| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Departments of Psychiatry and Family Medicine, University of Cincinnati, Cincinnati, Ohio (L.R.W.); The National Heart, Lung, & Blood Institute's Framingham Heart Study, Framingham, Massachusetts (J.C.E., R.S.V., J.M.M., M.K.-H., E.J.B.); The Division of Cardiology & Preventive Medicine and Neurology, Boston Medical Center, Boston, Massachusetts (J.C.E., R.S.V., M.K.-H., E.J.B.); The Section of General Internal Medicine, Boston Medical Center, Boston, Massachusetts (J.M.M.).
Address correspondence and reprint requests to Lawson R. Wulsin, MD, 231 Albert Sabin Way, ML 559, Cincinnati, OH 45267-0559. E-mail: Lawson.wulsin{at}uc.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Methods: Using multivariable-adjusted sex-stratified Cox proportional hazards regression, we examined the association between depressive symptoms and incident coronary disease and all-cause mortality in 3634 Framingham Heart Study original and offspring cohort participants (mean age 52 years, 55% women) attending a routine study examination between 1983 and 1994.
Results: Over 6 years of follow-up, 83 participants had a hard coronary heart disease event (myocardial infarction or coronary death), and 133 died. Depressive symptoms (Center for Epidemiologic Studies Depression Scale (CES-D) 
16) did not predict hard coronary disease events. All-cause mortality, however, was directly associated with depressive symptoms. Compared with the lowest tertile of CES-D score, multivariable-adjusted risks of death in the second and third tertiles were 33% and 88% higher, respectively (hazards ratio per tertile increment = 1.37, 95% confidence interval 1.10–1.71, p for trend = 0.005).
Conclusion: These findings underscore the importance of further research into the pathogenesis and prevention of excess mortality experienced with depressive symptoms.
Key Words: depressive symptoms • epidemiology • incidence • coronary heart disease • mortality • Framingham Heart Study
Abbreviations: 95% CI = 95% confidence interval; BMI = body mass index; CES-D = Center for Epidemiologic Studies Depression Scale; CHD = coronary heart disease; RR = relative risk.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
The public health importance of the relations between depression and coronary disease and depression and mortality calls for further observational as well as intervention studies. Community-based studies (6,7) have found that adjustment for confounding by physical illness attenuates the relations between depressive symptoms and coronary disease incidence or all-cause mortality. The role of gender in these relationships remains unclear, with no clear pattern across studies (8). The relationship between depression and coronary disease is complex, and several key questions remain unanswered: (1) does the effect of depression on the risk for incident coronary disease or total mortality vary according to age and sex?; and (2) does the severity of depression affect the risk for coronary disease incidence or total mortality?
Our observational study aimed to address these questions by prospectively testing the hypothesis that high levels of depressive symptoms predict both coronary disease incidence and all-cause mortality in a large community-based sample free of baseline CHD and followed for up to 6 years.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Depressive Symptoms
Depressive symptoms were evaluated using the CES-D scale (9). This 20-item, self-report scale, designed for use in community studies, has been validated in cohorts similar to the Framingham Heart Study (10–12). The CES-D scale was self-administered in the offspring cohort but technician-administered in the original cohort because of their advanced age. The CES-D scale asks participants how often they have been bothered by specific symptoms during the previous week. Total scores can range from 0 to 60, with a score of 16 or above generally considered to indicate clinical depression (9,10). We examined the CES-D data in 3 ways: as a continuous variable, a dichotomous variable (score <16 versus 16), and according to sex-specific tertiles.
For participants with 4 or fewer missing items on the CES-D scale, we took a conservative approach and set the missing values to zero. The internal consistency and reliability of the CES-D scale was very good (standardized Cronbach's 
= 0.86). One hundred ninety participants, 5.2% of the total study sample, had between 1 and 4 missing items from the CES-D scale and were retained in our sample. Among those with missing items, the vast majority had only 1 item missing (82%), 13% of them had 2, and 5% had 3 or 4 items missing.
Clinical Measures
Framingham Heart Study participants are reexamined approximately every 2 (original cohort) to 4 (offspring cohort) years. Examinations include routine medical history, physical examination, blood pressure measurements, 12-lead electrocardiogram, and laboratory assessment of cardiovascular risk factors. Covariates were defined at the baseline examinations. Hypertension was defined as a systolic blood pressure 140 mm Hg, a diastolic blood pressure 90 mm Hg, or current use of antihypertensive medication. Diabetes was defined as a fasting blood sugar 126 mg/dl, a random blood sugar 200 mg/dl, or use of oral hypoglycemic or insulin medication. Cigarette smoking (regularly within the past year) was self-reported. Alcohol (self-reported ounces per week), serum total cholesterol, and body mass index (BMI, kg/m2) were used as continuous variables in the models.
