Twenty-four embalmed adult human cadaver femurs were obtained from the Department of Anatomy, Southern Medical University in Guangzhou, China (average age: 77.8 years, range: 67–89 years; 15 males and nine females). The Institutional Review Board of Guangdong Provincial People’s Hospital waived the informed consent procedure for this portion of the study. After removing soft tissue, anterior–posterior and lateral radiographs were taken for each femur to exclude pre-existing disease, deformity, or trauma.

To evaluate the heterogeneity of bone density and geometry in the femur neck during the biomechanical testing, QCT scans were carried out using a clinical scanner (SIEMENS SOMATOM 64, 140 kV, 80 mAs, 0.5 * 0.5 mm/pixel resolution, and 1 mm slice thickness) for each cadaver femur. Using a calibration phantom (MINDWAYS Software, Inc., San Francisco, CA), grayscale values were mapped to K₂HPO₄ equivalent density (ρ_KHP) using five tubes with reference densities and Hounsfield Units (HUs) were calibrated. Segmentation of the bone hard tissue from its surroundings was performed for each slice. The raw QCT images (in DICOM format) were converted into a binary format, and a combination of user-defined threshold limits with an edge following scheme was used for generation of the hard-tissue contours and elimination of soft tissue. BMD (g/cm³) of the femoral head, femoral neck, and intertrochanteric region was measured (Figure 1A).

Specific to geometric measurements, the femoral neck–shaft angle was defined as the angle formed between the femoral neck axis and the femoral shaft axis, and this angle was measured on QCT scout images for each patient (Lee et al., 2017). Hip axis length was defined as the length along the femoral neck axis, from the lateral aspect of the greater trochanter to the inner pelvic rim (Maeda et al., 2011). Neck width was also measured as described by Maeda et al. (2011).

After QCT scanning, 24 specimens were randomly and equally allocated to the following four groups: control group, 25% defect group, 50% defect group, and 50% defect + screw group (Figures 2B1–B4). The area of the bone defect was defined according to the description outlined by Çaypınar et al. (2016). Briefly, the femur neck is most likely a cylinder, and we adopted a cylinder with a 10 cm base (the circumference of the circle) and a height of 7 cm. For example, a 25% bone defect was obtained when we create a rectangle anterior cortical bone defect using a bone drill, with a 2.5 cm base and a 7.0 cm height. In the 50% defect + screw group, two 6.5 mm AO cannulated screws were inserted in parallel into the coronal plane at a 125° angle from the femoral shaft over the guiding K-wires. The two cannulated screws were inserted close to the superior and inferior margin of the femoral neck, and both screws advanced to within 5 mm of the subchondral bone to meet the requirements of the tip apex distance. The area of bone defect and the position of cannulated screws were further confirmed by X-ray (Figures 2A1–A4).

The specimens were then mounted with a fixed shaft in a servo-hydraulic test frame (ElectroForce 3500, TA instrument, United States). The angle between the loading axis and the proximal shaft was 15° during the mechanical testing to mimic the weight of the femur during a one-legged stance (Figure 1C) (Dou et al., 2019). The load was applied to the most cranial portion of the femoral head in the plane, spanned by the neck axis and the proximal femur axis. An axle bearing was inserted between the embedded femoral head and the test frame to allow rotation orthogonally to the loading axis through the femoral head. Before each test, all specimens were manually preloaded with a maximum of 100 N to avoid play between the embedded specimens and the test setup. The axial force was recorded via a 100-kN load cell (U3 force transducer; HBM, Darmstadt, Germany).

To keep the femora remained in the linear–elastic regime before the ultimate axial loading failure testing, a displacement range of 1 mm was chosen based on pre-tests, ensuring test conditions would not initiate plastic deformation and irreversible damage to the specimen. The axial stiffness was determined by vertically applying a vertical displacement of 1 mm maximum using displacement control 5 mm/min during the axial loading testing (Dou et al., 2019). Following a short relaxation period, a second compression cycle was applied with the same speed until the failure of load.

For a more detailed evaluation of the biomechanical property of the two cannulated screws, we also performed FEA. A healthy volunteer (male, 30-year-old) without previous history of surgery, trauma, and deformity in the lower limbs was recruited. Computerized tomography (CT) scan (SIEMENS SOMATOM 64) was performed on the lower limbs of the volunteer, with the slice thickness set at 0.5 mm. Image data of the right femur were imported into Mimics (version 21.0, Materialise NV, Leuven, Belgium) for three-dimensional (3D) reconstruction, which was based on the gray value of the tissue and segmentation of the region (Figure 1D). Subsequently, the model in STL format was imported into the Geomagic Wrap (version 2017, Geomagic Corporation, United States) for smoothing, meshing, and fitting surface processing.

A 3D computer-aided design software Solidworks (version 2017, Dassault Systèmes, Waltham, MA, United States) was then used to create the cortical and cancellous bone defect caused by the lesion and mimic the implantation of two cannulated screws (Figure 1E). Four models including control (intact), 25% bone defect, 50% bone defect, and 50% bone defect + two 6.5 mm cannulated screws were developed (Figures 2D1–D4). The 25% and 50% bone defects were obtained as outlined previously . To simulate the characteristics of the surgery more accurately, the position of the screws was strictly followed as per conventions of clinical practice, which we described in detail in the biomechanical analysis section.

Data from the four models were imported into Abaqus 6.14 software (Dassault Systèmes S.A., France) to generate C3D10 tetrahedral elements. As per the study of Palumbo et al. (2014), Ti6Al-4V titanium was used as the internal fixation instruments, and the property parameters (Young’s Modulus and Poisson’s Ratio) of materials are listed in Table 1. The models were meshed to 1.0 mm, equal-sized facets, with >1,000,000 elements and ranging from 1,088,910 to 1,464,698. A mesh convergence test was conducted so that the deviation was less than 2%. In the current study, the cortical bone and cancellous bone interfaces were given with a tie feature. Regarding the interfaces between the screw thread and bone, the interfaces between the bone and metal were assigned as a sliding contact, with the frictional coefficient being 0.46 (Jiang et al., 2021).

