Download - Fibrinogen may mediate the association between long sleep duration and coronary heart disease

Fibrinogen may mediate the association between long sleepduration and coronary heart disease

LAUREN HALE 1 , V I C TOR I A PARENTE 2 , J ENN I F ER BEAM DOWD 3 ,MEGAN SANDS 4 , J E F FREY S . BERGER 5 , Y I Q I NG SONG 6 , L I S A W . MART I N 7

and MATTHEW A . A L L I SON 8

1Stony Brook University, Stony Brook, NY, USA, 2Children’s Health Center, Duke University, Durham, NC, USA, 3CUNY School of Public Healthat Hunter College, New York,NY, USA, 4Center for Sleep and Circadian Neurobiology, University of Pennsylvania Perelman School of Medicine,Philadelphia, PA, USA, 5Divisions of Cardiology and Hematology, New York University Medical Center, New York University, New York, NY,USA, 6Division of Preventive Medicine, Brigham Women’s Hospital, Brigham and Women’s Hospital, Boston, MA, USA, 7Cardiology, GeorgeWashington University School of Medicine and Health Sciences, George Washington University, Washington, DC, USA, 8University ofCalifornia, San Diego, La Jolla, CA, USA

Keywordscoagulation, coronary heart disease, fibrinogen,long sleep, mortality, sleep duration

CorrespondenceLauren Hale, PhD, HSC Level 3, Room 071,Stony Brook, NY 11794-8338, USA.Tel.:631 444 1007;fax : 631 444 3480;e-mail: [email protected]

Accepted in revised form 24 October 2012;received 11 July 2012

DOI: 10.1111/jsr.12020

SUMMARYLong sleep duration has been associated with increased risk ofcardiovascular disease (CVD) and all-cause mortality. Inflammationand coagulation have been hypothesized as possible physiologicalpathways to explain this association, although specific biomarkers havenot been studied. Using longitudinal data from 3942 postmenopausalwomen in the Women’s Health Initiative observational study and clinicaltrials, we investigated whether fibrinogen, an acute-phase inflammatoryprotein involved in blood clotting, mediates the associations betweensleep duration and coronary heart disease (CHD) and mortality amongwomen. Fibrinogen levels were associated positively with self-reportedlong sleep duration (9+ h per night), CHD and all-cause mortality, evenafter adjustment for a range of sociodemographic characteristics,cardiovascular risk factors and comorbidities.Compared with self-reported 7–8 h per night sleep duration, self-reported long sleep durationwas associated with increased odds of CHD [odds ratio (OR) = 2.05,95% confidence interval (CI): 1.02–4.11]. Adjustment for fibrinogen levelsreduced the increased odds of CHD associated with long sleep byapproximately 8 percentage points (OR = 1.97, 95% CI: 0.98–3.97).A similar reduction in the OR was observed with mortality. For bothoutcomes there is support for partial mediation of 6–7%, suggesting thatfibrinogen may be a mechanism through which long sleep duration isassociated with CHD and mortality.

INTRODUCTION

Long sleep duration has been associated with increasedcardiovascular disease (CVD) and all-cause mortality inwomen (Ayas et al., 2003; Cappuccio et al., 2010; Gallicchioand Kalesan, 2009; Grandner et al., 2009; Knutson andTurek, 2006; Kripke et al., 2002; Stamatakis and Punjabi,2007; Stranges et al., 2008; Tamakoshi and Ohno, 2004;Youngstedt and Kripke, 2004). The physiological pathwaysthrough which these associations occur have not beenelucidated fully (Grandner et al., 2009; Youngstedt andKripke, 2004). Inflammation and coagulation, which areinvolved in the pathophysiology of cardiovascular disease

through inflammation of arteries and clot formation onplaques, have been hypothesized as possible mediatingmechanisms, particularly among women (Miller et al., 2009;Opp, 2009; Patel et al., 2009; Simpson and Dinges, 2007;Suarez, 2008). However, as recently as, Cappuccio et al.,2010 stated that: ‘no studies published to date havedemonstrated a possible mechanism mediating the effectof long duration of sleep as a cause of morbidity andmortality’.A more recent study has shown that longer self-reported

sleep durations were associated with increased odds ofintima-media thickness (IMT) greater than 1.2 mm, even afteradjustment for demographic factors (e.g. age, sex), health

ª 2012 European Sleep Research Society 305

J Sleep Res. (2013) 22, 305–314 Extending sleep and the heart

behaviors (e.g. alcohol intake, smoking history) and otherbiomarkers (e.g. lipid profile, fasting plasma glucose) (Abeet al., 2011). Among various potential markers of inflamma-tion and coagulation, prior research has indicated consis-tently that fibrinogen is associated positively with IMT(Martinez-Vila et al., 2003; Temelkova-Kurktschiev et al.,2002) and both sleep duration and quality (Matthews et al.,2010; Suarez, 2008; Wessendorf et al., 2000). Thus, in thisanalysis, we looked specifically at the potential explanatoryrole of fibrinogen in the association between long sleepduration and cardiovascular health. Fibrinogen has been anunderstudied mechanism of the association between longsleep duration and poor CVD health.Fibrinogen, an acute phase inflammatory protein produced

