Do lifestyle behaviours explain socioeconomic differences in all-cause
mortality, and fatal and non-fatal cardiovascular events? Evidence
from middle aged men in France and Northern Ireland in the PRIME
Study
Woodside, J., Yarnell, J., Patterson, C., Arveiler, D., Amouyel, P., Ferrieres, J., ... Ducimetiere, P. (2012). Do
lifestyle behaviours explain socioeconomic differences in all-cause mortality, and fatal and non-fatal
cardiovascular events? Evidence from middle aged men in France and Northern Ireland in the PRIME Study.
Preventive Medicine, 0(3-4), 0-0. DOI: 10.1016/j.ypmed.2012.01.017
Published in:
Preventive Medicine
Queen's University Belfast - Research Portal:
Link to publication record in Queen's University Belfast Research Portal
General rights
Copyright for the publications made accessible via the Queen's University Belfast Research Portal is retained by the author(s) and / or other
copyright owners and it is a condition of accessing these publications that users recognise and abide by the legal requirements associated
with these rights.
Take down policy
The Research Portal is Queen's institutional repository that provides access to Queen's research output. Every effort has been made to
ensure that content in the Research Portal does not infringe any person's rights, or applicable UK laws. If you discover content in the
Research Portal that you believe breaches copyright or violates any law, please contact [email protected].
Download date:17. Jun. 2017
Preventive Medicine 54 (2012) 247–253
Contents lists available at SciVerse ScienceDirect
Preventive Medicine
journal homepage: www.elsevier.com/locate/ypmed
Do lifestyle behaviours explain socioeconomic differences in all-cause mortality, and
fatal and non-fatal cardiovascular events? Evidence from middle aged men in France
and Northern Ireland in the PRIME Study
J.V. Woodside a,⁎, J.W.G. Yarnell a, C.C. Patterson a, D. Arveiler b, P. Amouyel c, J. Ferrières d, F. Kee a,
A. Evans a, A. Bingham e, P. Ducimetière e
a
Queen's University Belfast, Belfast, United Kingdom
MONICA-Strasbourg, Strasbourg, France
c
MONICA-Lille, Lille, France
d
MONICA-Toulouse, Toulouse, France
e
INSERM U780, Villejuif, France
b
a r t i c l e
i n f o
Available online 28 January 2012
Keywords:
Mortality
Coronary heart disease incidence
Socioeconomic status
Lifestyle behaviours
a b s t r a c t
Objective. To examine the contribution of lifestyle behaviours to the socioeconomic gradient in all-cause
mortality, and fatal and non-fatal cardiovascular events.
Method. 10,600 men aged 50–59 years examined in 1991–1994 in Northern Ireland (NI) and France and
followed annually for deaths and cardiovascular events for 10 years. Baseline smoking habit, physical activity,
and fruit, vegetable, and alcohol consumption were assessed.
Results. All lifestyle behaviours showed marked socioeconomic gradients for most indicators in NI and
France, with the exception of percentage of alcohol consumers in NI and frequency of alcohol consumption
in NI and France.
At 10 years, there were 544 deaths from any cause and 440 fatal and non-fatal cardiovascular events. After
adjustment for country and age, socioeconomic gradients were further adjusted for lifestyle behaviours.
For total mortality, the median residual contribution of lifestyle behaviours was 28% and for cardiovascular
incidence, 41%. When cardiovascular risk factors were considered in conjunction with lifestyle behaviours
these percentages increased to 38% and 67% respectively.
Conclusion. Lifestyle behaviours contribute to the gradient in mortality and cardiovascular incidence between socioeconomic groups, particularly for cardiovascular incidence, but a substantial proportion of
these differentials was not explained by lifestyle behaviours and cardiovascular risk factors.
© 2012 Elsevier Inc. All rights reserved.
Introduction
Socioeconomic differentials in mortality and cardiovascular disease (CVD) incidence have been described in France (Lang and
Ducimetière, 1995, Lang et al., 1997; Yarnell et al., 2005) and other
European countries (Kunst et al., 1998). The possible contribution of
conventional risk factors and lifestyle behaviours to socioeconomic
gradients in cardiovascular mortality is estimated at 15–40% (Davey
Smith et al., 1994; Marmot et al., 2008). However, few studies have
comprehensively examined lifestyle behaviours, whilst it is also
unclear whether their contribution to the socioeconomic gradient is
similar in countries with different lifestyle behaviours. The PRIME
study prospectively examined middle-aged men in France and
Northern Ireland; countries with significant differences in culture
⁎ Corresponding author. Tel.: + 44 2890 632585; fax: + 44 2890 235900.
E-mail address: [email protected] (J.V. Woodside).
0091-7435/$ – see front matter © 2012 Elsevier Inc. All rights reserved.
doi:10.1016/j.ypmed.2012.01.017
and lifestyle. The contribution of lifetime smoking habit to the socioeconomic differential in mortality and CVD incidence has recently
been examined (Yarnell et al., 2011). The median residual contribution of lifetime smoking habit was 21% for total mortality and 18%
for CVD incidence. As a substantial proportion of the socioeconomic
gradient was not explained by smoking behaviour, we have now
examined the contribution of other lifestyle behaviours (alcohol
consumption, fruit, vegetable and juice consumption (FV), physical
activity). We also examined which of the individual lifestyle
behaviours contributed most to the socioeconomic gradient.
Methods
Study Population: The PRIME Study is a multi-centre, prospective cohort
study. Men aged 50–59 years from the general population were recruited
between 1991 and 1993 in Lille, Strasbourg and Toulouse in France, and in
Belfast, NI (Yarnell et al., 2005, 2011). A detailed medical and lifestyle
questionnaire was completed at the baseline examination, including lifetime
248
J.V. Woodside et al. / Preventive Medicine 54 (2012) 247–253
was compliant with the guidelines contained within the Declaration of
Helsinki.
