Fifty-eight patients with rheumatoid arthritis (RA) entered a double blind trial of auranofin (AF) designed to assess dose response relationships and longterm outcome. Multivariate analysis of repeated measures with trend analysis and discriminant function analysis of standard measures of RA activity were applied to a randomized double blind trial of AF at daily doses of 4, 6 and 8 mg over 6 months. Improvement occurred in each group. There was a highly significant (p less than 0.001) linear trend in the 6 mg group, 73% of whom showed linear improvement. A significant correlation (p less than 0.05) was found between response of individual patients and AF dose (mg/kg/day), but there was no significant correlation between dosage and mean steady state serum gold concentration. No significant correlation was seen between outcome and pretreatment demographic and disease variables. In a subsequent 6 month phase of dosage adjustment, aiming for optimal dosage, no advantage resulted from increasing the dose above 6 mg/day. Patients apparently benefiting from treatment continued an open long-term trial of AF. By 45 months, 33.5% had stopped treatment due to lack of efficacy and 14.5% due to toxicity, mainly rash and diarrhea.