Sleep Medicine 12 (2011) 748–753

Contents lists available at ScienceDirect

Sleep Medicine

journal homepage: www.elsevier .com/locate /s leep

Original Article

Sleep disordered breathing in community dwelling elderly:Associations with cardiovascular disease, impaired systolic function,and mortality after a six-year follow-up

Peter Johansson a,b,⇑, Urban Alehagen a,b, Martin Ulander c,d, Eva Svanborg c,d, Ulf Dahlström a,b,Anders Broström c,e

a Department of Cardiology, Linköping University Hospital, S-58183 Linköping, Swedenb Department of Medical and Health Sciences, Division of Cardiovascular Medicine, Faculty of Health Sciences, Linköping University, S-58183 Linköping, Swedenc Department of Clinical Neurophysiology, Linköping University Hospital, S-58183 Linköping, Swedend Institution of Clinical and Experimental Medicine, Linköping University, S-58183 Linköping, Swedene Department of Nursing Science, School of Health Sciences, Jönköping University, S-551 85 Jönköping, Sweden

a r t i c l e i n f o

Article history:Received 28 April 2010Received in revised form 25 March 2011Accepted 26 March 2011Available online 2 July 2011

Keywords:Sleep apnea syndromesAgingElderly personsMortalityCardiovascular diseaseCardiac function

1389-9457/$ - see front matter � 2011 Elsevier B.V. Adoi:10.1016/j.sleep.2011.03.012

⇑ Corresponding author at: Department of CardHospital, S-58183 Linköping, Sweden. Tel.: +46 13 22

E-mail addresses: peter.johansson@aries.vokby.segen@liu.se (U. Alehagen), Martin.Ulander@lio.se (M.Svanborg), Ulf.Dahlstrom@lio.se (U. Dahlström), andb

a b s t r a c t

Introduction: Sleep disordered breathing (SDB), cardiovascular disease (CVD) and impaired cardiac func-tion are common in elderly people. We investigated the association of SDB and mortality in a communitydwelling elderly population, considering CVD and objectively measured impaired cardiac function havebeen poorly studied thus far.Aim: To investigate whether SDB is a factor that affects mortality in elderly people, with a focus on thosewith CVD and/or signs of impaired cardiac function.Methods: A prospective cohort design was used and 331 community dwelling elderly aged 71–87 yearsunderwent one-night polygraphic recordings in the subjects’ homes. CVD and systolic function wereobjectively established. Mortality data were collected after 6 years.Results: In the total population there were no significant associations between mortality and SDB. Inthose with CVD and impaired systolic function, as measured by NT-proBNP, oxygen desaturation index(ODI) P10 was associated with mortality. The hazard ratio of 3.0 (CI 95% 1.1–8.6, p = 0.03) remained sta-tistically significant after adjustments for age, gender, diabetes and plasma values of NT-proBNP.Conclusion: SDB in community dwelling elderly has no overall association to mortality irrespective ofdegree of SDB. However, hypoxic events (i.e., ODI P10) were associated with mortality in the groupwho had CVD in combination with impaired systolic function.

� 2011 Elsevier B.V. All rights reserved.

1. Introduction

Sleep disordered breathing (SDB), characterized by repeatedcessations in breathing during sleep, affects approximately 9%and 4% of middle-aged men and women, respectively, and 25% ofpersons older than 65 years [1–3]. SDB may cause nightly de-creases in oxygen saturation (i.e., intermittent hypoxia) and in-creased sympathetic activity, which might contribute tocardiovascular stress [4,5]. Large-scaled epidemiological studieshave shown SDB to be independently associated with hypertension[6–8], as well as to heart failure (HF) and stroke [9]. A number of

ll rights reserved.

iology, Linköping University2223; fax: +46 13 222224.(P. Johansson), Urban.Aleha-Ulander), evasv@inr.liu.se (E.r@imv.liu.se (A. Broström).

studies with samples recruited at sleep clinics or hospitals havealso shown SDB to be an independent predictor of mortality [10–14]. The importance of SDB cannot be generalized to the generalpopulation based on these studies. However, two recent commu-nity based studies, including only middle aged men and women,showed that SDB was independently associated with mortality[15,16]. Despite the high prevalence of SDB in the elderly its asso-ciation with mortality has not been well studied. Mant et al. fol-lowed 163 non-demented elderly village residents for 4 years[17]. Their data could not establish SDB to be independently asso-ciated with all-cause mortality. However, the study reported thatdeaths were more frequent in people with moderate/severe SDBand a history of hypertension, but due to the small sample sizethey could not test this interaction effect. Ancoli-Israel et al. fol-lowed 426 elderly community residents for almost 10 years [18].Those with severe SDB had shorter survival, but the effect onmortality was not independent. CVD combined with higher age

P. Johansson et al. / Sleep Medicine 12 (2011) 748–753 749

and pulmonary disease were independent predictors of death in aCox proportional hazards analysis. Data regarding CVD were col-lected by patient interviews but not objectively verified [18].

