For centuries poets and folklore have asserted that there is a relation between the mind and body in general and human moods and the heart in particular. Almost 400 years ago Shakespeare wrote, "My life.sinks down to death, oppress'd with melancholy" (Sonnet 45). However, only in the last few years has this conviction been scientifically tested. Nevertheless, it is now abundantly clear that depression is associated with ischemic heart disease.
The first scientific effort to test this relationship was a study by Malzberg published in The American Journal of Psychiatry in 1937 (1). He compared the mortality rate of patients with involutional depression in the New York State civil hospital system and the rate in the general population of the state. As hypothesized, the mortality rate for the patients was markedly elevated, as it was for patients with both cardiovascular and infectious diseases. However, the study conspicuously confounded the influence of involutional depression with that of chronic institutionalization. Although the results were dramatic, they were not convincing, and the issues involved were not further tested for almost 40 years.
To some extent this hiatus was a consequence of World War II, but beyond the war, psychiatry was focused on psychoanalysis and personality and, with respect to heart disease, type A behavior. In the late 1970s, several investigators began to reexamine the impact of depressive diagnoses. One of the earliest and most impressive of these studies was that of Weeke (2), who took advantage of the Danish National Registries and identified all individuals with a diagnosis of either major depression or manic-depressive disease between 1974 and 1978. All causes of natural death were examined, and there was a 50% increase in deaths from cardiovascular disease among depressed patients compared with the general Danish population. This study followed 6,000 patients for an average of 5 years, and unlike the Malzberg study group, the vast majority of the patients were not institutionalized. The Weeke data raised the concern that in the 40 years since Malzberg's study, drug treatments (tricyclic antidepressants and lithium) had become common, and the association between death and depression could be confounded by these treatments. However, although the literature is limited, it suggests that adequate treatment—whether with tricyclics, lithium, or ECT—reduces rather than increases mortality (3–6).
Moreover, Weeke herself addressed this question in a second study (7) in which she and her associates compared the relative risk of mortality from cardiovascular disease in depressed populations before and after the introduction of antidepressant drug treatments. The relative risk of mortality was elevated in both periods, but the risk was actually lower during the era of drug treatment. This, again, tends to support the suggestion that appropriate treatment might reduce the cardiovascular risk associated with major depression.
Besides the two Weeke studies (2, 7), Tsuang et al. (8), Norton and Whalley (9), and Rabins et al. (10), using methods very similar to that of Weeke and colleagues, compared depressed patients with the general population. All three of these 1980s studies found an excess of deaths from cardiovascular disease among depressed patients. In the 1990s Sharma and Markar (11) reported an excess of deaths from cardiovascular disease in 472 manic-depressive patients over a 17-year period, while Vestergaard and Aagaard (12), in a smaller group of subjects with bipolar depression, found a very similar result. Earlier, Black et al. (13) had examined 5,412 psychiatric patients admitted to the University of Iowa Psychiatric Hospital. They found an elevated mortality risk in the female but not the male depressed patients. They did observe an elevated relative risk of mortality from cardiovascular disease for the whole sample but did not specifically examine the mortality rate from cardiovascular disease in the depressed group. In general, the findings of Black and associates, although statistically significant and in the same direction as those of the other seven studies, were not as dramatic.
The one study with negative results that compared patients with the general population used the Central Psychiatric Case Register of Israel. The observations of Zilber and colleagues (14) were based on a 5-year follow-up of depressed patients that involved 7,868 person-years. Total mortality was significantly elevated, but unlike others, Zilber et al. found that the bulk of this excess was due to infection rather than cardiovascular disease. They pointed out that a major difference between their data and that of others is that they studied patients admitted to psychiatric hospitals rather than to psychiatric services of general medical hospitals. This process would have limited their sample to very severe and refractory cases of affective disorder, but it is not clear why this should reduce the association with cardiovascular disease.