Follow-up: Men were followed up annually by questionnaire for 10 years
and possible cardiovascular events confirmed by a medical committee
(Ducimetière et al., 2001; Yarnell et al., 2011). A total of 317 (3.0%) men
were lost to follow-up after 10 years, 215 (2%) refused to continue participating in the study, and 653 (6.1%) died during this time. Cardiovascular events
included validated myocardial infarction and/or stroke (fatal and non-fatal).
Death certificates were obtained for all men who died and causes of death
were classified using the International Classification of Diseases (ICD) Ninth
Revision.
smoking, physical activity (Wagner et al., 2003) alcohol consumption,
(Ruidavets et al., 2010) and diet (Dauchet et al., 2004).
Socioeconomic variables
Socioeconomic data included length and type of education, home ownership, car ownership, marital and employment status, the number of people in
the household and a summary variable termed “material condition”, based
on home ownership and the number of cars, baths/showers and toilets
(Wagner et al., 2003). Low material condition was defined by rental accommodation with one or fewer cars, baths/showers, and toilets; high material
condition was defined as home ownership with two or more cars and, either
two or more baths/showers, or two or more toilets; the remaining subjects
were classified as living in a mid-(range) material condition.
Statistical methods
Since men with clinical evidence of CVD may have modified their
health-related behaviour, we excluded from further analyses 891
men with clinical evidence of CVD at baseline examination. Recurrent
events were excluded from the analysis by ending follow-up at the
time of a man's first event.
Lifestyle behaviours were compared across socioeconomic variables either by the chi-square test for categorical variables or by Kruskal–Wallis one way ANOVA for continuous variables.
Cox proportional hazards models were used for analysis of mortality
and first cardiovascular events. In light of small numbers, the socioeconomic factors marital status (married/cohabiting versus single/
widowed/divorced), living accommodation (mortgaged/co-owned/
owned versus rented) and economic activity (working versus sick/
disabled/retired/unemployed) were regrouped into two categories
prior to inclusion in the model. For the remaining socioeconomic
Lifestyle behaviour variables
The distribution of pack years of smoking was heavily skewed and was summarised using median and interquartile range. Lifetime smoking was expressed
in five categories: never smoked, smoked other than cigarettes, smokedb 15
cigarette pack years, smoked ≥15 but b 30 cigarette pack years and smoked
≥30 cigarette pack years. Alcohol intake was also expressed as five categories
of consumption; none, 1–128, 129–265, 266–461 and ≥462 ml/week, and
four frequency categories (non-consumers, consumes alcohol in 5–7 days,
3–4 days or 1–2 days per week). FV consumption (portions/day), and total
physical activity (in metabolic equivalent scores/week) were each positively
skewed, and a square root transformation applied to each.
Ethical approval for baseline examination and for follow-up was obtained
at each study centre; all participants gave informed consent and the study
Table 1
Smoking characteristics in subgroups defined by socioeconomic variable for 9709 men from Northern Ireland and France who were free of cardiovascular disease at entry (1991–4).
Northern Ireland
Overall
Marital Status
Married/Cohabiting
Single
Widowed/Divorced
People in household
1 person
2-3 people
4 + people
Accommodation
Mortgaged/Co-owned
Owned
Rented
Cars in household
None
1 car
2 + cars
Years in full-time education
b12 years
12–14 years
15 + years
Educational level
Primary
Secondary
Technical
Higher
Economic activity
Working
Sick/Disabled/Retired
Unemployed
Material condition
Low
Medium
High
France
Current smokers
Ex smokers
Pack years
Current smokers
Ex smokers
Pack years
n
(%)
n
(%)
Median (IQ range)
n
(%)
n
(%)
Median (IQ range)
747
(31.3%)
802
(33.6%)
27.6
(13.0–40.3)
1934
(26.4%)
3352
(45.8%)
17.4
(5.2–31.3)
611
53
83
(29.9%)
(32.3%)
(45.9%)
702
43
57
(34.3%)
(26.2%)
(31.5%)
27.2
30.9
29.6
(12.3–40.1)
(13.5–38.6)
(17.1–43.2)
1615
110
209
(25.1%)
(34.8%)
(37.2%)
3041
98
211
(47.2%)
(31.0%)
(37.5%)
17.0
17.5
23.3
(5.2–30.4)
(4.2–34.0)
(7.2–36.5)
81
409
257
(44.8%)
(30.9%)
(29.1%)
46
454
300
(25.4%)
(34.3%)
(34.0%)
31.5
28.4
24.9
(15.2–42.1)
(13.2–40.7)
(11.1–39.2)
215
1302
416
(38.7%)
(25.7%)
(24.5%)
193
2401
757
(34.7%)
(47.5%)
(44.5%)
21.5
17.6
15.6
(4.7–37.0)
(5.5–31.2)
(4.1–29.8)
368
178
201
(27.7%)
(26.8%)
(51.1%)
474
218
109
(35.6%)
(32.8%)
(27.7%)
24.5
27.0
34.1
(10.2–38.6)
(11.0–40.4)
(21.4–43.5)
237
1123
569
(26.0%)
(23.7%)
(34.3%)
445
2172
725
(48.7%)
(45.9%)
(43.7%)
17.0
15.9
22.8
(4.5–30.0)
(4.6–29.4)
(7.3–36.5)
178
379
190
(52.0%)
(32.9%)
(21.3%)
85
397
318
(24.9%)
(34.5%)
