Introduction

Recent research documents the extent to which health-care professionals identify orthorexic eating behavior as a concern within their practice and believe orthorexia should be a distinct diagnosis [1,2,3]. Although not an official diagnosis, authors have proposed criteria for orthorexia nervosa, which generally emphasize obsession and extreme preoccupation with healthy eating, distress resulting from failure to adhere to self-imposed criteria for healthy eating, and resulting impairment in areas of life to include health and psychosocial functioning (for a review see [4]). Bratman [5], the individual who first named the phenomenon of orthorexia, emphasized obsessive striving for dietary purity and feelings of superiority over those who do not espouse the dietary practices. Other suggested features include obsessive–compulsive traits [6], worry about food quality [7], phobic avoidance of “unhealthy” food [4, 8], overvaluing “healthy” dietary practices relative to empirically documented evidence of health benefits [8, 9], unusual concern with health [10], escalating dietary restrictions [11], distress and/or self-harm in response to breaking self-imposed dietary rules [4, 9, 11], intolerance of others’ food beliefs [9], impaired functioning and health [1, 4, 7, 9,10,11], and potential weight loss [10, 11]. Although some include a lack of body image disturbance among the key distinguishing features of orthorexia [4], Dunn and Bratman [11] suggest that positive body image is dependent on complying with the self-imposed rules.

Measurement of orthorexia

One area of the existing literature on orthorexia with particular limitations is measurement of the construct. Some prior measures yielded mixed psychometric findings (e.g., [12,13,14,15,16,17,18]), and the effort to identify strong measures is needed [14, 19,20,21] and ongoing (e.g., a recent revision of the ORTO-15 [22], the most commonly used measure in orthorexia research, illustrates these efforts towards continuous improvement [23]). One orthorexia measure with promising psychometric properties (i.e., αs > 0.80 and test–retest reliabilities > 0.70) is the Teruel Orthorexia Scale (TOS [17]). The TOS contains two subscales assessing healthy eating and negative consequences of the preoccupation with healthy eating—Healthy Orthorexia and Orthorexia Nervosa. Within the development study, the Orthorexia Nervosa factor related to measures of disordered eating, negative affect, obsessive–compulsive symptoms, and perfectionism, whereas the Healthy Orthorexia factor was less strongly related to these indicators of psychopathology and weakly correlated with adaptive characteristics when removing variance associated with unhealthy aspects of the construct. Thus, the latter scale illustrated a residual effect (see [24] for a discussion of residual effects) in that the pattern suggest that interest in healthy eating has the potential to serve as a measure of a protective factor once statistically removing the overlap it shares with disordered orthorexic eating. As such, the authors of the measure subsequently defined the subscale as assessing “healthy interest in diet, healthy behavior with regard to diet, and eating healthily as a part of one’s identity” [25] (p. 2). In a study investigating its psychometric properties [26], a confirmatory factor analysis (CFA) provided modest support for the bidimensional structure with some fit indices clearly supporting the model (e.g., CFI > 0.95) and others offering less support (e.g., RMSEA > 0.06 with several cross-loadings). Recent CFAs with Portuguese [27], Arabic [28], German [29] translations provided mixed support for the two-subscale structure in populations outside of Spain. Although the TOS has been translated from Spanish to English, its psychometric properties have yet to be examined with an English-speaking sample.

Despite the lack of formal diagnostic criteria and the measurement concerns, prevalence studies permeate the literature and yield point prevalence estimates ranging incredibly, from 3.3 [30] to 88.7% [31]. Methodological issues, especially the measurement challenges, render many of these studies unreliable [11]. Thus, the prevalence of orthorexia is quite uncertain and availability of sound measures is critical.

