Vol. 177, No.

9
American Journal of Epidemiology
DOI: 10.1093/aje/kws342
© The Author 2013. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of
Advance Access publication:
Public Health. All rights reserved. For permissions, please e-mail: [email protected].
April 14, 2013

Original Contribution

Increased Prevalence of Sleep-Disordered Breathing in Adults

Paul E. Peppard*, Terry Young, Jodi H. Barnet, Mari Palta, Erika W. Hagen, and Khin Mae Hla
* Correspondence to Dr. Paul E. Peppard, Department of Population Health Sciences, University of Wisconsin–Madison, WARF Building 685,
610 Walnut St., Madison, WI 53726 (e-mail: [email protected]).

Initially submitted May 11, 2012; accepted for publication August 7, 2012.

Sleep-disordered breathing is a common disorder with a range of harmful sequelae. Obesity is a strong causal
factor for sleep-disordered breathing, and because of the ongoing obesity epidemic, previous estimates of sleep-
disordered breathing prevalence require updating. We estimated the prevalence of sleep-disordered breathing in
the United States for the periods of 1988–1994 and 2007–2010 using data from the Wisconsin Sleep Cohort
Study, an ongoing community-based study that was established in 1988 with participants randomly selected from
an employed population of Wisconsin adults. A total of 1,520 participants who were 30–70 years of age had base-
line polysomnography studies to assess the presence of sleep-disordered breathing. Participants were invited for
repeat studies at 4-year intervals. The prevalence of sleep-disordered breathing was modeled as a function of
age, sex, and body mass index, and estimates were extrapolated to US body mass index distributions estimated
using data from the National Health and Nutrition Examination Survey. The current prevalence estimates of mod-
erate to severe sleep-disordered breathing (apnea-hypopnea index, measured as events/hour, ≥15) are 10%
(95% confidence interval (CI): 7, 12) among 30–49-year-old men; 17% (95% CI: 15, 21) among 50–70-year-old
men; 3% (95% CI: 2, 4) among 30–49-year-old women; and 9% (95% CI: 7, 11) among 50–70 year-old women.
These estimated prevalence rates represent substantial increases over the last 2 decades (relative increases of
between 14% and 55% depending on the subgroup).

adult; middle age; obesity; sleep

Abbreviations: AHI, apnea-hypopnea index; BMI, body mass index; NHANES, National Health and Nutrition Examination Survey;
SDB, sleep-disordered breathing.

The apnea and hypopnea events of sleep-disordered breath- more than a decade ago (12, 13). The most important modi-
ing (SDB) have substantial harmful health consequences. fiable causes of SDB in adult populations are overweight and
Immediate effects include intermittent hypoxia, fragmented obesity. Weight gain and loss have been consistently
sleep, and exaggerated fluctuations in heart rhythm, blood associated with increasing and decreasing SDB severity,
pressure, and intrathoracic pressure (1). In turn, these acute respectively, in observational and intervention studies (14–
physiologic disruptions evolve into long-term sequelae, such 16). Over the last few decades, the prevalence rates of over-
as hypertension and cardiovascular morbidities (1–3), decre- weight and obesity experienced epidemic trajectories in the
ments in cognitive function (4, 5), mood and quality of life United States (17–20), which is likely to have resulted in
(6, 7), and premature death (8, 9). increased occurrence of obesity-related outcomes, including
In 1993, data collected over a 4-year period from the SDB.
Wisconsin Sleep Cohort Study uncovered a high prevalence In the United States, a systematic program for monitoring
of SDB assessed using polysomnography in a working SDB prevalence over time does not exist. High-quality objec-
population–based sample of adults 30–60 years of age (10). tive assessments of SDB are time-consuming, expensive, and
The findings were corroborated by other US population- burdensome to subjects, typically requiring overnight
based studies (11), but these prevalence rates were estimated continuous monitoring of multiple physiologic processes,

