Between 1985 and 1987, 1837 primary total knee arthroplasty (TKA) prostheses were implanted in 1503 patients. Group I included 843 knee with metal-backed patellar components (MBPC), and group II included 994 knees with all polyethylene patellar components (APPC). Follow-up averaged 5.7 years (range, 2 to 11 years). Twenty-four MBPC (2.9%) and 16 APPC TKA cases (1.6%) developed deep infection. In the time interval between arthroplasty and 2-year follow-up, eight MBPC and 11 APPC knees developed deep periprosthetic infection. The difference in the cumulative probability of infection between the two groups during this time interval was not significant (relative risk, 0.9; 95% confidence interval [CI], 0.3-2.1; P = 0.73). However, after the 2-year follow-up, 16 MBPC and 5 APPC knees developed late infection, and the difference in the cumulative probability of infection between the MBPC and APPC knees during this time interval was significant (relative risk, 3.1; 95% CI, 1.1-8.5; P = 0.02). Although the mechanism for this increased risk of these late infections is not well understood, the attendant synovitis, effusion, and relative hyperemia of these knees in the presence of the particulate metal and polyethylene debris may increase the potential of bacterial seeding to these prostheses. Particulate metal debris has been previously shown to suppress bacterial phagocytosis and may play a role in the pathogenesis of these infections. We propose that the presence of metal-backed patellar failure represents a "prosthesis at risk" for the development of late prosthetic infection.