Content uploaded by Roger Tavendale
Author content
All content in this area was uploaded by Roger Tavendale on Mar 31, 2015
Content may be subject to copyright.
General practice
Comparison of the prediction by 27 different factors of
coronary heart disease and death in men and women of
the Scottish heart health study: cohort study
Hugh Tunstall-Pedoe, Mark Woodward, Roger Tavendale, Richard A’ Brook, Mary K McCluskey
Abstract
Objective: To compare prediction by 27 different
factors in men and women of coronary heart disease
events, coronary deaths, and deaths from all causes.
Design: Cohort study.
Setting: Scottish population study.
Subjects: In 1984-7 random sampling of residents
aged 40-59 produced 11 629 men and women who
generated survey clinic questionnaires, examination
findings, and blood and urine specimens.
Main outcome measures: Subsequent death,
coronary artery surgery, and myocardial infarction.
Risks were calculated for each category of factor or
fifth of continuous variables. 27 factors were ranked
by descending age adjusted hazard ratio of the top to
bottom class in each factor, by sex and end point.
Results: Follow up averaged 7.6 years, during which
the 5754 men had 404 coronary events, 159 coronary
deaths, and 383 deaths and the 5875 women 177, 47,
and 208 respectively. The rankings for factors for the
three end points were mainly similar in men and
women, although hazard ratios were often higher in
women. Classical risk factors ranked better for
predicting coronary risk than newer ones. Yet strong
prediction of coronary risk was no guarantee of
significant prediction of all cause mortality. Findings
included an anomalous coronary protective role for
type A behaviour in women; raised plasma fibrinogen
as a strong predictor of all end points; and an
unexpectedly powerful protective relation of dietary
potassium to all cause mortality.
Conclusions: These initial unifactorial rankings and
comparisons must be interpreted with caution until
potential interaction, confounding, and problems of
measurement and causation are further explored.
Introduction
With few exceptions,
1-3
studies identifying risk factors
for coronary heart disease have focused on men,
4-9
with
their higher incidence rates,
10
yet women live longer
and lifetime risk is almost equal.
11
All cause mortality is
now a standard end point in intervention studies but
relatively neglected in studies of risk factors. There is a
need for studies which include standardised
measurement of lifestyle and coronary risk factors in
men and women and where all cause mortality is
reported alongside coronary end points.
The Scottish heart health study
12
began in 1984,
when Scotland was in the premier league for death
from coronary heart disease in both men and women.
It reported lifestyle and risk factor status for
representative samples of men and women across
Scotland
13 14
and showed how regional variations in
risk factors correlated with mortality from coronary
heart disease.
15
We always intended to see how well older and new
factors compared in predicting coronary risk,
12
and
after eight years’ follow up we now compare 27 factors
in the two sexes for three end points
—
major coronary
events (non-fatal myocardial infarction, death from
coronary heart disease, or coronary artery surgery),
deaths from coronary heart disease, and all deaths.
Methods
Recruitment—Twenty five districts of Scotland were
visited in two contrasting seasons in November 1984 to
October 1987. General practitioners were recruited
randomly and their patients enumerated in the eight
five year age-sex bands 40-59. A constant percentage
in each district band was selected by random
sampling.
12 16
In 23 districts the target total was 450
people from 10 general practitioners but the
Edinburgh and north Glasgow MONICA (monitoring
trends and determinants in cardiovascular disease)
population surveys
17
each included 30 general
practitioners and 800 participants aged 40-59. Joint
letters were sent out from survey and practice
enclosing appointments for local clinics, plus a 20 page
personal health record for self completion. This
included several classical cardiovascular question-
naires,
18
a food frequency questionnaire adapted from
Caerphilly,
19
and the Bortner questionnaire for type A
personality.
20
Survey clinic procedures—Clinics were run by survey
nurses. Participants reported without fasting and
progressed through three stations. Firstly, after remov-
ing shoes and outer clothing, they were weighed, had
their height measured, and gave informed consent,
including to follow up of medical records. The
questionnaire was checked. Blood pressure was
measured twice by random zero sphygmomanometer
Cardiovascular
Epidemiology Unit,
Ninewells Hospital
and Medical School,
Dundee DD1 9SY
Hugh
Tunstall-Pedoe,
professor and director
Mark Woodward,
senior lecturer in
statistical
epidemiology
Roger Tavendale,
biochemist
Richard A’ Brook,
research assistant
Mary K McCluskey,
research assistant
Correspondence to:
Professor
Tunstall-Pedoe
h.tunstallpedoe@
dundee.ac.uk
BMJ 1997;315:722–9
722 BMJ VOLUME 315 20 SEPTEMBER 1997
seated after five minutes’ rest. Station two recorded a 12
lead electrocardiogram and measured expired air
carbon monoxide. At station three venepuncture and a
subcutaneous fat biopsy were performed and a 2.5 litre
container for 24 hour urine collection was supplied.
12
Processing and quality control—Serum was separated
within two hours and chilled at 4°C while plasma was
separated immediately and stored at −20°C before
both were transferred within five days to Dundee.
Urinecollectionswereweighedandanalysedforelectro-
lytes and for creatinine. The latter and most serum
analyses were analysed in duplicate shortly after
reaching Dundee. Serum cotinine and adipose tissue
fatty acids were analysed subsequently by gas chroma-
tography. After storage at −40°C plasma specimens
were assayed in Glasgow by coagulometer for fibri-
nogen.
21
Survey and laboratory procedures including
lipid analyses were standardised on the World Health
Organisation’s MONICA protocol and its lipid
laboratory.
