Psychosocial determinant of selected eating behaviours in adolescents

Psychospołeczne uwarunkowania wybranych zachowań związanych z jedzeniem u nastolatków

Joanna Mazur, Anna Dzielska, Agnieszka Małkowska-Szkutnik
Department of Child and Adolescent Health
Head of Department: dr hab. med. prof. IMiD K. Mikiel-Kostyra
Institute of Mother and Child, Warsaw
Director: S. Janus

Abstract
Background: In population studies, we can follow the occurrence of initial symptoms of abnormal eating behaviours which may, but do not have to, lead to clinical manifestation of serious associated health problems, such as overweight or obesity, anorexia nervosa or bulimia nervosa.
The main aim: of this study was to determine to what extent psychosocial factors are associated with selected eating behaviours, such as: dietary restrictions, uncontrolled eating and emotional eating. Material and methods: Data from the postal survey conducted in 2008 on the sample of 605 children born in Poland in January 1995 served as a basis of this study. A short, 13-item Polish version of TFEQ (Three Factor–Eating Questionnaire) instrument was applied, representing three factors: Cognitive Restraint (CR), Uncontrolled Eating (UE) and Emotional Eating (EE). Multivariate linear regression models were estimated with eleven psychosocial factors, gender and body mass index BMI (1 – excess; 0 – lack of excess weight) as potential independent variables.
Results: The following variables appeared to be independent predictors of eating behaviours: 1) for CR: three risk factors – high BMI, gender (female) and low self-esteem; 2) for UE: two risk factors – high stress and low emotional support and one protective factor – positive attitude towards school; 3) for EE: two risk factors – high stress, low coping and two protective factors – high social acceptance and instrumental support. Psychosocial factors showed the strongest association with emotional eating (for final four predictors R2=0,078), and the weakest with dietary restrictions (for the only psychosocial predictor ÄR2=0,014).
Conclusions: 1) Negative behaviours, such as: dietary restrictions, uncontrolled eating and emotional eating may have strong association with psychosocial factors in adolescence. 2) In this period of life, the strongest association of psychosocial factors with emotional eating, and the weakest with dietary restrictions was found. 3)The knowledge of disordered eating predictors may become the basis for providing more effective intervention programmes which should aim to make young people aware of the mechanisms leading to this kind of disorders and potential consequences for health at present and in the future.

Key words: eating behaviours, psychological distress, social support, multivariate analysis

Streszczenie
Tło: W badaniach populacyjnych możemy śledzić pojawianie się pierwszych symptomów nieprawidłowych zachowań związanych z jedzeniem, które mogą, choć nie muszą, prowadzić do poważniejszych stanów chorobowych, takich jak nadwaga lub otyłość, anoreksja lub bulimia.
Głównym celem opracowania było określenie, w jakim stopniu czynniki psychospołeczne mogą wpływać na zachowania związane z jedzeniem u nastolatków, takie jak: ograniczanie jedzenia, niekontrolowane objadanie się i objadanie się na tle emocjonalnym.
Próba i metody: Podstawę opracowania stanowią badania ankietowe przeprowadzone w 2008 r., wśród 605 dzieci urodzonych w Polsce w styczniu 1995 r. Wykorzystano skróconą polską skalę TFEQ (Three Factor – Eating Questionnaire), która zawiera 13 pytań indeksowanych w trzech wymiarach: Ograniczanie Jedzenia (OJ), Niekontrolowane Objadanie się (NO) i Emocjonalne Jedzenie (EJ).Oszacowano wielowymiarowe modele regresji liniowej, włączając do nich jedenaście czynników psychospołecznych oraz płeć i wskaźnik masy ciała BMI (1 – nadmiar; 0 – brak nadmiaru masy ciała) jako potencjalne zmienne objaśniające.
Wyniki: Niezależnymi predyktorami zachowań związanych z jedzeniem były: 1) dla OJ: trzy czynniki ryzyka – wysoki BMI, płeć żeńska i niskie poczucie własnej wartości; 2) dla NO: dwa czynniki ryzyka – dystres i niskie wsparcie emocjonalne oraz jeden czynnik chroniący – pozytywna postawa względem szkoły; 3) dla EJ: dwa czynniki ryzyka: dystres i słabe radzenie sobie w życiu oraz dwa czynniki chroniące – wysoka akceptacja społeczna i wsparcie instrumentalne. Najsilniejszy okazał się związek czynników psychospołecznych z emocjonalnym objadaniem się (w końcowym modelu z czterema predyktorami R2=0,078), najsłabszy zaś związek z restrykcjami dietetycznymi (dla jedynego predyktora psychospołecznego ?R2=0,014).
Wnioski: 1) Niekorzystne zachowania związane z jedzeniem, takie jak ograniczanie jedzenia, niekontrolowane objadanie się i objadanie się na tle emocjonalnym, mogą mieć w okresie wczesnej adolescencji silne uwarunkowania psychospołeczne. 2) W tym okresie życia zarysował się najsilniejszy związek czynników psychospołecznych z jedzeniem na tle emocjonalnym, a najsłabszy z ograniczeniami jedzenia. 3) Znajomość uwarunkowań zachowań związanych z jedzeniem powinna stanowić podstawę budowy skuteczniejszych programów interwencyjnych, których celem powinno być uświadomienie młodym ludziom mechanizmów, które prowadzą do wystąpienia tego typu zaburzeń oraz ich wpływu na kształtowanie zdrowia zarówno w okresie adolescencji, jak i w wieku późniejszym.
Słowa kluczowe: zachowania związane z jedzeniem, dystres psychologiczny, wsparcie społeczne, analiza wielowymiarowa

