lished CVD risk factors into consideration.
hagen Male Study, 3,323 men aged 54 to 74 years were
classified accordingto self-reported
decreased with age, with a 50 percent higher frequency of
snorers among the youngest quintile than among the oldest
(p<O.00001). Snoring, age adjusted, was positively associ
ated with tobacco smoking (p<0.O01), alcohol consumption
(p<O.0001), bodymass index
triglyceridelevel(p<O.O1),systolic blood pressure
and nearly significantly associated with diastolic blood
studieson the associationbetweensnoringand
disease (CVD) have only partly taken estab
In the Capon
snoring habit@ Eleven
The prevalenceof snoring
less physically activein
leisure time than others (p<O.Ol).The association between
self-reportedsnoring and blood pressure
other factors, including BMI, were taken into consideration.
No significantassociations were foundbetween snoring and
social class, snoring and low- or high-density lipoprotein or
betweensnoring and hypertension.
ing is associatedwith major cardiovascular
Accordingly,it is evidentthat in studies
CVD, proper controlling for the influence of potential
confounders is a sine qua non.
We concludethat snor
on snoring and
(Chest 1992; 102:1371-76)
CVD = cardiovascular
SBP=systolic blood pressure; DBPdiastollc
disease; BMI body mass index;
D uring the last decade,
mation and controlling for the influence of potential
confounders, eg, body mass index (BMI), BP, smoking,
and alcohol consumption,has been inadequate.1'9
To which extent snoring is associated
smoking habits, and physical
To obtain a betterunderstanding
and CYD, we have analyzed data from a large ongoing
a number of studies have
blood pressure(BP), and cardiovas
(CVD).'@ The associations
since validationof questionnaire
of the complex
risk factors, CVDsnoring,
The data used are derived
The study was initiated
cohort study, and included 5,249 male employees at 14 Copenhagen
public or private companies. Their mean age was 48 years (range,
40 to 59 years).@
Approximately15 years later, in 1985-1986,
established. All survivors (except
from the Copenhagen
in 1970 as a prospective
a new baseline was
were34 emigrants) traced by
*From the Department
Hvidovre University Hospital, Hvidovre, Denmark (Dr. Jennum);
the EpidemiologicalResearch Unit, Department
Medicine, Rigs-hospitalet, Copenhagen, Denmark (Drs. Hem,
Suadicani, and Gyntelbergj;and Division ofProspective
Department oflnternal Medicine C, Glostrup Hospital, Denmark
(Drs. Jennum and Rain).
Manuscript received November26, 1991; revision accepted
Reprintrequests:Dr.Jennum, DepartmentofClinical Neurophysi
ologtj,Hoidoore Hospital, DK 2650 Hvidovre, Denmark
ofNeurology and Clinical Neurophysiology,
means of the Danish
part in a clinical follow-up
percent agreed to participate and gave informed consent. Their
mean age was 63 years (range, 54 to 74 years).
The 1985-1986studytookplaceat the Divisionof Prospective
Medicine, Glostrup University Hospital. The examination corn
prised the following: (1) an interview
on a previously completedquestionnaire;
byaphysician(H.O.H.), includingmeasurementsofheight, weight,
and BP; and (3) a venous blood sample drawn for the determination
ofserumlipid and serum cotmnine values, following a fasting period
study: 3,387 men corresponding
Register and invited to take
by a physician
(2) a clinical examination
RISK FACTOR ASSESSMENT
From the Questionnaire
know or have you been told that you snore during the
night?â€• Answeroptions were
or â€œ¿?oftenor always:'
SleepingIbrtner: The question
you share your bedroom with your spouse or other
person with whom you cohabit?â€• Answer
Own Bedroom Dueto Snoring:
phrased: â€œ¿?ifyou sleepalone,
so much it would disturb others?â€•Answer options
were â€œ¿?yesâ€•or â€œ¿?no:'
Tobacco Consumption:Smokers gave information on
the number ofcigarettes, cheroots, cigars, and number
of grams of pipe tobacco smoked per day, and their
dailyuse of tobacco was estimated:
equaled1 g, one cheroot equaled 3 g and one cigar
equaled 4 g oftobacco.Also, use ofsnuffand
tobacco was reported.
