Does sleep duration predict metabolic risk in obese adolescents attending tertiary services? A cross-sectional study.

Abstract

To determine, in a clinical sample of obese adolescents, whether shorter sleep duration is associated with metabolic risk and obesity severity.
Cross-sectional study.
Tertiary care weight-management clinic in Cincinnati, OH, USA.
133 obese adolescents aged 10-16.9 years.
N/A.
Multifaceted sleep duration data were examined with fasting venipuncture and anthropometric data collected during clinical care. Primary Outcome: presence of metabolic syndrome. Secondary Outcomes: waist circumference, triglycerides, HDL-cholesterol, blood pressure, glucose, insulin resistance (HOMA-IR), and body mass index (BMI). Predictors: Sleep duration by (1) parent-report, (2) self-report, and (3) multi-night actigraphy. Analysis: Relationships between sleep duration and each outcome were examined via regression models, adjusted for potential confounders.
Regardless of how measured, sleep duration showed no strong association with metabolic syndrome (OR 1.1 to 1.5, P = 0.2 to 0.8), BMI (β -0.03 to -0.01, P = 0.2 to 0.8), or most other outcomes. Lower triglycerides were predicted by shorter sleep duration by self-report (β 12.3, P = 0.01) and actigraphy (β 13.6, P = 0.03), and shorter parent-reported sleep duration was associated with higher HDL-cholesterol (β = -2.7, P = 0.002).
Contrary to expectations, sleep duration was not associated with metabolic outcomes, and showed limited associations with lipid profiles. Although inadequate sleep may affect other areas of functioning, it appears premature to expect that lengthening sleep will improve BMI or metabolic outcomes in clinical samples of obese adolescents.

