**Influence of the Acetabular Cup Material on the Shell Deformation and Strain Distribution in the Adjacent Bone—A Finite Element Analysis**

#### **Danny Vogel \* , Matthias Klimek, Michael Saemann and Rainer Bader**

Biomechanics and Implant Technology Research Laboratory, Department of Orthopaedics, Rostock University Medical Center, Doberaner Straße 142, 18057 Rostock, Germany; matthias.klimek@gmx.de (M.K.); michael.saemann@med.uni-rostock.de (M.S.); rainer.bader@med.uni-rostock.de (R.B.)

**\*** Correspondence: danny.vogel@med.uni-rostock.de; Tel.: +49-381-494-9375

Received: 16 January 2020; Accepted: 13 March 2020; Published: 18 March 2020

**Abstract:** In total hip arthroplasty, excessive acetabular cup deformations and altered strain distribution in the adjacent bone are potential risk factors for implant loosening. Materials with reduced stiffness might alter the strain distribution less, whereas shell and liner deformations might increase. The purpose of our current computational study was to evaluate whether carbon fiber-reinforced poly-ether-ether-ketones with a Young´s modulus of 15 GPa (CFR-PEEK-15) and 23 GPa (CFR-PEEK-23) might be an alternative shell material compared to titanium in terms of shell and liner deformation, as well as strain distribution in the adjacent bone. Using a finite element analysis, the press-fit implantation of modular acetabular cups with shells made of titanium, CFR-PEEK-15 and CFR-PEEK-23 in a human hemi-pelvis model was simulated. Liners made of ceramic and polyethylene were simulated. Radial shell and liner deformations as well as strain distributions were analyzed. The shells made of CFR-PEEK-15 were deformed most (266.7 μm), followed by CFR-PEEK-23 (136.5 μm) and titanium (54.0 μm). Subsequently, the ceramic liners were radially deformed by up to 4.4 μm and the polyethylene liners up to 184.7 μm. The shell materials slightly influenced the strain distribution in the adjacent bone with CFR-PEEK, resulting in less strain in critical regions (<400 μm/m or >3000 μm/m) and more strain in bone building or sustaining regions (400 to 3000 μm/m), while the liner material only had a minor impact. The superior biomechanical properties of the acetabular shells made of CFR-PEEK could not be determined in our present study.

**Keywords:** modular acetabular cup; poly-ether-ether-ketone (PEEK); titanium; ceramics; ultra-high-molecular-weight polyethylene (UHMW-PE); implant deformation; strain distribution; bone stock

#### **1. Introduction**

The aseptic loosening of the acetabular cup due to stress shielding and altered strain distributions within the adjacent bone stock is a common cause for the failure of a total hip replacement [1]. If the modular cups are stiffer than the adjacent bone cavity, the strain distribution is altered, leading to high strains in some regions and low strains in others. It results in a potential risk of bone fracture on the one hand and bone atrophy on the other [2–4]. To avoid these effects, the stiffness of the acetabular cup should be similar to the adjacent bone [5]. Dickinson et al. investigated the influence of the material of monolithic acetabular cups on the pelvis cortex surface strains using a composite bone model. It was shown that monolithic cups made of ultra-high molecular weight polyethylene (UHMWPE) have less influence on the load transfer into the bone stock, when compared to metallic cups, as the reduced elastic modulus of UHMWPE bears more similarity to the bone stock [2]. In terms of the liner material, Kim et al. observed no significant differences in bone remodeling after five years postoperatively when

ceramic and UHMWPE liners combined with structurally identical metallic shells were compared [6]. Therefore, the shell material of modular acetabular cups might have a stronger influence on the strain distribution than the liner material and shells made of polymers might have a lower influence on strain distribution.

Besides UHMWPE, poly-ether-ether-ketone (PEEK), often applied with carbon fibers for reinforcement (CFR-PEEK), is a biocompatible polymer that has been introduced as a material for monolithic cups and liners, but not as a material for acetabular shells [7–11].

However, PEEK and CFR-PEEK might be suitable alternatives to metal shells in modular acetabular cups as well, in order to avoid adverse effects on stress and strain distribution within the bone stock. However, it was shown that the reduced stiffness of shells made of PEEK and CFR-PEEK leads to increased shell deformations in the case of press-fit fixation in a finite element analysis (FEA) [12]. In particular, shells made of pure PEEK without reinforcement were deformed excessively, resulting in strong deformations of the liners, which might cause further problems in vivo. Excessive liner deformations can lead to a reduced clearance between the liner and femoral ball head and subsequently to increased frictional torques and increased wear rates [13–16]. Moreover, the seating of the liner might be jeopardized by strong shell deformations [17,18], which might lead to increased peak stresses within ceramic liners and, therefore, an increased risk of fracture [19–21]. Thus, shells made of PEEK without carbon fiber reinforcement do not seem to be a suitable alternative, whereas the suitability of shells made of CFR-PEEK needs to be further evaluated.

Hence, the aim of this computational study was to evaluate whether acetabular shells made of CFR-PEEK can serve as suitable alternatives to shells made of titanium in terms of strain distribution in the adjacent bone and whether the stiffness of such shells is sufficient to withstand adverse shell and liner deformations when they are pushed into an under-reamed bone cavity.

#### **2. Materials and Methods**

In the present numerical study, the insertion of a modular acetabular cup into a hemi-pelvis was simulated using ABAQUS/CAE (v 6.12-2, Dassault Systèmes Simulia Corp., Providence, RI, USA). The modular cup was designed based on a commercially available cup consisting of a titanium shell for press-fit fixation and a ceramic liner. The shell was flattened and spherical, with an outer diameter of 54 mm and a wall thickness of approximately 5.8 mm in the region of press-fit contact. The liner had an inner diameter of 36 mm and a maximum wall thickness of approximately 3.9 mm. In terms of simplification and comparability, the liner geometry was identical for both liner materials (ceramic and polyethylene), even so the real parts would have geometric differences in the dependency of the material. To apply realistic joint loads, a ceramic ball head was simulated in the form of an ideal sphere, with a diameter of 36 mm.

The hemi-pelvis included an acetabulum, which was equipped with a bone cavity, which was adapted to the shells outer diameter to simulate a diametric press-fit of 1 mm. The modular acetabular cup was oriented at 45◦ of inclination and 15◦ of anteversion, relative to the pelvic plane.

Due to the complex geometry, the hemi-pelvis was meshed using tetrahedral elements (141,877 elements), while the shell, liner, and ball head were meshed with 5987, 6162, and 14,400 hexahedral elements, respectively. To avoid excessive computational time, the components had to be meshed using linear elements. The element length at the contact area of the hemi-pelvis was set to 1.0 mm and the global element length of the acetabular shell, liner and head were set to 1.5 mm, 1.5 mm and 1.25 mm, respectively. A convergence analysis was carried out previously in order to determine the sensitivity of the results in relation to the mesh density, the element length of each component was varied between 1 mm and 4 mm. To exclude the influence of contact definitions on the results of the convergence analysis, each component was evaluated separately by means of deformation analyses. Thereby, the individual components were deformed by a defined amount and the deforming force was evaluated. The deforming forces determined with the chosen mesh densities changed less than 6% compared to the finest mesh.

The shell material was varied between a titanium alloy and two different kinds of CFR-PEEK (CFR-PEEK-15 and CFR-PEEK-23), while the liner material was varied between alumina toughened zirconia (ATZ) ceramic and UHMWPE. All the chosen materials were defined to be homogenous and isotropic (Table 1). As the yield strengths of the polymers were not exceeded, all the materials were simulated to be linear elastic.

**Table 1.** Overview of the Young's modulus and Poisson's ratio defined for the various simulated materials.


The numerically simulated hemi-pelvis was reconstructed from computed tomography (CT) data of a human cadaveric hemi-pelvis (Ethics Committee of the University of Rostock; Reg. No. A 2009 38), using an algorithm introduced by Kluess et al. [23]. The local X-ray attenuation from the CT resulted in a distribution of Hounsfield units (HU) in the CT slices, which directly correlate with bone density and were therefore mapped onto the FE mesh using a previously described approach [24]. For this, the HU of the CT dataset were treated as temperatures of a temperature-dependent material model assigned to the FE nodes. This results in a realistic, heterogeneous distribution of material characteristics throughout the bone geometry. To assign a corresponding Young's modulus to the HU, the HU from the CT were correlated with the local apparent densities, and the apparent densities were correlated with the Young's modulus. Due to the fact that the existing CT data provided no scanned bone mineral density phantom, an apparent density of 1.8 g/cm<sup>3</sup> was assigned to the maximal HU value, as per Taddei et al. [25]. The HU values of the cancellous bone were averaged, and the averaged value was assigned with an apparent density of 0.425 g/cm3, which represents a reasonable value for cancellous pelvic bone [26]. A linear correlation between the HU and apparent density was assumed using the following equation:

$$\text{op}\_{\text{app}}\text{ (g/cm}^3) = 0.0007918^\ast \text{HU} + 0.4718988 \tag{1}$$

To calculate the Young's modulus from the apparent density, the following equation based on the equation of Carter and Hayes was used [27]:

$$\text{EE (GPa)} = 3.79 \text{e}^{0.06} \text{p}\_{\text{app}}^{\text{-3}} \tag{2}$$

Forty equally spaced reference points were selected between the maximum and minimum HU, at which the Young's modulus was calculated and assigned. Between the reference points, a linear interpolation was used. Moreover, a lower limit of 500 MPa and an upper limit of 20,000 MPa were defined as thresholds to adequately represent the stiffness limits of the cancellous and cortical bones [28]. The mapping of the HU values and correlation to local stiffness was achieved with the self-developed software script AbaCTMat, based on Python 2.6.2 and implemented as plug-in in ABAQUS/CAE.

The model included three sliding contact formulations, which were described by normal and tangential contact. A penalty friction model with Coulomb friction formulation was chosen. The first contact was defined between the outer surface of the shell and the hemi-pelvis. The friction coefficient was set to 0.6, which represents a reasonable value for contacts between bones and porous coatings [29]. The second contact was defined between the shell and the liner, where a friction coefficient of 0.16

was applied [30]. The third contact was defined between the liner and the ball head, with a friction coefficient of 0.05 [13].

The simulation was executed in subsequent steps. Initially, the constraints for the hemi-pelvis were applied at the sacroiliac joint and the pubic symphysis, whereby the translational degrees of freedom at the sacroiliac joint were completely fixed (BC-1), while the motion in the sagittal plane was enabled at the pubic symphysis (BC-2) [13] (Figure 1). In the first step, the acetabular shell was pushed under displacement-control into the cavity, until a predefined shell overhang was reached using a kinematic coupling definition (Figure 1b). In the following step no forces or displacements were applied, to enable an elastic relaxation of the shell and cavity. Thereafter, the liner was also moved in a displacement-controlled manner, until a first contact between the shell rim and the liner was achieved (Figure 1c), followed by a force-controlled insertion with 500 N using a kinematic coupling definition again (Figure 1d). The coupling was applied to the distal inner surface of the liner, to not bias the radial deformation. Subsequently, the femoral ball head was inserted in a displacement-controlled manner into the liner to initiate contact (Figure 1e), followed by the application of a realistic hip joint load (Figure 1f). A load during normal walking (resulting load: 784.8 N) was chosen in accordance with Bergmann et al. [31] and applied via a reference point to the ball head using a kinematic coupling, again. The load was divided into the different load compounds in the x-, y-, and z-directions (Fx = 502.3 N, Fy = −78.5 N and Fz = 2048.3 N).

**Figure 1.** Depiction of the assembly (**a**) of the 3D finite element model consisting of a hemi-pelvis with a prepared bone cavity (1), acetabular shell (2), liner (3) and femoral ball head (4). The simulation was executed in several steps, consisting of the displacement-controlled insertion of the shell (**b**), displacement-controlled motion of the liner until the first contact to the shell (**c**), force-controlled insertion of the liner (**d**), displacement-controlled insertion of the femoral ball head (**e**) and the final loading, by applying a realistic hip joint load in x-, y- and z-direction (Fx, Fy and Fz) via the femoral ball head (**f**). During all simulation steps, the translational degrees of freedom at the sacroiliac joint were completely fixed (BC-1) and the pubic symphysis was fixed in the y-direction (BC-2).

The numerical simulations were performed as non-linear static calculations with incremental loading conditions. The FEA was evaluated in terms of the maximum radial shell and liner deformation after insertion and subsequent relaxation. Therefore, polar coordinate systems were created along the rotational axis of the shell and liner, respectively, and circumferential node paths along the inner rim of the acetabular shell (62 nodes) and liner (47 nodes) were defined and evaluated.

Moreover, the strain distribution over the full surface of the hemi-pelvis was investigated at a total of 29,923 node points (Figure 2c). It should be noted that only the magnitudes of the strains were considered and therefore no distinction was made between the compression and tensile strains.

**Figure 2.** Depiction of the node paths at the inner rim of the shell (**a**) and liner (**b**), consisting of 62 and 47 nodes, to analyze the radial deformation in polar coordinate systems. Moreover, the analyzed node region in terms of strain distribution at the surface of the hemi-pelvis is depicted (**c**).

ǯ

The determined logarithmic strains were divided into different strain regions, dependent on the bone response, based on studies by Biewener [32] and Frost [33], as summarized in Table 2.

**Table 2.** Values set for bone thresholds and response to particular strain rates [32,33].


#### **3. Results**

#### *3.1. Shell and Liner Deformation*

After the insertion into the hemi-pelvis and subsequent relaxation of the shells, the shell made of CFR-PEEK-15 showed the highest deformation (266.7 μm), followed by the shell made of CFR-PEEK-23 (136.5 μm) and the shell made of titanium, which exhibited the smallest maximum deformation (54.0 μm). Therefore, the deformation of shells made of CFR-PEEK-15 was approx. two and five times the deformation of shells made of CFR-PEEK-23 and titanium, respectively, while the deformation of a shell made of CFR-PEEK-23 was approx. 2.5 times higher than the deformation of the titanium shell. The shells made of CFR-PEEK were compressed in the region of the Os ilium and Os ischium and expanded in the direction of the acetabular notch (Figure 3a).

**Figure 3.** Exaggerated radial deformation (mm) at the inner rim of the shell (**a**) and liners made of ATZ ceramic (**b**) and UHMWPE (**c**) dependent on the chosen shell material and the orientation of the hemi-pelvis (**d**). The deformations are depicted in a polar coordinate system and in relation to the initial radii (Ri).

The deformation of the shells influenced the radial deformation of the liners, with the largest liner deformation resulting in combination with the shell made of CFR-PEEK-15, followed by CFR-PEEK-23. The UHMWPE liners displayed greater deformation compared to ceramic liners. The maximum radial liner deformations have been summarized in Table 3.



Like the shells, the liners were compressed in the region of the Os ilium and Os ischium and expanded in the direction of the acetabular notch (Figure 3b,c).

#### *3.2. Strain Distribution*

The resulting strains in the hemi-pelvis were influenced by the shell material (Table 4). The strongest influence could be seen for strains beneath 3000 μm/m. The percentage of nodes in the region of expected bone loss (<400 μm/m) was higher in shells made of CFR-PEEK, while the percentage of nodes in a region of bone preservation and building (400–3000 μm/m) was lower. On the other hand, the percentage of nodes with strains above 20,000 μm/m was higher in shells made of titanium.

**Table 4.** Comparison of the percentage of analyzed nodes, categorized into the different strain regions after shell relaxation and dependent on the shell material (Titanium, CFR-PEEK-23 or CFR-PEEK-15).


On the macroscopic scale, there were almost no differences of the visible strain distribution (Figure 4). The strongest alteration of the strain was seen at the acetabular fossa, where the strain increased when CFR-PEEK was used instead of titanium. Minor deviations were seen at the Os ischium and Os pubis.

The liner material had a small influence on the strains in the hemi-pelvis after insertion and subsequent relaxation, with more nodes with a strain beneath 400 μm/m and less nodes between 400 μm/m and 20,000 μm/m, when UHMWPE was used (Table 5). Due to the joint load, the overall strains increased and the influence of the liner material was compensated (Table 5 and Figure 5). The percentage of nodes with strains beneath 400 μm/m was reduced up to 73.0%, due to the hip load, while the percentage of nodes with strains in the range from 400 μm/m to 3000 μm/m and from 3000 μm to 20,000 μm/m increased up to 21.0% and 19.8%, respectively.

