Objective       To evaluate the accuracy of different axes relative to surgical transepicondylar axis (STEA) by measuring the distal femur with different rotation axes for the patients suffering from severe osteoarthritis with varus deformity of knee joint on MR images.       Methods       A retrospective analysis was performed on 86 cases who had severe osteoarthritis in knee (86 knees) with varus deformity (varus deformity ≤15°) in our center between June 2013 and January 2014. They were 34 males (34 knees) and 52 females (52 knees)

at an aged from 51 to 70 years (mean 63.7 years). MR imaging was applied to locate the STEA

the clinical transepicondylar axis (CTEA)

3° extorsion against the posterior femoral condylar line (3° PCL)

and the anterior posterior line (APL) by 2 surgeons respectively. Then the deviation relative to the STEA among 3° PCL

the CTEA

and the perpendicular of APL were observed. Analysis of variance (ANOVA) was used to compare the accuracy of the angle close to STEA.       Results       3°PCL relative to STEA (Angel A) was out 0.63°±1.82°

CTEA to STEA (CSA) was out 3.78°±0.75°

and perpendicular of anterior posterior line to STEA (ASA) was out 5.38°±3.17°

in an accuracy of Angle A＞CSA＞ASA. There was statistical significance in any two-angle comparison (P0.01）.       Conclusion       In the patients suffering from severe osteoarthritis knee with mild varus deformity

except STEA

3° PCL is of the highest accuracy in the distal femoral bone cutting axis. When it is difficult to locate STEA during operation

3° PCL is helpful to obtain a good femoral rotation accuracy.