Agreement between clinical practice and trained central reading in reading of sacroiliac joints on plain pelvic radiographs. Results from the DESIR cohort.

OBJECTIVE: To investigate the degree of agreement between local rheumatologists/radiologists and central trained readers (external standard) on the presence/absence of sacroiliitis on radiographs of the sacroiliac (SI) joints.
METHODS: Patients with inflammatory back pain (duration ≥3 months but <3 years) suggestive of axial spondyloarthritis (SpA) were included in the Devenir des Spondylarthropathies Indifferérenciées Récentes (DESIR) cohort. Baseline radiographs of the SI joints were interpreted by 2 central readers (modified New York criteria); cases of disagreement were adjudicated by a third reader, yielding a positive or a negative result (central reading). The same radiographs were also interpreted by local radiologists/rheumatologists and were rated as "normal," "doubtful sacroiliitis," "obvious sacroiliitis," or "SI joint fusion" (local reading); positive findings were defined as "at least unilateral obvious sacroiliitis," "bilateral obvious sacroiliitis," or "at least unilateral fusion." Agreement and misclassifications between central readers and between central reading versus local reading were calculated (kappa values).
RESULTS: Interreader agreement between the central readers was moderate (κ = 0.54); 108 of 688 radiographs (15.7%) were adjudicated. According to local reading ("at least unilateral obvious sacroiliitis"), 183 of the 688 patients (26.6%) had sacroiliitis, whereas according to central reading, 145 of 688 patients (21.1%) had sacroiliitis. Agreement between local reading and central reading was also moderate (κ = 0.55); 76 of 183 patients (41.5%) with "at least unilateral obvious sacroiliitis" (positive by local reading) and 32 of 109 patients (29.4%) with "bilateral obvious sacroiliitis" or "at least unilateral fusion" (positive by local reading) were rated as "negative" by central reading, and 38 of 505 patients (7.5%) and 68 of 579 patients (11.7%), respectively, without sacroiliitis (negative by local reading) were interpreted as "positive" by central reading.
CONCLUSION: In patients with recent-onset inflammatory back pain, both trained readers and local rheumatologists/radiologists agreed only moderately on the recognition of radiographic sacroiliitis. A significant proportion of locally recognized ankylosing spondylitis (AS) patients were not confirmed as having AS by central reading (false positive), while a small minority of patients were false negative, indicating the necessity of reevaluating the role of radiographic sacroiliitis as diagnostic criterion for axial SpA.