Introduction

The alarmingly high rates of obesity, mainly observed in the Western world, have motivated more and more people to follow a healthy diet. In the last few years though, many specialists in eating disorders (EDs) have begun to discuss the construct of orthorexia nervosa (ON) in an effort to describe the phenomenon of obsessive and excessive adherence to consuming healthy food [1].

The term orthorexia was first used by Bratman [1]. According to him, a healthy diet evolves into ON when a person crosses the “borderline” and his/her concern about food begins to negatively affect essential aspects of life. It is not easy to define the point of transition, but it can be described as a state in which the pursuit of a healthy diet dominates the individual’s life and ceases to serve the purpose of improving health [2].

ON bears many similarities to bulimia (BN) and anorexia nervosa (AN). In fact, the weight loss that is often observed in ON may make it difficult to differentiate it from AN [3]. In both cases there is a clear orientation towards achieving the nutritional goal, accompanied by stress and individuals consider their adherence to diet as an indication of self-discipline, while interpreting any derogation as a failure of self-control [4]. Furthermore, similar psychological and socio-cultural factors seem to have an important role in the genesis, but also in the development of ON and EDs [5].

Several studies have also revealed that increased symptomatology of ON is associated with features of obsessive–compulsive disorder (OCD) [4, 6,7,8,9,10]. Ritualistic behaviors are common in OCD and EDs, accompanied by high levels of anxiety when disturbed and the expected effect is not achieved [11]. This feature is also present in ON, where individuals try to reduce their anxiety through the sense of superiority induced by consuming strictly healthy food [12].

The dysregulation of a person’s everyday life caused by the excessive concern about healthy eating, along with the growing number of people expressing this kind of behavior, have given rise to research and clinical interest in ON. Although there is an increasing number of studies concerning ON, it has not yet been included in the classification of mental disorders and therefore, there is neither an official definition of this phenomenon, nor specific diagnostic criteria. It is still under discussion whether ON is a distinct disorder or it should be incorporated in the field of eating disorders (EDs) or obsessive compulsive disorders (OCD). Futhermore, it is debatable if ON can be considered a disturbed eating habit [13] or, even simpler, a behavioral trait [14] as well as the extend of sociocultural influences 15].

Based on the necessity to reach a consensus on the definition and diagnostic criteria of ON and to validate new assessment instruments [16], the Orthorexia Nervosa Task Force (ON-TF) was established in 2016, including researchers actively involved with ON. After taking into consideration the proposals presented in several published articles, ON-TF summarized the following main diagnostic criteria for ON: (1) a pathological preoccupation with healthy nutrition, (2) emotional consequences (such as distress on anxiety) at any transgression from self-imposed dietary rules, and (3) psychosocial impairments in essential areas of life, weight loss and malnutrition [17].

As far as the diagnostic tools used for the assessment of ON, the most widely used—to date—is ORTO-15 [18] and its versions [4, 19]. Other questionnaires developed for the same purpose are the Eating Habits Questionnaire-EHQ [9], the Dusseldorf Orthorexia Scale-DOS [20] and the Teruel Orthorexia Scale-TOS [21].

The aim of this study was the adaptation and validation of the ORTO-15 questionnaire for the Greek language, and specifically, the exploration of its reliability, construct and converging validity; additionally, the exploration of the relation between orthorectic and disordered eating behaviors, so as the possible relation of age, gender, financial and marital status with ON symptomatology, in a sample of Greek university students. University students were chosen as the study group mainly for three reasons. The first reason is that students are at a stage where as emerging adults they are establishing for the first time in their life eating habits and preferences beyond their family legacy [22]. Most of these habits will probably remain unchanged through most of their adult life. The second reason was that although university students have been found to possess an increased awareness of the benefits of healthy eating and its effect on disease prevention, at the same time they suffer from an increased vulnerability to anxiety and EDs [23]. The third reason concerns the clinical considerations that arise from the previous mentioned reasons and the application of university driven prevention initiatives related to ON and EDs among students and young adults in general. ORTO-15 was used in its full form (including all 15 items) for the first research approach of ON in Greece as it is still the most widely used instrument for the study of ON despite the discussion and criticism in the literature conceiving its psychometric properties.

