Given the limited research on disordered eating among sexual and gender minority (SGM) individuals–particularly in Eastern European contexts–the present study aimed to clarify how gender identity, gender expression, and key psychosocial mechanisms contribute to disordered eating behaviors in a Romanian sample. Specifically, it compared SGM and non-SGM individuals and examined the roles of gender expression, minority status, social appearance anxiety, and thin-ideal internalization in shaping maladaptive eating patterns. Building on prior literature linking gender identity, body image concerns, and sociocultural appearance norms, two mediation models were tested to examine the psychological mechanisms contributing to disordered eating in this understudied population.

Based on these objectives, the following hypotheses were developed:

This study included 270 Romanian participants (M_age = 24.7 years, SD = 6.22; range: 18–63). Eligibility criteria required participants to be at least 18 years old and residing in Romania at the time of the study. Participants were recruited through convenience sampling between October 2024 and March 2025 using both online and offline strategies designed to ensure diverse representation within the Romanian sexual and gender minority (SGM) community. Online outreach included posts on social media platforms (e.g., Facebook, Instagram), as well as targeted messages shared by individuals with visibility in LGBTQ+ and activist circles. Notably, an environmental and human rights activist with a large Romanian following promoted the study via Instagram. Offline recruitment was conducted in two LGBTQ+ -affirming venues in Bucharest, where printed flyers containing the survey link and QR code were displayed. Members of local drag performance houses were also contacted directly and encouraged to share the study within their networks. Eligibility criteria included being at least 18 years old and residing in Romania at the time of participation. All participants provided informed consent prior to beginning the survey, in compliance with the General Data Protection Regulation [GDPR; Regulation (EU) 2016/679]. Participants were informed about the purpose of the study, the voluntary nature of their participation, and how their data would be processed and stored.

The anonymous survey was administered via Google Forms (Google LLC) and responses were stored securely on a password-protected Google Drive account. According to Google’s published policies, data storage and transfer comply with GDPR cybersecurity and privacy requirements. No compensation (monetary or otherwise) was provided for participation. The study protocol adhered to ethical standards for research involving human participants, including voluntary participation, anonymity, informed consent, and the right to withdraw at any time without penalty.

Of the 271 individuals who accessed the survey, one was excluded due to missing data on gender identity and sexual orientation. The final sample included diverse sexual orientations and gender identities: homosexual (N = 94; 34.8%) and bisexual (N = 93; 34.4%), followed by heterosexual (N = 51; 18.9%). Less frequently reported orientations included pansexual (N = 25; 9.3%), queer (N = 3; 1.1%), and asexual/demisexual (N = 3; 1.1%). One participant (0.4%) reported being in the process of exploring their sexual orientation and identified as questioning.

Regarding gender identity, the most frequently reported identity was cisgender woman (N = 156; 57.8%), followed by cisgender man (N = 75; 27.8%). Non-binary individuals represented 8.5% of the sample (N = 23), and transgender individuals 4.8% (N = 13), which included transgender women (N = 3; 1.1%) and transgender men (N = 10; 3.7%). Less frequently reported gender identities included gender fluid (N = 1; 0.4%), agender (N = 1; 0.4%), and genderqueer man (N = 1; 0.4%).

In terms of sex assigned at birth, 186 participants (68.9%) reported being assigned female at birth, and 84 (31.1%) reported being assigned male. Two participants (0.7%) reported being born with intersex characteristics. The mean age at which they became aware of these characteristics was 8 years. Both identified as cisgender women.

A total of 8 participants (3%) reported undergoing hormonal treatment to feminize or masculinize physical characteristics. One transgender woman reported feminizing hormone therapy, while the remaining seven reported masculinizing therapy and identified either as transgender men (N = 6) or non-binary (N = 1).

Based on participants’ gender identity, sexual orientation, and intersex status, two mutually exclusive groups were created: the sexual and gender minority (SGM) group and the non-SGM group. The non-SGM group included only cisgender, heterosexual individuals who were not born with intersex traits. The SGM group included all other participants - those identifying as sexual or gender minorities and/or born with intersex characteristics. The SGM group included 225 participants (83.3%), while the non-SGM group included 45 participants (16.7%). The mean age in the SGM group was 24.0 years (SD = 4.47; range: 18–52), with 59.5% having completed higher education. The non-SGM group had a mean age of 28.4 years (SD = 10.9; range: 18–63), and 64.4% reported completed higher education. Within the non-SGM group, 26 participants were cisgender women (M = 26.0 years, SD = 9.24; range: 21–63), and 19 were cisgender men (M = 27.8 years, SD = 13.0; range: 18–55).

An a priori power analysis was conducted using G*Power software (Faul et al., 2009) to determine the minimum sample size required for a Mann–Whitney U test. Assuming a medium effect size (d = 0.5), a significance level of α = 0.05, and a statistical power of .95, the recommended sample size was 184 participants.

