**2. Background**

A fractured stainless steel 3.5 mm LCP proximal humerus locking plate (PHILOS), originally manufactured by Synthes, is investigated for the failure mechanism(s) responsible for its failure. A typical LCP proximal humerus locking plate is shown in Figure 1. It can be seen that the plate was inverted to handle the anatomical shape of the tibiotalocalcaneal area, where the proximal part of the plate was fixed to the calcaneus and talus. The fractured pieces are analyzed to determine the mechanism of failure. In Paper I, the PHILOS plate failure analysis was carried out and a state of the art literature review of hybrid constructs is summarized [16]. This paper not only presents the results of finite element simulations of the PHILOS construct (plate and screws) but also examines the failure mechanisms of the screws that failed.

**Figure 1.** The fractured 3.5 mm LCP proximal humerus locking plate shaft showing the crack initiation and propagation in two directions (red arrows) (**left**). The proximal part of the plate is divided into six levels, from A to F (**middle**). It can be seen that the plate is inverted to align with the anatomical shape of the tibiotalocalcaneal area, where the proximal part of the plate was fixed to the calcaneus and talus (**right**).

The PHILOS plate and screws were used for a 68-year-old patient that had a right ankle plantar arthrodesis, deformation on the right ankle, and dislocation on the lateral subtalar joint. The distal part of the plate was fixed to the lateral part of tibia and calcaneus by 3.5 mm cortical screws. Additionally, 3.5 mm locking screws were used to fix the proximal part of the plate. Furthermore, a 6.5 mm cannulated screw was used to apply compression on tibiotalar and subtalar joints. In addition, 3.5 mm screws with washers were used to support the fusion of the talonavicular joint. After six years, the plate and the screws were removed from the body. Table 2 summarizes all the parts that are submitted for investigation.


**Table 2.** Pieces of the PHILOS system submitted for investigation.

Note: X-ray images with respect to dates are shown in [16].

#### **3. Visual Examination**

#### *3.1. The Plate*

Visual examination demonstrated that the LCP plate was fractured into three pieces. The anterior and posterior views of the plate are shown in Figure 1. The proximal part of the plate can be divided into six levels, from A to F [17], as shown in Figure 1.

The plate failed at levels B–E. Additionally, Figure 1 shows that the crack initiated from under one of the locking holes near the middle of the plate at level (E). This fracture progressed in two directions (the red arrows), one in the diagonal direction and another in the perpendicular direction, which caused the plate to fail into three pieces. The distal part of the plate shaft had been removed to perform some additional tests. Observing the plate under the optical microscopy showed a considerable amount of scratches and pitting. Details of the plate failure have been discussed in Paper I [16].

#### *3.2. The Screws*

Twelve screws were submitted for examination, as shown Figure 2. Four were 3.5 mm cortical non-locking screws, and three (75%) of these screws fractured into two pieces. The X-ray images showed that the first cortical screw failed after two years of implantation [16] and the other two screws failed after five to six years [16], visible in X-rays. One of the cortical screw heads remained on the plate after removal. Seven 3.5 mm locking screws were submitted for investigation, and three (42.8%) of these screws fractured into two pieces. It was difficult to visualize the time of failure of the locking screws after the implantation in the X-ray images because the screw density was high at the area of insertion of the locking screws. One of the locking screw heads and other undamaged locking screws remained in the plate after removal. In addition, one 6.5 mm cannulated screw fractured within five years of implantation, as was observed in the X-ray images [16]. No physical damage to the devices that may have led to the failure was observed. Radiographic evidence, as seen in Figure 1, shows that the screws failed ahead of the plate from the proximal end.

**Figure 2.** Screw visual observation showed the fractured screws (**upper**), some thread flattening (**middle**), and the fractured 6.5 mm cannulated screw (**below**).

#### **4. Fractographic Examination**

Optical and scanning electron microscopes were used to document the fracture surface characteristics such as extensive scratching, plastic deformation, rubbed surfaces, discoloration, and pitting, along with cleavage, cracking, deposits of debris, striations, and dimples. A PEMTRON scanning electron microscope (SEM PS-230, Seoul, Korea) with 10 KV was used to perform the fractographic examination. Fractographic analysis was performed on the plate and four screws in the fractured regions. By examining the fractured surface of the plate, quasi-cleavage fracture was

observed, as shown in Appendix A (Figure A1). The literature showed similar observations that the corrosion–fatigue mechanism was responsible for the failure of the plates [18–20]. However, another study showed that the failure of plates could be a result of improper material [21].

