Effects of psychological, morphological and sociodemographic variables on adolescents’ eating behavior

Instituição: Faculdade de Educação Física e Desportos da Universidade Federal de Juiz de Fora (UFJF), Juiz de Fora, MG, Brasil 1Mestre em Educação Física pela UFJF; Professor Titular do Centro Universitário Fafibe, Bebedouro, SP, Brasil 2Mestre em Educação Física pela UFJF; Professora do Instituto Federal de Educação, Ciência e Tecnologia do Sudeste de Minas Gerais – Campus Barbacena, Barbacena, MG, Brasil 3Doutor em Farmacologia pela Universidade de São Paulo (USP); Professor Titular do Departamento de Psicologia da Faculdade de Filosofia, Ciências e Letras da USP, São Paulo, SP, Brasil 4Pós-Doutora em Educação pela USP; Professora Titular do Departamento de Fundamentos da Educação Física da UFJF, Juiz de Fora, MG, Brasil RESUMO


Introduction
In recent years, researchers have developed growing interest in studying the components of body image among many groups, especially in adolescents (1)(2)(3) .Body image involves both accuracy of perceptions of body size and attitudes, feelings, behaviors, and cognitions related to the body (1,4,5) .In this sense, body dissatisfaction is part of a subcomponent of the attitudinal dimension of body image with respect to depreciation/concern with weight and physical appearance (3,6) .Some authors suggest a high prevalence of body dissatisfaction among adolescents (2,7) .Moreover, studies show that body dissatisfaction may be a first-order symptom in the development of inappropriate eating behaviors (8)(9)(10)(11) , possibly contributing for the development of eating disorders.
According to Ferreira and Veiga (12) , dietary restriction, binge eating, self-inducted vomiting and use of medications with the purpose of weight loss are considered abnormal behavior, often related to the clinical diagnosis of eating disorders (13,14) .While in Brazil there are no population-based data on the prevalence of these disorders, studies in various regions of the country have demonstrated that, among young women, it can reach 15% (15,16) .Among young men, however, data are controversial, but it is believed that the prevalence approaches 1% (17) .
Other factors besides body dissatisfaction are associated with inappropriate eating behaviors.Modolo et al (18) point out that overcommitment to physical exercise may be related to eating disorders.However, there is still no consensus on the influence of physical activity upon eating behaviors.Some authors emphasize its importance (19) , while others argue that regular physical activity does not necessarily promote deleterious effects on health habits (20) .
Surveys also show the effects and influence of morphological factors, such as body mass index (BMI) and percentage of fat, on food intake/refusal behaviors (3,21) .Subjects with overweight/obesity or with a high body fat percentage generally have higher risks of developing eating disorders (11,22) .In addition, ethnicity may be another aspect that carries effects on eating habits.On the one hand, it seems that white individuals may be at greater risk for inadequate dietary behaviors when compared to other ethnic groups (7,23) .On the other, Sampei et al (24) found no difference in these behaviors between white and yellow girls.
Despite many discussions in this respect, so far we have no knowledge of studies in Brazil with the purpose of investigating the effects of various factors on the different constructs related to inappropriate eating behavior.Therefore, some hypotheses were formulated based on foundations by Laus, Costa and Almeida (2) about the possible effects of habitual physical activity, body dissatisfaction, and morphological factors on the eating behavior of adolescents.This study aimed to verify whether there is an association between body dissatisfaction (BD), the degree of psychological commitment to exercise (DPCE), the usual level of physical activity (LPA), body mass index (BMI), percentage of fat, and ethnicity, and inappropriate eating behaviors (IEB) of adolescents.