Examining community epidemiological samples rather than study groups based on patients seeking treatment circumvents both the influence of institutionalization and any potential effect of treatment. In this situation the vast majority of cases of depression that are identified involve persons who have never had any antidepressant treatment. Using community-based samples also significantly shifts the focus from more severe to less severe cases. It would be easy to think that the phenomena responsible for the increased mortality seen in treatment samples would not be strong enough to manifest an effect in milder cases. Nevertheless, two studies based on community samples demonstrated a significant association between depression and mortality. Murphy et al. (15) used the data from 1,000 individuals who were followed for 16 years in Sterling County, Canada, and Bruce et al. (16) examined the data from 3,500 individuals followed for 9 years in New Haven, Conn. Murphy et al. investigated cause of death, and depression was explicitly associated with mortality from cardiovascular disease. The investigators in the New Haven study did not give information about cause beyond saying that the excess deaths were due to natural causes.
For more than 20 years, a major focus of our research group has been the cardiovascular effects of the various antidepressant drugs, and we have followed quite closely the possibility that depression itself could influence cardiovascular disease. By the late 1980s, it was the opinion of the first author (A.H.G.) that the evidence linking depression and mortality due to cardiac disease had become quite convincing. During that same period, another aspect of our work involved nicotine and resulted in the observation that the mere history of major depression, even in the absence of current illness, increased the likelihood that a person would smoke and, if that person attempted to quit, increased the chances that he or she would fail (17). When we began to study nicotine, it did not occur to us that smoking studies would be connected to our interest in depression and cardiovascular disease. However, even before the first replication was published, it was apparent that the association between depression and smoking represented a potentially serious confounding factor in the relation between depression and mortality. The relationship could simply be that depressed individuals were more likely to smoke and that smoking caused cardiovascular disease.
One of the earliest replications of our original observation came from the Centers for Disease Control (CDC). Using the National Health and Nutrition Epidemiological Study data, Anda et al. (18) found almost exactly what we had found using data from St. Louis (19). While discussing the differences in our data sets, Anda et al. described another CDC data set called the Medical Illness Follow-Up Study. We quickly realized that this data set contained not only information about depression and smoking but about depression and death, while controlling for smoking (20). Information was available on 2,832 individuals over the age of 45 years who were found to be free of any disease after a physical examination and laboratory tests and were then followed for 12½ years. To preclude the possibility that subjects had some awareness that they were ill which was not revealed by their history, laboratory tests, or physical examination, patients who developed their illness in the first 2½ years of the study were excluded. Nevertheless, after controlling for smoking and other known risk factors for cardiovascular disease (gender, weight, activity, blood pressure, and cholesterol), we found that the apparently healthy individuals who had elevated depression ratings were more likely both to develop and to die of ischemic heart disease. Even after separation of the sample into those who smoked and those who did not, the influence of depression persisted. In addition to the control for smoking, this is the first study to link depression with the dual risk of both developing and dying of ischemic heart disease.
In the last few years, six more epidemiological studies (21–26) have taken a similar approach to different data sets. All of these studies controlled for smoking, and five of the six studies found a relation between depression and cardiovascular disease. Ford et al. (26) reported on 1,198 male former Johns Hopkins University medical students followed for a median of 35 years. The men with depression had a higher relative risk of myocardial infarction (relative risk=1.68, 95% confidence interval=1.03–2.74) than the men free of depression. On average, the first report of depression preceded the first report of cardiovascular disease by 10 years. Aromaa et al. (21) reported on the Mini-Finland Health Survey follow-up of individuals 40–64 years of age. The study was based on 5,355 persons who had both medical and psychiatric evaluations at baseline and were followed an average of 6.6 years. The risks of developing and of dying of ischemic heart disease were both significantly elevated among depressed individuals after control for age, education, and traditional ischemic heart disease risk factors including smoking.
The Kuoppio Ischemic Heart Disease Study (22) followed a separate sample of 2,428 Finnish men over a 6-year period. In this study, very much like the findings of Anda et al. in the United States, men without any prior history of angina or myocardial infarction who had higher levels of depression were more likely to experience a first infarction, even after control for a wide array of biological, behavioral, and social risk factors. There was again an elevated risk of mortality from cardiovascular disease among the men with higher depression scores.
Another recently published long-term study that also controlled for smoking is that of the Glostrup cohort (23). This was a study of 730 persons born in Glostrup, Denmark, in 1915. They had both physical and psychological examinations in 1964 and again in 1974 and were followed an average of 27 years. After controlling for both smoking and physical health, the investigators found, as with the CDC and Finnish data, that individuals with elevated depression scores were 65% more likely to develop ischemic heart disease. Elevated depression scores were also associated with a significant increase in all natural causes of death; however, the investigators did not address the question of whether this increased mortality rate was specifically due to cardiovascular disease.