Orthorexia and related disorders

Relationships between orthorexic symptoms and other psychological conditions are also unclear, and disagreement exists as to whether orthorexia nervosa deserves its own classification (e.g., [5]) or constitutes a subtype of another disorder [32, 33]. For example, researchers theorize an overlap between obsessive compulsive disorder (OCD) and orthorexia nervosa in that both can include intrusive thoughts and rituals around food and health, exaggerated concern with impurity, and rules about food [18]. However, unlike with OCD, these experiences are thought to be ego-syntonic in orthorexia nervosa [8, 18, 34]. Similarly, orthorexia nervosa and eating disorders both include restrictive eating habits and the potential for weight loss as well as traits of perfectionism and OCD symptoms [18, 33, 35]; however, the motivation for dietary rules with orthorexia nervosa is thought to be health and purity [9, 18] rather than pursuit of thinness and body image disturbance (though studies with measures of orthorexic eating call this theoretical motivation into question [25, 36]).

Case studies (e.g., [37]) and quantitative research link orthorexic symptoms to somatic disorders [38, 39], and researchers have suggested that hypochondriasis should be a differential diagnosis for orthorexia nervosa [7, 10]. Despite the theorized connection (e.g., [8, 18, 40]), scant research explores the relationship between somatic concerns and orthorexia nervosa.

The articulated features (i.e., perfectionism, distress) of orthorexia nervosa along with the theorized relationships and distinctions between orthorexic symptoms and other disorders suggest that a valid measure of orthorexia nervosa would relate to measures of obsessive–compulsive and eating disorders symptomology, perfectionism, and distress. A valid measure should also produce patterns of relationships that support the distinction between orthorexic symptoms and disordered eating such that body image disturbance is less strongly associated with orthorexic symptoms. In addition, the theorized link between orthorexia nervosa and somatic concerns in the literature suggest greater endorsement of orthorexic symptoms might be found among those with more somatic concerns.

Purpose

Mixed performance of extant measures of orthorexia nervosa limits confidence in conclusions drawn from scholarly work. One purpose of our study is to examine psychometric properties of the TOS, a promising new instrument not yet examined with an English-speaking sample. We also evaluated the theorized, but rarely investigated, relationship between orthorexia and health anxiety, a key component of somatic symptom and illness anxiety disorders.

We expected that an English-speaking U.S. sample would replicate the TOS’s two-factor structure (Hypothesis 1). We expected the TOS subscales to yield good internal consistency (α ≥ 0.80; Hypothesis 2). Given the relationships between other measures of orthorexia nervosa and eating disorder symptoms, often above 0.75 (e.g., [41, 42]), as well as the initial psychometric findings for the TOS [17], we predicted the Orthorexia Nervosa subscale would yield a strong correlation with a measure of eating disorder symptoms (i.e., 0.70 > r > 0.50) and that Healthy Orthorexia would yield a significant, moderate correlation with a measure of disordered eating (i.e., 0.30 < r < 0.50; Hypothesis 3). We also expected that scores on the Orthorexia Nervosa subscale would moderately relate to scores on measures of obsessive–compulsive symptoms, negative reaction to mistakes, and subjective distress (i.e., 0.50 > rs > 0.30; Hypothesis 4). Given past findings where Healthy Orthorexia showed near-zero relationships (unless controlling for its overlap with Orthorexia Nervosa, in which case residual effects sometimes emerge; e.g., [17, 26]), we did not set any hypotheses for the relationship between Healthy Orthorexia and measures of obsessive–compulsive symptoms, negative reaction to mistakes (i.e., perfectionism), and subjective distress. In support of the validity of the subscales as sufficiently distinct from a measure of disordered eating, we expected both subscales of the TOS would be less strongly related to body image disturbance than would be a measure of eating disorder symptoms (Hypothesis 5). Finally, we expected symptoms of health anxiety would relate more to symptoms of pathological preoccupation with healthy eating than interest/engagement in healthy eating (Hypothesis 6) and that, together, the two dimensions of orthorexia measured by the TOS would account for a significant amount of variance in health anxiety after controlling for eating disorder symptoms (Hypothesis 7).

Methods

Participants

We set a minimum sample size of 200 to ensure 10 participants per parameter [43]. We restricted participation to college students enrolled at a large southeastern U.S. university who were at least 18 years of age.

