K. Halmi presents the salient components of a comprehensive service for eating disorders. In planning such a service, health policy makers need to understand the prevalence of eating disorders in the catchment area to be served. The size of an eating disorder service is not only dependent on the size of the catchment area, but also on the demographic characteristics of its population. Because eating disorders occur mainly among young females, it is important to know the proportion of this high risk group in the total population served. But characteristics of the catchment area itself are also important, such as the degree of urbanization. For example, in the Netherlands, the incidence of bulimia nervosa was found to be five times higher in cities than in rural areas 1. Also, most epidemiological data on eating disorders are derived from studies in Western countries, especially from Western Europe, because of the availability of comprehensive health registration systems and population statistics. With eating disorders, one has to be especially careful about applying the results from studies in Western countries for planning health care facilities in non-Western countries. In a comprehensive study on the Caribbean island of Curacao, the incidence of eating disorders among the minority mixedrace and white population was similar to the incidence in the United States and Western Europe, while no cases were found among the majority black population 2. In a meta-analysis providing one-year prevalence rates per 100,000 young females in the Netherlands at different levels of care 3, we reported that the rates for anorexia nervosa were 370 in the community, 160 at the primary care level, and 127 at the level of outpatient and inpatient mental health services. The corresponding rates for bulimia nervosa were 1,500, 170 and 87. These data confirm that only a minority of eating disorder patients in the community enters the mental health care system. This is particularly true for bulimia nervosa. These low rates of persons with eating disorders in care is likely to be the result of patient delay, due to denial or shame, and doctor delay, due to inadequate detection of these severe disorders. In a community study of Finnish twins 4, the lifetime prevalence of DSM-IV bulimia nervosa was 2.3%, but few of these women were identified by the health care system. The lifetime prevalence of DSM-IV anorexia nervosa was 2.2% and as high as 4.2% for broadly defined anorexia nervosa 5. Only half of the anorexia nervosa cases ascertained in this study had been detected in the health care system, though most of them had suffered from anorexia nervosa for several years. Another remarkable finding was that the lifetime prevalence of anorexia nervosa in young males in the Finnish community was 0.24%, also much higher than previously thought 6. Studies in the Netherlands and Sweden provide circumstantial evidence that early detection of anorexia nervosa is of major importance for a favorable outcome 7,8. Anorexia nervosa is still characterized by high lifetime mortality from both natural and unnatural causes 9. However, mortality among female patients with anorexia nervosa in hospital care in Sweden has decreased dramatically, which is probably related to the introduction of specialized care units 10. Eating disorder treatment programs are traditionally developed for patients with anorexia or bulimia nervosa. However, nowadays the majority of cases suffering from an eating disorder can only be classified in the DSM-IV category of eating disorders not otherwise specified (EDNOS). These include partial syndromes of anorexia or bulimia nervosa as well as binge eating disorder, a proposed new category in DSM-IV for research purposes 11. In a community study of eating disorders in Portugal, three quarters of young females with an eating disorder were classified as EDNOS 12. Even in outpatient settings, EDNOS cases account for an average of 60% of all cases 13. Although these patients are characterized by similar core cognitive psychopathology, they represent the least studied group of patients with eating disorders. Setting up an eating disorder service requires knowledge of its catchment area and of the epidemiology of eating disorders in general. But it is perhaps even more important to realize that most persons with eating disorders never come into treatment. In planning health services for persons with eating disorders, it is essential to take into account this enormous hidden morbidity of the disorders among the population and to develop methods to address it.