The most recently published data come from a 13-year follow-up of the Baltimore cohort of the Epidemiologic Catchment Area study (24). Among the 1,551 respondents who were apparently free of medical illness at baseline, a diagnosis of major depression increased the risk of myocardial infarction more than fourfold after control for both medical risk factors and other psychiatric diagnoses. This study is interesting because it used a DSM-III diagnosis of depression, it is the only study to control for other psychiatric diagnoses, and it included the category of dysphoria in the same population. Individuals who were dysphoric but never met lifetime criteria for major depression were intermediate in their risk for myocardial infarction compared with those with DSM-III major depression and those who were totally free of depression.
Only one study that started with healthy individuals and controlled for all of the usual medical risk factors failed to find a relation between depression and new ischemic heart disease and/or mortality due to cardiovascular disease. That study was the 15-year follow-up of 2,573 members of the Northwest Region of Kaiser Permanente (25). Only 1,399 of these persons were over the age of 45 years. Even in this study the authors referred to a "possible" relation between mortality and depression among men.
In 1988 Carney et al. (27) used a study strategy that started with patients who already had cardiovascular disease. They conducted structured psychiatric examinations of consecutive patients undergoing coronary angiography. Of the 52 patients whose studies confirmed the presence of coronary artery disease, just less than 20% met criteria for major depression. This depressed group turned out to be about 2½ times more likely to develop a serious adverse cardiac complication over the next 12 months. Although the study group was too small to control adequately for other risk factors, this was corrected in subsequent studies using the same strategy.
Ahern and colleagues (28) obtained baseline ratings of anxiety, anger, and depression in 350 patients with ventricular arrhythmia following a heart attack. The postinfarction patients who survived the first year had lower baseline depression scores than the nonsurvivors after the usual medical and social risk factors were controlled. Neither anger nor anxiety scores were associated with survival. Several other investigators studying post-myocardial-infarction patients have also found a higher mortality associated with depression (29–32). The most convincing of these studies was the one by Frasure-Smith et al. in 1993 (32).
Frasure-Smith and colleagues obtained structured psychiatric examinations of 222 patients 5–15 days after they suffered myocardial infarction. The patients were contacted again 6, 12, and 18 months after discharge from the hospital. Sixteen percent of these patients showed evidence of major depression while hospitalized for their index myocardial infarction. This is consistent with the other surveys that regularly show 15%–20% of postinfarction patients developing major depression (33, 34). At 6 months approximately 17% of the depressed patients in the study by Frasure-Smith et al. (32) had died, compared with 3% of the nondepressed patients. With control for other independent predictors of risk, the relative hazard for the depressed patients was almost 3½ times greater than for the nondepressed patients. The strongest medical predictor of mortality after myocardial infarction is generally heart failure, and it is associated with just about the same 3½-fold increase in mortality.
Although concern has regularly been expressed about the reliability of a diagnosis of depression made in a medically ill patient, and especially a post-myocardial-infarction patient, only 25% of the patients Frasure-Smith et al. identified as depressed were free of major depression at both the 6- and 12-month follow-up interviews (35). This compares with 81% of those who did not originally receive a diagnosis of depression who remained free of depression. Similarly, Hance et al. (34) found that one-half of patients diagnosed as having major depression in the immediate postinfarction period either remained depressed or relapsed within 12 months.
In addition to studying patients with the diagnosis of major depression, Frasure-Smith et al. also measured symptoms of depression with the Beck Depression Inventory (36). Six months after myocardial infarction, the patients who had elevated Beck inventory scores but were free of major depression had mortality rates much like those of the patients who were free of any depression. However, by 12 months, the patients with Beck inventory scores above 10 were dying at a rate significantly higher than that of the patients without depression, and by 18 months their mortality rate was not much less than that of the patients with major depression (37). It is not clear whether this increased mortality was really associated with lesser degrees of depression or whether these high Beck inventory scores were merely identifying a group at high risk for major depression. In the Hance et al. study (34), 42% of patients with minor depression developed major depression over the subsequent 12 months.