The cortical screws were marked as CS1, CS2, CS3, and CS4 from the proximal to the distal direction. The locking screws were marked depending on the level of insertion, shown in Figure 1, as LA, LB, LC, LD, and LE. Cortical screws CS1 and CS4 were examined with the SEM. Appendix A (Figures A2–A4) show the main details in CS1, where the presence of striations and microcracks can be noticed, while Appendix A (Figures A5–A7) shows the main details in CS4, where the presence of striations in different directions can be noticed. In addition, two locking screws were examined with the SEM. Appendix A (Figures A8–A11) shows the main details for the locking screws.

It can be noticed that the fractured surfaces of the screws were rubbed. Rubbing happened inside the body as the devices failed after two years of insertion and were kept inside the body for four more years, which caused the loss of some of the important features that caused the crack initiation. Additionally, it was observed that the cracks in the fatigue areas of the plate initiated from corrosion pits. In general, the investigation showed that the failure was a result of conjoint bending/torsion loading and corrosion–fatigue cracking that propagated from the bottom of pit(s). The pitting susceptibility of this material is widely documented in the literature arising from inclusions, which were reported in Paper I [16]. The mechanisms governing failure of the screws are similar to that found for the plate [16].

#### **5. Material Conformity**

The metallographic qualitative analysis was performed based on ASTM standards F138-03 and F139-03 [22]. A sample from the locking screw was taken. Cutting, mounting, and polishing were performed to prepare the sample. The polishing was done with different grades of silicon carbide papers (320–600 grit) and diamond abrasive solutions (9–0.01 μm). After the polishing, X-ray energy dispersive spectroscopy (EDS, Thermo Fisher Scientific, Hillsboro, OR, USA) was performed on the screw sample surface, and the results were compared with the ASTM standards F138-03 and F139-03 reported for stainless steel 316L to ensure that there is no significant difference that might have caused the failure [18,23]. Figure 3 shows the peaks of different element weight percent in the screw, and the data on each area is summarized in Table 3. Additionally, ASTM standard F139-03 has set additional requirements to ensure that the material meets the specifications, which is shown in Equation (1):

$$\text{\textquotedblleft Cr} + (3.3 \times \text{\textquotedblleft} \text{\textquotedblright}) \text{\textquotedblright} \geq 26.0 \tag{1}$$

**Figure 3.** Energy dispersive X-ray analysis shows the peaks of different element weight percent in the cortical screw sample.


**Table 3.** X-ray energy dispersive spectroscopy (EDS).

The screw confirmed to the ASTM standard and there is no significant difference that might have caused the failure.

#### **6. Computational Simulations of Failures**

#### *6.1. Finite Element Analysis*

Three-dimensional models of the plate and the screws: cortical, locking, and cannulated were constructed using SolidWorks (Dassault Systèmes SolidWorks Corp., Concord, MA, USA) and imported in ANSYS Workbench 16.2 (ANSYS Inc., Canonsburg, PA, USA) to simulate the loading conditions and regions of stress development. The dimensions of the cortical, locking, and cannulated screws were taken from the Synthes brochure [24], and the PHILOS plate was measured from the submitted samples, Figure 4. Models were generated depending on patient-specific geometric information that was obtained manually from the X-ray images. The modeling of each screw has depended on the angle of its insertion inside the bone of this patient. However, locking screw placement angles were not clearly visible in the X-ray due to the presence of too many screws that were used in the fixation, Figure 5. That being said, the locking screws placement angles were measured manually by inserting locking screws inside a sample plate. Figure 6 shows the angles of the locking screws at level A. Locking screws at level A were inserted slightly upward with a 30◦ angle above the neutral axis of the center of the screw hole and parallel to each other. The locking screws at level B were inserted perpendicular to the plate with an inward 30◦ angle. The locking screws at level C were inserted upward with a 40◦ angle above neutral axis and outward 30◦ angle. The locking screws at level D were inserted slightly upward with a 10◦ angle above the neutral axis. Finally, the locking screw at level E was inserted upward with a 30◦ angle above the neutral axis and slightly outward. The heads of the locking screws were fixed with no displacement allowed. Additionally, the material properties of stainless steel 316L were assigned according to the ASTM standard [22], shown in Table 4. The cortical and cancellous components were modeled with a 12.7 GPa and a 0.9 GPa modulus of elasticity and a 0.3 and a 0.2 Poisson's ratio, respectively [25]. Materials were treated as homogeneous, isotropic, and a linear elastic analysis was carried out. All the results were within the limitation of this assumption assuming all loads transferred from bone to the plate. Figure 7 shows the loading conditions that have been applied to the plate and the screws.