After the insertion of the liner, no relevant differences in strain distribution in the dependency of the shell or liner material were visible on the macroscopic scale. Only small areas near the acetabular rim showed altered strain distribution due to the shell material (Figure 5a,b). After loading with a hip joint load, the strains in the acetabulum and the wing of ilium increased strongly (Figure 5c,d). A small effect of the shell material was seen in the same regions as in the unloaded condition. In Figure 5, only the shells made of titanium and CFR-PEEK-15 with a liner made of UHMWPE were compared as an example. The results of the other simulations showed similar strain patterns.

**Figure 4.** Strain distribution in the adjacent bone of the hemi-pelvis after relaxation of the acetabular shell, dependent on the chosen shell material (**a**) titanium, (**b**) CFR-PEEK-23 and (**c**) CFR-PEEK-15. Differences visible on the macroscopic scale are highlighted in red.


**Table 5.** Comparison of the percentage of analyzed nodes, categorized into the different strain regions after liner insertion and after loading in dependency of the acetabular shell and liner material.

**Figure 5.** Comparison of strain distribution in the adjacent bone of the hemi-pelvis after the UHMWPE liner insertion into the titanium (**a**) and CFR-PEEK-15 (**b**) shell, followed by a relaxation. Additional the alteration of strain distribution after the application of a hip joint load in dependency of the shell material (**c**) titanium, (**d**) CFR-PEEK-15.

#### **4. Discussion**

In total hip arthroplasty, the stiffness of modular acetabular cups should be similiar to the adjacent bone, to avoid stress and strain shielding, that could otherwise lead to subsequent atrophy in the adjacent bone [5,34]. Compared to titanium, polymer materials are characterized by Young's modulus closer to bone and, therefore, might alter the stress and strain distribution to a lesser degree. Particularly, reinforced polymers, such as CFR-PEEK, might be a suitable alternative for acetabular cups, as their material properties can be altered to fit the bones' properties by varying the proportion of carbon fibers. However, the reduced stiffness compared to titanium shells might lead to increased acetabular shell and liner deformation, that can subsequently lead to a reduced clearance, possibly resulting in higher frictional torques and wear rates [13–16,19,20].

CFR-PEEK has previously been tested in numerical, experimental and in vivo studies as a material for monolithic cups and liners, but barely as a shell material [7–11,35,36]. Therefore, the purpose of our current computational study was to evaluate whether the stiffness of two different CFR-PEEK materials is sufficient to avoid excessive shell and liner deformation when they are used as shell materials for modular acetabular cups and whether these shells may reduce the strains in the adjacent bone.

Our study is restricted by certain limitations. First of all, only one hemi-pelvis was used in the current study and, thus, anatomical differences and differences in terms of age-related bone properties could not be taken into account. In particular, the bone density might be crucial in terms of shell deformation, even though previous studies concluded that altered bone properties do not influence the stress distribution greatly [36]. Moreover, the hemi-pelvis was CT-scanned without an additional bone mineral density phantom. Therefore, the equation to correlate the HU and apparent density had to be set manually, using additional data from the literature [25,26]. To calculate the Young's modulus from the apparent density, an equation based on an equation by Carter and Hayes [27], which was originally set up to calculate the compressive modulus depending on the apparent density and strain rate, was used as this equation can be used for cancellous as well as cortical bones. The numerically simulated hemi-pelvis showed unreasonably small HU in a region in the Os ilium; therefore, a minimum Young's modulus of 500 MPa was established, to avoid areas with excessively low modulus. The maximum Young's modulus of the cortical bone was set to 20,000 MPa, which is high compared to other studies, in which the maximal Young's modulus for cortical bones was set to between 8500 MPa and 18,000 MPa [20,23,36,37]. This is applicable, because the highest HU was defined as the point of maximum stiffness, but this point still includes naturally occurring deviations in x-ray attenuation which, however, are compensated over several HU. This still may lead to a slight overestimation of the bone stiffness, but is favored in order to consider a disadvantageous scenario in terms of acetabular shell and liner deformations. Moreover, these deformations were not experimentally validated in the current study, as the simulated hemi-pelvis was not accessible for experimental analysis. It is assumed that a qualitative comparison between the shells is legitimate, as the evaluated feature for shell and liner deformation, as well as strain distribution, was the material definition of the shell. Additionally, the material behavior of the CFR-PEEK was simplified. The CFR-PEEK was simulated as elastic with isotropic behavior, despite CFR-PEEK being an anisotropic material with its mechanical properties depending on the carbon fiber orientation. By adjusting the fiber orientation in the direction of strongly loaded areas, the shell could be strengthened, possibly leading to reduced shell deformations in these areas. Hence, an advanced material definition should be considered for future investigations. Moreover, it is known that implants made of CFR-PEEK have to be coated to enable osseointegration in vivo [38–40]. Even so, it was previously shown that the coating procedure can influence the mechanical properties (e.g., yield strength) of PEEK materials [41], the material properties of uncoated CFR-PEEK were used in the current study in terms of simplification.

Despite these limitations, the current study was suitable for providing an insight into whether CFR-PEEK can be used as an alternative shell material in the context of acetabular shell and liner deformation, as well as strain distribution in the adjacent bone. In terms of shell deformation, a correlation between the shell material stiffness and deformation value was determined, as shown previously [12]. The determined maximum radial deformation of the titanium shell after insertion into the bone cavity was 54.0 μm. When the shells made of CFR-PEEK-23 or CFR-PEEK-15 with a reduced stiffness were simulated, the radial shell deformation increased by about 2.5 times and 5 times, respectively. In a previous study, the same shell was numerically inserted into bone substitute cavities, with a diametric press-fit of 2 mm. The deformation of the titanium shell in the present study is about

three times, and the deformation of the shell made of CFR-PEEK-23 is still about 1.5 times higher compared to the data of the previous study, even so the diametric press-fit was only 1 mm, indicating that the simulated hemi-pelvis is stiffer compared to the bone substitute material used before [12]. In another study, Everitt et al. inserted monolithic cups made of CFR-PEEK with a Young's modulus of 15 GPa into cavities made of polyurethane foams (20 pcf or 30 pcf), with a diametric press-fit of 1 mm or 2 mm [42]. The monolithic cup was deformed by about 250 μm when inserted into a 1 mm under-reamed cavity made of 20 pcf polyurethane foams, which is in good agreement with the present deformation in the case of the shell made of CFR-PEEK-15. When the monolithic cups were inserted in 2 mm under-reamed cavities or cavities made of 30 pcf polyurethane foams, the deformation doubled [42]. In case of modular cups, an increased shell deformation results in greater liner deformations and stresses, harboring the risk of reduced clearance, increased frictional torques, incomplete seating and misalignment, as well as higher wear rates at the articulating surfaces [13–16]. Therefore, a diametric press-fit above 1 mm or the insertion of acetabular shells made of CFR-PEEK in young patients with high bone mineral density could be critical in terms of deformation, and the applicability of CFR-PEEK shells has to be cautiously questioned in these cases. In further investigations, the effect of an increased press-fit and increased bone mineral density on the implant deformation should be considered. On the other hand, one advantage of CFR-PEEK is the ability to adjust the material properties for a specific application. For example, the orientation of the carbon fibers could be altered to create an anisotropic material that can be used for acetabular shells. It would be conceivable to radially reinforce the shell in order to avoid strong deformations, while remaining flexible in the direction of the load application. Another opportunity to alter the mechanical properties of PEEK is by adding various additives (like TiO2) to the PEEK powder before molding, which can increase the bioactivity and therefore osseointegration at the same time [43].

The shells were mainly compressed in the contact area at the Os ilium and Os ischium, and expanded in the direction of the Os pubis and acetabular notch. Therefore, the shell deformation was consistent with the distribution of the bone density, as the deformation was initiated from the stiffest region. This is also in line with previous studies, in which similar deformation patterns were found [13,20,44]. In the present study, only the initial deformation of the shells was determined, even though creep deformation over time is a well-known problem of polymers. The creep deformation of shells made of polymers might lead to liner migration and, subsequently, to altered liner deformations and stresses. In worst-case scenarios, this could lead to increased rates of ceramic liner fractures, as reported for sandwich liners [45–51]. Therefore, the creep deformation of acetabular shells made of CFR-PEEK should be considered in future long-term experimental investigations.

The radial liner deformation in the present study was in a range between 149.0 μm (titanium) and 184.7 μm (CFR-PEEK-23) in case of the UHMWPE liner, and 1.8 μm (CFR-PEEK-23) and 4.4 mm (CFR-PEEK-15) in case of the ceramic liner after the load-controlled insertion and subsequent relaxation. Therefore, a linear correlation between shell and liner deformation was not shown, although previous studies have shown that increased shell deformations lead to the increased deformation of the inserted liners [12,13,46]. The non-uniform liner deformations might occur due to other factors, like the liner's initial seating depth after load-controlled insertion, and a varying bounce-back after the liner relief.

Compared to the above mentioned study [12] in which the same acetabular cups were numerically inserted into bone substitute cavities, the deformation of the UHMWPE liner was deformed six times more in case of the titanium shell and 1.5 times in case of the CFR-PEEK-23 shell during the load-controlled liner insertion. After subsequent relaxation, the liner deformation decreased, but was still three times and 1.5 times higher, respectively. However, when the determined liner deformations are compared to typical radial clearance of 300 μm for hard-on-soft bearings [52], the articulation between a femoral ball head and an UHMWPE liner would not be hampered in combination with the simulated shell materials. The deformations of the ceramic liner were also influenced by the chosen shell material, but were less critical regarding the clearance. The maximum radial deformation of the ceramic liner ranged from 1.8–4.4 μm and, thus, was approximately 20% of a typical radial

clearance for ceramic-on-ceramic bearings of 20 μm [53]. The maximum determined peak tensile stress of the ceramic liner (129.9 MPa), arising during load-controlled liner insertion, was nowhere near a critical value in terms of ceramic fracture as well [54]. However, under unfavorable conditions, such as an increased press-fit, a decreased wall thickness of the shell, a high bone mineral density, or increased loads, the radial deformation of the shell would increase, leading to increased liner deformation, increased stresses and a further decrease of the clearance, especially in the case of the UHMWPE liners [12–14,17,18,42,44,55–57]. Decreased clearances are known to increase the wear rates [15] and frictional torques at the articulating surfaces and therefore, might hamper the initial fixation of acetabular cups [16].

The magnitudes of the strains in the hemi-pelvis were slightly influenced by the stiffness of the shell material. With the decreasing stiffness of the shells, the proportion of nodes with a strain of less than 400 μm/m increased, indicating that more nodes are located in a bone resorbing range in the case of shells made of CFR-PEEK as compared to titanium, with the most affected region detected at the acetabular fossa. At the regions where the shell is in direct contact with the bone, the highest strains occurred (overloading), indicating a possible risk of bone fracture during shell insertion. In these regions, the strains were the lowest when the acetabular shell made of CFR-PEEK-15 was simulated. However, it is assumed that the determined strains would be reduced in vivo over time, due to the viscoelastic response of the adjacent bone stock, allowing the re-expansion of the shells [14,56].

Overall, only small effects of the shell material on the strains were determined, supporting the conclusions of other studies that a simple change in material stiffness is not sufficient to avoid the loss of acetabular bone [36,58]. The chosen liner material had nearly no impact on the strain distribution under loading, which is also in line with the findings of other studies [6,59,60].

The strain distribution pattern at the surface of the bone cavity was similar to previous studies [36,59]. The peak strains occurred at the rim of the reamed bone cavity, and there was nearly no strain at the acetabular fossa, as shown previously [36,59,61]. In our present FE study, only the magnitudes were evaluated. We found that the strains in the acetabulum were mainly tensile strains, except for the surfaces in direct contact to the shell, where compressive strains arose.

In the present study, only three loading states were analyzed: either directly after the shell and liner insertion with subsequent relaxation (no external loading), or at the peak load of a normal gait cycle [31,62]. We showed that nodes in the bone sustaining or building region of 400–3000 μm/m increased up to 21.0% during the loading, compared to the unloaded cup. Therefore, the loading condition was crucial in terms of strain distribution, as expected, and more loading conditions, such as stair-climbing, sitting down and standing up, or stumbling, with higher loads and altered loading directions should be considered in further investigations [63].

#### **5. Conclusions**

In conclusion, the deformation of acetabular shells made of CFR-PEEK is more pronounced as compared to titanium shells during intraoperative insertion, resulting in the increased radial deformation of liners. However, liner deformations were not critical in terms of the radial clearance in our present computational study. Particularly, the combination of shells made of CFR-PEEK and liners made of ATZ ceramic showed only minor differences in implant deformations, as compared to titanium shells. In terms of strain distribution, only small effects were observed to be dependent on the shell and liner material, with a greater area of bone being in a bone-sustaining or building region when CFR-PEEK was used. Therefore, clear advantages in terms of strain distribution were not seen in our computational study. Acetabular shells made of CFR-PEEK are not likely to replace titanium shells in case of standard THA. However, as the strain distribution was not impaired and the shell and liner deformations were not critical, shells made of CFR-PEEK might be a possible alternative in special cases. However, in our present study only shell and liner deformations as well as strain distribution in the adjacent bone were considered, but even so, more parameters have to be taken into account before a final statement can be made about the suitability of acetabular shells made of CFR-PEEK. Therefore, further experimental investigations are needed, e.g., to determine the creep and wear properties of acetabular shells made of CFR-PEEK.

**Author Contributions:** Conceptualization: D.V. and R.B.; methodology: D.V., M.K., M.S. and R.B.; software: D.V., M.K. and M.S.; data curation: D.V. and M.K.; formal analysis: D.V. and M.K.; investigation: D.V. and M.K.; resources: R.B.; writing—original draft preparation: D.V.; writing—review and editing: D.V., M.K., M.S. and R.B.; visualization: D.V. and M.K.; supervision: D.V. and R.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Third Body Wear of UHMWPE-on-PEEK-OPTIMA**™

#### **Raelene M. Cowie <sup>1</sup> , Naveen Manikya Pallem 1, Adam Briscoe <sup>2</sup> and Louise M. Jennings 1,\***


Received: 18 February 2020; Accepted: 9 March 2020; Published: 11 March 2020

**Abstract:** PEEK-OPTIMA™ is being considered as an alternative to cobalt chrome (CoCr) in the femoral component of total knee replacements. To date, investigations of ultra-high molecular weight polyethylene (UHMWPE)-on-PEEK have shown an equivalent wear rate to conventional implant materials under standard conditions. In this study, the third body wear performance of UHMWPE-on-PEEK was directly compared to UHMWPE-on-CoCr in a series of pin-on-plate studies using two approaches for third body damage. Damage simulation with particles of bone cement showed a significant (*p* < 0.001), four-fold increase in the mean surface roughness of PEEK plates compared to CoCr. However, wear simulation against the damaged plates showed no significant difference in the wear of UHMWPE pins against the different materials (*p* = 0.59), and a polishing effect by the pin against the PEEK plates was observed. Scratching PEEK and CoCr counterfaces with a diamond stylus to create scratches representative of severe third body damage (4 μm lip height) resulted in a significantly higher (*p* = 0.01) wear of UHMWPE against CoCr compared to PEEK and again, against PEEK plates, polishing by the UHMWPE pin led to a reduction in scratch lip height. This study shows that in terms of its wear performance under third body wear/damage conditions, UHMWPE-on-PEEK differs from conventional knee replacement materials.

**Keywords:** biomaterials; arthroplasty; orthopaedic tribology; experimental simulation; total knee replacement; PEEK-OPTIMA™; UHMWPE; third body wear

#### **1. Introduction**

Total knee replacement (TKR) is a highly successful procedure with >90% survival at 10 years [1]; however, up to 20% of patients may be dissatisfied with their procedure [2]. When considering how to improve patient satisfaction, the materials from which the implant is manufactured is one potential variable to be considered. PEEK-OPTIMA™ (poly-ether-ether-ketone) has recently been investigated as an alternative to cobalt chrome (CoCr) for use in the femoral component of TKR [3–5]. There are a number of potential advantages of using PEEK in this application. For example, the Young's modulus of the PEEK femoral component (~3.7 GPa) is more similar to that of bone (0.001–20 GPa) than a CoCr femoral component (~210 GPa), giving the potential for a reduction in stress shielding, which can lead to bone resorption particularly behind the anterior flange of the implant and, hence, increase the risk of implant loosening [6]. In addition, an all-polymer implant would be lighter weight than conventional materials and more similar to the weight of the natural joint.

