dia care 2009 vgontzas 1980 5

Upload: subli-muhammad

Post on 07-Jul-2018

217 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/19/2019 Dia Care 2009 Vgontzas 1980 5

    1/6

    Insomnia With Objective Short SleepDuration Is Associated With Type 2Diabetes

     A population-based studyALEXANDROS N. VGONTZAS,  MD1

    DUANPING LIAO,   MD, PHD2

    SLOBODANKA  PEJOVIC,   MD1

    SUSAN CALHOUN,   PHD1

    MARIA K ARATARAKI,   PSYCHD1

    EDWARD O. BIXLER,   PHD1

    OBJECTIVE — We examined the joint effects of insomnia and objective short sleep duration,the combination of which is associated with higher morbidity, on diabetes risk.

    RESEARCH DESIGN AND METHODS — A total of 1,741 men and women randomlyselected from Central Pennsylvania were studied in the sleep laboratory. Insomnia was definedby a complaint of insomnia with duration of  1 year, whereas poor sleep was defined as a

    complaint of difficulty falling asleep, staying asleep, or early final awakening. Polysomnographicsleep duration was classified into three categories: 6 h of sleep (top 50% of the sample); 5– 6h (approximately third quartile of the sample); and 5 h (approximately the bottom quartile of the sample). Diabetes was defined either based on a fasting blood glucose 126 mg/dl or use of medication. In the logistic regression model, we simultaneously adjusted for age, race, sex, BMI,smoking, alcohol use, depression, sleep-disordered breathing, and periodic limb movement.

    RESULTS — Chronic insomnia but not poor sleep was associated with a higher risk fordiabetes. Compared with the normal sleeping and6 h sleep duration group, the highest risk of diabetes was in individuals with insomnia and 5 h sleep duration group (odds ratio [95% CI]2.95 [1.2–7.0]) and in insomniacs who slept 5–6 h (2.07 [0.68–6.4]).

    CONCLUSIONS — Insomnia with short sleep duration is associated with increased odds of diabetes. Objective sleep duration may predict cardiometabolic morbidity of chronic insomnia,the medical impact of which has been underestimated.

    Diabetes Care 32:1980–1985, 2009

    Many studies have established thatinsomnia, the most commonsleep disorder, is highly comorbid

    with psychiatric disorders and is a riskfactor for the development of depression,anxiety, and suicide (1,2). In contrastwith sleep-disordered breathing (SDB),the second most common sleep disorder,chronic insomnia has not been associatedwith significant medical morbidity, e.g.,

    cardiovascular disorders (3,4).Recently, we demonstrated that in-

    somnia with objective short sleep dura-

    tion is associated with a high risk forhypertension (5). These data suggest thatobjective sleep measures in insomnia pro-vide an index of the severity of the disor-der and that the more severe form of insomnia is most likely associated withmorbidity and possibly mortality. Thishypothesis is further supported by phys-iological studies, which demonstratedthat activation of the hypothalamic-

    pituitary-adrenal (HPA) axis and auto-nomic system, including increased heartrate, 24-h metabolic rate, and impaired

    heart rate variability, is present in insom-niacs who meet both subjective and ob-

     jective polysomnographic criteria (6 –11).Given the association of the HPA axis andsympathetic system activation with thepathogenesis of metabolic disorders, in-cluding diabetes (12), we hypothesizedthat insomnia with objective short sleepduration will be associated with type 2diabetes.

    Previous studies have shown that

    sleep disturbances or complaints are asso-ciated with increased incidence of type 2diabetes (13–16). However, in these stud-ies, thepresenceof sleep disturbances wasbased only on a subjective questionnaireand did not control for obstructive sleepapnea, a sleep disorder whose associationwith diabetes and insulin resistance iswell established (12). Thus, it is notknown whether insomnia per se is asso-ciated with an increased risk for diabetes.

    To test this hypothesis, we examinedthe joint effects of the complaints of chronic insomnia and poor sleep (amilder form of insomnia) and objectivesleep duration on the prevalence of diabe-tes in a large cross-sectional population-based sample from Central Pennsylvania(Penn State Cohort).

