Background/Purpose: The diagnosis of ankylosing spondylitis
(AS) is based on pelvic radiographs plagued by poor sensitivity, specificity,
and reproducibility. Many AS patients, particularly those with inflammatory
bowel disease (IBD), may have CT scans performed for other clinical indications
and sacroiliitis may be incidentally noted. Though the modified New York (mNY) criteria
have never been validated in CT imaging, previous studies have used a radiologist’s
adaptation of the criteria as a gold standard for diagnosing sacroiliitis.

Our objective is to develop a validated scoring system for
sacroiliitis on CT that can ground future studies in prevalence and
pathogenesis.

Methods: Patients from the Toronto AS clinic meeting mNY
criteria for AS who had CT scans of the abdomen/pelvis were matched to controls
by age and gender. Control patients had their charts reviewed to ensure they
had no history of spondylitis, colitis, uveitis, or psoriasis. A training
exercise involving 10 CT scans (5 AS and 5 controls) was conducted to identify
candidate features and to optimize reliability. A derivation cohort of 24 CT
scans (12 AS and 12 controls) was used to test these features. Finally, 2
blinded readers performed a validation study on 68 CT scans (34 AS and 34
controls).

SI joints were divided into left and right as well as iliac
and sacral segments for a total for 4 segments. The maximum number of erosions
seen on a single slice was counted for each segment. The sum of these values
gave a total erosion score. Sclerosis was only measured on the slice with the
longest synovial length. Inter- and intra-observer values, sensitivity,
specificity, and likelihood ratios (LR) were calculated for variables that
correlated with AS. Combinations of variables were trialed to maximize
sensitivity and specificity.

Results: Features with the highest +LR included ankylosis,
number of erosions, iliac sclerosis >0.5cm, and sacral sclerosis >0.3cm. Inter-reader
reliability for these variables were 1.0 for ankylosis, 0.99 for number of
iliac erosions, 0.99 for number of sacral erosions, 0.58 for iliac sclerosis,
and 0.39 for sacral sclerosis. Fig 1A demonstrates the ROC curves for the total
erosion number as well as the increasing depth of sclerosis. A total erosion
number of ≥3 erosions was found to have the highest sensitivity and
specificity for AS. Fig 1B demonstrates the ROC curves for combinations of
ankylosis, sclerosis, and erosions for diagnosing AS. Sclerosis was defined as
either >0.5cm of iliac or >0.3cm of sacral sclerosis >1cm in length.
The presence of >1cm of ankylosis or ≥3 total erosions resulted in a sensitivity
of 91% and specificity of 91%. The addition of >0.5cm of iliac sclerosis or
>0.3cm of sacral sclerosis marginally increased the sensitivity to 94% but
decreased specificity to 86%.

Conclusion: It is proposed that the presence of ankylosis
>1cm or ≥3 total erosions has the greatest diagnostic utility for AS.