sleep quality and the metabolic syndrome: the role of sleep duration and lifestyle

Sleep quality and the metabolic syndrome: the roleof sleep duration and lifestyle

Arthur Eumann Mesas1

Pilar Guallar-Castillón2

Esther López-García2

Luz María León-Muñoz2

Auxiliadora Graciani2

José Ramón Banegas2

Fernando Rodríguez-Artalejo2*

1Department of Public Health,Health Sciences Center,Universidade Estadual de Londrina,Londrina, PR, Brazil2Department of Preventive Medicineand Public Health, School of Medicine,Universidad Autónoma de Madrid ⁄

IdiPAZ – CIBER in Epidemiology andPublic Health (CIBERESP),Madrid, Spain

*Correspondence to: FernandoRodríguez-Artalejo, Department ofPreventive Medicine and PublicHealth, School of Medicine,Universidad Autónoma de Madrid ⁄

IdiPAZ – CIBER in Epidemiology andPublic Health (CIBERESP), Madrid,Spain.E-mail: [email protected]

Abstract

Background This study examined the association between sleep quality andthe metabolic syndrome and whether if it is independent of sleep duration andif it can be explained by lifestyles linked to sleep quality.

Methods Cross-sectional study conducted from 2008 to 2010 with 10 342individuals representative of the population aged≥18 years in Spain. Poor sleepquality was ascertained through self-reported difficulty falling asleep, difficultymaintaining sleep and sleeping pill consumption. Metabolic syndrome wasdefined according to the recent harmonized definition. Analyses were conductedwith logistic regression and adjusted for the main confounders.

Results Difficulty falling asleep was associated with higher frequency ofmetabolic syndrome after adjustment for sociodemographic variables,lifestyle and diagnosed morbidity [odds ratio (OR)=1.25; 95% confidence in-terval (CI)=1.06–1.47]. The association was slightly attenuated after furtheradjusting for sleep duration (OR=1.23; 95% CI=1.04–1.46) and held afteradditional adjustment for energy intake, adherence to a Mediterranean dietarypattern, energy spent in physical activity and time watching TV (OR=1.20;95% CI=1.01–1.42). No associations were found between metabolic syndromeand other sleep quality indicators. Difficulty falling asleepwas associatedwith highblood pressure in the fully adjusted analyses (OR=1.17; 95% CI=1.00–1.37) butnot with the rest of components of metabolic syndrome.

Conclusions Difficulty falling asleep is associated with metabolic syndromeand, in particular, with high blood pressure. This association is independent ofsleep duration and is not due to lifestyles related to poor sleep. This findingshould be replicated in prospective studies using objective sleep measures; also,the influence of antihypertensive and lipid-lowering drug treatment on this asso-ciation should be further studied. Copyright © 2013 John Wiley & Sons, Ltd.

Keywords sleep quality; metabolic syndrome; population-based study; Spain

Introduction

The metabolic syndrome is a clustering of cardiometabolic risk factors,including abdominal obesity, dyslipidemia and elevated blood pressure (BP)and glycemia [1]. Metabolic syndrome is an important health problem becauseit affects a quarter to one third of the adult population in developed countries

RESEARCH ARTICLE

Received: 17 August 2013Revised: 18 September 2013Accepted: 29 September 2013

Copyright © 2013 John Wiley & Sons, Ltd.

DIABETES/METABOLISM RESEARCH AND REVIEWSDiabetes Metab Res Rev 2014; 30: 222–231.Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/dmrr.2480

[2–4], and it is associated with an increased risk of diabetes,cardiovascular disease and all-cause mortality [5]. It istherefore important to develop strategies to prevent meta-bolic syndrome on the basis of the control of its risk factors.

Obesity, physical inactivity, sedentariness and a poor diet(high in saturated fats and sugars and low in fibre) are knownrisk factors of metabolic syndrome [6–8]. Moreover, in recentyears, accumulatedevidencehas shownthat sleepdisorders, par-ticularly obstructive sleep apnea, are associated with increasedrisk ofmetabolic syndrome [9]. It also seems that short sleep du-ration is related to metabolic syndrome and some of its compo-nents, such as obesity, hypertension or dysglycemia [10,11].

Yet the evidence on the relationship of sleep quality withmetabolic syndrome is still scarce because, to our knowl-edge, only three studies have assessed this association[12–14]; the first one found that poor global sleep qualityscores were cross-sectionally associated with metabolicsyndrome [12], and the second study observed thatdifficulty falling asleep and ‘unrefreshing’ sleep predictedmetabolic syndrome over a 3-year follow-up [13]. Thethird study found that objective indices of sleep continuity,sleep depth and sleep disordered breathing but notsubjective sleep complaints were significant correlates ofmetabolic syndrome in middle-aged women [14]. However,neither study tested if the effect of sleep quality was indepen-dent of sleep duration. This is important because poor sleepquality is associated with extreme sleep duration [15]. More-over, the previous studies did not examine whether theassociation between poor sleep quality and metabolic syn-drome could be due to lower physical activity, a more seden-tary lifestyle or inadequate diet among those with worsesleep quality. The mechanisms of the association betweensleep duration or quality andmetabolic syndrome are insuffi-ciently understood, but it has been suggested that lifestylesrelated to sleep disturbances may exert a key role [11].

Accordingly, in a large sample representative of the adultpopulation of Spain, we assessed the association betweensleep quality and metabolic syndrome while accounting forthe effect of sleep duration.We also testedwhether this asso-ciation could be explained by lifestyles linked to sleep quality.

Methods

Study design and participants

The data were taken from the Study on Nutrition andCardiovascular Risk in Spain (ENRICA), whose methodshave been reported elsewhere [16]. In brief, this is a cross-sectional study conducted from June 2008 through October2010 in a representative sample of the noninstitutionalizedpopulation of Spain aged ≥18 years. Information was col-lected by telephone interview, a physical examination andby obtaining blood and urine samples in the households.

