EQUATION FOR EATING DISORDER RISK BEHAVIORS IN FEMALE ATHLETES

Introduction: It is difficult to detect athletes at risk of eating disorders. Objective: The aim of this study was to create a predictive equation for eating disorder risk behaviors (EDRB) in female athletes. Methods: 484 athletes participated in the study. The Eating Attitudes Test (EAT-26) was used to assess EDRB. The Sociocultural Attitudes Towards Appearance Questionnaire-3 was used to analyze the sociocultural influence on body image. The Body Shape Questionnaire and the Body Checking and Avoidance Questionnaire were used to assess body dissatisfaction in relation to thinness, and body checking, respectively. Skinfold thickness, body weight and height were also measured. The demographic data collected were age and weekly training regime. Results: The findings indicate that body mass index (BMI), body fat percentage (BF%), sociocultural influence, body checking and body dissatisfaction in relation to thinness explain 23% of the EDRB variance (p = 0.001). The results showed no difference between the EAT-26 scores and the predictive equation (p = 0.27). Conclusion: It was concluded that the predictive equation is reliable for evaluating EDRB in female athletes. Level of evidence III; study case-control.


INTRODUCTION
Eating restrictions for long periods, self-inducing vomiting, use of laxatives/diuretics and strenuous physical exercise for the purpose of "burning calories" are considered risk behaviors for eating disorders. 1 Studies indicate that approximately 30% of female athletes adopt risk behaviors for eating disorders with the premise of reducing body weight to optimize sports performance. [2][3][4][5] However, it appears that risk behaviors for eating disorders have an inversely proportional relationship with sports performance. 1,6 Although scientific investigations in the sports field show a prevalence of 30% of risk behaviors for eating disorders in female athletes, 2,3,5 researchers emphasize that this prevalence may be even higher due to their omission regarding such behaviors. 7 According to Fortes et al., 8 due to fear of being cut from the main team, athletes can offer false answers in subjective tools that seek to track risk behaviors for eating disorders. Therefore, it is extremely important to develop a new method that can indicate athletes who are vulnerable to eating disorders without them knowing that they are being screened for such psychiatric syndromes. In this sense, creating an equation which includes factors associated with risk behaviors for eating disorders may be feasible to analyze whether female athletes are vulnerable to such behaviors.
The scientific literature has pointed out personal (body dissatisfaction and drive for thinness and body checking), 3,9 morphological (body fat percentage and body mass index) 5 and demographic (age and weekly training regime) 10 variables as predictors of health risk behaviors for eating disorders in female athletes. Thus, the aim of this study was to create a predictive equation of risk behaviors for eating disorders, using body dissatisfaction and drive for thinness, body checking, body fat percentage (% F), body mass index (BMI), age and weekly training regime as predictors.

Ethical considerations
This study was approved by the Human Research Ethics Committee (CAAE -05166712.8.0000.5407). The guardians (in the case of athletes under 18 years old) signed an Informed Consent Form (ICF), which explained the objectives and procedures of the study. All athletes (minors and over 18 years old) signed the Assent Form, agreeing with their voluntary participation in the investigation. Participants were guaranteed anonymity and total confidentiality in the data treatment.

Participants
The population consisted of Brazilian female athletes, aged 12 years or over. The following inclusion criteria were adopted: a) present the ICF signed by the guardian (if she was under 18) and sign the Assent Form; b) systematically train her respective modality with a minimum training regime of six hours per week (minimum frequency of three days per week and duration of two hours/training, or frequency of six days per week and duration of one hour/training); c) have minimal participation in a regional competition in 2014 and; d) be available to answer questionnaires and participate in anthropometric assessments. In addition, the following exclusion criteria were adopted: a) present a physical or intellectual disability (indicated by the trainer) and; b) use psychoactive medications in the last six months (indicated by the athlete himself ).

Eating Attitudes Test (EAT-26)
EAT-26 was applied to assess the risk behavior for eating disorders. The questionnaire consists of 26 questions divided into three sub-scales: 1) diet -refers to the pathological refusal of foods with a high caloric content and concern about physical appearance; 2) bulimia and concern about food -refers to episodes of binge eating, followed by purgative behaviors for loss/control of body weight and; 3) oral self-control -reflects self-control in relation to food and assesses the environmental and social forces that stimulate food intake. EAT-26 final score is given by the sum of its items.
There are six answer options ranging from 0 to 3 points (always = 3, usually = 2, often = 1, sometimes = 0, rarely = 0 and never = 0). The version used was validated for the Brazilian population, 11 presenting a 0.82 internal consistency analysis. For the present study, Cronbach' s alpha of 0.89 was found.

