J Public Health DOI 10.1007/s10389-017-0818-z

ORIGINAL ARTICLE

Socioeconomic differences in diet composition of the adult population in southern Brazil: a population-based study Katia Jakovljevic Pudla Wagner 1 & Silvia Ozcariz 1 & Francieli Cembranel 1 & Antonio Fernando Boing 1 & Albert Navarro 2 & David Alejandro González-Chica 3

Received: 26 January 2017 / Accepted: 14 June 2017 # Springer-Verlag GmbH Germany 2017

Abstract Aim To describe the intake of macro- and micronutrients, verify its adequacy, and analyze their distribution by the socioeconomic and demographic profile of adults. Subject and methods Longitudinal population-based study with a sample of 1222 people aged 22–63 years from Florianópolis, Southern Brazil. The dietary intake data were collected with a 24-h recall in the total sample, plus a second recall applied on a sub-sample of participants with subsequent adjustment of regular consumption. Diet composition and adequacy of intake were compared between the different educational groups and stratified by sex. Results The mean energy consumption was 1851 Kcal [standard error (SE) = 15.0 Kcal] and 2259 Kcal (SE = 27.8 Kcal) in females and males, respectively (p-value <0.001). With increasing education, women showed an increasing trend in energy consumption, carbohydrates, proteins, lipids, saturated, polyunsaturated and monounsaturated fat, cholesterol, and fiber. In men, the same differences were found only in saturated fat and cholesterol (p-value <0.05 for all). Analyzing the adequate consumption of nutrients, differences were found Electronic supplementary material The online version of this article (doi:10.1007/s10389-017-0818-z) contains supplementary material, which is available to authorized users. * Katia Jakovljevic Pudla Wagner [email protected]

1

Department of Public Health, Federal University of Santa Catarina, Delfino Conti street, no number, Florianópolis, SC 88040-900, Brazil

2

Unitat of Biostatistics, Faculty of Medicine, Universitat Autónoma de Barcelona, 08193 Barcelona, Spain

3

School of Population Health, The University of Adelaide, Adelaide, Australia

regarding saturated fat and cholesterol in women and carbohydrates, saturated and polyunsaturated fat in men. The prevalence of inadequacy was up to 47% comparing the different educational groups. Conclusion A lower percentage of adequacies in food consumption was found in both sexes for those with more years of schooling. Public policy should also target the group with higher education. Keywords Dietary intake . Dietary fats . Dietary proteins . Dietary carbohydrates . Educational level . Income

Introduction Dietary intake has shown significant changes in the last 50 years. The increments in total food supply, falling prices and rising income levels are some of the factors contributing to the changes in both the quality and quantity of food consumption (Kearney 2010). The daily energy consumption has risen in this period, particularly the consumption derived from commodities. Studies show that in developing countries, between 1963 and 2003, large increases in the available calories were derived from animal products, sugar and vegetable oils, all of which are widely used by food industries in food processing (Kearney 2010; Lallukka et al. 2007). In Brazil, there is a lack of population-based food intake surveys, but national Household Budget Surveys (POF) comparing food purchases between 2002/03 and 2008/9 suggest significant changes in food acquisition patterns (Brasil-Ministério do Planejamento, Orçamento e Gestão 2010). In this country, the largest increase was in industrialized and ready-to-eat products (37% in this period from 2.6 to 3.5 kg per capita), while acquisition of typical Brazilian food such as rice and beans

J Public Health

declined over the same period (40.5% and 26.4%, respectively) (Brasil-Ministério do Planejamento, Orçamento e Gestão 2010). Food choices are also influenced by sociodemographic characteristics, and variables such as income, educational level and gender have demonstrated an association with diet composition (Darmon and Drewnowski 2008; Lallukka et al. 2007; Teixeira et al. 2016). Socioeconomically disadvantaged groups have traditionally made poorer food choices and also lack access to nutritious food, as these products are more expensive (Darmon and Drewnowski 2008). Moreover, people with a lower educational level have been shown to have less nutritional knowledge than the group with a high educational level, which contributes to the higher energy density and poor nutritional quality of their diet. Scientific literature also indicates differences in relation to gender as women have lower energy diets compared to men, attach more importance to healthy food habits and are more likely to avoid certain foods (Westenhoefer 2005). This difference between genders also seems to be associated with the motivational level for following a diet with better nutritional content, which seems to be higher in females than in males (Leblanc et al. 2015). Considering that the nutritional composition of the diet has a direct influence on cardiovascular and chronic diseases, including obesity and diabetes (Mente et al. 2009), the analysis of the diet composition of populations is extremely important for determining health policies. Furthermore, the prevalence of obesity in Brazil increased four fold in males and two fold in females from 1974/5 to 2008/9 (Brasil-Ministério do Planejamento, Orçamento e Gestão 2010). This increase in the prevalence of obesity affected both sexes differently according to socioeconomic status, and currently the obesity rate is higher among women of lower income and lower education level and in men with higher incomes (Brasil-Ministério do Planejamento, Orçamento e Gestão 2010; Sousa et al. 2011). Some studies conducted in Brazil indicate inadequacies in diet intake with a different distribution of nutrients depending on the population group (Costa et al. 2013; Molina et al. 2007; Morimoto et al. 2012; Sousa et al. 2011). However, there are few population-based studies in Brazil and in other low- and middle-income countries that have evaluated the diet composition of adults using 24-h recall (24HR) and have made the adjustment calculation for usual consumption, with correction for the within-person or day-to-day variability. The studies that have performed an adjustment of usual consumption differ in the age of participants, type of nutrient analyzed and data stratification regarding socioeconomic and demographic characteristics (Morimoto et al. 2012; Moshfegh et al. 2005; Tarasuk et al. 2010. Tarasuk et al. (2010) analyzed the consumption of different educational, income and age groups and found that individuals in disadvantaged groups (lower income and education level) had more inadequacies in consumption, but this differs from current research by analyzed nutrients. Moshfegh et al. (2005) only examined