SDB is known to be common among people with impaired sys-tolic function and heart failure [3]. But to diagnose heart failure isdifficult, so when evaluating the association of SDB to mortality inthe elderly it is important to have objective data (e.g., Dopplerechocardiography or natriuretic peptides) regarding cardiac func-tion besides subjective symptoms/signs of the patient. To ourknowledge no such studies have been performed in an elderlycommunity-dwelling population. The aim of this study was there-fore to investigate whether SDB is a factor that affects prognosis inelderly people, with special focus on those with a CVD and/or signsof impaired cardiac function.

2. Methods

2.1. Sample

The patient population was derived from the CoroKind studyand has been described previously [3,19]. In brief, the major aimwas to evaluate the prevalence of HF among community dwellingelderly. The study took place from year 1998 to year 2000. All in-cluded subjects were aged 65–82 years and lived in a rural com-munity with 10,300 inhabitants in southeast Sweden. Allinhabitants in the age span mentioned above were invited to clin-ical and echocardiographic examinations. Of 1130 individuals, 876agreed to participate (participation rate 78%). From January 2003to June 2005 the cohort was again contacted and invited to re-newed clinical and echocardiographic examinations. A total of675 subjects agreed to participate. The reasons for not participat-ing were death (12%), having moved to nursing homes or otherparts of Sweden (3%), declination (7%) or not showing up (1%).Out of the 675 participants investigated a total of 346 subjects(participation rate 51%) also agreed to have their breathing patternduring sleep recorded. The data in the present study are from those346 participants who had both their systolic cardiac function aswell as their sleep respiratory patterns investigated. Those partic-ipating in the sleep study did not differ regarding gender, CVD,diabetes, body mass index (BMI), smoking, and plasma values ofN-terminal fragment of proBNP (NT-proBNP) from those who didnot. But sleep study participants had more respiratory disease(asthma or chronic obstructive pulmonary disease) (17% vs. 12%,p = 0.04), less hypertension (73% vs. 82%, p = 0.007) and less leftventricular ejection fraction (LVEF) <50% (17% vs. 27%, p = 0.002)compared to non-participants.

2.2. Baseline assessment

Every participant was examined by an experienced cardiologist(UA), who took the patient’s history and performed a clinical exam-ination. The examination included electrocardiogram, measure-ment of blood pressure, body mass index, and auscultation of theheart. Doppler echocardiographic (Accuson XP-128c) examinationswere performed with the patient in the left-supine position. LVEFwas determined semiquantitatively and LVEF P50% correspondedto normal systolic function, whereas LVEF <50% corresponded to animpaired systolic function. Hypertension was defined as a previousdiagnosis or a blood pressure of more than 140/90 mm Hg. Ische-mic heart disease (IHD) was defined if the participants had a his-tory of angina pectoris, previous myocardial infarction orcoronary artery by-pass surgery. TIA/stroke was defined if the par-ticipants had a diagnosis of TIA or stroke. Diabetes mellitus was de-fined as ongoing treatment for diabetes or a fasting blood glucoseconcentration P7 mmol/L. A respiratory disease was established if

the participant had a diagnosis or was undergoing treatment forchronic pulmonary disease or asthma. Blood sampling of the NT-proBNP was drawn while the patients were fasting, sitting, andafter resting for 30 min. NT-proBNP was measured using an elec-trochemiluminescence immunoassay (Elecsys 2010, Roche Diag-nostics, Mannheim, Germany).

