Introduction

Orthorexia Nervosa (ON) is defined as an excessive preoccupation with healthy food, characterized by the adherence to strict dietary rules that might lead individuals to cut out some food or food categories of their diet, because they are considered not pure or bad for their health [1]. In a study in Belgium and the Netherlands, 66.7% of healthcare professionals reported to have observed this potential disorder in their own practice, and 68.5% expressed that the diagnosis of ON should deserve more attention [2]. To date, despite proposals of diagnostic criteria (i.e., obsessional preoccupation with eating “healthy food”, impairment of physical health, and impairment of social, academic or vocational functioning [3]), research has not yet allowed the categorization of ON as a separate mental disorder in classifications [4]. This is partly due to an overlap of certain ON features with Anorexia Nervosa (AN), but also with other eating disorders or obsessive–compulsive disorder [5, 6]. However, on a theoretical level, ON and AN can be distinguished from each other on several aspects. For example, individuals with ON are essentially focused on food quality and its impact on health, while, in AN, the obsession is exclusively orientated on food quantity and fear of obesity [7]. In addition, it has been suggested that ON could also serve as a healthier alternative or as a coping strategy for individuals suffering from AN, allowing them to eat more while maintaining control over food [8]. Regarding weight matters, it is unclear how far weight loss is intended in ON [9, 10] and there is a lack of a clear link between ON symptomatology and body mass index (BMI). Indeed, most studies found no significant relationship between ON and BMI, while others reported an increased ON symptomatology linked to greater BMI [11].

Besides weight loss, other negative effects of ON encompass different aspects, including nutritional deficiencies [9, 12], hypoestrogenism [13], depression [14], social isolation [1, 5] or stigmatization by others [15]. In community or student samples, positive correlations were found between ON symptomatology and depression symptoms, negative affects, or suicidal ideations [11, 16]. In addition, negative correlations were reported between ON and well-being, self-esteem, and ability to relax [17]. Moreover, children living with orthorexic parents can be exposed to harmful effects of malnutrition [18, 19].

ON prevalence in community or student samples was estimated between 6.9% [20] and 70% using ORTO-15 [9, 21], a widely used self-report questionnaire created to assess ON symptomatology [22]. Regarding gender, a large meta-analytic work compiling 67 publications showed that tendencies toward orthorexic behaviors were comparable between genders, but pathological healthy eating appeared to be more pronounced in women [23].

To date, the high prevalence rates of ON reported in some studies are thought to be due to psychometric limitations of ORTO-15. Indeed, ORTO-15 might also detect a construct other than pathological eating beliefs and behaviors [24], as it includes items seemingly unspecific to ON, e.g., “At present, are you alone when having meals?” [22]. It thus seems important to replicate those prevalence studies with more reliable measurement tools. Another instrument for assessing ON is the Eating Habits Questionnaire (EHQ) [16], but it has been criticized for not taking into account negative emotionality or other dimensions of ON [7, 25]. The EHQ was nonetheless used in this current study, because it has better psychometric properties than the ORTO-15 and seems, therefore, to be more appropriate to assess convergent validity. The Teruel Orthorexia Scale is a novel and promising tool as it attempts to distinguish ON from healthy orthorexia (interest in healthy eating) [25]. It remains to be validated in different languages.

One validated and reliable instrument that was recently published is the Düsseldorfer Orthorexia Skala (DOS) [17]. The DOS has been used in various samples in Germany, e.g., [8, 17, 26,27,28]. It was also translated in English [29], Chinese [30], and Spanish [31], showing good psychometric properties in original or translated versions.

The aim of this study was to adapt the DOS into French and examine its psychometric properties (internal consistency, factor structure, and convergent and divergent validities) using a university student sample. To assess the convergent and divergent validities of the translated DOS, participants also completed the EHQ and the Eating Attitudes Test evaluating eating disorders [32]. In addition, to address the links between ON and psychopathology, a correlational analysis was performed between ON symptomatology and markers of functioning and mental health status in students (i.e., results of the last semester, levels of anxiety, depression, and obsessional–compulsive symptoms).

Methods

Participants and procedure

The data were collected through an online survey that was distributed to students from different universities in France. The link was shared on social networks in groups specifically dedicated to students. To be eligible for participation, participants had to be university students and older than 18. Informed consent was obtained online before completing the survey and no compensation was offered to participate in the study. The study protocol was approved by the local ethics committee (Comité d'Ethique de la Recherche of Toulouse University).