It is claimed that when a PHILOS plate is closed with all the screw holes, the fixation is stable and fracture union takes place with time. However, with the current subject, the device started to fail before the bone union took place, which caused some displacement and led the screws to fail after two years, visible when X-rays were taken. We ignored the effect of screw failure on the adjacent level screws both in terms of construct plate-screw placement angles and load transfer. As there is no viable way to identify the amount of screw micro-motion in this construct, it was assumed that displacement in the cortical screws within the range of 0.05–0.25 mm occurring within the few months after load bearing based on our research on hybrid constructs where the stiffness drop occurs within the first 5000 fatigue cycles [7] resulting in altering the coefficients of friction between the screws and bone; 0.5 with all screws intact to 0.1 when screws began to fail. All the locking screws were assumed to be fixed, while some displacement was allowed for the cortical screws along the *x*-axis, as shown in Figure 8, simulating human gait to reflect loading and deformation. The loading was applied in two steps; (1) the preload torque for the installation of the screw, and (2) the axial loads. The applied load on the ankle joint might reach 3.5 times of the body weight in the gait cycle during the push-off stage [26]. The body weight of the patient was assumed to be 80 kg in this study. However, for patients with abnormal gait cycles, the applied load may reach four times of the body weight (3000 N) or higher. To simulate the contact analysis in the model, the target and contact surfaces between the plate and bone, the plate and screws, and bone and screws were defined by not merging the nodes between the components and assuring the union and the transformation of the loads and the deformation between the plate, screws, and bone with a coefficient of friction of 0.4 between the plate and the bone [27], 0.35 between the plate and the screws [28], and five different coefficients of friction were obtained between the screws and the bone (0.1, 0.2, 0.3, 0.4, and 0.5) [27]. Furthermore, five different axial loads were tested to observe the PHILOS device in normal and extreme situations with axial loads ranging from 500–2000 N with a torque load 5.0 Nm [29,30]. The loading in the screws had been inclined to match with the insertion angle of each of the screws. In addition, two screw pattern designs (Figure 8) were assumed in order to investigate the effect of screw position on the maximum stress of the plate: design (A) where the screws inserted at levels A, C, D, and E, and design (B) where the screws inserted at levels A, B, C, and E, as shown in Figure 8, and the loading and boundary conditions are shown in Figure 7. In addition, the tetrahedral element type was used in this study. The FE meshing was automatically adapted through Ansys, where convergence tools were used with 5% convergence, which resulted in approximately 50,000 elements and 85,000 nodes, as shown in Figure 7. Some studies suggested removing of unused screw holes from the model, however, in this study all the holes were included in the design, and only filled with screws per designs A and B, Figure 8.

