Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Chronobiology International

The Journal of Biological and Medical Rhythm Research

ISSN: 0742-0528 (Print) 1525-6073 (Online) Journal homepage: https://www.tandfonline.com/loi/icbi20

Relationships between chronotype, social jetlag,


sleep, obesity and blood pressure in healthy young
adults

Daria M. McMahon, James B. Burch, Shawn D. Youngstedt, Michael D. Wirth,


James W. Hardin, Thomas G. Hurley, Steven N. Blair, Gregory A. Hand, Robin
P. Shook, Clemens Drenowatz, Stephanie Burgess & James R. Hebert

To cite this article: Daria M. McMahon, James B. Burch, Shawn D. Youngstedt, Michael D.
Wirth, James W. Hardin, Thomas G. Hurley, Steven N. Blair, Gregory A. Hand, Robin P. Shook,
Clemens Drenowatz, Stephanie Burgess & James R. Hebert (2019): Relationships between
chronotype, social jetlag, sleep, obesity and blood pressure in healthy young adults, Chronobiology
International, DOI: 10.1080/07420528.2018.1563094

To link to this article: https://doi.org/10.1080/07420528.2018.1563094

View supplementary material

Published online: 21 Jan 2019.

Submit your article to this journal

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=icbi20
CHRONOBIOLOGY INTERNATIONAL
https://doi.org/10.1080/07420528.2018.1563094

Relationships between chronotype, social jetlag, sleep, obesity and blood


pressure in healthy young adults
Daria M. McMahona, James B. Burcha,b,c, Shawn D. Youngstedtd,e, Michael D. Wirtha,b,f, James W. Hardina,
Thomas G. Hurleyb, Steven N. Blaira,g, Gregory A. Handh, Robin P. Shooki, Clemens Drenowatzg,j,
Stephanie Burgessf, and James R. Heberta,b
a
Department of Epidemiology and Biostatistics, University of South Carolina, Columbia, South Carolina, USA; bCancer Prevention and Control
Program, University of South Carolina, Columbia, South Carolina, USA; cWJB Dorn Department of Veterans Affairs Medical Center Research
Department, Columbia, South Carolina, USA; dCollege of Nursing and Health Innovation, Arizona State University, Phoenix, Arizona, USA;
e
Department of Veterans Affairs, Phoenix Health Care System Research Department, Phoenix, Arizona, USA; fCollege of Nursing, University of
South Carolina, Columbia, South Carolina, USA; gDepartment of Exercise Science, University of South Carolina, Columbia, South Carolina,
USA; hSchool of Public Health, West Virginia University, Morgantown, West Virginia, USA; iDepartment of Pediatrics, Center for Healthy
Lifestyles and Nutrition, Children’s Mercy Hospital, Kansas City, Missouri, USA; jDivision of Physical Education, University of Education Upper
Austria, Linz, Austria

ABSTRACT ARTICLE HISTORY


Sleep disturbances, chronotype and social jetlag (SJL) have been associated with increased risks Received 28 September 2018
for major chronic diseases that take decades to develop, such as obesity, metabolic syndrome and Revised 17 December 2018
cardiovascular disease. Potential relationships between poor sleep, chronotype and SJL as they Accepted 20 December 2018
relate to metabolic risk factors for chronic disease have not been extensively investigated. This KEYWORDS
prospective study examined chronotype, SJL and poor sleep in relation to both obesity and Actigraphy; body mass
elevated blood pressure among healthy young adults. index; chronotype; repeated
SJL and objective sleep measures (total sleep time, sleep onset latency, wake after sleep onset measures latent class
and sleep efficiency) were derived from personal rest/activity monitoring (armband actigraphy) analysis; sleep efficiency;
among 390 healthy adults 21–35 years old. Participants wore the device for 6–10 days at 6-month social jetlag
intervals over a 2-year period (n = 1431 repeated observations). Chronotypes were categorized
into morning, intermediate and evening groups using repeated measures latent class analysis.
Means of SJL and sleep measures among latent chronotype groups were compared using partial
F-tests in generalized linear mixed models. Generalized linear mixed models also were used to
generate odds ratios (ORs) with 95% confidence intervals (CIs) examining the relationship
between repeated measures of chronotype, SJL, sleep and concurrent anthropometric outcome
measures (body mass index, percentage of body fat, waist-to-hip ratio, waist-to-height ratio),
systolic blood pressure and diastolic blood pressure.
Sleep latency ≥12 min was associated with increased odds of a high waist-to-height ratio
(OR = 1.37; CI: 1.03–1.84). Neither chronotype nor SJL was independently associated with anthro-
pometric outcomes or with blood pressure. Relationships between poor sleep and anthropometric
outcomes or blood pressure varied by chronotype. Morning types with total sleep time <6 h, sleep
efficiency <85% or wake after sleep onset ≥60 min were more likely to have an increased
percentage of body fat, waist-to-hip ratio and waist-to-height ratio relative to those with an
intermediate chronotype. Similarly, sleep latency ≥12 min was associated with increased odds of
elevated systolic blood pressure (OR = 1.90; CI: 1.15–3.16, pinteraction = 0.02) among morning versus
intermediate chronotypes. No relationships between poor sleep and obesity or elevated blood
pressure were observed among evening chronotypes.
The results from this study among healthy young adults suggest that poor sleep among
morning types may be more strongly associated with obesity and elevated blood pressure relative
to those with an intermediate (neutral) chronotype. Sleep-related metabolic alterations among
different chronotypes warrant further investigation.

CONTACT James B. Burch burch@mailbox.sc.edu Department of Epidemiology and Biostatistics, Cancer Prevention & Control Program, University of
South Carolina, 915 Greene St, Room 229, Columbia, SC 29208, USA
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/icbi.
Supplemental data for this article can be accessed here.
© 2019 Taylor & Francis Group, LLC
2 D. M. MCMAHON ET AL.

Introduction intermediate chronotype (Culnan et al. 2013; Wang


2014). However, not all studies have confirmed this
In 2013–2014, almost 38% of adults in the United
association; in one study, morning rather than evening
States were obese (Flegal et al. 2016). Obesity is
chronotype was associated with increased risk of
associated with many serious complications such
weight gain (Roenneberg et al. 2012), while in two
as metabolic syndrome, type 2 diabetes mellitus,
others body mass index (BMI) did not differ by chron-
arterial hypertension, cardiovascular disease and
otype (Maukonen et al. 2017; Pabst et al. 2009).
cancer (Malnick and Knobler 2006). Some of
Chronotype in combination with social schedule
these conditions begin to develop early in child-
can elicit an increase in SJL, particularly among
hood (Ogden et al. 2016).
those with an evening chronotype (Roenneberg et al.
One modifiable risk factor for weight gain,
2012; Wittmann et al. 2006). Thus, it is reasonable to
obesity and metabolic syndrome is short, dis-
hypothesize that SJL may mediate the relationship
rupted or inefficient sleep (Bailey et al. 2014;
between chronotype and obesity. One study that
Hasler et al. 2004; Vgontzas et al. 2014). Poor
tested this hypothesis produced conflicting results
sleep also has been associated with an increased
(Culnan and Kloss 2013), whereas other studies have
risk of hypertension and cardiovascular disease
assumed that chronotype, SJL and sleep duration may
(Gottlieb et al. 2006), diabetes (Cespedes et al.
confound each other’s association with anthropo-
2016) and cancer (Gu et al. 2016). Moreover,
metric indices of metabolic syndrome (Parsons et al.
sleep disturbances are prevalent and persistent
2015; Roenneberg et al. 2012).
in young adults (Hasler et al. 2004; McMahon
This study was originally designed to examine the
et al. 2018; Vgontzas et al. 2014).
effects of energy intake and expenditure on changes in
Modern life often causes persistent misalignment in
weight and body composition (Hand et al. 2013). In
sleep/wake timing between weekdays and weekends,
a cross-sectional analysis, relationships between sleep
a phenomenon referred to as “social jetlag” (SJL)
disturbances, BMI and percentage of body fat (BF%)
(Wittmann et al. 2006). SJL can start in teenage years
were evaluated at baseline (Wirth et al. 2015), although
and continue throughout working years until retire-
chronotype and SJL were not considered. More
ment, and thus may affect millions of people in indus-
recently, temporal patterns of sleep, chronotype and
trialized countries over long periods of time
SJL were characterized (McMahon et al. 2018),
(Roenneberg et al. 2012). SJL has been associated
although relationships between these factors and obe-
with sleep loss (Roenneberg et al. 2007; Wittmann
sity or elevated blood pressure were not evaluated.
et al. 2006), as well as increases in obesity and meta-
Young adults ~21–35 years of age are at risk for
bolic syndrome, biomarkers of systemic inflammation
developing metabolic changes that may predispose
and with mood disturbances (Levandovski et al. 2011;
them to an increased risk of chronic disease later in
Parsons et al. 2015; Roenneberg et al. 2012; Rutters
life (Ogden et al. 2016).
et al. 2014).
During the last decade, prevalence of obesity has
Preferred sleep timing and sleep efficiency (SE)
remained high (30%) in this age group in the United
are related to an individual’s diurnal preference or
States (Flegal et al. 2016). Adults in this age range also
chronotype (Juda et al. 2013; Roenneberg et al.
may experience SJL and have persistently poor sleep
2007). Individuals with an evening chronotype
(Hasler et al. 2004; McMahon et al. 2018; Vgontzas
tend to accumulate sleep debt during the week,
et al. 2014). The present analysis prospectively exam-
for which they often try to compensate with longer
ined independent associations between sleep, chron-
sleep and delayed sleep onset during the weekend
otype or SJL and measures of obesity and elevated
(Roenneberg et al. 2007; Valdez et al. 1996).
blood pressure among healthy young adults 21–-
Morning chronotypes, by contrast, are more likely
35 years of age. In addition to these direct associa-
to develop a sleep debt during the weekend
tions, this analysis also evaluated the potential
(Roenneberg et al. 2007; Valdez et al. 1996).
interactions between chronotype, sleep and SJL. It
Some studies suggest that individuals with an eve-
also was hypothesized that SJL may act as
ning chronotype are more likely to be overweight or
a mediator of the relationship between chronotype
obese compared to those with a morning or
and obesity or elevated blood pressure. In this
CHRONOBIOLOGY INTERNATIONAL 3