Methods

Design

The study protocol was approved by the Ethics Committee of Eginition University Hospital (registration number 449/18-6-2018). After acquiring permission from the authors of the ORTO-15, the original questionnaire was translated into Greek, using international guidelines and the internationally accepted translation method [24, 25]. The standard linguistic validation process started with the conceptual analysis of the original instrument’s items. Upon agreement, two professional translators, native speakers of the Greek language and fluent in the English language, undertook independent forward translations into the target language. Then, a backward translation of the reconciled version was performed by a professional translator; the necessary changes were made and the final Greek version of ORTO-15 was produced. The items and the categories of the Greek translation of ORTO were in the same grouping and order as in the original English version. Subsequently, the questionnaire’s test–retest reliability was checked and the statistical correlation between the two scores was explored.

Procedure

The present study is of a cross-sectional descriptive design. The Greek version of the ORTO-15 was firstly administered to 20 students and re-administered 2 weeks later, for the purposes of the test–retest procedure. The 2-week interval was chosen in order to balance the necessity to re-measure ON in a period of time where the symptomatology would not have changed considerably wlile at the same time allowing enough time to pass from the original testing so that the participants would not remember the answers that they had given. It should be noted that the 2-week interval is the most frequently recommended interval in the literature for the test–retest procedure [26]. Afterwards, for the purpose of the study 148 students of the Psychology Department of Panteion University of Social and Political Sciences were approached after lectures, out of which 120 (81%) agreed to participate in the study. After being informed of the main purposes of the study, participants signed a consent form and privacy statement, prior to the administration of the questionnaires. The sample was selected irrespective of sex, origin, marital and socioeconomic status. Age (18–35 years) and sufficient knowledge of the Greek language were set as inclusion criteria. Two participants were excluded as they were much older than 35 years. The age criterion was included as ON and ED symptomatology is often related to age groups and the study was focused mainly on young adults and their eating behaviors. It should be noted that the students who participated in the test–retest procedure were not included in the in the group of the 120 participants.

Measurements

ORTO-15: ORTO-15 is a self-reported questionnaire developed by Donini et al. [18] in order to determine perceptions related to: (a) benefits derived from eating healthy food; (b) attitudes guiding food selection; (c) habits related to food consumption and (d) the point to which food concern affects one’s daily routine.

The authors of the ORTO-15 assume that there are three underlying factors related to eating behavior: cognitive–rational (explored by questions 1, 5, 6, 11, 12, 14), clinical (questions 3, 7, 8, 9, 15) and emotional (questions 2, 4, 10, 13) [27]. Orthorexia nervosa is assessed in a four-point Likert scale (1 = always, 2 = often, 3 = sometimes, 4 = never). Lower score indicates higher ON symptomatology. For the original version of the questionnaire, the authors suggest a cut-off point of 40 for the diagnosis of ON [18].

Eating Attitudes Test-26 (EAT-26): EAT-26 is a self-reported questionnaire for the assessment of disordered eating attitudes [28]. It is probably the most widely used questionnaire for screening ED symptomatology. The EAT-26 items form 3 subscales: Diet, Bulimia/Food preoccupation and Oral control. The questionnaire consists of 26 questions with six possible answers from always to never. Three answers get a score of 1, 2 or 3 and the rest a score of 0. The EAT-26 total score ranges from 0 to 78. In the Greek version of the questionnaire, a score of equal to or more than 20 is considered an indicator for the diagnosis of EDs. EAT-26 has been translated into Greek and validated by Simos (doctoral thesis, 1996).

EAT 26 was used for two main reasons. The first reason was the lack of any other instrument, translated and validated in the Greek language, for the measurement of ON that could be used to compare its results with ORTO-15. The second reason was that, since this was the first attempt to adapt and use ORTO-15 in the Greek language, the methodology of other studies on validation and adaptation of ORTO-15 in various cultural contexts was replicated [6, 15, 29, 30].

Sociodemographic and clinical data: data concerning sociodemographic characteristics, such as age, gender, financial and marital status, educational level, body mass index (BMI) and medical history of EDs (whether they have tried to lose weight in the past, have visited a nutritionist or an ED expert, were satisfied with their weight and body image, were diagnosed with an ED) were collected through a self-constructed questionnaire.

Statistical analysis

The SPSS Software version 22.0 was used to perform statistical analysis. The results of the continuous variables are displayed as mean and standard deviation (SD), while categorical variables are displayed as percentages and frequencies.