Full text

Page 1

SLEEP, Vol. 34, No. 7, 2011
891
Sleep and Metabolic Risk in Obese Adolescents—Sung et al
in 25% to 50% of obese children and adolescents from clinical
samples,7,8 and in 2% to 10% of obese and non-obese adoles-
cents in population samples.9,10 Finding modifiable risk factors
for metabolic syndrome could therefore prevent substantial car-
diovascular morbidity, particularly in obese adolescents.
Sleep Duration, Childhood Obesity, and Metabolic Syndrome
A trend toward decreasing sleep duration has paralleled the
obesity epidemic. A 2005 US national survey showed that 45%
of adolescents have insufficient sleep (< 8 h/night) and a further
31% have borderline (8-9 h) sleep.11 Further, many studies now
support an inverse linear relationship between short sleep dura-
tion and body mass index (BMI) in preadolescent children.12-16
This raises the possibility of a causal relationship, and therefore
the potential for sleep related interventions.
However, results for adolescents have been less consistent,13
with some17-21 but not all22-24 reporting similar findings in popu-
lation-based samples. One case-control study found differences
between obese and demographically matched lean adolescents
across a number of sleep variables, including duration, but did
not report on associations between sleep duration and adiposity
within the obese subgroup.25 A major weakness of research in
this area is the exclusive reliance on subjective parent- or self-
reported sleep measures, with one study of 7-year-old children
and 3 studies of adolescents including actigraphy20,25-27 as objec-
tive measures of sleep.
INTRODUCTION
Childhood Obesity and the Metabolic Syndrome
The epidemic of childhood obesity presents a major public
health problem, with up to 32% of US children and adolescents
overweight or obese, and up to 16% obese (Centers for Dis-
ease Control [CDC] definitions).1 Complications from obesity
involve multiple organ systems, which further compound the
health risk to these children.
One cluster of obesity-linked complications, known as the
metabolic syndrome, strongly predicts diabetes and early car-
diovascular morbidity.2-5 In 2007, the International Diabetes
Foundation defined metabolic syndrome in children as the pres-
ence of central obesity together with any 2 of the following 4
age-dependent factors: hypertriglyceridemia, low high-density
lipoprotein (HDL) cholesterol, high blood pressure, and high
fasting blood glucose.6 Metabolic syndrome has been reported
SLEEP DURATION AND METABOLIC RISK IN OBESE ADOLESCENTS
DOI: 10.5665/SLEEP.1122
Does Sleep Duration Predict Metabolic Risk in Obese Adolescents Attending
Tertiary Services? A Cross-Sectional Study
Valerie Sung, MBBS(Hons), FRACP1,2; Dean W. Beebe, PhD3,4; Rhonda VanDyke, PhD3,4; Matthew C. Fenchel, MS3; Nancy A. Crimmins, MD3,4;
Shelley Kirk, PhD, RD3,4; Harriet Hiscock, MBBS, MD1,2; Raouf Amin, MD3,4; Melissa Wake, MBChB, MD1,2
1Centre for Community Child Health, Royal Children’s Hospital and Murdoch Children’s Research Institute, Royal Children’s Hospital, Parkville,
Victoria, Australia; 2The University of Melbourne, Parkville, Victoria, Australia; 3Cincinnati Children’s Hospital Medical Center, Cincinnati, OH;
4Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, OH
Submitted for publication December, 2010
Submitted in final revised form February, 2011
Accepted for publication February, 2011
Address correspondence to: Dr. Valerie Sung, Centre for Community
Child Health, Royal Children’s Hospital, Flemington Road, Parkville, Vic-
toria, 3052, Australia; Tel: +613 9345 6150; Fax: +613 9345 5900; E-mail:
Valerie.sung@rch.org.au
Study Objectives: To determine, in a clinical sample of obese adolescents, whether shorter sleep duration is associated with metabolic risk and
obesity severity.
Design: Cross-sectional study.
Setting: Tertiary care weight-management clinic in Cincinnati, OH, USA.
Participants: 133 obese adolescents aged 10-16.9 years.
Interventions: N/A.
Measurements: Multifaceted sleep duration data were examined with fasting venipuncture and anthropometric data collected during clinical care.
Primary Outcome: presence of metabolic syndrome. Secondary Outcomes: waist circumference, triglycerides, HDL-cholesterol, blood pressure,
glucose, insulin resistance (HOMA-IR), and body mass index (BMI). Predictors: Sleep duration by (1) parent-report, (2) self-report, and (3) multi-
night actigraphy. Analysis: Relationships between sleep duration and each outcome were examined via regression models, adjusted for potential
confounders.
Results: Regardless of how measured, sleep duration showed no strong association with metabolic syndrome (OR 1.1 to 1.5, P = 0.2 to 0.8),
BMI (β −0.03 to −0.01, P = 0.2 to 0.8), or most other outcomes. Lower triglycerides were predicted by shorter sleep duration by self-report (β 12.3,
P = 0.01) and actigraphy (β 13.6, P = 0.03), and shorter parent-reported sleep duration was associated with higher HDL-cholesterol (β = −2.7,
P = 0.002).
Conclusions: Contrary to expectations, sleep duration was not associated with metabolic outcomes, and showed limited associations with lipid
profiles. Although inadequate sleep may affect other areas of functioning, it appears premature to expect that lengthening sleep will improve BMI
or metabolic outcomes in clinical samples of obese adolescents.
Keywords: Sleep, obesity, metabolic syndrome, adolescent, cross sectional studies
Citation: Sung V; Beebe DW; VanDyke R; Fenchel MC; Crimmins NA; Kirk S; Hiscock H; Amin R; Wake M. Does sleep duration predict metabolic
risk in obese adolescents attending tertiary services? A cross-sectional study. SLEEP 2011;34(7):891-898.

