Study design and setting
This cross-sectional study was conducted among students of the Saudi Electronic University (SEU), a national public institution offering blended online and in-person education across 15 regional campuses in Saudi Arabia. The campuses included Dammam, Al-Qassim, Riyadh, Abha, Tabuk, Madinah, Al-Ahsa, Jeddah, Jazan, Al-Jubail, Yanbu, Al-Qurayyat, Al-Ula, Hail, and Najran [16]. The study aimed to estimate the prevalence and predictors of eating disorders and disordered eating behaviors among SEU students, aligning with Saudi Arabia’s Vision 2030 objective to promote mental and psychological well-being [12, 17]. Data were collected between February 20 and March 14, 2025, from a geographically and demographically diverse student population representing all university branches.
Participants
All actively enrolled students at the Saudi Electronic University (SEU) aged 18 years or older were eligible to participate, regardless of gender, academic program, or campus location. Students were excluded only if they were under 18 years of age or declined to provide informed consent. At the time of data collection, approximately 25,000 students were enrolled across SEU’s 15 campuses. Although convenience sampling was employed, the final sample of 771 participants represented students from all university branches and academic programs, aged 18 to 49 years.
Sample size calculation
The required sample size was estimated using an anticipated prevalence of disordered eating behaviors of 25%, based on prior regional studies, with a 5% margin of error and a 95% confidence level. This calculation yielded a minimum required sample of 451 participants. To account for potential nonresponse and ensure sufficient power for multivariable analysis, a 20% oversampling was applied, increasing the target sample to approximately 565 participants. Additionally, sample adequacy for logistic regression was verified using the standard criterion of at least 10 outcome events per predictor variable, considering factors such as age, gender, marital status, prior diagnosis, BMI, recent weight loss, and regional variation. With a prevalence exceeding 25%, the final achieved sample size of 771 participants exceeded the minimum requirement for both prevalence estimation and multivariable modeling. All calculations were conducted using OpenEpi version 3.01 (Emory University, Atlanta, GA, USA) [18].
Post-stratification weights were applied to improve representativeness and adjust for sample bias based on the official gender distribution of SEU’s student population. Weighting followed standard procedures to correct the overrepresentation of female students and produce gender-balanced estimates for both prevalence and regression analyses [19].
Sampling and recruitment
The study employed a convenience sampling approach. An Arabic-language online questionnaire, created using SEU’s official Google Forms platform, was distributed via institutional email lists to all currently enrolled students across the university’s 15 campuses. Eligible participants were students aged 18 years or older. Multiple reminder emails were sent during the three-week data collection period (February 20–March 14, 2025) to encourage participation. A total of 771 students completed the survey, exceeding the recommended minimum for both prevalence estimation and multivariable regression analysis. Because the exact number of eligible students who received or opened the survey link was unknown, a formal response rate could not be accurately calculated. Participation was anonymous and voluntary, with no incentives provided. To improve representativeness and correct for sampling bias, post-stratification weights based on the official gender distribution of SEU’s enrolled student population were applied during analysis [19].
Survey instrument
Disordered eating attitudes and behaviors were assessed using the Eating Attitudes Test-26 (EAT-26) and its Behavioral Questions (BQ) section—two standardized and widely used screening tools for identifying risk of eating disorders [20]. The EAT-26 evaluates attitudinal risk, while the BQ assesses behavioral symptoms such as binge eating, purging, and excessive exercise. These tools were selected for their strong psychometric properties, cross-cultural applicability, and extensive use in both community and student populations [21].
The EAT-26 measures attitudinal risk for eating disorders across three subscales—dieting, bulimia, and oral control—using a 6-point Likert scale ranging from Always to Never. Items 1–25 were scored as follows: Always = 3, Usually = 2, Often = 1, and Sometimes, Rarely, and Never = 0, while item 26 was reverse-coded (Never = 3, Rarely = 2, Sometimes = 1, Often/Usually/Always = 0). The total score ranges from 0 to 78, with a cutoff of >20 indicating high risk for eating disorders, in line with standard scoring guidelines. Internal consistency for the EAT-26 in this study was strong (Cronbach’s α = 0.87). While Cronbach’s α provides an estimate of internal consistency, it should be interpreted in light of the multidimensional nature of behavioral checklists such as the BQ [22].
