BACKGROUND: The value of computer-assisted surgery in total knee arthroplasty for arthritic knees continues to be debated. We hypothesized that the usefulness of computer assistance is related to the magnitude of the deviation from the preoperative mechanical axis and that computer-assisted surgery may be beneficial under certain circumstances. METHODS: Patients with bilateral knee osteoarthritis and genu varus deformity who were to have staged bilateral total knee arthroplasty were enrolled. The patients randomly underwent computer-assisted total knee arthroplasty in one knee and conventional total knee arthroplasty in the contralateral knee. The two methods were compared for accuracy of placement of the components and lower extremity alignment after total knee arthroplasty as determined by six radiographic parameters. RESULTS: One hundred and thirteen patients (226 knees) met the inclusion criteria. For patients with a preoperative mechanical axis deviation of <10° and those with a deviation of 10° to 14.9° in both knees, the postoperative radiographic parameters did not differ significantly between the two techniques. In patients with a preoperative mechanical axis deviation of >20°, the reconstructed mechanical axes were significantly closer to normal in the computer-assisted total knee arthroplasty group. Significant results were also noted in the anatomical axes, femoral valgus angle, and femoral flexion angle. Furthermore, a higher percentage of knees in which computer-assisted surgery was used had restoration of the mechanical axis within 3° of neutral. CONCLUSIONS: Computer-assisted surgery was a valuable adjunct for obtaining proper alignment during total knee arthroplasty in patients with knee osteoarthritis with severe varus deformity. Conventional total knee arthroplasty was as effective as computer-assisted total knee arthroplasty for obtaining proper alignment in patients with a minor to moderate deformity.