This study determined relationships among total energy intake, nutrient intake, body composition, exercise group status, and annual rates of change (slopes) in bone mineral density in 66 Caucasian premenopausal women (mean age, 34.4 +/- 2.7) taking calcium supplements. Body composition components measured by dual-energy X-ray absorptiometry included fat mass, soft tissue lean mass, and bone mineral density (g/cm2) of total body, spine (lumbar vertebrae 2-4), and three femur sites measured at baseline, 5, 12, and 18 mo. Nutrients were not significant variables in regression models predicting bone mineral density slopes (rates of change) at any femur site. The only significant variable in models predicting Ward's triangle bone mineral density slope was the initial fat mass and, for trochanter, exercise. Significant variables (P < 0.05) in models predicting total body bone mineral density slope included the initial fat mass and fat mass slope plus either vitamin A, carotene, fiber, magnesium, or phosphorus (R2 from 0.31 to 0.25) and fat mass slope plus sodium (R2 = 0.24). The significant variable in the model predicting L2-4 slope was energy intake (R2 = 0.17, P < 0.05). We conclude that nutrient intake, exercise, and body composition are related to bone mineral density rate of change and that relations among these variables vary by bone site.