It has been suggested that the negative association between rheumatoid arthritis (RA) and oral contraceptive (OC) use might be limited to the more severe forms of RA. To investigate this further, we studied 121 consecutive female patients with definite RA, 52 female patients with probable RA, and 378 female controls. All patients had RA symptoms of recent onset. After a mean followup period of 6 years, patients with definite RA were classified as having either a severe disease course (n = 76) or a mild disease course (n = 45). The negative association between OC use prior to the onset of RA symptoms and the development of RA was limited to those patients with definite RA who had a severe disease course. We therefore conclude that OC use prior to the onset of RA symptoms is only associated with a reduction in the incidence of severe RA. This may explain the divergent results of previous studies.