Figure 1. Analyses of the relationships between chronotype, social jetlag, sleep measures, obesity and elevated blood pressure. The
dashed lines represent relationships for which chronotype was assessed as a potential effect modifier.

manner, both the independent and combined effects considered ineligible, 66 declined to participate
of chronotype, SJL and sleep on obesity and elevated and 169 did not attend the baseline clinic and
blood pressure were evaluated in a single study (see therefore did not enter the study (Hand et al.
Figure 1). 2013). Thus, a total of 430 healthy men and
women residing in the Columbia, SC, region
were enrolled. The study was approved by the
Institutional Review Board at the University of
Methods
South Carolina, and all study participants pro-
Study population vided informed consent. The Energy Balance
Study was originally planned for 1 year; however,
The population characteristics and recruitment pro-
as additional funding was obtained, participants
cedures have been previously described (Hand et al.
were asked if they wished to continue for
2013). Briefly, eligible individuals were 21–35 years
another year.
of age, with a BMI of 20–35 kg/m2, no major health
After a comprehensive baseline assessment
conditions or large changes in body composition
(demographics, health history, anthropometric
during the previous 6 months and no pregnancy or
measurements, diet, actigraphy), participants
childbirth during the past 12 months (Hand et al.
were re-examined every 6 months to characterize:
2013). Individuals with major depression, anxiety
anthropometric outcomes, blood pressure, sleep/
disorder or panic attacks (including those who took
wake patterns via armband actigraphy and diet.
selective serotonin reuptake inhibitors) were
During the 2-year study period (2011–2013),
excluded. Those who met diagnostic criteria for
each participant completed between 1 and 5
arterial hypertension or diabetes – systolic blood
assessments.
pressure (SBP) ≥150 mmHg and/or diastolic blood
pressure (DBP) ≥90 mmHg, or a blood glucose level
>145 mg/dl – also were excluded (Hand et al. 2013).
Chronotype, SJL and sleep characteristics
Potential participants provided demographic
and health characteristics via the internet, followed Sleep characteristics were derived from the data
by a telephone interview to establish eligibility. collected using a SenseWear® Mini
Those considered eligible were invited for an Armband (BodyMedia Inc, Pittsburgh, PA) worn
orientation session and three baseline screenings. by participants over the triceps muscle of the left
Out of 1831 persons contacted, 1116 were arm (Sharif and Bahammam 2013). An armband
4 D. M. MCMAHON ET AL.

sensor records limb movements at a frequency of week day (Roenneberg et al. 2012; Wittmann et al.
32 times per second and data are averaged over 2006). Absolute values of SJL were used for all ana-
1-min intervals. Participants wore the armband for lyses (hereafter referred to as SJL). Information on
6–10 days (minimum of three weekdays and one work schedule was not available, and it was assumed
weekend day, 81% had ≥10 days) and kept a log of that free days occurred only on weekends.
their activities during non-wear periods. Observations were excluded from statistical
Participants wore armbands with different serial analyses if armband data had any of the following
numbers during different assessment periods. characteristics: missing bed- or wake-time (n = 40
Prior to calculating summary measures, infor- observations), <4 days of data in a given assess-
mation on sleep and physical activity from ment period (n = 51 observations), missing sleep
a personal log was used to estimate activity during data on the weekend (n = 38 observations),
non-wear periods (Hand et al. 2013). Activities extreme or implausible TST values (<4 h either
reported on the log were assigned Metabolic on a weekday, free day or on average) (n = 71
Equivalent of Task (MET) values from the 2011 observations), TST >11 h on a work day or on
Compendium of Physical Activities (Ainsworth average (n = 1 observation) or mid-sleep on a
et al. 2011) and used with measured resting meta- free day occurring in the afternoon (n = 2 obser-
bolic rate to calculate energy expenditure. vations) (Levandovski et al. 2011; Roenneberg
Analogously, periods of sleep or lying down for et al. 2012) (Note: ‘observation’ refers to one 6–10-
the non-wear periods were assigned either a 0 or day period of armband use for a single participant;
a 1 in the minute-by-minute data. A compliant day Fischer et al. 2017.) Due to a lack of information
was defined as having verifiable wear time (actual about work schedules, an average TST <4 h on
wear plus data from the log) for at least 80% of the a week day, weekend or total for a given week,
24-h day, and observations with wear time <80% and/or mid-sleep occurring in the afternoon on
were excluded. a week day or weekend, were used as criteria for
Minute-by-minute armband data were used to excluding those potentially involved in shift work.
calculate sleep characteristics. The following aver- Participants also were excluded if they regularly
age night-time sleep measures for work- and free used prescription or over-the-counter sleep-
days were calculated for every 6–10-day period of promoting medications (≥3 times per week,
armband wear: sleep onset and wake-up times, n = 22), self-reported working nights (n = 2) or
total sleep time (TST), sleep onset latency (SOL), traveled across time zones during periods of arm-
wake after sleep onset (WASO) and SE (Wirth band use (n = 1) (Levandovski et al. 2011;
et al. 2015). McMahon et al. 2018; Roenneberg et al. 2012).
The sleep onset time was defined as the first of 3 After exclusions, only five observations (0.35% of
minutes asleep that coincided with ≥10 minutes all observations) were based on six consecutive days
lying down. SOL was defined as the time between of armband wear, which is below a reliability
lying down and sleep onset. Morning wake-up threshold of ≥7 days for actigraphy-derived sleep
time was defined by the first out of 90 consecutive and circadian measures (van Someren 2007); the
minutes awake. TST was defined as the sum of all remainder wore the armbands for 7 days (N = 26,
minutes asleep from initiation of the sleep period 1.8%); 8 days (N = 57, 4.0%); 9 days (N = 191,
until wake-up. WASO was calculated as the sum of 13.3%) or ≥10 days (N = 1157, 80.6%).
wake periods of at least 2 minutes duration
between sleep onset and the final wake-up time.
Anthropometric outcomes and blood pressure
SE was calculated as a proportion of TST relative
to total time in bed. Height and weight were measured using tradi-
Chronotype was calculated as the time of mid-sleep tional stadiometer and electronic scales with
on free days (i.e. weekends) corrected for “make-up a precision of 0.1 cm and 0.1 kg, respectively
sleep” on free days (weekends) (Roenneberg et al. (Hand et al. 2013). The average of three measure-
2012). SJL was defined as the difference between ments was used to calculate an individual’s BMI
unadjusted midpoints of sleep on a weekend and (weight [kg]/height [m]2).
CHRONOBIOLOGY INTERNATIONAL 5

Waist and hip circumferences were measured least two interviews) were conducted by trained,
with a calibrated, spring-loaded tape measure. registered dietitians, two interviews for weekday
Waist circumference was measured at midpoint and one for weekend day consumption (Hand
between costal margin and iliac crest on the axillary et al. 2013). The 24HR is the method with the lowest
line on both sides of the trunk and 2 cm above the overall variance and, therefore, total error (Hebert
umbilicus (Hand et al. 2013). Hip circumference was et al. 1998) and 3 days is the optimal number of days
measured at the level of the greater trochanter, at the needed to compute total caloric intake (Ma et al.
widest point. The average of three measurements, 2009). The Nutrient Data System for Research (ver-
each rounded to 0.1 cm, was recorded and used to sion 2012: Nutrition Coordinating Center,
calculate waist-to-hip ratio (WHR) and waist-to- University of Minnesota, Minneapolis, Minnesota)
height ratio (WHtR). BF% was estimated based on was used to estimate nutrient and individual food
body composition measurements with dual x-ray intakes from the dietary recalls. Forty-three food
absorptiometry full-body Lunar fan-beam scanner parameters (including whole foods and nutrients)
(GE Healthcare model 8743, Waukesha, WI). were used to calculate the dietary inflammatory
Resting blood pressure was measured on the arm index (DII)™ score, which expresses the inflamma-
of a subject’s preference using Baumanometer® with tory potential of each individual’s diet (Shivappa
Calibrated® V-Lok® Complete Inflation System (W. et al. 2014). Lower DII scores are more anti-
A. Baum Co., Inc., NY) and a Littmann Classic II S. inflammatory while the higher scores are more
E. stethoscope (Hand et al. 2013). All readings were proinflammatory (maximum theoretical range is
taken and reported to the nearest even digit, −8.87 to 7.98). To account for individual differences
according to the top of the meniscus. Typically, in energy intake, the DII scores were calculated per
two measurements were taken with the interval of 1000 kcal (4184 kJ) (E-DII).
1 min between them. If the measurements differed
by more than 10 mmHg for SBP or more than
Statistical analyses
5 mmHg for DBP, the third measurement was
taken. Average value of two measurements was All statistical analyses were performed using SAS®
used for each time point in the statistical analyses. 9.4 software (Cary, NC). Continuous actigraphy
When three measurements were taken, two values variables were categorized using previously estab-
with the difference equal to or less than 10 mmHg lished cut-points: TST (<6, 6 to <7, ≥7 h) (Gottlieb
for SBP or 5 mmHg for DBP were averaged. If all et al. 2006), SOL (≥12 vs. <12 min) (Natale et al.
three measurements exceeded the limit of differ- 2009), WASO (≥60 vs. <60 min) (Kim 2015) and
ence between them, the summary value was set to SE (<85 vs. ≥85%) (Kim 2015).
missing. All available anthropometric and blood Relationships between SJL, chronotype, sleep
pressure measures (1–5 time points) were used in measures (TST, SOL, WASO, SE), obesity mea-
the analyses. sures and elevated blood pressure were analyzed
using generalized linear mixed models for repeated
measures (GLMMs) with an unstructured covar-
Covariates
iance matrix, binary distribution and logit link
Estimates of total daily physical activity were function (Figure 1, paths 1–3). Separate models
obtained from armband data. Information for the were fit for each predictor (SJL, chronotype, TST,
periods of non-wear was supplemented with values SOL, WASO, SE). BMI was categorized into two
for matching activity from the 2011 Compendium groups: underweight/normal weight (<25 kg/m2)
of Physical Activity (Ainsworth et al. 2011). Total and overweight/obese (≥25 kg/m2). Sex-specific
daily hours of physical activity were defined as cut-points were used to categorize BF% and
a sum of all activities of at least 3 MET. WHR. Obesity was defined as BF% ≥25% in men
Estimates of dietary intake were derived from and ≥30% in women (Shah and Braverman 2012).
random 24-h dietary recall interviews (24HR), col- WHR was dichotomized: <0.95 vs. ≥0.95 in men
lected on non-consecutive days by telephone inter- and <0.80 vs. ≥0.80 in women (Lean et al. 1995).
view. Up to three interviews (>97% completed at WHtR was dichotomized as <0.5 vs. ≥0.5
6 D. M. MCMAHON ET AL.

