From 1971 through 1987, 9200 total knee arthroplasties were performed at the Mayo Clinic. Actuarial analysis was used to estimate cumulative rates of survival. Use of a proportional-hazard, general linear model led to the identification of four independent variables that were associated with a significantly lower risk of failure: primary total knee arthroplasty, a diagnosis of rheumatoid arthritis, an age of sixty years or more, and use of a condylar prosthesis with a metal-backed tibial component. When all four of these favorable variables were present (without regard for radiographic changes and non-disabling symptoms), the probability of an implant being in situ was 97 per cent at both five and ten years.