Coronary Heart Disease (CHD) and Mortality Outcomes
Participants were followed for a mean of 5.9 years (range 0.1–6 years) after completing the baseline examination. Two principal study outcomes were evaluated: (1) hard CHD incidence, comprising recognized myocardial infarction and CHD death, (sudden or nonsudden CHD death) and (2) death from all causes. In secondary analyses, we additionally included angina pectoris and coronary insufficiency as CHD outcomes. Events were reviewed using all pertinent available medical records and adjudicated based on previously published criteria (13,14) by a panel of 3 physician investigators, blinded to the participant's CES-D score.
Statistical Analyses
Statistical analysis was performed using SAS software (version 6.12) (15). Differences in unadjusted baseline characteristics between participants with and without depressive symptoms were analyzed using 
2 tests or t tests, as appropriate. Cox proportional hazards regression was used to estimate the risks for incident CHD and all-cause mortality associated with our 3 methods of analyzing CES-D depressive symptoms adjusting for age, sex, smoking, hypertension, diabetes, BMI, total cholesterol and reported alcohol use. Both sex-specific and sex-stratified models were considered. We tested for effect modification by sex and by age (<65 versus 65 years).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
16) were reported more often in women, current smokers, and younger participants (Table 1). Hypertension, diabetes, total cholesterol, and BMI did not differ appreciably in those with and without reported depression.
|
The mean CES-D score for the total sample was 8.1 ± 7.7 (range 0–51). Overall, 14% of participants scored 16 on the CES-D, comprising 10% of men and 17% of women. Figure 1 shows the distribution of CES-D scores 16 according to age and sex; women's scores exceeded men's at all ages. Depressive symptoms declined with age in both men and women.
|
Over 6 years of follow-up, we observed 83 incident hard CHD events (58 men, 25 women), 121 incident total CHD events (91 men, 30 women), and 133 deaths from all causes (60 men, 73 women). Thirty-nine participants died from cardiovascular disease, 40 from cancer, 48 due to other causes (including 2 participants who committed suicide 4 years after the examination, who had CES-D scores of 0 and 2), and 6 for whom the cause of death was unknown. For both hard CHD and CHD including angina and coronary insufficiency, crude incidence rates per 10,000 person-years' follow-up were not higher in those reporting depressive symptoms, nor was there a trend toward higher incidence rates across CES-D tertiles (Table 2). Crude all-cause mortality rates, however, were higher in those reporting depressive symptoms. The length of follow-up was uniform across the range of CES-D scores. In subjects experiencing events, the average (mean) time to event across CES-D tertiles ranged from 3.0 to 3.1 years for hard CHD, 2.7 to 2.9 years for CHD including angina and coronary insufficiency, and 3.6 to 3.8 years for mortality. Over 90% of subjects without events completed 6 years of follow-up.
|
Effect modification by sex was evaluated using both sex-specific Cox proportional hazards regression models and sex-adjusted Cox models including interaction terms. The relation of CES-D depressive symptoms to CHD incidence and all-cause mortality was similar in men and women, and no statistically significant interactions were found. Therefore, both sexes were included in all Cox regression models presented, with stratification by sex.
Results of Cox proportional hazards regression models for mortality are shown in Table 3. The sex-stratified, age-adjusted risk for all-cause mortality in the highest tertile of CES-D was twice that of the lowest tertile (hazards ratio = 2.07, 95% confidence interval [CI] 1.34–3.18) and the trend for increased mortality across CES-D tertiles was statistically significant (p < .001). Multivariable-adjusted analysis only slightly attenuated these associations; the risk in the third tertile of CES-D was 1.88 (95% CI 1.22–2.91) times higher than the risk in the first tertile. CES-D analyzed as a continuous measure was also significantly related to increased all-cause mortality.
|
In sex-stratified, age-adjusted models, the test of trend across CES-D tertiles for increased hard CHD risk was of borderline statistical significance (p = .055, Table 4). Multivariable-adjusted models, however did not demonstrate a statistically significant increased risk for hard CHD with increasing CES-D. The results were not substantively different in a secondary analysis (data not shown), in which the definition of CHD was extended to include angina and coronary insufficiency.
|
We examined our power to detect associations between CES-D and the events analyzed. For hard CHD, we had 80% power to detect multivariable-adjusted hazards ratios of 1.5 or greater for trend across tertiles of CES-D. For all-cause mortality, however, we had 80% power across CES-D tertiles to detect a multivariable-adjusted hazards ratio as low as 1.37, the same as the point estimate we obtained. The association we observed between CES-D and all-cause mortality was not clearly driven by any particular cause of death. Of the 133 deaths over 6 years, cancer (n = 40) and cardiovascular disease (n = 39) each accounted for 30%, leaving 40% of all mortality attributed to other causes.