    RESEARCH DESIGN AND

    METHODS — The data were collectedas part of a two-phase protocol whose pri-mary purpose was to establish the age dis-tribution of SDB (17,18). In thefirst phaseof the study, a sample of adult men and

    women (aged

    20 years) was randomlyselected from local telephone householdsin two counties of Central Pennsylvania(Dauphin and Lebanon) using the Mitofsky-

     Waksberg two-stage random digit dialingprocedure. A within-household selectionprocedure described by Kish was used toselect the specific man or woman to beinterviewed. Telephone interviews wereconducted with 4,364 age-eligible menand 12,219 age-eligible women residingin the sample households for a total sam-ple of 16,583 with response rates of 73.5and 74.1%, respectively. The question-

    ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

    From the   1Sleep Research and Treatment Center, Department of Psychiatry, Pennsylvania State UniversityCollege of Medicine, Hershey, Pennsylvania; and the  2Department of Public Health Sciences, Pennsylva-nia State University College of Medicine, Hershey, Pennsylvania.

    Corresponding author: Alexandros N. Vgontzas, [email protected] 13 February 2009 and accepted 13 July 2009. Published ahead of print at http://care.

    diabetesjournals.org on 29 July 2009. DOI: 10.2337/dc09-0284.© 2009 by the American Diabetes Association. Readers may use this article as long as the work is properly

    cited, the use is educational and not for profit, and the work is not altered. See http://creativecommons.org/licenses/by-nc-nd/3.0/ for details.

    The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be herebymarked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

    E p i d e m i o l o g y / H e a l t h S e r v i c e s R e s e a r c hO R I G I N A L A R T I C L E

    1980   DIABETES CARE,  VOLUME 32,  NUMBER  11, NOVEMBER  2009 care.diabetesjournals.org

  • 8/19/2019 Dia Care 2009 Vgontzas 1980 5

    2/6

    naire used in this interviewincluded basicdemographic and sleep information.

    In the second phase of this study, asubsample of 741 men and 1,000 womenselected from those subjects previouslyinterviewed by telephone were studied inour sleep laboratory. The sample of menwas chosen by counting how many of four

    risk factors (snoring, daytime sleepiness,obesity, and hypertension) each inter-viewed subject reported. Subjects withhigher counts of risk factors were over-sampled (3, 9,32,45,and70%for 0,1, 2,3, and 4 symptoms reported, respec-tively). The sample of women was chosenby counting how many of five risk factors(snoring, daytime sleepiness, obesity, hy-pertension, and menopause) were re-ported by each interviewed subject.Those with higher counts of risk factorswere oversampled (2, 4, 10, 27, 51, and

    56% of those with 0, 1, 2, 3, 4, and 5symptoms reported, respectively). Theresponse rates for this phase were 67.8and 65.8% for men and women, respec-tively. There were no significant differ-ences between those subjects who wererecorded in the laboratory and those whowere selected but were not recorded interms of age, BMI, reported use of medi-cation for hypertension or diabetes, andprevalence of sleep disorders. Each sub-

     ject selected for laboratory evaluationcompleted a comprehensive sleep historyand physical examination and was con-

    tinuously monitored for one night for 8 husing 16-channel polygraphs includingan electroencephalogram, electrooculo-gram, and electromyogram. Bedtimeswere adjusted to conform to subjects’usual bedtimes, and recordings weredone between 10:00 and 11:00   P.M. and6:00 and 7:00  A.M. The sleep records weresubsequently scored independently ac-cording to standardized criteria. Percentsleep time is total sleep time (duration of sleep) divided by recorded time in bedand multiplied by 100. Respiration was

    monitored throughout the night by use of thermocouples at the nose and mouthand thoracic strain gauges. All-night re-cordings of hemoglobin oxygen satura-tion (SaO

    2) were obtained with an

    oximeter attached to the finger.

    Key measurementsDiabetes was defined as being medicallytreated for diabetes or having fastingblood glucose   126 mg/dl from blooddrawn the morning after the sleep labora-tory testing. Hypertension was defined asdiastolic blood pressure  90 mmHg or

    systolic blood pressure   140 mmHgat the time of the sleep laboratory evalua-tion or the use of antihypertensionmedication.

    The presence of sleep disorders wasbased on a standardized questionnairecompleted by the subjects on the eveningof their sleep laboratory visit. This ques-

    tionnaire consists of 53 questions (7 de-mographic, 20 sleep-related, and 26general health questions). In addition,women responded to eight questions re-lated to menstrual history, menopause,and hormone therapy (for more details,see ref. 5). The presence of sleep difficultywas established on three levels of severity.First, insomnia was defined by a com-plaint of insomnia with a duration of atleast 1 year. Second, poor sleep was de-fined as a moderate to severe complaint(based on a mild to severe scale) of diffi-

    culty falling asleep, difficulty stayingasleep, early final awakening, or unre-freshing sleep. To create two mutually ex-clusive categories, the insomnia groupcould include those who reported one ormore of the four symptoms of poor sleep,whereas none of the poor sleep group hada complaint of insomnia. Third, normalsleeping was defined as the absence of ei-ther of these two categories.

