We have recently proposed a new hypothesis to explain the association of Human Leukocyte Antigen (HLA) with rheumatoid arthritis (RA) predisposition. In this model, which challenges the Shared Epitope (SE) hypothesis, HLA-DQ predisposes while HLA-DR protects. In the present study, we have compared these two models in an Early Arthritis Clinic started in 1993 in the Department of Rheumatology at the Leiden University Medical Centre. Out of 524 patients who enrolled this programme in the period 1993-1998 and completed the one year follow-up, 155 have been classified as RA. These patients along with 306 consecutive cadaveric renal organ donors have been typed for HLA-DR and -DQ. The distributions of predisposing DR alleles according to SE, and predisposing DQ and protective DR according to our model were analysed. We found that two doses of predisposing DQ alleles strongly predisposed to RA, even in individuals with a single dose of SE while DRB1 alleles carrying the motif DERAA confered a dominant protection in DQ5-positive individuals. We conclude that the present findings are consistent with our previously described model of HLA and RA association. Using this new model, we have been able to characterise two novel groups of individuals on the basis of their HLA typing: one strongly predisposed to RA and one protected. Knowing the mechanism of HLA-related dominant natural protection may help in designing novel treatment modalities for RA.