Introduction

The concept of orthorexia nervosa was introduced by Steven Bratman in 1997 describing a “fixation on healthy food” [1]. Bratman most recently outlined the difference between choosing to eat a healthy diet on the one hand, and, subsequently, potentially developing an obsession-like pursuit of this choice [2]. Only the latter element makes some people with orthorexic eating behavior the focus of research on eating disorders. Orthorexia nervosa has not been recognized by international classification guidelines, such as the DSM-5 or ICD. To determine whether it is a pathological disorder, clinical significance, e.g. impairment in areas of functioning. must be investigated.

Following Bratman’s approach, when an “orthorexic acceleration” [2, p. 383] is set off, negative emotions, as well as clinical significance (e.g. depressive symptoms, lack of control) occur. However, most studies on the topic are flawed due to insufficient psychometric properties of the scales that were used [3]. In particular, the authors argue that clinical significance or pathological behavior has simply not been part of the instruments that are used to measure orthorexia.

The prevalence of orthorexia in previous studies ranged from 7 to 89% depending on the instrument and sample that was used in the study [3]. Most studies use the ORTO-15 scale or a derivate. The ORTO-15 was constructed to identify people with “health fanatic” eating behavior, as well as obsessive–compulsive traits and phobia. It has been criticized for a tendency for false positive cases when applying the cut-off of 40 as in the cited study. Additionally, psychometric limitations, especially concerning the construction of the scale have been discussed as well [3]. These limitations of the scale are reflected in large variation of prevalence rates, even when assessing orthorexia in the same country. A study from the general population of Italy found that more than half of the population (57.6%) seems to be affected [4]. An attempt to translate the ORTO-15 to German and conduct a validation, yielded a prevalence of 69.1%, but the authors concluded that the instrument was insufficient to measure orthorexia and more valid instruments were needed [5]. Likewise, in a US student sample, the estimated prevalence measured by ORTO-15 was 71%, but dropped to less than 1% when items on clinical significance (e.g. negative consequences for health and social relationships) were taken into account [6]. Results are mixed regarding characteristics that are associated with symptoms of orthorexia and mostly stem from highly selected samples. Following a vegetarian diet [7], lower body satisfaction and higher overweight preoccupation [8] were found to be associated with orthorexia in some studies. For other variables, such as body weight [9, 10], education and age [9], results are heterogenous [11]. Whether gender differences exist, remains unknown, with some studies reporting more orthorexic behavior in women [4, 11, 12]. The only other study from a population-based sample using the ORTO-15 found women to be more affected than men [4].

A different approach to measuring orthorexic behavior can be found in the Dusseldorf Orthorexia Scale (DOS) that was developed in Germany. It is a measure based on primary criteria that Bratman found “promising” [2, p. 384]. Prevalence rates using the DOS range from 2 to 3% [13, 14]. Population-based studies using purportedly validated instruments, however, are lacking. This study therefore set out to investigate the prevalence of orthorexic behavior in a large German representative sample. Additionally, it aims to identify groups at risk for orthorexic behavior by analyzing characteristics associated with the relevant DOS cut off.

Materials and methods

Sample

A randomly selected sample of the German general public was drawn within this telephone-based survey in December 2015 to January 2016. In this procedure, computer-generated telephone numbers were randomly called (including mobile numbers) and respondents were invited into the study (n = 2045) in order to achieve the sample goal of n = 1000 participants. Participants in households were chosen according to the Kish-Selection grid. The response rate was 49.8% (refusal: 27.2%, not being reached: 23%) and n = 1007 participants consented to participate in a 30-min interview on several aspects of health assessed by standardized questionnaire. The survey was conducted with the help of a large private German research institute (USUMA GmbH).