    From the objectively recorded sleeptime data, we regrouped the entire studysample into three ordinal groups based onstandard mathematical approaches using

    as cutoff points the 50th percentile andthe 25th percentile: the top 50% of per-sons above the median percent sleep time(“normal sleep duration group”), the 25%of persons in the third quartile (“moder-ately short sleep duration group”), andthe bottom 25% of persons (“severelyshort sleep duration group”). We thenrounded the quartile cutoff points topractically meaningful numbers. Thus,we created the following three relativesleep duration groups: the normal sleepduration group consisted of those who

    slept 

    6 h, the moderately short sleepduration group consisted of those whoslept 5–6 h, and the severely short sleepduration group consisted of those whoslept 5 h.

    To control for possible confoundingvariables influencing the relation betweeninsomnia and diabetes in the subsampleof 1,741, we ascertained whether the re-spondent was currently treated for de-pression (including a history of suicidalthoughts or attempts), had a history of smoking (current use of any type of to-bacco product), had a history of alcohol

    use (more than two alcohol drinks perday), and had sleep apnea or periodiclimb movements. For the purpose of thisstudy, sleep apnea was defined as an ob-structive apnea or hypopnea index of 5(5,17,18). The condition of periodic limbmovement was considered present whenthere were five or more movements per

    hour of sleep, and leg movements werescored according to standardized criteria(5). BMI was based on measured height(centimeters) and weight (kilograms)during the subjects’ sleep laboratory visit,and data are presented in terms of mean,percentile distribution, and prevalencewithin each category.

    Statistical analysesThe design of this study included over-sampling of those at higher risk for SDBand women with markedly higher levels

    of BMI to increase the precision of the riskestimates. Because of this sampling strat-egy, numeric sampling weights were de-veloped for the analysis so that theestimates could be inferred to the originaltarget population (5,17,18). Specifically,three weights were created for the men.First, in the telephone sample, 32 of the963 clusters of phone numbers in the firststage were “exhausted” before the targetsample size was obtained. A compensa-tory weight was computed that correctedfor this problem. A second weight wascomputed because the within-household

    screening deliberately introduced un-equal probabilities of selection across thethree age-groups to oversample the mid-dle age-group. The final weight for themen was computed to account for theoversampling of subjects for the sleep lab-oratory study (phase II); those with largercounts of the four possible risk factors,i.e., snoring, daytime sleepiness, obesity,and hypertension, had substantiallyhigher probability of being selected. Forthe women, the only weight required wasto account for the oversampling of sub-

     jects for the sleep laboratory study. Toeliminate any suggestion of possible sam-ple bias, we calculated 32 unique weightsfor the women and 16 unique weights forthe men corresponding to all possiblecombinations of the five risk factors forthe women and four for the men. Anyindividual weight that had too small of acell size was combined with adjacent cellsso that  10% of the cells had a sample25, and no cell had a size 10. Finally,we used the BMI and race distributions byage decade from the National Health andNutrition Examination Survey III labora-

    Vgontzas and Associates

    care.diabetesjournals.org DIABETES CARE,  VOLUME 32,  NUMBER  11, NOVEMBER  2009   1981

  • 8/19/2019 Dia Care 2009 Vgontzas 1980 5

    3/6

    tory data as the standard to adjust thenumbers of both the men and women interms of BMI and race to be more repre-sentative of the national population.

    Logistic regression models were usedto assess the independent associations of the three-level sleep difficulty complaintsand objective sleep duration with diabe-

    tes. We adjusted for major confoundingfactors expected to affect this relationship(i.e., age, race, sex, BMI, smoking status,alcohol consumption, depression, andSDB). We further tested the interactionbetween sleep difficulty complaints andobjective sleep duration using a 2 log-likelihood ratio test in logistic regressionmodels. Because the interaction betweensleep difficulty complaints and objec-tive sleep duration was not significant,we entered these two variables into themodel separately, except for the last

    model (Model 3, Table 2), in which theresults were adjusted for each other.Then we performed final logistic regres-sion models to include eight dummyvariables to represent all nine possiblecombinations of sleep difficulty andsleep duration and used persons with-out insomnia/poor sleep and with 6 hof sleep duration as a common referencegroup (Table 3).

