CLINICAL RESEARCH
European Heart Journal (2016) 37, 880–889
doi:10.1093/eurheartj/ehv630
Prevention and epidemiology
Heterogeneous impact of classic atherosclerotic
risk factors on different arterial territories: the
EPIC-Norfolk prospective population study
Robert M. Stoekenbroek1,2, S. Matthijs Boekholdt3, Robert Luben 4, G. Kees Hovingh 2,
Aeilko H. Zwinderman 5, Nicholas J. Wareham 6, Kay-Tee Khaw 4, and Ron J.G. Peters 3*
1
Department of Vascular Surgery, Academic Medical Center, Amsterdam, The Netherlands; 2Department of Vascular Medicine, Academic Medical Center, Amsterdam, The
Netherlands; 3Department of Cardiology, Academic Medical Center/University of Amsterdam, PO Box 22660, 1100 DD Amsterdam, The Netherlands; 4Department of Public Health
and Primary Care, Institute of Public Health, University of Cambridge, Cambridge, UK; 5Clinical Epidemiology and Biostatistics, Academic Medical Center, Amsterdam, The
Netherlands; and 6Medical Research Council Epidemiology Unit, University of Cambridge, Cambridge, UK
Received 12 February 2015; revised 22 October 2015; accepted 30 October 2015; online publish-ahead-of-print 17 December 2015
Aims
Particular atherosclerotic risk factors may differ in their association with atherosclerosis across vascular territories. Few
studies have compared the associations between multiple risk factors and cardiovascular disease (CVD) manifestations
in one population. We studied the strength of the associations between traditional risk factors including coronary artery
disease (CAD), ischaemic and haemorrhagic stroke, abdominal aortic aneurysms (AAAs), and peripheral arterial
disease (PAD).
.....................................................................................................................................................................................
Methods and
This analysis included 21 798 participants of the EPIC-Norfolk population study, without previous CVD. Events were
results
defined as hospitalization or mortality, coded using ICD-10. The associations between the risk factors, such as lowdensity lipoprotein cholesterol, systolic blood pressure (SBP), and smoking, and the various CVD manifestations
were compared using competing risk analyses. During 12.1 years, 3087 CVD events were recorded. The associations
significantly differed across CVD manifestations. Low-density lipoprotein cholesterol was strongly associated with
CAD [adjusted hazard rate (aHR) highest vs. lowest quartile 1.63, 95% CI 1.44 –1.86]. Systolic blood pressure was a
strong risk factor for PAD (aHR highest vs. lowest quartile 2.95, 95% CI 1.78–4.89) and ischaemic stroke (aHR highest
vs. lowest quartile 2.48, 95% CI 1.55–3.97), but not for AAA. Smoking was strongly associated with incident AAA (aHR
current vs. never 7.66, 95% CI 4.50 – 13.04) and PAD (aHR current vs. never 4.66, 95% CI 3.29 – 6.61), but not with
haemorrhagic stroke.
.....................................................................................................................................................................................
Conclusion
The heterogeneity in the risk factor – CVD associations supports the concept of pathophysiological differences
between atherosclerotic CVD manifestations and could have implications for CVD prevention.
----------------------------------------------------------------------------------------------------------------------------------------------------------Keywords
Cardiovascular disease † Risk factors † Pathophysiology † Prevention
Introduction
A substantial body of evidence has demonstrated extensive overlap
between the risk factors for various clinical manifestations of atherosclerotic cardiovascular disease (CVD), including coronary artery disease (CAD), stroke, abdominal aortic aneurysms (AAA),
and peripheral arterial disease (PAD).1 – 4 Although various studies
have quantified the strength of the associations between particular
risk factors and different clinical CVD manifestations, the results of
individual studies may not be directly comparable due to differences
in population and study characteristics.5 – 9 Notably, substantial differences in the strength of the associations between high blood
pressure and different CVD outcomes have recently been reported
in a large cohort study.9 Few studies have investigated relationships
between multiple risk factors and multiple CVD manifestations simultaneously within one population.1,10 Consequently, the relative
strength of the associations between conventional risk factors,
such as systolic blood pressure (SBP), low-density lipoprotein
* Corresponding author. Tel: +31 20 566 5971, Fax: +31 20 566 9343, Email: [email protected]
Published on behalf of the European Society of Cardiology. All rights reserved. & The Author 2015. For permissions please email: [email protected].
881
Impact of atherosclerotic risk factors on different arterial territories
cholesterol (LDL-c), and smoking, and different manifestations of
CVD, is largely unknown. Yet, insights in these associations may improve our understanding of differences in the pathophysiology of
the various atherosclerotic CVD manifestations. Moreover, knowledge on the relative contribution of particular risk factors to various
CVD manifestations is important to estimate the benefit that would
result from improving those risk factors.11
Our primary objective was to quantify the associations between
LDL-c, SBP, and smoking, and the incidence of CAD, ischaemic and
haemorrhagic stroke, AAA, and PAD among apparently healthy individuals in the general population.
Methods
Study design
The European Prospective Investigation of Cancer (EPIC)-Norfolk is a
prospective, population-based study of 25 639 inhabitants of Norfolk,
UK, aged between 45 and 79 years old. A detailed description of the
study design and population characteristics has been published previously.12 The EPIC-Norfolk study is part of the 10 country EPIC study
to investigate determinants of cancer. Additional data were obtained
to study the determinants of other diseases, including CVD. Eligible individuals were recruited by mail from registers of general practices. The
study population resembled the UK general population with regard to
many characteristics, including anthropometry, blood pressure, and lipid
profiles, but with a lower proportion of smokers.12 Participants completed a health and lifestyle questionnaire during the baseline survey between 1993 and 1997, and additional data were obtained by trained
nurses during a clinical visit. Participants with a history of myocardial infarction or stroke were excluded from the present analysis. All participants gave informed consent; the study was approved by the Norwich
District Health Authority Ethics Committee and complied to the Declaration of Helsinki.
Definitions
Smoking status was determined based on the answers to the questions
‘Have you ever smoked as much as one cigarette a day for as long as a
year?’ and ‘Do you smoke cigarettes now?’ Systolic blood pressure was
measured after the participant had seated for 5 min, and the average of
two measurements was used. Diabetes was defined by self-report, or
the use of a diet or glucose-lowering drugs. Serum total cholesterol,
high-density lipoprotein cholesterol (HDL-c), and triglycerides were
measured in fresh non-fasting samples using the RA 1000 (Bayer Diagnostics, Basingstroke, UK). Low-density lipoprotein cholesterol levels
were calculated using the Friedewald formula.
