The definition of anorexia nervosa was revised for the Fifth Edition of the Diagnostic and Statistical Manual (DSM-5). We examined the impact of these changes on the prevalence and prognosis of anorexia nervosa. In a nationwide longitudinal study of Finnish twins born 1975-1979, the women (N = 2825) underwent a 2-stage screening for eating disorders at mean age 24. Fifty-five women fulfilled DSM-IV criteria for lifetime anorexia nervosa. When we recoded the interviews using DSM-5 criteria, we detected 37 new cases. We contrasted new DSM-5 vs. DSM-IV cases to assess their clinical characteristics and prognosis. We also estimated lifetime prevalences and incidences and tested the association of minimum BMI with prognosis. We observed a 60% increase in the lifetime prevalence of anorexia nervosa using the new diagnostic boundaries, from 2.2% to 3.6%. The new cases had a later age of onset (18.8 y vs. 16.5, p = 0.002), higher minimum BMI (16.9 vs. 15.5 kg/m(2), p = 0.0004), a shorter duration of illness (one year vs. three years, p = 0.002), and a higher 5-year probability or recovery (81% vs. 67%, p = 0.002). Minimum BMI was not associated with prognosis. It therefore appears that the substantial increase in prevalence of anorexia nervosa is offset by a more benign course of illness in new cases. Increased diagnostic heterogeneity underscores the need for reliable indicators of disease severity. Our findings indicate that BMI may not be an ideal severity marker, but should be complemented by prognostically informative criteria. Future studies should focus on identifying such factors in prospective settings. (C) 2016 Elsevier Ltd. All rights reserved.