Diagnostic performance of sacroiliac and spinal MRI for the diagnosis of non-radiographic axial spondyloarthritis in patients with inflammatory back pain

Objective The lack of specificity of the ASAS MRI criteria for non-radiographic axial spondylarthritis (NR-axSpA) justifies the evaluation of the discriminatory capacity of other MRI abnormalities in the sacroiliac joints and dorsolumbar spine. Methods In patients hospitalized for inflammatory lumbar back pain, the diagnostic performance (sensitivity, specificity, positive likelihood ratio (PLR)) of MRI abnormalities was calculated using the rheumatologist expert opinion as a reference: (i) sacroiliac joints: Bone marrow edema (BME) (number and location), extended edema > 1 cm (deep lesion), fatty metaplasia (number), erosion (number and location), backfill. (ii) Dorsolumbar spine: BME (number and location), fatty metaplasia (number), posterior segment involvement. Results In this prospective cohort, 40 NR-axSpA cases and 79 other diagnoses were included. The presence of at least 3 inflammatory signals in the sacroiliac joints (PLR: 25.67 [95% CI: 3.48–48.9]), the presence of at least one sacroiliac erosion (PLR: 12.80 [3.04–54]), the combination of an inflammatory signal and sacroiliac erosion (PLR: 11.85 [2.79–50]), the combination of deep lesion and fatty metaplasia (PLR: 15.80 [2.05–121.9]) or erosion (PLR: 11.86 [1.47–95.01]) had the best diagnostic performance. The combination of spinal and sacroiliac MRI criteria significantly increased diagnostic performance for the diagnosis of NR-axSpA. Conclusion When NR-axSpA is suspected, in addition to the presence and number of inflammatory lesions, MRI interpretation should include the location and the extent of the sacroiliac lesions, the presence of erosion or fatty metaplasia, and anterior involvement of the lumbar spine.