Observer variation in computed tomography of the sacroiliac joints: a retrospective analysis of 1383 cases

Abstract

Background: Computed tomography (CT) for evaluation of sacroiliitis has a higher diagnostic accuracy than radiography. There is a high degree of interobserver variation in evaluating sacroiliitis on radiographs. Purpose: To evaluate interobserver variation in CT of the sacroiliac joints for evaluation of sacroiliitis in a large number of patients. Material and Methods: 1383 CT examinations of the sacroiliac joints were reviewed by two observers. The outcomes as originally reported and the findings from the reviews were classified as no sacroiliitis, equivocal, unilateral sacroiliitis, or bilateral sacroiliitis. The unweighted kappa statistic was used for assessment of observer agreement. Results: The interobserver agreement between the two reviewers was good (κ = 0.6724), with excellent agreement on cases of bilateral sacroiliitis and moderate agreement on cases of unilateral sacroiliitis. Excellent agreement was also reached in normal cases. Compared to the original reports, there were moderate interobserver agreements between both reviewers' findings and the original reports (κ = 0.4651 and κ = 0.4481, respectively). Conclusion: The interobserver variation for the diagnosis of sacroiliitis on CT between two reviewers in a study setting showed good agreement, with moderate agreement between each of the observers and the original clinical reports. CT is a reliable method for evaluating the sacroiliac joints for changes of sacroiliitis.