BACKGROUND Osteoporosis risk in middle-aged women is twofold greater than that in men, and the difference increases with age. Gender differences in bone mineral density, estimated rates of bone loss, and usefulness of markers of bone metabolism for predicting bone density have not been well described in healthy elders aged 65 and above. The purpose of this cross-sectional analysis was to describe associations of bone mineral density at the hip, spine, and whole body with age, serum osteocalcin, and urinary N-telopeptide crosslinks of Type I collagen in healthy elderly men and women. METHODS A total of 1,087 healthy adults (273 men and 814 women) aged 65 to 87 years were enrolled in a collaborative study at 3 sites: Tufts University (Boston, MA), University of Connecticut Health Center (Farmington, CT), and Creighton University (Omaha, NE). Bone mineral density (BMD) at three regions of the hip, the lumbar spine, and whole body was determined by dual-energy x-ray absorptiometry. Serum osteocalcin was measured by immunoassay, and measurement of N-telopeptide crosslinks (Ntx) in urine was made using an enzyme-linked radioimmunoassay (ELISA). RESULTS Among women, the age-related decline in BMD at all non-spine skeletal sites was significantly different from zero, with the largest decline seen at the femoral neck (-.0038 g/cm2/y, p < .001) and the smallest at the trochanter of the hip (-.0023 g/cm2/y, p = .03). Among men, the changes at all non-spine sites were not significant. In both sexes, spine BMD tended to increase with age (men, +.0045 g/cm2/y, women, +.0003 g/cm2/y). Serum osteocalcin and urinary Ntx were inversely related to BMD at all skeletal sites, but the weakest associations were observed at the spine. Individuals whose values of both osteocalcin and Ntx were in the lowest quartiles of the respective sex-specific distributions had mean femoral neck BMD that were 11% higher than individuals with marker values in the highest quartiles. CONCLUSIONS These findings suggest that age-related decreases in BMD may vary by gender and skeletal site. Determinations of osteocalcin and N-telopeptide crosslinks at a single point in time may potentially be used as indicators of current bone status, particularly at non-spine skeletal sites.