Study participants gave written informed consent. Thestudy was approved by the Clinical Research EthicsCommittees of the University Hospital La Paz in Madridand the Hospital Clinic in Barcelona.

Study variables

Main variablesPoor sleep quality was assessed with three self-reportedcomplaints: DFA, difficulty maintaining sleep andconsumption of sleeping pills. Participants reported thefrequency of each complaint with the following options:‘habitually’, ‘sometimes’ and ‘rarely or never’. For thiswork, a complaint was deemed to be present when itwas reported to occur habitually.

Height, weight and waist circumference were measuredtwice in each subject under standardized conditions. BPwas determined with validated automatic devices (modelOmron M6) and cuffs of three sizes according to armcircumference. Two sets of BP readings were made sepa-rated by 90 min. In each set, BP was measured three timesat 1–2 min intervals. In the analyses, BP was calculated asthe mean of at least three out of the last five readings[17]. Each study participant also provided 12-h fastingblood samples. Glucose, high-density lipoprotein choles-terol and triglycerides (TGL) were determined on theADVIA 2400 Chemistry System, Siemens. The laboratorytests were performed centrally at the Center of BiologicalDiagnosis of the Hospital Clínic in Barcelona, according tostandard procedures and appropriate quality controls.

On the basis of a recent harmonized definition [1], met-abolic syndrome was considered to be the presence of atleast three of the following risk factors: [1] abdominal obe-sity (waist circumference ≥102 cm in men or ≥88 cm inwomen), [2] high glucose (≥100 mg/dL or in diabetestreatment), [3] high BP (systolic BP ≥130 mmHg or dia-stolic BP ≥85 mmHg or in antihypertensive drug treat-ment), [4] high TGL (≥150 mg/dL), and [5] low high-density lipoprotein cholesterol (<40 mg/dL in men or<50 mg/dL in women).

Potential contributors to the association between sleepquality and metabolic syndromeInformation on total sleep duration was obtained with thequestion, ‘How many hours do you usually sleep per day(including sleep at night and during the day)?’ Partici-pants were asked to report the number of hours andminutes they slept. Given that both short and long sleepduration are associated with worse quality of life andhigher mortality [18,19], sleep duration was classifiedinto the following three categories: <7, 7–8, and >8 hper day.

Habitual food consumption in the previous year wasassessed with a computerized diet history developed from

Sleep Quality and the Metabolic Syndrome 223

Copyright © 2013 John Wiley & Sons, Ltd. Diabetes Metab Res Rev 2014; 30: 222–231.DOI: 10.1002/dmrr

the one used in the EPIC-Spain cohort study [20,21].Food consumption was converted into energy and nutri-ents using standard composition tables [16]. Accordanceof food consumption with the Mediterranean diet wasevaluated with the Mediterranean Diet AdherenceScreener (MEDAS) [22]. This instrument consists of 12items with targets for food consumption, and anothertwo items with targets for food intake habits characteristicof the Mediterranean diet in Spain. One point is given forachievement of each target. The MEDAS score rangesfrom 0 to 14, so that a higher score indicates betteraccordance with the Mediterranean diet.

Physical activity was assessed in leisure time and athome using a validated questionnaire developed for theEPIC study [23], which allowed for estimating energyexpenditure in MET-hour/week [24]. Lastly, as an indica-tor of sedentary lifestyle, we asked participants to reportthe average time (hours per week) they spent watchingTV in a typical week over the last year.

Other variablesWe collected data on other variables that might be relatedto sleep quality and metabolic syndrome. In particular, weobtained sociodemographic data such as age, sex, educa-tion and occupation-based social class (nonmanual, manualworkers). We also asked about tobacco smoking (never, for-mer and current smokers); regular alcohol consumption(never, former and current drinkers <1 time/week or≥1 time/week); binge drinking, defined as the intake of≥8 standard units of alcohol in men and ≥6 units or more inwomen, during a single drinking occasion in the last 30 days,and coffee intake (number of cups/day; one cup=150 mLof regular coffee or 30 mL of express coffee). Lastly, weobtained self-reported morbidity diagnosed by a physician(depression, diabetes, ischemic heart diseases, stroke andcancer at any site) and the use of antihypertensive (withfocus on beta blockers and diuretics) or lipid-lowering drugs.

Statistical analysis

The analyses were conducted with the 10 342 individualswith complete information on study variables. The associ-ations between indicators of sleep quality (DFA, difficultymaintaining sleep and use of sleeping pills) and metabolicsyndrome were summarized with odds ratios (OR) andtheir respective 95% confidence intervals (CIs) obtainedfrom logistic regression. We built several models with pro-gressive adjustment for potential confounders or media-tors of the study association. The first model adjustedfor sociodemographic variables; the second one addition-ally adjusted for lifestyles, including tobacco, alcohol andcoffee consumption; and the third one was further adjustedfor comorbidity. To assess whether the study association

was independent of sleep duration, the fourth model wasadditionally adjusted for this variable. Lastly, we examinedthe role of possible mediators of the association betweensleep quality and metabolic syndrome; thus, the fifth modelfurther adjusted for energy intake and the MEDAS score,and the sixth model was additionally adjusted for energyspent in physical activity and time watching TV. Given thatbeta blockers and diuretics may increase glucose levels andbecause lipid-lowering drugs may decrease TGL (therebyaffecting metabolic syndrome), we ran a final model withfurther adjustment for these drug treatments.

To test if the study associations varied with sex and age(<60,≥60 years), we used likelihood ratio tests to comparemodels with and without interaction terms (products of theage or sex categories by the sleep quality indicators).Because none of the interactions were statistically signifi-cant (two-tailed p> 0.05), we present results for the wholestudy sample.

Statistical analyses were performed with the surveyprocedure in STATA, version 11.1 (College Station, TX:StataCorp LP, 2009).