Outcomes
All participants were flagged for mortality at the UK Office of National
Statistics, and vital status was ascertained for the entire cohort. Death
certificates were coded by trained nosologists according to the International Classification of Diseases (ICD), 10th revision. Hospitalization
data were obtained using National Health Service numbers through linkage with the East Norfolk Health Authority (ENCORE) database, which
contains information on all hospital contacts throughout England and
Wales. Participants were identified as having experienced an event if
the corresponding ICD-10 code was registered on the death certificate
(as the underlying cause of death or as a contributing factor), or as the
cause of hospitalization. Five CVD outcomes were studied in this report:
CAD (ICD-10 I20 – I25), ischaemic stroke (ICD-10 I63 and I65 – I66),
haemorrhagic stroke (ICD-10 I60 – I62), AAA (ruptured or non-
ruptured; ICD-10 I71), and PAD (ICD-10 I70– I73). The current report
includes results for follow-up through 31 March 2008.
Statistical analysis
Data are presented as means and standard deviations (SDs), medians
and interquartile ranges, or percentages (numbers). Baseline characteristics were compared between participants in the highest vs. lowest
quartiles for LDL-c and SBP, and between current- and never-smokers,
using Student’s t-tests, Mann– Whitney U-tests or X 2 tests.
For each participant, the time interval was calculated between the
baseline health examination and the first CVD event. Individuals were
censored at the time of the first event, death, or the end of follow-up.
Only the first event was included in the current analysis, because risk
factors and the risk of subsequent events may change after having experienced a first CVD event. Only participants with multiple events
on the day of first occurrence were registered as having both outcomes.
The study outcomes were considered competing risks, as participants
were censored after the first event.13 We used the competing risks regression model described by Fine and Gray14 to estimate the subdistribution hazard for each outcome. Fatal and non-fatal events were
combined. Associations are expressed as hazard ratios (HR) and 95%
confidence intervals (95% CI). Low-density lipoprotein cholesterol
and SBP were analysed as quartiles, using the lowest quartile as the reference. In addition, significance of the associations was assessed for
LDL-c and SBP as continuous variables. Separate analyses were performed for current and previous smokers, using never-smokers as the
reference group. For each risk factor, three regression models were
used. The first model was unadjusted, the second model was adjusted
for age and sex, and the third model additionally adjusted for body
mass index (BMI), diabetes, HDL-c, smoking, and either SBP or LDL-c
(in the analyses for LDL-c and SBP, respectively). Likelihood ratio tests
were performed to determine the significance of the heterogeneity in
the associations between risk factors and various CVD manifestations.
A 95% CI excluding unity was considered statistically significant. All statistical analyses were performed using R (R-Project for Statistical Computing, GNU project), version 3.1.1, with the competing risks library
cmprsk and the mstate package.
Results
Of the 25 639 EPIC-Norfolk participants who completed the baseline examination, 1136 were excluded because of previous CVD and
2705 because of incomplete data, leaving 21 798 individuals for the
present analysis (Table 1). The mean age at entry was 59 years (SD 9
years). The various risk factors clustered among individuals, as summarized in Table 2. Male participants had a more adverse risk profile
than females (Table 1). At baseline, 2.5% of male participants and
1.3% of female participants had diabetes. Overall, 37% of the cohort
had an SBP of .140 mmHg, and 16% used antihypertensive drugs at
baseline. Low-density lipoprotein cholesterol .4 mmol/L was observed in 46% of the participants, and 1.1% used lipid-lowering drugs
at baseline. During 254 604 person-years at risk (median follow-up
12.1 years), 3087 incident cardiovascular events were recorded
(Table 3). Of these, 2332 were CAD events, 268 were ischaemic
stroke, 117 were haemorrhagic stroke, 143 were AAA, and 227
were PAD. The overall crude incidence rates per 1000 person-years
were 9.2 (95% CI 8.8 – 9.5) for CAD, 1.1 (95% CI 0.9 – 1.2) for ischaemic stroke, 0.5 (95% CI 0.4 –0.6) for haemorrhagic stroke, 0.6
(95% CI 0.5 – 0.7) for AAA, and 0.9 (95% CI 0.8 – 1.0) for PAD.
882
R.M. Stoekenbroek et al.
Table 1
Baseline characteristics
Characteristics
Males
(N 5 9576)
Females
(N 5 12 222)
59 + 9
................................................................................
Age (years)
59 + 9
Body mass index (kg/m2)
26.4 + 3.2
26.1 + 4.2
.25
Diabetes
65.2% (6246)
2.5% (242)
54.1% (6611)
1.3% (154)
Current
Previous
12.1% (1159)
53.5% (5129)
11.3% (1377)
31.8% (3882)
Never
34.3% (3288)
57.0% (6963)
137 + 18
39.8% (3809)
133 + 19
34.4% (4200)
Smoking
Systolic blood pressure (mmHg)
.140
Diastolic blood pressure (mmHg)
84 + 11
81 + 11
Total cholesterol (mmol/L)
LDL-c (mmol/L)
6.0 + 1.1
3.9 + 1.0
6.3 + 1.2
4.0 + 1.1
.4
HDL-c (mmol/L)
Triglycerides (mmol/L)
44.6% (4271)
47.1% (5751)
1.2 + 0.3
1.7 (1.2–2.4)
1.6 + 0.4
1.4 (1.0–1.9)
Data are presented as means with standard deviations for continuous variables
with a normal distribution, medians and interquartile ranges for continuous
variables with a non-normal distribution, and percentages and numbers for
categorical variables.
HDL-c, high-density lipoprotein cholesterol; LDL-c, low-density lipoprotein
cholesterol.
