Abstract
Objective: Midlife women are at risk of obesity. Poor sleep outcomes including inadequate sleep duration and variable sleep timing are risk factors for obesity, but there is a lack of understanding on how specific sleep constructs uniquely or concurrently are associated with weight outcomes in this population. This study examined the unique association of sleep timing with weight outcomes and how sleep timing works in conjunction with sleep duration to influence weight in midlife women.
Materials and Methods: An archival analysis was performed using the Midlife in the United States-II study (MIDUS-II). The sample consisted of 132 midlife women (40–64; M = 52.9, standard deviation = 6.94). Sleep timing (mean sleep time, variability) and duration were measured through actigraphy and daily sleep diaries. Weight was assessed using body mass index (BMI) and waist circumference measures.
Results: PROCESS mediation and moderation analyses assessed direct, indirect, and moderating pathways. Sleep duration emerged as an indirect link between sleep timing (mean and variability) and weight outcomes (95% CI = 0.0001–0.0123; 95% CI = 0.0007–0.0378; 95% CI = 0.0079–0.1006). Sleep timing (95% CI = −0.0144 to 0.0076; 95% CI = −0.0358 to 0.0219) and sleep time variability (95% CI = −0.0124 to 0.0438; 95% CI = −0.0533 to 0.0939) were not directly associated with BMI or waist circumference. Also, sleep timing and duration did not interact to influence weight outcomes.
Conclusions: Sleep duration, rather than sleep timing, is associated with weight outcomes, and is an indirect link in the sleep timing and weight outcomes association. Future work is needed to further disentangle the impact of sleep on weight in midlife women using prospective studies, implementing daily assessments of sleep behavior.
Keywords: sleep timing, sleep duration, midlife women, obesity, weight
Introduction
Obesity is a major public health issue in the United States.1 Women in midlife (i.e., 40–64 years old) are at particular risk with 42% categorized as obese indicated by having a body mass index (BMI) ≥30.1 In addition to having high rates of obesity measured through BMI, women in midlife have high rates of central adiposity (i.e., >88 cm).2 Both obesity and abdominal obesity are concerning given associated adverse physical health outcomes, including early mortality and serious health conditions such as metabolic syndrome.2,3
Given the negative health implications of obesity, there is a need to identify factors that put women at an increased risk. Biologically, women in midlife go through the menopausal transition, which is associated with significant weight gain and changes in body fat distribution.1 In particular, menopause is characterized by 12 months of amenorrhea (i.e., lack of menstrual cycle) following the final menstrual cycle.4 Menopause is also regarded as a process that encompasses various stages, including the early and late menopause transition (i.e., perimenopause) as well as the early and late postmenopause period. The stages are variable, but include changes in menstrual cycles and endocrine functioning with elevations in follicle stimulating hormone during the menopause transition as well as postmenopause.4 Although menopause is linked to weight gain due to hormonal changes and aging, these alone do not account for weight gain in this time period.5 Consequently, behavioral factors have been proposed as predictors of obesity in midlife, including poor sleep health.6,7
Poor and disrupted sleep is common in midlife women due to the menopausal transition.8 Between 33% and 51% of women report sleep problems during this time,9 and sleep issues are recognized as a core symptom of menopause.10 In addition, aging is associated with changes in sleep duration and timing with increasing age associated with decreased total sleep time as well as shifts to earlier sleep timing.11 Poor sleep leads to adverse physiological and psychological outcomes in midlife, including poor mood, reduced quality of life, disability, as well as an increased risk of obesity and abdominal adiposity.12,13
Although poor and disrupted sleep can be defined in a variety of ways, the specific sleep constructs of duration and timing are of particular relevance when examining weight outcomes in this population. Sleep duration, the total amount of sleep obtained per nocturnal sleep period or across the 24-hour period (e.g., night and daytime sleep), and sleep timing, the placement of sleep within the 24-hour day (e.g., time of initiation of nocturnal sleep or the clock time of sleep onset),14 have consistently been linked to higher rates of obesity in midlife women.7,15 Specifically, short sleep duration (i.e., <6 hours/night) has been linked to metabolic syndrome and higher weight outcomes in this population.12,16 In addition, sleep timing, predicated on both daily behavioral and circadian rhythms, impacts social and biological rhythms. Biologically, sleep timing is a key component in the regulation of metabolism and energy homeostasis, while socially, dysregulation of metabolism can, in turn, impact eating behaviors and weight outcomes.17,18
Although the specific attributes of sleep duration and sleep timing are associated with higher weight outcomes in the general population,7,12,15,16 research in midlife women has focused more on the unique influence of sleep duration,6,19 rather than sleep timing. The study of sleep timing in midlife women is warranted as women in midlife have issues initiating sleep, which is particularly true during the menopausal transition due to psychosocial factors and changes in estrogen and progesterone, which influence stress, mood, body temperature, and regulation of sleep timing.8,20,21 Although current research suggests a relation between late bed timing, variability in daily bedtimes, and weight outcomes in this population, only a small body of literature has investigated these associations with inconsistent findings.22
There is also a lack of understanding of how sleep duration and timing are uniquely or concurrently associated with weight outcomes in this population. Sleep duration and sleep timing are often associated with similar mechanisms (i.e., hormonal dysregulation of appetite, reward processing activation),23 making it unclear whether they independently or sequentially influence weight. Although sleep timing may uniquely influence weight outcomes through circadian misalignment effects on metabolism and increased opportunity to engage in poor health behaviors due to increased wake time,24 contrasting research oftentimes fails to differentiate the influence of timing and/or duration on weight outcomes.25,26 As such, it is unclear if the sleep timing, or its influence on sleep duration, independently contributes to weight gain. Differentiating the unique role of sleep timing versus duration could have significant implications for women's health behaviors in midlife. More information is needed to inform tailored sleep interventions for midlife women, which could focus on increasing sleep duration, achieving an earlier bedtime, or implementing consistent bedtimes.
Therefore, this study aimed to define the connection between sleep timing, sleep duration, and weight outcomes in midlife women by comprehensively investigating sleep timing (i.e., mean sleep time, sleep time variability), sleep duration, and weight outcomes (BMI, waist circumference) using daily, objective, and self-report measurement of sleep behavior and objective measurement of weight. This study had two aims: first, we examined the unique role of sleep timing in its association with weight outcomes in midlife women. Second, we investigated how sleep timing works in conjunction with sleep duration to influence weight. In particular, this study implements mediation and moderation analyses to assess study aims. Mediation will establish the extent to which sleep timing is associated with weight outcomes directly versus through sleep duration, whereas moderation will help determine whether the size of the effect between sleep timing and weight is affected by sleep duration. We hypothesized that (1) later and more variable sleep timing would be directly associated with higher levels of weight, (2) sleep duration would serve as an indirect pathway between sleep timing and weight, and (3) sleep duration would strengthen the association between sleep timing and weight outcomes.
