Sleep complaints predict coronary artery disease mortality in males: a 12-year follow-up study of a middle-aged Swedish population
Lena Mallon MD PhD Sleep Disorders Unit, Department of Neuroscience, Psychiatry, University Hospital, SE-751 85 Uppsala, Sweden (fax: +46 18 51 58 10; e-mail: email@example.com).
Abstract. Mallon L, Broman J-E, Hetta J (University Hospital, Uppsala, and Falun Hospital, Falun; and Institute of Clinical Neuroscience, Göteborg; Sweden). Sleep complaints predict coronary artery disease mortality in males: a 12-year follow-up study of a middle-aged Swedish population. J Intern Med 2002; 251: 207–216.
Objectives. Only a few prospective surveys have been performed to investigate the relationship between sleep complaints and coronary artery disease (CAD) mortality. This study was conducted to determine whether sleep complaints in a middle-aged population predicted total mortality and CAD mortality.
Design. A population-based prospective study.
Setting. The County of Dalarna, Sweden.
Subjects and methods. In 1983, a random sample of 1870 subjects aged 45–65 years responded to a postal questionnaire (response rate 70.2%) including questions about sleep complaints and various diseases. Mortality data for the period 1983–95 were collected, and Cox proportional hazard analyses were used to examine the mortality risks.
Results. At 12-year follow-up 165 males (18.2%) and 101 females (10.5%) had died. After adjustment for a wide range of important putative risk factors, difficulties initiating sleep (DIS) were related to CAD death in males [relative risk (RR), 3.1; 95% confidence interval (CI), 1.5–6.3; P < 0.01], but not in females. Short or long sleep duration did not influence risk of CAD mortality or total mortality for either gender. Depression in males increased the risk of death attributed to CAD (RR, 3.0; 95% CI, 1.1–8.4; P < 0.05) and total mortality (RR, 2.2; 95% CI, 1.1–4.5; P < 0.05).
Conclusion. These results provide evidence that there is an association between difficulties falling asleep and CAD mortality in males.
Prevalence estimates of insomnia in the general population are in the range of 10–50%, and variations are dependent on the definition of insomnia, the population examined and survey methods. Cross-sectional epidemiological studies have demonstrated that various psychosocial, medical and psychiatric conditions are correlated with insomnia [1–6]. The long-term effects of insomnia have, however, been little explored and knowledge of how insomnia may be associated with total mortality and especially coronary artery disease (CAD) mortality is still limited.
In the few prospective surveys hitherto performed to investigate the relationship between insomnia and total mortality, the results have yielded somewhat divergent results. An increased death rate has been reported in subjects with a complaint of insomnia . Increased risk ratios for mortality have also been found in males reporting bad sleep quality , and in older males with insomnia . In other surveys of the elderly insomnia was not, however, related to increased mortality [10–12].
An important question concerns whether the studies on the association between insomnia and mortality have either adjusted for or ruled out other risk factors as confounders. One important confounder is the occurrence of depression [13, 14]. Not only are sleep complaints and altered sleep patterns diagnostic criteria for major depression [15, 16], but depression also increases the risk of death from CAD, especially in males .
Studies have demonstrated an increased risk of CAD morbidity and death in subjects with sleep complaints. In males, reports of bad sleep quality increased the risk of death or hospitalization because of ischaemic heart disease , problems falling asleep increased the risk of myocardial infarction and cardiac death , and a higher risk of developing myocardial infarction was found in males with severe sleep disturbances . In a survey of females difficulty falling asleep predicted myocardial infarction or coronary death , and in older adults trouble falling asleep increased the risk of nonfatal and fatal infarction . In a recent review of insomnia and heart disease, Schwartz et al.  point out the importance of controlling for depression. However, only one of the epidemiological studies performed so far investigating the association between sleep complaints and CAD death simultaneously controlled for the presence of depression, and this study was carried out in adults aged 65 or older .