INTRODUCTION
Disordered eating refers to troublesome eating behaviours, such as restrictive dieting, bingeing, or purging, which are less severe than those required to meet the full criteria for the diagnosis of an eating disorder (1). Eating disorders often occur in adolescence, as well as weight-related problems and body dissatisfaction. It could be considered as a result of complex problems, such as unfavourable life events, family situation, failures at school or abnormalities in the areas of feeling and expressing emotions. Eating disorders may depend on many family, cultural and individual factors (2). Adolescents have a specific way of reacting to stimuli coming from the external environment as well as to internal changes raising numerous strong emotions which are a sign of developing identity. For adolescents, eating habits may become an easy way to achieve a feeling of control over the environment, over their bodies and emotions, and also a method for solving problems. In studies based on clinical data, risk factors for the development of diseases related to diagnosed eating disorders and factors assisting recovery are analysed (3, 4). In population-based studies, we can only follow the occurrence of initial symptoms of abnormal eating behaviours which may, but do not have to, lead to clinical manifestation of serious health problems, such as overweight or obesity, anorexia nervosa or bulimia nervosa (5, 6). As has been stressed in the title, we focus on disordered eating as a behaviour rather then on eating disorder as a disease.
Various tools are used for measuring behaviours related to eating. Among those used in Poland, the EDI (Eating Disorders Inventory) adapted for Polish population (7) and originally developed in Poland KZZJ (Kwestionariusz Zachowań Związanych z Jedzeniem) (8) questionnaires are worth a comment. In this study, the TFEQ-13 (Three-Factor Eating Questionnaire) scale was used, which was adapted to Polish conditions by the same team of researchers (9), and which measures the cognitive, behavioural and emotional aspects of behaviours related to the control over eating.
The studies presented below are part of a project aimed to identify the factors which lead to overweight and obesity among young people (project no NN 404 329933 funded by Ministry of Science and Higher Education for period 2007-2010, principal investigator: K.Mikiel-Kostyra). Earlier analyses, carried out on the basis of the same data from 2008, suggest that in early adolescence disordered eating behaviours measured with the TFEQ-13 questionnaire have limited impact on the occurrence of excess weight (10). The likelihood of excess weight and obesity was only significantly higher among adolescents who introduced greater dietary restrictions, without important differences resulting from uncontrolled binge eating and emotion-related eating. Overweight may be caused by a cycle of restrictive eating, followed by overeating or binge eating.
The analysis of the link between psychosocial factors and excess body weight also led to ambiguous conclusions. A stronger correlation between body weight and quality of life than between body weight and mental health was identified (10). It may be assumed that eating behaviours shaped in that period of life increase the risk of obesity in late adolescence and in adulthood. Rather than having a direct impact on the occurrence of excess weight and obesity, psychosocial factors are more likely to determine the development of unhealthy behaviours and habits. However, the problem should be examined not only in the context of the threat of obesity, but also in the context of other potential health problems mentioned earlier.