Several in vitro pin-on-plate and whole joint simulation studies have been carried out investigating the wear performance of the PEEK-on-ultra-high molecular weight polyethylene (UHMWPE) bearing couple and have demonstrated an equivalent rate of wear for this bearing couple compared to CoCr-on-UHMWPE for a well-positioned implant tested in a clean environment under standard conditions [3,7,8]. In addition, an in vivo large animal study has also investigated the potential to use PEEK in an all-polymer TKR [9,10]. Despite this animal study being relatively short-term, no problems were encountered with fracture or fixation of the device; there was, however, an inflammatory response but, due to the lack of a control group (metal-on-polyethylene implant), it is not known whether a similar response would occur if conventional TKR materials were implanted in this animal model. Having obtained promising initial results, prior to the clinical adoption of the device, it is important to consider how the implant will respond under a wider envelope of clinically relevant conditions [11]. In this study, the influence of third body wear on UHMWPE-on-PEEK-OPTIMA™ in a simple geometry pin-on-plate configuration has been considered.

Third body wear can occur when hard particles such as bone cement particles, bone fragments or other particulate materials become trapped between the articulating surfaces of an implant causing damage to the joint surfaces and accelerating implant wear. Two approaches have been taken for the simulation of third body wear of arthroplasty bearing materials. One approach has been to introduce clinically relevant particles such as polymethyl methacrylate (PMMA) cement into the articulating interface [12]; an alternative approach has been to inflict third body damage directly to the surface(s) either by scratching the surface of the implant directly using a diamond stylus, which gives close control of the position and geometry of the scratches [13], or by abrading the component bearing surface producing a random orientation of scratches [14]. There are advantages and limitations to these approaches. Protocols using particles to replicate third body damage can more closely represent the clinical situation; however, controlling the particles trapped between the articulating surfaces can be difficult especially in the knee where the low conforming nature of the articulating surfaces may lead to particles being ejected from the joint. In bearing couples containing polymeric implant materials, if particles become embedded in the polymer, assessing wear of the polymer gravimetrically can be unreliable. Scratching the surface of the implant directly can give more consistent and reproducible damage. Controlling the lip height of the scratches, which is the variable that most influences polyethylene wear in metal-on-UHMWPE configurations, using a method such as a diamond stylus allows the scratch geometry to be optimised to more closely replicate observations of retrieved implants [15]. However, this approach could be considered a less clinically relevant method for simulating third body damage than using particles.

Whilst carrying out whole joint wear simulation studies is optimal, valuable information can be gained from pin-on-plate studies. By simplifying component geometries and applied loads and motions, the influence of individual variables can be investigated [15]. The aim of this study was to investigate the influence of third body damage on the wear of UHMWPE-on-PEEK-OPTIMA™ in simple geometry pin-on-plate wear simulation. Third body damage was simulated using two approaches: (1) using particles of PMMA cement and (2) scratching the implants directly using a diamond stylus. When considered together, the outcomes of these two approaches can give a better understanding both of how third body particles can damage the articulating surfaces, and how different magnitudes of damage can influence wear. For all the studies, the wear of UHMWPE-on-PEEK was directly compared to conventional knee implant materials, UHMWPE-on-CoCr, which were tested in parallel. It was hypothesised that, because of the different material properties of PEEK compared to CoCr, the third body wear behaviour of the two material combinations would differ.

#### **2. Materials and Methods**

The studies were split into two phases, first carrying out damage simulation (using each method) before determining the wear factors of UHMWPE against the damaged surfaces. This two-phase approach has been adopted in previous third body wear simulation studies [15,16].

#### *2.1. Materials*

The pins used were GUR 1020 UHMWPE (conventional, non-sterile) machined into a truncated cone with either 3 mm or 8 mm flat contact face for damage or wear simulation, respectively. The plates were either injection moulded, implant grade, unfilled (natural) PEEK-OPTIMA™ (Invibio Ltd, UK), initial surface roughness (Ra) ~0.02 μm or CoCr, polished to an initial Ra < 0.01 μm. To create damage with particles, Palacos R + G PMMA cement (Heraeus, Germany) was mixed and cured in a block as per the manufacturers' instructions before turning and crushing with a mortar and pestle to create particles, which were sieved within a size range of 500–1000 μm diameter. The particle size range used was within a clinically relevant range [17].

#### *2.2. Methods*

#### 2.2.1. Damage Simulation: Third Body Damage with PMMA Cement Particles

The protocol used for damage simulation with particles was adapted from previously described studies [15,16]. In brief, the PEEK or CoCr plate was clamped onto a sliding table mounted on the platen of an Instron materials testing machine (Instron, MA, USA). The PMMA cement particles described above were trapped (in excess) between a UHMWPE pin (3 mm diameter contact face) and the plate; before a load of 120 N was applied axially through the pin. Then, using the Instron materials testing machine, the plate was pulled beneath the pin for 15 mm at a speed of 8 mm/s to create third body damage. Five regions of damage were created on each plate with a spacing of 3 mm, the particles were passed over the plate 5 times in each region of damage. Third body damage was created perpendicular to the direction of the subsequent wear test (Figure 1a).

**Figure 1.** (**a**) Schematic showing the respective directions of the damage simulation and wear simulation; (**b**) schematic showing the profile of a 1 μm scratch created with a diamond stylus and the lip height measurement taken.

#### 2.2.2. Damage Simulation: Third Body Damage Created Using a Diamond Stylus

To create a scratch in the plates with a reproducible geometry, the plate was set up on a sliding table as described for the particle method, a stylus with a 200 μm radius diamond tip was axially loaded and the plate pulled beneath the stylus to create scratches. On each plate, 5 scratches were created running perpendicular to the direction of the wear study (Figure 1a). The lip height of the scratches was adjusted to either 1, 2 or 4 μm by changing the load applied to the stylus (Table A1 in Appendix A) [18].

#### 2.2.3. Pin-on-Plate Wear Simulation

Wear simulation was carried out using a six-station multi-axial pin-on-plate reciprocating rig and aimed to replicate the average contact pressure and cross-shear ratio in a total knee replacement [19]. The plate was mounted in a lubricant-containing bath, which reciprocated over a length of 20 mm at 1Hz, and as the bath reciprocated, the UHMWPE pin rotated (±20◦) via a rack and pinion mechanism. a constant axial load of 160 N was applied through the pin (8 mm flat contact face)

to give an average contact pressure of 3.18 MPa consistent with previous wear simulation of the UHMWPE-on-PEEK bearing couple [8]. All wear simulation studies were carried out using 25% bovine serum supplemented with 0.03% sodium azide solution as a lubricant and were carried out under rig running (room) temperature conditions as previously described and optimised for the UHMWPE-on-PEEK bearing couple [8]. Prior to the start of the study, the UHMWPE pins were soaked for a minimum of 2 weeks to maximise water uptake prior to cleaning and weighing using an XP26 digital microbalance (Mettler Toledo Inc., OH, USA) with a resolution of 1 μg. Measurements of each pin were taken until 5 consecutive measurements were within ± 5 μg and an average of these 5 measurements taken. Studies were carried out for 1 million cycles (MC) with gravimetric analysis of the UHMWPE pins every 0.3 MC. The weight loss of the pins was converted to a volume loss (V) using two unloaded soak controls to compensate for the uptake of moisture by the polyethylene and a density of 0.934 mg/mm3 for GUR 1020 UHMWPE. The wear factor (k) was calculated using the sliding distance for the test (X) and the applied load (P) as shown in Equation (1).

$$\mathbf{k} = \mathbf{V}/\mathbf{P}\mathbf{X}\_{\prime} \tag{1}$$

For both methods of damage simulation, the wear of the UHMWPE pins against the plates which had undergone damage simulation was compared to that against undamaged, polished plates which served as a control. In the second approach (scratching using a diamond stylus), these plates are referred to as 0 μm lip height.

The surface topography of the plates was assessed pre- and post-test using a contacting Form Talysurf (Taylor Hobson, Leicester, UK) with a 2 μm conical tipped stylus. To analyse the surface roughness, a Gaussian filter was applied to the measurements and a 0.25 mm upper cut-off used as described in the ISO standard [20]. The surface roughness parameter of interest was the mean surface roughness (Ra). Measurements were taken both perpendicular to the direction of damage simulation (A) and perpendicular to the direction of the wear test (B) as shown in Figure 1a. The lip height (Figure 1b) of the scratches was also assessed by carrying out LS line form removal and primary analysis. For third body damaged plates, the density of the scratches within a given lip height range was calculated and expressed as the number of scratches per mm both after damage simulation and after the subsequent wear test. For this analysis, a threshold was applied whereby scratches with a lip height of <0.1 μm were not measured as they were considered too indistinct from the background topography of the plate to be reliably measured. For the plates scratched with a diamond stylus, the lip height of each scratch was measured at 5 points, scratches with a lip height less than 0.2 μm were not measured.

At the conclusion of the studies, images of the wear scars on the PEEK and CoCr plates were taken using an optical profiler, Alicona G5 IF (Graz, Austria) with 5× magnification.

A minimum of 3 sets of samples were used for each material configuration for each of the two approaches. The mean wear factor of the UHMWPE pins, the mean surface roughness (Ra) and the mean lip height of the scratches on the plates was determined with 95% confidence limits. Statistical analysis was carried out using a one-way ANOVA to compare the two configurations of UHMWPE-on-PEEK to UHMWPE-on-CoCr with significance taken at *p* < 0.05.

The data associated with this article is openly available from the University of Leeds Data Repository [18].

#### **3. Results**

#### *3.1. Third Body Damage with PMMA Cement Particles*

Having carried out third body damage simulation with PMMA cement particles, on the surface of all 3 PEEK plates, there was a high density of linear scratching visible in the region of damage; for CoCr, on one plate, scratching was visible, and on the other plates, there was little evidence of damage (Figure 2). Surface roughness measurements taken perpendicular to the direction of damage simulation

(A) showed a significant increase of more than four-fold (*p* < 0.001) in mean surface roughness of the PEEK plates compared to CoCr (Table 1). Inspection of the plates showed no evidence of PMMA cement particles becoming embedded in the surface of the PEEK.

**Figure 2.** Representative images of the CoCr and PEEK plates following damage simulation with PMMA cement particles and wear simulation taken using an Alicona G5 IF with 5× magnification. (**a**,**b**) CoCr plates. In (**b**), scratches caused by the PMMA cement particles are more clearly visible; in both images, isolated scratches from the wear simulation can be seen. (**c**) a PEEK plate. The polishing and scratching from the wear simulation is visible running perpendicular to the direction of damage simulation. Scale bars represent 2 mm, arrows denote directions of damage simulation and wear simulation.



In the control tests, after one MC wear simulation, the articulation of the pin against the polished plates caused discrete linear scratches on the CoCr plates and a higher density of scratches on the PEEK plates, resulting in a significantly higher post-test mean surface roughness of PEEK compared to CoCr (*p* = 0.03) (Table 1). Following wear simulation against the damaged plates, burnishing could be seen on the PEEK plates, where the scratches in the PEEK as a result of the damage simulation had been polished out leaving a defined wear scar (Figure 2c). However, within the wear scar, there was evidence of scratching from the articulation of the UHMWPE pin against the PEEK plate. On the CoCr plates, some discrete scratching was evident as a result of the wear simulation. At the conclusion of the study, there was no significant difference (*p* = 0.07) in the mean surface roughness of the PEEK and CoCr plates when measured either perpendicular to the damage simulation (A) or perpendicular to the direction of the wear test (B) (*p* = 0.09). However, despite there being no significant difference in the measurements, when measured perpendicular to the wear simulation, the mean surface roughness of the PEEK was much higher than that of CoCr (more than six-fold) with greater variability in the measurements.

In the control test, the mean wear factor with 95% confidence limits of the UHMWPE pins against smooth CoCr and PEEK plates was similar (*<sup>p</sup>* <sup>=</sup> 0.84) at 1.88 <sup>×</sup> <sup>10</sup>−<sup>7</sup> <sup>±</sup> 0.92 <sup>×</sup> <sup>10</sup>−<sup>7</sup> mm3/Nm and 1.97 <sup>×</sup> <sup>10</sup>−<sup>7</sup> <sup>±</sup> 1.52 <sup>×</sup> <sup>10</sup>−<sup>7</sup> mm3/Nm, respectively (Figure 3). Against third body damaged CoCr plates, the wear of the UHMWPE pins was 2.24 <sup>×</sup> <sup>10</sup>−<sup>7</sup> <sup>±</sup> 1.41 <sup>×</sup> <sup>10</sup>−<sup>7</sup> mm3/Nm; against third body damaged PEEK plates, wear was 1.93 <sup>×</sup> <sup>10</sup>−<sup>7</sup> <sup>±</sup> 1.82 <sup>×</sup> <sup>10</sup>−<sup>7</sup> mm3/Nm. There was no significant difference (*<sup>p</sup>* <sup>=</sup> 0.59) in the wear factors of the UHMWPE against the different material types and for all tests, the wear of the UHMWPE pins was linear.

**Figure 3.** Mean wear factor (mm3/Nm) of UHMWPE pins (<sup>±</sup> 95% confidence limits) against smooth (control) CoCr and PEEK-OPTIMA™ plates and plates which had undergone damage simulation with particles of PMMA cement (n = 3).

Analysis of the lip height of the scratches (Figure 4) showed that on the CoCr plates, lip heights were primarily in the 0.1–0.2 μm range both before and after damage simulation. For the PEEK plates, after damage simulation, most scratch lip heights were between 0.2 and 0.4 μm but some were as large as 0.8 μm, then after wear simulation, scratch lip heights were smaller and predominantly in the range of 0.1–0.2 μm.

**Figure 4.** Mean number of scratches per mm (± 95 % confidence limits) within given lip height ranges on cobalt chrome and PEEK plates. Measurements taken post damage simulation and post wear test, (n = 3).

#### *3.2. Third Body Damage Created Using a Diamond Stylus*

To adjust the lip height of the scratches created, the load applied to the diamond stylus was varied. To create scratches in the PEEK plates of the same magnitude as the CoCr, a lower load, approximately one-quarter of that required to create scratches in the CoCr was applied to the stylus (Table A1 in Appendix A) [18]. The magnitude of the lip heights created in the PEEK plates was more difficult to control than in the CoCr so variability in the lip heights was greater (Figure 5 and Table 2).

**Figure 5.** Mean wear factor (mm3/Nm) <sup>±</sup> 95% confidence limits of UHMWPE pins tested against PEEK-OPTIMA™ and cobalt chrome plates scratched with a diamond stylus as a function of pre-test (initial) lip height, μm (± 95% confidence limits).

**Table 2.** Mean lip height, μm (± 95% confidence limits) of scratches on PEEK-OPTIMA™ and cobalt chrome plates taken pre- and post- wear simulation (minimum N = 3). Statistical analysis to compare PEEK to CoCr, \* denotes *p* < 0.05.


After 1 MC wear simulation, discrete scratches were visible in the wear scars on the CoCr plates; whereas a polished region was evident and there was a visible reduction in the lip height of the scratches in the PEEK plates (Figure 6).

**Figure 6.** Representative images of CoCr (**a**) and PEEK (**b**) plates taken using an Alicona G5 IF with 5× magnification. The plates were scratched with a diamond stylus to create a scratch lip height of 4 μm, then 1 million cycles (MC) wear simulation was carried out perpendicular to the direction of the scratches. Scale bar represents 1 mm.

The mean wear factors of the UHMWPE pins against the scratched plates are shown in Figure 5. With no scratches on the plates (0 μm lip height), the wear of the UHMWPE against CoCr (3.4 <sup>×</sup> <sup>10</sup>−<sup>7</sup> <sup>±</sup> 8.2 <sup>×</sup> <sup>10</sup>−<sup>7</sup> mm3/Nm) and PEEK (3.9 <sup>×</sup> <sup>10</sup>−<sup>7</sup> <sup>±</sup> 5.3 <sup>×</sup> <sup>10</sup>−<sup>7</sup> mm3/Nm) was similar (*p* = 0.64); when scratches were created in the plates with a lip height of 4 μm, the wear of UHMWPE was significantly higher (*<sup>p</sup>* <sup>=</sup> 0.01) against CoCr than against PEEK at 9.7 <sup>×</sup> <sup>10</sup>−<sup>7</sup> <sup>±</sup> 4.3 <sup>×</sup> <sup>10</sup>−<sup>7</sup> mm3/Nm and 3.8 <sup>×</sup> <sup>10</sup>−<sup>7</sup> <sup>±</sup> 4.4 <sup>×</sup> <sup>10</sup>−<sup>7</sup> mm3/Nm, respectively, there were no significant differences (*<sup>p</sup>* <sup>&</sup>gt; 0.05) in wear factor for the 1 and 2 μm lip height conditions.

