Evidence based cardiology: Psychosocial factors in the aetiology and prognosis of coronary heart disease: systematic review of prospective cohort studies

Abstract

In healthy populations, prospective cohort studies show a possible aetiological role for type A/hostility (6/14 studies), depression and anxiety (11/11 studies), psychosocial work characteristics (6/10 studies), social support (5/8 studies). In populations of patients with coronary heart disease, prospective studies show a prognostic role for depression and anxiety (6/6 studies), psychosocial work characteristics (1/2 studies), and social support (9/10 studies); none of five studies showed a prognostic role for type A/hostility. Although this review can not discount the possibility of publication bias, prospective cohort studies provide strong evidence that psychosocial factors, particularly depression and social support are independent aetiological and prognostic factors for coronary heart disease.

Full text

Clinical review
Evidence based cardiology
Psychosocial factors in the aetiology and prognosis of
coronary heart disease: systematic review of prospective
cohort studies
Harry Hemingway, Michael Marmot
Do psychosocial factors cause coronary heart disease
or affect survival among patients with coronary heart
disease? Here we use an explicit methodological qual-
ity filter to review systematically the prospective
cohort studies testing specific psychosocial hypoth-
eses. This review of the epidemiological literature
identifies the psychosocial factors that have been most
rigorously tested. Only four psychosocial factors met
the quality filter: type A/hostility, depression and anxi-
ety, work characteristics, and social supports. The
importance of other study designs
—
for example, eco-
logical1or nested case-control2–4 studies
—
is acknowl-
edged. The review should be seen as complementary
to existing reviews5–8 on single psychosocial factors
and as a challenge to investigators in the field to
ensure that the systematic review is made unbiased,
kept up to date, and used to guide future hypothesis
testing.
What is a psychosocial factor?
A psychosocial factor may be defined as a measure-
ment that potentially relates psychological phenomena
to the social environment and to pathophysiological
changes. The validity and reliability (precision) of the
questionnaire based instruments used to measure psy-
chosocial factors has been improved through the use
of psychometric techniques. By avoiding the unhelpful
general term of “stress,” recent work has developed
theoretical models
—
for example, the job control-
demands-support model of psychosocial work
characteristics
—
which generate specific hypotheses
that can be tested.
How might psychosocial factors be
linked to coronary heart disease?
Evidence of mechanisms linking psychosocial factors
with coronary heart disease (reviewed elsewhere910
)is
important in making causal inferences and therefore
in designing preventive interventions. Psychosocial
factors may act alone or combine in clusters11 and may
exert effects at different stages of the life course.12
Broadly, three interrelated pathways may be consid-
ered. Firstly, psychosocial factors may affect health
related behaviours such as smoking, diet, alcohol
consumption, or physical activity, which in turn may
influence the risk of coronary heart disease.13 If such
behaviours do lie on the causal pathway between psy-
chosocial factors and coronary heart disease, then
treating them as confounding variables, as some stud-
ies do, must be questioned.Secondly,psychosocial fac-
tors may cause direct acute or chronic patho-
physiological changes. Thirdly, access to and content
of medical care may plausibly be influenced by, for
example, social supports (but there is little direct
evidence for this). Although it is beyond the scope of
this review to consider the determinants of adverse
psychosocial factors, socioeconomic status is inversely
associated with coronary heart disease14 and also with
certain psychosocial factors, and it has been proposed
that psychosocial pathways may play a mediating
role.15 16
website
extra
References in the
tables are given on
the BMJ website
www.bmj.com
Summary points
In healthy populations, prospective cohort
studies show a possible aetiological role for type
A/hostility (6/14 studies), depression and
anxiety (11/11 studies), psychosocial work
characteristics (6/10 studies), social support
(5/8 studies)
In populations of patients with coronary heart
disease, prospective studies show a prognostic
role for depression and anxiety (6/6 studies),
psychosocial work characteristics (1/2 studies),
and social support (9/10 studies); none of five
studies showed a prognostic role for type
A/hostility
Although this review can not discount the
possibility of publication bias, prospective cohort
studies provide strong evidence that psychosocial
factors, particularly depression and social support,
are independent aetiological and prognostic
factors for coronary heart disease
This is the third
of four articles
International
Centre for Health
and Society,
Department of
Epidemiology and
Public Health,
University College
London Medical
School, London
WC1E 6BT
Harry Hemingway,
senior lecturer in
epidemiology
Michael Marmot,
professor of
epidemiology and
public health
Correspondence to:
Dr H Hemingway,
Department of
Research and
Development,
Kensington and
Chelsea and
Westminster Health
Authority, London
W2 6LX
h.hemingway@
public-health.ucl.
ac.uk
BMJ 1999;318:1460–7
1460 BMJ VOLUME 318 29 MAY 1999 www.bmj.com

Method of systematic review
A methodological quality filter was used to select studies
for inclusion in the systematic review, so that the
strength of evidence could be compared across psycho-
social factors. Prospective cohort studies are the best
observational design for questions of aetiology and
prognosis. The studies included had a prospective
cohort design; a population based sample (aetiological
studies in healthy populations); at least 500 participants
(aetiological studies) or 100 participants (prognostic
studies in populations of patients with coronary heart
disease); measurements of a psychosocial factor used in
at least two different study populations; outcomes of
fatal coronary heart disease or non-fatal myocardial inf-
arction or (prognostic studies only) all cause mortality.
Articles were identified by Medline search (1966-
97), manual searching of the bibliographies ofretrieved
articles, previous review articles, writing to research-
ersin the field, and an in-house bibliographic database.