Page 2

SLEEP, Vol. 34, No. 7, 2011
892
Sleep and Metabolic Risk in Obese Adolescents—Sung et al
≥ 95th percentile for age and sex, using CDC definitions.42 Ex-
clusion criteria for current analyses included neurological his-
tory (e.g., seizures), craniofacial abnormalities, developmental
disorders (e.g., Down syndrome), conditions involving daytime
hypoxia (e.g., poorly controlled asthma), use of psychiatric or
neurological medication, and treatment for OSA in the past 2
years.
Overview of Procedures
At the initial visit to the clinic, demographics and anthro-
pometric data, including weight, height, waist circumference,
and blood pressure were routinely collected. Metabolic mark-
ers, including triglycerides, HDL-cholesterol, glucose, and
insulin, were routinely collected by a 12-h fasting blood test
performed within a month of the initial medical visit. All serum
analyses were conducted at the same clinically accredited lab
at CCHMC, and the procedure for obtaining the fasting labora-
tory measures was the same in all cases. During the initial clinic
visit, an investigator (DWB) briefly introduced himself and the
study, then followed up with interested families via telephone
to conduct a more detailed discussion and screening interview.
Once eligibility and participation were confirmed, a booking
was made for an overnight polysomnogram (PSG) and for ac-
tigraphy, usually in the week immediately before the PSG. The
PSGs were conducted at a mean of 14 weeks after the initial
clinic visit. Parent and participant questionnaires were indepen-
dently completed in the context of an afternoon neurobehavior-
al assessment (detailed in our prior report),41 which in nearly all
cases occurred the afternoon prior to the PSG, and in all cases
occurred within 1 month of the PSG.
MEASURES
Sleep Duration Predictor Measures
(1) Subjective sleep parameters: Parent-reported adoles-
cent sleep duration was recorded by the Child Sleep Habits
Questionnaire,43 and self-reported sleep duration by the Youth
Sleep Habits Survey.44,45 Both reporters were asked how long
the participant typically slept, as well as when he or she typi-
cally goes to bed and rises for the day, on school nights. Simi-
lar questions are widely used in sleep surveys, and have been
found to correlate strongly with objective actigraphy data, es-
pecially on school nights.44 For each reporter, we computed
sleep duration as the mean of the direct report of sleep dura-
tion and the arithmetic difference between reported bedtime
and rise time.
(2) Objective actigraphy: As previously described,25 ado-
lescent participants were asked to wear motion-sensitive ac-
tigraphs (Mini-Motionlogger, Ambulatory Monitoring, Inc.,
NY) the size of sports wristwatches on their non-dominant
wrists during evening and overnight hours for a week. The
units were provided in person by study staff or via overnight
mail, accompanied by detailed written instructions that were
also reviewed via telephone. Subjects were asked to wear the
units at home during evening and overnight hours during their
normal sleep routine. Raw data on participant movement pat-
terns were cross-checked against a nightly sleep diary to help
screen for artifacts (e.g., actigraph removal) prior to being
submitted to an automated scoring algorithm that estimated
Adult studies have shown sleep duration to be associated
with the metabolic syndrome and diabetes.28 The very few
metabolic studies exploring associations in children suggest
relationships between sleep duration and insulin resistance
and waist circumference,29-31 but are inconsistent. No study has
found associations between sleep duration and hypertension,
glucose levels, or lipid abnormalities in the pediatric popula-
tion. Further, although obstructive sleep apnea (OSA) is com-
mon among obese youth,25 there is considerable disagreement
as to whether its presence or severity does32-37 or does not38-40
contribute to adolescent metabolic syndrome, over and above
the contribution of obesity itself. Thus OSA could obscure or
confound the relationship between sleep duration and the meta-
bolic syndrome, especially in obese adolescents.
Aims and Hypotheses
Despite this conflicting literature on associations among
child and adolescent sleep duration, BMI, and metabolic risk,
sleep is being explored as a possible treatment target to mini-
mize the risk for cardiovascular morbidity. This seems pre-
mature until associations are confirmed and quantified and, if
found to be important, causal directions established. To address
these issues, we studied a large clinical sample of children and
adolescents attending a tertiary hospital outpatient weight man-
agement program for whom measured sleep duration, anthro-
pometry, and markers of metabolic syndrome were all available.
Specifically, we aimed to determine, in obese 10- to
16.9-year-olds, cross-sectional associations between sleep du-
ration and the metabolic syndrome. We also aimed to determine
if associations existed between sleep duration and individual
components of the metabolic syndrome (i.e., lipid markers,
blood pressure, fasting glucose), as well as insulin resistance,
and adiposity. We hypothesized that sleep duration would be
inversely associated with metabolic syndrome, its individual
components, adiposity, and insulin resistance in a manner that
is statistically independent of OSA.
METHODS
Design and Setting
A sample of obese adolescents was consecutively recruited
between 2003 and 2008 at intake into the Cincinnati Children’s
Hospital Medical Center (CCHMC) multidisciplinary pediat-
ric weight-management clinic, as part of a larger study inves-
tigating the relationship between sleep and neurobehavioral
functioning in adolescents.41 Demographic and sleep data were
prospectively collected, and anthropometric and metabolic
data were collected as part of each family’s comprehensive in-
take evaluation for the clinic. Both the original study and the
merging of research and clinical databases for current analyses
were approved and overseen by the Institutional Review Board
(IRB) of CCHMC. All parents provided written informed con-
sent, and all participants provided written assent for collection
of research-specific data and for review of the participant’s
medical record for data collected during clinical care.
Sample
Adolescents attending the clinic were eligible if they were 10
to 16.9 years of age and had a body mass index (BMI, kg/m2)

