The results of total knee replacement (TKR) are commonly assessed by survival analysis using revision as the endpoint. We have used the assessment of pain by a patient-based questionnaire as an alternative. In one hospital, 1429 TKRs were inserted by 66 surgeons between 1987 and 1993. The survival at seven years, with revision as the endpoint, was 97.5% (CI 94 to 100). There were no significant differences between the three different types of implant used, the AGC, the IB2 and the Nuffield Knee. When the endpoint was the development of moderate pain, the survival at seven years for the AGC knee was 72% and that for the IB2 was similar. Significantly more patients (p = 0.007) with the Nuffield Knee, however, had developed moderate pain. Using revision as the endpoint, it is difficult to discriminate between the various types of TKR, but this can be achieved using pain. In this investigation 30% of the patients reported moderate pain at some stage by seven years from operation.