Specimens preparation
Eight pairs of fresh-frozen proximal human cadaveric femora from one female and seven male donors aged 72.7 ± 14.9 years (mean ± standard deviation, SD; range 48–89 years) were used in this study. The specimens were collected from an accredited donation program (Science Care, Inc., Phoenix, AZ, USA). All donors gave their informed consent inherent within the donation of the anatomical gift statement during their lifetime. The specimens underwent computed tomography (CT) scanning (Revolution EVO, GE Medical Systems AG, Switzerland) at a slice thickness of 0.63 mm to ensure no evidence of any pathology and to measure their bone mineral density (BMD) within a cylinder of 20 mm diameter and 30 mm length, located in the center of the femoral head, with the use of a calibration phantom (European Forearm Phantom QRM-BDC/6, QRM GmbH, Möhrendorf, Germany). Subsequently, the femora were assigned in paired fashion to two study groups.
In Group 1, stable pertrochanteric fractures AO/OTA 31-A1 were created in each right femur. By definition, a pertrochanteric fracture line is located in the region between the greater and lesser trochanter [9]. A stable fracture A1 is defined by a lateral wall thickness of more than 20.5 mm as measured from a reference point located 3 cm below the innominate tubercle of the greater trochanter to the intersection with the fracture line in an angle of 135° from this point on the anteroposterior x-ray image [9]. Although several previous studies [10, 11] reported a high interobserver variability in defining the subgroups A1.1-A1.3 of this fracture type (A1), this definition corresponds to all of them.
All fractures were created with a custom-made blunt guillotine blade, which was connected to the actuator of a material testing machine (#5866, Instron, Norwood, MA, USA) equipped with a 10 kN load cell. The blade was positioned on the anterior cortex of the bone in an angle of 41° to the femoral shaft with the cutting direction being adjusted to the individual anteversion angle in the frontal plane by supporting the bone on three fixed points located at the lesser trochanter, the greater trochanter, and the distal condyle plane (Fig. 1). The determination of the fracture angle was based on findings from a previous radiologic study analyzing 164 anteroposterior radiographs of the hip and pelvis from patients with trochanteric fractures [12]. An angle of 41° ± 8° with respect to the femoral shaft axis was reported for the cases with two-part fractures among those patients. In order to create a stable fracture, the blade location was specified as starting from the innominate tubercle of the greater trochanter and ending in the region proximal to the lesser trochanter. A constant actuator displacement rate of 5 mm/s was applied up to a depth of 10 mm, which proved to be enough in conducted pilot tests to penetrate the first (anterior) cortex in the pertrochanteric region and initiate the artificial creation of standardized fracture type AO/OTA 31-A1. Consequently, the fracture line of each bone proceeded its natural propagation according to the trabecular architecture of the femur and resulted in a full AO/OTA 31-A1 fracture type pattern (Fig. 2).
In Group 2, an osteotomy, simulating stable femoral pertrochanteric fracture type AO/OTA 31-A1, was created in each left femur with an oscillating saw blade of 0.8 mm thickness. For that purpose, a custom-made cutting guide was designed and aligned with the lateral cortex of the femoral shaft. Starting from the innominate tubercle of the greater trochanter, the osteotomy was set in an angle of 41° to the shaft axis in the frontal plane and an angle of 15° in the sagittal plane, the latter chosen to consider the anteversion of the femoral neck.
All specimens were anatomically reduced and subsequently implanted with a Dynamic Hip Screw (DHS, DePuy Synthes, Zuchwil, Switzerland) according to the manufacturer’s guidelines. The lag screw was positioned in center-center position and a tip-apex distance of less than 25 mm was considered to minimize the risk of cut-out failure. No additional antirotation screw was used. All femora were cut distally at a length of 250 mm, measured from the tip of the greater trochanter, and the distal 65 mm were embedded in polymethylmethacrylate (PMMA, SCS-Beracryl D-28, Suter Kunststoffe AG, Fraubrunnen, Switzerland) in preparation for biomechanical testing. Finally, retro-reflective marker sets were attached to the shaft, femoral head fragment and DHS head element for optical motion tracking (Fig. 3).
Creation of a mean shape model
All sixteen specimens were scanned in both intact and fractured/osteotomized states using a clinical CT scanner (Revolution EVO, GE Medical Systems AG, Switzerland) at a slice thickness of 0.63 mm. Cortex masks of the different fragments were generated on the fractured/osteotomized scans in a semi-automatic procedure using in-house developed script in Matlab (R2019a, MathWorks, Natick, MA, USA) and C + + code [13]. These scans were then registered on the intact scans in order to locate the fracture lines using Amira software package (Amira 6.2, FEI Company, Hillsboro, Oregon, USA). Based on the scans of intact bones, a mean shape model was created by manual positioning of the anatomical and segmental landmarks as previously described [14]. The left bone sides were mirrored to create one mean shape model for both right and left femora. The fracture lines were automatically created using custom made Amira scripts and Visualization Toolkit (VTK, Kitware Inc., Clifton Park, NY, USA). The closest points of the registered fracture parts were automatically determined and interpolated to a single fracture line. Each individual fracture/osteotomy line was homologized and transferred to the mean shape model. One average line and two standard-deviation lines were calculated for all eight individual lines of each group and projected onto the surface of the mean shape model. The standard deviation values corresponding to each separate point on the average fracture/osteotomy line – calculated via the distances between the respective eight individual points and their projections on the average fracture/osteotomy plane – were considered for statistical equality/divergence comparison analysis of the two different procedures for creation of the stable pertrochanteric fractures by means of either constant force application (Group 1) or osteotomizing (Group 2).