    RESULTS — The demographic, clini-cal, and sleep characteristics of the entiresample and its subgroups, based on sleep

    difficulty, and the three levels of objectivesleep duration and diabetes, are pre-sented in Table 1. Insomnia was associ-ated with a significantly higher risk fordiabetes (odds ratio [OR] 1.84 [95% CI1.05–3.20], P    0.05) in the first basiccovariable adjusted model (Model 1, Ta-ble 2). With the increases in the numberof potential confounding factors in themodel, the OR of insomnia and diabetesremained very similar (changed   8%,from 1.84 to 1.69), but the precision of the estimation worsened, reflected by in-

    creased width of 95% CI. Poor sleep wasassociated with a slight, nonsignificantincrease of risk for diabetes. Finally, anobjective sleep duration of 5–6 h wasassociated with a slight, nonsignificantincrease of risk for diabetes (1.35[0.92–1.98]).

    The risk of diabetes was synergisti-cally and significantly increased amongpersons with both insomnia and shortsleep duration (Table 3). The presence of both insomnia and an objective sleep du-ration5 h increased the odds for diabe-tes by about 300% (OR 2.95 [95% CI

    T  a b l    e1 —D e  m o  gr  a  p h   i    c  , c  l    i    n i    c  a l    , a n d   s  l    e  e   p c  h   ar  a c  t   e r  i    s  t   i    c  s  o  f    t   h   e  s  t   u d    y  p o  p u l    a t   i    o n

    A l   l   

    Di   a  b  e  t   e  s 

     S l    e  e   p d i  f   fi  c  ul    t    y

     S l    e  e   p d  ur  a  t  i   on (  h  )  

     N o

    Y  e  s 

     N or m a l   

     s l    e  e   pi  n  g

    P  o or  s l    e  e   p

    I  n s  omni   a 

     5 

     5 – 6 

     6 

     n

    1  , 7  4 1 

    1  , 3 2  7 

     4 1  4 

    1  , 0 2 2 

     5 2  0 

    1  9  9 

     4  4  9 

     4  3  0 

     8  6 2 

    A   g e  (    y e  a r  s  )  

     4  8  . 7 

    1  3  . 5 2 

     4  7  . 3 

    1  4  .2 

     5  7  .1 

     8  . 6 

     4  9  . 3 

    1  4  . 9 

     4 

     6  . 5 

    1 1  . 5  6 

     4  9  . 9 

     9  . 8  8 

     5  8  . 3 

    1 1  . 7  8 

     5 1  . 4 

    1  3  .1 2 

     4  4  . 0 

    1 2  . 0 

    B MI  

    2  7  . 6 

     5  . 6  7 

    2  7  . 0 

     5  . 6 

     3 1  .2 

     5  .1 

    2  7  . 0 

     5  . 5 2 

     8  . 8 

     5  . 5  7 

    2  9  . 0 

     6  .1 1 

    2  8  . 0  6 

     5  . 4  0 

    2  7  . 8 

     5  . 4  4 

    2  7  .1 

    2  . 8  9 

    B MI    p e r  c  e n t  i  l    e 

    2  5  t  h 

    2  4  . 0 

    2  3 

     . 5 

    2  5  .2 

    2  3  . 8 

    2  4  . 6 

    2  3  . 4 

    2  3  . 8 

    2  4  .2 

    2  4  .2 

     5  0  t  h 

    2  6  .2 

    2  5 

     . 5 

    2  8  . 3 

    2  6  . 0 

    2  7  . 3 

    2  7  . 7 

    2  6  .2 

    2  6  . 6 

    2  6  . 5 

     7  5  t  h 

     3  0  .2 

    2  8 

     . 6 

     3 2  . 5 

    2  9  . 3 

     3 1  . 9 

     3 2  . 6 

    2  9  . 8 

     3  0  . 3 

     3  0  . 9 

     O b  e  s i   t    y (  B MI   3  0 k   g /  m2  )   (   %

     )  

    2  6 

    1  9 

     3  9 

    2 1 

     3  5 

     4  0 

    2  4 

    2  7 

    2  9 

     S  e x (   %m a l    e  )  

     4  8 

     4  4 

     7 2 

     5  3 

     3  7 

    2  6 

     5  9 

     5  0 

     4 2 

    E  t  h ni   c i   t    y (   % wh i   t   e  )  

     8  6 

     8  6 

     8  7 

     8  6 

     8  7 

     7  6 

     9  3 

     8  5 

     8  3 

    Di   a  b  e  t   e  s  (   % )  

    1  4 

    1  3 

    1  5 

    1  8 

    2 2 

    1  8 

     9 

     Gl    u c  o s  e  (  m  g /   d l    )  

    1  0  4 

     3  5 

     9  3 

    1  4 

    1  6  4 

     4  3 

    1  0  3 

     3  8 

    1  0  3 

    2  9 

    1  0  9 

     3 2 

    1 1  4 

     3  8 

    1  0  7 

     3  6 

     9  8 

     3 1 

     C  ur r  e n t   s m ok  e r  (   % )  

