Since 1991, 5760 knee arthroplasty procedures done by 53 surgeons have been registered in a community joint implant registry and were reviewed regarding initial revision done within the healthcare system. The 168 revisions done represented 2.9% of the knee arthroplasties between September 1991 and December 2002. Survival was defined as the absence of revision surgery. Death was considered a censored event. Cumulative survival rates for the different total knee arthroplasty configurations were: cemented total knee arthroplasty with all-polyethylene tibia, 99.2%; cemented total knee arthroplasty with metal-backed tibia, 96.3%; hybrid total knee arthroplasty, 89.3%; and unicondylar knee arthroplasty, 87.2%. Cemented total knee arthroplasty with metal-backed tibia had better survival than hybrid total knee arthroplasty, ingrowth total knee arthroplasty, and unicondylar knee arthroplasty. Cemented total knee arthroplasty with a metal-backed tibia did not have better survival than cemented total knee arthroplasty with an all-polyethylene tibia. Gender was not related to survival. Age was related to survival, with older patients' knees surviving longer. Aseptic loosening or wear was the cause of revision in 40.8% of patients having total knee arthroplasty and 46.6% of patients having unicondylar knee arthroplasty, whereas progression of arthritis necessitated unicondylar knee revision in 51.2% of patients having that procedure. This study presents further evidence of the value of and ongoing need for total joint registries. Cemented total knee arthroplasty with all-polyethylene tibia and with metal-backed tibia showed more than 95% 10-year cumulative survival. Hybrid total knee arthroplasty, ingrowth total knee arthroplasty, and unicondylar knee arthroplasties did not show such good results.