To assess agreement among methods for classifying patients with inflammatory back pain (IBP) after a 2-year follow-up.

Patients with IBP in the French nationwide, longitudinal, prospective cohort DESIR were classified after 2years based on imaging findings, rheumatologist's confidence in a diagnosis of spondyloarthritis, three classification criteria sets (axial Assessment of Spondyloarthritis international Society [ASAS], European Spondylarthropathy Study Group [ESSG], and Amor) and treatment (TNFα antagonists). Agreement among these methods was assessed by computing the percentage of concordant classifications and Cohen's kappa coefficient. Using logistic regression, we identified the items most strongly associated with rheumatologist's confidence.

Agreement among criteria sets was poor (kappa<0.6), even in the group with inflammation by magnetic resonance imaging. Of 708 patients, 541 had all available data including rheumatologist's confidence after 2years, which was 0/10 for 31 (5.7%) patients, 1/10 to 7/10 for 158 (29.2%) patients, 8/10 or 9/10 for 167 (30.9%) patients, and 10/10 for 185 (34.2%) patients. TNFα antagonists were used in 156/356 (43.8%) patients in the two highest confidence groups versus 53/188 (28.2%) patients in the two lowest confidence groups. Factors independently associated with confidence ≥8/10 were fulfilment of ASAS, ESSG, and Amor criteria.

Confidence of rheumatologists in the diagnosis of spondyloarthritis in patients with recent-onset IBP shows limited agreement with classification criteria. The best way to currently classify spondyloarthritis should be the association of both at least one classification criteria and a diagnosis of spondyloarthritis according to the rheumatologist.