Depressive Symptoms and Survival of Patients With Coronary Artery Disease
JOHN C. BAREFOOT, PHD, BEVERLY H. BRUMMETT, PHD, MICHAEL J. HELMS, BS, DANIEL B. MARK, MD,
ILENE C. SIEGLER, PHD, MPH, AND REDFORD B. WILLIAMS, MD
Objective: Multiple studies have shown that high levels of depressive symptoms increase the mortality risk of
patients with established coronary disease. This investigation divided depressive symptoms into groups to assess
their relative effectiveness in predicting survival. Methods: Questionnaires about the presence of depressive
symptoms were administered to 1250 patients with significant coronary disease while they were hospitalized for
diagnostic coronary angiography. Follow-up for mortality due to cardiac disease was conducted annually for up to
19.4 years. Factor analysis was used to divide items on the Zung Self-Rating Depression Scale into four groups:
Well-Being, Negative Affect, Somatic, and Appetite. In addition, responses to a single item regarding feelings of
hopelessness were available for 920 patients. Results: Well-Being and Somatic symptoms significantly predicted
survival (p ? .01). Negative Affect items were also related to survival (p ? .0001) and interacted with age. A 2-SD
difference in the Negative Affect term was associated with a relative risk of 1.29 for patients ?50 years old and 1.70
for younger ones. Only Negative Affect remained significant in a model with the other symptom groups. Hopeless-
ness also predicted survival with a relative risk of 1.5. Both the Hopelessness and Negative Affect items remained
as independent predictors in the same model. All models controlled for severity of disease and treatment. With one
exception (income and Hopelessness), results were essentially unchanged by additional controls for age, gender,
and income. Conclusions: Depressive symptoms differentially predicted survival, with depressive affect and
hopelessness being particularly important. These effects were independent of disease severity and somatic symp-
toms and may be especially important in younger patients. Key words: depressive symptoms, survival, coronary
CAD ? coronary artery disease; MMPI ? Minnesota
Multiphasic Personality Inventory; RR ? relative risk;
SDS ? Zung Self-Rating Depression Scale.
A number of studies have demonstrated that the
presence of depressive symptoms has an adverse im-
pact on the prognosis of patients with established CAD
(1–3). However, depression manifests itself in a variety
of symptoms that do not necessarily occur together.
The present study is an investigation of the importance
of different symptoms for the prediction of survival in
patients with coronary disease. There have been sev-
eral suggestions that some aspects of the syndrome of
depression are more important than others for patients
with CAD. Appels et al. (4, 5) have argued that feelings
of fatigue and demoralization, indicative of the state of
vital exhaustion, are potent precursors of myocardial
infarction. Others have suggested that feelings of hope-
lessness are particularly important, predicting coro-
nary events while controlling for other aspects of de-
components of depressive symptomatology has not
been systematically addressed.
A related issue has to do with the importance of so-
matic symptoms in the relationship between depression
and survival. Depression and heart disease share a num-
ber of physical symptoms, such as loss of energy, inabil-
ity to perform normal activities, and sleep disturbances.
Therefore, it is possible that depression scores predict
ifestations of underlying coronary disease. The impor-
tance of separating physical symptoms from other as-
pects of depression when studying chronically ill
patients has been recognized (7). Nevertheless, a number
of symptom checklists used in the literature on depres-
sion in patients with coronary disease contain somati-
cally related items. The prognostic significance of those
items should be examined separately.
This study is based on further analyses of data previ-
ously reported from a study in which it was shown that
scores on the SDS (8) predicted survival over an ex-
tended follow-up period (2). For the new analyses, the
SDS was divided into groups of symptoms with the aid
to predict survival while controlling for severity of dis-
ease and treatment was evaluated. A measure of hope-
lessness, which was available for a subsample of pa-
tients, was also examined in relation to survival.
