© International Epidemiological Association 2001
International Journal of Epidemiology 2001;30:248–255
Printed in Great Britain
Socioeconomic status and ischaemic heart
disease mortality in middle-aged men:
importance of the duration of follow-up.
The Copenhagen Male Study
Poul Suadicani,a Hans Ole Heina,b and Finn Gyntelberga
Objectives
The predictive value of some risk factors may diminish with increasing duration of
follow-up. This study was performed to elucidate the role of socioeconomic status
as a risk factor for ischaemic heart disease (IHD) mortality in middle-aged men,
testing the hypothesis that the role of mediators of the association of socioeconomic
status with risk of IHD would diminish with increasing length of follow-up.
Methods
A cohort of 5249 men aged 40–59 was established in 1971. Baseline data on
social class and other confounder variables were collected, and the cohort was
followed through registers for 8, 15, and 22 years. In all, 5028 without a history
of myocardial infarction or angina pectoris were included in the follow-up. Four
factors associated with either occupation or lifestyle were strong mediators of
the association found between social class and risk of fatal IHD, and were more
common in the lower social classes (classes IV and V): occasional demand for
vigorous activity at work, low leisure time physical activity level, high alcohol
consumption, and smoking.
Results
After the first 8 years, 78 men had died due to IHD, after 15 years: 222, and after
22 years: 411. Compared with social classes I, II, and III, the age-adjusted relative
risk (RR) with 95% CI for classes IV and V was 1.69, P , 0.05 after the first
8 years; adjusted for the above potential risk factors the RR dropped to 1.09,
P = NS. Corresponding RR after 15 years were 1.67, P , 0.001 and 1.33, P = NS;
and after 22 years, 1.59, P , 0.001 and 1.36, P , 0.05.
Conclusions Risk factors with an uneven social distribution related to occupation and lifestyle
were strong mediators of the association of socioeconomic status with risk of IHD.
A quite strong explanatory potential persisted but diminished with length of
follow-up.
Keywords
Myocardial ischaemia, social class, epidemiology, confounding factors, work, risk
factors
Accepted
28 June 2000
During the preceding three to four decades the association of
socioeconomic status with risk of ischaemic heart disease (IHD)
has changed in Western, industrialized countries; from being a
a Epidemiological Research Unit, Clinic of Occupational and Environmental
Medicine, H:S Copenhagen University Hospital, Bispebjerg, Bispebjerg
Bakke 23, DK-2400, Copenhagen, Denmark.
b The Glostrup Population Studies, Department of Internal Medicine C,
Glostrup Hospital, University of Copenhagen, Denmark.
Correspondence: Poul Suadicani, Epidemiological Research Unit, Clinic of
Occupational and Environmental Medicine, H:S Copenhagen University
Hospital, Bispebjerg, Bispebjerg Bakke 23, DK-2400, Copenhagen. E-mail:
poul.suadicani@jubiipost.dk
disease more common among affluent people, IHD has become
a disease of the poor. Several studies have attempted to disclose
relevant confounders, i.e. mediators, responsible for these differences in risk. Typically, it has been possible to explain less than
50% of the inequalities through an uneven social distribution of
cardiovascular risk factors.1–8
In the Copenhagen Male Study we have previously shown
that among men with a mean age of 63 years (range 54–75),
low socioeconomic status was associated with a significantly
increased risk of contracting a first IHD event; in a 6-year
follow-up study using a baseline established in 1985–1986, men
in social classes IV and V (basically unskilled and semiskilled
248
THE COPENHAGEN MALE STUDY
workers) had a 44% increased risk compared to men in social
classes I, II and III (men with an intermediate or high educational level). Approximately 70% of the excess risk could be
attributed to an uneven social distribution of risk factors, in
particular factors associated with lifestyle and physical work
environment. Social differences in fasting serum lipids, blood
pressure, and body mass index (weight in kg/height in m2)
contributed only little to the explanation.9
The predictive value of some risk factors may diminish with
increasing duration of follow-up, a suggestion supported in a
recent study by Wannamethee et al.10 A number of explanations
for this may exist, e.g. measurements, either of the risk factor
studied or of the mediating factors, become outdated because
they were taken a long time ago. Accordingly, the length of
follow-up may in itself be an important potential bias or effect
modifier in studies attempting to identify relevant mediators of
socioeconomic differentials in risk.
