Magnesium intake in a Longitudinal Study of Adult Health: associated factors and the main food sources

Resumo O estudo tem por objetivo identificar fatores sociodemográficos e de estilo de vida associados à ingestão de magnésio e descrever seus principais alimentos contribuintes no Estudo Longitudinal de Saúde do Adulto (ELSA-Brasil). Trata-se de um estudo observacional, transversal, desenvolvido com dados da linha de base do ELSA -Brasil (2008-2010). Associações entre a ingestão habitual de magnésio e fatores sociodemográficos e de estilo de vida foram testadas por regressão linear múltipla. Contribuintes alimentares foram identificados a partir do cálculo do porcentual de magnésio fornecido por cada item do QFA em relação quantidade total proveniente de todos os alimentos. Os principais alimentos contribuintes para a ingestão de magnésio foram: feijão, aveia, nozes, arroz branco, laranja, pão francês, peixe cozido, carne sem osso, leite integral e pão integral. Foram encontradas associações positivas entre consumo de magnésio e sexo feminino, faixa etária ≥ 60 anos, cor de pele autodeclarada como negra, indígena ou parda, renda “per capita” ≥ 3 salários mínimos e níveis de atividade física moderado ou vigoroso. Alimentos da dieta tradicional do brasileiro foram os maiores contribuintes para a ingestão de magnésio, que também foi influenciada por fatores sociodemográficos e de estilo de vida. Palavras-chave Magnésio, Fatores sociodemográficos, Estilo de vida, Alimentos contribuintes Abstract This study aimed to identify the sociodemographic and lifestyle factors associated with magnesium intake and describe the main food sources in the Brazilian Longitudinal Study of Adult Health (ELSA-Brazil). This observational, cross-sectional study was conducted using the baseline data from the ELSA-Brazil (2008–2010). Associations between usual magnesium intake and sociodemographic and lifestyle factors were analyzed using multiple linear regression. Food sources were identified by calculating the percentage contribution of each FFQ item to the amount of magnesium provided by all foods. The analysis was performed using Stata® software (version 12), assuming a statistical significance level of 5%. The top food sources to magnesium intake were as follows: beans, oats, nuts, white rice, orange, French bread, cooked fish, boneless meat, whole milk, and whole wheat bread. There were positive associations between magnesium intake and female sex; age ≥60 years; self-reported black, indigenous, or brown skin colors; per capita income ≥3 minimum wages, and moderate or vigorous physical activity levels. Sociodemographic and lifestyle factors were associated with magnesium intake among the evaluated individuals.


introduction
Magnesium is the second most abundant intracellular ion and is involved in many metabolic functions, being vital for the activity of more 300 enzymes 1 . It plays an important role in ATP synthesis and activates almost all glycolytic enzymes and those of citric acid cycle. It is related to cell membrane permeability and electrical activity, besides being important for bone mineralization, muscle relaxation, and neurotransmission [2][3][4] . Deficiency of this ion can favor the development of various chronic noncommunicable diseases (NCDs), such as metabolic syndrome [5][6][7] , type 2 diabetes mellitus 8,9 , fibromyalgia 10 , hypertension 8,11,12 , osteoporosis 13 , and cardiovascular diseases 14 .
The Estimated Average Requirement (EAR) 15 of magnesium is between 255 mg and 265 mg for women and between 330 mg/day and 350 mg/day for adult and elderly men. Magnesium is present in dark green vegetables, legumes, oilseeds, milk and dairy products, and whole grains. Fish, meat, and some fruits are the poorest sources of this mineral 2 . In the United States, 60% of the adult population have insufficient magnesium intake to attend the EAR 16 . This scenario was observed in more than 70% of the Brazilian adult population, according to the 2008-2010 National Food Survey (INA) 17 .
Food consumption of an individual or a population is strongly influenced by age, sex, income, and schooling [18][19][20] . In Brazil, family income is positively associated with the consumption of milk, meat, fruits, vegetables, and legumes; however, the consumption of vegetables and legumes is moderate even in the richest stratum of the population 17 . Furthermore, some studies reported that families in less favored socioeconomic strata and mothers with lower educational level consume more sweets and products rich in fat 20 .
Knowledge about food components of a population diet and the identification of the determinants of nutrient consumption can serve as subsidies for the formulation of public policies for the promotion of healthy eating and of combating NCDs. This study aimed to identify the sociodemographic and lifestyle factors associated with magnesium intake and describe the main foods that contribute to this nutrient among participants of the Longitudinal Study of Adult Health (ELSA-Brazil), the largest multicenter cohort ever recruited for research incidence and risk factors of NCD in the Brazilian population 21 .