Clustering of multiple, different event types during follow-up is
shown in Supplementary material online, Table S1. An additional
265 cases of unspecified stroke were reported, which were not included in the current analysis. For all of the studied CVD outcomes,
the proportion of patients who experienced an event increased
across LDL-c and SBP quartiles, as well as according to smoking status (never-, previous-, and current-smokers; Table 3). However, as
shown in Figure 1 and Tables 4 – 6, the strength of the associations
was highly variable. Supplementary material online, Table S2 presents the associations between defined increases in LDL-c (per
1 mmol/L increment) and SBP (per 10 mmHg increment) and the
CVD types. To analyse the impact of age on the associations, we
additionally stratified our analysis in three age categories. The results
are presented in Supplementary material online, Table S3.
diabetes, BMI, HDL-c, and LDL-c, significant differences were observed in the relative grade of association between SBP and the various studied CVD events (P , 0.001). Systolic blood pressure was
particularly strongly associated with PAD (aHR 2.95, 95% CI
1.78 – 4.89). A significant association was observed between SBP
and all of the studied outcomes except for AAA (aHR 1.30, 95%
CI 0.70 – 2.39). Moreover, for all outcomes other than AAA, the
HRs increased across SBP quartiles. The HR for any CVD event
(aHR 1.60, 95% CI 1.41 – 1.81) predominantly expresses the
strength of the association with CAD (Figure 1).
Smoking
Smoking was significantly associated with all event types, except for
haemorrhagic stroke (aHR for current- vs. never-smokers 1.45, 95%
CI 0.84 – 2.51; Table 6). After adjusting for age, sex, diabetes, BMI,
HDL-c, LDL-c, and SBP, there was a significant difference in the associations between current-smoking across the various CVD risk
factors (P , 0.001). A particularly strong association was observed
between smoking and AAA and PAD (aHRs 7.66, 95% CI 4.50 –
13.04 and 4.66, 95% CI 3.29 – 6.61, respectively). The HRs for all
event types were smaller for previous-smoking than for
current-smoking. In fact, CAD and AAA were the only studied outcome measures for which the association with previous-smoking remained significant. The HR for any CVD event is intermediate of the
HR for CAD and the large HRs for AAA and PAD (Figure 1).
Cardiovascular disease manifestations
When comparing the relative importance of the various risk factors
to the studied CVD manifestations, it becomes apparent that all of
the risk factors were positively associated with CAD, particularly
LDL-c. Although ischaemic and haemorrhagic stroke were both associated with SBP, their association with LDL-c did not reach statistical significance. When compared with LDL-c and SBP, AAA was
particularly strongly associated with smoking. Peripheral arterial disease was strongly associated with SBP and smoking, but the association with LDL-c was not statistically significant.
Discussion
Hazard ratios for the various event types according to LDL-c quartiles are provided in Table 4. After adjusting for age, sex, smoking,
diabetes, BMI, HDL-c, and SBP, a significant interaction between
LDL-c and the various studied CVD manifestations was observed
(P , 0.001). A particularly strong association was observed with
CAD [multiple-adjusted HR (aHR) 1.63, 95% CI 1.44 – 1.86]. The
HRs for all other CVD types did not reach statistical significance.
The association between LDL-c and any CVD event reflects the
predominance of CAD (aHR 1.50, 95% CI 1.35 –1.68; Figure 1).
In this analysis of 21 798 participants of the EPIC-Norfolk population
study, we observed substantial heterogeneity in the associations between the traditional atherosclerotic risk factors, such as LDL-c,
SBP, and smoking, and the risk of various CVD manifestations. Lowdensity lipoprotein cholesterol was a particularly strong risk factor
for CAD, but appeared to be less strongly associated with ischaemic
and haemorrhagic stroke, AAA, and PAD. Although SBP was a
strong predictor for incident PAD and ischaemic stroke, the association with AAA lost statistical significance after adjusting for potential confounders. Smoking showed a particularly strong association
with incident AAA and PAD, and the association was much stronger
for current- than for previous-smoking. The overall association between particular risk factors and CVD events is strongly affected by
the predominant occurrence of CAD.
Systolic blood pressure
Previous findings
Hazard ratios for the various event types according to SBP quartiles
are displayed in Table 5. After adjusting for age, sex, smoking,
The strength of the associations between multiple risk factors and
multiple types of atherosclerotic CVD has rarely been studied in a
Low-density lipoprotein cholesterol
Clustering of risk factors among individuals
LDL-c
..........................................
Quartile 1
P-value
Quartile 4
SBP
.......................................
Quartile 1
P-value
Quartile 4
Smoking
.......................................
Never
P-value
Current
.............................................................................................................................................................................................................................................
Cutoff values
,3.24 mmol/L
.4.60 mmol/L
N
5446
5511
Male
Age (years)
42.7% (2329)
56 + 9
39.3% (2168)
61 + 9
Body mass index (kg/m2)
25.5 + 3.9
26.7 + 3.7
,0.001
24.9 + 3.4
27.2 + 4.0
,0.001
Diabetes
Smoking
2.3% (127)
1.3% (74)
,0.001
0.7% (40)
3.4% (184)
,0.001
Current
11.0% (599)
11.8% (648)
12.9% (715)
10.4% (562)
,0.001
N/A
N/A
Previous
Never
40.9% (2227)
48.1% (2620)
41.4% (2280)
46.9% (2583)
37.4% (2071)
49.7% (2755)
44.9% (2430)
44.7% (2422)
N/A
N/A
N/A
N/A
N/A
,0.001
,0.001
0.297
,122 mmHg
.146 mmHg
5541
5414
33.7% (1865)
54 + 8
47.0% (2546)
64 + 8
N/A
,0.001
,0.001
N/A
N/A
10 251
2536
N/A
32.1% (3288)
58 + 9
45.7% (1159)
57 + 9
,0.001
,0.001
26.0 + 3.8
25.4 + 3.7
,0.001
1.5% (153)
1.0% (26)
0.073
Impact of atherosclerotic risk factors on different arterial territories
Table 2
N/A
Systolic blood pressure (mmHg)
131 + 18
139 + 18
,0.001
114 + 7
159 + 12
,0.001
134 + 18
133 + 19
,0.001
.140
Diastolic blood pressure (mmHg)
28.9% (1572)
80 + 11
44.8% (2467)
84 + 11
,0.001
,0.001
N/A
71 + 6
N/A
95 + 9
N/A
,0.001
35.6% (3645)
82 + 11
32.3% (820)
81 + 12
0.002
,0.001
Total cholesterol (mmol/L)
4.9 + 0.6
7.5 + 0.8
,0.001
5.9 + 1.1
6.4 + 1.2
,0.001
6.1 + 1.1
6.1 + 1.2
0.478
LDL-c (mmol/L)
.4
2.7 + 0.4
N/A
5.3 + 0.7
N/A
,0.001
N/A
3.7 + 1.0
36.6% (2028)
4.2 + 1.1
54.0% (2919)
,0.001
,0.001
4.0 + 1.0
45.6% (4673)
4.0 + 1.1
46.6% (1181)
0.285
0.373
HDL-c (mmol/L)
1.5 + 0.5
1.4 + 0.4
,0.001
1.5 + 0.4
1.4 + 0.4
,0.001
1.5 + 0.4
1.4 + 0.4
,0.001
Triglycerides (mmol/L)
1.2 (0.9–1.8)
1.8 (1.3– 2.4)
,0.001
1.2 (0.9– 1.7)
1.7 (1.2–2.4)
,0.001
1.4 (1.0–2.0)
1.5 (1.1–2.2)
,0.001
Data are presented as means with standard deviations for continuous variables with a normal distribution, medians and interquartile ranges for continuous variables with a non-normal distribution, and percentages and numbers for categorical
variables. All estimates presented are unadjusted.
