THU0359 WHAT IS THE LEVEL OF AGREEMENT BETWEEN LOCAL AND CENTRAL READERS IN THE DETECTION OF ACTIVE AND STRUCTURAL LESIONS ON MRI TYPICAL OF AXIAL SPONDYLOARTHRITIS? DATA FROM THE ASAS CLASSIFICATION COHORT STUDY

Abstract

Background There has been no central reader evaluation of MRI scans from the ASAS Classification Cohort (ASAS-CC)1 to compare detection of lesions in the sacroiliac joints (SIJ) between central and ASAS-CC local site readers. Active MRI lesions typical of axSpA were reported in 61.6% and 2.2% of patients from this cohort diagnosed with axSpA and non-axSpA back pain, respectively1. Structural lesions were recorded but not reported in the literature. Objectives We aimed to compare detection of active and structural lesions on MRI images of the SIJ from the ASAS-CC between ASAS-CC local site readers and central readers from the ASAS-MRI group. Methods MRI images were available from 258 of the 495 cases who had MRI performed in the ASAS-CC and also had a local rheumatologist diagnosis. Seven central readers recorded MRI lesions in an eCRF that included wording of lesions defining active and structural lesions typical of axSpA that was exactly the same as in the original ASAS-CC eCRF permitting comparisons between central and local site readers. In addition, lesions that met the criteria for an ASAS positive MRI were recorded by central readers. Active and structural lesion frequencies were assessed descriptively according to majority agreement (=4/7) of central reader data and also any 2 central readers. Reliability of detection of MRI lesions was compared between central and local readers using the kappa coefficient. Results Significant differences in lesion frequencies were observed according to diagnostic category (Table 1). The frequency of active lesions reported by local readers (61%)was greater than for central readers that agreed on the presence of an active lesion (49.7%). Structural lesions were reported less frequently by local readers (44.0%) compared to central readers that agreed on the presence of a structural lesion (54.9%). Reliability of local readers for detection of active lesions was good but only fair for structural lesions (Table 2). Conclusion Local readers may have overestimated the presence of active lesions and underestimated the presence of structural lesions in the ASAS-CC. Their reliability for detection of structural lesions was limited which could reflect lack of awareness of structural lesions related to axSpA.