Primary Preventive Potential for Stroke by Avoidance of
Major Lifestyle Risk Factors
The European Prospective Investigation Into Cancer and
Nutrition-Heidelberg Cohort
Kaja Tikk, PhD; Disorn Sookthai, MSc; Stefano Monni, PhD; Marie-Luise Gross, MD;
Christoph Lichy, MD, PhD; Manja Kloss, MD; Rudolf Kaaks, PhD
Downloaded from http://stroke.ahajournals.org/ by guest on June 15, 2017
Background and Purpose—Because primary prevention of stroke is a priority, our aim was to assess the primary preventive
potential of major lifestyle risk factors for stroke in middle-aged women and men.
Methods—Among 23 927 persons, 551 (195 women and 356 men) had a first diagnosis of stroke during an average followup of 12.7 years. Using Cox proportional hazards models, we estimated the associations of adiposity, smoking, physical
activity, alcohol consumption, and diet with risk of developing stroke. A competing risk model built from cause-specific
proportional hazards models accounting for concurrent risk of death was used to calculate relative and absolute reductions
in stroke occurrences that could have been achieved by maintaining a healthy lifestyle pattern.
Results—Obesity, smoking, alcohol consumption, diet, and physical inactivity were each identified as modifiable lifestyle risk
factors for stroke. About 38% of stroke cases were estimated as preventable through adherence to a healthy lifestyle profile
(never smoking, maintaining optimal body mass index and waist circumference, performing physical exercise, consuming
a moderate quantity of alcohol, and following a healthy dietary pattern). Age-specific estimates of 5-year incidence rates
for stroke in the actual cohort and in a hypothetical, comparable cohort of individuals following a healthy lifestyle would
be reduced from 153 to 94 per 100 000 women and from 261 to 161 per 100 000 men for the age group 60 to 65 years.
Conclusions—Our analysis confirms the strong primary prevention potential for stroke based on avoidance of excess body
weight, smoking, heavy alcohol consumption, unhealthy diet, and physical inactivity. (Stroke. 2014;45:2041-2046.)
Key Words: epidemiology ◼ lifestyle ◼ prevention ◼ prospective study ◼ risk factors ◼ stroke
L
arge epidemiological studies have identified several primary modifiable risk factors for stroke, including smoking, overall and abdominal obesity, alcohol consumption, diet,
and physical inactivity.1–5 In addition, prospective cohort studies have provided estimates on the overall reduction in stroke
risk that may be obtained by maintaining a healthy lifestyle.6,7
Because primary prevention of stroke is a priority, knowledge is required about the primary risk factors that should be
targeted through generalized prevention campaigns and about
the overall part of stroke occurrences that could be prevented.
In this context, it is relevant not only to estimate relative
increases or reductions in the occurrence of stroke, in terms
of relative risk and population attributable fractions, but also
to estimate the effects that prevention measures may have on
individuals’ absolute risks in terms of age-specific incidence
rates and lifetime cumulative risks. The estimation of agespecific incidence rates, overall or for individuals with lowor high-risk lifestyle profiles, allows comparisons not only
within a given study cohort but also across study populations
and can underscore to both policymakers and individuals the
importance that primary prevention can have for avoidance of
major disease outcomes.
Here, we present results from the European Prospective
Investigation into Cancer and Nutrition (EPIC)-Heidelberg, a
cohort of middle-aged men and women recruited around the
town of Heidelberg, in South-West Germany. In particular, we
examined associations of modifiable lifestyle factors with the
risk of stroke and quantified both relative and absolute reductions in stroke occurrences that could be achieved by maintaining a comprehensive, healthy lifestyle pattern.
Study Population and Methods
EPIC-Heidelberg is a prospective cohort study that is a part of a
large-scale Europe-wide study, the EPIC. The full EPIC-Heidelberg
cohort consists of 25 540 persons comprised of 11 928 men aged
40 to 64 years and 13 612 women aged 35 to 64 years, who were
Received January 30, 2014; final revision received April 8, 2014; accepted April 17, 2014.
From the Division of Cancer Epidemiology, German Cancer Research Centre, Heidelberg, Germany (K.T., D.S., S.M., M.-L.G., R.K.); Department
of Neurology, Memmingen Hospital, Memmingen, Germany (C.L.); and Department of Neurology, Neurological Clinic, University of Heidelberg,
Heidelberg, Germany (M.K.).
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.
114.005025/-/DC1.
Correspondence to Rudolf Kaaks, PhD, Division of Cancer Epidemiology, German Cancer Research Center, Im Neuenheimer Feld 581, 69120
Heidelberg, Germany. E-mail [email protected]
© 2014 American Heart Association, Inc.
Stroke is available at http://stroke.ahajournals.org
DOI: 10.1161/STROKEAHA.114.005025
2041
2042 Stroke July 2014
recruited between 1994 and 1998 from the general population around
Heidelberg. All participants gave informed consent at the study entry, which was approved by the ethical committee of Heidelberg
University Medical School. A detailed description of the study design, the data collected at recruitment, and during prospect followup for the EPIC-Heidelberg was previously published.8 Briefly, at
baseline, all individuals completed a general questionnaire and a
computer-guided interview about basic demographic factors, prevalent diseases, and lifestyle factors. For all study participants, weight,
height, hip, and waist circumferences were measured by trained study
nurses, following a standardized protocol. Detailed description of
lifestyle factors and classification of variables used in this study is
described in the online-only Data Supplement.
