The objective of the study assess the relationship between bone mineral density (BMD) loss over time and fracture incidence in postmenopausal women. This is a posthoc analysis that includes women from the placebo group of two large randomized controlled trials having assessed the efficacy of a new anti-osteoporotic drug. BMD was assessed every 6 months during 3 years at the lumbar spine, the femoral neck and the total proximal femur. Vertebral fractures were assessed using a semiquantitative method. Hip fractures were based on written documentation. All patients received calcium and vitamin D. In the present study that included 1,775 patients (with complete data at baseline and after 3 years), the logistic regression analysis, adjusted for covariates, showed that 3-year change in lumbar BMD was not statistically associated with the new vertebral fractures after 3 years. However, femoral neck and total proximal femur BMD changes was statistically correlated with the incidence of new vertebral fractures (P < 0.001). When considering change in BMD after the first year of follow-up, a decrease in total proximal femur BMD was statistically associated with an increase in the incidence of new vertebral fractures during the last 2 years of follow-up (P = 0.048). The 3-year change in femoral neck and total proximal BMD was statistically correlated with the incidence of hip and fragility fracture after 3 years (all P < 0.001). In this elderly osteoporotic population receiving calcium and vitamin D, a decrease in hip BMD after 1 or 3 year of follow-up, is associated with an increased risk of fracture incidence. However, spine BMD changes do not influence vertebral fracture incidence.