The study employed a cross-sectional observational design, with data collected at a single point in time without manipulation of variables. As several variables violated the assumption of normality, non-parametric and resampling-based methods were used for statistical analysis. Specifically, Mann–Whitney U tests were conducted to compare groups, and mediation analyses were performed using the percentile bootstrap method. For the planned mediation analyses using percentile bootstrap methods, prior literature was consulted. According to Fritz and MacKinnon (2007), a minimum of 558 participants is required to detect a small mediation effect (a = b = 0.14) with 80% power, while 71 participants are sufficient to detect medium effects (a = b = 0.39). The final sample included 270 participants, of whom 225 were SGM individuals, providing sufficient power to detect medium-to-large effects in the mediation models (Table 1).

Given the violation of the normality assumption for EAT-26 scores (Shapiro–Wilk W = 0.890, p < 0.001), the non-parametric Mann–Whitney U test was used to compare disordered eating levels between the SGM and non-SGM groups. A one-tailed test was conducted. Results showed that EAT-26 scores were significantly lower in the non-SGM group (Median = 8) compared to the SGM group (Median = 12), U = 4007, p = 0.014 (one-tailed), 95% CI [−∞, −1.00]. This result supports the research hypothesis, suggesting that disordered eating behaviors tend to be more elevated among SGM individuals compared to non-SGM individuals.

To assess the preliminary relationships among key variables within the SGM group, bivariate correlations were computed (Table 4). As expected, disordered eating (EAT-26) showed strong, statistically significant positive correlations with both social appearance anxiety (SAAS; r = 0.54, p < 0.001) and thin-ideal internalization (THIINA; r = 0.66, p < 0.001), supporting their role as potential mediators. SAAS and THIINA were also positively correlated (r = 0.55, p < 0.001). Gender variance (GVS) was weakly and positively correlated with SAAS (r = 0.18), but this association did not reach the conservative significance threshold of p < 0.001. Similarly, Femininity was weakly correlated with THIINA (r = 0.13) and with EAT-26 (r = 0.17), though neither correlation achieved statistical significance. Overall, these patterns provide preliminary support for the hypothesized indirect pathways, while underscoring the need for further testing through mediation analyses.

A percentile bootstrap mediation analysis was conducted to test whether social appearance anxiety (SAAS) mediated the relationship between gender variance (GVS) and disordered eating (EAT-26) among SGM participants (see Figure 3). As shown in Table 5, the total effect of gender variance on disordered eating was not statistically significant (β = −0.0428, z = −0.642, p = 0.521, 95% CI [−0.1301, 0.0656]). However, the direct effect of gender variance on disordered eating, in the absence of the mediator, was significant and negative (β = −0.1465, z = −2.615, p = 0.009, 95% CI [−0.1981, −0.0140]). The indirect effect through social appearance anxiety was significant and positive (β = 0.1036, z = 2.670, p = 0.008, 95% CI [0.0101, 0.1325]); the confidence interval for this effect did not include zero, indicating a statistically significant mediation. Gender variance was significantly associated with social appearance anxiety, although the effect was small (β = 0.1813, z = 2.766, p = 0.006, 95% CI [0.0258, 0.3462]). In turn, social appearance anxiety was strongly associated with disordered eating (β = 0.5715, z = 10.206, p < 0.001, 95% CI [0.3001, 0.4641]). These findings indicate a pattern of associations consistent with partial mediation, with the indirect and direct effects operating in opposite directions, effectively canceling each other out in the total effect.

Another percentile bootstrap mediation analysis was conducted to examine whether thin ideal internalization (THIINA) mediated the relationship between femininity and disordered eating (EAT-26) among SGM participants (see Figure 4). As shown in Table 6, the total effect of femininity on disordered eating was statistically significant (β = 0.1709, z = 2.60, p = 0.009, 95% CI [0.0334, 0.3480]). However, the direct effect was not significant (β = 0.0856, z = 1.72, p = 0.086, 95% CI [−0.0295, 0.2110]), and the indirect effect via thin ideal internalization approached the threshold for significance (β = 0.0852, z = 1.95, p = 0.051, 95% CI [−0.0094, 0.2020]). Thin ideal internalization showed a strong positive association with disordered eating, with a large effect size (β = 0.6549, z = 13.14, p < 0.001, 95% CI [0.3641, 0.5080]). Femininity was a marginally significant predictor of thin ideal internalization, with a small effect size (β = 0.1302, z = 1.97, p = 0.049), although the corresponding confidence interval included zero (95% CI [−0.0219, 0.4400]).

One key limitation of this study lies in the use of the EAT-26, an instrument not specifically designed or validated for diverse populations such as sexual and gender minorities (SGM). Like many legacy measures for eating disorders, the EAT-26 was developed to assess the symptomatology of cisgender heterosexual women from higher socioeconomic backgrounds. As a result, it may not fully capture the diverse expressions of disordered eating among SGM individuals. For instance, the measure tends to prioritize weight loss as a core indicator, which may not encompass other motivations such as body masculinization, gender congruence, or increased muscularity.