Prospective Ascertainment of Stroke Cases
Downloaded from http://stroke.ahajournals.org/ by guest on June 15, 2017
Since baseline recruitment, at regular 3-year intervals information
about incident diseases was collected by means of a self-administered
questionnaire, and the prospective ascertainment of incident stroke
cases was complemented by regular record linkages to the University
Clinics of Heidelberg and Mannheim. Information on vital status was
collected through municipal population registries. All cases of stroke
were systematically verified by a trained study physician and in collaboration with a neurologist, who jointly reviewed relevant medical
records and official death certificates. All cases were then coded according to the International Classification of Diseases, 10th Revision
(ICD-10) and classified as ischemic stroke (ICD-10 I63), intracerebral or subarachnoid hemorrhage (ICD-10 I61, ICD-10 I60), or undetermined stroke (ICD-10 I64). Only first-ever verified stroke cases
with a diagnosis date were considered as the confirmed incident cases
of stroke. This study reports follow-up data until December 2009 (4
rounds of follow-up). Participants who had reported a previous diagnosis of stroke at baseline recruitment were excluded (n=238).
Statistical Analysis
All analyses were performed separately for men and women. We estimated the preventive potentials of major lifestyle risk factors for
stroke by looking at the reduction in stroke risk, at both the individual
and the population level. This was done through fitting multivariable
Cox models, which were then combined in a competing risk framework, considering death before occurrence of stroke as the competing
event. Age was used as the underlying time scale, modeling delayed
entry by left truncation. A detailed description of the statistical analysis used in this study is reported previously9 and described in the
online-only Data Supplement. All analyses were performed with R
(the R foundation for Statistical Computing, Vienna, Austria).
Results
During an average follow-up period of 12.7 years, 551 persons including 356 men and 195 women developed stroke.
Furthermore, 814 cases of death occurred among men and 436
among women without previous occurrence of stroke. General
characteristics of the study population at baseline and number
of observed stroke events are presented in Table 1.
In Cox proportional hazards models, anthropometric indices for both general obesity (body mass index) and abdominal
obesity (waist circumference) were related with an increased
risk of stroke in men and women, although in multivariable
models these associations were no longer significant, with
the exception of abdominal obesity among women (Table 2).
Regarding physical activity, women engaging in any level of
physical activity, compared with the inactive category of our
combined physical activity index, showed a reduced risk of
stroke, which persisted after controlling for the other variables
in the study. In men, by contrast, the risk estimates by different levels of physical activity categories were weaker and
mostly not statistically significant. Irrespective of controlling
for other lifestyle factors, subjects who reported current smoking at the time of recruitment had an ≤2-fold risk increase in
risk of stroke compared with never smokers, among both men
and women. For alcohol, no statistically significant association between average lifetime consumption levels and stroke
risk was observed among women. Among men, those with the
highest level of lifetime mean alcohol consumption (>60 g/d)
had a significant 57% increase in risk (hazard ratio=1.57; 95%
confidence interval, 1.11–2.23) compared with the reference
group, although this association was reduced after adjustment
for the other lifestyle factors. The healthiest diet score was
inversely associated with stroke risk in men but not in women.
Absolute risk models, accounting for the competing risk
of death, predicted 316 stroke cases among the men in our
cohort (compared with 356 cases actually observed) and 180
stroke cases among the women (195 actually observed) during
the follow-up period. The time-truncated concordance index
C(t)—a measure for the predictive ability of our competing
risk model—at ages 50, 55, 60, and 65 was estimated as 66.9,
65.5, 62.1, and 60.2 for the men and 58.7, 57.6, 60.3, and 58.6
for the women, respectively (data not shown).
Figure 1A and 1B shows the cumulative incidence function
of stroke using the competing risk model. For a 42-year-old
man with a high-risk profile (current smoker, obese with large
waist circumference, physically inactive, with high alcohol
consumption, and unhealthy diet score), who had no stroke
to date, the model predicted a 13.1% absolute risk of developing stroke by the age of 75 years. Likewise, for a 38-year-old
woman with a high-risk profile, the predicted absolute risk
was 17.2%. By comparison, for a man or a woman with the
healthiest risk profile (never smoking, optimal body weight
[body mass index] and waist circumference, physically active,
consuming a moderately low quantity of alcohol, and having
a healthy diet score), the estimated absolute risks to develop
stroke by age 75 years were 5.3% and 2.7%, respectively.
In terms of overall preventable fraction, the predicted number of stroke cases would be reduced by 21.5% for men and
27.2% for women if all persons in our study population had
maintained an optimal body mass index and waist circumference (Table in the online-only Data Supplement). About
37.0% of stroke cases among men and 37.8% of stroke cases
among women could have been prevented if the study participants had lived with the healthy lifestyle profile as defined
above. It is important to point out that the respective reductions of stroke cases without inclusion of diet in our models
would be 32.6% for men and 42.8% for women, respectively.
Figure 2A and 2B shows the substantial disparity in the predicted 5-year incidence rates for stroke in the actual study population and in a hypothetical, comparable cohort with the healthy
lifestyle profile, which would be reduced from 261 to 161 per
100 000 (38.3% reduction) men and from 153 to 94 per 100 000
(38.5% reduction) women in the age group of 60 to 65 years.
