Comparison of methods for the visual identification of prevalent vertebral fracture in osteoporosis

The identification of vertebral fracture in osteoporosis is based mainly on the identification of abnormal variation in vertebral shape, but this can be misleading in the presence of a non-fracture deformity or normal variant of vertebral shape. Qualitative identification of vertebral fracture (Qual) is influenced by the subjectivity of the approach, and although more objective, the semiquantitative method (SQ) can be difficult to apply. In addition, there has been little independent evaluation of SQ in relation to other approaches. We aimed to evaluate a new algorithm-based approach for the qualitative identification of vertebral fracture (ABQ) and to compare it with SQ and Qual. Two radiologists reported spinal radiographs for 372 postmenopausal women using Qual (reader 1), and SQ and ABQ (reader 2). Non-fracture deformities and normal variants were also reported using Qual and ABQ. The prevalence of vertebral fracture by subjects was higher for SQ (24%) than for Qual (11%) and ABQ (7%). Agreement was poor between SQ and the other methods, and moderate between Qual and ABQ. Twenty-two women with vertebral fracture were agreed by all three methods, similar to the total identified by ABQ (25 women). Seventeen women diagnosed with fracture by Qual, had non-fracture deformity or normal variant (but no fracture) according to ABQ. Of the women with SQ fractures, 53% and 70% were identified negative for fracture but positive for non-fracture deformity or normal variant by ABQ and Qual. The main sources of discrepancy between SQ and the other methods were Scheuermann's disease, normal variation, and degenerative change accompanied by short anterior vertebral height. For all methods, bone mineral density (BMD) and BMD Z-scores were lower in women with vertebral fractures than in those with no fractures. Bone mineral density and BMD Z-scores were also lower at the lumbar spine and total body in women with vertebral fractures according to Qual and ABQ than they were for SQ, and were lower in women with SQ fractures agreed by Qual and ABQ, compared with those diagnosed negative for fracture by Qual and ABQ (p<0.01). We conclude that poor agreement between methods arises mainly from difficulties in differentiating true fracture from non-fracture deformity. Our new approach attempts to address this problem but requires further testing in a larger study population.