Of the 450 participants who responded to the study, 94.9% completed the study in full. We removed 123 participants who failed one or more attention checks and 23 who did not finish the study. This left 304 participants who responded to all questions, passed all attention checks, and were included in the analyses. Within our final sample, most participants (n = 261, 85.9%) identified as White/Caucasian (no other racial/ethnic identity exceeded 8% in the sample). Participants’ ages ranged from 18 to 45 (M = 20.63; SD = 2.52), with 94.7% under the age of 24. Most participants identified as cisgender women (n = 218, 71.7%) or men (n = 80, 26.3%). Undergraduate students comprised the vast majority (97.0%) of participants.

Participants’ mean BMI (calculated from self-reported height and weight) was 24.37 (SD = 4.61), falling within the “normal” range [44]. Few participants (11.8%) reported a professionally diagnosed food allergy. A history of dieting was common among participants, with 62.2% having engaged in at least one diet; rarely were the diets recommended by health-care professionals. Nearly 10% of participants reported a previous diagnosis of an eating disorder (4.9%), obsessive–compulsive disorder (4.0%), or both (0.1%); and 60% (n = 18) of these individuals had received treatment.

Measures

Teruel Orthorexia Scale (TOS)

Barrada and Roncero [17] created the two-subscale TOS to assess orthorexia symptoms. The Healthy Orthorexia subscale examines interest and engagement in a healthy diet (e.g., “I mainly eat foods that I consider to be healthy”). The Orthorexia Nervosa subscale assesses negative consequences of fixation on a healthy diet, including self-punishment, negative emotions, and interference with daily life (e.g., “If, at some point, I eat something that I consider unhealthy, I punish myself for it”). Participants rate 17 statements on a 4-point Likert scale ranging from 0 (completely disagree) to 3 (completely agree). Items are summed to create subscale scores, with higher scores indicating more orthorexic symptoms. The original measure (in Spanish) was validated on a sample of over 900 individuals aged 18 and above (mostly students). The authors provided an English translation of the TOS in which a psychologist and professional translator with 20 years of translation experience translated the measure. To our knowledge, the English version was not back-translated or evaluated psychometrically.

Measures used to evaluate concurrent validity

For all but the measure of body image disturbance, we created total scores, by summing across items, on the measures used to evaluate concurrent validity. We averaged items to compute the total score for the measure of body image disturbance. For all measures, higher scores indicate greater endorsement of symptomology/concerns.

We used the Eating Attitudes Test-26 (EAT-26 [45]) to assess eating disorder symptoms. The EAT-26 consists of 26 items rated on a 6-point Likert scale ranging from always to never (e.g., “I am preoccupied with a desire to be thinner”). The three most pathological responses receive scores of 3, 2, or 1, depending on severity, with the remaining responses all scored 0. Research supports the reliability (α > 0.90 [46]) and validity [47] of the EAT-26. Current sample alpha was 0.85.

We used the Obsessive–Compulsive Inventory-Revised (OCI-R [48]), an 18-item shortened version of the original OCI [49], to assess OCD symptoms. Participants rated how much 18 experiences (e.g., “I check things more often than necessary”) distressed or bothered them during the past month using a 5-point scale ranging from 0 (not at all) to 4 (extremely). Research supports the reliability and validity of the OCI-R [48, 50], and the OCI-R effectively discriminates between individuals with OCD, those with other anxiety disorders, and healthy controls [48]. Current sample alpha was 0.89.

We used the Frost Multidimensional Perfectionism Scale–Concern over Mistakes Scale (CMS [51]) to measure respondents’ perfectionism, specifically their tendency to experience negative emotions as a result of mistakes, equate mistakes to failure, and believe that mistakes cause a loss of respect. Participants rated their agreement with the 9 CMS items (e.g., “If I do not do well all the time, people will not respect me”) using a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). Research supports the reliability and validity of the CMS [17, 51, 52]. Current sample alpha was 0.90.

The Positive and Negative Affect Schedule–Negative Affect Scale (PANAS–NAS [53]) was used to measure subjective distress. Participants rated the extent to which they experienced 10 markers of negative affect (e.g., “distressed” and “ashamed”) during the past few weeks using a 5-point scale ranging from 1 (very slightly or not at all) to 5 (extremely). Research supports the reliability and validity of the PANAS-NAS in a college sample [53]. Current sample alpha was 0.88.