1006 Am J Epidemiol. 2013;177(9):1006–1014

Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest

on 05 April 2018
 Prevalence of Sleep-Disordered Breathing 1007

including sleep/wake state and respiratory functioning. As a sleep study (53% of those invited for baseline studies). The
consequence, despite the high prevalence of and broad range primary reported reason for nonparticipation was the burden
of negative health outcomes from SDB, there is presently no of sleeping overnight in a sleep laboratory.
national monitoring system for tracking SDB prevalence in a Sleep studies were performed at the Clinical Research Unit
fashion analogous to the manner in which iterations of the of the University of Wisconsin Hospital and Clinics. The
US National Health and Nutrition Examination Survey majority of participants had undergone 2 or more (up to 6)
(NHANES) are used to track the prevalence rates of health sleep studies, all of which were used for this analysis (416 par-
exposures and outcomes such as overweight and obesity, ticipants underwent exactly 1 study, 172 underwent 2 studies,
hypertension, and blood lead levels. However, if a robust 298 underwent 3 studies, 325 underwent 4 studies, 245 under-
SDB prevalence estimation model is developed that has as went 5 studies, and 64 underwent 6 studies). Not all subjects
input parameters factors that are routinely tracked and accu- hadthemaximumpossiblenumberofstudies(6) because: 1) not
rately estimated, that model could allow for serial estimations all subjects had yet had the opportunity to be invited for mul-
of SDB prevalence. tiple repeat studies ( protocols are ongoing and repeat studies
In the present study, we developed models of SDB preva- accrue at a rate of approximately 200 per year); 2) 83 subjects
lence as a function of sex, age group, and weight status cate- had studies that produced sleep data of insufficient quality
gories, the 3 most important factors in SDB prevalence in (e.g., in-laboratory sleep time <4 hours); 3) 112 subjects have
US populations (11). We applied the models to adult (30–70 died since the study began; and 4) 28 subjects who partici-
years of age) SDB prevalence in the early 1990s and in a pated in a baseline study refused future studies or were not
recent time period (2007–2010), framing a time interval that locatable. The participation rate for ongoing follow-up studies,
corresponded with a rapid expansion of the US obesity epi- that is, the proportion of those invited for follow-up studies
demic. This was performed by combining information from that agree to participate, is approximately 80%. For the present
the Wisconsin Sleep Cohort Study, which since 1988 has analysis, studies from 31 subjects older than 70 years of age
performed more than 4,500 overnight in-laboratory SDB were excluded because there were too few observations from
evaluations on 1,520 study participants, and the NHANES subjects in that age range.
in an approach that increased our new estimates’ generaliz-
ability and replaced previous, now-outdated prevalence esti- Measurements
mates (10, 16). We did this in 3 steps: 1) using data from
participants in the ongoing Wisconsin Sleep Cohort Study, Sleep studies included polysomnography and other
we modeled sex, age, and body mass index (BMI) strata– assessments, such as ascertainment of BMI, which was cal-
specific prevalence estimates of SDB; 2) we used NHANES culated as measured weight (in kilograms) divided by height
(21) data to estimate US population distributions of corre- (in meters) squared (23). Data on medical history, medica-
sponding sex, age, and BMI strata for 2 periods, the early tion use, alcohol use, smoking habits, and self-reported sleep
1990s (representing the initiation of the Wisconsin Sleep habits, problems, and daytime sleepiness were obtained using
Cohort) and the late 2000s; and 3) we applied the Wisconsin questionnaires. For this study, daytime sleepiness was
Sleep Cohort SDB prevalence estimates to the 2 periods. assessed using the Epworth Sleepiness Scale (24, 25), a
This process yielded estimates of US adult SDB prevalence widely used, validated questionnaire that was added to the
in the early 1990s and late 2000s. Wisconsin Sleep Cohort Study in 1993 and that asks sub-
jects to rate their likelihood of falling asleep in a variety of
common situations. Possible scores range from 0 (least
MATERIALS AND METHODS sleepy) to 24 (sleepiest), with daytime sleepiness defined as
Participants a score of >10, a commonly used clinical definition of exces-
sive sleepiness.
Informed consent documents and study protocols for the A polysomnography system (Grass Instruments, Quincy,
Wisconsin Sleep Cohort Study, which have been des- cribed Massachusetts) was used to assess sleep state and respiratory
previously (22), were approved by the University of and cardiac parameters. Sleep state was determined by elec-
Wisconsin-Madison Health Sciences Institutional Review troencephalography, electrooculography, and electromyog-
Board. All employees of 5 state agencies in south-central raphy. Arterial oxyhemoglobin saturation, oral and nasal
Wisconsin who were 30–60 years were mailed a survey con- airflow, nasal air pressure, and thoracic cage and abdominal
taining questions about sleep habits, health, and demo- respiratory motion were used to detect SDB events. Oxyhe-
graphic characteristics in 1988. Of the 6,947 state employees moglobin saturation was recorded using a pulse oximetry
who received the survey, 5,091 (73%) completed and device (Ohmeda 3740, Englewood, Colorado). Thermocou-
returned it. From these respondents, a sampling frame was ples (ProTec, Hendersonville, Tennessee) was used to detect
constructed from which a stratified random sample of 2,884 airflow. A pressure transducer (Validyne Engineering Corp.,
persons (nonpregnant and without recent airway cancer, Northridge, California) measured air pressure at the nares.
airway surgery, or decompensated cardiopulmonary disease) Respiratory inductance plethysmography (Respitrace, Ambu-
was selected for invitation to participate in overnight in- latory Monitoring, Ardsley, New York) was used to record
laboratory polysomnography (sleep) studies to be repeated at thoracic and abdominal excursions. Sleep state and respira-
4-year intervals as part of the Wisconsin Sleep Cohort. Data tory event scoring were performed by trained sleep tech-
for this report were collected between 1988 and June 2011 nicians. Each 30-second epoch of the polysomnographic
from 1,520 eligible participants with at least 1 adequate records was scored for sleep stage using conventional

Am J Epidemiol. 2013;177(9):1006–1014

Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest

on 05 April 2018
 1008 Peppard et al.