in the liver, promotes blood clotting and is associatedpositively with increased risks of mortality and incident CVD(Danesh et al., 1998, 2001; Ernst and Resch, 1993; Kannelet al., 1987). Fibrinogen was associated with more difficultyfalling asleep and longer sleep duration among 95 women ina sample of adults in North Carolina (Suarez, 2008). In across-sectional study of multi-ethnic women, however, therewere positive associations between sleep-disordered breath-ing and fibrinogen, but no association between fibrinogenand sleep duration (Matthews et al., 2010). That study,however, treated sleep duration as a continuous variable anddid not separate out categories of sleep duration, whichwould have allowed for a non-linear association betweensleep duration and fibrinogen (Matthews et al., 2010). In asample of more than 1000 Taiwanese adults age 53 yearsand over, there was a positive association between longsleep duration (>8 h per night) and higher levels of severalinflammatory markers (Dowd et al., 2011). With regard tofibrinogen, the study found that long sleepers had fibrinogenlevels 21.8 mg dL�1 (P < 0.01) higher than 6–8-h sleepers,even after adjustment for potential health conditions includingself-reported diabetes, arthritis, heart disease, waist circum-ference, depressive symptoms and health decline since theprevious wave. Consistent with the findings from the multi-ethnic sample (Matthews et al., 2010), no associationbetween fibrinogen and sleep duration was found whensleep duration was treated linearly. The potential mechanismlinking long sleep duration to increased fibrinogen levels maybe due directly to the act of long sleeping or indirectly throughincreased sleep fragmentation, which is more commonamong long sleepers (Grandner and Drummond, 2007).Given these prior findings, we investigated the possible

mediating role of fibrinogen, a marker of inflammation andcoagulation, in the associations between sleep duration andincident coronary heart disease (CHD) and all-cause mortal-ity in a longitudinal sample of post-menopausal women.Specifically, we hypothesized that long sleep duration wouldbe associated positively with fibrinogen and adverse healthoutcomes (CHD and all-cause mortality), even after adjust-ment for a range of demographic and health covariates. Inaddition, we expected to find that adjustment for fibrinogenattenuated the associations between long sleep duration,

subsequent CHD and mortality. Fig. 1 provides a heuristicdiagram of the hypothesized relationships.

METHODS

Sample

Data are from a subsample of the Women’s Health Initiative(WHI) clinical trial (CT) and observational study (OS).Descriptions of the study population, recruitment methodsand measurement protocols have been published previously(Matthews et al., 1997). The study enrolled postmenopausalwomen between 1993 and 1998 from 40 geographicallydiverse clinical centers throughout the United States, aged50–79 years. In total, 161 808 women met the WHI eligibilitycriteria and were enrolled into the study. All participants werefollowed to August 2009 for outcomes for these analyses. Allparticipants provided informed consent, which was approvedby the Institutional Review Boards of the participating studysites.Of this large sample, only a subset of women had serum

studies performed at baseline; fibrinogen measurementswere part of the ‘core blood analytes’ and were measuredprospectively in a 6% random subsample of CT participantsat baseline and in a 1% random subsample of OS partic-ipants at baseline. A total of 5016 women had baselinefibrinogen levels recorded. Women with a history of CHD atbaseline were removed from this sample, to leave a remain-ing sample size of 4206. Of these women, approximately 6%were missing information on variables to be included inregression analyses, and were removed from the sample fora final sample size of 3942 for complete case analysis. Therewere no statistically significant differences in long sleepduration among those who were included in the studycompared to those who were not (both around 4% of the

Sleep duration

Fibrinogen

CovariatesAge

Race/ethnicityEducation

IncomeCVD risk factors

Comorbidities

Health outcomesCHD

All-cause mortality

Figure 1. A heuristic diagram of the hypothesized relationships.

ª 2012 European Sleep Research Society

306 Lauren Hale et al.

population). There were, however, sociodemographic differ-ences between those who were included in the study. Forexample, the included sample were more likely to be black,Hispanic and Asian, and slightly less likely to be married,compared to the full WHI sample.

Outcome variables

Our primary endpoints were incident CHD and all-causemortality during follow-up. Incident CHD was defined by theWHI study as the first occurrence of clinical myocardialinfarction (MI), definite silent MI or a death due to CHD(defined before October 1999), definite CHD or possibleCHD. Overall mortality was defined as death from any cause.All outcomes were determined by a physician, and wereidentified initially through annual follow-up contacts and thenverified through medical records and death certificates.

Key predictor variables

Self-reported sleep duration at study baseline was theprimary predictor of interest. As part of the baseline exam-ination, each WHI participant was asked to report ‘hours ofsleep on a typical night during the past 4 weeks’ (� 5, 6, 7, 8,9, � 10). In our study sample, 450 (11.4%) women slept� 5 h, 1189 (30.2%) women slept 6 h, 1381 (35.0%) slept7 h, 770 (19.5%) slept 8 h, 133 (3.4%) slept 9 h and 19(0.5%) slept 10 h or more. Due to the very small number ofwomen at the longest sleep duration category, we combinedthe top two sleep durations and labeled the categories basedon typical descriptions in the literature: ‘long sleepers’ (9+ h),‘mid-range sleepers’ (7 or 8 h), ‘short sleepers’ (6 h) and‘very short sleepers’ (� 5 h).Fibrinogen levels at screening were evaluated as both a

predicting and mediating variable. The average fibrinogenlevel in the sample was 304.45 mg dL�1, with a standarddeviation of 61.68 mg dL�1. Fibrinogen levels were stan-dardized for ease of interpreting values such that the meanwas 0 and a standard deviation was 1 [standardizedlevel = (fibrinogen level: 304.45)/61.68]. The standardizedfibrinogen level was used for all multivariable modelspresented in the results.

Other pertinent covariates

Demographic covariates

All demographic characteristics were determined at the initialscreening visit and were based on self-report questionnaire.Age was treated as a continuous variable. Education wascategorized as less than high school, high school diploma orgeneral education diploma (GED), some college or voca-tional training and college degree or more. Annual householdincome was classified into the groups <$20 000, $20 000–34 999, $35 000–49 999, $50 000–74 999, $75 000–99 999 and � $100 000. Race/ethnicity was categorized as

white/Caucasian, black/African American, Hispanic/Latinaand Asian.

