The “Bankart knee”: high-grade impression fractures of the posterolateral tibial plateau lead to increased translational and anterolateral rotational instability of the ACL-deficient knee

In each of the tested specimens, the femur and tibia were cut approximately 20 cm from the joint line. The dissection was initiated through a midline incision through the cutaneous and subcutaneous tissue. All of the soft tissue structures up to approximately 10 cm from the joint line were removed, revealing the shaft of the tibia and femur. The remaining fascia, muscles and ligaments were left intact (Fig. 2). Special care was taken to leave a significant portion of the iliotibial tract (ITT), as well as the structures of the posterolateral corner of the knee, intact. The insertion of the posterior cruciate ligament (PCL) on the posterolateral margin of the tibia was identified, followed by a small horizontal incision (approx. 0.5 cm) through the posterolateral capsule just lateral to the tibial insertion of the PCL. This allowed for intraarticular measurement of the width of the PHLM using a standard small screw depth gauge. Once the testing was concluded, the measurements were reconfirmed using an electronic calliper (± 0.2 mm), and no measurement errors were found in any of the specimens. The head of the fibula was fixed at the proximal tibiofibular joint with a 4 mm cortical screw, and the remaining length of the bone was removed. The insertions of the LCL and biceps femoris muscle (BFM) on the fibular head were left intact. A modified patellar arthrotomy along the midline of the supero-inferior axis of the patella was performed using a fine oscillating saw. The arthrotomy was completed with a scalpel, and the split was extended into the patellar tendon. This approach allows direct visualization and palpation of the femoral insertions of the ACL and PCL, as well as access to and dissection of the ACL during testing. This longitudinal transpatellar approach was previously outlined as a reliable and convenient approach to the knee joint for in vitro studies, without any significant effect on the surrounding tissue and overall kinematics of the joint [29, 36, 38]. The bony approach to the intraarticular space was closed and fixed using two 4-mm cancellous screws. The tibia was then rigidly fixed in an aluminium tube (cylinder) using polymethyl methacrylate bone cement (PMMA). The tibia was centred in the cylinder so that its axis was aligned with the condylar eminence at the centre of the tibial plateau [23]. Additional 4-mm screws were added to prevent possible sliding out of the tibial shaft during testing (Fig. 2). The described preparation and fixation procedure have previously been published [23, 36, 38].

A six-degree-of-freedom (DOF) robot was used for biomechanical testing (6DOF; KR 125; KUKA Robotics, Augsburg, Germany). Additionally, a force/moment sensor (UFS; FTI Theta 1500–240; Schunk, Lauffen, Germany) was mounted on the robotic rig. The repeatability and accuracy of the KUKA robot system are 0.2 mm and 0.02° for the orientation and position of the end effector, respectively, whereas the maximum loading capacity is 2450 N [36, 38]. The UFS resolution was 0.25 N and 0.05 Nm for forces and torques, respectively. To track the kinematics of the knee movement through simulated stress tests, an optical tracking system with an accuracy of up to 0.1 mm and a resolution of 0.01 mm was used (Optotrak Certus Motion Capture, Northern Digital, Ontario, Canada). Two sensors were rigidly affixed to the tibial tuberosity and the shaft of the femur using custom-made pins (Fig. 2). The knee coordinate system was defined by six tibial (medial/lateral tibial shaft and medial/lateral/anterior/posterior tibial plateau) and four femoral landmarks (medial/lateral femoral shaft and medial/lateral femoral epicondyle). Each of the anatomic landmarks was digitized with a generic digitizing probe. This allowed 6-DOF kinematics to be computed in terms of rotations and translations. Visual 3D computer software (C-Motion Inc., Germantown, Maryland) was used for data processing. This software designs a digitized coordinate system of the tibia using the Cardan x–y–z axis model [36, 38]. The digitized landmarks were used to create static and moving knee models at each of the tested flexion angles. The difference in translation (mm) or rotation (degree) between the static and moving models at the end of each simulated laxity test in all degrees of flexion (0°, 30°, 60° and 90°) was recorded and compared [38]. The specimens were fixed in the robot in an upside-down orientation by attaching the aluminium tube with the fixed tibia to the movable component (arm) of the robot and fixing the femur in a tube placed in the stationary base of the machine (Fig. 2). The force vectors and moments were minimized to < 1 N and < 0.5 Nm, respectively. Prior to testing, each specimen was passively brought from full extension to 90° flexion in 1° increments. Throughout this passive motion, the robot was programmed to minimize forces acting upon the joint. The position of the knee at each flexion angle was recorded, and the passive path was repeated three times. Through the testing protocol, the specimens were placed under a constant 200 N axial compressive load. Once the passive path was established, the specimens could be moved automatically by the robotic arm to predetermined flexion angles and submitted to laxity tests while the remaining acting forces and torques remained at a minimum. The specimens were submitted to the following simulated knee laxity tests: The Lachman test (89 N), the posterior drawer test (89 N), internal rotation (IR; 4 Nm of rotational torque), and external rotation (ER; 4 Nm of rotational torque) were performed at 0°, 30°, 60° and 90° of knee flexion. The simulated pivot-shift test, corresponding to 4 Nm tibial internal rotation and 8 Nm valgus torque, was conducted at 0° and 30° of flexion. This method for robotic knee testing has been based on the overall principles of superposition as previously validated and published in studies by Guenther et al. [15, 16]. To differentiate the movement of the lateral tibial plateau in response to this dynamic multiplanar test, we distinguished between maximum anterolateral translation (ALT) of the midpoint of the lateral tibial plateau and anterolateral rotation (ALR) of the tibia, defined as the maximum IR occurring around the longitudinal axis of the tibia. The testing intervals were repeated three times for each of the 4 cutting states, i.e. the intact, ACL-deficient, 1st-degree depression fracture (Bankart 1) and 2nd-degree depression fracture (Bankart 2) states. The midpoint of the lateral tibial plateau was taken as ¼ of the maximum distance between the lateral and medial digitized points on the plateau, as previously published [38].

