There is a large published literature on associations between socio-economic position and chronic disease, with socioeconomically disadvantaged groups experiencing higher mortality and morbidity rates for coronary heart disease, noninsulin dependent diabetes mellitus and some cancers 1234. Chronic diseases are largely preventable diseases, and diet has been known fore many years to play a key role as a risk factor for chronic diseases. While age, sex and genetic susceptibility are non-modifiable, many of the risks associated with age and sex are modifiable. Such risks include a complex mixture of interacting socio-economic, cultural and other environmental factors 5678.

The relationship between socio-economic factors and diet has been examined on the basis of food and nutrient intake and the results are mixed, sometimes contradictory, and often the observed differences are small 910. When differences are found, it is usually the case that persons from socioeconomically disadvantage backgrounds have food intakes consistent with their higher rates of chronic diseases 11121314151617.

In the context of European countries, economic development and increased purchasing power have recently changed the food availability situation 18. On the other hand, socio-cultural influences may contribute, along with economic constraints, to particular food choices, which may explain the still substantial differences in food consumption across European countries 19. Understanding the influences of socio-economic variables on food consumption may be useful to predict the outcome of interventions, to change food behaviour, and generate hypotheses concerning food consumption in diverse circumstances, as well as to explain observations in epidemiological studies.

The aim of this study was to evaluate the importance of educational and economic factors in determining food choice in a representative sample of the Portuguese general adult population.

The study sample comprised 20977 women (52,9%) and 18663 men, with mean ages of 50.3 (±18.88) and 47.7 (±18.51) years, respectively; BMI was significantly lower in women than in men (25.1 ± 4.53 Kg/m² versus 25.6 ± 3.83, p < 0.001). There was a lower proportion of smokers among women compared to men (8.2% versus 30.5%, p < 0.001). General characteristics (gender, age, BMI, smoking status, and physical activity) by education and income categories are presented in Tables 1 and 2.

In women, the odds favouring milk, vegetable soup, vegetables, fruit, and fish consumption, increased with increasing education (p-values for trends were always ≤0.046), being the odds ratios, respectively, 2.60 (2.24–3.01), 1.20 (1.05–1.38), 1.75 (1.44–2.13), 1.92 (1.49–2.49), and 1.40 (1.23–1.60) for those having >12 years of education compared to those with ≤4 years, after adjusting for age, BMI, smoking habits, physical activity and income (Tables 3 and 4). The odds favouring bread, starchy foods (other than bread), wine, and spirits consumption in women decreased with increasing education (p trend ≤ 0.002), being the odds ratios, respectively, 0.44 (0.34–0.56), 0.68 (0.53–0.87), 0.51 (0.41–0.62), and 0.13 (0.03–0.53) for those having >12 years of education compared to those with ≤4 years (Tables 3 and 4).

In men, similar odds ratios were observed for milk, vegetable soup, vegetables, fruit, fish, bread, wine, and spirits (Tables 3 and 4). However, in men but not in women, odds favouring meat consumption increased with increasing education (OR = 1.16 (0.96–1.41) for those having >12 years of education compared to those with ≤4 years), while for beer consumption, odds decreased with increasing education (OR = 0.57 (0.48–0.68) for those having >12 years of education compared to those with ≤4 years).

No such significant trends were observed for these food groups and income with the exceptions of meat and starchy foods (other than bread) consumption, in men, which decreased with increasing income (p trend ≤ 0.022), and milk consumption which increased with increasing income (Tables 3 and 4).

The main finding of the present study is that educational attainment was more frequently associated with food choices than income. There is general agreement among researchers 20212223 that education and income are conceptually distinct, and that they are likely to make separate and unique contributions to health-related outcomes 24. In our study, the most educated consumed more frequently fruit, vegetables, milk and fish, and less wine and spirits, than their counterparts from less educated groups.

Over the last years, several studies have attempted to identify the influence of socioeconomic factors on individual's dietary intake 25262728. Our interest in educational and economic determinants of food choice in Portuguese adults relate to these particular characteristics in the population. Portugal, according European standards, is a small and relatively poor country, exhibiting the highest level of social inequalities in the European Union 29. Nevertheless, Portugal had significantly and positively changed in the last four decades, in several domains such as the economy and culture, although the census of 1991 revealed that 15.3% of the Portuguese were illiterate. That of 2000 showed that, despite the improvements and changes in the education of adults, 7% can still not read or write 29. This is a reality that classifies Portugal as the country with the higher percentage of individuals with low level of education in all the European Union 29. From the employment perspective, Portugal's unemployment rates in the last 25 years never surpassed 10% of the active population, which is a better indicator than the observed levels in the majority of the European countries. However, the percentage of individuals with low-remuneration in Portugal is much higher than the EU average 29.

Several studies have concluded that a strong relationship exists between countries' per capita national incomes and nutrition 303132. The economic issue is of considerable significance, and it is sometimes suggested that this is probably the key variable of all in influencing food choice 30. Household income is expected to influence food choices, especially for relatively high-priced food items such as fish, fresh fruit and vegetables 33. Nevertheless, this not seems to be the case when we compared income and education levels as determinants of intake of significant food groups in Portuguese adults. Our data shows, in both genders, a significant positive trend in the consumption of vegetables, vegetable soup, fruits, milk and fish, with higher levels of education, which did not occurred in relation to income with the exception of milk. In our study, education was adjusted for income and vice versa. While the majority of investigators use two or more indicators of socio-economic position, several 34353637 do not simultaneously adjust for the unmeasured effects of each indicator on the other. Two types of bias may result from this practice: (1) using a single indicator such as education may bias the point estimate (food choice) because the education variable is allowed to account for some of the variation that is actually the product of unmeasured socio-economic influences; as a result, if we did not simultaneously adjusted education for income and vice versa, our claims about the influence of education level on food choice probably would have been overestimated; (2) the use of a single indicator may result in the overall or total socio-economic effect being underestimated.

