count/day)	5372 ± 502	5244 ± 512	−128 ± 366	0.74

3.2. Insulin Sensitivity, Substrate Oxidation, and Energy Expenditure

Compared to habitual sleep, SR increased basal glucose (93±3 vs. 96±3 mg/dL; p=0.05) and reduced fasting insulin (8.3±1.1 vs. 6.5±1.0; p=0.04), respectively. No differences in fasting FFA or C-peptide were observed between habitual and SR conditions. Additionally, no differences in basal REE and RER, as well as basal carbohydrate, fat, and protein oxidation were observed (Table 4).

Table 4. Participant differences in metabolic characteristics.

Treatment effects for sleep restriction (reported as least squares mean ± SEM) and associated p-values were derived from linear mixed modelling. Endpoints significantly affected by sleep restriction (p≤0.05) are highlighted in grey. All data are presented as 9 paired sleep conditions.

Endpoint	Sleep Condition		Treatment Effect
	Habitual	Restricted	∆ ± SEM	p
Insulin sensitivity
Basal
Glucose (mg/dL)	93 ± 3	96 ± 3	3 ± 1	0.05
Insulin (μU/mL)	8.3 ± 1.1	6.5 ± 1.0	−1.8 ± 0.7	0.04
FFA (mmol/L)	0.539 ± 0.047	0.555 ± 0.047	0.017 ± 0.018	0.38
HOMA-IR	2.06 ± 0.38	1.97 ± 0.34	−0.10 ± 0.29	0.75
C-peptide (ng/mL)	2.1 ± 0.2	2.2 ± 0.2	0.1 ± 0.1	0.24
Low-Dose Insulin Infusion
Glucose (mg/dL)	88 ± 1	87 ± 1	−1 ± 1	0.48
Insulin (μU/mL)	24.5 ± 2.7	20.4 ± 2.8	−4.1 ± 1.2	0.01
FFA (mmol/L)	0.078 ± 0.020	0.078 ± 0.020	0.000 ± 0.010	0.99
GIR (mg/kg/min)	2.55 ± 0.22	2.03 ± 0.20	−0.52 ± 0.16	0.01
Oxidation GIR (mg/kg/min)	1.62 ± 0.09	1.53 ± 0.09	−0.09 ± 0.11	0.43
Non-oxidative GIR (mg/kg/min)	1.18 ± 0.20	0.72 ± 0.17	−0.46 ± 0.14	0.01
C-peptide (ng/mL)	1.1 ± 0.0	1.0 ± 0.0	−0.1 ± 0.1	0.15
MCRI (mL/min/m2)	1092 ± 269	1039 ± 213	−53 ± 188	0.79
High-Dose Insulin Infusion
Glucose (mg/dL)	95 ± 3	91 ± 3	−4 ± 4	0.40
Insulin (μU/mL)	132.1 ± 9.0	114.4 ± 9.6	−17.8 ± 6.2	0.02
FFA (mmol/L)	0.027 ± 0.005	0.019 ± 0.004	−0.007 ± 0.007	0.30
GIR (mg/kg/min)	10.48 ± 0.72	9.19 ± 0.72	−1.29 ± 0.23	<0.001
Oxidation GIR (mg/kg/min)	2.16 ± 0.13	2.13 ± 0.13	−0.04 ± 0.06	0.54
Non-oxidative GIR (mg/kg/min)	8.38 ± 0.70	7.07 ± 0.71	−1.31 ± 0.25	0.001
C-peptide (ng/mL)	1.1 ± 0.1	0.8 ± 0.1	−0.2 ± 0.2	0.18
MCRI (mL/min/m2)	762 ± 92	792 ± 76	30 ± 62	0.65
Standard meal test
Glucose AUC (mg/dL x hr)	102 ± 11	89 ± 11	−12 ± 10	0.25
Insulin AUC (µU/mL x hr)	111 ± 18	94 ± 17	−18 ± 8	0.07
C-peptide AUC (ng/mL x hr)	813 ± 102	770 ± 102	−43 ± 58	0.48
Energy expenditure and substrate oxidation
Basal
REE (kcal/d)	1371 ± 47	1310 ± 50	−62 ± 40	0.14
RER	0.79 ± 0.01	0.79 ± 0.01	0.00 ± 0.01	0.76
npRER	0.79 ± 0.01	0.80 ± 0.01	0.01 ± 0.01	0.47
Carbohydrate oxidation (g/d)	83.1 ± 9.6	88.1 ± 8.6	5.1 ± 10.7	0.65
Fat oxidation (g/d)	81.9 ± 5.7	77.3 ± 4.8	−4.6 ± 6.7	0.51
Protein oxidation (g/d)	58.3 ± 5.4	52.7 ± 6.3	−5.6 ± 5.0	0.30
Low-Dose Insulin Infusion
REE (kcal/d)	1313 ± 42	1267 ± 46	−46 ± 37	0.24
RER	0.865 ± 0.006	0.863 ± 0.006	−0.002 ± 0.006	0.81
High-Dose Insulin Infusion
REE (kcal/d)	1452 ± 50	1396 ± 54	−56 ± 42	0.20
RER	0.91 ± 0.01	0.94 ± 0.01	0.03 ± 0.01	<0.01