AIM OF THE STUDY
The main aim of study was to determine to what extent psychosocial factors are associated with selected eating behaviours, such as: dietary restrictions, uncontrolled eating and emotional eating. The specific objectives were defined as: - to test an interaction between determinants of eating behaviours, especially between risk and protecting factors; - to investigate the above relationships in adolescents with and without excess body weight.

MATERIAL AND METHODS
Population under study
The studied group comes from a cohort of all children born in Poland between 1 and 10 January 1995 who participated in a three-stage prospective study. In 1995, a comprehensive nationwide study of 11,937 infants was carried out, out of which a 20% sample was randomly chosen to be studied in 1998. During the second stage, 65% of the randomly chosen sample was finally studied (N=1.250), which provided a basis for the 2008 postal survey. The subject of analysis in the study is the data collected in 2008 on 605 adolescents, taken from the questionnaires for children and from the growth charts. One advantage of this sampling design is a greater regional diversity of subjects, compared to school studies usually conducted in this age group (10).
The response rate in relation to the number of families to which the questionnaires were sent was 51.4%. It was achieved thanks to two reminder letters. The studied group was balanced in terms of gender (50.4% of girls) and representative for Poland with respect to the place of domicile. Urban residents made up 53.5% of this age group nationwide while it was 48.7% of our sample.

Measures
Dependent variables
Three main outcome measures were defined as continuous variables. The research tool for assessing eating behaviours was the Three-Factor Eating Questionnaire (TFEQ-13) questionnaire. The questions were taken from the scale provided by Stunkard and Messick, initially comprising 51 items (11), but subsequently reduced to 18 items by Karlsson (12). Psychometric features and the process of language adaptation of the Polish version was discussed in an earlier paper (9). The following results relate to the first use of that scale in such a large epidemiological study in Poland.
The TFEQ-13 scale is composed of three subscales, examined separately:
- Cognitive Restraint of Eating (CR) subscale - comprises 5 items; it measures behaviours related to restricting the amount or types of foods in order to control body weight and the body image; scale reliability of Cronbach's ?=0.787 was achieved;
- Uncontrolled Eating (UE) subscale - comprises 5 items; it assesses the tendency to eat more than usual due to the loss of control over eating and uncontrollable hunger resulting in binge eating; scale reliability of Cronbach's ?=0.721 was achieved;
- Emotional Eating (EE) subscale - comprises 3 items; it measures binge eating episodes caused by low mood and anxiety; scale reliability of Cronbach's ?=0.764 was achieved.
The TFEQ-13 questionnaire includes four response categories in 12 questions and eight categories in one CR item. All the answers were re-coded to the range of 0-3 points so that a higher score would mean higher intensity of abnormalities. The regression method (principal components analysis) was used to obtain standardized indexes for the three TFEQ-13 subscales (mean=0; SD=1). This ensured the optimal scale continuity from the point of view of further analyses.