Sleep respiratory recordings were performed unattended forone night in the participants’ home using the Embletta PortableDiagnostic System (ResMed Trollhättan, Sweden) [20,21]. Record-ings included nasal air flow pressure, posture and body position,abdominal and thoracic movements, pulse, and oxygen saturation(SaO2). Apneas, hypopneas and SDB severity were scored manuallyaccording to American Academy of Sleep Medicine Task Forceguidelines [22]. A hypopnoea was defined as P50% in reductionairflow for P10 s accompanied by P4% desaturation. Absence ofSDB was defined as an apnea–hypopnea index (AHI) <5; mildSDB was defined as an AHI 5–15; moderate SDB as AHI 15–30;and severe SDB as AHI >30. Onset and end of the probable sleepperiod were determined with the combination of the recordedbreathing pattern, posture, movements, and information aboutthe subject’s self-estimated sleep onset and morning awakening.All recordings were scored by the main researcher (PJ). The studyprotocol was approved by the ethics committee at the faculty ofhealth sciences, University of Linköping, Sweden, in accordancewith the provisions of the Helsinki declaration.

2.3. Mortality follow-up

Death certificates were collected from the National Board ofHealth and Welfare, Stockholm, Sweden. No patient was lost dur-ing follow-up.

2.4. Data analysis

Categorical variables are presented as numbers and percentagesand were analyzed using the Chi-square test or Spearman rank cor-relations. Continuous variables were analyzed using the Student’st-test, and data are presented as mean and SD or median and25th–75th quartiles. Binary logistic regression analysis was usedto calculate univariate odds ratios for mortality. The associationbetween SDB and mortality was analyzed according to the stan-dard cut-off points of SDB severity. When associations of SDB tomortality in people with different types of CVD and measures ofsystolic function were evaluated the following variables wereused: hypertension, IHD, TIA/Stroke, LVEF <50%, plasma values ofNT-proBNP > median. In these analyses presence of SDB was de-fined according to the dichotomized variables, AHI P15, or oxygendesaturation index P10 (ODI). ODI was used as a complement toAHI since studies have shown stronger associations with hemody-namic stress and inflammation for ODI [23,24]. Since no estab-lished cut-off for ODI exists ODI P10 was used. In the next stepanalyses were made to evaluate if SDB was associated with moral-ity in those with CVD alone.

Since IHD and/or TIA/Stroke both were associated with mortal-ity and in order to increase the power in these analyses the partic-ipants with IHD and/or TIA/Stroke were amalgamated into onegroup, labeled the CVD. The CVD group was divided into two differ-ent groups in relation to the presence of the respective measure ofimpaired systolic function (i.e., CVD and LVEF <50%; CVD and plas-ma values of NT-proBNP > median). Mortality rates in these groupswere evaluated in relation to the presence of SDB, as measured byAHI P15 or ODI P10. Cox proportional hazard regression analyseswere used to analyse if SDB independently was associated withmortality in those with CVD and impaired systolic function. SPSSversion 18.0 (SPSS Inc., Chicago, Illinois, USA) was used for statisti-cal computing.

750 P. Johansson et al. / Sleep Medicine 12 (2011) 748–753

3. Results

Sleep recordings were performed over a median of 12 days(interquartile range 12 days) after the clinical and echocardio-graphic examination. Of the 346 sleep recordings, 15 were lostdue to technical failure. Thus, the final study population consistedof 331 persons. Baseline characteristics of the study population aredescribed in Table 1. There was no difference in the number ofmales and females. Mean age was 78 years. Hypertension wasfound in 73% of the study population and about one-quarter haddiabetes or IHD. Impaired systolic function, defined as LVEF <50%or NT proBNP > median (262 pg/ml), was found in 17% and 52%,respectively. The mean AHI and ODI in the population was 9.8(SD 10.7) and 8.4 (SD 9.6), respectively.

During the follow-up of a mean period of 1648 days (145–2263 days), a total of 47 deaths occurred (14%). The respectivebaseline characteristic univariate odds ratios (OR) to mortalityare presented in Table 1. Those who died were at baseline signifi-cantly older, more often had diabetes, IHD, TIA/Stroke, impairedsystolic function (LVEF <50%) and higher plasma levels of NT-proB-NP (Table 1). The univariate OR:s presented in Table 2 shows thatthe different AHI or ODI cut-offs were not associated with mortal-ity. Compared to the 14% of mortality rate in the total study popu-lation, the mortality rates in the different AHI and ODI groups (i.e.,no, mild, moderate and severe SDB, AHI P15 or ODI P10) rangedbetween 13% and 17%. No associations with mortality were foundin a comparison between those with AHI 0–5 and those in the AHIgroups AHI P15 (p = 0.86) or AHI >30 (p = 0.61).