Also, there is evidence that short or long sleep duration is associated with an increased risk of death [7, 23]. The associations between duration of sleep and total mortality were regardless of whether the subject reported insomnia or not, suggesting that duration of sleep is an independent and valid indicator of risk. Considering cause-specific mortality, the risk of stroke was increased in persons who slept more than 8 h in a survey of adults, but sleep duration was not related to hospitalization or death from CAD .
Given the paucity of data from prospective studies of sleep complaints and mortality in middle-aged populations, our purpose was to provide data on the relationship of sleep complaints to risk of CAD death and total mortality whilst controlling for depression, sleep duration and other important risk factors.
In 1983, a random sample of 2663 males and females aged 45–65 years was selected using the population register of the County of Dalarna in central Sweden. A postal questionnaire was sent and 1870 (70.2%) subjects responded after one reminder, 906 males and 964 females. The mean (±SD) age was 56 ± 6 years, and the responders were found representative of the total population sample selected .
The questionnaire contained questions about demographic and life style variables, medical conditions, sleeping difficulties, sleep habits and medication. A detailed description of the questionnaire has been published previously .
Marital status was defined as married or not married (i.e. single, divorced, widowed or separated), living conditions were defined as living alone or with someone else and smoking was defined as currently smoking or not. Questions were asked about body weight and height, and body mass index (BMI, in kg m–2) was calculated. Medical conditions were ascertained by asking subjects (in a yes/no question) if they had a history of heart disease, hypertension, respiratory disease (asthma, bronchitis), diabetes, joint pain, gastrointestinal disease or urogenital disease. Depression was assessed by the question ‘Do you feel depressed?’ to be answered by yes or no. The validity of single-item depression was tested against the HAD Depression Subscale (HAD-D)  amongst the survivors in 1995, and we found that 81.0% of subjects scoring above 8 on the HAD-D (possible depression) reported depressed mood in the single question .
Duration of sleep and time to fall asleep (sleep latency) were assessed by asking subjects to estimate total night sleep time and the average time it took them to fall asleep in the evening, and the answers were expressed as continuous variables. The questions about difficulties initiating sleep (DIS) and difficulties maintaining sleep (DMS) were adopted from the Uppsala Sleep Inventory (USI), which has been used previously in epidemiological studies [28, 29]. The questions concern severity of sleep complaints (1=no problems, 2=minor problems, 3=moderate problems, 4=severe problems, 5=very severe problems). At least moderate problems (scores 3–5) were considered to be a complaint. The questions about snoring and sleeping pill utilization were also from the USI, and they concern frequency of occurrence (1=never, 2=seldom, 3=occasionally, 4=often, 5=very often). Responders with scores of 4 or 5 were regarded as habitual snorers or sleeping pill users.
Mortality data for the 12-year period 1983–95 were collected from both the responders and the nonresponders of the questionnaire. By November 1995, 266 of the responders and 157 of the nonresponders had died. Deaths certificates were obtained from the National Cause of Death Register in Sweden and death certificates were available for all the 423 deceased.
Using the International Classification of Diseases, Ninth Revision (ICD-9), causes of death were coded and grouped as follows: CAD (codes 410–414), cancer (codes 140–239) and ‘all other causes’ being non-CAD, noncancer conditions and death from unnatural causes, i.e. injuries and poisonings (800–999) and E-codes.
The study was approved by the Ethics Committee of the Faculty of Medicine at Uppsala University in Sweden.
The statistical analyses were performed on a Macintosh computer using the statistical analysis program SPSS 6.1 (SPSS Inc., Chicago, IL, USA). Standard methods have been used to calculate means and standard deviations. For comparison between categorical variables in the baseline survey the chi-square test was used and when the comparison involved continuous variables the t-test was used. To study the influence of possible explanatory variables on mortality, the Cox proportional hazard model was used. When analysing determinants of cause-specific mortality, subjects who died of other causes were treated as censored observations.