In addition to including different risk factors, the above cited
studies1–8 also had different lengths of follow-up which may
also explain in part differences between the studies. This study
was performed to elucidate the role of socioeconomic status as
a risk factor for IHD mortality in middle-aged men. It uses the
initial study baseline in the Copenhagen Male Study which was
established in 1970–1971 and tests the hypothesis that the role
of mediators of the association of socioeconomic status with risk
of IHD would diminish with increasing length of follow-up.
Material and Methods
The 1970–1971 baseline
The Copenhagen Male Study was established in 1970–1971. At
14 companies in Copenhagen, covering the railway, public road
construction, military, post, telephone, customs, national bank,
and the medical industry, all men aged 40–59 years were invited
to take part in a study of fitness and risk of cardiovascular
disease; 5249 men, 87% of potential participants, agreed to
participate.
The examination consisted of a questionnaire, a short interview, and a clinical examination with measurement of systolic
and diastolic blood pressure, and height and weight. From the
questionnaire information was obtained about working conditions, lifestyle and general health factors. For diagnosis of angina
pectoris and myocardial infarction the Rose Questionnaire was
used.11 The information given in the questionnaire was clarified
with each subject in the ensuing interview. Details on the
questionnaire have already been published.12 A number of
these factors are elaborated in more detail below.
Smoking
The men were asked if they smoked currently, previously or
had never smoked. In addition, smokers gave information on
the number of cigarettes, cheroots, cigars and number of pipes
smoked a day. An estimated weekly amount of tobacco smoked
was calculated. One cigarette was regarded as equivalent to 1 g
of tobacco, one cheroot to 3 g of tobacco, one cigar to 4 g and a
pipe to 2 g of tobacco.
Alcohol
Participants reported their daily alcohol consumption as the
number of alcoholic beverages consumed per day. An estimated
249
weekly amount consumed was calculated. One beverage corresponded to 10–12 g of ethanol, and most of the alcohol consumed came from beer.
Physical activity in leisure time
The men classified themselves into one of four different groups
of leisure time physical activity as: predominantly sedentary,
slightly active, fairly active or very active. For analytical purposes
answers were dichotomized into the least active (predominantly
sedentary) versus the rest.
Physical activity at work
Physical activity at work was assessed from the questionnaire.
To a question phrased ‘Do you do heavy physical work?’, 4.9%
of subjects answered ‘often’, 27.2% ‘occasionally’ and 67.9%
‘rarely or never’.
Social class
The men were divided into five social classes according to a
system originally elaborated by Svalastoga,13 later adjusted by
Hansen.14 This system of classification is based on education
level, and job position in terms of number of subordinates.
Typical jobs in the study cohort were, in social class I: officer,
civil engineer, office executive, head of department; social class
II: head clerk, engineer, unqualified architect; social class III:
engine driver, train guard; social class IV: machine fitter in a
telephone company; social class V: unskilled labourer, mechanic, driver. For the analyses, study participants were divided
into two classes, social classes IV and V (basically unskilled and
semi-skilled workers), and social classes I, II and III, i.e. men
with an intermediate or high educational level. This dichotomy
made it possible to compare the present results with those from
previous work in the Copenhagen Male Study.9
Exclusion criteria
Men with a history of angina pectoris or myocardial infarction
were excluded for this prospective study together with eight
men with missing job title. A total of 5028 men were included
for study.
End-points
In 1995 a register follow-up on mortality between 1971 and
1993 was carried out. All men who had taken part in the study
were traced by means of the Danish Central Population
Register. Death certificate diagnoses within the follow-up period
were obtained from the National Health Service register and
from the Danish Institute of Clinical Epidemiology. Both registers
used International Classification of Diseases, 8th Revision, and
IHD diagnoses used were codes 410–12. To test the importance
of duration of follow-up, the association of socioeconomic
status with risk of IHD mortality was tested in three different
time periods: 8, 15, and 22 years.
