Nineteen revision total knee arthroplasties (TKAs) performed from March 1991 to March 1995 were evaluated to determine the clinical and roentgenographic results and the problems in revision TKA. The mean age was 63 years (range, 23-85 years) and the mean follow-up time was 19 months (range, 12 months-4.4 years). The primary revisions were done in eighteen knees and a re-revision, in one. Four revision TKAs were performed for failed unicompartmental arthroplasty and fifteen, for failed TKA. The mean interval from initial to revision TKA was 4.1 years (range, 3.3 years-5 years) in failed unicompartmental arthroplasty and 5.7 years (range, 4 months-14 years) in failed TKA. The initial diagnosis was degenerative arthritis in 16 knees, tuberculous arthritis in two and rheumatoid arthritis in one. The main cause of revision for both of the failed unicompartmental arthroplasties and failed TKAs was aseptic loosening. The HSS knee score improved from 43±14 to 82±7 in the failed unicompartmental arthroplasties and from 46±16 to 79±14 in the failed TKAs. The clinical success rate of revision TKA was 78%. The coronal tibiofemoral angle improved from valgus 0.2degrees to valgus 5.1degrees. At final follow-up, radiolucent line greater than 2mm in width was found around femoral component in one knee. In our series, four complications were occurred. One knee was fused due to mas-sive bone loss and ipsilateral femur shaft fracture. Deep wound infections were developed in two knees and loss of fixation after tibial tuberosity osteotomy in one. In conclusion, significant improvements of clinical and radiological results were achieved in revision TKA but the success rate was relatively low and the complication rate was relatively high compared with primary TKA. So, especially appropriate skillful surgical techniques and postoperative cares were required to overcome the soft tissue problem and infection which was encountered in revision TKA.