We have previously demonstrated that girls with anorexia nervosa (AN) have higher levels of GH and cortisol and lower levels of estradiol than healthy adolescents. The effects of endocrine alterations on regional body composition in AN, however, have not been reported. We, therefore, enrolled 23 adolescent girls with AN and 20 healthy girls of comparable maturity in a study examining regional body composition. Levels of estradiol and IGF-I, as well as measures of GH and cortisol concentration (using cluster analysis of data obtained from frequent sampling q30' for 12 h overnight) were examined to determine hormonal determinants of regional body composition in adolescent girls with AN and controls. Girls with AN were followed for 1 yr and examined again at weight recovery (10% increase in body mass index) (n = 11). Percent trunk fat and trunk to extremity fat ratio (T/E fat) were significantly reduced in girls with AN compared with healthy adolescents (P = 0.001 and 0.01, respectively). Percent trunk lean mass and trunk to extremity lean mass ratio (T/E lean) were higher in AN than in controls (P = 0.01 and 0.009); percent extremity lean mass was lower in AN (P = 0.009). In healthy controls, total area under the curve (AUC) for GH correlated inversely with percent trunk fat and T/E fat (r = -0.66, P = 0.002 and r = -0.62, P = 0.003). Similar correlations were observed between other measures of GH concentration (mean and nadir) and percent trunk fat and T/E fat. No relationship was observed between GH concentration and regional lean mass or between cortisol concentration and regional body composition. In contrast, GH concentration did not predict regional body composition in adolescents with AN on regression analysis. However, nadir cortisol concentration correlated inversely with percent extremity lean mass (r = -0.49; P = 0.02) and positively with percent trunk lean mass and T/E lean (r = 0.48, P = 0.03; and r = 0.49, P = 0.02) in girls with AN. A similar trend was observed between other measures of cortisol concentration (mean cortisol and AUC) and percent trunk lean mass and T/E lean in AN. Trunk fat was lowest in girls with AN who had high GH but low cortisol levels (based on median values), whereas some preservation of trunk fat was observed in girls with low GH and high cortisol levels. Weight recovery occurred in seven of 11 girls with low GH and high cortisol values; however, only two of the nine girls with high GH and low cortisol recovered weight. High GH with lower cortisol levels may thus be a marker of greater severity of AN. Our results suggest that in healthy controls, GH concentration predicts regional body composition and favors a redistribution of body fat such that T/E fat ratio decreases. In AN, however, high levels of GH and cortisol have contrasting associations with fat mass. High cortisol levels in AN predict a redistribution of lean body mass such that extremity lean mass decreases. Further studies are necessary to better understand the implications of these data.