In secondary analyses, effect modification by age was evaluated by comparing subjects younger than 65 years with those 65 years and older. For both hard CHD and mortality, the point estimates obtained from separate models were similar and, in models including age, the test for an interaction between age and CES-D was not statistically significant (data not shown). We also examined effect modification by cohort, comparing the older original cohort with the younger offspring cohort, and found no effect modification by cohort membership.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Our data demonstrated a dose-response relationship between severity of depressive symptoms and death from all causes over 6 years of follow-up. Mortality was 88% higher in the highest tertile of CES-D score compared with the lowest tertile adjusting for age, smoking, hypertension, diabetes, cholesterol, BMI, and alcohol consumption. This dose-response relationship shows that the effect on mortality is not limited to severe depression. The association between depressive symptoms and all-cause mortality was consistent with the majority of methodologically strong community studies (3–8), which have shown that depression contributes to death (5,16), primarily through pathways other than suicide. Our finding of a dose-response relationship between depressive symptoms and mortality also supports the findings of recent reports, one a large community study in Amsterdam (17), one a study of a large cardiac sample in Montreal (18), and one a large study of women (9), all of which reported markedly increased mortality risk with increasing severity of depression after adjusting for major cardiac risk factors.
Our finding that depressive symptoms did not predict incident CHD contrasts with the majority of comparable community studies, which have found that depressive symptoms predicted the onset of coronary disease over durations of follow-up ranging from 4 to 40 years (3). For example, in a study of 7800 adults in the National Health and Nutrition Examination (NHANES 1) followed for 10 years, Ferketich and colleagues (19) reported a statistically significant increased relative risk (RR) for incident CHD of 1.7 in both men and women with depressive symptoms. In the Ferketich et al. (19) study, depression was also associated with increased risk for CHD mortality in men, but not in women. However, consistent with our findings, other population-based studies have not found a significant association between depressive symptoms and CHD in persons free of the disease at baseline, adjusting for important risk factors and health status (10,11). Depressive symptoms may not be an independent risk factor for coronary disease in the general population but may be important in particular subgroups such as high functioning healthy older women (12).
There are several possible reasons why our results do not agree with the meta-analyses that found a positive relationship between depression and incident CHD, including a lack of statistical power in our study; differences in the composition of the study samples; misclassification of the exposure, depression; misclassification of the outcome, CHD; or positive publication bias. Compared with 10 similar incidence studies we reviewed (3), our sample was clearly younger at the beginning of follow up than 7 of the samples. Our follow-up period was also shorter than all but 2 of these studies. A combination of the relatively less sensitive measure of depression (CES-D versus structured interview), and short follow-up in a young sample may have limited our ability to detect incidence rates of the magnitude described elsewhere. In addition, we had limited power to detect a modest association (RR < 1.5) between CES-D and incident CHD. The CHD incidence rate in our relatively young sample (mean age 53) was low.
Two systematic reviews (3,4), using similar meta-analytic methodology but slightly different selection criteria, pooled 10 and 11 cohort studies, respectively, and both found nearly identical results, with depression contributing an estimated combined overall RR of 1.64 for incident CHD (95% CI 1.40–1.9 in one report (3) and 1.3–2.1 in the other (4)). There was an overlap of 7 studies between the 2 meta-analyses. These reviews suggested that depression may be a modest but significant risk factor for the onset of coronary disease.
The CES-D instrument may lack sensitivity for the diagnosis of depression, and may lead to misclassification of the exposure. Although the measure of depression in our study was limited to a single CES-D administration, 3 other studies (19–21) have reported significant risks for incident CHD using the same single administration of the CES-D questionnaire to define exposure while adjusting for traditional cardiac risk factors. The durations of follow-up in 2 of these studies were similar to ours, 4 to 6 years. Nonetheless, the type of depression measure undoubtedly alters the effect size of the relation. Rugulies (4) reported in his review that a sensitivity analysis showed that clinical depression assessed by a structured diagnostic interview (RR = 2.69, 95% CI 1.63–4.43, p < .001) was a stronger predictor of CHD onset than depressive symptoms assessed by self-report questionnaires (RR = 1.49, 95% CI 1.16–1.92, p = .02).
It is unlikely that our negative findings for CHD incidence are due to misclassification of the outcome. The coding of CHD events was made by a panel of 3 investigators using established written diagnostic guidelines, blinded to knowledge of CES-D data. We examined both hard CHD and all CHD events and failed to find an association.