The lip heights of the scratches were assessed pre- and post-test. For the pre-test measurements, there was no significant difference (*p* < 0.05) between the lip height of the scratches in PEEK and CoCr for 1 or 4 μm. Post-test, the lip height of the scratches in the PEEK was significantly lower (*p* < 0.05) than in the CoCr for all scratch lip heights investigated due to the polishing effect of the pin against the plate (Table 2).

Surface roughness measurements taken perpendicular to the direction of wear simulation (B) between the scratches created with a diamond stylus are shown in Table 3. Pre-test, the mean surface roughness of the PEEK plates was significantly higher (*p* > 0.05) than the CoCr plates. After one MC wear simulation, there was a high density of linear scratching on the PEEK plates, and discrete scratches visible on the CoCr plates, resulting in a significantly higher (*p* < 0.05) mean surface roughness of the PEEK plates compared to CoCr.


**Table 3.** Mean surface roughness (Ra), μm (± 95% confidence limits) of PEEK-OPTIMA™ and cobalt chrome plates taken perpendicular to the direction of wear simulation (B) pre- and post- wear testing. Statistical analysis to compare PEEK to CoCr, \* denotes *p* < 0.05.

In both studies, the PEEK plates were weighed at each measurement point; however, despite extensive pre-test soaking (>90 days in sterile water [21]), due to inconsistencies in moisture uptake by the PEEK [22], it was not possible to assess wear and the data was considered unreliable.

#### **4. Discussion**

The aim of this study was to investigate the influence of third body damage on the wear of UHMWPE-on-PEEK in a simple geometrical configuration. Two approaches were used to simulate third body damage: using particles of PMMA cement to create damage and scratching the implant surfaces directly with a diamond stylus prior to carrying out multi-directional pin-on-plate wear simulation against the damaged counterfaces. The wear of UHMWPE-on-PEEK was compared to UHMWPE-on-CoCr in studies carried out in parallel. It was hypothesised that the third body wear performance of UHMWPE-on-PEEK would differ to that of UHMWPE-on-CoCr.

Simulating third body damage with particles of PMMA cement caused light scratching on one of the three CoCr plates, which highlights the difficulty in controlling the particles trapped between the pin and plate to create reproducible third body damage using a particle method. Following damage simulation, the mean surface roughness of the plates was twice that reported by Cowie et al. [16] where third body damage simulation was carried out using a similar method. This higher surface roughness may be attributed to differences in the composition of the cement used. In the previous study, the cement used contained 10% barium sulphate as a radiopacifier, whilst in this study, the cement contained 14.75% zirconium dioxide. Whilst both barium sulphate and zirconium dioxide are commonly used additives to improve the visibility of bone cement under x-ray, previous third body wear simulation studies have shown the harder zirconium dioxide agglomerates to cause more damage to metallic counterfaces than barium sulphate [23]. This may account for the higher surface roughness of the plates following damage simulation in this study compared to previous work. Simulating third body wear with cement particles against PEEK plates gave a significantly higher mean surface roughness of the PEEK plates, which, therefore, indicates a reduced scratch resistance of PEEK compared to CoCr. Previous wear simulation studies of scratched metallic counterfaces have shown that it is the lip height of scratches in the metal that influences the wear of UHMWPE and that, to increase polyethylene wear, the scratch lip height must exceed a critical value [15,24]. When compared to control tests carried out against smooth plates, the damage simulation protocol used in this study did not generate scratches with a lip height of sufficient magnitude to influence the wear of UHMWPE against either damaged PEEK or CoCr plates. The wear factor of the UHMWPE pins in the control study was similar to that previously reported under comparable conditions [8]. It is interesting to note that the variability in the data was consistently larger in the UHMWPE-on-PEEK bearing couple compared to UHMWPE-on-CoCr and that following third body damage simulation, the variability in the data was further increased for both material combinations. This higher variability suggests that larger sample sizes and further testing under a wider range of conditions may need to be considered to fully understand the wear performance of the bearing couple.

Scratching the counterfaces with a diamond stylus allowed the geometry of the scratches to be more closely controlled. Previous studies have shown the lip height of scratches on retrieved hip and knee implants to be in the range of 0.01–4.1 μm [25,26] and experimental studies have shown larger scratch lip heights, up to 4 μm, to increase the wear of both hip [27] and knee implants [28]. To create a scratch of the same lip height in PEEK, a lower axial load was applied to the diamond stylus against the PEEK plates compared to CoCr due to the lower hardness of the PEEK polymer (Table A1) [18]. For both materials, there was a linear relationship between the load applied to the stylus and the lip height of the scratch created. Greater control of the lip height could be achieved in CoCr plates, where the scratch lip heights also had a more consistent geometry.

Against polished plates, there was no significant difference in the wear of UHMWPE against PEEK or CoCr. Scratches of 1 or 2 μm lip height in CoCr did not increase the wear of UHMWPE compared to polished control plates; however, with lip heights of 4 μm, the wear of UHMWPE was higher. This exponential relationship between scratch lip height and wear has previously been reported. Minakawa et al. showed scratches with lip heights below 0.5 μm to have a wear factor similar to that of polished plates; with larger lip heights (1 μm), a dramatic increase in wear factor was seen [15], and Lancaster et al. showed a similar relationship between Ra and wear factor [24]. In this study, a larger lip height (4 μm) was necessary to significantly increase the wear rate of the polyethylene, perhaps due to the differences in the grade of polyethylene used. When investigating the influence of scratch lip height in PEEK on the wear of UHMWPE, a different trend was seen. In the PEEK plates, a polishing effect of the pin against the plate was seen and the wear factor of the UHMWPE did not increase with lip height. When a 4 μm lip height was created in PEEK plates, the wear factor of the UHMWPE was significantly lower (*p* < 0.05) than when a scratch of similar magnitude was created in CoCr. This difference in wear performance can be attributed to the different hardness of the PEEK and CoCr materials.

Independent control tests were carried out for each of the damage simulation methodologies used and the different wear factors obtained may be attributable to the studies being carried out using different experimental rigs, slight differences in the composition of the cobalt chrome, the polymers used being from different batches and/or the studies being carried out by different researchers. It is, however, a strength of the study that independent control tests were carried out for each method.

There were a number of limitations that should be considered in the interpretation of this study. Firstly, this was a fundamental study carried out in a pin-on-plate configuration using simplified geometries of components and whilst the input kinematics replicated the average cross-shear and contact pressure in a total knee replacement [29]. The loading and motion profiles were simplified and consistent with previous studies [8], and it is not known whether the results would be replicated in a clinical setting. The components used were not sterilised or cross-linked; however, sterilisation by ethylene oxide, which is commonly used for TKRs, has been shown not to influence mechanical properties or induce crosslinking in polyethylene and, therefore, does not influence wear [30]. The sample size was small. In some cases, three repeats were carried out; however, it was deemed that a sufficiently large number of samples were tested to show the polishing effect of the UHMWPE pin against the PEEK plates scratched either with particles or a diamond stylus. This polishing effect did not occur in the UHMWPE-on-CoCr studies. It is also unclear whether the scratching on the PEEK plates after damage simulation or the polishing of the scratched regions of the plate by UHMWPE pins would contribute to an increase in PEEK wear or wear debris and whether in vivo this would accelerate osteolysis, as the inconsistent moisture uptake of PEEK made gravimetric analysis unreliable [8,22]. It is also likely that the relationship between scratch lip height and wear would be different for other grades or compositions of UHMWPE, particularly when articulating against CoCr or with the introduction of different kinematic conditions or geometries of components. Finally, this was a short-term study. When investigating damage simulation with particles, the particles were passed over the plate five times in each region of damage; and in the wear simulation, studies were limited to 1 MC.

#### **5. Conclusions**

Simulating third body damage with PMMA particles did not influence mean surface roughness enough to increase the wear of UHMWPE against either CoCr or PEEK. On the PEEK plates, a polishing effect of the UHMWPE pin against the scratched plate was observed.

When scratches with a lip height of greater magnitude were created directly on PEEK and CoCr plates, an exponential increase in UHMWPE wear factor was observed with increasing lip height against CoCr plates; whereas against PEEK plates, increasing lip height did not have a similar influence on the wear factor of UHMWPE.

**Author Contributions:** Conceptualisation, R.M.C., A.B. and L.M.J.; data curation, R.M.C. and N.M.P.; formal analysis, R.M.C., N.M.P. and L.M.J.; funding acquisition, L.M.J.; investigation, R.M.C. and N.M.P.; methodology, R.M.C. and L.M.J.; project administration, L.M.J.; resources, A.B. and L.M.J.; supervision, L.M.J.; writing—original draft, R.M.C.; writing—review and editing, N.M.P., A.B. and L.M.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by Invibio Knees Ltd. and the Innovation and Knowledge Centre in Regenerative Therapies and Devices funded by the EPSRC, TSB and BBSRC under grant number EP/J017620/1. It was partially funded through WELMEC, a centre of Excellence in Medical Engineering funded by the Wellcome Trust and EPSRC under grant number WT 088908/Z/09/Z and supported by the EPSRC Centre for Innovative Manufacturing in Medical Devices under grant number EP/K029592/1.

**Acknowledgments:** The PEEK-OPTIMA™ plates were provided by Invibio Knees Ltd. Thanks to Phil Wood and his team for technical assistance.

**Conflicts of Interest:** A.B. is a paid employee of Invibio Ltd. and had a role in the study design and writing the manuscript; L.M.J. has received research funding support from Invibio Ltd.

#### **Appendix A**


**Table A1.** Load applied to a 200 μm radius diamond stylus to create scratches of a known geometry in PEEK and cobalt chrome plates.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **How to Proceed with Asymptomatic Modular Dual Taper Hip Stems in the Case of Acetabular Revision**

#### **Thomas M. Grupp 1,2,\*,**†**, Marc Baxmann 1,**†**, Volkmar Jansson 2, Henning Windhagen 3, Karl-Dieter Heller 4, Michael M. Morlock <sup>5</sup> and Hanns-Peter Knaebel <sup>6</sup>**


Received: 28 November 2019; Accepted: 18 February 2020; Published: 2 March 2020

**Abstract:** How to proceed with a clinically asymptomatic modular Metha®Ti alloy stem with dual taper CoCr neck adapter in case of acetabular revision? To systematically answer this question the status of research and appropriate diagnostic methods in context to clinically symptomatic and asymptomatic dual taper stem-neck couplings has been evaluated based on a systematic literature review. A retrieval analysis of thirteen Metha®modular dual taper CoCr/Ti alloy hip stems has been performed and a rational decision making model as basis for a clinical recommendation was developed. From our observations we propose that in cases of acetabular revision, that for patients with a serum cobalt level of > 4 μg/L and a Co/Cr ratio > 3.6, the revision of the modular dual taper stem may be considered. Prior to acetabular revision surgery a systematic diagnostic evaluation should be executed, using specific tests such as serum metal (Co, Cr) ion analysis, plain antero-posterior and lateral radiographs and cross-sectional imaging modalities (Metal Artefact Reduction Sequence Magnetic Resonance Imaging). For an asymptomatic Metha®dual taper Ti alloy/CoCr stem-neck coupling at the stage of acetabular revision careful clinical decision making according to the proposed model should be followed and overreliance on any single examination should be avoided, considering the complete individual differential diagnosis and patient situation.

**Keywords:** total hip arthroplasty; dual taper modular hip stem; acetabular revision; asymptomatic stem modularity; decision making model; threshold

#### **1. Introduction**

Failure of total hip arthroplasty (THA) is a relatively rare condition and the most common causes for revision are aseptic loosening, dislocation, septic loosening and peri-prosthetic fractures [1,2]. For a better adaptation to different diaphyseal and extra-diaphyseal anatomical conditions a modular dual taper neck design was clinically introduced for primary THA by Toni et al. [3] in 1995. Using an anatomic cementless stem design in combination with a modular rectangular tapered neck adapter in a consecutive series of 216 hip arthroplasties they reported a survival rate of 98.6% at 5 years including all implant related complications [3]. Clinical benefits of modular dual taper neck adapter hip stems

are the restoration of the hip centre of rotation in combination with an adequate soft tissue balancing in complex anatomical, muscular and ligamenteous patient situations [4–6]. For anatomic differences between specific patients (e.g., dysplasia), as well as different hip morphologies in males and females in regard to femur size, Caput-Collum-Diaphysis (CCD) angle, femoral offset and anteversion [7], these designs allow to adjust the native femoral anteversion and offset to restore an adequate abductor muscles lever arm [5], limb length [7] and patients' original biomechanics [8], independently of the stem size and femoral fixation [6,9]. In an exploratory study on 684 mono-bloc and 125 THAs with a modular neck in patients dedicated to a rapid recovery programme based on selected cases with primary arthritis in an otherwise anatomically normal hip joint, Gerhardt et al. [10] did not find a clear benefit in restoring hip geometry and dislocation rate—only the abductor moment arm which is closely correlated to the femoral offset was better reconstructed in the modular neck adapter cohort. For this "straightforward" THA the exclusion criteria were profound acetabular dysplasia, discrepancy in leg length or anatomical deformities of the proximal femur [10].

The mid- to long-term clinical results for some modular dual taper stem designs are comparable to the implant survivorship of mono-bloc cementless stem designs [7,9,11–15]. Examining the effectiveness of neck modularity in THA considering sex differences in hip morphology, Traina et al. [7] retrospectively reviewed the clinical results of 2131 modular stems implanted in 1051 men and 1080 women. They reported an estimated Kaplan-Meier survival rate at 11 years of 97.6% for men and 96.0% for women without any modular Ti neck failures [7]. In a series of demanding THA procedures with developmental dysplasia of the hip in 47 patients with 61 modular neck stems with an average follow-up of 117.2 months (range 57–167) Traina et al. [9] reported a cumulative survival of 97.5% at 11 years (one ceramic liner fracture). Analysing the long-term survivorship and complication rate of a cementless modular neck primary stem design in the data base of the Emilia-Romagna registry of orthopaedic prosthetic implants, Fitch et al. [12] found for 692 modular THAs implanted in 26 orthopaedic units an overall Kaplan-Meier survivorship of 95.8% at 12 years follow-up, similar to the 96.1% for all mono-bloc cementless stem designs implanted during the same period.

A major disadvantage of dual taper modular hip stems is the additional interface at the neck-stem junction, increasing the risk for fatigue fracture, fretting, crevice corrosion and metal particle and ion release [6,16–22]. For Ti alloy hip stem designs with Ti alloy neck adapters, the main implant-related clinical failure mode is fatigue fracture of the neck taper in the highly stressed neck-stem junction due to fretting corrosion, local stress concentration and crack initiation by multi-directional bending and torsion [16,23–27]. For CoCr neck adapters only few fatigue failures have been reported clinically for a Profemur Z titanium alloy stem (Profemur Hip System, Wright, Arlington, TN, USA) with long 8◦ varus type CoCr alloy modular necks in overweight or obese patients with considerable physical activity [28–30]. For CoCr neck adapters, the main clinical failure mechanism is the generation of particulate Co and Cr wear debris and metal ion release from the neck adapter due to tribo-corrosion, that may cause adverse local tissue reactions (ALTR) [17,31–34] like osteolysis, metallosis and pseudo-tumor formation in the surrounding tissue [19,35,36] as well as elevated serum ion levels [17,33,37–39] and systemic toxicity [40,41]. This failure mode is predominantly occurs in mixed CoCr neck/Ti alloy stem junctions [31,33,39,40,42–44], but has also been described for Ti alloy/Ti alloy dual taper stems [37,38] and single alloy CoCr couplings [45,46].