Table 1Characteristics of the studied population (n = 331) and univariate odds

Baseline assessment Total N = 331

Demographics:Male gender % (n) 49 (164)Age, mean (SD) 78 (3)BMI, median kg/m2 (25th–75th quartiles) 27 (25–30)Current smoker % (n) 4 (12)Co-morbidities:Diabetes % (n) 23 (77)Hypertension % (n) 73 (241)IHD % (n) 25 (85)Respiratory disease % (n) 16 (54)TIA/stroke % (n) 9 (29)Left ventricular ejection fraction:<50% % (n) 17 (55)NT-proBNP, pg/ml, mean (SD) 532 (840)NT-proBNP > median (262 pg/ml) % (n) 52 (173)NTproBNP–Q4 (550 pg/ml) % (n) 25 (82)SBP, mean (SD) 148 (19)DBP, mean (SD) 74 (9)

Key: BMI, Body mass index; DBP, Diastolic blood pressure; IHD, IschproBNP, N-terminal fragment brain natriuretic peptide; SBP, Systoliischemic attack/stroke.

Table 2Prevalence of sleep disordered breathing (SDB), number of deaths and usleep disordered breathing and the cut-offs for apnea–hypopnea index

Numbers % (n)

No SDB (AHI <5) 45 (148)Mild SDB (AHI 5–15) 32 (107)Moderate SDB (AHI 15–30) 16 (53)Severe SDB (AHI >30) 7 (23)Dichotomized SDB categories:AHI P15 23 (76)ODI P10 29 (97)

Key: AHI, apnea–hypopnea index; ODI, oxygen desaturation index.

Fig. 1 shows that the proportion of CVD (TIA/stroke and IHD)significantly (p = 0.05) increased with the severity of SDB. In a bin-ary logistic regression analysis AHI P15 remained associated withCVD after control for BMI and male sex (OR 1.7, p = 0.049). The AHIand ODI correlated with LVEF <50% (r = 0.23, p < 0.001; r = 0.25,p < 0.001) but not with plasma levels of NT-proBNP. Analyses werethen performed to examine if SDB was associated with mortality ina subset with CVD and/or impaired systolic function. Table 3 de-scribes SDB, as defined by the AHI P15 or ODI P10 in those withCVD, LVEF <50% or NT-proBNP > median, as having no significantimpact on mortality. Analyses with AHI >30 were also performedbut this did not change the results. Of notice, in those with ODIP10 and with NT-proBNP > median the mortality rate was 26%compared to the rate of 16% in those with ODI <10 (p = 0.14). Ina next step we analyzed if SDB (i.e., AHI P15 or ODI P10) wasassociated wit mortality in those with CVD and who also had signsof impaired systolic function (CVD + LVEF <50% or CVD + NT-proB-NP > median). We found that ODI P10 significantly increased themortality rate in those with CVD + NT-proBNP > median(p = 0.03). In those with CVD + NT-proBNP > median and ODI P10mean AHI was 22.5 (SD 9)—significantly (p < 0.001) different com-pared to the mean AHI of 5.6 (SD 4.6) found in those withCVD + NT-proBNP > median and ODI <10.

To analyse if ODI P10 was an independent predictor of mortal-ity, a Cox proportional hazard regression analysis was performed inthe CVD group who had a plasma value of NT-proBNP > median(n = 64). A total of 17 deaths were registered in this group andsince the limiting sample size in survival analyses is the number

ratios for mortality.

Mortality odds ratio (CI 95%) p

1.3 (0.7–2.4) 0.391.2 (1.1–1.3) <0.0011.5 (0.8–2.9) 0.181.2 (0.6–5.7) 0.68

2.6 (1.4–5.1) 0.031.4 (0.7–3.1) 0.332.0 (1.1–3.8) 0.031.3 (0.6–2.8) 0.572.4 (1.01–5.9) 0.04

2.2 (1.1–4.5) 0.0284.3 (2.1–8.8) <0.0012.4 (1.2–4.7) 0.0093.6 (1.9–6.9) <0.0011.0 (0.9–1.03) 0.220.98 (0.95.1.02) 0.42

emic heart disease; LVEF, Left ventricular ejection fraction; NT-c blood pressure; SD, Standard Deviation; TIA/stroke, Transient

nivariate odds ration to mortality in the different severity levels of(AHI) P15, and oxygen desaturation index (ODI) P10.