Results

In total, 14.8% of individuals reported having DFA, 15.9%difficulty maintaining sleep and 6.9% using sleeping pills.As compared with those with no sleep complaints, thosewho reported sleep complaints were older, morefrequently women, had lower educational level and socialclass, less frequently consumed tobacco and alcohol andhad a lower coffee intake. They also showed a higherprevalence of diagnosed depression, diabetes, ischemicheart disease and cancer and a higher use of antihyper-tensive and lipid-lowering drug treatment. Lastly, theyreported shorter sleep duration and lower energy intakeand spent more time watching TV (Table 1).

A total of 2297 (22.2%) individuals had metabolic syn-drome. Table 2 shows that DFA was associated with higherfrequency of metabolic syndrome after adjustment forsociodemographic variables, lifestyle and diagnosed mor-bidity (OR 1.25; 95% CI 1.06–1.47). The association wasslightly attenuated after further adjusting for sleep dura-tion (OR 1.23; 95% CI 1.04–1.46). DFA remained associ-ated with metabolic syndrome even after additionaladjustment for energy intake, the MEDAS score, energyspent in physical activity and time watching TV (OR 1.20;95% CI 1.01–1.42). However, the association was some-what reduced and lost statistical significance in analyseswith further adjustment for antihypertensive and lipid-low-ering drug treatment (Table 2).We found no association be-tween difficulty maintaining sleep and metabolicsyndrome. Moreover, the use of sleeping pills was

224 A. E. Mesas et al.


associated with metabolic syndrome only in analyses ad-justed for sociodemographic variables, smoking, alcoholand coffee consumption (Table 2).

Of note is that themagnitude of the association did not varyacross categories of sleep duration (p for interaction=0.83),although it remained statistically significant only for thosesleeping 7–8 h/day (Table 3). Given that, in this group ofindividuals, the association between DFA and metabolicsyndrome lost statistical significance after adjustment forantihypertensive and lipid-lowering drug treatment, we strat-ified the analyses according to thesemedications (Table 4). Ineach stratum, the OR of metabolic syndrome associated withDFAwas higher than 1 in the fully adjustedmodels, althoughit did not reach statistical significance (Table 4).

To reduce the number of statistical comparisons, theassociations between sleep quality and the components ofmetabolic syndrome were examined only for DFA, because

it was the single indicator of poor sleep associated withmetabolic syndrome. Table 5 shows that DFA was associ-ated with high BP in the analyses adjusted for mostvariables, including sleep duration, diet, physical activityand time watching TV (OR 1.17; 95% CI 1.00–1.37). TheOR for the associations with the rest of the components ofmetabolic syndrome were higher than 1, although theywere substantially attenuated after adjusting for potentialconfounders and did not reach statistical significance(Table 5).

Discussion

This study shows that DFA is associated with metabolicsyndrome and, in particular, with high BP. This

Table 1. Characteristics of study participants by sleep quality indicators (n=10 342)

Difficulty falling asleepDifficulty

maintaining asleep Sleeping pill consumption

No(n=8815)

Yes(n=1528)

No(n=8703)

Yes(n=1639)

No(n=9626)

Yes(n=716)

Women (%) 47.2 65.2 48.0 60.0 48.2 72.3Age (years), mean (SE) 45.2 (0.3) 51.1 (0.5) 44.6 (0.3) 53.8 (0.5) 45.0 (0.3) 59.6 (0.6)Education (%)Primary or lower 26.0 42.1 25.2 45.2 26.5 54.4Secondary 43.2 38.0 44.1 34.0 43.5 29.3University 30.8 19.9 30.7 20.8 30.0 16.3

Social class based on occupation (%)I–II (nonmanual workers) 66.0 60.3 65.8 62.1 65.4 62.1II–IV (manual workers) 34.0 39.7 34.2 37.9 34.6 37.9

Smoking (%)Never a smoker 46.7 51.1 46.4 52.0 46.5 57.7Former smoker 25.4 21.6 24.3 27.8 25.1 21.7Current smoker 27.9 27.3 29.3 20.2 28.4 20.6

Alcohol consumption (%)Never drinkers 19.3 32.1 19.7 29.1 19.9 38.8Former drinkers 5.7 7.6 5.8 7.3 5.7 9.9Average drinkers <1 time/week 28.4 23.5 28.6 22.6 28.2 20.0Average drinkers ≥1 time/week 46.6 36.7 45.9 41.0 46.2 31.3

Binge drinking in the last montha (%) 6.1 4.3 6.1 4.6 6.1 2.1Coffee intake (cups/day), mean (SE) 1.31 (0.02) 1.22 (0.05) 1.31 (0.03) 1.21 (0.05) 1.31 (0.02) 1.03 (0.06)Comorbidity (%)Depression 4.4 15.9 4.7 13.3 3.8 36.0Diabetes 5.9 8.9 5.8 9.1 5.9 11.9Ischemic heart disease 0.5 0.9 0.5 0.9 0.5 1.6Cerebrovascular disease 0.2 1.1 0.2 1.1 0.3 1.0Cancer at any site 0.7 1.6 0.7 1.3 0.8 1.6

Antihypertensive drug use (%) 13.2 21.7 12.7 23.9 13.2 31.3Lipid-lowering drug use (%) 9.3 15.2 8.9 16.6 9.1 24.8Sleep durationb (hours), mean (SE) 7.38 (0.01) 6.50 (0.04) 7.38 (0.01) 6.58 (0.04) 7.28 (0.01) 6.89 (0.06)Total energy intake (kcal/d), mean (SE) 2,210 (10) 2,060 (20) 2,210 (10) 2,070 (20) 2,200 (10) 1,970 (30)Mediterranean diet accordance, MEDAS score,mean (SE)

6.33 (0.03) 6.36 (0.05) 6.28 (0.03) 6.60 (0.05) 6.32 (0.03) 6.48 (0.07)

Leisure time and household physical activity,METs-h/week, mean (SE)

68.7 (0.6) 71.1 (1.3) 68.5 (0.6) 71.8 (1.3) 69.1 (0.6) 68.6 (1.9)

Time watching TV, h/week, mean (SE) 12.9 (0.1) 16.1 (0.3) 12.9 (0.1) 15.9 (0.3) 13.0 (0.1) 18.1 (0.5)

MEDAS, Mediterranean Diet Adherence Screener (score range: 0–14); SE, standard error.aIntake of ≥8 standard units of alcohol in men and ≥6 in women, during a single drinking occasion in the last 30 days.bSelf-reported habitual hours of sleep per 24-h period.