Materials and Methods
Design
A secondary data analysis was performed using cross-sectional data from the Midlife in the United States-II study (MIDUS-II) dataset. MIDUS-II is funded by the National Institute on Aging at the University of Wisconsin Madison (P01-AG020166). MIDUS data collection was reviewed and approved by the Education and Social/Behavioral Sciences and the Health Sciences International Review Board (IRB) at the University of Wisconsin-Madison.
Participants
Participants completed the MIDUS-II, a longitudinal telephone and paper-and-pencil follow-up study of the original MIDUS-I study (N = 7,108). MIDUS-I participants were recruited through a Random Digit Dial national sample of American adults aged 24–74 years. Data collection for MIDUS-II took place from 2004 to 2006. Participants of MIDUS-I were invited to participate in MIDUS-II data collection. All participants were consented on the study purpose and confidentiality precautions. Eligible participants were noninstitutionalized, English-speaking adults between 35 and 86 years of age.
MIDUS-II included ∼4,963 participants and consisted of 4 subprojects. This study consists of a subsample of 139 female MIDUS-II participants between the ages of 40–64 years who completed Project 1 and Project 4 of the MIDUS-II study. Midlife status was identified as being within the 40–64 year age range. Menopausal stage could be calculated for a subset of women using participant responses to a medical history questionnaire specific to women's health. Participants answered the following questions: “Have you have a menstrual period in the last year?” and “Do you know if your menstrual period stopped due to menopause?” However, menopause status was not included in this study due to an inadequate sample of participants (i.e., 75 of 132 midlife women) who completed all necessary measures. To provide an estimate of menopause status, women in the sample were characterized by age to approximate status. The age ranges of 40–48, 49–53, and 54–64 were utilized for descriptive purposes to capture pre-, peri-, and postmenopause status, whereas for specific analyses age was dichotomized as “younger” (40–52 years) and “older” (53–64 years) to represent premenopause and postmenopause status, respectively.
On average, participants were 52.9 years old (standard deviation [SD] = 6.94), primarily White, with an average annual income of $66,449 (SD = $53,563; Table 1). Women in the sample were generally overweight and obese, with high levels of abdominal obesity (BMI = 31.04 ± 6.96, waist circumference = 99.61 ± 17.64; Table 2). Levels of obesity were slightly higher in the current sample (i.e., 48.6%) compared with the national average of midlife women in 2014 (i.e., 42.1%).1 Waist circumference values were also slightly higher (i.e., 99.45) than the national average for women ≥18 years in the United States as of 2012 (i.e., 96.0).27 Additional information regarding participant weight and sleep characteristics is included in Table 2.
Table 1.
Participant Sociodemographic and Health Characteristics (M, SD)
| Variable | N | % |
|---|---|---|
| Age (52.90, 6.94) | ||
| 40–48 years | 41 | 28.9 |
| 49–53 years | 31 | 21.8 |
| 54–64 years | 70 | 49.3 |
| Race | ||
| White/Caucasian | 131 | 87.9 |
| Black/African American | 7 | 4.7 |
| Asian American | 2 | 1.3 |
| Native American/Alaska Native | 1 | 0.7 |
| Other/don't know | 8 | 5.4 |
| Income ($66,449, 53,563) | ||
| 0–$20,000 | 24 | 19.4 |
| $20,000–$39,999 | 20 | 16.1 |
| $40,000–$59,999 | 18 | 14.5 |
| $60,000–$79,999 | 19 | 15.3 |
| $80,000–$100,000 | 9 | 7.3 |
| >$100,000 | 34 | 27.4 |
| Education | ||
| Professional degree | 4 | 2.7 |
| Master's degree | 4 | 2.7 |
| Some graduate school | 4 | 2.7 |
| Bachelors | 30 | 20.1 |
| Two-year | 15 | 10.1 |
| 3+ college | 9 | 6.0 |
| 1–2 college | 29 | 19.5 |
| High school | 35 | 23.5 |
| GED | 7 | 0.7 |
| Did not graduate high school | 7 | 4.7 |
| Physical activity (244.79, 414.52) | ||
| <150 minutes | 70 | 53.4 |
| 150–299 minutes | 34 | 26.0 |
| ≥300 minutes | 27 | 20.6 |
| Self-rated health (2.37, 0.97) | ||
| Excellent (1) | 25 | 16.8 |
| Very good (2) | 67 | 45.0 |
| Good (3) | 38 | 25.5 |
| Fair (4) | 15 | 10.1 |
| Poor (5) | 4 | 2.7 |
| Depressive symptoms (14.68, 5.19, 4.00–31.20) | ||
Age was trichotomized to approximate age ranges for pre, peri, and postmenopause time periods. Self-rated health consists of scale ranging from 1 to 5 with items reverse scored, such that 1 indicates “excellent” and 5 indicates “poor.” Physical activity variable indicates sum of daily physical activity in minutes per week.
Table 2.
Participant Weight and Sleep Characteristics (M, SD, range)
| Variable | N | % |
|---|---|---|
| BMI (31.05, 6.89, 17.13–52.00) | ||
| <18.5 | 1 | 0.7 |
| 18.6–25 | 24 | 16.4 |
| 25.0–29.9 | 50 | 34.2 |
| ≥30 | 71 | 48.6 |
| Waist circumference (99.45, 17.52, 66.00–158.00) | ||
| <80.00 cm | 16 | 10.7 |
| 80.0–87.99 | 24 | 16.1 |
| ≥88 cm | 109 | 73.2 |
| Mean sleep time (11:36 pm, 84 minutes, 7:48 pm–3:54 am) | ||
| Sleep time variability (43, 51, 0–412 minutes) | ||
| Sleep duration (369, 63, 385 minutes) | ||
| Sleep efficiency (79%, 11.5%, 59.8%) | ||
| Sleep onset latency (28, 36, 278 minutes) | ||
| Sleep quality (2.5, 0.72, 3.4) | ||
A BMI level <18.5 indicates underweight status, a level ≥25 indicates overweight status, and a level ≥30 indicates obese status. A waist circumference ≥88 cm indicates central obesity. Sleep efficiency is the average percentage of total sleep time subtracted by the total wake time for each evening rest interval. Sleep quality is rated on a 1–5 scale, with “1” indicating “very good” and “5” indicating “very poor.”
BMI, body mass index.
Exclusionary criteria
To represent a normal range of sleeping behavior, this study controlled for possible effects of shiftwork and abnormal sleeping behavior due to irregular work schedules. We excluded participants if they indicated idiosyncratic sleep patterns due to work schedule, illness, or travel. Thus, we excluded participants if they indicated “yes” on the item: “Does participant show idiosyncratic sleep pattern due to work schedule, illness or travel.” This item was obtained from the MIDUS-II actigraphy dataset and was calculated to assess irregular sleeping patterns. Participants were also excluded if they scored >1 on the item: “How many nights in the past 12 months did your work require you to be away from home overnight?” This question was taken from a self-administered phone questionnaire from MIDUS-II Project 1. As a result of excluding irregular sleep patterns due to irregular work and sleep schedules, seven participants were excluded from the study. The final sample size included 132 participants.
