We compared quantitative computed tomography (QCT) and dual X-ray absorptiometry (DXA) with respect to their ability to discriminate subjects with and without prevalent vertebral fractures. In 240 post-menopausal women (mean age 63.7 +/- 6.9 years) lateral spine radiographs (T4-L4) were reviewed for the presence of vertebral fracture. Using a semiquantitative technique to grade the severity of vertebral deformities, we classified fractures as mild, moderate or severe (grade 1 to 3, respectively). Postero-anterior DXA (PA-DXA) and lateral DXA (L-DXA) measurements (L2-4) as well as QCT measurements of the lumbar spine (T12-L3 or L1-14) were obtained in all women. Seventy-two women were diagnosed with at least one fracture, and of these 40 were graded as mild. Comparing normal women with fractured women, we found the area under the receiver operating characteristics (ROC) curves to be greatest for QCT (0.81), followed by L-DXA (0.72) and PA-DXA (0.65). The differences among all three techniques were significant. Comparing the normal women with women having only mild fractures, the areas under the ROC curves were 0.79, 0.73 and 0.63 for QCT, L-DXA and PA-DXA, respectively. Significant differences existed between QCT and PA-DXA as well as between L-DXA and PA-DXA. Logistic regression analysis also revealed the highest age-adjusted odds ratios for QCT (3.67; 2.25-5.97) while L-DXA and PA-DXA showed substantially lower odds ratios (2.00; 1.39-2.87, and 1.54; 1.11-2.15, respectively). We conclude that low bone density as measured by QCT, PA-DXA or L-DXA is significantly associated with the prevalence of vertebral fractures. Of the methods studied, QCT of trabecular bone offered the best discriminatory capability. L-DXA proved to be superior to PA-DXA in its diagnostic sensitivity, particularly in women with mild fracture. Mild vertebral fractures are associated with decreased spinal bone density and may be regarded as osteoporotic deformities.