The two transepicondylar axes (the clinical and surgical epicondylar axes),
the posterior condylar axis, and the anteroposterior axis were constructed
using computed tomography scans in 111 (66 patients) knees with symptomati
c arthritis, The relationships between angles made by these reference axes
and two angles indicating frontal knee alignment (the tibiofemoral valgus a
ngle and the femoral valgus angle) were investigated. In (1)/(4) of the kne
es, the surgical epicondylar axis could not be constructed because the sulc
us of the medial epicondyle was not recognizable. The condylar twist angle
was almost constant and averaged 6 degrees when the femoral valgus angle wa
s 9 degrees or less, but increased gradually when the angle was greater tha
n 9 degrees, The difference between the condylar twist angle and the poster
ior condylar angle was constantly 3 degrees, The anteroposterior axis was a
lmost at right angles to the clinical epicondylar axis, and the relationshi
p between these axes was constant, independent of the femoral valgus angle.
With 3 degrees to 6 degrees external rotation relative to the posterior co
ndylar axis, the femoral component could be set parallel to the transepicon
dylar axis in common varus or neutral knees. In cases with a larger femoral
valgus angle, the anteroposterior axis would be a more reliable reference
axis. Preoperative computed tomography scans are recommended for patients w
ith knees with severe valgus deformity or severe hypertrophic osteoarthriti
s.