Patients entering Duke University Medical Center for diagnostic
coronary angiography between October 1974 and February 1980
From the Behavioral Medicine Research Center and the Depart-
ment of Psychiatry and Behavioral Sciences (J.C.B., B.H.B., M.J.H.,
I.C.S., R.B.W.) and the Department of Medicine (D.B.M.), Duke Uni-
versity Medical Center, Durham, North Carolina.
Address reprint requests to: John C. Barefoot, PhD, Duke Univer-
sity Medical Center, Box 2969, Durham, NC 27710. Email:
Received for publication November 30, 1999; revision received
May 10, 2000.
790Psychosomatic Medicine 62:790–795 (2000)
Copyright © 2000 by the American Psychosomatic Society
were administered a battery of psychosocial measures, including the
SDS and the MMPI (9). The measures were administered after the
angiographic procedure but before the results were known to pa-
tients. Patients were recruited if they were admitted for their first
cardiac catheterization, were medically stable, and were able to read
at least at a sixth grade level. It was not necessary to exclude patients
on the basis of comorbidity because, at the time, cardiac catheter-
ization was generally not performed on those with major comorbid-
ity (except for cardiac risk factors such as diabetes mellitus or
systemic hypertension). Only 2% of the patients approached who
met the above criteria declined to participate. The 1568 patients who
were found to have significant CAD (?75% narrowing in diameter of
at least one coronary artery) were enrolled. An additional 68 patients
were excluded because data on key medical variables were missing.
The SDS was not completed by 226 patients, leaving a sample of
1031 men and 219 women, who are the focus of the present inves-
tigation. The MMPI, the source of the hopelessness measure, is long
and burdensome with 566 items. Not surprisingly, only 757 men and
163 women completed the hopelessness measure, which was near
the end of the questionnaire. Patient characteristics are described in
Those who did not complete the SDS or the hopelessness mea-
sure were compared with those who completed both measures.
There were no differences between groups in gender or the number
of significantly narrowed coronary arteries, but those who did not
complete the measures did have more severe illness, as indicated by
the hazard score (p ? .002), a summary prognostic index (see below).
Those with missing depression scores were also older (53 vs. 51
years, p ? .002) and had lower incomes (p ? .001). However,
survival among these patients was not significantly poorer after
controlling for illness severity and treatment status.
Correlates of missing data patterns in this sample have been
studied extensively (10). Those who completed relatively few items
on the questionnaires were characterized by low levels of education
and high levels of depressive symptoms. Therefore, those excluded
from analyses because of missing data are likely to be among the
more depressed patients. This should serve to restrict the range of
the independent variable and make it more difficult to detect an
effect of depression on survival. Thus, any bias introduced by the
missing data should be a conservative one.
Data collection procedures are described in more detail else-
where (11, 12). All procedures used in this study were approved by
the Duke University Medical Center Institutional Review Board.
SDS Symptom Groups
The SDS (8) has 20 items describing depressive symptoms. Re-
spondents describe how frequently they experience each symptom
on a four-point scale ranging from “a little of the time” to “most of
the time.” The SDS is widely used and has been shown to have
satisfactory reliability and validity (13). In this sample, 11% of
patients had SDS scores in the range indicative of moderate or severe
depression, and another 26% had scores indicative of mild
Groups of symptoms were identified with the aid of a principal
components analysis with a varimax rotation. This analysis was
performed on the 1933 patients who had complete data on all SDS
items. This sample included some patients who were not included
in the follow-up analyses because they did not meet the criterion for
significant disease. A SCREE test revealed the presence of four
factors, accounting for 46% of the variance. The items in each group
and their loadings on the factor to which they were assigned are
presented in Table 2. With one exception, the magnitudes of the
loadings were high. The first group contains items describing posi-
tive experiences and feelings of Well-Being. The absence of these
experiences constitutes a high score on this variable. The second
contains symptoms of Negative Affect, and the third contains So-
matic symptoms. The final factor contained only two items, both
TABLE 1. Patient Characteristics
Median age (y)
Median education (y)
Income ?$20,000 (%)
History of smoking (%)
Family history of CAD (%)
Typical angina (%)
Recent myocardial infarction (?30 d) (%)
Number of diseased vessels (75% stenosis) (%)
Median ejection fraction
Median pain (frequency/wk)
Nocturnal pain (%)
Numbers in parentheses refer to 25th and 75th percentiles.