The results are presented as adjusted relative risks (RR) with 95% confidence interval (CI). The minimum statistical significance level for all analyses was P < 0.05. Separate analyses were developed for males and females in order to explore potential gender differences in the nature of associations, and estimates were adjusted for age in 5-year strata. Univariate Cox proportional hazard models were used to estimate the risk of death for the covariates as independent factors. Multivariate Cox proportional hazard models were used to assess whether socio-demographic factors, medical conditions or mood disturbance were potential effect modifiers or confounders of the relationship between sleep complaints and mortality. Variables were added to proportional hazard model of the relationship using a forward step-wise approach. The final model included those variables that were significantly related to risk of dying (P < 0.05).
Amongst the 1870 responders to the questionnaire 266 deaths occurred during the follow-up period, 165 males (mortality rate 18.2) and 101 females (mortality rate 10.5). Death certificates were available for all the deceased, and were based on autopsies in 118 cases (44.4%). The mean age at death was 66.1 ± 6.6 years. The primary cause of death was CAD in 34.2% (91 subjects) and cancer in 31.2% (83 subjects). Of ‘all other causes’, 30.1% (80 subjects) died of non-CAD or noncancer conditions, and 4.5% (12 subjects) died of unnatural causes.
Amongst the 793 nonresponders there were 157 deaths, 95 males (mortality rate 24.7) and 62 females (mortality rate 15.8). The mean age at death was about the same as for responders, 66.5 ± 6.7 years. The primary cause of death was CAD in 33.1% (52 subjects) and cancer in 28.0% (44 subjects). Of ‘all other causes’, 33.1% (52 subjects) died of non-CAD or noncancer conditions, and 5.7% (nine subjects) died of unnatural causes. There was no significant difference in primary cause of death between the responders and the nonresponders.
Table 1 shows the distribution of background characteristics and risk factors for males and females at baseline. There was a significant male preponderance in reports of cardiac disease, diabetes, snoring and urogenital disease, the latter as a result of prostate complaints amongst males. Females reported more often DIS, DMS and depression. Reported duration of sleep was about the same for males and females, 7.0 ± 1.1 h per night. Females had longer sleep latencies, 24 ± 37 min, compared with males, 18 ± 29 min (t=3.5; P < 0.001). Subjects reporting DIS had longer sleep latencies than subjects not reporting DIS, 53 ± 50 vs. 10 ± 12 min (t=27.6; P < 0.001).
Prevalence of variables in the baseline survey in males (n
=906) and females (n
Table 2 shows that after adjustment for age, sleeping more than 8 h per night, DIS, DMS and habitual sleeping pill usage were amongst those variables related to total mortality in males. In regard to cause-specific mortality in males we found that DIS was related to CAD death. Habitual sleeping pill usage was related to cancer death, DIS and DMS were associated with death from ‘all other causes’. After further adjustments for the influence of potential risk factors in multivariate survival analysis, none of the sleep-related variables remained significantly related to total mortality in males (Table 3). When examining the relationship between sleep-related variables and cause-specific mortality we found that DIS were related to an increased risk of death from CAD, habitual sleeping pill usage increased the risk of cancer death, and DMS were related to death from ‘all other causes’ in males. Furthermore, depression in males was associated with excess total mortality and CAD mortality.
Relative risk (RR) [95% confidence interval (CI)] of age-adjusteda
mortality in males (n
=906) by using the Cox proportional hazard model
Relative risk (RR) [95% confidence interval (CI)] of multivariate adjusteda
mortality in males (n
=906) by using the Cox proportional hazard model, forward stepwise procedure
In females DIS and habitual sleeping pill usage were related to total mortality in age-adjusted analyses, whereas habitual sleeping pill usage was associated with death from ‘all other causes’ (Table 4). In females we made adjustments for the same potential risk factors as in males with the exception of urogenital disease, because of the low prevalence (0.4%) in the baseline survey. When adjusting for putative risk factors in females, death from ‘all other causes’ was increased in females reporting DIS or habitual sleeping pill usage (Table 5).