differences related to gender and age, and the results of macronutrient and fiber analysis indicate that higher inadequacies were seen among women. In a study conducted in São Paulo, Morimoto et al. (2012) were the first to use an adjustment for within-person variation to evaluate food intake in a population-based sample in Brazil. Nevertheless, they did not evaluate inadequacies according to sociodemographic variables. Thus, there is a lack of population-based studies investigating food consumption and inadequacies in macro- and micronutrients in low- and middle-income countries such as Brazil. The main objective of this study was to describe the intake of energy, macronutrients, fibers, monounsaturated, saturated and polyunsaturated fats, and cholesterol and trans fats, as well as to verify its adequacy and to analyze their distribution by the socioeconomic and demographic profiles of males and females who participated in a population-based cohort study in Florianópolis, Southern Brazil.

Methods The investigated sample consisted of individuals aged 22– 63 years living in the city of Florianópolis, Southern Brazil, who participated in the second wave of a longitudinal population-based study (EpiFloripa Adult Study) for whom detailed information on food consumption was collected. This cohort aimed to examine living conditions, general and oral health, nutritional status, food consumption and lifestyle in a population-based sample of adults in the city. Florianópolis is the capital of the state of Santa Catarina (Southern Brazil) and has the third highest Human Development Index of all Brazilian municipalities (0.847) (PNUD 2013). The baseline survey was conducted in 2009 (first wave), and in 2012 the first follow-up (second wave) was conducted (Boing et al. 2014). In the first wave, the study population included adults residing in the city aged 20–59 years, with a total of 1720 individuals interviewed. Details about the sampling, study population and methodological aspects of the research have been published previously (Boing et al. 2014). In 2012, the same adults were traced and interviewed at home (Supplementary Graph 1). Losses were considered as the adults who refused to participate or when interviewers were unable to locate them after at least four attempts to contact them by phone and four attempts to visit them at home. Data collection Data collection was carried out by trained interviewers in the two waves by way of individual face-to-face interviews conducted at the homes of the participants. The food consumption data were only collected in 2012 using 24-h recall (24HR) in the total sample, plus a second

J Public Health

24HR applied by telephone to a sub-sample of 40% of these participants to obtain an estimate of usual consumption adjusted by intra- and interindividual variability (Dodd et al. 2006). The selection of participants for the application of the second 24HR was made with a systematic sampling of those who answered the first 24HR (one in every three participants). This selection procedure was carried out at the end of every week of data collection. The second 24HR was applied over a period of 15–30 days after the first survey, always alternating between weekdays and weekends. The method adopted for the implementation of 24HR was the Bmultiple pass method,^ collected by trained interviewers, and the information was recorded using the Nutrition Data System for Research (NDSR) software from the University of Minnesota, which provides information for more than 150 nutrients. Educational level was recorded in 2009, while the other socioeconomic (per capita family income) and demographic variables (gender, age, marital status and skin color) used in the analysis were obtained in 2012 using a pre-tested questionnaire. All the variables were self-reported by the participants. Educational level was calculated based on the completed years of study and per capita family income by dividing the total family income by the number of members. Age was calculated as the difference between the interview data and the date of birth. Skin color was collected as a nominal variable (white, brown, black, Asiatic, indigenous) based on the classification system of the Brazilian Institute of Geography and Statistics (Brasil-Instituto Brasileiro de Geografia e Estatística 2015). Anthropometric measurements (weight and height) were collected by the interviewers according to standardized procedures and categorized according to the literature. Body mass index (BMI) was categorized as obese (BMI ≥30 kg/m2), overweight (BMI of 25.0 to 29.9 kg/m2), eutrophic (BMI of 18.5 to 24.9 kg/m2) and low weight (BMI <18.5 kg/m2) (WHO 1995). This variable was used to compare the participants in the two waves. In both waves, the project was approved by the Ethics Committee on Human Research of the Federal University of Santa Catarina (351/08 and 1772/11), and all respondents signed an informed consent form. Data analysis The NDSR software database uses the food composition table of the US Department of Agriculture (USDA), but the data entry followed standardized procedures included in a manual created by researchers from the Health Survey (Inquérito de Saúde em São Paulo-ISA) in São Paulo, Brazil. Although the USA table is the basis of the NDSR software, this software was chosen not only to allow comparability with other Brazilian studies, but also because of the quantity of information provided and the possibility to incorporate food products and food preparations.