Page 3

SLEEP, Vol. 34, No. 7, 2011
893
Sleep and Metabolic Risk in Obese Adolescents—Sung et al
Finally, OSA was entered as a covariate because OSA se-
verity may independently influence metabolic markers, is as-
sociated with obesity, may affect sleep duration,25,41 and may
introduce variance into the quality of sleep of any duration.
OSA, as indexed by the apnea+hypopnea index (AHI), was
measured during overnight PSG, conducted during a single
night in an inpatient pediatric sleep unit.25,41 The AHI was com-
puted as the sum of obstructive apneas and hypopneas divided
by hours slept. Obstructive apnea was defined as > 80% decline
in airflow over 2 breaths, despite continued chest/abdominal
wall movement. Obstructive hypopneas were defined as a de-
crease of 50% to 80% in airflow over ≥ 2 breaths associated
with (a) paradoxical respiration and (b) oxyhemoglobin desatu-
ration (≥ 4%) or a subsequent arousal.
Statistical Analysis
For our primary analysis, 3 logistic regression models ex-
amined the change in odds of participants displaying metabolic
syndrome by sleep duration for each of the 3 sleep duration
measures (parent-report, self-report, and actigraphy). Simi-
larly, the secondary outcomes (all continuous variables) were
examined using multiple linear regression models. All regres-
sion models included the covariates identified above, with AHI
log-transformed to approximate normality, and BMI z-score
included only in the metabolic outcome analyses only.
Analyses were conducted using SAS 9.2 software (SAS In-
stitute, Cary, NC). Post hoc power analyses were conducted
based upon the available sample size and our observed distri-
butions of the predictors, outcomes, and covariates. Assum-
ing a population base rate of metabolic syndrome of 25% in
obese adolescents (the observed rate in this sample was 23%),
α = 0.05, and 120 subjects, our primary analyses had 80% pow-
er to detect an odds ratio of 1.95 for a 1-h increase in sleep dura-
tion. Power would be even higher if the base rate of metabolic
syndrome in obese adolescents exceeded 25, as suggested by
some authors.7,8,54,55
RESULTS
Sample Description
Figure 1 shows the participant flow and reasons for non-par-
ticipation. Of the 380 families attending clinic, 247 were ap-
proached and deemed to be eligible for the study, of whom 133
(54%) participated and provided anthropometric and metabolic
data. Of these, 92% of parents and adolescents also reported
sleep duration, and 91% had actigraphy data (worn on average
5.8 ± 1.3 nights).
Table 1 summarizes the demographic, sleep, anthropomet-
ric, and metabolic features of the sample. Adolescents who
participated in the study did not significantly differ from non-
participants in age or sex (P = 0.4-0.5). Because we did not have
consent to access additional information on non-participants, it
is not known whether participants and non-participants differed
on other study variables. By design, the sample was obese, with
a mean BMI of 37 kg/m2; and participants’ mean sleep duration
was < 9 h. More than half of the sample (56%) experienced
some degree of sleep disordered breathing, and 23% met cri-
teria for the diagnosis of metabolic syndrome. A minority had
individually significant markers of metabolic risk, except for
sleep onset and offset based upon movement patterns; the re-
sults of this algorithm have been shown to correlate highly
with PSG-determined sleep.46 Sleep duration was calculated
as the offset minus onset times averaged over the nights the
unit was worn.
Laboratory Measures
All serum analyses were conducted at the same clinically
accredited lab at CCHMC, and the procedure for obtaining
the fasting blood work and laboratory measures was the same
in all cases. Plasma glucose was measured using a Hitachi
model 704 glucose analyzer. Plasma insulin was measured
by RIA using an anti-insulin serum raised in guinea pegs, 125I
labeled insulin (Linco, St. Louis, MO), and a double antibody
method to separate bound from free tracer. Assays of fasting
plasma lipid profiles were carried out in a laboratory that is
National Heart, Lung, and Blood Institute/Centers for Dis-
ease Control and Prevention standardized, with the low-den-
sity lipoprotein (LDL) cholesterol concentration calculated
using the Friedewald equation.
Anthropometric Measures
Weight was measured by a digital scale, and height measured
using a calibrated wall-mounted stadiometer while subjects
were in light clothing and without shoes. Waist circumference
was measured using a standardized protocol where the circum-
ference measurement is taken just above the uppermost lateral
border of the right iliac crest.47 Blood pressure was measured
using an appropriate-sized cuff, and the mean blood pressure
value was taken from 2 readings.
Definitions and Calculations
The presence of metabolic syndrome was defined accord-
ing to the 2007 International Diabetes Foundation criteria6,48
as: waist circumference ≥ 90th percentile for age/gender49 and
2 of the following: (a) fasting triglycerides ≥ 150 mg/dL, (b)
fasting HDL-cholesterol < 40 mg/dL, (c) systolic or diastolic
BP ≥ 90th percentile,50 and (d) fasting glucose ≥ 100 mg/dL.
HOMA-IR was calculated by [fasting serum insulin (μU/mL)
× fasting plasma glucose (mmol/L)] divided by 22.5.51-53 BMI
(kg/m2) was transformed to BMI z-scores using published CDC
tables that adjust for age and sex.42
Potential Confounders/Covariate Measures
All potential confounders were identified a priori. Parent-
reported demographic variables included age, gender, race
(coded as white/non-white, with 96% of non-white partici-
pants being African American) and socioeconomic status, all
of which are known to be associated with sleep duration, met-
abolic risk, or both. Socioeconomic status was represented as
a SES composite index, computed as the mean z-score within
the sample across 4 variables: maternal education, paternal
education, annual family income and median local income
based upon the 2000 US Census for the subject’s postal code
of residence, prorating for any missing fields (e.g., paternal
education unreported).41 For the metabolic analyses, BMI z-
score was also included because of its previously reported
independent associations with both sleep duration and meta-
bolic syndrome.