No register of published and unpublished studies with
Table 1 Studies of type A behaviour, hostility, and coronary heart disease. References in this table are given on the BMJ website
Author, year, country
Total sample
(% women)
Age at
entry Exposure
Follow up
(years)
No of
events Type of events Adjustments Relative risk* Summary†
Prospective aetiological studies
Jenkins 1974w1 USA 2750 (0) 39-59 Type A 4 120 Non-fatal MI +
angina
Age 1.8* +
Rosenman 1976w2 USA 3154 (0) 39-59 Type A 8.5 257 Fatal CHD +
non-fatal MI
Age, smoking, cholesterol,
family history, corneal arcus,
schooling, â:álipoprotein ratio
2.16* ++
Haynes 1980w3 USA 1674 (57) 45-77 Type A
(Framingham)
8 170 Fatal CHD +
non-fatal MI +
coronary
insufficiency +
angina
Age, smoking, blood pressure,
cholesterol, glucose intolerance
and other psychosocial factors
1.8*; among
men, the effect
was confined
to white collar
workers
+
Shekelle 1983w4 USA 1877 (0) 40-55 Hostility
(MMPI)
10 139 Fatal CHD +
non-fatal MI
Age, smoking, blood pressure,
cholesterol, alcohol
1.47*, but
effect not linear
+
Cohen 1985w5 USA 2187 (0) 57.8
(mean)
Type A (JAS) 8 190 Fatal CHD +
non-fatal MI +
angina
Smoking, blood pressure,
cholesterol, body mass index,
alcohol, and other biological
factors
1.43, Type A
associated with
prevalence, not
incidence or
postmortem
findings
0
Shekelle 1985w6 USA 3110 (0) 46 (mean) Type A (JAS) 7.1 554 Fatal CHD +
non-fatal MI
Age, smoking, blood pressure,
cholesterol, alcohol, education
0.87 0
Johnston 1987w7 UK 5936 (0) 40-59 Type A
(Bortner)
6.2 255 Fatal CHD +
non-fatal MI
Age, social class 0.89 0
Ragland 1988w8 USA 3154 (0) 39-59 Type A (SI) 22 214 Fatal CHD Age, smoking, blood pressure,
cholesterol
0.98 0
Hearn 1989w9 USA 1399 (0) 19 Hostility
(MMPI)
33 54 Fatal CHD +
non-fatal MI +
angina + coronary
surgery
Smoking, hypertension, family
history
1.1; no
association in
crude or risk
factor adjusted
analyses
0
Barefoot 1995w10 USA 730 (44) 50 Hostility
(Cook-Medley)
27 122 Non-fatal MI Age, sex, smoking, blood
pressure, triglycerides, exercise
1.26 (men)
2.95* (women)
0 (men)
++(women)
Bosma 1995w11
Lithuania and
Netherlands
5817 (0) 45-60 Type A (JAS) 9.5 394 Fatal CHD +
non-fatal MI
Age No association 0
Kawachi 1996w12 USA 1305 (0) 40-90 MMPI-2 anger
content scale
7 110 Fatal CHD +
non-fatal MI +
angina
Age, smoking, blood pressure,
cholesterol, body mass index,
family history, alcohol
2.66* ++
Everson 1997w13 Finland 1599 (0) 42-60 Cynical hostility
(Cook-Medley)
6 60 First MI Age, biological, socioeconomic,
behavioural, social support,
prevalent diseases
1.43 (2.18*
when adjusted
for age only)
0
Tunstall-Pedoe 1997w14
Scotland
11659 (50) 40-59 Type A
(Bortner)
7.6 581 Fatal CHD +
non-fatal MI +
coronary surgery
Age 0.82* in
women,
ie type A
protective
0
Prognostic studies
Case 1985w15 USA 516 (18) patients
<14 days post MI
<70 Type A (JAS) 2 53 Fatal CHD and all
cause mortality
Age, sex, education, rales,
ejection fraction, New York Heart
Assocation functional class,
ventricular premature beats
0.8 0
Shekelle 1985w16 USA 2314 (11) patients
post MI
30-69 Type A (JAS) 3 294 Non-fatal MI and
fatal CHD
Smoking, previous MI, angina,
fasting glucose
No association 0
Ragland 1988w17 USA 257 (0) with MI or
angina
39-70 Type A (SI) 11.5 91 Fatal CHD Age at initial event, follow up
time, type of initial coronary
event, smoking, blood pressure,
cholesterol
0.58*; type A
protective
0
Barefoot 1989w18 USA 1467 (18) patients
with angiographic
disease
mean 52
(SD 9)
Type A (SI) 5 315 Fatal CVD +
non-fatal MI
Stratified on clinical prognostic
factors
No association
with non-fatal
MI
0
Jenkinson 1993w19 UK 1376 (22) 7 days
post-MI
25-84 Type A 3 247 All cause mortality Age, previous MI,hospital
complications, diabetes,
hypertension, car ownership, sex
No association 0
CHD=coronary heart disease; MI=myocardial infarction; SI=structured interview; JAS=Jenkins activity survey; MMPI=Minnesota multiphasic personality inventory.
*P<0.05. †0=no association—that is, relative risk not significantly different from unity; +=moderate association (relative risk >1<2.0); ++=strong association (relative risk >2.0).
Clinical review
1461BMJ VOLUME 318 29 MAY 1999 www.bmj.com

psychosocial exposures exists, and hand searching of
journals was not performed, so there is a serious
potential for publication bias. For this reason as well as
the lack of standardised methods of measurement of
psychosocial factors, we carried out a narrative, rather
than quantitative, systematic review. Given that
randomised controlled trials, at least for primary
prevention, are rarely feasible,observational studies are
likely to remain the main type of evidence on which to
base preventive action.
Evidence for specific psychosocial factors
Largely on the basis of studies in middle aged men
(table 1), four groups of psychosocial factors were
identified by using the predefined quality filter:
psychological traits (type A behaviour, hostility),
psychological states (depression, anxiety), psychologi-
cal interaction with the organisation of work (job
control-demands-support), and social networks and
social support. In simple terms this corresponds to a
spectrum with mainly psychological components at
one end and a stronger social component at the other.