In a previous study [16] clinical fatigue failures for a short modular Ti alloy stem in combination with a Ti alloy neck adapter and the influence of the implant material on the endurance behaviour have been described. The current study is focused on the evaluation of possible clinical failure modes related to the effects of particulate debris and ion release due to mechanically assisted fretting and crevice corrosion from the same stem in combination with a CoCr neck adapter in order to derive a clinical decision making model.

#### **2. Objectives**

The objectives of our study therefore to attempt to answer the following questions:


#### **3. Literature on Symptomatic and Asymptomatic Dual Taper Hip Stems**

Searching for total hip arthroplasty with dual taper or bi-modular neck adapters in combination with crevice or tribo-corrosion, particle release, metallosis or adverse local tissue reactions we performed a systematic review in PubMed and EMBASE to present an actual literature overview (time frame 1 January 2006 to 31 January 2020). We found after removal of duplicates n = 281 publications, within n = 166 are not related to the topic of dual taper hip stems. Undergoing a systematic full text review based on n = 115 publications we found n = 75 suitable publications and identified additional n = 20 publications (registry reports, conference proceedings, et al.) from our internal database including in total n = 95 studies into the meta-analysis (Scheme 1).

**Scheme 1.** Systematic literature review (Embase n = 284 publications) and studies included in meta-analysis (n = 95).

In a review, Mistry et al. [19] postulated etiologies like modular head-neck or neck-stem junctions wear, corrosion damage and metal ion release as "trunnionosis", a cause of failed total hip arthroplasties [47]. They described the effects of the femoral head size as well as the trunnion design and localized biological reactions associated with trunnionosis [19]. Jacobs et al. [17] described a variety of factors, like head size, taper geometry, material composition, metallurgical processing, surface finish, neck offset and length, design-related factors and a contamination of the taper interface during assembly which may contribute to mechanically assisted crevice corrosion (MACC) in modular junctions involving at least one CoCr component. They found adverse local tissue reactions (ALTR) with clinically symptoms in eleven patients due to MACC with 19-fold elevated mean serum cobalt levels for a beta-titanium alloy (TiMo12Zr6Fe2) modular stem with CoCr neck compared to well-functioning

primary THA patients with metal-on-polyethylene articulations at a relatively early post-operative period (mean 7.9 months) [17]. Dimitriou et al. [33] reported about the early outcomes of revision surgery for taper corrosion of two dual taper THA designs (Rejuvenate & ABG II) with CoCr neck and beta-titanium alloy stem based on 198 revision surgeries in 187 patients. They described a significant decline of patients serum ion levels for cobalt from 5.3 μg/L (range 2.3 to 48.5 μg/L) to 1.4 μg/L (range 0.2 to 8.8 μg/L) and for chromium from 2.6 μg/L (range 0.2 to 64 μg/L) to 0.7 μg/L (range 0.1 to 3.9 μg/L) after revision surgery, with a half-life of 3.2 months for cobalt and of 5 months for chromium. The cobalt/chromium ratio also significantly decreased from 4.7 (range 2.1 to 35) to 2.2 (range 0.4 to 8.8) [33].

For the Rejuvenate dual taper modular neck stem consisting of a CoCr neck and a beta-titanium alloy stem, Bernstein et al. [44] described a revision rate of 86% (63 of 73 hip stems) at a mean follow-up of 4.2 years (range 3 to 5.5) with indications for revision surgery being a serum cobalt ion level > 4 μg/L, persistent pain or abnormal MRI (Magnetic Resonance Imaging) findings. They reported mean serum cobalt and chromium ion levels prior to revision of 10.0 μg/L (range 0.3 to 40.0 μg/L) and of 2.3 μg/L (range 1.0 to 7.4 μg/L), respectively and found a substantial decrease of cobalt levels post-operatively. The unrevised group had a serum cobalt level of 2.1 ± 2.0 μg/L and chromium of 1.2 ± 0.4 μg/L, whereas the patients with abnormal MRI findings had 8.1 μg/L (0.3 to 28.9) and 2.0 μg/L (1.0 to 7.0), respectively [44]. Meftah et al. [31] examined the rate of corrosion-related failure and survivorship of the dual taper Rejuvenate stem (n = 97) and correlated implant and patient factors with serum cobalt ion levels and revisions. The Kaplan-Meier survivorship was 40% at four years and the mean cobalt and chromium levels related to metal corrosion in symptomatic patients were 8.1 μg/L (range 0.4 to 31 μg/L) and 2.5 μg/L (range 0.2 to 4.3 μg/L), respectively.

In contrast to that Vundelinckx et al. [39] described for the ABG II dual taper stem design, which was also made of beta-titanium alloy with CoCr neck, only one revision out of a cohort of 306 THAs consisting of a ceramic-on-polyethylene or a ceramic-on-ceramic bearing interface implanted between 2007 and 2011. The patient undergoing a revision developed intermittent pain in the trochanteric area 2 years post-operatively and showed after 4 years peri-prosthetic fluid accumulation and a soft tissue mass around the proximal stem and neck region and an increased serum level of 7.4 μg/L for cobalt. Taking a randomized sample of 19 asymptomatic patients, 9 patients presented a cobalt level > 4 μg/L with a maximum of 7.5 μg/L.

Walsh et al. [48] studied 103 THA revision cases (78 Rejuvenate; 25 ABG II) with dual taper modular neck at a mean time of 2.4 years (range 0.66 to 5) from index surgery to revision and found a mean serum cobalt level of 7.6 μg/L (range 1.1 to 23 μg/L) and a mean serum chromium level of 1.8 μg/L (range 0.1 to 6.8 μg/L). They performed an aspiration of the hip for synovial fluid metal content prior revision THA in 40 patients and found a mean cobalt level of 916 μg/L (range 12 to 3900 μg/L) and a mean chromium level of 599 μg/L (range 3.4 to 3300 μg/L) in the synovial fluid [48].

Barlow et al. [49] reported about 54 patients undergoing revision surgery with 59 revised Rejuvenate dual taper stems based on a cohort of 199 THAs implanted between 2010 and 2012 by a senior surgeon. They analysed the serum ion levels prior to revision and the decline of serum levels at 6 weeks and 12 weeks after revision surgery. For 49 patients with unilateral THA they found a mean serum cobalt ion level of 8.19 ± 5.54 μg/L, which significantly decreased in all patients to 2.68 ± 2.67 μg/L at 6 weeks and to 1.58 ± 1.57 μg/L at 3 months. In five patients with bilateral modular Rejuvenate hip arthroplasty they measured a pre-revision serum cobalt level of 13.33 ± 6.45 μg/L and also a significant drop to 3.73 ± 2.19 μg/L at six weeks post-revision. An overview of additional studies analyzing serum levels of cobalt, chromium and Ti alloy is given for modular dual taper hip designs with TiAl6V4/TiAl6V4 and CoCr29Mo6/TiMo12Zr6Fe2 neck stem couplings [50–54] (Table 1).

*Materials* **2020**, *13*, 1098 **Table 1.**

Literature overview about clinical cohorts with dual taper neck adapter.


In addition, Hussey and McGrory [56] performed a ten years cross-sectional study including 1352 consecutive patients with metal-on-polyethylene THA combined with 12/14 trunnion Ti alloy stem types and found symptomatic MACC present in 43 cases (3.2%). A dual taper Ti alloy neck and Ti alloy stem design (M/L Taper) showed a higher prevalence (4.9%) of MACC than all other Ti alloy stems combined (1.2%) of the same manufacturer. For these stem design (M/L Taper Kinectiv) with a bi-modular Ti alloy neck on a Ti alloy stem and a CoCr 40 mm head, Canham et al. [57] described in a case study a characteristic peri-prosthetic pseudo-tumor formation as an ALTR and an elevated serum cobalt level of 12.3 μg/L and chromium of 1.8 μg/L whereby the only potential source was the head-neck junction.

For symptomatic patients with dual taper hip arthroplasty Kwon et al. [58] propose a differential diagnosis of the temporal onset, duration, severity and location of pain and recommend that also additional symptoms as a swelling or feeling of fullness around the hip should be elicited. To evaluate dual-taper modular implants with corrosion related metal debris contamination, the analysis of serum inflammatory markers as erythrocyte sedimentation rate and C-reactive protein as well as hip aspiration for synovial fluid counts are described to differentiate and exclude periprosthetic joint infection [59]. The elevation of metal ion serum levels and an increase in the cobalt/chromium ratio is also observed [18,40,59].

Adverse local tissue reactions associated with metal ion and debris released by a dual taper Rejuvenate stem were identified in 36 revised stems out of a cohort of 118 THAs by Ghanem et al. [43]. The symptoms of the 36 THAs which were considered as failed began at a mean post-operative time of 14.8 months (range 2.8–34.8) and the average time to revision was 24.1 months (range 8.8–50.2). The authors described higher cobalt serum levels in the failed group of 9.5 ± 6.8 μg/L (range 1.9–24.7) compared to the asymptomatic group of 4.9 ± 3.6 μg/L (range 0.1–15.7) and higher cobalt/chromium ratios of 5.2 ± 3.2 for the failed compared to 3.6 ± 2.3 for the asymptomatic group. However, they reported no correlation in diagnostic accuracy for ALTR, while MRI scan considering pseudo-tumor size was more sensitive [43].

Kwon et al. [35] evaluated 97 consecutive dual taper modular stem THAs in a retrospective study by MARS-MRI and stratified 83 of these patients in pseudo-tumor absent (n = 53) and pseudo-tumor present (n = 30). In the pseudo-tumor present group they found no substantial difference between symptomatic (n = 21) and asymptomatic patients (n = 9) in the serum cobalt level (sympt.: 7.6 μg/L (range 3.3–14.4) versus asympt. 6.2 μg/L (range 3.4–11.7)) or cobalt/chromium ratio (sympt.: 8.25 (range 4.5–68) versus asympt. 10.6 (range 4.8–29.5)). For the THA patient group with a pseudo-tumor the cobalt serum level and cobalt/chromium ratio (8.0 μg/L; 10.3) were significantly higher than for those without a pseudo-tumor (2.0 μg/L; 2.4) and based on this the authors suggest cross-sectional imaging (MARS-MRI) for THA patients with elevated metal ion serum levels [35]. In a nano-analysis of wear particles from retrieved peri-prosthetic tissue, Xia et al. [60] could clearly distinguish between metal-on-metal (MoM) surface replacement (n = 12; implantation time 51.6 months), MoM large head THA (n = 18; implantation time 59.9 months) and non-MoM dual modular neck hip arthroplasty (Rejuvenate) (n = 23; implantation time 30.9 months). They found that the particle physical characteristics and metal composition are consistent in each implant category and concluded that substantial differences in size, shape and element composition of the metallic particles correlate with the histological features of severity of ALTR and variability in implant performance, indicating that the immunogenicity and toxicity of the released particles is a leading factor in the specific onset and severity of the reaction [60,61].

#### **4. Clinical Case Presentations and Retrieval Analysis**

#### *4.1. Metha*®*Dual Taper CoCr*/*Ti Alloy Couplings with Adverse Local Tissue Reactions*

Data on all known cases (n = 5) with soft tissue reactions due to debris and metal ion release following THA with the dual modular short stem hip prosthesis Metha®(Aesculap AG, Tuttlingen, Germany; Figure 1) with CoCr neck adapter in the period 2007 to 2019 (Table 2). In single cases (#3 & #5) microbiology and histological analysis was performed and showed a wear particle induced peri-prosthetic interface membrane type I according to the classification of Krenn and Morawietz [62,63]. Average duration between index procedure and revision was 92.6 months (61–128 months).

**Figure 1.** Modular Ti alloy short hip stem with a dual taper CoCr neck—Metha® modular.

All retrieved CoCr neck components were characterized by means of light microscopy (M165c, Leica Microsystems GmbH, Wetzlar, Germany), scanning electron microscopy (SEM, EVO 50, Carl Zeiss Microscopy GmbH, Jena, Germany) and energy dispersive X-ray analysis (EDS, X-Max 50, Oxford Instruments plc, Abingdon, UK). Volumetric material loss of the neck adapters was quantified according to a previously published algorithm by Buente et al. [64] using a tactile co-ordinate measurement machine (Figures 2–4).

**Figure 2.** Macroscopic view (left) and volumetric material loss (right) of 5 symptomatic CoCr dual taper neck adapter retrieved after 61–128 months in vivo. The absence of oxygen in the energy dispersive X-ray measurements on characteristic taper surface points indicated that electrochemical processes due to contact, crevice and fretting corrosion affected the ability of the CoCr alloy to form a passive oxide layer in the physiological environment, resulting in increased material loss and metal ion release.




**Figure 3.** SEM and EDS analysis of the medial contact area of one symptomatic retrieved CoCr neck adapter (Case #3). The SEM and EDS-analysis indicated that the local damage of the protective oxide layer and the suppressed repassivation process led to fretting wear and contact corrosion with metal dissolution in the medio-proximally and latero-proximally contact area of the retrieved neck adapter. Furthermore, local material deposition on the neck adapter with a high element concentration of titanium was observed at the circular medial taper interface caused by local adhesion and cold welding in the mixed CoCr/Ti neck-stem junction.

**Figure 4.** Macroscopic view (left) and volumetric material loss (right) of eight asymptomatic CoCr dual taper neck adapter retrieved after 13–111 months in vivo..

Statistics were performed with SPSS 24 (IBM Inc., Armonk, NY USA). Normality was checked using Shapiro-Wilk test. The homogeneity of variances between the groups was checked using Levene's test. Univariate analysis of variance (ANOVA) was performed to compare the two cohorts linear regression analysis to investigate the dependency of material loss with time in situ. The type 1 error level was set to 5%.

Case 1 (exemplified):

In March 2011, a fifty-one years old woman underwent an uncomplicated total hip arthroplasty of the right hip and received a Metha®(Ti6Al4V, stem size 2) with a CoCr neck adapter (CCD-angle of 135◦, 0◦ neutral version) and a 40 mm ceramic-on-ceramic articulation with a titanium sleeve (DeltaMotion®, Finsbury Orthopaedics, Leatherhead, UK). At 61 months after this index procedure, she presented with progressive pain and swelling localized to the right proximal thigh. The clinical and radiographical evaluation showed no pathologic findings with regard to the range of motion, component position and osseointegration. Magnetic resonance imaging (MRI) demonstrated a large heterogeneous fluid collection (7 cm × 3 cm × 3.4 cm) in the ventral aspect of the iliopsoas muscle. Serum cobalt ion level was elevated at 31.3 μg/L and chromium level was normal at 0.3 μg/L. The positive MRI findings and the high Co/Cr-ratio (ratio = 104) were in keeping with adverse local tissue reactions and pseudo-tumor formation and revision total hip arthroplasty was performed in April 2016. The retrieved modular neck exhibited local surface changes and black deposits on the taper interface. Material analysis showed characteristic fretting and corrosion wear patterns concentrated on the medial and lateral contact region of the neck adapter (Figure 2, neck adapter #1). The largest amount of material loss was observed proximal at the medial taper interface with a maximum wear depth of 37 μm. The total volumetric material loss of the neck adapter was 2.6 mm<sup>3</sup> (see also Table 2 cases #2–5 and Figure 2, symptomatic neck adapter #2–5).

#### *4.2. Retrieval Analysis of CoCr Neck Adapters Revised for Other Reasons than Adverse Local Tissue Reactions*

A total of eight asymptomatic Metha®modular CoCr neck adapters revised in the period of 2007 to 2019 for other reasons than adverse local tissue reactions, listed in our database (Table 3).

The reasons for revision were insufficient osseointegration with migration, cup malpositioning, patient discomfort, cup loosening, luxation, acetabular fracture and periprosthetic fracture. The mean age at time of revision was 68.7 years (63 to 77 years) and the mean period of implantation was 62 months (13 to 111 months). The time to revision for the asymptomatic group is comparably shorter than for the symptomatic group (92.6 months).



Visual inspection of the modular neck adapters of the asymptomatic group (Figure 4) by light microscopy and SEM/ EDS-analysis, demonstrated the same characteristic signs of corrosion and debris concentrated on the medial and lateral surface of the neck/stem interface, but qualitatively less pronounced than for the symptomatic group (Figure 2). One patient (Table 3, #4) with signs of a cup loosening revised after 54 months in situ showed some local tissue reactions, which, however, were classified as asymptomatic due to the low neck adapter wear and material loss (3.0 mm3) (Figure 4, neck adapter #4).