Discussion
The present prospective study confirms that obesity, smoking, heavy alcohol consumption, unhealthy diet, and physical
inactivity are significant lifestyle risk factors for stroke among
middle-aged men and women. For the actual study population,
Tikk et al Primary Preventive Potential of Stroke 2043
Table 1. Baseline Characteristics and Observed Stroke
Events in the Study Population by Sex (Median [10th Lowest
and Highest Percentile Values] or n [%])*
Men (n=11 062)
Women (n=12 865)
52.8 (42.3–62.1)
48.9 (37.7–61.5)
BMI, kg/m2
26.5 (22.7–31.6)
24.5 (20.4–31.6)
Optimal weight <25.0
3461 (31%)
7110 (55%)
Overweight 25.0–29.9
5667 (51%)
3836 (30%)
Obesity ≥30.0
1934 (17%)
1919 (15%)
Age at baseline, y
Excess body weight
Abdominal obesity
Waist circumference, cm
95.0 (84.0–109.0)
79.0 (68.0–97.3)
Optimal <91 (m) <74.7 (w)
3775 (34%)
4294 (33%)
Moderate 91–99.4 (m)
74.7–84.9 (w)
3679 (33%)
4513 (35%)
Large ≥99.5 (m) ≥ 85 (w)
3608 (33%)
4058 (32%)
Downloaded from http://stroke.ahajournals.org/ by guest on June 15, 2017
Alcohol consumption
Lifetime mean alcohol
consumption, g/d
20.1 (4.6–57.8)
4.4 (0.6–17.4)
Low <12 (m) >6 (w)
3271 (30%)
7684 (60%)
Moderately low
12–24 (m) 6–11 (w)
3436 (31%)
2799 (22%)
Moderately high
25–59 (m) 12–23 (w)
3341 (30%)
1735 (13%)
High ≥60 (m) ≥24 (w)
1014 (9%)
647 (5%)
Never smoker
3372 (30%)
6198 (48%)
Former smoker
5143 (46%)
3749 (29%)
Current smoker
2547 (23%)
2918 (23%)
1152 (10%)
1551 (12%)
Moderately inactive
3702 (33%)
4642 (36%)
Moderately active
3215 (29%)
3677 (29%)
Active
2993 (27%)
2995 (23%)
Unhealthy diet score
<20 (m) <23 (w)
3893 (35%)
4307 (33%)
Intermediate diet score
20–23 (m) 23–27 (w)
3501 (32%)
4849 (38%)
Healthy diet score
>24 (m) >28 (w)
3668 (33%)
3709 (29%)
Smoking status
Physical activity index
Inactive
DASH-style diet score
Stroke subtypes
Ischemic
277 (78%)
143 (73%)
Hemorrhagic
59 (17%)
39 (20%)
Unknown
20 (5%)
13 (7%)
BMI indicates body mass index; DASH, Dietary Approach to Stop Hypertension;
m, men; and w, women.
*The percentages might not add ≤100% because of rounding.
and for a hypothetical, comparable population where all individuals would have avoided excess body weight, abdominal
obesity, smoking, high alcohol consumption, physical inactivity, and unhealthy diet, differences in the predicted 5-year
incidence rates for stroke were observed. Overall, our estimates indicate that ≈38% of stroke cases could have been
prevented if the study participants had lived with the healthy
lifestyle profile as defined above.
Our basic findings for the 2 strongest lifestyle risk factors—
smoking and excess body weight—are much in line with those
from previous studies, which also showed 2-fold and higher
increases in risk among current smokers compared with nonsmokers1,2,5 and increased risks among both men and women
with abdominal obesity, especially, as defined by large waist
circumference.1,3 Furthermore, being a former smoker was not
associated with stroke risk, showing that cessation of smoking
is effective in stroke prevention.5 In contrast to the well-documented influence of smoking and obesity, the relationship of
alcohol consumption with stroke risk is more controversial.
Existing evidence suggests a J-shaped association, where the
consumption of >60 g of alcohol per day may increase the
risk of stroke by >60%, whereas moderate consumption levels of <12 g/d has been associated with a 15% to 20% risk
reduction in comparison with abstainers.1,10 Our data showed
a tendency toward a risk increase for stroke among men with a
history of relatively heavy alcohol consumption (>60 g/d) but
no clear association among women. By contrast, no evidence
for a possible protective effect of light alcohol consumption
against stroke was found, contrary to findings from some previous studies.6,10
The association between stroke risk and physical activity has been studied extensively but with inconsistent results
varying from no or a weak association7,11 to moderately strong
relationships.1,12,13 Some studies also showed sex-specific
associations.12,13 Taken together, however, the results from
these studies do suggest that lack of any physical activity is a
relevant modifiable risk factor for stroke. In our study, women
who reported any level of physical activity above the lowest
inactive level showed a major reduction in the risk of stroke,
in the order of 50%. Our results thus seem to confirm earlier
interpretations that moderately intense levels of physical activity are sufficient to achieve a risk reduction among women and
that more intense activity may not confer any further benefit.12
In line with a previous study, we observed that adherence to a
Dietary Approach to Stop Hypertension–style diet contributes
to the prevention of stroke risk.6 Interestingly, this association
was not present among women in our cohort.