The box summarises the key results.
Hostility and type A behaviour
Type A behaviour pattern
—
the only personality trait
which met the criteria of our review
—
is characterised
by hard driving and competitive behaviour, a potential
for hostility, pronounced impatience, and vigorous
speech stylistics. The instruments for measurement of
type A behaviour and hostility
—
the Jenkins activity
scale, the structured interview, the Minnesota multi-
phasic personality inventory (MMPI), the Bortner hos-
tility scale
—
have been subjected to psychometric
testing and incorporated into many cardiovascular
cohort studies, including some that have not reported
results. Unlike other psychosocial factors, type A is dis-
tinguished by being the subject of numerous interven-
tion trials.17 On the basis of early positive findings in
the Framingham study18 and the Western Collaborative
Group’s eight year follow up,19 among other evidence,
the National Institutes of Health declared type A an
independent risk factor for coronary heart disease.
However, with the publication of negative findings20–22
it was proposed that a more specific component of
type A, namely hostility,might be aetiological, although
there are conflicting studies. None of the five studies
that examined type A or hostility in relation to
prognosis among patients with coronary heart disease
showed an increased risk; indeed, one suggested a
protective effect.
Depression and anxiety
The relation between depression and anxiety and cor-
onary heart disease differs from those of other psycho-
social factors for several reasons. Firstly, unlike other
psychosocial factors, depression and anxiety represent
well defined psychiatric disorders, with standardised
instruments for measurement. Secondly, depression
and anxiety are commonly the consequence of coron-
ary heart disease, and the extent to which they are also
the cause poses important methodological issues.
Thirdly, the ability to diagnose and treat such disorders
makes them attractive points for intervention. Finally,
depression and coronary heart disease could share
common antecedents
—
for example, environmental
stressors and social supports.
Table 2 shows the results from the 11 prospective
studies that investigated depression or anxiety in the
aetiology of coronary heart disease, all of which were
positive. All three of the prospective studies examining
the effect of anxiety in the aetiology of coronary heart
disease had positive results. Intriguingly, there is some
evidence that this effect is strongest specifically for
phobic anxiety and sudden cardiac death. Wassertheil-
Smoller23 reported the effect of depression in relation
to cardiovascular events among 4367 healthy older
people. An increase in depression symptoms (but not
the baseline scores) predicted events, even when multi-
ple covariates were controlled for. Such findings are
compatible with the hypothesis that premonitory signs
of coronary heart disease such as angina or breathless-
ness may have led to the increase in depression.
Studies with longer periods of follow up are less likely
to be confounded by the possibility of early disease
causing depression, but raise further questions about
Studies showing role of psychosocial factors
In healthy populations, prospective cohort studies
suggest a possible aetiological role for:
•Type A/hostility (6/14 studies)
•Depression and anxiety (11/11 studies)
•Psychosocial work characteristics (6/10 studies)
•Social support (5/8 studies)
In coronary heart disease patient populations,
prospective studies suggest a prognostic role for:
•Type A/hostility (0/5 studies)
•Depression and anxiety (6/6 studies)
•Psychosocial work characteristics (1/2 studies)
•Social support (9/10 studies)
Although this review cannot discount the possibility of
publication bias, prospective cohort studies provide
strong evidence that psychosocial factors, particularly
depression and social support, are independent
aetiological and prognostic factors for coronary heart
disease.
SUE SHARPLES
Clinical review
1462 BMJ VOLUME 318 29 MAY 1999 www.bmj.com

Table 2 Studies of depression and anxiety and coronary heart disease. References in the table are given on the BMJ website
Author, year,
country
Total sample
(% women)
Age at
entry Exposure
Follow
up
(years)
No of
events Type of events Adjustments Relative risk* Summary†
Prospective aetiological studies
Hallstrom 1986w20
Sweden
795 (100) 38-54 Depression
(Hamilton and
psychiatric
interview)
12 75 Non-fatal MI +
angina + ischaemic
changes on
electrocardiograph
Age, social class, marital
status, conventional risk
factors
5.4* severity of depression
predicted angina but not other
outcomes
++
Hagman 1987w21
Sweden
5735 (0) 55
(mean)
Anxiety
(“stress”)
2-7 162 Angina with or
without MI
Age, smoking, blood
pressure, cholesterol,
relative weight
Strong predictor for angina
alone
+
Haines 1987w22 UK 1457 (0) 40-64 Phobic anxiety
(Crown-Crisp)
10 113 Fatal CHD +
non-fatal MI
fibrinogen, cholesterol, factor
VII, systolic blood pressure
3.77* for fatal CHD ++
Appels 1990w23
Netherlands
3877 (0) 39-65 Depression 4.2 59 Non-fatal MI +
unstable angina +
angina
Age, smoking, blood
pressure, cholesterol
1.86* for unstable angina for
combination of low mood,
low energy, hopelessness,
poor sleep (termed “vital
exhaustion”)
+
Anda 1993w24 USA 2832 (52) 45-77 Depression
(General Well
Being)
12 394 Fatal CHD +
non-fatal CHD
hospitalisations
Age, sex, race, education,
marital status, smoking,
blood pressure, cholesterol,
body mass index, alcohol,
exercise
1.6* +
Aromaa 1994w25
Finland
5355 (55) 40-64 Depression (GHQ
and PSE)
6.6 91 Fatal CHD Age, pre-existing
cardiovascular disease
3.36* (5.52 in those with
pre-existing cardiovascular
disease)
++
Kawachi 1994w26
USA
33999 (0) 42-77 Phobic anxiety
(Crown Crisp)
2 168 Fatal CHD +
non-fatal MI
Age, smoking, blood