TABLE 2. SDS Symptom Groups, Factor Loadings, and Item-
Well-Being (all reverse scored)
I feel that I am useful and needed.
My life is pretty full.
I find it easy to make decisions.
I am hopeful about the future.
I still enjoy the things that I used to.
My mind is as clear as it used to be.
Morning is when I feel the best.
I feel downhearted, blue, and sad.
I have crying spells or feel like it.
I feel that others would be better off if
I were dead.
I am more irritable than usual.
I am restless and can’t sit still.
I get tired for no reason.
My heart beats faster than usual.
I find it easy to do the things I used to.
I have trouble sleeping through the
I have trouble with constipation.
I notice that I am losing weight.
I eat as much as I used to. (reverse
* p ? .05; ** p ? .01; *** p ? .005.
DEPRESSIVE SYMPTOMS AND SURVIVAL
791Psychosomatic Medicine 62:790–795 (2000)
dealing with Appetite. Scores on each symptom group were calcu-
lated with unit weightings summed across items.
There is no standard scale for hopelessness in the MMPI, so it
was assessed with one item that most clearly reflected that construct:
“I find it hard not to give up hope for the future.” Although not ideal
from a psychometric perspective, brief measures of hopelessness
have been used successfully in other studies (eg, Ref. 6).
Patients were contacted 6 and 12 months after their hospitaliza-
tion and annually thereafter. March 10, 1994, was the end of fol-
low-up for the present analyses. As of that date, follow-up was 97%
complete with only 25 (1.6%) lost and 26 (1.7%) withdrawn from
the study. Follow-up times ranged up to 19.4 years, with a median of
15.2 years for patients still living.
Deaths were classified by a mortality committee into cardiovas-
cular and noncardiovascular categories on the basis of information
provided by the patient’s physician. These procedures have been
described elsewhere (14). Cardiac deaths occurred in 488 patients
and there were 116 deaths from other causes. Cardiac death was the
outcome for these analyses.
It is necessary to control for CAD severity to determine whether
any association between depressive symptoms and survival is due to
confounding with clinical characteristics at baseline. As in previous
studies (2, 12), disease severity was summarized with a “hazard
score” assigned on the basis of a formula devised in analyses based
on the entire population of Duke University Medical Center patients
with CAD from 1969 to 1984 (15). Baseline clinical and anatomic
data obtained during diagnosis were combined into the hazard score
using weights derived from those analyses. Primary components of
the hazard score include age, left ventricular ejection fraction, elec-
trocardiographic abnormalities, number of vessels with ?75% nar-
rowing, and various indicators of myocardial damage (12). Hazard
scores have been shown to be accurate predictors of observed sur-
vival (16) and useful as summary indices of prognostic information.
In addition, treatment status (medical management vs. surgery) was
included in all models as a time-dependent covariate.
Cox proportional hazards survival analyses were performed sep-
arately for each SDS symptom, each symptom group, and hopeless-
ness. Hazard scores, income, gender, and age were evaluated as
potential confounders and moderators. All significant SDS symptom
groups were then combined into the same model. Finally, we fit a
model that combined hopelessness with the significant effects from
the SDS model.
Table 3 presents the intercorrelations of the various
symptom groups and their associations with patient
characteristics. With the exception of the Appetite fac-
tor, the symptoms groups were moderately interre-
lated. Women scored higher on the SDS than did men,
a commonly observed finding. The negative associa-
tions between income and depressive symptoms were
also expected. The finding of a negative association
between age and affective symptoms was less ex-
pected, but it has been observed in other data sets (17;
J. C. Barefoot, unpublished data). The small but signif-
icant associations of hazard scores with somatic symp-
toms and hopelessness suggests that disease severity is
affecting physical symptomatology and that feelings of
hopelessness may partially reflect the patient’s knowl-
edge of the probable extent of their disease.