Independent variables
In search of the potential psychosocial determinants of eating behaviours, three research tools known from the literature were used along with two questions from the HBSC (Health Behaviour in School-aged Children) study protocol (13). With regard to multi-item scales, standardized indexes were calculated in an identical way as for the TFEQ scale. The structure of all the scales used in the study was checked by factor analysis. Individual questions were categorized into two or three groups, depending on the responses and value distribution.
The first instrument used was GHQ (General Health Questionnaire) developed by D. Goldberg, which is a tool frequently used in epidemiological studies to diagnose mental health problems (14). The scale is composed of 12 questions with four response categories. Two responses have positive meaning and two have negative. There are many alternative methods for scoring and interpreting the GHQ-12 result. Usually, persons with three or more negative responses are identified as experiencing increased stress levels (psychological distress). It was emphasized in many studies that GHQ does not have a single-factor structure, which encourages the use of two or three indexes (15, 16). In the end, the experiences of Spanish authors (16) were applied, who (on the basis of confirmatory factor analysis) suggested a three-factor structure of GHQ-12, identifying subscales such as: Coping (6 items), Stress and Self-esteem (three items each). High score always indicates a negative state, i.e. respectively: low coping abilities, high stress levels and low self-esteem. Reliability coefficients for the three GHQ subscales were: 0.612, 0.642 and 0.725 respectively.
Another research tool was one of the modules of the SPPA (Self-Perception Profile for Adolescents) questionnaire proposed by S. Harter (17), which measures the level of social acceptance. It includes five questions and reliability of this scale was at ?=0.690.The high score was considered as high self-esteem in social relationships.
The third, widely-known tool was the social support scale developed by O. Dalgard (18), often cited as Oslo 3-Item Social Support Scale. It is composed of three items relating to various aspects of support, including: structural, emotional and instrumental. Due to such a construction, weaker psychometric characteristics (?=0.573) and greater interpretative possibilities, each of the three questions was analysed separately. In the question on structural support, two levels were defined: the absence of a close person and the presence of at least one close person who they could count on in case of need. In the question on emotional support, three levels were defined: positive – other people care about and take interest in the respondent, neutral – hard to say, and negative – other people do not care about and do not take interest in the respondent. Likewise, the question on the third type of support was re-coded to three levels, where positive level indicated ease of receiving help should the respondent need it. The questions were encoded in a way where higher value indicated positive state (better support).
Described elsewhere HBSC questions on general attitude towards school and the quality of family relations were also used (19). In the question on school, by combining two categories of responses, persons who like school and do not like school were identified. The question on family was treated as a continuous scale, adopting a 0-10 range. Young people used a visual scale, based on Cantril’s ladder, for their assessment of satisfaction from relations between family members.
It was considered necessary to adjust multidimensional analyses for gender of 13-year-olds and their BMI. The 2007 WHO cut-off points for 13-year-old boys and girls were applied here. Due to the small number of extreme BMI groups, underweight (3.1%) together within BMI in the range of standards (76.9%), as well as excess weight (11.4%) together with obesity (8.6%) were combined into one category. Every fifth adolescent was categorized as having excess weight (coded as "1").

Statistical analysis
Pearson’s correlation coefficients were calculated among the three scales of disordered eating behaviours and the five scales related to psychosocial factors. The mean values of TFEQ-13 subscales were compared using the t-Student and ANOVA tests in groups rhich differed tn terms of social support level and attitude towards school.
Multivariate linear regression models, predicting three types of behaviours, were estimated. Stepwise forward method of variable selection was used, which is considered the most reliable and enables achieving the optimal set of predictors. A positive regression coefficient means that the increase in independent variable will cause less favorable behaviours (risk factor), while a negative regression coefficient means better outcome (protective factor).
As regards the variables encoded as three categories, two dichotomous dummy variables were introduced to the model, which indicated the effect of extremely low and high levels.
It was checked whether or not model specification would change in the event of: 1) eliminating 18 underweight individuals; 2) limiting the sample to 445 adolescents with normal body weight.
A supplement to the analyses was an attempt to present the interaction between variables selected for the final models, which was carried out using the GLM (General Linear Model) method.
All analyses were performed using PASW Statistics 17.0. software.

RESULTS
Dependent and independent variables by gender and BMI
No significant differences dependent on gender were found when comparing the mean values of TFEQ-13 subscales. In the CR and EE scales, an explicit tendency for higher (i.e. worse) scores in girls was noted. A significant difference in the mean values of the CR scale dependent on body weight was observed. In overweight and obese adolescents, the mean CR values were much higher than those in adolescents without excess body weight.
Only significantly higher stress levels among girls were noted when comparing the mean values of the three GHQ subscales. Although young people with excess body weight achieved worse results in all the GHQ subscales, the differences were not statistically significant in any case. The mean results in Harter’s scale, as well as the assessment of family relations, did not greatly differ in terms of gender or the BMI.
The percentage of young people who positively assess school was slightly greater in girls than in boys (p=0.061) with no distinctive differences dependent on body weight. When analysing the level of social support depending on gender and body weight, an important difference in the perception of emotional support was found only between boys and girls. That type of support was assessed more favourably by girls (p=0.034).

