European Journal of Clinical Nutrition (2001) 55, 366±373
ß 2001 Nature Publishing Group All rights reserved 0954±3007/01 $15.00
www.nature.com/ejcn
Original Communication
Social position and nutrition: a gradient relationship in Canada
and the USA
L Dubois1,2* and M Girard1
1
Canadian Institute for Advanced Research, Toronto, Ontario, Canada; and 2DeÂpartement de MeÂdecine Sociale et PreÂventive, Faculte de
MeÂdecine, Universite Laval, Sainte-Foy, QueÂbec, Canada
Obejctive: To study the existence of a graded relationship between the nutrient content of the diet and some
measures of social position.
Design and setting: The graded relationship hypothesis was veri®ed by secondary analysis performed on two
different nutrition surveys: the Quebec Nutrition Survey (QNS) and the NHANES III, both based on a single 24 h
recall. The data of these surveys were collected on a representative sample of two different populations, the ®rst
(n ˆ 2103) in the province of Quebec (Canada) in 1990 (QNS) and the second in the US population (n ˆ 14 877)
between 1988 and 1994 (NHANES III).
Main outcomes measures: The social gradient hypothesis was tested with four different social position indicators.
The analyses were performed separately for men and women aged from 18 to 74 y, for all the nutrients available in
the databanks. For the USA, the graded relationship for the total population, for the non-Hispanic white population
and for respondents sometimes or often experiencing a lack of food was also measured.
Results: A graded relationship between almost all nutrients and the studied social position measurements is
observed for the consumption of total calorie-adjusted nutrients, for the proportion of people eating in accordance
with dietary guidelines and for the proportion of individual not meeting 75% of the respective Recommended
Dietary Allowances for their country. The direction of these relationships (positive or negative) is mainly in
accordance with the clinical known impacts of nutrients on chronic diseases.
Conclusions: It is possible that nutrition plays a role in the graded distribution of social health inequalities in
North America even in the magnitude of this contribution remains to be evaluated.
Descriptors: nutrition survey; socio-economic gradient; health inequalities; chronic diseases; population health;
North America
European Journal of Clinical Nutrition (2001) 55, 366±373
Introduction
Marmot (1994) was the ®rst to demonstrate a graded
relationship between socio-economic status and several
chronic diseases. This researcher put into perspective the
fact that health inequalities are not just the result of
material deprivation because even relatively well-off persons have higher rates of disease. This gradient has now
been observed in different countries and for almost all
causes of morbidity and mortality as well as for important
health risk factors such as tobacco, obesity and blood
*Correspondence: L Dubois, DeÂpartement de MeÂdecine Sociale et
PreÂventive, Faculte de MeÂdecine, Pavillon de l'Est, bureau 1105,
Universite Laval, 2180 Chemin Sainte-Foy, Sainte-Foy, QueÂbec, Canada
G1K 7P4.
E-mail: [email protected]
Guarantors: L Dubois.
Contributors: LD was the principal investigator. MG was the statistician.
Received 24 February 2000; revised 30 November 2000;
accepted 6 December 2000
cholesterol levels (Power et al, 1991; Evans, 1994;
Marmot, 1994; Acheson et al, 1998; Kubzansky et al,
1998; Anan et al, 1999; Martinez et al, 1999; Woo et al,
1999). This gradient persists even after these risk factors
have been taken into account, suggesting that other risk
factors have been overlooked. Nutrition is one of the
factors not yet studied extensively with regard to the
gradient. Previous research had put in perspective the fact
that different social position indicators are positively
related with different aspects of the quality of the diet
(Bolton-Smith et al, 1991; Hulshof et al, 1991; Smith &
Baghurst, 1992; Shea et al, 1993; Krebs-Smith et al, 1996;
Bowman et al, 1998; Lino et al, 1998; Roos et al, 1998;
Shi, 1998; Woo et al, 1999). These studies generally did
not measure the existence of a social gradient in all the
nutrient content of the diet for various social position
measurements. The aim of this research is to verify the
existence of such a gradient at the population level and to
evaluate how the use of different socio-economic indicators
shapes the results of the analysis.
Social position and nutrition
L Dubois and M Girard
Methods
Study design and sample
The purpose of this research was to verify the existence of a
socio-economic gradient for various aspects of the quality
of the diet on two different populations. The analyses were
performed on data previously collected in two nutrition
surveys, the 1990 Quebec Nutrition Survey (QNS) and the
NHANES III (1988 ± 1994). Both surveys, based on a
representative sample of a population, collected a 24 h
recall Ð repeated on a sub-sample Ð to evaluate the content of the diet. In Quebec, 1023 men and 1080 women
aged 18 ± 74 were surveyed. For the United States, the same
age group (18 ± 74 y) was studied for the total sample
(n ˆ 6962 men and 7915 women) and for the non-Hispanic
white population (NH White) (n ˆ 2499 men and 2911
women). This choice was made because there was not a
signi®cant presence of black and Hispanic people in the
QNS. Both surveys had high response rates (69% for QNS,
73% for NHANES III). Analyses were also performed for
people experiencing food insecurity. More details and
previous results on these surveys have been published
elsewhere (Federation of American Societies for Experimental Biology, 1995; Sante QueÂbec & Bertrand, 1995).
