Objective. To examine the effect of alleles encoding the 'shared'/'rheumatoid' epitope on rheumatoid arthritis (RA) disease severity in patients who participated in the minocycline in RA (MIRA) trial. Methods. Of 205 patients with a week-48 visit, blood was available for typing of HLA- DRB1 and HLA-DQB1 in 174 (85%) and successfully completed in 169 (82%). Baseline erosions were used to assess disease severity and new erosions at the last visit served as a proxy for progression. Results. At baseline, there was no association between the presence of erosive disease or rheumatoid factor status and the dose of rheumatoid epitope (homozygous, heterozygous, none) or the specific alleles identified. At the final visit, a gradient was observed for the 3 allelic subgroups (and their gene doses) in the occurrence of new erosions among the Caucasian placebo-treated, but not the minocycline- treated, patients. A treatment group/HLA-DR4 epitope interaction was demonstrated in multivariate analyses. Approximately two-thirds of African- American patients did not have the rheumatoid epitope. Conclusion. HLA-DRB1 oligotypig may be useful in predicting the progression of disease in some Caucasian patients. Our study corroborates the infrequency of the epitope among African-American patients with RA.