    2  6 

    2  5 

    2  7 

    2  7 

    2  7 

    2 1 

    2  6 

    2 1 

    2  7 

     C  ur r  e n t   a l    c  oh  ol    c  on s  um  p t  i   on (   % )  

    1  6 

    1  7 

    1  5 

    1  9 

     9 

    1  4 

    1  6 

    1  9 

    1  5 

    D e   pr  e  s  s i   on (   % )  

    1  7 

    1  7 

    1  9 

    1  0 

    2  0 

     4 2 

    1  7 

    1  7 

    1  8 

    H  y  p e r  t   e n s i   on (   % )  

     3  5 

     3 1 

     6  4 

     3  3 

     3  6 

     5 2 

     5  6 

     4  0 

    2  5 

    A HI   5  (   % )  

    1 1 

     8 

    1  5 

    1 1 

    1 1 

     9 

     7 

    1 1 

    2  0 

     S l    e  e   p d  ur  a  t  i   on (   % )  

     5 h 

    2 1 

    1  9 

     3  4 

    2 1 

    1  9 

     3  0 

     5 – 6 h 

    2  3 

    2 1 

     3  0 

    2 1 

    2  6 

    2  5 

     6 h 

     5  6 

     5  9 

     3  6 

     5  8 

     5  5 

     4  4 

     S l    e  e   p d i  f   fi  c  ul    t    y

     N or m a l    s l    e  e   pi  n  g

     7  0 

     7 1 

     6  6 

     7  0 

     6  6 

     7 2 

    P  o or  s l    e  e   p

    2 2 

    2 2 

    2  4 

    2  0 

    2  6 

    2 2 

    I  n s  omni   a 

     8 

     7 

    1  0 

    1 1 

     8 

     6 

    D a  t   a  a r  e m e  a n s 

     S D unl    e  s  s i  n d i   c  a  t   e  d  o t  h  e r  wi   s  e  a n d  a r  e  a  d   j   u s  t   e  d f    or  s  a m  pl   i  n  g w e i    gh  t   .A HI   , a   pn e  a -h   y  p o  pn e  a i  n d  e x .

    Insomnia, objective sleep duration, and diabetes

    1982   DIABETES CARE,  VOLUME 32,  NUMBER  11, NOVEMBER  2009 care.diabetesjournals.org

  • 8/19/2019 Dia Care 2009 Vgontzas 1980 5

    4/6

    1.24–7.05]) compared with the groupwho have no insomnia/poor sleep com-plaint and slept for6 h. In addition, the

     joint effect of insomnia and a sleep dura-tion of 5–6 h increased the odds for dia-betes by200% (2.07 [0.68– 6.37]). Theassociation in subjects with poor sleepand a short sleep duration was notsignificant.

    Finally, objective short sleep durationin the absence of a sleep complaint wasassociated with nonsignificant increasedodds for diabetes. The ORs remained verysimilar to those reported in Table 3 afterwe adjusted for number of wakes, num-ber of sleep stage changes, percentage

    of stage 1 sleep, and periodic limbmovements.

    CONCLUSIONS — This is the firststudy to demonstrate that chronic insom-nia associated with objectively measuredshort sleep duration is a clinically signifi-cant risk factor for type 2 diabetes. Thisincreased risk is independent of comor-bid conditions frequently associated withinsomnia or diabetes, such as age, race,obesity, alcohol consumption, smoking,SDB, periodic limb movements, or de-pression. Furthermore, our findings sug-gest that objective measures of sleepduration in insomnia may be a usefulmarker of thebiological severity and med-ical impact of the disorder.

    Several studies have examined the

    association of “sleep disturbances” withdiabetes with inconsistent findings (13–16,19). However, none of these studiesobtained objective sleep data or con-trolled for sleep apnea, a major con-founder, both for sleep disturbance andincreased incidence of type 2 diabetes.

    In our study, the more severe insom-nia (i.e., complaint of insomnia for at least1 year) was significantly associated withhigher odds of diabetes in the basic ad-

     justed model. Most important, severe in-somnia in combination with an objectivesleep duration of   5 h was associatedwith a 300% higher odds for diabetesthan the subjects who did not have a sleepcomplaint and slept for6 h. The secondhighest odds ratio was found in the groupof insomniacs who slept 5– 6 h, althoughthe association did not reach significance.These findings are consistent with thedata on insomnia and hypertension, forwhich the strongest associations werefound in these two groups (5). Further,these data areconsistent with our hypoth-esis and previous physiological studies,which showed that the HPA axis and sym-

    pathetic system activity is elevated ininsomniacs with objective short sleep du-ration (6 –11).

