From 1971 to 1985, 393 total knee arthroplasties (TKAs) were performed in patients for rheumatoid arthritis. Of these, 112 used polycentric prostheses and 131 used total condylar prostheses. One hundred seven of the patients with polycentric prostheses and 102 of the patients with total condylar prostheses were followed for a minimum of two years (average, 61 and 55 months, respectively). At the time of the latest examination, nine knees with polycentric prostheses required revision surgery: six for tibial component loosening and instability and three for patellofemoral pain. Five knees with total condylar prostheses required revision: one for chronic instability, one for a late posttraumatic patellar fracture, and three for late hematogenous infections. The total condylar prostheses with patellar resurfacing had better scores for pain (8.5 points) than the polycentric prosthesis without patellofemoral resurfacing (7.0 points). The total condylar prosthesis appeared more durable in terms of fixation, with no evidence of aseptic loosening of femoral or tibial components noted in this series thus far.