We used the Situational Inventory of Body-Image Dysphoria-Short Form (SIBID-SF [54]) to assess a component of body image disturbance—dysphoric body image emotions. The SIBID-SF, based on the original 48-item version [55], contains 20 items. Participants rated each item based on how often they experienced negative feelings about their appearance in the given situation (e.g., “When I look at myself in the mirror”) using a 5-point Likert scale ranging from 0 (never) to 4 (always or almost always). Research supports the reliable and valid use of the SIBID-SF [54]. Current sample alpha was 0.97.

Short Health Anxiety Inventory (SHAI)

We used the SHAI [56], an 18-item short form of the Health Anxiety Inventory, to assess health anxiety. Each item contains four statements (a–d). Participants were instructed to select the statement which best captures their feelings over the prior six months [e.g., “(a) I usually feel at very low risk for developing a serious illness; (b) I usually feel at fairly low risk for developing a serious illness; (c) I usually feel at moderate risk for developing a serious illness; (d) I usually feel at high risk for developing a serious illness”). Items are scored from 0 to 3, where a = 0, b = 1, and so on. Participants can select more than one statement and the statement with the highest score is used in summing items to compute the total score. Higher scores represent greater levels of health anxiety. Research supports the internal consistency (α ≥ 0.86 [56, 57]), and correlations with a measure of attitudes towards illness and medical utilization support the validity [57], of the SHAI. In addition, the SHAI effectively differentiates between people diagnosed with hypochondriasis and comparison groups [56]. Current sample alpha was 0.90.

Demographics

The survey included a demographics questionnaire to gather background information. Participants indicated their age, gender identity, racial/ethnic identity, height, weight, food allergies, dieting history, and history of obsessive–compulsive and eating disorders diagnoses.

Procedure

To complete a back-translation of the TOS, we provided the English version of the TOS to a second-year Spanish graduate student teaching assistant with a bachelor’s degree from Spain. The individual was blinded to the original Spanish version of the TOS, per the accepted procedure [58] and translated the English version of the TOS to Spanish. A second graduate student in the same department, who was a native of Spain, compared that back-translated version to the original to determine comparability of language and ensure that any differences would result in equivalent interpretation [58]. Most items were considered equivalent in meaning despite minor differences in language choice. However, slight differences in meaning for items 3 and 13 were found. We chose to use the published English TOS as is given its existence in the literature and our goal to evaluate the psychometric properties of the published measure.

Once Institutional Review Board approval was gained, we recruited participants through SONA (a research participation system used at the University). Upon selecting the study, participants rerouted to the survey in Qualtrics, an online software, where they viewed a letter containing information for informed consent. Those who consented proceeded with the study and completed the measures in a randomized sequence followed by the demographics questionnaire. We embedded nine attention check items in measures throughout the study (e.g., “I am reading carefully and therefore will select agree as my response”). We carefully worded attention checks to mirror length of items in our measures and match the structure of the items within the measure (e.g., if most items in the measure started with “I,” the attention check also started with “I”). Thus, our attention checks avoided easy detection by a visual scan and required reading each item. We did not include attention checks on some measures due to the difficulty to structure our attention checks to align with the measure and the goal to embed them within measures. In no case did we include more than two attention checks within any measure. Participants received extra credit in a course in exchange for participation.

Statistical analyses

We used R to conduct a parallel analysis and exploratory factor analysis (EFA) and used MPlus to conduct exploratory structural equation modeling (ESEM) to test the factor structure of the English-language TOS in a U.S. sample. Specifically, in evaluating the internal structure of the TOS, we followed the process used by Sánchez-Carracedo et al. [59] by first conducting an EFA prior to the ESEM. We used parallel analysis and extant research to extract factors for the EFA. For our EFA, we used weighted least squares (WLS) extraction with oblimin rotation. We evaluated model fit using the Tucker–Lewis index (TLI), standardized root mean squared residual (SRMR), and root mean square error approximation (RMSEA). For our ESEM, we used WLS extraction with geomin rotation. We evaluated fit using the same indices as for the EFA and with the comparative fit index (CFI). Although recommendations for cutoffs vary, we set the criteria for good fit as RMSEA and SRMR < 0.06 and 0.08, respectively, and values of > 0.95 for the CFI and TLI [60]. In addition, in evaluating factor structure, we considered items with loadings > 0.30 as factor indicators and sought clean factor loadings (differentiation > 0.20). Using SPSS 27, we tested our remaining hypotheses using Cronbach’s alphas, correlations, z tests of the beta weights for dependent correlations, and a hierarchical regression.