    Insomnia based solely on clinical cri-teria and after adjustment for multiplevariables was not associated with signifi-cant odds for diabetes. This finding isconsistent with the results of the Sleep

    Heart Health Study (SHHS) (20). Further,the milder form of insomnia, i.e., poorsleepcombined with short sleepduration,was not associated with a significant riskfor diabetes. This lack of significance mayreflect, given the large confidence inter-vals, a relatively small number of patientswith diabetes in each subgroup and/orlack of use of more sensitive measures of glycemic status, i.e., a standard oral glu-cose challenge, in our study. An alterna-tive explanation is that a larger degree of sleep disturbance is required to affect gly-

    cemic status compared with blood pres-sure, as is the case with sleep apnea forwhich an association with diabetes hasbeen reported only in those with moresevere apnea (21).

    Experimental studies have shownthat acute or short-term modest sleep lossis associated with impaired glycemic con-trol (22) and inflammation, a conditionthat predisposes an individual to diabetes(23). In this study in a general populationsample, we did not observe an associationbetween objective sleep duration and di-abetes. Several limitations may have re-

    sulted in this lack of significance, such asa relatively small number of patients withdiabetes in each of the nine subgroups(combination of insomnia complaintsand objective sleep measures), lack of more sensitive measures of glycemic sta-tus (i.e., oral glucose tolerance test), and asingle night sleep recording, which maynot be representative of the subjects’ typ-ical sleep patterns. Notably, all previousstudies that have reported an associationbetween sleep duration and diabetes werebased on subject self-report (22). Addi-

    tional studies that include both subjectiveand objective sleep measures and moresensitive methods of glycemic statusshould be conducted to address the issueof long-term sleep restriction anddiabetes.

    The data on the association of insom-nia with hypertension and diabetes, aswell as previous reports on insomnia andthe stress system (6–8) and the auto-nomic system (9–11), provide the basisfor a meaningful subtyping of chronic in-somnia based on objective duration of sleep. One subtype is associated with

    Table 2—Multivariable adjusted ORs (95% CIs) of diabetes associated with insomnia or 

    objective sleep duration

    Model 1 Model 2 Model 3

    Sleep difficulty

    Normal sleeping 1.00 1.00 1.00Poor sleep 1.31 (0.91–1.90) 1.23 (0.84–1.80) 1.22 (0.83–1.78)

    Insomnia 1.84 (1.05–3.20) 1.70 (0.95–3.02) 1.69 (0.95–3.02)Sleep duration (h)

    6 1.00 1.00 1.005–6 1.38 (0.94–2.02) 1.37 (0.93–2.00) 1.35 (0.92–1.98)

    5 1.15 (0.77–1.71) 1.11 (0.74–1.65) 1.08 (0.72–1.61)

    Model 1: adjusted for age, race, sex, BMI, and sampling weight. Model 2: adjusted for age, race, sex, BMI,sampling weight, smoking, alcohol consumption, depression symptoms, and SDB. The interaction betweeninsomnia and sleep duration was not statistically significant. Thus, these two variables were entered into themodel separately, except for model 3, in which the results were adjusted for each other.

    Table 3—Multivariable adjusted ORs (95%

    CIs) of diabetes associated with insomnia andobjective sleep duration

    Sleep difficulty

    and duration

     Adjusted OR 

    (95% CI)*

    Normal sleeping6 h 1.00

    5–6 h 1.45 (0.91–2.30)5 h 1.10 (0.68–1.79)

    Poor sleep6 h 1.52 (0.87–2.65)

    5–6 h 1.55 (0.80–3.01)5 h 1.06 (0.53–2.15)

    Insomnia

    6 h 1.10 (0.40–3.03)5–6 h 2.07 (0.68–6.37)

    5 h 2.95 (1.24–7.03)

    Interaction betweeninsomniaand sleep durationis notstatistically significant, P 0.75. *Adjusted for age,race,sex, BMI,sampling weight, smoking,alcoholcon-sumption, depression symptoms, and SDB.

    Vgontzas and Associates

    care.diabetesjournals.org DIABETES CARE,  VOLUME 32,  NUMBER  11, NOVEMBER  2009   1983

  • 8/19/2019 Dia Care 2009 Vgontzas 1980 5

    5/6

    physiological hyperarousal, i.e., shortsleep duration, activation of the stress sys-tem, and significant medical sequelae,such as hypertension and/or diabetes.The other subtype is not associated withphysiological hyperarousal, i.e., normalsleep duration, normal activity of thestress system, and lack of significant med-

    ical sequelae. The diagnostic validity andutility of this subtyping should be testedin future studies.