22
Electrocardiograms were coded by two
*Previous angina+01
Total cholesterol+02
HDL cholesterol–03
Diastolic blood pressure+04
Systolic blood pressure+05
Triglycerides+06
Serum cotinine+07
Plasma fibrinogen+08
*Smoke cigarettes+09
Body mass index+10
Ascorbic acid–11
Alcohol–12
Height–13
Urine potassium–14
*Housing tenure+15
*Leisure inactive+16
Carotenoid–17
Weight+18
*Work inactive+19
Uric acid+20
*Diabetes+21
Tocopherol–22
Urine sodium+23
Energy intake–24
Blood glucose+25
Bortner score+26
*Previous angina+01
Triglycerides+02
HDL cholesterol–03
Serum cotinine+04
Total cholesterol+05
Systolic blood pressure+06
*Diabetes+07
*Smoke cigarettes+08
Plasma fibrinogen+09
*Housing tenure+10
Bortner score–11
*Work inactive+12
Diastolic blood pressure+13
Ascorbic acid–14
Urine sodium+15
Height–16
Alcohol–17
Uric acid+18
*Leisure inactive+19
Urine potassium–20
*Menopause+21
Weight+22
Body mass index+23
Tocopherol–24
Energy intake+25
Blood glucose+26
Carotenoid–27
0
NS
NS
NS
NS
NS
NS
NS
NS
NL
NS
NS
NS
NL
NS
123456
012345678
All coronary heart disease: women
All coronary heart disease: men
NS = not significant, NL = not linear
Fig 1 Ranking of risk factors for all coronary events in men and
women by age adjusted hazard ratios between highest and lowest
category (fifths unless indicated). *Factors with fewer than five
classes
*Previous myocardial infarction+01
Plasma fibrinogen+02
Serum cotinine+03
*Smoke cigarettes+04
Urine potassium–05
*Diabetes+06
*Leisure inactive+07
*Housing tenure+08
*Work inactive+09
Systolic blood pressure+10
Urine sodium–11
Bortner score–12
Tocopherol–13
Diastolic blood pressure+14
Height–15
HDL cholesterol–16
Triglycerides+17
Ascorbic acid–18
Alcohol+19
Uric acid+20
Total cholesterol+21
Weight–22
Carotenoid–23
Energy intake+24
Body mass index+25
Blood glucose–26
*Previous myocardial infarction+01
*Work inactive+02
Serum cotinine+03
*Smoke cigarettes+04
Urine potassium–05
Plasma fibrinogen+06
*Leisure inactive+07
Systolic blood pressure+08
*Housing tenure+09
Uric acid+10
Bortner score–11
Triglycerides+12
*Diabetes+13
Diastolic blood pressure+14
HDL cholesterol–15
Ascorbic acid–16
Energy intake+17
*Menopause+18
Height–19
Urine sodium–20
Total cholesterol–21
Weight+22
Alcohol–23
Blood glucose+24
Tocopherol–25
Carotenoid–26
Body mass index–27
01
NS
NS
NS
NS
NS
NL
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NL
NL
NL
NL
234
0 12 3456
All deaths: women
All deaths: men
NS = not significant, NL = not linear
Fig 2 Ranking of risk factors for all cause mortality in men and
women by age adjusted hazard ratios between highest and lowest
category (fifths unless indicated). *Factors with fewer than five
classes
General practice
723BMJ VOLUME 315 20 SEPTEMBER 1997
observers with a third adjudicating discrepancies.
23
Questionnaires were part coded by nurses before
transfer to Dundee, where they were checked, double
coded, keyed, and verified on microcomputer. Data
underwent range and logic checks and were
assembled by individual anonymous code on the
mainframe. Personal identifiers were stored elsewhere.
Follow up—Participants were flagged on the Scottish
NHS Register, which forwarded copies of death certifi-
cates. The Scottish heart health study register of
participants, updated with information from health
boards, was run against a central file of hospital
discharge data by the information and statistics
division of the Scottish Common Services Agency,
generating a list of all hospital admissions from before
the study began to the end of 1993,
24
which was also
used as the cut point for mortality follow up.
End points—Case notes were requested for all
hospital episodes of myocardial infarction and other
emergency admissions for coronary heart disease.
These were extracted and coded according to
MONICA project criteria as definite, possible, or no
myocardial infarction, the first two categories being
included as end points.
10 22 25
Hospital diagnoses of
coronary artery surgery (coronary artery bypass grafts
or percutaneous transluminal coronary angioplasty)
were accepted without verification, as were out of
hospital coronary deaths, as further inquiry in the
Scottish MONICA project rarely changes the diag-
nosis.
25
We report three overlapping end points.
Participants progressing to myocardial infarction,
coronary artery surgery, or death from coronary
disease qualified for all coronary heart disease,
barring those few who had non-fatal recurrence of a
prerecruitment myocardial infarction. Two fatal end
points were deaths from coronary heart disease and
deaths from all causes.
Table 1 Numbers of people at risk and numbers experiencing a qualifying end point
event during follow up
Age group
40-44 45-49 50-54 55-59 All ages
Men:
All people at baseline 1398 1370 1521 1465 5754
All coronary heart disease* 66 68 106 164 404
Fatal coronary heart disease 21 15 45 78 159
Deaths from all causes 42 54 110 177 383
Women:
All persons at baseline 1503 1438 1521 1413 5875
All coronary heart disease* 17 32 43 85 177
Fatal coronary heart disease 0 5 13 29 47
Deaths from all causes 21 32 54 101 208
*Endpoint of all coronary heart disease includes coronary death, or coronary artery surgery, or non-fatal definite or possible myocardial
infarction but the latter only qualifies in the absence of a previous history of myocardial infarction.
Table 2 Age adjusted hazard ratios for two and three class categorical data
Class
Men Women
123
Missing
12 3
Missing
Multiplicative constant
(95% CI) P value
Multiplicative constant
(95% CI) P value
Housing tenure (owned/rented)
Tenure Own Rent Own Rent
% 53.5 46.5 19 51.4 48.6 25
All CHD 1 1.48 1.48 (1.21 to 1.80) *** 1 2.64 2.64 (1.89 to 3.68) ***
CHD deaths 1 1.56 1.56 (1.13 to 2.14) ** 1 2.42 2.42 (1.28 to 4.59) **
All deaths 1 1.55 1.55 (1.26 to 1.90) *** 1 2.12 2.12 (1.58 to 2.84) ***
Diabetes mellitus (no/yes)
Diabetes No Yes No Yes
% 98.4 1.6 65 98.5 1.5 87
All CHD 1 1.31 1.31 (0.67 to 2.53) NS 1 2.91 2.91 (1.43 to 5.92) **
CHD deaths 1 2.14 2.14 (0.95 to 4.85) BS 1 4.03 4.03 (1.25 to 12.99) *
All deaths 1 2.08 2.08 (1.22 to 3.55) ** 1 1.50 1.50 (0.62 to 3.66) NS
Menopausal status (no/yes)
Menopause No Yes
% 42.3 57.7 36
All CHD — — 1 1.48 1.48 (0.92 to 2.40) NS
CHD deaths — — 1 1.07 1.07 (0.36 to 3.19) NS
All deaths — — 1 1.35 1.35 (0.86 to 2.11) NS
Cigarette smoking (never/ex/current smoker)
Status Never Ex Current Never Ex Current
% 25.7 34.8 39.5 41 41.6 20.2 38.2 39
All CHD 1 1.38 2.02 1.43 (1.25 to 1.63) *** 1 1.16 2.70 1.69 (1.42 to 2.02) ***
CHD deaths 1 1.54 2.50 1.60 (1.28 to 1.98) *** 1 1.73 3.24 1.81 (1.27 to 2.57) ***
All deaths 1 1.54 2.34 1.53 (1.33 to 1.75) *** 1 1.56 2.80 1.68 (1.43 to 1.99) ***
Physical inactivity in work (active/average/inactive)
Status Active Average Inactive Active Average Inactive
% 42.4 44.4 13.2 62 47.4 48.1 4.5 25
All CHD 1 1.18 1.35 1.17 (1.01 to 1.34) * 1 1.13 2.45 1.35 (1.05 to 1.74) *
CHD deaths 1 1.57 1.84 1.39 (1.11 to 1.34) ** 1 0.82 3.79 1.44 (0.88 to 2.35) NS††
All deaths 1 1.32 1.55 1.26 (1.09 to 1.45) ** 1 1.10 3.50 1.54 (1.22 to 1.94) ***††
Leisure physical inactivity (active/average/inactive)
Status Active Average Inactive Active Average Inactive
% 22.8 59.0 18.2 27 19.2 62.1 18.7 23
All CHD 1 1.11 1.45 1.21 (1.03 to 1.41) * 1 1.26 1.51 1.23 (0.96 to 1.56) BS
CHD deaths 1 1.28 2.07 1.46 (1.14 to 1.87) ** 1 0.78 2.22 1.68 (1.04 to 2.70) *†
All deaths 1 1.08 1.66 1.31 (1.11 to 1.54) ** 1 0.99 2.20 1.61 (1.29 to 2.02) ***††
CHD=coronary heart disease.
