Interreader agreement in the assessment of magnetic resonance images of rheumatoid arthritis wrist and finger joints--an international multicenter study.

Abstract

Magnetic resonance imaging (MRI) allows direct visualization of inflammation and destruction in rheumatoid arthritis (RA) joints. However, MRI scoring methods have not yet been standardized or appropriately validated. Our aim was to examine interreader agreement for a simple system of scoring RA changes on MRI among 5 centers that had not undertaken intergroup calibration. MRI of RA wrist and metacarpophalangeal (MCP) joints were scored by experienced readers in 5 centers in different countries. In substudy 1, 5 sets of 2nd-5th MCP joints from UK [Technique A: 1.5 T, coronal and axial T1 and T2 spin-echo, -/+ fat saturation (FS), -/+ iv gadolinium (Gd)] were scored for synovitis (score 0-3) and bone lesions (0-3). In substudy 2, we evaluated 19 sets of 2nd-5th MCP joints [10 sets from UK (Technique A) and 9 sets from the US (Technique B: 1.5 T; coronal T1 spin-echo and T2* gradient-echo + FS, no Gd)] and 19 wrist joints [9 from the US (Technique B) and 10 from Denmark (Technique C: 1.0 T; coronal and axial T1 spin-echo, no FS, -/+ Gd)]. Synovitis (0-3), bone lesions (0-3), and joint space narrowing (JSN, 0-3) were scored in each MCP joint and in 3 different regions of the wrist. Bone erosions and lesions in each bone were scored 0-5. Substudy 1 served to test and redesign the score sheets. In substudy 2, the scores of synovitis and bone lesions by the 5 groups were the same or differed by only one grade in 73% and 85% of joints, respectively. On MRI that included 2 imaging planes and iv Gd (Techniques A and C), these rates were 86% (synovitis) and 97% (bone lesions). Corresponding intraclass correlation coefficients (quadratic weighted kappas) were 0.44-0.68, mean 0.58 (synovitis), and 0.44-0.69, mean 0.62 (bone lesion), i.e., in the moderate to good range. Unweighted kappa values were in the low to moderate range, generally lowest for JSN (< 0.20), better for synovitis and bone erosions, and best for bone lesions, being generally highest for MRI with 2 planes pre- and post-Gd and in MCPjoints compared with wrists. These preliminary results suggest that the basic interpretation of MRI changes in RA wrist and MCP joints is relatively consistent among readers from different countries and medical backgrounds, but that further training, calibration, and standardization of imaging protocols and grading schemes will be necessary to achieve acceptable intergroup reproducibility in assessing synovitis and bone destruction in RA multicenter studies.