FRI0592 Scoring mri inflammation and structural lesions in sacroiliac joints of patients with axial spondyloarthritis: is inter-reader reliability dependent on the number of mri slices?

Background The SPARCC sacroiliac joint (SIJ) scoring system assesses 6 semicoronal MRI slices for inflammation and 5 slices for structural lesions in patients with axial spondyloarthritis (axSpA). 1,2 However, the cartilaginous SIJ compartment may show 1–2 additional slices anteriorly or posteriorly, depending on body size and scan orientation/tilt. Objectives To investigate inter-reader reliability of an “all slices” approach versus the standard SPARCC scoring of 6/5 slices. Methods Fifty-three patients with axSpA were treated with TNF inhibitor and had MRIs obtained at weeks 0/4/16/52. An experienced (UW) and two newly trained (GK, SK) blinded readers independently scored 199 SIJ MRI scans in chronological order. The cartilaginous SIJ compartment was scored slice by slice by the SPARCC 6/5 slices approach and by all available cartilaginous slices. Initially, the most anterior and posterior slices covering the cartilaginous compartment and the transitional slice were identified. The transitional slice was defined as the most anterior cartilaginous slice with the first portion of the ligamentous compartment, clearly visible on the left and/or right side. We scored SIJ inflammation, fat metaplasia, erosion and backfill, and a combined erosion and backfill score was created. Inter-reader reliability for reader pairs SK-UW/GK-UW/SK-GK was assessed using percent agreement (for individual scores) and intra-class correlation coefficients for sum scores. Results Pairwise percent agreement was 67%/63%/79% for identification of anterior slice, 47%/56%/44% for posterior slice and 69%/68%/72% for transitional slice. Using the “all slices” approach, readers UW/SK/GK scored mean 7.2/7.7/7.0 slices per MRI scan. “6/5 slices” and “all slices” correlated closely with each other for status scores at baseline/status scores at week 52, and change scores at week 52; BME 0.983/0.985/0.983; fat metaplasia 0.994/0.982/0.953; erosion 0.981/0.974/0.957; backfill 0.993/0.983/0.978; combined erosion and backfill 0.983/0.971/0.919. Conclusions The standardised 6/5 slices SPARCC methods had equal reliability as compared to evaluation of all cartilaginous slices. There was limited reliability to identify the posterior slice in the ”all slices” approach, as opposed to good reproducibility to determine the transitional slice in the ”6/5 slices” approach. Combining erosion and backfill scores tended to result in superior reliability compared to the 2 lesions separately, indicating a challenge to identify the transition from erosion to backfill. References [1] Arthritis Care Res2005;53:703–9. [2] J Rheumatol2015;42:79–86. Disclosure of Interest None declared