Risk factors

Behavioral risk factors were also measured at baseline andincluded body mass index (BMI), physical exercise, alcoholintake and smoking. Smoking was categorized as a dichot-omous variable: ever smoke ‘yes’ or ‘no’. Alcohol consump-tion was measured as servings per week. High alcohol intakewas defined as being in the top third of the sample—approximately three or more drinks per week—and wastreated as a dichotomous variable. Physical activity wasmeasured as met-hours per week (Ainsworth et al., 2011).This estimate was based on nine questions related toexpenditure of energy from recreational activity (includingwalking, mild, occasional and strenuous activity). Low phys-ical activity was defined as being in the bottom third of thesample. BMI was measured in kg m�2, and assessed at thebaseline visit.

Comorbidities

We included three dichotomous measures of comorbidities:elevated blood pressure, diabetes and depression. Bloodpressure was measured over the brachial artery usingstethoscope bell and mercury manometer at the baselinevisit. All participants were required to bring their currentprescription medications to the baseline visit, and thisinformation was recorded in a medications database. Hyper-tensive participants included individuals whose systolic bloodpressure was >140 mm Hg and/or diastolic blood pressurewas >90 mm Hg, and/or were taking antihypertensive med-ication. Individuals with diabetes included those participantswith physician-diagnosed diabetes who were taking diabetesmedications. Depression was measured at baseline amongthe WHI-OS participants using the Burnam screening algo-rithm, a measure that includes six items from the 20-itemCenter for Epidemiologic Studies-Depression Scale (CES-D)and two items from the National Institute of Mental Health’sDiagnostic Interview Schedule (DIS). The Burnam screenwas developed initially for the Medical Outcomes Study anduses a logistic regression algorithm to develop a compositescore between 0 and 1. A cutoff point of 0.06 was used toidentify current depressive symptomatology. This instrumenthas shown good sensitivity (74%) and specificity (87%) fordetecting depressive disorder (major depressive disorder anddysthymia) in both primary care and mental health settingsamong postmenopausal women (Tuunainen et al., 2001).We also included covariates of self-rated health (‘In

general, would you say your health is: excellent, very good,good, fair or poor?’) and a measure of life satisfaction. Self-rated health was grouped into a dichotomous variable withexcellent and very good (=1) compared to all else. The lifesatisfaction variable came from the question: ‘Overall, howwould you rate your quality of life?’ based on a scale of 1–10


Long Sleep, Fibrinogen, and CHD 307

with 10 being the best. We dichotomized this variable at nineor better (=1), compared to all those scores at eight or below.

Statistical analyses

First, we tested the association between fibrinogen and self-reported sleep duration using nested ordinary least-squaresregression models. In the crude model, we utilized a four-level categorical variable for sleep duration (reference group7–8 h) to predict continuous fibrinogen (in mg dL�1), unad-justed for other covariates. The next model added adjustmentfor demographic variables (age, education, income andethnicity) and other risk factors and comorbidities.We then investigated two sets of nested logistic regression

models with fibrinogen predicting incident CHD and all-causemortality, respectively. Crude and adjusted models wereestimated as described above.Finally, we estimated two different sets of nested models

for incident CHD and all-cause mortality. In model A, weadjusted only for sleep duration and demographic factors toidentify the baseline association of long sleep duration (9+ h)on the outcome variables. In model B, we included fibrinogento test whether the odds ratio for long sleep duration wasreduced with the addition of fibrinogen. Model C includedadjustment for cardiovascular risk factors and comorbidities.Finally, we conducted tests of mediation using thebinary_mediation command in STATA, which is used forassessing mediation dichotomous outcomes (MacKinnonand Dwyer, 1993; Sobel, 1982). All analyses were conductedin STATA version 11.0 (STATA Corp., College Station, TX,USA).

RESULTS

Descriptive statistics of the analytical sample are shown inTable 1, stratified by sleep duration category. The all-femalesample was primarily white, age 60 or older, with at least withsome college education. Approximately one-third of thesample had low physical activity levels, 47% had a historyof smoking and 17% consumed three or more drinks perweek. In terms of health conditions, 40% had elevated bloodpressure (>140/90), 5% were categorized as diabetic and10% were depressed. In addition, a total of 55% of thesample reported ‘excellent’ and ‘very good’, and 44%reported having life satisfaction at level of 9 or higher.Long sleep duration (9+ h per night) was associated with

increased fibrinogen, in both the unadjusted and adjustedmodels (Table 2), compared to those in the 7–8-h sleepduration. The magnitude of the effect was such that 9+-hsleepers had fibrinogen levels that were 0.20 (P < 0.05)standard deviations (approximately 12 mg dL�1) higher than7–8-h sleepers. There were no differences in fibrinogenlevels between the 7–8-h sleepers compared to the 6-h or� 5-h sleepers.Increased fibrinogen was also associated with increased

odds of incident CHD [odds ratio (OR) = 1.24, 95% confi-

dence interval (CI): 1.06–1.45] and overall mortality(OR = 1.19, 95% CI: 1.07–1.33), such that every onestandard deviation increase in fibrinogen was associatedwith approximately 24% increased odds of CHD and 19%increased odds of overall mortality (Table 3). This associa-tion persisted after adjustment for demographic characteris-tics, risk factors and comorbidities, with only mild attenuation(OR = 1.19, 95% CI: 1.00–1.41 and OR=1.15, 95% CI: 1.03–1.29, respectively).Self-reported long sleep duration (9+ h) was associated

