National registers compare implants by their revision rates, but the validity of the method has never been assessed. The New Zealand Joint Registry publishes clinical outcomes (Oxford knee scores, OKS) alongside revision rates, allowing comparison of the two measurements. In the two types of knee replacement, unicompartmental (UKR) had a better knee score than total replacement (TKR), but the revision rate of the former was nearly three times higher than that of the latter. This was because the sensitivity of the revision rate to clinical failure was different for the two implants. For example, of knees with a very poor outcome (OKS < 20 points), only about 12% of TKRs were revised compared with about 63% of UKRs with similar scores. Revision therefore is not an objective measurement and should not be used to compare these two types of implant. Furthermore, revision is much less sensitive than the OKS to clinical failure in both types and therefore exaggerates the success of knee replacements, particularly of TKR.