criteria (26) and for breathing events. Cessation of airflow groups. Individual subjects could meet none, some, or all of
lasting 10 seconds or longer defined an apnea event. A dis- the 4 definitions.
cernible reduction in the sum of thoracic plus abdomen Extrapolation of Wisconsin Sleep Cohort Study estimates
respiratory inductance plethysmography amplitude associ- to the US population. We estimated the prevalence of people in
ated with a 4% or more reduction in oxyhemoglobin satura- the United States in each BMI stratum (<25, 25–29, 30–39,
tion defined a hypopnea event. The average number of apnea and ≥40) from the entire NHANES III (28) data set (1988–
plus hypopnea events per hour of sleep were used to 1994) and from combined data from NHANES 2007–2008
determine the apnea-hypopnea index (AHI), our summary (29) and 2009–2010 (30). We used sampling weights
parameter of SDB. From 1988 to 2000, sleep studies were provide by NHANES and included only subjects who were
scored using a paper-based recording system; from 2000 on, 30–70 years of age and not pregnant. We then esti- mated US
studies were scored by sleep technicians using a computer- prevalence of SDB in age- and sex-specific strata by
screen system. All statistical modeling of SDB prevalence applying Wisconsin Sleep Cohort Study–estimated age, sex,
adjusted for the pre- and post-2000 instrument changes; this and BMI strata–specific SDB prevalence rates to NHANES-
effectively removed instrumentation-related influences on estimated age and sex strata–specific BMI distri- butions.
SDB assessments after 2000. This produced US estimates of SDB prevalence rates in each
age and sex stratum.
Statistical analyses Confidence interval estimation. A bootstrap procedure
was used to generate 95% confidence intervals for the preva-
Modeling SDB Prevalence in the Wisconsin Sleep Cohort lence estimates: 999 samples of subject identifiers (for Wis-
Study. SAS software, version 9.2 (SAS Institute, Inc., Cary, consin Sleep Cohort subjects) or NHANES clusters (using
North Carolina) and SUDAAN software, version NHANES strata and population sampling units reweighted
10.0.1 (Research Triangle Institute, Research Triangle Park, using the Rao-Wu rescaling bootstrap method (31)) were
North Carolina) were used for analyses. To efficiently use all selected using the SAS Surveyselect procedure, version 9.2.
available data, up to 6 sleep studies per subject were used for For each of the 999 samples, stratum-specific prevalence
prevalence estimation. To produce robust standard errors, estimates were calculated as described above. The 2.5th and
generalized estimating equation (27) logistic models 97.5th percentiles for each estimate were used to construct
(SUDAAN logistic procedure) were used to regress the confidence intervals. These confidence intervals reflected
absence or presence of 4 distinct definitions of SDB on age, the variability in both the strata-specific estimates of SDB
sex, and BMI. Because SDB prevalence is highly dependent prevalence from the Wisconsin Sleep Cohort and in the esti-
on age, sex, and BMI, and to better accommodate applica- mates of the proportions of the US population represented
tion of SDB prevalence estimates from the Wisconsin Sleep by each age, sex, and BMI stratum for both examined time
Cohort to adult populations with varying age, sex, and BMI periods. The confidence intervals also accounted for the
distributions, we chose the largest number of strata for which varying number of observations (sleep studies) per subject
we could measure age × sex × BMI strata–specific pre- by bootstrap sampling of subjects rather than individual
valence estimates with acceptable precision. BMI was cate- sleep studies.
gorized into 4 strata (<25, 25–29, 30–39, and ≥40) and age
was categorized into 2 strata (30–49 and 50–70 years). For
the regression modeling, BMI strata were assigned the strata RESULTS
mean sample values (23.0, 27.6, 33.9, and 45.4, respec-
tively). Age, sex, BMI, and BMI squared, as well as pairwise Key descriptive statistics for 1,520 study participants
and higher-order interactions among these factors, were (96% non-Hispanic white) who were assessed for SDB at a
tested for statistical significance (2-sided P < 0.05); signifi- total of 4,563 sleep studies performed between 1988 and
cant terms were retained in the final models. Supplementary 2011 are presented in Table 1. Women made up 45% of the
models included alcohol use, cigarette smoking, and race/ sample. Most sleep studies (62%) were undertaken by older
ethnicity as covariates. However, inclusion of those factors (50–70 years of age) participants, which reflects the fact that
did not substantially affect estimates of SDB prevalence many follow-up sleep studies were performed after a major-
rates and were not included in final models. ity of the baseline cohort (ages 30–60 years in 1988) aged in
Four separate regression models were fitted as described to the 50–70-year-old range. A large minority (47%) of
above, 1 for each of the 4 SDB definitions, all of which are sleep studies were provided by obese (BMI ≥30) partici-
commonly used in clinical settings or in epidemiologic pants. One third of the sleep studies identified mild to severe
studies. The 4 SDB outcome definitions are: 1) mild to SDB (AHI ≥5).
severe SDB (AHI ≥5); 2) moderate to severe SDB (AHI Modeled estimates of the prevalence of mild to severe SDB
≥15); 3) mild to severe SDB with daytime sleepiness (AHI (AHI ≥5) and of moderate to severe SDB (AHI ≥15) by sex,
≥5 and Epworth Sleepiness Scale score >10); and 4) moder- age, and BMI strata are presented in Tables 2 and 3,
ate to severe SDB with daytime sleepiness (AHI ≥15 and respectively. Analogous prevalence estimates for SDB co-
Epworth Sleepiness Scale score >10). Subjects who were occurring with daytime sleepiness (Epworth Sleepiness Scale
diagnosed with and being treated for SDB at the time of the score >10) are presented in Table 4 (mild to severe SDB with
sleep studies were classified as having an AHI >15. Note that sleepiness) and Table 5 (moderate to severe SDB with sleepi-
each of the 4 separately modeled SDB definitions is binary, ness). The prevalence of SDB was substantially higher in men,
partitioning all subjects in the sample into 1 of 2 older subjects, and subjects with higher BMIs.

Am J Epidemiol. 2013;177(9):1006–1014

Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest

on 05 April 2018
 Prevalence of Sleep-Disordered Breathing 1009

Table 1. Descriptive Characteristics of Participants, Wisconsin Table 2. Model-baseda Prevalence Estimates of Mild to Severe
Sleep Cohort Study, Wisconsin, 1988–2011 Sleep-Disordered Breathing, Wisconsin Sleep Cohort Study,
Wisconsin, 1988–2011