Log linear significance testing: NS=not significant (P>0.10), BS=borderline (0.05<P < 0.1), *=significant (0.01<P< 0.05), **=highly significant (0.001<P<0.01), ***=very highly significant (P<0.001). Residual
non-linear significance testing uses † instead of *.
General practice
724 BMJ VOLUME 315 20 SEPTEMBER 1997
Statistical procedures
Cox’s proportional hazards model
26
allows for the differ-
ent follow up times from attendance at the initial clinic
until the end of 1993. Survival was counted to the first
qualifying event. Loss to follow up for coronary end
points occurred through death from a non-coronary
cause and for all end points through emigration.
For continuously distributed factors the population
was partitioned into fifths, up to and including the
20th, 40th, 60th, and 80th centiles. The tables specify
the quintile values, but also those of the 1st and 99th
centiles, rather than extreme readings. Values of zero
identified the lowest class for alcohol consumption and
for serum cotinine concentration; partitioning at and
below the 25th, 50th, and 75th centiles classified others
with positive values into the four remaining classes. For
categorical factors class sizes were uneven and their
number, 2-5, implicit.
For each factor, end point, and sex, after age adjust-
ment, the risk in the lowest class was set at unity and the
Table 3 Age adjusted hazard ratios for five class discontinuous factors
Class
Men Women
12345
Missing
12345
Missing
Multiplicative constant
(95% CI) P value
Multiplicative constant
(95% CI) P value
Previous coronary heart disease
None Rose+ ECG+ Angina MI None Rose+ ECG+ Angina MI
% 78.5 6.4 6.9 3.4 4.8 134 79.1 7.7 8.2F 3.4 1.6 135
All CHD 1 1.57 2.69 5.83 NA 1.75 (1.60 to 1.92) *** 1 1.75 2.67 7.31 NA 1.86 (1.63 to 2.12) ***
CHD deaths 1 1.90 3.783 5.41 8.04 1.70 (1.54 to 1.87) *** 1 1.88 5.44 3.77 25.32 2.10 (1.74 to 2.54) ***
All deaths 1 1.53 2.15 2.36 3.39 1.36 (1.27 to 1.46) *** 1 1.60 2.19 2.21 5.23 1.44 (1.29 to 1.60) ***
Bortner score (type A personality)
<134 135- 160- 179- >203 <138 139- 162- 178- >199
% 19.9 20.1 19.8 20.0 20.0 481 19.9 20.5 19.6 20.3 19.8 670
All CHD 1 1.08 0.87 0.78 1.04 0.98 (0.91 to 1.05) NS 1 0.52 0.64 0.57 0.38 0.82 (0.73 to 0.93) **
CHD deaths 1 1.16 0.90 0.76 1.08 0.98 (0.87 to 1.10) NS 1 0.39 0.62 0.34 0.35 0.77 (0.60 to 0.99) *
All deaths 1 0.86 0.69 0.73 0.74 0.92 (0.85 to 0.99) * 1 0.66 0.75 0.84 0.59 0.92 (0.82 to 1.02) NS
Serum cotinine (ng/ml)
0 0.01- 2.17- 59.31- > 285.0 0 0.01- 1.40- 14.21- >248.1
% 13.9 21.5 21.5 21.5 21.6 1409 22.5 19.4 19.3 19.5 19.3 1656
All CHD 1 1.19 1.13 1.72 2.18 1.23 (1.13 to 1.35) *** 1 0.91 0.96 2.61 4.48 1.59 (1.38 to 1.83) ***
CHD deaths 1 1.39 0.89 2.01 2.91 1.32 (1.14 to 1.54) *** 1 1.20 1.65 2.90 10.73 2.05 (1.46 to 2.87) ***
All deaths 1 1.18 1.43 2.09 2.56 1.29 (1.18 to 1.42) *** 1 0.52 1.17 1.92 3.30 1.48 (1.30 to 1.68) ***†
Alcohol (units/week)
Units 0 1-7 8-15 16-29 >30 0 1-2 3-5 6-9 >10
% 20.0 20.4 19.7 20.1 19.7 15 38.0 414.6 17.8 14.5 15.0 22
All CHD 1 1.00 0.59 0.82 0.62 0.89 (0.83 to 0.95) **† 1 0.70 0.64 0.56 0.64 0.86 (0.77 to 0.96) ***
CHD deaths 1 0.87 0.63 0.94 0.62 0.92 (0.82 to 1.03) NS 1 0.49 0.58 0.36 0.52 0.79 (0.62 to 1.00) *
All deaths 1 0.90 0.84 0.83 1.17 1.02 (0.95 to 1.10) NS 1 0.64 0.85 0.79 0.93 0.97 (0.88 to 1.07) NS
CHD=coronary heart disease. ECG=electrocardiogram. MI=myocardial infarction. See footnote to table 1 for P values.