The specimens were initially tested in the intact state, followed by sequential cutting of the ACL and an additional 2 degrees of posterolateral depression fractures (marked Bankart 1 and 2) of the tibial plateau. The same cutting sequence was applied to all tested specimens. During cutting procedures, the specimens were kept mounted on the robot. Initial cutting of the ACL was performed through the preformed patellar split at 90° of knee flexion. The ACL was separated from its femoral insertion under direct visualization using a scalpel and scissors. Once the cut was made, the patella was fixed with two 4-mm screws. An initial fracture labelled Bankart 1 was made using a sharp 15-mm-wide chisel just lateral to the insertion of the PCL on the posterior edge of the tibial plateau. Standardized proportions for the plateau depression were applied to all tested specimens. The height of the depression was held constant at 10 mm, and the width was 15 mm starting from the point just lateral to the tibial insertion of the PCL (Fig. 3B). The intraarticular depth of the initial fracture, labelled Bankart 1, corresponded to half of the predetermined width of the PHLM, whereas the depth of the second depression fracture, labelled Bankart 2, corresponded to the full width of the PHLM (Figs. 1 and 3A).

Before the statistical tests were performed, the data were checked for normality (using the Shapiro–Wilk normality test and by visual inspection using a QQ plot), and extreme outliers were removed (based on the boxplot method, i.e. values above Q3 + 3 × IQR or below Q1–3 × IQR).

Two-way repeated-measures ANOVA with post hoc Bonferroni corrections for multiple comparisons was performed to evaluate the effect of different cutting states over different knee flexion angles on tibial translation/rotation. The two independent factors were the four states (intact, ACL-deficient, Bankart 1 and 2) and the knee angle (0°, 30°, 60° and 90°) within the same specimen, while the dependent variable was the resulting translation/rotation. In the cases where the two-way repeated-measures ANOVA did not detect a statistically significant interaction effect of the cutting state and the flexion angle on translation or rotation of the tibia, one-way repeated-measures ANOVA was performed to evaluate the effect of the different flexion angles or different cutting states on the translation/rotation separately. One-tailed paired Student’s t tests were used to compare the mean translation/rotation between the fixed state/angle combinations. The significance level was set to 0.05. This analysis was performed using R (R Core Team, 2020) with additional packages (see Supplement file).

A post hoc power analysis was performed using G*Power according to the study by Faul et al. [11] using the results of the simulated pivot shift test at 0° and 30° of flexion in all specimens. Based on the data reported in that study, effect sizes of 0.5 and 0.8 were obtained at 0° and 30°, respectively. With these effect sizes and an alpha of 0.05, the power was 0.85 and 0.99, respectively.

Pairwise comparisons showed that the mean ATT was statistically significantly different between all cutting states at angles of 0° and 30°, between the ACL-deficient and Bankart 1 states at 60°, between the intact and ACL-deficient states at 90°, and between the Bankart 1 and Bankart 2 states at 90° (Fig. 4 and Table 1).

Regarding posterior tibial translation, there was no statistically significant interaction between effect of state and angle on translation (n.s.). Furthermore, the effect of the cutting state was not statistically significant at any fixed flexion angle (n.s.).

Regarding ER, there was no statistically significant interaction between effect of cutting state and knee flexion angle on the ER of the tibia (n.s.). The effect of the cutting state was not statistically significant at any flexion angle, and pairwise tests did not show any difference between the mean rotation values for different states at any fixed angle (n.s.).

Regarding the internal rotation of the tibia, there was a statistically significant interaction between effect of cutting state and flexion angle on the rotation of the tibia (p = 0.0002, eta2[g] = 0.01). Pairwise comparisons showed that the mean IR of the tibia was significantly different between all states at 0°, with the highest significance between the Bankart 1 and Bankart 2 states, meaning that both grades of PLTP fractures led to a statistically significant increase in internal angulation of the tibia. At 30° and 90° of knee flexion, the Bankart 1 impression fracture yielded a further increase in internal angulation, as shown in Fig. 5 and Table 2.