Characteristic %a
Body Mass Index b Estimated Prevalence of AHI c ≥5
Sex
by Age, years %d 95% CI
Male 55.1
Female 44.9 Men
Age, years 30–49
30–49 37.7 <25 7.0 5.0, 9.3
50–70 62.3 25–29.9 18.3 15.2, 21.6
30–39.9 44.6 48.7, 50.2
Body mass index b ≥40 79.5 71.1, 86.2
<25 19.5
25–29.9 33.5 50–70
30–39.9 36.2 <25 18.9 14.8, 23.8
≥40 10.8 25–29.9 36.6 32.8, 40.3
Apnea-hypopnea indexc 30–39.9 61.4 57.0, 65.5
≥40 82.8 77.1, 87.7
<5 67.3 Women
5–14 21.4 30–49
≥15 11.3 <25 1.44 0.82, 2.23
≥5 and ESS score >10 d
14.5
25–29.9 4.2 2.7, 5.8
Abbreviation: ESS, Epworth Sleepiness Scale.
a
For sex, the value is the percentage of participants (n = 1,520); 30–39.9 13.5 9.8, 17.7
for age, body mass index, and apnea-hypopnea index, the value is ≥40 43.0 33.0, 54.2
the percentage of polysomnography studies (n = 4,563; individual 50–70
participants contributed multiple observations, potentially ranging
<25 9.3 6.8, 12.3
across age, body mass index, and sleep-disordered breathing
categories). 25–29.9 20.2 16.4, 24.4
b
Weight (kg)/height (m)2. 30–39.9 41.1 35.6, 46.7
c
Events/hour. ≥40 67.9 60.6, 75.1
d
The ESS was added to Wisconsin Sleep Cohort Study protocols
in 1993 and was available from 3,623 polysomnography studies by Abbreviations: AHI, apnea-hypopnea index; CI, confidence in-
1,291 participants. terval; SE, standard error.
a
Estimated logistic regression model coefficients: intercept =
−10.3 (SE, 1.3); βSex = −1.6 (SE, 0.2; female = 1, male = 0);
βAge = 2.1 (SE, 0.5; 30–49 years of age = 0, 50–70 years of age = 1);
Two interactions on the logistic scale involving age (sex × βBMI = 0.41 (SE, 0.07; <25 = 23.0, 25–29.9 = 27.6, 30–39.9 = 33.9,
≥40 = 45.4); βBMI2= −0.003 (SE, 0.001); βAge×BMI = −0.041 (SE,
age and BMI × age) were statistically significant (P < 0.05)
0.015); and βAge×Sex = 0.82 (SE, 0.22).
in prevalence models and are reflected in the prevalence esti- b
Weight (kg)/height (m)2.
mates shown in Tables 2–5. For all definitions of SDB, there c
Events/hour.
was a stronger relationship between age and SDB in women d
Among 1,520 participants who contributed a total of 4,563 sleep
than in men. For example, as shown in Table 2, among over- studies.
weight men (BMI 25–29.9), the prevalence of mild to severe
SDB was approximately 2-fold higher in older men than in
younger men (37% vs. 18%, respectively); among over-
weight women, the prevalence was approximately 5-fold Summing over age × sex × BMI strata, we estimated the
higher in older women than in younger women (20% vs. 4%, overall prevalence of mild to severe SDB (AHI ≥5) to be
respectively). Also, for all definitions of SDB (Tables 2– 26% (95% confidence interval: 24, 28) among person 30–70
5), BMI was more strongly related to SDB preva- lence in the years of age for the recent time period (2007–2010). Simi-
younger age stratum. larly, the estimated overall prevalence of moderate to worse
Table 6 presents age- and sex-specific SDB prevalence SDB (AHI ≥15) was 10% (95% confidence interval: 8, 11).
estimates that were determined by extrapolating Wisconsin
Sleep Cohort Study age-, sex-, and BMI-specific SDB prev- DISCUSSION
alence estimates to US age, sex, and BMI distributions
estimated from 1988–1994 NHANES data and recent (2007– The ongoing obesity epidemic in the United States is
2010) NHANES estimates. All age and sex strata were likely to result in “offspring epidemics” of obesity-related
estimated to have increased prevalence rates of SDB, with conditions. Overweight and obesity are strong casual factors
younger age categories of men and women experiencing the for SDB, and in tandem with an escalation in the prevalence
largest relative prevalence increases. of obesity in the United States, the prevalence of SDB
Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest
on 05 April 2018
 Am J Epidemiol. 2013;177(9):1006–1014

Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest

on 05 April 2018
 1010 Peppard et al.

Table 3. Model-baseda Prevalence Estimates of Moderate to Table 4. Model-baseda Prevalence Estimates of Mild to Severe
Severe Sleep-Disordered Breathing, Wisconsin Sleep Cohort Study, Sleep-Disordered Breathing With Concomitant Daytime Sleepiness
Wisconsin, 1988–2011 Wisconsin Sleep Cohort Study, Wisconsin, 1993–2011

Estimated Prevalence Estimated Prevalence of

Body Mass Index b of AHI c ≥15 Body Mass Index b AHIc ≥5 and ESS Score >10
by Age, years by Age, years
%d 95% CI %d 95% CI

Men Men
30–49 30–49
<25 0.93 0.49, 1.59 <25 2.7 1.5, 4.2
25–29.9 3.8 2.5, 5.3 25–29.9 6.8 4.7, 9.1
30–39.9 16.6 12.5, 21.7 30–39.9 18.9 14.3, 24.0
≥40 55.0 41.8, 69.2 ≥40 52.9 39.3, 66.6
50–70 50–70
<25 3.6 2.0, 5.4 <25 7.7 4.9, 10.9
25–29.9 10.6 8.2, 13.0 25–29.9 13.8 11.0, 16.1
30–39.9 29.0 24.6, 33.9 30–39.9 24.9 20.7, 29.4
≥40 56.0 46.3, 65.3 ≥40 43.0 33.8, 53.9
Women Women
30–49 30–49
<25 0.18 0.07, 0.34 <25 0.49 0.19, 0.91
25–29.9 0.73 0.34, 1.23 25–29.9 1.26 0.58, 2.20
30–39.9 3.6 2.0, 5.7 30–39.9 3.9 2.1, 6.2
≥40 18.6 11.7, 27.0 ≥40 16.4 10.3, 23.8
50–70 50–70
<25 1.44 0.74, 2.52 <25 2.8 1.6, 4.4
25–29.9 4.5 3.0, 6.3 25–29.9 5.3 3.8, 7.1
30–39.9 13.9 10.4, 17.4 30–39.9 10.5 7.8, 13.2
≥40 33.5 25.7, 40.8 ≥40 20.9 15.4, 27.8