Table 4 Age adjusted hazard ratios by fifths of physical attributes
Centile
Men Women
1 20 40 60 80 99 Missing 1 20 40 60 80 99 Missing
Fifth 12345
Multiplicative constant
(95% CI) P value 12345
Multiplicative constant
(95% CI) P value
Height (m)
1.57F 1.67 1.71 1.74 1.78 1.88 27 1.46 1.55 1.58 1.61 1.65 1.75 9
All CHD 1 0.94 0.76 0.78 0.66 0.90 (0.84 to 0.97) ** 1 0.65 0.62 0.61 0.63 0.89 (0.80 to 0.98) *
CHD deaths 1 0.92 0.75 0.69 0.80 0.93 (0.83 to 1.03) NS 1 0.82 0.82 0.88 0.91 0.98 (0.80 to 1.20) NS
All deaths 1 1.06 0.76 0.78 0.78 0.92 (0.86 to 0.99) * 1 0.86 1.11 1.10 0.84 1.00 (0.91 to 1.10) NS
Weight (kg)
53 68 74 80 87 111 6 44 56 61 66 74 103 3
All CHD 1 1.01 1.23 1.12 1.40 1.08 (1.01 to 1.16) * 1 0.79 0.69 0.75 1.40 1.08 (0.97 to 1.20) NS†
CHD deaths 1 0.88 0.90 0.77 1.31 1.05 (0.94 to 1.18) NS 1 0.42 0.85 0.63 1.40 1.11 (0.91 to 1.37) NS
All deaths 1 0.83 0.66 0.64 0.92 0.96 (0.89 to 1.03) † 1 0.75 0.78 0.67 1.13 1.02 (0.92 to 1.12) †
Body mass index (kg/m
2
)
18.8 23.3 25.1 26.7 28.7 36.2 27 18.0 22.1 24.0 26.0 28.9 41.1 10
All CHD 1 1.71 1.47 1.52 1.97 1.12 (1.05 to 1.20) ** 1 0.69 0.71 0.98 1.37 1.13 (1.01 to 1.25) *
CHD deaths 1 1.77 1.11 1.25 1.68 1.06 (0.95 to 1.19) NS 1 0.40 0.29 0.68 1.20 1.13 (0.92 to 1.40) NS†
All deaths 1 0.98 0.86 0.78 1.05 0.99 (0.92 to 1.06) NS 1 0.57 0.69 0.77 0.97 1.03 (0.93 to 1.13) †
Systolic blood pressure (mm Hg)
98 118 127 136 148 190 6 95 113 123 133 148 192 6
All CHD 1 1.07 1.54 1.65 2.25 1.23 (1.15 to 1.33) *** 1 1.22 1.38 1.84 3.10 1.35 (1.20 to 1.52) ***
CHD deaths 1 0.54 1.26 0.98 1.62 1.18 (1.05 to 1.32) **† 1 1.65 6.48 5.08 13.1 1.76 (1.33 to 2.33) ***
All deaths 1 0.77 1.20 1.13 1.43 1.12 (1.04 to 1.20) ** 1 1.02 0.84 1.07 2.18 1.25 (1.12 to 1.39) ***††
Diastolic blood pressure (mm Hg)
59 74 80 86 93 115 7 58 71 78 83 90 113 7
All CHD 1 1.52 1.47 1.51 2.47 1.21 (1.12 to 1.30) *** 1 0.98 1.13 1.08 1.85 1.17 (1.05 to 1.30) **
CHD deaths 1 1.40 1.54 1.06 2.43 1.19 (1.06 to 1.34) **† 1 0.94 3.25 0.60 3.63 1.31 (1.05 to 1.63) *††
All deaths 1 0.82 1.06 0.96 1.29 1.08 (1.00 to 1.16) * 1 0.73 1.23 0.85 1.45 1.11 (1.00 to 1.22) *
CHD=coronary heart disease. See footnote to table 1 for P values.
General practice
725BMJ VOLUME 315 20 SEPTEMBER 1997
hazard ratios relative to that calculated for the remain-
ing one to four classes. Significance tests were applied
both for a linear trend in the logarithm of hazard ratios
and for a residual non-linear effect where appropriate.
Rather than calculating confidence intervals for each
class, we calculated an estimate and 95% confidence
limits for the multiplicative constant of risk across con-
secutive classes. After converting hazard ratios of < 1.0
to their reciprocals for protective factors we ranked
those for the top class of each factor in bar charts for all
coronary heart disease and all deaths, although this
meant mixing results for continuous and categorical
factors. Complex issues of interaction, confounding,
measurement, and causation are deferred to later
analyses.
Results
After allowing for wrong addresses, the 11 629 men
and women in the Scottish heart health study follow up
study were 72% of those originally invited. Numbers at
risk and numbers of end points by age and sex are
given in table 1. Annual event rates in men were 9.6 per
1000 for all coronary heart disease, 3.7 for coronary
heart disease deaths, and 8.9 for all deaths. In women
the equivalent rates were 4.0, 1.1, and 4.7.
Tables 2-6 shows the relation of different types of
factors to the risk of the three end points. Fuller details
of the categories are referenced where necessary.
Housing tenure
27
—
Renters had highly significant
excess hazard ratios than owner occupiers for all end
points, higher for women than men.
Diabetes mellitus had the same low prevalence of
1.5% in both sexes. A highly significant excess of all
deaths existed in men and of all coronary heart disease
in women; other hazard ratios were less significant.
Women who underwent the menopause
28
before
recruitment, after age adjustment, had insignificantly
raised hazard ratios.
Current cigarette smokers
29-32
had very highly signifi-
cant increases in hazard ratios over never smokers for
all end points, greater in women.
Physical inactivity at work
33
was associated with
increased risk for all end points, but the proportion
admitting to it was small, particularly in women.
Self reported leisure physical inactivity
33
was also asso-
ciated with significantly increased risks for all end points.
Previous coronary heart disease
34
was analysed for four
categories compared with those with no evidence of it
from their medical history, Rose questionnaire, or
electrocardiogram.
18
These were: a positive score on the
Rose questionnaire alone; an electrocardiogram positive
for ischaemia without a diagnosis; diagnosed angina
without myocardial infarction; diagnosed myocardial
infarction. Those with past myocardial infarction could
not score for non-fatal recurrence, hence the missing
cells. Hazard ratios were greatly increased for all catego-
ries of coronary heart disease and for all cause mortality.
A high Bortner score
20 35
for type A personality in
men showed no significant effect in predicting
coronary end points but was significantly protective
against all deaths. In women it showed a protective
effect for coronary end points too.
Serum cotinine values,
29-32
measuring exposure to
tobacco smoke, showed a two to threefold increase in
hazard ratio in men for all end points; hazard ratios in
women were even higher.