Statistical analyses
All analyses were performed using the SPSS statistical software
for Windows.15,16 Student’s t-test and χ2 analyses were used for
comparing unadjusted characteristics of men who died due to
IHD during the first 8 years of follow-up with men who did not,
and for comparing men in lower social classes with men
in higher social classes (Tables 1 and 2). Also in Tables 1 and 2
250
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 1 Characteristics of men who suffered a fatal ischaemic heart diseasae (IHD) event during the first 8 years of follow-up and characteristics
of others. Values presented are mean (SD) or frequency in per cent together with the age-adjusted odds ratio (OR) with 95% CI
Fatal IHD
(n = 78)
Others
(n = 4950)
Age-adjusted OR
(95% CI)
67.9%
54.7%
1.69 (1.05–2.7)*
15.4 (17.2)
11.2 (12.6)
1.02 (1.01–1.04)**
9.1%
3.0%
3.4 (1.5–7.6)**
85.9%
71.8%
2.3 (1.2–4.4)**
14.7 (8.6)
11.5 (9.7)
1.04 (1.01–1.06)**
5.2%
9.0%
0.6 (0.2–1.6)
27.8%
16.9%
1.7 (1.0–2.8)*
5.1%
6.6%
0.8 (0.3–2.2)
Social class IV or V
Lifestyle factors
Alcoholic beverages/week
Alcoholic beverages . 5/week
Smoking
Grams of tobacco smoked/day
Never smoking
Low leisure time physical activity
Frequently under psychological strain in leisure time
Occupational factors
Demand for vigorous physical activity at work
Often
2.6%
4.8%
0.5 (0.1–2.1)
Occasionally
43.6%
27.0%
2.1 (1.4–3.4)***
Seldom/never
53.8%
68.2%
0.5 (0.3–0.8)**
16.7%
21.9%
0.7 (0.4–1.4)
0.98 (0.95–1.02)(P = 0.07)
Frequently under psychological strain at work
Clinical parameters
Height, cm
173.2 (6.4)
174.6 (6.5)
Weight, kg
79.6 (10.9)
77.1 (10.7)
1.02 (1.00–1.04)**
Systolic blood pressure, mmHg
148.9 (24.4)
134.7 (19.0)
1.02 (1.01–1.03)***
Diastolic blood pressure, mmHg
88.0 (12.9)
83.0 (11.5)
1.02 (1.01–1.04)***
Hypertension treatment
6.4%
1.5%
3.2 (1.2–8.2)*
Non-insulin dependent diabetes
1.3%
0.8%
1.2 (0.2–9.3)
11.5%
5.3%
2.0 (1.0–4.1)*
Health parameters
Frequent use of sedatives
Other characteristics
Car ownership
Age, years
62.8%
69.0%
0.87 (0.54–1.39)
51.4 (5.3)
48.0 (5.2)
1.13 (1.08–1.17)***
* P , 0.05, ** P , 0.01, *** P , 0.001.
age-adjusted odds ratios with 95% CI are given based on the
results of multiple logistic regression analyses using maximum
likelihood estimation and backward elimination of variables.
To estimate the explanatory potential, i.e. their role as mediators,
of the above identified potential confounders, they were
included in three different analyses using 8, 15, and 22 years of
follow-up. Relative risk (RR) comparing social classes IV and V
with classes I, II and III, and RR of risk factors significant after
adjustment in a backward elimination model were estimated
by taking the natural log e raised to the hazard coefficient for
the variable of interest in Cox proportional hazards regression
models using the maximum likelihood ratio method17 and
backward elimination of variables (Table 3). The RR associated
with low social class was calculated from the hazards regression
coefficient and the standard error of the dichotomous social class
variable at the time of its removal from the stepwise model. The
explanatory role of each factor in the final model was analysed
by calculating how large a proportion of the ‘explained’ excess
risk in social classes IV and V could be attributed to each factor.
Variables with P-values ,0.10 were allowed to remain in the
final multivariable models (Tables 3 to 5).
To examine the predictive value of socioeconomic status among
men with and without occasional demand for heavy physical
activity at work, two separate analyses were performed using
Cox proportional hazards regression with statistical criteria similar
to those described above (Table 4). To obtain a sufficient number
of end-points in these stratified analyses a 10-year follow-up
period was used. All potential risk factors presented in Table 1
are included in these subsample analyses.
Finally, a Cox proportional hazards regression model including main effects and five interaction terms was performed to test
if the predictive value of smoking, weight, non-insulin dependent diabetes (NIDDM), high alcohol intake, and frequent use of
sedatives, was modified by occupational physical activity category
(Table 5). The rationale of the use of interaction terms has been
described by Kleinbaum.18
For all analyses a two-sided probability-value of P , 0.05 was
taken as statistically significant.
Ethics
The study was approved by The Ethics Committee for Medical
Research in the County of Copenhagen.