(Browning et al. 2010). SBP was dichotomized were not used to adjust group trajectories.
using cut-points for prehypertension ≥120 vs. Participants with complete armband data for all
<120 mmHg, and DBP ≥80 vs.<80 mmHg since available time points (1–5) were included in these
only a few participants met clinical criteria for analyses (N = 390, 1431 observations). GLMMs
hypertension (SBP ≥ 140 mmHg, 11%; DBP ≥ were estimated to compare means of SJL, sleep or
90 mmHg, 4%) (Chobanian et al. 2003). chronotype among latent chronotype groups using
Potential covariates included age, sex, race the F-tests in PROC GLIMMIX and the intermedi-
(European American [EA], African American ate chronotype group as the referent (Figure 1,
[AA], Other [Hispanic, Asian, Native American paths 4–5). The p value was penalized for multiple
and mixed race]), education, annual income, comparisons using the Bonferroni method.
employment, marital status, having children (yes/ Because chronotype may influence anthropo-
no), physical activity (hours/day), caffeine intake metric measures and blood pressure in several
(g/day), napping (yes [>0 min/day]/no [0 min/ ways, additional analyses were conducted using
day]), season (winter [November–January], spring chronotype as either a continuous or categorical
[February–April], summer [May–July], autumn (i.e. latent group identified using RMLCA) vari-
[August–October]) and current dieting (yes/no). able. When examining relationships between SJL,
The assessment sequence number (1–5 at sleep measures, obesity measures and blood pres-
6-month intervals) was treated as a categorical sure, chronotype (continuous) was considered as
variable and included in crude and adjusted statis- a potential confounder along with other covariates
tical models. To select potential covariates, bivari- (Figure 1, paths 1, 3). It was not included in the
ate relationships between each sleep measure and final models because tests for confounding (10%
the anthropometric outcomes were presented as rule) were negative. The potential role of chrono-
odds ratios (OR) and 95% confidence intervals type as an effect modifier was tested using two
(95% CI), and variables with a p value <0.20 approaches (Figure 1). In the first approach, an
were selected for further evaluation in multivari- interaction term between chronotype and a main
able statistical models. Final variable selection was independent variable (e.g. each sleep measure) was
performed by identifying covariates that changed added to the multivariable statistical model, and its
an estimate of the main exposure variable by significance was assessed using the Wald test. The
>10%. The final model included only covariates continuous or categorical version of chronotype
that were identified using the 10% rule or if they that was used for the interaction term was chosen
were statistically significant (p ≤ 0.05). to match the form of the other independent vari-
Repeated measures latent class analysis able. The second approach examined the relation-
(RMLCA) was performed to identify latent groups ship between absolute SJL or sleep characteristics
for chronotype (morning, intermediate and eve- and each outcome after stratification by latent
ning) using PROC TRAJ in SAS® (Jones et al. chronotype group (Figure 1). Neither chronotype
2001; McMahon et al. 2018). This analysis assumes nor SJL was directly associated with obesity mea-
a mixture model to define the trajectories of sures or elevated blood pressure; thus, a potential
unique subgroups within a population that does mediating role of SJL in these relationships was
not change group membership over time (Jones not evaluated.
and Nagin 2007; Nagin 2005). The best-fitting
model was selected using the Bayesian
Results
Information Criterion (Jones and Nagin 2007)
and a priori minimum group sizes containing The final analytical data set consisted of 390 parti-
≥10% of the study population (McMahon et al. cipants with a total of 1431 repeated observations.
2018). RMLCAs were adjusted for race (EA vs. Complete data were available at other time points
AA/other) and sex, which were assumed not to as follows: 6 months (n = 341), 12 months
change over time. Because the group trajectories (n = 317), 18 months (n = 206) and 24 months
for all three latent chronotype groups were linear, (n = 177). Note that the 18- and 24-month mea-
other potential covariates that change over time surements were beyond the initial design of this
CHRONOBIOLOGY INTERNATIONAL 7

study. Subjects measured at those time points (BMI≥30 kg/m2, 14%) at baseline (Table 2).
agreed to additional involvement after the study Among males, 37% had BF% ≥25% and 3% had
was extended. A majority of the participants had at high WHR (≥0.95). Among females, 71% had BF%
least three repeated assessments (79%), including ≥30% and 15% had high WHR (≥0.80). Twenty-
those who had four (15%) and five (39%). Only two percent of participants of both sexes had high
10% of participants had two, and 11% had only WHtR (≥0.50).
one assessment. The average age at baseline was SBP was elevated (≥120 mmHg) among 59% of
28 ± 4 years, and the distribution of sex was participants (Table 2), including those who met
approximately equal (51% women, Table 1). Most a criterion for hypertension (≥140 mmHg, 11%,
participants were EA (68%) and had at least not shown). Only 23% had elevated DBP
4 years of college education (84%). Fifty-five per- (≥80 mmHg) (Table 2), including those with hyper-
cent of participants were employed, and 45% were tension (≥90 mmHg, 4%, not shown) at baseline. By
college students (19% undergraduate, 81% gradu- the end of 2-year follow-up, there were negligible to
ate). A majority of participants (71%) had an moderate increases in the proportion of obese par-
annual income <$60,000. Thirty-two percent of ticipants (Table 2). Alternatively, the proportion of
participants were married, and 14% had children. those with elevated SBP decreased after 2 years of
Forty-five percent of participants were over- follow-up (Table 2).
weight (BMI 25–29.9 kg/m2, 31%) or obese At baseline, chronotype was normally distribu-
ted and moderately correlated with SJL (Pearson
r = 0.46, p < 0.001; Spearman r = 0.43, p < 0.001).
Table 1. Characteristics of the participants at baseline, Energy When analyzed as a continuous variable, chrono-
Balance Study, Columbia, SC, 2011–2014. type was not associated with any of the outcomes
All participants Women Men (Table 3). The RMLCA identified three latent
Variable (n = 390) (n = 198) (n = 192)
Age (years, mean±SD) 27.6 ± 3.8 27.7 ± 3.7 27.4 ± 3.9
chronotype groups: morning (mean ± SE:
Race, n (%) 3.0 ± 0.04 h, 32.5% of participants), intermediate
European American 264 (68) 131 (66) 133 (70) (4.4 ± 0.03 h, 53%) and evening (6.1 ± 0.1 h,
African American 47 (12) 31 (16) 16 (8)
Hispanic/Latino 11 (3) 7 (3) 4 (2) 14.5%) (McMahon et al. 2018). Based on the base-
Asian 42 (11) 15 (8) 27 (14) line data, the cut-points between chronotype
Native American 12 (3) 8 (4) 4 (2)
Mixed race 14 (3) 6 (3) 8 (4)
groups were approximately at 2:40 a.m. and 4:20
Education, n (%) a.m. Chronotype defined by latent group member-
HS graduate/GED 62 (16) 18 (9) 44 (23) ship was not associated with any of the outcomes
Some college
College (4+ years) 328 (84) 180 (91) 148 (77) (Table 3).
Income ($), n (%) Each sleep measure except SOL was associated
<20 000 65 (17) 34 (17) 31 (16)
20 000 to <40 000 137 (35) 76 (39) 61 (32)
with chronotype (Table 4). Participants with an
40 000 to <60 000 74 (19) 37 (19) 37 (19) evening chronotype had shorter TST, lower SE
60 000 to <80 000 50 (13) 23 (12) 27 (14) and larger WASO than those with an intermediate
≥80 000 62 (16) 26 (13) 36 (19)
Employment, n (%) chronotype.
Employed and self- 214 (55) 113 (57) 101 (53) Approximately 50% of participants had more
employed
Student/othera 176 (45) 85 (43) 91 (47)
than 1 h of SJL at baseline (>1 to <2 h: 33%; ≥2 h:
Marital status, n 17%). Morning types had less SJL (0.9 ± 0.04,
Marriedb 178 (46) 90 (45) 88 (46) p< 0.001), whereas those with an evening chrono-
Singlec 212 (54) 108 (55) 104 (54)
Children, age type had more SJL (1.5 ± 0.1 h, p < 0.001) compared
<18 years, n (%) to intermediate chronotypes (1.2 ± 0.03 h). SJL was
0 336 (86) 173 (88) 163 (85) not associated with any of the outcome measures
1 29 (7) 13 (7) 16 (8)
≥2 24 (7) 11 (5) 13 (7) (Table 5).
Source: McMahon et al. (2018), doi: 10.1080/07420528.2017.1405014. Among the sleep measures, SOL ≥12 min was
a
Out of work (<1 year), homemaker, unable to work (n = 4, 1%). associated with an increased odds of elevated WHtR
b
Includes members of unmarried couples (n = 55, 14%).
c
Includes divorced and separated (n = 7, 1.7%). (OR: 1.37, 95% CI: 1.03–1.84, Table 6). Stratification
SD: standard deviation. by latent chronotype group indicated that morning
8 D. M. MCMAHON ET AL.