(35.6%)
34.5
28.0
21.2
(22.0–44.1)
(14.2–40.3)
(7.8–37.3)
127
842
960
(39.6%)
(28.0%)
(24.2%)
116
1364
1863
(36.1%)
(45.3%)
(46.9%)
27.3
19.8
15.4
(10.8–44.7)
(6.6–33.1)
(3.9–28.5)
541
137
68
(33.8%)
(27.7%)
(23.1%)
545
170
87
(34.1%)
(34.3%)
(29.6%)
29.6
24.1
18.2
(15.3–41.7)
(7.2–36.4)
(5.8–37.4)
1080
520
334
(26.6%)a
(26.4%)a
(26.0%)a
1820
912
619
(44.8%)a
(46.2%)a
(48.2%)a
20.0
16.7
10.1
(7.3–33.5)
(5.1–30.1)
(1.1–23.3)
249
119
216
140
(40.7%)
(34.6%)
(29.0%)
(22.3%)
205
131
240
207
(33.5%)
(38.1%)
(32.2%)
(32.9%)
34.4
28.4
26.4
19.5
(19.2–44.1)
(13.8–39.6)
(11.6–37.8)
(5.6–36.5)
424
203
668
585
(27.9%)a
(26.5%)a
(27.4%)a
(24.8%)a
664
369
1084
1129
(43.7%)a
(48.2%)a
(44.5%)a
(47.8%)a
23.0
16.4
17.7
13.6
(10.5–36.5)
(4.6–29.3)
(6.2–31.2)
(2.3–27.2)
619
35
93
(29.5%)
(36.1%)
(48.2%)
707
37
58
(33.7%)
(38.1%)
(30.1%)
26.2
31.8
37.2
(10.8–39.2)
(18.4–43.6)
(25.0–49.0)
1480
251
203
(26.2%)
(23.5%)
(34.4%)
2587
505
259
(45.7%)
(47.2%)
(43.9%)
16.5
20.2
24.1
(4.4–30.0)
(8.3–33.9)
(8.8–38.0)
371
144
232
(41.0%)
(27.5%)
(24.2%)
300
184
316
(33.1%)
(35.2%)
(33.0%)
31.4
23.9
23.9
(17.6–42.7)
(9.1–37.7)
(9.3–38.7)
507
277
1138
(35.4%)
(29.2%)
(23.2%)
609
448
2279
(42.5%)
(47.2%)
(46.5%)
23.5
18.9
15.6
(8.0–36.6)
(4.3–32.7)
(4.5–29.1)
Lifestyle behaviours were compared across socioeconomic variables either by the chi-square test for categorical variables (% current or ex smokers) or by Kruskal–Wallis one way
ANOVA for continuous variables (pack years). All comparisons were significant (P b 0.05) except where indicateda.
J.V. Woodside et al. / Preventive Medicine 54 (2012) 247–253
249
than France and pack-years of smoking higher in NI than France. A
higher proportion of French men consumed alcohol, although weekly
consumption was similar in the two countries. Alcohol consumers in
NI consumed it in fewer days than in France. NI men were more physically active, but consumed fewer daily FV portions.
Lifestyle behaviours showed significant socioeconomic gradients
(Tables 1–3). These were similar in both countries, except for the percentage of alcohol consumers, which was less associated with socioeconomic variables in NI than France, and alcohol frequency, which
was less associated with socioeconomic variables in both countries,
but particularly in France.
Risks of death or of a first cardiovascular event are shown in
Table 4 by individual socioeconomic categories in three columns;
in model 1, adjustment has been made for age and country only; in
model 2, data are adjusted for all lifestyle behaviours, in model 3,
further adjustment has been made for conventional risk factors.
Socioeconomic factors were more closely associated with all-cause
mortality than with CVD. Interactions between country and each of
the socioeconomic variables were fitted but none attained
significance (P b 0.05) suggesting that socioeconomic factors had
broadly the same effect on the risk of death or of CVD in France and
NI.
In the analysis of all-cause mortality the contribution of all seven
socioeconomic indicators remained statistically significant in model 2.
The median percentage of the country/age-adjusted effect explained
by adjustment for lifestyle behaviours was 28%, similar to the analysis
factors, a test for trend across the categories was used. For two categories, the hazard ratios represent a straightforward comparison of two
categories while for the latter they represent the change in hazard associated with one step up the scale.
The extent to which gradients in mortality or cardiovascular
incidence (first event) between socioeconomic groups could be
explained by lifestyle behaviours was estimated using the expression
100(b0 − b1) / b0, where b0 was the coefficient for country or socioeconomic group on a log hazard scale in the Cox regression model
without lifestyle behaviours and b1 the coefficient in the corresponding
model with lifestyle behaviours (Hardoon et al., 2008). This
approach provided an estimate of the residual percentage contribution
of lifestyle behaviours after adjustment for age and country. These
residual contributions were only calculated if the socioeconomic gradient coefficient, b0, was significant before lifestyle and conventional risk
factors were added to the model; otherwise these residual contributions could simply reflect random variation. Bootstrap re-sampling
was used to obtain a 95% confidence interval (CI) for each residual
contribution estimate. This analysis was performed using Stata release
9 (College Station, Texas) while the remainder was performed using
SPSS version 17 (SPSS Inc., Illinois).
Results
Differences in lifestyle behaviours between countries are shown
(Tables 1–3). Baseline cigarette smoking was more common in NI
Table 2
Alcohol intake in subgroups defined by socioeconomic variable for 9709 men from Northern Ireland and France who were free of cardiovascular disease at entry (1991–4).