Results
Baseline characteristics
Table 1 shows characteristics of men who died due to IHD during
the first 8 years of follow-up and men who did not. Lifestyle
factors, occupational factors, and other factors separated the
THE COPENHAGEN MALE STUDY
251
Table 2 Characteristics of men in social classes IV and V and of men in social classes I, II, and III. Values presented are mean (SD) or frequency in
per cent together with the age-adjusted odds ratio (OR) with 95% CI
Social class IV/V
(n = 2763)
Social class I, II, III
(n = 2265)
Age-adjusted OR
(95% CI)
13.1 (14.1)
9.0 (10.1)
1.03 (1.02–1.035)***
4.5%
1.3%
3.5 (2.4–5.3)***
75.3%
68.0%
1.4 (1.3–1.6)***
12.0 (9.5)
11.0 (9.9)
1.01 (1.00–1.02)***
7.9%
10.3%
0.7 (0.6–0.9)**
19.4%
14.4%
1.4 (1.2–1.7)***
5.8%
8.0%
0.7 (0.6–0.9)**
15.5 (8.8–27.2)***
Lifestyle factors
Alcoholic beverages/week
Alcoholic beverages .35/week
Smoking
Grams of tobacco smoked/day
Never smoking
Low leisure time physical activity
Frequently under psychological strain in leisure time
Occupational factors
Demand for vigorous physical activity at work
Often
8.2%
0.6%
Occasionally
44.3%
6.5%
11.6 (9.6–13.9)***
Seldom/never
47.5%
93.0%
0.07 (0.06–0.08)***
14.3%
31.1%
0.37 (0.32–0.42)***
Frequently under psychological strain at work
Clinical parameters
Height, cm
173.3 (6.4)
176.1 (6.3)
0.93 (0.92–0.94)***
Weight, kg
77.3 (10.5)
77.1 (10.2)
1.00 (0.99–1.01)
Systolic blood pressure, mmHg
135.2 (19.5)
134.4 (18.7)
1.00 (0.99–1.01)
Diastolic blood pressure, mmHg
82.9 (12.0)
83.2 (11.5)
1.00 (0.99–1.01)
Hypertension treatment
1.7%
1.5%
1.1 (0.7–1.8)
Non-insulin dependent diabetes
0.8%
0.8%
0.94 (0.51–1.75)
Frequent use of sedatives
5.0%
5.9%
0.8 (0.7–1.1)
61.3%
78.3%
0.44 (0.39–0.50)***
48.1 (5.4)
47.9 (5.2)
1.01 (0.997–1.018)
Health parameters
Other characteristics
Car ownership
Age, years
** P , 0.01, *** P , 0.001.
Table 3 Relative risk (RR) of ischaemic heart disease (IHD) comparing social classes IV and V with classes I, II, and III using three different
lengths of follow-up 8, 15, and 22 years. Relative risk is presented with and without adjustment for confounders
Class IV/V versus I,II,III
Age-adjusted RR
Adjusted RR
8-year follow-up (n = 78)
15-year follow-up (n = 222)
22-year follow-up (n = 411)
1.69 (1.05–2.7)*
1.09 (0.6–1.9)
1.67 (1.3–2.2)***
1.33 (0.96–1.8)
1.59 (1.3–2.0)***
1.36 (1.09–1.77)**
RR
E.F.a
RR
E.F.a
RR
E.F.a
Occasional demand for heavy physical activity versus not
1.9 (1.1–3.2)
47%
1.4 (1.03–2.0)*
Low leisure-time physical activity versus more
1.6 (0.9–2.7)
15%
1.3 (0.96–1.9)
34%
1.3 (0.99–1.6)
23%
6%
1.4 (1.1–1.9)**
2.9 (1.3–6.5)**
10%
5%
1.4 (0.7–2.7)
2%
1.3 (0.8–2.2)
3.5%
1.13 (1.8–1.19)*
2.1 (1.1–4.2)*
9%
1.10 (1.07–1.13)***
6%
1.09 (1.07–1.1)***
5.5%
6%
1.6 (1.1–2.3)**
3%
1.5 (1.2–2.0)**
Predictors in the final model
Alcohol consumption .35 beverages/week versus less
Age (risk increase associated with a 10-year increase
in age)
Current smoking versus not
Unexplained
13%
49%
2%
61%
a E.F. = explanatory fraction, i.e. the relative proportion in per cent of the ‘explained’ excess risk in social classes IV and V which can be attributed to each
factor.
* P , 0.05, ** P , 0.01, *** P , 0.001.
two groups. Compared to others, among those who died, a larger
proportion belonged to the lower social classes, drank more
alcohol, smoked more tobacco, and were less physically active
in their leisure time. Conspicuous and highly significant associations were found for the phenomenon ‘occasional demand
for vigorous physical activity at work’ and for systolic as well as
diastolic blood pressure. Also, those who died due to IHD comprised a larger proportion receiving antihypertensive medication,
and even those using sedatives on a regular basis. On average
those who died were 3 years older than controls.
