Our objective was to estimate the relationship between longitudinal change in BMD and fragility fractures. We studied 3635 women and 1417 men 50-85 yr of age in the Canadian Multicentre Osteoporosis Study who had at least two BMD measurements (lumbar spine, femoral neck, total hip, and trochanter) within the first 5 yr of the study and fragility fractures (any, main, forearm/wrist, ribs, hip) within the first 7 yr. Multiple logistic regression was used to model the relationship between baseline BMD, BMD change, and fragility fractures. We found that, among nonusers of antiresorptives, independent of baseline BMD, a decrease of 0.01 g/cm(2)/yr in total hip BMD was associated with an increased risk of fragility fracture with ORs of 1.15 (95% CI: 1.01; 1.32) in women and 1.34 (95% CI: 1.02; 1.78) in men. The risk of fragility fractures in subgroups such as fast losers and those with osteopenia was better estimated by models that included BMD change than by models that included baseline BMD but excluded BMD change. Although the association between baseline BMD and fragility fractures was similar in users and nonusers of antiresorptives, the association was stronger in nonusers compared with users. These results show that BMD change in both men and women is an independent risk factor for fragility fractures and also predicts fracture risk in those with osteopenia. The results suggest that BMD change should be included with other variables in a comprehensive fracture prediction model to capture its contribution to osteoporotic fracture risk.