Low bone mineral density (BMD) is correlated with increased fracture risk. Whether greater BMD increases induced by osteoporosis drugs are related to greater decreases in fracture risk is controversial. We analyzed the relationship between BMD change and fracture risk in postmenopausal osteoporotic women receiving antiresorptive treatment. The analysis combined data from three pivotal risedronate fracture end-point trials. Women received risedronate (n = 2047) or placebo (n = 1177) daily for up to 3 yr. The BMD and vertebral radiographs were assessed periodically during 3 yr. The estimated risk of new vertebral fracture was compared between patients whose BMD increased and those whose BMD decreased. Risedronate-treated patients whose BMD decreased were at a significantly greater risk (p = 0.003) of sustaining a vertebral fracture than patients whose BMD increased. The fracture risk was similar (about 10%) in risedronate-treated patients whose increases in BMD were < 5% (the median change from baseline) and in those whose increases were >/= 5% (p = 0.453). The changes in lumbar spine BMD explained only 18% (95% confidence interval [CI], 10%, 26%; p < 0.001) of risedronate's vertebral fracture efficacy. Although patients showing an increase in BMD had a lower fracture risk than patients showing a decrease in BMD, greater increases in BMD did not necessarily predict greater decreases in fracture risk.