Background/Purpose: Reading of MRIs of the sacroiliac joints (MRI-SI) in clinical trials is usually performed by ≥1 trained readers while in daily practice this is done by local radiologists/rheumatologists. However, this varies in cohorts and in the DEvenir des Spondylarthropathies Indifferenciées Récentes (DESIR)-cohort, MRIs-SI at inclusion were first read by the local radiologist/rheumatologist, then by central readers. The impact of reading by multiple readers in various centres as in daily practice, instead of a centralized reading, is unknown. We compared the local reading (LocR) to centralized reading (CentR) regarding the presence or absence of inflammation on MRI-SI.

Methods: The 25 participating centers included patients aged 18-50 with inflammatory back pain (IBP; ≥3 months, ≤3 years) in the DESIR-cohort (n=708). Available baseline MRIs-SI were read by local radiologists/rheumatologists with access to clinical and laboratory data, on the presence of inflammatory lesions in both SI-joints. A grade 0 corresponds to ‘normal’, a grade 1 to ‘doubtful’, and a grade 2 to ‘definite inflammatory lesions’. For this analysis, a positive MRI was defined as at least one SI-joint marked grade 2. Next, 2 well-calibrated central readers independently read all MRIs-SI according to the ASAS definition¹, blinded for clinical and laboratory data. In case the readers disagreed, an experienced radiologist served as adjudicator. An MRI-SI was marked positive if 2/3 readers agreed. Agreement between the 2 central readers, between LocR and CentR and between LocR and the central readers separately was calculated (Kappa; % agreement).

Results: In this analysis patients with complete MRI-SI data (n=663) were included. Inter reader agreement between the 2 central readers is acceptable (Kappa 0.73), and the percentage agreement (87.5%) is good (table). The adjudicator scored 84/663 (12.7%) MRIs-SI because of disagreement between the 2 central readers. Comparison between CentR (2/3) and LocR shows the same levels of agreement (kappa 0.70, % agreement 86.6%; table). In 38/663 patients (5.7%), the MRI-SI was positive by LocR but negative by CentR; in 51 patients (7.7%) it was the other way around. There was no difference in agreement between LocR and CentR if MRIs-SI were scored by local rheumatologists or by local radiologists (data not shown). Comparisons of LocR versus the separate readers show very similar results (table).

Conclusion: Both inter reader agreement between the 2 central readers and agreement between the local and centralized readings is acceptable to good. This indicates that local rheumatologists/radiologists perform as good as trained readers in identifying inflammation on MRI-SI in patients with recent onset IBP, thereby suggesting that MRI-SI is a reliable assessment in diagnosing and classifying the majority of patients with spondyloarthritis.

References: ¹Rudwaleit ARD 2009;68:1520-7

		Central reader 2
Central reader 1		MRI-SI+ (ASAS)	MRI-SI- (ASAS)
	MRI-SI+ (ASAS)	200	56
	MRI-SI- (ASAS)	28	380
	Kappa (95% CI) / Agreement (%)	0.73 (0.67-0.78)	87.5
		Centralized reading (2/3)
Local reading		MRI-SI+ (ASAS)	MRI-SI- (ASAS)
	MRI-SI+	185	38
	MRI-SI –	51	389
	Kappa (95% CI)  / Agreement (%)	0.70 (0.65-0.76)	86.6
		Central reader 1
Local reading		MRI-SI+ (ASAS)	MRI-SI- (ASAS)
	MRI-SI+	180	43
	MRI-SI –	76	364
	Kappa (95% CI) / Agreement (%)	0.61 (0.55-0.67)	82.1
		Central reader 2
Local reading		MRI-SI+ (ASAS)	MRI-SI- (ASAS)
	MRI-SI+	177	46
	MRI-SI –	51	389
	Kappa (95% CI)  / Agreement (%)	0.67 (0.62-0.73)	85.4

---

« Back to 2013 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/concordance-between-a-positive-mri-of-the-sacroiliac-joints-based-on-the-local-reading-versus-a-centralised-reading-experience-from-the-desir-cohort/