For the comparison of the continuous variables between two groups, we used the Student’s t test. To assess the structure of the ORTO-15 a factor analysis (rotation method: Varimax) was applied. We chose Varimax rotation rather than oblimin, because the correlation of the items was not high. The questionnaire’s construct validity was assessed by means of Pearson correlation coefficients between factors, as obtained from the factor analysis.

A confirmatory factor analysis (CFA) using maximum likelihood procedure was conducted in order to examine how well the ORTO model fitted the data. The fit of the CFA model was assessed using the comparative fit index (CFI), the goodness of fit index (GFI) and the root mean square error of approximation (RMSEA) [31]. For the CFI and GFI indices, values close to or greater than 0.95 are taken to reflect a good fit to the data [32]. RMSEA values of less than 0.05 indicate a good fit and values as high as 0.08 indicate a reasonable fit [32].

For the evaluation of the questionnaire’s internal consistency, we calculated Cronbach’s alpha coefficient. Furthermore, correlation coefficients within groups (ICCs) were used to explore the answers’ agreement during the test–retest procedure. The agreement is considered to be low when the coefficient is up to 0.4, medium when it ranges from 0.41 to 0.6, high when it scores from 0.61 to 0.80 and very high when it is greater than 0.8.

Pearson correlations were used for ORTO dimensions, because they were normally distributed. For the correlation of ORTO with EAT-26 dimensions, we used Spearman’s coefficients, since EAT-26 dimensions were not approximately normally distributed. The significance level for all the cases was set at p < 0.05. Finally, three cut-off points of 40, 35 and 30 were tested following the methodology of other studies [10, 15, 18, 19, 27, 30, 33].

Considering the power of the study, a posterior analysis showed that 120 participants were an adequate sample. According to Bruce Thompson’s Exploratory and Confirmatory Analysis [34], if factors are defined by four or more measured variables with structure coefficients less than 0.6, then the sample size of 60 is adequate. If factors are defined with 10 or more structure coefficients, each with loading around 0.4, then the sample size should be around 150. Similarly, Comrey and Lee state that 5–10 observations per scale item are necessary, in order to have a powered factor analysis result [35].

Results

Sociodemographic characteristics of the participants

The mean age of the participants was 21.6 years (SD = 2.1) while the mean BMI was calculated at 22.2 (SD = 3.0). 97.5% of the participants were single, 53.3% lived with their parents and 55.5% considered their financial status as moderate. Additionally, 78.3% of the participants reported that during the last year they had tried to lose weight, 28.3% had visited a nutritionist and 5.8% a mental health EDs expert. Finally, 67.5% reported that they were satisfied with their weight and 70% with their body image.

Test–retest reliability

The test–retest procedure was performed on 20 students prior to the beginning of the study, with an interval of two weeks between the first and the second measurement. The data obtained from the first measurement of these 20 students were not included in the final database. There was significant agreement in all the questions. ICC was calculated for each ORTO-15 item and ranged from 0.62 to 0.87. The ICC of ORTO-15 total score was 0.75 (p < 0.001).

Validation of the Greek ORTO-15

Mean and item total correlations of ORTO-15 items (Table 1) were more than 0.3, indicating that the correlation between each item and the remaining was not weak. A four-factor grouping of the ORTO-15 items produced a group with only one question, so a three-factor analysis was pursued (Fig. 1). A CFA was conducted to estimate if the model fitted the data well. The CFA indicated an adequate fit of the three-factor model (RMSEA = 0.077, CFI = 0.951 and GFI = 0.942). Based on the factor analysis, the items of the questionnaire were grouped in 3 factors: (a) behavioral, (b) emotional, (c) cognitive/rational (Table 2). The behavioral factor described attitudes associated with food selection (questions 2, 3, 7, 13). The emotional factor was related to worries and emotions concerning healthy nutrition (questions 4, 9, 10, 11, 12, 14, 15). The cognitive/rational factor was associated with cognitions about nutrition (questions 1, 5, 6, 8) (Table 2). The 3 factors model explained 46.3% of the total variance (Emotional: 18.0%, Behavioral: 15.5%, Rational: 12.8%). The factor loads ranged between 0.40 and 0.74.

