Twenty-four cases of septic arthritis in rheumatoid arthritis patients were compared with 99 cases of septic arthritis in patients without rheumatoid arthritis. In addition, 238 previously published cases of septic arthritis with rheumatoid arthritis were analyzed. Fifteen percent of our patients with septic arthritis had rheumatoid arthritis, which was typically of long duration (mean 15 years), erosive, and seropositive. Fifty-four per cent (28% in the literature) and 9% of patients with and without rheumatoid arthritis, respectively, had pyarthrosis of multiple joints. The knee represented one-third of infected joints and the elbows and wrists were more often infected in patients with than without rheumatoid arthritis. S. aureus was recovered in 80% versus only 60% of patients with and without rheumatoid arthritis, respectively. The source of sepsis was often a skin lesion, in particular at the foot, emphasizing the need for early orthopedic treatment of deformities responsible for skin lesions. Monoarticular infection was more likely to be due to an intraarticular injection. Mortality rate was 17% in patients with rheumatoid arthritis (23% in the literature) versus 7% in patients without rheumatoid arthritis. Staphylococcal infection and infection of multiple joints were associated with higher mortality rates (35% and 49%, respectively). The mortality rate in polyarticular infections has failed to decline over the last 35 years. Initial failure to distinguish septic arthritis from an exacerbation of rheumatoid arthritis contributes to the high mortality rate. The diagnosis of septic arthritis rests on a high index of suspicion. Septic arthritis cannot be ruled out based on absence of local inflammation, fever, or hyperleukocytosis or on presence of inflammation of multiple joints. Joint fluid specimens should routinely be sent to the microbiological laboratory and should be inoculated in blood culture bottles at the least suspicion.