Northern Ireland
Overall
Marital Status
Married/Cohabiting
Single
Widowed/Divorced
People in household
1 person
2-3 people
4 + people
Accommodation
Mortgaged/co-owned
Owned
Rented
Cars in household
None
1 car
2 + cars
Years full-time education
b12 years
12–14 years
15 + years
Educational level
Primary
Secondary
Technical
Higher
Economic activity
Working
Sick/Disabled/Retired
Unemployed
Material condition
Low
Medium
High
n
% alcohol
consumers
2390
60.4%
2045
164
181
59.9%
55.5%
70.7%
181
1324
882
68.0%
60.3%
59.0%
1330
665
393
France
n
% alcohol
consumers
Alcohol (g/week)
in consumers
Frequency per
wk in consumers
(%)
19%
7319
90.7%
Median
(IQ Range)
1-2
3-4
5+
213
(102, 354)
13%
8%
79%
18%
24%
28%
6439
316
562
91.3%
85.4%
86.5%
211a
226a
232a
(101, 352)
(112, 363)
(115, 375)
13%a
11%a
11%a
8%a
7%a
8%a
79%a
82%a
81%a
556
5060
1700
84.5%
92.4%
87.6%
238
214
202
(127, 389)
(102, 354)
(96, 346)
9%
13%
13%
7%
7%
9%
84%
80%
78%
913
4732
1658
90.0%
92.9%
84.7%
201
216
216
(92, 362)
(106, 346)
(96, 378)
15%a
12%a
14%a
7%a
8%a
8%a
78%a
80%a
78%a
19%a
18%a
20%a
321
3012
3970
68.5%
90.0%
93.0%
252
218
209
(96, 443)
(103, 365)
(102, 343)
13%a
13%a
13%a
7%a
7%a
8%a
80%a
80%a
79%a
26%
31%
29%
17%
22%
25%
4060
1973
1285
89.6%
91.9%
92.5%
230
211
178
(109, 377)
(106, 343)
(86, 299)
12%
13%
16%
7%
7%
9%
81%
80%
75%
56%
45%
58%
49%
26%
30%
28%
29%
18%
25%
14%
22%
1520
766
2435
2360
84.5%
92.6%
92.3%
92.3%
252
206
223
194
(112, 390)
(110, 336)
(105, 358)
(96, 332)
14%a
13%a
12%a
13%a
5%a
8%a
8%a
8%a
81%a
79%a
80%a
79%a
(82, 316)
(104, 338)
(162, 479)
53%a
50%a
55%a
28%a
33%a
23%a
19%a
17%a
22%a
5659
1069
590
91.9%
87.5%
85.1%
206a
238a
256a
(101, 346)
(112, 378)
(104, 418)
13%a
10%a
12%a
8%a
7%a
8%a
79%a
83%a
80%a
(117, 392)
(68, 266)
(78, 305)
57%
57%
47%
25%
28%
30%
18%
15%
23%
1432
949
4906
83.9%
89.5%
92.9%
226
204
212
(103, 408)
(94, 350)
(103, 346)
13%a
14%a
12%a
7%a
7%a
8%a
80%a
79%a
80%a
Alcohol (g/wk)
in consumers
Frequency per wk
in consumers
(%)
Median
(IQ range)
1–2
3–4
5+
182
(89, 328)
53%
28%
174
250
257
(84, 307)
(106, 439)
(133, 441)
54%
45%
51%
28%
31%
21%
259
177
181
(123, 422)
(91, 321)
(82, 317)
50%a
54% a
53% a
24%
28%
28%
a
61.4%
53.4%
69.2%
160
181
269
(80, 282)
(86, 341)
(156, 462)
54%a
50% a
57% a
29%
29%
23%
a
342
1151
894
72.5%
60.5%
55.8%
312
188
132
(184, 527)
(97, 327)
(67, 242)
53%a
56%a
51%a
28%a
26%a
29%a
1599
495
294
60.9% a
60.6%a
57.8%a
207
149
115
(106, 374)
(68, 266)
(60, 223)
57%
47%
46%
612
344
746
629
63.2%a
59.0%a
61.5%a
57.7%a
246
184
182
127
(123, 439)
(103, 340)
(100, 318)
(61, 230)
2100
97
193
60.2%a
59.8%a
63.2%a
173
211
282
905
523
959
66.7%
60.0%
54.7%
225
155
159
a
a
a
a
a
a
a
26%
18%
19%
a
17%
21%
20%
a
a
a
a
a
Lifestyle behaviours were compared across socioeconomic variables either by the chi-square test for categorical variables (% alcohol consumers or frequency per week in consumers) or by Kruskal–Wallis one way ANOVA for continuous variables (alcohol consumption per week). All comparisons were significant (P b 0.05) except where indicateda.
250
J.V. Woodside et al. / Preventive Medicine 54 (2012) 247–253
Table 3
Other lifestyle behaviours in subgroups defined by socioeconomic variable for 9709 men from Northern Ireland and France who were free of cardiovascular disease at entry (1991–1994).
