Measurement of vertebral area on spine x-rays in osteoporosis: reliability of digitizing techniques

Much of the clinical research in osteoporosis is directed toward documenting a reduction in vertebral fracture rate, but there is considerable disagreement about defining and quantifying vertebral fractures. We have evaluated the technique of digitizing landmarks identified on lateral radiographs of thoracic and lumbar vertebrae and computing vertebral body area. Reduction in area indicates that fractures occurred. Radiographs from 10 patients with osteoporosis and vertebral fractures were obtained from each of two centers. Henry Ford Hospital (HFH) and Mayo Clinic (MC), and vertebral area for each individual in the complete set of 20 radiographs was calculated at each center. Measurements at the two centers differed by a multiplicative constant related to the method of recording landmarks on the radiographs that was estimated using 300 x-rays from HFH. After adjusting the MC areas for this multiplicative relationship, the average ratio of the HFH areas to the transformed MC areas of individual vertebrae (T4-L5) ranged from 0.98 to 1.06. The correlation between HFH and transformed MC areas for individual vertebrae averaged 0.85, with slopes between 0.87 and 1.00, intercept average -0.57. Within-patient rank correlation averaged 0.97. We conclude that radiographic digitization is a reliable and reproducible method of determining vertebral body dimensions that is suitable for evaluating radiographs obtained at different clinical sites and for comparison with normal data. This technique should prove useful for documenting the presence of a vertebral fracture that may not be readily apparent on visual inspection of radiographs and for monitoring serial changes in vertebral body dimensions in long-term epidemiologic and therapeutic studies.