Procedure
In MIDUS-II, Project 1, participants completed a phone interview and self-administered questionnaires measuring psychological constructs, demographic variables, and mental and physical health. Project 4 involved the completion of self-administered questionnaires, an in-person physical examination, and the completion of a daily sleep diary and wrist actigraphy protocol.
Measures
Sleep
Daily sleep behaviors were assessed using self-report daily sleep diaries and wrist actigraphy (Actiwatch-64®; Philips Respironics). The sleep diary had participants answer questions regarding bedtime (i.e., “What time did you go to bed and begin trying to go to sleep”), amount of time to fall asleep (i.e., “How long did it take you to get to sleep last night?”), wake time (i.e., “What time did you wake up for the day and not return to sleep?”), out of bedtime (i.e., What time did you get out of bed for the day?), and sleep quality ratings on a 1–5 scale with 1 indicating “very good” and 5 indicating “very poor” (i.e., overall quality of your sleep last night). Sleep diary entries were recorded every morning and evening across a 7-day time period. Participants also wore actigraphy watches for the same time period where they indicated when they tried to fall asleep (bedtime) and when they awoke in the morning (wake time) using an event marker button. Actigraphy watches were programmed for medium wake threshold selection using a 30-second sampling epoch. All participants wore the watches for a 7-day time period, including weekdays and weekends. One week of actigraphy is regarded as the ideal minimum to capture in vivo sleep and circadian rhythm patterns.28 Event marker and sleep diary information assisted in the automatic detection of sleep intervals by Actiwatch® algorithms.28 Both actigraphy and sleep diaries are primary forms of sleep assessment, and have been found to be both reliable and valid.29,30 In addition, sleep diaries and actigraphy provide a repeated assessment of sleep behavior, which can incorporate variability across days.30 All participants in this study completed all 7 days of sleep diaries and actigraphy.
Sleep timing
Sleep timing was assessed by calculating mean sleep time values and sleep time variability. Mean sleep time was calculated as the mean time of sleep start (clock time of sleep onset) across 7 days. Participants reported their bedtime using sleep diaries (e.g., lights out), and then the clock time of sleep onset was determined as the first actigraphy interval. Specifically, sleep start time was defined as the clock time after the first 10-minute period in which no more than one epoch (i.e., 30-second time period) was scored as mobile using Actiwatch software. Variability in sleep time was quantified by calculating intraindividual variability in sleep timing for each participant's sleep start time (clock time of sleep onset) across 7 days using intraindividual SDs. Intraindividual SDs describe the extent to which an individual's daily sleep timing varies around their mean score. In this study, a large intraindividual SD would indicate that an individual's sleep timing is highly flexible, whereas a small intraindividual SD would indicate more consistent sleep timing.31 All intraindividual SDs variables were detrended for time to control for any variations due to the effects of observing behaviors over time. Specifically, detrending for time produces a variable that precisely reflects variability over time that is not due to the effects of time, per se (e.g., practice effects, participation fatigue, etc.), but rather due to inherent variations in the behavior.31 Detrending was conducted through linear regression analyses for all participants with time (linear, quadratic, and cubic functions) as the independent variable and the sleep timing variables as the dependent outcomes. Intraindividual SD values were then calculated for the sleep timing variables using the time-independent residuals from the aforementioned linear regression analyses. This detrending process resulted in a sleep time variability variable of within-person SDs, which was independent of any influences of time. The intraindividual SD has been utilized to determine variability in sleep variables in additional research investigating sleep behaviors, and is a well-validated methodology to determine variability in behavior.22,31,32
Sleep duration
Sleep duration was calculated using actigraphy. Sleep duration was defined as the amount of time between sleep start (i.e., clock time of sleep onset as defined by actigraphy) and sleep end (i.e., clock time of final awakening as defined by actigraphy) excluding time spent awake after sleep onset.
Sleep covariates
Sleep covariates included sleep onset latency, sleep quality, and sleep efficiency. Sleep onset latency was determined by calculating the mean amount of minutes it took participants to fall asleep and consisted of the time from self-reported bedtime reported using the daily sleep diary to objectively assessed first sleep interval measured through actigraphy. This sleep onset interval was then averaged across 7 days to get an average sleep onset latency variable. Sleep quality was determined from the sleep diary, and was calculated by taking the mean score of how participants rated their overall sleep quality on a 1–5 (1 = very good, 5 = very poor) scale using sleep diaries. Finally, sleep efficiency was calculated using actigraphy. In particular, sleep efficiency was calculated by taking the percentage of total sleep time subtracted by the total wake time throughout the evening for each rest interval. An average sleep efficiency was calculated by averaging the values across the 7 days of data collection.
Weight
Weight was analyzed as two continuous variables in statistical models: (1) BMI defined as the calculation of body weight (in kilograms) by height (in meters squared)3 and (2) waist circumference defined as the measurement of the narrowest point between the ribs and iliac crest.2 In addition, data describing the prevalence of obesity and central obesity were calculated to describe the sample. Obesity and central obesity were measured during a physical examination using a standardized procedure. Obesity was defined as having a BMI ≥30,3 whereas central obesity was defined as having a waist circumference >88 cm.2
Additional covariates
In addition to the sleep covariates, age, race, educational level, annual income, self-rated health, depressive symptoms, and physical activity were included as covariates, given their known associations with sleep and weight in this population.16,20,33–35 Age, race, educational level, annual income, self-rated health, and depressive symptoms were obtained from self-administered questionnaires. Self-rated health was assessed by a self-report measure (1–5 scale) where participants were asked to rate their general physical health. Depressive symptoms were measured with the Center for Epidemiologic Studies Depression Scale.36 Physical activity was assessed by daily sleep diaries. A score of weekly physical activity was obtained by summing minutes of physical activity across the 7 days of data collection. Finally, for follow-up analyses, age was dichotomized as “younger” (40–52 years) and “older” (53–64 years) to approximate premenopause and postmenopause status.
Statistical analyses
Descriptive statistics and covariate evaluation
Pearson's correlation analyses identified significant bivariate associations between covariates and main variables of interest (i.e., mean sleep time, sleep time variability, sleep duration, BMI, waist circumference). To increase power and degrees of freedom for analyses, only variables with significant bivariate associations were included in final statistical models (p < 0.05). In addition, the specific covariate of sleep efficiency was not included, given its high correlation with the main variables of interest. This variable was omitted from mediation analyses to avoid multicollinearity. In final analyses, covariates for BMI analyses included educational level. No covariates were included in waist circumference analyses given the lack of significant correlations (Table 3).
Table 3.