De®nition of variables
A gradient relationship is observed when the studied
variable increases or decreases gradually with the improvement of social position. Therefore, a positive gradient
means a higher nutrient density for higher social status
individuals and a negative gradient implies lower nutrient
density for lower social status individuals. Analyses were
performed using four different social position indicators:
relative education, income adequacy, working class and a
global socio-economic status (SES) scale. Education was
grouped by percentiles in three categories for each age
group and for each databank. Income used for QNS data
was calculated from household gross income adjusted for
household size, while the poverty index in three categories,
based on household size gross income and income threshold, was used for NHANES III data. These scales are not
identical but are based on the same process of classi®cation, both taking into account the household size. Five
categories based on the Pineo classi®cation (in 16 categories; Pineo et al, 1979) for both datasets were used to
classify the respondents in accordance with their reported
working class. Non-workers (eg students, retired, housewives, etc), classi®ed in a sixth category, were not included
in the analysis. As suggested by Willms and Shields
(1996), the global SES score is derived from the level of
education (last grade attended for QNS and years of
schooling for NHANES III), the modi®ed Pineo's prestige
score (work classi®cation in QNS and type of work for
NHANES III), and last year household gross income. Each
component of the global score was transformed to achieve
the continuity criteria and was standardized, making it
centred at 0 with a variance at 1. These elements were
then added up to give the global measure as a continuous
variable, before being regrouped in three categories. Dietary supplements are included in the intake ®gures. Table 1
gives the proportion of the studied population in each
social position categories.
The quality of the diet was evaluated by the nutrient
content of the diet. The existence of a social gradient was
367
Table 1 Characteristics of the studied populations for each social position indicators, in percentage (weighted
values)
QNS
Social position
indicators and categories
NH III-All
NH III-NH White
Men
(n ˆ 1023)
Women
(n ˆ 1080)
Men
(n ˆ 6962)
Women
(n ˆ 7915)
Men
(n ˆ 2499)
Women
(n ˆ 2911)
Relative education
Low
Middle
High
39.0
38.8
22.2
41.9
33.0
25.1
17.5
47.3
35.2
23.4
38.2
38.3
12.1
48.3
39.6
18.2
38.9
42.9
Income adequacy
Low
Middle
High
16.1
42.9
41.0
23.6
42.5
34.0
16.0
45.9
38.1
20.6
44.3
35.1
10.6
45.1
44.3
14.1
44.5
41.4
Working class
Not working
Blue collar
White collar
Of®ce worker
Semi-professional
Professional
23.5
10.9
28.7
15.4
11.4
10.1
43.0
2.6
5.6
33.7
9.7
5.4
17.9
26.3
17.2
10.1
10.3
18.2
34.6
9.0
5.7
28.6
5.9
16.2
16.5
23.8
17.2
9.7
11.9
21.0
32.6
7.8
5.2
29.3
6.9
18.3
Global SES score
Low
Middle
High
33.4
33.6
33.0
34.0
33.8
32.2
34.9
32.9
32.2
33.7
32.9
33.5
26.8
35.6
37.6
26.9
34.0
39.1
European Journal of Clinical Nutrition
Social position and nutrition
L Dubois and M Girard
368
measured for energy and 31 nutrients for the QNS and for
energy and 26 nutrients for the NHANES III. Analyses
were performed for men and women separately because of
the known differences in nutrient needs and health outcomes by sex, and of the complexity of the study of social
position indicators for women in comparison with men.
Outlier values (less than 1% of all data) were omitted using
the extreme studentized deviate procedure (Rosner, 1995)
from each nutrient variable before being standardized.
Even though they were not of®cially identi®ed as data
entry errors, examination of these outliers suggested their
elimination from the analyses. To use a measure independent of total caloric intake, the analyses were performed or
energy-adjusted nutrients as suggested by Willett (1990).
We did not control for the possibility of selective underreporting of energy intake between the subgroups of social
status because of the lack of adequate data to do so.
To evaluate a possible role of nutrition in social
health inequalities, it was insuf®cient to measure only
the existence of a graded relationship between the
consumption of different nutrients and some indicators
of social position. A graded relationship in the case of a
population able to cover its basic needs for nutrients
would not necessarily indicate a possible role for nutrition in the social health inequalities observed in the
same population. For this reason, the existence of a
social gradient for people not meeting 75% of the
respective recommended dietary allowances (RDAs) for
each nutrient (not energy-adjusted) and for people not
meeting some nutrition recommendations were also
measured. Although the dietary method used in the
QNS and the NHANES III was a 24 h recall, which
could not be considered as representative of the average
nutrient intake for each individual, this method is suitable for group comparison. The Canadian RDAs (Health
and Welfare Canada, 1990) were used for the QNS and
the USA RDAs (National Research Council, 1989) for
the NHANES III. These RDAs vary slightly but their
scienti®c basis is similar. Moreover, these recommendations were the ones being transmitted to these respective
populations at the period of data collection. In our
analyses, we were also preoccupied by the people in
Table 2 Graded relationships of relative education (®rst symbol), family income (second symbol), working class (third symbol) and SES (fourth symbol)
(positive :, negative ;, or no association s; P 0.05) with total nutrients (energy adjusted), people aged 18 ± 74 y, by sex for Quebec Nutrition Survey (QNS)
and NHANES III (NH III)
QNS
Energy and nutrients
Energy (kcal)
Carbohydrates (g)
Total dietary ®bre (g)
Proteins (g)
Total fats (g)
Cholesterol (mg)
Saturated fatty acids (g)
Monounsaturated fat. acids (mg)
Polyunsaturated fatty acids (mg)
Omega-3 fatty acids (mg)
Omega-6 fatty acids (mg)
Alcohol (g)
Caffeine (mg)
Carotene (RE)
Vitamin A (RE)
Vitamin D (mg)
Vitamin E (a-tocopherol eq.)
