Background Depression has been shown to adversely affect the prognosis of patients with established coronary artery disease, but there is comparatively little evidence to document the role of depression in the initial development of coronary disease.
Methods and Results Study participants were 409 men and 321 women who were residents of Glostrup, Denmark, born in 1914. Physical and psychological examinations in 1964 and 1974 established their baseline risk factor and disease status and their level of depressive symptomatology. Initial myocardial infarction (MI) was observed in 122 participants, and there were 290 deaths during follow-up, which ended in 1991. A 2-SD difference in depression score was associated with relative risks of 1.71 (P=.005) for MI and 1.59 (P<.001) for deaths from all causes. These findings were unchanged after we controlled for risk factors and signs of disease at baseline. There were no sex differences in effect sizes.
Conclusions High levels of depressive symptomatology are associated with increased risks of MI and mortality. The graded relationships between depression scores and risk, long-lasting nature of the effect, and stability of the depression measured across time suggest that this risk factor is best viewed as a continuous variable that represents a chronic psychological characteristic rather than a discrete and episodic psychiatric condition.
While there is evidence that depression worsens the prognosis of patients after MI,1 2 the role of depression in the pathogenesis of coronary artery disease is less well established. This issue has taken on added significance with recent suggestions that depression may be an especially important coronary disease risk factor for women.3 4
Because coronary events often lead to heightened depression,5 cross-sectional designs are not desirable for the study of this question; prospective studies are necessary. The relevance of prospective studies of psychiatric samples6 for the general population is limited. Few longitudinal studies of depression and coronary disease have been conducted in nonclinical samples, and many of them have yielded no evidence of an association7 8 or have been inconclusive.9 However, a recent follow-up of a large national sample observed an elevated risk of ischemic heart disease among those individuals with high scores on measures of depressed affect and hopelessness.10
The present study addresses this issue by evaluating the ability of a measure of depressive symptoms from the MMPI11 to predict AMI over a 27-year period in a community sample of aging Danish men and women. Total mortality also was considered an outcome because some evidence suggests that depression is associated with increases in all-cause mortality.9 12 Baseline physical examinations provided information to be used as control for risk factors and initial disease status.
Glostrup is a municipality west of Copenhagen, Denmark. In 1964, all residents of Glostrup who were born in 1914 (n=975) were invited to participate in an epidemiological study of cardiovascular disease that would involve extensive physical examinations and psychological testing. At that time, this target population was representative of the Danish 1914 birth cohort in terms of sex, education, and occupational indicators.13
The 1964 examinations were performed on 436 men and 366 women. The physical examination involved a full day of testing at Glostrup Hospital. A detailed description of the procedures is available elsewhere.14 The psychological examination, including the MMPI, was conducted on the second day of testing. Some participants did not return for the psychological examination, so 1964 MMPI measures were gathered from 384 men and 289 women.
In 1974, the 890 surviving members of the target population were invited to participate in a second round of examinations. Physical examinations were administered to 666 participants, including 39 who were not included in the 1964 examinations. MMPI data were obtained from 570 participants in 1974, including 57 who did not take part in the 1964 psychological examination. For these participants, 1974 was considered to be the baseline, and exposure was computed from that point. Because this group was older at the time of their entry into the study, age at baseline was covaried in the analyses.
The sample of the present study is made up of the 409 men and 321 women for whom psychological measures were available from at least one examination. Comparisons were performed between the members of this sample and the 111 people who did not take the MMPI in either examination but participated in the other parts of the study. Those who did not take the MMPI were more likely to be female and to be among the 39 people with physical data from 1974 who did not participate in the 1964 examination. The two groups differed on only one other baseline variable considered in the study: those who did not take the MMPI had poorer pulmonary function at baseline. Cox proportional hazards models showed that those with no MMPI data did not differ from those with MMPI data in the incidences of AMI or death during the follow-up period after we accounted for the effects of age at baseline, sex, and pulmonary function.