Table 2. Descriptive outcomes at baseline and after 2 years of follow-up, Energy Balance Study, Columbia, SC, 2011–2014.
Baseline 24 months
All participants Women Men All participants Women Men
(n = 390) (n = 198) (n = 192) (n = 177) (n = 82) (n = 95)
Outcome
measure Category n (%) n (%) n (%) n (%) n (%) n (%)
Body mass index Overweight/obese 175 (45) 84 (42) 91 (47) 90 (51) 39 (48) 51 (54)
(kg/m2) Normal 215 (55) 114 (58) 101 (53) 87 (49) 43 (52) 44 (46)
Percent body fat
Obese 210 (54) 140 (71) 70 (37) 110 (62) 65 (79) 45 (47)
Non-obese 179 (46) 58 (29) 121 (63) 67 (38) 17 (21) 50 (53)
Waist-to-hip ratio
Obese 34 (9) 29 (15) 5 (3) 20 (11) 17 (21) 3 (3)
Non-obese 356 (91) 169 (85) 187 (97) 157 (89) 65 (79) 92 (97)
Waist-to-height Ratio
Obese 87 (22) 47 (24) 40 (21) 48 (27) 22 (27) 26 (27)
Non-obese 303 (78) 151 (76) 152 (79) 129 (73) 60 (73) 69 (73)
Systolic blood pressure (mmHg)
Elevated 232 (59) 93 (47) 139 (72) 58 (33) 12 (15) 46 (48)
Normal 158 (41) 105 (53) 53 (28) 119 (67) 70 (85) 49 (52)
Diastolic blood pressure (mmHg)
Elevated 91 (23) 23 (12) 68 (35) 47 (27) 11 (13) 36 (38)
Normal 299 (77) 175 (88) 124 (65) 130 (73) 71 (87) 59 (62)
Cut-points for outcomes: body mass index (≥25 vs. <25 kg/m2), percent body fat (males ≥0.25 vs. <0.25, females ≥30 vs. <30), waist-to-hip ratio
(males ≥0.95 vs. <0.95, females ≥0.8 vs. <0.8), waist-to-height ratio (≥0.5 vs. <0.5), systolic blood pressure (≥120 vs. <120 mmHg) and diastolic
blood pressure (≥80 vs. <80 mmHg).

Table 3. Relationship between chronotype and outcome measures, Energy Balance Study, Columbia, SC, 2011–2014.
Chronotype Chronotype group
(continuous, hours) (referent: intermediate)
Outcome Morning Evening
measure Model OR (95% CI) p Value OR (95% CI) OR (95% CI) p value
Body mass index Crude 0.91 (0.79–1.06) 0.24 1.56 (0.81–2.98) 0.95 (0.39–2.28) 0.36
(kg/m2) Adjusteda 0.86 (0.76–1.03) 0.12 1.49 (0.80–2.78)c 1.22 (0.50–2.97)c 0.45
Percent body fat Crude 1.01 (0.86–1.17) 0.94 0.67 (0.35–1.30) 0.45 (0.18–1.12) 0.17
Adjusteda 1.11 (0.94–1.31) 0.22 0.77 (0.40–1.50)c 1.41 (0.54–3.69)d 0.48
Waist-to-hip ratio Crude 0.83 (0.68–1.01) 0.07 1.24 (0.64–2.44) 0.58 (0.20–1.66) 0.38
Adjustedb 0.86 (0.69–1.07) 0.17 1.18 (0.56–2.48)e 1.66 (0.46–6.06)e 0.71
Waist-to-height ratio Crude 0.96 (0.82–1.13) 0.64 1.30 (0.67–2.52) 0.70 (0.30–1.98) 0.55
Adjustedb 0.98 (0.83–1.17) 0.86 1.38 (0.70–2.73)b 1.36 (0.50–3.72)b 0.60
Systolic blood Crude 1.00 (0.90–1.11) 0.90 1.21 (0.81–1.81) 1.81 (1.04–3.13) 0.10
pressure (mmHg) Adjusteda 0.96 (0.87–1.07) 0.47 1.16 (0.79–1.72)a 1.10 (0.63–1.90)a 0.74
Diastolic blood Crude 1.07 (0.96–1.20) 0.22 0.95 (0.62–1.44) 1.83 (1.06–3.16) 0.07
pressure (mmHg) Adjusteda 1.04 (0.93–1.16) 0.49 0.89 (0.59–1.34)a 1.19 (0.68–2.07)a 0.61
Cut-points for outcomes: body mass index (≥25 vs. <25 kg/m2), percent body fat (males ≥0.25 vs. <0.25, females ≥30 vs. <30), waist-to-hip ratio
(males ≥0.95 vs. <0.95, females ≥0.8 vs. <0.8), waist-to-height ratio (≥0.5 vs. <0.5), systolic blood pressure (≥120 vs. <120 mmHg) and diastolic
blood pressure (≥80 vs.<80 mmHg).
a
Adjusted for time, sex and physical activity.
b
Adjusted for time, age, sex and physical activity.
c
Adjusted for time, sex, physical activity and income.
d
Adjusted for time, sex, race and physical activity.
e
Adjusted for time, age, sex, race, physical activity and employment status.
OR: odds ratio; 95% CI: 95% confidence interval.

types who also had short TST (<6 h), elevated WASO have elevated SBP (95% CI: 1.15–3.16,
(≥60 min) or low SE (<85%) were more likely to have pinteraction = 0.02, Table 7). Elevated WASO
increased BF%, WHR or WHtR (Table 7). (≥60 min) was associated with increased odds of
Participants with both morning chronotype and elevated DBP among those with an intermediate
sleep latency ≥12 min were 1.90 times more likely to chronotype (OR = 1.68, 95% CI: 1.05–2.71,
CHRONOBIOLOGY INTERNATIONAL 9

Table 4. Relationship between sleep characteristics and latent chronotype group (n = 390, 1431 observations), Energy
Balance Study, Columbia, SC, 2011–2014.
Morning Intermediate Evening
(32.5%) (53%) (14.5%) F-test,
Variable Mean±SE Mean±SE Mean±SE p value†
Chronotype (h)a 3.0 ± 0.05* 4.4 ± 0.04 6.1 ± 0.08* <0.01
Total sleep time (h)b 6.5 ± 0.13 6.5 ± 0.13 6.1 ± 0.16* 0.01
Total sleep time, work days (h)c 6.2 ± 0.14 6.3 ± 0.13 6.0 ± 0.17* 0.04
Sleep onset latency (min)c 11.1 ± 0.04 11.6 ± 0.03 10.6 ± 0.10 0.35
Sleep efficiency (%)d 81.4 ± 0.50 81.3 ± 0.50 78.8 ± 0.80* 0.02
WASO (min)e 49.6 ± 2.41 52.3 ± 2.36 62.9 ± 4.20* 0.01
Absolute SJL (h)f 1.0 ± 0.10* 1.3 ± 0.09 1.6 ± 0.11* <0.01
*p <0.025 vs. intermediate chronotype, with Bonferroni adjustment for multiple comparisons.
†Partial F-test in generalized linear mixed models with repeated measures and unstructured covariance matrix.
a
Adjusted for time.
b
Adjusted for time, sex, race, having children, employment status, physical activity and caffeine intake.
c
Adjusted for time, sex, race, employment status and physical activity.
d
Adjusted for time, sex, race, employment status and caffeine intake.
e
Adjusted for time, sex and having children.
f
Adjusted for time and employment status.
SE: standard error of the mean.

Table 5. Relationship between absolute social jetlag (continu-


ous, hours) and outcome measures, Energy Balance Study, Discussion
Columbia, SC, 2011–2014.
To the authors’ knowledge, this is the first study to use
Outcome F-test,
measure OR (95% CI) p value† repeated actigraphy measures to examine relation-
Body mass indexa (kg/m2) 0.93 (0.73–1.16) 0.71 ships between chronotype, SJL, sleep and measures
Percent body fata 0.90 (0.70–1.15) 0.41 of obesity and elevated blood pressure. The results
Waist-to-hip ratioa 0.81 (0.59–1.12) 0.20
Waist-to-height ratiob 1.00 (0.77–1.29) 0.98 suggest that disrupted sleep was associated with mod-
Systolic blood pressurea (mmHg) 1.12 (0.95–1.31) 0.19 est increases in total and abdominal obesity, as well as
Diastolic blood pressurea (mmHg) 1.02 (0.86–1.21) 0.83 with elevated blood pressure, particularly among
Cut-points for outcomes: body mass index (≥25 vs. <25 kg/m2), percent
body fat (males ≥0.25 vs. <0.25, females ≥30 vs. <30), waist-to-hip
morning types. These observations are consistent
ratio (males ≥0.95 vs. <0.95, females ≥0.8 vs. <0.8), waist-to-height with some previous studies that found an association
ratio (≥0.5 vs. <0.5), systolic BP (≥120 vs. <120 mmHg) and diastolic between sleep disruption and weight gain or obesity in
BP (≥80 vs.<80 mmHg).
†Partial F-test in generalized linear mixed models with repeated mea- young adults (Bailey et al. 2014; Hasler et al. 2004;
a
sures and an unstructured covariance matrix. Lauderdale et al. 2009; Theorell-Haglow et al. 2012;
Adjusted for time, sex and physical activity.
b
Adjusted for time, age, sex and physical activity.
Vgontzas et al. 2014). However, none of the men-
OR: odds ratio; CI: confidence interval. tioned studies considered individual chronotype.
One relatively novel aspect of this study was the
inclusion of both the independent effects of chron-
pinteraction = 0.10, Table 7). No statistically significant otype, SJL and sleep, as well as their combined
association was detected between evening chrono- effects. Most previous studies investigated only the
type and any sleep measure (Table 7). separate effects of chronotype or sleep on obesity
Results obtained when the data were ana- (Bailey et al. 2014; Culnan et al. 2013; Hasler et al.
lyzed using sleep measures as continuous vari- 2004; Lauderdale et al. 2009; Theorell-Haglow et al.
ables were generally consistent with those 2012; Vgontzas et al. 2014; Wang 2014), whereas
obtained using categorical sleep variables. For a minority of studies adjusted for the effects of
example, among morning types, the odds of chronotype and sleep duration while examining the
a high WHR increased per 1% decrease in SE relationship between absolute SJL and BMI, percent
(OR = 1.11; 95% CI: 1.03–1.20, pinteraction fat mass, waist circumference or obesity (Parsons
= 0.01, data not shown). et al. 2015; Roenneberg et al. 2012). Some previous
10 D. M. MCMAHON ET AL.

