Comparison of three classification criteria of rheumatoid arthritis in an inception early arthritis cohort.

The aim of this study is to compare the three classification criteria for rheumatoid arthritis (RA) in a large cohort of early arthritis patients. Patients who had at least one clinically swollen joint with disease duration no more than 1 year and age more than 18 years were enrolled. The clinical and laboratory parameters were recorded. The patients were diagnosed by two experienced rheumatologists. Undiagnosed patients were followed up every 3 months until 1 year. The sensitivity, specificity, and predictive value were compared among the early RA (ERA) criteria, the 1987 ACR criteria, and the 2010 ACR/EULAR criteria in this inception cohort of early arthritis patients. A total of 417 patients with inflammatory arthritis were recruited. By the end of 1 year follow-up, there were 399 patients (95.7 %) with a definitive diagnosis and 18 (4.3 %) patients remained as undifferentiated arthritis. Among the patients with definitive diagnosis, 202 (50.6 %) patients were diagnosed with RA and 197 (49.4 %) with non-RA. The sensitivity of ERA criteria was equal to 2010 ACR/EULAR criteria (both were 72.3 %), but much higher than 1987 ACR criteria (72.3 vs. 39.1 %, P < 0.001); the specificity of ERA criteria was comparable to 2010 ACR/EULAR criteria (87.8 vs. 83.2 %) and slightly lower than 1987 ACR criteria (87.8 vs. 92.4 %, P < 0.001). Unlike the complicated scoring system of 2010 criteria, the ERA criteria were more feasible to use in practice with five criteria only. The ERA criteria have a high sensitivity and more clinically feasibility in daily practice for early RA diagnosis.