Northern Ireland
Overall
Marital Status
Married/Cohabiting
Single
Widowed/Divorced
People in household
1 person
2–3 people
4 + people
Accommodation
Mortgaged/co-owned
Owned
Rented
Cars in household
None
1 car
2 + cars
Years full-time education
b12 years
12–14 years
15 + years
Educational level
Primary
Secondary
Technical
Higher
Economic activity
Working
Sick/Disabled/Retired
Unemployed
Material condition
Low
Medium
High
France
FVJ consumption
(portions/day)
Physical activity score
FVJ consumption
(portions/day)
Physical activity score
Median
(IQ range)
Median
(IQ range)
Median
(IQ range)
Median
(IQ range)
2.1
(1.3, 3.0)
87.5
(49.3, 135.5)
2.5
(1.7, 3.5)
81.3
(40.7, 134.9)
2.1
2.0
1.8
(1.4, 3.0)
(1.1, 3.1)
(1.1, 2.6)
87.9a
77.8a
87.2a
(50.7, 135.6)
(36.9, 127.9)
(42.1, 141.2)
2.6
2.5
2.2
(1.8, 3.5)
(1.7, 3.4)
(1.4, 3.1)
82.7
67.5
71.0
(41.9, 136.1)
(30.7, 121.4)
(29.8, 118.3)
1.9
2.1
2.1
(1.1, 2.8)
(1.4, 3.1)
(1.3, 3.0)
82.0a
90.0a
84.4a
(42.1, 131.5)
(50.9, 138.9)
(48.8, 130.7)
2.3
2.6
2.5
(1.4, 3.1)
(1.8, 3.5)
(1.7, 3.5)
68.1
82.8
81.3
(26.8, 119.0)
(42.0, 136.3)
(40.0, 134.5)
2.2
2.1
1.6
(1.5, 3.1)
(1.4, 3.1)
(0.9, 2.5)
87.1a
92.4a
78.8a
(53.1, 130.2)
(52.3, 137.9)
(25.3, 147.8)
2.6
2.6
2.4
(1.7, 3.5)
(1.8, 3.5)
(1.5, 3.3)
79.7
83.2
76.0
(41.0, 129.1)
(43.3, 135.0)
(30.4, 136.1)
1.5
2.0
2.4
(0.9, 2.3)
(1.3, 2.9)
(1.7, 3.3)
76.5
92.9
84.6
(23.8, 149.0)
(52.0, 137.4)
(54.4, 124.4)
2.3
2.5
2.6
(1.3, 3.1)
(1.6, 3.5)
(1.8, 3.5)
64.5
81.5
82.1
(21.3, 132.5)
(37.2, 139.7)
(44.0, 132.0)
1.9
2.3
2.6
(1.2, 2.8)
(1.6, 3.1)
(1.9, 3.5)
95.5
79.9
71.3
(49.0, 145.0)
(51.5, 115.2)
(47.7, 106.5)
2.5
2.5
2.8
(1.6,3.4)
(1.7, 3.5)
(2.0, 3.8)
93.5
76.5
59.2
(43.7, 151.8)
(40.8, 123.5)
(35.0, 98.8)
1.6
2.1
2.1
2.6
(1.0,
(1.4,
(1.4,
(1.8,
2.6)
2.9)
2.9)
3.4)
93.5
82.2
98.8
76.2
(31.3, 151.3)
(48.6, 126.5)
(62.9, 144.2)
(48.5, 110.6)
2.5
2.4
2.4
2.6
(1.6, 3.5)
(1.7, 3.4)
(1.6, 3.3)
(1.9, 3.6)
93.8
71.4
94.5
66.5
(42.0,
(36.7,
(48.0,
(36.5,
2.1
2.0
1.4
(1.4, 3.1)
(1.1, 3.0)
(0.8, 2.3)
96.3
20.0
20.0
(61.2, 141.5)
(9.0, 40.2)
(9.9, 39.8)
2.6
2.4
2.4
(1.7, 3.5)
(1.7, 3.5)
(1.6, 3.3)
94.6
41.9
35.0
(53.4, 146.3)
(19.7, 79.8)
(15.0, 67.4)
1.9
2.1
2.3
(1.1, 2.6)
(1.4, 3.0)
(1.5, 3.2)
95.0a
82.3a
84.7a
(46.6, 143.9)
(51.5, 129.4)
(50.9, 127.3)
2.4
2.5
2.6
(1.5, 3.3)
(1.6, 3.5)
(1.8, 3.5)
76.0
80.5
83.0
(30.5, 138.2)
(41.6, 135.4)
(43.1, 134.1)
155.1)
115.6)
153.0)
107.0)
Lifestyle behaviours were compared across socioeconomic variables by Kruskal–Wallis one way ANOVA.
All comparisons were significant (P b 0.05) except where indicateda.
of material condition (34%). Additional adjustment for conventional
risk factors (model 3) increased the median percentage of the
country/age-adjusted effect explained to 38% (41% for material
condition).
For CVD only four of the seven socioeconomic indicators were significant after adjustment for age and country (model 1) and only two
remained significant in model 2. For the four significant indicators in
model 1, the median percentage of effect explainable by lifestyle
behaviours was 41%; the influence of material condition was no
longer significant in model 2, with adjustment for lifestyle behaviours
explaining 51% of the country/age-adjusted estimate. After additional
adjustment for conventional risk factors (model 3), none of the
socioeconomic indications were significantly associated with cardiovascular incidence and many of the hazard ratios lay close to 1.0. Adjustment for lifestyle behaviours and conventional risk factors
increased the median percentage of the age/country-adjusted effect
explained to 67% (65% for material condition). However, some of
the confidence intervals in these analyses exceeded 100% indicating
that, in a number of bootstrap samples, the socioeconomic effect
would not only be abolished but even reversed after adjustment for
lifestyle and conventional risk factors. This emphasises the potentially
large sampling variability for some of the estimates of percentage
explained.
Fig. 1 shows the effects of adjustment for the lifestyle behaviours,
both individually and in combination. For all socioeconomic variables,
and for both all-cause mortality and CVD, the largest contributor to
the socioeconomic gradient was smoking, followed by FV intake and
finally physical activity and alcohol.
Discussion
Despite clear differences in lifestyle behaviours between NI and
France, there were significant socioeconomic gradients in lifestyle behaviours in both countries. Combined lifestyle behaviours (smoking,
physical activity, FV intake and alcohol consumption, considered as
both amount and frequency of consumption) explained 28% and
41% (median) of the socioeconomic gradient in total mortality and
cardiovascular incidence respectively. The addition of known cardiovascular risk factors increased these percentages to 38% and 67%.
Healthier combined lifestyle behaviours have consistently been
associated with lower total mortality (Khaw et al., 2008; van Dam
et al., 2008), disease-specific mortality (van Dam et al., 2008), and
risk of CHD (Akesson et al., 2007), and stroke (Myint et al., 2009),
although the lifestyle behaviours included and measurement
methods differ between studies. Fewer studies have, however,
examined the ability of combined lifestyle behaviours to explain
socioeconomic differences in mortality or disease incidence. Most
studies suggest that a large proportion (often >50%) of the socioeconomic gradient in mortality or CVD incidence remains unexplained by
lifestyle behaviours alone. For example, in a 13 year follow-up of the
Whitehall study (Marmot et al., 2008), current smoking habit, physical activity, diet and alcohol consumption explained 30% of the
inequality index in occupational grade for CHD incidence. Similarly,
in an analysis of the Americans’ Changing Lives Survey, smoking, alcohol, BMI and physical activity together only accounted for 12–13% of
the predictive effect of income on mortality (Lantz et al., 1998,
2010). In fact, Lynch et al. (1996), in analyses of the Kuopio Ischemic
J.V. Woodside et al. / Preventive Medicine 54 (2012) 247–253
251
Table 4
Hazard ratios for death from any cause and for a cardiovascular event by socioeconomic gradient, and estimated proportion of that gradient explained by lifestyle factors alone, and
lifestyle factors and conventional risk factors in men from NI and France (assessed in 1991–1994).