Study population
This observational, cross-sectional study was developed using the baseline data from the EL-SA-Brazil. ELSA-Brazil participants were recruited between August 2008 and December 2010. ELSA-Brazil is a cohort of 15,105 participants of both genders, aged 35-74 years, and are active and retired workers from six different states of Brazil: Espírito Santo, Minas Gerais, Bahia, São Paulo, Rio de Janeiro, and Rio Grande do Sul. Data were collected by trained and certified personnel under strict quality control [21][22][23] . Those without food consumption information (n = 24) were excluded from this study, totaling 15,081 participants. Individuals below the 1st percentile and above the 99th percentile of the total energy intake estimates (n = 362) were also disregarded in order to exclude possibly invalid food intake data. Thus, the final study sample consisted of 14,719 individuals.
The ELSA-Brazil was approved by the research ethics committees of all its research centers. All individuals voluntarily participated in this study and signed an informed consent form.

Food consumption assessment
The food frequency questionnaire (FFQ) developed and validated for ELSA-Brazil was used to evaluate the habitual food consumption of participants in the last 12 months 24 . This semiquantitative FFQ has 114 food items and is answered by interview. The questions are structured into 3 sections: (1) food/preparations, (2) consumption portion measures, and (3) consumption frequencies, with 8 response options: "more than 3 times/day," "2-3 times/day, ""once a day,""5-6 times a week,""2-4 times a week, ""once a week,""1-3 times a month," and "never/ almost never." At the end of the FFQ, participants were asked if they changed their dietary intake or if they did a restrictive diet over the past six months, being the participants able to answer yes or no to this question.
To evaluate energy and nutrient intakes, we used the United States Department of Agriculture (USDA) Food Composition Database, except when its values were outside of the range of 80% to 120% from those described in the Brazilian Table of Food Composition, which cases the latter database was used 24 . To reduce the errors associated with dietary measurement, magnesium intake was adjusted by total energy intake using the residue method 25 . Energy-adjusted values were employed both in the stratification of quantiles and linear regression analysis.

Sociodemographic and lifestyle factors
The choice of sociodemographic and lifestyle factors that could influence the dietary pattern was based on previous studies that addressed the determinants of food intake in the Brazilian adult population 18,19 . Therefore, sex, age, schooling, income, self-reported skin color, smoking and alcohol habits, nutritional status, and physical activity level were selected for this study.
Participants were classified according to sex as male and female) and according to age as adults (34-59 years) and elderly (≥ 60 years). Schooling was categorized as "complete elementary school," "complete high school," and "higher education or postgraduate." The family income per capita was initially calculated as equivalent to the average minimum wage in the period between 2008 and 2010 (R$ 463.33) and then stratified into < 3 or ≥ 3 minimum wages. The following categories of self-reported skin color proposed by the Brazilian Institute of Geography and Statistics in the demographic census were questioned: "white," "black," "brown or mixed," "yellow," and "indigenous" 26 . Due the low frequency of yellow and indigenous reporters, these two categories were collapsed for analysis.
Smoking was evaluated using a semi-structured questionnaire about smoking habits at the time of the interview and in the past. Based on this questions, participants were categorized as "non-smokers," "former smokers," or "smokers." Alcohol consumption data (grams of ethanol/ day) were obtained from the FFQ. Participants were classified as alcohol "non-consumers" or "consumer" based on the reporting of consumption of any alcoholic beverages in the previous 12 months, irrespective of its frequency or amount.
To assess nutritional status, body mass index (BMI) was calculated and classified according to the World Health Organization criteria: low weight (< 18.5 kg/m 2 ), eutrophia (18.5-24.9 kg/ m 2 ), overweight (25-29.9 kg/m 2 ), and obesity (≥ 30 kg/m 2 ) 27 . For the evaluation of physical activity level, we used the International Physical Activity Questionnaire (IPAQ) 28 , which consist in predetermined questions on frequency and duration of walking as well as moderate and vig-orous physical activities at work, commuting, home and leisure times 29 . For the purposes of this study, we used only the domain of physical activity during leisure time, considering that these types of activities has been more consistently associated with socio-demographic factors, such as income, age, schooling and sex 30 . Moreover, physical activity in leisure is most frequently studied in epidemiological surveys 31,32 .