Measurement of Depressive Symptoms
The most commonly used measure of depression on the MMPI is the D scale, which has its origin in work with psychiatric samples. Items on the D scale are heterogeneous, and many of them are not face-valid reflections of depression.15 However, the scale has been divided on the basis of face validity into obvious and subtle subscales.16 The 40-item OBD subscale is a more straightforward measure of depressive symptoms as experienced outside the psychiatric context. It was therefore chosen as the measure for the present study to increase the relevance of the findings for nonpsychiatric populations. The 10-year test-retest correlation for the OBD scale was .67 in those 513 individuals who took the test at both the 1964 and 1974 examinations. Analyses with the complete D scale also were performed but are not reported because they led to the same conclusions as the OBD analyses.
As in most depression scales, some of the items on OBD pertain to somatic complaints (eg, “I feel weak all over much of the time”). Studies of depression and physical health must consider the possibility that such complaints might have an organic basis such as subclinical coronary disease rather than psychological origins. If so, relations between depression scores and disease outcomes might be spuriously inflated. Therefore, a second depression score was calculated with omission of the nine somatic items (MMPI items 9, 18, 23, 36, 51, 153, 154, 189, and 285) on OBD, and analyses were repeated to see whether the results with the full scale were due solely to the reports of somatic symptoms.
Baseline Risk Indicators
Blood pressure was measured with an arm cuff, Korotkoff phases I and V, while the participant was supine after 10 minutes of rest. Only the systolic reading was used in the analyses.
Blood samples were drawn at the beginning of the examination (8:15 am) after a 13-hour fast that included abstinence from smoking. These samples yielded measures of total serum cholesterol, triglycerides, and insulin.
Smoking was measured with a five-level variable: nonsmoker, ex-smoker, and smoker of 1 to 14, 15 to 24, or 25 g/d or more of tobacco. This index was coded into four dichotomous variables for the analyses.
There were two physical activity measures, physical activity at work and amount of activity in leisure. Both were four-level ordinal measures (sedentary, moderately active, active, and heavy work or competitive sport) that were coded into three dichotomous variables for analyses.
Body mass index was calculated as weight in kilograms divided by height in meters squared.
Signs of Disease
The possible existence of myocardial ischemia at baseline was assessed in two ways, the Rose questionnaire17 to test for symptoms of angina pectoris and ECG. ECGs were taken after 10 minutes of rest and defined as positive according to the Whitehall criteria as Minnesota codes 1.1 to 2.18 A dichotomous index combining these two measures was created that was positive if either of the two assessment methods suggested the presence of ischemia.
A second important indicator of disease in this age group is pulmonary function. It was evaluated with a Godart bell respirator and a Wright peak flowmeter. Forced expiratory volume (liters) in the first second (Fev1) was taken as the indicator. Participants had one practice and two test trials, with the highest value used in the analysis.
Follow-up of the participants has been aided by Danish national registries for deaths, hospitalizations, and changes of address. Additional data collection in 1984 and 1989 (not discussed in this article) also helped ensure the completeness of the follow-up. Except for four participants who emigrated, follow-up is complete up to the end of this study, December 1991.
The incidence of fatal AMI was determined from death certificates and supporting evidence from hospital records or autopsy. The occurrences of nonfatal AMI were detected from hospital records, with the diagnosis confirmed on two of three criteria: typical pains, characteristic ECG changes, and changes in enzyme levels.
AMI occurred in 91 men and 31 women during the course of follow-up. Five men and one woman were excluded from all analyses of this end point because they suffered nonfatal AMI before the 1964 examination, and 3 participants were excluded because they were lost to follow-up for this outcome.
At the end of 1991, 196 men and 94 women, 40% of those with MMPI data, were deceased. Causes of death were determined from death certificates, autopsies, and hospital records. Death certificates from the National Board of Health were coded by the Danish Institute for Clinical Epidemiology. Ischemic heart disease and MI (ICD codes 410 through 414) accounted for 88 deaths. Other cardiovascular diseases (ICD codes 390 through 409 and 415 through 458) were cited as the cause of 38 deaths. Malignant neoplasms (ICD codes 140 through 239) were implicated as the primary cause of death for 90 people. Only 10 of the deaths appear to have been accidents or suicides. Cause of death has yet to be determined for 22 people.