Recently, Barefoot and his co-workers at Duke University (38) published a 17-year follow-up of patients admitted to the cardiology service with a diagnosis of coronary artery disease. This study combined the strategies of using a high-risk population and using a long-term follow-up. Once again, depression was associated with diminished survival among patients with preexisting cardiovascular disease.
Taken together, these studies constitute a convincing body of evidence linking depression, cardiovascular disease, and mortality. Moreover, this relationship seems to have at least two components. All six of the studies that specifically examined patients with preexisting cardiovascular disease found higher mortality rates over the next 6–12 months among the depressed patients than among the nondepressed patients. These studies were well controlled for both medical and social risk factors, and the evidence is particularly strong in the postinfarction patients. The second component involves the relative risk of experiencing a first heart attack among persons initially free of heart disease. Here again, the evidence is strong and generally well controlled. Six of the seven studies that identified healthy individuals at baseline, controlled for cardiovascular risk factors including smoking, and recorded new episodes of cardiovascular morbidity found a significantly elevated risk among the persons who had higher depression scores at baseline (20–24, 26). That increased risk of morbidity is generally from 1.5 to 2.0 times and is about one-half the magnitude of the increase in mortality that depression confers in post-myocardial-infarction patients.
Although there is no question that depression is associated with both developing cardiovascular disease and death, there are a number of issues that remain to be clarified. One is whether the association with cardiovascular disease and death is unique to depression or whether it exists with other negative mood states. The next best body of information concerns anxiety, but it is neither as extensive nor as sophisticated as the literature on depression. It is beyond the scope of this article to examine these data in detail, but it is our opinion that they suggest that phobic anxiety or panic states are associated with death, most clearly sudden death, over the long term (39, 40). However, unlike depression, higher levels of general anxiety in initially healthy individuals do not seem to be associated with an increased rate of developing or dying of cardiovascular disease (39–42). If there is a short-term effect of anxiety on mortality from cardiovascular disease, it is not as robust or as easy to demonstrate as that of depression (28, 43, 44).
Another unresolved issue is precisely what it is that carries the risk. In phenomenological terms, the studies that have been mentioned up to this point have identified populations at increased risk by using a diagnosis of major depression, various scales that measure symptoms of depression, or the hopelessness item of the General Well-Being Schedule (45). The obvious question is whether the increased risk is associated specifically with the diagnosis of major depression, whether it varies with the severity or chronicity of a depressive state, or whether it can be associated with symptoms of depression in the absence of a diagnosis.
There are only very limited data available to address these distinctions. Lesperance et al. (35) found that in the short term following a heart attack, the risk of death was strongly associated with recurrent major depression, but that a single episode of major depression occurring for the first time after a myocardial infarction was not associated with any increased risk of mortality. Because of the manner in which Frasure-Smith, Lesperance, and colleagues collected their follow-up data, it is not possible to say whether the excess deaths occurred among the patients who were persistently depressed after infarction, whether those who died had a return of their depression before their deaths, or whether the mere occurrence of depression after infarction in those with a history of prior depression marked a group that were at higher risk of death regardless of whether they continued to be depressed. In the long-term studies of Anda et al. (20), Everson et al. (22), and Barefoot and Schroll (23), the higher the symptom scores at baseline, the higher the risk of both developing ischemic heart disease and dying of it. However, it is again not clear whether this increased risk was the result of these individuals merely having more severe symptoms of depression. Alternatively, the higher symptom scores may have identified individuals who were more likely to develop major depression, and the risk was concentrated in those persons who developed major depressive episodes. Everson et al. asserted that hopelessness is a state that can be separated from depression and that this state confers a risk independent of the risk of depression. In general, these remain open issues.
In addition to the studies that have been described to this point, a series of studies from the Netherlands have examined a condition described as vital exhaustion. This state is characterized by lack of energy, increased irritability, and feelings of demoralization (46). Vital exhaustion sounds similar to depression and seems to demonstrate a similar association with cardiovascular disease. In long-term prospective studies it was associated with an increased relative risk of initial myocardial infarction (47), and it was associated with an increase in adverse cardiac events in the year and a half following angioplasty (48). It seems very likely that recurrent major depression and vital exhaustion tap the same underlying process, but the data available do not allow us to test that hypothesis.