Abbreviations: FFA, free fatty acids; HOMA-IR, homeostatic model assessment for insulin resistance; GIR, glucose infusion rate; MCR_I, metabolic clearance rate of insulin; AUC, area-under-the-curve; REE, resting energy expenditure; RER, respiratory exchange ratio; npRER, non-protein respiratory exchange ratio.

At both low- and high-dose insulin infusions, insulin sensitivity was reduced with SR. As shown in Figure 3, SR reduced both low-dose insulin GIR by 20% (2.55±0.22 vs. 2.03±0.20 mg/kg/min; p=0.01) and high-dose insulin GIR by 12% (10.48±0.72 vs. 9.19±0.72 mg/kg/min; p<0.001). These reductions in GIR were attributed to reductions in non-oxidative GIR, which was reduced by 39% at the low-dose (1.18±0.20 vs. 0.72±0.17 mg/kg/min; p=0.01) and 16% at the high-dose (8.38±0.70 vs. 7.07±0.71 mg/kg/min; p=0.001) insulin infusions. No difference in oxidative GIR was observed. Despite similar insulin infusion during both sleep conditions, insulin concentrations were significantly lower with SR at both the low-dose and high-dose clamp periods. Lower insulin concentrations did not appear to be caused by differences in insulin clearance (MCR_I) or insulin secretion (as assessed by C-peptide) (Table 4).

Figure 3. Insulin Sensitivity.

Sleep restriction dramatically lowered both low-dose (left) and high-dose (right) insulin GIR by 20% and 12%, respectively. These alterations were the result of reduced non-oxidative component of GIR by 39% during low-dose and 16% during high-dose clamp segments. Oxidative GIR was unaffected. All data are presented as 9 paired sleep conditions. Data are presented as least squares mean ± SEM. (**p≤0.01 and ***p≤0.001).

Compared to habitual sleep, SR increased RER during high-dose insulin infusion (0.91±0.01 vs. 0.94±0.01; p<0.01) reflecting reduced fat oxidation. No differences in REE were observed.

3.3. Glycemic Control

As shown in Figure 4, SR did not affect glucose concentrations at any time point or overall glucose AUC following the standard meal test. Compared to habitual sleep, there was a trend towards a reduction in overall insulin AUC with SR (111±18 vs. 94±17 uU/mL x hr; p=0.07) despite no effect on individual insulin concentrations at any time points (Table 4). This trend toward a reduction in overall insulin AUC appeared to be driven by a single time point (120 minute) and, therefore, should be interpreted with caution. Furthermore, this trend towards a reduction in overall insulin AUC was not explained by differences in insulin secretion, as C-peptide AUC was not different between sleep conditions following the standard meal (p=0.48).

Figure 4. Glycemic Control.

In response to a standard meal test, (A) sleep restriction did not affect individual glucose concentrations (main) or overall glucose AUC (inset); (B) sleep restriction did not affect individual insulin concentrations (main), but overall insulin AUC trended towards being reduced by 15% (p=0.07) (inset); and (C) sleep restriction did not affect individual C-peptide concentrations (main) or overall C-peptide AUC (inset). All data are presented as 9 paired sleep conditions. Data are presented as least squares mean ± SEM. (^†0.05<p≤0.10).