Table

2.Associationbetwee

nslee

pqualityindicators

andth

emetab

olic

syndrome(n

=10

342)

Diffi

cultyfalling

asleep

Diffi

cultymaintaining

asleep

Slee

ping

pillco

nsum

ption

No(n

=88

14)

Yes(n

=15

28)

No(n

=87

03)

Yes(n

=16

39)

No(n

=96

26)

Yes(n

=71

6)

Metab

olic

synd

rome

%(95%

CI)

21.0

(19.9–

22.1)

29.0

(26.2–

31.8)

20.7

(19.6–

21.8)

29.8

(27.2–

32.4)

21.2

(20.2–

22.2)

34.9

(30.8–

39.1)

OR(95%

CI)

adjusted

forsex,

age,

educ

ationa

llevel

andoc

cupa

tion

-based

social

class

Ref.

1.31

(1.12–

1.53

)**

Ref.

1.15

(0.99–

1.33

)Re

f.1.27

(1.04–

1.54

)*

OR(95%

CI)ad

dition

allyad

justed

forsmok

ing,

alco

hol

intake

,binge

drinking

andco

ffee

intake

Ref.

1.28

(1.09–

1.49

)**

Ref.

1.15

(0.99–

1.33

)Re

f.1.23

(1.01–

1.50

)*

OR(95%

CI)ad

dition

ally

adjusted

forco

morbidity

Ref.

1.25

(1.06–

1.47

)**

Ref.

1.16

(0.99–

1.36

)Re

f.1.16

(0.93–

1.45

)OR(95%

CI)ad

dition

ally

plus

adjusted

for

slee

pdu

ration

Ref.

1.23

(1.04–

1.46

)*Re

f.1.14

(0.97–

1.35

)Re

f.1.15

(0.92–

1.44

)

OR

(95%

CI)

addition

ally

adjusted

foren

ergy

intake

andMED

ASscore

Ref.

1.23

(1.04–

1.46

)*Re

f.1.15

(0.97–

1.35

)Re

f.1.15

(0.92–

1.44

)

OR(95%

CI)ad

dition

ally

adjusted

forph

ysical

activity

andtimewatch

ingTV

Ref.

1.20

(1.01–

1.42

)*Re

f.1.13

(0.96–

1.33

)Re

f.1.11

(0.88–

1.39

)

OR(95%

CI)ad

dition

allyad

justed

foran

tihy

perten

sive

orlip

id-lo

weringdrug

trea

tmen

tRe

f.1.16

(0.97–

1.39

)Re

f.1.09

(0.92–

1.30

)Re

f.1.02

(0.81–

1.29

)

CI,co

nfide

nceinterval;O

R,od

dsratio;

MED

AS,

Med

iterrane

anDietAdh

eren

ceScreen

er(sco

rerang

e:0–

14).

* p<0.05

.**p<0.01

.Fu

llde

scriptionof

adjusted

variab

les(categ

ories):sex

(male,

female),a

ge(18–

29,3

0–44

,45–

64,≥

65years),e

ducation

allevel(prim

aryan

dlower,secon

dary,u

niversity),o

ccup

ation-

basedsocial

class(non

man

ual,man

ualw

ork),s

mok

ing(never

asm

oker,former

smok

er,c

urrent

smok

er),alco

holintak

e(abs

temious,former

drinke

r,<1time/wee

k,≥

1time/wee

k),

bing

edrinking

(no,

yes),coffeeintake

(num

bero

fcup

s/da

y;on

ecu

p=15

0mL,co

ntinuo

us),slee

pdu

ration

(<7,

7–8,

>8h/24

-hpe

riod

),totalene

rgyintake

(kcal/d

,con

tinu

ous),M

EDAS

score(ran

ge:0

–14

points,c

ontinu

ous),leisure

timeph

ysical

activity

[qua

rtile

s(M

ETS/h/wee

k),c

ontinu

ous],tim

ewatch

ingTV

(h/w

eek,

continuo

us),de

pression

(no,

yes),d

iabe

tes(no,

yes),ische

mic

heartdisease(no,

yes),c

ereb

rovascular

disease(no,

yes),c

ancerat

anysite

(no,

yes),a

ntihyp

ertensivedrug

use(no,

yes)

andhy

poch

olesterolemic

drug

use(no,

yes).



association is independent of sleep duration and is notexplained by lifestyles related to poor sleep quality; specif-ically, it is not due to the longer time watching TV associ-ated with DFA.

