Operative materials and pathway
The hospital ethics committee approved all study-related procedures. We used 18 synthetic left human humerus bones (SAWBONES; Pacific Research Labs) with specially created structural defects  (Figure 1A, 1B). All proximal humerus fractures were fixed using six sets of locking plates and matching screws (Depuy Synthes). The 18 bones were randomly divided among three groups, resulting in three groups of six bones each. Then, we created a comminuted (two-part) fracture model of the humerus surgical neck for each bone. All the specimens were taken from the humeral heads and cut while preserving 220 mm. We established a horizontal line 10 mm below the humerus surgical neck and used a Stryker oscillating saw (saw blade thickness 1 mm) to cut the bone along this line. The saw was able to cut through the entire cortex, creating a greater tuberosity osteotomy along the 50° humerus coronal oblique line.
The medial cortical defect model was completed according to the Sanders method, as follows. An osteotomy was created 5 mm parallel to the distal end of the fracture line. We preserved the 1/3-peripheral cortex of the lateral greater tuberosity of the proximal humerus for plate fixation. The plate was placed on the lateral side of the humerus. The upper end was 8-10 mm from the apex of the greater tuberosity, and the medial side was 5 mm from the outer side of the intertubercular groove. The posterior plate was placed at the junction of the posterior metaphysis and the humeral surgical neck.
Group A bones underwent fixation using a Proximal Humerus Internal Locking System (PHILOS) plate support alone (Figure 1C, 1D). Group B bones were fixed using a PHILOS plate and posterior plate, but without medial column support screws (Figure 1E, 1F). Group C specimens were fixed using a PHILOS plate, a posterior plate, and medial column support screws (Figure 1G, 1H). A 20 cm long distal humerus specimen was resected 15 cm from the fracture line. The distal clamp of the specimen was fixed and embedded at a depth of 12 cm within denture base resin.
Axial compression test
The top of the humeral head of each bone was subjected to vertical and vertical-downward pressures . Each test was completed in triplicate using a preload of 50 N, a loading speed of 5 mm/min, and a maximum displacement of 1 mm. After each test we recorded the maximum load and created a load curve by recording the test data and calculating the compressive stiffness. The average value of the maximum load and compressive stiffness was also calculated.
We used a biomechanical tensile torsion test fixture with evenly distributed clamps and two circular holes that were 8 mm in diameter . Eight semi-threads that were 5.7 cm long and had a diameter of 8 mm were used with universal friction bolts that were passed through the circular holes to fix each humerus head. Each test was completed in triplicate using a preload of 0 N•m, a rate of 12°/min, and a maximum torsion angle of 120°. Maximum torque values were recorded after each test and used to draw the loading curve and calculate the torsional stiffness. We used average values of maximum torque and torsional stiffness.
Shear compression test
Each humerus was placed in 20° abduction to simulate upper limb support when falling. In this position, the proximal humerus receives shear weight forces which easily lead to fracture . We applied vertical down-pressure on the top of the humeral head, with a preload of 50 N, a rate of 5 mm/min, and a maximum displacement of 1 mm. Each test was completed in triplicate, and we recorded the maximum load each time. Each loading curve was constructed according to load data and used to calculate compression stiffness of each bone using the average maximum load and compression stiffness values.
Model failure test
We tested the tibial position using shear compression, with a preload of 50 N and a rate of 5 mm/min . The test was stopped when the bone fractured, the plate or the screw broke, and the load reached its peak value. We considered the maximum load to be the model failure load.
We preloaded each bone with 50 N. According to the anatomical structure of the shoulder joint and Poppen and Walker’s method, forces are largest when a normal shoulder joint is abducted 90° under physiological conditions. We used a load of 600 N, a frequency of 1 Hz, and 10,000 loading repetitions and compared displacement size before and after loading.
Resistance strain gauge test
Four strain gauges were placed on the medial and lateral sides of the proximal and distal ends of the fracture. We then completed the axial compression test, torsion test, shear compression test, and fatigue test. We recorded the maximum strain observed during the axial compression test (displacement 1 mm), the torsion test (torsion angle 5°), the shear compression test, and the fatigue test (where 600 N loads were applied 10,000 times). We additionally recorded changes in maximum strain associated with 1 mm displacements.
Computer-based three-dimensional finite element analysis
After providing written informed consent, the first author of this study (age 28 years, height 174 cm, and body mass 70 kg) underwent an x-ray examination to exclude shoulder joint lesions and injuries, and a shoulder joint CT scan. Imaging data obtained by CT scan was entered into Materialise's Interactive Medical Image Control System (MIMICS) 17.0 software (Materialise, Belgium) in DICOM 3.0 format to create a three-dimensional model of a proximal humeral fracture and internal fixation. The solid geometry model was then mesh-optimized using the MIMICS 17.0 FEA software module. Cortical bone, cancellous bone, and titanium were regarded as isotropic materials. The mesh node model information was exported, and we used ANSYS (Canonsburg, PA) software to read and generate a three-dimensional finite element model for static structural analyses. The material properties included cortical bone (elastic modulus 2000 MPa, Poisson's ratio 0.3), cancellous bone (elastic modulus 100 MPa, Poisson's ratio 0.26), and the plate (elastic modulus 120,000 MPa, Poisson's ratio 0.3). For the distal end of the humerus fixed boundary conditions, we applied 600 N axial pressure to the humeral head, and the specimen was abducted by 20°. The 600 N axial load was also used to simulate shear forces on the humerus that are sustained during a typical fall. The head were combined to perform 5×N•m torsional loading and we observed the model’s stress distribution characteristics under maximum physiological and experimental stressors.
The data were analysed using SPSS 19.0 statistical software (IBM Corp., Armonk, NY). Data are expressed as means ± standard deviations. We compared measures across the three groups using one-way ANOVA and the Student–Newman–Keuls (SNK) method, as appropriate. P < 0.05 indicated statistical significance.