Deaths % (n) Univariate odds ratio. CI 95%, p

14 (20) REF15 (16) 1.1(0.6–2.3), 0.7413 (7) 0.97 (0.4–2.4),0.9517 (4) 1.4 (0.4–4.4), 0.62

14 (11) 1.03 (0.5–2.1), 0.9316 (16) 1.3 (0.7–2.5), 0.44

Fig. 1. Percentages of persons with cardiovascular diseases (i.e., IHD and TIA/stroke) across the severity of sleep disordered breathing (SDB). Overall Chi-squaretest v2 = 7.7, df 3, p = 0.05.

P. Johansson et al. / Sleep Medicine 12 (2011) 748–753 751

of deaths, we choose to adjust for only those variables that inTable 1 were associated with mortality (i.e., age, diabetes and log-arithmic transformed plasma values plasma values of NT-proBNP).After adjustments, ODI P10 increased the risk for mortality bythree times (HR 3.0 CI 95% 1.1–8.2, p = 0.03). The proportion of par-ticipants with a respiratory disease was not different in those withCVD + NT-proBNP > median with ODI P10 compared to those ODI<10 (14% vs. 27%, p = 0.19).

4. Discussion

In this study of 331 communities dwelling elderly SDB had nooverall association with mortality. In those with no SDB (AHI <5)the mortality rate was 14% compared to the 13% and 17% foundin the groups with moderate (AHI 15–30) or severe SDB (AHI>30). However, in those with CVD and impaired systolic function(i.e., NT-proBNP > median) SDB, as measured by an ODI P10, wefound significantly increased mortality rates after 6 years of fol-low-up. The adjusted HR showed that ODI P10 increased the riskfor mortality by three times (HR 3.0 CI 95% 1.1–8.2) in the CVDgroup with impaired systolic function. To our knowledge, this isa new finding and these results therefore add important informa-tion concerning the impact of SDB on community dwelling elderly.

No overall association between SDB and mortality in a commu-nity dwelling elderly population was found (Table 2). Other studiesfocusing on younger patients have found increased mortality rates,especially for moderate/severe SDB [15,16]. A methodologicalconcern of the present study is that all participants have beenexamined by a cardiologist and also had echocardiography at leasttwo times within 5 years [3,19], the present population may beregarded as a survivor population and, from a cardiovascularperspective, probably better treated compared to other elderly

Table 3The impact of sleep disordered breathing (SDB), in relation to different cut-offs for apnea–<50% or NT-proBNP > median and CVD + LVEF <50% or CVD + NT-proBNP > median.

AHI P15

No YesDeath % (n) Deat

CVD, (n = 103) 22 (17) 19 (5LVEF <50%, (n = 55) 21 (7) 29 (6NT-proBNP > median (262 pg/ml), (n = 173) 19 (24) 20 (9CVD + LVEF <50%, (n = 32) 28 (5) 14 (2CVD + NT-proBNP > median (262 pg/ml), (n = 64) 26 (11) 27 (6

Key: C AHI, Apnea–hypopnea index; CVD, Cardiovascular disease; LVEF, Left ventricularOxygen desaturation index.

persons. Our mortality rates could therefore be an underestimationof the true effect of SDB. At present only a few studies have inves-tigated SDB relation to mortality in community dwelling elderly,and the results presented in these studies confirm our results[17,18,25]. Our sample size is larger than the that of Mant et al.[17] (n = 163) and comparable to the study of Ancoli-Israel et al.[18] (n = 426). In Punjabi et al.’s [25] large scaled study the totalsample size was n = 6294, of which 3165 were >70 years. Interest-ingly, significant differences for mortality in those aged below70 years was found in that study. Elderly persons seem to exert lessrespiratory efforts as well as lesser oxygen desaturations in theirresponses to apneas and hypopneas [26,27]. This may explain theweak associations of SDB with BMI, as well as with subjective com-plaints of insomnia and daytime sleepiness in the elderly [2,28].Taken all together this suggest that SDB in community dwelling el-derly, compared to middle aged people, may be seen as a relativelybenign sleep disturbance. However, this does not exclude olderadults from receiving the best medical care for SDB or other sleep-ing problems.