Table 6. Relationship between sleep characteristics and outcome measures, Energy Balance Study, Columbia, SC, 2011–2014.
Total sleep time
(referent: 6 to <7 h) Sleep efficiency Sleep onset latency Wake after sleep onset
Outcome
measure Model OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)
Body mass index <6 h ≥7 h <85 vs. ≥85% ≥12 vs. <12 min ≥60 vs. <60 min
(kg/m2) Crude 1.28 (0.77–2.10) 1.01 (0.65–1.57) 1.03 (0.68–1.54) 1.06 (0.72–1.55) 1.04 (0.67–1.60)
Adjusted 1.30 (0.77–2.19)a 0.87 (0.55–1.37)a 0.96 (0.64–1.48)a 0.99 (0.67–1.45)b 0.97 (0.63–1.51)b
Continuous Continuous† Continuous Continuous
Crude 0.88 (0.67–1.14) 1.00 (0.97–1.04) 1.01 (0.98–1.05) 1.00 (0.99–1.01)
Adjusted 0.81 (0.62–1.06)a 1.00 (0.96–1.03)a 1.00 (0.97–1.04)b 1.00 (0.99–1.01)a
<6 h ≥7 h <85 vs. ≥85% ≥12 vs. <12 min ≥60 vs. <60 min
Crude 0.88 (0.53–1.49) 1.02 (0.64–1.64) 0.98 (0.65–1.48) 1.41 (0.94–2.10) 1.35 (0.86–2.14)
Percent body fat Adjusted 1.00 (0.57–1.77)c 0.75 (0.45–1.24)c 0.95 (0.60–1.51)d 1.41 (0.91–2.16)c 1.31 (0.80–2.14)c
Continuous Continuous† Continuous Continuous
Crude 1.07 (0.81–1.41) 1.03 (0.99–1.07) 1.04 (1.00–1.07) 1.01 (1.00–1.02)
Adjusted 0.83 (0.62–1.12)c 1.04 (1.00–1.08)e 1.03 (1.00–1.07)c 1.01 (1.00–1.02)c
<6 h ≥7 h <85 vs. ≥85% ≥12 vs. <12 min ≥60 vs. <60 min
Crude 1.16 (0.62–2.18) 1.17 (0.66–2.09) 1.22 (0.72–2.08) 1.25 (0.76–2.06) 1.31 (0.76–2.24)
Waist-to-hip ratio Adjusted 1.34 (0.68–2.63)e 0.98 (0.53–1.80)e 1.20 (0.68–2.12)e 1.23 (0.73–2.08)e 1.23 (0.69–2.18)c
Continuous Continuous† Continuous Continuous
Crude 1.09 (0.79–1.52) 1.02 (0.98–1.06) 1.00 (0.96–1.04) 1.00 (1.00–1.01)
Adjusted 0.93 (0.63–1.31)e 1.02 (0.98–1.06)e 0.97 (0.94–1.03)e 1.00 (0.99–1.01)c
<6 h ≥7 h <85 vs. ≥85% ≥12 vs. <12 min ≥60 vs. <60 min
Crude 1.16 (0.62–2.18) 1.17 (0.66–2.09) 1.14 (0.89–1.47) 1.29 (1.01–1.65) 1.15 (0.71–1.87)
Waist-to-height Adjusted 1.58 (0.88–2.83)e 0.76 (0.45–1.30)e 1.16 (0.86–1.55)e 1.37 (1.03–1.84)f 1.30 (0.96–1.75)e
ratio Continuous Continuous† Continuous Continuous
Crude 0.83 (0.62–1.12) 1.02 (0.98–1.06) 1.03 (1.01–1.04) 1.00 (1.00–1.01)
Adjusted 0.76 (0.56–1.04)e 1.02 (0.99–1.06)f 1.02 (1.00–1.04)f 1.00 (1.00–1.01)e
<6 h ≥7 h <85 vs. ≥85% ≥12 vs. <12 min ≥60 vs. <60 min
Systolic blood Crude 1.36 (0.95–1.96) 0.90 (0.64–1.25) 0.98 (0.96–1.01) 1.13 (0.85–1.50) 1.32 (0.96–1.82)
pressure (mmHg) Adjusted 1.26 (0.87–1.81)c 0.95 (0.68–1.33)c 1.15 (0.85–1.55)c 1.05 (0.79–1.40)c 1.31 (0.95–1.79)e
Continuous Continuous† Continuous Continuous
Crude 0.80 (0.66–0.97) 1.03 (1.00–1.05) 1.02 (0.99–1.04) 1.00 (1.00–1.01)
Adjusted 0.86 (0.71–1.04)c 1.02 (0.99–1.04)c 1.01 (0.98–1.03)c 1.00 (1.00–1.01)e
<6 h ≥7 h <85 vs. ≥85% ≥12 vs. <12 min ≥60 vs. <60 min
Crude 1.26 (0.85–1.86) 1.18 (0.83–1.70) 1.12 (0.81–1.54) 1.16 (0.86–1.58) 1.24 (0.89–1.73)
Diastolic blood Adjusted 1.18 (0.80–1.75)c 1.28 (0.89–1.84)c 1.04 (0.75–1.43)c 1.08 (0.79–1.46)c 1.23 (0.88–1.73)e
pressure (mmHg) Continuous Continuous† Continuous Continuous
Crude 0.86 (0.70–1.05) 1.03 (1.00–1.05) 1.02 (0.99–1.04) 1.00 (1.00–1.01)
Adjusted 0.92 (0.75–1.12)c 1.02 (0.99–1.05)c 1.01 (0.98–1.03)c 1.00 (1.00–1.01)e
Cut-points for outcomes: body mass index (≥25 vs. <25 kg/m2), percent body fat (males ≥25 vs. <25%, females ≥30 vs. <30%), waist-to-hip ratio
(males ≥0.95 vs. <0.95, females ≥0.8 vs. <0.8), waist-to-height ratio (≥0.5 vs. <0.5), systolic blood pressure (≥120 vs. <120 mmHg) and diastolic
blood pressure (≥80 vs. <80 mmHg).
† - OR per 1 unit decrease in sleep efficiency.
a
Adjusted for time, sex, income and physical activity.
b
Adjusted for time, sex, employment status and physical activity.
c
Adjusted for time, sex and physical activity.
d
Adjusted for time, sex, physical activity and sleep latency.
e
Adjusted for time, age, sex and physical activity.
f
Adjusted for time, age, sex, race and physical activity.

studies have reported associations between anthro- confounders correlated with a main predictor, may
pometric indicators of obesity and either sleep, SJL lead to overadjustment bias (Breslow 1982). Thus, in
or chronotype, although inconsistencies in both the the present analysis, relationships between chrono-
analytical approaches and the results are apparent type, SJL or sleep measures and the outcomes were
(Parsons et al. 2015; Roenneberg et al. 2012). It has first examined independently. Chronotype was asso-
been suggested that SJL may partially mediate the ciated with sleep measures, but not with obesity or
relationships between chronotype and BMI (Culnan elevated blood pressure; thus, it did not meet criteria
and Kloss 2013). Adjustment for an intermediate for confounding. Instead, relationships between
variable on a causal path, or adjustment for sleep and outcomes were modified by chronotype.
CHRONOBIOLOGY INTERNATIONAL 11

Table 7. Relationships between sleep characteristics and outcomes measures stratified by latent chronotype group, Energy Balance
Study, Columbia, SC, 2011–2014.
Total sleep time
(referent: 6 to <7 h) Sleep efficiency Sleep onset latency Wake after sleep onset
<6 h ≥7 h <85% vs. ≥85% ≥12 vs. <12 min ≥60 vs. <60 min
Outcome Chronotype
measure group OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)
Body mass index Morning 0.98 (0.34–2.82)a 1.01 (0.42–2.44)b 1.12 (0.52–2.43)b 1.17 (0.55–2.51)c 0.88 (0.37–2.10)a
(kg/m2) Intermediate 1.63 (0.79–3.37)a 0.81 (0.45–1.47)a 0.99 (0.57–1.71)b 1.04 (0.62–1.75)c 1.26 (0.69–2.29)a
Evening 1.31 (0.33–5.27)a 1.77 (0.28–11.18)a 0.60 (0.17–2.11)b 1.26 (0.43–3.74)c 1.38 (0.39–4.86)a
pinteraction† 0.99 0.69 0.97 0.32
Percent body fat Morning 0.88 (0.30–2.57)b 0.69 (0.27–1.73)b 1.22 (0.53–2.84)d 1.73 (0.81–3.72)b 3.43 (1.35–8.72)b
Intermediate 0.83 (0.37–1.86)b 0.79 (0.40–1.56)b 0.80 (0.43–1.50)d 1.33 (0.73–2.41)b 0.71 (0.37–1.38)b
Evening 1.23 (0.28–5.48)b 0.59 (0.09–3.79)b 0.68 (0.14–3.30)d 1.18 (0.29–4.82)b 1.38 (0.30–6.26)b
pinteraction 0.92 0.64 0.84 0.03
Waist-to-hip ratio Morning 3.91 (1.19–12.84)e 1.39 (0.42–4.54)e 3.38 (1.07–10.69)e 2.13 (0.83–5.47)b 3.72 (1.38–10.07)b
Intermediate 0.66 (0.25–1.78)e 0.82 (0.37–1.80)e 0.95 (0.46–1.95)e 1.06 (0.54–2.10)b 0.67 (0.30–1.51)b
Evening 2.01 (0.19–21.28) 1.36 (0.16–11.59)e
e
0.20 (0.03–1.30)e 0.19 (0.03–1.20)b 0.42 (0.06–2.90)b
pinteraction 0.15 0.03 0.27 0.01
Waist-to-height Morning 1.59 (0.47–5.39)e 0.60 (0.21–1.72)e 1.89 (1.10–3.25)e 1.63 (0.94–2.84)f 2.21 (1.27–3.84)e
ratio Intermediate 1.76 (0.79–3.84)e 0.85 (0.42–1.70)e 0.87 (0.58–1.30)e 1.44 (0.96–2.18)f 0.81 (0.53–1.24)e
e
Evening 1.05 (0.22–5.09) 1.65 (0.26–10.31)e 0.95 (0.39–2.32)e 0.87 (0.37–2.09)f 1.28 (0.55–2.94)e
pinteraction 0.74 0.05 0.13 0.01
Systolic blood Morning 0.70 (0.36–1.38)e 0.65 (0.37–1.16)e 1.05 (0.63–1.75)b 1.90 (1.15–3.16)b 1.19 (0.67–2.09)e
pressure (mmHg) Intermediate 1.45 (0.85–2.46)e 1.09 (0.68–1.75)e 1.24 (0.81–1.89)b 0.90 (0.59–1.35)b 1.38 (0.87–2.18)e
e
Evening 2.51 (0.99–6.38) 2.31 (0.74–7.27)e 0.96 (0.41–2.22)b 0.63 (0.29–1.35)b 1.21 (0.55–2.66)e
pinteraction 0.17 0.95 0.02 0.93
Diastolic blood Morning 0.83 (0.38–1.81)b 1.51 (0.79–2.90)b 0.67 (0.38–1.20)b 0.83 (0.47–1.45)b 0.94 (0.50–1.78)e
pressure (mmHg) Intermediate 1.44 (0.81–2.54)b 1.46 (0.88–2.40)b 1.41 (0.90–2.22)b 1.40 (0.90–2.16)b 1.68 (1.05–2.71)e
Evening 1.15 (0.49–2.72)b 0.63 (0.20–2.01)b 0.72 (0.30–1.74)b 0.79 (0.36–1.73)b 0.73 (0.32–1.64)e
pinteraction 0.50 0.13 0.33 0.10
Cut-points for outcomes: body mass index (≥25 vs. <25 kg/m2), percent body fat (males ≥0.25 vs. <0.25, females ≥30 vs. <30), waist-to-hip ratio
(males ≥0.95 vs. <0.95, females ≥0.8 vs. <0.8), waist-to-height ratio (≥0.5 vs. <0.5), systolic blood pressure (≥120 vs. <120 mmHg) and diastolic
blood pressure (≥80 vs.<80 mmHg).
†pinteraction: p value for the interaction term between a sleep variable and chronotype group in adjusted statistical model.
a
Adjusted for time, sex, income and physical activity.
b
Adjusted for time, sex, physical activity.
c
Adjusted for time, sex, employment status and physical activity.
d
Adjusted for time, sex, physical activity and sleep latency.
e
Adjusted for time, age, sex and physical activity.
f
Adjusted for time, age, sex, race and physical activity.
OR: odds ratio; CI: confidence interval.

