An empirical study of the classification of eating disorders.

OBJECTIVE: The nosology for eating disorders, despite having been extensively revised over time, may not capture the natural clustering of eating-related pathology as it occurs in general population samples.

METHOD: Detailed information about anorectic and bulimic behaviors was assessed through personal interviews of 2,163 Caucasian female twins from a population-based registry. Latent class analysis was applied to nine eating disorder symptoms to develop an empirically based typology. Demographic, comorbidity, personality, and co-twin diagnosis data were used to validate the resultant classes.

RESULTS: A six-class solution provided the best fit. One class displayed distorted eating attitudes without low body weight. Two classes demonstrated low weight without the psychological features of eating disorders. Three classes broadly resembled the DSM-IV classifications of anorexia nervosa, bulimia nervosa, and binge-eating disorder. For all classes, and especially for the three that reflected current diagnoses of eating disorders, monozygotic twins resembled one another much more in terms of class membership than did dizygotic twins.

CONCLUSIONS: The authors found within a community sample, and through an empirical method, classes of eating-related pathology that broadly resembled the current classifications of anorexia nervosa, bulimia nervosa, and binge-eating disorder. Additional classes were marked by either the psychological features of eating disorders or low body weight. Individuals in the three eating-disorder classes had similar personality profiles but displayed differences in symptom expression and co-twin risk for anorexia nervosa, bulimia nervosa, and obesity.