Abbreviations: AHI, apnea-hypopnea index; CI, confidence inter- Abbreviations: AHI, apnea-hypopnea index; CI, confidence inter-
val; SE, standard error. val; ESS, Epworth Sleepiness Scale; SE, standard error.
a
Estimated logistic regression model coefficients: intercept = a
Estimated logistic regression model coefficients: intercept = −9.9
−15.2 (SE, 1.8); βSex = −1.7 (SE, 0.3; female = 1, male = 0); (SE, 1.6); βSex = −1.7 (SE, 0.3; female = 1, male = 0); βAge = 2.6 (SE,
βAge = 2.7 (SE, 0.7; 0–49 years of age = 0, 50–70 years of age = 1); 0.7 ; 0–49 years of age = 0, 50–70 years of age = 1); βBMI = 0.33 (SE,
βBMI = 0.58 (SE, 0.10; <25 = 23.0, 25–29.9 = 27.6, 30–39.9 = 33.9, 0.09; <25 = 23.0, 25–29.9 = 27.6, 30–39.9 = 33.9, ≥40 = 45.4);
≥40 = 45.4); βBMI2= −0.005 (SE, 0.001); βAge×BMI = −0.059 (SE, βBMI2= −0.002 (SE, 0.001); βAge×BMI = −0.07 (SE, 0.02); and
0.021); and βAge×Sex = 0.75 (SE, 0.33). βAge×Sex = 0.70 (SE, 0.33).
b b
Weight (kg)/height (m)2. Weight (kg)/height (m)2.
c c
Events/hour. Events/hour.
d d
Among 1,520 participants who contributed a total of 4,563 sleep Among 1,291 participants who contributed a total of 3,623 sleep
studies. studies.

increases translate to millions of additional afflicted persons

among adults has increased substantially. We estimate that
currently, among adults 30–70 years of age, approximately
13% of men and 6% of women have moderate to severe SDB
(AHI ≥15). We also estimate that 14% of men and 5% of
women have an AHI ≥5 plus symptoms of daytime sleep-
iness. Either of these SDB definitions fit the Medicare crite-
ria for a positive indication of obstructive sleep apnea (32).
These prevalence estimates represent double-digit relative
percentage increases, ranging from approximately 14% up
to 55% depending on age group, sex, and SDB severity
level, as evidenced in Table 6 by comparing data from
1988–1994 with that from 2007–2010. The estimated
Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest
on 05 April 2018
 excess body mass and SDB, including increased upper
in the United States, and, consequently, incidences of seque- airway collapsibility and impaired neuromuscular control of
lae of SDB, such as cardiovascular morbidities and stroke upper airway patency due to local fat deposition (35). Excess
(1, 33, 34), cognitive decline (4), depression (7), and prema- fat deposited outside of the upper airway may also contribute
ture death (8, 9), are likely accelerated relative to levels that to breathing events via several mechanisms, including
would otherwise be expected. reduced lung volume, greater whole-body meta- bolic
Multiple mechanisms likely explain the associations of demand, and increased effort of breathing (24, 36).

Am J Epidemiol. 2013;177(9):1006–1014

Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest

on 05 April 2018
 Prevalence of Sleep-Disordered Breathing 1011

Table 5. Model-baseda Prevalence Estimates of Moderate to the prevalence of SDB (10, 16). Important advances in prev-
Severe Sleep-Disordered Breathing With Concomitant Daytime alence estimation include the fact that thousands more obser-
Sleepiness, Wisconsin Sleep Cohort Study, Wisconsin, 1988–2011 vations were used to model prevalence; prevalence estimates
Estimated Prevalence of presented in Table 6 were estimated using US distributions