4. DISCUSSION

This is the first study in postmenopausal women to assess the impact of SR on both insulin sensitivity and energy metabolism using gold-standard methods. We found that short-term SR (4 nights) significantly reduced whole-body insulin sensitivity and fat oxidation during insulin infusion. These findings partially confirm our hypotheses and emphasize the deleterious effect of sleep disturbances―a highly prevalent menopausal symptom―on cardiometabolic risk in post-menopausal women. To date, the role of sleep disturbances on whole-body metabolism in post-menopausal women remains largely unexplored. Because sleep disturbances are known to impair glucose tolerance and reduce insulin sensitivity in men and premenopausal women (5, 6, 7, 8, 9, 10, 11, 12), perimenopausal and postmenopausal women with highly prevalent sleep disturbances may be disproportionately impacted.

Findings from cross-sectional studies linking menopause with the development of insulin resistance are mixed. One of the earliest studies to investigate the impact of menopause on insulin sensitivity was conducted exclusively in White women without obesity (29). Compared to premenopausal women, this study observed that postmenopausal women were paradoxically more insulin sensitive yet had 35% lower C-peptide response and 51% lower pancreatic insulin secretion during a frequently sampled intravenous glucose tolerance test (FSIVGTT) (29). Although the physiological effects of ovarian hormones on insulin metabolism are unclear, a role for estrogens (particularly estradiol) in maintaining glucose homeostasis through effects on insulin secretion and clearance has been suggested (30). Other cross-sectional studies have reported that postmenopausal women had a 12% lower glucose disposal rate by hyperinsulinemic-euglycemic clamp compared to premenopausal women (31), and that increased abdominal fat is associated with decreased tissue insulin sensitivity and glucose tolerance (32). Again, none of these cross-sectional studies in menopausal women measured sleep characteristics.

Longitudinal studies that have examined intraindividual changes in insulin sensitivity across the menopause transition are scarce but do not support that menopause progression causes insulin resistance (33, 34, 35). Although these longitudinal studies provide a wealth of data about metabolic health in women as they age, only surrogate measures of insulin sensitivity (e.g., HOMA-IR, quantitative insulin sensitivity check index [QUICKI], etc.) are reported and detailed measurement of sleep characteristics are again lacking.

Our findings that 4 nights of SR reduced fat oxidation only during insulin infusion, and not under resting conditions, was contrary to our hypothesis. To our knowledge, only one experimental study has examined the effect of sleep disturbances during menopause on energy expenditure and substrate oxidation (15). In one arm of this study, short-term (3 nights) sleep fragmentation was administered to 20 premenopausal women and resulted in decreased fat oxidation (by indirect calorimetry). In the second arm of this study, a smaller subset of 9 premenopausal women underwent monthly injection of the GnRH agonist leuprolide acetate to induce menopause (hypo-estrogenized state), followed next by short-term sleep fragmentation to study its effect on energy metabolism. They observed that a hypo-estrogenized (postmenopausal) state reduced fat oxidation, but the addition of sleep fragmentation did not further alter fat oxidation (15). Although our findings suggest that SR (rather than sleep fragmentation) reduces fat oxidation in postmenopausal women, this may be due in part to the fact that women in our study had previously undergone a natural menopause transition years prior rather than a more recent experimental menopause transition (15).

Other studies among men and premenopausal women exposed to SR have instead observed increases in fat oxidation (via reduced RER) rather than a decrease (36, 37). In one study, healthy men and premenopausal women who underwent 5 nights of normal sleep (8 hours of time in bed: 11:00 PM–7:00 AM) versus SR (4 hours of time in bed: 1:00–5:00 AM) had reduced basal RER (increased fat oxidation) with SR (36). Another study conducted in premenopausal women using a similar sleep paradigm also found that 3 nights of SR (4 hours/night) reduced basal RER, but not postprandial RER, compared to normal sleep (8 hours/night) (37). Indeed, differences in study populations (e.g., young vs. old, men vs. women, or pre-menopause vs. post-menopause), small sample sizes, lack of diverse populations (in each study), and technical limitations associated with assessment of metabolism using different techniques may explain these differences in fat metabolism observed with SR. Our findings that SR does not alter basal and/or postprandial energy expenditure are, however, consistent with previous studies using indirect calorimetry (7, 15, 36, 37, 38, 39, 40).

