Revisable criteria for vertebral deformity.

In order to study vertebral fractures in various study populations, we earlier prepared a database of vertebral dimensions derived from spinal radiographs of 191 normal women seen regularly over 25 years. In this report we have expanded the range of measurements to include vertebral levels T3 to L5. We report means and standard deviations on anterior and posterior heights, on wedge shape and on heights relative to adjacent vertebrae. When one or both of the latter two quantities are 'far' below the mean, a vertebra is called deformed. We also describe a more flexible way of expressing damage using the number of deformed vertebrae, the degree of deformity of individual vertebrae, or the total damage to the entire spine. In assessing damage we use criteria for deformity adjusted to the limits detected by an experienced diagnostician, replacing an earlier approach based on 95% probability limits of normal variation. The normal women from whom these variations are ascertained are a low-prevalence group with respect to vertebral deformity, with prevalence of 2.8%. When the criteria developed from these women were applied to a moderate-prevalence group (37%) the model had a sensitivity of 97%, a specificity of 89% and an accuracy of 92% as regards the identification of subjects with damaged vertebrae. When used epidemiologically for a moderate-prevalence group the model has a known overestimation of 15%. the model is compared with other schemes for identifying vertebral deformities.