This prospective study was aimed at determining the risk factors for the development of fractures in perimenopausal women. The study group (n = 3068) was comprised of a stratified population sample of women aged between 47 and 56 years. During the follow-up period of 3.6 years, 257 (8.4%) of the women sustained a total of 295 fractures. After adjustment for covariates, the relative risk (RR) of sustaining a fracture was found to be 1.4 [95% confidence interval (CI) 1.2-1.6] for a 1 standard deviation (SD) decrease in the spinal and femoral neck bone mineral density (BMD). Women with a previous fracture history were found to have an increased risk of fracture [RR 1.7 (95% CI 1.3-2.2)] and those reporting three or more chronic illnesses exhibited a RR of 1.4 (95% CI 1.0-1.9). Women not using hormone replacement therapy (HRT) had a RR of 1.5 (95% CI 1.1-2.2) for all fracture types. When osteoporotic fractures (vertebral, hip, proximal humerus and wrist fractures; n = 98) were used as an endpoint, the independent risk factors were found to be a low BMD (RR for a 1 SD decrease in both spinal and femoral neck BMD was 1.6, 95% CI 1.3-2.0), a previous fracture history (RR 1.9, 95% CI 1.3-2.9) and nonuse of HRT (RR 2.2, 95% CI 1.3-4.0). The independent risk factors for all other fractures (n = 158) were a low BMD (RR for a 1 SD decrease in the spinal BMD was 1.4, 95% CI 1.2-1.6 and in the femoral neck BMD was 1.3, 95% CI 1.1-1.5), a previous fracture history (RR 1.6, 95% CI 1.1-2.2), smoking (RR 1.8, 95% CI 1.1-2.7) and having had three or more chronic illnesses (RR 1.6, 95% CI 1.1-2.2). Weight, height, age, menopausal status, maternal hip fracture, use of alcohol, coffee consumption or dietary calcium intake were not independently associated with the development of any particular type of fracture. We conclude that the independent risk factors for perimenopausal fractures are a low bone density, previous fracture history, nonuse of HRT, having had three or more chronic illnesses and smoking, the gradient of risk being similar for spinal and femoral neck BMD measurements in the perimenopausal population. The risk factors are slightly different for perimenopausal osteoporotic than for other types of fractures.