pressure, cholesterol, body
mass index, diabetes,
parental history of MI,
alcohol, exercise
3.01* (6.08 when sudden
cardiac death examined)
++
Everson 1996w27
Finland
2428 (0) 42-60 Hopelessness 6 95 Non-fatal MI Age, smoking, blood
pressure, cholesterol
education, income, exercise,
alcohol, lipids, social
supports, depression
2.05* ++
Wassertheil-Smoller
1996w28 USA
4367 (53) 72
(mean)
Depression
(CES-D)
4.5 321 Non-fatal MI +
non-fatal strokes
Age, smoking, baseline
depression, sex, race,
randomisation group,
education, history of stroke,
MI, diabetes, and baseline
ADL
1.18* per 5 unit increase in
depression score (baseline
scores alone did not predict
events)
+
Barefoot 1996w29
Denmark
730 (44) 50 or 60 Depression
(MMPI-obvious
depression scale)
27 122 Non-fatal MI Age, conventional CHD risk
factors, baseline CHD
1.7* for 2 SD difference in
depression score
+
Kubzansky 1997w30
USA
1759 (0) 21-80 Social conditions
worry scale
20 323 Fatal CHD +
non-fatal MI +
angina
Age, smoking, blood
pressure, cholesterol, body
mass index, family history,
alcohol
1.23* per 1 point increase in
social conditions worry scale
+
Prognostic studies
Ahern 199031 USA 353 Depression
(Beck), anxiety
(Spielberger)
12 Fatal CHD Age, left ventricular
dysfunction and previous MI
1.3* for depression +
Kop 199432
Netherlands
127 (17)
patients 2
weeks after
coronary
angioplasty
56
(SD 9)
Maastricht
questionnaire for
vital exhaustion
1.5 29 Fatal CHD +
non-fatal MI +
further
revascularisation +
increase in coronary
atherosclerosis +
new angina
Age, sex, smoking, blood
pressure, cholesterol,
severity of coronary artery
disease, clinical presentation
2.34 (P=0.06) +
Ladwig 199433
Germany
377 (0) 17-21
days after
acute MI
29-65 Depression
(interview)
0.5 Angina, not
returning to work,
continuing to smoke
Age, social class, recurrent
infarction, rehabilitation,
cardiac events and
helplessness
2.31* for the effect on angina;
depression predicted all
outcomes
++
Frasure-Smith
1995w34 USA
222 (21)
patients 5-15
days after
acute MI
24-88 Depression
(diagnostic
interview
schedule)
1.5 21 All cause mortality
and fatal CHD
Age, Killip class, premature
ventricular contractions and
previous MI
6.64* effect of depression
higher in those with (10
premature contractions per
hour
++
Barefoot 1996w35
USA
1250 (18)
patients with
angiographic
disease
52
(mean)
Depression
(Zung)
19.4 604 All cause mortality
and fatal CHD
Disease severity, initial
treatment
1.66*, 1.84* and 1.72* in
three follow up periods (year
1, 5-10 and >10 respectively)
+
Denollet 1996w36
Belgium
303 (12)
patients with
angiographic
disease
31-79 Type D
personality
(suppression of
emotional
distress),
depression,
social alienation
7.9 38 All cause mortality
and fatal CHD
Left ventricular function,
number of diseased vessels,
low exercise tolerance, lack
of thrombolytic therapy
4.1* for type D and 2.7* for
depression
++
CHD=coronary heart disease; MI=myocardial infarction; CES-D=Center for Epidemiological Studies-Depression scale; GHQ=general health questionnaire; PSE=present state examination;
MMPI=Minnesota multiphasic personality inventory.
*P<0.05. †0=no association—that is, relative risk not significantly different from unity; +=moderate association (relative risk >1<2.0); ++=strong association (relative risk >2.0) association.
Clinical review
1463BMJ VOLUME 318 29 MAY 1999 www.bmj.com

the time course of exposure. For example, it is possible
that there is a common trigger (such as viral illness)
that precipitates both symptoms of depression and
atherothrombotic processes. By examination of sub-
clinical manifestations of coronary heart disease (using
non-invasive measures of arterial structure and
function, for example) before the onset of symptoms,
the temporal sequence of the relation might be better
understood.
Depression in patients after myocardial infarction
seems to be of prognostic importance beyond the sever-
ity of coronary artery disease. Although discrete major
depressive episodes are not uncommon after a myocar-
dial infarction, depressive symptoms are more prevalent.
Given the graded relation between depression scores
and risk, the long lasting nature of the effect, and the
stability of the depression measured across time, it has
been proposed that depression is a continuously
distributed chronic psychological characteristic.
Psychosocial work characteristics
The longstanding observation that rates of coronary
heart disease vary markedly among occupations
—
more than can be accounted for by conventional risk
factors for coronary heart disease
—
has generated a
quest for specific components of work that might be of
aetiological importance. The dominant “job strain”
model of psychosocial work characteristics, as pro-
posed by Karasek and Theorell, grew out of secondary
analyses of existing survey data on the labour force.
This model proposes that people in jobs characterised
by low control over work and high conflicting demands
might be high strain. A subsequent addition to the
model was that social support might buffer this effect.
The advantage of the model is that it generates specific
hypotheses for testing.
Table 3 shows prospective cohort studies that have
examined the relation between job strain and coronary
heart disease. Both self reports and ecological
Table 3 Studies of psychosocial work characteristics and coronary heart disease. References in table are given on BMJ website
Author, year,
country
Total sample
(% women)
Age at
entry Exposure
Follow
up
(years)
No of
events Type of events Adjustments Relative risk* Summary†
Prospective aetiological studies
LaCroix 1984w37
USA
876 (37) 45-64 Job control and
demands (individual
and ecological)
10 Not
stated
Fatal CHD + non-fatal MI
+ coronary insufficiency
+ angina
Age, smoking, blood
pressure, cholesterol
2.9* all women (clerical
women RR=5.2) no
association in men.