Pearson's Correlation Coefficients Among Weight, Sleep, and Covariate Variables
| BMI | Waist circumference | Mean sleep time | Sleep time variability | Sleep duration | |
|---|---|---|---|---|---|
| BMI | — | ||||
| Waist circumference | 0.847** | — | |||
| Mean sleep time | −0.004 | 0.017 | — | ||
| Sleep time variability | 0.133 | 0.164* | 0.187* | — | |
| Sleep duration | −0.200* | −0.283** | −0.264** | −0.529** | — |
| C1 depressive symptoms | 0.087 | 0.083 | 0.227** | 0.204* | −0.190* |
| C2 education | −0.101* | −0.082 | −0.023 | 0.097 | 0.007 |
| C3 sleep onset latency | −0.057 | 0.007 | 0.337** | 0.406** | 0.315** |
| C4 sleep quality | 0.051 | 0.088 | 0.148 | 0.144 | 0.322** |
| C5 sleep efficiency | −0.131 | −0.203* | −0.236** | −0.639** | 0.652** |
| C6 physical activity | −0.063 | −0.107 | −0.034 | −0.027 | −0.261** |
| C7 self-rated health | 0.139 | 0.078 | −0.045 | −0.131 | −0.083 |
| C8 annual income | −0.104 | −0.139 | −0.102 | 0.056 | −0.021 |
| C9 age | 0.033 | 0.076 | −0.135 | −0.120 | −0.150 |
| C10 race | −0.145 | −0.110 | 0.130 | 0.022 | 0.060 |
C1–C10 indicate covariates.
p < 0.05, **p < 0.001.
Mediation and moderation analyses
Hayes' SPSS PROCESS macro37 ran four mediation models to test the direct association of sleep timing (mean sleep time, sleep time variability) and weight outcomes (BMI, waist circumference; aim 1), and the mediating role of sleep duration in the sleep timing–weight outcomes association (aim 2). In addition, four moderation analyses were conducted using Hayes' SPSS PROCESS37 to test for an interaction between sleep duration and timing (aim 3). Finally, three moderated mediation models were included in a follow-up analysis to assess the moderation by age of the associations between sleep timing, sleep duration, and weight outcomes. Specifically, moderated mediation models tested for sleep duration as an indirect pathway between sleep timing and weight outcomes with age as a moderation in the sleep timing–sleep duration pathway (Figs. 3 and 4). All indirect and interaction effects were tested using a nonparametric, bias-corrected bootstrapping procedure that provided an empirical approximation of the sampling distribution of the product of the estimated coefficients in the indirect paths using 5,000 resamples from the dataset.
FIG. 3.
Moderated mediation model for the association between mean sleep time and BMI as mediated by sleep duration and moderated by age. *p < 0.05.
FIG. 4.
Moderated mediation model for the association between (B) mean sleep time and waist circumference as mediated by sleep duration and moderated by age, and (C) sleep time variability and waist circumference as mediated by sleep duration and moderated by age. *p < 0.05.
Results
Data preparation and data cleaning
SPSS 23.0 was used for all data analyses. Data were cleaned, and descriptive statistics (means, SDs, and frequencies) were calculated to verify that data met the assumptions of the planned analyses. A review was conducted to assess skewness, kurtosis, and outliers for all main variables and covariates of interest. Skewness and kurtosis values for mean sleep time, BMI, and waist circumference were close to or below an absolute value of 1, indicating that they were approximately normally distributed. Sleep time variability and sleep duration were positively skewed and had positive kurtosis (i.e., values >1). Outliers were removed for these variables to reduce skewness, and a square root transformation was performed to create a normal distribution. Assumptions of independence, normality, multicollinearity, and homoskedasticity were also assessed, and criteria were sufficiently met. Power calculations using G*Power38 suggested that for regression-based analyses with three predictors, a sample size of ∼77 participants was needed to find an association with an R2 of ∼0.15 at an alpha level of 0.05, with a power of 0.80. In this study, assuming a medium effect size, 132 participants were sufficient to detect an effect.
Direct association between sleep timing and BMI
Although the overall model after controlling for covariates was significant, F(3, 128) = 5.00, p = 0.003, R2 = 0.105, mean sleep time was not significantly associated with BMI (95% CI = −0.0144 to 0.0076). In the investigation of sleep time variability and BMI, the overall model was significant after controlling for covariates, F(3, 128) = 5.32, p = 0.002, R2 = 0.111; however, sleep time variability was not significantly associated with BMI (95% CI = −0.0124 to 0.0438).
Direct association between sleep timing and waist circumference
Although the overall model was significant after controlling for sleep duration, F(2, 129) = 5.74, p = 0.004, R2 = 0.082, mean sleep time was not significantly associated with waist circumference (95% CI = −0.0358 to 0.0219). Similarly, although the overall model of the association between sleep time variability and waist circumference was significant after controlling for sleep duration, F(2, 129) = 5.78, p = 0.003, R2 = 0.082, sleep time variability was not significantly associated with waist circumference (95% CI = −0.0533 to 0.0939).
Sleep duration as a pathway linking sleep timing and weight outcomes
Next, four PROCESS mediation models were used to examine the indirect association of sleep timing (i.e., mean sleep time and sleep time variability) with weight outcomes (i.e., BMI, waist circumference) through sleep duration, and four PROCESS moderation models tested whether sleep duration strengthens or weakens the association between sleep timing and weight outcomes.
Sleep duration as a mediator in the sleep timing–weight outcomes association
Four PROCESS mediation models were run with BMI and waist circumference as outcomes to examine sleep duration as an indirect pathway between sleep timing and weight outcomes. Bootstrapping analyses were conducted to assess the indirect effect of sleep duration as a mediator of the sleep timing–weight outcomes association. Results indicated that sleep duration was a significant mediator in the mean sleep time–BMI (95% CI = 0.0001–0.0123; Fig. 1 and Table 4), mean sleep time–waist circumference (95% CI = 0.0007–0.0378; Fig. 2 and Table 4), and sleep time variability–waist circumference associations (95% CI = 0.0079–0.1006; Fig. 2 and Table 5), meaning that later sleep timing was associated with shorter duration in sleep, which in turn led to higher BMI and waist circumference. In addition, more variability in sleep time led to shorter sleep duration and thus, higher waist circumference. Furthermore, the mean sleep time–BMI model accounted for 11% of the variance in BMI, whereas the mean sleep time–waist circumference and sleep time variability–waist circumference models both accounted for 8% of the variance in waist circumference. However, sleep duration was not found to be a significant mediator in the sleep time variability–BMI association (95% CI = −0.0124 to 0.0438; Fig. 1 and Table 5).
FIG. 1.
Simple mediation models for the association between (A) mean sleep time and waist circumference as mediated by sleep duration, and (B) sleep time variability and BMI as mediated by sleep duration. *p < 0.05. Mean sleep timing is defined as the mean clock time of sleep onset across 7 days determined through actigraphy. Sleep duration is defined as the amount of time between clock time of sleep onset and clock time of sleep end through actigraphy. BMI, body mass index.
Table 4.
