The nineteenth-century comments by Dr. Louis-Victor Marce are relevant to both articles on eating disorders in this issue of The American Journal of Psychiatry. The rapid developments in molecular genetics are forcing the field of psychiatry to more specifically define the phenotypes of the various mental disorders. The eating disorders anorexia nervosa, bulimia nervosa, and variants thereof are latecomers to the arena of systematic investigations. The article by von Ranson et al. is a contribution to further defining the phenotype of bulimia nervosa.

The references by Marce to obstinacy in his anorexia nervosa patients are the first descriptions of the inflexible, rigid, and stubborn characteristics of the personalities of anorexia nervosa patients. Further case descriptions and systematic investigations in the twentieth century (2, 3) have confirmed Marce’s early description.

The clinical phenotype of bulimia nervosa patients is more complex. The bingeing and purging behaviors of these patients are suggestive of problems with impulse control and satiety regulation. A series of systematic studies with structured interviews (4–6) showed that about one-fourth to one-third of bulimia nervosa patients met the threshold criteria for a cluster B (impulsive) personality disorder. For example, the study by Braun et al. (6) showed that 26% of bulimia nervosa patients had a cluster B personality disorder diagnosis and 19% had a diagnosis of borderline personality disorder. These findings seem to confirm the phenotype of bulimia nervosa as an impulsive disorder. Neuroendocrine studies added further confirmation. Low levels of CSF 5-hydroxyindoleacetic acid (5-HIAA) with impulsive, suicidal, and aggressive behavior have been demonstrated by Asberg et al. (7) and Brown et al. (8). Animal studies have shown that impaired serotonergic function leads to overeating and obesity (9). Several studies have indicated deficient serotonergic function in bulimic patients. Jimerson et al. (10) showed that patients with more severe binge eating have lower CSF 5-HIAA levels than do control subjects. Brewerton et al. (11) demonstrated that bulimic patients have a blunted prolactin response to m-chlorophenylpiperazine (m-CPP), and McBride et al. (12) showed that these patients have a reduced prolactin response to the serotonin agonist fenfluramine. Unfortunately, the clinical phenotype in bulimia nervosa is more complex.

In the Braun et al. study (6), 26% of the bulimia nervosa patients had a diagnosis of cluster C (anxious) personality disorder, and 29% had an axis I anxiety disorder diagnosis. Thirteen percent actually had a comorbid diagnosis of obsessive-compulsive disorder (OCD). The study by von Ranson et al. in this issue further confirms an obsessive-compulsive aspect to bulimia nervosa. In that study the Yale-Brown Obsessive Compulsive Scale (an interviewer scale that measures obsessive-compulsive symptoms) was used to compare ill and recovered bulimic subjects with healthy women. Both ill and recovered women with bulimia nervosa had higher scores on the Yale-Brown Obsessive Compulsive Scale than the nonbulimic women. The main shortcoming of this study is one the authors do not acknowledge—the interviews both for psychiatric diagnoses and for the Yale-Brown Obsessive Compulsive Scale were given by the authors, who were not blind to the hypothesis of the study. The question remains: Do OCD symptoms contribute to the pathogenesis of bulimia nervosa? If OCD symptoms do not change with recovery from bulimia nervosa, a perfectly reasonable hypothesis is that they have nothing to do with the pathogenesis of bulimia nervosa.

In the nineteenth century Gull (13) and Marce (1) advised that the seriously ill anorexia nervosa patient cannot be treated advantageously if she remains in the midst of her own family. Both physicians recommended changing the "habitation" and entrusting the care of the patient to strangers. In the twentieth century a variety of behavioral and cognitive techniques have been developed to treat these patients in both an inpatient hospital and an outpatient clinic setting (14, 15). With the increasing refusal of managed care and insurance companies to support hospitalization of adequate length, day programs for eating disorders were developed (16). These programs are often used as a transition from inpatient treatment to outpatient therapy. Admission to a day program is usually based on economic necessities. There are no controlled studies available to determine criteria for transition from an inpatient to a day treatment program for anorexia nervosa. The retrospective chart review by Howard et al. in this issue yielded predictors for failure of day treatment that are similar to the predictors of long-term outcome for anorexia nervosa. Theander’s 30-year follow-up study (17) also showed that a duration of illness longer than 6 years predicts a poor outcome.

Of greater interest is the finding that patients forced to make the transition to the day hospital program with a body mass index less than 19 or a weight less than 90% of an established normal target weight had a significantly greater chance of treatment failure or readmission than did patients entering the day program within their target weight range. More than one-third of those entering the day program at a body mass index less than 19 experienced treatment failure. These data are very similar to the 5-year outcome data for hospitalized anorexia nervosa patients reported by Commerford et al. (18). This study compared the outcome of patients who met discharge criteria—achieving a target weight and maintaining it for 2 weeks while independently selecting foods from family-style table servings—with the outcome of patients who had to be discharged earlier because of lack of insurance coverage or patient unwillingness to comply with the treatment program. Those who met the discharge criteria were significantly closer to their ideal body weight at the 5-year follow-up than were those who had not achieved this goal during inpatient treatment. Forty-two percent of those discharged at their target weight were within 10% of that target weight at 5 years, whereas only 10% of those discharged below their target weight were within 10% of the target at 5 years. In addition, those who met the discharge criteria were more likely to be menstruating 5 years later.

Now at the end of the twentieth century the managed care and insurance companies are ignoring sound clinical observations over the past century and one-half. In the twenty-first century we need to add to the wisdom of the earlier centuries with data from prospective controlled studies to carefully establish criteria for discharge from inpatient treatment programs for the eating disorders.

Address reprint requests to Dr. Halmi, Weill Medical College of Cornell University, Cornell Medical Center–Westchester Division, 21 Bloomingdale Rd., White Plains, NY 10605; kah29@cornell.edu (e-mail).