The absence of specific criteria for the definition of vertebral fracture has major implications for assessing the apparent prevalence and incidence of vertebral deformity. Also, little is known of the effect of using different criteria for new vertebral fractures in clinical studies. We therefore developed radiological criteria for vertebral fracture in women for assessing both the prevalence and the incidence of vertebral osteoporosis in population and in prospective studies and compared these with several other published methods. Normal ranges for vertebral shape were obtained from radiographs in 100 women aged 45-50 years. These included ranges for the ratios of anterior/posterior, central/posterior and posterior/predicted posterior vertebral heights from T4 to L5. The predicted posterior height was calculated from adjacent vertebrae. In contrast to other methods, our definition of fracture required the fulfillment of two criteria at each vertebral site, and was associated with a lower apparent prevalence of fracture in the control women due to a lower false positive rate. The prevalence and incidence of vertebral deformity using different criteria were then compared in a series of women with skeletal metastases from breast cancer in whom radiographs were obtained 6 months apart. The prevalence of vertebral deformity and the specificity for deformity varied markedly with differing criteria. Using a cut-off of 3 standard deviations the prevalence of vertebral deformity in the women with breast cancer was 46%. Using other methods, the prevalences of deformity ranged from 33% to 74%. Over a 6-month interval 25% of patients with breast cancer sustained 61 deformities using our method, of which only 8% resulted from errors in reproducibility. The number of patients sustaining new deformities was increased twofold when assessed by other methods (45%-53%), but errors of reproducibility may have accounted for 21% of the new deformities. The magnitude and distribution of these errors have important implications for the apparent therapeutic efficacy of agents in clinical trials of osteoporosis. The rapid semi-automated technique for assessing vertebral deformities on lateral spine radiographs that we have developed has a high specificity, and reduces the impact of errors of reproducibility on estimates of prevalence and incidence. The method should prove a value in assessing vertebral deformity both in population studies and in prospective clinical trials.