Coefficients for Mean Sleep Time and Sleep Time Variability (IIV) Mediation Models with Body Mass Index as Outcomes
| Predictor | M (sleep duration) | M (sleep duration) | Y (BMI) | Y (BMI) | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean sleep time–sleep duration pathway | IIV–sleep duration pathway | Mean sleep time–BMI pathway | IIV–BMI pathway | |||||||||
| Coefficient | SE | p | Coefficient | SE | p | Coefficient | SE | p | Coefficient | SE | p | |
| Constant | 594.29 | 74.72 | 0.000 | 391.52 | 15.90 | 0.000 | 49.66 | 9.85 | 0.000 | 41.56 | 4.52 | 0.000 |
| X (mean sleep time) | −0.16 | 0.05 | 0.002 | — | — | — | −0.00 | 0.01 | 0.547 | — | — | — |
| X (IIV) | — | — | — | −0.72 | 0.10 | 0.000 | — | — | — | 0.02 | 0.01 | 0.271 |
| M (sleep duration) | — | — | — | — | — | — | −0.02 | 0.01 | 0.016 | −0.02 | 0.01 | 0.156 |
| C1 education | 0.14 | 2.30 | 0.952 | 1.35 | 2.03 | 0.507 | −0.72 | 0.24 | 0.003 | −0.75 | 0.24 | 0.002 |
| R2 = 0.07 | R2 = 0.28 | R2 = 0.11 | R2 = 0.11 | |||||||||
| F(2, 129) = 4.85, p = 0.009 | F(2, 129) = 25.31, p ≤ 0.001 | F(3, 128) = 5.00, p = 0.003 | F(3, 128) = 5.32, p = 0.002 | |||||||||
p-Values in bold indicate a statistically significant effect.
IIV, IIV indicates sleep time variability which was calculated using intraindividual variability in sleep timing for each participants sleep start time.
FIG. 2.
Simple mediation models for the association between (C) mean sleep time and waist circumference as mediated by sleep duration, and (D) sleep time variability and waist circumference as mediated by sleep duration. *p < 0.05; **p < .001. Mean sleep timing is defined as the mean clock time of sleep onset across 7 days determined through actigraphy. Sleep duration is defined as the amount of time between clock time of sleep onset and clock time of sleep end through actigraphy.
Table 5.
Coefficients for Mean Sleep Time and Sleep Time Variability (IIV) Mediation Models with Waist Circumference as Outcomes
| Predictor | M (sleep duration) | M (sleep duration) | Y (WC) | Y (WC) | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean sleep time–sleep duration pathway | IIV–sleep duration pathway | Mean sleep time–WC pathway | IIV–WC pathway | |||||||||
| Coefficient | SE | p | Coefficient | SE | p | Coefficient | SE | p | Coefficient | SE | p | |
| Constant | 595.33 | 72.38 | 0.000 | 401.18 | 6.48 | 0.000 | 140.02 | 24.68 | 000 | 125.21 | 11.23 | 0.000 |
| X (mean sleep time) | −0.16 | 0.05 | 0.002 | — | — | — | −0.01 | 0.01 | 0.635 | — | — | — |
| X (IIV) | — | — | — | −0.71 | 0.10 | 0.000 | — | — | — | 0.02 | 0.04 | 0.586 |
| M (sleep duration) | — | — | — | — | — | — | −0.08 | 0.024 | 0.001 | −0.07 | 0.03 | 0.012 |
| R2 = 0.07 | R2 = 0.23 | R2 = 0.08 | R2 = 0.08 | |||||||||
| F(1, 130) = 9.76, p = 0.002 | F(1, 130) = 50.40, p ≤ 0.001 | F(2, 129) = 5.74, p = 0.004 | F(2, 129) = 5.78, p = 0.004 | |||||||||
p-Values in bold indicate a statistically significant effect.
WC, waist circumference.
Sleep duration as a moderator of the sleep timing–weight outcomes association
Next, four moderation analyses were conducted to assess whether sleep duration strengthens or weakens the association between sleep timing and weight outcomes. No significant moderating effects of sleep duration were found for the mean sleep time–BMI (−0.0002, 0.0001), mean sleep time–waist circumference (−0.0006, 0.0002), sleep time variability–BMI (−0.0002, 0.0004), and sleep time variability–waist circumference associations (−0.0005, 0.0010).
Follow-up analysis: age as a moderator of the sleep timing–sleep duration pathway
Given the significant mediation by sleep duration in the indirect association between sleep timing and weight outcomes (i.e., BMI and waist circumference), it was of interest to further explore this association in midlife women. As women in midlife experience menopause that can impact sleep timing and sleep duration,7,39 age was added as a moderator to the three significant mediation models (i.e., mean sleep time–BMI, mean sleep time–waist circumference, and sleep time variability–waist circumference) to determine whether models differed depending on age. Age was dichotomized as “younger” (40–52 years) and “older” (53–64 years) to represent premenopause and postmenopause status, respectively.
Age as a moderator of mean sleep timing–sleep duration associations
First, age was assessed as a moderator of the mean sleep timing–sleep duration associations (i.e., BMI model, waist circumference model) using moderated mediation. A significant moderation was present in both mean sleep time–sleep duration analyses, b = −0.26, p < 0.05. For younger women, there was no association between mean sleep timing and sleep duration, b = −0.00, 95% CI = −0.01 to 0.01. However, for older women, there was a significant association between mean sleep timing and sleep duration, b = 0.01, 95% CI = 0.00–0.01, where older women who had later sleep timing were more likely to have shorter sleep duration compared with the younger women. In addition, the mediation of the mean sleep timing and BMI association by sleep duration was contingent on age, such that the indirect pathway of sleep duration held for the “older” women but not for the “younger” women, b = 0.01, 95% CI = 0.00–0.02 (Fig. 3 and Table 6). Furthermore, the mean sleep time–BMI model accounted for 11.5% of the variance. However, the mediation of the mean sleep timing and waist circumference association by sleep duration was not contingent on age. The mean sleep timing–sleep duration–waist circumference mediation was no longer upheld when age was added to the model, b = 0.02, 95% CI = −0.00 to 0.05 (Fig. 4 and Table 6).
Table 6.
Coefficients for Moderated Mediation Models with Body Mass Index and Waist Circumference as Outcomes
| Predictor variables | Outcome variables | |||||
|---|---|---|---|---|---|---|
| M (sleep duration) | Coefficient | Y (BMI) | p | |||
| Coefficient | SE | p | SE | |||
| Mean sleep time–BMI model | ||||||
| Constant | 0.30 | 250.12 | 0.999 | 49.656 | ||
| X (mean sleep time) | 0.27 | 0.18 | 0.131 | — | — | — |
| M (sleep duration) | — | — | — | −0.023 | ||
| W (age) | 366.50 | 148.26 | 0.015 | — | — | — |
| Sleep time × status | −0.26 | 0.10 | 0.013 | — | — | — |
| C1 education | −0.35 | 2.28 | 0.878 | −0.724 | 0.242 | 0.003 |
| R2 = 0.116 | R2 = 0.15 | |||||
| F(4, 127) = 4.15, p < 0.05 | F(3, 128) = 5.00, p < 0.05 | |||||
| M (sleep duration) | Y (WC) | |||||
|---|---|---|---|---|---|---|
| Mean sleep time–WC model | ||||||
| Constant | 1.09 | 249.11 | 0.997 | 140.02 | 24.68 | 0.000 |
| X (mean sleep time) | 0.27 | 0.18 | 0.132 | — | — | — |
| M (sleep duration) | — | — | — | −0.082 | 0.02 | 0.001 |
| W (age) | 364.38 | 147.06 | 0.015 | −0.007 | 0.02 | 0.635 |
| Sleep time × status | −0.26 | 0.10 | 0.013 | — | — | — |
| R2 = 0.115 | R2 = 0.08 | |||||
| F(3, 128) = 5.56, p < 0.05 | F(2, 129) = 5.74, p < 0.05 | |||||
| M (sleep duration) | Y (WC) | |||||
|---|---|---|---|---|---|---|
| Sleep time variability–WC model | ||||||
| Constant | 426.43 | 21.63 | 0.000 | 125.21 | 11.14 | 0.000 |
| X (sleep time variability) | −1.02 | 0.34 | 0.003 | 0.02 | 0.04 | 0.012 |
| M (sleep duration) | — | — | — | −0.07 | 0.03 | 0.586 |
| W (age) | −16.05 | 13.13 | 0.223 | — | — | — |
| Sleep time × status | 0.19 | 0.20 | 0.352 | — | — | — |
| R2 = 0.288 | R2 = 0.08 | |||||
| F(3, 128) = 17.24, p < 0.001 | F(2, 129) = 5.78, p < 0.05 | |||||
Labels within the table reflect the following: X refers to the independent variable (mean sleep time, sleep time variability), Y refers to the dependent variable (BMI, waist circumference), M refers to the mediator (sleep duration), W refers to the moderator (age), and Cx refer to covariates. p-Values in bold indicate a statistically significant effect (p < .05).
