Objective: To evaluate the construct validity of the Three-Factor Eating Questionnaire (TFEQ) in obese men and women.
Subjects: A total of 4377 middle-aged, obese subjects in the Swedish Obese Subjects (SOS) study.
Methods: The total sample was randomly split into two data subsets and psychometric testing was performed separately in each sample. Multitrait/multi-item analysis was conducted to test scaling assumptions and factor analysis was used to test the factor structure. Measures of mental well-being (MACL, HAD) were used for testing criterion-based validity.
Results: The Cognitive Restraint factor was consistently reproduced and scaling analysis demonstrated strong item-scale discriminant validity, while the item-scale convergent validity was unsatisfactory. The internal structure of the Disinhibition scale was weak. Most Disinhibition and Hunger items grouped in one global factor labeled Uncontrolled Eating. A third cluster containing items on Emotional Eating was also identified. The obtained three-factor structure was cross-validated and replicated across subgroups by gender, age and BMI.
Conclusion: The original TFEQ factor structure was not replicated. A short, revised 18-item instrument was constructed, representing the derived factors of Cognitive Restraint, Uncontrolled Eating and Emotional Eating. The most efficient items were used to boost both the convergent and discriminant validity of the scales.