Vitamin C (mg)
Thiamin (mg)
Ribo¯avin (mg)
Niacin (mg)
Vitamin B6 (mg)
Folate (mg)
Vitamin B12 (mg)
Pantothenic (mg)
Calcium (mg)
Phosphorus (mg)
Magnesium (mg)
Iron (mg)
Zinc (mg)
Sodium (mg)
Potassium (mg)
NH III-All
Men
(n ˆ 1023)
Women
(n ˆ 1080)
Men
(n ˆ 6962)
Women
(n ˆ 7915)
Men
(n ˆ 2499)
Women
(n ˆ 2911)
ss;s
ssss
:s:s
ssss
s:s:
;s;;
s:s:
s:;:
s:ss
s:s:
ssss
::::
::::
ssss
ss:s
ssss
sss:
:s::
ssss
ssss
ssss
ssss
:s::
ssss
:sss
::::
::::
::::
ssss
ssss
ssss
:s::
sss:
ssss
:sss
s:s:
s:s:
sss:
sss:
s:s:
s:ss
s:s:
:sss
s:s:
::::
:sss
:sss
ssss
s:s:
::s:
s:ss
s:s:
s:ss
ssss
::::
ssss
::ss
::::
::::
:::s
ssss
;sss
s:s:
:sss
ss;:
;;;;
::::
;;ss
;sss
;;;;
;;;;
;s;s
::::
Ð
Ð
s:ss
Ð
:s::
::::
Ð
::::
::::
:sss
:s::
s:::
::::
::::
ss:s
Ð
:sss
;;ss
::::
::::
;s;s
;;ss
::::
;ss:
;;;;
::::
::s:
;sss
;;s;
;sss
;sss
;:::
Ð
Ð
::::
Ð
::::
::::
Ð
::::
:ss:
:sss
::::
::::
::s:
::::
:sss
Ð
::::
::::
::::
::::
::s:
;:ss
::::
;s;s
;;;;
::::
;sss
;sss
;;;;
;;;;
;s;s
::::
Ð
Ð
ssss
Ð
:s::
:s::
Ð
::::
::::
:sss
s;ss
s:::
::::
::::
;sss
Ð
s;ss
;;ss
::::
::::
ssss
;sss
::::
sss:
;;s;
::::
::::
;sss
s;ss
;sss
;sss
s:::
Ð
Ð
s:s:
Ð
::::
::::
Ð
::::
::::
:ss:
:ss:
::::
::s:
::::
ssss
Ð
:s::
::::
::::
::s:
::s:
;:ss
::::
s: no graded relationship;:: positive graded relationship;;: negative graded relationship; Ð : no data.
European Journal of Clinical Nutrition
NH III-NH White
Social position and nutrition
L Dubois and M Girard
material deprivation. A proportion of 7% of the interviewed people in NHANES III declared that they `sometimes' or `often' do not have enough food to eat. These
respondents were not removed from the analyses because
their distribution in the sample was not exclusive to the
lower social position categories. Moreover, the results of
the analyses do not differ if they are excluded or not
from the statistical calculations. Nevertheless, speci®c
analyses were performed to evaluate the existence of an
educational gradient within this subgroup of the population (NHANES-All). Data related to food insecurity
were not available in the QNS.
Trend analysis was carried out to evaluate the presence
of a gradient using the Cochran-Armitage trend test (Fleiss,
1981). Statistical analysis were performed with SAS software (version 6.12).
Results
Table 2 shows the existence of graded associations between
the studied social position indicators and the nutrients
available in each databank. For example, for total dietary
®bre, the QNS data shows for men a positive graded
relationship for education, no graded relationship for
family income, a positive graded relationship for working
class and no graded relationship for the global SES score.
For women of the QNS, a graded relationship is only
observed with education and this relationship is positive.
For the NHANES III dataset, the relationships between
®bre and the studied social position indicators are all
gradual and positive.