Table 5 Age adjusted hazard ratios by fifths of metabolic and haemostatic factors
Centile
Men Women
1 20 40 60 80 99 Missing 1 20 40 60 80 99 Missing
Fifth 12345
Multiplicative constant
(95% CI) P value 12345
Multiplicative constant (95%
CI) P value
Total serum cholesterol (mmol/l)
3.96 5.41 6.01 6.56 7.31 9.44 424 3.98 5.47 6.16 6.80 7.65 10.25 733
All CHD 1 1.13 2.05 2.15 3.15 1.34 (1.34 to 1.44) *** 1 2.43 2.97 3.51 3.94 1.28 (1.12 to 1.45) ***
CHD deaths 1 1.14 1.50 1.74 2.21 1.23 (1.09 to 1.38) *** 1 1.10 3.23 1.87 2.27 1.14 (0.88 to 1.48) NS
All deaths 1 0.88 1.04 0.93 1.13 1.03 (0.96 to 1.11) NS 1 0.83 1.00 0.73 0.86 0.96 (0.86 to 1.07) NS
HDL cholesterol (mmol/l)
0.68 1.06 1.23 1.40 1.64 2.43 676 0.86 1.32 1.54 1.74 2.00 2.90 926
All CHD 1 0.65 0.54 0.42 0.40 0.79 (0.73 to 0.85) *** 1 0.59 0.44 0.35 0.22 0.69 (0.61 to 0.78) ***
CHD deaths 1 0.71 0.65 0.43 0.44 0.80 (0.71 to 0.91) *** 1 0.87 0.53 0.53 0.32 0.76 (0.60 to 0.97) *
All deaths 1 0.81 0.91 0.81 0.82 0.95 (0.88 to 1.02) NS 1 0.74 0.59 0.35 0.71 0.87 (0.78 to 0.97) *†
Serum triglycerides (mmol/l)
0.57 1.23 1.68 2.30 3.25 7.27 430 0.50 0.91 1.22 1.59 2.23 5.55 735
All CHD 1 1.48 1.65 2.11 2.25 1.21 (1.12 to 1.30) *** 1 2.21 1.86 3.90 5.23 1.46 (1.28 to 1.67) ***
CHD deaths 1 1.69 1.36 2.18 1.70 1.13 (1.00 to 1.27) * 1 3.50 4.11 5.25 6.21 1.33 (1.01 to 1.74) *
All deaths 1 1.29 0.79 1.27 1.20 1.03 (0.96 to 1.12) † 1 1.01 1.24 1.50 1.61 1.15 (1.02 to 1.29) *
Blood glucose (mmol/l)
3.28 4.35 4.69 5.02 5.54 11.24 421 3.29 2.44 4.62 4.80 5.21 9.55 731
All CHD 1 1.11 0.68 0.73 1.12 0.99 (0.92 to 1.06) NS†† 1 1.01 0.87 0.79 1.14 1.01 (0.90 to 1.13) NS
CHD deaths 1 1.12 0.63 0.66 1.07 0.97 (0.86 to 1.09) NS 1 1.14 0.90 0.71 1.60 1.09 (0.86 to 1.37) NS
All deaths 1 0.89 0.72 0.64 0.96 0.97 (0.90 to 1.04) † 1 0.88 0.78 0.69 1.06 1.00 (0.89 to 1.11) NS
Uric acid (ìmol/l)
185.6 264.2 295.7 327.4 367.3 488.2 421 127.6 193.9 221.5 248.3 285.9 426.8 730
All CHD 1 1.02 1.10 0.97 1.32 1.06 (0.98 to 1.14) NS 1 1.17 1.20 1.10 1.54 1.09 (0.97 to 1.23) NS
CHD deaths 1 0.57 1.05 0.81 1.26 1.08 (0.96 to 1.22) NS 1 1.71 0.58 1.48 1.84 1.14 (0.89 to 1.46) NS
All deaths 1 0.96 1.01 0.95 1.16 1.03 (0.96 to 1.11) NS 1 0.95 1.11 1.05 1.72 1.15 (1.03 to 1.29) *
Plasma fibrinogen (g/l)
1.06 1.83 2.10 2.36 2.75 4.61 659 1.07 1.90 2.17 2.44 2.82 4.51 1015
All CHD 1 1.21 1.30 1.30 2.16 1.19 (1.10 to 1.29) *** 1 1.59 1.66 1.91 2.70 1.24 (1.09 to 1.40) ***
CHD deaths 1 1.38 1.36 1.49 3.01 1.30 (1.14 to 1.48) *** 1 0.88 1.39 2.15 3.42 1.45 (1.11 to 1.91) **
All deaths 1 1.40 1.34 1.56 2.59 1.25 (1.15 to 1.35) *** 1 0.87 1.09 1.20 2.20 1.26 (1.12 to 1.41) ***
CHD=coronary heart disease. See footnote to table 1 for P values.
General practice
726 BMJ VOLUME 315 20 SEPTEMBER 1997
Alcohol consumption
36
differed in dosage between the
sexes. In women the large group of abstainers did
worst for all end points: alcohol had a very highly sig-
nificant protective effect against all coronary heart dis-
ease. Men showed similar protection against all
coronary heart disease. The male risk curve for all
deaths was a shallow non-significant “U” shape, with
the higher limb in the high consumption group (>30
units/week), whereas in women the curve was tilted
marginally the other way.
Height was significantly protective against all
coronary heart disease in both sexes and all deaths in
men.
Weight showed a “U” shaped relation to death from
all causes in both sexes. For other end points the curve
was not significant or was “J” shaped.
Body mass index produced a significant excess risk of
all coronary heart disease with obesity but a “U”
shaped curve for all deaths, which was shallow in
women and non-significant in men.
The mean of the two systolic blood pressure
37
readings
showed highly significant gradients for all end points in
both sexes. Gradients were steeper in women.
For diastolic blood pressure
37
gradients in men were
similar to those for systolic pressure, but in women they
were somewhat reduced.
The gradient with total serum cholesterol concentra-
tion
38
was very highly significantly positive for all
coronary heart disease, weaker for coronary heart dis-
ease deaths, and undetectable for all deaths.
By contrast, high density lipoprotein cholesterol
concentration showed a protective negative gradient
for coronary end points, more marked in women, in
whom it also related significantly to all deaths; men
also showed a negative tendency, albeit insignificant.
Serum triglyceride concentrations (measured with-
out fasting) showed significant positive gradients for
five of the six end points, steeper in women. All deaths
in men was the exception.
Blood glucose (measured without fasting) showed
shallow insignificant “U” shaped risk curves for all end
points in women and significantly non-linear effects for
two end points in men.
There seemed to be no general pattern of risk for
uric acid concentration beyond a slightly raised risk in
the highest class for all end points in both sexes.
Plasma fibrinogen
21 28 39
showed highly significant
positive trends in both sexes for all end points.
Sodium excretion
40
did not predict coronary heart
disease in men and showed a borderline negative
gradient for all deaths, whereas in women it was just
positive for all coronary heart disease.
In contrast, potassium excretion
40
showed a highly
significant protective gradient for all deaths in both
sexes and significantly protected against all coronary
heart disease in men.
Estimated energy intake
41
from the food frequency
questionnaire showed no significant effects.
Carotenoids
42
(part of vitamin A) were significantly
protective for all coronary heart disease in men.
That finding was even stronger for ascorbic acid
42
(vitamin C).
Tocopherol
42
(vitamin E) intake was just significantly
protective for male end points.
