To the Editor:

The article by Walter H. Kaye, M.D., et al. (1) on comorbid anxiety disorders in eating disorders is a considerable contribution to this research area. However, other studies on this topic (2–4) were not referred to. Important information for several points of discussion is raised by these unmentioned studies.

We (2) found that 71% of 271 current subjects with eating disorders had lifetime comorbidity with at least one anxiety disorder (64% for Dr. Kaye et al.). The proportion of generalized anxiety disorder that was reported by Dr. Kaye et al. (10%) appears lower than our findings (anorexia nervosa: 45.6%, bulimia nervosa: 31.4%; all current). Converse to their finding, the eating disorders in our study group were all current, which may have affected the comorbidity rates. Given that subjects with a lifetime eating disorder (who are not currently ill) have a ratio of having no anxiety disorder to having an anxiety disorder significantly higher than for people who are currently ill (1), we wonder whether this discrepancy reflects a diagnostic bias instead of a bias of recall or a weak association with recovery. Indeed, high levels of anxiety and depressive symptoms (due to denutrition [5] or other factors, such as duration of illness, social disability, or preexisting trait anxiety) could lead to excessive current diagnoses of anxiety disorder.

Obsessive-compulsive disorders (OCDs) were nearly twice as frequent in the study of Dr. Kaye et al. (41%) as in our study (anorexia nervosa: 24.1%) and that of Iwasaki et al. (3). Although we did not use a symptomatic scale and thus may have missed some cases, the study by Iwasaki et al. suggests that it may rather be because the participants in the study by Dr. Kaye et al. "came from enriched pedigrees," leading to higher rates of comorbidity than in the community (1) or in other eating disorders groups.

Dr. Kaye et al. found that 66% of their comorbid cases and 42% of their entire study group had an onset of at least one anxiety disorder before the onset of an eating disorder. Our rates were, respectively, 50% and 33% (2). Although OCD and generalized anxiety disorder usually preceded the onset of an eating disorder in the study by Dr. Kaye et al., we observed the inverse pattern (2). This discrepancy could be due to some memory bias (i.e., people who have durably been characterized by obsessive-compulsive traits may have difficulties in remembering the exact time of the onset of OCD) or to a selection bias. Knowing that unusually precocious age at the onset of OCD is a risk factor for the development of eating disorders (6) and that the group selection of Dr. Kaye et al. was specific, we wonder whether the rate of early-onset OCD in their group of "enriched pedigrees" might have been unusually high.

Dr. Kaye et al. reported no differences in the rates of OCD between the patients with anorexia nervosa and those with bulimia nervosa, converse to another of their studies (4) in which they observed higher rates of OCD in patients with anorexia nervosa than in those with bulimia nervosa. In another of our studies (7), current diagnoses of agoraphobia and OCD were significantly more frequent in patients with anorexia nervosa than in those with bulimia nervosa. These contradictory results stress the need for developing further research on the comorbidity between eating disorders and anxiety disorders.