Results

Internal structure of the TOS

The parallel analysis indicated a three-factor solution. Given the TOS authors developed the measure with two factors, we conducted a set of EFAs with two and three-factor solutions. Bartlett’s Test of Sphericity (χ2 = 2366.67; df = 136; p < 0.001) and the KMO of 0.91 supported use of EFA with the data. See Table 1 for EFA factor loadings. Initially, we included all 17 TOS items in the EFA. The three-factor EFA accounted for 51% of the variance and fit the data well (SRMR = 0.04; RMESA = 0.05, CI = 0.04–0.07; and TLI = 0.95). However, there were several cross-loadings between factors, only three items loaded cleanly onto factor 3 and five items lacked clean loadings (differentiation < 0.20; items 1, 9, 13, 15, and 16). Correlation between factors ranged from 0.32 to 0.58. The two-factor EFA accounted for 47% of the variance and produced mixed fit indices (SRMR = 0.05; RMSEA = 0.07, CI = 0.06–0.08; and TLI = 0.90). In this case, only one item demonstrated cross loading (i.e., item 13) and two items (i.e., 13 and 15) lacked clean loadings. The correlation between the factors was 0.41.

Table 1 Factor loadings from exploratory factor analysis of TOS items
Full size table

Next, we used ESEM to evaluate model fit, following procedures reported by the TOS authors [26]. Given the number of cross-loadings on the three-factor EFA and the theoretical structure for the TOS, we tested the two-factor model published elsewhere. Table 2 contains the factor loadings for our ESEM analyses. The 2-factor ESEM with all 17 TOS items fit the data well (CFI = 0.98, TLI = 0.97; SRMR = 0.04; RMSEA = 0.06 [CI = 0.05–0.07). In this case, only one item produced a loading > 0.30 on the non-primary factor. However, item 13 continued to show poor differentiation between the factors, even with this more robust analysis, and produced a loading on its primary factor that was weaker than the correlation between the factors (r = 0.51, p < 0.001). In addition, the wording used in item 13 confounds quantity and health when quantity is said to be unimportant in orthorexia nervosa (e.g., [7]). With the concerns with poor differentiation with item 13 (drawing on problems also found with the EFA) and the confounding of two constructs within the item wording, we ran a second ESEM eliminating that item. This modified ESEM with 16 TOS items fit the data well (CFI = 0.98; TLI = 0.97; SRMR = 0.04; RMSEA = 0.06 [CI = 0.05–0.08]). One item produced a loading > 0.30 on the non-primary factor but demonstrated good differentiation, and all loadings on their primary factor exceeded the correlation between the factors (r = 0.50, p < 0.001).

Table 2 Factor loadings from exploratory structural equation modeling
Full size table

We used this 16-item version of the measure, the TOS-16, with 2 subscales, Healthy Orthorexia and Orthorexia Nervosa, when testing the remaining hypotheses.

We examined the reliability of each TOS-16 subscale. The Cronbach’s alpha was 0.88 for the eight-item Healthy Orthorexia subscale and was 0.86 for the eight-item Orthorexia Nervosa subscale. As predicted, both alpha values were satisfactory.

