The results of 57 revision total knee arthroplasties performed for aseptic failure between 1984 and 1992 with a cemented posterior-stabilized or constrained condylar prosthesis were reviewed at follow-up examinations at a minimum of 36 and an average of 62 months (range, 36-120 months). The reason for revision was aseptic loosening of 1 or both components in 32 knees (56%), instability in 16 knees (28%), polyethylene wear and osteolysis in 4 knees (7%), supracondylar femur fracture in 2 knees (4%), and a failed allograft, pain, and arthrofibrosis in 1 knee each (5% total). The average age of the patients at the time of the revision was 74 years (range, 38-90), and the original diagnosis for the majority of patients was osteoarthritis (74%). All of the revision prostheses were cemented posterior stabilized or constrained condylar-type implants. Bone deficiencies were grafted with cancellous allograft in contained defects and cortical allograft in noncontained defects. Five knees were reconstructed with allograft-prosthesis composites. The average modified Hospital for Special Surgery knee score improved from 49 to 82 (100 points possible) at final follow-up evaluation (P < 0.001). Seventy-nine percent of knees were graded as good or excellent. Kaplan-Meier survivorship analysis predicted 94%+/-6.2% survival at 40 months and 75%+/-25% at 99 months. There were 4 clinical failures, 3 of which were related to residual instability in patients with a posterior-stabilized prosthesis. Complications (3 knees) were exclusively related to the extensor mechanism. Radiographically, overall knee alignment improved from 0.3 degrees varus to 3.0 degrees valgus. Fifty-six percent of tibial components were placed in slight varus alignment. Radiolucent lines occurred in 33% of knees, but there were no complete or progressive radiolucencies. Radiolucent lines were more prevalent adjacent to press-fit intramedullary femoral stems compared with cemented stems (P < .02), but the difference did not correlate with clinical or radiographic failure. The median bone defect score, as proposed by the Knee Society Committee on Bone Defects, was significantly greater in knees that were revisions of a failed cemented total knee arthroplasty compared with revision of a failed cementless total knee arthroplasty (P = .02) but was not correlated with clinical or radiographic outcome (P > .05).