Few studies have compared the long-term success of different surgical strategies in prosthetic knee-associated infection. Accordingly, a retrospective cohort study was performed of 40 episodes in 35 consecutive patients undergoing revision surgery for prosthetic knee-associated infection at a single centre between 1988 and 2003. The median patient age was 70 (44-90) years; the median follow-up period was 28 (2-193) months; 45% of infections were early, 23% were delayed, and 32% were late; and 55% of infections were caused by staphylococci. The probability of survival without prosthesis failure was 92.4% (95% CI, 84.1-100) after 1 year, and 88.7% (95% CI, 78-99.4) after 2 years. Recurrence-free survival was observed in 20 (95%) of 21 patients treated with debridement and retention, in both patients with one-stage exchange, and in 11 (85%) of 13 patients with two-stage exchange. Patients with delayed infection had a worse outcome than those with early or late infection (67% vs. 97%; p < 0.03). Patients with at least partially adequate antimicrobial therapy had a higher success rate than those with inadequate treatment (94% vs. 60%; p 0.069). The outcome was similar for patients with a duration of therapy of 3 to < 6 months, and those with a duration of therapy of > or = 6 months (91% vs. 87% success). Different surgical procedures had similar success rates, provided that the type of infection, the pathogen, the stability of the implant and the local skin and soft-tissue condition were considered. Adherence to an algorithm defining a rational surgical and antibiotic treatment strategy contributed to a favourable outcome.