The associations of OBD scores to each outcome were evaluated with Cox proportional hazard survival models. OBD scores were treated as continuous variables. Effect sizes are reported as RRs that were calculated by exponentiating the β from the model after it was multiplied by the number of points equivalent to 2 SDs of the scale (10.78 for the full OBD scale). Thus, the RR can be interpreted as the risk of a person 1 SD above the mean compared with the risk of a person 1 SD below the mean. As stated above, the analyses also were performed with the full D scale of the MMPI, but results did not differ from those reported here.
Previous multivariate analyses based on the total Glostrup sample have established those risk indicators that are independent predictors of the primary end points, AMI and mortality.19 Risk factor controls in the present study were based on those analyses. Only those baseline risk factors that were significant in final versions of those models were included as covariates. By limiting the covariates to those that are independent predictors, problems of multicollinearity are minimized.
Four models were fitted to assess the ability of OBD scores to predict the AMI outcome. These analyses omitted the data of the 6 individuals who had an AMI before 1964 and 3 individuals lost to follow-up. The first model (n=721) contained controls for age and sex. The second model contained additional controls for risk factors shown to be independently related to the outcome19 : age, sex, systolic blood pressure, triglycerides, smoking, sedentary work, and sedentary leisure. There were 675 individuals in the second model owing to missing values on some of the covariables. The final two models were fitted to reduce the possibility that the prospective results were contaminated by the presence of prevalent cases at baseline. The third model (n=679) repeated the first model but eliminated the data of those participants who had signs of possible myocardial ischemia at baseline as evidenced by ECG or questionnaire. The fourth model was based on the same reduced sample but used an abbreviated version of the OBD scale that did not contain the nine items reflective of somatic complaints. Potential sex differences in effect sizes were evaluated by tests for interactions between OBD scores and sex. Potential deviations from linearity were evaluated with quadratic and cubic terms.
A similar set of models was performed for the mortality outcome. The first (n=730) controlled for age and sex only. The second (n=671) added controls for those risk factors found to be independently predictive of mortality (systolic blood pressure, triglycerides, and smoking) and pulmonary function, an important disease indicator. The third (n=682) eliminated the data of those individuals with possible ischemia or prior AMI at baseline. The fourth model repeated the same analysis with the abbreviated OBD scale that did not contain somatic items. Finally, separate models were fitted to predict deaths from ischemic heart disease and deaths from all other causes.
Table 1⇓ characterizes the sample on the OBD scale and the background variables in the study. The levels of depressive symptoms are somewhat high according to US norms but were well within the normal range.15 The observed sex difference in OBD scores (P<.001) is in keeping with the common observation of higher prevalence of depression among women.20
Table 2⇓ shows the partial correlations of OBD scores with baseline risk variables by sex, controlling for age at baseline. Among women, depressive symptoms were positively associated with amount smoked and negatively associated with body mass index and systolic blood pressure. These relationships were absent in men, and the sex-by-OBD interactions were significant (P≤.02) for all three variables.
Cross-sectional analyses also evaluated the association of OBD scores to signs of disease at baseline. Those participants who reported more depressive symptoms were more likely to be among those to show evidence of possible ischemia on the ECG or the Rose questionnaire than those who described fewer symptoms. A logistic model controlling for sex showed that the odds of a person having signs of ischemia were 1.9 times greater for a person 1 SD above the OBD mean than for a person 1 SD below the mean (χ2 1 df=4.3; P<.05). Those participants with more symptoms of depression also had poorer pulmonary function after we controlled for other correlates of Fev1: sex, age, systolic blood pressure, and smoking. The partial correlation between OBD scores and Fev1 was −.09 (P<.02).