Ultimately, it will be established whether these descriptive entities are measuring the same or different processes by identifying the mechanism or mechanisms that lie behind these associations. In the study by Frasure-Smith et al. (37), the excess mortality among the post-myocardial-infarction patients with depression was due almost exclusively to sudden death. Because sudden death is almost always the result of ventricular arrhythmia, Frasure-Smith et al. investigated whether there was an interaction between depression and postinfarction arrhythmia. There was, in fact, a striking increase in mortality among the postinfarction patients who had both depression and even mild baseline ventricular arrhythmia (37). There is considerable evidence that fluctuations in autonomic nervous system tone influence the risk of ventricular fibrillation and sudden death (49). Changes in autonomic tone have long been considered an integral part of serious depression, and the direction of those changes is such that one would anticipate an increase in sudden death (50, 51). This could easily explain a good part of the increased mortality associated with depression following infarction. However, it would be an unlikely explanation for the increased rate of new infarctions seen in depressed individuals initially free of any cardiac disease (20–24, 26).
Increasingly, platelet function has been seen as playing a crucial role in coronary disease, leading from plaque to coronary occlusion and infarction. Noting that platelet receptors were studied as markers for depression, Anda et al. (20) hypothesized that platelets could represent a mechanism explaining the association between depression and new ischemic disease. This provided the impetus for Musselman et al. (52) to study platelet function in depression. Laghrissi-Thode and colleagues (53) were interested in the potential of selective serotonin-inhibiting antidepressants to influence platelet function because of the ability of these drugs to deplete platelets of serotonin. For this reason they obtained baseline measures of platelet function. Both groups of investigators observed significant abnormalities among drug-free depressed patients. These abnormalities, although not identical, were all associated with an increased propensity for platelets to aggregate and could easily be seen as a reason for the increased rate of new ischemic events associated with depression. There are also studies of cholesterol and high-density lipoprotein cholesterol which suggest that alterations in lipid metabolism in depressed patients may increase the risk of vascular disease (54).
Before the studies of Musselman and colleagues and Laghrissi-Thode and colleagues, studies of platelet function examined the influence of psychological state (stress, anger, or anxiety) over minutes or hours (55, 56). This distinction between the influence of mood states over very short, intermediate, and long time periods has undoubtedly been a source of some of the confusion that might appear to exist among studies that relate moods and cardiovascular function. Acute mood states are difficult to approach epidemiologically, and the information that does exist tends to be anecdotal. Retrospective studies of individuals who died suddenly suggest that almost one-quarter had experienced substantial emotional distress in the period immediately preceding their deaths (57). The Northridge earthquake in Los Angeles provided an opportunity to collect epidemiological evidence about the influence of acute stress on a large number of individuals (58). There was a significant rise in the number of sudden cardiac deaths on the day of the quake. However, there was then an unusually low number of such deaths in the week following the quake, suggesting that acute emotional stress may precipitate cardiac events in people who are already predisposed.
Although data on humans are difficult to collect, there are excellent animal models for examining the influence of short-term emotional states on the outcome of myocardial infarction. In animals with deliberately occluded coronary arteries, acute states of anger, stress, and anxiety have all been shown to increase dramatically near-term ventricular arrhythmia and mortality (59–61). There is little reason to doubt that these same intense negative emotional states can have adverse consequences for human beings over the short term. However, the long-term consequences of more chronic anger and anxiety are not entirely clear. As indicated earlier, chronic anxiety, if not specifically panic or phobic anxiety, does not seem to be associated with long-term cardiovascular morbidity. If there is any long-term effect that is independent of depression, the effect is very small (28, 42). The picture with chronic states of anger is less clear. The findings of the available studies have been contradictory and difficult to reconcile (42, 62, 63). Some aspects of hostility probably do relate to the progression of coronary heart disease. What those aspects are, and whether they exert an effect independent of depression, remains speculative.