Page 4

SLEEP, Vol. 34, No. 7, 2011
894
Sleep and Metabolic Risk in Obese Adolescents—Sung et al
duration and triglycerides/cholesterol displayed heterosce-
dasticity—i.e., there was greater variability at one end of the
sleep duration than at the other. Re-analysis of each secondary
outcome using multiple logistic regression and dichotomizing
according to cutpoints used to determine metabolic syndrome
yielded nearly identical conclusions to the linear analyses.
After controlling for covariates, sleep duration did not pre-
dict BMI, BMI z-score, or waist circumference, nor did it pre-
dict insulin resistance as measured by HOMA-IR and fasting
insulin levels (all P > 0.05). AHI was a significant covariate
for modeling waist-circumference against self-reported sleep
duration (β estimate for log-AHI = 5.68, P = 0.03) and baseline-
BMI z-score against actigraphy sleep duration (β estimate for
log-AHI = 0.14, P = 0.04). AHI was not a significant covariate
in any of the other models for secondary outcomes.
DISCUSSION
Shorter sleep duration did not predict the presence of the met-
abolic syndrome, adiposity, or insulin resistance in this sample
of obese children and adolescents. Moreover, short sleep dura-
tion was unexpectedly associated with better triglyceride and
HDL-cholesterol levels, although the strength and statistical
low HDL-cholesterol, which occurred in more than half of the
sample, and high waist circumference, which was present in
almost the entire sample.
Primary Outcome
There were no significant differences in sleep duration
among adolescents with a diagnosis of metabolic syndrome and
those without (Table 2), although small increases in the odds of
having metabolic syndrome were observed for each h increase
in sleep duration (parent-report OR = 1.2, 95% CI = 0.7-2.0;
self-report OR = 1.1, 95% CI = 0.6-1.8; actigraphy OR = 1.5,
95% CI = 0.8-2.7, P = 0.2-0.8). Further, AHI was not a signifi-
cant covariate in any of the logistic regression models.
Secondary Outcomes
There was a lack of relationship between sleep duration and
all secondary outcomes, except for triglycerides and HDL-cho-
lesterol that were in the opposite direction to that hypothesized
(Table 3). Longer sleep duration by self-report and actigra-
phy predicted worse (i.e., higher) triglyceride levels (β = 12.3
[P = 0.01] and 13.6 [P = 0.03], respectively; Figs. 2A and 2B),
while longer sleep duration by parent-report predicted worse
(i.e., lower) HDL-cholesterol levels (β = -2.7 [P = 0.002];
Figure 2C). These unanticipated relationships between sleep
Table 1—Demographic, sleep and metabolic characteristics of the
sample (n = 133)
Characteristics
Child
Male sex (%)
Age (years), mean (SD)
Race: white (%)
BMI (kg/m2), mean (SD)
BMI-z score, mean (SD)
Sleep duration (hours per night), median (range)
Parent-report
Self-report
Actigraphy
Obstructive sleep apnea: AHI ≥ 1 (%)
Metabolic Syndrome present (%)
Markers of metabolic risk (%)
Waist circumference ≥ 90th percentile
Triglycerides ≥ 150 mg/dL
HDL cholesterol < 40 mg/dL
Systolic blood pressure ≥ 90th percentile
Diastolic blood pressure ≥ 90th percentile
Fasting glucose levels ≥ 100 mmol/L
% or Mean (SD)
or Median (range)
34
13.2 (1.8)
32
37.0 (7.2)
2.5 (0.3)
8.5 (6.3 – 11.8)
8.3 (5.5 – 13.2)
7.8 (4.9 – 9.5)
56
23
98
17
52
13
12
11
Table 2—Sleep duration amongst adolescents with and without the
diagnosis of metabolic syndrome
Sleep duration
(hours per night),
median (range)
Parent-report
Self-report
Actigraphy
With metabolic
syndrome
(n = 28)
8.5 (6.5 – 11.5)
8.5 (6.2 – 9.7)
7.9 (5.6 – 9.5)
Without metabolic
syndrome
(n = 105)
8.5 (6.3 – 11.8)
8.1 (5.5 – 13.2)
7.8 (4.9 – 9.3)
Figure 1—Participant flow chart
Between 2003 and 2008 approached 380 obese
(BMI ≥ 95th percentile) adolescents aged 10 to 16.9
years attending HealthWorks! clinic
• 75 failed to meet
inclusion criteria
• 68 unable to be
contacted
• 96 declined or failed
follow-up
• 8 missing metabolic
data
133 adolescents included in sample
Anthropometric
& metabolic
data
(n = 133)
Self-report
of sleep
duration
(n = 122)
Parent-report
of sleep
duration
(n = 123)
Actigraphy
(n = 122)