More graded relationships are observed with the
NHANES III databank than with the QNS and these
relationships are generally stronger. For indication only,
the proportion of relationships at P 0.001 is 37% for
QNS, 83% for NHANES-All and 69% for NHANES-NH
White. These differences may be due to the small numbers
in some categories in the QNS, such as the proportion of
women in the `professional category'. Nevertheless, in the
Quebec population, 16 elements for men and 26 for women
(out of 32) show a positive graded relationship with at least
one of the social position indicators. A negative graded
relationship is observed with energy and cholesterol for
men and with zinc for women, while monounsaturated fat
shows an unstable direction for men. The NHANES III data
shows a more constant portrait, presenting a positive
relationship for ®bre, proteins (women only), polyunsaturated fat (except with education for women), alcohol and
almost all of the vitamins and minerals studied, and a
negative relationship for carbohydrates, proteins (men
only), total fats, cholesterol, saturated fat, monounsaturated
fatty acids and phosphorus (men only). The SES scale
shows the same kind of relationships. A smaller number
of graded relationships is observed for working class
women than for men, perhaps due to the fact that `working
class' is not as good an indicator of social position for
women as it is for men. In NHANES III, a graded relationship is more commonly observed for the total population
than for the NH White population, and for education in
comparison with the other measurements of social position,
for men and for women. Only education was used to
measure the gradient for the individuals having experienced
`sometimes or often' a lack of food in NHANES III. For
369
Table 3 Proportion of the population not meeting 75% of the RDAs (18 ± 74 y) by sex, for Quebec Nutrition
Survey (QNS) and NHANES III (NH III)
QNS
Energy and nutrients
Energy
Proteins
Omega-3
Omega-6
Vitamin A
Vitamin E
Vitamin C
Thiamin
Ribo¯avin
Niacin
Vitamin B6
Folate
Vitamin B12
Calcium
Phosphorus
Magnesium
Iron
Zinc
NH III-All
NH III-NH White
Men
(n ˆ 1023)
Women
(n ˆ 1080)
Men
(n ˆ 6962)
Women
(n ˆ 7915)
Men
(n ˆ 2499)
Women
(n ˆ 2911)
24
5
24
27
40
31
18
8
8
1a
32
35
8
26
7
10
31
25
38
6
30
37
40
32
15
12
11
3a
40
30
20
46
14
30
32
32
34
11
Ð
Ð
57
45
31
18
18
14
33
20
11
41
8
37
10
47
44
16
Ð
Ð
55
48
36
21
24
21
42
31
23
57
21
44
44
58
29
9
Ð
Ð
46
40
29
13
11
8
31
16
9
32
6
30
7
44
41
15
Ð
Ð
46
46
36
18
18
16
40
26
20
50
16
37
37
55
Ð : no data. aLess than 5% of the sample.
European Journal of Clinical Nutrition
Social position and nutrition
L Dubois and M Girard
370
Table 4 Graded relationships for relative education (®rst symbol), relative family income (second symbol), working class (third symbol) and SES (fourth
symbol) (positive :, negative ;, or no association, s; P 0.05) for the proportion of the population not meeting 75% of the RDAs, for people aged 18 ± 74 y,
by sex, for Quebec Nutrition Survey (QNS) and NHANES III (NH III)
QNS
Energy and nutrients
Energy
Proteins
Omega-3
Omega-6
Vitamin A
Vitamin E
Vitamin C
Thiamin
Ribo¯avin
Niacin
Vitamin B6
Folate
Vitamin B12
Calcium
Phosphorus
Magnesium
Iron
Zinc
NH III-All
NH III-NH White
Men
(n ˆ 1023)
Women
(n ˆ 1080)
Men
(n ˆ 6962)
Women
(n ˆ 7915)
Men
(n ˆ 2499)
Women
(n ˆ 2911)
ssss
ssss
ssss
ssss
s;;;
ssss
ssss
ssss
ssss
s;ssa
ssss
;;;;
ssss
s;s;
;;s;
;ss;
ssss
ssss
;ss;
;;s;
ss;;
;ss;
;ss;
;;s;
;ss;
ssss
;;s;
sss;a
;;s;
;ss;
ssss
;s;;
;;;;
;ss;
ssss
s;s;
;;s;
;;;;
Ð
Ð
;;;;
;;;;
;;;;
;;s;
;;;;
;;;;
;;;;
;;;;
s;s;
;;;;
;;s;
;;;;
;;;;
s;s;
s;s;
;;s;
Ð
Ð
;;;;
;;;;
;;;;
;;s;
;;;;
;;s;
;;s;
;;;;
s;s;
;;;;
;;;;
;;;;
;;s;
s;s;
;;s;
;;;;
Ð
Ð
;;;;
;;;;
;;;;
;;s;
;;;s
;;;;
;;;;
;;;;
ssss
;ss;
;;s;
;;;;
;;;;
s;s;
s;s;
;;s;
Ð
Ð
;;;;
;;;;
;;;;
;;s;
;;;;
;;s;
;;s;
;;;;
sss;
;;;;
;;;;
;;;;
;;s;
;;s;
s: no graded relationship;:: positive graded relationship;;: negative graded relationship, Ð : no data. aLess than 5% of the sample.
men, a negative gradient is observed for proteins and
monounsaturated fat, and a positive gradient appears for
carotene, vitamin A, vitamin C, thiamin, riboblavin, niacin,
folate, calcium, iron and sodium. For women, the analyses
show a negative gradient for carbohydrates and a positive
one for vitamin A, niacin, vitamin B6, magnesium and
potassium (data not shown). These results indicate that,
even for people experiencing material dif®culties, the
quality of the diet improves with education.