with a more than twofold increased risk of CHD (OR = 2.05,95% CI: 1.02–4.11) compared to 7–8-h sleep duration(Table 4). With the addition of fibrinogen to the model, theassociation of long sleep duration with CHD was attenuatedand no longer statistically significant at the P < 0.05 level(OR = 1.97, 95% CI: 0.98–3.97), with an eight percentagepoint (105% down to 97%) decline in increased odds of CHD.The magnitude of this attenuation accounted for approxi-mately half the total attenuation observed when all cardio-vascular risk factors (e.g. BMI, low physical exercise andever smoking) and comorbidities (e.g. elevated bloodpressure, diabetes, depression, general health and qualityof life) are added to the model (OR = 1.89, 95% CI: 0.92–3.84), suggesting a relatively strong mediating role offibrinogen compared to the other covariates. A similar patternwas seen in Table 5, where long sleep duration wasassociated with 71% increased odds (OR = 1.71, 95% CI:1.04–2.83) of mortality in crude models, which is reduced to1.65 (95% CI: 1.00–2.72) after inclusion of fibrinogen. Theaddition of cardiovascular risk factors and comorbiditiesfurther attenuated the association (OR = 1.57, 95% CI: 0.93–2.63).Tests of binary mediation indicated support for fibrinogen

partially mediating the associations between long sleepduration and both CHD and overall mortality. The proportionof the total effect that is mediated was about 6% for theassociation with CHD and 7% for the association with overallmortality.

DISCUSSION

Consistent with prior studies, this epidemiological studyfound that fibrinogen levels were highest among longsleepers (9+ h). Secondly, we found that the associationsbetween long sleep and incident CHD and overall mortalitywere mediated partially by fibrinogen, providing evidence thatinflammation and procoagulation activity may be one path-way linking long sleep duration with increased risk of CHDand overall mortality. Ancillary analyses not shown indicatedthat if the CVD risk factors were added first to the analyses,and then fibrinogen is added, we again saw that approxi-mately half the overall attenuation was accounted for by theother CVD risk factors, with the remaining due to adjustmentfor fibrinogen. While the magnitude of the partial mediationhere was not particularly large (6–7%), it is comparable to theexplanatory power of a range of other health behaviors and



health variables with regard to sleep duration and CHD andmortality, as evidenced by the change in the long sleep oddsratio after adding in the additional variables in model C.Moreover, the causes of CHD and mortality are undoubtedlycomplex, and there is unlikely to be a single explanation as towhy long sleepers have higher rates of CHD and mortality.However, given that the current evidence is relatively limitedwith regard to the potential pathways explaining the associ-ation between long sleep and mortality (see Cappuccio et al.,

2010), this observed mediating pathway makes an importantcontribution to the current knowledge and should stimulatefuture research on this topic.Our findings regarding the association between sleep

duration and fibrinogen are consistent with prior researchshowing an association between long sleep duration andincreased fibrinogen (Dowd et al., 2011; Suarez, 2008).Similarly, others have found that fibrinogen was associatedwith sleep continuity, sleep architecture, sleep-disordered

Table 1 Descriptive characteristics of the sample (n = 3942) by hours of sleep duration

Hours of sleep

Total (all sleepcategories)

5 h or lessn = 450

6 hn = 1189

7–8 hn = 2151

9 or more hn = 152

P-value

Mean age (years) 62.1 61.5 62.0 62.2 62.5 0.221Ethnicity <0.001

Asian (%) 8.1 10.7 11.4 6.0 4.6Black (%) 21.8 35.6 27.7 15.6 21.7Hispanic (%) 12.3 11.8 14.1 11.2 14.5White (%) 54.0 36.0 43.2 63.7 55.3Other (%) 3.8 6.0 3.7 3.4 3.9

Education <0.001< High school (%) 7.6 10.9 8.4 6.5 7.2High school (%) 17.7 16.2 17.8 18.0 17.8Some college (%) 39.2 44.0 41.3 37.0 40.8College+ (%) 35.5 28.9 32.6 38.5 34.2

Income <0.001<20 k (%) 21.1 29.6 20.9 19.1 26.320–35 k (%) 24.4 26.4 25.4 23.3 25.035–50 k (%) 21.0 17.8 22.6 21.1 16.550–75 k (%) 18.8 16.0 18.8 19.3 21.175–100 k (%) 8.3 6.7 7.6 9.4 4.0100 k+ (%) 6.4 3.6 4.8 7.9 7.2

BMI 0.107�25 (%) 25.8 24.2 27.2 29.9 25.0>25–30 (%) 2.5 35.8 33.9 35.7 35.5>30–35 (%) 35.1 24.0 22.8 21.3 23.7>35 (%) 36.6 16.0 16.2 13.1 15.8

Energy expenditure per week 0.010Low physical exercise (%) 32.6 36.0 30.8 32.2 42.8

Smoke ever 0.077Yes (%) 47.0 51.3 45.95 46.3 53.3

Alcohol intake <0.001>3 drinks/week (%) 17.3 13.1 14.8 19.2 23.7

Elevated blood pressure 0.095Yes (if >140/90) (%) 39.4 29.1 27.9 24.8 29.0

Diabetic 0.132Yes (%) 4.5 11.6 8.9 8.7 5.9

Depressed <0.001Yes (%) 10.2 23.1 10.9 7.1 9.9

General health <0.001Excellent–very good (versus good,fair, poor) (%)

55.2 40.2 52.5 60.2 48.7

Life satisfaction <0.0019–10 (versus 1–8) (%) 44.2 32.9 42.1 47.9 42.8

Inflammatory markersAverage fibrinogen (mg/dL) 304.4 305.2 303.6 303.9 316.5 0.045

Chi-squared tests were used to evaluate all categorical variables (ethnicity, education, income, body mass index (BMI), energy expenditureper week, smoke ever, alcohol intake, elevated blood pressure, diabetic, depressed, general self-rated health and life satisfaction) andanalysis of variance test was used to evaluate the two continuous variables (age and average fibrinogen level at baseline).