The Behavioral Questions (BQ) section, part of the original EAT-26 developed by Garner et al. [23], includes four items assessing disordered eating behaviors such as binge eating, self-induced vomiting, use of laxatives or diet pills, and excessive exercise for weight control. These items were scored using the same 6-point Likert format as the EAT-26. Total scores ranged from 0 to 12, with higher scores indicating more frequent engagement in maladaptive behaviors. Because the BQ represents a checklist of heterogeneous behaviors rather than a single latent construct, internal consistency (α) was not interpreted as a measure of reliability. Instead, BQ data were analyzed descriptively, and any non-zero response (“Often,” “Usually,” or “Always”) on any item was classified as indicative of disordered eating behavior. This inclusive operationalization was adopted to capture a broader range of subclinical behaviors that may be underreported in conservative cultural settings.
Both instruments were administered in Arabic-adapted versions, developed through forward–backward translation and pilot testing to ensure linguistic and cultural appropriateness for Saudi university students. The Arabic version of the EAT-26 has previously demonstrated strong psychometric validity in Middle Eastern populations [24], and minor contextual refinements were introduced for this study to enhance clarity and relevance.
Additional self-reported variables included gender, age, marital status, academic program, SEU campus (city), prior eating disorder diagnosis, and anthropometric measures. Body Mass Index (BMI) was calculated using self-reported height and weight, collected via direct survey questions (“Please enter your current weight in kilograms” and “Please enter your height in centimeters”), and computed using the standard formula: weight (kg) ÷ height (m²). Participants also reported their highest and lowest lifetime weights and recent weight loss (≥ 10 kg in the past six months).
Survey piloting (internal validity)
Although the Arabic EAT-26 has been previously validated, minor linguistic refinements were made following pilot testing to ensure clarity and cultural appropriateness for Saudi university students without altering item meanings. Recent reviews have emphasized the need for validated Arabic measures and more robust epidemiological data on eating disorders within the Saudi population [25]. Prior to formal data collection, the Arabic survey was pilot-tested among a convenience sample of 20 SEU students representing both genders, various academic programs, and multiple campuses. The pilot aimed to evaluate item clarity, cultural relevance, and response comprehension for all components, including the EAT-26 and Behavioral Questions (BQ) sections. Participants completed the survey and provided structured feedback regarding language clarity, item relevance, and any points of confusion. Based on this feedback, minor wording modifications were implemented to improve sentence flow and remove ambiguities. Specifically, select Likert-scale options were rephrased for more precise differentiation, and specific psychological or medical terms were simplified to align with common terminology used among young adults in Saudi Arabia. No major content revisions were required, indicating the adapted instrument’s strong face validity and acceptability. The finalized version incorporated these minor adjustments to enhance participant comprehension and response accuracy.
Handling missing data
All submitted questionnaires were screened for completeness prior to analysis. Surveys with more than 20% missing responses across core sections (EAT-26 or BQ items) were excluded. For participants with partial responses within scales, domain-level scores were computed if at least 80% of items were completed, using prorated scoring based on available data. Missing responses for independent variables (e.g., BMI, marital status) were minimal (
BMI values were screened for plausibility using established cutoffs ( 65 kg/m² considered implausible). Data were rechecked for potential entry errors, including reversed height and weight units or misplaced decimals. Implausible cases were corrected where verifiable and excluded if uncertain. Despite these checks, BMI values remained right-skewed due to a small number of extreme but plausible values; therefore, BMI was summarized using both mean (SD) and median (IQR) to provide a more accurate representation of central tendency. All data cleaning and verification procedures were conducted in SAS version 9.4 (SAS Institute, Cary, NC, USA), with double-entry verification performed for categorical variables.
Ethical considerationss