SBP, systolic blood pressure; HDL-c, high-density lipoprotein cholesterol; LDL-c, low-density lipoprotein cholesterol.
883
884
All estimates presented are unadjusted. Depicted are the incidence rates per 1000 person-years and the number of events in each category. For mean and range, the units for LDL-c are mmol/L, and for systolic blood pressure mmHg.
PAD, peripheral arterial disease; CAD, coronary artery disease; LDL-c, low-density lipoprotein cholesterol.
0.59 (17)
1.40 (40)
2.34 (67)
0.78 (81)
0.93 (97)
0.46 (48)
0.43 (52)
0.18 (22)
0.52 (63)
0.92 (56)
1.91 (116)
0.76 (46)
0.55 (35)
0.64 (41)
0.79 (51)
0.49 (31)
0.62 (39)
0.28 (18)
0.69 (44)
1.16 (74)
0.23 (15)
0.32 (21)
0.50 (32)
0.58 (36)
1.05 (65)
0.42 (27)
0.70 (45)
Abdominal aneurysm (n ¼ 143)
PAD (n ¼ 227)
0.56 (36)
0.67 (43)
0.45 (29)
Haemorrhagic stroke (n ¼ 117)
0.43 (28)
0.45 (28)
0.27 (18)
11.46 (328)
1.40 (40)
11.37 (1181)
1.15 (120)
6.74 (823)
0.85 (104)
14.96 (909)
2.16 (131)
10.89 (699)
1.09 (70)
6.49 (410)
0.68 (43)
12.62 (802)
1.40 (89)
5.77 (372)
0.67 (43)
CAD (n ¼ 2332)
Ischaemic stroke (n ¼ 268)
8.24 (532)
1.05 (68)
10.09 (626)
1.10 (68)
4.72 (314)
0.36 (24)
.............................................................................................................................................................................................................................................
N/A
N/A
28 629
N/A
103 898
N/A
N/A
N/A
122 077
147– 232
60 769
159
139
134– 146
64 172
123 –133
63 199
4.60 –10.30
63 574
128
114
5.32
3.90 –4.59
62 053
0.20 –3.24
64 431
Range
Person-years
3.25– 3.89
64 547
2.73
Mean value
3.57
4.22
83– 122
66 464
Current
Previous
Never
4
3
2
1
4
3
2
1
...................................................................
Systolic blood pressure quartiles
...................................................................
LDL-c quartiles
.............................................................................................................................................................................................................................................
Table 3
Incidence (per 1000 person-years) per risk factor quartile and history of smoking
Smoking status
...................................................
R.M. Stoekenbroek et al.
single population. Moreover, relatively few studies have included the
outcomes PAD and AAA. In agreement with our findings, the Framingham Study, a population-based prospective cohort study of
5209 individuals, demonstrated a strong associations between blood
pressure and stroke, blood cholesterol and CAD, and smoking and
PAD.3 The initial analysis showed a weak relationship between SBP
and PAD. However, in line with our results and those from various
other studies, SBP was a strong risk factor for PAD at longer followup.9,15 Differences in the association between a wide variety of risk
factors and PAD and CAD were investigated in the Edinburgh Artery Study, a cross-sectional survey of 1592 individuals from the general population. Consistent with our findings, smoking was a strong
risk factor for PAD, and to a lesser extent for CAD.10 Most other
reports focused on single risk factors or CVD types.6 – 9,16,17
Although the number of participants who experienced some of the
specific CVD outcomes in the current study are too small to draw
conclusions on the shape of the associations and possible threshold
effects, previous studies have indicated that differences may exist
across CVD types.9 For example, Rapsomaniki et al. 9 reported a linear
association between SBP and CAD, but a log-linear association between SBP and ischaemic and haemorrhagic stroke, and PAD, in individuals aged 30 – 79. Such differences are potentially important to
estimate the effect of modifying particular risk factors on various
CVD outcomes. Results of individual studies are difficult to compare
due to differences in outcome assessment and population characteristics. For example, incidence rates of different CVD manifestations
may differ according to age (e.g. mean age at haemorrhagic stroke
is lower than that at ischaemic stroke). Our cohort comprised relatively young individuals. Because participants were censored after
the occurrence of the first event, it is possible that our study does
not adequately reflect the associations between particular risk factors
and CVD types which tend to occur only at older ages. However, the
Oxford Vascular Study, a prospective population-based cohort study
of over 90 000 individuals from the UK, did not support the assumption of differences in the age at which patients experience a first occurrence of CAD, stroke, or PAD events.18
Implications for clinical trials
Knowledge of the relative importance of particular risk factors to
various CVD types is important in selecting appropriate outcome
measures for clinical trials. Currently, most large-scale clinical trials
employ composite endpoints such as major adverse cardiac event
or major cardiovascular event. Heterogeneity in the definitions of
these composite outcome measures may lead to different results
and conclusions on the efficacy of study interventions and could
lead to over- or underestimation of the effect on specific CVD
types. This observation is particularly relevant given the fact that
several large-scale clinical trials are underway on new classes of
drugs with a potentially large impact on clinical practice, such as
PCSK9 inhibitors. Our results on the substantial differences in the
strength of the associations between particular risk factors and specific CVD outcomes could suggest that differences may exist across
specific CVD outcomes depending on the specific risk factors targeted by the study intervention. For example, although our results
demonstrate that SBP is particularly strongly associated with PAD,
this outcome measure is only rarely included in clinical trials on
the effect of antihypertensive therapies.
