Fifty-six constrained total knee replacements (forty Guepar and sixteen Herbert prostheses) were performed in forty-nine severely disabled arthritic patients and they were followed for two to four years. Seventeen of these knees were operated on to salvage a failed prosthesis. Good relief of pain was achieved in 64 per cent of the knees but there was less improvement in walking and function. The over-all range of motion increased because of a reduction in preoperative flexion contractures. The results in the salvage group were less successful than in those patients who underwent a primary operation. Between six months and one year postoperatively, pain had developed in 17 per cent of the fifty-six knees, 9 per cent had walking difficulties, and 7 per cent lost function. Patellofemoral pain accounted for the unfavorable pain ratings in over half of the knees.

Zonal roentgenographic analysis of each knee indicated high incidences of radiolucent lines at the cement-bone interface (75 per cent), cement-metal lucencies (68 per cent), and cement deficiencies (73 per cent). Forty-three per cent of the knees showed excessive posterior placement of the prosthesis of more than five millimeters, resulting in flexion contractures of as much as 5 degrees.

Complications requiring reoperation developed in nine knees (16 per cent). There was one case of loosening and three knees had deep infections. Chronic postoperative effusions were present in 48 per cent of the knees.

The use of either a cemented metal-on-metal hinged knee replacement or a metal-on-polyethylene hinge type of prosthesis was found to result in a relatively high incidence of failures and complications, and did not solve the problem of treating patients with a failed knee prosthesis.