Development and validation of the spondyloarthritis radiography module for calibration of readers using the modified Stoke Ankylosing Spondylitis Spine Score.

Abstract

To develop and validate a reference image module aimed at calibration of readers using the modified Stoke Ankylosing Spondylitis Spine Score (mSASSS) to assess radiographic progression in spondyloarthritis. Our working group comprised 6 rheumatologists and 3 musculoskeletal radiologists. The following developmental steps were conducted: (1) review of the literature to identify aspects of the mSASSS requiring methodologic clarity; (2) independent assessment of baseline and 2-year radiographs from 25 patients using the mSASSS (pilot exercise); (3) development of a training module (the Spondyloarthritis Radiography [SPAR] module) that clarifies definitions, rules, and scoring methodology and a set of reference radiographic images; (4) scoring exercise 1 by 6 readers on 39 patients, where baseline and 2-year radiographs were scored blinded to time point; and (5) revision of the SPAR module followed by scoring exercise 2 conducted by the same 6 readers on 35 patients. Reliability of status and 2-year change scores was assessed by the intraclass correlation coefficient (ICC) method. ICCs for change scores for the radiologist reader pair improved from 0.46 to 0.62 after minimal calibration with the SPAR module. Recalibration from exercise 1 to exercise 2 with the SPAR module led to substantial improvement in interreader reliability for change in mSASSS score from ICC 0.44 (range 0.31-0.62) to ICC 0.62 (range 0.34-0.84). Simultaneous assessment of anteroposterior and lateral lumbar radiographs did not enhance reliability or detection of progression. Calibration according to the SPAR module led to improved reliability in the scoring of the mSASSS, even for expert readers.