Habits: The questionwas phrased:â€œ¿?Do you
â€œ¿?rarely or hardly everâ€•
was phrased: â€œ¿?Do
is it becauseyou snore
CHESTI 102 I 5 I NOVEMBER,19921371
74 Years: The Copenhagen
M.D.; Hans Ole Hem, M.D.; Poul Suadicani,
M.D. , Ph.D.
Validation of Self-ReportedSnoring
sumption was calculated from the questionnaire.
Physical Activityin Leisun@ Thne: According
questionnaire information, the men were divided into
two groups: (1) those physically active less than 4 h/
wk, and (2) those physically active 4 h or more/wk.
Consumption:Average daily alcohol con
to approximately 10 g of alco.
From the Clinical Examination
Blood I@ssure: BP was measured
with the subject seated using a manometer developed
by London School of Hygiene.tm
Body Mass Index: BMI was calculated
kilogramsper height in square meters.
on the right arm
as weight in
From the %I@nous Blood Sample
were analyzed using standardized methods.@
low-density lipoprotein was estimated from the above.
Serum Cotinine: Radioimmunoassay
of serum cotinine was performed
penhagen) using the method developed
and serum high-density
by Knight et
Other Risk Factors
and BP measurements,
150/100 mm Hg.
Social Class: According
five social classes, based on level of education and job
profile. Men in social class 1 were academics or other
well-educatedadministrators and executives.
social class 5 were unskilled and semiskilled workers.
For use in the analyses, we made a binary variable by
pooling social classes 1, 2, and 3, referred to as â€œ¿?higher
social classesâ€•and classes4 and 5, referred
Based on questionnaire
or having BP
to a system by Svalastoga,tm
by Hansen,@ the men were divided into
Vali&ition of Tobacco Use
90 ng/ml. All who reported
levels below 90 ng/ml. Among those who reported to
be previous smokersapproximately
levels above 90 ng/ml. Four percent of the population
were users of chewing tobacco or snuff and had the
highest values ofserum cotinine. Based on the cotinine
determinations, we accepted
with levels above 90 ng/ml as smokers. Self-reported
nonsmokers with levels above 90 ng/ml were reclassi
fled as smokers ifthey were not users ofchew tobacco
of those who reported
never to have smoked had
5 percent had
ers randomly sampled from the population were asked
to take part in a night recording.
nonsnorers served as control subjects. The abdominal
and thoracic respiratory movements were determined
using inductive plethysmography.3'
was determined using a microphone
placed over the larynx. The amplifiers were calibrated
(Brtiel and Kjmer,B&K 4230, Copenhagen,
The frequency band was 250
laryngeal sound was ifitered
0.3 s. The resulting data were sampled on a portable
computertogether with the respiratory
using a 12 bit analog-digital-converter
vices)with a sampling rate oflO Hz. Data were visually
analyzed on the computer
correctly sampled and ofsufficient
and peak laryngeal sound were calculated.
cordings were made during two nights. No differences
were found betweennight 1 and night 2 (mean: 44.8
[range, 29 to 63] vs 45.4 [range, 30 to 66] db [A], not
significant; peak: 84.6 [range, 35 to 93] vs 84.8 [range,
30 to 94] db [A], not significant). We used the following
peak laryngeal sound exceed 60 db (A).
of the questionnaire
was tested by
Hz. to 3,000 The
with a time constantof
ensuringthat the data were
quality. The average
35 db (A) or
In the examination of possible differences
nonsnorers, a Student's t test was applied for testing
age means of the two groups. Other risk factors were
included,one at a time, in a forward stepwise
multiple regression analysis adjusting for age with
snoring as the binary outcome variable. Probability
to-enter was for all analyses p<O.05.