Ecological exposure was
associated with risk in men
and women
+
Alfredsson 1985w38
Sweden
958 096 (51) 20-64 Hectic work and few
possibilities for
learning (ecological)
1 1201 Non-fatal MI
(hospitalisation)
Age, 10 sociodemographic
factors, smoking, heavy
lifting
1.5* +
Haan 1988w39
Finland
902 (33) factory
workers
20-62 Job control, physical
strain, variety
(individual)
10 60 Fatal CHD + and
non-fatal CHD
Age, smoking, blood
pressure, cholesterol,
alcohol, relative weight
4.95* for low control, low
variety, high physical strain
++
Reed 1989w40
Hawaii
(Japanese
ancestry)
4737 (0) 45-65 Job control, demands
and their interaction
(ecological)
18 359 Fatal CHD and non-fatal
MI
Age No effect of control,
demands or their interaction
(ns trend for lower strain
men to have higher CHD)
0
Netterstrom
1993w41
Denmark
2045 (0) bus
drivers
21-64 Job variety,
satisfaction
10 59 Fatal CHD Age 2.1*—high job variety and
satisfaction associated with
CHD risk
0
Suadicani 1993w42
Denmark
1752 (0) 59
(mean)
Job influence,
monotony, pace,
satisfaction, ability to
relax
3 46 Fatal CHD + non-fatal MI None Only inability to relax after
work associated with CHD
0
Alterman 1994w43
USA
1683 (0) 38-56 Job control, demands
and their interaction
(ecological)
25 283 Fatal CHD Age 1.4 for job strain 0
Bosma 1997w44 UK 10 308 (33) civil
servants
35-55 Job control, demands
(individual, assessed
twice 3 years apart,
and ecological)
5 654 Angina + doctor
diagnosed ischaemia
Age, smoking, blood
pressure, cholesterol, body
mass index, employment
grade
1.93* self reported or
externally assessed low job
control predicted CHD
+
Lynch 1997w45
Finland
1727 (0) 42-60 Job demands,
resources, income
8.1 89 Fatal CHD + non-fatal MI Age, behavioural, biological
and psychosocial covariates
1.57* for the effect of high
demands, low resources and
low income; 2.59 when
adjustment made for age only
+
Steenland 1997w46
USA
3575 (0) 25-74 Job control and
demands (ecological)
14 519 Fatal CHD + non-fatal MI Age, smoking, blood
pressure, cholesterol,
education, body mass
index, self reported
diabetes
1.41* for low control +
Prognostic studies
Hlatky 1995w47
USA
1489 (24)
employed
patients
undergoing
coronary
angiography
41-59 Job control, demands
(individual)
5 112 Fatal CHD + non-fatal MI
prevalence of coronary
artery disease
Ejection fraction, extent of
coronary atherosclerosis,
myocardial ischaemia
0.96 for effect of job strain
on events. Job strain was
associated with normal
coronary arteries
-
Hoffmann 1995w48
Switzerland
222 (0) after
first MI
30-60 Job work load, locus
of control, social
supports
1 19 All cause mortality,
reinfarction, severe
symptoms, or poor
exercise capacity
Age, severity of MI,
exercise
High workload and low
external locus of control
associated with outcome
+
CHD=coronary heart disease; MI=myocardial infarction.
*P<0.05. †0=no association—that is, relative risk not significantly different from unity; +=moderate association (relative risk >1<2.0); ++=strong association (relative risk >2.0) association.
Clinical review
1464 BMJ VOLUME 318 29 MAY 1999 www.bmj.com

Table 4 Studies of social networks and social supports and coronary heart disease. References in table are given on BMJ website
Author, year,
country
Total sample
(% women)
Age at
entry Exposure
Follow
up
(years)
No of
events Type of events Adjustments Relative risk* Summary†
Prospective aetiological studies
Medalie 1976w49
Israel
10 000 (0) >40 Perceived love and
support from
spouse
5 300 Angina Age, blood pressure, cholesterol,
diabetes, ECG abnormalities
1.8* +
House 1982w50
USA
2754 (52) 35-69 Social relationships
and activities
11 114 Fatal CHD Age, baseline CHD, smoking, forced
expiratory volume at 1 second
Not stated +
Berkman 1983w51
USA
4725 (53) 30-69 Social network
index
9 120 Fatal CHD Age 2.13* ++
Reed 1983w52
USA
4653 (0) 52-71 Social network
score
6 218 Fatal CHD +
non-fatal
Age, blood pressure, cholesterol,
glucose, uric acid, forced vital
capacity, body mass index, exercise,
alcohol, complex carbohydrate
Social network associated
with CHD prevalence, but
not incidence
0
Kaplan 1988w53
Finland
13301 39-59 Social network
index
5 223 Fatal CHD Age, smoking, blood pressure,
cholesterol, prevalent illness, urban/
rural residence
1.34 for men but not
women
0
Vogt 1992w54 USA 2603 (54) 18-75+ Network scope,
network frequency,
and network size
15 not
stated
Fatal CHD +
non-fatal CHD
Age, sex, SES, smoking and
subjective health status at baseline
1.5* for effect of network
scope on CHD incidence;
all 3 measures predicted
survival in those with
CHD
+
Orth-Gomer
1993w55
Sweden
736 (0) 50 Emotional support
from close people
and support from
extended network
(social integration)
6 25 Fatal CHD +
non-fatal CHD
Age, cholesterol treatment of
hypertension, diabetes, body mass
index< smoking, physical activity
3.8* for social integration
3.1 for emotional support
++
Kawachi 1996w56
USA
36 624 (0) 42-77 Social network
index
4 403 Fatal CHD +
non-fatal MI
Age, time period, smoking, blood
pressure, cholesterol, diabetes,
angina, body mass index, family
history, alcohol, exercise
1.14. Some evidence for
association with fatal CHD
(particularly non-sudden
cardiac death) rather than
non-fatal MI
0
Prognostic studies
Chandra 1983w57
USA
1401 Not stated Marital status 10 Not
stated
All cause
mortality
Age, race, smoking, severity of MI,
medical care factors
Married men and women
had better in hospital and
10 year survival
+
Ruberman
1984w58 USA
2320 (0) patients
with MI
30-69 Social support, life
stress
3 128 All cause
mortality, sudden
cardiac death
Age, myocardial function,
ventricular arrhythmia, smoking
4.5* for the effect of
social isolation + high life
stress on all cause
mortality; 5.62 for
sudden cardiac death
++
Wiklund 1988w59
Sweden
201(0) patients
with first MI
32-60 Social support,
depression and
other psychosocial
factors
