Do lifestyle, health and social participation mediate educational inequalities in frailty worsening?
European Journal of Public Health
Do lifestyle, health and social participation mediate educational inequalities in frailty worsening?
Astrid Etman 2
Carlijn B. M. Kamphuis 2
Tischa J. M. van der Cammen 0 1 3
Alex Burdorf 2
Frank J. van Lenthe 2
0 Department of Medicine, Brighton and Sussex Medical School , Brighton , UK
1 Department of Internal Medicine, Section of Geriatric Medicine, Erasmus MC, University Medical Center Rotterdam , Rotterdam , The Netherlands
2 Department of Public Health, Erasmus University MC , Rotterdam , The Netherlands
3 Faculty of Industrial Design Engineering, Delft University of Technology , Delft , The Netherlands
4 United Nations, Department of Economic and Social Affairs. World Population Prospects, the 2012 Revision [database]. Available at:
5 Mitchell TD, Carter TR, Jones PD, et al. A Comprehensive Set of High-resolution Grids of Monthly Climate for Europe and the Globe: The Observed Record (1901-2000) and 16 Scenarios (2001-2100). Tyndall Centre Working Paper 55, Tyndall Centre for Climate Change Research. Norwich: University of East Anglia , 2004. Dataset available at:
20 Eng H, Mercer JB. Mortality from cardiovascular diseases and its relationship to air temperature during the winter months in Dublin and Oslo/Akershus. Int J Circumpolar Health 2000;59:176-81. 24 United Nations Statistics Division. Composition of macro geographical (continental) regions, geographical sub-regions, and selected economic and other groupings. Available at: http://millenniumindicators.un.org/unsd/methods/m49/m49regin.htm (22 March 2013, date last accessed).
railty develops as a consequence of age-related decline in many
Fphysiological systems, which collectively results in vulnerability to
sudden health status changes.1 Due to ageing in Western
populations, an increased number of older persons will become
frail in the upcoming years. According to the often-used definition
of Fried,2 currently, 37% of community-dwelling persons aged >55
years are pre-frail and about 4% are frail,3 with percentages
increasing to 51 and 26% respectively for those aged >70 years.4
Among those aged 55 years, almost one quarter of the population
in Western countries worsened in frailty over a relatively short
period of 2 years.5 Because frailty can lead to falls, hospitalization,
nursing home placement and death,1 it is important to find out how
the frailty process develops to prevent or slow down this process
from onset. Moreover, because the development of frailty is found to
be a reversible process, appropriate interventions may contribute to
frail older persons becoming pre-frail or even non-frail.5,6
Frailty is more prevalent among lower educated as compared with
higher educated persons.7 Two recent longitudinal studies suggested
a causation mechanism as lower educated persons aged 55 years
showed an increased risk to worsen in frailty over time, compared
with higher educated persons.4,6 Potential factors contributing to
educational inequalities in worsening in frailty are largely
unknown, but can be derived indirectly. There is evidence that an
unhealthy lifestyle (e.g. smoking), limited social participation and
health conditions are related to the frailty development process,2,8–11
although reverse causality cannot always be excluded. For example,
frailty is associated with the onset of depression, but depression may
also result in a worsening of frailty.12 Because educational differences
in lifestyle,13,14 health15,16 and social participation15,17 are well
known, these factors may likely contribute to the educational
inequalities in frailty worsening; a quantification of their
contribution, however, is currently lacking.
The Survey of Health, Ageing, and Retirement in Europe
(SHARE) aims at investigating population ageing processes across
European countries. Longitudinal data on frailty and underlying
determinants make the study suitable for research aimed at
improving the understanding of educational inequalities in the
frailty process. When investigating the role of lifestyle, health and
social participation in educational inequalities in frailty worsening,
possible differences between European countries in the extent to
which potential mediators may contribute to inequalities in frailty
worsening should be acknowledged. Therefore, this study adds
knowledge by exploring whether lifestyle, health and social
participation mediate the relationship between educational level and frailty
worsening among community-dwelling elderly in 11 European
Data of persons participating in the SHARE in both 200
4 (wave one)
(wave two) were used. The SHARE study was designed to
investigate population-ageing processes by looking at changes in
health, economic situations and social networks of individuals
aged 50 years. Nationally representative samples of 11 European
countries (Sweden, Denmark, Germany, The Netherlands, Belgium,
Switzerland, Austria, France, Italy, Spain and Greece) were
interviewed face-to-face with structured computerized questionnaires.
