This is the second-year part of a two-phase prospective epidemiological study as well as a community case-controled study. The aims of the second year study are firstly, to investigate the clinical course and 1-year outcome of eating disorders in the community, as well as the prospective risk factors of eating disorders; secondly, a case-control design involving with other psychiatric disorders from the same study population is used for risk factors study.

The study subjects are Taiwanese gifted dance high school students.Parental informed consent for nonpartipation was delivered 3 days before study started. Screening questionnaires for eating disorders and general psychiatric disorders included EAT(Eating Attitude Test), BITE(Bulimic Investigatory Test, Ediburgh BITE) and CHQ-12(Chinese Health Questionnaire). Other self-report questionnaires include perfectionism subscale of Eating Disorder Inventor-I (EDI), Rosenberg Self-esteem Scale (RSE), Eyseck Personality Inventory, Pubertal Develoment Scale, and body figure test. Subjects who received interviews at the first-year study and those whose scores are above the threshold of the screening questionnaires need to receive interviews at 2nd phase. The criteria of interview threshold included EAT≧20+ BMI<18.5Kg/m2, symptom scale of BITE ≧15 and/or severity scale of BITE≧5, and CHQ ≧4. The content of interview included structured diagnostic interview (SCID-IP 1995) as well as risk factors assessment. Each time a subject with bulimia nervosa was recruited, one matched control subject with no mental illness general psychiatric diagnosis was sought from the same class or school. Potential psychiatric control subjects were identified from their scores on the CHQ. The psychiatric control subjects were required to have no current or past eating disorders. The contents of risk factor assessments include personal vulnerability and family problems.

The relationships between individual risk factors and case status were assessed by univariate analysis. Conditional logistic analysis was used for multivariate analysis of risk factors prediction for eating disorders. To assess the relative importance of different types of exposure, the relationship between case status and exposure in each domain were first assessed by univariate analysis and then in multivariate stepwise regression analyses.