Instruments

After general sociodemographic variables (age, gender, occupation, education), participants were asked to indicate their height and weight. If body weight was not reported, the computer-assisted interview calculated weight ranges corresponding to BMI categories of under or normal weight, overweight or obesity according to WHO standards [15]. These ranges were proposed after a respondent reported height, but refused to give information on weight. The computer-assisted telephone interview then calculated weight ranges. For instance, if a participant reported a height of 1.68 m, the interviewer would then read weight ranges (e.g. for normal weight: “Is your weight between 54-69kg?”). A weight category was therefore available for all respondents. Additionally, participants were asked to indicate whether they were omnivore, vegetarian, vegan or adhered to a different kind of diet due to allergies (specific needs). Lastly, depression was assessed using the Patient Health Questionnaire (PHQ-9, [16]).

Orthorexic behavior was measured with the Dusseldorf Orthorexia Scale. The long version of the DOS consists of three subscales (orthorexic eating behavior, avoidance of additives, supply of minerals). For this study, the short version with ten items only covering the scale orthorexic eating behavior was used, as validity studies were also done with this version. The presence of Orthorexia is determined by using this ten-item short version, at a cut-off value of 30. It has shown excellent psychometric properties as well as validity [14]. Validity was concluded from significant correlations with the Orthorexia Self-Test (r = 0.72), as well as sub-scales of the Eating Disorders Inventory (drive for thinness, bulimia, body dissatisfaction [17]). Further validation studies were conducted in samples with patients at risk for orthoretic eating behavior (e.g. Diabetes Mellitus Type 1 patients) and patients with eating and compulsive obsessive disorders. The cut-off was chosen since at this cut-off participants in the norm sample reported dissatisfaction with eating behavior and rated their eating behavior as “unhealthy”. Its reliability in the present sample was Cronbach’s alpha = 0.80.

Statistical analysis

All data were analyzed using the survey data package in Stata 13.0 [18]. The procedures then include a weighing variable (age, gender, region of Germany and education) that was provided by the research institute. This provides the basis for a representative sample.

Education was recoded and was dichotomized to either 10 years or less compared with 12 years or more of schooling. The PHQ-9 score was used continuously to reflect depressive symptoms. Major depression was determined following the proposed coding scheme (presence of at least five symptoms, of which at least one is found in symptom loss of interest or depressed mood). The presence of orthorexia was determined with the proposed cut-off of 30 points or more. The prevalence rate across socio-demographics as well as mean DOS scores were estimated and compared using t-test and Chi square proportion tests. Logistic regression analysis was used to further investigate determinants of orthorexia. Interactions of depressive symptoms and BMI as well as socio-economic status and BMI were tested and only included when the interaction term became significant. The level of significance was set to p < 0.05.

Results

The sociodemographics in the weighted sample were as follows: The mean age was 50.6 years (SE: 0.76) with a proportion of women of 48.6% in the sample. About three quarters (70.7%) reported an educational attainment of under 12 years of schooling. While 2% of all participants reported a BMI under 18.5 (underweight), 39.9% were normal weight. The prevalence of overweight was 35.5, 22.5% of the sample were obese. The mean BMI was 26.44 kg/m2 (SD = 5.30). These data support the representative nature of the sample, although our sample is slightly younger. In the German general public, 50.7% are women, and the mean age is 44.2 years. Educational attainment below 12 years of schooling is documented in 66.9% of all Germans [19].

Nine out of ten participants indicate an omnivore diet. Only 5.9% report to eat vegetarian, and another 3.2% adhere to specific needs, due to allergies and so on. The mean sum score of the PHQ was 3.58 (SD = 4.74). The prevalence of major depression (PHQ-9) was 11.12%.

The mean DOS score was 19.25 (SD = 6.11). Orthorexic behavior in the sample was reported by 6.9% (5.9% in men, 7.9% in women). There was no significant difference in these proportions (p = 0.392). When comparing age groups along the Median split of the sample (M = 51), the prevalence did not differ between the two groups (3.4 vs. 3.5%, p = 0.875). Table 1 summarizes differences in prevalence across socio-demographic variables.