The nutritional composition of the Brazilian nutritional table (NEPA 2015) was also taken into account to select the best food options in the NDSR program for all food products and/or food preparations by comparing the composition of the foods that were in the software with those existing on the Brazilian table. The serving sizes were also carefully converted into grams, given that portion sizes of industrialized products and some kitchen utensils could differ. These measurements were taken using standardized manuals of serving measures and kitchen utensils (Pinheiro 2004). For food preparations not available in the NDSR program, standardized recipes using Brazilian manuals were added (Institute of Medicine 2005). The energy consumption of the individuals included alcohol consumption, as this contributed to a lower percentage of total calories in the sample [1.5% of the total energy intake (TEI)]. After inputting all the data into the NDSR software, specific nutritional intake information was obtained from each participant. To estimate the usual intake of the participants, information on both 24HRs was used following the steps set out by the National Research Council and the Institute of Medicine (Dodd et al. 2006). Data on TEI, macronutrients (carbohydrates, protein and total lipids), and monounsaturated, polyunsaturated and trans fat and cholesterol intake were symmetrically distributed, and no transformation was necessary. Only fiber and saturated fat intake showed an asymmetrical distribution, and thus a log transformation was performed prior to the next steps. Subsequently, to reduce the individual daily variation in nutrient intake, and to correct the information for intra- and interindividual variability, an adjustment was made using the following formula: adjusted intake = [(participant mean group mean) × (standard deviation interpersonal/standard deviation observed)] + mean group (Dodd et al. 2006). Information on fiber and saturated fat intake was backtransformed to natural numbers after this adjustment. To assess the adequacy of macronutrients, primarily the percentage contribution of TEI was calculated. The percentages of adequacy of macronutrients followed the guidelines of the Acceptable Macronutrient Distribution Ranges (AMDR) (Institute of Medicine 2005), while for the other nutrients analyzed the reference of adequacy used was the World Health Organization (WHO) (2003). To calculate adequacy, a specific cutoff was used, which for the majority of nutrients takes into account the percentage contribution of TEI. In the case of carbohydrates, for example, adequate consumption should be in the range of 45–65% of TEI. All the other contribution percentages are presented in Table 4 with their respective references. For cholesterol and fiber intake, the adequacy values were defined as <300 mg and >25 g/day, respectively. For analysis, mean and standard errors or median and interquartile range (IQR) was used for numerical variables based on the normality of the distribution. Relative frequencies were used to describe categorical variables.

J Public Health

Analysis of TEI, carbohydrates, lipids and fractions (monounsaturated, saturated, polyunsaturated, cholesterol and trans fats) and fiber was performed according to educational level and family income, and stratified by gender. The results for these outcomes were expressed as means with their respective standard errors. Adjustments for possible sociodemographic confounders (according to the literature: age, skin color and marital status) (Araujo et al. 2014) were performed using multiple linear regression. Instead of presenting adjusted regression coefficients, adjusted means were estimated to make the description of the association more tangible and comparable with the crude means. Family income and educational level were not mutually adjusted because of collinearity. Tests for the association between gender and socioeconomic variables were also conducted. The prevalence of food intake adequacy (yes/no) was also evaluated according to socioeconomic variables (stratified by sex) using Poisson regression, and the results were expressed as prevalence ratios. Similar adjustments for confounders and interaction tests used in linear regression were made in this case. Data analysis was carried out using STATA version 13.0 software (Stata Corp., College Station, TX, USA), and the value of statistical significance was set at p ≤0.05. All analyses were carried out using probability weights, taking into account the effect of sample design (2009) and the probability of location in the second wave (2012) according to the census tract (Boing et al. 2014).

Results The sample of the 2012 wave of the EpiFloripa study included 1222 adults (71% of the sample in 2009), and the response rate for the first and second 24HRs was 93.4% and 98.9%, respectively. The median time of education in the sample was 11 years (IQR = 6 years), and the median per capita income was R$1125 (IQR = R$1393) per month. A comparison between the characteristics of the participants in the two waves is presented in Table 1. The average daily energy consumption was 1851 Kcal (SE = 15.0 Kcal) and 2259 Kcal (SE = 27.8 Kcal) for females and males, respectively (p-value <0.001). Among women, the intake of all macronutrients was lower than among men (Table 2). Table 2 also shows the association between the educational level of the participants and TEI, macronutrients and fiber intake, stratified by sex. There was a direct association between educational level and TEI for both sexes, but only among women was the trend statistically significant in crude and adjusted analysis (p-value <0.001 in both cases). The same direct trend with the educational level was observed for carbohydrates and protein intake in women, among whom there was a difference in the average consumption of

28.2 g (95% CI 13.0; 43.3) and 6.8 g (95% CI 1.5; 12.0), respectively, between the extremes of educational levels. In men, there was no association between these variables. The high consumption among those with a high educational level was observed again in relation to lipid consumption by both sexes, but in this case the differences between the extreme categories of the educational level were higher in women (11.4 g; 95% CI 7.0; 15.9) than in men (7.03 g 95% CI -2.5; 17.3), and once again it was statistically significant only among the former. For fiber, the pattern was similar among women, with an intake 4.4 g higher among those with a high educational level (95% CI 2.4; 5.7), but among men there was no association. It is noteworthy that the tests related to the association between gender and education level indicated that there was only a significant modification effect for fiber (pvalue = 0.006). Table 3 shows the analysis of each type of lipid separately according to educational level. The same direct trend found with respect to total lipids was also observed for saturated fats and cholesterol in women and in men in both crude and adjusted analyses. The differences between the extremes of educational level showed similar mean differences in both sexes for saturated fat (about 5 g), but for cholesterol the difference among men was almost twice as high as among women (mean difference 25 g in women and 38 g in men). Consumption of unsaturated fats (mono- or polyunsaturated) was also lower among women with a low educational level, but no association was found for males. Trans fat consumption was not associated with gender or educational level for either sex. For per capita family income (Supplementary Tables 1 and 2), the pattern of associations was very similar to the results found for educational level. In analyzing the adequate consumption of macronutrients (Fig. 1), over 95% of respondents had adequate protein intake. The inadequacy of carbohydrate consumption has been related more to values lower than those recommended (21.8%), while for lipids an excess consumption was found (16.0%). The prevalence of adequacy for the three macronutrients was similar in both sexes (p-value > 0.05 in all cases). Due to a low percentage of individuals having values in a category of inadequacy for all macronutrients, adequacy was categorized as dichotomous (yes/no) for the analyses presented in Table 4. The prevalence of adequacy for the other nutrients ranged from 12.8% (fibers) to 79.5% (polyunsaturated fats, inadequacy group with consumption below the recommendation). The remaining adequacy percentages were 13.8% for monounsaturated fats (with inadequacy group below the recommendation), 23.3% for trans fats, 38.8% for saturated fats and 72.6% for cholesterol (data not shown in tables). Table 4 shows the prevalence ratios for adequate nutrient intake in the category B12 or more^ years of education considered as the reference group. Among females, the adjusted adequate intake of saturated fat

