A cross-sectional study of a random sample of 649 girls, aged 12-14 years (mean+/-SD: 12.9+/-0.6 years), in the Beijing area examined the relationship between diet and bone mineral status. Food and nutrient intakes over the past year were estimated by means of a semiquantitative food frequency questionnaire. Bone mineral content (BMC) and bone width (BW) at the distal one-third and one-tenth radius and ulna were measured by single-photon absorptiometry. Results showed Beijing pubertal girls had a low mean milk consumption (fresh and powdered milk, vitamin D-fortified milk, and yogurt) at 50 g/day (95% confidence interval [CI] 44-55 g/day whereas one-third consumed no milk at all. Mean calcium intake was 356+/-97 mg/day of which only 21% was provided by milk and milk products. Milk intake varied by region (rural, suburban, and urban: 9, 36, and 83 g/day, respectively, p<0.0005) as did the proportion of milk consumers in the three areas (30%, 64%, and 91%, p<0.0005). Bone mineral density (BMD) at the distal one-third and one-tenth radius and one-tenth ulna was positively associated with milk consumption (p<0.05). Multiple regression analysis of BMC on foods and nutrients as well as confounding factors, including weight, bone age, Tanner stage, and School Physical Activity Score (SPAS), showed that milk intake was the only dietary factor included in the models for BMC at the four bone sites measured. The model explained 54%-65% of the variation in BMC, and milk alone accounted for up to 3.2% of the variation. Milk was the only food group with significant partial correlation with BMC. SPAS, weight, bone age, and Tanner stage each accounted for a smaller variation in BMC (<1.8%). The results indicate that milk (presumably as an integrated source of nutrients) had a beneficial effect on bone mass of Beijing pubertal girls and was a better nutritional determinant of BMC than intake of any milk nutrient alone. Promotion of milk consumption should be considered for achieving optimal bone mass in this population group. (Bone 30:521-528; 2002) (C) 2002 by Elsevier Science Inc. All rights reserved.