885
Impact of atherosclerotic risk factors on different arterial territories
Figure 1 Forest plot of HRs (95% CIs) for highest vs. lowest quartiles of LDL-c and SBP, and current- vs. never-smoking. Depicted are the HRs
(95% confidence intervals), adjusted for age, sex, HDL-c, BMI, diabetes status, LDL-c, SBP, and smoking status. HR, hazard ratio; SBP, systolic blood
pressure; CI, confidence interval; CAD, coronary artery disease; CVD, cardiovascular disease; LDL-c, low-density lipoprotein cholesterol; PAD,
peripheral arterial disease.
Implications for cardiovascular disease
prevention
The heterogeneity in the association between particular risk factors
and specific atherosclerotic CVD types demonstrated in the current
study could improve the selection of high-risk patients for
population-based screening programmes. For example, the particularly strong association between current-smoking and incident AAA
supports the notion that AAA screening may be particularly worthwhile in smoking patients. Furthermore, although we did not study
the effect of modifying risk factors, our results could suggest that differences may exist in the efficacy of improving specific risk factors
across CVD types. Population-based prevention programmes commonly address one particular risk factor. Favourably altering one
atherosclerotic risk factor would decrease the incidence of several
associated CVD types. However, based on the substantial heterogeneity in the associations demonstrated in our study, it could be
speculated that the effects of interventions on different atherosclerotic CVD types may vary largely depending on the risk factors addressed. Future intervention studies are required to test this
hypothesis.
could possibly be explained by differences in the relative importance of particular risk factors in the pathophysiology of specific
CVD types. For example, the strong association between SBP
and ischaemic stroke might be explained by the relationship
between hypertension and cerebral small vessel disease or atrial
fibrillation.19,20 In addition, the weak associations between AAA
and both SBP and LDL-c could imply that different mechanisms
may be involved in the pathogenesis of aneurysms and atheroocclusive disease.5,21,22 Furthermore, the different associations
between LDL-c and CAD vs. ischaemic stroke could support
the hypothesis from autopsy studies which have suggested morphological differences in plaques between intra- and extracranial
arteries.23 Smoking was a particularly strong risk factor for AAA
and PAD. Recent studies indicated that nicotine may upregulate
matrix metalloproteinases, which are implicated in aneurysm
formation.24,25 Also, the relatively strong association with both
AAA and PAD could indicate a more profound effect of smoking
on the peripheral arteries when compared with the coronary and
cerebral circulation.10
Strengths and weaknesses
Pathophysiology
The observed heterogeneity in the associations between particular risk factors and various atherosclerotic CVD manifestations
To our knowledge, our study is the first to investigate the strength of
the associations between LDL-c, SBP, and smoking and the various
studied CVD outcomes in a single cohort. Studying associations
886
R.M. Stoekenbroek et al.
Table 4
Hazard ratios for cardiovascular outcomes by LDL-c quartiles
LDL-c quartiles
..................................................................................................
P-value*
P-value**
1
2
3
4
<3.24
3.25–3.88
3.89–4.59
>4.60
Model 1
Model 2
1.00
1.00
1.43 (1.26– 1.64)
1.22 (1.07– 1.39)
1.73 (1.53– 1.97)
1.45 (1.27– 1.65)
2.19 (1.94– 2.48)
1.71 (1.51– 1.94)
,0.001
,0.001
,0.001
,0.001
Model 3
1.00
1.20 (1.05– 1.37)
1.40 (1.23– 1.59)
1.63 (1.44– 1.86)
,0.001
,0.001
Model 1
Model 2
1.00
1.00
1.57 (1.07– 2.30)
1.27 (0.87– 1.86)
1.57 (1.07– 2.31)
1.23 (0.84– 1.80)
2.01 (1.39– 2.89)
1.32 (0.91– 1.91)
,0.001
0.21
,0.001
0.40
Model 3
1.00
1.26 (0.86– 1.85)
1.21 (0.83– 1.78)
1.28 (0.88– 1.86)
0.28
0.53
Haemorrhagic stroke (n ¼ 117)
Model 1
1.00
Range, mmol/L
...............................................................................................................................................................................
CAD (n ¼ 2332)
Ischaemic stroke (n ¼ 268)
0.89 (0.52– 1.51)
1.02 (0.62– 1.71)
1.06 (0.64– 1.76)
0.79
0.72
1.00
0.82 (0.49– 1.37)
0.80 (0.48– 1.34)
0.78 (0.47– 1.29)
0.37
0.41
1.00
0.79 (0.47– 1.33)
0.76 (0.45– 1.28)
0.72 (0.43– 1.22)
0.26
0.28
Model 1
1.00
1.33 (0.81– 2.19)
1.33 (0.81– 2.19)
1.63 (1.01– 2.63)
0.06
0.15
Model 2
Model 3
1.00
1.00
1.09 (0.67– 1.80)
1.05 (0.64– 1.74)
1.12 (0.68– 1.84)
1.09 (0.66– 1.81)
1.24 (0.76– 2.02)
1.23 (0.75– 2.00)
0.39
0.40
0.68
0.71
Model 1
Model 2
1.00
1.00
0.91 (0.59– 1.38)
0.80 (0.53– 1.22)
1.48 (1.02– 2.16)
1.21 (0.83– 1.77)
1.60 (1.11– 2.32)
1.24 (0.85– 1.81)
0.002
0.072
,0.001
0.02
Model 3
1.00
0.80 (0.53– 1.22)
1.20 (0.82– 1.77)
1.24 (0.84– 1.82)
0.08
0.02
Any CVD event (n ¼ 3087)
Model 1
1.00
Model 2
Model 3
AAA (n ¼ 143)
PAD (n ¼ 227)
1.37 (1.22– 1.54)
1.66 (1.49– 1.85)
2.07 (1.86– 2.30)
,0.001
,0.001
Model 2
1.00
1.15 (1.03– 1.29)
1.36 (1.21– 1.52)
1.57 (1.41– 1.75)
,0.001
,0.001
Model 3
1.00
1.13 (1.01– 1.27)
1.32 (1.18– 1.48)
1.50 (1.35– 1.68)
,0.001
,0.001
Depicted are the hazard rates (95% CIs) from the unadjusted, the age- and sex-adjusted and the multiple-adjusted model, using the lowest exposure quartile as a reference. Model 1
is unadjusted, Model 2 is age- and sex-adjusted, and Model 3 is multiple-adjusted for age, sex, smoking, BMI, diabetes, HDL-c, and systolic blood pressure.
