Orthorexia nervosa, eating patterns and personality traits: a cross-cultural comparison of Italian, Polish and Spanish university students

Background The amount of research about orthorexic attitudes and behaviours has increased in the last five years, but is still mainly based on descriptive and anecdotal data, yielding a variety of prevalence data and inconsistent results. The interplay between socio-cultural context and orthorexia has been poorly investigated and is still far from being understood. Method Multicentre, cross-sectional study involving Italian (N = 216), Polish (N = 206) and Spanish (N = 242) university students, assessed through a protocol including informed consent, socio-demographic and anamnestic data sheet and self-administered questionnaires (ORTO-15, Eating Attitudes Test- 26 [EAT-26], Temperament and Character Inventory [TCI]). Results Higher prevalence of orthorexia (as described by the ORTO-15 cutoff) was found in Poland. Female gender, Body Mass Index (BMI), current Eating Disorder, dieting, EAT-26 score ≥ 20 and low/medium Persistence were associated with orthorexia in the whole sample. The cross-cultural comparison showed several differences among the three subgroups of students. Conclusions The associations found between orthorexic attitudes, self-reported current eating disorder, BMI and adherence to a dieting need to be supported by further research. The differences among students from the three countries seem to suggest a possible rolve for cultural elements in the construct of orthorexia.

The interplay between socio-cultural context and orthorexia has been poorly investigated and still needs to be better understood. The associations found between orthorexic attitudes, self-reported current eating disorder, BMI and dieting may be explained by the great attention paid to the individual responsibility for health, typical of Western societies. Differences in the frequency of orthorexic attitudes and behaviors among the Italian, Polish and Spanish samples can be related to the role of gastronomic culture, Mediterranean diet, convivial and social value attributed to eating and main approaches aimed at improving one's health.

