We tested the Herbert knee prosthesis, which was designed to restore
intrinsic stability with limited degrees of freedom for rotation to
severely damaged or deformed knees, in a knee simulator. These tests
indicated a tendency for the prosthesis to fracture through the medial
femoral housing after cycling for the equivalent of one to three years of
normal use. In a clinical series of thirty-five knees treated with the
prosthesis, there were five failures similar to those produced by
laboratory testing. On the basis of this combined study, the Herbert
prosthesis appears to have design characteristics that seriously limit its
usefulness for long-term knee replacement. Testing in a knee simulator in
this case appeared to be a valid predictor of clinical failure.