Although age was related to CHD incidence and overall mortality, we found no interaction between age and sex for depression's relation to either outcome. This finding suggests that in this sample over this period of follow-up, age and sex did not alter the outcomes. Gender differences in depression and heart disease remain a puzzle. With respect to severity of depression, however, the dose-response relationship between depressive symptoms and overall mortality in our study also suggests that risk for death becomes significant when the level of severity of depressive symptoms is in the highest tertile of the CES-D score (9 in men and 11 in women), which includes a substantial number of people with mild to moderate depressive symptoms. That is, as others have shown (17,18), the increased mortality risk is not limited to those with severe depressive symptoms, although in general the incidence and mortality risks for major depression exceed those for minor depression (4,17).
Because our assessment of depressive symptoms was made at a single point in time, this study cannot address the question of the effect of duration of depression on CHD incidence or overall mortality. Future research must account for duration, as well as severity of depressive symptoms, because the allostatic load of clinical depression over time is likely to play an important role in both the onset and the progression of coronary disease. The best method for assessing duration, as well as severity, of depression is repeated measures of depressive symptoms in combination with structured interviews of the person's lifetime history of depressive disorders (8).
However, our finding that depressive symptoms were related to mortality but not CHD incidence raises the question of whether the mechanisms that contribute to the development of CHD differ from the mechanisms that contribute to cardiac mortality or mortality by other causes. For example, depression that alters heart rate variability or platelet aggregation may contribute to sudden cardiac death, whereas depression that alters inflammatory response (22,23) or the hypothalamic-pituitary adrenal axis stress response may contribute more to the gradual development of early CHD (5,24–26). No studies have yet examined the relative importance of the effects of these mechanisms in the development of CHD compared with the progression or mortality of CHD. Furthermore, few studies have examined the mechanisms by which depression contributes to mortality by noncardiac causes (5).
Limitations and Strengths
Our study sample is predominantly Caucasian, so the generalizability to other ethnicities/races is uncertain. In addition, we lacked power to detect a very modest effect of depressive symptoms on risk of incident CHD (hazard ratio <1.5). It will be possible to examine the relation between depressive symptoms and CHD more comprehensively and with greater statistical power as the offspring cohort ages, assessments of depressive symptoms are repeated, and the incidence of coronary disease increases. Our study's strengths include the community-based cohort design, the routine assessment of depressive symptoms, frequent updating of covariate data, routine longitudinal follow-up, and rigorous coding of cardiovascular disease and mortality outcomes.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
This work was supported by grants NIH/NHLBI N01-HC-25195, NINDS 5RO1-NS-17950, K24-HL-04334 (to R.S.V.).
DOI:10.1097/01.psy.0000181274.56785.28
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  G. A. Wellenius, K. J. Mukamal, A. Kulshreshtha, S. Asonganyi, and M. A. Mittleman Depressive Symptoms and the Risk of Atherosclerotic Progression Among Patients With Coronary Artery Bypass Grafts Circulation, May 6, 2008; 117(18): 2313 - 2319. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W. F. Pirl, J. S. Temel, A. Billings, C. Dahlin, V. Jackson,, H. G. Prigerson, J. Greer, and T. J. Lynch Depression After Diagnosis of Advanced Non-Small Cell Lung Cancer and Survival: A Pilot Study Psychosomatics, May 1, 2008; 49(3): 218 - 224. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. G. Surtees, N. W.J. Wainwright, R. N. Luben, N. J. Wareham, S. A. Bingham, and K.-T. Khaw Depression and Ischemic Heart Disease Mortality: Evidence From the EPIC-Norfolk United Kingdom Prospective Cohort Study Am J Psychiatry, April 1, 2008; 165(4): 515 - 523. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Ryan, I. Carriere, K. Ritchie, R. Stewart, G. Toulemonde, J.-F. Dartigues, C. Tzourio, and M.-L. Ancelin Late-life depression and mortality: influence of gender and antidepressant use The British Journal of Psychiatry, January 1, 2008; 192(1): 12 - 18. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. J. Salaycik, M. Kelly-Hayes, A. Beiser, A.-H. Nguyen, S. M. Brady, C. S. Kase, and P. A. Wolf Depressive Symptoms and Risk of Stroke: The Framingham Study Stroke, January 1, 2007; 38(1): 16 - 21. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Darensburg, M. E. Andrew, T. A. Hartley, C. M. Burchfiel, D. Fekedulegn, and J. M. Violanti Gender and Age Differences in Posttraumatic Stress Disorder and Depression Among Buffalo Police Officers Traumatology, September 1, 2006; 12(3): 220 - 228. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