Body Shape Questionnaire (BSQ)
BSQ was applied to assess body dissatisfaction and drive for thinness. It is a self-assessment test with 34 questions that seek to assess the subjects' concern about their weight and body shape. The version used for subjects aged 18 years and younger was validated for Brazilian adolescents, 12 and its internal consistency analysis revealed a 0.96 α for both sexes and a significant correlation coefficient between test-retest scores. The version used for athletes over 18 years old was validated for the young Brazilian population. 13 BSQ score is given by the sum of its items. The higher the score, the greater the dissatisfaction with the body. It should be noted that the items in BSQ versions for adults and adolescents are identical. For the sample of this study, Cronbach's alpha of 0.95 was identified.

Body Checking and Avoidance Questionnaire (BCAQ)
BCAQ assessed body checking behaviors. BCAQ consists of 22 items, subdivided into 5 sub-scales, namely: 1) tweak -reflects the frequency with which subjects evaluated "tweak" themselves to check or highlight  Incomplete anthropometric data (n = 49) parts of the body; 2) use of mirror -refers to the frequency with which subjects look in the mirror to check some part of the body; 3) palpation -refers to the frequency that subjects feel or touch parts of the body in order to identify body saliences; 4) use of measuring tape -indicates the frequency of using measuring tape to check the size of the perimeters of body parts; 5) self-weighing and social comparisons -reflects the frequency with which subjects weigh themselves and/or compare themselves with other women. BCAQ was validated for the young Brazilian female population and showed good psychometric properties. 14 For this sample, internal consistency was calculated, represented by Cronbach's alpha of 0.93.

Sociocultural Attitudes Towards Appearance Questionnaire-3 (SATAQ-3)
In order to assess the internalization of the ideal of thinness, the athletes answered SATAQ-3. This tool is scored on a Likert scale with five response options ("Strongly disagree" to "Strongly agree"). SATAQ-3 total score is calculated by the sum of the answers, with the highest score representing a greater influence of socio-cultural aspects on body image. The questionnaire consists of 30 questions designed to assess the general internalization of socially established standards (nine items), including the ideal athletic body (five items), the pressure exerted by these standards (seven items), and the media as a source of information about appearance (nine items). SATAQ-3 version used in this study was translated and adapted to the young Brazilian population, with its internal consistency established for this population. 15 For this sample, the internal consistency was calculated by Cronbach's alpha, obtaining a satisfactory value of 0.86.

Anthropometry
To calculate the percentage of body fat (% F), a LANGE® adipometer (Cambridge Scientific Industries Inc.) was used, with an accuracy of 1 mm, in order to verify the thickness of the skinfolds. The procedures for the collection of skinfolds were performed according to the standards determined by the International Society for Advancement for Kineanthropometry. 16 The literature recommends that there should be no differences greater than 10% between one measurement and the others in the same place, 17 therefore, the calculation of the technical error of measurement (TEM) proposed by Silva et al., 17 was performed excluding data with variance greater than 10%.
The % F was estimated by the doubly indirect method, using the equation developed by Siri. 18 For the calculation of body density, the protocols of Slaughter et al. 19 were used [considering ethnicity and maturation stage according to chronological age (pubescent -12 to 14 years old, and post-pubescent -15 to 17 years old)], and Jackson, Pollock and Ward, 20 created for adolescents and adult females, respectively. Triceps and subscapular skinfolds were measured in athletes under 18 years old. In athletes aged 18 years or older, tricipital, suprailiac and abdominal skinfolds were measured. These measurements were taken in a rotational manner and collected three times, considering the mean of the values. It should be noted that no skinfold datum was excluded due to ETM, since all variances were less than 5% for all athletes.
Body mass was measured using a Tanita® portable digital scale (São Paulo, SP, Brazil) with 0.1 kg accuracy and a maximum capacity of 220 kg. A Tonelli® portable stadiometer (São Paulo, SP, Brazil) with a precision of 0.1 cm and a maximum height of 2.20 m was used to measure athletes' height. The body mass index (BMI) was obtained using the calculation: BMI = body mass (kg)/height (m²).

Demographic data
Demographic data [age and weekly training regime (weekly frequency x daily hours)] were collected by means of a questionnaire created by the researchers.

Procedures
The coaches of several teams of diverse sports were sought to explain the procedures, as well as the objectives of the study. After coaches' consent, a meeting was arranged with each team to talk to the athletes and explain all the ethical procedures of the research.
The study was divided into two moments: The first was intended for the application of EAT-26, BSQ, BCAQ and SATAQ-3 questionnaires. It is worth mentioning, above all, that only one researcher was responsible for their application. This precaution was taken so that there were no differences in explanations and influence from other researchers on the responses to self-administered questionnaires.
During the collections, the athletes received the same verbal guidance, and written guidance on appropriate procedures was also present in the questionnaires. There was no communication between the study subjects. The questionnaires were distributed at the moment the athletes entered the room, and filling them out was configured as a voluntary character. There was no time limit for filling them out.
In the second stage, which was carried out immediately after completing the questionnaires, anthropometric measurements (body mass, height and skinfolds) were performed. The clubs that participated in the research provided adequate rooms for the evaluations. The measurements were performed individually in the environment to avoid interference from teammates.
These procedures were carried out in clubs and/or competitive events in 11 cities in eight Brazilian states, during the ten-month period, until the desired sample was reached.