Bivariate analysis
The correlation coefficients among the subscales of eating behaviours and the five continuous scales related to psychosocial factors are presented in table I. All the analysed psychosocial factors showed a link with at least one TFEQ subscale. The values for correlation coefficients were rather small; the maximum value of Pearson’s r was 0.218 for the link between stress and EE. Emotional eating was significantly correlated with all the five psychosocial scales. Dietary restrictions and uncontrolled binge eating were correlated mostly with the GHQ subscales; only the link with the coping subscale was weaker. This result confirms the legitimacy of analysing GHQ in three dimensions instead of one. Had the traditional approach been applied, the link between the general GHQ index and the CR scale would have not been revealed.
When comparing the mean values of the TFEQ-13 subscales, significant differences were noted with regard to school (fig. 1). UE-type behaviours were more common among young people who do not like school.
When analysing the mean values for the three TFEQ-13 scales depending on the level of social support, significantly higher UE values with regard to low emotional support and significantly higher EE values with respect to low instrumental support were observed (tab. II). If those three types of support had not been singled out, the impact of support on UE would have not been revealed. The correlation between UE and Dalgard’s continuous scale was only – 0.078 (p=0.069).

Multivariate analysis
A range of multivariate linear regression models were estimated for three types of TFEQ-13 eating behaviours. Based on the above methodological assumptions, 13 potential explanatory variables were added to them (11 psychosocial risk or protective factors, gender and BMI).
Three variables were independent predictors of CR-type eating behaviours: BMI, gender and GHQ low self-esteem. Table III features the results of the model estimation, showing the order of variable introduction and the impact of new predictors on the goodness-of-fit statistics. BMI explains 3.9% of the CR variability, while the combined value for the two other factors was only 2.2%. Stress was not included in the final model even though it showed an important correlation with CR in a simple correlation analysis.
Removing underweight individuals from the sample did not change model specification. After limiting the sample to individuals with normal body weight, GHQ self-esteem became the only CR predictor.
Three variables were independent predictors of UE-type eating behaviours, according to the order of their introduction to the model: stress, low emotional support and attitude towards school (tab. IV). In total, they explained 7.0% of UE variability. The two other GHQ subscales were not included in the final model, even though they showed an important correlation with UE in a simple correlation analysis.
After removing underweight individuals from the sample, the set of explanatory variables remained the same, they were only introduced to the model in a different order. Attitude towards school remained the most significant predictor. However, the level of emotional support was the main predictor in the model assessed for individuals with normal body weight.
Four variables (tab. V) were independent predictors of EE-type eating behaviours, according to the order of their introduction to the model: stress, social acceptance, coping and high instrumental support. In total, they explained 7.8% of EE variability. Self-esteem and the quality of family relations were not included in the final model, even though those factors showed an important correlation with EE in a simple correlation analysis.
After removing underweight individuals from the sample, the set of independent variables remained the same, social acceptance transpired to be the most important predictor. Conversely, only two variables: stress and high instrumental support remained in the model assessed for individuals with normal body weight.
Therefore, multivariate analysis showed that the majority of analysed psychosocial factors may have impact on at least one type of eating behaviours in early adolescence. It was only the quality of family relations that was not included in any final multivariate model. It has to be pointed out that both risk factors and protective factors were included in the models. Looking at the variables definition and the sign of regression parameter, social acceptance, high level of support and positive attitude towards school may be considered protective factors. Taking into consideration R2 coefficient of determination and it’s change after introduction a new factor to each model (tab.III-tab.V) we can assess to what extent psychosocial factors influence TFEQ-13 eating behaviours. Psychosocial factors have the strongest impact on emotional eating (R2=0,078 for the final model with 4 factors), and the weakest on dietary restrictions (?R2 =0,014 for the only psychosocial factor: self-esteem).