Initial Survival Models
The results of survival analyses testing each SDS
item individually are summarized in Table 2. The RRs
in Table 2 are the increases in risk associated with a
change of one point on the four-point Likert scale used
as the response format.
Table 4 presents the results of survival analyses for
each symptom group and the entire SDS. The RRs were
calculated as the increase in risk associated with a
2-SD increase in scores on the relevant scale. All vari-
ables but the Appetite items of the SDS significantly
predicted survival, although there was some variation
in effect sizes.
The hopelessness item was a strong predictor of
survival, although its RR cannot be directly compared
with those for the SDS groups because it is a dichoto-
mous measure. Only 88 patients (9.2%) answered the
TABLE 3. Correlations of Symptom Groups With Patient Characteristics and Each Other
Well-Being Negative AffectSomaticAppetiteHopelessness
* p ? .01; ** p ? .001.
J. C. BAREFOOT et al.
792Psychosomatic Medicine 62:790–795 (2000)
hopelessness question affirmatively, but 55.7% of
them died during follow-up, compared with 36.2% of
those who answered it negatively.
Further analyses fitted models designed to evaluate
the independent predictive abilities of the symptom
groups. The first simultaneously included the three
SDS symptom groups found to be predictive in the
initial models. In this model the ? value for the Well-
Being variable was reduced by 60% (RR ? 1.10) com-
pared with the model in Table 4, and it became non-
significant. The ? value for the Somatic term was also
reduced substantially, by 56% (RR ? 1.10), and was
not significant. However, the ? value for the Negative
Affect symptoms was reduced by only 26% (RR ?
1.30), and it retained its statistical significance (p ?
.01). Thus, both the performance of the Negative Affect
measure in the multivariable model and its strength
when evaluated alone suggest that it is the most im-
portant component of the SDS for predicting survival.
We examined the joint effects of hopelessness and
Negative Affect to evaluate their independent contri-
butions to the prediction of survival. Only 867 patients
were included in this model because data were miss-
ing for the others. Inclusion of both terms resulted in
only a 28% decrease in the magnitude of ? (RR ? 1.40)
for hopelessness and a 19% decrease in the Negative
Affect ? (RR ? 1.33). Both were significant or nearly
significant (p ? .06 for hopelessness and p ? .02 for
Negative Affect). Therefore, there is evidence to sug-
gest that hopelessness and depressive affect were in-
dependent predictors of survival.
Role of Demographic Factors
The associations of the Negative Affect symptoms
with age, gender, and income (Table 3) led us to eval-
uate whether its ability to predict survival might be
due to confounding with those characteristics and
whether it might interact with those characteristics.
Controls for age, gender, and income left the magni-
tude of the Negative Affect ? value essentially un-
changed. Negative Affect did not interact with income
or gender, but it did interact with age (Figure 1). For
those older than 50 years, the RR associated with a
2-SD difference in Negative Affect scores was 1.29. For
younger patients it was 1.70. Thus, younger patients
reported more depressive affect and it may have had
more impact on them as well.
Fig. 1. Survival probability by Negative Affect scores (median
split) and age. Upper pair of curves are for younger patients,
and lower pair of curves are for older patients.
TABLE 4.Symptom Groups and Survival
Scale Deaths/NUnadjusted RRp Adjusted RRp
DEPRESSIVE SYMPTOMS AND SURVIVAL
793Psychosomatic Medicine 62:790–795 (2000)
The hopelessness effect was not substantially al-
tered by controls for sex and age, nor did it interact
significantly with those variables. The effect did be-
come nonsignificant (p ? .12) when income was con-
trolled in a model with a considerably smaller sample
(N ? 861). The ? value for hopelessness was reduced
by a modest 33% (RR ? 1.31) in that model.