We found that SDB was associated with CVD and with impairedsystolic function. In middle-aged sleep clinic patients free fromheart diseases SDB has been found as an independent risk factorfor developing coronary heart disease [29]. Findings from the com-munity based study by Shahar et al. [9] showed an increased OR forthe presence of self-reported CVD according the degree of AHI (25,50, and 75 percentiles of AHI 1.3, 4.4, and 11.0). This supports thefindings in our study despite that the mean age of their study’s par-ticipants was approximately 15 years younger. In contrast toShahar et al. [9], who found the highest risk of having CVD in thehighest AHI quartile, our study found that the percentage of per-sons belonging to the CVD group with severe SDB was 39% com-pared to the 44% found in those with moderate SDB. Theassociation between SDB and CVD in the present study could bethe result of a process that has evolved over many years. The pop-ulation might have been overweight or obese and therefore devel-oped SDB during their middle age, which in turn led to increasedrisk for hypertension [30], and as a consequence a higher risk fora future CVD. Hypothetically, the low number with severe SDB inthe CVD group in our study could therefore be due to higher inci-dence of premature deaths because of CVD. This could also serve asanother possible explanation for the absent association betweenSDB, BMI, and mortality in our study. Irrespective of this, we founda threefold increased risk for mortality in a subset of participantswith CVD + NT-proBNP > median and who had ODI P10. Ancoli-Israel et al. [18] found more cardiovascular deaths in elderly withrespiratory disturbance index (RDI) P15 and P30, compared tothose with RDI <15, but this association was not independent aftercontrolling for age. In the present study ODI P10 in those withCVD + NT-proBNP > median remained associated with mortalityafter adjustments for age, diabetes and plasma levels of NT-proB-NP. The number of participants with CVD + NT-proBNP > median

hypopnea index (AHI) P15, and oxygen desaturation index (ODI) P10 on, CVD, LVEF

ODI P10

No Yesh % (n) p Death % (n) Death % (n) p

) 0.50 18 (11) 28 (11) 0.32) 0.50 18 (5) 30 (8) 0.30) 0.85 16 (19) 26 (14) 0.14) 0.43 22 (3) 22 (4) 1.0) 0.93 16 (6) 41 (11) 0.03

ejection fraction; NT-proBNP, N-terminal fragment brain natriuretic peptide; ODI,

752 P. Johansson et al. / Sleep Medicine 12 (2011) 748–753

was small (n = 64) and this finding should therefore be interpretedcautiously. However, our data indicate that SDB may be a poorprognostic sign in a subset of elderly who suffer from CVD and im-paired systolic function.

Surprisingly it was the ODI and not the AHI that was associatedwith mortality in those with CVD and impaired systolic function.The effects of SDB on the cardiovascular system are complex,multi-factorial and not yet fully understood. But possible mecha-nistic explanations for this finding may be due to oxidative stressand systemic inflammation. In SDB oxidative stress and inflamma-tion occurs as a consequence of the frequent cycles of desatura-tions followed by rapid reoxygenation (i.e., intermittent hypoxia)during apneas and hypopneas [31,32]. This promotes adverse car-diovascular effects such as sympathetic activation, atherosclerosis[31,32], and systolic function impairment [33]. Yumino et al. [34]reported that SDB significantly increased mortality rates in pa-tients with ischemic HF, but not in those with non-ischemic HF.A recent study on patients with HF showed evidence for the impor-tance of hypoxia on hemodynamic stress measured by BNP [24].Changes in BNP during the night were related to the severity ofhypoxemia, but not to the AHI of the examined patients. Further-more, ODI and not AHI has been found to be an independent pre-dictor of elevated plasma levels of the pro-inflammatory cytokinetumor-necrosis-a [23]. AHI in the elderly with CVD and impairedsystolic function may therefore not always be the most relevantpathophysiological aspect of SDB.

Despite the prospective study design, causality may not bedetermined since SDB and the disease variables were measuredonly at baseline. We can therefore not evaluate whether SDB isthe cause or the effect of a CVD or impaired systolic function.Out of 675 participants in the cohort, 51% choose to participatein the sleep study. As this population was elderly (mean age78 years) and concomitant diseases were common, it is not sur-prising that several felt that the sleep study was burdensomeand choose not to participate. This is also confirmed by the factthat a larger proportion of those with LVEF <50% did not volunteerfor the sleep study. When sub-grouping our population, such as thecombination of CVD with the different measures of impaired sys-tolic function, this lead to small sample sizes which limit thepower of the analyses. Only a few studies have investigated theassociation between SDB and mortality in community dwelling el-derly [17,18,25]. An advantage of the present study, compared tothese studies, is that all participants were clinically examined bya cardiologist, including Doppler echocardiography, as well as theinclusion of biochemical marker of impaired systolic function,NT-proBNP. This strengthens the diagnostic procedures for CVDand impaired systolic function. We therefore believe that resultspresented in this study, despite our limitations, add importantinformation concerning the association between SDB and mortalityin community dwelling elderly.

