Rotational malalignment has been demonstrated as an important cause of failure in total knee arthroplasty. While there is consensus about the transepicondylar axis as a reference for the femoral rotational alignment [4, 9–13] there is no comparable agreement for the tibial rotational alignment.
To the author's knowledge, there has been no CT-controlled study performed to measure the variance between femoral and tibial component after TKA with correction of a femoral rotational malalignment.
Eckhoff et al.  implanted TKA in seven fresh frozen anatomic specimen knees and found as much as 19° of external rotation of the tibial component relative to the femoral component using the tibial tubercle as a reference for rotational alignment of the tibia. However, this study was performed on non-osteoarthritic knees.
Bindelglass et al.  measured intraoperatively the distance between the medial third of the tibial tubercle and the alignment of the tibial tray when seeking its own position during trial reduction. This resulted in 25 of 30 TKAs with an internally rotated tibial component relative to the tibial tubercle. It must be mentioned that a 3° of femoral external rotation relative to the posterior condylar line was used as a reference in the aforementioned study and not the transepicondylar axis. No correlation to the transepicondylar axis was performed. Therefore this relative internal rotation may be caused by a transferred femoral malrotation.
Incavo et al.  studied MRI scans of 30 normal knees. He templated an symmetric tibial baseplate using the femoral epicondylar axis as reference and measured the intersection of the midaxis line and the patellar ligament. This resulted in 22 of 30 knees with an optimal position of the tibial baseplate, defined as intersection of the midaxis line of the tibial template between the midpoint of the ligament and the midpoint of the medial third of the patellar ligament, which is consistent with our findings. However, this study was done on non-osteoarthritic knees and the patellar ligament was used as reference which may be difficult to obtain intraoperatively when everting the patella.
Akagi et al.  measured the angle between a line perpendicular to the transepicondylar axis and different landmarks on healthy subjects. They found the lowest variability in an axis from the medial border of the ligamentum patellae to the posterior cruciate ligament. However, this measurement was done on non-osteoarthritic knees and it is difficult to define the middle of the posterior cruciate ligament on CT scans after total knee arthroplasty. Furthermore the medial border of the ligamentum patellae could be difficult to identify after medial arthrotomy.
Uehara et al.  studied osteoarthritic knees with a varus malalignment before TKA. This study demonstrated a mean external rotation of 2,6° of the tibia relative to the transepicondylar axis (range: 16° external to 10° internal rotation) using a line from the medial third of the tibial tubercle to the midpoint of the longest medio-lateral distance.
Huddleston et al.  measured intraoperatively during trial reduction the deviation of the trial insert from the tibial tubercle. This angle was 5,2° external rotation (range: 10° internal to 15° external rotation) of the trial insert relative to the medial border of the tibial tubercle. For femoral rotation they used a mix of flexion-gap blancing, transepicondylar axis, Whiteside's line and posterior condylar line as reference points. No correlation to a possible femoral malrotation was done and the deviation was measured in steps of 5°.
Recently Cobb et al.  described the "anatomical tibial axis", which is the perpendicular to the mid-point of the line joining the medial and lateral condylar centres. This line was more reliable than the posterior condylar line or a line from the lateral tibial spine to the center of the tibial tubercle. However, this "anatomical axis" touches the tibial tubercle lateral of the medial border.
This investigation demonstrates that using an axis from the medial third of the tibial tubercle to the center of the tibial tray as a reference for the tibial rotational alignment leads to a better femoro-tibial alignment in extension than using the medial border of the tibial tubercle. This corresponds with the results of Huddleston  where the trial insert was rotated to a point 5° lateral to the medial border of the tibial tubercle, and results from Uehara  and Cobb's "anatomical axis"  which was lateral of the medial border of the tibial tuberosity.
While the deviation from the transepicondylar axis was low in our study group, there was a considerable range for the tibial rotational alignment as already demonstrated in previous studies [24, 25]. This may be due to difficulties to define the correct landmarks on the tibial tubercle during surgery as well as during measurement. Because rotational alignment was calculated relative to the short axis (tibial tubercle to the centre of the tibial tray), a deviation of one millimeter at the tibial tubercle resulted in a rotational deviation of about five degrees (depending on the size of the tibia). Another limitation of this study is the retrospective design because the original aim was the precision of implant position in navigated and conventional TKA. However, the statistical power appeared to be sufficient to support the results.
However, it must be realized that there is a notable rotational variance between femoral and tibial component using a fixed bone landmark as a reference (even if both components are ideally implanted). This could possibly be compensated by using a rotating tibial platform or a prosthesis with a fixed bearing design, which allows a certain amount of rotational freedom between the femoral and the tibial component. Fixed bearing ultra-congruent inlays restrict the rotation between femoral and tibial component and must be critically discussed when using a fixed landmark for tibial rotational alignment. The rotational restriction of ultra-congruent inlays may result in premature polyethylene wear. Thus using the "self-seeking method" for rotational alignment of the tibial component could help prevent femoro-tibial mismatch. However, this induces the risk of transferring a femoral malrotation to the tibia and caution must be taken with the patellar tracking.
There was no difference in patellar tracking between patients with a "tolerable" rotational mismatch (± 10°) and patients with more than 10° of mismatch between the femoral and the tibial component. This might be due to the use of the rotating platform and could be different with a fixed bearing.