Osteopenia is a typical finding in patients suffering from anorexia nervosa. Unfortunately, available longitudinal studies are limited by a relatively short follow-up period. Therefore cross-sectional long-term followup studies may help to determine both the outcome of this bone lesion and variables that influence its subsequent development. Of an initial 66 consecutive patients with anorexia nervosa, 51 (77.3%) could be further evaluated. After an average of 11.7 years following first admission, cross-sectional measurements of lumbar and proximal radial bone mineral density (BMD) were performed. The ability to predict BMD using variables obtained from anamnestic and clinical data was then determined by multiple-regression analysis. The BMD of both radial and lumbar bone in anorexic patients with poor disease outcome (as defined by the Morgan-Russell general outcome categories) deviated by -2.18 and -1.73 SD (Z score), respectively. In patients with a good disease outcome lumbar BMD was significantly less reduced compared with radial BMD (-0.26 versus -0.68 SD). Variables reflecting estrogen deficiency and nutritional status in the course of the disease, that is, relative estrogen exposure (for lumbar BMD) and years of anorexia nervosa (for radial BMD), allowed the best prediction of BMD. A marked reduction in cortical and trabecular BMD in anorexic patients with poor disease outcome suggests a higher risk of fractures in these patients. Furthermore, the finding of a persistently reduced cortical and a slightly reduced trabecular BMD, even in patients with good disease outcome, suggests that a recovery of trabecular BMD might be possible, at least in part. Recovery of cortical bone, if possible at all, seems to proceed more slowly.