J Public Health Table 1 Comparison of the participants’ characteristics in two waves of the EpiFloripa study cohort, Florianópolis, Brazil

Variable

Original sample (2009; N = 1720)%

Second wave (2012; N = 1222)%

Sex Male Female

44.2 55.8

42.7 a 57.3

Age 20–29 years 30–39 years

31.4 22.8

26.8 a 22.7

25.5 20.3

28.2 22.3

40–49 years 50 or more years Skin color White

89.6

89.7

10.4

10.3

Married/living with partner Single/divorced/widowed Educational level 0–4 years

60.6 39.4

63.7 a 36.2

9.2

8.9

5–8 years 9–11 years 12 or more years Per capita income (median and ICQ) BMI (body mass index) Low weight

14.7 33.1 43.0 R$866.7(1166.7) [U$473.6]

14.3 32.3 44.5 R$1125.0 (1383.4) a [$554.2]

Black and others Marital status

Eutrofic Overweight Obesity a

2.0

2.1

50.1 31.7 16.1

48.9 31.7 17.4

p < 0.05 of the comparison between located and non-located

and cholesterol was higher for those with a lower educational level. In males, there was an inverse relationship in the prevalence of adequacy with educational level for carbohydrates and polyunsaturated and saturated fat, with an adjusted prevalence of adequacy between 31% and 124% higher in those with a lower educational level compared to the reference category. None of the other nutrients analyzed had the adequacy associated with educational level, not in men or women. The same analysis of adequacy was conducted according to family per capita income, but in this case only cholesterol adequacy was inversely associated in women, while in men the adequacy of saturated fat consumption was higher among those with the lowest income (Supplementary Table 3).

Discussion The results of this study indicate that, despite the direct association between socioeconomic variables and the mean consumption of macronutrients, lipid fractions and fibers (with a dose-response effect of consumption according to educational

level and family income), the prevalence of adequacy for saturated fat and cholesterol (above the cutoff) was consistently lower among males and females with a higher educational level. In men, the higher educational level groups also showed a lower prevalence of adequacy for carbohydrates and polyunsaturated fat (both below the recommendation). Another paper using data of the same cohort showed that the prevalence of abdominal obesity is 37% lower among those with the highest income when compared to the less educated women, while in males this association was not observed (Sousa et al. 2011). Thus, particularly among females, the direct association observed between socioeconomic variables and saturated fat and cholesterol consumption was contrary to our original hypothesis of higher risk among the poorest. Nevertheless, our results also show that consumption of protective nutrients (fiber, monounsaturated and polyunsaturated fat) was also higher among the more educated women. Furthermore, in this cohort, the prevalence of physical inactivity is 53% lower than among the poorest (Silva et al. 2013), suggesting a complex relationship between risk and protective habits in the development of obesity.

J Public Health Table 2 Description of the crude and adjusted mean values of intake for energy, carbohydrates, proteins, lipids and fibers according to sex and educational level among adults aged 22 to 63 years, Florianópolis, Brazil, 2012 Energy (Kcal) N Sex and educational level Female

Crude mean (SE)

Carbohydrates (g)

Proteins (g)

Adjusted a mean (SE)

Crude mean Adjusted a (SE) mean (SE)

Crude Adjusted Crude mean mean (SE) a mean (SE) (SE)

699 1851.0 (15.0)

231.1 (2.0)

Lipids (g)

79.8 (0.7)

Fiber (g) Adjusted Crude a mean mean (SE) (SE)

67.2 (0.7)

Adjusted mean (SE)

a

16.5 (0.3)

0–4 years

65 1616.6 (44.8)

1642.9 (48.1)

206.2 (6.4)

206.5 (6.5)

74.3 (2.3)

75.3 (2.5) 57.0 (1.5) 58.4 (1.7) 14.4 (0.6) 13.4 (0.4)

5–8 years 9–11 years

107 1820.3 (52.0) 215 1843.0 (26.9)

1811.2 (50.2) 1830.5 (27.6)

229.3 (6.6) 233.7 (3.7)

228.4 (6.6) 231.8 (3.6)

78.7 (2.1) 78.6 (1.2)

78.3 (1.8) 66.0 (1.9) 65.4 (1.9) 15.8 (0.8) 15.5 (0.8) 78.3 (1.2) 66.5 (1.2) 66.2 (1.2) 16.4 (0.5) 16.5 (0.5)

≥12 years P value Males

312 1911.5 (25.5) <0.001 b 520 2259.0 (27.8)

1910.1 (24.6) <0.001 b

234.5 (3.4) 234.7 (3.5) 0.010 b 0.007 b 273.2 (73.6)

82.1 (0.9) 0.007 98.1 (1.4)

82.1 (0.8) 70.0 (1.2) 69.9 (1.2) 17.3 (0.4) 17.5 (0.5) 0.003 <0.001 b <0.001 b 0.001 b <0.001 b 77.9 (1.1) 18.6 (0.4)