The effect on interaction
At the final stage of the analyses, with the use of the previous results, an assessment was attempted of the effect of the interaction between the risk factors and the protective factors against harmful behaviours. A hypothesis was made about the existence of an interaction between the level of stress and social support. For the purposes of this analysis, three questions on stress from the GHQ questionnaire were re-coded to a dichotomous variable. The value of “1” (increased stress) was assigned to individuals who responded negatively to at least one out of the three questions.
Figure 2 relates to the GLM model explaining UE-type eating behaviours. Regardless of the stress levels, adolescents with low level of emotional support achieved higher, i.e. less favourable UE values. The effect of a accumulation of the risks of negative behaviours was observed. Very high values on the UE scale were noted when increased stress and low emotional support co-existed. The graphic presentation of the mean values is interesting, but it is difficult to speak of interaction effect which would be statistically significant (p=0.172), which may partially result from too small number of the analysed cases.
Figure 3 relates to the GLM model explaining EE-type eating behaviours. In this case, statistical significance achieved was closer to the borderline value (p=0.093). The effect of accumulation of protective factors can also be observed. Very low, and therefore favourable means for EE were noted in the cases where stress was nonexistent and high instrumental support was present. The observation that in young people with low level of support the effects of stress are less visible seems interesting.

DISCUSSION
The results presented in the study concern 605 adolescents surveyed in 2008. The subject of the analysis was the probable impact of psychosocial factors on the occurrence of the first disordered eating symptoms as early as at the age of 13 yrs. A hypothesis was also made about the existence of an interaction between the risk factors and the factors protective against the occurrence of those symptoms. Eating-related problems were identified using the TFEQ-13 questionnaire. The following unfavorable behaviours were singled out: dietary restrictions, binge eating episodes and emotional eating. The GHQ-12 questionnaire, the S. Harter Social Acceptance Scale and the Oslo 3-Items Social Support Scale were also used in the studies. The presented results relate to population-based studies carried out in healthy adolescents. The terms “eating disorders”, “disordered eating symptoms” and “disordered eating behaviours” used in the paper with regard to the results of our study do not relate to descriptions of clinical cases. It may be assumed that disordered eating behaviours may lead to more serious mental disorders or may result in overweight and obesity in the future.
The results of multivariate analyses showed that almost all the psychosocial factors included in the studies were associated with the eating-related habits. The following protective factors were identified: the feeling of social acceptance, a high level of social support, satisfaction with school, along with the risk factors for those disorders: the strain of psychological stress, lack of the ability to cope in difficult situations. It was also demonstrated that the coexistence of positive and negative levels of the analysed variables causes a distinct fall or increase in symptom strength. For instance, a high level of social support and the absence of psychological distress are related to low (and therefore positive) results on the scale of emotional eating. However, cross-sectional design cannot be used to prove the directionality of the any association detected. The work presented here is part of the world research on the psychosocial determinants of eating disorders and disordered eating behaviours. The literature includes descriptions of population studies of healthy people, along with clinical cases (obese individuals as well as those suffering from bulimia or anorexia). The prospective studies by R. Isomaa et al. proved that eating disorders were fifteen times more likely to occur in adolescents on a diet who have previously signalled depressive moods than among their peers who did not manifest such problems (20). Furthermore, A. Doyle et al. distinguished a group of young people with problems related to social relations and feeling negative emotions. Their tendency for uncontrolled binge eating during the previous month was put under study. Binge eating episodes occurred significantly more often in this group than among their peers who were not at risk from the risk factors mentioned earlier (21). In other studies, a greater tendency to engage in abnormal eating behaviours was identified in young people who had poorer relations with their peers and family. This concerned also adolescents who had suffered abuse or bullying from peers or family members (22).
The link between eating behaviours and emotions concerns mostly individuals described in the literature as emotional eaters. The risk group includes individuals who make use of food to cope with excessive stress or negative emotions, such as annoyance or anger (23, 24). Changes in eating behaviours may be a symptom of a regulatory process. Emotions regulate eating (I eat when I am sad) and eating regulates emotions (I am less sad when I eat something sweet) (25).
The results we obtained did not show statistically significant differences dependent on the gender, which may result from too small a sample or the indicators used. For instance, other studies proved that a relationship exists between eating disorder symptoms and externalising behaviours, such as violence or psychoactive substance use. Moreover, the relationship between eating disorders and internalising behaviours were more often associated with girls (26). It may be assumed that the role of perceived social support in the context of symptoms heralding eating-related problems was, until now, considered marginal by researchers. Social support and its positive effects are usually mentioned with reference to individuals with clinical eating disorders. In most cases, it is the support received in the course of treatment/ therapy of a particular disorder (27).
This study considers social support as an important component of prevention. The outlined interaction between social support and stress confirmed the importance of social support as a factor which may lower the risk of those disorders. In context of the discussed relationships, the results of a study by C. Limbert are worth a mention; she analysed the perception of social support in a population of healthy young females who described themselves as more or less at risk from anorexia or bulimia. The differences suggested that individuals included in the eating disorder risk group perceive social support as lower than people who do not manifest such problems. A low level of perceived social support should be considered as a kind of a cry for help (28).
The presented results should be of particular importance to parents and primary care physicians, as well as for individuals working with young people: counsellors, teachers, psychologists and, in particular, people in charge of prevention. Seemingly insignificant symptoms and behaviour changes may not be ignored. People from the closest circle of adolescents have a daily opportunity to observe their health condition and attitudes towards eating and their own bodies. The correlation between observed abnormalities and psychosocial factors should encourage a greater care for young people’s mental health. Support provided for adolescents in this regard mostly means the time we devote to them, the confidence that they have in us, and their willingness to trust us with their problems and difficulties. Adolescents are more likely to step out of the vicious circle of influenced eating if they can share stress with a supportive person.
Above all, research results should be presented to the adolescents in order to enhance their knowledge and awareness of the mechanisms which link eating and emotions. This would be an attempt to prevent a further spread of negative role models, and it could encourage young people to timely rectify eating behaviour patterns.