Table 3 presents the proportion of the population not
meeting 75% of the RDAs and Table 4 shows the existence
of a graded relationship with social position for these
individuals. Again, a larger number of graded relationships
appears with the NHANES III data than with the QNS data.
All vitamins and minerals show a negative association,
meaning that gradually more individuals in each category
are not reaching 75% of the RDAs with the lowering of
social position. This important observation means that
nutrition could play a role in social health inequalities not
only for people experiencing material deprivation, but also
for the whole population. In the USA, more graded relationships are observed with education and income than with
working class.
Table 5 shows the proportion of the population not
meeting some nutrition recommendations and Table 6
presents the graded relationships with social position for
these individuals. These associations, even if less stable,
are positive for carbohydrates (except for QNS men),
alcohol and sodium (except for QNS men), and negative
for total fat (except QNS women), saturated fat and
cholesterol (except QNS women). These results suggest
that going down the social hierarchy, gradually more
people eat more total fat, cholesterol and saturated fat
than is recommended.
Table 5 Proportion of the population not meeting the nutrition recommendations (18 ± 74 y) by sex, for
Quebec Nutrition Survey (QNS) and NHANES III (NH III)
QNS
NH III-All
NH III-NH White
Men
Women
Men
Women
Men
Women
(n ˆ 1023) (n ˆ 1080) (n ˆ 6962) (n ˆ 7915) (n ˆ 2499) (n ˆ 2911)
Carbohydrates (less than 55% of energy)
Total fat (more than 30% of energy)
Cholesterol (more than 300 mg)
Saturated fat (more than 10% of energy)
Alcohol (more than 5%=2.5% of energy)
Sodium (more than 2400 mg)
European Journal of Clinical Nutrition
80
70
53
73
44
76
74
65
30
68
56
49
77
64
49
58
77
77
67
63
28
57
12
56
77
68
43
64
48
81
69
63
21
59
52
58
Social position and nutrition
L Dubois and M Girard
371
Table 6 Graded relationship for relative education (®rst symbol), relative family income (second symbol), working class (third symbol) and SES (fourth
symbol) (positive :, negative ;, or no association s; P < 0.05) for the proportion of the population not meeting the nutrition recommendations, for people
aged 18 ± 74 y, by sex, for Quebec Nutrition Survey (QNS) and NHANES III (NH III)
QNS
Carbohydrates (less than 55% of energy)
Total fat (more than 30% of energy)
Cholesterol (more than 300 mg)
Saturated fat (more than 10% of energy)
Alcohol (more than 5 ± 2.5% of energy)
Sodium (more than 2400 mg)
NH III-All
NH III-NH White
Men
(n ˆ 1023)
Women
(n ˆ 1080)
Men
(n ˆ 6962)
Women
(n ˆ 7915)
Men
(n ˆ 2499)
Women
(n ˆ 2911)
;sss
;sss
ss;s
;sss
::::
;s;s
s:s:
sss:
sss:
ssss
::s:
s:s:
ssss
ssss
;;;;
;s;:
s:s:
::s:
s:s:
ssss
;;s;
;sss
::::
s:s:
ssss
ssss
;;;;
;;;;
::s:
s:s:
s:s:
;s;s
s;ss
;s;s
::s:
s:s:
s: no graded relationship;:: positive graded relationship;;: negative graded relationship.
Discussion
The role of social position in the quality of the diet
The results of this research put into perspective two major
elements. First, the graded in¯uence of social position on
the quality of the diet indicates a possible intermediate role
for nutrition in the distribution of social health inequalities
in North America. This relationship should then be more
extensively studied in line with the known clinical role of
nutrition in chronic diseases, which is beyond the purpose
of this paper. The second important element is the existence
of inequalities even within the materially deprived subgroup of the population. This should not only be linked
with the quality of the diet, but also with the social
comparison process involved in the population health
gradient (Wilkinson, 1996). In that context, material deprivation for people unable to afford the growing diversity of
foods with added value (prepared, transformed, engineered,
etc) highly marketed by the mass-media is compounded by
social deprivation in regards to the majority who can afford
them in wealthy societies. Four different social position
indicators were studied separately in this research because
these indicators are not totally overlapping in the study of
health and socio-economic status (Adler et al, 1994). As
put in perspective by Adler et al (1994) in a more general
way, the fact that a graded association was found with each
studied socio-economic indicators indicates that a `broader
dimension of social strati®cation or social ordering is the
potent factor'. Moreover, the in¯uence of different social
position measurements on the quality of the diet is probably
affected in different ways and the use of indicators at the
individual and family level is essential to nutrition studies.
For example education level is important for the comprehension of the information regarding the relationship
between diet and health on a long-term basis. Family
income plays a direct role in food expenditures in stores
and restaurants, while the type and place of work could
relate to food availability at lunch time and time allowed
for meal preparation and consumption. This is of particular
interest for working women, where having a job means
more family money to buy food but less time to prepare it.