Relative risk (RR) [95% confidence interval (CI)] of age-adjusteda
mortality in females (n
=964) by using the Cox proportional hazard model
Relative risk (RR) [95% confidence interval (CI)] of multivariate adjusteda
mortality in females (n
=964) by using the Cox proportional hazard model, forward stepwise procedure
The main finding of the present study is that in males DIS is an independent risk factor of CAD death. Short or long sleep duration did not influence risk of CAD mortality or total mortality for either gender. Another important confirmative finding was that depression in males was associated with CAD mortality and total mortality.
To our knowledge, this is the first prospective epidemiological study of middle-aged subjects in which the association between sleep complaints and CAD mortality has been examined whilst controlling for depression as well as other important risk factors. The strengths of this study include the representativeness of the baseline sample and the fact that the follow-up period is longer than in most epidemiological surveys within this field. As Kripke et al. have pointed out, selecting variables to be included in multivariate analysis requires some considerations. Controlling for risk factors might lead to both over- and underestimation of associated mortality risks . In this study, variables were included in the survival analysis if they were considered to be relevant confounders and sufficiently frequent in the baseline survey.
In the analyses we used available information about health status obtained from the questionnaire. Risk factors examined were: age in 5-year strata, marital status, living arrangements, smoking status, BMI, the occurrence of medical conditions (hypertension, cardiac disease, respiratory disease, diabetes, joint pain, gastrointestinal disease, urogenital disease) and mood disturbance. These factors are widely considered to be correlates or risk factors for premature death, and on a priori grounds these were taken as being potentially related to mortality. Sleep complaints were measured using two items that inquired about DIS and DMS. Also, short and long sleep duration, snoring and sleeping pill usage were examined as they all have been found to relate to excess mortality [7, 23, 29, 30]. In common with the dataset described by Kripke et al. , responses to the item on sleeping pills were taken at face value.
A limitation is that all data were collected in 1983 only, and any changes in risk factors during the follow-up period could not be assessed. Another limitation is that we did not have objective data on actual sleep in relation to complaints. However, the definition of insomnia is dependent on the subjective experience of disturbed sleep . Subjective sleep complaints will also decide if the person will seek medical advice or treatment. Furthermore, self-reports of sleep are the measures most widely used in community-based epidemiological surveys and there is evidence that subjective evaluations correlate adequately with objective laboratory data [32–34].
The mortality rate was higher amongst the nonresponders, indicating that responders were healthier than nonresponders. Higher mortality amongst nonresponders has also been reported from other surveys conducted in Sweden . Well known risk factors for premature mortality such as smoking, diabetes and cardiovascular disease were verified. As reported by others we found gender differences in mortality and risk factors . A female advantage in respect to mortality was found. Risk factors in which gender differences were seen included medical conditions, lifestyle characteristics as well as sleep-related variables.
The insomnia variables, DIS and DMS, were not related to total mortality when adjustment was made for other potential risk factors, indicating that the associations were mediated through other risk factors. However, males with DIS had a threefold increased risk of death from CAD, and this increased risk was similar to that conferred by other CAD risk factors. Contrary to a study of older subjects  this relationship remained significant after controlling for the effect of important CAD risk factors including depression.
Studies have suggested that long-lasting psychological stress predict coronary events, probably mediated through increased sympathetic nervous system activity . Data have also demonstrated that individuals with chronic insomnia suffer from increased arousal and increased sympathetic activation, implying that they could be at increased risk for the development of CAD [38–40]. The activity of the stress system has been found to relate positively to the degree of objective sleep disturbance . Based on epidemiological data it has been shown that individuals with insomnia complaints continue to have this problem for a long time, and thus can be considered as a chronic problem for many sufferers [42, 43].
It has been suggested that DIS are part of a vital exhaustion syndrome that is distinct from depression [18, 44] or an indicator of possible adjimive breakdown in males . Our finding of DIS being predictive of death from CAD in males provides additional evidence for a link between enduring mental stress and CAD, and suggests that DIS are an important marker for psychological stress possibly related to increased sympathetic nervous system activity.
