Osteoporosis in men is becoming a public health problem in developed countries. Fracture incidence increases with age, and the number of fractures increases because of the ageing of the population. We assessed the predictive value of bone mineral density (BMD) for osteoporotic fractures evaluated prospectively in a large cohort of elderly men and assessed the sensitivity of the T-score =-2 to detect men who will sustain a fracture. Fracture incidence was evaluated for 90 months in 759 men from the MINOS cohort aged 50 and over at baseline. In 74 men, 77 incident vertebral and peripheral fractures occurred. BMD was measured at baseline at the lumbar spine, hip, whole body and distal forearm. The incidence of osteoporotic fractures increased with age and with decreasing body weight. In men with low BMD (T-score <-2), fracture incidence varied from 2.26 to 3.07 fractures per 100 person-years and was 2.1 to 3.6 times higher than in men with normal BMD. After adjustment for age, body weight and height, baseline BMD was 3.7 to 7.9% ( P <0.05-0.0001) lower at all the sites of measurement in men who sustained a fracture. After adjustment for age, weight and prevalent fractures, BMD was predictive of osteoporotic fractures at all the sites. Odds ratios varied from 1.28 to 1.89 per 1 SD decrease in BMD ( P <0.05-0.0001). The predictive accuracy of BMD for fractures (area under the curve of the receiving operator characteristics adjusted for age, weight and prevalent fractures) varied from 0.643 to 0.712 according to the skeletal site and was higher for the whole body than for other sites. Thus, BMD itself has a limited value for determining men at an increased risk for fracture. The percentage of incident fractures occurring in men with low BMD (T-score <-2) ranged from 13.7% at the trochanter to 44.6% at the ultradistal radius. Conversely, 27 to 45% of incident fractures occurred in men with mildly decreased BMD (T-score between -1 and -2). In conclusion, BMD predicts osteoporotic fractures in men independently of age, body weight and prevalent fractures. However, the sensitivity of BMD to detect men at high risk of fracture is low. More studies on the predictors of fractures in men, such as bone architecture, morphology, biochemical markers of bone turnover and hormonal levels, are necessary.