Money also makes the difference for the socially deprived
elderly unable to buy food services or to go out at lunch
time (Dubois et al, 1999). More globally, food choices
could also refer to a status symbol (Bourdieu, 1979) to be
put in relation with the health food movement in North
America (McIntosh, 1996). These elements could explain
the fact that even the same nutritional program does not
have the same impact on different social groups, as
observed by Whitehead (1995).
The results of this study are in accordance with other
research showing a relationship between some nutrients
and different social position indicators. Smith and Baghurst
(1992) studied three social status indicators (occupation,
education and income) and some elements of the diet for an
Australian urban population. In their study, a higher social
status was generally associated with a better diet, especially
higher ®bre and alcohol densities and lower fat and re®ned
sugar densities. They found no association between social
status and salt, polyunsaturated fat, protein and complex
carbohydrates. Moreover, for some elements of the diet,
they found an association with only one or two of the three
studied social position indicators, as it is the case in our
study. The analysis of the global Healthy Eating Index
(HEI) developed by the UDSA also show that higher
education and income are associated with a higher HEI
score (Variyam et al, 1998). The analyses were generally
performed using only one social position indicator, such as
education. Murphy et al, (1992) found an association
between a low level of education in USA and the number
of nutrients below two-thirds of the RDA level. Woo et al
(1999), in the Hong Kong Dietary and Cardiovascular Risk
and Prevention Survey, also observed that a higher level of
education was associated with a higher proportion of
protein, ®bre and calcium intake for men, and a higher
proportion of ®bre, calcium, fat, niacin, vitamin D and
polyunsaturated fat intake for women. Hupkens et al (1997)
also observed a positive association between education
level and fat and ®bre consumption for women living in
two European cities, but they found no association for
women living in a third city. In their study, percentage of
calories for one nutrient was sometimes associated with
European Journal of Clinical Nutrition
Social position and nutrition
L Dubois and M Girard
372
education, but not total grams. They also observed that
education is a better predictor of food consumption for
women than for men (Hupkens et al, 1997).
These results illustrate the complexity of the study of
social factors in relation to the quality of the diet. The
content of the diet should be studied separately for men and
women. Moreover, other elements should be part of the
study, such as eating out, which is increasing in developed
countries (Eck Clemens et al, 1999). This social variable
could explain part of the inconsistency in the relationship
for some nutrients between QNS and NHANES III. In fact,
eating out is more common in higher social class and
affects negatively the quality of the diet (Dubois et al,
2000), especially total energy, fat, saturated fat, sodium and
proteins (Haines et al, 1992; LeFrancËois et al, 1996; Eck
Clemens et al, 1999). These elements are important since
the recent rise in energy consumption observed in the USA
induced an increase in absolute fat consumption even if
relative fat consumption is declining (Anan & Basiotis,
1998; Katz et al, 1998). As stated by Roos et al (1998)
following their analysis of the FinMonica study, structural
position and family status are of greater in¯uence on
women's food choices, whereas education level and marital
status are more important for men. It is also important to
follow the trends of inequalities in food consumption over
the years. PraÈttaÈllaÈ et al (1991) found that Finns from lower
education levels followed the trends in the quality of the
diet of people with higher levels of education with a 10 y
lag, in a study done between 1979 and 1990. Popkin et al
(1996) for the US population also indicates that social
inequalities in the quality of the diet had narrowed up to
1990.
Therefore, analyses of the relationship between social
position and the quality of the diet could only bene®t from
improved indicators. For example, the absence of a graded
relationship with some dietary variables could be related to
the lack of precision in the nutrients measured in the
studied surveys (eg vitamin D and table salt in the QNS).
As mentioned by Hulshof et al (1991) in their study (with a
2 day food record), the use of a 1 day record probably
weakens the strength of the relationships observed in our
analyses. A more representative data collection method
could reveal even more graded relationships. The same
applies to socio-economic indicators. For example, working position is not necessarily the best indicator for women
in older age groups as an important proportion of them do
not work. A socio-economic indicator taking into account
the diversity of family structures in North America could
re¯ect more adequately the situation, as food is a resource
shared by all members of a family living in a given social
context.
Public health and policy implications
Over the last 30 y, nutrition has been classi®ed as a lifestyle. Consequently, education remains the main way for
public health nutritionists to intervene. That is not to say
that nutrition studies have never focused on social inequalities, but as is often the case in the study of socio-economic
European Journal of Clinical Nutrition
status and health (Adler et al, 1994), previous nutrition
studies were more interested in the lower SES Ð still a
fundamental topic of interest Ð in comparison with the
other SES categories. This way of thinking is quite
normal in nutrition since developed countries are experiencing food abundancy. Eating is then considered as a basic
need that is easily met by the majority. Yet the results of
this research show the existence of social inequalities in the
content of the diet within all population sub-groups. It
should then be taken into consideration in public health
interventions intended to eradicate social inequalities in
food access as well as in cultural food choices. At a policy
level, the harmonization of nutrition recommendations in
North America would bene®t from a better understanding
of the relationship between SES and nutrition. These
elements should be analysed in relation with population
health, especially because the scale of inequalities is
different in Canada and in the USA (Wolfson & Murphy,
1998). In the years to come, we have to make sure that
these differences will not be ampli®ed by the continuing
transformation of work, growing family instability, the
aging of the population and the rapid development of
new nutraceuticals, those high-tech foods engineered in a
health perspective (Dubois, 1996). Good indicators of diet
quality and adequate social position measurements should
then be part of a monitoring system oriented towards the
reduction of nutritional health inequalities in both countries. The reduction of these inequalities will contribute to
achieve Health for All in the twenty-®rst century.