    The objective sleep duration in thisstudy was based on one night of poly-somnography, which may not be repre-sentative of the subjects’ habitual sleepduration. However, in our previous stud-ies, the association between objectivesleep duration and hypercortisolemia wasbased on a four consecutive night sleeplaboratory protocol, which should betterrepresent the typical sleep profile of the

    subjects (6,7). It should be noted that, inour study, we investigated the relativesleep duration measured objectively (i.e.,5 h of objective sleep is relativelyshorter than   6 h of sleep). Objectivesleep duration was used as an internallyvalid marker of the severity of insomniaand not as a recommended optimal sleepduration for the general population,which is beyond the scope of our study.Finally, the proposed criterion of 7 h asthe cutoff point for “normal” sleep dura-tion is based on self-reported duration,whereas in large epidemiological studies,

    i.e., the SHHS or Coronary Artery RiskDevelopment in Young Adults (CARDIA)study, the average objective sleep dura-tion is 6 h, which is very similar to thatof our sample (24,25). This duration isindependent of whether sleep was re-corded athome (SHHS) (24) or for 3 con-secutive nights with actigraphy (CARDIAstudy) (25) or in the sleep laboratory(Penn State Cohort) (17,18). The consis-tency of the findings on the role of objec-tive sleep duration in predicting insomniaseverity between the physiological studies

    with multiple night recordings (6,7) andthe current epidemiological study basedon a single night recording increases ourconfidence about the replicability andgeneralizability of the present findings.Future researchers should explore the as-sociation between insomnia, sleep dura-tion, and diabetes using multiple nightrecordings. Finally, our study is cross-sectional and does not provide causalityin terms of the direction of the associa-tion. However, based on large amounts of clinical and research data, which havedocumented that insomnia is associated

    with physiological hyperarousal (2,6–11), the most likely direction is that in-somnia leads to diabetes.

    In summary, insomnia with shortsleep duration is associated with a signif-icant risk for diabetes to a degree compa-rable to that for the other most commonsleep disorder, i.e., SDB (21). Given the

    high prevalence of thedisorderin thegen-eral population and the widespread mis-conception that this is a disorder of the“worried well,” its diagnosis and appro-priate treatment should become the targetof public health policy. Objective mea-sures of sleep duration of insomnia mayserve as clinically useful predictors of themedical severity of chronic insomnia, andthere is a need for validation of practical,easy-to-use, inexpensive methods, e.g.,actigraphy, to measure sleep durationoutside of the sleep laboratory. Finally,

    insomnia with objective short sleep dura-tion may represent a phenotype withinchronic insomnia that may respond dif-ferentially to treatment.

    Acknowledgments— T h i s res earch w asfunded in part by the National Institutes of Health (Grants R01HL 51931, R01HL 40916,and R01HL 64415).

    No potential conflicts of interest relevant tothis article were reported.

     We thank Barbara Green for overall prepa-ration of the manuscript.

    References1. National Institutes of Health. NIH state

    of the science statement on manifesta-tions and management of chronic in-somnia in adults. J Clin Sleep Med2005;1:412–421

    2. Basta M, Chrousos GP, Vela-Bueno A, Vgontzas AN. Chronic insomnia and thestress system. Sleep Med Clin 2007;2:279–291

    3. Phillips B, Mannino DM. Do insomniacomplaints cause hypertension or cardio-

    vascular disease? J Clin Sleep Med 2007;3:489–494

    4. Phillips B, Bůžková, Enright P, Cardio-vascular Health Study Research Group.Insomnia did not predict incident hyper-tension in older adults in the Cardiovas-cular Health Study. Sleep 2009;32:65–72

    5. Vgontzas AN, Liao D, Bixler EO, Chrou-sos GP, Vela-Bueno A. Insomnia with ob- jective short sleep duration is associatedwith a high risk for hypertension. Sleep2009;32:491–497

    6. Vgontzas AN, Tsigos C, Bixler EO,Stratakis CA, Zachman K, Kales A, Vela-BuenoA, ChrousosGP. Chronic insomnia

    and activity of the stress system: a prelim-inary study. J Psychosom Res 1998;45:21–31

    7. Vgontzas AN, Bixler EO, Lin HM, Prolo P,Mastorakos G, Vela-Bueno A, Kales A,Chrousos GP. Chronic insomnia is asso-ciated with nyctohemeral activation of thehypothalamic-pituitary-adrenalaxis: clin-ical implications.J Clin Endocrinol Metab

