Ninety-six patients who had a total of 108 replacements with an uncemented porous-coated anatomic knee prosthesis were followed for an average of sixty-four months (range, thirty-nine to ninety-three months). At the time of the most recent follow-up, twenty-one implants (19 per cent) had failed, all due to problems with the tibial component. A patellar component was not used, and no noteworthy patellar problems were encountered after the operation. There were no infections about the prostheses, and no femoral implant was revised. The most common cause of failure was collapse of the anteromedial part of the tibial plateau, which occurred in fourteen knees. The prosthesis loosened without collapse of bone in two knees, and five knees were revised because of gross wear of the polyethylene. When the time of failure was defined as the point at which revision of the prosthesis was recommended, the cumulative rate of survival was 84 per cent (95 per cent confidence interval, +/- 7 per cent) at five years and 77 +/- 10 per cent at six years. When the time of failure was defined as the point at which the knee replacement was actually done, the five-year rate of survival was the same. No significant differences were demonstrated between groups that were stratified by age, sex, weight, or primary diagnosis. We therefore do not recommend the use of an uncemented porous-coated anatomic knee replacement of the design that was evaluated in this study.