Fifty-two patients with severe rheumatoid arthritis (RA) from four Australian centres were randomised to receive cyclosporin A (CSA) (n = 25) or azathioprine (AZA) (n = 27) for six months. Initial mean doses of CSA and AZA were 4.2 mg/kg and 1.7 mg/kg respectively. The mean doses of CSA and AZA at six months were 3.4 mg/kg and 1.9 mg/kg. Assessments of side-effects and outcomes of benefit were made monthly by independent, blinded observers. Both treatment groups exhibited statistically significant improvement in standard outcome parameters when compared with baseline values. However, there were no statistically significant differences in these parameters between the two groups. There was a mean increase in serum creatinine concentration associated with CSA; no persons were withdrawn from the study for this reason. Seven CSA recipients (three gastrointestinal symptoms, two neurological symptoms, two other) and 12 AZA recipients (six gastrointestinal symptoms, four inefficacy, two other) withdrew from treatment prematurely. Seven CSA recipients became hypertensive and four required anti-hypertensive therapy. Adverse events not requiring cessation of therapy were more commonly seen among CSA patients. In this group of severely affected patients with RA both cyclosporin and azathioprine were effective therapies. CSA toxicities were predictable and manageable but required close monitoring.