8.3 85 All cause
mortality +
recurrent
non-fatal MI
Hypertension, smoking, angina Being single increased
risk of death
+
Case 1992w60
USA
1234 (38)
participants in
diltiazem post-MI
trial
25-75 Living alone,
disrupted marriage
2 226 Fatal CHD +
recurrent
non-fatal MI
New York Heart Assocation
functional class, ejection fraction,
education, no âblockers,
ventricular premature complexes,
prior infarction
1.54* for effect of living
alone. No effect of
marital disruption
+
Hedblad 199261
Sweden
98 (0) men with
ischaemic 24
hour ECG
68 Social support and
social network
5 17 Fatal CHD +
non-fatal MI
Age, smoking, blood pressure,
cholesterol, alcohol, exercise, body
mass index, triglycerides
5.6* and 4.1* for low
informational support
and low emotional
support respectively
++
Williams 1992w62
USA
1368 (18)
patients with
angiographicdis
ease
52
(median)
Structural social
support (marital
status) and function
social support
9 249 All cause
mortality
Age, ejection fraction, non-invasive
myocardial damage index,
conduction disturbance, pain/
ischaemic index, mitral regurgitation,
number of diseased vessels, %
stenosis of left main stem and left
anterior descending artery
3.34* for effect of
unmarried patients
without confidant
++
Berkman 1992w63
USA
194 (48) patients
with acute MI
65-85+ Emotional support 0.5 76 All cause
mortality
Age, sex, Killip class, ejection
fraction, reinfarction, comorbidity,
functional disability, previous MI,
ventricular tachycardia
2.9* for lack of
emotional support
+
Gorkin 1993w64
USA
1322 (17)
patients with
previous MI +
ventricular
premature
complexes
60.8
(SD 9.9)
Social support 0.8 Not
stated
All cause
mortality
Ejection fraction, arrhythmia rates,
CHD risk factors,
1.46* for 1 point
decrease in social
support
+
Jenkinson
1993w19 UK
1376 (22) 7 days
after MI
25-84 Social isolation, life
stress, depression,
type A
3 247 All cause
mortality
Age, previous MI, hospital
complications, diabetes,
hypertension, car ownership, sex
1.33 for social support;
no effect of type A or
depression
0
Friedman 1995w65
USA
369 (15) patients
after acute MI
with ventricular
arrhythmias in
the CAST
63 (SD 9) Social support, life
events, depression,
anxiety, type A,
anger
1 20 All cause
mortality
Physiological severity, demographic
and other psychosocial factors
Not stated +
CHD=coronary heart disease; MI=myocardial infarction; CAST=cardiac arrhythmia suppression trial.
*P<0.05. †0=no association—that is, relative risk not significantly different from unity; +=moderate association (relative risk >1<2.0); ++=strong association (relative risk >2.0).
Clinical review
1465BMJ VOLUME 318 29 MAY 1999 www.bmj.com

measurements (assigning a score on the basis of job
title) of job strain have been made. Self reports may be
biased by early manifestations of disease, and ecologi-
cal measurements may lack precision. The finding that
these methods tend to give reasonably consistent
results suggests that they are complementary. Six of the
10 studies were had positive results. There is growing
emphasis on the importance of low job control rather
than on conflicting demands,24 and it seems likely that
these empirical results will lead to a reformulation of
the model. Alternative models of psychosocial work
characteristics involve an imbalance between the effort
at work and rewards received.25 26
Social network structure and quality of social
support
Social supports and networks relate to both the
number of a person’s social contacts and their quality
(including emotional support and confiding support).
Marital status
—
information routinely sought in clinical
practice
—
is a simple measure of social support, and the
ability of low social support to predict all cause
mortality has long been recognised. It has been
proposed that social supports may act to buffer the
effect of various environmental stressors and hence
increase susceptibility to disease,27 but most of the evi-
dence supports a direct role.
Five of the eight prospective cohort studies that
investigated aspects of social support in relation to the
incidence of coronary heart disease were positive (table
4). Nine of the 10 prognostic studies were positive,and
the relative risks for three of these studies exceeded 3.
Despite the strength and consistency of these findings,
the relative effect of structural and functional aspects of
social supports has yet to be delineated.
Modification of psychosocial factors
The main implications of these findings for clinical
practice are summarised in the box. A recent
meta-analysis found that psychosocial interventions
are associated with improved survival after myocardial
infarction.28 However, two recent large randomised
controlled trials of psychological rehabilitation after
myocardial infarction found no difference in anxiety
and depression, and this may in part explain the lack of
effect on mortality.29 30 Randomised controlled trials of
modification of social supports after myocardial infarc-
tion show a decrease in cardiac death or reinfarction
rates.31 A patient’s social circumstances should be
elicited as part of the history, and the doctor may have
a role in mobilising social support. A multicentre trial
of 3000 patients after myocardial infarction
(ENRICHD
—
enhancing recovery in coronary heart
disease) is currently under way in the United States. It
will target patients at high psychosocial risk (those who
are depressed or socially isolated) and enrol large
numbers of women and ethnic minorities.
The potential for primary prevention in relation to
psychosocial factors lies largely outside the remit of cli-
nicians. Psychosocial factors themselves are deter-
mined largely by social, political, and economic factors
and it is therefore policy makers who influence the
structure and function of communities
—
in the public
and private domains
—
who may have scope for primary
prevention.