BMI, body mass index.
Age as a moderator of sleep timing variability–sleep duration associations
Next, age was assessed as a moderator of the sleep timing variability–sleep duration association using moderated mediation. A significant moderation was not present in the sleep time variability–sleep duration analyses, b = 0.19, p = 0.35. However, the mediation of the sleep time variability and waist circumference association by sleep duration was significant, 95% CI = 0.0079–0.1006, thus indicating that the sleep duration was a significant indirect pathway between sleep time variability and waist circumference when accounting for age (Fig. 4). The sleep time variability–waist circumference model accounted for 28.8% of the variance in waist circumference. Additional information regarding the association between all variables and age is found in Table 6.
Discussion
This study examined the association between sleep timing, sleep duration, and weight outcomes in midlife women using prospective measures and objective assessments. Mediation analyses established that sleep timing was not directly associated with weight in midlife women. Rather, sleep timing was indirectly associated with weight outcomes through sleep duration, indicating that later and more variable sleep timing was associated with shorter sleep duration, which was linked to higher BMI and waist circumference. Sleep duration did not moderate the association between sleep timing and weight, indicating that the effect of sleep timing on weight was not exacerbated by sleep duration. Finally, age influenced the association among sleep timing and weight outcomes where older women with later bed timing were more likely than younger women to report shorter sleep duration, thus influencing weight outcomes (i.e., BMI).
The finding that sleep timing indirectly, rather than directly, contributes to weight outcomes in this sample of midlife women is novel as previous research has investigated only direct pathways between sleep timing and weight in this population.7,15,24,40 However, the existing small body of literature that has explored the direct association has produced inconsistent results. Baron et al. found that sleep timing was associated with BMI in a sample of adults only when sleep duration was not included as a covariate.24 Conversely, in research specifically focused on midlife women, Taylor et al. found that sleep timing was not directly associated with BMI outcomes in both cross-sectional and prospective analyses after controlling for covariates including sleep duration.7
There are multiple theoretical and methodological reasons to explain the discrepancy in findings. Theoretically, sleep timing and duration share similar biological and behavioral mechanisms that could explain their association with weight outcomes. In particular, delayed or more variable sleep timing may create less opportunity for sleep (i.e., shorter sleep duration).26,41 In addition, dysregulated sleep timing as well as decreases in sleep duration can increase appetite and preference for energy-dense food, which can influence eating behavior and weight gain.42–44 Shorter sleep duration can also impair glucose tolerance and lead to increases in insulin resistance, which influences progression to negative disease outcomes including type II diabetes.45 As such, although both later and more variable sleep timing and shorter sleep duration can influence weight, the association between sleep timing, sleep duration, and weight was driven primarily through sleep duration, not sleep timing in the current sample of midlife women.
Methodologically, differences in the investigation of the association among sleep timing and weight exist in terms of (1) the measurement of sleep and (2) the operationalization of sleep timing. First, research that has found an independent and direct association between sleep timing and weight outcomes when controlling for factors such as sleep duration has used self-report measures of average bedtime or general bedtime preference.15,40 Conversely, research that has not found an independent association has used daily measures of sleep time in the form of daily sleep diaries or actigraphy.7,24 Sleep diaries and actigraphy are the gold standards of ambulatory and self-report sleep assessment given their ability to decrease recall bias, provide a more precise estimate of sleep behaviors, and capture fluctuation that can exist across daily sleep behaviors.46 This study explored the association between sleep timing and weight outcomes in midlife women using actigraphy. Current results are consistent with other findings investigating sleep timing using prospective daily sleep methodology.
There is also methodological variability in the operationalization of sleep timing. Sleep timing can be measured through mean sleep time (i.e., an individual's average bedtime) or variability in sleep time (i.e., daily fluctuation in sleep time). In samples including midlife women, the majority of work in this area has investigated mean sleep time,15,24 and only one study has investigated sleep time variability in relation to weight.7 Variability in sleep timing encompasses day-to-day factors that can influence sleep–wake cycles, which are obscured when only the average sleep time is measured.22 Despite research suggesting that irregularity in sleep time increases risk of health problems such as obesity,47 variability in sleep time is underinvestigated in midlife women. As women in midlife may be particularly vulnerable to fluctuation in night-to-night sleep timing behavior due to various biopsychosocial demands,20,21,48 it is of value to comprehensively and accurately understand the link between sleep timing and weight outcomes to inform clinical recommendations.
In addition to sleep duration serving as an indirect link between sleep timing and weight outcomes, this study highlights that this association is also contingent on age. Whereas this study was not able to determine menopause status, midlife women in the “older” age range (53–64 years) were more likely to have later sleep timing, which was associated with shorter sleep duration and higher weight outcomes. However, age appeared to be less impactful in the relation between sleep timing, sleep duration, and waist circumference, as sleep duration served as an explanatory mechanism between sleep timing and waist circumference regardless of age. These findings highlight the importance of addressing developmental stage when examining sleep and weight outcomes in this population. Previous research highlights that women of particular menopause statuses (i.e., perimenopause, postmenopause) are more likely to report shorter sleep duration in comparison with women of premenopause status.49 In addition, weight outcomes such as obesity are more prevalent when women are undergoing the menopause transition in comparison with premenopause women.50 As such, differentiating results based on developmental status can aid in the assessment of who is at risk of poor sleep as well as promote understanding of factors that influence weight outcomes in this population.