    2001;86:3787–37948. Rodenbeck A, Cohrs S, Jordan W,

    Huether G, Rüther E, Hajak G. The sleep-improving effects of doxepin are paral-leled by a normalized plasma cortisolsecretion in primary insomnia. Psychop-harmacology 2003;170:423– 428

    9. Stepanski E, Glinn M, Zorick F, Roehrs T,Roth T. Heart rate changes in chronic in-somnia. Stress Med 1994;10:261–266

    10. Bonnet MH, Arand DL. 24-hour meta-bolic rate in insomniacs andmatched nor-mal sleepers. Sleep 1995;18:581–588

    11. Bonnet MH, Arand DL. Heart rate vari-ability in insomniacs and matched normalsleepers. Psychosom Med 1998;60:610–615

    12. Aurora RN, Punjabi NM. Sleep apnea andmetabolic dysfunction: cause or co-rela-tion? Sleep Med Clin 2007;2:237–250

    13. Meisinger C, Heier M, Loewel H,MONICA/KORA Augsburg Cohort Study.Sleep disturbance as a predictor of type 2diabetes mellitus in men and women fromthe general population. Diabetologia 2005;48:235–241

    14. Mallon L, Broman JE, Hetta J. High inci-dence of diabetes in men with sleep com-plaints or short sleep duration: a 12-year

    follow-up study of a middle-aged popula-tion. Diabetes Care 2005;28:2762–276715. Kawakami N, Takatsuka N, Shimizu H.

    Sleep disturbance and onset of type 2diabetes. Diabetes Care 2004;27:282–283

    16. Nilsson PM, Rööst M, Engström G, Hed-blad B, Berglund G. Incidence of diabetesin middle-aged menis relatedto sleep dis-turbances. Diabetes Care 2004;27:2464–2469

    17. Bixler EO, Vgontzas AN, Ten Have T, Ty-son K, Kales A. Effects of age on sleepapnea in men: I. Prevalence and severity. Am J Respir Crit Care Med 1998;

    157:144–14818. Bixler EO, Vgontzas AN, Lin HM, Ten

    Have T, Rein J, Vela-Bueno A, Kales A.Prevalence of sleep-disordered breathingin women: effects of gender. Am J RespirCrit Care Med 2001;163:608– 613

    19. Björkelund C, Bondyr-Carlsson D, Lapi-dus L, Lissner L, Månsson J, Skoog I,Bengtsson C. Sleep disturbances inmidlife unrelated to 32-year diabetes in-cidence. Diabetes Care 2005;28:2739–2744

    20. Gottlieb DJ, Punjabi NM, Newman AB,Resnick HE, Redline S, Baldwin CM, Ni-eto FJ. Association of sleep time with dia-

    Insomnia, objective sleep duration, and diabetes

    1984   DIABETES CARE,  VOLUME 32,  NUMBER  11, NOVEMBER  2009 care.diabetesjournals.org

  • 8/19/2019 Dia Care 2009 Vgontzas 1980 5

    6/6

    betes mellitus and impaired glucosetolerance. Arch Intern Med 2005;165:863–868

    21. Punjabi NM, Shahar E, Redline S, Got-tlieb DJ, Givelber R, Resnick HE, SleepHeart Health Study Investigators. Sleep-disordered breathing, glucose intoler-ance, and insulin resistance: the SleepHeart Health Study. Am J Epidemiol

    2004;160:521–53022. Spiegel K, Knutson K, Leproult R, Tasali

    E, Van Cauter E. Sleep loss: a novel riskfactor for insulin resistance and type 2 di-abetes. J Appl Physiol 2005;99:2008–2019

    23. Vgontzas AN, Zoumakis E, Bixler EO, LinHM, Follett H, Kales A, ChrousosGP. Ad-verse effectsof modestsleep restriction onsleepiness, performance, and inflamma-tory cytokines. J Clin Endocrinol Metab

    2004;89:2119–212624. Silva GE,Goodwin JL,Sherrill DL,Arnold

     JL, Bootzin RR, Smith T, Walsleben JA,Baldwin CM, Quan SF. Relationship be-tween reported and measured sleeptimes:the Sleep Heart Health Study (SHHS). J Clin Sleep Med 2007;3:622–630

    25. Lauderdale DS, Knutson KL, Yan LL,Rathouz PJ, Hulley SB, Sidney S, Liu K.Objectively measured sleep characteris-tics among early-middle-aged adults: the

    CARDIA study. Am J Epidemiol 2006;164:5–16

    Vgontzas and Associates

    care.diabetesjournals.org DIABETES CARE,  VOLUME 32,  NUMBER  11, NOVEMBER  2009   1985