Abstract

Objective: To assess performance of radiologists and rheumatologists in detecting sacroiliitis

Methods: 100 rheumatologists and 23 radiologists participated. One set of films was used for each assessment, another for training, and the third for confidence judgment. Films of HLA-B27+ patients with AS were used to assess sensitivity. For specificity films of healthy HLA-B27− relatives were included. Plain sacroiliac (SI) films with simultaneously taken computed tomographic scans (CTs) were used for confidence judgment. Three months after reading the training set, sensitivity and specificity assessments were repeated. Next, participants attended a workshop. They also rated 26 SI radiographs and 26 CTs for their trust in each judgment. Three months later final assessments were done.

Results: Sensitivity (84.3%/79.8%) and specificity (70.6%/74.7%) for radiologists and rheumatologists were comparable. Rheumatologists showed 6.3% decrease in sensitivity after self education (p=0.001), but 3.0% better specificity (p=0.008). The decrease in sensitivity reversed after the workshop. Difference in sensitivity three months after the workshop and baseline was only 0.5%. Sensitivity <50% occurred in 13% of participants. Only a few participants showed changes of >5% in both sensitivity and specificity. Intraobserver agreement for sacroiliitis grade 1 or 2 ranged from 65% to 100%. Sensitivity for CT (86%) was higher than for plain films (72%) (p<0.001) with the same specificity (84%). Confidence ratings for correctly diagnosing presence (7.7) or absence (8.3) of sacroiliitis were somewhat higher than incorrectly diagnosing the presence (6.6) or absence (7.4) of sacroiliitis (p<0.001).

Conclusion: Radiologists and rheumatologists show modest sensitivity and specificity for sacroiliitis and sizeable intraobserver variation. Overall, neither individual training nor workshops improved performance.