43 2143.3 (133.5) 2180.6 (115.5) 279.2 (19.9) 280.7 (14.9) 89.6 (3.8) 67 2208.8 (90.7) 2250.8 (93.6) 278.3 (11.4) 282.4 (11.5) 97.4 (5.2)

91.0 (3.6) 69.1 (4.4) 71.6 (4.3) 20.7 (1.5) 20.0 (1.4) 99.4 (5.6) 73.7 (3.1) 75.4 (3.3) 19.4 (1.0) 19.4 (0.9)

0–4 years 5–8 years 9–11 years ≥12 years

179 2269.7 (39.7) 231 2282.0 (48.8)

P value

0.273 b

2257.1 (41.1) 2266.6 (43.4)

273.9 (4.5) 270.3 (5.7)

272.5 (4.6) 268.9 (5.5)

97.6 (2.0) 97.2 (2.0) 79.0 (1.7) 78.6 (1.7) 17.9 (0.6) 17.9 (0.6) 100.1 (2.3) 99.4 (2.2) 79.7 (1.9) 78.9 (1.7) 18.7 (0.4) 18.6 (0.5)

0.545 b

0.912 b

0.669

0.106 b

a

Adjusted for age, marital status and skin color

b

Linear trend test

A longitudinal study performed in Spain between 1992 and 2003 was unable to find any association between change in energy intake and the increase in the prevalence of obesity in the sample. The authors suggested that food consumption would be associated with an increase in overweight only when combined with a change in the pattern of physical activity levels (Serra-Majem et al. 2007). Differences in food consumption associated with sociodemographic variables can be explained by several factors. First, foods with low energy content and higher amounts of vitamins and minerals may be more expensive, and people with lower socioeconomic status have less access to them (Darmon and Drewnowski 2008; Ferreira et al. 2010). Second, population groups living in the wealthiest areas have more access to supermarkets and other places that sell healthier foods and a greater knowledge of nutrition, and more accurate body weight perception may also contribute to better food choices among those with a higher educational level (Darmon and Drewnowski 2008). This explanation is consistent with our findings in females of a direct association between socioeconomic status and the consumption of fiber and unsaturated fat, but not with the results regarding saturated fat and cholesterol. Nevertheless, according to the Brazilian Household Budget Surveys, which have analyzed the acquisition of food products by Brazilian families since the 1970s, there have been important changes in food acquisition in the last 30 years according to socioeconomic status. Those with higher incomes are currently purchasing more pre-prepared and industrialized foods (food products with large amounts

0.258

0.027 b

0.143 b

0.445

0.486

of saturated and trans fat, salt and sugar) than those with lower incomes, but at the same time they are buying more fruits and vegetables, which contributes to higher quantities of fiber, vitamins and minerals (Brasil-Ministério do Planejamento, Orçamento e Gestão 2010), corresponding to the results of this research. Contrary to our results on the association between socioeconomic conditions and total and saturated fat intake, studies with adults from high-income countries such as Australia (Mishra et al. 2010) and Belgium (Mullie et al. 2010) have revealed a higher prevalence of high-fat diets among lower educational level and income groups. However, a population-based study performed in Ireland (McCartney et al. 2013) like our study also found higher consumption of proteins and lipids among the richest, while carbohydrate intake was higher among the poorest. Although our results showed a direct association between socioeconomic status and carbohydrate intake among females, in males the prevalence of adequacy was higher among the less educated. A study from Japan (Fukuda and Hiyoshi 2012) also found higher protein consumption in the higher household expenditure group, but there was no association for fat and carbohydrate intake. According to Rombaldi et al. (2014), this contrast in dietary intake according to socioeconomic status can be explained by the nutritional transition process. In some high-income countries with a more advanced level of nutritional transition, the highest prevalence of high-fat diets is found among the poorest. However, in low- and middle-income economies at

J Public Health Table 3 Description of the crude and adjusted mean values of intake for monounsaturated, saturated, polyunsaturated, cholesterol and trans fats according to sex and educational level among adults aged 22 to 63 years, Florianópolis, Brazil, 2012 Monounsaturated (g) Sex and educational level

N

Crude mean (SE)

Females

699 23.2 (0.2)

Adjusted a mean (SE)

Saturated (g)

Polyunsaturated (g)

Adjusted a mean (SE)

Crude mean (SE)

Crude mean (SE)

20.1 (0.2)

15.4 (0.2)

Adjusted a mean (SE)

0–4 years

65 19.8 (0.5) 20.3 (0.6)

16.2 (0.7) 16.6 (0.8)

13.0 (0.4) 13.4 (0.4)

5–8 years

107 22.6 (0.7) 22.3 (0.7)

19.2 (0.7) 19.0 (0.7)

15.6 (0.5) 15.5 (0.5)

9–11 years

215 23.0 (0.4) 22.9 (0.4)

19.7 (0.4) 19.6 (0.4)

15.5 (0.4) 15.4 (0.4)

≥12 years

312 24.2 (0.5) 24.2 (0.5)

21.4 (0.4) 21.4(0.3)

15.6 (0.3) 15.7 (0.3)

P value Males

<0.001b <0.001 b 520 26.9 (0.4)

<0.001 b <0.001 b 23.5 (0.4)

<0.001 b 0.002 17.6 (0.3)

0–4 years

43 24.0 (1.6) 25.0 (1.6)

19.1 (1.4) 19.9 (1.4)

17.0 (1.1) 17.3 (1.0)

5–8 years

67 25.5 (1.2) 26.0 (1.2)

21.4 (1.1) 22.0 (1.1)

17.3 (0.8) 17.6 (0.7)

9–11 years

179 27.4 (0.6) 27.3 (0.6)

23.5 (0.6) 23.3 (0.6)