breathing and sleep quality (Matthews et al., 2010). Onemight hypothesize that the long sleepers are simply maskingother comorbidities or sleep disorders (obstructive sleepapnea or periodic leg movement syndrome), because longsleeping has been associated with increased sleep fragmen-tation and more frequent awakenings (Grandner and Drum-mond, 2007). However, a recent study using actigraphy andpolysomnography data indicated that self-reported longsleepers spend more time in bed and more time asleep thanself-reported normal duration sleepers, associations notexplained by differences in comorbidity of sleep disorders(Patel et al., 2012).The present study improves upon prior studies of fibrino-

gen and sleep in several ways. First, our longitudinal designallowed analysis of incident events (i.e. coronary heartdisease and overall mortality), increasing confidence thathigh fibrinogen levels preceded these health events. Secondly,

compared to previous studies, our study used a much larger(nearly 4000 postmenopausal women), well-characterizednational sample. Finally, we used mediation to test statisti-cally whether fibrinogen is partially mediating the associationbetween long sleep and adverse outcomes.While this analysis of high-quality longitudinal data pro-

vides suggestive evidence of the possible mediating role offibrinogen in the association between long sleep duration andadverse health outcomes, more research is necessary. Forexample, we have some concerns about the generalizabilityof the findings in this all-female and older sample. Futureresearch should include more variation in demographiccharacteristics, and should study possible interaction effects(e.g. by age, sex, race/ethnicity). Another important issue isthe bidirectionality of the association between long sleep andpoor health. While long sleep may promote inflammation andcoagulation, it is also likely that long sleep, especially at olderages, is a symptom of underlying inflammatory diseaseprocesses. Increased inflammation and procoagulation activ-ity may induce longer sleep or reflect other conditions in thebody that are sleep-promoting. Another limitation of this studyis the reliance on self-reported sleep duration. Although thiscost-effective assessment of sleep duration is correlatedmoderately with actigraph-measured sleep, it has been foundto over-report sleep duration systematically (Lauderdaleet al., 2008). Misclassification of normal sleepers as long-sleepers thus may have led to an underestimate of associ-ations between long sleep and both fibrinogen and healthoutcomes, so future studies with objectively measured sleepshould be used to confirm these results. Further, self-reported sleep duration data fails to include information onawakenings during the night, which may be part of themechanism. Additional studies need to investigate if there isa threshold effect at which too much sleep becomes high risk.In particular, there may be differential thresholds by demo-graphic characteristics (e.g. age, sex, race/ethnicity). Futureresearch should explore data-driven cutpoints for sleepduration, as conducted by Kripke et al. (2011), instead ofrelying on a priori sleep duration categories for predictinghealth outcomes.

Table 3 Standardized fibrinogen as predictor of adverse health outcomes (n = 3942)

CHD All-cause mortality

Cases = 132 Cases = 335

Standardized fibrinogen OR (95% CI) P-value OR (95% CI) P-valueUnadjusted model 1.24 (1.06, 1.45) ** 1.19 (1.07, 1.33) **Adjusted modela 1.19 (1.00, 1.42) * 1.15 (1.03, 1.29) *

*P-value <0.05; **P-value <0.01.aModel is adjusted for age (continuous), education (<high school, high school diploma, some college, college+), income (<20 k, 20–35 k, 35–50 k, 50–75 k, 75–100 k, 100 k+), ethnicity (white, black, Hispanic, Asian), body mass index (BMI) (continuous), low physical exercise (yes if<1.5 met-hours per week), high alcohol intake (>3 drinks/week), ever smoke (yes/no), elevated blood pressure (>140/90), diabetes (yes/no),depression (� 0.06 on depression scale), general health (excellent–very good/good–fair–poor) and life satisfaction scale (9–10/1–8). CI:confidence interval; OR: odds ratio.

Table 2 Results of ordinary least-squares (OLS) regressionmodel of hours of sleep as a predictor of standardizedcontinuous fibrinogen level (n = 3942)

Fibrinogena

Beta-coefficient(95% CI)

Fibrinogenb

Beta-coefficient(95% CI)

5 h or less(n = 450)

0.02 (�0.08, 0.12) �0.06 (�0.16, 0.04)

6 h (n = 1189) �0.01 (�0.08, 0.07) �0.06 (�0.13, 0.01)7–8 h (n = 2151) Ref Ref9 h or more(n = 152)

0.20 (0.04, 0.37)* 0.17 (0.004, 0.33)*

*P-value <0.05.aUnadjustedmodel.bModel is adjusted for age (continuous), education (<high school,high school diploma, some college, college +), income (<20 k, 20–35 k, 35–50 k, 50–75 k, 75–100 k, 100 k+), ethnicity (white,black, Hispanic, Asian), low physical exercise (yes if <1.5 met-hours per week), high alcohol intake (>3 drinks/week), ever smoke(yes/no), elevated blood pressure (>140/90), diabetes (yes/no),depression (�0.06 on depression scale), general health (excel-lent–very good/good–fair–poor) and life satisfaction scale (9–10/1–8).



Surprisingly, in Tables 4 and 5, we failed to find astatistically significant association between short sleep andboth CHD and all-cause mortality in our data, despite the bulkof prior studies (Grandner et al., 2009; Kripke et al., 2002).This inconsistency may be due to a relatively small number ofparticipants having the adverse health outcomes of interest(n = 132 and n = 335 for CHD and mortality, respectively), orbecause of having self-reported sleep measured at only onepoint in time (Cappuccio et al., 2010). Alternatively, the

association between short sleep duration and health may notbe persistent across all populations. In fact, a recent meta-analysis shows a number of studies that do not findsignificant associations for short sleep duration with CVDand mortality outcomes (Cappuccio et al., 2010). Futureresearch should also seek to reconcile more clearly oursomewhat conflicting findings with regard to short sleepduration and CHD and all-cause mortality that were observedin this study.