AAA, abdominal aortic aneurysm; SBP, systolic blood pressure; CI, confidence interval; CAD, coronary artery disease; LDL-c, low-density lipoprotein cholesterol; PAD, peripheral
arterial disease; CVD, cardiovascular disease; BMI, body mass index; HDL-c, high-density lipoprotein cholesterol.
*P-value across quartiles.
**P-value for LDL-c as a continuous variable, per 1 mmol/L increment.
within a single cohort allows for direct comparison of the impact of
the risk factors on the different CVD outcomes. The large sample
size and long duration of follow-up enabled us to study a large number of CVD events. To illustrate the differences in the associations
between particular risk factors and specific CVD outcomes, we divided the CVD outcomes into sub-categories. Dividing the outcomes into their components reduces the number of patients
experiencing particular event types. For example, only 117 participants had a haemorrhagic stroke and 143 participants had an abdominal aneurysm. Although some analyses may have therefore
been underpowered, resulting in wide CIs, not splitting the outcomes may introduce heterogeneity and over- or underestimated
associations for particular event types. However, the low number
of participants experiencing each of the component CVD outcomes
also means that statistical comparison of the differences in the
strengths of the associations across the specific CVD outcomes
was considered inappropriate. Although the results of the likelihood
ratio tests indicate that differences in the strength of the associations
across CVD types do exist, independent confirmation in other cohorts is required, at least for some of the observed associations.
Both sex and age are known to be associated with both the presence of risk factors and CVD outcomes, and are therefore important potential confounders. Ideally, all analyses would have been
stratified according to sex and age categories. Stratified analysis suggests that differences in risk factors—CVD associations may exist
across age groups for some of the studied CVD types (see Supplementary material online, Table S3). However, the limited number of
participants in each stratum seriously undermines the accuracy of
the estimates and precludes definite conclusions. We have therefore not stratified the main analysis for age, but did include age
and sex as covariates in the analyses.
Although other studies demonstrated differences in the association between diabetes mellitus (DM) and various atherosclerotic
CVD types, we did not include DM in the current analysis because
of the small number of diabetic patients.26 Diabetes was ascertained by means of self-report, or diabetes-specific diet or
887
Impact of atherosclerotic risk factors on different arterial territories
Table 5
Hazard ratios for cardiovascular outcomes by systolic blood pressure quartiles
SBP quartiles
..................................................................................................
P-value*
P-value**
1
2
3
4
<122
122– 133
133–146
>146
Model 1
Model 2
1.00
1.00
1.39 (1.20–1.60)
1.05 (0.90–1.22)
2.34 (2.04– 2.67)
1.42 (1.24– 1.63)
3.26 (2.87–3.71)
1.60 (1.39–1.84)
,0.001
,0.001
,0.001
,0.001
Model 3
1.00
0.98 (0.84–1.14)
1.29 (1.12– 1.48)
1.39 (1.21–1.60)
,0.001
,0.001
Model 1
Model 2
1.00
1.00
1.88 (1.14–3.10)
1.42 (0.86–2.35)
2.96 (1.86– 4.71)
1.69 (1.04– 2.74)
5.80 (3.75–8.96)
2.49 (1.57–3.94)
,0.001
,0.001
,0.001
,0.001
Model 3
1.00
1.41 (0.85–2.35)
1.69 (1.04– 2.77)
2.48 (1.55–3.97)
,0.001
,0.001
Haemorrhagic stroke (n ¼ 117)
Model 1
1.00
Range, mmol/L
...............................................................................................................................................................................
CAD (n ¼ 2332)
Ischaemic stroke (n ¼ 268)
1.05 (0.55–2.02)
1.96 (1.11– 3.47)
2.69 (1.56–4.65)
,0.001
,0.001
1.00
0.90 (0.47–1.73)
1.44 (0.79– 2.61)
1.65 (0.93–2.92)
0.03
0.06
1.00
0.96 (0.49–1.85)
1.59 (0.86– 2.93)
1.89 (1.03–3.45)
0.01
0.03
Model 1
1.00
2.17 (1.17–4.02)
2.76 (1.53– 4.99)
3.93 (2.22–6.95)
,0.001
,0.001
Model 2
Model 3
1.00
1.00
1.41 (0.76–2.62)
1.45 (0.78–2.70)
1.23 (0.67– 2.26)
1.29 (0.69– 2.38)
1.26 (0.69–2.30)
1.30 (0.70–2.39)
0.74
0.68
0.66
0.60
Model 1
Model 2
1.00
1.00
1.95 (1.15–3.32)
1.52 (0.90–2.59)
2.45 (1.48– 4.07)
1.55 (0.92– 2.60)
5.83 (3.66–9.27)
3.00 (1.84–4.88)
,0.001
,0.001
,0.001
,0.001
Model 3
1.00
1.50 (0.88–2.55)
1.56 (0.92– 2.63)
2.95 (1.78–4.89)
,0.001
,0.001
Any CVD event (n ¼ 3087)
Model 1
1.00
Model 2
Model 3
AAA (n ¼ 143)
PAD (n ¼ 227)
1.47 (1.29–1.67)
2.43 (2.16– 2.74)
3.74 (3.34–4.19)
,0.001
,0.001
Model 2
1.00
1.10 (0.97–1.25)
1.44 (1.27– 1.62)
1.77 (1.57–1.99)
,0.001
,0.001
Model 3
1.00
1.04 (0.92–1.19)
1.34 (1.18– 1.51)
1.60 (1.41–1.81)
,0.001
,0.001
Depicted are the hazard rates (95% CIs) from the unadjusted, the age- and sex-adjusted and the multiple-adjusted model, using the lowest exposure quartile as a reference. Model 1
is unadjusted, Model 2 is age- and sex-adjusted, and Model 3 is multiple-adjusted for age, sex, smoking, BMI, diabetes, HDL-c, and LDL-c.
