Background/Purpose: There is growing recognition
of the importance of structural lesions of the sacroiliac joints (SIJ) on MRI
to the diagnostic evaluation of axial spondyloarthritis
(axSpA) but it is unclear to what degree current
definitions for such lesions and methods for evaluation of the SIJ allow
reliable assessment between different centers and readers. We aimed to assess
reliability of evaluation of structural SIJ lesions on MRI without conducting
calibration exercises between readers at two centers, readers at one center
being internally calibrated and having extensive experience in evaluation of
structural lesions in the SIJ.  

Methods: All readers simultaneously evaluated T1-weighted
and short tau inversion recovery (STIR) scans from 9 patients with non-specific
back pain and 23 with axial SpA from 2 centers
(Leiden=12, Edmonton=20). All readers were internally calibrated, and readers
in one center have extensive experience in evaluation of structural lesions in
the SIJ. Scans were read blinded to patient characteristics, diagnosis, and
scan origin, and data entered directly online using a web-based schematic of
the SIJ.  Bone marrow edema (BME)
and structural lesions (erosion, fat metaplasia, sclerosis, ankylosis)
were scored on consecutive semi-coronal slices through the SIJ using a binary
method based on presence/absence of the lesion in SIJ quadrants except for ankylosis, which was assessed according to SIJ halves.  Slices were selected according to the
methodology used in the Spondyloarthritis Research
Consortium of Canada (SPARCC) MRI SIJ inflammation scoring method.  Interobserver
reliability was calculated by intra-class correlation coefficient (ICC) and
comparisons made with pre-specified expert readers.

Results: Mean (SD) scores for BME, erosion, fat
metaplasia, sclerosis, and ankylosis were 5.3 (7.1),
6.1 (5.5), 6.5 (7.5), 2.5 (3.4), 0.3 (1.2) with
significantly higher scores being recorded for BME, erosion, and fat metaplasia
by expert readers.  Mean (range) for
sensitivity/specificity for expert clinician diagnosis of SpA
was 0.85 (0.75-0.91)/0.88 (0.79-1.00), sensitivity of expert readers being
higher (0.90 vs 0.81) though specificity was lower
(0.85 vs 0.92). 
Interobserver reliability for structural
lesions was comparable between groups of readers and approached the reliability
of BME for erosion, and fat metaplasia (Table). Reliability for sclerosis and ankylosis varied although few cases demonstrated
significant involvement with these lesions.

Conclusion: The presence and extent of erosion and
fat metaplasia can be reliably assessed by simultaneous evaluation of T1W and
STIR scans even without calibration between readers and seems promising for
diagnostic evaluation. Evaluation of ankylosis and
sclerosis, on the other hand, requires further study in cohorts that have more
cases with these lesions. 

Table. ICC [95% confidence intervals]