The household response in wave 1 varied from 38.8% in
Switzerland to 79.2% in France. Details on this survey are
described by Bo¨rsch-Supan et al. (2008, available online at http://
Subjects were eligible for the analyses if they were
communitydwelling, aged 55 years at wave one, and participated in wave
two as well. A total of 14 477 European community-dwelling
persons fulfilled these inclusion criteria; however, 395 were
excluded due to missing values for educational level, for
1 mediator(s), or because they had 3 missing Fried items at
one or both waves. This resulted in a study population of 14 082
Educational level was measured at wave one and was defined as the
number of years a person received full time education (i.e. receiving
tuition, engaging in practical work or supervised study or taking
examinations). For international comparisons of education,
SHARE used the 1997 International Standard Classification of
Education (ISCED-97). Educational level was dichotomized in
0–10 years (which corresponds with ISCED level 0–2; ‘lower
educated’) and 11–25 years (which corresponds with ISCED level
3–6; ‘higher educated’).
Physical frailty was based on the Fried’s criteria, i.e. weakness,
slowness, low activity, weight loss and exhaustion. To make
optimal use of the data available in the SHARE survey, we
measured frailty level with an adapted version of Fried’s frailty
scale as developed by Santos-Eggimann and colleagues.3 Weakness
was defined as being below cut-off points (stratified by sex and body
mass index2) for the highest of four measurements of hand grip
strength. Participants were classified positive for slowness when
mentioning having difficulty walking 100 meter or climbing one
flight of stairs. Participants were classified as positive for low
activity when answering the question ‘How often do you engage
in activities that require a low or moderate state of energy, such as
walking, gardening, cleaning the car, or doing a walk?’ with ‘one to
three times a month’ or ‘hardly ever or never’. Unintentional weight
loss was based on the answers ‘less’ or ‘diminution in desire for food’
to the question ‘what has your appetite been like?’ or the answer
‘less’ to the question ‘So you have been eating more, or less than
usual?’. Exhaustion was based on the question ‘In the last month,
have you had too little energy to do the things you wanted to do?’.
Answering ‘yes’ was considered as being positive for exhaustion.
Frailty states were based on the total number of criteria met: ‘frail’
( 3 criteria), ‘pre-frail’ (1–2 criteria), ‘non-frail’ (0 criteria).
Worsening in frailty was defined as changing from a lower to a
higher frailty state after two years (i.e. from non-frail to pre-frail
or frail, or from pre-frail to frail) with ‘no change in frailty’ as the
reference group. Additional analyses were performed for improving
in frailty, which was defined as changing from a high to a low frailty
state after two years (results in Supplementary Appendix).
Potential mediators: Lifestyle, health and social participation
Self-reported lifestyle (smoking and alcohol consumption), health
(presence of chronic diseases, memory function and depression)
and social participation were measured at baseline. Smoking
behaviour was measured with the question ‘Do you smoke?’
(current, former or never smoker). Alcohol consumption was
based on the number of days per week participants were drinking
alcohol during the last six months (<1 day, 3–4 days, 5 days per
week). Chronic diseases were measured by questioning ‘Has a doctor
ever told you that you have any of the following conditions?’,
followed by a list of 14 chronic conditions, e.g. hypertension,
arthritis, osteoporosis (none, 1 chronic diseases). Memory
function was based on the maximum number of words (out of a
10-words list) a respondent was able to recall after a verbal and a
numeric test [‘impaired’ ( 4 words), ‘good’ (>4 words)].
Depression was measured based on the EURO-D scale with 12
items on e.g. depression, pessimism, appetite and fatigue [‘not
depressed’ (0–4 items), ‘probably depressed’ ( 5 items ‘yes’)].18
Social participation was measured with participating in social
activities over the last month, e.g. voluntary work, cared for a sick
person, participation at sports club (‘none’, ‘one or more’).
When scores for one or two of the five frailty criteria were missing,
values were imputed through single random imputation, using
software package R V.2.7.1. The scores of the population without
missing values were used to replace missing values through a logistic
regression model. Using this model, the probability of scoring
‘positive for frailty’ on a frailty indicator for every individual (with
one or more missing values) was predicted and a random draw from
the binomial distribution with that probability was made. To check
the influence of random imputation, the procedure was repeated
and no essential differences were found. Furthermore, a sensitivity
analysis was conducted in which participants with missing outcome
data were excluded (results available upon request). No substantial
differences were found. Data were imputed for 2080 (14.8%) and
2312 (16.5%) individuals in waves one and two, respectively.