AAA, abdominal aortic aneurysm; SBP, systolic blood pressure; CI, confidence interval; CAD, coronary artery disease; PAD, peripheral arterial disease; CVD, cardiovascular
disease.
*P-value across quartiles.
**P-value for SBP as a continuous variable, per 10 mmHg increment.
medication. We cannot exclude the possibility that this may have
caused some underestimation of the actual prevalence of diabetes
in our sample. Event identification through hospital admission data
and death certificates could be prone to misclassification. However, the ascertainment of CAD and stroke has been validated
for this cohort, and imaging studies were performed to ascertain
stroke subtype in the majority of cases.27,28 Specifically, examination of hospital records of EPIC-Norfolk participants with recorded stroke confirmed a definite or probable diagnosis in 93%
of cases, with radiological evidence present in 74% of cases.27 In
addition, inspection of medical records of a sample of patients
with recorded CAD confirmed the diagnosis in 97% of patients
with recorded coronary death and 100% of patients with CAD according to hospital admission data. 28 We only recorded PAD
events leading to hospitalization or death. As only ICD-10-coded
data were available, we could not establish whether validated diagnostic methods were used. The strength of the associations between particular risk factors and different CVD types may have
been influenced by the use of antihypertensive- or lipid-lowering
medication. The effect of reducing blood pressure or LDL-c would
likely be most pronounced on the CVD types with which the risk
factor is most strongly associated. It may therefore be speculated
that the contrasts between the associations might be larger when
taking treatment effects into account.
Patients with previous CVD were excluded from the present analysis, primarily because any risk factor—CVD associations would be
largely affected by therapeutic risk factor modulation or by the
event itself. This makes our findings applicable only to individuals
without previous CVD.
In conclusion, our results demonstrate that the traditional risk
factors, such as LDL-c, SBP, and smoking, show substantial heterogeneity in the associations with different atherosclerotic CVD manifestations. This observation supports the concept of differences in
the pathophysiology of different atherosclerotic CVD manifestations, and insights in these differences are important in estimating
the benefit that could be achieved in prevention programmes.
888
R.M. Stoekenbroek et al.
Table 6
Hazard ratios for cardiovascular outcomes by smoking status
Previous-smoking
N 5 9011
P-value
Current-smoking
N 5 2534
P-value
...............................................................................................................................................................................
CAD (n ¼ 2332)
Model 1
1.72 (1.57–1.88)
,0.001
1.71 (1.51– 1.95)
,0.001
Model 2
Model 3
1.21 (1.10–1.33)
1.19 (1.09–1.31)
,0.001
,0.001
1.76 (1.54– 2.00)
1.77 (1.55– 2.03)
,0.001
,0.001
Model 1
Model 2
1.34 (1.03–1.74)
1.04 (0.79–1.37)
0.03
0.77
1.77 (1.24– 2.52)
2.12 (1.48– 3.04)
0.002
,0.001
Model 3
1.05 (0.80–1.39)
0.71
2.14 (1.49– 3.06)
,0.001
Haemorrhagic stroke (n ¼ 117)
Model 1
1.07 (0.72–1.59)
Ischaemic stroke (n ¼ 268)
Model 2
Model 3
AAA (n ¼ 143)
0.73
1.37 (0.79– 2.36)
0.26
0.93 (0.62–1.39)
0.73
1.52 (0.88– 2.65)
0.13
0.97 (0.65–1.44)
0.87
1.45 (0.84– 2.51)
0.19
Model 1
4.28 (2.67–6.85)
,0.001
7.58 (4.51– 12.74)
,0.001
Model 2
Model 3
2.43 (1.52–3.89)
2.47 (1.54–3.96)
,0.001
,0.001
7.76 (4.57– 13.18)
7.66 (4.50– 13.04)
,0.001
,0.001
Model 1
Model 2
1.79 (1.30–2.45)
1.26 (0.91–1.73)
,0.001
0.16
4.46 (3.16– 6.29)
4.67 (3.31– 6.58)
,0.001
,0.001
Model 3
1.27 (0.92–1.75)
0.14
4.66 (3.29– 6.61)
,0.001
PAD (n ¼ 227)
Any CVD event (n ¼ 3087)
Model 1
1.74 (1.61–1.88)
,0.001
2.07 (1.86– 2.30)
,0.001
Model 2
1.23 (1.13–1.33)
,0.001
2.25 (2.02– 2.51)
,0.001
Model 3
1.22 (1.13–1.33)
,0.001
2.30 (2.06– 2.56)
,0.001
Depicted are the hazard rates (95% CIs) from the unadjusted, the age- and sex-adjusted and the multiple-adjusted model, using non-smoking status as a reference. Model 1 is
unadjusted, Model 2 is age- and sex-adjusted, and Model 3 is multiple-adjusted for age, sex, BMI, diabetes, LDL-c, HDL-c, and SBP.
AAA, abdominal aortic aneurysm; SBP, systolic blood pressure; CI, confidence interval; CAD, coronary artery disease; PAD, peripheral arterial disease; CVD, cardiovascular
disease.
Author’s contributions
R.M.S., S.M.B., G.K.H., A.H.Z., R.J.G.P.: performed statistical analysis.
S.M.B., R.L., N.J.W., K.-T.K.: handled funding and supervision. S.M.B.,
R.L., N.J.W., K.-T.K.: acquired the data. S.M.B., R.L., N.J.W., K.-T.K.:
conceived and designed the research. R.M.S., S.M.B., G.K.H.,
R.J.G.P.: drafted the manuscript. S.M.B., R.L., G.K.H., A.H.Z.,
N.J.W., K.-T.K., R.J.G.P.: made critical revision of the manuscript
for key intellectual content.
Supplementary material
Supplementary material is available at European Heart Journal
online.
Funding
EPIC-Norfolk is supported by program grants from the Medical
Research Council UK (MRC G0401527, MRC G0701863, and
MRC G1000143) and Cancer Research UK (CRUK 8257).
Conflict of interest: none declared.
References
1. D’Agostino RB Sr, Vasan RS, Pencina MJ, Wolf PA, Cobain M, Massaro JM,
Kannell WB. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation 2008;117:743 –753.
