Introduction

Meat avoidance might act as a socially acceptable means to restrict food, disguising pathological eating [e.g., 1]. It is unclear, however, which components of meat avoidance relate to problematic eating behaviours. Research simultaneously suggests disordered eating is higher in vegetarians than omnivores [1], semi-vegetarians versus vegans [2] and omnivores [2, 3], and omnivores versus vegans [4]. These contradictory findings could be explained by grouping meat avoiding subtypes together (i.e., vegetarians with vegans), masking eating behaviour differences specific to vegans. Studies in which meat avoiders are grouped together also have limited representativeness due to small sample sizes [2, 3] and the recruitment of female-only [3, 5] or university student samples [3, 5]. Thus, there is a clear need to investigate whether veganism is associated with disordered eating in a more representative sample, which this study addresses.

Risk for engaging in disordered eating might lie in people’s motivation for being vegan [i.e., animal welfare, health, religion; [4]. Specifically, health-motivated vegans may be attempting to disguise pathological eating behaviours more so than animal-welfare motivated vegans. This link has been partially supported in vegetarian samples [6, 7], but not vegans [2, 4]. Vegetarians who selected weight concerns as their main dietary motivation reported significantly higher levels of dietary restraint than those who selected ethical, religious, or taste preferences [6, 7]. However, Heiss and colleagues [4] found, in a sample of 358 vegans, no significant differences among primary motivations to going vegan and measures of disordered eating. This combination of findings supports the need to further characterise the association between meat avoidance and diet motivations.

As the prevalence of veganism grows [8], so does the need for research examining the association between veganism and disordered eating. This study aimed to establish whether vegans displayed greater levels of disordered eating than omnivores in an online sourced sample, and whether the relationship between diet-type and disordered eating is affected by gender. We also re-tested the idea that diet motivations influence disordered eating [4], with health-motivated vegans expected to have greater levels of disordered eating than vegans motivated by other reasons.

Method

This study was preregistered on Open Science Framework (https://osf.io/sz7re/, https://osf.io/qz8e6/).

Participants

Participants were recruited from Amazon Mechanical Turk (MTurk; > 100 HITS approved, > 95% approval rate). MTurk provides rapid and cost-effective access to a wide age range of adults [9]. To prevent bot and server farmers, participants had to complete a Captcha, a simple arithmetic question presented as an image, and score at least 6/10 on an English proficiency test to enter the survey [10]. To ensure data quality, we included three attention checks, an open-ended question, and a duplicate question within the survey. These steps ensured we received reliable data, demonstrated by our sample showing high reliability on all eating disorder questionnaires [9].

Vegan sample

Participants were classified as vegan if they are self-reported to routinely eliminate red meat, poultry, fish/seafood, dairy products, and eggs from their diet. Participants who did not meet our definition of veganism were exited (n = 215), debriefed, and paid $US0.20. Although participants saw a study advertisement called Health and Lifestyle Preferences in Vegans, we reduced the likelihood of participants saying they were vegan purely to be eligible for the study by assessing veganism based on certain food avoidance, rather than self-classification. A further nine participants were excluded: six did not complete all eating disorder questionnaires and three reported non-binary gender. The final vegan sample comprised 110 participants (50% female). Most were aged 25–34 years (46.4%) and mean BMI was 24.4 (SD = 4.5). Participants resided in the United States (78.2%), Canada (17.3%), the UK (2.7%), and Australia (1.8%), and varied in length of vegan adherence (e.g., 4–12 months: 17.3%, 1–2 years: 24.6%, 3–4 years: 23.6%).

Omnivore sample

Participants were classified as omnivore if they did not self-report eliminating any animal products from their diet. We excluded participants who reported eliminating an animal product or fruit, legumes, and wheat or grains from their diet (i.e., common food intolerances; n = 73) to ensure we did not capture higher levels of disordered eating caused by avoiding foods for allergies. We excluded a further 17 participants: 15 did not complete all eating disorder questionnaires and two failed all attention checks. The final omnivore sample comprised 118 participants (53.4% female). Most were aged 25–34 years (35.5%) and mean BMI was 26.2 (SD = 7.4). Participants resided in the United States (91.5%), Canada (7.6%), and the UK (0.9%). The vegan and omnivore sample sizes provided sufficient power to detect a medium effect size (Cohen’s f = 0.25) with 0.80 power and p < 0.05 [11].

Materials

Motivations for diet adherence

Participants were selected from a list of motivations [e.g., animal welfare, health; 2, 4] for diet adherence. Motivation options matched between samples, except we exchanged “animal welfare” for “muscle building” and added “no specific reason” for the omnivore sample. Participants who selected more than one motivation were asked to select their primary motivation. We classified participants, based on their primary motivation, as “health followers” if they indicated “my health”, “weight control”, or “muscle building” as their primary reason for diet adherence, “ideological followers” if they indicated “animal welfare”, “the environment”, or “my spiritual beliefs”, “other” if they indicated another reason (e.g., food insensitivity/intolerance), or “no reason” [4].To ensure the samples were comparable, we categorised motivations the same for both samples, despite these categories often not applying to omnivores.