All causes of death
Marital Status
People in household
Living accommodation
No of cars in household
Years of education
Educational level
Economic activity
Material condition
Hard CHD or CVD
Marital Status
People in household
Living accommodation
No of cars in household
Years of education
Educational level
Economic activity
Material condition
Model 1
Model 2
Model 3
Adjusted for age,
country
Adjusted for age,
country and lifestyle
behaviours
Adjusted for age,
country, lifestyle
behaviours and
conventional risk
factors
Percentage of country/age
adjusted effect
explainable by lifestyle
behaviours
Percentage of country/age
adjusted effect
explainable by lifestyle
behaviours and
conventional risk factors
HR
(95% CI)
HR
(95% CI)
HR
(95% CI)
Explained
(95% CI)
Explained
(95% CI)
2.04***
0.72***
1.92***
0.59***
0.79***
0.85***
1.68***
0.72***
(1.66, 2.51)
(0.62, 0.85)
(1.59, 2.31)
(0.51, 0.67)
(0.69, 0.89)
(0.79, 0.92)
(1.38, 2.05)
(0.66, 0.80)
1.80***
0.78**
1.53***
0.68***
0.85*
0.91**
1.45***
0.81***
(1.45,
(0.67,
(1.26,
(0.59,
(0.75,
(0.84,
(1.16,
(0.73,
2.22)
0.92)
1.86)
0.78)
0.97)
0.98)
1.82)
0.89)
1.75***
0.79**
1.47***
0.71***
0.88
0.93
1.38*
0.83***
(1.42,
(0.67,
(1.21,
(0.62,
(0.77,
(0.86,
(1.10,
(0.75,
2.16)
0.92)
1.79)
0.82)
1.00)
1.00)
1.72)
0.92)
18%
23%
35%
28%
34%
41%
28%
34%
(11%, 31%)
(14%, 52%)
(23%, 52%)
(18%, 42%)
(15%, 71%)
(23%, 85%)
(7%, 57%)
(23%, 51%)
22%
26%
41%
36%
44%
52%
38%
41%
(14%, 37%)
(15%, 59%)
(22%, 60%)
(24%, 54%)
(20%, 95%)
(32%, 100%)
(16%, 69%)
(28%, 62%)
1.14
0.97
1.40**
0.80**
0.81**
0.92
1.58***
0.86**
(0.86, 1.50)
(0.81, 1.15)
(1.13, 1.75)
(0.69, 0.93)
(0.71, 0.93)
(0.84, 1.00)
(1.26, 1.98)
(0.77, 0.96)
1.05
0.99
1.20
0.90
0.86*
0.97
1.35*
0.93
(0.79,
(0.83,
(0.95,
(0.77,
(0.75,
(0.89,
(1.05,
(0.83,
1.39)
1.18)
1.51)
1.06)
1.00)
1.06)
1.74)
1.04)
1.03
1.01
1.12
0.96
0.94
1.00
1.22
0.95
(0.77,
(0.85,
(0.89,
(0.82,
(0.81,
(0.91,
(0.94,
(0.84,
1.36)
1.21)
1.42)
1.13)
1.08)
1.09)
1.57)
1.06)
47%
55%
29%
(22%, 163%)
(24%, 170%)
(7%, 80%)
66%
83%
68%
(32%, 216%)
(43%, 252%)
(35%, 207%)
34%
51%
(6%, 81%)
(23%, 158%)
57%
65%
(26%, 132%)
(30%, 195%)
*P b 0.05, **P b 0.01, ***P b 0.001, HR—hazard ratio, CI—confidence interval.
Conventional risk factors—systolic blood pressure, diabetes, body mass index, cholesterol, HDL cholesterol, height.
Lifestyle behaviours—smoking, fruit vegetable and juice consumption, alcohol intake, total physical activity.
Alcohol intake (five consumption categories and 4 frequency categories) and Smoking (never, non-cigarettes and three pack-year categories) were fitted separately to French and
Northern Ireland data.
Heart Disease Risk Factor Study, showed that, after adjustment for biologic risk factors, further adjusting for behavioural risk factors including alcohol, smoking and physical activity, explained little of the
excess relative risk of myocardial infarction in the lowest income
group.
In this study, despite significant and consistent socioeconomic
gradients in lifestyle behaviours, differences in lifestyle behaviours
account for approximately 30% of inequalities in mortality, and 40%
in cardiovascular incidence. Even after inclusion of cardiovascular
risk factors, 62% and 33% of the socioeconomic gradient in all-cause
mortality and cardiovascular incidence, respectively, were still
unexplained. Other factors must therefore contribute to the observed
socioeconomic gradient, perhaps related to material factors such as
exposure to occupational and environmental hazards or access to
healthcare, or psychosocial factors such as lack of social support,
self-esteem, and exposure to stress. The data from the current study
suggest that policies simply concentrating on improving lifestyle
behaviours in lower socioeconomic groups are unlikely to eradicate
the socioeconomic gradient in cardiovascular incidence, and, in
particular, all-cause mortality.
Although the main focus of this analysis was an examination of
combined lifestyle behaviours, after our previous focus on smoking
(Yarnell et al., 2011), we were also interested in which of the individual
lifestyle behaviours explained most of the socioeconomic gradient.
