Objective To evaluate the surgical skill, cause of revision, compl ications, prosthetic survival and postoperative function in revision of custom-made tumor prosthesis replacement of knee joint. Methods The cl inical data of 33 patients who received prosthetic revision surgery between June 2002 and June 2007 were reviewed. There were 17 malesand 16 females with an average age of 33.1 years (range, 16-67 years). The pathological diagnosis included 17 osteosarcomas, 11 giant cell tumors, 2 mal ignant fibrous histocytomas, 1 chondrosarcoma, 1 synovial sarcoma, and 1 l iposarcoma. The involved locations were distal femur in 22 cases and proximal tibia in 11 cases. The average interval between first prosthetic replacement and revision surgery was 45.3 months (range, 6-180 months). The reason for revision included local recurrence in 2 cases, deep infection in 8 cases, aseptic loosening in 7 cases, peri prosthetic fracture in 1 case, prosthetic stem fracture in 6 cases, and prosthetic hinge failure in 9 cases. Six patients with deep infection received two-stage revision surgery, while the other 27 patients received one-stage revision. Cemented prostheses were used in all patients. Allograft prosthetic composite and revisions were used in 2 patients who had deficit of diaphysis for stem fixation. Results In 17 patients who received both primary prosthetic replacement and revision, the operative time was (149.8 ± 40.5) minutes and (189.9 ± 43.8) minutes, and the blood loss was (605.2 ± 308.0) mL and (834.1 ± 429.9) mL for primary prosthetic replacement and revision, respectively; all showing statistically significant differences (P lt; 0.05). The mean time of follow-up was 45.1 months (range, 12-76 months). Heal ing between allograft and host bone was obtained in 2 patients with allograft prosthetic composite and revision after 1.5 years and 2 years, respectively. After revision surgery, 3 patients died of lung metastasis after 12-24 months, and other 3 patients havinglung metastasis were al ive with disease. Nine (30%) compl ications occurred in 30 patients who were al ive at last follow-up. The compl ications included wound infection in 2 patients, deep infection in 5 patients, mechanical problems in 2 patients. Prosthetic failure occurred in 7 patients (23.3%). The 5-year survival rate of revised prosthesis was 68.6%. The Musculoskeletal Tumor Society (MSTS) score at 6 months after revision (73.6% ± 14.4%) was significantly improved (P lt; 0.01) when compared with before revision (57.1% ± 10.6%). Conclusion The main reasons for revision of custom-made tumor prosthesis of knee joint were mechanical problems and deep infection. Although revision surgery of knee is relatively compl icated and has some compl ications, a functional l imb could be maintained in most tumor patients.