The British Orthopaedic Association assessment questionnaire for knee replacements was adapted to allow comparison of the severity of underlying polyarthritis with the benefits of geometric knee replacement in a retrospective study of 150 knees between six months and six years after operation. Total or partial relief of pain was achieved in 81 per cent of the operation, and changes in mobility occurred in fewer patients. Late sepsis remained a serious complication of nine per cent of the operations and one patient died from septicaemia. Late sepsis was associated with previous synovectomy or osteotomy. Retropatellar pain rarely interfered with the mobility of the patient. There was no association of operations that failed with a high erythrocyte sedimentation rate, a high platelet count, a low haemoglobin level or with a strongly positive rheumatoid factor but pain in the contralateral knee was associated with a diminished functional capacity.