Summary and implications
This study did not find support that sleep timing is directly associated with weight outcomes in this population. Rather, sleep timing indirectly contributed to poor weight outcomes through sleep duration, suggesting that delayed or variable bedtimes have important implications for weight through sleep duration. In addition, age differentiated the association among certain aspects of sleep timing and negative weight outcomes where older women in midlife with later bed timing were more likely to report shorter sleep duration and higher weight outcomes in comparison with younger women in midlife. Future research is needed to help inform clinical recommendations regarding how sleep behaviors in midlife women may serve as preventative or intervention approaches for maintaining a healthy weight. In particular, longitudinal designs would be beneficial for investigating the direct association between sleep timing and weight outcomes. Specifically, repeated concurrent assessment of daily sleep behavior and objectively measured weight across time would be particularly useful.
Strengths and limitations
Although this study has many strengths, several limitations must be addressed. First, the current sample lacks racial and ethnic diversity. The majority of the women are White, thus preventing broad generalizability of results. In addition, due to the use of secondary data, we could not control for potentially relevant covariates, including menopausal status, psychosocial factors (i.e., eating behavior, age of children living in the home, sleep of bed partners), and daily behaviors (i.e., work schedule, hours worked/week), which can influence sleep and weight outcomes. In particular, menopausal status is an important variable to include or control for in investigations of sleep and weight as sleep concerns are prevalent in menopause and can influence weight and eating behavior.51,52 Although we could not specify menopausal status of the participants, we used age as an approximate measurement to determine whether age may account for differences in the association among sleep timing, sleep duration, and weight. As age accounted for differences in study results, there is preliminary support that menopause may, in part, account for the observed age-related differences. However, this cannot be determined in this study. Finally, as this study is cross-sectional, causality cannot be determined.
Although the study contains limitations, there are several strengths. First, MIDUS-II represents a national sample of women, allowing us to examine a broad range of women in the United States. Second, actigraphy and daily sleep diaries enabled the examination of sleep timing across multiple days, including weekdays and weekends. As such, we assessed sleep with reduced recall bias. We also provided a comprehensive account of sleep through the use of actigraphic measurements (i.e., mean sleep time, sleep time variability). Third, we used objective measures of BMI and waist circumference to explore the association between sleep timing and waist circumference in midlife women. As BMI tends to be used more frequently in studies investigating weight outcomes in this population, this study was able to extend knowledge regarding the association between sleep timing and abdominal circumference in this population. Finally, this project extends the current literature by investigating multiple aspects of sleep timing in relation to weight outcomes. As sleep timing is a relatively new sleep construct with sparse research examining its impact on weight outcomes in midlife women, this study serves to add to the existing literature and provides recommendations for future research.
Acknowledgments
D.R.S. contributed to the study concept, and D.R.S. and N.D.D. contributed to study design. Testing and data collection were performed by MIDUS-II study staff who were supported by the John D. and Catherine T. MacArthur Foundation Research Network on Successful Midlife Development. D.R.S. analyzed the data and drafted the article, which was critically reviewed by N.D.D. All authors approved the final version of the article for submission. The MIDUS 1 study (Midlife in the United States) was supported by the John D. and Catherine T. MacArthur Foundation Research Network on Successful Midlife Development. The MIDUS 2 research was supported by a grant from the National Institute on Aging (P01-AG020166) to conduct a longitudinal follow-up of the MIDUS 1 investigation.
Author Disclosure Statement
D.R.S. has no conflicting financial interests. N.D.D. serves as a sleep consultant for the National Sleep Foundation and Merck Sharp & Dohme Corp.
References
- 1. Ogden CL, Carroll MD, Fryar CD, Flegal KM. Prevalence of obesity among adults and youth: United States, 2011–2014. NCHS Data Brief 2015;1–8 [PubMed] [Google Scholar]
- 2. Grundy SM, Cleeman JI, Daniels SR, et al. Diagnosis and management of the metabolic syndrome. Circulation 2005;112. Available at: http://circ.ahajournals.org/content/112/17/2735.short Accessed June21, 2017 [DOI] [PubMed]
- 3. World Health Organization. Obesity: Preventing and managing the global epidemic. Exec Summ WHO Tech Rep Ser 2010;1997:5–8 [PubMed] [Google Scholar]
- 4. Soules MR, Sherman S, Parrott E, et al. Stages of reproductive aging workshop (STRAW). Available at: www.liebertpub.com Accessed October12, 2018 [DOI] [PubMed]
- 5. Davis SR, Castelo-Branco C, Chedraui P, et al. Understanding weight gain at menopause. Climacteric 2012;15:419–429 [DOI] [PubMed] [Google Scholar]
- 6. Gangwisch JE, Malaspina D, Boden-Albala B, Heymsfield SB. Inadequate sleep as a risk factor for obesity: Analyses of the NHANES I. Sleep 2005;28:1289–1296 [DOI] [PubMed] [Google Scholar]
- 7. Taylor BJ, Matthews KA, Hasler BP, et al. Bedtime variability and metabolic health in midlife women: The SWAN Sleep Study. Sleep 2016;39:457–465 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Polo-Kantola P. Sleep problems in midlife and beyond. Maturitas 2011;68:224–232 [DOI] [PubMed] [Google Scholar]
- 9. Ford K, Sowers M, Crutchfield M, Wilson A, Jannausch M. A longitudinal study of the predictors of prevalence and severity of symptoms commonly associated with menopause. Menopause 2005;12:308–317 [DOI] [PubMed] [Google Scholar]
- 10. National Institutes of Health. National Institutes of Health State-of-the-Science Conference statement: Management of menopause-related symptoms. Ann Intern Med 2005;142:1003–1013 [PubMed] [Google Scholar]
- 11. Feinberg I. Changes in sleep cycle pattern with age Measurement of sleep need in afolescemce View project theories of schizophrenic mind View project. Artic J Psychiatr Res 1974. DOI: 10.1016/0022-3956(74)90011-9 [DOI] [PubMed]
- 12. Hall MH, Okun ML, Sowers M, et al. Sleep is associated with the metabolic syndrome in a multi-ethnic cohort of midlife women: The SWAN Sleep Study. Sleep 2012;35:783–790 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Strauss JR. The reciprocal relationship between menopausal symptoms and depressive symptoms: A 9-year longitudinal study of American women in midlife. Maturitas 2011;70:302–306 [DOI] [PubMed] [Google Scholar]