252
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 4 Predictors of fatal ischaemic heart disease (IHD) during the first 10 years of follow-up among men with and without occasional demand
for vigorous physical activity on the job. Risk factors are presented according to their relative strength of association with fatal IHD after
multivariable adjustment. Relative risk (RR) with 95% CI is estimated from the hazard ratios of Cox proportional hazards regression models
No. fatal
IHD
Occasional demand for heavy physical activity at work (n = 1369)
RR (95% CI)
49
Age (risk increase associated with a one-year increase in age)
1.14 (1.07–1.22)***
Systolic blood pressure (risk increase associated with a 10-mmHg increase)
1.29 (1.16–1.43)***
Current smoking versus not
6.2 (1.5–26.1)**
Weight (risk associated with a one-unit change in weight in kg)
0.97 (0.94–0.998)*
Others (n = 3659)
70
Systolic blood pressure (risk increase associated with a 10-mmHg increase)
1.21 (1.10–1.31)***
Age (risk increase associated with a one-year increase in age)
1.09 (1.04–1.14)***
Alcoholic beverages >35/week versus less
3.8 (1.6–9.3)**
NIDDMa versus not
5.0 (1.3–19.5)*
Frequent use of sedatives versus not
2.1 (0.96–4.6)
a Non-insulin dependent diabetes.
Covariates included in both analyses: social class, age, high alcohol consumption, smoking, leisure time physical activity, NIDDM, systolic blood pressure,
diastolic blood pressure, weight, height, frequent use of sedatives, car ownership, antihypertensive medication, work related stress, perceived stress in leisure
time.
* P , 0.05, ** P , 0.01, *** P , 0.001.
Table 5 Predictors of fatal ischaemic heart disease (IHD) (n = 119) during the first 10 years of follow-up among all men. Risk factors are
presented according to their relative strength of association with fatal IHD after multivariable adjustment. Relative risk (RR) with 95% CI
estimated from the hazard ratios of Cox proportional hazards regression models
Predictors
RR (95% CI)
Systolic blood pressure (risk associated with a 10-mmHg increase)
1.24 (1.16–1.32)***
Age (risk increase associated with a one-year increase in age)
1.11 (1.06–1.15)***
NIDDMa versus not
3.8 (1.4–10.5)**
Interaction term: occasional demands of job-related heavy physical activity * smoking
Alcoholic beverages .35/week versus less
5.7 (1.3–23.9)*
2.4 (1.1–5.1)*
Interaction term: occasional demand for job-related heavy physical activity * weight (lowest fifth, <69 kg, versus rest)
Frequent use of sedatives versus not
2.0 (1.01–4.1)*
1.8 (0.94–3.5)
a Non-insulin dependent diabetes.
Covariates included in the analysis:
Main effects: social class, age, high alcohol consumption, smoking, leisure time physical activity, NIDDM, systolic blood pressure, diastolic blood pressure,
weight, height, frequent use of sedatives, car ownership, antihypertensive medication, work-related stress, perceived stress in leisure time, frequent demand
for heavy physical activity on the job, occasional demands of heavy physical activity on the job.
Interaction terms:
(1) Occasional demand for heavy physical activity on the job * smoking
(2) Occasional demand for heavy physical activity on the job * weight
(3) Occasional demand for heavy physical activity on the job * NIDDM
(4) Occasional demand for heavy physical activity on the job * high alcohol consumption
(5) Occasional demand for heavy physical activity on the job * frequent use of sedatives
* P , 0.05, ** P , 0.01, *** P , 0.001.
Table 2 illustrates that several of the factors separating those
who died due to IHD from those who did not, also significantly
separated men in the lower social classes from men in classes I,
II, and III, with a clustering of adverse factors in the lower
classes: they drank more, they smoked more, they were less
active in their leisure time, and demand for physical activity at
work was very different between high and low social classes.
More than half of the men in the lower classes reported having
frequent or occasional demand for vigorous activity at work
compared to only 7% among classes I, II and III.
Identification of potential mediators of the
association of socioeconomic status with risk
of fatal IHD
Based on the results presented in Tables 1 and 2 potential
confounders or mediators were identified. Four factors were
significantly associated with fatal IHD as well as socioeconomic
status on the 0.05 level, and were regarded as relevant: low
leisure time physical activity, smoking, occasional demand for
vigorous physical activity at work, and alcohol consumption.