Examination of groups of depressive symptoms in
patients with CAD revealed that most were predictive
of survival, but some were better than others. Depres-
sive affect and hopelessness had somewhat larger ef-
fect sizes than other types of symptoms, and they
generally retained their strength in multivariable mod-
els that controlled for other symptom and demo-
The impact of depressive affect was especially ap-
parent among younger patients. Elsewhere (18) we
have observed that the influence of social support on
the course of depressive symptoms is more potent in
younger patients. This led us to suggest that the nature
and severity of depression in patients with CAD may
be affected by the age at which they develop the dis-
ease. The occurrence of a coronary event may have
more significant meaning in those for whom it is more
unexpected and likely to result in major lifestyle
changes. This is consistent with the observation that
stressors may have more psychological impact if they
occur at an age that is not “on time” developmentally
(19). The higher level of depressive affect in younger
patients and its greater prognostic importance in that
group support this line of reasoning.
Hopelessness was a potent predictor of survival
even though the measure was based on only one ques-
tion that was endorsed by a relatively small number of
patients. As in other studies (6), hopelessness seemed
to have an effect over and above that of other depres-
sion measures. A question was raised by the finding
that hopelessness was not significant when we con-
trolled for income. However, the size of the ? value
was not substantially reduced in that analysis. The
reduced power of that model due to missing data sug-
gests that the p value of the effect may not be a good
guide to its importance. Furthermore, it is not possible
to tell whether the effect of hopelessness is due to
confounding with some factor associated with income
or whether the effect of income is due to its association
with hopelessness. Certainly there should be further
investigations of the correlates (particularly income)
and consequences of hopelessness in patient samples
using more sophisticated measures.
In addition to identifying the potential importance
of depressive affect and hopelessness, these data show
that it is not necessary to include somatic symptoms to
demonstrate a relationship between depressive symp-
toms and survival. This, coupled with controls for
clinical indicators of disease by means of the hazard
score, further supports the notion that depressive
symptoms are not simply surrogates for underlying
physical disease. Other mechanisms must be sought to
explain their effects.
Many studies have demonstrated that depressed pa-
tients with coronary disease are at increased risk for
mortality (1–3, 20), other coronary events (21), disabil-
ity (22), and high levels of medical care utilization
(23). It is now appropriate to move beyond demonstra-
tions of these effects to a more detailed understanding
of the phenomena. Investigation of potential behav-
ioral and physiological mechanisms is in order (24). A
significant part of this strategy could be the identifica-
tion of those depressive symptoms that are most im-
portant and the types of patients whose prognoses are
most affected by them. These findings could have
treatment implications by identifying the classes of
symptoms that need to be ameliorated and the types of
patients most likely to benefit from interventions. The
data of the present study call attention to depressive
affect and hopelessness as potentially central to the
phenomena and suggest that depressive affect may be
especially significant in the prognosis of younger
This research was supported in part by Grants P01
HL36587, R01 HL45702, and R01 HL54780 from the
National Heart, Lung, and Blood Institute; Grant R01
AG12458 from the National Institute on Aging; and
Grants T32 MH19109 and R05 MH70482 from the Na-
tional Institute of Mental Health.
1. Frasure-Smith N, Lesperance F, Talajic M. Depression and 18-
month prognosis after myocardial infarction. Circulation 1995;
2. Barefoot JC, Helms MJ, Mark DB, Blumenthal JA, Califf RM,
Haney TL, O’Connor CM, Siegler IC, Williams RB. Depression
and long-term mortality risk in patients with coronary artery
disease. Am J Cardiol 1996;78:613–7.
3. Ahern D, Gorkin L, Anderson J, Tierney C, Hallstrom A, Ewart C,
Capone R, Schron E, Kornfeld D, Herd J, Richardson D, Follick
M. Biobehavioral variables and mortality or cardiac arrest in the
Cardiac Arrhythmia Pilot Study (CAPS). Am J Cardiol 1990;66:
4. Appels A, Mulder P. Fatigue and heart disease: the association
between vital exhaustion and past, present and future coronary
heart disease. J Psychosom Res 1989;33:727–38.