Our results are only partly comparable with those fromthe three previous studies in this field. In the first one,conducted among 210 middle-aged individuals, a worsescore on the Pittsburgh Sleep Quality Index (PSQI) wasassociated with higher odds of having metabolicsyndrome and larger waist circumference and higherglucose level [12]. However, the association lost statisticalsignificance when age-adjusted and sex-adjusted analyseswere further adjusted for depression. Moreover, the PSQIincludes a number of components of sleep quality in addi-tion to sleep duration, so that it is not possible to know the

specific sleep disturbances associated with metabolic syn-drome [12]. In a subsequent work, the same group of re-searchers examined the prospective association between anumber of sleep quality indicators (DFA, difficulty stayingasleep, frequent awakening from sleep, unsound sleepand unrefreshing sleep) and metabolic syndrome in 812middle- aged subjects (36% African American, 67%women) [13]. Only DFA and unrefreshing sleep were as-sociated with metabolic syndrome, although the associa-tion lost statistical significance when adjusted for thenumber of metabolic syndrome components present atbaseline. Moreover, neither DFA nor unrefreshing sleeppredicted any of the individual components of metabolicsyndrome [13]. Finally, a third study on 370 middle-agedwomen investigated the cross-sectional association of

Table 3. Association between difficulty falling asleep and the metabolic syndrome, by sleep duration (n=10 342)

Difficulty falling asleep

No (n=8815) Yes (n=1528) p for interaction

Sleep duration (hours per 24-h period) 0.83<7 h/day (n=2573)

% (95% CI) 25.5 (23.0–27.9) 30.7 (27.0–34.4)OR (95% CI) adjusted for sex, age, educational level andoccupation-based social class

Ref. 1.24 (0.98–1.58)

OR (95% CI) additionally adjusted for smoking, alcohol intake,binge drinking and coffee intake

Ref. 1.24 (0.98–1.58)

OR (95% CI) additionally adjusted for comorbidities Ref. 1.22 (0.94–1.57)OR (95% CI) additionally adjusted for total energy intake andMEDAS score

Ref. 1.22 (0.95–1.57)

OR (95% CI) additionally adjusted for physical activity and timewatching TV

Ref. 1.17 (0.91–1.52)

OR (95% CI) additionally adjusted for antihypertensive orlipid-lowering drug treatment

Ref. 1.18 (0.90–1.54)

7–8 h/day (n=6613)% (95% CI) 18.9 (17.7–20.2) 26.5 (22.3–30.8)OR (95% CI) adjusted for sex, age, educational level andoccupation-based social class

Ref. 1.39 (1.09–1.78)**


Ref. 1.34 (1.04–1.72)*

OR (95% CI) additionally adjusted for comorbidities Ref. 1.29 (1.01–1.66)*OR (95% CI) additionally adjusted for total energy intake andMEDAS score

Ref. 1.30 (1.01–1.67)*


Ref. 1.28 (1.00–1.65)*


Ref. 1.19 (0.92–1.55)

>8 h/day (n=1156)% (95% CI) 24.9 (21.9–27.8) 31.6 (21.1–42.1)OR (95% CI) adjusted for sex, age, educational level andoccupation-based social class

Ref. 1.25 (0.73–2.14)


Ref. 1.23 (0.72–2.08)

OR (95% CI) additionally adjusted for comorbidities Ref. 1.29 (0.74–2.25)OR (95% CI) additionally adjusted for total energy intake andMEDAS score

Ref. 1.30 (0.75–2.27)


Ref. 1.21 (0.67–2.18)


Ref. 1.20 (0.67–2.14)

CI, confidence interval; OR, odds ratio; MEDAS, Mediterranean Diet Adherence Screener (score range: 0–14); SE, standard error.*p<0.05.**p<0.01.Adjustment variables are the same as in Table 2.



PSQI and in-home polysomnography with metabolic syn-drome. Although PSQI was not associated with metabolicsyndrome in univariate analyses, decreased sleep effi-ciency and increased non-rapid eye movement beta powerand apnea–hypopnea index were associated with meta-bolic syndrome regardless of sociodemographic, lifestyleand health status [14].

Difficulty falling asleep was the indicator of poor sleepassociated with metabolic syndrome in both our studyand a previous one [13]. One possible reason is that itreflects underlying sleep disturbances more precisely thanother variables, such as difficulty maintaining sleep,which may also result from stressful life events thatproduce early awakening and inability to go back to sleep.Also, consumption of sleeping pills might reflect a combi-nation of sleep problems, mental health disorders andaddiction to the medication itself, which leads to continu-ous use. Moreover, previous research has observed thatDFA was associated with increased risk of diabetes anddeath [25,26].

In any case, sleep quality should be assessed by a set ofindicators rather than by a single one. Thus, it is notewor-thy that, as yet, only a few isolated indicators of sleepquality have been linked to metabolic syndrome and thatthe only study that used a syndromal definition of insom-nia found no association with metabolic syndrome [13].Moreover, in both our and previous studies, the associa-tion between indicators of sleep quality and metabolicsyndrome was of small magnitude. Therefore, more re-search is needed before concluding that sleep quality isclearly associated with metabolic syndrome.

Our study does not concur with previous research onthe individual components of metabolic syndrome associ-ated with poor sleep quality. One study observed an asso-ciation with obesity and high glucose [12], another onedid not observe association with any individual compo-nent [13] and ours found that DFA was associated withhigh BP only. We do not know the reasons for these dis-crepancies. However, it should be noted that the main un-derlying components of metabolic syndrome vary acrosspopulations. For instance, in our study, 88.9% of subjectswith metabolic syndrome had high BP; whereas in otherpopulations, waist circumference and fasting insulin arethe factors most strongly correlated with metabolic syn-drome [27]. Also, it is no surprise that DFA was notassociated with all components of metabolic syndrome.In fact, diagnosis of metabolic syndrome is based on justthree out of five cardiometabolic risk factors; moreover,metabolic syndrome is indeed a syndrome resulting fromone underlying factor that shows correlations of varyingmagnitude with its component risk factors [28].