Since SJL was not associated with obesity or elevated analysis. In prior cross-sectional studies, SJL was
blood pressure, the potential mediating role of SJL associated with elevated BMI, fat mass and waist
was not assessed. circumference (Roenneberg et al. 2012; Valdez et al.
A lack of association between chronotype and 1996; Wong et al. 2015). In another study, SJL
BMI in the current study is consistent with find- predicted increases in BMI only among those who
ings reported by others (Maukonen et al. 2017; were already overweight or obese (Roenneberg et al.
Parsons et al. 2015). Most other studies investigat- 2012).
ing relationships between circadian preference and The present study used quantitative actigraphic
obesity found that evening types were more likely armband data to characterize SJL, whereas others
to be obese compared to morning types (Culnan typically used self-report information to ascertain
et al. 2013; Wang 2014). However, in one large sleep/wake timing. The present study had fewer
cross-sectional study, morning circadian prefer- individuals with elevated absolute SJL relative to
ence was associated with increased BMI other studies (≤1 h 49% vs. 31–37%, >1 h 51% vs.
(Roenneberg et al. 2012). 63–69%, ≥2 h 16% vs. 26–30%, respectively)
Similarly, SJL was not associated with any (Roenneberg et al. 2012; Rutters et al. 2014).
anthropometric indices of obesity in the current Thus, it is possible that the 2-year study duration
12 D. M. MCMAHON ET AL.

was insufficient to detect changes in outcomes that observed among evening chronotypes despite pre-
may have been related to SJL or chronotype. vious evidence that poor sleep was more common
In this population of relatively young healthy among evening types compared to others
adults, the increase in the proportion of those with (McMahon et al. 2018; Merikanto et al. 2012;
obesity relative to baseline was modest (~10%– Patterson et al. 2018). The results are similar to
15% or less), while the proportion of participants those from a recent population-based study where
with elevated SBP decreased. Participants in the morning types who had short or long sleep dura-
present study gained an average of ~3 lbs over two tion had an increased odds of being overweight or
years, consistent with the Centers for Disease obese relative to morning types with normal sleep
Control and National Longitudinal Survey of (7–8 h) (Patterson et al. 2018). However, increased
Youth 1979 (1990–2008), which reported that odds of being overweight or obese also were
young adults gain, on average, 1–2 lbs each year observed among evening types with long sleep
(Fryar et al. 2012; Malhotra et al. 2013). durations (Patterson et al. 2018). Participants in
Discrepancies between this and other studies may that study were older (mean age = 56.5,
have been due to unexplained methodological dif- SD = 8.1 years) and the cut-point for long sleep
ferences, residual confounding, differences in the was ≥9 h vs. ≥7 h in the current study (Patterson
age distribution, or other population characteris- et al. 2018).
tics. For example, young adults tend to have an Increased vulnerability to the detrimental effects
increased tolerance to shift work (Saksvik et al. of sleep among morning types may be mediated by
2011). Reasons for the discrepancies between this physiological mechanisms controlling stress and
analysis and some other studies also may have fat deposition. It has been reported that morning
been due in part to the accuracy of the indicators. chronotypes had higher baseline and postprandial
For example, BMI likely underestimates obesity total cholesterol, lower low-density lipoprotein
among those with BMI <30 kg/m2, particularly and a higher adiponectin compared to intermedi-
among white and Hispanic women (Frankenfield ate or evening chronotypes (Merikanto et al.
et al. 2001; Rahman and Berenson 2010), while it 2013). In the same study, participants with a
also can overestimate obesity among some popula- morning chronotype also had higher baseline glu-
tions relative to percent body fat (Ode et al. 2007). cose levels and greater increases in postprandial
In a prior analysis of this population at baseline, blood glucose levels compared to other chrono-
measures of sleep disturbance (lower SE, higher types (Merikanto et al. 2013). Others have
WASO, longer TST) were associated with elevated observed that morning chronotypes had a higher
mean BMI (Wirth et al. 2015). The present analy- cortisol awakening response relative to evening
sis employed conventional BMI categories widely types (Bailey and Heitkemper 2001; Kudielka
used to diagnose obesity along with sex-specific et al. 2006; Randler and Schaal 2010).
cut-points for %BF and WHR, and the effects of Short or fragmented sleep may initiate
chronotype and SJL also were considered. The a neuroendocrine stress response leading to
repeated measures design used data from the increased activity of the sympathoadrenal and
entire 2-year period and provided greater statisti- hypothalamic-pituitary axes, which in turn can
cal power, although only modest increases in BF% lead to increased blood pressure, impaired glucose
or WHtR were observed among those with sleep metabolism and abdominal obesity (Meerlo et al.
disturbances (increased SOL, or decreased SE). 2008; Spiegel et al. 1999). People with disrupted or
When an individual’s chronotype was incorpo- short sleep had increased cortisol levels at night,
rated into the analysis, relationships between poor delayed cortisol response in the morning and
sleep and anthropometric measures were accentu- increased cortisol levels the following evening
ated among morning types; and those with dis- (Meerlo et al. 2008; Spiegel et al. 1999). Based on
rupted sleep had increased odds of total or these observations, it is plausible that people with
abdominal obesity. Morning types with a SOL a morning chronotype who also experience poor
≥12 min also had increased odds of elevated SBP. sleep may have augmented cortisol secretion with
No increases in obesity or blood pressure were its accompanying metabolic consequences (e.g.
CHRONOBIOLOGY INTERNATIONAL 13

greater fat deposition) relative to those with only (McMahon et al. 2018). To diminish the possible
one or the other of these conditions. influence of shift or weekend work on the study
Recent research shows that various lifestyle fac- results, participants with extreme or implausible
tors such as diet composition, meal timing and sleep schedules (TST values <4 h either on week
physical activity may differ by chronotype days, free days, or on average, or mid-sleep time in
(Maukonen et al. 2017; Mota et al. 2016; Munoz the afternoon) were excluded (Levandovski et al.
et al. 2017). Evening types were less likely to 2011; McMahon et al. 2018; Merikanto et al. 2012;
adhere to healthy diet, preferred energy-dense Roenneberg et al. 2012).
foods over fruit and consumed significantly higher There is a possibility that objectively measured
amounts of energy, sucrose and saturated fat later wake-up time could have led to some overestima-
in a day relative to morning types (Maukonen tion of morningness among those who woke up
et al. 2017; Munoz et al. 2017). On weekends, earlier relative to their preferred time due to
these differences were even more pronounced insomnia. However, prevalence of morning and
with evening types having more frequent meals evening chronotypes in this study population was
at irregular times (Maukonen et al. 2017). In one consistent with several populations in Europe and
study, differences in meal content and timing in the United States within the same age range
between chronotypes were modified by an indivi- (Allebrandt et al. 2014; Fischer et al. 2017;
dual’s BMI (Munoz et al. 2017). Overweight or McMahon et al. 2018). Also, participants of the
obese participants with a morning chronotype current study did not have any major acute or
consumed high-calorie meals for dinner while eve- chronic health conditions that could potentially
ning types consumed them at breakfast and lunch, affect sleep, and those who regularly used sleep
a pattern opposite to that observed in the normal promoting medications, both over the counter
weight stratum (Munoz et al. 2017). In the current and prescription, were excluded. Therefore, con-
study, E-DII scores among morning types were founding by somatic conditions or sleep disorders
lower (anti-inflammatory) than in intermediate was unlikely to bias the results.
and evening types, although among overweight The generalizability of these results is limited to
and obese participants, they were higher (more healthy adults ages 21–35 years, half of whom were
pro-inflammatory) than among those with college or graduate students with relatively low
a normal BMI. Obesity among morning chrono- incomes (<$40 000). Loss to follow-up by the end
types may result from complex interactions of 2 years (19% and 44% during the first
between several factors including diet, meal timing and second years, respectively) may have intro-
and sleep disturbances. duced selection bias. However, those who ended
This study had a few noteworthy limitations. the study early did not differ from those who
Participants’ work schedules and alarm clock use completed all five assessments with respect to sev-
were not available; thus, it was assumed that they eral demographic characteristics (Table S1). It is
worked only on weekdays and were free on week- possible that some individuals who were excluded
ends. This assumption also was used in a recent, from the study, such as those with BMI >35 kg/m2
nationally representative, cross-sectional study of or who had arterial hypertension, had evening
chronotype in US population (Fischer et al. 2017), chronotypes and poor sleep. This may explain, at
but it may have led to non-differential misclassifi- least partially, why no increased risk for obesity or
cation of chronotype and SJL and likely biased elevated blood pressure was observed among eve-
measures of association toward the null. ning chronotypes. The study population was rela-
Approximately 45% of the participants in the cur- tively young and healthy, thus they also may have
rent study were graduate students who typically been more resilient to the potential effects of poor
attend classes and work part-time during the day. sleep or changes in sleep/wake timing on the study
Regular night work in this population is less likely outcomes.
since it could potentially interfere with course- One strength of this study was the use of acti-
work. Approximately 5–7% of the study sample graphic measurements to define SJL, chronotype
may have used an alarm clock on a weekend and sleep disruption, which avoids issues inherent
14 D. M. MCMAHON ET AL.