18.6 (0.5) 18.5 (0.5)

≥12 years

231 27.4 (0.6) 27.1 (0.6)

24.9 (0.6) 24.6 (0.5)

17.1 (0.5) 17.0 (0.4)

<0.001 b

0.278

P value

0.177

0.545

a

Adjusted for age, marital status and skin color

b

Linear trend test

0.002 b

an early stage of transition, the consumption of high-fat diets is higher among the richest. As incomes rise, diets high in complex carbohydrates and fiber give rise to denser energy and diets high in saturated and trans fats. In this sense, evidence from other studies in low- and middle-income countries is consistent with our results of higher

Crude mean (SE) 256.6 (3.4) 239.3 (8.8) 249.9 (5.8) 253.0 (6.1) 264.7 (3.5) 0.002 b 294.9 (4.8) 256.6 (8.9) 284.0 (13.7) 293.1 (8.1) 306.0 (7.1) 0.003 b

Adjusted a mean (SE)

Trans (g) Crude mean (SE)

Adjusted a mean (SE)

3.4 (2.0) 240.0 (8.9)

3.9 (0.1)

3.8 (0.2)

246.2 (5.6)

3.6 (0.1)

3.7 (0.2)

253.3 (5.9)

3.4 (0.1)

3.4 (0.1)

265.4 (3.3)

3.3 (0.2)

3.3 (0.2)

<0.001 b

0.077 b 3.0 (2.0)

0.081 b

264.8 (10.1) 3.1 (0.4)

3.0 (0.4)

288.5 (14.2) 3.0 (0.3)

2.9 (0.3)

292.9 (8.0)

2.8 (0.2)

2.8 (0.2)

303.1 (7.0)

3.1 (0.1)

3.0(0.1)

0.028 b

0.545

0.464

energy consumption in the categories of high educational level and family income, as was found in Malaysia in 2008 (Mirnalini et al. 2008) and in São Paulo (Brazil) in 2007 (Molina et al. 2007). However, these results differ from a systematic review of European Union studies in which there was no evidence of an association between energy consumption and

Women

Lipids

Proteins Carbohydrates Lipids

Men

Fig. 1 Description of the sample stratified by sex according to adequacy of macronutrients by AMDR reference. Florianópolis, Brazil, 2012

0.233

Cholesterol (mg)

Proteins Carbohydrates 0%

10%

20%

30%

40% 50% 60% Percentage

70%

Relationship with AMDR reference values Below Within Above

80%

90% 100%

J Public Health Table 4 Description of the adequacy of nutrients analyzed according to sex and educational level among adults aged 22 to 63 years, Florianópolis, Brazil, 2012 Prevalence ratio of adequacy according to educational level Females

Males Adjustedd

Crude Nutrient Carbohydrates Proteins Lipids Saturated Monounsaturated Polyunsaturated Trans Cholesterol Fiber

Recommendation 45–65%a 10–35% a 20–35% a <10%b 15–20% b 6–10% b <1% b <300 mg b >25 g/day b

0–4 1.12 1.00 1.05 1.68 0.78 1.07 0.27 1.15 0.19

5–8 1.12 1.00 0.96 1.52 0.73 1.07 0.83 1.07 1.00

9–11 1.11 1.00 0.98 1.28 0.85 1.00 1.24 1.05 0.86

a

AMDR reference

b

WHO (2003) reference

c

Reference group: 12 or more years of schooling

d

Adjusted for age, marital status and skin color

P value c 0.049 1.000 0.408 0.005 0.579 0.469 0.131 0.076 0.264

0–4 1.09 1.00 1.04 1.74 0.94 1.09 0.28 1.18 0.15

5–8 1.12 1.00 0.92 1.63 0.76 1.07 0.74 1.12 0.93

socioeconomic status (Giskes et al. 2010). Therefore, this evidence reinforces the existence of a different pattern in low- and middle-income countries in comparison to high-income countries with regard to inequalities in food intake, in which the high socioeconomic position group that lives in low-income countries has a diet with higher consumption of energy. According to Darmon and Drewnowski (2008), analysis of energy intake according to socioeconomic status should be made with caution, since there seems to be underreporting of intake among the lower socioeconomic groups, which may introduce bias in the results. In our study, when assessing the relationship between the TEI and the Estimated Energy Requirement to estimate the probability of underreporting, the percentage was −16% for men (95% CI -18%, −14%) and −10% for women (95% CI -11%, −8%), and among the former this percentage was no different with regard to education or family income. However, among women, underreporting was lower only by those with a lower educational level (−19% 95% CI -24%, −14%), while for other educational levels the value of TEI was similar among them. Thus, it is unlikely that underreporting affected the results, as the energy intake and prevalence of adequacy were higher among the less educated groups. The effects of family income and educational level on diet composition were consistent, but the effects were greater for the latter. This finding may be related to the fact that family income is associated with access and purchase of food, while educational level is linked to knowledge about food consumption (Darmon and Drewnowski 2008; Lallukka et al. 2007). This suggests that in middle-income countries such as Brazil, investments in health policies that aim to improve educational levels can contribute to the adoption of healthier eating habits.

