Visceral adiposity is a strong determinant of growth hormone (GH) secretion, and states of GH deficiency are associated with increased visceral adiposity and decreased lean body mass. The purpose of our study was to determine the sensitivity and specificity of different methods of assessing body composition [anthropometry, dual-energy X-ray absorptiometry (DXA), and computed tomography (CT)] to predict GH deficiency in premenopausal women and threshold values for each technique to predict GH deficiency, using receiver operator characteristic (ROC) curve analysis. We studied a group of 45 healthy lean, overweight, and obese premenopausal women who underwent anthropometric measurements (body mass index, waist and hip circumferences, skin fold thickness), DXA, CT, and a GH-releasing hormone-arginine stimulation test. ROC curve analysis was used to determine cutoff values for each method to identify GH deficiency. Visceral adiposity measured by CT showed the highest sensitivity and specificity for identifying subjects with GH deficiency with a cutoff of >9,962 mm(2) [area under the curve (AUC), 0.95; sensitivity, 100%; specificity, 77.8%; P = 0.0001]. Largest waist circumference showed high sensitivity and specificity with a cutoff of >101.7 cm (AUC, 0.89; sensitivity, 88.9%; specificity, 75%; P = 0.0001). When the ROC curves of visceral fat measured by CT and largest waist circumference were compared, the difference between the two methods was not statistically significant (P = 0.36). Our study showed that the largest waist circumference predicts the presence of GH deficiency in healthy premenopausal women with high sensitivity and specificity and nearly as well as CT measurement of visceral adiposity. It can be used to identify women in whom GH deficiency is likely and therefore in whom formal GH stimulation testing might be indicated.