Examining the TOS-16 and theoretically related psychological constructs

Table 3 contains means, standard deviations, and intercorrelations for each measured variable. Our third hypothesis was that scores on measures of disordered eating (EAT-26), obsessive–compulsive symptoms (OCI-R), perfectionism (CMS), and emotional distress (PANAS-NAS) would moderately to strongly correlate (i.e., r would exceed 0.30, but would not exceed 0.70) with scores on the TOS-16 Orthorexia Nervosa. As expected, all relationships with the Orthorexia Nervosa subscale fell within the expected, moderate range, lending support to both the convergent and discriminant validity of the TOS-16 by demonstrating that the instrument is measuring a construct that is related, but distinct from disordered eating, obsessive–compulsive disorder, perfectionism, and general emotional distress, respectively. Except for eating disorder symptoms, TOS-16 Healthy Orthorexia was not significantly related to the indicators of psychopathology, and consistent with our third hypothesis, the relationship between TOS-16 Healthy Orthorexia and a measure of disordered eating was < 0.50.

Table 3 Means, standard deviations, reliability coefficients, and intercorrelations for TOS and theoretically related constructs
Full size table

Body image disturbance, eating disorder symptoms and orthorexia

Because body image disturbance is said to be less relevant for orthorexia (e.g., [7, 11]), we hypothesized that scores on the SIBID-SF (i.e., body image disturbance) would relate more strongly to scores on the EAT-26 (i.e., eating disorder symptoms) than to scores on both TOS-16 subscales. As anticipated and supporting discriminant validity, the z tests of the beta weights showed the relationship between body image disturbance and eating disorder symptoms was significantly stronger than that between body image disturbance and interest/engagement in (z = 5.91, p < 0.001) or pathological preoccupation with (z = 1.98, p = 0.024) healthy eating.

Health anxiety and orthorexic eating

For Hypothesis 5, we predicted that symptoms of health anxiety would relate more to symptoms of pathological preoccupation with healthy eating than interest/engagement in healthy eating. We found a moderate correlation between scores on the SHAI and scores on the Orthorexia Nervosa subscale and a near-zero relationship between scores on the SHAI and scores on the Healthy Orthorexia subscale (see Table 3). A z test of the beta weights supported our hypothesis (z = 7.34, p < 0.001).

We regressed health anxiety on a measure of eating disorder symptoms in Step 1, R2 = 0.20, p < 0.001, and orthorexia (using both TOS-16 subscales) in Step 2, R2Δ = 0.05, p < 0.001, to test our sixth hypothesis. TOS-16 subscales explained a significant amount of variance in health anxiety after accounting for the relationship between eating disorder symptoms and health anxiety, with both interest/engagement in healthy eating (i.e., TOS-16 Healthy Orthorexia; sr = − 0.16, p = 0.002) and pathological preoccupation with healthy eating (i.e., TOS-16 Orthorexia Nervosa; sr = 0.22, p < 0.001) each explaining a significant amount of unique variance.

Exploratory analyses

Although we made no hypotheses about the relationship between TOS-16 Healthy Orthorexia and measures of obsessive–compulsive symptoms, perfectionism, or negative affect, we examined the partial correlations in light of prior research sometimes illustrating a residual effect. Specifically, the residual effect reflects a statistically pure construct in which the significant shared variance (i.e., r = 0.49) between TOS-16 Orthorexia Nervosa and TOS-16 Healthy Orthorexia is removed (essentially, holding participants’ scores on TOS-16 Orthorexia Nervosa constant). In such case, we too found residual relationships between TOS-16 Healthy Orthorexia and measures of obsessive–compulsive symptoms (r = − 0.23, p < 0.01), perfectionism (r = − 0.15, p = 0.009), and negative affect (r = 0.32, p < 0.001). Similarly, partialing out the shared variance with TOS-16 Orthorexia Nervosa resulted in a negative relationship between TOS-16 Healthy Orthorexia and a measure of body image concerns (r = − 0.35, p < 0.001) and a near-zero relationship with a measure of disordered eating (r = 0.01, p = 0.853).

Discussion

Implications of findings

The English version of the TOS was translated from Spanish to English by an experienced professional before its publication in 2018 [17], but never back-translated. We followed the recommended procedure for back-translation [58] and found slight differences in meaning for items 3 and 13. We used the published version of the TOS to evaluate its psychometric properties given its availability in the literature. However, the discrepancies raise questions about what might be measured across the two versions of the TOS.