Page 5

SLEEP, Vol. 34, No. 7, 2011
895
Sleep and Metabolic Risk in Obese Adolescents—Sung et al
There are some limitations to this study. As this was a cross-
sectional study, it remains possible that short sleep increases the
risk for the initial development of obesity and/or precursors of
metabolic syndrome, which are then maintained and exacerbated
by other factors. Such an effect would only be evident in a lon-
gitudinal model, perhaps complemented by within-subjects ex-
perimental sleep restriction studies that allow for better control
of between-subjects variance. Measures for the current study
were obtained at different time-points, and there was a general
time-lag between the metabolic and sleep measures; insofar as
sleep or metabolic functioning might have fluctuated over time
this could have attenuated associations. While our recruitment
from a weight management clinic allowed us to efficiently en-
roll significantly obese adolescents, these findings may not gen-
eralize to obese, non–treatment-seeking individuals. Detailed
information on physical activity, diet, medication, smoking, and
alcohol drinking were not systematically collected in the study,
and therefore could not be considered as potential confounders.
Although we observed considerable variability in sleep duration
and the large majority of participants obtained less than the rec-
ommended 9 hours of sleep, very few had sleep duration < 6
hours (< 2% per parent-report or self-report, 6% per actigraphy).
Thus it remains possible that markedly short sleep duration is as-
sociated with metabolic syndrome. Finally, this was not an inter-
vention study that could more definitively speak to whether sleep
duration modification alters metabolic risk factors within indi-
viduals; however, across obese adolescents, there appears to be
minimal relationship between sleep duration and metabolic risk.
Some of our results are similar to, and some differ from, pre-
vious studies in the pediatric population. We did not find shorter
sleep duration to be associated with insulin resistance (in con-
trast to Flint’s retrospective chart review of 40 obese 3.5- to
18.5-year-olds undergoing overnight PSG29) or waist circum-
ference (in contrast to Verhulst’s PSG study of 104 obese 6- to
17-year-olds30 and Hitze’s population study of 414 6- to 20-year-
olds31). However, in common with all three of these studies,
sleep duration was not associated with blood pressure.29-31
To our knowledge, this is the only study that has found an
association between longer sleep duration and worse triglyc-
significance of these effects varied across three different mea-
sures of sleep duration.
This study is the first to comprehensively describe the rela-
tionship between measured sleep duration and all aspects of the
metabolic syndrome, while adjusting for potential confounders,
in a group of high-risk obese children and adolescents. While
conventional statistical methods are not designed to accept the
null hypothesis, this study had a number of strengths that lend