A unique feature of our study is the simultaneous use ofinformation on sleep quality and duration. First, our studyconfirmed that poor sleep quality is associated withshorter sleep duration; specifically, those with DFA sleptalmost 1 h less than those without DFA. Second, in fullyadjusted analyses including sleep quality, sleep durationwas not associated withmetabolic syndrome. Compared withthose sleeping 7–8 h, the OR of metabolic syndrome in thosesleeping <7 h was 1.08 (95% CI 0.93–1.25) and in thosesleeping >8 h the OR was 1.11 (95% CI 0.91–1.35). Thiscontrasts with a previous report of short sleep associated

Table 4. Association between difficulty falling asleep and the metabolic syndrome among individuals sleeping 7–8 h/day,stratified by antihypertensive or lipid-lowering drug treatment (n=6613)


No (n=5973) Yes (n=640)

Sleep duration of 7–8 h/day

Users of antihypertensive or lipid-lowering drugs (n=1149)% (95% CI) with metabolic syndrome 47.2 (43.4–50.9) 54.8 (46.4–63.2)OR (95% CI) adjusted for sex, age, educational level and occupation-basedsocial class

Ref. 1.28 (0.88–1.85)

OR (95% CI) additionally adjusted for smoking, alcohol intake, bingedrinking and coffee intake

Ref. 1.18 (0.81–1.73)

OR (95% CI) additionally adjusted for comorbidities Ref. 1.12 (0.76–1.67)OR (95% CI) additionally adjusted for total energy intake and MEDAS score Ref. 1.12 (0.75–1.68)OR (95% CI) additionally adjusted for physical activity and time watching TV Ref. 1.11 (0.75–1.66)

Not users of antihypertensive or lipid-lowering drugs (n=5464)% (95% CI) with metabolic syndrome 13.5 (12.4–14.7) 15.8 (11.6–19.9)OR (95% CI) adjusted for sex, age, educational level and occupation-basedsocial class

Ref. 1.20 (0.84–1.73)

OR (95% CI) additionally adjusted for smoking, alcohol intake, bingedrinking and coffee intake

Ref. 1.18 (0.82–1.71)

OR (95% CI) additionally adjusted for comorbidities Ref. 1.17 (0.82–1.67)OR (95% CI) additionally adjusted for total energy intake and MEDAS score Ref. 1.18 (0.83–1.68)OR (95% CI) additionally adjusted for physical activity and time watching TV Ref. 1.17 (0.82–1.66)

CI, confidence interval; OR, odds ratio; MEDAS, Mediterranean Diet Adherence Screener (score range: 0–14).Adjustment variables are the same as in Table 2.



with metabolic syndrome; however, this study did not ac-count for sleep quality [10]. And third, our study shows forthe first time that the association between DFA andmetabolicsyndrome holds after adjustment for sleep duration.

Those with DFA watched TV for 3 h/week more thanthose with no DFA, although no substantial differences

were observed between the two groups in energy intake,diet quality or physical activity. In our study, DFA waslinked to metabolic syndrome even after adjustment forlifestyles that could mediate the association, in particulartime watching TV. Thus, mechanisms other than theselifestyles should account for the study association; it has

Table 5. Association between difficulty falling asleep and components of the metabolic syndrome (n=10 342)


No (n=8815) Yes (n=1528)

Central obesity (men: WC ≥102 cm; women: WC ≥88 cm)% (95% CI) 33.1 (31.8–34.4) 43.9 (41.0–46.9)OR (95% CI) adjusted for sex, age, educational level and occupation-based social class Ref. 1.15 (1.00–1.31)*OR (95%CI) additionally adjusted for smoking, alcohol intake, bingedrinking and coffee intake Ref. 1.15 (1.01–1.32)*OR (95% CI) additionally adjusted for comorbidities Ref. 1.11 (0.97–1.27)OR (95% CI) additionally adjusted for sleep duration Ref. 1.07 (0.93–1.25)OR (95% CI) additionally adjusted for total energy intake and MEDAS score Ref. 1.08 (0.93–1.25)OR (95% CI) additionally adjusted for physical activity and time watching TV Ref. 1.05 (0.91–1.22)OR (95% CI) additionally adjusted for antihypertensive or lipid-lowering drug treatment Ref. 1.03 (0.89–1.20)

High glucose level (glucose ≥100 mg/dL or in diabetes drug treatment)% (95% CI) 21.0 (19.9–22.1) 26.4 (23.7–29.1)OR (95% CI) adjusted for sex, age, educational level and occupation-based social class Ref. 1.10 (0.94–1.29)OR (95% CI) plus adjusted for smoking, alcohol intake, binge drinking and coffee intake Ref. 1.09 (0.93–1.28)OR (95% CI) plus adjusted for comorbidities Ref. 1.06 (0.89–1.26)OR (95% CI) plus adjusted for sleep duration Ref. 1.05 (0.88–1.26)OR (95% CI) plus adjusted for total energy intake and MEDAS score Ref. 1.04 (0.87–1.26)OR (95% CI) plus adjusted for physical activity and time watching TV Ref. 1.02 (0.85–1.23)OR (95% CI) plus adjusted for antihypertensive or lipid-lowering drug treatment Ref. 1.01 (0.84–1.21)

High BP (systolic BP ≥130 mmHg or diastolic BP ≥85 mmHg or in hypertension drug treatment)% (95% CI) 46.1 (44.6–47.5) 54.1 (51.1–57.1)OR (95% CI) adjusted for sex, age, educational level and occupation-based social class Ref. 1.22 (1.05–1.41)**OR (95% CI) additionally adjusted for smoking, alcohol intake, binge drinking andcoffee intake

Ref. 1.22 (1.05–1.41)**

OR (95% CI) additionally adjusted for comorbidities Ref. 1.21 (1.05–1.42)*OR (95% CI) additionally adjusted for sleep duration Ref. 1.18 (1.01–1.39)*OR (95% CI) additionally adjusted for total energy intake and MEDAS score Ref. 1.18 (1.01–1.39)*OR (95% CI) additionally adjusted for physical activity and time watching TV Ref. 1.17 (1.00–1.37)*OR (95% CI) additionally adjusted for antihypertensive or lipid-lowering drug treatment Ref. 1.12 (0.94–1.34)

High triglycerides level (triglycerides ≥150 mg/dL)% (95% CI) 17.4 (16.4–18.4) 18.9 (16.7–21.2)OR (95% CI) adjusted for sex, age, educational level and occupation-basedsocial class