2. Fifth Joint Task Force of the European Society of Cardiology. European Guidelines
on cardiovascular disease prevention in clinical practice (version 2012): the Fifth
Joint Task Force of the European Society of Cardiology and Other Societies on
Cardiovascular Disease Prevention in Clinical Practice (constituted by representatives of nine societies and by invited experts). Atherosclerosis 2012;223:1– 68.
3. Kannel WB, McGee D, Gordon T. A general cardiovascular risk profile: the Framingham Study. Am J Cardiol 1976;38:46 –51.
4. Jackson R, Lawes CM, Bennett DA, Milne RJ, Rodgers A. Treatment with drugs to
lower blood pressure and blood cholesterol based on an individual’s absolute cardiovascular risk. Lancet 2005;365:434 –441.
5. Lederle FA, Nelson DB, Joseph AM. Smokers’ relative risk for aortic aneurysm
compared with other smoking-related diseases: a systematic review. J Vasc Surg
2003;38:329 – 334.
6. Lewington S, Clarke R, Qizilbash N, Peto R, Collins R; Prospective Studies Collaboration. Age-specific relevance of usual blood pressure to vascular mortality: a
meta-analysis of individual data for one million adults in 61 prospective studies. Lancet 2002;360:1903 –1913.
7. Lewington S, Whitlock G, Clarke R, Sherliker P, Emberson J, Halsey J, Qizilbash N,
Peto R, Collins R, Prospective Studies Collaboration. Blood cholesterol and vascular mortality by age, sex, and blood pressure: a meta-analysis of individual data from
61 prospective studies with 55 000 vascular deaths. Lancet 2007;370:1829 –1839.
8. Lu L, Mackay DF, Pell PJ. Meta-analysis of the association between cigarette smoking
and peripheral arterial disease. Heart 2014;100:414 –423.
889
Impact of atherosclerotic risk factors on different arterial territories
9. Rapsomaniki E, Timmis A, George J, Pujades-Rodriguez M, Shah AD, Denaxas S,
White IR, Caulfield MJ, Deanfield JE, Smeeth L, Williams B, Hingorani A,
Hemingway H. Blood pressure and incidence of twelve cardiovascular diseases: lifetime risks, healthy life-years lost, and age –specific associations in 1.25 million people. Lancet 2014;383:1899 – 1911.
10. Fowkes FG, Housley E, Riemersma RA. Smoking, lipids, glucose intolerance, and
blood pressure as risk factors for peripheral atherosclerosis compared with ischaemic heart disease in the Edinburgh Artery Study. Am J Epidemiol 1992;135:
331– 340.
11. Qizilbash N. Are risk factors for stroke and coronary disease the same? Curr Opin
Lipidol 1998;9:325 –328.
12. Day N, Oakes S, Luben R, Khaw KT, Bingham S, Welch A, Wareham N. EPICNorfolk: study design and characteristics of the cohort. European Prospective Investigation of Cancer. Br J Cancer 1999;80(Suppl 1):95 –103.
13. Gooley TA, Leisenring W, Crowley J, Storer BE. Estimation of failure probabilities in
the presence of competing risks: new representations of old estimators. Stat Med
1999;18:695 –706.
14. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999;94:496 –509.
15. Kannel WB, McGee DL. Update on some epidemiologic features of intermittent
claudication: the Framingham Study. J Am Geriatr Soc 1985;33:13–18.
16. Peters SA, Huxley RR, Woodward M. Smoking as a risk factor for stroke in women
compared with men: a systematic review and meta –analysis of 81 cohorts, including 3 980 359 individuals and 42 401 strokes. Stroke 2013;44:2821 –2828.
17. Wang X, Dong Y, Qi X, Huang C, Hou L. Cholesterol levels and risk of hemorrhagic
stroke: a systematic review and meta –analysis. Stroke 2013;44:1833 –1839.
18. Rothwell PM, Coull AJ, Silver LE, Fairhead JF, Giles MF, Lovelock CE, Redgrave JN,
Bull LM, Welch SJ, Cuthbertson FC, Binney LE, Gutnikov SA, Anslow P, Banning AP,
Mant D, Mehta Z. Population-based study of event-rate, incidence, case fatality, and
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
mortality for all acute vascular events in all arterial territories (Oxford Vascular
Study). Lancet 2005;366:1773 –1783.
Sörös P, Whitehead S, Spence JD, Hachinski V. Antihypertensive treatment can
prevent stroke and cognitive decline. Nat Rev Neurol 2013;9:174 –178.
Benjamin EJ, Levy D, Vaziri SM, D’Agostino RB, Belanger AJ, Wolf PA. Independent
risk factors for atrial fibrillation in a population-based cohort. The Framingham
Heart Study. JAMA 1994;271:840–844.
Nordon IM, Hinchliffe RJ, Loftus IM, Thompson MM. Pathophysiology and epidemiology of abdominal aortic aneurysms. Nat Rev Cardiol 2011;8:92–102.
Sakalihasan N, Limet R, Defawe OD. Abdominal aortic aneurysm. Lancet 2005;365:
1577 –1589.
Ritz K, Denswil NP, Stam OC, van Lieshout JJ, Daemen MJ. Cause and mechanisms
of intracranial atherosclerosis. Circulation 2014;130:1407 –1414.
Curci JA. Effect of smoking on abdominal aortic aneurysms: novel insights through
murine models. Future Cardiol 2007;3:457 –466.
Wang S, Zhang C, Zhang M, Liang B, Zhu H, Lee J, Viollet B, Xia L, Zhang Y,
Zou MH. Activation of AMP-activated protein kinase alpha2 by nicotine instigates
formation of abdominal aortic aneurysms in mice in vivo. Nat Med 2012;18:
902 –910.
Beckman JA, Paneni F, Cosentino F, Creager MA. Diabetes and vascular disease:
pathophysiology, clinical consequences, and medical therapy: part II. Eur Heart J
2013;34:2444 –2452.
Sinha S, Myint PK, Luben RN, Khaw KT. Accuracy of death certification and hospital
record linkage for identification of incident stroke. BMC Med Res Methodol 2008;8:
74.
Boekholdt SM, Peters RJ, Day NE, Luben R, Bingham SA, Wareham NJ, Hack CE,
Reitsma PH, Khaw KT. Macrophage migration inhibitory factor and the risk of
myocardial infarction or death due to coronary artery disease in adults without
prior myocardial infarction or stroke: the EPIC-Norfolk Prospective Population
study. Am J Med 2004;117:390 –397.