To study the effects of prolonged anorexia nervosa on bone density (BD) and to determine whether estrogen administration prevents bone loss in women with this disorder, 48 amenorrheic women with anorexia nervosa (mean age, 23.7 yr) were randomized to receive estrogen and progestin replacement (n = 22) or no replacement (n = 26). Clinical variables, biochemical indices, and spinal trabecular BD were measured every 6 months for a mean of 1.5 yr. Initial mean BD (130 +/- 27 mg K2HPO4/cm3, +/- 1 SD) was significantly (P < 0.001) less than normal (176 +/- 26 mg K2HPO4/cm3) and less than 2 SD below normal in 21 of the 48 women. Forty-four women completed the study (19 in the estrogen group and 25 in the control group). The mean duration of follow-up was comparable in the estrogen-treated (1.57 +/- 0.89 yr) vs. the control group (1.41 +/- 0.69 yr). The estrogen-treated group had no significant change in BD compared with the control group; however, there was a 4.0% increase in mean BD in patients with an initial ideal body weight of less than 70% who were treated with estrogen. In contrast, control patients with comparably low initial weight had a 20.1 % decrease in BD. Women in the control group with spontaneous resumption of menses, all of whom had an initial percent ideal body weight of greater than 70%, had a 19.3% increase in bone mass. It is concluded that: 1) estrogen replacement cannot prevent progressive osteopenia in young women with anorexia nervosa; 2) a subset of patients may have improved BD with estrogen and progestin administration depending on initial body weight; and 3) recovery from anorexia nervosa is associated with significantly improved BD.