- 14. Buysse DJ. Sleep health: Can we define it? Does it matter? Sleep 2014;37:9–17 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Lee S, Shin C. Habitual sleep initiation time and metabolic syndrome in middle-aged and elderly adults. Sleep Biol Rhythms 2015;13:371–379 [Google Scholar]
- 16. Choi J-K, Kim M-Y, Kim J-K, et al. Association between short sleep duration and high incidence of metabolic syndrome in midlife women. Tohoku J Exp Med 2011;225:187–193 [DOI] [PubMed] [Google Scholar]
- 17. Delezie J, Challet E. Interactions between metabolism and circadian clocks: Reciprocal disturbances. Ann N Y Acad Sci 2011;1243:30–46 [DOI] [PubMed] [Google Scholar]
- 18. Gonnissen HKJ, Hulshof T, Westerterp-Plantenga MS. Chronobiology, endocrinology, and energy- and food-reward homeostasis. Obes Rev 2013;14:405–416 [DOI] [PubMed] [Google Scholar]
- 19. Cappuccio FF, Taggart FFM, Kandala NN-B, et al. Meta-analysis of short sleep duration and obesity in children and adults. Sleep 2008;31:619–626 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Owens JF, Matthews KA. Sleep disturbance in healthy middle-aged women. Maturitas 1998;30:41–50 [DOI] [PubMed] [Google Scholar]
- 21. Hislop J, Arber S. Sleepers wake! the gendered nature of sleep disruption among mid-life women. Sociology 2003;37:695–711 [Google Scholar]
- 22. Bei B, Wiley JF, Trinder J, Manber R. Beyond the mean: A systematic review on the correlates of daily intraindividual variability of sleep/wake patterns. Sleep Med Rev 2016;28:108–124 [DOI] [PubMed] [Google Scholar]
- 23. Vgontzas AN, Fernandez-Mendoza J, Miksiewicz T, et al. Unveiling the longitudinal association between short sleep duration and the incidence of obesity: The Penn State Cohort. Int J Obes 2014;38:825–832 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Baron KG, Reid KJ, Kern AS, Zee PC. Role of sleep timing in caloric intake and BMI. Obesity 2011;19:1374–1381 [DOI] [PubMed] [Google Scholar]
- 25. Knutson KL, Spiegel K, Penev P, Van Cauter E. The metabolic consequences of sleep deprivation. Sleep Med Rev 2007;11:163–178 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Roepke SE, Duffy JF. Differential impact of chronotype on weekday and weekend sleep timing and duration. Nat Sci Sleep 2010;2010:213–220 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Ford ES, Maynard LM, Li C. Trends in mean waist circumference and abdominal obesity among US adults, 1999–2012. JAMA 2014;312:1151–1153 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Ancoli-Israel S, Cole R, Alessi C, Chambers M, Moorcroft W, Pollak CP. The role of actigraphy in the study of sleep and circadian rhythms. Sleep 2003;26:342–392 [DOI] [PubMed] [Google Scholar]
- 29. Carney CE, Lajos LE, Waters WF. Wrist actigraph versus self-report in normal sleepers: Sleep schedule adherence and self-report validity. Behav Sleep Med 2004;2:134–143 [DOI] [PubMed] [Google Scholar]
- 30. Carney CE, Buysse DJ, Ancoli-Israel S, et al. The consensus sleep diary: Standardizing prospective sleep self-monitoring. Sleep 2012;35:287–302 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Ram N, Gerstorf D. Time-structured and net intraindividual variability: Tools for examining the development of dynamic characteristics and processes. Psychol Aging 2009;24:778–791 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Dillon HR, Lichstein KL, Dautovich ND, Taylor DJ, Riedel BW, Bush AJ. Variability in self-reported normal sleep across the adult age span. J Gerontol Ser B Psychol Sci Soc Sci 2015;70:46–56 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Campbell L V, Samaras K. What is the evidence, reasons for and impact of weight gain during menopause? Med J Aust 2000;173 Suppl:S100–S101 [DOI] [PubMed] [Google Scholar]
- 34. Jennings JR, Muldoon MF, Hall M, Buysse DJ, Manuck SB. Self-reported sleep quality is associated with the metabolic syndrome. Sleep 2007;30:219–223 [DOI] [PubMed] [Google Scholar]
- 35. Lahmann P, Lissner L, Gullberg B, Berglund G. Sociodemographic factors associated with long- term weight gain, current body fatness and central adiposity in Swedish women. Int J Obes 2000;24:685–694 [DOI] [PubMed] [Google Scholar]
- 36. Radloff LS. The CES-D scale: A self-report depression scale for research in the general population. Appl Psychol Meas 1977;1:385–401 [Google Scholar]
- 37. Hayes AF. Introduction to mediation, moderation, and conditional process analysis: a regression-based approach. 2012. Available at: http://library.wur.nl/WebQuery/clc/2063623 Accessed June21, 2017
- 38. Faul F, Erdfelder E, Buchner A, Lang A-G. Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behav Res Methods 2009;41:1149–1160 [DOI] [PubMed] [Google Scholar]
- 39. Lampio L, Polo-Kantola P, Himanen S-L, et al. Sleep during menopausal transition: A 6-year follow-up. Sleep 2017;40 DOI: 10.1093/sleep/zsx090 [DOI] [PubMed] [Google Scholar]
- 40. Yu JH, Yun C-H, Ahn JH, et al. Evening chronotype is associated with metabolic disorders and body composition in middle-aged adults. J Clin Endocrinol Metab 2015;100:1494–1502 [DOI] [PubMed] [Google Scholar]
- 41. Wittmann M, Dinich J, Merrow M, Roenneberg T. Social jetlag: Misalignment of biological and social time. Chronobiol Int 2006;23:497–509 [DOI] [PubMed] [Google Scholar]
- 42. Benedict C, Brooks SJ, O'Daly OG, et al. Acute sleep deprivation enhances the brain's response to hedonic food stimuli: An fMRI study. J Clin Endocrinol Metab 2012;97:E443–E447 [DOI] [PubMed] [Google Scholar]
- 43. Nedeltcheva AV, Kilkus JM, Imperial J, Kasza K, Schoeller DA, Penev PD. Sleep curtailment is accompanied by increased intake of calories from snacks. Am J Clin Nutr 2008;89:126–133 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Spiegel K, Tasali E, Penev P, et al. Brief communication: sleep curtailment in healthy young men is associated with decreased leptin levels, elevated ghrelin levels, and increased hunger and appetite. Ann Intern Med 2004;141:846. [DOI] [PubMed] [Google Scholar]
- 45. Spiegel K, Knutson K, Leproult R, Tasali E, Cauter EV. Sleep loss: A novel risk factor for insulin resistance and Type 2 diabetes. J Appl Physiol 2005;99:2008–2019 [DOI] [PubMed] [Google Scholar]
- 46. Martin JL, Hakim AD. Wrist actigraphy. Chest 2011;139:1514–1527 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Patel SR, Hayes AL, Blackwell T, et al. The association between sleep patterns and obesity in older adults. Int J Obes 2014;38:1159–1164 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. Shaver JLF, Johnston SK, Lentz MJ, Landis CA. Stress exposure, psychological distress, and physiological stress activation in midlife women with insomnia. Psychosom Med 2002;64:793–802 [DOI] [PubMed] [Google Scholar]
- 49. Woods NF, Mitchell ES. Sleep symptoms during the menopausal transition and early postmenopause: Observations from the Seattle Midlife Women's Health Study. Sleep 2010;33:539–549 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Pasquali R, Casimirri F, Labate AM, et al. Body weight, fat distribution and the menopausal status in women. The VMH Collaborative Group. Int J Obes 1994;18:614–621 [PubMed] [Google Scholar]
- 51. Polo-Kantola P. Sleep and menopause. Women's Health 2007;3:99–106 [DOI] [PubMed] [Google Scholar]
- 52. Kapoor E, Collazo-Clavell ML, Faubion SS. Weight gain in women at midlife: A concise review of the pathophysiology and strategies for Management. Mayo Clin Proc 2017;92:1552–1558 [DOI] [PubMed] [Google Scholar]