Only minor traces of wear and corrosion were seen at the neck adapter of patient #8 (Table 3), based on 111 months of implant in service. The maximum wear depth was 14 μm observed at the proximal lateral taper interface, resulting in a very low total volumetric material loss of 0.3 mm3 (Figure 4, neck adapter #8). This may be related to parameters like a low demanding biomechanical loading of the hip, low tribo-corrosion due to a less aggressive joint fluid composition or unknown co-diseases or morbidity of the patient. The adapter was classified as an outlier and not included into the analysis.

The volumetric material loss of the CoCr neck adapters increased for both cohorts with time in situ (Figure 5; adjusted r<sup>2</sup> = 0.73, *p* < 0.001). A strong trend for a lower wear rate for the asymptomatic cohort compared to the symptomatic cohort was observed (*p* = 0.059). The total volume loss for the symptomatic group was 7.16 <sup>±</sup> 3.73 mm3 and for the asymptomatic 3.69 <sup>±</sup> 2.52 mm3 (*<sup>p</sup>* <sup>=</sup> 0.082). Time in situ tended to be longer for the symptomatic group (92.6 ± 24.2 months) compared to the asymptomatic group (55.0 ± 32.4 months) (*p* = 0.054). The two cohorts did not differ with respect to gender distribution or any other patient or implant specific variable shown in Tables 2 and 3 (*p* > 0.1 for all analysis).

**Figure 5.** The volumetric material loss versus time in situ of seven asymptomatic and five symptomatic retrieved modular CoCr neck components. Note: asymptomatic neck adapter #8 was excluded from the regression analysis.

Due to multiple factors of influence like the patient specific loading situation, weight, activity level, implant orientation and muscular situation, as well as varying physiological surrounding lubricant conditions the comparably low number of only thirteen neck adapters does not allow to estimate a specific patient profile or phenotype. In a descriptive manner of the symptomatic and asymptomatic

cases the patient age at revision was 57 to 79 years, their were 12 females and one male, BMI was between 20.2 and 32 (unknown in 6 cases), neck adapter CCD angles were 130◦ and 135◦, and the head material was ceramic in all cases (2/13 ceramic heads with Ti-Sleeve).

#### **5. Decision Making Model for Asymptomatic Dual Taper Stems in Case of Acetabular Revision**

In the present study, metal ion concentrations were not determined. Therefore the available data from the literature were utilized to develop a decision making model for the case of acetabular revision for pre- and intra-operative decision making, as an orientation how to decide whether to maintain or revise an asymptomatic dual taper stem (Scheme 2).

**Scheme 2.** Practical guide how to proceed with an asymptomatic dual taper modular hip stem in case of acetabular revision.

Cobalt and chromium ions are released from modular dual taper stem connections as consequence of mechanically assisted crevice corrosion [17]. In several retrieval studies elevated cobalt ion levels were associated with adverse local tissue reactions in THA patients with dual taper stems [18,39,45] and a cobalt value of ≥ 8 μg/L serum concentration has been documented for patients having a symptomatic dual taper stem and or a pseudo-tumor present [31,35,39,43,44,49,65,66]. An additional important diagnostic parameter is differential elevation of cobalt relative to chromium [58,65], originated by a predominant cobalt ion release at modular taper connections related to a chemical corrosion process that involves more soluble cobalt dissipating as free ions [40,45,67].

In a series of 447 consecutive patients tested for serum levels Fillingham et al. [65] identified 64 patients with a metal-on-polyethylene THA bearing (12 with a dual taper modular neck), whereas 44 were positive for an adverse local tissue reaction. The diagnosic measures showed a mean serum cobalt level of 8.58 μg/L and a Co/Cr ratio of 11.56. Kwon et al. [35] performed a retrospective study of 97 consecutive patients with a dual taper femoral stem and found substantially elevated cobalt levels of 8.0 μg/L (3.3–14.4) and an elevated Co/Cr ratio of 10.3 (4.5–68.0) in their pseudo-tumor present group. Ghanem et al. [43] identified 107 patients who underwent 118 THAs (11 bilateral cases) with a Rejuvenate dual taper femoral stem and proposed a decision tree to detect whether or not symptoms were present. They found that patients with a serum cobalt level < 6.25 μg/L had a chance of 82% to stay without symptoms, while those with ≥ 18.5 μg/L had a very high risk of failure. Patients in the

failure group had a mean cobalt concentration of 9.5 μg/L and a mean Co/Cr ratio of 5.2, whereas the asymptomatic group had a concentration of 4.9 μg/L and a ratio of 3.6.

In a recent study 148 patients with dual taper modular THA (110 Rejuvenate, 38 ABG II) were examined for pseudo-tumors (n = 90) on MARS-MRI and the severity of intra-operative tissue reactions was correlated with pre-operative cobalt ion levels [36]. The occurrence of pseudo-tumors was associated with significantly elevated cobalt levels (5.0 mg/L vs 3.7 mg/L), a higher Co/Cr ratio (6.0 vs 3.7) and also higher intra-operative tissue damage grades demonstrated substantially elevated Co/Cr ratios (8.6 vs 3.4). These findings on cobalt values and Co/Cr ratios were also underligned by a comparison of cobalt and chromium level diagnostic measurements between positive and negative ALTR groups for a specific metal-on-polyethylene head-neck modularity THA design without dual-taper neck (cohort n = 62) [68]. In 43 THA patients with ALTR a mean cobalt level of 8.92 μg/L and a Co/Cr ratio of 5.91 were found.

In a consensus statement of the American Association of Hip and Knee Surgeons, the American Academy of Orthopaedic Surgeons and the Hip Society, Kwon et al. [58] defined a "high risk" group stratification combining a serum cobalt level > 5 μg/L and a Co/Cr ratio > 5 as factors for diverse modular taper junctions associated with adverse local tissue reactions.

Investigating a cohort of 123 Rejuvenate dual taper THAs, Meftah et al. [31] described a comparably high revision free probability for patients with cobalt serum levels < 4 μg/L. In addition their patients with a Co/Cr ratio of < 3.6 had a likelihood to stay within the asymptomatic group as it has been similarly analysed by Ghanem et al. [43] and Kwon et al. [36].

Due to the fact that serum cobalt ion levels and Co/Cr ratio are confounded in patients having a contra-lateral head-neck-trunnionosis, a bilateral dual taper stem, a metal-on-metal bearing or another joint replacement (knee, shoulder, ankle), surgeons should not solely rely on ion serum concentration factors to determine a clinical recommendation for stem revision [58,65].

An aspiration of the hip to rule out peri-prosthetic infection [69] and to perform intra-articular synovial fluid collection for cobalt and chromium ion content analyses may be a considerable diagnostic option to detect symptomatic MACC prior acetabular revision THA [48,70]. McGrory et al. [70] examined the relationship between serum and intra-articular (IA) cobalt and chromium levels in a cohort of 16 patients with symptomatic MACC undergoing hip revision and they concluded that intra-articular joint fluid levels (IA cobalt 940 μg/L; IA chromium 491 μg/L) positively correlated with serum levels (cobalt 5.1 μg/L; chromium 1.3 μg/L), but intra-articular levels were on average 100-fold higher.

As an important diagnostic tool in detection of adverse tissue reactions due to dual-taper fretting wear and corrosion, cross sectional imaging modality by metal artifact reduction sequence magnetic resonance imaging (MARS-MRI) [35,48,58,59,71–75] and musculoskeletal ultrasound (US) [76] have been qualified.

Walsh et al. [75] described the incidence of different pathologies based on MARS-MRI images in a retrospective cohort of 312 THAs in 272 patients with a dual taper CoCr neck and a beta-titanium alloy stem implanted between 2007 and 2012. They noted synovitis in 167 hips (53.5%), osteolysis in 18 hips (5.8%) and an effusion or fluid collection in 194 hips (62.3%), whereas 52 (29.1%) of these fluid collections were noted to contain debris. Solely intra-capsular effusion and fluid collection was found in 127 hips (40.7%) and combined intra- and extra-capsular in 52 hips (16.7%) [48]. Tendinopathy of one of the related muscle groups (glutaeus minimus, glutaeus medius, iliopsoas or hamstrings) was seen in 250 (80.1%) of the hips and in 87 (27.9%) some tendon disruption occurred.

The presence of a thickened capsule in association with an effusion is a common MARS-MRI abnormality often accompanied by findings like iliopsoas and abductor tendinopathy, peri-tendinous collections and also the presence of metallic debris [59]. For the detection of adverse local tissue reactions like solid or cystic pseudo-tumors MARS-MRI is a highly sensitive modality [59].

Barlow et al. [71] performed in a revised cohort of 90 THA patients with 98 Rejuvenate modular femoral neck stems MRI and serum cobalt and chromium ion level analysis before revision and used

histologic samples from revision surgery to score for synovial lining, inflammatory infiltrate and tissue organization according to Campbell et al. [77]. They found that MRI enables to accurately describe ALTR in dual taper modular neck THA patients and they predicted histologic severity particularly based on maximal synovial thickness and synovitis volume.

To identify femoral osteolysis, loosening and erosions in trochanteric or calcar regions possibly associated with taper corrosion, a focused review of a series of plain antero-posterior and lateral radiographs is proposed [59,69].

Werner et al. [78] reported about adverse inflammatory soft tissue reactions as a consequence of enhanced wear and corrosion of a dual taper neck-stem interface. At time of revision they describe extensive debridement of pseudo-tumor and necrotic bursal and capsular tissue encapsulating the hip joint, as well as corrosion at the neck-stem interface with significant black corrosive debris throughout the modular neck and the soft tissues [78]. Walsh et al. [48] published a study on a cohort of 99 patients including 103 revisions of a dual taper modular neck stem (78 Rejuvenate, 25 ABG II) at a mean time of 2.4 years from primary surgery. They reported intra-operative findings of the 103 revised hips, observing a black metallic sludge associated from corrosion and wear debris in all hips (100%), bony calcar erosion in 88/103 (85.4%), pseudo-tumor formation in 26/103 (25.2%), peri-capsular necrosis in 84/103 (81.6%), tissue necrosis in 80/103 (77.7%) and synovitis in 101/103 (98.1%) of the cases.

Dimitriou et al. [33] evaluated 198 revision surgeries of a dual taper modular femoral stem in 187 THA patients, by an intra-operative tissue damage grading system and observed adverse tissue reactions in 178 (89%), a large amount of fluid entering the capsule in 160 (81%) and particulate wear debris in 103 (52%) of the hips, whereby in all cases a black metallic corrosion debris was found.

If a disassembling of the modular neck takes place intra-operatively during removal of the femoral head within an acetabular revision procedure, the modular hip stem shall be considered for revision. The reason for this is, that a not firmly fixed modular neck/stem taper connection is of high risk for mechanically assisted crevice corrosion. A loosened neck/stem taper connection can be originated during index surgery by an insufficient assembling force or can be caused by macroscopic visible material loss at the medial neck taper interface due to corrosion resulting in a toggling of the modular CoCr neck relatively to the Ti alloy stem [79].

During acetabular cup revision, it may be necessary to place the new cup in an anatomically different position compared to index surgery and this may impact the necessary offset and neck length of the femoral head. In addition, the restoration of the centre of rotation and related ligament balancing during trial head reduction possibly requires a higher offset or longer neck length [4,7,80,81]. Increasing the offset or neck length to adapt for a different cup orientation or sufficient soft tissue balancing may create a more demanding biomechanical loading situation [7,16,32,82–84] for the so far asymptomatic dual taper neck adapter, possibly triggering MACC [17,32,79]. Therefore an indication may be given for revision of the modular stem.

#### **6. Discussion**

In an attempt to find an answer to the question—How to proceed with a clinically asymptomatic modular Metha®stem with dual taper CoCr neck adapter in case of acetabular revision?—following systematic methods have been applied:


A limitation may arise by the fact that the literature review about serum ion levels, radiographic and clinical findings and the retrieval analysis was based in the vast majority on two recalled dual taper stems in the material combination CoCr29Mo6 (neck) and TiMo12Zr6Fe2 (stem), and hence a generalization of the findings and transfer to other implant designs and materials of dual taper stems may be limited [64,85,86]. Meftah et al. [31] used persistent pain and high cobalt levels as predictors for revision surgery of the dual taper Rejuvenate stem and reported a Kaplan-Meier survivorship of 40% at four years with revision related to neck taper corrosion as the end point. They calculated a revision-free probability of 93% for patients with cobalt levels of less than 4.0 μg/L compared with 45% for those with cobalt levels of 4.0 μg/L and found that significantly higher metal ion levels correlated with younger age and a higher femoral offset [31]. Bernstein et al. [44] described a corrosion-related revision rate of 28% at a mean follow-up of 2.7 years in a cohort of 81 Rejuvenate modular hip stems. They prospectively followed this cohort of patients with elevated serum cobalt ion levels (> 4 μg/L), persistent pain, or abnormal MRI findings as indications for revision and observed a clinical failure rate of 86% at a mean of 4.2 years [44]. Koziara et al. [87] reported a study group of 66 out of a cohort of 156 patients who underwent modular Rejuvenate THA with an average follow-up of 55 months (range 22–89) with a revision rate of 31.8% (21 of 66 THAs). They found in the non-revision group a mean serum cobalt ion level of 3.48 μg/L and a Visual Analog Scale (VAS) pain score of 2.4, whereas for the revision group a level of Co = 5.05 μg/L and a VAS pain score 5.1 was present. From the revised group, 18 patients were undergoing MARS-MRI. The THA patients who did not have reactive tissue showed a mean serum cobalt ion level of 4.15 μg/L and a VAS pain score 3.8 and in the group with reactive tissue formation cobalt ion level was Co = 5.01 μg/L and a VAS pain score of 5.63 [87].

For a cohort of 36 patients who underwent uni-lateral primary THA with Profemur®Preserve Ti6Al4V alloy femoral stems and ceramic-on-ceramic bearings, Barry et al. [88] determined the impact of the modular neck material Ti alloy (n = 22) or CoCr (n = 14) with no patient being revised. With a comparably short-term follow-up of 20 months (range 9–44), they observed significantly higher cobalt ion serum concentrations in the CoCr neck group (0.46 vs 0.26 μg/L) and higher titanium ion serum concentrations in the Ti alloy neck group (1.98 vs 1.59 μg/L), but on a comparably low level. Laurencon et al. [89] reported serum and whole blood metal ion levels of a prospective cohort study on 40 THA patients with a cementless anatomic SPS stem made of Ti6Al4V alloy with modular CoCr necks with a mean follow-up of 23 months (range 12–28) and found in 6 of 40 (15%) serum cobalt ion levels > 2 μg/L and in 3 of 40 (7.5%) values > 4 μg/L. Applying MARS-MRI in all THA patients with a serum ion level > 2 μg/L, they detected a pseudo-tumor in one patient having a serum level of 5.21 μg/L for cobalt, 3.51 for chrome and 42 μg/L for titanium [89]. Using a nationwide retrospective cohort of 324,108 THA patients from the French health insurance system, Colas et al. [90] described a cumulative revision incidence of 6.5% for exchangeable neck THAs (n = 8,931) versus 4.7% for fixed neck THAs (n = 315,177) and a significantly increased adjusted hazard ratio of revision of 1.26.

In the National Joint Replacement Registry Report 2018 of the Australian Orthopaedic Association for exchangeable femoral neck adapters the cumulative percent revision for primary THA was reported to be 4.9% at 5 years and 6.8% at 10 years for Ti alloy-Ti alloy stem-neck couplings [91]. For Ti alloy-CoCr modular neck couplings they reported a cumulative percent revision of 9.6% at 5 years and of 16.6% at 10 years [91].

In THA femoral stems with modular exchangeable neck components had significantly lower 10-year survival rates in literature reviews and in registry data compared to primary THA implant survivorship for femoral mono-bloc stems [92,93].

Su et al. [94] performed a retrieval analysis for neck fretting and corrosion on 60 Rejuvenate modular stem designs and compared those to 26 retrieved implants from seven other modular CoCr and Ti alloy neck designs. For the Rejuvenate design they found significantly higher damage and corrosion scores and a 20-fold increased likelihood to show ALVAL based on histologic samples, than for the other designs. As a relevant parameter they stated—beyond design aspects—the lower Youngs modulus of 80 GPa for the TiMo12Zr6Fe2 stem material (Ti6Al4V; 110 GPa), being responsible for increased metal transfer and surface damage in coupling with a CoCr neck, which could account for the higher ALVAL and corrosion scores [94].