When adjustments were made individually, smoking behaviours
contributed most, with FV intake contributing to a lesser extent, and
alcohol consumption and physical activity having little impact. This is
in agreement with previous work, which found smoking to be the
strongest contributor, with physical activity and dietary pattern, but
not alcohol consumption, making modest contributions (Marmot et
al., 2008). As drinking patterns differ in France and NI, and these patterns are associated with CHD risk (Ruidavets et al., 2010), our analysis
incorporated both alcohol consumption and the number of days over
which alcohol was consumed.
Strengths of the current analysis are that we were able to consider
a panel of lifestyle behaviours, including lifetime smoking behaviour,
and that we included both an estimate of total alcohol intake and frequency of consumption. The analysis has been carried out in two
countries with significant differences in lifestyle behaviours, but the
observations made were similar in both countries. The bootstrapping
method also allowed the presentation of confidence intervals around
the estimate of the percentage of the socioeconomic gradient
explained by the various adjustments. Finally, the 95% complete
follow-up at 10 years is also a strength.
Weaknesses of the current analysis include the self-reported
assessment of lifestyle behaviours when objective measures, e.g.
accelerometers for physical activity, would have been preferable.
Lifestyle behaviours were only assessed at one time-point, and it is
possible that these lifestyle behaviours change over time. In an analysis of civil servants (Stringhini et al., 2010), which assessed health
behaviours at 4 time-points over a 24 year period, health behaviours
explained only 42% of social inequalities in all-cause mortality when
baseline health behaviours were used, but this increased to 72%
when the repeated measures were included. The difference between
the baseline only and repeated assessments of health behaviours in
the study of civil servants was mostly due to an increased explanatory
power of diet, physical activity and alcohol, with the role of smoking
remaining similar. By having only a single measure of diet, physical
activity and alcohol in the current study, the contribution of these
health behaviours to the socioeconomic gradient may therefore lead
to an underestimation of their effects. However, we have excluded
those with clinical evidence or history of CVD at baseline, and
therefore major lifestyle changes motivated by illness are likely to
have been minimised. The scope for reverse causality also increases
when using repeated measures if people change lifestyle in periods
of ill-health occurring prior to death, although this possibility can be
reduced by considering only repeated lifestyle behaviours prior to
the outcome event. A final weakness is that we focused on men
252
J.V. Woodside et al. / Preventive Medicine 54 (2012) 247–253
Mortality
Hazard ratio (95%CI)
0.25
0.5
1
2
Cardiovascular disease
Hazard ratio (95%CI)
4 0.25
0.5
1
2
4
Smoking
Alcohol
Exercise
Fruit, vegetable & juice
All of the above
Marital Status
(Single/Divorced/Widowed
versus Married)
Smoking
Alcohol
Exercise
Fruit, vegetable & juice
All of the above
People in household
(4+ versus 1 persons)
Smoking
Alcohol
Exercise
Fruit, vegetable & juice
All of the above
Living accommodation
(Rented versus Mortgaged/
Co-owned/Owned)
Smoking
Alcohol
Exercise
Fruit, vegetable & juice
All of the above
No of cars in household
(2+ versus 0)
Smoking
Alcohol
Exercise
Fruit, vegetable & juice
All of the above
Years of education
(per three years)
Smoking
Alcohol
Exercise
Fruit, vegetable & juice
All of the above
Educational level
(Higher versus Primary)
Smoking
Alcohol
Exercise
Fruit, vegetable & juice
All of the above
Economic activity
(Sick/Disabled/Retired/
Unemployed versus Working)
Smoking
Alcohol
Exercise
Fruit, vegetable & juice
All of the above
Material condition
(High versus Low)
Fig. 1. Hazard ratios for death from all causes and for a cardiovascular event by socioeconomic gradient, and effect of adjusting for lifestyle factors individually and in combination
(as coded in Table 4) on that estimate.
aged 50–59 years, and therefore our findings are only applicable to
that specific cohort.
The current report may also underestimate the contribution of
lifestyle behaviours to the health outcomes presented since men
with evidence or history of CVD at baseline (9%) were excluded
from the present analyses. Such men have a much greater mortality
and cardiovascular incidence than the largely healthy remaining cohort; lifestyle behaviours would be expected to have contributed to
premature disease in these men, and may have increased the socioeconomic gradient. However, re-analysis of the data to include
these subjects resulted in only small alterations in hazard ratios for
all socioeconomic variables for mortality and cardiovascular incidence, and the values for the percentage of these gradients
‘explained’ by lifestyle and conventional risk factors remained largely
unchanged.
Conclusion
The substantial gradients in mortality and in cardiovascular incidence between socioeconomic groups in NI and France are only partially explained by lifestyle behaviours and cardiovascular risk
factors. Substantial proportions of these gradients remain unexplained, particularly for overall mortality.
Conflict of interest statement
The authors declare that they have no conflict of interest.
Acknowledgements
The PRIME Study is organised under an agreement between
INSERM and the Merck, Sharpe and Dohme-Chibret Laboratory, with
the following participating Laboratories:
The Strasbourg MONICA Project, Strasbourg, France (D. Arveiler, B.
Haas).
The Toulouse MONICA Project, INSERM U558, Toulouse, France
(J. Ferrières, JB. Ruidavets).
The Lille MONICA Project, INSERM U744, Lille, France (P. Amouyel,
M. Montaye).
The Department of Epidemiology and Public Health, Queen's
University Belfast, NI (A. Evans, J. Yarnell, F.Kee).
The Department of Atherosclerosis, INSERM U545, Lille, France
(G. Luc, JM. Bard).
The Laboratory of Hematology, La Timone Hospital, Marseille,
France (I. Juhan-Vague).
The Laboratory of Endocrinology, INSERM U326, Toulouse, France
(B. Perret).
The Vitamin Research Unit, The University of Bern, Bern, Switzerland
(F. Gey).
