Abstract
OBJECTIVES: To determine and compare the diagnostic accuracy of imaging tests for the prediction of RA progression in people with inflammatory joint pain or clinically suspect arthralgia (CSA). METHODS: We searched MEDLINE, Embase and Web of Science from 1987 to March 2024. Studies evaluating any imaging tests in participants with inflammatory joint pain or CSA without clinical synovitis were eligible. Reference standards included RA classification criteria, methotrexate initiation or development of inflammatory arthritis (IA). Two authors independently extracted data and assessed validity according to QUADAS-2. We estimated summary sensitivities and specificities for each imaging characteristic and fitted bivariate and hierarchical SROC models for meta-analysis where possible. RESULTS: We found 39 eligible studies including 42 cohorts, of which 12 evaluated MRI (n = 2782; 19% with RA/IA), 26 evaluated ultrasound (US) (n = 6805; 25% with RA/IA) and 10 evaluated other imaging tests (n = 3362; 20% with RA/IA). Summary sensitivity and specificity for US Power Doppler ≥1 in at least one joint were 37% (95%CI 18%-60%) and 90% (95%CI 82%-94%), respectively (seven studies). Summary sensitivity and specificity for MRI synovitis in at least one joint were 45% (95%CI 29%-62%) and 84% (95%CI 66%-94%), respectively (four studies). Lack of consensus regarding positive threshold definitions limited meta-analysis for other imaging features. CONCLUSION: Evidence for MRI or US in predicting RA/IA in people with CSA is heterogeneous and of variable quality. Further studies with larger sample sizes, longer follow-up times and uniform imaging test scoring are warranted to determine whether imaging characteristics, in combination with clinical information, can predict RA in this population. SYSTEMATIC REVIEW REGISTRATION: PROSPERO: https://www.crd.york.ac.uk/prospero, CRD42024501243.