Table 4 Odds ratio (95% CI) of baseline sleep duration predicting incident coronary heart disease mediated by fibrinogen levels at baseline(n = 3942)

CHD

Model A Model B Model C

Sleep duration9 h or more 2.05 (1.02, 4.11)* 1.97 (0.98, 3.97)+ 1.88 (0.92, 3.83)+

7–8 h Ref Ref Ref6 h 0.64 (0.40, 1.00)+ 0.65 (0.41, 1.02)+ 0.66 (0.42, 1.04)+

5 h or less 1.13 (0.66, 1.94) 1.16 (0.68, 1.98) 1.09 (0.63, 1.89)Demographics

Age 1.09 (1.06, 1.12)** 1.09 (1.06, 1.12)** 1.09 (1.06, 1.12)**EthnicityWhite Ref Ref RefBlack 1.07 (0.68, 1.68) 1.01 (0.64, 1.60) 0.92 (0.57, 1.48)Hispanic 0.56 (0.27, 1.17) 0.54 (0.26, 1.13) 0.56 (0.27, 1.16)Asian 1.00 (0.50, 1.99) 1.03 (0.50, 2.00) 1.02 (0.50, 2.08)

Education< High school Ref Ref RefHigh school 0.49 (0.25, 0.99)* 0.51 (0.25, 1.02)+ 0.55 (0.27, 1.11)+

Some college 0.82 (0.45, 1.47) 0.84 (0.47, 1.51) 0.89 (0.50, 1.63)College+ 0.42 (0.21, 0.83)* 0.43 (0.22, 0.85)* 0.49 (0.24, 0.97)*

Income< 20 k Ref Ref Ref20–35 k 0.64 (0.40, 1.03)+ 0.65 (0.40, 1.04)+ 0.66 (0.41, 1.07)+

35–50 k 0.92 (0.56, 1.51) 0.93 (0.57, 1.53) 0.99 (0.60, 1.64)50–75 k 0.35 (0.17, 0.72)* 0.36 (0.17, 0.75)* 0.38 (0.18, 0.80)*75–100 k 0.68 (0.29, 1.61) 0.70 (0.30, 1.64) 0.77 (0.32, 1.84)100 k+ 0.39 (0.11, 1.32) 0.40 (0.12, 1.36) 0.41 (0.12, 1.40)

FibrinogenStandardized fibrinogen 1.19 (1.01, 1.41)* 1.17 (0.99, 1.39)+

Cardiovascular risk factorsBMI 0.98 (0.95, 1.02)Low physical exercise 1.36 (0.93, 1.98)+

Smoke ever 1.14 (0.79, 1.63)High alcohol intake 1.06 (0.64, 1.77)

ComorbiditiesElevated blood pressure 1.21 (0.83, 1.75)Diabetes 2.22 (1.22, 4.01)**Depression 1.38 (0.78, 2.43)

General healthFair/poor/good RefVery Good/excellent 0.85 (0.57–1.27)

Life satisfaction1–8 Ref9–10 1.08 (0.74–1.57)

+P-value <0.10; *P-value <0.05; **P-value <0.01.Descriptions of covariates: body mass index (BMI) (continuous), low physical exercise (yes if <1.5 met-hours per week), high alcohol intake(>3 drinks/week), ever smoke (yes/no), elevated blood pressure (>140/90), diabetes (yes/no), depression (� 0.06 on depression scale),general health (excellent–very good/good–fair–poor) and life satisfaction scale (9–10/1–8). CHD: coronary heart disease; CI: confidenceinterval.



Table 5 Odds ratio (95% CI) of sleep duration at screening as a predictor of all-cause mortality mediated by fibrinogen levels at screening(n = 3942)

Death


(a) Sleep duration9 h or more 1.71 (1.04, 2.83)* 1.65 (1.00, 2.72)+ 1.55 (0.92, 2.60)+

7–8 h Ref Ref Ref6 h 0.89 (0.67, 1.16) 0.89 (0.68, 1.18) 0.94 (0.71, 1.24)5 h or less 1.01 (0.69, 1.48) 1.00 (0.68, 1.46) 1.01 (0.68, 1.51)

DemographicsAge 1.10 (1.08, 1.12)** 1.10 (1.08, 1.12)** 1.09 (1.07, 1.11)**

EthnicityWhite Ref Ref RefBlack 0.85 (0.62, 1.15) 0.80 (0.59, 1.10) 0.64 (0.46, 0.89)**Hispanic 0.57 (0.36, 0.90)* 0.55 (0.35, 0.87)* 0.60 (0.37, 0.95)*Asian 0.53 (0.32, 0.90)* 0.54 (0.32, 0.92)* 0.59 (0.34, 1.01)+

Education< High school Ref Ref RefHigh school 0.92 (0.57, 1.48) 0.94 (0.58, 1.51) 1.04 (0.64, 1.70)Some college 0.92 (0.61, 1.42) 0.94 (0.63, 1.47) 1.00 (0.63, 1.58)College+ 0.79 (0.49, 1.28) 0.81 (0.52, 1.30) 0.97 (0.59, 1.59)

Income<20 k Ref Ref Ref20–35 k 0.92 (0.67, 1.22) 0.93 (0.68, 1.26) 0.91 (0.67, 1.25)35–50 k 0.61 (0.43, 0.89)* 0.62 (0.43, 0.90)* 0.60 (0.41, 0.88)**50–75 k 0.57 (0.38, 0.86)* 0.58 (0.39, 0.88)* 0.57 (0.38, 0.87)**75–100 k 0.55 (0.27, 1.00)+ 0.57 (0.28, 1.03)+ 0.57 (0.31, 1.05)+