Depressive Symptoms and AMI Incidence
Higher OBD scores were associated with increased risk of AMI in all models (see Table 3⇓). Controls for risk factors and elimination of those with possible ischemia had virtually no impact on the size of the OBD effect. Removal of the items dealing with somatic complaints reduced the effect size somewhat, but it remained significant. Tests for an OBD-by-sex interaction were not significant, indicating the absence of sex differences. The Figure⇓ illustrates a gradation of risk across the range of OBD scores in both sexes.
Depressive Symptoms and Total Mortality
Reports of depressive symptoms were positively associated with the risk of mortality from all causes (see Table 4⇑). The effect was significant in all models. As in the AMI analyses, there were no sex differences in effect size, and tests for departures from linearity were not significant.
Further analyses divided deaths into two categories: those attributed to ischemic heart disease and those due to all other causes. Models fitted on the full sample showed that OBD scores predicted both ischemic heart disease mortality (RR=1.62; χ2 1 df=4.7; P=.03) and mortality from all other causes (RR=1.57; χ2 1 df=9.6; P=.002). Therefore, it does not appear that the deleterious effects of depression on health are limited to ischemic heart disease.
High scores on a measure of depressive symptoms were associated with increased risks for AMI and early mortality during a 27-year follow-up. These findings remained essentially unchanged after we controlled for traditional risk factors and signs of disease at baseline. The impact of depressed affect on health did not differ between men and women. These results are among the first to demonstrate that depression can play a role in the development of initial coronary events.
Some research has shown that depression predicts total mortality12 and the prognosis of patients after MI,1 2 but evidence concerning the role of depression in the pathogenesis of coronary disease has been less clear. Several theoretical and methodological issues have contributed to the difficulty in establishing this relationship. One problem that has complicated research in this area is the influence of personality on the perception of symptoms. Because those individuals with depression and related psychological characteristics are more likely to report chest pain in the absence of coronary artery disease, they are more likely to be falsely diagnosed with angina pectoris.21 22 Therefore, it is important for studies to rely on “hard” outcomes such as documented AMI.
Another factor that has complicated research in this area is the similarity between the symptoms of depression and the symptoms of coronary disease. Because both illnesses result in tiredness, feelings of weakness, etc, it is possible that manifestations of subclinical heart disease could be mistaken for depressive symptomatology. Therefore, the apparent relation between depressive symptoms and coronary events could be spuriously inflated. There are several ways to address this problem. One is to exclude from analysis all coronary events that occur soon after baseline measurement on the assumption that subclinical coronary disease will become evident within the first few years. This strategy was adopted by Anda et al.10 In the present study, no coronary events occurred in the first year after baseline, and only 4 of the 122 infarctions occurred in the first 5 years, supporting the independence of the OBD measure from subclinical disease. A second solution is to eliminate somatic symptoms from consideration when depression is assessed. The present study found that the relationship was not dependent on the presence of somatic symptoms in the criteria for depression. A third approach is to include only those individuals who were symptom-free at baseline in the analyses. In the present study, the OBD effect was unchanged when those participants with symptomatic or ECG evidence of ischemia were eliminated from consideration.
The timing of events during the course of follow-up is also important because it helps distinguish the influence of depression from that of vital exhaustion, a related phenomenon that precedes MI.23 The risk associated with vital exhaustion is highest in the first few months after assessment and decreases over time.24 Because the short-term risk in the present study was low, it appears that OBD scores reflect depressive symptoms rather than vital exhaustion.
It is important to note that a graded increase of risk was observed across the distribution OBD scores. Anda et al10 observed a similar relationship. These findings, coupled with the fact that both studies were performed on representative samples of the normal population, suggest that it is appropriate to view this risk factor as a continuum of depressive symptoms rather than a psychiatric disorder of major depression. Signs of depression can show stability over time.25 This can be seen in the high correlation (.67) between OBD scores in the 1964 and 1974 Glostrup examinations. Such stability probably is necessary for depression to exert the type of chronic influence necessary for the development of coronary artery disease.
