The risk of fragility fractures in elderly men is only partly explained by areal bone mineral density (aBMD) measured by dual X-ray absorptiometry (DXA). Several studies suggest the importance of bone morphology for the risk of fracture. The aim of this study was to assess the value of bone size and estimated structural parameters for the prediction of incident fractures in a large cohort of men. This study was made in 759 men aged 50-85 from the MINOS cohort. During a 90-month follow-up, 74 men sustained incident vertebral and peripheral fractures. Areal BMD was measured by DXA at femoral neck, distal radius and distal ulna. Estimates of structural bone parameters and volumetric BMD (vBMD) were derived from aBMD measured by DXA. Given the limited number of fractures, the predictive value of investigated parameters was assessed for peripheral and vertebral fractures jointly by using logistic regression. Men who sustained the fractures had, at baseline, lower aBMD (3.5-6.5%), lower bone mineral content (BMC 5.4-8.7%) and lower cortical thickness (3.5-6.9%) compared with the men without fracture. At all the three skeletal sites, aBMD, BMC, width, cortical area and thickness, cross-sectional moment of inertia (CSMI), and section modulus predicted incident fractures (O.R. = 1.28-1.92 per 1 SD decrease, P < 0.05-0.0001). Fracture risk was weakly associated with vBMD for ulna (O.R. = 1.25 per 1 SD decrease, P < 0.05) but not for femoral neck or radius. After adjustment for aBMD, bone width remained a significant predictor of fractures (O.R. = 1.37-1.48 per 1 SD decrease, P < 0.02-0.01). Men with osteopenia (BMD T score < -1) and low bone width (T score < -1) had the fracture incidence similar to that observed in men with BMD T score < -2. Bone width and aBMD of the femoral neck and radius were predictive of fractures in 49 men with the incident peripheral fractures, whereas their O.R. did not attain the level of statistical significance in 25 men with the incident vertebral fractures. Men, who had both low aBMD and low CSMI ( both T scores < -1), had the fracture risk 3.8 to 4.2 higher than the reference group (both T scores >or= -1). Men, who had both low aBMD and low section modulus (both T scores < -1), had the fracture risk 2.1 to 4.1 higher than the reference group (both T scores >or= -1). In conclusion, men who sustained a fragility fracture during a 90-month follow-up had, at baseline, lower BMC because they had narrower bones but not necessarily less dense. In elderly men, small bone width, low BMC and poor resistance to bending may increase bone fragility. Low bone width seems to be associated with an increased fracture risk in elderly men regardless of aBMD.