Observer variation in plain radiography of the lumbosacral spine

In this study, interobserver and intraobserver variations in the interpretation of plain radiographs of the lumbosacral spine were evaluated. Three radiologists independently interpreted the radiographs from 200 consecutive outpatients, aged 13-93 years, mostly referred from general practitioners. Interobserver agreement was best for vertebral fractures, osteopenia, spondylolisthesis at L5-S1, lumbosacral junctional vertebra, reduced disc height at L4-S1 and osteophytes at L2-S1 (kappa 0.61-0.95), and poorest for spina bifida of S1, degenerative spondylolisthesis and facet joint arthrosis at T12-L4, sacroiliac joint arthrosis, narrow central spinal canal, film quality, and for decisions concerning evaluation of facet joints and spinal canal (kappa < 0.34). For several diagnoses, the number of abnormal findings differed significantly between observers (p < 0.05, McNemar's test), indicating different diagnostic thresholds. Intraobserver agreement in 36 reevaluated patients was fair to excellent for almost all variables (kappa > 0.46). Although some diagnoses related to low back pain were quite consistently evaluated, the substantial disagreement on many findings should alert clinicians and radiologists against overestimating the validity and usefulness of the examinations. To improve diagnostic consistency, it is important to reduce variation caused by different thresholds for abnormality.