A prospective cohort study of 1098 postmenopausal Japanese-American women evaluated the relationship between baseline non-spine fractures and new (incident) spine fractures. At the baseline examination in 1981, prevalent non-spine fractures were ascertained by interview, and prevalent spine fractures by radiograph. Bone mass measurements of the distal radius, proximal radius, calcaneus (1981), the lumbar spine (1984) were obtained and repeated at 1- to 2-year intervals. Women with existing non-spine fractures have a threefold greater risk of subsequent spine fractures, independent of bone mass, and independent of the known association between prevalent spine fractures and subsequent spine fractures. Women with both a prevalent non-spine fracture and low bone mass (50th percentile or lower) have an eightfold greater risk of new spine fractures compared with women above the 50th percentile of bone mass and no prevalent fractures. In addition to low bone mass, both prevalent spine fractures and prevalent non-spine fractures are strong risk factors for subsequent spine fracture. These data suggest that not all osteoporotic risk factors are expressed via bone mass, and that other, unmeasured risk factors, such as bone quality defects, may explain these results. In clinical terms, women with both prevalent fractures and low bone mass should be recognized as being at extremely high risk, and treatment potency should be commensurate with this level of risk.