Background
Eating styles and behaviors have always been and are still currently deeply shaped by the context where people live, encompassing cultural, social and environmental factors [1]. For instance, Western societies have long been influenced by theoretical constructs on the purity of the body obtained through food restrictions, in a mind over matter framework that extends from classical culture and early Christian beliefs to contemporaneity [2].
In contemporary society, the interest in the body and the will to keep it healthy have become "obsessive" and sometimes leading to eating behaviors based on the compulsive search for natural and "pure" foods [3]. Recently, a new health consciousness", called "healthism" [4] suggests that health can be accomplished in an unproblematic way through individual discipline and conduct, mainly by doing regular exercise and healthy eating [5]. Further context-related elements which have gained importance and become an important part of this phenomena over the last years [6,7] include the massmedia and the Internet.
The concept of orthorexia nervosa (ON) defined as as a persistent fixation on healthy eating [8] has developed in this cultural and temporal context. Currently, though ON is not classified as a formal diagnostic category and its grading is still a matter of debate [9], several suggestions have been made about it being a distinct subtype of the Avoidant/Restrictive Food Intake Disorder (ARFID) described by the Diagnostic and Statistical Manual of Mental Disorders [10], sharing features with anorexia nervosa (AN) [11] or overlapping with obsessive-compulsive disorder [2]. While Dunn and Bratman [8] pointed out that ON could be considered as a distinct condition, the current state of research does not allow the final categorization of ON as a separate mental disorder [12]. Actually, despite the increase in the amount of research about this topic in the last five years [11,13], the literature on ON is still mainly represented by descriptive and anecdotal data, often with inconsistent results [14], and its prevalence in the general population has been recently estimated to be less than 1% [8].
Regarding the shared features between ON and AN, there are cognitive fixation on nutrition, perfectionism, high anxiety and need to exert control, achievement orientation, guilt over food transgressions, value adherence to dieting as a marker of self-discipline, limited insight, cognitive rigidity, denial of the functional impairments associated with disorder [9]. Therefore, it is likely, but not yet consistently supported by the existing literature, that some AN features, including personality traits, could be involved in the development and maintenance of ON as well [12,15].
The personality correlates of ON have been addressed only by a few studies. Higher levels of neuroticism [16], narcissism and perfectionism [17] were described in individuals with ON -related tendencies, cognitions, behaviors, and feelings. A recent study [18], performed with a widely used assessment tool for personality, i.e. the Temperament and Character Inventory (TCI) [19], has investigated personality traits in ON individuals, showing high Harm Avoidance, high Self-Trascendence and low Self-Directedness. According to these dimensions, an "orthorexic personality" has been described as characterized by excessive preoccupation and shyness in social situations, combined with the desire to be perfect and to feel accepted [18]. Further conclusions of this study suggested that strictly planned, excessively "pure" dieting would result from an intense need for control and as a compensation of poor self-esteem and feelings of ineffectiveness in managing harmful events [18].
According to these premises, we identified three main areas of interest: 1) a deeper understanding of the orthorexia construct, considering its possible and currently debated relation with full-blown eating disorders; 2) the investigation of the personality correlates of orthorexic attitudes and behaviors; 3) the role of the cultural context on orthorexia.Regarding the first area, there is still much inconsistency in the existing literature. A previous study [20] by our research group investigated orthorexia in a sample of patients with a clinical diagnosis of eating disorders, showing that more pathological ON attitudes and behaviors (as measured by the ORTO-15 test) were related to lower degrees of disordered eating (as assessed with the Eating Attitudes Test-26). While another study has supported a negative correlation between ON and disordered eating patterns [21], others (using the Bratman Orthorexia Test and the Eating Habits Questionnaire, respectively) found increased ON features corresponding to more severely disordered eating behaviors (particularly calories restriction and weight concerns) [2].
As far as personality is concerned, considering its role in eating disorders [22], the widespread use of the TCI and the familiarity we have with this assessment tool [23,24], we decided to include it in our research.
Last, the interplay between socio-cultural context and ON has been poorly investigated. We already addressed this topic in a previous study [20] involving Italian and Polish females with AN and healthy controls. Less orthorexic features, as assessed with the ORTO-15, emerged in the Italian than in the Polish samples (both AN and healthy control group). Research involving nonclinical populations of students have reported a wide variability in the prevalence of orthorexic attutides and behaviors, being up to 68.55% in Polish University students [26], and57.6% in a previous Italian study [25].
Therefore, the aim of the present study was: 1) to assess the prevalence of orthorexic attitudes and behaviours (as measured by the ORTO test; later on referred to in the text as ON, for the sake of conciseness) in a sample of university students enrolled in different European countries (Italy, Poland and Spain); 2) to investigate the association between ON, socio-demographic features, eating patterns (measured with the EAT-26) and personality traits (assessed with the TCI) within the whole sample; and 3) to investigate the possible crosscultural differences in ON and eating patterns in the three samples, subdivided according to nationality.
We put forward the hypothesis that ON prevalence and the differences in eating patterns could reflect eating habits in Italy, Poland and Spain as a mirror of cultural normative pressures.

Study design
Multicentre, cross-sectional study analyzing the ON construct, eating habits and personality traits in a sample of European university students from Italy, Poland and Spain. We recruited a convenience sample, based on previous collaboration among some of the authors (P.Z. and C.G., A.B.M., M.R.). Each of these authors proposed the research protocol to the students attending their University courses (different types of university courses were involved, as detailed below, thus ensuring a good representativity of the University population), whowere assessed with a protocol including informed consent, socio-demographic and anamnestic data sheet and selfadministered questionnaires.

Sample
The sample was composed by Italian, Polish and Spanish university students enrolled by the Psychiatry Institute, Department of Translational Medicine, Università del Piemonte Orientale at the Faculty of Medicine and Nursing Sciences, in Novara, Italy; by the Psychology Faculty of Katowice at different universities in Silesia, Lower Silesia, Lesser Poland and Mazovia regions (philosophy, dietetics, psychology, sociology students) in Poland; and by psychology students from the Faculty of Ciencias Sociales y Humanas of the Universidad de Zaragoza, in Teruel, Spain, and from the Faculty of Psycholgy, Universitat de València, in Valencia, Spain. Recruitment took place from January 1st 2016 to December 31st 2017. No inclusion/exclusion criteria were applied. Students were approached during lesson time (hence it was not possible to retrieve the exact number of those approached, nor that of those who declined); they were explained the study protocol and asked about their willingness to participate and to provide their informed written consent. In case they accepted to take part in the study, they were e-mailed a link with the study protocol, allowing anonymous compilation. The data sheet gathered information about gender, age, ethnicity, marital status, weight and height (in order to allow the calculation of BMI), use of substances and/or alcohol, cigarette smoking, sporting activity, dieting, use of nutritional supplements and past or current ED diagnosis. All the information gathered was self-reported, including that about anthropometric measures and ED diagnosis.