One way that depression affect might act to foster coronary disease and early mortality is through health habits such as smoking. Smoking was positively correlated with depressed affect among the women in the Glostrup sample, and Anda et al26 have observed positive associations between depressed affect and smoking in both men and women in a large national sample of the United States. However, OBD scores continued to predict both coronary events and total mortality after we controlled for a variety of risk factors (including smoking) and signs of disease at baseline. Some other mechanism, eg, heightened sympathetic reactivity, must be responsible for the association.
It has been suggested that depression might play an especially important role in the health of women.3 4 We found that OBD scores were equally predictive in men and women. However, the higher prevalence of depression in women20 could still mean that depressed affect is more significant for them.
Evidence from this and other studies is emerging to demonstrate the adverse health consequences of chronic depression. It is time to search for ways to ameliorate its effects.
Selected Abbreviations and Acronyms
|AMI||=||acute myocardial infarction|
|ICD||=||International Classification of Diseases|
|MMPI||=||Minnesota Multiphasic Personality Inventory|
This research was supported by fellowship F06 TW01952 from the Fogarty International Center and by grants AG-09276 5P60, AG-11268, and PO2 AG-12058 from NIA and HL-36587 from NHLBI.
Reprint requests to John C. Barefoot, PhD, Box 2969, Duke University Medical Center, Durham, NC 27710. E-mail firstname.lastname@example.org.
- Received July 5, 1995.
- Revision received September 29, 1995.
- Accepted October 6, 1995.
- Copyright © 1996 by American Heart Association
Chesney MA. Social isolation, depression, and heart disease: research on women broadens the agenda. Psychosom Med. 1993;55:434-435.
Dimsdale JE. Coronary heart disease in women: personality and stress-induced biological responses. Ann Behav Med. 1993;15:119-123.
Brozek J, Keys A, Blackburn H. Personality differences between potential coronary and noncoronary subjects. Ann N Y Acad Sci. 1966;10:57-64.
Aromaa A, Raitasalo R, Reunanen A, Impivaara O, Heliovaara M, Knekt P, Lehtinen V, Joukamaa M, Maatela J. Depression and cardiovascular disease. Acta Psychiatr Scand. 1994;377(suppl):77-82.
Hathaway SR, McKinley JC. Manual for the Minnesota Multiphasic Personality Inventory. New York, NY; Psychological Corp; 1943.
Hagerup LM. Coronary heart disease, risk factors in men and women, from the Population Study in Glostrup, Denmark. Acta Med Scand. 1974;(suppl 557):I-116.
Greene RL. The MMPI-2/MMPI: An Interpretive Manual. Boston, Mass: Allyn and Bacon; 1991.
Wiener DN. Subtle and obvious keys for the MMPI. J Consult Clin Psychol. 1948;12:164-170.
Rose G, Blackburn H. Cardiovascular Survey Methods. Geneva, Switzerland: World Health Organization; 1968. WHO monograph series No. 56.
Thomsen KK, Larsen S, Schroll M. Do CVD risk factors still count in the elderly? A follow-up until 1990 of the 1914 population in Glostrup surveyed at the age of 50, 60 and 70. Am J Ger Cardiol. In press.
McGrath E, Keita GP, Strickland BR, Russo NF. Women and Depression: Risk Factors and Treatment Issues. Washington, DC: American Psychological Association; 1990.
Shekelle RB, Vernon SW, Ostfeld AM. Personality and coronary heart disease. Psychosom Med. 1991;53:176-184.
Appels A. Mental precursors of myocardial infarction. Br J Psychiatry. 1990;156:465-471.
Appels A, Otten F. Exhaustion as precursor of cardiac death. Br J Clin Psychol. 1992;31:351-356.
Costa PT Jr, McCrae RR. Depression as an enduring disposition. In: Schnieder LS, Reynolds CF, Lebowitz BD, Friedhoff AJ, eds. Diagnosis and Treatment of Depression in Late Life. Washington, DC: American Psychiatric Press; 1993.