Abstract
BACKGROUND: The factorial structure of the Eating Attitudes Test (EAT-26) has been found to be inconsistent across studies and samples. This study aimed to resolve inconsistencies in the factorial structure of the Arabic version of the EAT-26 by identifying the best-fitting model and test its measurement invariance across sexes and BMI categories in a large non-clinical Saudi sample. METHODS: 1,734 Saudi adults (M(age) 26.88 and SD 9.13), predominantly female, completed an online survey. Several existing models were tested (e.g., original 26-item three-factor model, second order 26-item three-factor model, 20-item four-factor model, and 16-item four-factor model) using confirmatory factor analysis (CFA). Fit indices including the CFI, SRMR, and RMSEA were used to identify the best-fitting model for Arabic version of the EAT-26. Multi-group confirmatory factor analysis (MGCFA) was used to test measurement invariance. RESULTS: The original three-factor model and two other common models demonstrated poor fit (e.g., CFI = 0.727; SRMR = 0.0911; RMSEA = 0.085 [90% CI 0.082-0.087] for the original three-factor model). Instead, a 16-item, four-factor structure [(Self-Perceptions of Body Weight), (Dieting), (Awareness of Food Contents), and (Food Preoccupation)] showed acceptable fit ([CFI = 0.904; SRMR = 0.0554; RMSEA = 0.073 [90% CI 0.068- 0.077]). Internal consistency was good (α and ω = 0.88), and measurement invariance was supported across sex (male and female) and BMI categories (underweight, normal weight, overweight, and obese). CONCLUSIONS: These findings underscore the need for culturally relevant validation of the EAT-26 among Arabic-speaking populations, as the revised factorial structure diverged from previously established models. Future research should further examine this revised 16-item, four-factor structure in clinical settings.