RESUMO
OBJECTIVE: To evaluate the Chinese Eating Attitudes Test (EAT-26) in screening patients with anorexia nervosa (AN) and bulimia nervosa (BN) in Hong Kong. METHOD: A consecutive series of Chinese patients with BN (N = 67) and typical (fat phobic; N = 65) and atypical (nonfat phobic; N = 44) AN underwent clinical assessment and completed the EAT-26. Results were compared with those of Chinese female undergraduates (N = 646). RESULTS: The mean EAT scores for bulimic and typical AN patients were significantly higher than those of undergraduates, but the scores of atypical AN patients were anomalously low. The dieting and bulimia factor, scores and body mass indices entered the classification tree. When compared with using the conventional EAT-26 cutoff, the misclassification rate for typical AN, atypical AN, and BN changed from 41.4% to 52.3%, 88.6% to 43.2%, and 23.9% to 29.9%, respectively. DISCUSSION: Using the EAT-26 in the conventional manner would lead to an underestimate of atypical AN in community surveys. Complementary use of a classification tree improved the prediction of atypical AN, but the EAT-26 remains a suboptimal screening instrument for the community epidemiological study of AN.