Differences in sex, age and educational level between the study
population and the excluded sample were investigated using Chi2
tests (sex, educational level) and a t-test (mean age). The association
of educational level and frailty worsening was based on odds ratios
[ORs, 95% confidence interval (CI)] from multinomial logistic
regression analyses. Following conventional rules of mediation
analysis,19 the associations of educational level with the possible
mediators, and of the possible mediators with frailty worsening,
were explored by binominal and or multinomial logistic regression
analyses (depending on the number of categories of the mediating
factor) among the total study sample. Finally, in multinomial logistic
regression analyses, potential mediators were successively added to a
model in which educational level was associated with frailty
worsening, with ‘no change in frailty’ as the reference group, for
each country separately. All analyses were adjusted for age and sex.
Analyses concerning frailty changes were adjusted for baseline frailty
state which has been found to be associated with subsequent changes
in frailty.3,20 In all analyses, P-values of <0.05 were considered
significant using SPSS V.20.0. To reduce potential selection bias
generated by non-response, analyses were performed with
individual longitudinal weights (SHARE Release guide 2.5.0).
The study sample was younger and higher educated than those
excluded from the analyses (P < 0.01, not tabulated). Within the
study sample, most were women, in the younger age categories,
and non-frail at wave one and wave two. After two years of
follow-up, 22.1% worsened, 16.0% improved and 61.9% showed
no change in frailty state. Lower educated persons (59.3%) were
older, more often frail at both waves, and more often worsened in
frailty after two years compared with higher educated persons.
Among the higher educated persons 19.2% worsened compared
with 24.0% among lower educated persons (table 1). The absolute
prevalence of worsening in frailty during a 2-year period was up to
9.5% higher among lower educated persons compared with those
with higher education (Germany) (figure 1).
As shown in table 2, in the total study sample, lower educated
persons had a lower probability to be a current or former smoker, to
drink alcohol, or to participate in social activities as compared with
higher educated persons. Furthermore, lower educated persons had
a higher probability to be depressed, to have impaired memory
or have one or more chronic diseases (table 2). Cross-national
differences in the pattern of inequalities were found for smoking
behaviour: in Sweden and The Netherlands, lower educated
Table 3 Associations (odds ratios, 95% confidence intervals)
between lifestyle, health and social participation and frailty
worsening over 2 years follow-up, adjusted for age, sex,
educational level, baseline frailty state and country (N = 14 082)
persons more often were current smokers [ORs 1.62 (95% CI:
1.19–2.22) to 2.00 (95% CI: 1.45–2.76)], whereas in France, Spain,
and Greece, lower educated persons less often were current
smokers compared with higher educated persons [ORs varying 0.65 (95%
CI: 0.47–0.88) to 0.54 (95% CI: 0.35–0.84)] (Supplementary
Being a current smoker, drinking no alcohol, having a depression,
or one or more chronic diseases, and no social participation
increased the likelihood of frailty worsening (table 3). Although
not always significant for the separate countries, the direction of
these associations was mostly comparable with that in the total
study sample (Supplementary Appendix 2).
Because alcohol consumption, chronic diseases, depression and
social participation were associated with both educational level
and frailty worsening, these factors were added as mediators to the
explanatory models for educational inequalities in frailty worsening.
Smoking was not added to the explanatory models, because smoking
did increase the risk of worsening in frailty, but lower educated were
doing better on this risk factor (i.e. less likely to be a current smoker)
than higher educated.
An increased probability of worsening in frailty in lower as
compared with higher educated persons was found in 10
countries, but was only statistically significant in five countries
[ORs varying from 1.40 (95% CI: 1.06–1.84) to 1.61 (95% CI:
1.21–2.14), see table 4]. Inclusion of lifestyle, health and social
participation separately resulted in only a minor attenuation of the odds
ratios which varied between 4.9% in The Netherlands to 31.5% in
Germany (table 4).
