Objective: To determine how well the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) 1987 classification criteria for rheumatoid arthritis (RA), when used at study inclusion in a cohort of 270 patients with early (<1 year) arthritis, predicted a diagnosis of RA 2 years later and how well they classified these patients at the end of the 2 years.
Methods: Patients were evaluated during 1995-1997 at 7 hospitals in the Brittany region of France. Patients were evaluated at 6-month intervals until November 1999. The diagnosis made by a panel of 5 rheumatologists (P5R) after the last visit was used as the "gold standard." The ACR 1987 criteria for RA were applied prospectively, without taking into account the initial diagnosis.
Results: At the last visit (mean +/- SD followup 29.1 +/- 11.8 months; median 30 months), the P5R diagnosed RA in 98 patients. At the last visit, classification by the ACR criteria was satisfactory, and the combination of an office-based rheumatologist's (OBR's) diagnosis of RA and fulfillment of the ACR criteria was sensitive (87%; 85 of 98 RA patients had both) and highly specific (99%; 170 of 172 non-RA patients did not have both). Application of the criteria at the first visit was of limited value for predicting a diagnosis of RA 2 years later.
Conclusion: After a 2-year followup, the ACR 1987 classification criteria used in combination with an OBR's diagnosis were effective in distinguishing patients with and without RA. The criteria were not useful for predicting RA in patients with arthritis onset within the previous year. Some patients who met the criteria at baseline and after 2 years did not have RA, suggesting that incorporating exclusion criteria may improve the performance of the ACR criteria when used without taking into account the diagnosis by a rheumatologist, particularly in early arthritis.