We retrospectively reviewed the records of 1150 computer-assisted total knee replacements and analysed the clinical and radiological outcomes of 45 knees that had arthritis with a pre-operative recurvatum deformity. The mean pre-operative hyperextension deformity of 11° (6° to 15°), as measured by navigation at the start of the operation, improved to a mean flexion deformity of 3.1° (0° to 7°) post-operatively. A total of 41 knees (91%) were managed using inserts ≤ 12.5 mm thick, and none had mediolateral laxity > 2 mm from a mechanical axis of 0° at the end of the surgery. At a mean follow-up of 26.4 months (13 to 48) there was significant improvement in the mean Knee Society, Oxford knee and Western Ontario and McMaster Universities Osteoarthritis Index scores compared with the pre-operative values. The mean knee flexion improved from 105° (80° to 125°) pre-operatively to 131° (120° to 145°), and none of the limbs had recurrent recurvatum. These early results show that total knee replacement using computer navigation and an algorithmic approach for arthritic knees with a recurvatum deformity can give excellent radiological and functional outcomes without recurrent deformity.