Somers et al. [95] stated for the market withdrawn Rejuvenate and ABG II hip systems that the different design features and the stem material TiMo12Zr6Fe2 show more fretting corrosion and it is possible that the different metal trace elements (molybdenum, zirconium and iron) might lead to a more pronounced toxic local tissue reaction.

Lewinski and Floerkemeier [13] described their 10-year experience with short stem TH based on 1953 Metha®short stem procedures with an overall aseptic stem revision rate of 1.3% and 1.9% including 12 modular Ti alloy neck adapter failures based on 190 modular stems with Ti alloy necks implanted before the product recall in November 2006 [16]. Schnurr et al. [14] recorded data for 1888 Metha®short stem implantations from 2004 to 2014 with three implanted versions: Modular Ti6Al4V alloy stems with Ti alloy (n = 314) or CoCr (n = 230) neck adapters and mono-bloc Ti alloy stems (n = 1090) with a mean follow-up of 6 years (range 1–11). They found a 7-year revision rate for mono-bloc of 1.5%, for modular cobalt-chrome of 1.8% and for modular Ti alloy adapter stems of 5.3%, including 15 modular Ti alloy neck fractures.

Apart of the promising 7 to 10 years clinical experiences with the modular Metha®short stem with CoCr necks [13,14], a limitation of our retrieval analysis study is the small number of symptomatic (n = 5) and asymptomatic (n = 8) retrieved CoCr neck adapters out of a cohort of 25,177 dual taper modular stems with CoCr neck, implanted from January 2007 until January 2017. On the other hand side compared to the mostly short-term follow-up in the literature [31,43,44,70,87–89] the retrievals in our study have an average follow-up of 73.8 months and 6 of them have been in patients service for more than 7 years (range 84–128 months).

#### **7. Conclusions**

Based on the analysis of the literature it is suggested that in case of required acetabular revision in patients with a serum cobalt level of > 4 μg/L [31,33,35,36,43,44,58,87] and a Co/Cr ratio > 3.6 [33,35,36,42,43,58] revision of the modular dual taper stem may be considered.

Prior acetabular revision surgery in patients with dual taper modular neck stem THA [59], a systematic diagnostic evaluation has to be executed, using specific tests such as serum metal (Co, Cr) ion analysis, plain antero-posterior and lateral radiographs [59,69] and cross-sectional imaging modalities (MARS-MRI, US) [48,55,59,77]. The patient's stated pain level (e.g., VAS pain score) should also be included as an important factor and measurements of IA cobalt and chromium levels may be meaningful [70].

For an asymptomatic Metha®dual taper Ti alloy/CoCr stem-neck coupling at stage of acetabular revision, careful clinical decision making according to the proposed model should be followed and overreliance on any single examination should be avoided, considering the complete individual differential diagnosis and patient situation.

**Author Contributions:** Conceptualization, T.M.G., M.B., V.J., H.W., K.-D.H., and H.-P.K.; methodology, T.M.G., M.B. and H.-P.K.; validation, T.M.G., M.B., V.J., H.W., K.-D.H., M.M.M. and H.-P.K.; formal analysis, T.M.G., M.B., V.J., H.W., K.-D.H., M.M.M. and H.-P.K.; investigation, T.M.G., M.B. and M.M.M.; data curation, T.M.G., M.B. and M.M.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors would like to thank Sabine Rusch-Zieris for her performance of the systematic literature review in PubMed and EMBASE and Moritz Wente and Thomas Guettler for their fruitful input and critical review of the decision making model.

**Conflicts of Interest:** Two of the authors (T.M.G, M.B.) are employees of Aesculap AG Tuttlingen a manufacturer of orthopaedic implants. Three of the authors (V.J., H.W. and K.-D.H.) are advising surgeons in Aesculap research projects and three of the authors (V.J., H.W. and M.M.M.) have got institutional support by Aesculap AG Tuttlingen.

**Ethical Approval:** Not required.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Biological Reactions to Metal Particles and Ions in the Synovial Layer of Mice**

#### **Xiangyun Cheng <sup>1</sup> , Sabine C. Dirmeier 1, Sandra Haßelt 1, Andrea Baur-Melnyk 2, Jan Philippe Kretzer 3, Rainer Bader 4, Sandra Utzschneider 1,\* and Alexander C. Paulus 1,\*,**†


Received: 29 October 2019; Accepted: 24 February 2020; Published: 26 February 2020

**Abstract:** Metal particles and ions released from implants not only have a fundamental effect on the longevity of total joint replacements, but can also be disseminated to remote organs. Periprosthetic tissues harvested during revision surgeries mainly reflect end-stage failure but may not adequately reveal initial biological reactions and systemic side effects. Therefore, primary reactions caused by metal particles and ions were investigated in an established murine model. Left knee joints in three groups, each consisting of ten female BALB/c mice, received injections of metal ions (MI), metal particles (MP) and phosphate-buffered saline (PBS) (control). Seven days after the injection, immunohistochemical analyses of the synovial layer were performed with respect to some biological markers including Tumor necrosis factor -α (TNF-α), Interleukin-6 (IL-6), Interleukin-1β (IL-1β), Cluster of Differentiation 45 (CD45), Cluster of Differentiation 68 (CD68) and Cluster of Differentiation 3(CD3). The MP group showed significantly enhanced proinflammatory cytokine expression (TNF-α, IL-6 and IL-1β) compared with the other groups (*p* < 0.05). Interestingly, CD3, as a marker for T lymphocytes, did not increase in any of the groups. The MI group showed a significantly increased expression of CD45 compared with the control group (*p* < 0.05). Therefore, during the primary process, metal particles have stronger pro-inflammatory potential than metal ions, and T lymphocytes did not seem to be activated in our murine model. Systemic reactions caused by metal particles and ions were found by observing the untreated right knees.

**Keywords:** inflammation; cytokines; metal particles; metal ions; synovium

#### **1. Introduction**

Medical cobalt-chromium-molybdenum (CoCrMo) alloys are mainly classified by the International Organization for Standardization (ISO) or American Society for Testing and Materials (ASTM) [1] and can be of a cast (ASTM F75, ISO 5832-4) or wrought content (ASTM F1537, ISO 5832-12) [2,3]. Due to their excellent biocompatibility and mechanical properties, these alloys have been widely employed in implant devices that replace hard tissue in the human body [4]. Since the mid-1980s, over one million metal-on-metal (MoM) hip endoprostheses made from a CoCrMo alloy have been implanted worldwide [5]. However, during the 2000s, issues of aseptic loosening due to the release of metal particles and ions were found, and the use of MoM replacements was almost completely stopped [6]. Subsequently, in joint arthroplasties, hybrid combinations were mainly recommended, such as a polyethylene inserts and CoCrMo heads (MoP). Unfortunately, metal wear particles and released metal ions have also been found in patients with MoP prostheses due to the mechanically assisted crevice corrosion of modular taper junctions, including the head–neck and neck–stem taper interfaces [7,8]. Some related studies have been conducted, and researchers have found that CoCrMo particles in revision patients are quite small (< 60 nm) and numerous. In general, high levels of metal ions are also generated from MoM implants [9]. Undoubtedly, the generation of degradation products and the subsequent biological reactions to the metal particles and ions have a fundamental effect on the longevity of total joint replacement [10]. However, to date, the effects of small wear particles and metal ions on local or systemic biological reactions remain complex and are still not fully understood in detail.

Commonly, metal debris, numerous inflammatory cells and increased proinflammatory mediators, such as IL-1β, TNF-α and IL-6, are found in periprosthetic tissues from patients with aseptic loosening [11]. Additionally, a synovial-like membrane with aggressive granulomatous lesions, in which many macrophages are infiltrated, is usually observed around aseptically loosened MoM implants [12]. Therefore, many researchers believe that small metal particles and high levels of metal ions cause an aseptic inflammatory response, which probably leads to periprosthetic osteolysis and, eventually, results in aseptic implant loosening [13]. However, most of the histological evidence for aseptic inflammatory reactions in periprosthetic tissues is from tissues of revision surgeries [14,15], which might not reflect the important initial reactions caused by metal particles and ions. Meanwhile, in terms of metal particles and metal ions, it is still unclear which factor plays a more critical role in aseptic inflammatory reactions.

In addition to aseptic inflammatory reactions, it has been speculated for a long time that type IV hypersensitivity reactions via T lymphocyte activation could play a role in aseptic loosening [13,16]. The assumption is mainly based on the presence of T lymphocytes in periprosthetic tissues from some studies and on the ability of metal ions to activate type-IV hypersensitivity by acting as haptens [17]. However, reactions against metal haptens are mainly determined by individual hereditary factors, with only some patients being genetically susceptible to developing type IV hypersensitivity against metals. Additionally, in some studies, patients with metal-on-metal implants displayed a significant decrease in the number of T lymphocytes and a significant increase in the level of metal ions after the hip replacement [18]. Therefore, it is still somewhat controversial whether type IV hypersensitivity reactions contribute to the aseptic loosening induced by metal particles and metal ions, especially in the short term after joint arthroplasties.

Primary biological reactions around periprosthetic tissues need to be precisely elucidated, not only to improve diagnoses but also to provide valid evidence for subsequent treatment [19–21]. Therefore, the objective of the present study was to evaluate and compare the primary aseptic inflammatory reactions to metal ions and wear particles in an established murine model. In addition, to better understand the role of type IV hypersensitivity reactions caused by degeneration products, T lymphocytes were also evaluated in the established model. The first hypothesis is that the metal particle (MP) group and metal ion (MI) group can both induce enhanced proinflammatory cytokine expression (TNF-α, IL-6 and IL-1β), and that T lymphocytes would be recruited when particles and ions were released in vivo. The second hypothesis was that the MI group have a stronger pro-inflammatory potential and recruit more T lymphocytes than the MP group.

#### **2. Materials and Methods**

#### *2.1. Metal Particle*/*Ion Generation*

Standard wrought CoCrMo alloys (ISO 5832 -12, ASTM F1537) [20] for hip implants were used to generate metal ions and wear particles in this study.

To obtain metal particles that were similar to clinically produced particles, a newly developed pin-on-disc simulator was used to perform wear tests at a frequency of 1 Hz. The wear volume was determined in the test medium by a high resolution inductively coupled plasma mass spectrometry (HR-ICP-MS) instrument (Thermo Scientific, Bremen, Germany). After performing an analysis via scanning electron microscopy (SEM; Zeiss EVO 50, Carl Zeiss NTS GmbH, Oberkochen, Germany), the following parameters were recorded: the equivalent circular diameter (ECD), the aspect ratio (AR) and the roundness (R) of wear particles.

To induce the release of metal ions, the CoCrMo alloys were anodized in a corrosion chamber and phosphate-buffered saline (PBS) was used as the surrounding medium. Using an HR-ICP-MS instrument (Thermo Scientific, Bremen, Germany), a total metal ion content (cobalt, chromium, molybdenum and nickel) of 20.5 mg/L was determined, which was adjusted to the desired target concentration of 200 μg/L using PBS. The target concentration of 200 μg/L was based on a study in which the metal ion level of a joint puncture of patients was analyzed prior to revision surgery, which showed average concentrations in the range of 200–250 μg/L in the joint fluid [22].

#### *2.2. Elimination of Endotoxin*

To avoid the influence of adherent endotoxins on the results, the generated particles and ions had to be free of endotoxins. To remove endotoxins, metal particles were cleaned by an ethanol washing process, and the metal ion solution was heat shocked. The elimination of endotoxins was proven by the Limulus Amebocyte Lysate (LAL) test (Lonza, Cologne, Germany).

#### *2.3. Animals and Intraarticular Injection*

Forty female BALB/c mice (Charles River Wiga Company, Sulzbach, Germany), weighing 23.5 ± 1.9 g, were used to establish an animal model in which biological reactions could be evaluated. These mice were randomly assigned to three groups: the PBS control group (*n* = 10), MP group (*n* = 10) and MI group (*n* = 10). To exclude the possibility that the process of the intra-articular injection itself produces an inflammation reaction in subsequent experiments, the remaining untreated mice (*n* = 10) were regarded as the negative control (NC) group. All experimental steps involving animals were performed according to the rules and regulations of the Animal Protection Laboratory Animal Regulations (2013) and European Directive 2010/63/EU Act, which is in accordance with the National Animal Protection Law (protocol number 55.2-1-54-2532-82.12, Government of Bavaria, Germany).

Prior to the intra-articular injection, all solutions were sonicated for at least 60 min to prevent possible agglomeration and precipitation. Under sterile conditions, 50 μL PBS-suspension, 50 μL of a 0.1 vol% CoCrMo particle suspension and 50 μL of 200 μg/L CoCrMo ions were injected into the left knees of the mice. After a 7-day incubation period, animals were euthanized by an intracardial pentobarbital injection (Merial GmbH, Hallbergmoos, Germany).

#### *2.4. Immunohistochemistry*

Both knees (left treated and right untreated) of all groups were removed and fixed in 4% paraformaldehyde for 24 h. All knee joints were decalcified using Osteosoft solution (Merck KGaA, Darmstadt, Germany) for four days at room temperature. Decalcification of the knees was followed by dehydration in a Spin Tissue Processor-120 (Especialidades Médicas Myr, S.L., Tarragona, Spain) in an ascending alcohol series (7%, 96%, 100%, and xylene) and, finally, a transfer to two successive paraffin baths at 60 ◦C. After being embedded in paraffin, solidified blocks were cut into 2 μm thick sections. Subsequently, the prepared tissue sections were immunochemically analyzed using six different

monoclonal mouse antibodies: TNF-a (1: 200 dilution), IL -1b (1: 150 dilution), IL -6 (1: 200 dilution), CD68 (1: 100 dilution), CD45 (1: 400 dilution) and CD3 (1: 100 dilution) (Biorbyt Ltd., Cambridge, United Kingdom). Factors from the spleen (TNF-a), lung (IL -1β and IL-6) and tonsil (CD68, CD45 and CD3), which always react positively with the corresponding antibodies, served as positive controls in our study (data not shown).

As shown in many studies, the cytokines TNF-α, IL-1β and IL-6 are important inflammatory triggers in peri-implant tissues and are even involved in subsequent prosthetic loosening [23,24]. Therefore, in this study, TNF-α, IL-1β and IL-6 were used as inflammatory markers. According to the pro-inflammatory cytokine expression (TNF-α, IL-6 and IL-1β) in immunohistochemistry, we assessed the degree of inflammatory response in each group.

CD45 can be expressed in various immunocompetent cells, including dendritic cells, T lymphocytes, B lymphocytes, macrophages, etc. [25,26]. CD45 was used as a general marker of immunocompetent cells in the synovial layer in this study. The CD3 antigen binds to the membranes of all T lymphocytes and virtually no other cell types, which makes it a useful immunohistochemical marker of T lymphocytes in tissue sections [27]. The CD68 antigen is mainly expressed by monocytes and macrophages. When monocytes migrate into local tissues, they differentiate into macrophages [28,29]. Therefore, we used CD68 as a marker of macrophages in this study. To analyze the systemic biological reactions caused by metal particles and metal ions in our murine model, the immunohistochemical markers mentioned above were analyzed in the right knees of all groups.

Images of the stained samples were collected under a light microscope at 200× magnification (Carl Zeiss Micro Imaging GmbH, Oberkochen, Germany). Every image included the most synovial tissue (region of interest) of each sample. If cells or tissues were stained from light yellow to brown, positive immunostaining was recorded. Two observers independently performed manual counts of the obtained images with the assistance of Image J software (National Institutes of Health, Bethesda, MD, USA) (Ver.1.43, available at rsbweb.nih.gov/ij). Preceding manual counting, images were cropped, scaled to μm and separated by a color channel, and artifacts were removed. An area tool was used to select the synovial region and calculate the area. The Image J cell counter tool recorded mouse clicks on cells that were labeled with colored dots. The results were saved to a spreadsheet and screen shots were used to record the session. Positive cells from the synovial membrane were counted by each observer. If the results were inconsistent, specific samples were collected and counted by both observers for a third time.

#### *2.5. Statistics*

The data obtained from the immunohistochemical evaluation were evaluated using IBM SPSS®Statistics 22 (IBM Deutschland GmbH, Ehningen, Germany). Statistical analyses were carried out with the non-parametric Kruskal–Wallis test for independent samples. *P*-values below 0.05 were considered statistically significant and were adjusted by Bonferroni correction. A graphical presentation of the results was produced using box plots.

