Objective: We assessed levels of agreement between a diagnosis of rheumatoid arthritis (RA) at inclusion in a recent-onset arthritis cohort, then 2 and 10 years later. Performance of American College of Rheumatology (ACR) criteria alone or combined with rheumatologist diagnosis, and of recent new criteria adding antibodies to cyclic citrullinated peptides ("anti-CCP-revised criteria") to existing ACR criteria, was evaluated.
Methods: In total, 270 patients with recent-onset arthritis of less than 1 year duration were included between 1995 and 1997 and followed for 2 years. A diagnosis was recorded by an office-based rheumatologist (OBR) at inclusion, then 2 years later. In 2007, a questionnaire was sent to each rheumatologist to collect the final diagnosis, which was considered the reference.
Results: Final diagnosis was available for 164 patients: 57 had RA. Agreement was low (kappa = 0.27) between the baseline and final diagnoses, and substantial (kappa = 0.69) between the 2-year and final diagnoses. Anti-CCP-revised criteria had sensitivity of 65% to 81% and specificity of 55% to 75%. Sensitivity and specificity of ACR criteria were 57.9% (44.1%-70.9%) and 74.8% (65.5%-82.7%) at inclusion, 80.7% (70.5%-90.0%) and 63.6% (54.5%-72.7%) at 2 years. The combination OBR diagnosis/ACR criteria after 2 years showed considerably increased specificity (87% vs 64%) and slightly decreased sensitivity (77% vs 81%).
Conclusion: ACR criteria for RA showed poor performance even at 2 years. The absence of exclusion criteria may explain the lack of specificity, which improved when combined with the OBR diagnosis. Adding anti-CCP criteria to the existing criteria could help in diagnosing RA.