OBJECTIVE--To evaluate the clinical usefulness of genomic HLA typing during the first five years of established rheumatoid arthritis (RA). METHODS--The HLA-DRB and -DQB alleles were determined by restriction length polymorphisms and polymerase chain reaction amplification with sequence specific primers in 99 Swedish patients with RA. Clinical features after two and five years disease duration were related to the genetic pattern. Seventy four patients were seropositive, 25 had nodules, 90 developed erosions, and 15 required joint replacements. Twelve patients were in remission after five years. Disability was assessed by health assessment questionnaire, and radiographic damage in hands and feet by the Larsen method. RESULTS--Eighty seven per cent of the patients carried the conserved third hypervariable region sequence (HVR3), 32% had DRB1*04 on one allele, and 26% had DRB1*04 on both alleles (all frequencies significantly greater than in controls). Frequencies of DRB1*04 associated DQB*0301 and *0302 were normal. Patients carrying DRB1*04 on both alleles tended to have more radiographic changes after two years, but this difference had diminished after five years. Disability did not vary with regard to the genotype. Homozygous HVR3 patients had about three times greater risk of undergoing joint replacement. Homozygosity for HVR3 and presence of DQB*0302 both tended to be associated with erosive disease. CONCLUSIONS--We confirmed a strong association of disease with the presence of the shared epitope on one or two alleles. However, genotype was not strongly associated with disease severity after two and five years disease duration, and thus the value of genomic typing to select patients for early aggressive therapy is questionable.