Our combined risk factor analysis indicated that ≈38% of
primary stroke occurrences could have been prevented in our
study population if all study participants had maintained the
healthiest risk profile (optimal body weight/waist circumference, not smoking, moderate alcohol consumption, physically
active, and following a healthy dietary pattern). Similar and
sometimes even greater estimates of lifestyle-related attributable risks have been reported from other prospective studies,
and it has been proposed that ≤60 or even 90% of stroke cases
might be preventable through lifestyle modifications (eg, not
smoking, optimal body mass index, physical activity equivalent to >30 minutes/d of walking, moderate alcohol consumption, and healthy diet), in combination with the avoidance of
other important risk factors, such as hypertension and type-2
diabetes mellitus.2,4,14 For comparison, however, it should be
mentioned that our analysis did not include any measures of
blood pressure, blood lipids, or other preclinical indicators
of stroke risk because such factors would most likely reflect
2044 Stroke July 2014
Table 2. Risk of Stroke Associated with Lifestyle Factors Among Men (m) and Women (w)
Men
Women
No. of
Stroke
Univariable Model
HR (95% CI)
Multivariable
Model*
HR (95% CI)
<25.0
77
1.00
1.00
25.0–29.9
206
1.43 (1.10–1.86) 1.33 (0.96–1.83)
65
1.07 (0.77–1.49) 0.82 (0.54–1.24)
≥30
73
1.44 (1.05–1.99) 1.22 (0.79–1.89)
49
1.57 (1.09–2.25) 1.03 (0.63–1.76)
No. of Univariable Model
stroke
HR (95% CI)
Multivariable
Model*
HR (95% CI)
BMI categories, kg/m2
81
1.00
1.00
Waist circumference, cm
<91 (m) <74.7 (w)
87
1.00
1.00
35
1.00
1.00
91–99.4 (m) 74.7–84.9 (w)
119
1.21 (0.92–1.60) 1.04 (0.75–1.43)
63
1.24 (0.82–1.88) 1.35 (0.87–2.10)
≥99.5 (m) ≥85 (w)
150
1.45 (1.11–1.90) 1.18 (0.82–1.70)
97
1.79 (1.20–2.67) 1.92 (1.11–3.30)
Physical activity index
Downloaded from http://stroke.ahajournals.org/ by guest on June 15, 2017
Inactive
54
1.00
1.00
52
1.00
1.00
Moderately inactive
109
0.71 (0.51–0.98) 0.76 (0.54–1.05)
62
0.47 (0.33–0.69) 0.49 (0.34–0.71)
Moderately active
95
0.74 (0.53–1.04) 0.81 (0.57–1.13)
47
0.49 (0.33–0.73) 0.53 (0.35–0.79)
Active
98
0.84 (0.60–1.17) 0.92 (0.66–1.29)
34
0.47 (0.30–0.73) 0.50 (0.32–0.77)
Smoking status
Never smoker
101
Former smoker
154
1.00 (0.77–1.28) 0.98 (0.76–1.27)
1.00
1.00
39
98
1.00 (0.69–1.45) 0.87 (0.59–1.26)
1.00
1.00
Current smoker
101
2.15 (1.63–2.85) 1.63 (1.23–2.17)
58
2.61 (1.87–3.65) 2.04 (1.46–2.88)
Lifetime mean alcohol consumption, g/d
<12 (m) >6 (w)
100
12–24 (m) 6–11 (w)
111
1.04 (0.79–1.36) 1.00 (0.76–1.31)
1.00
1.00
117
42
1.03 (0.72–1.47) 1.09 (0.76–1.56)
1.00
1.00
25–59 (m) 12–23 (w)
99
0.96 (0.73–1.27) 0.88 (0.66–1.16)
23
0.98 (0.63–1.54) 1.01 (0.64–1.59)
≥60 (m) ≥24 (w)
46
1.57 (1.11–2.23) 1.30 (0.91–1.86)
13
1.40 (0.79–2.49) 1.31 (0.73–2.34)
DASH-style diet score
Unhealthy diet score
139
Intermediate diet score
125
0.91 (0.72–1.16) 0.97 (0.76–1.23)
1.00
1.00
117
78
1.00 (0.72–1.40) 1.14 (0.82–1.60)
1.00
1.00
Healthy diet score
92
0.58 (0.45–0.75) 0.68 (0.52–0.89)
55
0.89 (0.62–1.27) 1.16 (0.79–1.68)
BMI indicates body mass index; CI, confidence interval; DASH, Dietary Approach to Stop Hypertension; and HR, hazard ratio.
*Multivariable model adjusted for BMI, waist circumference, physical activity, smoking status, alcohol consumption, and diet
where appropriate.
intermediate factors on the pathway between lifestyle and disease development, whereas our primary aim was to estimate
the potential for stroke prevention specifically with respect the
up-front, primary lifestyle risk factors.
With regard to incidence rates, we estimated average agespecific risks in our study cohort to be ≈261/100 000 for
men and 153/100 000 for women in the age range of 60 to
65, which is relatively comparable to the rates reported from
regional stroke registries in different German subpopulations
(eg, Ludwigshafen and Erlangen). Within Germany, estimates
from these registries indicate corresponding rates for 55- to
64-year-old persons varying from 188 to 368 for men and
from 203 to 240 for women.15,16 Studies in Europe and Asia
indicate ≤8-fold variation in 1-year incidence rates ranging
from ≈100/100 000 in Italy17 to 330 to 433/100 000 in The
Netherlands and Japan18,19 and with relatively extreme values
as high as from 500 to 800/100 000 in Ukraine and Russia20,21
for men and women aged 55 to 64 years. The incidence rates
in Germany and in our cohort thus lay within the midrange of
the rates reported worldwide.