Fig. 1
figure 1

Factor analysis of ORTO-15 with Varimax rotaion

Full size image
Table 1 Mean and item total correlations for ORTO-15 items
Full size table
Table 2 Factor analysis of ORTO-15
Full size table

The results from the application of the Bartlett’s sphericity test (p < 0.001) and the Kaiser–Meyer–Olkin index (KMO = 0.8) indicated that data were suitable for exploratory factor analysis. Cronbach’s alpha coefficient was 0.7, so the questionnaire demonstrates acceptable internal consistency. The Cronbach’s a coefficient for each factor separately was: Emotional 0.74, Behavioral 0.71 and Rational 0.70. The construct validity assessment by means of Pearson correlation coefficients among factors showed that there was significant positive correlation between almost all the factors of the ORTO-15 scale. The only exemption was that the Behavioral ORTO-15 factor did not correlate with the Rational ORTO-15 factor (Table 3).

Table 3 Pearson correlation among the 3 factors of ORTO-15
Full size table

We considered 3 different cut-off points (< 40, < 35, < 30) for the Greek version of the ORTO-15, which gave us a result of 55, 29.1 and 2.5% positive cases, respectively.

Relation between orthorexia and eating disorder symptomatology

In order to investigate if orthorexic tendency could be related to pathological eating attitude, we examined the relationship between the ORTO-15 scores and EAT-26 scores (Table 4). There was a significant negative correlation between the Emotional and Behavioral factor of ORTO-15 and two of the three dimensions of EAT-26 (Dieting and Bulimia/Food preoccupation), confirming the convergent validity. Therefore, more disturbed attitudes regarding dieting and bulimia were associated with more severe orthorexic symptoms, especially behaviors and emotions related to “healthy” eating.

Table 4 Spearman correlation between ORTO-15 and EAT-26
Full size table

Moreover, those participants who reported difficulties related to their weight regulation, body image and mental health scored lower in the ORTO-15 (Table 5). Finally, no significant relation was found between gender (p = 0.61), age (p = 0.93), marital (p = 0.25), financial (p = 0.77) and residential status (p = 0.41), or reported BMI (p = 0.98) and ON symptomatology.

Table 5 ORTO-15 and clinical data
Full size table

Discussion

The study’s findings suggest an acceptable validity of the Greek version of the ORTO-15 questionnaire. One puzzling problem that arose during the study was the use of an appropriate cut-off point. As reported in the literature, this is one of the main concerns with the application of ORTO-15 and its versions [15, 27, 36]. Since no consensus has emerged yet, it was decided to follow the same methodology with other studies [10, 15, 18, 19, 27, 30, 33] and applied the cut-off points of 40 and 35, which produced a result of 55 and 29.1%, respectively. A threshold of 30, produced a more sensible result of 2.5% positive cases, so it was hypothesized that a cut-off point of 30 might be more appropriate for the Greek population.

Another issue that had to been taken under consideration was the cultural and linguistic adaptation of the questionnaire in the Greek language. ORTO-15 has been adapted and validated in several languages [6, 27, 37,38,39]. In most studies some of the items of the original tool have been removed, in an effort to improve its internal validity and reliability [5, 6, 38], taking also into account the socio-cultural features of each country regarding food and eating [38]. The results regarding internal validity among these versions of ORTO-15 vary, with the Cronbach a ranging from 0.30 [37] to 0.82 [38]. In our study, we found a Cronbach a of 0,7 which indicates an acceptable internal consistency of the questionnaire. In the Greek version of ORTO-15 all items of the original questionnaire were included. It should be noted though that, during the initial test–retest reliability procedure, it became apparent that the translation of question 15 (“at present, are you alone when having meals?”) did not make clear to that participants that they were asked whether they ate alone by choice. As similarly reported by Haddad et al. [30], in the Greek culture common meals with family and friends are part of everyday life and that is probably the reason why there was a confusion to the participants whether the question was referring to their choice of eating alone or to the actual fact of being obliged to eat alone due to the long hours that a student has to spend outside his/her home. So, item 15 was reformed to: “At present, do you choose to be alone when having meals?” After rephrasing the item, it had a satisfying loading, so we kept it in the analysis.

As mentioned in the results, the analysis produced a three-factor solution. Interestingly, various linguistic adaptations of ORTO-15 have reported a variety of results concerning ORTO-15 factors. A three-factor solution similar to the one produced in this study has been reported by the Spanish [27] and Lebanese [30] versions of ORTO-15. Contrary to the above results, the Turkish [6], Hungarian [38] and German [37] versions of ORTO-15 reported a one-factor solution and the Polish version a two-factor solution [39].