Body Mass Index b AHIc ≥15 and ESS Score >10 of BMI strata and are more generalizable than are those cal-
by Age, years
%d 95% CI culated directly from the Wisconsin Sleep Cohort Study
Men
population; and the age distribution for which estimates
were calculated was broader in the current analysis (ages
30–49
30–60 years in the 1993 report vs. ages 30–70 years in the
<25 0.39 0.10, 0.87 present report). Modeled prevalence estimates for the 1988–
25–29.9 1.6 0.74, 2.66 1994 period are highly consistent with the directly calculated
30–39.9 7.7 4.7, 11.3 (not modeled) previous estimates, providing a validation of
≥40 35.9 20.2, 56.6 our present models. For example, the 1993 report found that
9% of men and 4% of women 30–60 years of age had an
50–70
AHI ≥15. Here, we estimate that in the 1988–1994 period,
<25 1.2 0.44, 2.5 6% of men 30–49 years of age and 14% of men 50–70 years
25–29.9 3.7 2.3, 5.1 of age had an AHI ≥15; the analogous 1988–1994 estimates
30–39.9 11.5 8.4, 14.8 for women were 2% (30–49 years) and 7% (50–70 years).
≥40 28.7 19.7, 38.1 However, the prevalence estimates for SDB co-occurring
Women
with sleepiness presented here are not comparable with
those estimated in 1993 because of an alternative definition
30–49
of “daytime sleepiness” used in the present study. To be
<25 0.045 0.008, 0.140 classified as having “daytime hypersomnolence” in the 1993
25–29.9 0.19 0.048, 0.44 study, subjects had to report “frequent” or “habitual” feel-
30–39.9 0.95 0.36, 1.9 ings of excessive sleepiness in addition to waking up unre-
≥40 6.0 2.8, 10.6 freshed regardless of sleep quantity, as well as having
50–70
uncontrollable daytime sleepiness that caused daytime
impairment. In the present study, we have updated the occur-
<25 0.48 0.15, 1.09
rence of daytime “sleepiness” to be defined using the
25–29.9 1.46 0.77, 2.33 Epworth Sleepiness Scale (24), an instrument in wide use in
30–39.9 4.7 3.1, 6.6 both clinical and research settings.
≥40 13.3 9.1, 18.6 Our findings benefited from important methodological
strengths, including the use of a large nonclinical sample and
Abbreviations: AHI, apnea-hypopnea index; CI, confidence laboratory-based polysomnographic SDB assessment.
interval; ESS, Epworth Sleepiness Scale; SE, standard error. Nevertheless, there are study limitations that, in aggregate,
a
Estimated logistic regression model coefficients: intercept = may result in underestimation of SDB prevalence. The base-
−15.9 (SE, 3.1); βSex = −2.2 (SE, 0.5; female = 1, male = 0); βAge = 2.7
line population was recruited out of a working sample, and it
(SE, 1.1; 0–49 years of age = 0, 50–70 years of age = 1); βBMI = 0.57
(SE, 0.17; <25 = 23.0, 25–29.9 = 27.6, 30–39.9 = 33.9, ≥40 = 45.4);
is possible that persons with very severe SDB might have
βBMI2= −0.005 (SE, 0.002); βAge×BMI = −0.07 (SE, 0.03); been less likely to be present in the sampling frame.
βAge×Sex = 1.2 (SE, 0.5). However, we included data from follow-up sleep studies
b
Weight (kg)/height (m)2. regardless of subjects’ subsequent employment status, and
c
Events/hour. thus our estimates include information from postemploy-
d
Among 1,520 participants who contributed a total of 4,563 sleep ment observations. Also, although we did not find that race/
studies. ethnicity affected our SDB prevalence estimates, our sample
lacked racial/ethnic diversity (96% non-Hispanic white).
Some studies have included large samples of multiple racial/
ethnic groups and have found that prevalence rates of SDB
vary among white, black, Asian, or Hispanic subsamples
Studies have demonstrated that weight loss has the potential (39–41). However, prevalence variations may have reflected
to reduce and sometimes eliminate SDB in overweight systematic differences in BMI distributions observed among
patients (11). However, intentional weight loss with long- the race/ethnic groups (40, 41); if so, our method of extrapo-
term weight maintenance is typically not sustained (37), and lating BMI category–specific prevalence rates from the Wis-
overweight persons with SDB in whom sufficient weight consin Sleep Cohort Study to the more representative
loss cannot be attained may require direct treatments of SDB NHANES-estimated US BMI distributions may reduce the
(e.g., positive airway pressure therapy) to mitigate SDB- impact of this limitation. Additionally, BMI is an imperfect
related outcomes. Unfortunately, like weight loss, high rates proxy for underlying parameters (e.g., fat deposition in the
of long-term adherence to positive pressure therapy have upper airway) that may be more germane to the development
been difficult to achieve (38). of SDB; thus, the relation of weight and SDB may be under-
The present analysis necessarily updates and substantially estimated. However, the use of BMI was necessary because
extends earlier Wisconsin Sleep Cohort Study estimates of US population estimates of BMI distributions are available,

Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest

on 05 April 2018
 Am J Epidemiol. 2013;177(9):1006–1014

Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest

on 05 April 2018
 1012 Peppard et al.

Table 6. Sleep-Disordered Breathing Prevalence Estimates for 4 Definitions of Sleep-Disordered Breathing in US

Adults Estimated Using US Body Mass Index Population Distributions, Wisconsin Sleep Cohort Study, Wisconsin

1988–1994, Estimated 2007–2010, Estimated

Sleep-Disordered Breathing Prevalence by Sex and Prevalence by Sex and
Definition by Age, years Age Strata Age Strata

% 95% CI % 95% CI

Men
AHIa≥5
30–49 20.0 17.2, 23.1 26.6 22.8, 30.5
50–70 38.5 34.9, 42.4 43.2 39.4, 47.4
30–70 26.4 23.9, 28.9 33.9 30.8, 37.0
AHI ≥15
30–49 6.2 4.4, 8.1 9.5 7.0, 12.1
50–70 13.9 11.5, 16.8 17.4 14.5, 20.6
30–70 8.8 7.3, 10.5 13.0 10.8, 15.2
AHI ≥5, ESS score >10
30–49 8.5 6.3, 10.8 11.7 9.0, 14.7
50–70 15.3 12.6, 17.8 17.6 14.7, 20.3
30–70 10.8 9.0, 12.6 14.3 12.0, 16.4
AHI ≥15, ESS score >10
30–49 3.1 1.8, 4.4 4.8 3.1, 6.9
50–70 5.4 4.0, 6.8 7.0 5.3, 8.9
30–70 3.8 2.9, 4.9 5.8 4.5, 7.2
Women
AHI ≥5
30–49 6.6 4.9, 8.6 8.7 6.5, 11.3
50–70 24.4 20.8, 28.2 27.8 24.0, 31.6
30–70 13.2 11.4, 15.3 17.4 15.2, 20.0
AHI ≥15
30–49 1.9 1.2, 3.0 2.7 1.7, 4.0
50–70 7.4 5.5, 9.5 9.1 6.8, 11.4
30–70 3.9 3.1, 5.0 5.6 4.4, 7.0
AHI ≥5, ESS score >10
30–49 2.1 1.2, 3.3 2.9 1.7, 4.3
50–70 6.6 5.1, 8.6 7.5 5.9, 9.7
30–70 3.8 2.9, 4.9 5.0 3.9, 6.3
AHI ≥15, ESS score >10
30–49 0.55 0.24, 0.99 0.79 0.35, 1.34
50–70 2.6 1.8, 3.6 3.2 2.3, 4.4
30–70 1.3 0.9, 1.8 1.9 1.4, 2.6

Abbreviations: AHI, apnea-hypopnea index; CI, confidence interval; ESS, Epworth Sleepiness Scale.
a
Events/hour.