Ref. 1.22 (1.02–1.45)*

OR (95% CI) additionally adjusted for smoking, alcohol intake, binge drinking andcoffee intake

Ref. 1.19 (0.99–1.42)

OR (95% CI) additionally adjusted for comorbidities Ref. 1.14 (0.96–1.37)OR (95% CI) additionally adjusted for sleep duration Ref. 1.14 (0.95–1.37)OR (95% CI) additionally adjusted for total energy intake and MEDAS score Ref. 1.13 (0.94–1.37)OR (95% CI) additionally adjusted for physical activity and time watching TV Ref. 1.11 (0.92–1.34)OR (95% CI) additionally adjusted for antihypertensive or lipid-lowering drug treatment Ref. 1.09 (0.91–1.32)

Low HDL cholesterol level (men: HDL <40 mg/dL; women: <50 mg/dL)% (95% CI) 25.7 (24.6–26.9) 28.2 (25.5–30.8)OR (95% CI) adjusted for sex, age, educational level and occupation-based social class Ref. 1.12 (0.97–1.30)OR (95% CI) additionally adjusted for smoking, alcohol intake, binge drinking andcoffee intake

Ref. 1.07 (0.93–1.24)

OR (95% CI) additionally adjusted for comorbidities Ref. 1.05 (0.91–1.22)OR (95% CI) additionally adjusted for sleep duration Ref. 1.05 (0.91–1.23)OR (95% CI) additionally adjusted for total energy intake and MEDAS score Ref. 1.05 (0.90–1.23)OR (95% CI) additionally adjusted for physical activity and time watching TV Ref. 1.03 (0.89–1.21)OR (95% CI) additionally adjusted for antihypertensive or lipid-lowering drug treatment Ref. 1.02 (0.88–1.19)

BP, blood pressure; CI, confidence interval; HDL, high-density lipoprotein; OR, odds ratio; MEDAS, Mediterranean Diet Adherence Screener(score range: 0–14); WC, waist circumference.*p<0.05.**p<0.01.Adjustment variables are the same as in Table 2.



been suggested that reduced brain glucose utilization,increased nocturnal release of cortisol, augmented day-time release of growth hormone and higher sympatheticnervous activity may contribute to the associationbetween disturbed sleep and metabolic syndrome [11].

Lastly, the association between DFA and metabolicsyndrome lost statistical significance after adjustment forantihypertensive and lipid-lowering drug treatment,which may affect some components of metabolic syn-drome. We believe, however, that these results shouldnot be overemphasized because the OR was attenuatedonly slightly (1.20 before and 1.16 after adjustment).Moreover, we found a specific association between DFAand high BP, which suggests that antihypertensive treat-ment plays only a small role in the association betweenDFA and metabolic syndrome.

This work has several strengths and limitations. Amongits strengths are the large sample size, that it is represen-tative of the adult population of an entire country andthat all laboratory determinations were performedcentrally under the appropriate quality controls; more-over, the analyses were adjusted for a large number ofpotential confounders. Among its limitations is that thecross-sectional design precludes a causal interpretationof the association between sleep quality and metabolicsyndrome. Another limitation is that both sleep qualityand duration were self-reported; however, there isevidence of a certain correlation between self-reportedand objective measures of sleep [29]. Also, we lacked in-formation on obstructive sleep apnea and loud snoring,which are associated with worse sleep quality and meta-bolic syndrome. However, it is unlikely that our resultswould be entirely because of failure to account for thesevariables. First, because obstructive sleep apnea is muchless frequent than poor sleep quality and metabolic syn-drome; in fact, the estimated frequency of sleep apnea is6% [30], whereas the frequency of symptoms of poorsleep and of metabolic syndrome is 15% and 22%,

respectively. And second, because our results also heldeven after adjustment for body mass index (data notshown), which is a correlate of sleep apnea and snoring.Finally, a further limitation is the lack of statistical powerto detect relatively small associations (e.g. OR <1.40) instratified analyses, such as those in Table 4.

In conclusion, this study confirms that symptoms ofpoor sleep quality, specifically DFA, are associated withmetabolic syndrome. The current work extends previ-ous knowledge by showing that this association isindependent of sleep duration and is not entirely be-cause of lifestyles linked to poor sleep such as longertime watching TV. However, this finding should bereplicated in prospective studies using objective sleepmeasures. Studies with larger samples and greater statis-tical power are also needed to examine the influence ofantihypertensive and lipid-lowering drug treatment onthe association between DFA and metabolic syndrome.Lastly, the mechanisms of the association between poorquality sleep and metabolic syndrome should be inves-tigated further.

Acknowledgements

Data for this work have been taken from the ENRICA study, whichwas funded by Sanofi-Aventis. Specific funding for this work wasobtained from FIS grant PI12/1166 and from the ‘Cátedra UAMde Epidemiología y Control del Riesgo Cardiovascular’. Fundershad no role in the design, implementation, analysis or interpreta-tion of the data.

Conflict of interest

No conflicts of interest or financial disclosures were reportedby the authors of this article.

References

1. Alberti KG, Eckel RH, Grundy SM,et al. Harmonizing the metabolic syn-drome: a joint interim statement ofthe international diabetes federationtask force on epidemiology andprevention; national heart, lung, andblood institute; American heart associa-tion; world heart federation; internationalatherosclerosis society; and interna-tional association for the study ofobesity. Circulation 2009; 120(16):1640–1645.

2. Fernandez-Berges D, Cabrera de Leon A,Sanz H, et al. Metabolic syndrome inSpain: prevalence and coronary risk

associated with harmonized definitionand who proposal. Darios study. RevEsp Cardiol 2012; 65(3): 241–248.

3. Ervin RB. Prevalence of metabolicsyndrome among adults 20 years of ageand over, by sex, age, race and ethnicity,and body mass index: United States,2003–2006. Natl Health Stat Report2009; 13: 1–7.