Table 1 Prevalence of orthorexic behavior across socio-demographics
Full size table

The single item results for the DOS reveal that four items are more often endorsed than the others: having established nutrition rules, paying attention to healthy foods, preference of healthiness to indulgence and enjoying foods that are healthy (Table 2). Most other items were endorsed by 5–7% of the respondents and closely correspond to the overall prevalence.

Table 2 Response pattern in single items on the DOS
Full size table

Logistic regression revealed an association of a orthorexic behavior with lower education, higher depressive symptoms, the presence of obesity and a vegetarian diet in univariate analysis (Table 3).

Table 3 Logistic regression, presence of orthorexia (no/yes)
Full size table

In the multivariate model, only educational attainment, higher depressive symptoms, and a meat-free diet remain significantly associated to a higher proportion of orthorexic behavior. The diagnosis of major depression was also associated with a higher probability of reaching the DOS cut-off. This association was significant- in univariate analysis (OR = 2.71, p = 0.026) as well as in multivariate analysis (OR = 2.59, p = 0.044, data not shown, all other associations remained as described above).

Discussion

This study set out to report prevalence of orthorexic behavior and its determinants in a large German representative sample. It found a prevalence of 6.9% and documents that lower educational attainment, higher depressive symptoms and a vegetarian diet are associated with orthorexic behavior.

The DOS aims to identify people that are at a higher risk of developing pathological eating with clinical relevance [20]. In the publication describing the construction of the DOS, a thorough test construction procedure was described (selection and pre-testing of items). The sample was drawn in an online survey representative for the younger population of Germany [17]. In this sample, 70% were women and they reported a higher prevalence of orthorexic behavior than men (4.1 vs. 1.6%). However, the overall prevalence in this sample was only 3%. Our results therefore could cautiously be interpreted as showing a rise in the prevalence of orthorexic behavior in Germany since the original study in 2012 [14]. In contrast to the study in 2012, we observe a higher prevalence overall (6.9% compared with 3% in 2012), also in men (5.9%) and in women (7.9%). Other research using the DOS has also not shown gender differences. A study in two German student populations reports no significant differences between men and women in a student sample [21]. A predominance of orthorexia in men has been reported in other studies [11], but findings are extremely heterogenous.

Some studies reported a notion of pride when interviewing participants who scored high on orthorexia dimensions [1]. A recent review of the literature finds that most studies on the topic have framed orthorexia nervosa in the concept of healthism. Healthism entails a health consciousness [22] that motivates individuals to thrive for best possible eating and exercise behavior. The slim and fit body is then seen as a marker of good health, while the obese body is perceived as a sign of personal weakness and illness [23]. An optimized eating behavior indicated by specific food choices, such as choosing home-made and organic food, avoidance of certain foods, including alcohol, is part of this healthism movement observed in the general public [24]. In this recent review, studies were summarized that investigated the association of orthorexia nervosa, body image and weight. It was found that high internalization of a thin body ideal as well as previous dieting are closely related to higher scores of orthorexia [24].

The observed change in prevalence in our study coincides with another trend in the German public: The rise of the prevalence of obesity [25]. We did not find a higher prevalence in those who are underweight compared with normal and overweight participants as expected under the healthism frame. Likewise, other studies for example have also documented higher orthorexia scores among people with higher BMIs [26,27,28]. Two studies were conducted in student samples (USA/Turkey), one in a an adolescent sample [26], using the Bratman Orthorexia test or the ORTO-11/15. One possible explanation can be a mediated relationship of obesity and hence low SES, body image and orthorexia.

