Which definitions for MRI sacroiliac joint lesions predict the diagnosis of early axial spondyloarthritis best? A 2-year follow-up in the SPACE cohort

Background: The Leiden group and the Assessment of Spondyloarthritis International Society (ASAS)-MRI study group have proposed definitions for structural (SL) and active (AL) lesions typical for axial spondyloarthritis (axSpA) on MRI of the sacroiliac joints (MRI-SIJ). Objectives: To analyse the predictive validity of proposed SL and AL MRI-SIJ definitions in early axSpA and compare proposed AL with the current ASAS-MRI-SIJ+ definition. Methods: Patients with chronic back pain (≤2 years) from the Spondyloarthritis Caught Early cohort were diagnosed as axSpA or non-axSpA after 2 years follow-up. Three central readers scored baseline MRI-SIJ for SL (erosions and fat lesions) and AL (bone marrow oedema). Validation required specificity and positive predictive value (PPV) ≥95%. Results: Among 643 patients (52% axSpA), SL were infrequent (2%-14%). All Leiden and most MRI study group SL definitions met the validation threshold, except for 'erosion in ≥2 consecutive slices' and the overall MRI study group definition (PPV <95%). The ASAS-MRI-SIJ+ definition had a higher sensitivity than the MRI study group AL (40% vs 31%) with similar specificity (98% vs 99%). Combining SL and AL, the Leiden SL with ASAS-MRI-SIJ+ definition met the validation threshold with the highest sensitivity (46%). SL increased sensitivity beyond AL alone by 6%-11%. Conclusions: The ASAS-MRI-SIJ+ definition outperforms the MRI study group AL. The Leiden SL combined with the ASAS-MRI-SIJ+ definition is validated, most sensitive and feasible as it simultaneously upholds lesion quantification and detection precision, making it the preferred approach. Nevertheless, SL were uncommon in early axSpA, contributing only marginally beyond AL definitions in early diagnosis.