Acknowledgements Ð This work has bene®ted from a research grant from
the Conseil QueÂbeÂcois de la Recherche Sociale (CQRS). We also extend
our gratitude to the Canadian Institute for Advanced Research for its
continuous support.
References
Acheson D, Barker D, Chambers J, Graham H, Marmot M & Whitehead M
(1998): Independent Inquiry into Inequalities in Health Report.
London: The Stationary Of®ce.
Adler NE, Boyce T, Chesbey MA et al (1994): Socioeconomic status and
health. The challenge in the gradient. Am. Psychol. 49, 15 ± 24.
Anan RS & Basiotis PP (1998): Is total fat consumption really decreasing?
Family Econ. Nutr. Rev. 11, 58 ± 60.
Anan RS, Basiotos PP & Klein BW (1999): Pro®le of overweight children.
USDA Center for Nutrition Policy and Promotion, Nutrition=Insights
13.
Bolton-Smith C, Smith WCS, Woodward M & Tunstall-Pedoe H (1991):
Nutrient intakes of different social-class groups: results from the
Scottish Heart Health Study (SHHS). Br. J. Nutr. 65, 321 ± 335.
Bourdieu P (1979): La Distinction. Critique Sociale du Jugement. Paris:
Les eÂditions de Minuit.
Bowman SA, Lino M, Gerrior SA & Basiotis PP (1998): The Healthy
Eating Index, 1994 ± 1996. Family Econ. Nutr. Rev. 11, 2 ± 14.
Dubois L (1996): L'aliment, un futur miracle de la biotechnologie?
Sociolo. Soc. XXVIII, 45 ± 57.
Dubois L, Labrecque J, Girard M, Grignon R & Damestoy N (1999):
DeÂterminants des dif®culteÂs relieÂes aÁ l'alimentation dans un groupe de
personnes aÃgeÂes non-institutionnaliseÂes du QueÂbec. Nutrition et vieillissement. L'anneÂe geÂrontologique, Serdi EÂdition, p 21 ± 52. Paris.
Social position and nutrition
L Dubois and M Girard
Dubois L, Beauchesne EÂ, Girard M, BeÂdard B, Bertrand L & Hamelin AM
(2000): Alimentation: perceptions, pratiques et inseÂcurite alimentaire,
In EnqueÃte Sociale et de Sante 1998, Chap 6, pp 149 ± 170, Institut de
la statistique du QueÂbec.
Eck Clemens LH, Slawson DL & Klesges RC (1999): The effect of eating
out on quality of diet in premenopausal women. J. Am. Diet. Assoc. 99,
442 ± 444.
Evans RG (1994): Introduction. In Why are Some People Healthy and
Others not? The Determinants of Health of Populations, ed. RC Evans,
ML Barer, TR Marmor, pp 3 ± 26. New York: Aldine de Gruyter.
Federation of American Societies for Experimental Biology, Life Sciences
Research Of®ce (1995): Third Report on Nutrition Monitoring in the
United States. Washington, DC: Government Printing Of®ce.
Fleiss JL (1981): Statistical Methods for Rates and Proportions. New
York: John Wiley.
Haines PS, Hungerford DW, Popkin BM & Guilkey DK (1992): Eating
patterns and energy and nutrient intakes of US women. J. Am. Diet.
Assoc. 92, 698 ± 704, 707.
Health and Welfare Canada (1990): Nutrition Recommendations. The
Report of the Scienti®c Review Committee 1990. Ottawa.
Hulshof KFAM, LoÈwik MRH, Kok FJ, Wedel M, Brants MAM, Hermus
RJJ & ten Hoor F (1991): Diet and other life-style factors in high and
low socioeconomic groups (Dutch Nutrition Surveillance System). Eur.
J. Clin. Nutr. 45, 441 ± 450.
Hupkens C, Knibbe RA & Drop MJ (1997): Social class differences in
women's fat and ®bre consumption. A cross-national study. Appetite 28,
131 ± 149.
Katz DL, Brunner RL, St-Jeor Sachiko T, Scott B, Jekel JF & Brownell
KD (1998): Dietary fat consumption in a cohort of American adults,
1985 ± 1991: covariates, secular trends and compliance with guidelines.
Am. J. Health Prom. 12, 382 ± 390.
Krebs-Smith SM, Cook A, Subar AF, Cleveland L, Friday J & Kahle LL
(1996): Fruit and vegetable intakes of children and adolescents in the
United-States. Arch. Pediatr. Adolesc. Med. 150, 81 ± 86.
