Although bone mineral density (BMD) is causally related to fracture, the burden of fractures attributable to low BMD is not clear. In this study, we estimated the fraction of different fracture types occurring in old people that can be attributed to low BMD.

The study involved 2264 women and 1339 men aged 50 years and older, whose bone health has been continuously monitored for up to 20 years. During the follow-up period, the incidence of total fracture, hip fracture, clinical vertebral fracture, and wrist fracture was ascertained by X-ray report. Femoral neck BMD was measured at baseline, and was converted to T-scores. We defined "low BMD" as femoral neck T-scores being less than -2.5 (i.e., osteoporosis). The prevalence of osteoporosis in women and men was 21% and 11%, respectively. The estimation of time-dependent attributable fraction (AF) was based on the Cox's proportional hazards model which allows for confounding factors to be taken into account.

During the follow-up period, the incidence rate of fracture was 3.8% (n=687) and 1.7% (n=240) per 100 person-years for women and men, respectively. For hip fracture, the 5-year AF was 55% in women and 40% in men, while 20-year AF was 42% in women and 30% in men. For clinical vertebral fracture, the 5-year AF was 27% in women and 20% in men; 20-year AF was 20% in women and 17% in men. For wrist fracture, the 5-year AF was 14% in women and 15% in men; 20-year AF was 13% in women and 15% in men. Thus, the burden of fractures occuring in older individuals attributable to osteoporosis is modest. The decline in the attributable fraction with time implies that repeat BMD measurements can be valuable in accounting for a substantial proportion of hip fractures.