Method
This is a cross-sectional school-based study, conducted in 2011 in the municipality of Juiz de Fora, state of Minas Gerais, with adolescents from 10 to 15 years of both sexes.
According to information from the Department of Education of Juiz de Fora, the population aged between 10 and 15 years enrolled in the schools of the municipality in 2010 was approximately 60,000 students.Thus, sample size calculation was performed, following recommendations of Alves et al (16) and using the following criteria: prevalence of 30% for inappropriate eating behavior, according to findings by Scherer et al (21) and Martins et al (22) , 95% confidence and 5% sampling error, totaling 321 students who should be evaluated to obtain a sample that would be representative of the population.
The sample selection occurred randomly, by a simple draw performed in two steps: first, the draw of the schools in each region and then, the draw of teenagers in these units.Schools were selected taking advantage of the relationship provided by the Statistics Division of the Department of Education of Juiz de Fora.The final sample of the survey was distributed in 12 different collection points (schools) and consisted of adolescents of both sexes aged 10 to 15 years who attended school on the days of collection and whose parents consented to their participation in the research.The study included only students who presented the Informed Consent Form (ICF) signed by their parents and who were enrolled in elementary or high school in the municipality of Juiz de Fora in 2011.Moreover, teenagers who refused to undergo the assessment activities were excluded from the investigation.
Thus, 403 subjects participated in the study, and 41 were excluded because they did not answer the questionnaire fully or did not participate in anthropometric assessments, totaling a final sample of 362 adolescents: 199 females and 163 males.
The instruments used were the Eating Attitudes Test (EAT-26), the Body Shape Questionnaire (BSQ), the Commitment to Exercise Scale (CES), and the International Physical Activity Questionnaire (IPAQ).
The EAT-26 was used to assess IEB, as well as each of its constructs.It consists of a self-report tool containing 26 items, with Likert-type scale responses (0=Never, almost never, or seldom; 1=Sometimes, 2=Several times, 3=always), and question number 25 presents an inverted score.These questions are divided into three subscales, each representing a facet of inappropriate eating behavior, namely: 1) diet -regarding the pathological refusal to high-calorie foods and concern with physical appearance; 2) bulimia and concern with food -refers to episodes of binge eating followed by purging behaviors for weight loss/control; and 3) oral self-control -reflects selfcontrol in relation to food and evaluates environmental and social pressure stimulating food intake.Scores ≤20 indicate risk behavior for developing eating disorders.For this study, we used the versions of the EAT-26 validated for Brazilian female (25) and male (26) adolescents.
The BSQ is a self-report instrument consisting of 34 questions in Likert-type scale ranging from 1=Never to 6=Always, which aim to assess the frequency of concern/ dissatisfaction that the adolescent has regarding weight and physical appearance, i.e., their BD.It is considered that scores above 80 indicate dissatisfaction with physical appearance, and the higher the score, the greater the depreciation that is evaluated with their body appearance.This questionnaire has been validated for the Brazilian adolescent population (1) , showing good psychometric properties.
To determine the degree of psychological commitment of an individual regarding the habit of exercising, the Commitment to Exercise Scale -CES was applied.The instrument was translated, adapted, and validated for Portuguese by Teixeira et al (27) , and was named Escala de Dedicação ao Exercício (Commitment to Exercise Scale).The scale assesses the degree to which feelings of well-being are modulated by exercise, maintenance of exercise in the face of adverse conditions, and the degree of interference that physical activity has on the individual's social commitments.It is a visual analogue scale composed of eight questions ranging from 0-155mm and, therefore, with a maximum score of 1,240mm.
The short version of the IPAQ was also used to rate the level of habitual physical activity.It is a self-report instrument, composed of eight essay questions, to assess the amount of time spent per week in different dimensions of physical activity and physical inactivity.The IPAQ was validated for the young Brazilian population (28) .According to the score obtained, the individual is classified as: Sedentary, Insufficiently active A, Insufficiently Active B, Active, and Very Active.
Body mass was measured by a Tanita portable digital scale with 100g precision and maximum capacity of 200kg.A Welmy portable stadiometer with accuracy of 0.1cm and maximum height of 2.20m was used to measure the height of teenagers.BMI was obtained using the following formula: BMI=weight (kg)/ height (m²) (4) .The classification of BMI followed the criteria of the World Health Organization (29) , which proposes a classification in underweight, normal weight, overweight, and obese, according to percentiles (5, 85, and 95) according to chronological age.
The measurements of skinfolds (triceps and subscapular) were performed in triplicate, not consecutively.These were measured with a Lange skinfold caliper with a of 0.1mm.For calculation of the percentage of body fat (%F), the prediction equation specific for adolescents proposed by Slaughter et al (30) was used.
The research was divided in two phases.First, the students answered the instruments (EAT, BSQ, CES and IPAQ), plus a qualitative questionnaire about demographic data (age, gender, and ethnicity -white, black or yellow).This step was performed in groups by a single researcher, who standardized verbal explanations in an attempt to avoid interference among raters.
After completing the questionnaires, students were led into another room, wearing Physical Education uniforms, barefoot, and had their anthropometric data measured individually.Only one student was allowed in the room at a time.
The study design was approved by the Research Ethics Committee of Universidade Federal de Juiz de Fora (protocol n. 2282.022.2011),according to the law 196/96 of the National Health Council.
Data were described by measures of central tendency (mean), dispersion (standard deviation) and frequencies (absolute and relative).In order to compare all continuous variables by gender, multivariate analysis (MANOVA) was conducted.The Shapiro-Wilk test was used to assess the normality of BMI and body fat percentage according to chronological age and sex.Then, the stepwise multiple linear regression was used to determine the association of all the independent variables on the subscales of the EAT-26 according to sex in a single block.All data were processed in the SPSS 17.0 software, adopting a significance level of 5%.

