There is a current debate about the extent to which antifracture efficacy of antiresorptive drugs are related to changes in bone mineral density (BMD). In vitro studies show that most of the variability in bone strength is related to BMD, and prospective studies have shown that low BMD is an important predictor of fracture risk. It seems that higher levels of bone turnover are also associated with increased fracture risk. Over the short term, a reduction in activation frequency or resorption depth would lead to fewer (and/or shallower) resorption sites and refilling of existing sites initially. There is also evidence that inhibiting resorption allows bone to respond to mechanical demands, preferentially thickening critical trabeculae, and this may help compensate for reduced connectivity. Each of these mechanisms would increase BMD and would disproportionately improve bone strength. Over the long term, maintaining bone mass and preventing loss of structural elements would result in progressively greater differences in BMD and fracture risk over time, relative to untreated women. The conceptual model predicts that both the short- and long-term antifracture efficacy of antiresorptive drugs will depend on the extent to which treatment can increase and maintain BMD. To examine this issue, we compiled data from clinical trials of antiresorptive agents and plotted the relative risk of vertebral fractures against the average change in BMD for each trial. The confidence intervals are large for individual trials, and there was substantial variability in antifracture efficacy at any given level of change in BMD. Overall, however, trials that reported larger increases in BMD tended to observe greater reductions in vertebral fracture risk. Poisson regression was used to quantify this relationship. The model predicts that treatments that increase spine BMD by 8% would reduce risk by 54%; most of the total effect of treatment was explained by the 8% increase in BMD (41% risk reduction). These findings are consistent with the short-term predictions of the conceptual model and with reports from randomized trials. The small but significant reductions in risk that were not explained by measurable changes in BMD might be related to publication bias, measurement errors, or limitations of current BMD technology.