Statistical analysis
The consumption of energy-adjusted magnesium was stratified in quintiles in order to better represent the ranking of dietary magnesium, and sociodemographic and lifestyle factors were described according to the lowest (1st quintile) and highest (5th quintile) levels of its intake. Sociodemographic and lifestyle factors were presented as frequencies and percentages according to the sex of the participants. Pearson's chi-squared test was used to evaluate the significant associations between variables.
The contribution of food to magnesium intake was calculated according to the methodology proposed by Block et al. 33 . Magnesium provided by each food item was divided by the total population magnesium intake to obtain the contribution of each food item. Then, the foods were listed according to the contribution ranking 34 .
The associations between energy-adjusted magnesium intake (mg/day, dependent variable) and sociodemographic and lifestyle factors (predictors) were tested by multiple linear regression analysis using the stepwise backward method. The energy-adjusted magnesium consumption variable approaches normality, according to the Shapiro-Wilk test and the use of histogram and Q-Q plot graphs, thus meeting this assumption for multiple linear regression.The sociodemographic and lifestyle factors included in the model were sex (reference: male), age (reference: adults), income (reference: < 3 minimum salaries), skin color (reference: white), schooling (reference: complete primary school), smoking (reference: non-smoker), alcohol consumption (reference: non-consumer), assess nutritional status (reference: eutrophy) and physical activity (reference: light).
The multiple model was further adjusted by self-reported change in dietary habits over the past 6 months. All analyses were performed using the Stata® (version 12) software, assuming a level of statistical significance of 5%.

Results
The sample consisted of 14,719 participants, predominantly adults (78.5%), female sex (54.6%), non-smokers (57.1%), self-reported as white (52.6%), and with a higher education level or a post-graduate level (53.2%). As regards nutritional status, 40.3% of the population was classified as overweight and 22.8% were obese.
The distribution of sociodemographic and lifestyle characteristics according to the magnesium intake of men and women is presented in Table 1. Higher proportions of the elderly and individuals with a higher education or who achieved a postgraduate level, with income ≥ 3 minimum wages, who are former or non-smoker, with eutrophia, and with moderate or vigorous physical activity level had magnesium intake in the last quintile, compared with the first ( Table  1).
Except for schooling, all other sociodemographic and lifestyle variables investigated were independently associated with magnesium intake. As shown in Table 3, positive and significant correlations were found between intake of magnesium and female gender; age ≥60 years; skin color self-declared as black, brown, or indigenous; income ≥3 minimum wages; and moderate or vigorous physical activity levels. By contrast, smoking, alcohol consumption, and overweight or obesity were negatively associated with magnesium intake (Table 3).