Moreover, the EAT-26 does not assess compensatory behaviors such as laxative use, diet pill consumption, or substance use, behaviors that may be particularly salient among SGM populations. Some research (Marshal et al., 2008; Warren et al., 2012; Demant et al., 2016; Hill et al., 2022) has highlighted that individuals in these groups are more likely to engage in substance use for appetite suppression and body control, including energy drinks, smoking, or appearance-and performance-enhancing drugs (APEDS). These behaviors may co-occur with or be shaped by experiences of gender dysphoria or sociocultural pressures around body presentation, yet remain unmeasured by conventional instruments.

Relatedly, the present study did not assess muscle dysmorphia, which may be highly relevant in LGBTQ+ populations. Although the EAT-26 and THIINA address thinness-oriented concerns, they fail to account for muscularity-driven body image disturbances. Muscle dysmorphia, is conceptually linked to eating disorders through shared features such as perfectionism, compulsive exercise, and emotional avoidance. Emerging literature suggests this construct is particularly salient among gay and bisexual men and may be underrecognized in standard assessments (Fabris et al., 2020).

Although the study was conducted in Romania, where the majority of the population identifies as ethnically Romanian, the lack of data on participants’ ethnic backgrounds remains a limitation. This omission precludes a more nuanced intersectional analysis, particularly with respect to minority ethnic groups such as Roma individuals, who may face additional layers of marginalization. Likewise, although educational attainment was recorded, no indicators of income or occupational status were collected. Given the established links between food insecurity, financial strain, and disordered eating–especially among LGBTQ+ populations (Choi et al., 2019)–the absence of detailed socioeconomic data is a notable constraint.

Language and cultural adaptation posed further challenges. Feedback from participants highlighted concerns with the Romanian translation of key terms, such as the rendering of thin in the THIINA scale. While subţire was used to avoid connotations of weakness (slab) or athleticism (suplu), some respondents found it vague or unfamiliar. These nuances might have impacted the way items were interpreted, potentially affecting response validity.

The sample’s demographic composition also limits the generalizability of findings. In both SGM and non-SGM groups, cisgender women were overrepresented, and the distribution of SGM subgroups was uneven. The majority of SGM participants were assigned female at birth, and the most commonly reported gender identity was cisgender woman. This skew is consistent with known biases in online and LGBTQ+ -friendly recruitment channels. As such, findings may disproportionately reflect the experiences of certain subgroups while underrepresenting others, such as transgender men or non-binary individuals with low visibility.

Finally, the statistical power of the mediation analyses may have been insufficient to detect small indirect effects. While 225 SGM participants were included in these models, this number falls short of sample sizes typically required to detect subtle mediation effects–especially when the independent variable exerts a weak influence on the mediator but the mediator strongly predicts the outcome. Non-significant or marginally significant mediation effects, such as those involving thin-ideal internalization, may therefore be influenced by limited power, raising caution about interpreting these results as a lack of association. Nevertheless, observed effect sizes suggest underlying patterns that warrant further investigation.

Future research should adopt intersectional and longitudinal designs to clarify how gender identity, body image concerns, and disordered eating interact over time, particularly within marginalized populations. In Eastern European contexts, such as Romania, where LGBTQ+ individuals face unique cultural, political, and social challenges, these dynamics may unfold differently than in Western societies. The mechanisms linking gender variance, appearance anxiety, and eating pathology remain insufficiently understood, and context-specific factors such as stigma, identity integration, and access to gender-affirming care must be more thoroughly investigated.

To address current measurement gaps, future research should consider using or adapting assessment tools that have been previously applied in studies with sexual and gender minority (SGM) populations, even if not originally developed for these groups. These instruments should move beyond thinness-focused models to capture a broader range of appearance-related concerns, such as muscularity, gender congruence, and the use of appearance- and performance-enhancing substances. For example, the Eating Disorder Examination Questionnaire (EDE-Q) has been employed in studies involving LGBTQ+ participants (Nagata et al., 2019) and may offer a more nuanced assessment than the EAT-26. Additionally, constructs like muscle dysmorphia and gender-specific drives for body modification, particularly relevant for transgender and non-binary individuals, could be better assessed using tools such as the Muscle Dysmorphic Disorder Inventory (MDDI; Zeeck et al., 2018). Incorporating such instruments may help capture the diverse experiences of SGM individuals more accurately.

Recruitment strategies should aim for more balanced and representative samples. Stratified or randomized sampling across diverse geographic, socioeconomic, and gender identity groups would enhance generalizability. Future studies should actively include underrepresented subpopulations, such as transgender men, rural SGM individuals, ethnic minorities, or those with lower levels of identity disclosure, who are often excluded from convenience samples.

Finally, evolving beauty standards, particularly those amplified by social media, warrant deeper investigation. Future studies should explore how platform-specific content (e.g., Instagram’s fitness culture vs. TikTok’s body positivity trends) interacts with gender identity and internalized appearance ideals to shape eating behaviors. Qualitative and mixed-methods approaches could provide richer insight into lived experiences and support the development of culturally responsive, inclusive prevention and intervention strategies tailored to Eastern European SGM populations.