to self-reported data. Armband actigraphy is accounted for correlated data within individuals
a convenient, non-invasive and relatively inexpen- and bolstered statistical power. The RMLCA was
sive method for quantifying sleep relative to the used to categorize participants into latent chron-
“gold standard” method of polysomnography otype groups based on individual trajectories over
(PSG) (Soric et al. 2013). Research suggests that time. This “data-driven” approach provided clear
armband actigraphy is comparable with PSG at the contrasts among those with different chronotypes.
group level and at least as accurate as other acti- In summary, results from this study indicated
graphic devices including that disrupted sleep had a modest association with
Motionlogger® (Ambulatory Monitoring, Inc. anthropometric measures of obesity and arterial
Ardsley, NY) and Actiwatch2 (Philips Respironics, hypertension in a population of healthy, young
Murrysville, PA) (O’Driscoll et al. 2013; Roane et al. adults over a 2-year period, predominantly
2015; Sharif and Bahammam 2013; Shin et al. 2015). among those with a morning chronotype. The
Compared to PSG, the armband underestimated complex relationships between sleep, chronotype,
SOL while Actiwatch2 overestimated WASO at SJL, obesity and arterial hypertension deserve care-
ambient temperatures (17 °C and 22 °C) (Shin ful examination in future studies.
et al. 2015). At the individual level, armband-
derived sleep measures may deviate from PSG
Acknowledgment
(Roane et al. 2015; Soric et al. 2013). The misclassi-
fication bias resulting from armband inaccuracy The authors thank the study participants and the Energy
would likely have been non-differential, leading to Balance Study team.
an underestimation of the observed associations;
this also may have contributed to the null findings. Disclosure statement
Nonetheless, associations between some actigraphic
J.B.B. was supported by a Dept. of Veterans Affairs Office of
measures and several outcomes were identified.
Research and Development grant (Merit Award: I01CX001182-
In this study, the armband measurements over 01A1). J.R.H. was supported by an Established Investigator
6–10 days were obtained, and information for non- Award in Cancer Prevention and Control from the Cancer
wear periods was augmented using logs completed Training Branch of the National Cancer Institute (K05
by participants. Only days with at least 80% of CA136975). M.D.W. and J.R.H. were supported by grant num-
verifiable time (based on data from armband and ber R44DK103377 [N. Shivappa and M.D. Wirth, Multiple PIs]
from the National Institute of Diabetes and Digestive and
activity logs) were included (McMahon et al. 2018),
Kidney Diseases. S.D.Y. was supported by an NIH grant
thus minimizing potential measurement error. In (R01HL095799) and a Dept. of Veterans Affairs Office
addition, use of repeated sleep measures throughout ofResearch and Development grant (Merit Award:
the study increases their reliability within partici- 5I01CX000898). S.N.B. serves on the scientific advisory boards
pants. Armbands with different serial numbers of Technogym, Clarity, and Santech. He has received research
were worn during different assessment periods. funding from BodyMedia, Technogym, the U.S. Department of
Defense, and the NIH, and he receives book royalties from
Therefore, it is plausible that the measurement
Human Kinetics. G.A.H. has no disclosures. The Energy
error varied between time points within a given Balance Study was funded by an unrestricted grant from Coca-
participant. A previous analysis in this sample indi- Cola Company. Coca-Cola representatives did not participate
cated that sleep characteristics overall remained in the data analyses, interpretation of the results or manuscript
stable within 2 years, and that participants did not preparation.
switch between distinct subgroups with different
sleep quality characteristics (McMahon et al. 2018), Funding
suggesting that measurement error did not result in
significant misclassification. This work was supported by the National Institute of
Other study strengths included the use of gen- Diabetes and Digestive and Kidney Diseases
[R44DK103377]; National Institutes of Health
der-specific cut-points for obesity, which decreases
[R01HL095799]; VA Merit [5I01CX000898]; Established
potential misclassification of outcomes (Browning Investigator Award in Cancer Prevention and Control from
et al. 2010; Lean et al. 1995; Shah and Braverman the Cancer Training Branch of the National Cancer Institute
2012), and a statistical modeling strategy that [K05 CA136975]; Dept. of Veterans Affairs Office of
CHRONOBIOLOGY INTERNATIONAL 15

Research and Development grant [Merit Award: Fryar CD, Gu Q, Ogden CL. 2012. Anthropometric reference
I01CX001182-01A1]. data for children and adults: United States, 2007-2010. In:
Vital and Health Statistics Series, Number 11. Data from
the National Health Survey. Washington DC: US
Department of Health and Human Services, Centers for
References Disease Control and Prevention, National Center for
Health Statistics, p. 1–48. Available at: https://www.cdc.
Ainsworth BE, Haskell WL, Herrmann SD, Meckes N, gov/nchs/data/series/sr_11/sr11_252.pdf.
Bassett DR, Tudor-Locke C, Greer JL, Vezina J, Whitt- Gottlieb DJ, Redline S, Nieto FJ, Baldwin CM, Newman AB,
Glover MC, Leon AS. 2011. 2011 compendium of physical Resnick HE, Punjabi NM. 2006. Association of usual sleep
activities: a second update of codes and MET values. Med duration with hypertension: the Sleep Heart Health Study.
Sci Sports Exerc. 43:1575–81. Sleep. 29:1009–14.
Allebrandt KV, Teder-Laving M, Kantermann T, Peters A, Gu F, Xiao Q, Chu LW, Yu K, Matthews CE, Hsing AW,
Campbell H, Rudan I, Wilson JF, Metspalu A, Caporaso NE, Chang JS. 2016. Sleep duration and cancer
Roenneberg T. 2014. Chronotype and sleep duration: the in the NIH-AARP diet and health study cohort. PLoS One.
influence of season of assessment. Chronobiol Int. 11:e0161561.
31:731–40. Hand GA, Shook RP, Paluch AE, Baruth M, Crowley EP,
Bailey BW, Allen MD, LeCheminant JD, Tucker LA, Jaggers JR, Prasad VK, Hurley TG, Hebert JR,
Errico WK, Christensen WF, Hill MD. 2014. Objectively O’Connor DP, et al. 2013. The Energy Balance Study: the
measured sleep patterns in young adult women and the design and baseline results for a longitudinal study of
relationship to adiposity. Am J Health Promot. 29:46–54. energy balance. Res Q Exerc Sport. 84:275–86.
Bailey SL, Heitkemper MM. 2001. Circadian rhythmicity of Hasler G, Buysse DJ, Klaghofer R, Gamma A, Ajdacic V,
cortisol and body temperature: morningness-eveningness Eich D, Rössler W, Angst J. 2004. The association between
effects. Chronobiol Int. 18:249–61. short sleep duration and obesity in young adults: a 13-year
Breslow N. 1982. Design and analysis of case-control studies. prospective study. Sleep. 27:661–66.
Annu Rev Publ Health. 3:29–54. Hebert JR, Hurley TG, Chiriboga DE, Barone J. 1998.
Browning LM, Hsieh SD, Ashwell M. 2010. A systematic review A comparison of selected nutrient intakes derived from
of waist-to-height ratio as a screening tool for the prediction three diet assessment methods used in a low-fat mainte-
of cardiovascular disease and diabetes: 0.5 could be a suitable nance trial. Public Health Nutr. 1:207–14.
global boundary value. Nutr Res Rev. 23:247–69. Jones BL, Nagin DS. 2007. Advances in group-based trajec-
Cespedes EM, Dudley KA, Sotres-Alvarez D, Zee PC, tory modeling and an SAS procedure for estimating them.
Daviglus ML, Shah NA, Talavera GA, Gallo LC, Mattei J, Sociol Method Res. 35:542–71.
Qi Q, et al. 2016. Joint associations of insomnia and sleep Jones BL, Nagin DS, Roeder K. 2001. A SAS procedure based
duration with prevalent diabetes: the Hispanic Community on mixture models for estimating developmental
Health Study/Study of Latinos (HCHS/SOL). J Diabetes. trajectories. Sociol Methods Res. 29:374–93.
8:387–97. Juda M, Vetter C, Roenneberg T. 2013. Chronotype modu-
Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, lates sleep duration, sleep quality, and social jet lag in
Izzo JL, Jones DW, Materson BJ, Oparil S, Wright JT, et al. shift-workers. J Biol Rhythms. 28:141–51.
2003. The seventh report of the joint national committee on Kim M. 2015. Association between objectively measured
prevention, detection, evaluation, and treatment of high sleep quality and obesity in community-dwelling adults
blood pressure: the jnc 7 report. JAMA. 289:2560–71. aged 80 years or older: a cross-sectional study. J Korean
Culnan E, Kloss JD (2013). Social jetlag as a mediator of the Med Sci. 30:199–206.
relationship between chronotype and body mass index Kudielka BM, Federenko IS, Hellhammer DH, Wüst S. 2006.
[PhD diss.]. College of Arts and Sciences, Drexel Morningness and eveningness: the free cortisol rise after
University, p. 103. awakening in “early birds” and “night owls”. Biol Psychol.
Culnan E, Kloss JD, Grandner M. 2013. A prospective study 72:141–46.
of weight gain associated with chronotype among college Lauderdale DS, Knutson KL, Rathouz PJ, Yan LL, Hulley SB,
freshmen. Chronobiol Int. 30:682–90. Liu K. 2009. Cross-sectional and longitudinal associations
Fischer D, Lombardi DA, Marucci-Wellman H, between objectively measured sleep duration and body mass
Roenneberg T, Tosini G. 2017. Chronotypes in the US– index: the CARDIA sleep study. Am J Epidemiol. 170:805–13.
influence of age and sex. PLoS One. 12:e0178782. Lean ME, Han TS, Morrison CE. 1995. Waist circumference
Flegal KM, Kruszon-Moran D, Carroll MD, Fryar CD, as a measure for indicating need for weight management.
Ogden CL. 2016. Trends in obesity among adults in the BMJ. 311:158–61.
United States, 2005 to 2014. JAMA. 315:2284–91. Levandovski R, Dantas G, Fernandes LC, Caumo W, Torres I,
Frankenfield DC, Rowe WA, Cooney RN, Smith JS, Becker D. Roenneberg T, Hidalgo MPL, Allebrandt KV. 2011.
2001. Limits of body mass index to detect obesity and Depression scores associate with chronotype and social
predict body composition. Nutrition. 17:26–30. jetlag in a rural population. Chronobiol Int. 28:771–78.
16 D. M. MCMAHON ET AL.