Females with DIS had an almost threefold increased risk of death from ‘all other causes’, suggesting that also in females DIS could cause excessive mortality. Moreover, DMS in males were related to death from ‘all other causes’ thereby demonstrating that disruption of sleep during the night increases the risk of insufficient sleep periods  that appear to have an impact on future health.
We found no relationship between short or long sleep and survival when the confounding effect of depression and other risk factors were controlled for. Whether deviant sleep duration is an independent risk factor for excess mortality is still controversial. Kripke et al.  reported that both short and long sleep duration was associated with increased mortality in a 6-year follow-up study, and Wingard and Berkman  reported similar results from a 9-year follow-up. However, the 17-year follow-up data of the latter revealed no association between sleep duration and mortality . Also, in the only survey that controlled for depression Pollack et al.  were unable to find an association between sleep duration and survival in a 3.5-year follow-up of elderly people.
Consistent with previous surveys conducted in Sweden, we found no significant association between reported snoring and mortality . Lindberg et al. found, however, an increased mortality in males reporting both snoring and excessive daytime sleepiness, suggesting that snoring without excessive daytime sleepiness does not appear to influence the mortality rate.
Habitual sleeping pill usage was related to death from cancer in males and death from ‘all other causes’ in females. Both the first and second Cancer Prevention Study (CPS I and II) have reported associations between early death and sleeping pill utilization [7, 30]. It has also been demonstrated that regular users of hypnotics often have disabling psychiatric and somatic conditions [48, 49], and are frequently treated with other psychoactive drugs [48, 50]. In one survey more than one-third of female hypnotic users were receiving health care not for sleep problems but for depression, anxiety and other reasons . It has been observed that the sales of tranquillizers and hypnotics/sedatives are a marker for poor socio-economic conditions and correlate with both mortality and suicide . Furthermore, males using hypnotics habitually were found to have higher rates of psychiatric admissions as a result of substance abuse whilst females had an increased risk of suicide .
We cannot determine what ‘sleeping pills’ the responders were taking. It is possible that some of the ‘sleeping pills’ were not hypnotics or sedatives but other drugs, taken to promote sleep (e.g. antipsychotics or antidepressives). Moreover, we cannot rule out the possibility of habitual sleeping pill utilization being a proxy for potentially important confounding variables that were not included in the model.
It is possible that the deaths statistically associated with sleeping pill usage are not caused by sleeping pill usage because evidence of association does nor prove causality. Furthermore, it is much more difficult to analyse treatment variables than symptom variables. One would anticipate that subjects with serious illnesses would be more likely to use sleeping pills than healthy subjects and an association of sleeping pills with excess mortality would be expected even if sleeping pills have no effect whatsoever on mortality. Thus, sleeping pill usage is an indicator of illness rather than its cause. Sleeping pill usage may be related to so many comorbidities that the epidemiological method could never conceivably control for all possible confounders .
Furthermore, depression was related to excess total mortality and death from CAD in males. A limitation is that no explicit depression screening scale was included in the questionnaire. Depression was defined by answering ‘yes’ to the question ‘Do you feel depressed?’ In a review Schade et al.  found that brief instruments perform about as well as longer questionnaires for screening depression in general populations. The value of single-item depression screening has previously been noted in primary care patients , in older adults , in patients with anxiety disorders  and in the terminally ill . Our measurement of depressed mood also gave similar associations with mortality as other studies using various screening methods. In a recent review of the mortality of depression, Wulsin et al.  found that depression seems to increase the risk of death attributable to CAD, especially in males.
In conclusion, our findings demonstrate that DIS are an important risk factor for death from CAD in males. More knowledge within this field can possibly give opportunities for intervention.
This study was supported financially by the Swedish Medical Research Council (Project 06869), the Dalarna Research Institute, the Swedish Psychiatric Association, the Marta and Nicke Nasvells Foundation, Bror Gadelius Memory Fund, Gun and Bertil Stohnes Foundation and the Swedish Lundbeck Foundation.