includingall factors was performed
The above analyses were carried out using specific
software (SPSSPC + basic statistical software, version
The test for linear trend in the association between
snoring and age (in quintiles)@ and the relative risk
calculationsas an expression
were calculatedusing specffic
softwareversion 5).@ The
using the Mantel-Haenszel
confidencelimits as calculated
posed by Greenland and Robins.'Â°
A p value of <0.05was regarded
significant for all analyses.
in self-reportedsnorers and
A final analysis
using the same
methodas with 95 percent
half of the
50-@@@ ::::::::. â€¢¿?:
40::..:::l::.::.: :.:.:::@.:::::. :::::.:i::.::::
54-57 58-6162-64 65-6768-74
Ficuan1. Snoring frequency by age (quintiles).
Ficuax 2. Snoring frequency by smoking habits.
cHEST I 102 I 5 I NOVEMBER,19921373
beverages per week18.7
classi (percent education)49.0%50.2%NSNSLDL4
BP4 mm Hg121.7
BP,@ mm Hg73.1
with no professional
had a higher serum triglyceride
higher systolic BP (p<0.05),
hypertension was slightly but not signfficantly
in snorers than in nonsnorers.
potentialrisk factors in a forward
regressionmodel indicated which factors were most
likely to be genuinelyassociated
physical activity, serum triglyceride
BP became nonsignificant.
Snorers who had their own bedroom were signifi
candy older than other snorers, 63.5 years vs 62 years
(p<0.001),with a significantly
men in lower social classes, 45 percent vs 53 percent
(p<0.05). No other differences were found (not shown
Table 2 shows the exposure rate ratio for snoring,
denominated as relative risk, for various subdivisions
of the population. Obesity,
alcohol consumptionwere associated
percent increasedrisk of being a snorer. Men who
smoked15 g/clay, drank
week, and had a BMI 28 had an approximately
percenthigher risk of snoring than men who smoked
<15 g/day, drank <28 alcoholic beverages
and had a BMI <28;RR was 1.62 (1.37 to 1.91);
The above analyses were performed
those who reported that they slept with a partner (not
shown). There were no discrepancies
were less physically active (p<0.01),
and a nearly significant
(p = 0.07).The
BP prevalence of
of all the
with snoring. Due to
level, and systolic
Age, BMI, smoking, and
with a 20 to 30
28 alcoholic beverages per
The study has been approved by the Copenhagen
CountyEthics Committeefor Medical Research.
Some 1,670 (50.3 percent) subjects reported to be
snorers, ie, snoring always or often. Ofthose
ing, 1,653 (49.7 percent) reported to be nonsnorers,
1â‚¬, snoringnever or occasionally.
not answer the question on snoring habits. Responders
to the question were included in the study (n = 3,323).
Some 2,587 men (80 percent) reported that they slept
with a partner, and 212 men reported that they had
separatebedrooms due to snoring.
The prevalenceof snoring decreased with age (Fig
1), with a 50 percent higher snoring frequency among
the youngest quintilethan among the oldest; x2 for
linear trend was 57.934(p<0.00001).
was consistentand even slightly stronger
those who reportedthat they slept with a partner
were analyzed (not shown); x2 for linear trend was
Snoring prevalencewas signfficantly
smokinghabits (Fig2) (x2= 14.807,
Never smokers had the smallest prevalence
snorers; cigarette and mixed smokers had the highest.
The majorityof mixed smokers
cigarettes and pipe.
Table 1 showsthe distribution
factors in self-reported snorers and in self-reported
nonsnorers. Snorers were significantly
nonsnorers(p'(O.OOl),had a significantly
(p<0.001), they smokedmore
Table 1â€”Prevalence ofCardiovasctdar RiskThetorsin Relation to Occurrence of Snoring
*p values of linear regression adjusted for age.
tp values ofthe forward stepwise multiple regression.
jMean (standard deviation).
Â§Percent reporting physical leisure time activity >4 b/wk.
