Reduced bone density is observed in over half of women with anorexia nervosa (AN), in whom the risk of fracture is significantly increased even at a young age. It is unknown to what extent low bone density in AN differs from other conditions of premenopausal osteoporosis and is related to estrogen deficiency and/or other factors, such as nutritional status. We therefore investigated bone loss in nutritionally replete and nutritionally deplete amenorrheic women by comparing patients with AN (n = 30) to age-matched subjects with hypothalamic amenorrhea (HA; n = 19) in whom duration of amenorrhea, prior estrogen use, and age of menarche were comparable. Healthy, age-matched, eumenorrheic women were studied as a control group (NL; n = 30). Weight and nutritionally dependent factors including (body mass index, 20.7 +/- 0.3 vs. 16.7 +/- 0.3 kg/m2; P < 0.0001), insulin-like growth factor I (270 +/- 18 vs. 203 +/- 17 ng/mL; P < 0.01), percent body fat (26% vs. 19%; P < 0.0001), and lean body mass (38.7 +/- 1.1 vs. 34.3 +/- 0.8, P < 0.01) were significantly different between the HA and AN groups, respectively. The bone densities of the anterior-posterior (AP) spine, total hip, and total body measured by dual energy x-ray absortiometry were reduced in both amenorrheic groups compared to those in control subjects, but were significantly lower in women with AN than in those with HA. The t scores for AP spine and hip were -1.80 +/- 0.15 (AN), -0.80 +/- 0.22 (HA), and 0.28 +/- 0.19 SD (NL) for the AP spine and -1.62 +/- 0.17 (AN), -0.51 +/- 0.21 (HA), and 0.25 +/- 0.16 (NL) for the total hip, respectively (P < 0.01 for all comparisons). Among the amenorrheic subjects, duration of amenorrhea, age of menarche, and N-telopeptide were inversely correlated with bone density at all sites, whereas body mass index, insulin-like growth factor I, lean body mass, and fat intake were positively correlated with bone density at all sites measured. In multivariate regression analyses, bone density was most significantly related to lean body mass (P = 0.05 and P = 0.03 for the spine and hip, respectively), but not to the duration of amenorrhea or other indexes of estrogen status among patients with AN. In contrast, bone density of the lumbar spine was significantly related to weight and duration of amenorrhea among patients with HA. These data demonstrate that the severity of osteopenia in AN is greater than that in patients with HA and is critically dependent upon nutritional factors in addition to the degree or duration of estrogen deficiency itself. Lean body mass, independent of the duration or severity of estrogen deficiency, is an important predictor of bone loss among women with AN.