Conclusion
Of the large number of psychosocial factors that have
been studied, only four met the quality filter: type
A/hostility, depression and anxiety, work characteris-
tics, and social supports. While this review cannot
discount the possibility of publication bias, the
prospective observational studies show aetiological
roles for social supports, depression and anxiety, and
work characteristics and prognostic roles for social
supports and depression. Further evidence of a causal
role is provided by human and other primate evidence
of biological and behavioural pathways mediating
these effects. However, conflicting data exist on
whether psychosocial interventions reduce mortality
after myocardial infarction. This systematic review
should be updated and expanded to include other
observational study designs and other endpoints (for
example, all cause mortality) in order to focus future
research and, ultimately, policy. In this expanding area,
future primary research might investigate the:
xInterrelationships between different psychosocial
factors
xEffect of change in and cumulative exposure to
psychosocial factors
xShort and long term effects thoughout the life
course
xDifferences by sex, ethnic group,and country
xBehavioural and biological mechanisms involved
xEffect of psychosocial factors on different clinical
and subclinical outcomes
xAppropriate primary and secondary preventive
measures.
The comments of Lisa Berkman, Ichiro Kawachi, Redford
Williams, and Mandy Feeney are gratefully acknowledged. To
keep this review complete and up to date, we would be grateful
to be informed of studies which we may have missed.
Funding: MM is supported by an MRC research professor-
ship; his group investigating pyschosocial factors and health has
been supported by the Agency for Health Care Policy and
Research (5 RO1 HS06516); the New England Medical Centre-
Division of Health Improvement; the National Heart Lung and
Blood Institute (2RO1 HL36310); National Institute on Aging
(RO1 AG13196-02); the John D and Catherine T MacArthur
Foundation Research Network on Successful Midlife Develop-
ment; the Institute for Work and Health, Ontario, Canada; the
Volvo Research Foundation, Sweden; Medical Research
Council; Health and Safety Executive; and British Heart
Foundation (RG/28).
Contributors: HH designed and conducted the review in
discussion with MM. HH wrote the first draft of the manuscript
and incorporated MM’s revisions.
1 Marmot MG. Improvement of social environment to improve health.
Lancet 1998;351:57-60.
2 Hecker MHL, Chesney MA, Black GW.Coronar y prone behaviour in the
Western Collaborative Group study. Psychosom Med 1988;50:153-64.
3 Dembroski TM, MacDougall JM, Costa PT Jr. Components of hostility as
predictors of sudden death and myocardial infarction in the multiple risk
factor intervention trial. Psychosom Med 1989;51:514-22.
Psychosocial components of secondary
prevention
Clinicians should consider:
•Detecting and treating depression
•Mobilising social support
•Using socioeconomic status and psychosocial factors
to risk stratify patients
Clinical review
1466 BMJ VOLUME 318 29 MAY 1999 www.bmj.com

4 Johnson JV, Stewart W, Hall EM, Fredlund P, Theorell T. Long-term
psychosocial work environment and cardiovascular mortality among
Swedish men. Am J Public Health 1996;86:325-31.
5 Kubzansky LD, Kawachi I, Weiss ST, Sparrow D. Anxiety and coronary
heart disease: a synthesis of epidemiological, psychological and
experimental evidence. Ann Behav Med 1998;20:47-58.
6 Hayward C. Psychiatric illness and cardiovascular disease risk. Epidemiol
Rev 1995;17:129-38.
7 Miller TQ, Smith TW, Turner CW, Guijarro ML, Hallet AJ. A
meta-analytic review of research on hostility and physical health. Psychol
Bull 1996;119:322-48.
8 Schnall PL, Landsbergis PA. Job strain and cardiovascular disease. Ann
Rev Public Health 1994;15:381-411.
9 Schneiderman N, Skyler JS. Insulin metabolism, sympathetic nervous
system regulation, and coronary heart disease prevention. In:
Orth-Gomer K, Schneiderman N, eds. Behavioural medicine approaches to
cardiovascular disease prevention. Mawah, NJ: Lawrence Erlbaum Associ-
ates, 1996.
10 Brunner EJ. Stress and the biology of inequality. BMJ 1997;314:1472-6.
11 Williams RB, Barefoot JC, Blumenthal JA, Helms MJ, Luecken L, Pieper
CF, et al. Psychosocial correlates of job strain in a sample of working
women. Arch Gen Psychiatry 1997;54:543-8.
12 Kuh D,Ben-Shlomo Y. A life course approach to chronic disease epidemiology.
New York: Oxford University Press, 1997.
13 Pieper C, Lacroix AZ,Karasek RA. The relation of psychosocial dimensions
of work with coronary heart disease risk factors: a meta-analysis of five
United States data bases. Am J Epidemiol 1989;129:483-94.
14 Kaplan GA, Keil JE. Socioeconomic factors and cardiovascular disease:
a review of the literature. Circulation 1993;88:1973-98.
15 Marmot M, Bosma H, Hemingway H, Brunner EJ, Stansfeld SA. Contri-
bution of job control and other risk factors to social variations in coron-
ary heart disease incidence. Lancet 1997;350:235-9.
16 Mittleman MA, Maclure M, Nachnani M, Sherwood JB, Muller JE.
Educational attainment, anger and the risk of triggering myocardial
infarction onset. The Determinants of Myocardial Infarction Onset Study
Investigators. Arch Intern Med 1997;157:769-75.
17 Nunes EV, Frank KA, KornfieldDS. Psychologic treatment for the type A
behaviour pattern and for coronary heart disease: a meta-analysis of the
literature. Psychosom Med 1987;48:159-73.
18 Haynes SG, FeinleibM, Kannel WB. The relationship of psychosocial fac-
tors to coronary heart disease in the Framingham study: 3. Eight year
incidence of coronary heart disease. Am J Epidemiol 1980;111:37-58.