Another concept of the relation between psychological states and cardiovascular disease involves the so-called type A personality. In the late 1960s and early 1970s, Rosenman, Friedman, and colleagues (64, 65), among others, developed a body of evidence linking a competitive, time-urgent, irritable personality with an increased incidence of myocardial infarction. By the late 1970s, investigators began to have difficulty replicating these earlier observations. In 1987 Booth-Kewley and Friedman (42) analyzed some 55 studies that had examined the relation between personality and cardiovascular disease. Using meta-analysis techniques, they found considerable statistical evidence for the concept of type A personality over a 30-year period. However, the strength of the relationship was modest at best, and there was evidence for a decline in the association over time. They speculated on a number of issues that might have served to weaken the relationship. We would wonder about the increasing use of β blockers moderating processes that were primarily mediated by increased sympathetic outflow. It is interesting that Booth-Kewley and Friedman found that the strongest effect size in the 55 studies they analyzed was not for type A personality or anger but for depression. They commented that "depression appears to be reliably associated with cardiovascular heart disease.but has generally been overlooked" (42). Again, the evidence for an effect of anxiety was weak.
Regardless of what made the type A effect weaken, or whether some form of anger has long-term implications, depression is unquestionably associated with cardiovascular disease. It is hard not to think of this association in terms of depression causing heart disease. The presence of depression preceding the onset of ischemic heart disease in individuals initially free of disease, the greater risk of sudden death among post-myocardial-infarction patients with both depression and arrhythmia, and the predisposition to increased platelet aggregation among depressed patients all seem to point to a causal relationship. Yet it is important to remember that what has been demonstrated is an association and not causality. It is conceivable that atherosclerosis could be a cause of both depression and heart disease. Indeed, there is evidence that late-onset depression may be secondary to arteriosclerotic disease in the brain (66, 67). At this point, the source or sources of these associations remain to be determined. Although we suspect that the autonomic nervous system and platelet changes, as well as the health behaviors that are associated with depression, contribute significantly to the relationship with ischemic heart disease, the relationship could easily be a two-way street, where the arteriosclerotic process could also increase the risk of depression.
The other question raised by the relation between depression and mortality is whether successfully treating the depression would reduce the associated mortality. It almost seems obvious that if depression increases the risk of postinfarction mortality, then treating depression should reduce the risk. However, experience has taught that treatments which seem rational can produce very unexpected results. It was clear to cardiologists that ventricular arrhythmia after myocardial infarction predicted death and that antiarrhythmic drugs greatly reduced these arrhythmias. It seemed logical that these antiarrhythmic drugs should reduce this mortality. Nevertheless, when tested, although they eliminated arrhythmias, antiarrhythmic drugs increased mortality (68). Even before we can test whether antidepressants are able to reduce mortality after infarction, we need to test whether these drugs are safe and efficacious in these patients. It is already evident that the tricyclic antidepressants are too risky in this group (69). Tricyclic antidepressant pharmacology strongly resembles that of the recently incriminated antiarrhythmic drugs. The serotonin reuptake inhibitors seem more benign, but the evidence available at this time is scanty at best (70).
Although it remains to be clarified whether antidepressant treatment will alter this post-myocardial-infarction mortality, it is amply clear that depression is strongly associated with more frequent and more malignant cardiovascular disease. In fact, it is likely that depression's effect is not limited to cardiovascular disease but involves all vascular disease including stroke (71, 72). Psychiatry has been increasingly focused on the social and financial costs of depression. Data from the Medical Outcomes Study (73) are frequently cited as evidence that depression is as much a cause of disability as any major medical illness except heart disease. Our review of the evidence suggests that depression is actually a major contributor to the course and severity of heart disease and that the two conditions in later life may well be synergistic. There is recently published evidence that depression increases osteoporosis (74), and our own earlier studies have made obvious the link between depression and nicotine addiction (75). Although it would be a mistake to minimize the psychological cost of chronic depression, it is becoming increasingly clear that depression is an illness with very real and dangerous physical concomitants.
Received Jan. 10, 1997; revision received July 14, 1997; accepted July 16, 1997. From the Department of Psychiatry and the Department of Clinical Psychopharmacology, College of Physicians and Surgeons, Columbia University, New York. Address reprint requests to Dr. Glassman, New York State Psychiatric Institute, 722 West 168th St., New York, NY 10032. Supported in part by the Suzanne C. Murphy Foundation and the Nathaniel Wharton Fund.