Kubzansky LD, Berkman LF, Glass TA & Seeman TE (1998): Is educational attainment associated with shared determinants of health in
elderly? Findings from the MacArthur Studies of Successful Aging.
Psychosom. Med. 60, 578 ± 585.
LeFrancËois P, Calamassi-Tran G, HeÂbel P, Renault C, Lebreton S &
Volatier JL (1996): Food and nutrient intake outside the home of 629
French people of 50 y and over. Eur. J. Clin. Nutr. 50, 828 ± 831.
Lino M, Gerrior SA, Basiotis PP & Anan RS (1998): Report card on the
diet quality of children. USDA Center for nutrition policy and promotion, Nutrition=Insights 9.
Marmot MG (1994): Social differentials in health within and between
populations. Daedalus Health Wealth 123, 197 ± 216.
Marmot MG & Mustard JF (1994): Coronary heart disease from a
population perspective. In Why are Some People Healthy and Others
Not? The Determinants of Health of Populations. pp 189 ± 214. New
York, Aldine de Gruyter.
Martinez JA, Hearney JM, Kafatos A, Paquet S & Martinez-Gonzales MA
(1999): Variables independently associated with self-reported obesity in
the European Union. Public Health Nutr. 2, 125 ± 133.
McIntosh WA (1996): Sociologies of Food and Nutrition. New York:
Plenum Press.
Murphy SP, Rose D, Hudes M & Viteri FE (1992): Demographic and
economic factors associated with dietary quality for adults in the 1987 ±
1988 Nationwide Food Consumption Survey. J. Am. Diet. Assoc. 92,
1352 ± 1357.
National Research Council (1989): Recommended Dietary Allowances,
10th edn. Washington, DC: National Academy Press.
Pineo PC, Porter J & Roberts HA (1979): The 1971 census and the
socioeconomic classi®cation of occupations. Can. J. Soc. Anthrop. 14,
91 ± 102.
Popkin BM, Siega-Riz AM & Haines PS (1996): A comparison of dietary
trends among racial and socioeconomic groups in the United States.
New Engl. J. Med. 335, 716 ± 720.
Power C, Manor O & Fox J (1991): Health and Class, the Early Years.
London: Chapman and Hall.
PraÈttaÈllaÈ R, Berg MA & Puska P (1991): Diminishing or increasing social
contrasts? Social class variation in Finnish food consumption patterns,
1979 ± 1990. Eur. J. Clin. Nutr. 46, 279 ± 287.
Roos E, Lahelma E, Virtanen M, PraÈtaÈllaÈ R & Pietinen P (1998): Gender,
socioeconomic status and family status as determinants of food behaviour. Soc. Sci. Med. 46, 1519 ± 1529.
Rosner BA (1995): Fundamentals of Biostatistics, 4th edn. Boston:
Duxbury Press.
Sante Quebec & Bertrand L (sous la direction de) (1995): Les QueÂbeÂcoises
et les QueÂbeÂcois mangent-ils mieux? Rapport de l'EnqueÃte QueÂbeÂcoise
sur la Nutrition 1990. MontreÂal: MinisteÁre de la Sante et des services
sociaux, gouvernement du QueÂbec.
Shea S, Melnik TA, Stein AD, Zansky SM, Maylahn C & Basch CE
(1993): Age, sex, educational attainment, and race=ethnicity in relation
to consumption of speci®c foods contributing to the atherogenic
potential of diet. Prev. Med. 22, 203 ± 218.
Shi L (1998): Sociodemographic characteristics and individual health
behaviors. S. Med. J. 91, 933 ± 941.
Smith AM & Baghurst KI (1992): Public health implications of dietary
differences between social status and occupational category groups. J.
Epidemiol. Community Health 46, 409 ± 416.
Variyam JN, Blaylock J, Smallwood D & Basiotis PP (1998): USDA's
Healthy Eating Index and Nutrition Information. Economic Research
Service=USDA-TB 1866.
Willet W (1990): Nutritional Epidemiology. New York: Oxford University Press.
Willms DJ & Shields M (1996): A Measure of Socioeconomic Status for
the National Longitudinal Survey of Children. Atlantic Center for
Policy Research in Education, Universite du Nouveau-Brunswick et
Statistique Canada.
Whitehead M (1995): Tackling inequalities: a review of policy initiatives.
In Tackling Inequalities in Health. An Agenda for Action, ed. M
Benzeval, K Judge & M Whitehead, pp 22 ± 51. London: King's Fund.
Wilkinson RG (1996): Unhealthy Societies. The Af¯ictions of Inequalities.
New York: Routledge.
Wolfson MC & Murphy BB (1998): New views on inequality trends
in Canada and the United States. Monthly Labor Rev. April, 3 ± 23.
Woo J, Leung SSF, Ho SC, Sham A, Lam TH & Janus ED (1999):
In¯uence of educational level and marital status on dietary intake,
obesity, and other cardiovascular risk factors in a Hong Kong Chinese
population. Eur. J. Clin. Nutr. 53, 461 ± 467.
373
European Journal of Clinical Nutrition