In conclusion, this study showed that SDB in community dwell-ing elderly, in general may not be seen as a hazardous state, whichis in contrast to the impact of SDB in middle-aged persons. How-ever, in a subgroup of participants who had CVD combined withbiochemical signs of impaired systolic function, as measured byNT-proBNP, SDB was associated with mortality. More studies areneeded to establish the association of SDB with mortality in com-munity dwelling elderly with cardiovascular diseases and impairedsystolic function.

Conflict of interest

The ICMJE Uniform Disclosure Form for Potential Conflicts ofInterest associated with this article can be viewed by clicking onthe following link: doi:10.1016/j.sleep.2011.03.012.

Acknowledgment

Financial support was provided by the Health Research Councilin the Southeast of Sweden grant No. F2004-233 and Lisa and JohanGrönbergs foundation. Peter Johansson, Urban Alehagen, MartinUlander, Eva Svanborg Ulf Dahlström and Anders Broström haveno research grant of contract support, personal compensations orpersonal financial investments in relation to this manuscript todeclare.

References

[1] Young T, Palta M, Dempsey J, Skatrud J, Weber S, Badr S. The occurrence ofsleep-disordered breathing among middle-aged adults. N Engl J Med1993;328(17):1230–5.

[2] Young T, Shahar E, Nieto FJ, et al. Predictors of sleep-disordered breathing incommunity-dwelling adults: the Sleep Heart Health Study. Arch Intern Med2002;162(8):893–900.

[3] Johansson P, Alehagen U, Svanborg E, Dahlstrom U, Brostrom A. Sleepdisordered breathing in an elderly community-living population:relationship to cardiac function, insomnia symptoms and daytime sleepiness.Sleep Med 2009;10(9):1005–11.

[4] Malhotra A, White DP. Obstructive sleep apnoea. Lancet 2002;360(9328):237–45.

[5] Bradley TD, Floras JS. Sleep apnea and heart failure: part I: obstructive sleepapnea. Circulation 2003;107(12):1671–8.

[6] Nieto FJ, Young TB, Lind BK, et al. Association of sleep-disordered breathing,sleep apnea, and hypertension in a large community-based study. Sleep HeartHealth Study. Jama 2000;283(14):1829–36.

[7] Haas DC, Foster GL, Nieto FJ, et al. Age-dependent associations between sleep-disordered breathing and hypertension: importance of discriminating betweensystolic/diastolic hypertension and isolated systolic hypertension in the SleepHeart Health Study. Circulation 2005;111(5):614–21.

[8] Bixler EO, Vgontzas AN, Lin HM, et al. Association of hypertension and sleep-disordered breathing. Arch Intern Med 2000;160(15):2289–95.

[9] Shahar E, Whitney CW, Redline S, et al. Sleep-disordered breathing andcardiovascular disease: cross-sectional results of the Sleep Heart Health Study.Am J Respir Crit Care Med 2001;163(1):19–25.

[10] Marin JM, Carrizo SJ, Vicente E, Agusti AG. Long-term cardiovascular outcomesin men with obstructive sleep apnoea–hypopnoea with or without treatmentwith continuous positive airway pressure: an observational study. Lancet2005;365(9464):1046–53.

[11] He J, Kryger MH, Zorick FJ, Conway W, Roth T. Mortality and apnea index inobstructive sleep apnea. Experience in 385 male patients. Chest1988;94(1):9–14.

[12] Lavie P, Herer P, Peled R, et al. Mortality in sleep apnea patients: a multivariateanalysis of risk factors. Sleep 1995;18(3):149–57.

[13] Campos-Rodriguez F, Pena-Grinan N, Reyes-Nunez N, et al. Mortality inobstructive sleep apnea–hypopnea patients treated with positive airwaypressure. Chest 2005;128(2):624–33.

[14] Yaggi HK, Concato J, Kernan WN, Lichtman JH, Brass LM, Mohsenin V.Obstructive sleep apnea as a risk factor for stroke and death. N Engl J Med2005;353(19):2034–41.