Biomechanical testing
Biomechanical testing was performed on a servohydraulic test system (Bionix 858, MTS Systems, Eden Prairie, MN, USA) equipped with a 4 kN load cell at room temperature (20 °C) in dry environment. Both test setup and loading protocol were adopted from previous studies [15,16,17] (Fig. 3). The specimens were tested in 20° lateral angulation (adduction) of the femoral shaft [15]. Load transfer between the load cell – attached to the machine actuator – and femoral head was ensured via an interconnected PMMA shell, simulating the acetabulum. A special custom-made foil from electro-conducting material was implemented at the corresponding articular surface of the cup and used for immediate automatic detection of implant cutting through the femoral head (cut-out) during testing and interruption of the test procedure to prevent implant damage as soon as an electric contact occurred. The cranial area for proximal load transfer to the bone was localized at the superior aspect of the femoral head. Distally, the specimen was attached to the machine frame via a cardan joint.
Each specimen was loaded in compression along the machine axis, starting with a quasi-static ramp from 50 to 200 N at a rate of 15 N/sec, followed by progressively increasing cyclic loading at 2 Hz with physiologic profile of each cycle [15, 17]. Keeping the valley load of each cycle at 200 N, its peak load started at 500 N and then increased at a rate of 0.1 N/cycle until failure of the bone-implant construct. The application of progressively increasing cyclic loading allows to achieve construct failure of specimens with different bone quality within a predefined number of cycles and has been demonstrated as useful in previous studies [17, 18]. The test stop criteria considered either cut-out of the implant through the femoral head, 30 mm vertical displacement of the machine actuator relative to the test beginning, or reaching an axial load of 4 kN.
Data acquisition and evaluation
Machine data in terms of axial displacement (mm) and axial load (N) were recorded from the machine controllers at 128 Hz. Based on this, initial axial construct stiffness was calculated from the ascending slope of the load-displacement curve from the initial quasi-static ramp within the linear range between 100 and 200 N.
Three-dimensional coordinates of the retro-reflective markers attached to the bone and implant were collected at 100 Hz using 5 infrared cameras (ProReflex MCU 120, Qualisys AB, Gothenburg, Sweden) to investigate the femoral head movements with respect to the shaft and implant in all six degrees of freedom. Based on the motion tracking data, varus deformation was defined as the relative femoral head-to-shaft rotational movement in the coronal plane. Furthermore, leg shortening was derived from the movement of the head center along the shaft axis. Finally, the rotation of the femoral head around the neck axis was evaluated. For that purpose, the neck axis was reconstructed via virtual rotation of the shaft coordinate system around its anteroposterior axis by the amount given from the measured caput-collum-diaphyseal (CCD) angle of each femur. The outcome values of these parameters were analyzed after 1000, 5000 and 10,000 cycles in peak loading conditions to evaluate the degradation of the construct stability over the course of cycles.
Furthermore, 15° varus deformation, 10 mm leg shortening and 15° femoral head rotation around the neck axis – considered with respect to the beginning of the cyclic test – were defined as clinically relevant failure criteria, and the numbers of cycles until fulfilment of each of these criteria in peak loading condition were calculated for each specimen separately.
Anteroposterior radiographic images were taken at the beginning (50 N) and the end (200 N) of the quasi-static ramp, and then at timed intervals every 250 cycles during the cyclic test at valley loading (200 N) using a triggered C-arm (Siemens ARCADIS Varic, Siemens Medical Solutions AG, Erlangen, Germany). X-ray images taken at the end of each test served to evaluate the catastrophic failure modes of the specimens.
Statistical analysis was performed with SPSS software package (IBM SPSS Statistics, V23, IBM, Armonk, NY, USA). Shapiro-Wilk test was conducted to screen and prove the normality of data distribution. Independent-Samples t-test was applied to detect significant differences during the equality/divergence analysis for comparison of the two different procedures for creation of the stable pertrochanteric fractures in the two groups. Significant differences between the groups regarding BMD, axial stiffness and cycles to 15° varus deformation, 10 mm leg shortening and 15° rotation of the femoral head around the neck axis were identified with Paired-Samples t-tests. General Linear Model Repeated Measures test was applied to detect significant differences between the groups with regard to the parameters evaluated over the three time points after 1000, 5000 and 10,000 cycles. The homogeneity of biomechanical fracture stability was compared between the two groups by conducting a Wilcoxon-Signed Rank test over the pooled SDs of the outcomes derived from the three motion tracking parameters of interest investigated after 5000 test cycles, and the cycles to clinically relevant failure. The latter non-parametric paired approach was applied only to relate the corresponding ranks within the pooled SDs of the different parameters, because such an approach is anticipated to be sound when comparing values emerging from different physical variables. Level of significance was set to 0.05 for all statistical tests.