CONCLUSIONS
1. Negative behaviours, such as: dietary restrictions, uncontrolled eating and emotional eating may have strong association with psychosocial factors in adolescence.
2. In adolescence, the strongest association of psychosocial factors with emotional eating, and the weakest with dietary restrictions was found.
3. The knowledge of disordered eating predictors may become the basis for providing more effective intervention programmes which should aim to make young people aware of the mechanisms leading to this kind of disorders and potential consequences for health at present and in the future.
4. The TFEQ-13 questionnaire is a tool worth recommending for epidemiological research on the determinants and effects of disordered eating behaviours.

REFERENCES
1. http://www.womenshealth.gov/bodyimage/bodyworks/ CompanionPiece.pdf
2. Józefik B. (red.): Anoreksja i bulimia psychiczna. Rozumienie i leczenie zaburzeń odżywiania się, Wydawnictwo Uniwersytetu Jagiellońskiego, Kraków 1999, s. 29-29.
3. Kirkcaldy B.D., Siefen G.R., Kandel I., Merrick J.: A review on eating disorders and adolescence. Minerva Pediatr., 2007, 59(3), 239-248.
4. Treasure J., Claudino A.M., Zucker N.: Eating disorders. Lancet, 2010, 375(9714), 583-593.
5. Espíndola C.R., Blay S.L.: Anorexia nervosa treatment from the patient perspective: a metasynthesis of qualitative studies. Ann. Clin. Psychiatry, 2009, 21(1), 38-48.
6. Trent M., Jennings J.M., Watereld G., Lyman L.M., Thomas H.: Finding targets for obesity intervention in urban communities: school-based health centers and the interface with affected youth. J. Urban. Health, 2009, 86(4), 571-583.
7. Zechowski C.: Polska Wersja Kwestionariusza Zaburzeń Odżywiania (EDI) – adaptacja i normalizacja. Psychiatr. Pol., 2008, XLII(2), 179-193.
8. Oginska-Bulik N.: Konstrukcja narzędzi badawczych. Kwestionariusz Zachowań Związanych z Jedzeniem (KZZJ). [W:] Ogińska-Bulik N.: Psychologia nadmiernego jedzenia.
Wyd Uniwersytetu Łódzkiego, Łódź 2004, 97-102.
9. Dzielska A., Mazur J., Małkowska-Szkutnik A., Kołoło H.: Adaptacja polskiej wersji kwestionariusza Three-Factor Eating Questionnaire (TFEQ-13) wśród młodzieży szkolnej w badaniach populacyjnych. Probl. Hig. Epidemiol., 2009, 90(3), 362-369.
10. Mikiel-Kostyra K., Oblacinska A. (red.): Czynniki biologiczne, behawioralne i psychospołeczne kształtujące masę ciała (BMI) 13-latków. Instytut Matki i Dziecka, Warszawa 2010.
11. Stunkard A.J., Messick S.: The Three-Factor Eating Questionnaire to measure dietary restraint, disinhibition and hunger. J. Psychosom. Res., 1985, Vol. 1, No. 1, 71-83.
12. Karlsson J., Persson L.O., Sjöström L., Sullivan M.: Psychometric properties and factor structure of the Three-Factor Eating Questionnaire (TFEQ) in obese men and women. Results from the Swedish Obese Subjects (SOS) study. Int. J. Obes., 2000, 24, 1715-1725.