Our estimated stroke incidence rates in a hypothetical population of men and women with the healthy lifestyle profile
were on average 30% to 40% lower compared with the rates
in the actual study population. The 5-year incidence rates for
women and men with the healthiest lifestyle profile are comparable to the mean rates observed in populations with low
incident rates, such as low-risk regions in Italy17 and France.22
With regard to long-term absolute risks as a function of lifestyle profiles, the absolute risk to develop stroke by age 75
years would diminish to 5.3% for men and 2.7% for women
if lifestyle factors were modified to the healthiest level. It is
worth mentioning that, in a parallel analysis, we and others
found similar or even greater differences in the long-term
absolute risks of developing myocardial infarction, essentially
depending on the same risk factors as those considered in the
present analysis.9,23 Taken together, the high absolute lifetime
risks for stroke, myocardial infarction, and other chronic diseases among subjects with unhealthy lifestyle patterns, and
the large difference with those who had a low-risk pattern,
provide a compelling argument to health policymakers, as
Tikk et al Primary Preventive Potential of Stroke 2045
Downloaded from http://stroke.ahajournals.org/ by guest on June 15, 2017
Figure 1. Absolute risk to develop stroke for a man with a given lifestyle profile who is disease free at age 42 years (A) and for a woman
who is disease free at age 38 years (B). Highest risk: current smoker, obese with large waist circumference, physically inactive, high
alcohol intake, and unhealthy diet score. Average risk for men (based on typical risk profile in our study population): former smoker,
overweight with moderate waist circumference, moderately physically inactive, moderate alcohol consumption, and unhealthy diet score.
Average risk for women: never smoker, optimal weight with moderate waist circumference, moderately physically inactive, low alcohol
consumption, and intermediate diet score. Optimal risk: never smoker, optimal weight and waist circumference, moderately low alcohol
consumption, physically active, and healthy diet score. Obese: with average risk profile but obese and with large waist circumference.
Current smoker: with average risk profile for the other variables.
well as single individuals, to increase efforts for primary prevention and to maintain healthy habits.
Although our study has several strengths—in particular,
its prospective design, a reasonably large number of incident
cases, and detailed clinical verification and coding of stroke
end points—it also has several limitations. In line with conclusions and interpretations from many previous studies, we
assumed that to a large extent the risk factors retained for our
overall risk modeling were genuine and independent primary
determinants of stroke incidence. However, although mutual
adjustments between the risk factors could be made in our
multivariable risk models, residual confounding biases cannot
ever be ruled out entirely. Possible confounding factors that
we could not adjust for in our analysis include, for example,
chronic (eg, work-related) psychological stress as source of
hypertension, socioeconomic determinants of having access
to or making use of regular healthcare and health surveillance,
and use of antihypertensive drugs or other forms of medication (eg, nonsteroidal anti-inflammatory drugs, lipid-lowering
drugs), which, in subsets of cohort, may have been recommended to specifically target increased risk states that may be
intermediate between primary lifestyle factors and stroke as
an end point.
Finally, as in most observational studies, it is likely that
measurement errors caused substantial regression dilution
bias in effect estimates of each of the basic lifestyle factors
considered in this analysis. The assessments of alcohol consumption, diet, smoking habits, and physical activity levels
Figure 2. Five-year stroke incidence rates for men (A) and women (B) predicted in the actual study population and in a hypothetical population with a healthier lifestyle profile (never smoker, optimal body weight and waist circumference, physically active and moderate low
alcohol consumption, healthy diet score).
2046 Stroke July 2014
Downloaded from http://stroke.ahajournals.org/ by guest on June 15, 2017
were all based on self-reports and may not be fully accurate
with respect to true exposure levels. The same applies to our
anthropometric indices of excess weight, which provide only
approximate measures of overall and abdominal adiposity,
which are likely to be the true risk determinants. Also, as in
many other studies, we did not account for possible changes
in smoking or other lifestyle factors during follow-up. Besides
a non-negligible underestimation of relative risks and attributable fractions, random assessment errors would also lead to an
incomplete differentiation between highest and lowest absolute risk estimates for subjects classified by high- and low-risk
factor scores.
In conclusion, our results support previously established
classical lifestyle risk factors for stroke and emphasize the
importance of avoiding high-risk lifestyle factors for the preventions of stroke. Our estimates show that incidence rates for
men and women who adhere to a healthy lifestyle pattern can
be as low as those documented for typical low-risk regions
in Europe. For increased differentiation between the absolute
risks of low- and high-risk individuals, future prospective
studies should invest in increasing the accuracy of assessments of lifestyle factors and body composition, including
repeat measurements over time. From our and other analyses,
it seems that especially smoking and excess body weight are
the 2 major risk factors that should be targeted with greatest
priority for primary prevention strategies.
Acknowledgments
We would like to thank Marcus von Hornung and Christoph Neumann
for their valuable work to database management in the European
Prospective Investigation into Cancer and Nutrition (EPIC)Heidelberg cohort and all the EPIC-Heidelberg cohort participants
for their active participation in the study.