Statistical analysis
Kolmogorov-Smirnov test was conducted to assess data distribution. Due to the result of this test for EAT-26 scores, parametric techniques were chosen. Therefore, Pearson's correlation was used to analyze the relationship between the independent variables (BSQ, BCAQ, SATAQ-3, % F, BMI and demographic data) with EAT-26 scores. Then, the backward method was used to create the formula for predicting risk behaviors for eating disorders (EAT-26 score), inserting in the model only variables that demonstrated r ≥ 0.30 with a 5% significance. In this sense, the following blocks were conducted: 1) BMI; 2) % F; 3) sociocultural influence (SATAQ-3); 4) body check (BCAQ), and 5) body dissatisfaction drive for thinness (BSQ). Finally, the dependent Student's t test was used to compare EAT-26 scores and the scores obtained by the predictive equation. This analysis was conducted with the purpose of evaluating the predictive equation validity. The data treatment was carried out using SPSS 21.0 software, with a 5% significance level.
Concerning the multiple regression model, the results indicated that BMI,% F, sociocultural influence, body checking and body dissatisfaction and drive for thinness explained 43% of the variance of risk behaviors for eating disorders (F (5, 479)  Regarding the comparison of EAT-26 scores to the predictive equation scores, the findings showed no difference (F (2, 482) = 1.83, t = 1.03, p = 0.27).

DISCUSSION
The research was based on the premise of building a predictive equation of risk behaviors for eating disorders for Brazilian female athletes. The findings indicated the predictive equation: Many coaches and sports professionals stress the difficulty of identifying athletes susceptible to an eating disorder. The possibility of false answers is pointed out in psychometric tools that seek to assess the frequency of risk behaviors for eating disorders as one of these difficulties, for example, EAT-26. Although EAT-26 is an instrument with good psychometric properties, 11 in the case of athletes, it is possible that they omit some responses due to the fear of being cut from the team. In this sense, it is necessary to develop new methods to screen athletes who are vulnerable to development of eating disorders.
The results of the this investigation indicated that chronological age and training regime did not related to risk behaviors for eating disorders. In fact, other studies have not shown a relationship between chronological age and frequency of risk of eating disorders. 1,10 This demonstrates that there is no specific chronological age for an athlete to adopt risk behaviors for eating disorders as a strategy for reducing body weight. However, although there is still no scientific data to prove it, some researchers highlight that the training regime may be related to the frequency of risk behaviors for eating disorders. 8,21 According to these researchers, young women at risk of developing eating disorders seem to seek the practice of sports with a strong training regime. Therefore, more research needs to be conducted to better clarify this relationship.
The findings of this study showed that BMI, % F, body check and body dissatisfaction and drive for thinness related to risk behaviors for eating disorders and, therefore, were included in the regression model for the construction of the predictive equation. It should be noted that 43% of the variance in risk behaviors for eating disorders was explained by the predictors (BMI, % F, body check and body dissatisfaction and drive for thinness). In addition, it should be noted that because all the predictors adhere to the model, it was possible to build an equation with the premise of predicting Brazilian female athletes' susceptibility to eating disorders.
Concerning the comparison between EAT-26 scores and the scores obtained in the predictive equation, the results showed no statistically significant difference. In this sense, the predictive equation is capable of reliably predicting EAT-26 scores. For example, an athlete who has the following characteristics (BMI = 25.00 kg/m²,% F = 25.00, SATAQ-3 = 90, BCAQ = 10 and BSQ = 80) will obtain a score of 21.21 regarding EAT-26, which indicates vulnerability to eating disorders, considering the cutoff established by Bighetti et al. 11 Therefore, after SATAQ-3, BCAQ, BSQ being answered, together with the anthropometric assessment (BMI and% F), it will be possible to track a female athlete who frequently uses risk behaviors for eating disorders.
Although this study indicates results with great practical applicability for sports coaches, it has limitations that deserve to be highlighted. The use of questionnaires is pointed out, as researchers emphasize that athletes may omit some behaviors/feelings. 8 In addition, another limitation was the use of a doubly indirect method to assess % F. According to some authors, 5 using skinfolds can be limited to assess body composition. However, TME was calculated, which increases measurement reliability. Although there are limitations, it is believed that this investigation brings a new method to screen female athletes susceptible to developing eating disorders.

CONCLUSIONS
The results allowed concluding that the predictive equation is reliable to assess risk for eating disorders in Brazilian female athletes. The need to conduct new scientific investigations in order to assess the external validity of the created predictive equation is emphasized. In addition, it is necessary to create a risk predictive equation for eating disorders also regarding Brazilian male athletes.
All authors declare no potential conflict of interest related to this article