Following this line of thought, we did not find a higher prevalence in participants with higher educational attainment compared with participants with lower education status. We would expect a higher prevalence in participants with high educational attainment as a close correlation of educational attainment and a vegetarian diet has been reported, and people following a vegetarian diet reported orthorexic behavior more commonly, but instead observe that participants with lower educational attainment are more affected by orthorexic behavior, going align with another study that links orthorexic behavior to lower socioeconomic status [29]. Since obesity is often linked to low economic status, the mediated relationship with orthorexia seems comprehensible. Moreover, previous studies show a relationship of body dissatisfaction and obesity [30]. Higher body dissatisfaction is found in individuals with obesity and previous research concludes that higher body dissatisfaction may also be correlated to orthorexia. Although we cannot postulate causality, it seems reasonable that the strains that accompany body dissatisfaction and obesity are reflected in symptoms of impaired mood and drive. Within this frame, higher depressive symptoms may therefore be linked to orthorexia as indicated by our study. This might be the case, as orthorexic behavior that results in obsessive eating habits, can lead to social isolation and strong feelings of guilt. The reverse association—that people cope with depressive symptoms by following orthorexic dietary patterns—seems less likely as depressive symptoms are mainly linked to a diminished interest in almost all activities [31]. If this line of thought is followed through, the pattern observed by clinicians that orthorexia is indeed accompanied by psychological distress and strain can be supported [32].

Further studies on variables associated with orthorexia are needed, as the current overviews document very mixed results regarding almost all associations [11]. These also result from very different samples, as many studies are done in student samples or samples with a specific occupational background (such as nutritionists or doctors). This population-based study therefore adds evidence using a different approach to capture orthorexic behavior.

The DOS captures the proposed new criteria for orthorexia nervosa more closely than previous measures. Dunn and Bratman call for a modification (such as the one by Barthels et al. [14]). The concept proposed by Dunn and Bratman includes the role of weight loss [3] that covers the stepwise escalation of orthorexic behavior by including more food groups and becoming even more obsessive. These variables have not been included in research and should be the focus to determine the line that Bratman draws between food healthism and a potential pathological eating disorder named orthorexia nervosa. The DOS items may still represent mostly healthy eating, with only one item capturing social impairment. In one of the more recent studies, the attempt to additionally consider impairment due to orthorexic eating behavior led to a significant drop in prevalence to less than 1% [6]. Our study also shows that only about 3% of the respondents endorse social interaction impairment due to their eating behavior. To sum up, this indicates that when considering impairment more precisely, prevalence rates of orthorexia nervosa may be more similar to other eating disorders. Dunn and Bratman [3] argue that orthorexia nervosa is currently reflected the closest in the DSM diagnosis “Avoidant/Restrictive Food Intake Disorder,” (ARFID), however, the concern of unhealthiness of is not adequately covered in its diagnostic criteria [3] and further investigations are needed. An overlap with Anorexia nervosa may exist, but also needs further investigation [33, 34].

Limitations

This study also has limitations. For one, no questions on actual eating behavior or body image distortion were included. The DOS was administered orally via telephone for the first time and data on the validity of this approach is lacking. Research comparing orally and written questionnaires to capture other psychological constructs, however, shows that both approaches are equally appropriate [35, 36]. BMI was only derived from self-report data—a method known to underestimate the prevalence of obesity. Yet, the prevalence reported in our sample corresponds closely to the prevalence in the German public [25]. Another point of critique can be found in the absence of a measurement for social desirability. It could be the case that groups that are stigmatized, such as persons with obesity and lower education, feel the need to demonstrate positive and social desired behavior to a higher extend than those non-stigmatized groups. These findings, however, would underline the importance of addressing weight stigma in the general public, where weight stigma is still common [37] and comes with negative effects for those affected by it [38]. Lastly, we acknowledge that the sample size may still be too small to detect and report sub-group differences. This is the first investigation in a representative study and as such it can provide the basis for future research; however, clearly, a larger number of those affected will give way for a more detailed subgroup analysis.

Conclusions

This study provides data from the general public in Germany and gives important insight in the nature of orthorexic behavior. Contrary to reports that did not show significant strain in those affected by orthorexia nervosa, we find a close correlation to higher depressive symptoms. Furthermore, the association of this extreme dietary pattern and dieting behavior, body dissatisfaction and body weight needs further investigation, as our results show a surprising association to obesity, but not underweight. Lastly, we did not find a gender difference in prevalence rates. Due to the small number of affected individuals in the sample, investigations to clarify potential gender differences are needed.