In this form questions related to the general characteristics of the participants such as age, gender, height, body weight and and Body Mass Index (BMI) were included.

The Eating Habits Questionnaire (EHQ) was developed to measure healthy eating as a pathological condition labeled “Orthorexia Nervosa” and to determine emotions associated with cognition, behavior, and obsession with healthy eating.

It was developed as a self-report inventory consisting of 21 items for evaluation. The subscales of the EHQ are as follows; 1) problems associated with healthy eating, 2) knowledge of healthy eating, and 3) feeling positively about healthy eating [10]. Each item in the questionnaire was rated on a 4-point Likert scale (False, Not at all true, Slightly tr ue, Mainly True, Very true), where higher scores are indicative of ON tendencies. The answers to each question were given as 1, 2, 3 or 4 points respectively. The total subscale score was obtained by summing the scores per item. While questions 2, 4, 6, 7, 8, 10, 13, 14, 16, 17, 18 and 20 are related to problems; questions 1,3, 5, 11 and 21 are related to knowledge. In addition, questions 9, 12, 15, and 19 are about subscales measuring feelings.

The EAT-26 is a test that allows individuals to evaluate their eating behavior and attitudes. In our study, we used the Turkish version of the EAT-26 for convergent validity. The scale consists of 22 questions. The EAT-26 is a sensitive and specific scale for identifying individuals at high risk for eating disorders. Increasing scores indicate a deterioration in eating attitudes [12].

In the process of adapting the EHQ into Turkish, the authors got in contact with Dr. Gleaves (Professor of Psychology (Clinical) School of Psychology, Social Work and Social Policy University of South Australia) via e-mail, who is one of the original scale owners. The necessary approval and details about the scale were obtained from him. During the translation of the scale into Turkish, standard procedures that are recommended in the literature were applied [13,14].

The scale was translated into Turkish separately by three English language and literature expert translators. Barriers in linguistic comprehension, contextualized meaning and possible interpretations of the translated instrument were taken into account in the adaptation process. The expressions in the Turkish questionnaire were compared and reviewed, and the most appropriate expressions were selected in terms of intelligibility. The Turkish scale was back-translated by an English language and literature expert. The researchers compared the original scale items with the form translated from Turkish into English. No further modifications were necessary.

A pilot study, was carried out with 15 students and they were asked to evaluate the scale in terms of language and expression intelligibility. As they stated that the scale was easy and understandable, no changes were applied in to Turkish form, and the researchers subsequently applied the scale to the main sample.

In the study, the obtained data set was analyzed using the IBM®SPSS® (version 25) package program for Explanatory Factor Analysis (EFA) (to determine the item–factor relationship). Within the scope of reliability analysis, Pearson correlation analysis for test-retest reliability, t-test and Mann-Whitney U test for dependent groups, the Cronbach’s alpha coefficient for internal consistency analysis, and Pearson correlation analysis for item-total score analysis were envisaged. The Bartlett test of sphericity was used to statistically test the correlation between the variables in the data matrix. The Kaiser-Meyer-Olkin (KMO) criterion, which is obtained by using the correlation and partial correlation coefficients, is also an important criterion in evaluating the suitability of the data for factor analysis. KMO, which is a sample adequacy criterion, takes a value between 0-1. Confirmatory Factor Analysis (CFA) was also applied to test the suitability of the factors obtained by exploratory factor analysis to hypothetical or theoretical factor structures. IBM®SPSS® (version 25) Statistics for Windows and Amos (Version 24.0) package program were used for Confirmatory Factor Analysis in the study (to determine whether the items and sub-dimensions explain the originalstructure of the scale). Descriptive data were expressed in means, standard deviations, medians and frequencies. The p-value was <0.05 was considered significant.

The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethics Committee of the Ethics Committee of the University of Health Sciences (decision no: 21/505).

In the study, whether there is a correlation between the items, which are the prerequisites of factor analysis, was tested with the Bartlett sphericity test; the adequacy of the sample size was examined by the Kaiser–Meyer–Olkin (KMO) measure. As a result of the EFA, KMO coefficient was found to be 0.82. Bartlett’s test χ² value was 1253.058 (p=0.001). Regarding reliability, the Cronbach’s α coefficient of the whole scale was 0.827.

Factor variance and factor loadings of the EHQ-T were given in Table 1. The factor loadings of the questions in the problems, knowledge, and emotions ranged between 0.470-0.737, 0.499-0.833, and 0.500, 0.837, respectively. Croncbach’s Alpha (α) was considered sufficient as it was above 0.700. The Croncbach’s Alpha of the sub dimensions were 0.742, 0.794, and 0.784, respectively. It was determined that the items were gathered under three dimensions as stated in the original EHQ scale. Some items in the “problems” subscale in the original scale were loaded onto the “knowledge” subscale in the Turkish version, and some items in the “knowledge” subscale were loaded onto the “feelings” subscale.

As the 16th question on the scale had a correlation value of -0.008, it was excluded from the scale. The total correlation values of the remaining questions ranged from 0.379 to 0.722. Therefore, it did not require any reductions in the dimensions of the EHQ-TR (Table 2).

The lowest score of the scale was 20, and the highest score was 80. The mean score of the scale was found to be 44.03±8.21. The mean scores of the problems, knowledge and feelings sub-dimension were 12.90±4.01, 13.73±3.49 and 17.40±3.69, respectively. The relationships between the scale and its sub-dimensions were evaluated as statistically significantly positive.

On the basis of the CFA results, fit indices were determined as follows: (χ²=287.365 df=165), χ²/df =1.742, Root Mean Square Error of Approximation (RMSEA)=0.060, Goodness of Fit Index (GFI) = 0.884, Comparative Fit Index (CFI)=0.900, Incremental Fit Index (IFI )= 0.901, Standardized Root Mean Square Residual (SRMR)=0.080 and Tucker–Lewis Index (TLI)= 0.902.

For Confirmatory Factor Analysis, covariance connections were conducted as a result of the similarities between the 2nd and 4th questions in the knowledge dimension, and between the 3rd and 21st questions in the emotions dimension. The tested model was shown in Figure 1. Furthermore evaluation of effects between questions and sub-dimensions for the EHQ-TR model was given in Table 3.

Each of the sub-dimensions’ path coefficients on the 21 questions was found to be statistically significant (p<0.05). Accordingly, the knowledge sub-dimension consisted of questions 1, 2, 4, 5, 11 and 20; the problems sub-dimension consisted of questions 6, 7, 8, 10, 13, 14, 17 and 18, and the feelings sub-dimension consisted of questions 3, 9, 12, 15, 19 and 21. All sub-dimensions were highly statistically significant on the questions (p=0.001). Also, it was found that each of the coefficients on the EHQ three sub-dimensions was statistically significant (p<0.05).

The mean score of the EAT-26 scale was found to be 27.66 ± 12.82. As presented in Table 4, the convergent validity of the EHQ-TR was significantly correlated with EAT-26. The correlations between the three subscales of the EHQ (problems, knowledge and feelings) and EAT-26 were significantly positively related with a confidence level of p<0.05.

As a limitation of our study, it cannot be generalized to the population because it was conducted on a small group of university students. Second, the current study was conducted with Nutrition and Dietetics students, therefore, knowledge about healthy nutrition may indirectly affect their responses. A strength of the present study is that it is the frst study that has examined the psychometric properties of a Turkish version of the EHQ which measures ON.