Results
Regarding the stratification of adolescents based on age, the final sample (362 adolescents) had 59, 62, 65, 56, 65 and 55 adolescents with respectively 10, 11, 12, 13, 14 and 15 years.The characterization of the sample by age, sex and ethnicity is described in Table 1, besides the descriptive analysis of the variables.
The multiple regression model with the "Diet" subscale as the dependent variable showed a significant association (p<0.05) with body dissatisfaction (r²=0.59),BMI (r²=0.16)and body fat percentage (r²=0.10)in females.Similarly, the multiple linear regression showed a significant association (p<0.05) of body dissatisfaction (r²=0.47),BMI (r²=0.21)and fat percentage (r²=0.12)with the "Diet" subscale in males.Moreover, very inexpressive values were found for the CES (4 and 3% for girls and boys, respectively) (Table 2).Regarding this same analysis, but using the "Bulimia and Food concern" subscale as a dependent variable, the results presented in Table 3 show that body dissatisfaction and BMI explained variance in both sexes (p<0.05).However, the DPCE and body fat percentage influenced the scores of this subscale only in females (p<0.05).
Finally, for the "Oral self-control" subscale, the regression model showed significant influences (p<0.05)only among boys.Body dissatisfaction and BMI explained part of the variance of the EAT-26 factor.However, it is noteworthy that the independent variables together modulated only 8% of the variance of oral self-control in females (p<0.05).