Data from the Portuguese Household Budget Surveys (using the DAta Food Networking – DAFNE – classification system), shows similar results to ours in relation to the positive association between education attainment and the availability of fruits, fish, milk and alcoholic beverages but some different data in regard to other foods (availability of vegetables and cereal products is fairly stable or tends to decline with education) 18. Curiously, we found that meat consumption in men was positively associated with level of education, as in the DAFNE study 18, although our study showed a significant reduction of meat consumption with higher categories of income, in men. In Portugal 18 fish is more available among the trend-leading educated individuals which may be more advantageous to their cardiovascular health 38.

In our study, there was also a significant trend in the consumption of milk in men, being more frequently ingested with increasing income. As suggested by Axelson, 39 positive health relationships between dietary patterns and income may reflect a growing concern about health in the higher socio-economic groups.

The association between milk consumption and socio-economic position is sometimes contradictory. Cristofar and Basiotis 40, for example, reported lower intake of milk among low-income women, while Roos et al. 41 found that higher educational and income groups from both genders consumed less milk.

Consumption of alcoholic beverages, such as wine and spirits, in both genders, and beer in men, exhibited significant decreases in their frequency of intake, with increasing education levels. In Portugal, alcoholic beverages consumption is a major public health problem 42. In DAFNE study, 18 using Portuguese Household Budget Surveys, alcoholic beverages availability was also higher in the lower educated households.

Interestingly, in our study, the consumption of bread (in both genders) and other starchy foods (in women), decreased with increasing number of years of education; men seem to abandon starchy foods (other than bread) consumption under condition of higher incomes. It is possible that higher educated individuals tend to avoid foods that are considered as being more fattening or rich in energy, such as bread and other starchy foods 4344. Research has demonstrated that for a given body size, higher educated women are more dissatisfied with or concerned about their bodies and are more likely to have dieted in the past than lower educated women 4546.

One of the most interesting findings in terms of economic constraints and food consumption relationship in our study, is the few significant associations between income and food choices, even though the well established links between economic and material resources, food availability and dietary quality 47. By contrast to our results, Turrell et al. 23 showed household income to be the strongest and most robust independent predictor of food purchasing behaviour, and the effects of education to be substantially attenuated (to non-significance or marginal significance). In our study, the specificity of the relationship between education and food choice probably reflect each respondent's individual contribution to food choice, whereas household income was possibly capturing the combined contextual effects of numerous individuals, as well as many other within-household processes 23, and thus showed a weaker relationship with food choice. Our results may also reflect lesser difficulties faced by low-income groups when selecting the food groups that we studied. In several urban and rural areas of Portugal, there are many people who own plots of land that are too small to make a living, but allow them to work the land and produce foods (e.g., fruit, vegetables and poultry) for their own consumption. Although they produced a limited range of foods that is not accounted in official agricultural statistics, probably, if they stopped working the land they would experience greater difficulties in obtaining access to those particular foods. A potential limitation of our study and most nationwide population surveys is that the poor are usually not well presented. We know from previous research into survey participation that population-based samples typically under-represent the most socio-economically disadvantaged and over-represent the advantaged 4849, because homeless and unemployed may be difficult to reach, and this may debilitates the interpretation of our results.

In our study, it remains to be explained the different pattern of associations between income and important food groups (milk versus fruit and vegetables, for example) and the different pattern of associations between food choices and income in each gender (e.g., milk and starchy foods). Several reasons may explain specific differences in the findings of our study compared with those of previous mentioned studies, including differences in populations sampled (e.g., both genders versus women only, different cultural backgrounds ranges), differences in assessment of education or income, differences in dietary assessment (e.g., qualitative food data versus 24-h dietary recalls or food frequency questionnaires) and differences in analytic methods (e.g., covariates included in statistical models). Nonetheless, results from our study indicate that the associations between food choice were stronger in relation to educational attainment than income categories. Differences in food choices according the level of education reflect that more knowledge may influence the perceived relationship between diet and health as well as the perceived outcomes of following a healthy diet 33.

Despite differences in food consumption according education and income, in our study we could not assess if these differences were also evident on the energy and nutrient level, which was a limitation. British data point to micronutrient and antioxidant intakes as the most likely nutritional influences on health inequalities 50. Nevertheless, according Galobardes et al., 10 it is also possible that despite differences in food consumption, nutrient intake is similar among socio-economic groups, as these may not be substantial enough to translate into differences in nutrient intake.

If a country like Portugal wants to change the adult food choice behaviour, or in other words, wants to reach certain dietary goals, the support of applied research like ours is needed in order to plan the right strategies for promoting healthy diets. Confidence in a significant positive causal link between per capita national income and individual nutrition reinforces the importance of economic growth 51 but also implies that public policy should stress education as a mean for improving healthy food choices. Education might influence food choice by facilitating or constraining one's ability to understand the information communicated in nutrition education or in food labels 5253.

Whereas income-related dietary differences suggest ameliorative responses through the potential of the economic system, differences based on education point to initiatives such as nutrition education programmes 2254. According to Geraldes,55in Portugal it may sometimes be more appropriate to correct inequalities in the domains of education or nutrition than that of health. Given the poor education level of the majority of Portuguese adults, a move towards an increased acquisition of general knowledge and personal development through compulsory and higher education, lifelong learning and improved qualifications of the population, is desirable to promote the development of a knowledge society and improve the level and quality of national education which, in turn, may relate to healthier food choices. It is well recognized that changes in dietary behaviour may be brought about, not by direct modification of food habits, but by alteration or manipulation of the education and culture 8.