Background Sacroiliitis on MRI and X-rays play an important role in the ASAS axial spondyloarthritis (axSpA) criteria1. Though, recognition of sacroiliitis on X-rays (X-SI) and MRI of the sacroiliac joints (MRI-SI) can be challenging, resulting in misinterpretations. Usually the reading in clinical trials is done by ≥1 trained readers. In cohorts it varies and in the DEvenir des Spondylarthropathies Indifferenciées Récentes (DESIR)-cohort, X-SI and MRI-SI at inclusion are read by a local radiologist/rheumatologist. The impact on classification of patients (pts) by local read (LocR) instead of centralized read (CentR) is unknown.
Objectives To investigate the difference in classification of pts (ASAS axSpA) using LocR versus CentR as external standard.
Methods In the DESIR-cohort, pts aged 18-50 with inflammatory back pain (IBP; ≥3 months, ≤3 years) were included (n=708). Local radiologists/rheumatologists read all baseline X-SI and MRI-SI; X-SI according to a method derived from the modified New York (mNY) criteria2 (grade 2 and 3 pooled in one combined grade “DESIR-2”). Sacroiliitis was defined by at least unilateral ≥DESIR-2. Sacroiliitis on MRI was defined as definite inflammatory lesions in ≥1 SI-joint. Next, 2 well-calibrated central readers independently read all X-SI (original mNY) and MRI-SI (ASAS3). An experienced radiologist was adjudicator in case the 2 readers disagreed. An image was marked positive if 2/3 readers agreed. Subsequently, LocR was compared to CentR and to the reads by the central readers separately; pts were classified (ASAS axSpA), using both LocR and CentR (external standard).
Results Pts with onset IBP <45 and complete X-SI and MRI-SI (n=582) were included. LocR and CentR differed in 163/582 pts (28%; 91 X-SI; 59 MRI-SI; 13 both X-SI and MRI-SI). In 46/582 pts (7.9%), a different read resulted in a different classification; 18 no-SpA pts (3.1%) based on CentR were classified as axSpA using LocR (14 with positive X-SI); 28 axSpA pts (4.8%; 13 mNY+) based on CentR were classified as no-SpA using LocR (table). Among the patients classified as axSpA, additional discrepancies occurred if fulfilling the imaging arm was considered; 16 axSpA/582 pts (2.7%; 8 mNY+) fulfilled the imaging arm based on CentR but fulfilled the clinical arm based on LocR, and 29 axSpA pts (5.0%) fulfilled the clinical arm based on CentR but fulfilled the imaging arm based on LocR (table). Comparisons of LocR versus the separate readers show very similar results (table).
Conclusions Looking at the complete ASAS axSpA criteria, the classification changed in 7.9% of the pts when using LocR instead of CentR. However, when interested in whether pts fulfil the imaging arm or not, changes are seen in an additional 8.2% of the pts resulting in 15.6% of the pts classified differently.
Rudwaleit ARD 2009;68:777-83.
van der Linden A&R 1984;27:361–8.
Rudwaleit ARD 2009;68:1520-7.
Disclosure of Interest None declared