Among European community-dwelling older persons aged 55
years, lower educated were found to be at an increased risk of
worsening in frailty over a 2-years follow-up. While low alcohol
consumption, chronic diseases, depression and less social
participation increased the probability of frailty worsening and were more
prevalent among the lower educated, only small to modest parts of
the educational inequalities in frailty worsening were explained by
Our findings are in line with research in which lower educated
older persons were found to be at an increased risk of worsening in
frailty over an average 6.4 years period.7 Lifestyle factors are reported
to be on the pathway of educational inequalities in health,21–23
which is supported by our finding that alcohol consumption
contribute to educational differences in frailty changes. Our
finding that the presence of chronic diseases explained part of
educational inequalities in frailty worsening is supported by findings of
Gobbens et al.24 who found that multimorbidity partly mediates the
relationship between income and frailty. As mentioned, associations
between health factors such as depression and frailty may be due to
reverse causality.12 Our longitudinal approach, however, strengthens
the evidence of the association between health factors and the
development of frailty.
Our finding that alcohol consumption was associated with a lower
probability of frailty worsening is supported by studies in which
alcohol consumption was found to protect against coronary heart
disease and dementia.25,26
Smoking behaviour is associated with both the onset11,13 and
worsening in frailty. Overall, associations with frailty worsening
were found for certain health conditions, i.e. presence of depression
and the presence of chronic diseases, which is supported by earlier
research on the presence of frailty.2,8,13,27 Furthermore, persons who
were not socially participating showed an increased risk of worsening
in frailty. These results fit well with the findings of Cramm et al.28 who
showed that the social environment (e.g. social cohesion, social
support, contact with neighbours) plays an important role for the
well-being of older persons.
There is a possibility that two years is too short to detect an effect
of the possible mediators on the frailty worsening process. Future
research in this field should focus on follow up periods longer than 2
years, and search for additional explanations for the educational
inequalities in frailty worsening. Previous studies addressed the
importance of material (e.g. financial situation, housing
conditions), psychosocial (e.g. life events, external locus of
control) and environmental factors (e.g. neighbourhood
characteristics) when studying educational inequalities in health.18,29–31 It
therefore seems legitimate to further investigate how and when
differential exposure to material circumstances, psychosocial factors
and characteristics of the built environment over the life course
between educational groups may translate into an increased risk of
worsening in frailty among the lower educated.
Some limitations of this study should be mentioned. First, frailty
state was measured via self-report. Differential misclassification of
frailty state by educational level may have led to incorrect
associations. It is unknown, however, whether this would result in
underor overestimations of the educational inequalities in frailty
development. Second, the self-reported nature of the mediators may have
resulted in an underestimation of their contribution to educational
inequalities in frailty worsening. There is evidence of larger
underreporting of chronic conditions32 and over-reporting of a healthy
lifestyle33,34 among persons with lower as compared with higher
educational levels. Furthermore, higher educated persons are more
likely to participate in surveys as compared with lower educated
persons which may also have resulted in an underestimation of
educational inequalities in frailty worsening. Third, alcohol
consumption was measured by the number of days drinking alcohol per week,
without asking for the number of glasses per day. Therefore, this
measure does not allow to differentiate between binge drinkers and
regular drinkers. This may have underestimated the contribution of
alcohol consumption to inequalities in frailty worsening, as binge
drinking increases the likelihood of unfavourable health outcomes
(e.g. functional limitations and death),35–37 and for example in
Dutch persons, may be more common among lower than higher
educated persons.38 Fourthly, among the non-responses at wave 2,
some passed away between wave 1 and 2 (exact number is unclear).
As some deaths could have been due to worsening in frailty, this may
have resulted in an underestimation of the prevalence of frailty
worsening. As lower educated persons were more likely to worsen
in frailty, it may also have resulted in an underestimation of the
educational inequalities in frailty worsening.
In conclusion, our study showedthatalthoughlifestyle, healthandsocial
participation were associated with the frailty development process, only
small to moderate parts of educational inequalities in frailty worsening
among older European persons were explained by these factors.
Supplementary data are available at EURPUB online.
This paper uses data from SHARELIFE release 1, as of November
, and SHARE release 2.3.1, as of July 29th 2010. The
SHARE data collection has been primarily funded by the
European Commission through the 5th framework programme
(project QLK6-CT-2001- 00360 in the thematic programme
Quality of Life)
, through the 6th framework programme (projects
SHARE-I3, RII-CT- 2006-062193; COMPARE,
CIT5-CT-2005028857; and SHARELIFE, CIT4-CT-2006-028812) and through the
7th framework programme (SHARE-PREP, 211909 and
SHARELEAP, 227822). Additional funding from the U.S. National
Institute on Aging (U01 AG09740-13S2, P01 AG005842, P01
AG08291, P30 AG12815, Y1-AG-4553-01 and OGHA 04-064, IAG
BSR06-11, R21 AG025169) as well as from various national
sources is gratefully acknowledged (see www.share-project.org/t3/
share/index.php for a full list of funding institutions). We would
like to thank Caspar Looman for his statistical advice.
