We have compared the survival of 67 revision arthroplasties of the knee undertaken for aseptic loosening with and without the retention of a secure, cemented femoral component. All the patients had undergone a single primary procedure at a mean of nine years previously. In group I (25 knees) the original femoral component was secure and was retained. There were no abrasions or osteolysis. The knees were stable, normally aligned, with minimal bone loss. In group II 42 knees did not fulfil these criteria and underwent revision of both components. The mean follow-up was four years.

Re-revision for loosening was required in seven knees (28%) in group I and three (7%) in group II (p < 0.01). The remaining knees function well with Knee Society scores averaging 84/69 and no radiological evidence of osteolysis. When revising cemented implants, retention of a secure femoral component cannot be recommended even when conditions appear to be suitable.