Adjustedd

Crude 9–11 1.11 1.00 0.97 1.29 0.86 0.99 1.16 1.06 0.86

P value c 0.118 0.527 0.461 0.007 0.805 0.298 0.161 0.009 0.184

0–4 1.34 0.97 1.05 2.47 0.41 1.35 1.03 1.45 1.28

5–8 1.12 1.00 1.06 1.71 0.70 1.18 1.18 1.15 1.34

9–11 1.04 0.99 0.96 1.41 0.95 1.17 1.12 1.15 0.83

P value c <0.001 1.000 0.297 <0.001 0.654 0.005 0.911 0.005 0.453

0–4 1.31 0.97 1.02 2.24 0.44 1.34 1.10 1.34 1.10

5–8 1.10 1.00 1.05 1.62 0.72 1.16 1.30 1.10 1.33

9–11 1.02 0.99 0.95 1.40 0.99 1.17 1.14 1.12 0.82

P value c 0.004 0.423 0.254 <0.001 0.686 0.006 0.790 0.057 0.566

It is noteworthy in this study that gender was not an effect modifier in the evaluated associations, except for fiber consumption. We stratified the results by gender following evidence in the literature, suggesting differences in food intake in men and women as a consequence of different social pressures and intuitive eating behavior depending on the socioeconomic position (Eicher-Miller et al. 2015; Gast et al. 2012). Finally, the confounding factors that were analyzed (skin color, age and marital status) did not appear to influence the outcomes, as there were confounders only in the analyses involving protein, cholesterol and carbohydrates. This study has some limitations. First, the assessment of food consumption by a 24HR applied across the sample and a second applied to only 40% of the sample cannot take into account all the consumption variations. In addition, this method depends on the memory of individuals, which cannot establish the total food intake with accuracy. However, these possible limitations were addressed by using the multiple pass method, formulated to assist the participants in remembering and giving more detailed information about the meals consumed the day before. The interviewers were also trained to reduce the possibility of error in data collection. The use of the USDA table as the reference method with the NDSR software is also a possible limitation, since data could represent the American rather than the Brazilian diet. With this in mind, some procedures detailed in the methods section were used to minimize errors, such as the correct selection of the recipes in the software, the use of a standardized process for data entry and the inclusion of specific Brazilian recipes. In addition, the choice of the NDSR software was made because of the extensive range of nutrients evaluated as well as to allow for comparison with other international studies.

J Public Health

Another possible limiting factor is the analysis of the adequacy according to the caloric contribution percentage of each nutrient. In this regard, there may be subjects whose percentage of nutrient adequacy is considered normal based on the total energy contribution, but outside the recommended amount in grams for that nutrient. Nevertheless, when some of the analyses were repeated using the recommendation in grams per kg (i.e., for protein 0.8 g per kg), the results were consistent, showing a direct association with educational level in women and a lack of association in men. On the other hand, it should be emphasized that the study was conducted with a population-based sample of adults from a Brazilian capital, in addition to the high response rate, including the sample weights with the probability of location in 2012 to reduce the possibility of bias in the follow-up. Finally, the technique for measuring consumption using the intra- and inter-individual variation represents a strength of the study, as few population studies make this adjustment, especially in middle-income countries such as Brazil.

Conclusions Among the adult population of Florianópolis, Southern Brazil, a lower percentage of adequacy in food consumption was found in both sexes for those of a higher socioeconomic status. Considering that the assessment of dietary intake is essential for directing public policy, both for the prevention of various deficiency diseases as well as chronic non-communicable diseases, the results of this study suggest that for the implementation of measures to promote health and a healthy diet, subjects with a higher level of education and a higher income should also be the target of such initiatives.

Compliance with ethical standards Conflict of interest The authors declare that they have no conflict of interests.

Ethical approval In both waves, the project was approved by the Ethics Committee on Human Research of the Federal University of Santa Catarina (351/08 and 1772/11), and all the respondents signed an informed consent form.

Funding This work was supported by the Brazilian National Council for Scientific and Technological Development (CNPq) nos. 485327/ 2007–4 and 477,061/20109.