This study’s participants were 3235 college students (10.32% men, 89.67% women), aged between 18 and 29 years (mean age = 21.13, SD = 2.23). Based on self-reported height and weight, calculated BMI ranged from 13.67 to 59.12 (mean = 22.04, SD = 4.06) kg/m2. Other characteristics of participants are shown in Table 1. The majority of participants had a normal weight (68.9%) and did not follow a particular diet (79.3%), while 9.6% considered themselves as vegetarian, 3.9% flexitarian (i.e., a person with a primarily vegetarian diet with the occasional inclusion of meat or fish), and 2.9% vegan. Regarding fields of study, 40.9% of participants were in human sciences, 9.9% in law, 9.7% in literature and foreign language, 8.9% in sciences or engineering, 6.3% in history, geography, or political sciences, 5.2% in economics, 4.5% in medical/paramedical courses, 2.9% in art or design, 2.9% in art history or archeology, 2.6% in educational sciences, 2.2% in communication and information, and 4% in other fields.

Table 1 Participants' demographics and dietary style characteristics (N = 3235)
Full size table

Measures

Sociodemographic information was obtained including age, gender, height, weight, field of study, current year of study (first year’s bachelor, second year’s bachelor, etc.), and result of the last semester of participant's major course. Personal dietary style (vegetarian, vegan, etc.) was also assessed.

The Düsseldorfer Orthorexia Skala (DOS)

The DOS is a self-report questionnaire measuring orthorexic behaviors using 10 items, each rated on a 4-point Likert scale (from 1 = “this does not apply to me” to 4 = “this applies to me”) [17]. Higher scores suggest the presence of more pronounced orthorexic behaviors. A score ≥ 30 is considered as a cut-off to indicate the presence of ON, and a score ≥ 25 and ≤ 29 suggests a risk of ON [17, 33]. The DOS has shown high internal consistency, with Cronbach’s ⍺ ranging from 0.8 to 0.88 [8, 17, 26,27,28,29,30,31]. In this current study, the DOS was translated from German to French using a back translation procedure [34], i.e., the original version was translated into French and then back into German by two different translators. Differences were then discussed in a research meeting to decide the best translation for each item. The final version of the French version of the DOS (F-DOS) is presented in Table 2. For reviewing purpose, the corresponding items of the English version (E-DOS) [29] are also presented in Table 2.

Table 2 Mean, SD, and item selectivity
Full size table

The Eating Habits Questionnaire (EHQ)

The EHQ assesses cognitions, behaviors, and feelings related to an extreme focus on healthy eating [16]. Participants filled in the 21-item version [11]. Each item is rated on a 4-point scale (from 1 = “false, not at all true” to 4 = “very true”). The EHQ is composed of 3 subscales: (1) problems associated with healthy eating (e.g., “I spend more than 3 h a day thinking about healthy food”); (2) knowledge of healthy eating (e.g., “I know more about healthy eating than do other people”); (3) feeling positively about healthy eating (e.g., “I feel in control when I eat healthily”). As no ON measurement tool existed in French at the time the current study began, the EHQ was translated by the authors for the purposes of the study and used as the gold standard for the assessment of convergent validity. Since the study was conducted, the EHQ has been validated in French [35]. The EHQ demonstrated good internal consistency with a Cronbach’s ⍺ = 0.90 in the English version [16] and an Ordinal ⍵ = 0.92 in the current study.

The Eating Attitudes Test-26 (EAT-26)

EAT-26 is a standard assessment tool to measure pathological characteristics associated with anorexic and bulimic eating behavior [32]. It is composed of 26 items and 3 subscales (Dieting, Bulimia and food preoccupation, and Oral control). An example item is “I avoid eating when I am hungry”. The French adaptation [36] was used. The internal consistency of the French version was demonstrated with a Cronbach’s ⍺ = 0.86 [36] and an Ordinal ⍵ = 0.94 in the present study.

Psychopathology measures

Anxiety and depression symptoms were assessed using the Generalized Anxiety Disorder-7 and the Patient Health Questionnaire-9, respectively [37, 38]. Levels of obsessive–compulsive symptoms were evaluated with the Obsessive Compulsive Inventory-Revised [39]. For these 3 scales, high scores suggest more pronounced symptoms. For the present study, the French versions [40,41,42] of these questionnaires were used. Good internal consistency of the French versions was demonstrated with Cronbach’s of 0.90, 0.90, and 0.86, respectively, in the previous studies [40,41,42] and Ordinal ⍵ of 0.92, 0.89, and 0.90, respectively, in the current study.