This study was financially supported by the Erasmus MC, University
Medical Center Rotterdam.
Conflicts of interest: None declared.
Lifestyle, health and social participation were associated with
Small to moderate parts of educational inequalities in frailty
worsening were explained by lifestyle, health and social
To reduce educational inequalities in the development of
frailty, intervention development should take lifestyle,
health and social participation into account. However,
more research is needed to search for other related factors.
1 Clegg A , Young J , Lliffe S , et al. Frailty in elderly people . Lancet 2013 ; 381 : 752 - 62 .
2 Fried LP , Tangen CM , Walston J , et al. Frailty in older adults: evidence for a phenotype . J Gerontol A Biol Sci Med Sci 2001 ; 56 : 146 - 56 .
3 Santos-Eggimann B , Cuenoud P , Spagnoli J , et al. Prevalence of frailty in middleaged and older community-dwelling Europeans living in 10 countries . J Gerontol A Biol Sci Med Sci 2009 ; 64 : 675 - 81 .
4 Gill TM , Gahbauer EA , Allore HG , et al. Transitions between frailty states among community-living elderly . Arch Intern Med 2006 ; 166 : 418 - 23 .
5 Etman A , Burdorf A , Van der Cammen TJM , et al. Socio-demographic determinants of worsening in frailty among community-dwelling older persons in 11 European countries . J Epidemiol Community Health 2012 ; 66 : 1116 - 21 .
6 Espinoza SE , Jung I , Hazuda H . Frailty transitions in the San Antonio Longitudinal Study of Aging. J Am Geriatr Soc 2012 ; 60 : 652 - 60 .
7 Szanton SL , Seplaki CL , Thorpe RJ , et al. Socioeconomic status is associated with frailty: the Women's Health and Aging Studies . J Epidemiol Community Health 2010 ; 64 : 63 - 7 .
8 Boyle PA , Buchman AS , Wilson RS , et al. Physical frailty is associated with incident mild cognitive impairment in community-based older persons . J Am Geriatr Soc 2010 ; 58 : 248 - 55 .
9 Hubbard RE , Searle SD , Mitnitski A , et al. Effect of smoking on the accumulation of deficits, frailty and survival in older adults. A secondary analysis from the Canadian study of health and aging . J Nutr Health Aging 2009 ; 13 : 468 - 72 .
10 Woods NF , LaCroix AZ , Gray SL , et al. Frailty: emergence and consequences in women aged 65 and older in the women's Health Initiative Observational Study . J Am Geriatr Soc 2005 ; 53 : 1321 - 30 .
11 Woo J , Goggins W , Sham A , et al. Social determinants of frailty . Gerontology 2005 ; 51 : 402 - 8 .
12 Mezuk B , Edwards L , Lohman M , et al. Depression and frailty in later life: a synthetic review . Int J Geriatr Psychiatry 2012 ; 16 : 167 - 72 .
13 Cavelaars AE , Kunst AE , Geurts JJ , et al. Educational differences in smoking: international comparison . BMJ 2000 ; 320 : 1102 - 7 .
14 Droomers M , Schrijvers CTM , Mackenbach JP . Educational level and decreases in leisure time physical activity: predictors from the longitudinal GLOBE study . J Epidemiol Community Health 2001 ; 55 : 562 - 68 .
15 Vonneilich N , Jo¨ckel KH , Erbel R , et al. The mediating effect of social relationships on the association between socioeconomic status and subjective health - results from the Heinz Nixdorf Recall cohort study . BMC Public Health 2012 ; 12 : 285 .
16 Mackenbach JP , Stirbu I , Roskam AJ , et al. Socioeconomic inequalities in health in 22 European countries . N Engl J Med 2008 ; 358 : 2468 - 81 .
17 Weyers S , Dragano N , Mobus S , et al. Low socio-economic position is associated with poor social networks and social support: results from the Heinz Nixdorf Recall Study . Int J Equity Health 2008 ; 7 : 13 .