19 Rosenman RH, Brand RJ, Sholtz RI, Friedman M. Multivariate prediction
of coronary heart disease during 8.5 year follow-up in Western Collabo-
rative Group Study.Am J Cardiol 1976;37:903-9.
20 Shekelle RB, Hulley SB, Neaton JD, Billings J, Borhani NO, Gerace TA, et
al. The MRFIT behavior pattern study.II. Type A behavior and incidence
of coronary heart disease. Am J Epidemiol 1985;122:559-70.
21 Johnston DW, Cook DG, Shaper AG. Type A behaviour and ischaemic
heart disease in middle-aged British men. BMJ 1987;295:86-9.
22 Hearn M, Murray DM, Luepker RB. Hostility, coronary hear t disease and
total mortality: a 33 year follow up study of university students. JBehav
Med 1989;12:105-21.
23 Wassertheil-Smoller S, Applegate WB, Berge K, Chang CJ, Davis BR,
Grimm R Jr, et al. Change in depression as a precursor of cardiovascular
events. Arch Intern Med 1996;156:553-61.
24 Bosma H, Marmot MG, Hemingway H, Nicholson A, Br unner EJ, Stans-
feld S. Low job control and risk of coronary heart disease in the White-
hall II (prospective cohort) study. BM J 1997;314:558-65.
25 Siegrist J, Peter R, Junge A, Cremer P, Seidel D. Low status control, high
effort at work and ischemic heart disease: prospective evidence from
blue-collar men. Soc Sci Med 1990;31:1127-1134.
26 Bosma H, Peter R, Siegrist J, Marmot MG. Alternative job stress models
and the risk of coronary heart disease: the effort-reward imbalance
model and the job strain model. Am J Public Health 1998;88:68-74.
27 Alloway R. The buffer theory of social support-a review of the literature.
Psychol Med 1987;17:91-108.
28 Linden W, Stossel C, Maurice J. Psychosocial interventions in patients
with coronary artery disease: a meta-analysis. Arch Inter n Med
1996;156:745-52.
29 Jones DA, West RR. Psychological rehabilitation after myocardial
infarction: multicentre randomised controlled trial. BMJ 1996;313:
1517-21.
30 Frasure Smith N, Lesperance F, Prince RH, Verrier P, Garber RA, Juneau
M, et al. Randomised trail of home-based psychosocial nursing interven-
tion for patients recovering from myocardial infarction. Lancet
1997;350:473-9.
31 Bucher HC. Social support and prognosis following first myocardial
infarction. J Gen Int Med 1994;9:409-17.
Studying for the MRCP—was it worth while?
When in 1996 I decided to do part of my further vocational
training in Britain, I also decided to attempt the membership
examination of the Royal College of Physicians. I knew it was said
to be difficult,but as I was advanced in my training and always
had passed examinations easily, I thought of it as a challenge.
I started studying and soon was annoyed by the multiple choice
questions. So many of them were irrelevant to clinical medicine.
Does it really matter whether I know all the rare associations of
rare diseases? In a preparation course for part 1 I learnt that
questions starting “there is a recognised association” almost
always are true while questions containing the words “always” or
“never” almost always are wrong. This knowledge helped me to
pass the MRCP1
—
it did not enable me to treat my patients any
better.
Part 2 had the same preoccupation with rare diseases. To pass
the MRCP, it is more important to know about secondary
hypertension caused by endocrine disorders than to know about
idiopathic hypertension. You are more likely to encounter a
question about pseudopseudohypoparathyroidism than about
osteoporosis. Did you ever meet a patient with
pseudopseudohypoparathyroidism in real life? Simply to learn
how to pronounce this word took me hours of hard work. In the
MRCP, a patient with neuropathy will have porphyria or lead
intoxication in real life the diagnosis will be diabetes or
alcoholism. In Germany I have done close to 1000
echocardiography studies, but I never saw an atrial myxoma until
I started learning for the MRCP.
The training of an examination routine for the short cases had
some advantages, but once again, the emphasis was placed on
details of little meaning. The examination routine taught in
Germany is slightly different from the one taught for the MRCP.
I had to readjust to the British routine, not because it was better
but because it was what the examiners were looking for. By the
way, I never saw one of my consultants perform the complete
MRCP routine while examining a patient in outpatients or during
a ward round.
I tried part 2 three times and failed. I then returned to
Germany where I passed the German examination and now hold
the German equivalent of the certificate of completion of
specialist training.
Has all the studying been worth while? I learnt hundreds of
small facts I have already forgotten again
—
if I ever need them I
know where to look them up; I got proficient at a routine I have
shed again; and whenever I notice my consultant’s vitiligo I tell
myself he is perfectly healthy and unlikely to develop an
autoimmune disease. This is real life, after all.
“Non scholae sed vitae discimus”
—
we learn not for school, but
for life
—
was written above the main entrance of my school. Much
of what I learnt in those two years in Britain was for the MRCP
and not for practice. Instead of studying most of my time off duty,
I could have gone out with friends, visited famous places, enjoyed
walking in the Midlands. I would not have worked so hard if I had
not wanted to pass the MRCP. I failed, but yes, I would do it again.
And I will come back to Britain in my holidays to enjoy all those
things I missed because of my studies.
Waltraud Finckh, specialising in internal medicine,Bad Kissingen,
Germany
We welcome articles up to 600 words on topics such as
A memorable patient, A paper that changed my practice,My most
unfortunate mistake, or any other piece conveying instruction,
pathos, or humour. If possible the article should be supplied on a
disk. Permission is needed from the patient or a relative if an
identifiable patient is referred to.We also welcome contributions
for “Endpieces,”consisting of quotations of up to 80 words (but
most are considerably shorter) from any source, ancient or
modern, which have appealed to the reader.
Clinical review
1467BMJ VOLUME 318 29 MAY 1999 www.bmj.com