Data analysis

RStudio (Version 1.2.5033) for Macintosh was used for the analysis. Prior to statistical analysis, the assumption of normality was checked using an examination of skewness and kurtosis cues, as well as Shapiro–Wilk and Levene tests for normality and homoscedasticity. Due to significant deviation from normality in all the measures used in this study (Shapiro–Wilk < 0.001), we chose to implement nonparametric statistics. A Confirmatory Factor Analysis (CFA) based on the single-factor model of DOS was performed using the Lavaan and SemPlot packages from RStudio [43, 44]. Given the DOS 4-point Likert scale, we considered the data as categorical and we used the Weighted Least Squares with Mean and Variance (WLSMV) estimation method for the CFA [45]. F-DOS missing values (0.49%) were handled using listwise deletion before computing CFA, which is an acceptable procedure in regards with Tabachnick and Fidell’s recommendations [46] that removing missing data was acceptable when 5% or less of participants have missing data. To evaluate model fit we relied on the following recommendations: values of Comparative Fit Index (CFI) and Tucker Lewis Index (TLI) greater than about 0.9 or greater indicate good fit of the model [47], values of Root Mean Square Error of Approximation (RMSEA) about 0.08 or less suggest a reasonable model-data fit [48], and values of Root Mean Square Residual (RMSR) equal or less than 0.08 indicate a well-fitting model [49]. Reliability analysis was assessed using calculation of Ordinal ⍵ [50, 51]. To analyze convergent validity, Spearman’s Rho correlations were examined between F-DOS and EHQ. To test divergent validity, these coefficients were examined between F-DOS and EAT-26.

Results

Confirmatory factor analysis (CFA)

A CFA was performed based on the single-factor model of DOS. The computation of the fit indices showed that the model had a good fit: χ2(35) = 703.979, p < 0.001; CFI = 0.941; TLI = 0.924; RMSEA = 0.077; SRMR = 0.065. The graphical representation of the CFA with standardized factor loadings is represented in Fig. 1.

Fig. 1
figure 1

Path diagram of the tested F-DOS with standardized factor loadings

Full size image

Validity and reliability of F-DOS

The Ordinal ⍵ value of the F-DOS was 0.87, showing strong internal consistency. For each F-DOS item, mean, SD, and item selectivity are presented in Table 2.

The total scores of the F-DOS and the EHQ were significantly correlated (rs = 0.74, p < 0.001), suggesting good construct validity. Significant correlation coefficients were also found with the three subscales of the EHQ (knowledge: rs = 0.65; problems: rs = 0.69; feelings: rs = 0.62, p < 0.001 for all).

Regarding divergent validity, the relationship of each F-DOS item with the EAT-26 total or subscale scores revealed that coefficient correlations were in the 0.05–0.45 range. The correlation coefficient of F-DOS and EAT-26 total scores was 0.39 (p < 0.001).

Prevalence and psychopathology of ON

The mean total score of F-DOS for the sample was 19.2 (range 10–39; SD = 4.95). To assess the validity of the cut-offs previously reported, we applied the same method used in the original study [17]. In this method, the 95th percentile corresponds to 29 and a score equal or above 30 was suggested on the basis of this 95th percentile as a preliminary cut-off value for the presence of orthorexia [17]. A score between 25 and 29 was suggested as an “at risk” preliminary cut-off based on the 90th percentile [30, 33, 52]. With this method, in our sample, the 95th percentile corresponds to 28, suggesting a score equal or above 29 as a cut-off value for the presence of orthorexia, and the 90th percentile corresponds to 25, suggesting a score between 26 and 28 as an “at risk” cut-off. Considering that 29 is very close to 30, that 26 is very close to 25, and that in every other article validating the DOS in different languages [29, 30, 52], the preliminary recommended cut-offs of 30 and 25 were used, we chose to also use the cut-offs of 30 and 25. On this basis, 3.28% of participants could be considered having ON, while an additional 11.31% could be at risk of developing ON. No risk of ON was observed in 85.41% of the sample. No statistically significant relationships between the F-DOS categories and gender (Cramer’s V = 0.013, p = 0.766) or between the F-DOS and BMI categories (rs = 0.01, p = 0.39) were observed. However, a positive significant correlation, yet very small, was observed between BMI and the F-DOS total score (rs = 0.04, p < 0.01). Participants who indicated following a certain diet (vegetarian, vegan, flexitarian, etc.) were more likely to display ON or be at risk of developing ON (certain diet: ON: 8.4%, risk of ON: 27.7%; no certain diet: ON: 1.9%, risk of ON: 7.1%, Cramer's V = 0.311, p < 0.001). Considering only vegetarians or vegans (n = 404), they were more likely to have ON (6.19% vs. 1.95%) or to be at risk of developing ON (29.95 vs. 7.05%) compared with students who did not follow a particular diet (Cramer's V = 0.281, p < 0.001). Similarly, looking only at flexitarians (n = 126), they were more likely to have ON (4.76% vs. 1.95%) or to be at risk of developing ON (23.81 vs. 7.05%) compared with participants who did not follow a specific diet (Cramer's V = 0.14, p < 0.001).