Risk factor ranking—Figs 1 and 2 display the ranking
in men and women for the different factors for all
Table 6 Age adjusted hazard ratios by fifths of dietary factors
Centile
Men Women
1 20 40 60 80 99 Missing 1 20 40 60 80 99 Missing
Fifth 12345
Multiplicative
constant (95% CI) P value 12345
Multiplicative
constant (95% CI) P value
Urinary sodium ion excretion (mmol/day)
46.8 129.6 168.4 204.1 251.3 416.7 1634 37.8 98.0 123.4 149.0 187.3 319.3 1906
All CHD 1 1.18 1.11 1.26 1.23 1.05 (0.96 to 1.14) NS 1 0.93 01.97 1.09 1.76 1.16 (1.00 to 1.33) *
CHD deaths 1 0.96 0.62 0.97 10.92 0.98 (0.86 to 1.13) NS 1 1.36 0.41 0.85 2.05 1.14 (0.87 to 1.49) NS
All deaths 1 0.99 0.65 0.86 0.71 0.92 (0.84 to 1.00) BS 1 0.61 0.82 0.67 0.85 0.97 (0.86 to 1.10) NS
Urinary potassium ion excretion (mmol/day)
17.6 47.2 59.5 71.3 86.3 138.1 1633 15.3 39.7 49.4 58.5 70.2 116.4 1906
All CHD 1 0.62 0.87 0.58 0.66 0.91 (0.83 to 0.99) * 1 0.91 0.57 0.79 0.67 0.90 (0.79 to 1.04) NS
CHD deaths 1 0.57 0.76 0.59 0.60 0.89 (0.77 to 1.03) NS 1 0.73 0.51 0.62 0.45 0.83 (0.63 to 1.10) NS
All deaths 1 0.57 0.58 0.58 0.45 0.84 (0.77 to 0.92) *** 1 0.74 0.86 0.48 0.41 0.81 (0.81 to 0.92) **
Dietary energy intake (kcal/day)
1174 1845 2148 2433 2802 4135 12 812 1380 1611 1834 2107 3142 18
All CHD 1 1.04 1.01 0.89 0.85 0.95 (0.89 to 1.02) NS 1 1.05 0.80 0.84 1.23 1.03 (0.91 to 1.14) NS
CHD deaths 1 0.90 0.69 1.01 0.77 0.96 (0.86 to 1.07) NS 1 0.69 0.53 0.94 0.75 0.97 (0.79 to 1.19) NS
All deaths 1 0.82 0.86 1.10 1.07 1.04 (0.97 to 1.12) NS 1 1.01 0.97 0.95 1.38 1.07 (0.97 to 1.18) NS
Carotenoid intake (mg/day)
0.12 1.51 2.04 3.27 4.64 10.06 319 0.16 1.59 2.94 3.35 4.74 10.12 224
All CHD 1 1.11 1.03 0.73 01.70 0.90 (0.84 to 0.96) ** 1 0.75 1.07 0.73 0.89 0.97 (0.87 to 1.08) NS
CHD deaths 1 1.09 1.26 0.87 0.87 0.95 (0.85 to 1.07) NS 1 0.54 1.32 0.87 0.93 1.01 (0.82 to 1.24) NS
All deaths 1 1.11 0.99 0.95 0.93 0.97 (0.90 to 1.04) NS 1 0.83 0.87 0.81 0.96 0.99 (0.90 to 1.09) NS
Ascorbic acid intake (mg/day)
14.9 34.7 44.2 54.5 70.2 116.1 50 10.9 30.5 42.1 55.2 72.3 118.5 30
All CHD 1 0.87 0.83 0.63 0.52 0.85 (0.80 to 0.91) *** 1 0.59 0.85 0.81 0.56 0.92 (0.82 to 1.02) BS
CHD deaths 1 0.98 1.03 0.86 0.72 0.93 (0.83 to 1.04) NS 1 1.17 2.12 1.61 0.76 0.99 (0.81 to 1.21) NS
All deaths 1 1.09 1.00 0.87 0.85 0.95 (0.88 to 1.02) NS 1 0.98 0.79 1.03 0.72 0.95 (0.86 to 1.04) NS
Tocopherol intake (mg/day)
2.2 4.0 5.1 6.3 9.0 27.4 51 1.9 3.5 4.4 5.4 7.8 23.7 32
All CHD 1 0.88 0.72 0.76 0.79 0.94 (0.87 to 1.00) BS 1 0.79 0.71 0.75 0.75 0.94 (0.84 to 1.04) NS
CHD deaths 1 0.76 0.75 0.58 0.67 0.89 (0.80 to 1.00) * 1 0.87 1.10 0.37 1.06 0.96 (0.78 to 1.18) NS
All deaths 1 0.81 0.86 0.64 0.76 0.92 (0.86 to 0.99) * 1 1.30 1.18 1.12 0.94 0.97 (0.88 to 1.07) NS
CHD=coronary heart disease. See footnote to table 1 for P values.
General practice
727BMJ VOLUME 315 20 SEPTEMBER 1997
coronary heart disease and for all deaths, in
descending order of hazard ratios for the top class.
Harmful factors have a plus sign. For protective factors
(minus sign) the reciprocal was used.
Discussion
Our 27 factors cover the full range of classical and can-
didate factors included when the Scottish Heart Health
Study was planned, but to prevent overload we chose
single exemplars for social status
27 43
and tobacco smoke
inhalation,
29 32
restricted diet to six factors, and left fat
biopsy analyses for later. The three end points are also
a subset. Early cohort studies emphasised first develop-
ment of coronary heart disease,
14
but table 3 shows that
over 20% of the participants had evidence of coronary
heart disease when seen, and 8.2% of men and 5.0% of
women had been already diagnosed. We found similar
risk factor rankings to those in fig 1 in people with no
evidence or history of previous coronary heart disease
but the number of end points was halved. The total
study population is relevant to reality, where middle
aged people are at all stages of coronary heart disease
but the same risk factors operate across them. Our
results emphasise that existing coronary heart disease is
the most powerful of its own predictors
—
and therefore
the importance of secondary prevention. Death from
coronary heart disease provides a specific, severe end
point linking all coronary heart disease with all deaths,
the ultimate arbiter.
Failure of some coronary risk factors to predict all
cause mortality has been explained previously by sepa-
rating early from late deaths.
44
Our results for four and
eight years were similar. Alternatives to the Cox
model
26
produced similar findings, showing that our
results are robust.
We have not preselected risk factors by results.
Twenty seven of our own and others’ favourite factors
were entered as starters, without bias, handicap, or dis-
qualification, into three competitions for rankings run
in parallel in each sex.
Men and women, coronary end points and all cause
mortality
Results of comparing men and women are illustrated
in fig 3, which shows confidence limits for hazard ratios
in the different fifths of total cholesterol concentration.
Limits were wider in women, as the numbers of end
points were smaller, but the extremes of risk, as in
many factors, appeared greater. Risk overall is lower in
women so a larger hazard ratio may conceal a smaller
excess risk between top and bottom fifth, a situation
analogous to the effect of age in factors such as smok-
ing, where relative risk is very high in young people, in
whom risk is low.
45
Rankings used in the figures, based
on the extreme classes, and the multiplicative constant
between adjoining classes, both summarise the results.
The tables sometimes suggest other complex distribu-
tions and thresholds not simply summarised. Compre-
hensive ranking of factors in figs 1 and 2 was feasible
only by mixing together the categorical factors with
continuous ones.