In preliminary analyses we found that there was no association
THE COPENHAGEN MALE STUDY
between alcohol consumption and risk of fatal IHD during the
first 8 years of follow-up among men consuming ,35 beverages/
week; accordingly, the dichotomous variable alcohol consumption .35 beverages/week versus less was considered the more
relevant and was used for further analyses. The dichotomous
variable smoking and smoking in g/day were highly correlated.
Which of the two was used had no influence on the results
presented. For simplicity of presentation we used the dichotomous variable in the multivariable analyses.
Prospective analyses
Table 3 illustrates the association of social class with risk of
fatal IHD in 8, 15 and 22 years of follow-up, including the
four factors mentioned above together with age. The joint
explanatory potential of these factors was very strong in the
8-year follow-up, reducing the age-adjusted only RR from 1.69
to 1.09. An additional analysis including all factors presented in
Table 1 did not further reduce the strength of the association of
social class with IHD (not shown).
The relative contribution of each factor included in the final
model is depicted in the column denoted E.F. for explanatory
fraction. With increasing length of follow-up the role of the
factors included in the analyses diminished. However, almost
half of the excess risk was ‘explained’ in the 15-year follow-up
which included almost three times as many end-points, and
almost 40% could be attributed to an uneven social distribution
of baseline risk factors in the 22-year follow-up which had more
than five times the number of endpoints included in the 8-year
incidence study.
Table 4 shows which factors were associated with risk of fatal
IHD during 10 years of follow-up among men with and without
occasional demand for heavy physical activity at work. Within
both groups there was a strong predictive value of blood pressure and age. However, only among men with occasional demand
for heavy physical activity was smoking highly significantly
associated with IHD mortality, and also a significant inverse
association was found between weight and risk of IHD. Among
the large group of men without these demands, the association
between smoking and risk of fatal IHD was quite weak and far
from reaching statistical significance (RR = 1.06, P = 0.85), at
the time of removal from the multivariable model. In contrast,
only among men without occasionally physically demanding
work were a high alcohol consumption level, NIDDM, and
frequent use of sedatives, significantly associated with risk of
fatal IHD.
In order to test the relevance of these differences in predictors
between men with and without work-related occasional demand
for heavy physical activity, a final analysis was performed
including all men. Five multiplicative interaction terms were
added to the model together with all main effects included in
the analyses presented in Table 4. The association with IHD risk
was equally strong whether weight was included as a continuous or a dichotomous variable. To simplify interpretation of
the model, for the analysis presented, the variable weight was
dichotomized (arbitrarily using the lowest quintile, 69 kg, as
cut-off point).
Table 5 shows the result of the final model including only
factors significantly associated with fatal IHD after multivariable
adjustment. Systolic blood pressure and age were the strongest
risk factors, followed by the interaction of occasional demand
253
for heavy physical work and smoking, NIDDM, high alcohol
consumption, and the interaction of occasional demand for
heavy physical work and weight. In this 10-year follow-up RR
associated with low social class was 1.60 before adjustment, and
1.17 after adjustment.
Discussion
Two measures of socioeconomic status were analysed in this
study—hierarchical position and a measure of relative material
wealth—car ownership. Since car ownership was neither
an independent predictor of fatal IHD nor a mediator of the
association of social class with risk of fatal IHD, the analyses in
this study were mainly based on the hierarchical classification.
The Copenhagen Male Study cohort consists of a majority
of men employed in public service. In the Danish public
services salary and hierarchical status are closely related, and
also the relative status differences remain fairly stable with
time.
Irrespective of the length of follow-up, the lower social classes
had a clearly increased risk of fatal IHD. The excess RR during
the first 8 years of follow-up, 69%, was more than 50% higher
than the 44% risk increase reported in our study of the cohort
when it had a mean age of 63 years and was followed for
6 years.9 The results of the present study support the hypothesis
that the role of mediators of the association of socioeconomic
status with risk of IHD mortality diminish with increasing length
of follow-up. The excess RR in the lower classes remained quite
high with a 59% excess risk after 22 years of follow-up. This
observation may indicate that either the lower classes were less
likely to change their lifestyle in a positive direction (e.g to
reduce or quit smoking, and maintain a reasonable degree of
leisure time physical activity), or that other risk factors with
an uneven socioeconomic distribution were responsible for the
excess risk in the later part of the follow-up period. This latter
suggestion is supported by our previous work identifying
mediators of socioeconomic inequalities when the cohort was
15 years older and using the 1985–1986 baseline. For instance,
social differences in the wine drinking pattern of low and high
social classes with a much larger proportion of wine drinkers in
classes I,II, and III than in classes IV and V contributed strongly
to explaining differences in the incidence of IHD.9
Among employed men aged 40–59 years, four factors related
to either occupation or lifestyle were strong mediators of
the association of socioeconomic status with risk of fatal IHD.