Orto-15
The ORTO-15 is a self-administered scale developed in 2004 by Donini and coworkers, modeled on the Bratman test [30,31]. It is the most frequently used tool for the assessment of orthorexic symptoms in various populations [26,32,33], even though it is clearly not a diagnostic tool.
The ORTO-15 was developed in Italian [26] and later translated in other language. Several studies have eliminated one or more of its items, based on confirmatory analyzes and the adequacy of the instrument [34].
The Polish version of the ORTO-15 has a maximum score of 36 and adopts a 24-points threshold cut-off [25].
Since our study involved samples recruited in Italy, Poland and Spain, the cutoff of 35 was used for Italy and Spain, and the cutoff of 24 for Poland, to identify possible ON, as described in the literature [36]: therefore, students were classified dichothomically, as scoring under or above the specific cutoff scores.

Eating attitudes Test-26 (EAT-26)
The EAT-26 is the most common self-report measure assessing symptoms and concerns typical of EDs [27]. The scale consists of 26 items and includes three basic dimensions: dieting (which reflects restricting intake of high caloric foods and preoccupation with body image/ shape), bulimia and food preoccupation (which describes thoughts regarding food, binging and self-induced vomiting) and oral control (which is about the ability to regulate food intake and perceived pressure from others to gain weight). Albeit it is not a diagnostic tool, it has been suggested that the EAT-26 might identify cases at risk for EDs clinical spectrum [37]. A score of 20 or higher on the EAT-26 indicates a high level of concern about dieting, body weight or problematic eating behaviors. The questionnaire has been validated and translated into Italian [37], Polish [38], and Spanish [39] and used extensively in clinical settings as well as in the general population.

Temperament and character inventory (TCI)
Developed by Cloninger [19], it is a self-report scale consisting of 240 questions with yes/no answers. It is based on the biopsychosocial personality theory developed by Cloninger [29] allows graded answers from 1 (absolutely false) to 5 (absolutely true) on a Likert scale. The questionnaire has been validated and translated into Italian [40], Polish and Spanish [28] and used extensively in clinical settings as well as in the general population. Due to availability reasons, the TCI was used in Italy and Poland, while the TCI-R was used in Spain. To overcome the possible problems entailed by the different scoring methods of the scales used, we adopted the cutoffs described in the literature [19]. Scores in each temperament and character dimension were thus classified accordingly as low (low/very low scores); medium (medium-low/medium/medium-high scores); high (high/very high scores).

Ethics
The study was approved by the local ethics committees (Comitato Etico Interaziendale, Novara, Italy, protocol UNI-ORTO, no. 1/2014; University of Silesia in Katowice Human Research Ethics Committee, no. 14/2015; Comité Ético de Investigación en Humanos de la Universitat de València, no. H1409824786250). All procedures performed in the study were in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki). Participation was anonymous and informed written consent was obtained from participants. No fee or reimbursement was offered for participation in the study.