[15] Young T, Finn L, Peppard PE, et al. Sleep disordered breathing and mortality:eighteen-year follow-up of the Wisconsin sleep cohort. Sleep2008;31(8):1071–8.

[16] Marshall NS, Wong KK, Liu PY, Cullen SR, Knuiman MW, Grunstein RR. Sleepapnea as an independent risk factor for all-cause mortality: the BusseltonHealth Study. Sleep 2008;31(8):1079–85.

[17] Mant A, King M, Saunders NA, Pond CD, Goode E, Hewitt H. Four-year follow-up of mortality and sleep-related respiratory disturbance in non-dementedseniors. Sleep 1995;18(6):433–8.

[18] Ancoli-Israel S, Kripke DF, Klauber MR, et al. Morbidity, mortality andsleep-disordered breathing in community dwelling elderly. Sleep 1996;19(4):277–82.

[19] Alehagen U, Goetze JP, Dahlstrom U. Reference intervals and decision limits forB-type natriuretic peptide (BNP) and its precursor (Nt-proBNP) in the elderly.Clin Chim Acta 2007;382(1–2):8–14.

[20] Christ M, Sharkova Y, Fenske H, et al. Brain natriuretic peptide for prediction ofCheyne-Stokes respiration in heart failure patients. Int J Cardiol2007;116(1):62–9.

[21] Oldenburg O, Lamp B, Faber L, Teschler H, Horstkotte D, Topfer V. Sleep-disordered breathing in patients with symptomatic heart failure: acontemporary study of prevalence in and characteristics of 700 patients. EurJ Heart Fail 2007;9(3):251–7.

[22] AASMT. Sleep-related breathing disorders in adults: recommendations forsyndrome definition and measurement techniques in clinical research. TheReport of an American Academy of Sleep Medicine Task Force. Sleep1999;22(5):667–89.

[23] Ryan S, Taylor CT, McNicholas WT. Predictors of elevated nuclear factor-kappaB-dependent genes in obstructive sleep apnea syndrome. Am J RespirCrit Care Med 2006;174(7):824–30.

P. Johansson et al. / Sleep Medicine 12 (2011) 748–753 753

[24] Gottlieb JD, Schwartz AR, Marshall J, et al. Hypoxia, not the frequency of sleepapnea, induces acute hemodynamic stress in patients with chronic heartfailure. J Am Coll Cardiol 2009;54(18):1706–12.

[25] Punjabi NM, Caffo BS, Goodwin JL, et al. Sleep-disordered breathing andmortality: a prospective cohort study. PLoS Med 2009;6(8):e1000132.

[26] Krieger J, Sforza E, Boudewijns A, Zamagni M, Petiau C. Respiratory effortduring obstructive sleep apnea: role of age and sleep state. Chest1997;112(4):875–84.

[27] Bixler EO, Vgontzas AN, Ten Have T, Tyson K, Kales A. Effects of age on sleepapnea in men: I. Prevalence and severity. Am J Respir Crit Care Med1998;157(1):144–8.

[28] Johansson P, Alehagen U, Svensson E, Svanborg E, Dahlstrom U, Brostrom A.Determinants of global perceived health in community-dwelling elderlyscreened for heart failure and sleep-disordered breathing. J Cardiovasc Nurs2010;25(5):E16–26.

[29] Peker Y, Carlson J, Hedner J. Increased incidence of coronary artery disease insleep apnoea: a long-term follow-up. Eur Respir J 2006;28(3):596–602.

[30] Peppard PE, Young T, Palta M, Skatrud J. Prospective study of the associationbetween sleep-disordered breathing and hypertension. N Engl J Med2000;342(19):1378–84.

[31] Lavie L, Lavie P. Molecular mechanisms of cardiovascular disease in OSAHS:the oxidative stress link. Eur Respir J 2009;33(6):1467–84.

[32] Ryan S, McNicholas WT. Intermittent hypoxia and activation of inflammatorymolecular pathways in OSAS. Arch Physiol Biochem 2008;114(4):261–6.

[33] Sun M, Chen M, Dawood F, et al. Tumor necrosis factor-alpha mediates cardiacremodeling and ventricular dysfunction after pressure overload state.Circulation 2007;115(11):1398–407.

[34] Yumino D, Wang H, Floras JS, et al. Relationship between sleep apnoea andmortality in patients with ischaemic heart failure. Heart 2009;95(10):819–24.