Eating attitudes test-26 (EAT-26)

The EAT-26 [12] measures dieting, bulimia and food preoccupation, and oral control. Participants respond to 26 items using a 6-point scale (from always to never). Reliability (α) was high (vegan = 0.85, omnivore = 0.84).

Eating disorder examination questionnaire (EDE-Q)

The EDE-Q [version 6.0; 13] assesses restraint, eating concern, shape concern, and weight concern over the past 28 days along 28-items. Reliability (α) was excellent (vegan = 0.95, omnivore = 0.94).

Eating pathology symptoms inventory (EPSI)

The EPSI [14] measures disordered eating over the past four weeks. Participants indicate the frequency that 45-items have applied to them from 0 = never to 4 = very often. Lending to no global score, reliability along the eight subscales (α) ranged from acceptable to excellent (vegan = 0.66–0.90, omnivore = 0.72–0.91).

Procedure

Following approval from the Monash University Human Research Ethics Committee, participants responded via an online survey to demographic characteristics, diet motivations, and the EAT-26, EDE-Q, and EPSI in randomised order. Following a debriefing, participants were compensated $US1.30.

Results

Diet type and disordered eating

Table 1 displays scores on the EAT-26 and EDE-Q global, by diet type and gender.

Table 1 EAT-26 and EDE-Q global scores, by diet type and gender
Full size table

To determine whether diet type and gender influenced disordered eating, measured by EAT-26 and EDE-Q global scores, we conducted 2 (diet type: vegan, omnivore) × 2 (gender: female, male) ANOVAs. For the EAT-26, vegans scored higher than omnivores; a main effect of diet type, F(1, 224) = 8.94, p = 0.003, ηp2 = 0.038. Males and females did not significantly differ (F(1, 224) = 0.04, p = 0.84, ηp2 = 0.000), nor did diet type and gender interact, F(1, 224) = 0.03, p = 0.87, ηp2 = 0.000. These results demonstrate that EAT-26 scores differ by diet type, irrespective of gender. However, there was no significant difference between diet type for the EDE-Q, F(1, 224) = 2.36, p = 0.13, ηp2 = 0.01. EDE-Q scores significantly differed by gender with females scoring higher, F(1, 224) = 12.10, p = 0.004, ηp2 = 0.036, but there was no diet type × gender interaction, F(1, 224) = 0.33, p = 0.57, ηp2 = 0.001. Thus, EDE-Q scores differ by gender, irrespective of diet type.

We subsequently conducted three binomial logistic regressions using EAT-26, EDE-Q, and EPSI subscales (see Supplementary Information for descriptive statistics) to predict the probability of diet type (vegan, omnivore; Table 2). For the EAT-26, the overall model was statistically significant, χ2 (224) = 9.54, p = 0.02, Cox and Snell R2 = 0.04, with no specific subscale significantly predicting a vegan diet. For the EDE-Q, the overall model was statistically significant, χ2 (223) = 11.08, p = 0.03, Cox and Snell R2 = 0.05. Specifically, eating concern was a significant positive predictor of a vegan diet. For the EPSI, the overall model was also significant, χ2 (219) = 34.29, p < 0.001, Cox and Snell R2 = 0.14 (Table 2). Body dissatisfaction was a significant negative predictor, and binge eating and cognitive restraint were significant positive predictors of a vegan diet.

Table 2 EAT-26, EDE-Q, and EPSI predictor coefficients for the model predicting vegan diet type
Full size table

Diet motivations and disordered eating in the vegan sample

Finally, to examine whether health motivated vegans to engage in more disordered eating than other types of vegans, we ran two 3 (primary motivations: ideological, health, other) × 2 (gender: male, female) ANOVAs on the EAT-26 and EDE-Q. Motivations did not impact EAT-26 or EDE-Q scores among vegans (see Supplementary Information for results). A one-way ANOVA revealed diet length (< 2 years, 2–10 years, and > 10 years) was also not associated with disordered eating as measured by the EAT-26 (F(2,107) = 0.42, p = 0.66, ηp2 = 0.008) or the EDE-Q (F(2,107) = 0.64, p = 0.53, ηp2 = 0.012).

Discussion

This study aimed to establish whether disordered eating differs between vegans and omnivores using samples with similar numbers of males and females. Partially in line with hypotheses, vegans displayed higher levels of disordered eating than omnivores, but only when measured by the EAT-26 and not the EDE-Q. As expected, vegans showed higher levels of cognitive restraint than omnivores. Consistent with prior research [2, 4], diet motivations among vegans did not predict disordered eating, suggesting any disordered eating risk may be related to veganism itself. Below we offer several interpretations for these findings.