References Araujo MC, Verly Junior E, Junger WL, Sichieri R (2014) Independent associations of income and education with nutrient intakes in Brazilian adults: 2008-2009 National Dietary Survey. Public Health Nutr 17:2740–2752 Boing AC, Peres KG, Boing AF, Hallal PC, Silva NN, Peres MA (2014) EpiFloripa health survey: the methodological and operational aspects behind the scenes. Rev Bras Epidemiol 17:146–162 Brasil – Instituto Brasileiro de Geografia e Estatística (2015) Conceitos h t t p : / / w w w. i b g e . g o v. b r / h o m e / e s t a t i s t i c a / p o p u l a c a o / condicaodevida/indicadoresminimos/conceitos.shtm. Accessed November 06, 2015 Brasil-Ministério do Planejamento, Orçamento e Gestão (2010) Pesquisa de Orçamentos Familiares 2008–2009: Aquisição alimentar domiciliar per capita. Instituto Brasileiro de Geografia e Estatística-IBGE, Rio de Janeiro Costa AF, Yokoo EM, Antonio dos Anjos L, Wahrlich V, Olinto MTA, Henn LR, Waissmann W (2013) Seasonal variation of food intake of adults from Niteroi, Rio de Janeiro, Brazil. Rev Bras Epidemiol 16: 513–524 Darmon N, Drewnowski A (2008) Does social class predict diet quality? Am J Clin Nutr 87:1107–1117 Dodd KW, Guenther PM, Freedman LS, Subar AF, Kipnis V, Midthune D, Tooze JA, Krebs-Smith SM (2006) Statistical methods for estimating usual intake of nutrients and foods: a review of the theory. J Am Diet Assoc 106:1640–1650 Eicher-Miller HA, Fulgoni VL, Keast DR (2015) Energy and nutrient intakes from processed foods differ by sex, income status, and race/ethnicity of US adults. J Acad Nutr Diet 115:907–918 e6 Ferreira VA, Silva AE, Rodrigues CA, Nunes NLA, Vigato TC, Magalhães R (2010) Inequality, poverty and obesity. Cien Saude Colet 15:1423–1432 Fukuda Y, Hiyoshi A (2012) High quality nutrient intake is associated with higher household expenditures by Japanese adults. Biosci Trends 6:176–182 Gast J, Madanat H, Nielson AC (2012) Are men more intuitive when it comes to eating and physical activity? Am J Mens Health 6:164–171 Giskes K, Avendano M, Brug J, Kunst AE (2010) A systematic review of studies on socioeconomic inequalities in dietary intakes associated with weight gain and overweight/obesity conducted among European adults. Obes Rev 11:413–429 Institute of Medicine (2005) Dietary Reference Intakes for Energy, Carbohydrate, Fiber, Fat, Fatty Acids, Cholesterol, Protein, and Amino Acids (Macronutrients). Dietary Reference Intakes for Water, Potassium, Sodium, Chloride, and Sulfate, http:// publications.gc.ca/site/eng/278308/publication.html. Accessed November 06, 2015 Kearney J (2010) Food consumption trends and drivers. Philos Trans R Soc Lond Ser B Biol Sci 365:2793–2807 Lallukka T, Laaksonen M, Rahkonen O, Roos E, Lahelma E (2007) Multiple socio-economic circumstances and healthy food habits. Eur J Clin Nutr 61:701–710 Leblanc V, Bégin C, Corneau L, Dodin S, Lemieux S (2015) Gender differences in dietary intakes: what is the contribution of motivational variables? J Hum Nutr Diet 28:37–46 McCartney DM, Younger KM, Walsh J, O’Neill M, Sheridan C, Kearney J (2013) Socio-economic differences in food group and nutrient intakes among young women in Ireland. Br J Nutr 110:2084–2097 Mente A, de Koning L, Shannon HS, Anand SS (2009) A systematic review of the evidence supporting a causal link between dietary factors and coronary heart disease. Arch Intern Med 169:659–669 Mirnalini K Jr, Zalilah MS, Safiah MY, Tahir A, Siti Haslinda MD, Siti Rohana D, Khairul Zarina MY, Mohd Hasyami S, Normah H (2008)

J Public Health Energy and nutrient intakes: findings from the Malaysian adult nutrition survey (MANS). Malays J Nutr 14:1–24 Mishra GD, McNaughton SA, Ball K, Brown WJ, Giles GG, Dobson AJ (2010) Major dietary patterns of young and middle aged women: results from a prospective Australian cohort study. Eur J Clin Nutr 64:1125–1133 Molina MC, Bettiol H, Barbieri MA, Silva AAM, Conceição SIO, Santos JE (2007) Food consumption by young adults living in Ribeirao Preto, SP, 2002/2004. Braz J Med Biol Res 40:1257–1266 Morimoto JM, Marchioni DM, Cesar CL, Fisberg RM (2012) Statistical innovations improve prevalence estimates of nutrient risk populations: applications in Sao Paulo, Brazil. J Acad Nutr Diet 112:1614–1618 Moshfegh A, Goldman J, Cleveland L (2005) What we eat in America, NHANES 2001–2002: usual nutrient intakes from food compared to dietary reference intakes. US Department of Agriculture, Agricultural Research Service, Maryland Mullie P, Clarys P, Hulens M, Vansant G (2010) Dietary patterns and socioeconomic position. Eur J Clin Nutr 64:231–238 NEPA 2015 Tabela de Composição de Alimentos. TACO-UNICAMP (2004). Available in: < http://www.unicamp.br/nepa/taco/ >. Accessed November 05 Pinheiro AB (2004) Tabela para avaliação de consumo alimentar em medidas caseiras, 5th edn. Atheneu, São Paulo Programa das Nações Unidas para o Desenvolvimento (PNUD) 2013 Atlas de desenvolvimento humano no Brasil. Brasília: PNUD. Available at

Rombaldi AJ, da Silva MC, Neutzling MB, Azevedo MR, Hallal PC (2014) Factors associated with the consumption of high-fat foods among adults in a southern Brazilian city. Ciênc saúde coletiva 19: 1513–1521 Serra Majem L, Ribas Barba L, Salvador Castell G, Roman Viñas B, Castell Abat C, Cabezas Peña C, et al. (2007) Trends in the nutritional status of the Spanish population: results from the Catalan nutrition monitoring system (1992-2003). Rev Esp Salud Publica 81(5):559–570 Silva DA, Peres KG, Boing AF, Gonzalez-Chica DA, Peres MA (2013) Clustering of risk behaviors for chronic noncommunicable diseases: a population-based study in southern Brazil. Prev Med 56(1):20–24 Sousa TF, Nahas MV, Silva DAS, Duca GFD, Peres MA (2011) Factors associated with central obesity in adults from Florianopolis, Santa Catarina: a population based-study. Rev Bras Epidemiol 14:296–309 Tarasuk V, Fitzpatrick S, Ward H (2010) Nutrition inequities in Canada. Appl Physiol Nutr Metab 35:172–179 Teixeira MG, Mill JG, Pereira AC, Molina MCB (2016) Dietary intake of antioxidant in ELSA-Brasil population: baseline results. Rev Bras Epidemiol 19:149–159 Westenhoefer J (2005) Age and gender dependent profile of food choice. Forum Nutr 57:44–51 World Health Organization (1995) Physical status: the use and interpretation of anthropometry. Geneva: World Health Organization. (Technical Report Series, 854) World Health Organization (2003) Technical Report Series. Diet, nutrition and the prevention of chronic diseases: report of a joint WHO/ FAO expert consultation. WHO technical report series; 916 Geneva