Men and women show different rankings for the
factors but considerable agreement. For all coronary
heart disease the same eight factors appear in the top
12 for both sexes, for coronary heart disease death the
same nine, and for all deaths the same 10
—
which is
surprising in view of the heterogeneity in cause of
death between men and women. Random variation
could account for much of the differences in ranking
between the sexes. It would be reduced with larger
numbers. The multiplicative constants in tables 2-6 are
often very close, and their confidence limits overlap.
When we tested the hazard ratios for evidence of
differences between the two sexes a few were of
borderline statistical significance, but the most extreme
was the Bortner score, which is ranked 11th in women
and 26th in men for all coronary heart disease, a find-
ing consistent with reported sex differences in
responses.
46
Association, causation, interaction, confounding,
and regression dilution
Observational epidemiological studies are better
placed to show association than causation. Early
disease may change factors so that associations are
concealed or causation reversed. Factors may be asso-
ciated sufficiently closely to cause confounding
47
and
lack of independence. Difficulties of measurement and
within person variability may conceal or minimise true
effects
—
so called regression dilution.
48
There may be a
threshold. High ranking of a factor does not guarantee
causation, nor low rank lack of it.
Nevertheless, the strength of associations, or
hazard ratios, is important evidence for causation. The
associations help show relevance in a British
population targeted for change whose pattern of risk
factors overlaps that in many industrialised countries.
49
Our results emphasise the importance of coronary
heart disease itself as a marker for risk warranting
intervention, and also that of the classical coronary risk
factors in coronary heart disease risk. Reasons why
other factors have done unexpectedly well or badly
warrant investigation both within the dataset of the
Scottish Heart Health Study and elsewhere, as do the
discrepancies between risk factors for coronary heart
disease and all cause mortality.
We thank the 300 general practitioners and the survey nurses,
technicians, data clerks, and research assistants; the agencies for
providing death certificates and data on hospital admissions;
other past contributors (referenced coauthors); and, above all,
11 629 men and women across Scotland who willingly
participated.
10
5
2
1
0.5
Hazard ratio and 95% confidence interval
12345 12345
Fifths of cholesterol
Men Women
Fig 3 Age adjusted hazard ratios (and 95% confidence limits) for all coronary heart disease
events by fifths of total cholesterol in men and women (lowest fifth=1)
General practice
728 BMJ VOLUME 315 20 SEPTEMBER 1997
Funding: Scottish Office, British Heart Foundation, and
BUPA (an earlier version of this paper, confined to coronary end
points won the 1996 BUPA epidemiology prize).
Conflict of interest: None.
1 Dawber TR.The Framingham study. The epidemiology of atherosclerotic disease.
Cambridge MA: Harvard University Press, 1980
2 Isles CG, Hole DJ, Hawthorne VM, Lever AF. Relation between coronary
risk and coronary mortality in women of the Renfrew and Paisley survey:
comparison with men. Lancet 1992;339:702-6.
3 Rich-Edwards JW, Manson JE, Stampfer MJ, Colditz GA, Willett WC, Ros-
ner B, et al. Height and the risk of cardiovascular disease in women.AmJ
Epidemiol 1995;142:909-17.
4 Pooling Project Research Group. Relationship of blood pressure, serum
cholesterol, smoking habit, relative weight and ECG abnormalities to
incidence of major coronary events: final report on the Pooling Project. J
Chron Dis 1978;31:201-6.
5 Keys A. Seven countries: a multivariate analysis of death and coronary heart
disease. Cambridge, MA: Harvard University Press, 1980.
6 Shaper AG, Pocock SJ, Walker M, Phillips AM, Whitehead TP, Macfarlane
PW. Risk factors for ischaemic heart disease: the prospective phase of the
British Regional Heart Study. J Epidemiol Community Health 1985;39:197-
209.
7 Sweetnam PM, Bolton CH, Yarnell JW, Bainton D, Baker IA, Elwood PC,
et al. Associations of the HDL2 and HDL3 cholesterol subfractions with
the development of ischaemic heart disease in British men. The
Caerphilly and Speedwell Collaborative Heart Studies. Circulation
1994;90:769-74.
8 Meade TW, Ruddock V, Stirling Y, Chakrabarti R, Miller GJ. Fibrinolytic
activity, clotting factors, and long term incidence of ischaemic heart dis-
ease in the Northwick Park Heart Study. Lancet 1993;342:1076-9.
9 Smith GD, Shipley MJ, Marmot MG, Rose G. Plasma cholesterol concen-
tration and mortality. The Whitehall Study. JAMA 1992;267;70-6.
10 WHO MONICA Project, prepared by Tunstall-Pedoe H, Kuulasmaa K,
Amouyel P, Arveiler D, Rajakangas AM, Pajak A. Myocardial infarction
and coronary deaths in the World Health Organization MONICA Project
Registration procedures, event rates, and case fatality rates in 38 popula-
tions from 21 countries in 4 continents. Circulation 1994;90:583-612.
11 World health statistics annual 1995. Geneva: WHO, 1996.
12 Smith WCS, Crombie IK, Tavendale R, Irving JM, Kenicer MB, Tunstall-
Pedoe H. The Scottish Heart Health Study: objectives and development
of methods. Health Bull 1987;45:211-17.
13 Smith WCS, Tunstall-Pedoe H, Crombie IK, Tavendale R. Concomitants
of excess coronary deaths major risk factor and lifestyle findings from
10,359 men and women in the Scottish Heart Health Study. Scot Med J
1989;34:550-5.
14 Tunstall-Pedoe H, Smith WCS, Crombie IK, Tavendale R. Coronary risk
factor and lifestyle variation across Scotland: results from the Scottish
Heart Health Study. Scot Med J 1989;34:556-60.
15 Crombie IK, Smith WCS, Tavendale R, Tunstall-Pedoe H. Geographical
clustering of risk factors and lifestyle for coronary heart disease in the
Scottish Heart Health Study. Br Heart J 1990;64:199-203.
16 Crombie IK, Smith WCS, Irving JM, Tunstall-Pedoe H. Experience with
general practitioner lists as a sampling frame for a survey of
cardiovascular risk factors. Statistician 1989;38:25-31.
17 WHO MONICA Project Principal Investigators (prepared by Tunstall-
Pedoe H). The World Health Organization MONICA Project
(Monitoring trends and determinants in cardiovascular disease): a major
international collaboration. J Clin Epidemiol 1988;41:105-14.
18 Rose GA, Blackburn H, Gillum RF, Prineas RJ: Cardiovascular survey
methods. Geneva: World Health Organisation, 1982.
19 Yarnell JWG, Fehily AM, Milbank ME, Sweetnam PM, Walker CC. A short
dietary questionnaire for use in epidemiological surveys: a comparison
with weighed records. Hum Nutr:Appl Nutr 1983;37:103-12.