Vegans reported higher disordered eating when measured by the EAT-26, but not the EDE-Q. This inconsistency could be explained by the EDE-Q’s stronger emphasis on weight and shape concerns, compared to the EAT-26’s emphasis on dieting and food concerns. For example, the EAT-26 asks respondents to rate the extent they display self-control around food or feel that others pressure them to eat. Such items could be capturing vegan-motivated food choices, rather than eating disorder-motivated food choices, thus overinflating vegans EAT-26 scores. In line with this idea, EDE-Q eating concern and EPSI cognitive restraint were significantly related to an increased probability of being vegan, while the EPSI body dissatisfaction was significantly related to a decreased probability of being vegan. This finding suggests vegans may feel more positive and compassionate towards their bodies than omnivores [15, 16].

We unexpectedly found that EPSI binge eating predicted increased probability of being vegan. It may be possible that vegans believe they can consume higher quantities of vegan foods which are perceived to be “healthier”. It may also be possible that for some people, a vegan diet does not always provide a sense of fullness and thus the person may progress to a binge eating episode [17]. Another potential explanation for our finding is that vegans may have a lower threshold for what they consider to be binge eating behaviours, compared to omnivores, owing to their more considered eating habits (i.e., higher levels of cognitive restraint). As far as we are aware, the association between binge eating and veganism is a novel finding that requires further investigation.

To the best of our knowledge, this study is the first on disordered eating and veganism to include a comparable sample of men and women. Previous research on disordered eating in vegans has generally used female-only samples [e.g., 1, 5] or small proportions of males [e.g., 2, 3]. While we acknowledge our sample does not reflect eating disorders being more common in women than men, a supplementary aim of this study was to establish whether gender differences interacted with diet type to predict disordered eating. In addition to ensuring sufficient sample size for these analyses, the equal gender distribution extends understanding of disordered eating in men, which is typically understudied. Counter to the idea that women always show higher disordered eating than men, our findings showed no gender difference in EAT-26 scores, only on the EDE-Q. This pattern could be driven by females exhibiting more dissatisfaction with their shape, typically driven by a stronger sociocultural emphasis on appearance and thinness [18]. However, further research is required to understand diet type and disordered eating in men, as well as gender-diverse people.

Strengths and limits

We recruited a large online sample with an even male/female distribution, which extends our understanding of disordered eating in men. We attempted to reduce the likelihood of participants responding purely to be eligible for the study by assessing veganism based on certain food avoidance, rather than self-classification. In doing so, we provide a methodological advancement to this field which typically relies on self-classification [4]. With strengths in mind, the cross-sectional design limits our understanding of the development of disordered eating in vegans. It remains unclear whether veganism increases the risk of developing an eating disorder, or whether developing an eating disorder increases the chances of becoming vegan. Future research should track the progression of eating-related behaviours before and after the transition to veganism. Second, this study is based on self-report data that is prone to socially desirable responding. Finally, participants were recruited from MTurk and do not represent the general population [9]. This study should be replicated in a community-based sample including vegans and vegetarians to determine how disordered eating varies across meat-avoiders.

Clinical implications

Our results suggest that commonly used eating disorder measures in clinical settings (e.g., EDE-Q) include items, particularly around food exclusion and following food rules, which may unfairly pathologise vegans. Future research should develop additional subscales or entirely novel measures to cater to the growing vegan community [8]. In the meantime, clinicians should ask additional probing questions during initial assessments to ascertain how vegan patients arrived at their answers on these commonly used measures, particularly around food rules. Such questions would prevent the patient’s eating habits from being over-pathologised by following a vegan diet.

Conclusion

Veganism has permanently transitioned from the fringe to mainstream [8]. Through relying on high cognitive restraint and food preoccupation, veganism might predispose individuals to develop an eating disorder. Alternatively, the inclusion of cognitive restraint and food preoccupation items in eating disorder measures might be over-pathologising rates of disordered eating in vegans. There is a clear need to better understand the nuances of the relationship between veganism and pathological eating behaviours to help provide evidence-based diagnoses and treatments for vegans with a suspected eating disorder.

What is already known on the subject?

It has been suggested that veganism may serve as a socially acceptable means to restrict food intake, therefore hiding pathological eating behaviours. Previous research typically groups vegans with other meat avoiders (e.g., vegetarians), potentially masking unique vegan risk factors.

What this study adds?

We found greater disordered eating in vegans than omnivores on some measures. Diet motivations did not predict greater disordered eating in vegans. Our data suggests the relationship between disordered eating and veganism may be driven by elevated levels of eating concern and cognitive restraint. Cognitive restraint and food preoccupation items in eating disorder measures might be over-pathologising rates of disordered eating in vegans.