4. Riediger ND, Clara I. Prevalence of met-abolic syndrome in the Canadian adultpopulation. CMAJ 2011; 183(15):E1127–E1134.

5. Mottillo S, Filion KB, Genest J, et al. Themetabolic syndrome and cardiovascular

risk a systematic review and meta-analysis. J Am Coll Cardiol 2010; 56(14):1113–1132.

6. Lakka TA, Laaksonen DE. Physicalactivity in prevention and treatment ofthe metabolic syndrome. Appl PhysiolNutr Metab 2007; 32(1): 76–88.

7. Feldeisen SE, Tucker KL. Nutritionalstrategies in the prevention andtreatment of metabolic syndrome. ApplPhysiol Nutr Metab 2007; 32(1): 46–60.

8. Kastorini CM, Milionis HJ, Esposito K,et al. The effect of Mediterranean dieton metabolic syndrome and its compo-nents: a meta-analysis of 50 studies



and 534 906 individuals. J Am CollCardiol 2011; 57(11): 1299–1313.

9. Calvin AD, Albuquerque FN, Lopez-Jimenez F, et al. Obstructive sleepapnea, inflammation, and the metabolicsyndrome. Metab Syndr Relat Disord2009; 7(4): 271–278.

10. Hall MH, Muldoon MF, Jennings JR,et al. Self-reported sleep duration isassociated with the metabolic syndromein midlife adults. Sleep 2008; 31(5):635–643.

11. Knutson KL. Sleep duration andcardiometabolic risk: a review of theepidemiologic evidence. Best Pract ResClin Endocrinol Metab 2010; 24(5):731–743.

12. Jennings JR, Muldoon MF, Hall M, et al.Self-reported sleep quality is associatedwith the metabolic syndrome. Sleep2007; 30(2): 219–223.

13. Troxel WM, Buysse DJ, Matthews KA,et al. Sleep symptoms predict the devel-opment of the metabolic syndrome.Sleep 2010; 33(12): 1633–1640.

14. Hall MH, Okun ML, Sowers M, et al.Sleep is associated with the metabolicsyndrome in a multi-ethnic cohort ofmidlife women: the swan sleep study.Sleep 2012; 35(6): 783–790.

15. Mesas AE, Lopez-Garcia E, Leon-MunozLM, et al. The association between habit-ual sleep duration and sleep quality inolder adults according to health status.Age Ageing 2011; 40(3): 318–323.

16. Rodriguez-Artalejo F, Graciani A,Guallar-Castillon P, et al. Rationale andmethods of the study on nutrition and

cardiovascular risk in Spain (ENRICA).Rev Esp Cardiol 2011; 64(10): 876–882.

17. Leon-Munoz LM, Guallar-Castillon P,Graciani A, et al. Dietary habits of thehypertensive population of Spain: accor-dance with the dash diet and theMediterranean diet. J Hypertens 2012;30(7): 1373–1382.

18. Faubel R, Lopez-Garcia E, Guallar-CastillonP, et al. Sleep duration and health-relatedquality of life among older adults: Apopulation-based cohort in Spain. Sleep2009; 32(8): 1059–1068.

19. Mesas AE, Lopez-Garcia E, Leon-MunozLM, et al. Sleep duration and mortalityaccording to health status in older adults.JAmGeriatr Soc 2010;58(10): 1870–1877.

20. Relative validity and reproducibility ofa Diet History Questionnaire in Spain.I. Foods. Epic group of Spain. Europeanprospective investigation into cancerand nutrition. Int J Epidemiol 1997;26(Suppl 1): S91–S99.

21. Relative validity and reproducibility ofa Diet History Questionnaire in Spain.II. Nutrients. Epic group of Spain.European prospective investigationinto cancer and nutrition. Int J Epidemiol1997; 26(Suppl 1): S100–S109.

22. Schroder H, Fito M, Estruch R, et al. Ashort screener is valid for assessingMediterranean diet adherence amongolder Spanish men and women. J Nutr2011; 141(6): 1140–1145.

23. Pols MA, Peeters PH, Ocke MC, et al.Estimation of reproducibility andrelative validity of the questions in-cluded in the epic Physical Activity

Questionnaire. Int J Epidemiol 1997;26(Suppl 1): S181–S189.

24. Ainsworth BE, Haskell WL, Whitt MC,et al. Compendium of physical activities:an update of activity codes and METintensities. Med Sci Sports Exerc 2000;32(9 Suppl): S498–S504.

25. Dew MA, Hoch CC, Buysse DJ, et al.Healthy older adults’ sleep predicts all-cause mortality at 4 to 19 years offollow-up. Psychosom Med 2003; 65(1):63–73.

26. Cappuccio FP, D’Elia L, Strazzullo P,et al. Quantity and quality of sleep andincidence of type 2 diabetes: a system-atic review and meta-analysis. DiabetesCare 2010; 33(2): 414–420.

27. Li C, Ford ES. Is there a single underly-ing factor for the metabolic syndromein adolescents? a confirmatory factoranalysis. Diabetes Care 2007; 30(6):1556–1561.

28. Pladevall M, Singal B, Williams LK,et al. A single factor underlies the meta-bolic syndrome: a confirmatory factoranalysis. Diabetes Care 2006; 29(1):113–122.

29. Lockley SW, Skene DJ, Arendt J.Comparison between subjective andactigraphic measurement of sleep andsleep rhythms. J Sleep Res 1999; 8(3):175–183.

30. Duran J, Esnaola S, Rubio R, et al.Obstructive sleep apnea–hypopnea andrelated clinical features in a population-based sample of subjects aged 30 to70 years. Am J Respir Crit Care Med2001; 163(3 Pt 1): 685–689.



sleep quality and the metabolic syndrome: the role of sleep duration and lifestyle

Documents