Sources of Funding
This study was funded by the German Federal Ministry of Education
and Research.
Disclosures
None.
References
1. Boden-Albala B, Sacco RL. Lifestyle factors and stroke risk: exercise,
alcohol, diet, obesity, smoking, drug use, and stress. Curr Atheroscler
Rep. 2000;2:160–166.
2. O’Donnell MJ, Xavier D, Liu L, Zhang H, Chin SL, Rao-Melacini P,
et al; INTERSTROKE Investigators. Risk factors for ischaemic and
intracerebral haemorrhagic stroke in 22 countries (the INTERSTROKE
study): a case-control study. Lancet. 2010;376:112–123.
3. Suk SH, Sacco RL, Boden-Albala B, Cheun JF, Pittman JG, Elkind
MS, et al. Abdominal obesity and risk of ischemic stroke. Stroke.
2003;34:1586–1592.
4. Weikert C, Berger K, Heidemann C, Bergmann M, Hoffmann K, KlipsteinGrobusch K, et al. Joint effects of risk factors for stroke and transient ischemic attack in a German population. J Neurol. 2007;254:315–321.
5. Wannamethee SG, Shaper AG, Whincup PH, Walker M. Smoking cessation and the risk of stroke in middle-aged men. JAMA. 1995;274:155–160.
6. Chiuve SE, Rexrode KM, Spiegelman D, Logroscino G, Manson JE,
Rimm EB. Primary prevention of stroke by healthy lifestyle. Circulation.
2008;118:947–954.
7. Kurth T, Moore SC, Gaziano JM, Kase CS, Stampfer MJ, Berger K, et
al. Healthy lifestyle and the risk of stroke in women. Arch Intern Med.
2006;166:1403–1409.
8. Boeing H, Wahrendorf J, Becker N. EPIC-Germany—A source for studies
into diet and risk of chronic diseases. Ann Nutr Metab. 1999;43:195–204.
9. Li K, Monni S, Hüsing A, Wendt A, Kneisel J, Groß ML, et al. Primary
preventive potential of major lifestyle risk factors for acute myocardial
infarction in men: an analysis of the EPIC-Heidelberg cohort. Eur J
Epidemiol. 2014;29:27–34.
10. Reynolds K, Lewis LB, Nolen JD, Kinney GL, Sathya B, He J. Alcohol
consumption and risk of stroke. JAMA. 2003;289:579–588.
11. Evenson KR, Rosamond WD, Cai J, Toole JF, Hutchinson RG, Shahar E,
et al. Physical activity and ischemic stroke risk. The atherosclerosis risk
in communities study. Stroke. 1999;30:1333–1339.
12. Kiely DK, Wolf PA, Cupples LA, Beiser AS, Kannel WB. Physical activity
and stroke risk: the Framingham Study. Am J Epidemiol. 1994;140:608–620.
13. Myint PK, Luben RN, Wareham NJ, Welch AA, Bingham SA, Day NE, et
al. Combined work and leisure physical activity and risk of stroke in men
and women in the European prospective investigation into Cancer-Norfolk
Prospective Population Study. Neuroepidemiology. 2006;27:122–129.
14. Willett WC. Balancing life-style and genomics research for disease prevention. Science. 2002;296:695–698.
15. Kolominsky-Rabas PL, Sarti C, Heuschmann PU, Graf C, Siemonsen S,
Neundoerfer B, et al. A prospective community-based study of stroke in
Germany–the Erlangen Stroke Project (ESPro): incidence and case fatality at 1, 3, and 12 months. Stroke. 1998;29:2501–2506.
16. Palm F, Urbanek C, Rose S, Buggle F, Bode B, Hennerici MG, et al.
Stroke Incidence and Survival in Ludwigshafen am Rhein, Germany: the
Ludwigshafen Stroke Study (LuSSt). Stroke. 2010;41:1865–1870.
17. Janes F, Gigli GL, D’Anna L, Cancelli I, Perelli A, Canal G, et al. Stroke
incidence and 30-day and six-month case fatality rates in Udine, Italy: a population-based prospective study. Int J Stroke. 2013;8(suppl A100):100–105.
18. Wieberdink RG, Ikram MA, Hofman A, Koudstaal PJ, Breteler MM.
Trends in stroke incidence rates and stroke risk factors in Rotterdam, the
Netherlands from 1990 to 2008. Eur J Epidemiol. 2012;27:287–295.
19. Morikawa Y, Nakagawa H, Naruse Y, Nishijo M, Miura K, Tabata M, et
al. Trends in stroke incidence and acute case fatality in a Japanese rural
area: the Oyabe study. Stroke. 2000;31:1583–1587.
20. Mihalka L, Smolanka V, Bulecza B, Mulesa S, Bereczki D. A population
study of stroke in West Ukraine. Stroke. 2001;32:2227–2231.
21. Feigin VL, Wiebers DO, Whisnant JP, O’Fallon WM. Stroke incidence
and 30-day case-fatality rates in Novosibirsk, Russia, 1982 through
1992. Stroke. 1995;26:924–929.
22. Béjot Y, Osseby GV, Aboa-Eboulé C, Durier J, Lorgis L, Cottin Y, et
al. Dijon’s vanishing lead with regard to low incidence of stroke. Eur J
Neurol. 2009;16:324–329.