18 Prince MJ , Reischies F , Beekman AT , et al. Development of the EURO-D scale: a European, Union initiative to compare symptoms of depression in 14 European centres . Br J Psychiatry 1999 ; 174 : 330 - 8 .
19 Baron RM , Kenny DA . The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations . J Pers Soc Psychol 1986 ; 51 : 1173 - 82 .
20 Fallah N , Mitnitski A , Searle SD , et al. Transitions in frailty status in older adults in relation to mobility: a multistate modeling approach employing a deficit count . J Am Geriatr Soc 2011 ; 59 : 524 - 9 .
21 Barbeau EM , Krieger N , Soobader MJ . Working class matters: socioeconomic disadvantage, race/ethnicity, gender, and smoking in NHIS 2000 . Am J Public Health 2004 ; 94 : 269 - 78 .
22 Richardson CR , Kriska AM , Lantz PM , Hayward RA . Physical activity and mortality across cardiovascular disease risk groups . Med Sci Sports Exerc 2004 ; 36 : 1923 - 9 .
23 Feinglass J , Lin S , Thompson J , et al. Baseline health, socioeconomic status, and 10-year mortality among older middle-aged Americans: findings from the Health and Retirement Study, 1992 2002. J Gerontol B Psychol Sci Soc Sci 2007 ; 62 : 209 - 17 .
24 Gobbens RJ , Van Assen MA , Luijkx KG , et al. Determinants of frailty . J Am Med Dir Assoc 2010 ; 11 : 356 - 64 .
25 Tolstrup J , Jensen MK , Tjønneland A , et al. Prospective study of alcohol drinking patterns and coronary heart disease in women and men . BMJ 2006 ; 332 : 1244 - 8 .
26 Ruitenberg A , van Swieten JC , Witteman JC , et al. Alcohol consumption and risk of dementia: the Rotterdam Study . Lancet 2002 ; 359 : 281 - 6 .
27 Newman AB , Gottdiener JS , Mcburnie MA , et al. Associations of subclinical cardiovascular disease with frailty . J Gerontol A Biol Sci Med Sci 2001 ; 56 : 158 - 66 .
28 Cramm JM, van Dijk HM , Nieboer AP . The importance of neighborhood social cohesion and social capital for the well being of older adults in the community . Gerontologist 2013 ; 53 : 142 - 52 .
29 Van Oort FVA , Van Lenthe FJ , Mackenbach JP . Material, psyochosocial, and behavioural factors in the explanation of educational inequalities in mortality in the Netherlands . J Epidemiol Commun Health 2005 ; 59 : 214 - 20 .
30 Schrijvers CTM , Strinks K , van de Mheen HD , et al. Explaining educational differences in mortality: the role of behavioral and material factors . Am J Public Health 1999 ; 89 : 535 - 40 .
31 Pickett KE , Pearl M. Multilevel analyses of neighbourhood socioeconomic context and health outcomes: a critical review . J Epidemiol Commun Health 2001 ; 55 : 111 - 22 .
32 Mackenbach JP , Looman CW , Van der Meer JB . Differences in the misreporting of chronic conditions, by level of education: the effect on inequalities in prevalence rates . Am J Public Health 1996 ; 86 : 706 - 11 .
33 Dowler E. Inequalities in diet and physical activity in Europe . Public Health Nutr 2001 ; 4 : 701 - 9 .
34 Watkinson C, Van Sluijs EM , Sutton S , et al. Overestimation of physical activity level is associated with lower BMI: a cross-sectional analysis . Int J Behav Nutr Phys Act 2010 ; 7 : 68 .
35 Lin JC , Guerrieri JG , Moore AA . Drinking patterns and the development of functional limitations in older adults: longitudinal analyses of the health and retirement survey . J Aging Health 2011 ; 23 : 806 - 21 .
36 Jin M , Cai S , Guo J , et al. Alcohol drinking and all cancer mortality: a meta-analysis . Ann Oncol 2013 ; 24 : 807 - 16 .
37 Kamphuis CBM , Turrell G , Giskes K , et al. Socioeconomic inequalities in cardiovascular mortality and the role of childhood socioeconomic conditions and adulthood risk factors: a prospective cohort study with 17-years of follow up . BMC Public Health 2012 ; 5 : 1045 .
38 Droomers M , Schrijvers CT , Stronks K , et al. Educational differences in excessive alcohol consumption: the role of psychosocial and material stressors . Prev Med 1999 ; 29 : 1 - 10 .