RESUMEN
OBJECTIVE: Studies of the construct validity of the Eating Disorder Examination (EDE) show mixed results and none have included samples of adults with a sole diagnosis of either anorexia nervosa or bulimia nervosa. Thus, we examine the EDE's construct validity in a Danish clinical sample of adolescents and adults with eating disorders. METHOD: Confirmatory factor analyses of the four-factor model indicated by the original four subscales and subsequent ad hoc exploratory factor analyses were performed in a sample of patients with eating disorders (N = 1,586) divided into five subsamples based on age and diagnosis: (a) adolescents with anorexia nervosa, (b) women with anorexia nervosa, (c) women with bulimia nervosa, and women with atypical versions of (d) anorexia nervosa, and (e) bulimia nervosa. RESULTS: The four-factor model was not confirmed in these five subsamples. Subsequent exploratory factor analyses did not reveal a single model fit for all five groups. Rather, eating-disorder symptoms can be understood through a one-, two-, or three-factor model depending on the subsample. DISCUSSION: The four-factor model indicated by the original EDE subscales cannot be supported. The four subscales should be used with great care, if used at all, in trying to understand specific symptoms of eating disorders.