In women with osteoporosis, each 1% improvement in spine BMD (by DXA) is expected to reduce vertebral fracture risk by about 4%. However, randomized trials of antiresorptive agents show that 1 to 6% improvements in spine BMD reduce vertebral fracture risk by 35 to 50%. Less 20% of the decreased spine fracture risk produced by alendronate or raloxifene be explained by improvement in spine BMD. The discrepancy is even greater during the first year or two of treatment when 1 to 4% improvements in BMD are associated with 65-68% decreases in spine fracture risk. Bisphosphonates continue to increase BMD but the reduction in fracture risk wanes to 20 to 45%. DXA underestimates the change in bone density of spinal trabecular bone and this might explain part of the discrepancy between expected and observed reductions in spine fracture risk. Even more accurate measurement of BMD would not explain the rapid onset and later waning of effect despite gradually increasing BMD. The biomechanical effects inhibiting bone resorption could explain the early onset but not the waning effectiveness. The waning effectiveness of antiresorptives raises concerns that prolonged inhibition of remodeling may weaken bone by allowing microdamage to accumulate. The effect of drugs on nonspine fracture risk is more complex and cannot be predicted from changes in DXA BMD. For example, Beck showed that long-term users of estrogen increase section modulus vs. nonusers with a net increase in section modulus and predicted femoral neck strength despite losing about 0.4% per year in femoral neck BMD. PTH reduces spine fracture risk and this effect is more completely explained by improvement in spine BMD. This suggests that sustaining the increased BMD produced by PTH may maintain long-term reductions in fracture risk.