Measures of muscle mass, strength, and function predict risk of incident fractures, but it is not known whether this risk information is additive to that from FRAX (fracture risk assessment tool) probability. In the Osteoporotic Fractures in Men (MrOS) Study cohorts (Sweden, Hong Kong, United States), we investigated whether measures of physical performance/appendicular lean mass (ALM) by DXA predicted incident fractures in older men, independently of FRAX probability. Baseline information included falls history, clinical risk factors for falls and fractures, femoral neck aBMD, and calculated FRAX probabilities. An extension of Poisson regression was used to investigate the relationship between time for five chair stands, walking speed over a 6 m distance, grip strength, ALM adjusted for body size (ALM/height²), FRAX probability (major osteoporotic fracture [MOF]) with or without femoral neck aBMD, available in a subset of n = 7531), and incident MOF (hip, clinical vertebral, wrist, or proximal humerus). Associations were adjusted for age and time since baseline, and are reported as hazard ratios (HRs) for first incident fracture per SD increment in predictor using meta-analysis. 5660 men in the United States (mean age 73.5 years), 2764 men in Sweden (75.4 years), and 1987 men in Hong Kong (72.4 years) were studied. Mean follow-up time was 8.7 to 10.9 years. Greater time for five chair stands was associated with greater risk of MOF (HR 1.26;​ 95% CI, 1.19 to 1.34), whereas greater walking speed (HR 0.85;​ 95% CI, 0.79 to 0.90), grip strength (HR 0.77;​ 95% CI, 0.72 to 0.82), and ALM/height² (HR 0.85;​ 95% CI, 0.80 to 0.90) were associated with lower risk of incident MOF. Associations remained largely similar after adjustment for FRAX, but associations between ALM/height² and MOF were weakened (HR 0.92;​ 95% CI, 0.85 to 0.99). Inclusion of femoral neck aBMD markedly attenuated the association between ALM/height² and MOF (HR 1.02;​ 95% CI, 0.96 to 1.10). Measures of physical performance predicted incident fractures independently of FRAX probability. Whilst the predictive value of ALM/height² was substantially reduced by inclusion of aBMD requires further study, these findings support the consideration of physical performance in fracture risk assessment.