To address the links between ON and psychopathology, a correlational analysis was performed between the F-DOS total score, markers of functioning, and mental health status in students: results of the last semester, levels of anxiety, depression, and obsessional–compulsive symptoms. While no significant correlations were revealed between F-DOS and symptomatology of anxiety (rs = − 0.005, p = 0.76), depression (rs = − 0.008, p = 0.64), and obsessive–compulsive disorder (rs =− 0.022, p = 0.23), a small positive correlation was revealed between F-DOS and results of the last semester (rs = 0.071, p < 0.001).

Discussion

This study is the first to describe a French version of the DOS. The so-called F-DOS showed good validity and reliability. Its Ordinal ⍵ value (0.87) indicated a good internal consistency, in the same range (0.8 to 0.88) as the Cronbach’s ⍺ of other DOS versions in different languages [8, 17, 26,27,28,29,30,31]. In addition, the mean total score of F-DOS in this study (M = 19.2, SD = 4.95) was very close to that of the DOS obtained in young adults in Germany (M = 17.75, SD = 5.44) [17] or to that of the E-DOS in a US student sample with a women majority (M = 20.02, SD = 5.95) [29]. Likewise, the ranges of mean score per item were highly similar between DOS (1.23–2.43), E-DOS (1.42–2.59), and F-DOS (1.27–2.65), further strengthening the validity of F-DOS. Similarly, the range of item selectivity values was also very comparable between DOS (0.49–0.63), E-DOS (0.56–0.73), and F-DOS (0.52–0.67) [17, 29].

Regarding the factor analysis of the F-DOS, the single component model had a good fit. Regarding divergent validity, the correlational analysis of F-DOS and EAT-26 showed that these instruments evaluate distinct constructs, even though, as expected, some relations exist [8, 53]. Altogether, these analyses indicate that the F-DOS displays significant reliability and validity, as well as specificity for ON among eating disorders.

With 3.28% of participants scoring above the cut-off, the prevalence of ON may appear very low in this study, compared to reports of ON prevalence rates as high as 30–70% in community or student samples [9, 21]. However, these high rates are currently thought to reflect significant psychometric limitations when measuring ON with ORTO-15, rather than attesting a widespread prevalence of ON [7]. Indeed, when additional criteria were used to assess ON using ORTO-15, a prevalence rate below 1% was found in a US sample [54]. Moreover, using the DOS or its adaptations rather than ORTO-15, the ON prevalence rate in community or student samples varied from 1 to 6.9% in Germany [17, 27], and it was around 8% in samples of US or China university students composed with a majority of women [29, 30]. Therefore, the ON rate in French university students appeared to be within the range reported in comparable samples assessed using the DOS or one of its adaptations.

Regarding BMI, this study found a small positive correlation between BMI and ON symptomatology, which is in line with recent reports showing positive links between ON and drive for thinness, a common underlying feature of AN [8, 17]. However, other reports showed no major relation between ON and BMI in students or community samples [25, 29, 30, 55]. Additional research is thus required to clarify this area.

Looking at the general psychopathology of the sample, the results did not reveal any correlation between F-DOS total score and levels of anxiety, depression, and obsessive–compulsive symptoms; accordingly, it cannot be assumed that general psychopathology increases with higher levels of orthorexic eating behavior measured by the F-DOS. A possible explanation might be that people with slightly elevated DOS levels, without displaying any pathological features, feel better when they eat healthier, and thus, they display lower levels of general psychopathology. It could be further assumed that insight into illness might not be given in ON [56]. Hence, some individuals with high levels of orthorexic eating behavior, supposedly displaying a pathological behavior, might not be willing to admit that their well-being is impaired. These two possible explanations might then lead to an unclear pattern of the relationship between general psychopathology and F-DOS total scores. Yet, the F-DOS score was positively correlated with academic results. A possible reason of this association could be perfectionism, a trait linked with orthorexic behaviors [7, 31, 57] but also with higher performance and academic achievement [58]. Also, regarding these aspects, more research is needed to further analyze how orthorexic eating behavior elicits psychological distress.

This study is not without some limitations. First, the BMI was calculated with self-reported weight and size. The results regarding BMI should thus be considered with caution. In addition, the use of self-report questionnaires could be viewed as a weakness in trustworthiness of results, and semi-structured interviews could have strengthened the findings. Second, due to the transversal design of this study, test–retest reliability was not examined.

In conclusion, this study suggests that the F-DOS is a psychometrically adequate instrument to measure ON symptomatology in French-speaking populations, thereby providing a reliable instrument to assess ON in individuals in these populations. This report also shows an ON prevalence of 3.28% in young adults, while an additional 11.31% could be at risk of developing ON. The existence of a reliable assessment tool is important for future research, notably for identifying ON risk factors and protective elements, as well as for developing potential interventions.

What is already known on this subject?

The aim of this study was to adapt the DOS into French and examine its psychometric properties.

What this study adds?

This study provides an instrument to measure ON symptomatology in French-speaking populations, allowing future research on risk and protective factors, and the development of potential interventions.