Abstract 20 — Performance of Various Definitions for Active MRI: Lesions in Sacroiliac Joint and Spine in Discriminating Patients with Axial Psoriatic Arthritis

Background Unlike axial spondyloarthritis, no classification criteria exist for axial psoriatic arthritis (axPsA). In 2021, the Assessment of SpondyloArthritis International Society (ASAS) revised the magnetic resonance imaging (MRI) criteria for active sacroiliitis (2021 criteria: bone marrow edema [BME] present in [Formula: see text]4 sacroiliac joint [SIJ] quadrants or [Formula: see text]3 consecutive SIJ slices[1]). This study aimed to compare the utility of this new cut-off in discriminating PsA patients with/without axPsA versus the 2009 ASAS criteria[2] for active-MRI-SIJ (BME [Formula: see text] 2 consecutive slices or [Formula: see text]1 location in a single slice). Methods Consecutive patients who fulfilled the classification criteria for PsA were recruited into this cross-sectional study, regardless of back pain. Sixty-seven patients underwent radiography (including pelvis, cervical/thoracic/lumbar-spine) and MRI-SIJ. Additionally, 47 underwent whole-spine MRI. AxPsA diagnosis was based on clinical information and imaging findings, as determined by an expert rheumatologist and a radiologist, and used as the reference standard. Two independent readers evaluated the MRI images based on two criteria for active sacroiliitis (BME cut-off: [Formula: see text] 4[1] vs [Formula: see text] 2[2]) and spondylitis (BME cut-off: [Formula: see text] 5[3] vs [Formula: see text] 3[4]). The agreement between the two MRI BME cut-offs for active sacroiliitis/spondylitis and the reference standard was evaluated. Results Sixty-seven patients (mean age: 47±12 years, 44 (65.7%) male, psoriasis and PsA disease duration: 13.5±10.3 and 3.8±6.1 years respectively) were recruited (Table 1). Twenty-three (34.3%) were diagnosed with axPsA, including 13 (56.5%) with radiographic sacroiliitis and 10 (43.5%) with non-radiographic axPsA. 12/67 (17.9%) had active MRI-sacroiliitis based on the 2021 ASAS criteria, while 4/47 (8.5%) had spondylitis based on the 2016 proposed definition[3]. Compared with the reference standard, the agreement increased after applying a more stringent threshold to define active sacroiliitis (BME cut-off: [Formula: see text] 4 vs [Formula: see text] 2; Kappa: 0.514 vs 0.392, respectively; Fig. 1A-B). The agreement with the reference standard further increased by applying a more stringent criteria for active spondylitis in addition to active sacroiliitis (BME cut-off for MRI-SIJ and spine: [Formula: see text] 4 and [Formula: see text] 5 vs [Formula: see text] 2 and [Formula: see text] 3; Kappa: 0.717 vs 0.342, respectively; Fig. 1C-D), resulting in higher specificity (active-sacroiliitis: 97.7% vs 81.8%; active-sacroiliitis and/or spondylitis: 100% vs 72.7%) and higher positive predictive value (91.7% vs 61.9%; 100% vs 50.0%, respectively). Conclusion At least four BME lesions on MRI-SIJ and five inflammatory lesions on MRI-spine allow acceptable discrimination of axPsA and no axPsA while assuring [Formula: see text]95% specificity.