100 k+ 0.57 (0.30, 1.11)+ 0.59 (0.30, 1.14)+ 0.54 (0.28, 1.06)+

Inflammatory markersStandardized fibrinogen 1.17 (1.04, 1.30)* 1.14 (1.02, 1.28)*

Cardiovascular risk factorsBMI 0.98 (0.96, 1.00)+

Low physical exercise 1.55 (1.21, 1.99)**Smoke ever 2.03 (1.59, 2.60)**High alcohol intake 1.09 (0.79, 1.50)

ComorbiditiesElevated blood pressure 1.66 (1.29, 2.12)**Diabetes 1.90 (1.34, 2.70)**Depression 1.03 (0.67, 1.58)

General healthFair/poor/good RefVery good/excellent 0.73 (0.57–0.95)*

Life satisfaction1–8 Ref9–10 1.47 (1.07–2.0)*

(b) Sleep duration9 h or more 1.71 (1.04, 2.83)* 1.65 (1.00, 2.72) 1.57 (0.93, 2.63)7–8 h Ref Ref Ref6 h 0.89 (0.67, 1.16) 0.89 (0.68, 1.18) 0.94 (0.71, 1.24)5 h or less 1.01 (0.69, 1.48) 1.00 (0.68, 1.46) 1.05 (0.71, 1.56)

DemographicsAge 1.10 (1.08, 1.12)** 1.10 (1.08, 1.12)** 1.09 (1.07, 1.11)**

EthnicityWhite Ref Ref RefBlack 0.85 (0.62, 1.15) 0.80 (0.59, 1.10) 0.68 (0.49, 0.94)*Hispanic 0.57 (0.36, 0.90)* 0.55 (0.35, 0.87)* 0.65 (0.51, 1.03)+

Asian 0.53 (0.32, 0.90)* 0.54 (0.32, 0.92)* 0.62 (0.36, 1.06)+

Education< High school Ref Ref RefHigh school 0.92 (0.57, 1.48) 0.94 (0.58, 1.51) 1.03 (0.63, 1.68)Some college 0.92 (0.61, 1.42) 0.94 (0.63, 1.47) 0.99 (0.63, 1.56)College+ 0.79 (0.49, 1.28) 0.81 (0.52, 1.30) 0.92 (0.57, 1.51)



Despite these limitations, however, we believe that ourfindings provide an important new potential pathway forunderstanding the widely observed association between longsleep duration and mortality. Specifically, fibrinogen, as amarker of inflammation and coagulation, may be playing apartially mediating role in increasing the risk of CHD and all-cause mortality.

REFERENCES

Abe, T., Aoki, T., Yata, S. and Okada, M. Sleep duration is significantlyassociated with carotid artery atherosclerosis incidence in aJapanese population. Atherosclerosis, 2011, 217: 509–513.

Ainsworth, B. E., Haskell, W. L., Herrmann, S. D. et al. Compendiumof physical activities: a second update of codes and MET values.Med. Sci. Sports Exerc., 2011, 43: 1575–1581.

Ayas, N. T., White, D. P., Manson, J. E. et al. A prospective study ofsleep duration and coronary heart disease in women. Arch. Intern.Med., 2003, 163: 205–209.

Cappuccio, F. P., D’Elia, L., Strazzullo, P. and Miller, M. A. Sleepduration and all-cause mortality: a systematic review and meta-analysis of prospective studies. Sleep, 2010, 33: 585–592.

Danesh, J., Collins, R., Appleby, P. and Peto, R. Association offibrinogen, C-reactive protein, albumin, or leukocyte count withcoronary heart disease: meta-analyses of prospective studies.JAMA, 1998, 279: 1477–1482.

Danesh, J., Whincup, P., Walker, M. et al. Fibrin D-dimer andcoronary heart disease: prospective study and meta-analysis.Circulation, 2001, 103: 2323–2327.

Dowd, J. B., Goldman, N. and Weinstein, M. Sleep duration, sleepquality, and biomarkers of inflammation in a Taiwanese population.Ann. Epidemiol., 2011, 21: 799–806.

Ernst, E. and Resch, K. L. Fibrinogen as a cardiovascular risk factor:a meta-analysis and review of the literature. Ann. Intern. Med.,1993, 118: 956–963.

Gallicchio, L. and Kalesan, B. Sleep duration and mortality: asystematic review and meta-analysis. J. Sleep Res., 2009, 18:148–158.

Grandner, M. A. and Drummond, S. P. Who are the long sleepers?Towards an understanding of the mortality relationship. Sleep Med.Rev., 2007, 11: 341–360.

Grandner, M. A., Hale, L., Moore, M. and Patel, N. P. Mortalityassociated with short sleep duration: the evidence, the possiblemechanisms, and the future. Sleep Med. Rev., 2009, 14:191–203.

Kannel, W. B., Wolf, P. A., Castelli, W. P. and D’Agostino, R. B.Fibrinogen and risk of cardiovascular disease The FraminghamStudy. JAMA, 1987, 258: 1183–1186.

Knutson, K. L. and Turek, F. W. The U-shaped association betweensleep and health: the 2 peaks do not mean the same thing. Sleep,2006, 29: 878–879.

Kripke, D. F., Garfinkel, L., Wingard, D. L., Klauber, M. R. and Marler,M. R. Mortality associated with sleep duration and insomnia. Arch.Gen. Psychiatry, 2002, 59: 131–136.