It was reported that during static balance, approximately one-third of a person’s body weight falls on each hip vertically (Çaypınar et al., 2016). Accordingly, a load of 700 N was applied straight down on the femoral head to simulate the one-leg-stance of an obese adult. Subsequently, the load force and boundary conditions were established for a one-leg-stance (Figure 1F). To mimic biomechanical testing, the area on the weight-bearing region of the femoral head, which intersects the mechanical axis, was set as the loading position, and the distal end of the femur was fixed. Measurements of von Mises stress were queried at 14 points of interest on the femoral neck that lay within the mid-coronal plane (Figure 5A). Von Mises stresses on the superior margin of the femoral neck were defined as the mean of von Mises stress from point 1 to point 7. The mean of von Mises stress from point 8 to point 14 was calculated as the Von Mises stresses on the inferior margin.

In order to test the failure loads, four models were imported to Hypermesh14.0 software (Altair, United States) to be meshed as tetrahedral with a size of 2 mm²¹. The mimic of the femoral neck fracture was established by Hypermesh and LS-DYNA software. An elastic–plastic material model (*MAT_03) in Hypermesh14.0 was used to simulate cortical and trabecular bone behaviors. According to previous studies (Li et al., 2010; Khor et al., 2018), the femoral fracture was defined by detecting a failure strain that was initiated and propagated by element deletion as long as the strain of an element reaches the limit. The fracture analyses were performed by LS-DYNA (LSTC, Livermore, CA) software to investigate the biomechanical effect of the failure test (Zheng et al., 2021). Material properties applied for fracture analysis are shown in Table 1. The density of cortical bone, cancellous, and Ti6Al–4V titanium was obtained from previous reports (Dhanopia and Bhargava, 2017; Zheng et al., 2021). According to the equations between bone density (ρ), Young’s modulus (E), and yield stress (σ) (Morgan and Keaveny, 2001; Duchemin et al., 2008), cortical and cancellous bone’s yield stress was calculated respectively. The contact settings between bone and screws were the same as those in the static analysis. Meanwhile, the loading settings were applied according to the biomechanics test of the cadaveric femurs until the failure of load.

Continuous data were expressed as mean ± standard deviation or median with an interquartile range. Pearson’s correlation analysis was conducted to test the correlation between maximum displacements of FEA and the stiffness of the mechanical test. The data of QCT measurement, biomechanical test, and FEA analysis of different groups were compared with one-way ANOVA followed by the least significant difference (LSD) test for multiple comparisons. The two-sided Student’s t-test was used to compare the von Mises stresses of the inferior margin and the superior margin of the femoral neck within the groups. Bonferroni correction was performed for multiple testing. Differences were considered significant if p < 0.05. SPSS 20.0 statistical software (Chicago, IL, United States) was used for these analyses.

No significant difference in BMD was measured at the femoral neck (Figure 3A), the femoral head (Figure 3B), or the intertrochanteric region (Figure 3C). Furthermore, no significant difference was observed between femoral geometry parameters including neck-shaft angle (Figure 3D), width of the femoral neck (Figure 3E), and length of the femoral neck (Figure 3F) of the cadaver femurs within the four groups. Thus, any potential bias caused by heterogeneity of bone density and geometry in the femur neck during the biomechanical testing is likely limited.

As illustrated by the cadaveric biomechanical testing, marked decreases in the failure load were observed in the 25% defect (−41.1%, p = 0.038) and 50% defect groups (−56.8%, p = 0.004) when compared with the intact femur (Figure 4A). There was a significant improvement in the failure load after insertion of two 6.5-mm cannulated screws in the 50% bone defect model (+95%, p = 0.048), and no significant difference was found when this group was compared to that of the intact femur (Figure 4A). As for stiffness measurements, a similar trend was found among the groups but did not reach statistical significance (Figure 4B). Moreover, as can be seen in Figures 2C1–C4, all of the fractures occurred in the femoral neck region.

The maximum displacement of the intact femur, 25% defect model, 50% defect model, and 50% defect + two cannulated screw model was 0.885 mm, 0.940 mm, 0.952 mm, and 0.938 mm, respectively. A significant linear correlation between the FEA stiffness and the stiffness of the cadaveric test (r = 0.957, p = 0.04) was observed, suggesting that our static analysis FEA models were valid.

The von Mises stress distributions of the four models are shown in Figures 5A–D. The quantitative von Mises stresses of selected points on the inferior and superior margin of the femoral neck are given in Figures 5E,F. When compared with the intact femur, the 50% defect model showed significantly higher mean von Mises stresses on both the inferior aspect (+140%, p = 0.001) and superior aspect (+180%, P<0.001) of the femoral neck (Figure 5G). Insertion of the two screws significantly reduced the mean von Mises stresses on the inferior aspect in the 50% bone defect model (−71%, p = 0.043), while no significant difference was found when compared with the control group (Figure 5G). As shown in Figure 5G, there were more significant increases in the von Mises stress measurements on the inferior aspect of the femoral neck than on the superior aspect, in both the 25%-bone defect group (p = 0.006) and 50%-bone defect group (p = 0.024).

Regarding the failure load testing, the fracture lines of the four models are shown in Figures 6A–D. As illustrated in Figure 6E, there was a marked improvement in the failure load after insertion of two 6.5-mm cannulated screws in the 50% bone defect model. A statistically significant linear correlation of the failure load between FEA and cadaveric tests was also found (r = 0.953, p = 0.047).