Cross-sectional studies on the effects of depot-medroxyprogesterone acetate (DMPA) on bone mineral density (BMD) have been controversial. The present longitudinal cohort study on 59 Chinese women over a period of 3 years has shown that their annual rate of bone loss at 3 sites (0.44% in lumbar spine, 0.40% in neck of femur, 1.05% in Ward's triangle) was substantially less than the projected values (1.1% in lumbar spine, 2.3% in neck of femur, 3.5% in Ward's triangle) in a cross-sectional study that had demonstrated a significant reduction in BMD in DMPA users than the non-user population. The trochanter BMD measurement did not show the projected annual bone loss of 2.4%. The rate of bone loss is probably non-linear, with a rapid loss in the first 5 years and a leveling off afterwards. The duration of DMPA use was not significantly correlated with the rate of bone loss. Multiple linear regression analysis demonstrated that age and body mass index were significant variables in modeling the rate of bone loss in the lumbar spine and neck of femur, but not in the trochanter and Ward's triangle areas. The Z scores also suggested a retardation in bone loss with time and potentially due to the effect of progesterone in decreasing bone turnover that is similar to the situation in postmenopausal women. The present data provide another aspect of reassurance to the long-term use of DMPA.