IPercentbelonging to social classes 1, 2, or 3 according
IlPercent receiving antihypertensive treatment and/or with blood pressure exceeding 150/100mm Hg.
to designation of occupation.
Cardiovascular RiskF@tors in Snorers (Jennum eta!)
28 kg/sq m vs rest1.29
per week vsrestTobacco
smoking15 g/day and1.32
smoking <15 g/day andalcohol
per week and bodymass
15 g/day and1.62
smoking <15 g/day andalcohol
per week and bodymass
Table 2â€”Exposure Rate Ratiofor Snoring as Calculated
by a Mantel-Haenszel Age-A4justed
Equation (with 95% Confidence
Subdivisions ofthe Population*
Relative Risk (RR)
Limits) by ilzriou.
analysis, it was shown that the association
snoring and serum triglyceride levels was the result
of the association between
and BMI, physical activity, and smoking. The associ
ation between snoring and BMI, physical activity, and
smoking has been found previously by others.7,@hl.17,18
In accordance with previous studies,'@7@Â°'2â€•7
ers had a higher BP than nonsnorers.
ments were made for relevant confounders associated
withBP, ie,BMI, alcohol
activity, and smoking, the difference
and nonsnorers disappeared.
could not support that a true independent
between snoring and BP exists, an association proposed
in some previous stu&es.'@ Since we found a specificity
of classification of snorers of only 58 percent,
negative associationmay be incorrect.
BP and snoring are in agreement
al,7.10.11 except thatthey found
age of 40 years. This age group was not included
Our finding of a much higher prevalence
among the youngest quintile compared with the oldest,
60 percent as compared with less than 40 percent,
interesting. At least two explanations seem reasonable.
First, the decrease in snoring
ing age may be the result of selection.
with the adverse CVD risk factor profile of snorers,
snorers must be at higher risk ofpremature death, and
accordinglythe relative fraction of men with a profile
associationwithlow risk of CVD
of snoring, will increaseover the years. Second,
with increasing age may be a possible explanation. As
previously shown by Bloom et al,'@with increasing
age, fewer men sleep with a partner. Self-reported
snoring habits will undoubtedly
no partner tells them that they actually snore. We
controlled for this possibility
analyses ofmen with a sleeping partner. The same age
trend was found.
The main conclusion to be drawn from this study is
that snoring is strongly correlated
CVD risk factors. Accordingly,
studies on snoring and CVD, proper controlling
the influenceof potential confounders
Accordingly, this study
Our results on
with Gislason et
a relationto BP at the
as well as a low risk
of snorers as nonsnorers
be less precise, since
with well established
it is evident thatin
is a sine qua
Christian X's Foundation,
the Danish Heart Foundation, and the Else and Mogens
This study received grants from the King
the DanishMedicalResearch Council,
1 Lugaresi E, Cirigaotta J, Coccagna
miologicdata on snoring and cardiocirculatorydisturbances.
Sleep 1980; 3:319-21
2 Norton PC, Dunn EV. Snoring as a risk factor for disease: an
C, Piana C. Some epide
Haenszel 2x 2 age-stratified analysis (with age dichotomized into
youngest and oldest half).
representsthe probabilityoutcomeof a Mantel
results of analyses
percent who reported that they had a sleeping partner.
Table 3 shows the association
and objectively determined
respiration and laryngeal sounds were recorded
their sleep. There was a highly significant
recorded by microphone(p<0.0001)
of94 percentand a specificity
of the entire cohort and the 80
snoring in 50 men whose
objectively and snoring
with a sensitivity
The finding that snorers had serum
previously(to our knowledge).
serum triglyceride level, whereas low-density lipopro
tein and high-densitylipoprotein
were similar in snorers and nonsnorers. In multivariate
Snorers had a higher
Table 3â€”Association between Self-RepOrted
*p<O.0001; sensitivity,94 percent;specificity, 58 percent.
CHESTI 102 15 I NOVEMBER.1992 1375