The incidence of deep infection after 12,118 primary knee arthroplasties performed in Sweden from October 1, 1975 through 1985 with a median follow-up of 6 years was 1.7 percent for arthrosis and 4.4 percent for rheumatoid arthritis. Risk factors for infection were large prostheses, postoperative wound-healing complications, rheumatoid arthritis, a prior deep infection, and skin infections. We have analyzed the treatment of 357 knee arthroplasties with a deep infection. Systemic antibiotics alone were primarily used in 225 knees, with healing of the infection in 44 knees, 20 of which had a functioning prosthesis at the final follow-up; the treatment did not compromise later revision surgery. Soft-tissue surgery was used in 154 knees--37 healed, 15 of which had a functioning prosthesis. Resection arthroplasty resulted in healing of the infection in 11 of 22 knees. Revision arthroplasty was performed in 107 knees, with eventual healing of the infection in 81 knees, 36 of which had a functioning prosthesis; there were no differences in the outcome of one-stage and two-stage procedures. Arthrodesis was attempted in 135 knees, with eventual healing of the infection in 120 knees and fusion in 105. Twenty-two patients were amputated. Thus, the infection healed in 315 knees (88 percent), but only 71 (20 percent) recovered with a functioning prosthesis, and 8 patients died of the infection. Attention should therefore focus on prophylactic measures directed towards the soft-tissue problems--by avoiding conflicting skin incisions, by gentle handling of the periarticular soft tissues, by avoiding the use of constrained prostheses and oversized compartmental prostheses, by letting wound healing take priority over motion in knees with compromised soft tissues, and by using prophylactic antibiotic treatment for skin ulcers until these have healed.