whereas estimates for more pathophysiologically proximal on the stratification factors (age, sex, and BMI), the baseline
parameters are not. Finally, it is possible that subjects that AHIs were not significantly different (P > 0.2) in subjects
provided more data points (i.e., participated in a larger who only provided a single observation compared with those
number of follow-up studies) had systematically lower or who provided multiple observations; similarly, mean AHI
higher AHIs. To assess this possibility, we compared base- for the nth sleep study was not significantly different when
line mean AHIs for the 1,104 subjects who provided multi- comparing subjects who only had n sleep studies with
ple observations with those from the 416 subjects who subjects who provided more than n sleep studies (n = 2–5;
provided only a single (baseline) observation. Conditional all P > 0.4). Thus, we found no evidence that SDB severity

Am J Epidemiol. 2013;177(9):1006–1014
Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest
on 05 April 2018
 Prevalence of Sleep-Disordered Breathing 1013

was associated with the number of observations that subjects 5. Kim HC, Young T, Matthews CG, et al. Sleep-disordered
provided. breathing and neuropsychological deficits. A population-based
In the present study, we provided up-to-date estimates of study. Am J Respir Crit Care Med. 1997;156(6):1813–1819.
US SDB prevalence by age, sex, and BMI strata, and our 6. Baldwin CM, Griffith KA, Nieto FJ, et al. The association of
sleep-disordered breathing and sleep symptoms with quality of
findings suggest that prevalence rates in middle-aged adults
life in the Sleep Heart Health Study. Sleep. 2001;24(1): 96–
have risen substantially in recent decades. In addition to 105.
current SDB prevalence, our results and approach can be 7. Peppard PE, Szklo-Coxe M, Hla KM, et al. Longitudinal
used to estimate future SDB prevalence if there are further association of sleep-related breathing disorder and depression.
alterations in population distributions of overweight and Arch Intern Med. 2006;166(16):1709–1715.
obesity. This information is essential for the clinical and 8. Young T, Finn L, Peppard PE, et al. Sleep disordered
public health sectors because of the growing high population breathing and mortality: eighteen-year follow-up of the
burden of SDB, the treatability of SDB, and the myriad neg- Wisconsin sleep cohort. Sleep. 2008;31(8):1071–1078.
ative health consequences of untreated SDB. 9. Marshall NS, Wong KK, Liu PY, et al. Sleep apnea as an
independent risk factor for all-cause mortality: the Busselton
Health Study. Sleep. 2008;31(8):1079–1085.
10. Young T, Palta M, Dempsey J, et al. The occurrence of sleep-
ACKNOWLEDGMENTS disordered breathing among middle-aged adults. N Engl
J Med. 1993;328(17):1230–1235.
Author affiliations: Department of Population Health Sci- 11. Young T, Peppard PE, Gottlieb DJ. Epidemiology of
ences, School of Medicine and Public Health, University of obstructive sleep apnea: a population health perspective. Am
J Respir Crit Care Med. 2002;165(9):1217–1239.
Wisconsin–Madison, Madison, Wisconsin (Paul E. Peppard,
12. Bixler EO, Vgontzas AN, Ten Have T, et al. Effects of age on
Terry Young, Jodi H. Barnet, Mari Palta, Erika W. Hagen); sleep apnea in men: I. Prevalence and severity. Am J Respir
Department of Biostatistics and Medical Informatics, School Crit Care Med. 1998;157(1):144–148.
of Medicine and Public Health, University of Wisconsin– 13. Bixler EO, Vgontzas AN, Lin HM, et al. Prevalence of sleep-
Madison, Madison, Wisconsin (Mari Palta); and Department disordered breathing in women: effects of gender. Am J Respir
of Medicine, School of Medicine and Public Health, Univer- Crit Care Med. 2001;163(3 Pt 1):608–613.
sity of Wisconsin–Madison, Madison, Wisconsin (Khin 14. Newman AB, Foster G, Givelber R, et al. Progression and
Mae Hla). regression of sleep-disordered breathing with changes in
This work was supported by the National Heart, Lung, weight: the Sleep Heart Health Study. Arch Intern Med.
and Blood Institute (grant R01HL62252), National Institute 2005;165(20):2408–2413.
15. Peppard PE, Young T, Palta M, et al. Longitudinal study of
of Aging (grants 1R01AG036838 and R01AG14124), and
moderate weight change and sleep-disordered breathing.
the National Center for Research Resources (grant JAMA. 2000;284(23):3015–3021.
1UL1RR025011) at the National Institutes of Health. 16. Young T, Peppard PE, Taheri S. Excess weight and sleep-
We thank the following people for their valuable assis- disordered breathing. J Appl Physiol. 2005;99(4):1592–1599.
tance: Diane Austin, Laurel Finn, Amanda Rasmuson, 17. Flegal KM, Carroll MD, Ogden CL, et al. Prevalence and
Kathryn Pluff, Robin Stubbs, Nicole Salzieder, Kathy Stan- trends in obesity among US adults, 1999–2008. JAMA.
back, Mary Sundstrom, Dr. Kathryn M. Cacic, Dr. Steven 2010;303(3):235–241.
Weber, and Dr. Steven R. Barczi. 18. Ogden CL, Carroll MD, Curtin LR, et al. Prevalence of
Conflict of interest: none declared. overweight and obesity in the United States, 1999–2004.
JAMA. 2006;295(13):1549–1555.
19. Hedley AA, Ogden CL, Johnson CL, et al. Prevalence of
overweight and obesity among US children, adolescents, and
REFERENCES adults, 1999–2002. JAMA. 2004;291(23):2847–2850.
20. Flegal KM, Carroll MD, Ogden CL, et al. Prevalence and
1. Somers VK, White DP, Amin R, et al. Sleep apnea and trends in obesity among US adults, 1999–2000. JAMA.
cardiovascular disease: an American Heart Association/ 2002;288(14):1723–1727.
American College of Cardiology Foundation scientific 21. US National Center for Health Statistics. National Health and
statement from the American Heart Association Council for Nutrition Examination Survey (NHANES). Hyattsville, MD:
High Blood Pressure Research Professional Education National Center for Health Statistics; 2012. (http://www.cdc.
Committee, Council on Clinical Cardiology, Stroke Council, gov/nchs/nhanes.htm). (Accessed August 6, 2012).
and Council on Cardiovascular Nursing. J Am Coll Cardiol. 22. Young T, Palta M, Dempsey J, et al. Burden of sleep apnea:
2008;52(8):686–717. Rationale, design, and major findings of the Wisconsin
2. Shamsuzzaman AS, Gersh BJ, Somers VK. Obstructive sleep Sleep Cohort Study. Wisconsin Med J. 2009;108(5): 246–
apnea: implications for cardiac and vascular disease. JAMA. 249.
2003;290(14):1906–1914. 23. Lohman TG, Roche AF, Martorell R, eds. Anthropometric
3. Peppard PE, Young T, Palta M, et al. Prospective study of the Standardization Reference Manual. Champaign, IL: Human
association between sleep-disordered breathing and Kinetics; 1988:1–55.
hypertension. N Engl J Med. 2000;342(19):1378–1384. 24. Johns MW. A new method for measuring daytime sleepiness:
4. Yaffe K, Laffan AM, Harrison SL, et al. Sleep-disordered the Epworth Sleepiness Scale. Sleep. 1991;14(6):540–545.
breathing, hypoxia, and risk of mild cognitive impairment and 25. Johns MW. Daytime sleepiness, snoring, and obstructive sleep
dementia in older women. JAMA. 2011;306(6):613–619. apnea. The Epworth Sleepiness Scale. Chest. 1993;103(1):
30–36.