#### **3. Results**

#### *3.1. Characterization of Metal Particles and Ions*

For the generated metal particles, the mean size was in the nanometer range (ECD: 61.25 ± 18.47 nm). Meanwhile, their aspect ratio was 1.69 ± 0.66 and roundness was 0.64 ± 0.16 (Table 1). The particle shape was predominantly round and oval; in addition, a small proportion of acicular particles were formed (shape: round, 44%; oval, 49%; needle, 7%). This size and morphology of the particles were consistent with clinically found metal particles after a MoM total hip replacement [30,31].

**Table 1.** Morphological parameters of particles. The CoCr29Mo6 particle shape is predominantly oval and round with a small proportion of acicular particles (shape: round, 44%; oval, 49%; needle, 7%). Equivalent circular diameter (ECD).


The CoCr29Mo6 particle shape was predominantly oval and round with a small proportion of acicular particles (shape: round, 44%; oval, 49%; needle, 7%).

Using HR-ICP-MS, a total metal ion content (cobalt, chromium, nickel and molybdenum) of 20.5 mg/L was determined, which was adjusted to the desired target concentration of 200 μg/L using PBS (Table 2) [32]. The target concentration of ions was based on the concentration measured in the synovial fluid of patients with endoprosthesis during revision surgery [22].



Total ion concentrations of the CoCrMo stock solution as well as in the experimental solution (200 μg/L in total) according to Jonitz-Heincke et al. [32].

#### *3.2. Results of the Left Knee Joints*

#### 3.2.1. Expression of TNF-α, IL-1β and IL-6

With regard to the staining results of TNF-α, IL-1β and IL-6, the MP group showed a significantly increased number of positive cells compared with the PBS group (*p* < 0.05). However, the MI group did not express significantly more TNF-α (*p* = 0.056), IL-1β (*p* = 0.420) and IL-6 (*p* = 0.124) than the PBS group. Therefore, according to these results, the MP group had stronger inflammatory reactions in the synovial layers of the left knees (Figure 1).

**Figure 1.** Expression of inflammatory markers (IL-1β, IL-6 and TNF-α) in the synovial layer of left murine knees. The staining results of TNF-α in the negative control (NC), phosphate-buffered saline (PBS), metal ion (MI) and metal particles (MP) group. Both the strongest expression and a thickened synovial layer are found in the MP group. (**A**) The expression of TNF-α in the synovial layer of left murine knees. (**B**) The expression of IL-1β in the synovial layer of left murine knees. (**C**) The expression of IL-6 in the synovial layer of left murine knees. The staining results of IL-1β, TNF-α and IL-6 are consistent. The MP group had the strongest inflammatory reactions. (Magnification: 200x; \* *p* < 0.05).

#### 3.2.2. Expression of CD68, CD3 and CD45

The MP group showed a significantly higher number of CD45-positive cells compared with the PBS group (*p* < 0.05). Likewise, the MI group had a significantly different number of CD45-positive cells compared with the PBS group (*p* < 0.05). Interestingly, in terms of CD45, there was no significant difference between the MP group and the MI group (Figure 2A).

**Figure 2.** Expression of CD45, CD3 and CD68 in the synovial layer of left murine knees. The staining results of CD45 in the NC, PBS, MI and MP group. Numerous CD45 positive cells were found in the MP and MI group. MP and MI groups had significantly increased positive cells compared with the PBS group (*p* < 0.05). (**A**) The expression of CD45 in the synovial layer of left murine knees. (**B**) The expression of CD3 in the synovial layer of left murine knees. No significant difference was found in all groups. (**C**) The expression of CD68 in the synovial layer of left murine knees. (Magnification: 200x; \* *p* < 0.05).

With respect to the expression of CD3 positive cells in our study, there was no statistically significant difference among the three groups (*p* = 0.45) (Figure 2B).

The MP group showed a significantly increased number of CD68 positive cells compared with the PBS group (*p* < 0.05), and the MI group (*p* < 0.05) in the synovial layer of the left knees. There was no significant difference between the MI group and the PBS group (*p* = 1.0). Considering that macrophages play an important role in inflammatory reactions, these results indicate that the MP group had the strongest inflammatory reactions, which is consistent with the TNF-α, IL-1β and IL-6 results (Figure 2C).

#### *3.3. Results of the Right Knee Joints*

#### 3.3.1. Expression of TNF-α, IL-1β and IL-6

For TNF-α and IL-1β, the number of positive cells in the MP group was only significantly increased compared with the PBS group (*p* < 0.05). There was no significant difference among all groups in the expression of IL-6 in the synovial layer of the right murine knee joints, (*p* = 0.13) (Figure 3A–C).

**Figure 3.** Expression of all biological markers in the synovial layer of right murine knees. The aim of examining the right knees was to evaluate systemic reactions caused by the dissemination of metal particles and metal ions. (**A**) The expression of TNF-α in the synovial layer of right murine knees. (**B**) The expression of IL-6 in the synovial layer of right murine knees. (**C**) The expression of IL-1β in the synovial layer of right murine knees. (**D**) The expression of CD45 in the synovial layer of right murine knees. The MI group exhibited a significant difference compared with the PBS groups (*p* < 0.05). (**E**) The expression of CD3 in the synovial layer of right murine knees. (**F**) The expression of CD68 in the synovial layer of right murine knees. (\* *p* < 0.05).

#### 3.3.2. Expression of CD68, CD3 and CD45

The expression of CD45-positive cells in the MI group was significantly increased compared with all other groups (PBS: *p* < 0.05, MP: *p* < 0.05). No statistically significant difference was found in the direct comparison of the PBS and MP groups (*p* = 1.0). Meanwhile, there was no significant difference among all groups in the expression of CD68 and CD3 in the synovial layer of the right knees (CD68: *p* = 0.55, CD3: *p* = 0.91) (Figure 3E–G).

#### **4. Discussion**

Our initial hypothesis for this study could not be proved. No increased T-lymphocyte markers were found in either the MP or MI group. However, in terms of proinflammatory reactions (IL-1β, IL-6 and TNF-α), the MP group had stronger reactions than the MI group. In addition, in terms of systemic reactions, the MP group and the MI group both had significantly increased biological reactions in the synovial membrane of right-sided knee joints compared to the PBS (control) group.

Histopathological studies can decisively contribute to the determination of the main cell populations underlying the biological mechanisms of aseptic inflammation and osteolysis [33]. In addition, the key protein molecules (such as IL-1β, IL-6 and TNF-α) involved in aseptic loosening can also be precisely detected by immunohistochemistry [34,35]. Therefore, periprosthetic tissues harvested during revision surgery, especially the bone–implant interface membrane and pseudo-synovial tissues, are frequently examined via histological techniques in clinical work [14,36]. However, periprosthetic tissues from revision surgeries mainly reflect end-stage failure and may not adequately reveal the primary biological reactions caused by metal particles and ions released immediately after the initial surgery. Nevertheless, the primary biological reactions caused by metal products in vivo may be crucial to elucidate the evolution of the pathophysiological events that lead to prosthetic osteolysis. In light of the limitations mentioned above, an established murine model was used in this study, which mimics the initial biological reactions in the synovial-like tissues around prostheses caused by metal particles and ions [35,37–39]. In contrast to various in vitro cell culture studies that focus on one cell type [32,40], the murine model can not only reflect complex cellular and tissue interactions, but can also mimic the dynamic process of joints. Additionally, compared with other models, such as hamsters' skinfold-chamber models [41] and "air pouch" models [42], results from this model, as the suspensions were injected intraarticularly, are easier to translate into a clinical scenario because the generated wear debris primarily accumulates in the joint that has been replaced. In terms of the characteristics of metal materials, the CoCrMo particles and ions used in our study were consistent with those found in some clinical studies.

CD45 can be expressed on many types of cells in the immune system, such as lymphocytes, natural killer cells, granulocytes, dendritic cells and monocytes/macrophages [26,43]. Therefore, CD45 was chosen as a general marker of immunocompetent cells in the synovial layer. For the left (treated) mice knees, the MP group and MI group both showed significantly increased CD45-positive cells compared with the PBS group, assuming that some immunocompetent cells were recruited to the synovial layer as a result of the metal particles and metal ions. However, it is difficult to distinguish specific immunocompetent cells and biological reactions solely from marking CD45. CD3 and CD68 were used as specific markers for T lymphocytes and macrophages [27], while IL-1β, IL-6 and TNF-α were used as common inflammatory markers [16].

For the groups of left knees that received an intra-articular injection, the MP group's inflammatory reactions (IL-1β, TNF-α, IL-6 and CD68) increased considerably more than those of the other groups. Considering the physical properties of CoCrMo particles, released nanometer-sized particles probably cause physical harm, especially when the knee joint is in motion [21]. This may be one reason why CoCrMo particles had very intensive inflammatory reactions in this in vivo study. Additionally, via various released cellular mediators, damaged cells can lead to the activation and recruitment of immunocompetent cells, especially macrophages. Subsequently, recruited macrophages phagocytize some small CoCrMo particles, while foreign body multinucleated giant cells surround very large particles [44]. Inside the cells, the particles are exposed to oxidative attacks, and consequently, high levels of ions might be released during the chemical corrosion processes [45]. Therefore, CoCrMo particles not only have a destructive effect because of their physical properties—the corrosion process and the presence of metal ions can also cause some biological reactions in the surrounding tissues [46]. Furthermore, because of the presence of phagocytized debris, cellular necrosis can occur and many mediators of inflammation are released by macrophages, which can recruit more macrophages, amplifying the inflammatory cascade [20]. Numerous macrophages (CD68 positive cells) were found in the MP group in this study. Although metal ions can also initiate oxidation-reduction reactions, the metal ions were rapidly quenched, while metal particles offered a reservoir of redox-reactive metals for continuous Reactive oxygen species (ROS) generation, which could further explain why metal particles resulted in the strongest local inflammatory reactions in this study.

However, in contrast with the inflammatory reaction results in our murine model, some in vitro studies have shown different results. Chamaon et al. indicated that the treatment of a human monocytic cell line with ionic cobalt led to a decrease in metabolic activity [Water Soluble Tetrazolium-1(WST-1) assay], while CoCrMo particles had no effect [47]. They explained that the rather large abrasive particles (from 200 nm to several micrometers) used in their study might have resulted in the relatively low impact of CoCrMo particles [47]. Because particle size is considered a critical parameter that influences the biological reactions to wear particles, the majority of MoM-produced particles are usually from 40 to 60 nm [2,48]. One easily neglected factor should also be considered: the in vitro cell cultivation process was static and could not reflect physical injury from metal particles in vivo. Moreover, Caicedo et al. showed that similar increases of IL-1β, TNF-α and IL-6 were evoked by cobalt, molybdenum ions, and Co-Cr-Mo alloy particles in human monocytes/macrophages [49]. Meanwhile, they found that ions and cobalt alloy particles induced inflammasome activation in vitro in a dose-dependent manner [50]. In our murine model, metal ions might be unavoidably systemically disseminated via lymphatics and blood vessels and then circulated throughout the host body, probably leading to a more general distribution than the metal particles. The gradually decreased metal ion level could attenuate local inflammatory stimulation effects, which might be one important reason why the level of inflammatory mediators (IL-1β, IL-6 and TNF-α) and the number of macrophages (CD68+ cells) did not significantly increase in the MI group in our study.

According to the literature, a cell-mediated (type-IV delayed hypersensitivity) response, which is mainly characterized by the activation of T lymphocytes, can be triggered by metal particles and ions during the process of aseptic loosening [19,51]. At least theoretically, metal ions are able to activate type-IV hypersensitivity by forming haptens with host proteins [16,52]. Compared with toxic reactions, type-IV hypersensitivity reactions do not have a simple dose–response relationship, with higher doses being more potent than lower doses, because even small amounts of antigens can cause strong reactions [17]. In our study, we attempted to count T lymphocytes (CD3-positive cells) to verify the existence of a type-IV hypersensitivity response. However, unexpectedly, regarding T lymphocytes, there were no significant differences among the groups. A delayed hypersensitivity response typically occurred between 24 and 48 h after exposure to an antigen [53]. One possibility was that a delayed-type hypersensitivity response did not play an important role in aseptic implant loosening, and that T lymphocytes were scarce in the primary process. Many researchers support this opinion, because there is a relatively low revision rate related to "allergic" loosening [20]. Approximately 10% of the population is hypersensitive to the materials found in jewelry and joint replacements [54]. If metallic debris and ions can induce a serious hypersensitivity response in the periprosthetic tissues, then more revision surgeries might have been performed as a result of "allergic osteolysis and aseptic loosening"; however, this is not actually the case [20,54].

Some studies have indicated that small nanoparticles and high levels of metal ions from implants can be detected in remote organs, including the liver, lung, spleen and kidney [55]. Whether the systemic dissemination of wear particles and metal ions can cause side effects is a matter of debate [20,23]. As such, in our study, the synovial membrane of each right-sided knee joint after the intra-articular particle or ion injection (left knee) was also analyzed for the presence of biological markers (IL-1β, IL-6 TNF-α, CD45, CD68 and CD3). For IL-1β and TNF-α, only the right knees of the MP group had a stronger response than the PBS group, which means that a systemic response caused by metal particles occurred. By contrast, IL-6 was not significantly increased in the right knees of the MP group. One possible explanation for this result is the presence of a temporal factor, because IL-6 secretion in the periprosthetic membrane is preceded by the expression of TNF-α and IL-1β [56]. With regard to CD45, the results of the right knees in the MI group were consistent with those of the left knees, indicating that some immunocompetent cells (CD45-positive cells) were activated by metal ions, and that metal ions also led to systemic reactions. In terms of CD3 and CD68, there were no significant differences in the right knees of all groups, which means that no macrophages or T lymphocytes were recruited in the right knees of all groups.

There are some limitations to this study. Although the murine model can reflect inflammatory reactions caused by implant materials, it cannot imitate the osteolysis process around implants. Additionally, the single-injection murine model used in this study might not completely reflect the chronic production of wear debris in patients with joint replacements. Some researchers have used osmotic pumps in a murine model to achieve a continuous infusion of degeneration products [57]. This model is more like the clinical scenario, but irritation from osmotic pumps might be an interference factor. In a further study, we will aim to establish one continuous infusion model in combination with existing techniques. To investigate the systemic reactions caused by particles and ions, we used untreated knees. In the future, the impact of metal particles and ions on other organs, such as the liver, spleen, heart and kidney, will also be evaluated, which would further reflect the overall reactions caused by metal debris. Additionally, only one concentration of metal particles and ions was used, referring to the national animal laws. In a future study, different metal particle and ion concentrations, and the impact of different concentrations on aseptic inflammation, will be investigated.

#### **5. Conclusions**

The results of this study clearly demonstrate that the primary process which occurs after the release of metal particle and ions, especially CoCrMo particles, can lead to an intensive proinflammatory response in vivo. Metal ions can also cause the recruitment of immunocompetent cells but, in view of local inflammatory reactions, macrophages and inflammatory mediators were scarce in vivo. During the process by which metal particles and ions were released in present study, T lymphocytes were not recruited in our murine model. Systemic reactions by metal particles and ions were found according to the observation of untreated right knees.

**Author Contributions:** Data curation, X.C. and S.C.D.; formal analysis, X.C. and S.C.D.; funding acquisition, J.P.K., R.B., S.U. and A.C.P.; investigation, S.H.; methodology, S.H. and A.C.P.; project administration, S.U. and A.C.P.; resources, J.P.K., S.U. and A.C.P.; supervision, A.B.-M. and S.U.; validation, S.C.D.; writing—original draft, X.C.; writing—review & editing, S.U. and A.C.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by German Research Foundation, grant number UT 119/3-1, BA 3347/12-1, KR 3755/5-1. And Xiangyun Cheng's doctoral work is funded by the China Scholarship Council (CSC) under the State Scholarship Fund (Grant No.201708140085).

**Acknowledgments:** The authors would like to thank the German Research Foundation (DFG) and China Scholarship Council (CSC) for funding of this project. We also gratefully thank Dr. Paul Johan Høl for his assistance in ICP-MS techniques.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


proinflammatory cytokines critical to metal-induced lymphocyte reactivity. *J. Biomed. Mater. Res. A* **2010**, *93*, 1312–1321. [CrossRef]


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