20 Bortner RW. A short rating scale as a potential measure of pattern A
behaviour. J Chron Dis 1969;22:87-91
21 Lee AJ, Smith WCS, Lowe GDO, Tunstall-Pedoe H. Plasma fibrinogen
and coronary risk factors : the Scottish Heart Health Study. J Clin Epide-
miol 1990;43:913-9.
22 WHO MONICA Project: MONICA manual (revised edition). Geneva: World
Health Organisation, 1990.
23 Prineas RJ, Crow R.S, Blackburn H: The Minnesota code manual of
electrocardiographic findings: standards and procedures for measurement and
classification. Bristol: John Wright, 1982.
24 Kendrick S, Clarke J. The Scottish record linkage system. Health Bull
1993,51:72-79.
25 Tunstall-Pedoe H, Morrison C, Woodward M, Fitzpatrick B, Watt G. Sex
differences in myocardial infarction and coronary deaths in the Scottish
MONICA population of Glasgow 1985 91: presentation, diagnosis, treat-
ment and 28 day case fatality of 3991 events in men and 1551 events in
women. Circulation 1996;93:1981-92.
26 Collett D. Modelling survival data in medical research. London: Chapman
and Hall, 1994.
27 Woodward M, Shewry MC, Smith WCS, Tunstall-Pedoe H. Social status
and coronary heart disease : results from the Scottish Heart Health Study.
Prev Med 1992;21:136-48.
28 Lee AJ, Lowe GDO, Smith WCS, Tunstall-Pedoe H. Plasma fibrinogen in
women: relationships with oral contraception, menopause and hormone
replacement therapy. Br J Haematol 1993;83:616-21.
29 Woodward M, Tunstall-Pedoe H, Smith WCS, Tavendale R. Smoking
characteristics and inhalation biochemistry in the Scottish population. J
Clin Epidemiol 1991;44:1405-10.
30 Tunstall-Pedoe H, Woodward M, Brown CA. Tea drinking, passive smok-
ing, smoking deception and serum cotinine in the Scottish Heart Health
Study. J Clin Epidemiol 1991;44:1411-4.
31 Woodward M, Tunstall-Pedoe H. Self titration of nicotine: evidence from
the Scottish Heart Health Study. Addiction 1993;88:821-30.
32 Tunstall-Pedoe H, Brown CA, Woodward M, Tavendale R. Passive smok-
ing by self report and serum cotinine and the prevalence of respiratory
and coronary heart disease in the Scottish Heart Health Study. J
Epidemiol Community Health 1995;49:139-43.
33 Crombie IK, Lee AJ, Smith WCS, Tunstall-Pedoe H. Levels and social
patterns of self reported physical activity in Scotland. Health Ed J
1990;49:71-4.
34 Smith WCS, Kenicer MB, Tunstall-Pedoe H, Clark EC, Crombie IK.
Prevalence of coronary heart disease in Scotland: Scottish Heart Health
Study. Br Heart J 1990;64:295-8.
35 Woodward M, Shewry MC, Smith WCS, Tunstall-Pedoe H. Coronary
heart disease and socioeconomic factors in Edinburgh and north
Glasgow. Statistician 1990;39:319-29.
36 Woodward M, Tunstall-Pedoe H. Alcohol consumption, diet, coronary
risk factors and prevalent coronary heart disease in men and women in
the Scottish Heart Health Study. J Epidemiol Community Health
1995;49:354-62.
37 Smith WCS, Lee AJ, Crombie IK, Tunstall-Pedoe H. Control of blood
pressure in Scotland : the rule of halves. BMJ 1990;300:981-3.
38 Tunstall-Pedoe H, Smith WCS, Tavendale R. How-often-that-high graphs
of serum cholesterol. Findings from the Scottish Heart Health and Scot-
tish MONICA studies. Lancet 1989;i:540-2.
39 Lee AJ, Lowe GDO, Woodward M, Tunstall-Pedoe H. Fibrinogen in rela-
tion to personal history of prevalent hypertension, diabetes, stroke, inter-
mittent claudication, coronary heart disease and family history: the
Scottish Heart Health Study. Br Heart J 1993;69:338-42.
40 Smith WCS, Crombie IK, Tavendale RT, Gulland SK, Tunstall-Pedoe H.
Urinary electrolyte excretion, alcohol consumption, and blood pressure
in the Scottish heart health study. BMJ 1988;297:329-30.
41 Bolton Smith C, Woodward M, Tunstall-Pedoe H. The Scottish Heart
Health Study. Dietary intake by food frequency questionnaire and odds
ratios for coronary heart disease risk. I. The macronutrients. Eur J Clin
Nutr 1992;46:75-84.
42 Bolton Smith C, Woodward M, Tunstall-Pedoe H. The Scottish Heart
Health Study. Dietary intake by food frequency questionnaire and odds
ratios for coronary heart disease risk. II. The antioxidant vitamins and
fibre. Eur J Clin Nutr 1992;46:85-93.
43 Morrison C, Woodward M, Leslie W, Tunstall-Pedoe H. Effect of
socioeeconomic group on incidence of, management of, and survival
after myocardial infarction and coronary death: analysis of community
coronary event register. BMJ 1997;314:541-6.
44 Rose G, Shipley M. Plasma lipids and mortality: a source of error. Lancet
1980;i:523-6.
45 Parish S, Collins R,Peto R, Youngman L, Barton J,Jayne K, et al. Cigarette
smoking, tar yields and non fatal myocardial infarction in 14 000 cases
and 32 000 controls in the United Kingdom. BMJ 1995;311:471-7.
46 Knox SS, Follmann D. Gender differences in the psychosocial variance of
Framingham and Bortner Type A measures. J Psychosom Res 1993;709-16.
47 Smith GD, Phillips AN. Confounding in epidemiological studies: why
independent effects may not be all they seem. BMJ 1992;305:757-9.
48 Hughes MD. Regression dilution in the proportional hazards model. Bio-
metrics 1993;49: 1056-66.
49 The WHO MONICA Project: a worldwide monitoring system for cardio-
vascular diseases. WHO statistics annual 1989. Geneva: World Health
Organisation, 1989: 27-139.
(Accepted 17 June 1997)
Key messages
x Among Scottish men and women studied for 27
risk factors for coronary heart disease and
followed up for eight years classical risk factors
scored strongly in predicting coronary risk but
the performance of new ones was more variable
x Risk factors for coronary disease, and also for
death, showed few, albeit interesting, differences
between men and women
x Relative risk was often higher for risk factors in
women but they had low levels of absolute risk
when risk factor levels were low
x Smoking, blood pressure, and fibrinogen
predicted coronary disease and also death, but
other factors are less consistent
x Unifactorial results should not be
overinterpreted, but the protective effect of
potassium consumption is of particular interest
General practice
729BMJ VOLUME 315 20 SEPTEMBER 1997