Statistical analyses
To quantify categorical variables, frequency distribution tables were constructed for categorical variables and mean and standard deviation were calculated for continuous variables.
Univariable and multivariable logistic regression models were used to study the association between the study variables and the presence of ON (as measured by an ORTO-15 score below the cutoff ). Odds ratios and 95% confidence intervals (95% CI) were calculated. The significance of each individual variable was assessed using the likelihood ratio test (LRT). The Kruskal-Wallis test was used to evaluate the difference in continuous variables between the groups of students from Italy, Poland and Spain and for categorical variables the association with the nationality was evaluated by the chisquare test or Fisher test.
A p-value < 0.05 was considered statistically significant. Statistical analyses were performed with STATA v14.
Regarding questionnaire scores in the whole sample, the mean EAT-26 scores were 5.04 ± 6.06 for Dieting, 1.78 ± 2.93 for Oral Control, 0.99 ± 2.23 for Bulimia and Food Preoccupation; the total EAT-26 score was 7.89 ± 9.51. ORTO-15 and EAT-26 scores dichotomized according to scoring below or above the respective cutoff, and TCI scores classified in low, medium, high according to the mean value of each dimension (as described in the methods section) are reported in Table 1 for the whole sample.
Univariable and multivariable logistic regression: associations between ORTO-15 scores, the variables assessed and questionnaire scores (EAT-26, TCI) The results of the univariable and multivariable logistic regression analyses performed to assess the association of ON (as described by an ORTO 15 score below the cutoff; dependent variable in the multivariable analysis) with the self-report variables gathered from participants and questionnaire scores (EAT-26 and TCI) (independent variables in the multivariable analysis) are reported in Table 2 and Table 3, respectively.
Details about the descriptive data of the self-report variables and questionnaire scores with a statistically significant association with the ORTO-15 score are reported in Table 4.

Comparison among the Italian, Polish and Spanish samples
The results of the statistical analyses performed to compare the Italian, Polish and Spanish samples are reported in Table 5. While no statistically significant result was found from the chi-suqare comparison of those scoring higher/lower that the EAT-26 cutoff, the Kruskal-Wallis comparison of the EAT-26 total mean scores in the three countries yielded the following result: it was higher in the Italian than in the Polish and Spanish Sample (9.56 ± 12.28; 8.52 ± 8.18; 5.87 ± 7.11; p = 0.0001).

Descriptive analysis
The first aim of the present study was to investigate the prevalence of ON (as suggested by an ORTO-15 score below the cutoff ) in a sample composed by 664 university students enrolled in Italy (N = 216), Poland (N = 206) and Spain (N = 242).
The ORTO-15 score suggested ON in more than a third of the whole sample of students (score < cutoff in 37.05%). The quite high value we found is consistent with reports from previous studies with the same assessment tool, which reported a prevalence of ON up to more than 50% in university students from different countries [41][42][43]. More specifically, some at-risk groups have been described, including students attending university courses focused on nutrition (Dietetics) [44,45] and body care (Exercise and Sport Sciences) [45]), as well as in medical students and residents [32,46].
Anyway, the interpretation of the result about ON prevalence is currently difficult, because ON is not a clinical diagnosis and the ORTO-15 is far from being a diagnostic tool. Moreover, the possible implications of ON for the general population still need to be better understood (for instance, eating healthy may be desirable until it does not become clinically impairing).
Overall, although more studies on ON are required, in our sample the prevalence of ON, self-reported current/ previous EDs (6.48 and 3.01%, respectively), and the percentage of students scoring above the EAT-26 cutoff (suggestive of high-risk for EDs, 8.43%), suggest that attention should be paid to eating-related problems in university students, and that psychoeducational interventions targeting this topic may be warranted also for this population [47].