**Figure 4.** The dimensions of the 3.5 mm standard locking compression plates (LCP) proximal humerus locking plate.

**Figure 5.** X-ray image showing the screws used in the tibiotalocalcaneal fixation.

**Figure 6.** Variable angle placement of the locking screws at level A.


**Table 4.** Stainless Steel 316L material properties [22].

**Figure 7.** The loading and boundary conditions applied on the model.

**Figure 8.** Screw pattern designs. Design A (**left**) and design B (**right**) [17].

#### *6.2. The Results*

The maximum von Mises stress in the plate was obtained with two screw pattern designs, five axial loads and coefficients of friction between the screw and the plate, and five cortical screw displacements, and the results are summarized in Table 5. Figure 9 shows the von Mises stresses of the 316L SS plate with screw design B, 500 N axial load, 0.5 coefficient of friction between the screw and the plate, and 0.1 mm cortical screw displacement. It was observed that the maximum stresses are in the levels B–E, which validates the visual and fractography examinations. Additionally, the maximum stresses were between 201 and 346 MPa for 500–750 N axial loading, which is significantly lower than the yield strength (<sup>σ</sup>y = 690 MPa) of the material. However, for the fatigue limit at 10<sup>7</sup> cycles, the maximum stresses should be equal to, or less than, 440 MPa to be in the safe region [31]. On the other hand, the maximum stresses were between 504 and 692 MPa for 1500–2000 N axial loading, which is significantly higher than the failure criteria used. At this level, the developed stresses are likely to cause plastic deformation and failure prematurely. If the environmental effects were included, the corrosion fatigue

crack propagation may activate at this level of loading. Since the ankle deformity gives rise to the increased level of force development in the range of 4–8 times the body weight, the extrapolated von Mises stress may reach as high as 900 MPa. In addition, the simulations show that higher coefficients of friction reduce the von Mises stress for both pattern designs and for different loading conditions. As a result estimation of the actual loading behavior of the plate from the activities of daily living of the individual, together with mechanical parameters such as the coefficient of friction, loosening, and axial loading in the screw and plate, will be very difficult to isolate and predict the combined mechanisms and failure behavior.


**Table 5.** The maximum von Mises stresses of the plates were obtained for five different loads, five coefficients of friction, five cortical screws displacement, and two screw pattern designs.

On the other hand, the maximum von Mises stresses increased with the increase in cortical screw displacements, as the screws began to fail, as shown in Figures 10 and 11. The maximum von Mises stresses of the cortical screws occurred near the head of the screw when the screw was perpendicular to the plate, as shown in Figure 12 (will be further discussed in Section 7). However, the maximum von Mises stress of the cortical screw was near the middle of the screw when the screw was at an upward angle, which will be further discussed in Section 7. In addition, the maximum von Mises stresses of the locking screws were distributed across the screws and were higher at levels A, C, and E, where the screws have slightly upward angles. Figure 13 shows the von Mises stress of the locking screw at level C and the screw insertion angles. Table 6 shows the maximum von Mises stresses of the cortical and locking screws. Finally, the von Mises stress of the cannulated screw is shown in Figure 14, and it is observed that the stresses are higher in the shaft of the cannulated screw.

**Figure 9.** The von Mises stresses in the 316L SS plate with screw design B, 500 N axial load, 0.5 coefficient of friction between the screw and the plate, and 0.1 mm cortical screw displacement.

**Figure 10.** The surface plot of the maximum von Mises stresses in the 316L SS plate with screw design B and 0.25 mm cortical screw displacement vs. the coefficient of friction (COF) and the axial force.

**Figure 11.** The surface plot of maximum von Mises stresses in the 316L SS plate vs. the coefficient of friction (COF) and the cortical screw displacement (mm) for different axial loads (**a**) 500 N, (**b**) 750 N, (**c**) 1000N, (**d**) 1500 N, (**e**) 2000 N.

**Figure 12.** Comparing a two-year post operation X-ray image with the von-Mises stress of the 316L SS cortical screw (CS3).

**Figure 13.** The von Mises stress of the 316L SS locking screw at level C and the screw insertion angles.

**Table 6.** The maximum von Mises stresses in the screws.


**Figure 14.** The von-Mises stresses in the cannulated screw.

## **7. Discussion**

This paper not only examines the PHILOS plate computational simulations, but also examines the failure mechanisms of all the screws that failed. On visual examination, it can be noticed that the crack initiated from under one of the locking holes near the middle of the plate. Observing the plate under optical microscopy indicated that the surface of the plate had a considerable amount of scratches and pitting. Seventy-five percent of the cortical screws and 42.8% of the locking screws fractured into two pieces. Chemical, qualitative analysis of our results shows that the screw confirmed to the ASTM standards F138-03 and F139-03 and there is no significant difference that may have caused the failure.

