Anorexia nervosa (AN) is associated with low bone density in adolescents and adults. Hypercortisolemia has been reported in adults with this disorder and has been hypothesized to be a factor in bone loss. However, the secretory dynamics of cortisol in adolescents with AN and the contribution of alterations in cortisol secretion to bone metabolism in AN have not been examined. We examined the dynamics of cortisol secretion by Cluster and deconvolutional analysis in 23 girls with AN and 21 healthy adolescents of comparable age and maturity. Cortisol sampling was performed every 30 min for 12 h overnight. Twenty-four-hour urinary free cortisol (UFC) and creatinine (cr) were obtained for all subjects. The surface area (SA) of the subjects was calculated. Markers of bone turnover (type 1 procollagen, osteocalcin, and N-telopeptide) were examined. Subjects with AN were prospectively followed over 1 yr, and those who recovered weight (defined as a 10% increase in body mass index) were again studied. On Cluster analysis, girls with AN had significantly higher mean cortisol (8.6 +/- 2.0 vs. 5.9 +/- 1.1 microg/dl; P < 0.0001), nadir cortisol (5.5 +/- 2.3 vs. 3.4 +/- 1.2 microg/dl; P = 0.0008), valley mean cortisol (7.0 +/- 2.7 vs. 4.7 +/- 1.5 microg/dl; P = 0.001), peak amplitude (12.6 +/- 4.4 vs. 7.8 +/- 3.0 microg/dl; P = 0.0004), peak area (652 +/- 501 vs. 340 +/- 238 microg/dl; P = 0.02), and total area under the curve (6112 +/- 1467 vs. 4117 +/- 802 microg/dl; P < 0.0001) than healthy adolescents. On deconvolutional analysis, the frequency of nocturnal secretory bursts (7.0 +/- 1.2 vs. 5.8 +/- 1.3 /12 h; P = 0.001), total nocturnal pulsatile cortisol secretion (69.3 +/- 14.7 vs. 53.9 +/- 11.1 microg/dl; P = 0.0003), and total cortisol secretion (89.6 +/- 18.8 vs. 71.2 +/- 17.6 microg/dl; P = 0.002) were significantly higher in girls with AN than in healthy controls. Cortisol half-life trended higher in girls with AN. However, basal cortisol secretion and approximate entropy did not differ between the groups. UFC/cr and UFC/cr.SA were significantly higher in girls with AN than in controls [0.050 +/- 0.028 vs. 0.036 +/- 0.017 (P = 0.04) and 0.035 +/- 0.020 vs. 0.023 +/- 0.012 (P = 0.03)]. Six of 23 girls with AN had UFC/cr.SA values that were more than 2 sd above those in healthy controls. An inverse correlation was noted between measures of cortisol concentration as well as pulsatile secretion and measures of nutritional status (body mass index, fat mass, leptin, insulin, and IGF-I). An oral glucose load suppressed cortisol levels in healthy adolescents, but not in AN patients. Weight recovery was associated with a significant decrease in the number of secretory bursts. In girls with AN, strong inverse correlations were noted between levels of cortisol (mean, nadir, and total area under the curve) and levels of markers of bone formation (C-terminal propeptide of type 1 procollagen and osteocalcin). Conversely, in healthy controls, cortisol values did not predict levels of markers of bone turnover. Adolescent girls with AN have significantly higher serum cortisol concentrations and UFC/cr.SA values than healthy adolescents. This increased cortisol concentration is a function of increased frequency of secretory bursts, resulting in increased pulsatile secretion. Hypercortisolemia appears to be a direct consequence of undernutrition and is associated with a decrease in markers of bone formation. Therefore, high cortisol values in AN may contribute to the low bone density observed in adolescents with this disorder by decreasing bone formation.