Rheumatoid arthritis is a frequently and potentially severe disease which causes a functional handicap in nearly half the patients 10 years after the identification of the first clinical manifestations. Some patients develop very severe forms with joint destruction and multiple organ involvement while in others the disease remains benign, even after a long clinical course of several years. Theoretically, the future intensity of rheumatoid arthritis in a given patient cannot be predicted at the time of early diagnosis. No prognosis factor has been identified and universally accepted and validated. A marker of prognosis would be highly appreciated by clinicians who could then more closely adapt their management decisions to the disease potential. Clinical and biological data collected to date have provided a limited amount of prognostic information but recent progress in molecular biology suggests that genetic markers could be correlated with disease severity. Several HLA-DRB1 alleles including DR1*0401, DRB1*0404, DRB1*0101 and sometimes DRB1/1001 and DRB1/1402 are potential markers. In France 85% of the patients with rheumatoid arthritis have one of these genes compared with 25% in the general population. In patients with a "high risk" alleles, the second haplotype could also have prognostic value. It would appear possible to distinguish immunogenetically homogeneous subpopulations corresponding to the more severe forms of the disease. It is still too early to propose therapeutic strategies based on current prognosis markers, but a combination of the most pertinent markers should be already used to select homogeneous subsets of patients in fundamental research and clinical trials.