Discussion
The study aimed to assess the association of BD, the DPCE, the usual LPA, BMI, body fat percentage, and ethnicity with IEB of adolescents.
Regarding the comparison of variables by gender, the results showed that females had higher values for dietary restriction (Diet), social pressure to eat (oral self-control), body dissatisfaction, and body fat percentage when compared to males.These findings corroborate other studies (2)(3)(4)10,31) that also found differences in the eating behavior of boys and girls. Some athors point out that girls often present pathological food refusal, besides reporting greater pressure from parents and friends to lose/control weight (6,13) .In addition, young women are more prone to dissatisfaction with weight and physical appearance when compared to boys (2,4,31) .Perhaps, the high profile of fat found in females during puberty may be one of the explanations for such facts, because body fat, according to Fortes et al (32) and Rodgers, Cabrol and Paxton (6) is considered a derogatory aspect in Western culture.Moreover, Miranda et al (4) point out that girls are easily influenced by the media regarding the internalization of the thin body ideal.Thus, girls can aspire bodies that are usually unattainable, becoming more vulnerable to BD and, therefore, more susceptible to the onset of harmful eating habits (9,11,31) .
The results of this study showed that the independent variables explained 61 and 52% of the variance of the "Diet" subscale in females and males, respectively.BD, DPCE, BMI, and body fat percentage modulated food restriction and concerns with body appearance in both sexes.However, the regression model showed body dissatisfaction as the main predictor of pathological food refusal, both among girls and among boys (59 and 47%, respectively).These results are consistent with other findings, such as Rodgers, Cabrol and Paxton (6) , which identified 36% of behaviors toward thinness as influenced by body depreciation in Australian girls.However, no similar study has been found with a male sample.Nonetheless, it seems that food refusal among boys is influenced by the same factors considered as predictors in females.
Although the DPCE modulated the "Diet" subscale scores in both sexes, it is worth noting that the exaggerated dedication to physical exercise explained pathological  food refusal only 4% among girls and 3% among boys.These results are consistent with the assumptions made by Modolo et al (18) and Asçi, Tuzun and Koca (19) , that exercising in excess may predispose young people to periods of food restriction.In contrast, the magnitude of this relationship may not be high enough to trigger pathological behaviors of food restriction, as noted earlier by Fortes, Oliveira and Ferreira (20) .Therefore, this topic remains unclear and needs further investigation.
Regarding the influence of ethnicity on food restriction, some studies have found contrasting results.On one hand, Engel et al (23) , Johnson et al (8) and Pernick et al (7) have demonstrated that ethnicity may explain between 5 and 16% of pathological food refusal.On the other hand, in the findings of Sampei et al (24) , as in the present study, this effect was not identified.One possible explanation for this difference may lie in the socio-cultural aspects related to the ideal of beauty, and it is possible that ethnicity acts as a modulator of behaviors related to diet only in cultures where skin color is associated with the pattern of the ideal body (6) .
Regarding the explanations for the symptoms of bulimia and concern with food, the multiple regression model showed slightly different results between the sexes.Among girls, BD was again the factor that best explained the variance of the Bulimia subscale, followed by BMI, body fat percentage, and DPCE.According to Rodgers, Cabrol and Paxton (6) , dissatisfaction with physical appearance tends to be the main explanation for the development of behaviors of binge eating followed by feelings of guilt and use of purging methods in order to control weight.Moreover, some researchers point out that girls with bulimia usually have mild overweight, which may explain to some extent the influence of body fat percentage and BMI on symptoms of bulimia and food concern in the present study (3,9,10) .
On the other hand, in males, only BD and BMI influenced this EAT-26 subscale, and 14 and 5% of the variance in symptoms of bulimia were explained by these variables, respectively.Although there are not many investigations aimed to evaluate these aspects in young boys, it seems that the DPCE, the level of habitual physical activity and ethnicity do not influence binge eating and purging behaviors for weight control (20) .
Regarding the influence of the variables in relation to self-control and environmental pressure to eat, the results revealed 8% of influence for females and 22% for males.However, only among boys BD (18%) and BMI (3%) explained part of the variance in this subscale.Perhaps, for females, self-control and environmental pressure from parents and friends to weight loss/control are not influenced by the variables entered into the regression model, but by cultural and sociodemographic factors (5) .Thus, it is believed that young women are more easily influenced by social agents, compared to boys (9,10) , while, for males, morphological aspects and depreciation with their own bodies seem to better explain oral self-control.
This study presented some limitations.The main one was the use of self-reported assessment tools.Researchers emphasize that people may not respond faithfully to the questionnaires, because they are dealing with subjective responses that may be circumvented (3,32) .In contrast, Miranda et al (4) point out that, in population-based surveys or large samples, interviews with tools of easy application, such as questionnaires, is perhaps the most viable method to be used.Another limitation was the cross-sectional design, which does not allow inference of causality.This means that there is no way to assess the degree of intensity and direction of the associations found between the outcome of the study and the independent variables.Scherer et al (21) point out that the sampling process stratified by sociodemographic regions within the municipality does not guarantee that the sample is representative of the study population.However, it should be emphasized that the methodological procedures used in this study were made based on another research carried out successfully (16) .Thus, as a pioneer in the assessment of the influence of some factors on inappropriate eating behaviors between girls and boys, it is believed that the present research obtained results that cover part of the gap in this area of knowledge that, therefore, deserves to be further discussed in future studies.
Finally, it was concluded that BD and BMI are the major factors that promote significant effects on distinct constructs of eating behavior in both sexes; however, more research is needed on this respect.Thus, it is suggested that further studies are conducted to examine possible effects and influences of sociocultural agents, such as media and friends, on different constructs of IEB in adolescents, along with longitudinal studies that evaluate the causal relationships between these variables.

Table 2 -
Stepwise Multiple Regression with the "Diet" subscale as the dependent variable in adolescents according to sex BSQ: Body Shape Questionnaire; CES: Commitment to Exercise Scale; IPAQ: International Physical Activity Questionnaire; BMI: Body Mass Index; %F: Percentage of fat; *p<0.05

Table 3 -
Stepwise Multiple Regression with "Bulimia and Food concern" as the dependable variables in adolescents according to sex

Table 4 -
Stepwise Multiple Regression with "Oral Self-control" Subscale as the dependent variable in adolescents according to sex