Interobserver variation in the detection of osteopenia by radiography and comparison with dual X-ray absorptiometry of the lumbar spine

In 100 patients (20 male, 80 female) radiographs of the lumbar spine were obtained in both planes, anteroposterior and lateral. Nine readers independently and without specific criteria or training assessed the radiographs for presence of osteopenia in the form of a binary decision. A posteranterior dual x-ray absorptiometry (PA DXA) measurement of the lumbar spine was performed in all patients using the Hologic QDR 1000 bone densitometer. A bone mineral density (BMD) of 0.83 g/cm2 (T-score about 2 SD and 2.5 SD lower than BMD in normal young female and male subjects respectively) was used as a threshold for the diagnosis of osteopenia. Complete agreement amongst the 9 readers was achieved in 43 patients. In 26 more patients at least 8 readers agreed. kappa-coefficients for interobserver variation ranged from 0.458 to 0.691 for reader pairs. For agreement between the observer ratings and the DXA results, kappa-coefficients ranging between 0.347 and 0.555 were found. The vast majority of readers agreed in the diagnosis of osteopenia in cases where the BMD was less than 0.73 g/cm2. Where the BMD was between 0.73 and 1.03 g/cm2 a substantial disagreement was found between reader evaluation and DXA measurement, and also amongst the readers. We conclude from our results that osteopenia can reliably be detected from lumbar spine radiographs by all readers only after a substantial amount of BMD is lost. On the other hand, a diagnosis based solely on PA DXA measurement of the spine may also lack accuracy, due to a substantial influence of degenerative changes of the lumbar spine and aortic calcification.