Ma Y, Olendzki BC, Pagoto SL, Hurley TG, Magner RP, Parsons MJ, Moffitt TE, Gregory AM, Goldman-Mellor S,
Ockene IS, Schneider KL, Merriam PA, Hébert JR. 2009. Nolan PM, Poulton R, Caspi A. 2015. Social jetlag, obesity
Number of 24-hour diet recalls needed to estimate energy and metabolic disorder: investigation in a cohort study.
intake. Ann Epidemiol. 19:553–59. Int J Obes (Lond). 39:842–48.
Malhotra R, Ostbye T, Riley CM, Finkelstein EA. 2013. Patterson F, Malone SK, Grandner MA, Lozano A, Perkett M,
Young adult weight trajectories through midlife by body Hanlon A. 2018. Interactive effects of sleep duration and
mass category. Obesity. 21:1923–34. morning/evening preference on cardiovascular risk factors.
Malnick SD, Knobler H. 2006. The medical complications of Eur J Public Health. 28:155–61.
obesity. Qjm. 99:565–79. Rahman M, Berenson AB. 2010. Accuracy of current body mass
Maukonen M, Kanerva N, Partonen T, Kronholm E, index obesity classification for white, black, and Hispanic
Tapanainen H, Kontto J, Männistö S. 2017. Chronotype reproductive-age women. Obstet Gynecol. 115:982–88.
differences in timing of energy and macronutrient intakes: Randler C, Schaal S. 2010. Morningness-eveningness, habi-
A population-based study in adults. Obesity. 25:608–15. tual sleep-wake variables and cortisol level. Biol Psychol.
McMahon DM, Burch JB, Wirth MD, Youngstedt SD, Hardin 85:14–18.
JW, Hurley TG, Blair SN, Hand GA, Shook RP, Drenowatz Roane BM, Van Reen E, Hart CN, Wing R, Carskadon MA.
C, et al. 2018. Persistence of social jetlag and sleep disruption in 2015. Estimating sleep from multisensory armband mea-
healthy young adults. Chronobiol Int. 35:312–328. surements: validity and reliability in teens. J Sleep Res.
Meerlo P, Sgoifo A, Suchecki D. 2008. Restricted and disrupted 24:714–21.
sleep: effects on autonomic function, neuroendocrine stress Roenneberg T, Allebrandt KV, Merrow M, Vetter C. 2012.
systems and stress responsivity. Sleep Med Rev. 12:197–210. Social jetlag and obesity. Curr Biol. 22:939–43.
Merikanto I, Kronholm E, Peltonen M, Laatikainen T, Roenneberg T, Kuehnle T, Juda M, Kantermann T,
Lahti T, Partonen T. 2012. Relation of chronotype to Allebrandt K, Gordijn M, Merrow M. 2007.
sleep complaints in the general Finnish population. Epidemiology of the human circadian clock. Sleep Med
Chronobiol Int. 29:311–17. Rev. 11:429–38.
Merikanto I, Lahti T, Puolijoki H, Vanhala M, Peltonen M, Rutters F, Lemmens SG, Adam TC, Bremmer MA, Elders PJ,
Laatikainen T, Vartiainen E, Salomaa V, Kronholm E, Nijpels G, Dekker JM. 2014. Is social jetlag associated with
Partonen T. 2013. Associations of chronotype and sleep with an adverse endocrine, behavioral, and cardiovascular risk
cardiovascular diseases and type 2 diabetes. Chronobiol Int. profile? J Biol Rhythms. 29:377–83.
30:470–77. Saksvik IB, Bjorvatn B, Hetland H, Sandal GM, Pallesen S.
Mota MC, Waterhouse J, De-Souza DA, Rossato LT, Silva CM, 2011. Individual differences in tolerance to shift work–a
Araújo MBJ, Tufik S, de Mello MT, Crispim CA. 2016. systematic review. Sleep Med Rev. 15:221–35.
Association between chronotype, food intake and physical Shah NR, Braverman ER. 2012. Measuring adiposity in
activity in medical residents. Chronobiol Int. 33:730–39. patients: the utility of body mass index (BMI), percent
Munoz JSG, Canavate R, Hernandez CM, Cara-Salmerón V, body fat, and leptin. PLoS One. 7:e33308.
Morante JJH. 2017. The association among chronotype, Sharif MM, Bahammam AS. 2013. Sleep estimation using
timing of food intake and food preferences depends on BodyMedia’s SenseWear armband in patients with
body mass status. Eur J Clin Nutr. 71:736–42. obstructive sleep apnea. Ann Thorac Med. 8:53–57.
Nagin D. 2005. Group-based modeling of development. Shin M, Swan P, Chow CM. 2015. The validity of Actiwatch2 and
Cambridge (MA): Harvard University Press. SenseWear armband compared against polysomnography at
Natale V, Plazzi G, Martoni M. 2009. Actigraphy in the assess- different ambient temperature conditions. Sleep Sci. 8:9–15.
ment of insomnia: a quantitative approach. Sleep. 32:767–71. Shivappa N, Steck SE, Hurley TG, Hussey JR, Hébert JR.
O’Driscoll DM, Turton AR, Copland JM, Strauss BJ, 2014. Designing and developing a literature-derived,
Hamilton GS. 2013. Energy expenditure in obstructive sleep population-based dietary inflammatory index. Public
apnea: validation of a multiple physiological sensor for deter- Health Nutr. 17:1689–96.
mination of sleep and wake. Sleep Breath. 17:139–46. Soric M, Turkalj M, Kucic D, Marusic I, Plavec D, Misigoj-
Ode JJ, Pivarnik JM, Reeves MJ, Knous JL. 2007. Body mass Durakovic M. 2013. Validation of a multi-sensor activity
index as a predictor of percent fat in college athletes and monitor for assessing sleep in children and adolescents.
nonathletes. Med Sci Sports Exerc. 39:403–09. Sleep Med. 14:201–05.
Ogden CL, Carroll MD, Lawman HG, Fryar CD, Kruszon- Spiegel K, Leproult R, Van Cauter E. 1999. Impact of sleep debt
Moran D, Kit BK, Flegal KM. 2016. Trends in obesity on metabolic and endocrine function. Lancet. 354:1435–39.
prevalence among children and adolescents in the United Theorell-Haglow J, Berglund L, Janson C, Lindberg E. 2012. Sleep
States, 1988-1994 through 2013-2014. JAMA. 315:2292–99. duration and central obesity in women - differences between
Pabst SR, Negriff S, Dorn LD, Susman EJ, Huang B. 2009. short sleepers and long sleepers. Sleep Med. 13:1079–85.
Depression and anxiety in adolescent females: the impact of Valdez P, Ramirez C, Garcia A. 1996. Delaying and extending
sleep preference and body mass index. J Adolesc Health. sleep during weekends: sleep recovery or circadian effect?
44:554–60. Chronobiol Int. 13:191–98.
CHRONOBIOLOGY INTERNATIONAL 17

van Someren EJW. 2007. Improving actigraphic sleep esti- Wirth MD, Hebert JR, Hand GA, Youngstedt SD, Hurley TG,
mates in insomnia and dementia: how many nights? Shook RP, Paluch AE, Sui X, James SL, Blair SN. 2015.
J Sleep Res. 16:269–75. Association between actigraphic sleep metrics and body
Vgontzas AN, Fernandez-Mendoza J, Miksiewicz T, composition. Ann Epidemiol. 25:773–78.
Kritikou I, Shaffer ML, Liao D, Basta M, Bixler EO. Wittmann M, Dinich J, Merrow M, Roenneberg T. 2006. Social
2014. Unveiling the longitudinal association between jetlag: misalignment of biological and social time. Chronobiol
short sleep duration and the incidence of obesity: the Int. 23:497–509.
Penn State Cohort. Int J Obes (Lond). 38:825–32. Wong PM, Hasler BP, Kamarck TW, Muldoon MF,
Wang LL. 2014. Body mass index, obesity, and self-control: Manuck SB. 2015. Social jetlag, chronotype, and cardio-
a comparison of chronotypes. Soc Behav Personal. 42:313–20. metabolic risk. J Clin Endocrinol Metab. 100:4612–20.

You might also like