Spinal bone mineral density (BMD) is traditionally measured by dual-energy X-ray absorptiometry (DXA) in the anteroposterior (AP) projection which includes both the vertebral body and the posterior elements in the measurement. The posterior elements, however, contribute little to the compressive strength of the spine. It has therefore been suggested that spinal BMD measured in the lateral projection, including only the vertebral body in the measurement, might be more appropriate for the prediction of fracture risk. To date little clinical evidence has been presented to support this assumption. To address the issue, we measured vertebral, hip and forearm BMD in situ in 14 human cadavers and remeasured BMD in vitro in excised vertebrae. Lateral spinal measurements were performed in the decubitus position. Fracture force and other biomechanical measures were determined for 32 vertebrae in a mechanical testing machine and compared with BMD values in situ and in vitro. Correlations of BMD with vertebral fracture force were r = 0.48/0.51 (in situ/in vitro) for the AP spinal measurements, r = 0.45/0.71 (in situ/in vitro) for the lateral spinal measurements, and r = 0.64 and r = 0.53 for total hip and forearm measurements in situ, respectively. Thus, despite an apparent diagnostic advantage in vitro, lateral spinal BMD measurement was not superior to AP measurement when performed in situ. This observation corresponds well with previous clinical findings and is probably due to the larger accuracy error in the lateral than in the AP projection resulting from a lower ratio of bone to soft tissue. The high correlation between hip BMD and vertebral fracture force suggests that hip measurement may prove as useful for vertebral fracture risk assessment as spinal measurement in any projection, especially in the elderly with a high prevalence of degenerative changes in the spine.