Optical and scanning electron microscopes were used to document the fracture surface characteristics, such as extensive scratching, plastic deformation, rubbed surfaces, discoloration, and pitting, along with cleavage, cracking, deposits of debris, striations, and dimples. Fractographs show a quasi-cleavage fracture. Azevedo [19], Kanchanomai et al. [20], and Majid et al. [32] showed similar observations that the corrosion–fatigue mechanism was responsible for the failure of plates. However, Sivakumar et al. [33] showed that the failure of plates could be a result of improper fixation. On the fracture surface of the cortical screws, we noticed the presence of striations, microcracks, and rubbed surfaces. The presence of the inclusions, as presented in paper I [16] for this material, increased the pitting susceptibility of SS316L and possible transitioning to corrosion fatigue cracking. FEM analysis clearly shows that at any level of load bear, the developed stresses will cause plastic deformation at the sites of screw holes and this plasticity will initiate a crack. Thapa et al. showed a similar fractographic examination for the locking plate [34]. Indications of these features show that the plate failed by corrosion fatigue. However, overloading separated the screws into two parts. Radiographic evidence shows that the screws failed ahead of the plate from the proximal end. Finally, fractographic examination of the cortical and locking screws supports the mechanism of corrosion–fatigue fracture from crack initiation sites due to the presence of inclusion bodies and pits.

In addition, two screw pattern designs were assumed in order to investigate the effect of the screw position on the maximum stress of the plate. Some studies suggested the removal of unused screw holes from the model, however, in this study all the holes used in the simulation provided higher fixation strength and stability. The results of the computational simulation showed that the maximum stresses simulated the loading conditions and regions of stress development. The simulation showed that having higher coefficients of friction reduces the von Mises stresses for both pattern designs and for different loading conditions. On the other hand, the maximum von Mises stresses of the PHILOS plate increased with the increase in cortical screw displacements. The results for the visual, fractography, and quantitative examinations of the cortical, locking, and cannulated screws are summarized in Table 7.


The maximum von Mises stresses of the cortical screws were near the head of the screw when the screw was perpendicular to the plate and near the middle when the screws were at upward angles. If we compare the von Mises stresses of the four cortical screws across the cortical screws length, it was observed that the maximum stresses are almost similar for all the screws, but the position of the maximum stresses varied depending on the angle of the screws, as shown in Figure 15. Additionally, the graph shows that the stresses are, at a minimum, at the head of the C1 and C2 screws. Then, a sharp increase in the stresses can be noticed at the shaft of the screw near the head, as this is the area where the screw cuts through the cortical bone and is exposed to the axial load of the bone. Then the stress distribution decreased gradually. On the other hand, C3 and C4 cortical screws were at an upward angle. This angulation increased the stress by 165% on the shaft of the screw than the head of the screw. In the case of locking screws, the maximum von Mises stresses were distributed across the screws. However, the maximum von Mises stresses were only 110% higher at levels A, C, and E where the screws have slightly upward angles. Figure 16 shows a comparison between the von Mises stresses across the locking screw length. It can be noticed from the graph that the maximum stresses were near the heads of the screws and as the screws entering the cortical bone. Finally, it was observed that the stresses were 170% higher in the shaft of the cannulated screw than the head and the tip of the screw. In general, the maximum von Mises stresses experienced by the screws were lower than the fatigue stress limit. However, corrosion in this case may drastically affect the lifetime of the screws by reducing the number of cycles to failure. The numerical analysis was conducted to understand the impact of different factors on the maximum von Mises stresses of the locking compression plate. The statistical analysis showed that the applied load has a significant effect on the maximum stresses, but the screw pattern design does not have a significant effect. In summary, the fractographic examination of the cortical and locking screws supports the mechanism of corrosion–fatigue fracture from crack initiation sites due to the presence of pits and/or high plasticity regions. Additionally, the cracking followed by the rotating bending mechanisms since the axial load and torsion control that behavior. This study validates the physical failure with the computational simulations.

**Figure 15.** The von-Mises stresses of the 316L SS cortical screws across the screw length.

**Figure 16.** The von-Mises stresses of the 316L SS locking screws across the screw length.
