The vast majority of current sarcopenia definitions use DXA-derived appendicular lean mass (DXA ALM) as an approximation of muscle mass. However, there is increasing evidence that ALM is poorly predictive of incident fractures. In this editorial, we describe findings from several cohorts suggesting that DXA ALM gives only limited information on the risk of incident fracture, and that in the US MrOS, WHI and Health ABC cohorts, accounting for femoral neck bone mineral density may entirely remove ALM-fracture associations. These observations raise important questions about the role of DXA ALM both in sarcopenia definitions and as a potential input variable for FRAX. We conclude that DXA ALM is unlikely to be a useful addition to the FRAX tool, but that other means of estimating muscle mass, such as those derived from creatine dilution or peripheral quantitative CT, might offer more value for fracture risk assessment.