Am J Epidemiol. 2013;177(9):1006–1014

Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest

on 05 April 2018
 1014 Peppard et al.

26. Rechtscahaffen A, Kales A. A Manual of Standardized Obstructive+Sleep+Apnea+(OSA)&NCDId=226&ncdver=3&

Terminology, Techniques and Scoring System for Sleep Stages IsPopup=y&bc=AAAAAAAAIAAA&). (Accessed August 6,
of Human Subjects. National Institutes of Health publication 2012).
no. 204. Washington, DC: US Government Printing Office; 33. Yaggi HK, Concato J, Kernan WN, et al. Obstructive sleep
1968. apnea as a risk factor for stroke and death. N Engl J Med.
27. Liang K-Y, Zeger SL. Longitudinal data analysis using 2005;353(19):2034–2041.
generalized linear models. Biometrika. 1986;73(1):13–22. 34. Gottlieb DJ, Yenokyan G, Newman AB, et al. Prospective
28. US National Center for Health Statistics. National Health and study of obstructive sleep apnea and incident coronary heart
Nutrition Examination Survey (NHANES) III Series 11 Data disease and heart failure: the sleep heart health study.
Files. Hyattsville, MD: National Center for Health Statistics; Circulation. 2010;122(4):352–360.
2012. (http://www.cdc.gov/nchs/nhanes/nh3data.htm). 35. Schwartz AR, Patil SP, Laffan AM, et al. Obesity and
(Accessed August 6, 2012). obstructive sleep apnea: pathogenic mechanisms and
29. US National Center for Health Statistics. 2007–2008 National therapeutic approaches. Proc Am Thorac Soc. 2008;5(2):
Health and Nutrition Examination Survey (NHANES): Survey 185–192.
Data, Documentation, Codebooks, SAS Code. Hyattsville, 36. Peppard PE, Ward NR, Morrell MJ. The impact of obesity on
MD: National Center for Health Statistics; 2012. (http://www. oxygen desaturation during sleep-disordered breathing. Am J
cdc.gov/nchs/nhanes/nhanes2007-2008/nhanes07_08.htm). Respir Crit Care Med. 2009;180(8):788–793.
(Accessed August 6, 2012). 37. Wing RR, Phelan S. Long-term weight loss maintenance. Am
30. US National Center for Health Statistics. 2009–2010 National J Clin Nutr. 2005;82(suppl):222S–225S.
Health and Nutrition Examination Survey (NHANES): Survey 38. Sawyer AM, Gooneratne N, Marcus CL, et al. A systematic
Data, Documentation, Codebooks, SAS Code. Hyattsville, review of CPAP adherence across age groups: clinical and
MD: National Center for Health Statistics; 2012. (http://www. empiric insights for developing CPAP adherence
cdc.gov/nchs/nhanes/nhanes2009-2010/nhanes09_10.htm). interventions. Sleep Med Rev. 2011;15(6):343–356.
(Accessed August 6, 2012). 39. Redline S, Tishler PV, Hans MG, et al. Racial differences
31. Rao JNK, Wu CFJ. Resampling inference with complex in sleep-disordered breathing in African-Americans and
survey data. J Am Stat Assoc. 1988;83(401):231–241. Caucasians. Am J Respir Crit Care Med. 1997;155(1):
32. Centers for Medicare and Medicaid Services, US Department 186–192.
of Health and Human Services. Decision Memo for 40. Young T, Shahar E, Nieto F, et al. Predictors of sleep-
Continuous Positive Airway Pressure (CPAP) Therapy for disordered breathing in community-dwelling adults: the Sleep
Obstructive Sleep Apnea (OSA) (CAG-00093R2). Baltimore, Heart Health Study. Arch Intern Med. 2002;162(8):893–900.
MD: Centers for Medicare and Medicaid Services; 2008. 41. Yamagishi K, Ohira T, Nakano H, et al. Cross-cultural
(http://www.cms.gov/medicare-coverage-database/details/nca- comparison of the sleep-disordered breathing prevalence
decision-memo.aspx?NCAId=204&NcaName=Continuous+ among Americans and Japanese. Eur Respir J. 2010;36(2):
Positive+Airway+Pressure+(CPAP)+Therapy+for+ 379–384.

Am J Epidemiol. 2013;177(9):1006–1014

Downloaded from https://academic.oup.com/aje/article-abstract/177/9/1006/145450 by guest

on 05 April 2018