Using EFA and ESEM, we identified a 16-item 2-factor solution congruent with theory and supported by the data. These two subscales yielded reliability coefficients that suggest good reliability and equaled or exceeded those published for other orthorexia measures. The strong internal consistencies of the TOS-16 subscales suggest that the TOS-16 could serve as a high-quality assessment tool, and thus, is appropriate for use in future research in U.S. samples.

Consistent with our hypothesis and past research [17], the TOS-16 Orthorexia Nervosa subscale correlated moderately with obsessive–compulsive and eating disorder symptoms, perfectionism, and emotional distress. This supports the convergent validity of the TOS-16. Although the TOS-16 Orthorexia Nervosa subscale significantly related to the EAT-26 (r = 0.61), the magnitude of the relationship was at least as modest (suggesting assessment of related but distinct constructs) as found with other promising measures (e.g., Gleaves et al. [41] found the EHQ and EAT-26 were significantly correlated; r = 0.79). The TOS-16 Healthy Orthorexia subscale was relatively unrelated to obsessive–compulsive disorder symptoms, perfectionism, and emotional distress except when examining residual effects as described below. Although there was a moderate correlation between TOS-16 Healthy Orthorexia and eating disorder symptoms, it was not as strong as the relationship between the subscale that assesses pathological preoccupation with healthy eating and eating disorder symptoms. Segura-García et al. [40] proposed an interesting theory that orthorexia nervosa represents a “residual” symptom of anorexia and bulimia that allows individuals to have a continued sense of control after treatment. Our results, in which both the TOS-16 subscales measuring pathological preoccupation with (i.e., Orthorexia Nervosa) and interest/engagement in (i.e., Healthy Orthorexia) healthy eating related to a measure of disordered eating, are not inconsistent with this possibility. If orthorexia nervosa is residual syndrome, it could interfere with eating disorder recovery. This possibility also suggests precaution against use of the TOS-16 Healthy Orthorexia subscale in isolation as an indicator of recovery among individuals with documented eating disorders (i.e., without a low TOS-16 Orthorexia Nervosa score, the TOS-16 Healthy Orthorexia score may reflect overlap with pathological preoccupation with healthy eating; see [61] for consideration of potential types of TOS responders).

Theorists contend that body image disturbance is absent or unimportant in orthorexia [7, 11], distinguishing it from anorexia and bulimia. Supporting discriminant validity of the TOS-16, the relationship between body image disturbance and disordered eating symptoms was significantly stronger than that between body image disturbance and TOS-16 subscale scores. The moderate association between the TOS-16 Orthorexia Nervosa subscale and body image disturbance does call into question the notion that body image disturbance is absent in orthorexia. Indeed, other scholars suggest that thinness, weight, and appearance concerns seem to be a motivator in orthorexia nervosa (e.g., [62]). Such findings suggest a need to further determine whether orthorexia nervosa is a subtype of an existing disorder or warrants its own diagnosis, and whether a lack of body image concerns should be included in differential diagnosis [62]. Of course, without a recognized diagnosis, studies on orthorexia nervosa tend to be correlational and samples may include individuals with other, established eating disorders (e.g., anorexia nervosa).

We found a moderate correlation between pathological preoccupation with healthy eating and health anxiety, and this relationship was significantly stronger than that between interest/engagement in healthy eating and health anxiety. This is consistent with the results of the known studies that previously examined the relationship between the two constructs [38, 39, 63, 64]. We also found that the TOS-16 subscales together accounted for a significant amount of variance in health anxiety after controlling for variance accounted for by eating disorder symptoms with each explaining a significant amount of unique variance. The finding that the relationship between a focus on healthy eating and health anxiety is significant and negative when controlling for measures of disordered eating and preoccupation with healthy eating suggest a residual relationship. In that case, holding pathological interest in healthy eating and eating disorder symptoms constant, those with an interest in engaging in healthy eating may be less likely to experience health anxiety. Taken together, these relationships suggest that the pathological focus on healthy eating is the factor that may reflect an attempt to manage health anxiety, but a general interest in healthy eating does not seem to relate to health anxiety or could be protective. These findings are some of the first to provide research support for the long-hypothesized connection between orthorexia and health anxiety, and the first to do so in the U.S.