Associations between orthorexia nervosa, the variables assessed and questionnaire scores (EAT-26, TCI)
The second objective was to investigate the association between ON, socio-demographic features, eating patterns (measured with the EAT-26) and personality traits (assessed with the TCI) within the whole sample.
The univariable analyses shown an increased risk of ON in female students, with higher BMI values, self- reporting themselves as dieting and having a current ED, and in those scoring above the EAT-26 cutoff. The multivariables analyses confirmed an increased risk of ON in females and in students self-reporting themselves as dieting. An increased likelihood of ON was found also in students with low or medium Persistence (as measured by the TCI).
Regarding gender, while it is widely acknowledged that the prevalence of EDs like AN has a male to female ratio 1:9 [48], gender differences in ON are still a matter of debate; results are inconsistent across studies [9,49], and some did not even show any significant gender difference [25,32,42].
Literature findings are inconsistent also about the relationship between ON and BMI. In the current research, the likelihood of ON increased with higher BMI values. Some studies have not found any significant relationship, while others described an increasing trend of ON with increasing BMI values [32]; no study has found yet a negative correlation between BMI and orthorexia. Nonetheless, the low OR does not allow to draw clear conclusions about the clinical relevance of this association.
Students who declared to be dieting were more likely to display ON than those who were not following a diet. ON has already been described as being more frequent in individuals who strictly adhere to dieting habits, but once again results are inconsistent and another study [50] failed to identify any correlation between being dieting and ON.
Last, there was a higher risk of ON in students who self-reported having a current ED at the time of assessment, but not in those with a self-reported history of EDs. Furthermore, in the current study we also found a positive correlation between ON and high risk of EDs as suggested by an EAT-26 score above the cutoff.  The relationship between ON and EDs is complex and still far from being thoroughly understood. Afive-fold increase in the risk of having ON was identified in individuals with a disordered eating behavior compared to those with normal eating attitudes [21], and a frequent comorbidity, either current or lifetime (from 28 to 53% in a 3year follow-up study), has been described between ON and EDs [35,36]. Orthorexia can precede the onset of an ED, or it can represent its evolution in the phase of remission and recovery, representing a condition with the "advantage" of making the invidual with an ED feels once again accepted and part of society; this is one reason why ON can be described as "a disease masked by virtue" [3].
Moreover, EDs have been identified as risk factors for ON by some authors [34], while others have found EDs and weight concern as negative predictors of ON [36,51]. Also, we cannot exclude that in some cases inconsistent results in the literature could be due to the use of different instruments, which may likely measure different facets of orthorexia [52].
Despite the intrinsic limitations of self-report information about previous/current ED, and those retrieved from the EAT-26 scoring, overall, from our results described above and from the inconsistency of those available in the existing literature, it emerges the importance of fully defining the construct and implications of ON, and its possible relation with EDs, in order to better understand how to approach it in a more scientifically sound way.
An interesting result was found regarding the correlation between ON and personality features as assessed by the TCI. ON was more likely in students with low and medium scores on the P scale of the TCI than in those with high P scores. Persistence reflects the capacity for perseveration, determination and constancy in spite of frustration and fatigue, and predicts resistance to the extinction of behavior in the face of intermittent reinforcements [19]; it is usually characterized by high scores in typical descriptions of AN, restricter sybtype [53]. To date, only one study has investigated personality traits in ON with the TCI [24], finding a profile characterized by high HA, low SD and high ST, while no significant result emerged regarding P. The "orthorexic personality" has been previously described as characterized by an excessive preoccupation that shows itself with shyness in social situations, combined with the desire to be perfect and to feel accepted. In this context, "pure" dieting would result from an intense need for control to compensate for low self-esteem, feelings of ineffectiveness and inability to manage harmful events [18]. The low P we found in ON subjects may reflect a poor ability in adopting appropriate coping strategies to finalize action in everyday life. While this result, again, is difficult to contextualize given the shortage of available studies in the literature, further research about the personality features of ON individuals are warranted to identify possible targets for psychoeducational and Table 4 Descriptive data of the statistically significant associations between ORTO-15 score (> or < cutoff), self-report variables and questionnaire scores in the whole sample ORTO also therapeutic interventions (such as perfectionism, need for control, etc.).

