Résumé
Objectives: to compare sacroiliac joint (SIJ) lesions observed on MRI in women with versus without axial spondyloarthritis (ax-SpA), and to establish an algorithm to determine whether such lesions are due to ax-SpA, with better specificity than the current ASAS MRI criteria. Methods: the following lesions were assessed at the SIJ in two groups of women, one with and another without confirmed ax-SpA: bone marrow oedema (BME), sclerosis, erosions, osteophytes, and ankylosis. Using the ASAS criteria as gold standard for diagnosis of axSpA, the sensitivity and specificity were calculated for different combinations and characteristics of lesions. Results: compared to women without ax-SpA, women with ax-SpA had more BME (61%vs17%, p<0.001), sclerosis (40%vs22%, p<0.001), erosions (35%vs5%, p<0.001), and ankylosis (2%vs0%, p=0.007), but less osteophytes (5%vs33%, p<0.001). The ASAS MRI criteria had 59% sensitivity and 88% specificity, while a new algorithm achieved 56% sensitivity and 95% specificity using the following criteria: no osteophytes at the SIJ and either (i) BME at the SIJ with at least one dimension ≥8mm or (ii) at least one erosion at the SIJ. Conclusions: we recommend the following new algorithm for diagnosis of ax-SpA using MRI: no osteophytes at the SIJ and either (i) BME at the SIJ with at least one dimension ≥8mm or (ii) at least one erosion at the SIJ. The false positive rate when using our algorithm (3.3%) is less than half than when using the ASAS MRI criteria (7.7%), and thus its application in clinical practice could reduce overdiagnosis and prevent overtreatment of ax-SpA.