Table 5 Continued

Death


Income<20 k Ref Ref Ref20–35 k 0.92 (0.67, 1.22) 0.93 (0.68, 1.26) 0.90 (0.66, 1.23)35–50 k 0.61 (0.43, 0.89)* 0.62 (0.43, 0.90)* 0.60 (0.41, 0.87)*50–75 k 0.57 (0.38, 0.86)* 0.58 (0.39, 0.88)* 0.56 (0.37, 0.86)*75–100 k 0.55 (0.27, 1.00)+ 0.57 (0.28, 1.03)+ 0.57 (0.31, 1.05)+

100k+ 0.57 (0.30, 1.11)+ 0.59 (0.30, 1.14)+ 0.53 (0.27, 1.05)+

Inflammatory markersStandardized fibrinogen 1.17 (1.04, 1.30)* 1.14 (1.02, 1.28)*

Cardiovascular risk factorsBMI 0.99 (0.96, 1.00)Low physical exercise 1.57 (1.23, 2.01)**Smoke ever 2.03 (1.59, 2.59)**High alcohol intake 1.09 (0.79, 1.49)

ComorbiditiesElevated blood pressure 1.73 (1.35, 2.21)**Diabetes 2.17 (1.40, 3.35)**Depression 1.07 (0.70, 1.64)

General healthFair/poor/good RefVery good/excellent 1.02 (0.71–1.45)

Life satisfaction8–10 Ref0–7 0.72 (0.53–0.98)*

+P-value <0.10; *P-value <0.05; **P-value <0.01.Descriptions of covariates: body mass index (BMI) (continuous), low physical exercise (yes if <1.5 met-hours per week), high alcohol intake(>3 drinks/week), ever smoke (yes/no), elevated blood pressure (>140/90), diabetes (yes/no), depression (� 0.06 on depression scale),general health (excellent–very good/good–fair–poor) and life satisfaction scale (a: 0–7/8–10; b: 9–10/1–8). CI: confidence interval.



Kripke, D. F., Langer, R. D., Elliott, J. A., Klauber, M. R. and Rex, K.M. Mortality related to actigraphic long and short sleep. SleepMed., 2011, 12: 28–33.

Lauderdale, D. S., Knutson, K.L., Yan, L. L., Liu, K. and Rathouz, P.J. Self-reported and measured sleep duration: how similar arethey? Epidemiology, 2008, 19: 838–845.

MacKinnon, D. P. and Dwyer, J. H. Estimating mediated effects inprevention studies. Eval. Rev., 1993, 17: 144–158.

Martinez-Vila, E., Paramo, J. A., Beloqui, O. et al. Independentassociation of fibrinogen with carotid intima-media thickness inasymptomatic subjects. Cerebrovasc. Dis., 2003, 16: 356–362.Matthews, K. A., Shumaker, S. A., Bowen, D. J. et al. Women’shealth initiative Why now? What is it? What’s new?. Am. Psychol.,1997, 52: 101–116.

Matthews, K. A., Zheng, H., Kravitz, H. M. et al. Are inflammatory andcoagulation biomarkers related to sleep characteristics in mid-lifewomen? Study of Women’s Health Across the Nation Sleep Study.Sleep, 2010, 33: 1649–1655.

Miller, M. A., Kandala, N. B., Kivimaki, M. et al. Gender differences inthe cross-sectional relationships between sleep duration and mark-ers of inflammation: Whitehall II study. Sleep, 2009, 32: 857–864.

Opp, M. R. Sleeping to fuel the immune system: mammalian sleepand resistance to parasites. BMC Evol. Biol., 2009, 9: 8.

Patel, S. R., Zhu, X. and Storfer-Isser A.etal. Sleep duration andbiomarkers of inflammation. Sleep, 2009, 32: 200–204.

Patel, S. R., Blackwell, T., Ancoli-Israel, S. and Stone, K. L. Sleepcharacteristics of self-reported long sleepers. Sleep, 2012, 35:641–648.

Simpson, N. and Dinges, D. F. Sleep and inflammation. Nutr. Rev.,2007, 65: S244–S252.

Sobel, M. E. Asymptotic intervals for indirect effects in structuralequation models. Sociol. Methodol., 1982, 13: 290–312.

Stamatakis, K. A. and Punjabi, N. M. Long sleep duration: a risk tohealth or a marker of risk? Sleep Med. Rev., 2007, 11: 337–339.

Stranges, S., Dorn, J. M., Shipley, M. J. et al. Correlates of short andlong sleep duration: a cross-cultural comparison between theUnited Kingdom and the United States: the Whitehall II Study andthe Western New York Health Study. Am. J. Epidemiol., 2008, 168:1353–1364.

Suarez, E. C. Self-reported symptoms of sleep disturbance andinflammation, coagulation, insulin resistance and psychosocialdistress: evidence for gender disparity. Brain Behav. Immun.,2008, 22: 960–968.

Tamakoshi, A. and Ohno, Y. Self-reported sleep duration as apredictor of all-cause mortality: results from the JACC study,Japan. Sleep, 2004, 27: 51–54.

Temelkova-Kurktschiev, T., Koehler, C., Henkel, E. and Hanefeld, M.Leukocyte count and fibrinogen are associated with carotid andfemoral intimal-media thickness in a risk population for diabetes.Cardiovasc. Res., 2002, 56: 277–283.

Tuunainen, A., Langer, R. D., Klauber, M. R. and Kripke, D. F. Shortversion of the CES-D (Burnam screen) for depression in referenceto the structured psychiatric interview. Psychiatry Res., 2001, 103:261–270.

Wessendorf, T. E., Thilmann, A. F., Wang, Y., Schreiber, A.,Konietzko, A. and Teschler, H.Fibrinogen levels of obstructivesleep apnea in ischemic stroke. Am. J. Respir. Crit. Care Med.,2000, 162: 2039–2042.

Youngstedt, S. D. and Kripke, D. F. Long sleep and mortality:rationale for sleep restriction. Sleep Med. Rev., 2004, 8: 159–174.



Download - Fibrinogen may mediate the association between long sleep duration and coronary heart disease

Top Related