Comparison among the Italian, Polish and Spanish samples
The third aim of our study was to highlight any crosscultural difference in eating attitudes (as measured by the EAT-26) and ON in the three sub-groups of students (notwithstanding the limitations of the comparability of the different university populations recruited in the three countries due to the convenience sampling procedure). Differences among the three subgroups were found for all the self-reported information we gathered. While less Polish students reported practicing physical/sport activity, they were more frequently dieting and using food supplements (e.g. vitamins), compared to their peers in Italy and Spain.
A few studies are available about eating habits in Polish university students [54,55]; with more detail, a recent research found almost a third of the studied population following restrictions in the type and amount of food intake [56], consistent with our findings on dieting.
To contextualize these results (dieting, use of food supplements), it should be considered that in Poland, in recent decades, economic and political changes have had a significant impact on the lifestyle of various social groups, especially on young people, including university students [56].
Regarding self-reported current and past ED, the only students declaring a current ED belonged to the Polish sample, and the higher rate of students with a self-reported history of ED was found in the Polish sample, as well, even though no actual difference in the BMI of students belonging to the three subgroups emerged. While all this information is intrinsically limited by its self-report nature, we cannot exclude that a different perception of EDs may underlie these findings. The opposite trend was found for sport activity practiced by students in the three subgroups, which was less frequent in the Polish sample (only in the Polish sample the percentage of students who did not practice sports was more than 50%). Nonetheless, another study, involving a different population of university students (from medical/scientific studies rather than humanities faculties as in the current one) showed that the majority of Polish students rated their activity level as medium (68.5%) and high (25.3%) and that 80.5% of female students and 74.3% of men practiced active sports every day [54].
Despite inconsistencies and gaps in the available literature, and the limitations of our results, it will certainly be interesting to better understand whether cultural background differences among the three countries exist in terms of what students consider a healthy lifestyle and in how they believe it can be achieved.
To our knowledge, in the last years in Poland good consumer practices have gained greater popularity (for example in the control of the composition of food products) and healthy eating has become the most important among the strategies aimed at improving one's state of health [57].
On the other hand, Italy has a widely recognized culture about eating and nutrition and the Mediterranean diet has long been praised for its general health benefits and a strong attention paid to taste [58]; a similar situation is found in the Northeast (Basque Country) region of Spain.
Good consumer practices, control of the composition of food products and the belief that healthy dieting is the most important approach for the improvement of ones' health could explain the high percentage of ON found in the Polish sample. On the other hand, it is likely that the lowest frequency of ON and attitudes found in the Spanish sample depends on an enogastronomic culture that includes the Mediterranean diet style, similar to the Italian one, but with an even stronger convivial and social value of the meal.
Moreover, since the importance attributed to conviviality and to the eating-related social dimension could penalize orthorexic individauls, we cannot exclude that our results depend on under-reporting on behalf of students. A clearer definition of ON and its diagnostic criteria, and further studies using clinical interview and assessment rather than self-report information will allow a more comprehensive approach to the intriguing topic of the impact of socio-cultural differences on the ON construct.

Limitations
Some limitations of the current study should be underscored. As already stated, an assessment based on self-administered questionnaires entails problems in terms of realibility and possible underestimation/overestimation. The ORTO-15 Questionnaire does not allow to make a diagnosis of ON, and has intrinsic limitations already described by the scientific literature [59], with contradictory results regarding its psychometric properties, including construct validity [9,[60][61][62][63].
The possible problems due to the convenience sampling procedure and to different cut-offs and scoring methods of the scales used in the three countries have been overcome as described above in the methods section.
On the other hand, a strength of the study is that it represents the first one in the scientific literature specifically focused on the comparison of three countries as far as ON and eating patterns are concerned.

Conclusions
The positive association between ON, a self-reported current ED, presence of food psychopathology as suggested by the EAT-26 score, BMI and adherence to a dieting that emerged in the present study needs to be supported by further research in order to better understand the relationship between ON and ED. The current knowledge suggests that ON tendencies are observed in those who expect benefits for their health, society and the environment, deriving from food, but may also be found in individuals who may rather try to disguise disordered eating attitudes with a more socially acceptable eating habit [8,19,64].
Regarding the differences in the frequency of ON found among the Italian, Polish and Spanish samples, we cannot exclude that a role is played by gastronomic culture, Mediterranean diet, convivial and social value attributed to eating and main approaches aimed at improving one's health typical of each socio-cultural context.
Further studies with consistent and sound methodological approaches will help to shed light on the several gaps that still exist in this field of research, including that about whether ON is mediated by socio-cultural factors and to what extent.

Funding
The author(s) received no specific funding for this work.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Ethics approval and consent to participate Approval for the research was obtained by the local ethics committees (Comitato Etico Interaziendale, Novara, Italy, protocol UNI-ORTO, no. 1/2014; University of Silesia in Katowice Human Research Ethics Committee, no. 14/ 2015; Spain Comité Ético de Investigación en Humanos de la Universitat de València, no. H1409824786250). All procedures performed in the study were in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki). Participation was anonymous and informed written consent was obtained from participants. No fee or reimbursement was offered for participation in the study.