The role of each factor is addressed in order of importance as
mediators of the inverse association of social class with IHD.
The strongest mediating factor of social inequalities in risk of
IHD mortality was the issue ‘occasional demand for vigorous
activity at work’, the main effect of which explained approximately half of the excess risk in classes IV and V in the 8-year
follow-up. Considering the established association of a low
physical activity level with risk of IHD19 the observation was
somewhat surprising. This finding may have several implications.
It is consistent with the observation that the association of
socioeconomic status with risk of IHD was reversed in Western,
industrialized countries after World War II.20 From being a
disease of affluent people, IHD became a disease more dominant
among those socioeconomically underprivileged. An increasing
254
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
proportion of various forms of labour previously calling for
physical activity has been taken over by machines, and regular
demand for physically demanding work in the lower, manual
classes, has become rarer. Thus, nowadays a relatively larger
proportion of men in lower social classes have a low physical
training effect from their work. The results indicate that untrained men are at increased risk when they have to carry out
physically strenuous work for which they are not properly
trained. The observation is consistent with the results of a
recent paper on sudden cardiac death.21 The authors found that
‘undetected cardiac lesions cause unexpected sudden cardiac
death during occasional sport activity’. From the register followup data available to us it was not possible to obtain information
on how many men died suddenly while performing strenuous
tasks at work. In parallel with occasional strenuous sporting
activity, occasional strenuous work activity may have been
triggering a myocardial infarction event or lethal rhythm disturbances with manifest symptoms occurring during, as well as
after, work.
The above considerations were supported by a number of
further analyses. We analysed the predictive value of the factors
included in the present study of risk of fatal IHD in separate
analyses of those who reported having occasionally physically
demanding work, and among others (Table 4). One observation
was particularly noteworthy: only among those with occasional
demand for physically strenuous activity at work was smoking
significantly, and strongly, associated with risk of fatal IHD. This
finding, suggestive of a dangerous interplay of smoking and
physical demands, is interesting in the light of a recent in-depth
review of the pathology of the coronary arteries in smokers and
non-smokers considering the pathophysiological mechanism of
smoking.22 It concluded that ‘both clinical and experimental
data strongly support the notion that thrombogenetic factors
are responsible for the increased occurrence of ischaemic heart
disease, and particularly acute coronary syndromes, among
chronic smokers.’ Also in support of a possible causal association
between occasional demand for heavy physical work and
increased risk of fatal IHD was the observation that men with
a low weight belonging to the group occasionally exposed to
heavy work had an excess risk of fatal IHD. A weight load of a
given magnitude no doubt has a larger impact on the body of a
man of low weight than on a heavy man.
Low leisure time physical activity level and heavy alcohol
consumption were almost equally strong mediating factors;
as point estimates explaining 15% and 10% of the excess IHD
mortality in classes IV and V. The lower classes had an approximately 30% larger proportion of subjects who were largely
sedentary in their leisure time, and as a group they reported
consuming between 30% and 40% more alcohol per week.
Although the age difference between lower and higher classes
was small, differences in age explained 9% of the excess risk,
and differences in smoking habits explained 6%.
In previous work from the Whitehall Study, social class differentials expressed as employment grade and car ownership—a
marker of income and relative wealth—were independently
associated with total mortality and mortality from the major
cause groups.23 Although strongly associated with social class,
in the study presented here, car ownership was not associated
with risk of fatal IHD. Also height was included and scrutinized
in the Whitehall Study. Height differences between the socioeconomic groups explained little of the differential mortality, a
result in accordance with the results on IHD death presented
here.
Summing up, although the risk factors were not identical, in
essence the results of this study agree with our previous finding
—that a socioeconomically heterogeneous distribution of
occupational and lifestyle factors, and to a lesser extent personal
clinical features, are strong mediators of social inequalities in
risk of IHD.9 The results of this study also emphasize that the
predictive value of risk factors diminishes with increasing length
of follow-up, and that waiting for more end-points in follow-up
studies may be futile.
