Background/Purpose: The sacroiliac joint (SIJ) has a distinct anatomical shape that can pose challenges to standard planar imaging techniques making the detection of subtle changes difficult¹. SIJ radiographs are a very important tool for the evaluation of disease in patients with suspected axial spondyloarthritis (axSpA). With lack of a reliable biomarker and limitations in physical examination, clinicians are reliant on the interpretation of plain films by radiologists, who may or may not be trained in musculoskeletal radiology². This study characterizes the details reported by radiologists in their reports of SIJ radiographs and compares their scoring with formal interpretations by trained readers.

Methods: SIJ 3-view (Ferguson and obliques) radiographs of US Veterans enrolled in the Program to Understand Long-term Outcomes of Spondylarthritis (PULSAR) registry were evaluated by three trained readers using the scoring from the modified New York (mNY) criteria. Corresponding radiology reports were then evaluated for the presence of terms used to apply mNY scores (erosions, sclerosis, joint narrowing or widening, ankylosis, and blurring), descriptions regarding the severity of any abnormality (no abnormality, mild/possible abnormality, clear abnormality), and the presence of sacroiliitis or axSpA. Classification of sacroiliitis based on consensus mNY grade by trained readers was compared with corresponding radiologist interpretation. Radiograph interpretations were evaluated for association with HLAB27 status.

Results: Reports from 90 radiographs (180 joints) revealed that among patients described as having sacroiliitis or axSpA by the radiologist report, no particular finding was described more frequently than 20% when considering laterality separately and no finding was present more than 31% of the time when assessing either side for abnormalities. Similarly, among patients classified as axSpA by consensus of the three readers, no particular finding was described more frequently than 32% of the time when assessing either side. The most frequent described findings were ankylosis (31.2%) and erosions (29.0%). Compared with the gold-standard consensus interpretation by three trained readers, the radiologist report demonstrated a sensitivity of 57.6% (95% CI 44.1-70.4) and specificity of 93.3% (95% CI 68.1-99.8) for identifying sacroiliitis/axSpA. The strength of the association between HLAB27 status and classification of AS was also greater for the consensus of three readers (ß=0.465, R²=0.288) than the radiologist report (ß=0.319, R²=0.090).

Conclusion: In patients likely to have axSpA, radiology reports only describe specific findings suggestive of axSpA in a minority of cases.  Classification of axSpA by radiology reports is less well associated with HLAB27 status than a consensus of trained readers.  Rheumatologists should consider training to read SIJ radiographs themselves, rather than relying exclusively upon radiology reports.

« Back to ACR Convergence 2020

ACR Meeting Abstracts - https://acrabstracts.org/abstract/findings-compatible-with-axial-spondyloarthritis-in-radiologist-reports/