Concerning the second aim of the study, that is the investigation of a possible relationship between ON and EDs, the correlation between ORTO-15 and EAT-26 showed that dieting and bulimia were related with more increased ON symptoms, so disordered eating attitudes seem to have a significant correlation with orthorexia. Furthermore, the study results suggest that students who scored lower in ORTO-15 reported more often that they were dieting, were not satisfied with the body shape and weight and that they were diagnosed with a mental disorder. This finding is in agreement with previous findings in the literature, where participants with higher scores in the EAT-26 or EAT-40 appeared to score lower in the ORTO-15 questionnaire [6, 15, 29, 30]. Regarding BMI, no significant relation was found with ON symptomatology. The results of other studies are mixed as far as this parameter is concerned. Some studies have found no positive relationship between BMI and ON [5, 6] while others report opposite findings [40, 41].

Although this study provides useful information about ON-related behaviors in the Greek population, it is not exempt from limitations. The most important limitation concerns ORTO-15 drawbacks. ORTO-15 has been the target of substantial concern in the literature regarding its unstable factorial structure and the wide range of prevalence rates, which vary from 6.9% [18] to 75.2% [42]. These highly discrepant rates are mostly confusing rather than helpful, because it is rather difficult to understand how a phenomenon of restrictive eating behavior appears to have such a variable nature. Dunn et al. find it more likely these rates probably reflect the fact that ORTO-15 incorrectly identifies conscientious dieting as being harmful, while at the same time it fails to confirm any association with psychopathology. The authors estimate the real prevalence of ON to be less than 1% [43].

A second limitation was that data collection was based on self-administered questionnaires. Furthermore, the population consisted of university students residing in Athens, the capital of Greece, so the results cannot be generalized to the entire population. It is noticeable that most of the studies in ON have been conducted among university students. Parra Fernandez et al. [5] mention that, from a nutritional perspective, this population is particularly vulnerable because, as individuals, they are just forming and establishing their eating habits and perceptions of their body image, so they are at greater risk of developing an ED.

A third limitation of the study concerning the comparison of the Greek version of ORTO-15 with the original questionnaire and its derivatives was the rephrasing of item 15. Although the rephrasing was deemed necessary to include item 15 in the final Greek version of the questionnaire altering the translation of item 15 might prove an obstacle in a future comparison among the various linguistic and cultural adaptations of ORTO-15.

Finally, a fourth limitation of the study was the relatively small sample size of the test–retest procedure. From the 40 university students that were originally approached unfortunately only 20 of them agreed to provide us with their personal contact data, so that we could reach them after 2 weeks for the retest procedure. Although according to the literature [44], a sample size of 10–40 subjects can be considered as adequate, a group of 20 participants is still relatively small for the test–retest reliability measurement.

Concluding, the measurement of ON in a Greek student population produced similar to other European countries results. Although ORTO-15 is far from being the perfect measurement of ON, it is a widely used questionnaire and its introduction in Greece can pave the way for more research in the area of orthorexia. Furthermore, two recently published studies introduced two novel instruments for the study of ON. The first study, conducted by Rogoza and Donini [45] researched a revised edition (ORTO-R) of the original ORTO-15, aiming mainly at the assessment of orthorectic thoughts and behaviors. Containing six items out of the 15, ORTO-R showed an acceptable reliability. The second study, conducted by Oberle, De Nadai and Madrid [46] introduced the Orthorexia Nervosa Inventory (ONI), a 24-item scale that includes items assessing physical impairment due to nutritional deficiencies. ONI demonstrated quite good internal consistency and test–retest reliability. Although these new measurement seem to be quite promising for the assessment of orthorectic symptomatology, more cross-national studies are needed to establish transcultural reliable measurements for ON.

What is already known on this subject?

ORTO-15 is—to date—the mostly used tool for the measurement of orthorexia. It has been translated from English and used in many countries. In most of the studies, though, several items have been removed, in an effort to improve its psychometric flaws. However, no commonly accepted version of this instrument has yet emerged.

“What does this study add?”

This study is the first attempt to approach orthorexia in Greece. The Greek version of ORTO-15 was found to be a reliable questionnaire for the measurement of orthorexic symptomatology, thus providing researchers a translated, adapted and validated tool to study ON in Greece, and probably Cyprus, two Greek speaking countries in North-eastern of Europe. The study also confirmed findings from other studies that disturbed eating attitudes, and especially dieting and bulimia, are related to orthorexia.