KEY MESSAGES
• The association of low social class with risk of fatal ischaemic heart disease is strongly mediated by occupational
and lifestyle factors in middle-aged men.
• With increasing duration of follow-up occupational and lifestyle factors explain less of the association of low social
class with risk of fatal ischaemic heart disease.
• Occasional occupational exposure to vigorous physical activity is strongly associated with risk of fatal IHD in men
who smoke and in men with a low weight.
References
1 Rose G, Marmot MG. Social class and coronary heart disease. Br Heart J
1981;45:13–19.
2 Holme I, Helgeland A, Hjermann I, Leren P. Socio-economic status
as a coronary risk factor: the Oslo Study. Acta Med Scand 1982;
660(Suppl.):147–51.
3 Marmot MG, Shipley MJ, Rose G. Inequalities in death—specific
explanations of a general pattern? Lancet 1984;i:1003–06.
4 Pearce NE, Davis PB, Smith AH, Foster FH. Social class, ethnic group,
and male mortality in New Zealand, 1974–8. J Epidemiol Community
Health 1985;39:9–14.
5 Pocock SJ, Shaper AG, Cook DG, Phillips AN, Walker M. Social class differ-
ences in ischaemic heart disease in British men. Lancet 1987; ii:197–201.
6 Rosengren A, Wedel H, Wilhelmsen L. Coronary heart disease and
mortality in middle aged men from different occupational classes in
Sweden. BMJ 1988;297:1497–500.
7 Pekkanen J, Tuomilehto J, Uutela A, Vartiainen E, Nissinen A. Social
class, health behaviour, and mortality among men and women in
eastern Finland. BMJ 1995;311:589–93.
8 Lynch JW, Kaplan GA, Cohen RD, Tuomilehto J, Salonen JT. Do
cardiovascular risk factors explain the relation between socioeconomic status, risk of all-cause mortality, cardiovascular mortality,
and acute myocardial infarction? Am J Epidemiol 1996;144:934–43.
THE COPENHAGEN MALE STUDY
255
9 Suadicani P, Hein HO, Gyntelberg F. Strong mediators of social
16 Norusis MJ, SPSS Inc. SPSS Advanced Statistics 6.1. Chicago: SPSS, 1994.
inequalities in risk of ischaemic heart disease: a six-year follow-up in
the Copenhagen Male Study. Int J Epidemiol 1997;26:516–22.
17 Cox DR. Regression models and life-tables. J R Stat Soc 1972;34:
10 Wannamethee SG, Shaper AG, Whincup PH, Walker M. Role of risk
18 Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiological Research.
factors for major coronary heart disease events with increasing length
of follow-up. Heart 1999;81:374–79.
Principles and Quantitative Methods. London: Lifetime Learning
Publications, 1982, Ch. 19.
11 Rose G, McCartney P, Reid DD. Self-administration of a questionnaire
19 Lee IM, Paffenbarger RS Jr, Hennekens CH. Physical activity, physical
on chest pain and intermittent claudication. Br J Prev Soc Med 1977;
31:42–48.
20 Thelle S. The influence of socioeconomic status on cardiovascular
12 Hein HO, Suadicani P, Gyntelberg F. Ischaemic heart disease incidence
by social class and form of smoking: the Copenhagen Male Study:
17 years’ follow up. J Intern Med 1992;231:477–83.
13 Svalastoga K. Prestige, Class and Mobility. Copenhagen: Munksgaard,
1959.
14 Hansen EJ. Socialgrupper i Danmark. Copenhagen: Socialforsknings-
instituttet, 1984.
15 Norusis MJ, SPSS Inc. SPSS for Windows: Base System Users Guide, Release
6.0. Chicago: SPSS, 1993.
187–220.
fitness and longevity. Aging (Milano) 1997;9:2–11.
diseases. G Ital Cardiol 1999;29:680–83.
21 Tabib A, Miras A, Taniere P, Loire R. Undetected cardiac lesions cause
unexpected sudden cardiac death during occasional sport activity.
A report of 80 cases. Eur Heart J 1999;20:900–03.
22 Bøttcher M, Falk M. Pathology of the coronary arteries in smokers
and non-smokers. J Cardiovasc Risk 1999;6:299–302.
23 Smith GD, Shipley MJ, Rose G. Magnitude and causes of socioeco-
nomic differentials in mortality: further evidence from the Whitehall
Study. J Epidemiol Community Health 1990;44:265–70.