This study examined correlations among anthropometric parameters, body composition, bone parameters and predictive factors of bone mass in adolescent girls with different body fat percentages (%fat). A total of 129 females aged 15-18 years were categorized into 3 groups using %fat-for-age at the 50th and 95th percentiles as cutoff points (normal, over, and excess %fat groups). We recorded anthropometric data and measured the speed of sound at the tibia and radius using quantitative ultrasound. Dual-energy x-ray absorptiometry (DXA) was used to measure body composition and bone parameters, including bone mineral density (BMD), bone mineral content (BMC), and the BMD-z-score (z-score) in the lumbar spine (LS) and whole body (WB). These parameters were compared among the 3 groups using bivariate and multivariate correlation analyses. There were strong correlations among all anthropometric parameters, body composition, and DXA in the over %fat group. Lean parameters strongly correlated with LS and WB in the normal %fat group, whereas both lean mass (LM) and fat mass (FM) were positively correlated with BMC in the excess %fat group. The predictive factors of bone mass differed among the groups, as follows: lean body mass was predictive of BMD and BMC at both sites in the normal and over %fat groups; LM and body weight were predictive of LS-BMC and WB-bones, respectively, in the over %fat group; and FM was predictive of WB-bones in the excess %fat group. Body fat and waist circumference were negative predictors of bone mass. Predictive factors of bone strength appear to depend on the amount of body fat in adolescent girls.