23.Lloyd-Jones DM, Leip EP, Larson MG, D’Agostino RB, Beiser A,
Wilson PW, et al. Prediction of lifetime risk for cardiovascular disease
by risk factor burden at 50 years of age. Circulation. 2006;113:791–798.
Primary Preventive Potential for Stroke by Avoidance of Major Lifestyle Risk Factors:
The European Prospective Investigation Into Cancer and Nutrition-Heidelberg Cohort
Kaja Tikk, Disorn Sookthai, Stefano Monni, Marie-Luise Gross, Christoph Lichy, Manja Kloss
and Rudolf Kaaks
Downloaded from http://stroke.ahajournals.org/ by guest on June 15, 2017
Stroke. 2014;45:2041-2046; originally published online May 29, 2014;
doi: 10.1161/STROKEAHA.114.005025
Stroke is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright © 2014 American Heart Association, Inc. All rights reserved.
Print ISSN: 0039-2499. Online ISSN: 1524-4628
The online version of this article, along with updated information and services, is located on the
World Wide Web at:
http://stroke.ahajournals.org/content/45/7/2041
Data Supplement (unedited) at:
http://stroke.ahajournals.org/content/suppl/2014/05/29/STROKEAHA.114.005025.DC1
Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published
in Stroke can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial Office.
Once the online version of the published article for which permission is being requested is located, click
Request Permissions in the middle column of the Web page under Services. Further information about this
process is available in the Permissions and Rights Question and Answer document.
Reprints: Information about reprints can be found online at:
http://www.lww.com/reprints
Subscriptions: Information about subscribing to Stroke is online at:
http://stroke.ahajournals.org//subscriptions/
ONLINE SUPPLEMENT
Primary preventive potential for stroke by avoidance of major lifestyle risk
factors: the EPIC-Heidelberg cohort
Supplemental Methods
Classification of lifestyle variables used in the study
Lifestyle data on smoking, body height and weight (BMI), waist circumference, alcohol
drinking, physical activity and diet were considered in this analysis. All participants were
divided into three body-mass-index (BMI) categories according to the standard WHO
definition of optimal weight (<25 kg/m²), overweight (25-29.9 kg/m²), and obesity (≥ 30
kg/m²). In addition, population, sex-specific tertiles were employed to classify subjects into
categories of optimal, middle and high waist circumference (for men < 91, 91-99.4, ≥ 99.5
cm; for women <74.7, 74.7-84.9, ≥85 cm). For physical activity, an overall activity index,
internally referred to as the “Cambridge Index”, was used to calculate a Metabolic Equivalent
(METs) score combining both occupational and recreational activity (hours spent with cycling
and sports), described in greater detail previously1. Based on this MET score, subjects were
assigned to four categories named “inactive”, “moderately inactive”, “moderately active”, and
“active”. The lifetime mean alcohol intake was calculated as a weighted average of alcohol
intake at different ages, with weights equal to the individuals’ average alcohol intakes during
different age periods and four sex-specific categories were derived from baseline data adapted
from the alcohol intake categories reported in the general guidelines for stroke prevention2:
for men; <12, 12-24, 25-59, ≥ 60 g/day: for women; <6, 6-11, 12-23, ≥ 24 g/day. Study
participants were categorized into three groups as never smokers, former smokers and current
smokers.
Regarding habitual diet, all cohort participants filled out a detailed food frequency
questionnaire (FFQ) at baseline. To reduce the number of parameters associated with the
dietary information, several food item categories (eg. consumption of processed meat, red
meat, vegetables, fruits, whole-grain, etc) were created, which were then combined into a
dietary score. A criterion for variable selection was the availability of the necessary data to
construct the score. We have selected to use a diet-quality score, the DASH diet (The Dietary
Approaches to Stop Hypertension). This dietary pattern has been shown to be relevant for
stroke and is recommended to be followed for stroke prevention2-4, mainly through its effect
on blood pressure5,6. The DASH-style diet score was constructed based on diet emphasized or
minimized in the DASH diet, focusing on 8 components: high intake of fruits, vegetables,
nuts and legumes, dairy products, whole grains and low intake of sodium, sweetened
beverages, and red and processed meats. For each component, participants were classified into
quintiles according to their intake ranking and the component scores were summed up to
overall DASH score, ranging from 8 to 40. Based on sex-specific tertiles of this score,
participants were assigned to the following groups: unhealthy, intermediate or healthy dietary
score (for men 8-20, 20-23, 24-40; for women 8-23, 23-27, 28-40, respectively).
Because of the insufficient data among stroke cases, it was not possible to adjust the
models for the use of aspirin and NSAID. As our focus was explicitly on directly modifiable
lifestyle factors, we did not consider hypertension in our risk profile analyses.
1
Statistical analysis
We excluded individuals who were lost to follow-up and had no information on stroke status
(n=1,134), or had incomplete data on any variables required for this analysis (n=241). The
final data set thus consisted of 23,927 individuals (11,062 men and 12,865 women). Cox
proportional hazards models were used to estimate the association of adiposity as expressed
both via the BMI and waist circumference, physical activity, alcohol consumption, diet, and
smoking with the risk of developing stroke. For each variable, proportional hazard models
were first fitted to the data to estimate relative difference in risk among groups of individuals
defined by that variable. All variables were then considered in one Cox model. Age was used
as the underlying time scale, modelling delayed entry by left truncation. The proportional
hazards (PH) assumption was found to hold for all variables in the model (according to the
analysis of the scaled Schoenfeld residuals), except for the univariable model with smoking as
the predictor in the men cohort.
