Cadaveric lower extremities had femoral osteotomies at clinically useful proximal, mid-shaft and distal locations. External rotation was performed at each level, measured by pin rotation between bone segments. CT images were reconstructed in 3D and custom software used to semi-automatically compute actual amount of rotation induced (pin angle change) and femoral torsion following rotation at each level.
Specimens and preparation
Ten fresh-frozen adult cadaveric lower extremities (hemi-pelvis to foot) were acquired as power analysis determined that n = 9 would be needed in order to determine population differences in femoral torsion (data used from Kaiser et al. [12]). Specimens with no hip, knee or ankle arthritis, osteoporosis, prior extremity or joint infection, prior extremity surgery, previous femur, tibia or fibular fracture or other known systemic musculoskeletal disorder affecting bone or soft tissue anatomy were requested. Initially, each lower extremity underwent torsional profile CT scanning, following our institutional standard clinical technique of scanning the hip, knee and ankle (0.625 mm thick cuts, GE LightSpeed VCT 64-Slice, Piscataway, NJ, USA). CT images were viewed to confirm lack of joint and bone pathologies.
Lower extremities were then defrosted for 48 h at room temperature. The entire femur was exposed with a lateral incision, and the femoral length was measured from the greater trochanter to the most distal point between the femoral condyles. A mark was made in each femur indicating half of its length. Femoral osteotomies were performed perpendicular to the anatomic axis at all three levels: proximal (just distal to the lesser trochanter), midshaft (at the half-length mark) and distal (in the supracondylar region). Each osteotomy site was fixed into the native anatomic alignment (0 degrees of rotation) using a plate (4.5 mm narrow stainless steel locking compression plate, 7 holes, Synthes, West Chester, PA, USA) held in place with 2 or 3 self-tapping cortical screws (stainless steel 5.0 mm locking and 4.5 mm non-locking × 38-46 mm, Synthes, West Chester, PA, USA) on either side of the osteotomy location.
Surgical rotation technique and measurement
Four 5.0 mm self-drilling Schanz screws (pins, DePuy Synthes, New Brunswick, NJ, USA) were placed parallel to each other into each femur, one pin in each of the four bone segments created by the osteotomies (Fig. 1). One of the plates was removed and, mimicking rotational correction techniques used in the operating room, the more distal portion of the femur was rotated externally approximately 15 to 20 degrees (clinically applicable amounts of rotation) as viewed along the long axis of the femur, aligning two adjacent pins to a goniometer. Cortex screw holes were made in order to secure the osteotomy site with a plate and screws. The rotation was returned to neutral, and this was repeated for the remaining two other osteotomy levels. CT scans were then performed of the entire femur in a lateral position with the knee flexed 30 degrees as measured using a goniometer placed along the length of the femur and tibia. An initial CT scan was completed with plates securing all osteotomy levels in the neutral position. Then the femur was rotated externally through one osteotomy level, fixed in place with plates and screws using pre-drilled holes (Fig. 2), CT scan was repeated, then the rotation at that level was returned to the neutral position, prior to rotation at the next level. This was repeated at each osteotomy level with the sequence of first, second and third level to be rotated systematically alternated between proximal, midshaft and distal levels to remove effects of previous rotations.
Femoral torsion measurement technique
Mimics software (v.19, Materialise, Leuven, Belgium) was used to generate true 3D reconstructions of the proximal and distal ends of each femur from CT DICOM images. The pins placed into each of the four bone segments were also modeled using Mimics. All 3D reconstructions were exported as stereolithography (STL) files. Femoral heads were then separated from the proximal femur reconstruction using 3-matic Medical (Materialise NV, Leuven, Belgium). All 3D models were imported into custom MATLAB (Mathworks, Natick, MA, USA) software for femoral torsion and pin angle calculations. Femoral torsion was automatically calculated using a custom a MATLAB program, thus eliminating human measurement error [13]. Femur models were reoriented based on the mechanical axis (MA) and the anatomical axis (AA). MA was between the center of the femoral head best fit sphere and midpoint between the distal femoral condyles. AA was a center line along the shaft of the femur, drawn between points 10 cm distal to the head and 10 cm proximal to the condyles.
Femoral torsion was calculated as the angle between the femoral neck axis and the posterior condylar axis in the transverse plane. Torsion was set positive when the femoral neck was anteverted (in front of the posterior condylar axis) and negative when the femoral neck was retroverted (behind the posterior condylar axis) to represent clinical understanding of femoral torsion. Pin position was also determined in the MA and AA transverse planes by identifying centerlines along each of the screws from centroid positions. Pin angle was calculated using the angle change between the most proximal and distal pins (Fig. 3). External rotation of the pins was set positive.
For each osteotomy site, the change in femoral torsion and change in pin angle were calculated and compared between osteotomy levels (proximal, midshaft and distal) by the computer modeling and the intended amount of rotation during the surgery. The difference in angle calculated from 3D CT reconstructions between adjacent pins in each osteotomy site was considered variance related to technique rather than affect due to location of osteotomy.
Statistical analysis
Basic descriptive statistics are presented. The Shapiro-Wilk test of normality and Levene’s test of homogeneity of variance was performed on all continuous data prior to analysis. All data was found to be normally distributed. Analysis of variance was used to evaluate differences in torsion between osteotomy locations. Simple linear regression analysis was used to evaluate the effect of pin angle change on torsion. All statistical analysis was performed using SPSS (version 12; SPSS, Chicago, IL, USA). Statistical significance was defined as p < 0.05.