Interestingly, we observed paradoxical changes in insulin concentrations that were unexpected. First, we detected a trend towards reduced overall insulin AUC with SR following a standard meal test that was driven by a single time point (120 minute). We also observed that SR reduced insulin concentrations under basal, as well as during low-dose and high-dose insulin infusion (despite similar units of clamp insulin infused during both sleep conditions). Our observations are in contrast to our hypothesis that a compensatory increase in insulin concentration would occur in response to insulin resistance to maintain glucose homeostasis (41). Among adults with normal glucose tolerance, plasma insulin response is the aggregate of both insulin secretion by the pancreatic β-cells and the MCR_I working together to offset insulin resistance. Unfortunately, we did not observe any differences in insulin clearance or secretion. To the best of our knowledge, only one study has examined the relationship between insulin sensitivity, secretion, and elimination during menopause―reporting that postmenopausal women had 35% lower C-peptide response and 51% lower pancreatic insulin secretion compared to premenopausal women (29). Although we are unable to explain the reason for these lower insulin concentrations under basal and insulin-stimulated conditions at this time, a larger sample size would likely provide more confidence in our estimates.

The strengths of our study include: (1) the use of the gold standard method of assessing insulin sensitivity via hyperinsulinemic-euglycemic clamp (primary endpoint); (2) tightly controlled feeding of eucaloric meals to maintain energy balance throughout each sleep condition; (3) inpatient residence during nights where SR was administered; and (4) blinding of the study PI (KLM) who conducted all the clamps to the randomization assignments.

This study has several limitations. First, our final analysis included only 9 postmenopausal women and the randomization order was not balanced. Although our sample size is like other well-controlled or inpatient sleep trials, our results need to be replicated in a larger trial that includes more postmenopausal women and expands enrollment to women across the menopause transition (i.e., pre-, peri-, and post-menopause). This will allow clear determination of the effects of menopause during SR. Second, although acute inpatient studies of short-term, inadequate sleep have the advantage of controlling key confounding variables, results from such studies may not directly translate (and may even be underestimated) when compared to real-world metabolic adaptations that women with chronic sleep disturbances likely experience. Third, our standard meal test was conducted in the evening (as a dinner) rather than in the morning following an overnight fast. Although we were unable to accommodate an extra day of testing, we did standardize the time of day that lunch was eaten within each participant to minimize any variability in fasting duration. However, testing and meals occurred at different times relative to habitual sleep and wake, resulting in different durations of overnight fasting. Finally, our trial did not measure salivary cortisol or catecholamines. Indeed, one potential mechanism for reduced insulin secretion with SR may be due to increased sympathetic activity at the level of the β-cells.

5. CONCLUSION

Four nights of SR reduced insulin sensitivity and fat oxidation during insulin infusion in postmenopausal women. These findings underscore how sleep disturbances contribute to metabolic dysfunction in post-menopausal women. Larger trials investigating the degree to which sleep disturbances contribute to the metabolic dysfunction that occurs with menopause are needed across all stages of menopause (pre, peri, post).

Study Importance.

What is already known?

• Sleep disturbances are associated with adverse cardiometabolic risk.

What does this manuscript add?

• Our study is the first to investigate how sleep restriction alters metabolism in post-menopausal women.

How might these results change the direction of research?

• Shortened sleep duration contributes to some of the metabolic dysfunction observed in post-menopausal women.

• Larger trials are needed to assess how sleep disturbances cause metabolic dysfunction across all stages of menopause (pre, peri, post).

ABBREVIATIONS

AUC

Area-under-the-curve

BMI

Body mass index

BP

Blood pressure

FFA

Free fatty acids

GIR

Glucose infusion rate

HDL

High-density lipoprotein

HOMA-IR

Homeostatic model assessment for insulin resistance

LDL

Low-density lipoprotein

MCR_I

Metabolic clearance rate of insulin

npRER

Non-protein respiratory exchange ratio

REE

Resting energy expenditure

RER

Respiratory exchange ratio

VAT

Visceral adipose tissue

DATA AVAILABILITY

The dataset pertaining to the current study is available from the corresponding author in accordance with appropriate data use agreements and IRB approvals for secondary analyses.