13. Roberts C., Freeman J., Samdal O. et al.: The Health Behaviour in School-aged Children (HBSC) study: methodological developments and current tensions. Int. J. Public Health, 2009, Sep; 54 Suppl 2, 140-150.
14. Goldberg D., Williams P.: Podręcznik dla Użytkowników Kwestionariusza Ogólnego Stanu Zdrowia [W:] Dudek B. (red.) Ocena zdrowia psychicznego na podstawie badań kwestionariuszami Davida Goldberga. Instytut Medycyny Pracy, Łódź 2001.
15. Hankins M.: The factor structure of the twelve item General Health Questionnaire (GHQ-12): the result of negative phrasing? Clin. Pract. Epidemiol. Ment. Health, 2008, 4, 10.
16. Sanchez Lopez M., Dresch V.: The 12-item General Health Questionnaire (GHQ-12): Reliability, external validity and factor structure in the Spanish population. Psicothema, 2008, 20(4), 839-843.
17. Harter S.: Self-Perception Profile for Adolescents. University of Denver, Denver CO, 1988.
18. Dalgard O., Björk S., Tambs K.: Social support, negative life events and mental health. Brit. J. Psych., 1996, 166, 29-34.
19. Mazur J., Woynarowska B., Kołoło H.: Zdrowie subiektywne, styl życia i środowisko psychospołeczne młodzieży szkolnej w Polsce. Raport techniczny z badań HBSC 2006, Instytut Matki i Dziecka, Warszawa 2007.
20. Isomaa R., Isomaa A.L., Marttunen M. et al.: Psychological distress and risk for eating disorders in subgroups of dieters. Eur. Eat Disord. Rev., 2010, 18 (4), 296-303.
21. Doyle A.C., Ie Grange D., Goldschmidt A. et al.: Psychological and physical impairment in overweight adolescents at high risk for eating disorders. Obesity, 2007, 15 (1), 145-154.
22. Libbey H.P., Story M.T., Neumark-Sztainer D.R. et al.: Teasing, disorder eating behaviors, and psychological morbidities among overweight adolescents. Obesity, 2008, 16 suppl. 2, 24-29.
23. Canetti L., Bachar E., Berry E.M.: Food and emotion. Behavioural Processes 2002, 60, 157-164.
24. Wallis D.J., Hetherington M.M.: Emotions and eating. Selfreported and experimentally induced changes in food intake under stress. Appetite 2009, 52, 355-362.
25. Match M.: How emotions affect eating: A five-way model. Appetite 2008, 50, 1-11.
26. Aime A., Craig W.M., Pepler D. et al.: Developmental pathways of eating problems in adolescents. Int. J. Eat Disord., 2008, 41 (8), 686-696.
27. Pasold T.L., Boateng B.A., Portilla M.G..: The use of a parent support group in the outpatient treatment of children and adolescents with eating disorders. Eat Disord., 2010, 18(4), 318-332.
28. Limbert C.: Perception of Social Support and Eating Disorder Characteristics. Health Care for Women International. Online publication date: 07 January 2010.

Adres do korespondencji / Address for correspondence:
Joanna Mazur
Institute Mother and Child
ul. Kasprzaka 17a, 01-211 Warszawa
tel: (+48 22) 32-77-459
fax: (+48 22) 32-77-370
[email protected]