Consistent with prior research, we found residual effects in which the TOS-16 Healthy Orthorexia related to reduced concerns with obsessive compulsive symptoms, perfectionism, negative affect, and body image disturbance, as well as being unrelated with disordered eating, when we held participants’ scores constant on the TOS-16 Orthorexia Nervosa subscale. This suggests that the TOS-16 Healthy Orthorexia subscale assesses, in part, the potential for an interest in healthy eating to serve as a protective factor. At the same time, we caution clinicians in using this subscale in isolation or as an indicator that symptoms of orthorexia nervosa are not problematic. Individuals who endorse items on the TOS-16 Healthy Orthorexia subscale also tend to endorse items on the TOS-16 Orthorexia Nervosa subscale (there is 25% shared variance between the scales in our sample), suggesting that for an important subgroup, the focus on healthy eating is a part of their pathological orthorexic eating.

Limitations and strengths

Limitations to this study warrant acknowledgment. First, our sample was relatively homogenous, limiting the generalizability of results. Furthermore, we did not assess socioeconomic status which can have a significant impact on one’s ability to access foods considered “healthy” and “pure.” Second, the cross-cultural validity of the TOS remains uncertain. We cannot be sure items mean the same thing to U.S. and Spanish participants. Our findings support the need for evaluating measurement invariance across samples. Third, this study cannot address causation between symptoms of orthorexia and other psychological conditions, such as health anxiety. In addition, it is impossible to know whether the relationships found between symptoms of orthorexia and other relevant constructs (e.g., perfectionism, emotional distress) were independent of other conditions such as anorexia or OCD.

The key strengths of our study were examining the back-translation and evaluating the TOS in an English-speaking sample in the U.S. which was similar to samples used in studies with the Spanish version. We found psychometric support for the theorized relationship between orthorexic eating, particularly orthorexia nervosa, and health anxiety.

Conclusions

We found support for use of a modified English version of the TOS with 16 items comprising two subscales. We found a pattern of moderate relationships between the TOS-16 subscale measuring pathological preoccupation with healthy eating with related constructs such as perfectionism, symptoms of eating and obsessive–compulsive disorders, and negative affect, whereas the measure of interest and engagement in healthy eating only related to eating disorder symptoms. Therefore, the TOS-16 has psychometric support and appears to be a promising measure for use in English-speaking U.S. samples. Thus, the TOS-16 warrants further research, and such research is especially needed with clinical populations, in U.S. samples.

Overall, our results also speak to the distinctiveness of the subscales and how orthorexic eating goes beyond interest/engagement in healthy eating—it represents a level of preoccupation that is unhealthy and associated with psychological distress and impairment. For example, orthorexia nervosa may impact one’s ability to engage socially or to allow for flexibility in one’s diet without subsequent shame, anxiety, or self-punishment. For those practicing in health-related fields, this is an important distinction when considering whether a person’s behavior may be problematic.

Ours was also one of the first studies to examine the relationship between symptoms of health anxiety and symptoms of orthorexia, and was the first known study to do so with an English-speaking, U.S. sample. We found a moderate association between health anxiety and orthorexia, and demonstrated that symptoms of orthorexia accounted for a significant proportion of the variance in health anxiety. Thus, it is possible that pathological preoccupation with healthy eating reflects, at least to some extent, anxiety about health.

What is already known on this subject?

The TOS produced moderately good, relative to other measures of orthorexia, psychometric properties in Spanish-, Portuguese-, and Arabic-speaking samples, including yielding a pattern of relationships with measures of related constructs (e.g., obsessive–compulsive and eating disorder symptoms) fitting with theory. In addition, scant research examined the theoretical relationships between orthorexia and somatic symptoms.

What does this study add?

The English-language TOS required slight modification in an English-speaking, U.S. sample. Our study provided support for use of the English-language TOS-16 as a valid and reliable measure of orthorexic symptoms in a U.S. sample. In addition, somatic symptoms related to orthorexic symptoms in our sample.