The Trace Element Laboratory, Department of Medicine, Queen's
University Belfast, NI (J. Woodside, I. Young).
The DNA Bank, INSERM U525, Paris, France (F. Cambien).
The Coordinating Center, INSERM U909, Villejuif, France
(P. Ducimetière, A. Bingham).
J.V. Woodside et al. / Preventive Medicine 54 (2012) 247–253
The PRIME Study is supported by grants from INSERM, Merck,
Sharpe and Dohme-Chibret Laboratory, the French Research Agency
and the Foundation Heart and Arteries.
We also acknowledge the contribution of the members of the
event validation Committees: Pr L. Guize ( deceased), Dr C. Morrison,
Dr M-T. Guillanneuf, Pr M. Giroud; and the Alliance Partnership
Programme for its financial support.
References
Akesson, A., Weismayer, C., Newby, P.K., Wolk, A., 2007. Combined effect of low-risk
dietary and lifestyle behaviors in primary prevention of myocardial infarction in
women. Arch. Intern. Med. 167, 2122–2127.
Dauchet, L., Ferrieres, J., Arveiler, D., et al., 2004. Frequency of fruit and vegetable
consumption and coronary heart disease in France and NI: the PRIME study. Br. J.
Nutr. 92, 963–972.
Davey Smith, G., Blane, D., Bartley, M., 1994. Explanations for socioeconomic differentials
in mortality: evidence from Britain and elsewhere. Eur. J. Public Health 4, 131–144.
Ducimetière, P., Ruidavets, J.-B., Montaye, M., Haas, B., Yarnell, J., on behalf of The
PRIME Study Group, 2001. Five-year incidence of angina pectoris and other
forms of coronary heart disease in health men aged 50–59 in France and NI: The
PRIME Study. Int. J. Epidemiol. 30, 1057–1062.
Hardoon, S.L., Whincup, P.H., Lennon, L.T., Wannamethee, S.G., Capewell, S., Morris, R.W.,
2008. How much of the recent decline in the incidence of myocardial infarction in
British men can be explained by changes in cardiovascular risk factors? Evidence
from a prospective population-based study. Circulation 117, 598–604.
Khaw, K.T., Wareham, N., Bingham, S., Welch, A., Luben, R., Day, N., 2008. Combined impact of health behaviours and mortality in men and women: the EPIC-Norfolk prospective population study. PLoS Med. 5, e12.
Kunst, A.E., Groenhof, F., Mackenbach, J.P., the EU Working Group on Socioeconomic
Inequalities in Health, 1998. Occupational class and cause specific mortality in
middle aged men in 11 European countries: comparison of population based
studies. BMJ 316, 1636–1642.
Lang, T., Ducimetière, P., 1995. Premature cardiovascular mortality in France: divergent
evolution between social categories from 1970 to 1990. Int. J. Epidemiol. 24, 331–339.
253
Lang, T., Ducimetière, P., Arveiler, D., et al., 1997. Incidence, case fatality, risk factors of
acute coronary heart disease and occupational categories in men aged 30–59 in
France. Int. J. Epidemiol. 26, 47–57.
Lantz, P.M., House, J.S., Lepkowski, J.M., Williams, D.R., Mero, R.P., Chen, J., 1998.
Socioeconomic factors, health behaviors, and mortality: results from a nationally
representative prospective study of US adults. JAMA 279, 1703–1708.
Lantz, P.M., Golberstein, E., House, J.S., Morenoff, J., 2010. Socioeconomic and
behavioural risk factors for mortality in a national 19-year prospective study of
U.S. adults. Soc. Sci. Med. 70, 1558–1566.
Lynch, J.W., Kaplan, G.A., Cohen, R.D., Tuomilehto, J., Salonen, J.T., 1996. Do
cardiovascular risk factors explain the relation between socioeconomic status,
risk of all-cause mortality, cardiovascular mortality, and acute myocardial infarction? Am. J. Epidemiol. 144, 934–942.
Marmot, M.G., Shipley, M.J., Hemingway, H., 2008. Biological and behavioural
explanations of social inequalities in coronary heart disease: the Whitehall II
study. Diabetalogia 51, 1980–1988.
Myint, P.K., Luben, R.N., Wareham, N.J., Bingham, S.A., Khaw, K.T., 2009. Combined
effect of health behaviours and risk of first ever stroke in 20,040 men and
women over 11 years' follow-up in Norfolk cohort of European Prospective
Investigation of Cancer (EPIC Norfolk): prospective population study. BMJ 338,
b349.
Ruidavets, J.B., Ducimetière, P., Evans, A., et al., 2010. Patterns of alcohol consumption
and ischaemic heart disease in culturally divergent countries: the Prospective
Epidemiological Study of Myocardial Infarction (PRIME). BMJ 341, c6077.
Stringhini, S., Sabia, S., Shipley, M., et al., 2010. Association of socioeconomic position
with health behaviors and mortality. JAMA 303, 1159–1166.
van Dam, R.M., Li, T., Spiegelman, D., Franco, O.H., Hu, F.B., 2008. Combined impact of
lifestyle factors on mortality: prospective cohort study in US women. BMJ 337,
a1440.
Wagner, A., Simon, C., Evans, A.E., et al., 2003. Physical activity patterns in 50–59 year
old men in France and NI. Associations with socio-economic status and health
behaviour. Eur. J. Epidemiol. 18, 321–329.
Yarnell, J., Yu, S., McCrum, E., et al., 2005. Education, socioeconomic and lifestyle
factors, and risk of coronary heart disease: the PRIME Study. Int. J. Epidemiol. 34,
268–275.
Yarnell, J.W.G., Patterson, C.C., Arveiler, D., et al., 2011. Contribution of lifetime smoking
habit in France and NI to country and socioeconomic differentials in mortality and
cardiovascular incidence: the PRIME study. J. Epidemiol. Community Health April
17, epub ahead of print.