Furthermore, we estimated the preventive potential of stroke by looking at the
reduction in stroke risk, at both the individual and the population level. This was done by
building a competing risk model, considering death without occurrence of stroke as the
competing event. Cause-specific hazards were obtained from multivariable Cox models and
then combined to give estimates for the cumulative incidence function (the absolute risk of
stroke)7. To assess the predictive discrimination of our competing risk model, we employed
the time-truncated concordance index C(t), adapted for competing risk settings8. Specifically,
the inverse probability of censoring weighted (IPCW) estimators of C(t) were evaluated using
Kaplan-Meier estimates of the censoring times. We present estimates obtained by bootstrapcross validation.
The predicted number of incident stroke cases in our cohort was calculated by
summing up each person’s absolute risk to develop stroke in his/her entire actual follow-up
period. We then repeated the above analysis with all lifestyle risk factors being set to their
healthy level to estimate the reduction in the number of stroke cases due to this modification.
Standardized incidence rates were computed from the average time-standardized
probability of developing stroke (per 1 year) over all men and women observed within the
presented 5-year age groups multiplied by 100,000, assuming the person to be alive and
disease free at the beginning of the models’ time range. This analysis was repeated in a
hypothetical, comparable cohort having a healthy lifestyle profile.
All analyses were performed with R (the R foundation for Statistical Computing,
Vienna, Austria), using the libraries rms and survival. For the computation of the C-index, we
adapted functions of the library pec.
Supplemental references
1. Wareham NJ, Jakes RW, Rennie KL, Schuit J, Mitchell J, Hennings S, et al. Validity and
repeatability of a simple index derived from the short physical activity questionnaire used
in the European Prospective Investigation into Cancer and Nutrition (EPIC) study. Public
Health Nutrition. 2003;6:407-413.
2. Goldstein LB, Bushnell CD, Adams RJ, Appel LJ, Braun LT, Chaturvedi S, et al.
Guidelines for the primary prevention of stroke. Stroke. 2011;42:517-584.
2
3. Fung TT, Chiuve SE, McCullough ML, Rexrode KM, Logroscino G, Hu FB. Adherence to
a dash-style diet and risk of coronary heart disease and stroke in women. Arch Intern Med.
2008;168:713-720.
4. Chiuve SE, Rexrode KM, Spiegelman D, Logroscino G, Manson JE, Rimm EB. Primary
Prevention of Stroke by Healthy Lifestyle. Circulation. 2008;118:947-954.
5. Appel LJ, Moore TJ, Obarzanek E, Vollmer WM, Svetkey LP, Sacks FM, et al. A Clinical
Trial of the Effects of Dietary Patterns on Blood Pressure. N Engl J Med. 1997;336:11171124.
6. Sacks FM, Obarzanek E, Windhauser MM, Svetkey LP, Vollmer WM, McCullough M, et
al. Rationale and design of the Dietary Approaches to Stop Hypertension trial (DASH): A
multicenter controlled-feeding study of dietary patterns to lower blood pressure. Ann
Epidemiol. 1995;5:108-118.
7. Andersen PK, Geskus RB, de Witte T, Putter H. Competing risks in epidemiology:
possibilities and pitfalls. Int J Epidemiol. 2012;41:861-870.
8. Wolbers M, Blanche P, Koller MT, Witteman JC, Gerds TA. Concordance for prognostic
models with competing risks. [published online ahead of print February 2]. Biostatistics.
2014. doi:10.1093/biostatistics/kxt059. Access date Feb 2.
3
Supplemental Table I
Expected reduction in the number of incident stroke by modifying lifestyle risk factors
to healthier levels for men (A) and women (B).
Number of incident
stroke
Proportion (%)*
A. Lifestyle modification for men
cases reduced in our
population
Obesity/overweight and large/moderate waist
circumference  optimal weight and waist
68
Obesity and large waist  overweight and moderate
6
waist circumference
21.5%
1.9%
Current /former smokers  never smokers
22
7.0%
Current smoker  former smoker
28
8.9%
High alcohol intake  moderately low alcohol
consumption
5
1.6%
Physically inactive  physically active
1
0.3%
Physically inactive  moderately inactive
9
2.8%
Healthy level without diet**
103
32.6%
Healthy level**
117
37.0%
49
27.2%
Obesity and large waist  overweight and moderate
32
waist circumference
17.8%
B. Lifestyle modification for women
Obesity/overweight and large/moderate waist
circumference  optimal weight and waist
Current /former smokers  never smokers
20
11.1%
Current smoker  former smoker
29
16.1%
High alcohol intake  moderately low alcohol
consumption
1
0.6%
Physically inactive  physically active
21
11.7%
Physically inactive  moderately inactive
22
12.2%
Healthy level without diet**
77
42.8%
Healthy level**
68
37.8%
4
*The predicted total number of incident stroke cases, for the given study cohort without
lifestyle modifications, is 316 for men and 180 for women.
** Never smoker, optimal body weight and waist circumference, physically active,
moderately low alcohol consumption, healthy diet score.
5