confidence that, at least within this subpopulation, any relation-
ship between short sleep duration and the presence of metabolic
syndrome or its elements is, at most, quite small. A post hoc
power analysis suggested that the sample was of adequate size
to detect even a relatively small effect at a liberal significance
level of 0.05. For our primary outcome and several secondary
outcomes, the nonsignificant trends ran opposite to predictions,
suggesting that increasing the sample size would be extremely
unlikely to find effects in the direction hypothesized. Further,
the failure to find significant effects across multiple measures of
sleep duration, including multi-night actigraphy as well as par-
ent- and self-report, strongly argues against measure invalidity
as a cause of nonsignificant results. The only other commonly
used sleep measure in this age range, PSG, is the gold standard
for measuring OSA, but is a poor measure of typical free-living
sleep duration because it introduces artificial influences that can
fundamentally alter sleep patterns (e.g., presence of monitoring
leads and equipment).
Complementing our strong sleep measurement, this study
adopted an internationally recognized classification of metabol-
ic syndrome. Individual metabolic markers were also analyzed
to ensure that potentially important findings did not emerge for
specific elements of the metabolic syndrome, suggesting that
alternate metabolic syndrome classifications would not change
our conclusions. Further, aside from waist circumference, there
was good variability in both the predictors and outcomes, sug-
gesting that associations were not attenuated by a restriction
in range. Finally, potential confounders were identified a priori
and included in the analyses. In particular, OSA, which was
prevalent in this sample, was objectively measured and includ-
ed as a covariate.
Table 3—Regression of (Secondary) BMI and metabolic outcomes on sleep duration
Sleep duration (hours/night)
Self-report
Beta Coeff95% CI
-0.01-0.05, 0.04
Outcome VariableParent-report
95% CI
-0.07, 0.02
Actigraphy
95% CI
-0.09, 0.02
Beta Coeff
-0.02
PP
0.8
Beta Coeff
-0.03
P
BMI z-scorea
Metabolic markersb
Waist circumference (cm)
Triglycerides (mg/dL)
HDL-cholesterol (mg/dL)
Systolic blood pressure (mmHg)
Diastolic blood pressure(mmHg)
Fasting glucose (mg/dL)
HOMA-IR
0.4 0.2
-1.6
2.3
-2.7*
0.05
0.01
-0.5
0.2
-3.6, 0.5
-9.6, 14.1
-4.4, -1.0
-0.2, 0.3
-0.2, 0.2
-2.1, 1.1
-0.8, 1.2
0.1
0.7
0.002
0.6
1.0
0.5
0.7
0.6
12.3*
-0.9
-0.04
-0.03
-1.3
-0.02
-1.2, 2.3
2.5, 22.1
-2.6, 0.7
-0.2, 0.1
-0.2, 0.1
-2.6, 0.1
-0.8, 0.8
0.5
0.01
0.3
0.7
0.7
0.07
1.0
0.3
13.6*
-0.8
0.05
-0.2
-0.8
0.8
-1.8, 2.4
1.1, 26.1
-2.8, 1.2
-0.2, 0.3
-0.2, 0.02
-2.6, 0.9
-0.2, 1.8
0.8
0.03
0.4
0.7
0.1
0.4
0.1
*Significant association in opposite direction to hypothesized relationship. aAdjusted for age, gender, race, socio-economic status, and obstructive apnea.
bAdjusted for age, gender, race, socio-economic status, BMI z-score, and obstructive apnea.