Discussion
In this study, variations in magnesium intake among ELSA-Brazil participants (2008-2010) were explained by sociodemographic characteristics that influence food sources, such as sex, age, race/ethnicity, and family income. In addition, smoking and consume alcohol were lifestyle habits that were negatively associated with mineral intake, as did obesity and overweight, while the opposite was evidenced in relation to the level of leisure physical activity, independently of other factors evaluated. Food sources that contributed to more than a half of total magnesium consumption included beans, cereals (oats, rice, and French bread), nuts, oranges, meats (fish and cattle), and milk, although dark green vegetables, almonds, nuts, and legumes had little expressive participation, suggesting a possible dietary inadequacy 17,35,36 .
Consistent with our observations, in a population-based study, Sales et al. 35 reported that more than a quarter of the magnesium in the diet of São Paulo inhabitants came from beans, rice, and French bread, confirming the important contribution of typical Brazilian food standards. Moreover, age had a positive effect on the intake of magnesium and other minerals, such as calcium, phosphorus, and potassium, signaling better quality of diet among the elderly, in relation to adults and adolescents 35 . In our study, female gender, as well as age, was also associated with higher magnesium intake. In a previous analysis performed with the same ELSA-Brazil sample, Cardoso et al. 37 revealed that women and elderly had higher adherence to a "healthy" diet characterized by vegetables and fruits 38 , which could be related to the higher intake of magnesium among these individuals.
According to the data from POF 2008-2009, in the Brazilian population, schooling and income are indicators of socioeconomic status independently associated with the higher consumption of saturated fat, sodium, and lower consumption of fiber, indicating that purchasing power and educational level do not necessarily determine better food choices in our social context 33 . Furthermore, analyses showed that income, not schooling, was associated with higher magnesium intake, after adjusting for demographic and lifestyle characteristics. These findings may be due to factors related to access, availability, and prices of magnesium food sources (dairy products, fresh meats, and vegetables) 18,38 . Notably, ELSA-Brazil participants, linked to teaching and research institutions, present a higher level of education than the general Brazilian population, which could make income a stronger determinant of food consumption. In fact, we notice a relatively higher contribution of oats, walnuts, cooked fish and whole grain bread, but lower of beans to the total magnesium intake among participants with an income per capita ≥ 3 minimum wages, suggesting a different pattern of this mineral food sources consumption among the richer participants (Supplementary Table 1  The values presented are minimum and maximum for the continuous variable and frequencies and percentages for the categorical variables. Associations between categories were analyzed with the Pearson Chi-square test.
A value of P < 0.05 was considered statistically significant.
presented higher values of dietary magnesium. Due to ethnic miscegenation in Brazil, it is a fundamental element to understand the association of race/ethnicity with food consumption, given the recognized role of cultural heritage and historical value of food in the construction of traditional and healthy eating habits. In the National Health Survey (PNS, 2013), for example, black and brown skin colors were associated with a significantly higher frequency of regular bean consumption (≥ 5 times/week) 18 . As already commented, almost a quarter of the total intake of magnesium in ELSA-Brazil was attributed to this legume. Together with rice, beans make up the basis of traditional Brazilian lunch and dinner, and this combination has been shown to be a protective factor for obesity and other NCDs [41][42][43] . Changes in the gustatory ability of foods due to smoking and the recognized negative effect of excessive alcohol consumption on appetite and food consumption could explain the inverse correlation between these two lifestyle habits and the intake of magnesium. On the other hand, as already evidenced by another study 35 , higher values of dietary magnesium were estimated among the participants classified in the levels of moderate and vigorous physical activity. As characteristics of the nutritional transition faced by the country, urbanization and the adoption of unhealthy lifestyle habits have accompanied the increase in the consumption of ultra-processed foods and of low nutritional value 18,22 . The findings indicate the importance of promoting diet quality, with a stimulus to the consumption of magnesium sources, especially among subgroups at risk for NCDs. That way, there is an inverse association between magnesium intake and excessive body weight, that is, the worse the nutritional status the lower the consumption of magnesium, even after adjustment for energy consumption, physical activity level, and other sociodemographic and lifestyle characteristics evaluated. Some authors, based on evidence of a deleterious role of magnesium deficiency on insulin resistance, inflammation, and oxidative stress, support the hypothesis of a causal relationship between the inadequacy of the mineral and the aggravation of weight gain and expansion of body adiposity, a characteristic of obesity [44][45][46] . Although other population studies, such as ours, reported a lower intake of magnesium among obese individuals 16,44 , it is still uncertain whether these findings reflect a poor overall quality of the diet or if the inadequacy of its consumption would be a risk factor for the disease 45,46 . Due to the transversal design of the study, inferences of causality are not possible; however, they can be explored with a longitudinal follow up of these individuals.
Furthermore, we estimated magnesium intakes with a FFQ, which is a method widely used in large epidemiological studies to rank individual according to their levels of dietary intakes in the previous twelve months 45 . However, its use can be considered another study limitation since this method is not consider the most appropriate for the quantitative analysis of micronutrients, given its inherent inaccuracy, that preclude the evaluation of individual or population nutrients intake adequacy. However, ELSA-Brazil FFQ was previously validated and performed well in classifying individuals according to magnesium intake levels, allowing their use in our comparative analysis between groups 24,47 .
To evaluate the energy and nutrient intake, the Food Composition Database of the United States Department of Agriculture (USDA) or the Brazilian Food Composition trient values between databases, similarly to an approach employed by an American Latin multicentric study 48 . conclusion Foods from the traditional Brazilian diet were the largest contributors of dietary magnesium among the evaluated participants. In addition, not only sociodemographic but also lifestyle factors were associated with the ingestion of this mineral.

collaborations
The first author of the article, J Levy, done design and design of the study, analysis, interpretation of data and review of the article; AAM Miranda, JA Teixeira and E De Carli contributed analysis, interpretation of data and revision of the article; IJM Benseñor and PA Lotufo contribuited article revision and DML Marchioni contribuited review of the article and approval of the final version.