Results: There were more D8/17-positive individuals among those with PANDAS anorexia nervosa (81%) than among the comparison subjects (12%). The subjects with PANDAS anorexia nervosa had a higher percentage of D8/17+ cells (mean=27.1%, SD=17%) than the comparison subjects (mean=5.3%, SD=7.4%). Conclusions: A larger study is needed to determine whether D8/17 serves as a marker for susceptibility to a type of anorexia nervosa. (Am J Psychiatry 2002; 159:1430-1432) A norexia nervosa is a serious disorder with a mortal- ity rate among the highest of the psychiatric disorders, poorly understood pathogenesis, and no definitive treat- ment (1). Clinical and research observations led to the hy- pothesis that a postinfectious process, similar to that in rheumatic fever, may cause or exacerbate certain cases of anorexia nervosa (2). When the pathogen is group A β- hemolytic streptococcus, these disorders are called pedi- atric autoimmune neuropsychiatric disorders associated with streptococcus (PANDAS). Swedo et al. (3) identified a group of children with ob- sessive-compulsive disorder (OCD) and/or tics who had PANDAS characteristics. These children often expressed the lymphocyte B marker D8/17, apparently a trait marker for susceptibility to rheumatic fever and Sydenham's chorea (4). OCD and anorexia nervosa may be related, as evidenced by phenomenology, comorbidity, neurotransmitters, and CNS functional metabolism (5). Many patients with ano- rexia nervosa have obsessions (about food and weight) and compulsions (overexercise and eating behaviors). This, along with observation of cases of anorexia nervosa whose onset appeared temporally related to infection, led to the hypothesis of a type of anorexia nervosa similar to PANDAS OCD (2). The purpose of this project was to study the possibility of an infection-triggered type of anorexia nervosa and to determine whether D8/17 identifies PANDAS anorexia nervosa. We hypothesized that patients with PANDAS an- orexia nervosa would have more D8/17+ cells than com- parison psychiatric patients with no eating disorder or PANDAS characteristics. Method