Osteoporosis is the most common skeletal disease in the world, leading to fractures of the hip, spine and wrist. Current methods for detecting patients at risk for developing osteoporotic fractures involve the non-invasive measurement of bone mineral content at a central or peripheral site. The measured bone density or bone content is compared against that of age-matched controls and a fracture risk is estimated. This technique has been unsuccessful for accurate fracture prediction because of the large overlap in bone density values between osteoporotic and normal patient groups. To assess individual patient vertebral strength, a series of software routines has been developed which allow the conversion of quantitative computed tomography (QCT) studies of the spine into vertebral finite element models (FEM). The QCT studies of both normal and osteoporotic patients are loaded onto a Sun workstation for analysis. Each study is interpolated to cubic volume elements and then rotated to align the vertebrae to approximate anatomical loading conditions. The vertebral bodies of LI and L2 are extracted from the data using interactive thresholding software. The QCT volume elements representing the vertebral body are converted into a computer model of three-dimensional finite elements for structural analysis . The material properties for each element of the computer mesh are derived from the average bone mineral density of the element. The FEM are analyzed using a commercially available finite element program and a fracture load (yield stress) of the model is calculated. FEM estimates of vertebral yield for 36 normal patients are compared to actual compression tests done by other researchers, showing the FEM yield stress to underestimate the in vitro compressive strength due to limitations of the finite element analysis technique. The yield stress for 59 models from normal and osteoporotic patients is compared to the QCT measured total vertebral body bone content and trabecular bone density using receiver operator characteristics (ROC analysis). The FEM calculated yield stress is shown to be a more sensitive and accurate test for osteoporosis than the QCT trabecular mineral density by virtue of a larger area under the ROC curve. The area under the FEM curve was 0.964, and the areas under the trabecular and total vertebral bone and curves were 0.907 and 0.871, respectively (p < 0.05 for both cases). Based upon duplicate studies on the same patients, the reproducibility of the FEM calculated yield stress is determined to be 13%. From vertebral FEM with and without the cortical shell, normal patients were found to rely on the cortex for 12.4% of their strength, while osteoporotic patients showed a cortical strength contribution of 56.2% (p < 0.001 for both oases). The importance of maintaining both the trabecular and cortical vertebral bone is shown. Suggestions are made for improving the model in order to create a clinically useful tool for vertebral fracture prediction.