J. C. BAREFOOT et al.
794Psychosomatic Medicine 62:790–795 (2000)
5. Appels A, Kop W, Meesters C, Markusse R, Golombek B, Falger Download full-text
P. Vital exhaustion and the acute coronary syndromes. Int Rev
Health Psychol 1994;3:65–95.
6. Everson SA, Goldberg DE, Kaplan GA, Cohen RD, Pukkala E,
Tuomilehto J, Salonen JT. Hopelessness and risk of mortality
and incidence of myocardial infarction and cancer. Psychosom
7. Koenig H, Meador K, Cohen H, Blazer D. Self-rated depression
scales and screening for major depression in the older hospitalized
patient with medical illness. J Am Geriatr Soc 1988;36:699–706.
8. Zung W. A self-rating depression scale. Arch Gen Psychiatry
9. Hathaway S, McKinley J. Manual for the Minnesota Multiphasic
Personality Inventory. New York: Psychological Corp; 1943.
10. Barefoot JB, Helms MJ, Mark DB, Siegler IC, Williams RB. Re-
search participation as a predictor of survival in cardiac
patients: possible mechanisms. Clearwater (FL): American Psy-
chosomatic Society; 1998.
11. Williams R, Haney T, Lee K, Kong Y, Blumenthal J, Whalen R.
Type A behavior, hostility, and coronary atherosclerosis. Psy-
chosom Med 1980;42:539–49.
12. Barefoot JC, Peterson BL, Harrell FE, Hlatky MA, Pryor DB,
Haney TL, Blumenthal JA, Siegler IC, Williams RB. Type A
behavior and survival: a follow up study of 1,467 patients with
coronary artery disease. Am J Cardiol 1989;64:427–32.
13. Zung W. Zung Self-Rating Depression Scale and Depression
Status Inventory. In: Satorious N, Ban T, editors. Assessment of
depression. New York: Springer-Verlag; 1986. p. 221–31.
14. Harris P, Lee K, Harrell F, Behar V, Rosati R. Outcome in medically
treated coronary artery disease. Circulation 1980;62:718–26.
15. Harrell FE, Lee KL, Califf RM, Pryor DB, Rosati RA. Regression
modeling strategies for improved prognostic prediction. Stat
16. Pryor DB, Lee K, Harris PJ, Harell FE, Rosati RA. The effect of
crossovers on estimates of survival in medically treated patients
with coronary artery disease. J Chronic Dis 1984;37:521–9.
17. Aromaa A, Raitasalo R, Reunanen A, Impivaara O, Heliövaara
M, Knekt P, Lehtinen V, Joukamaa M, Maatela J. Depression and
cardiovascular diseases. Acta Psychiatr Scand Suppl 1994;377:
18. Barefoot JC, Brummett BH, Clapp-Channing NE, Siegler IC, Vi-
taliano PP, Williams RB, Mark DB. Moderators of the effect of
social support on depressive symptoms in cardiac patients. Am J
19. Neugarten BL. Time, age, and the life cycle. Am J Psychiatry
20. Ladwig K, Konig J, Breithardt G, Borggrefe M. Affective disor-
ders and survival after acute MI. Eur Heart J 1991;12:959–64.
21. Carney RM, Rich MW, Freedland KE, Saini J, teVelda A, Sime-
one C, Clark K. Major depressive disorder predicts cardiac
events in patients with CAD. Psychosom Med 1988;50:627–33.
22. Hlatky MA, Haney T, Barefoot JC, Califf RM, Mark DB, Pryor DB,
Williams RB. Medical, psychological, and social correlates of
work disability among men with coronary artery disease. Am J
23. Frasure-Smith N, Lesperance F. Depression and anxiety increase
physician costs during the first post-MI year (Abstract). Psycho-
som Med 1998;60:99.
24. Carney R, Freedland K, Rich M, Jaffe A. Depression as a risk
factor for cardiac events in established coronary heart disease: a
review of possible mechanisms. Ann Behav Med 1995;17:142–9.
DEPRESSIVE SYMPTOMS AND SURVIVAL
795Psychosomatic Medicine 62:790–795 (2000)