**Retrieval Analysis of Modern Knee Tumor Megaendoprosthesis Shows Considerable Volumetric Metal Wear Generated at the Rotating Hinge**

**Therese Bormann 1,\*, Sebastian Jäger 1, J. Philippe Kretzer 1, Laura Nebel 1, Lucas Clarius 1, Georg Omlor 2, Rudi Bitsch <sup>3</sup> and Burkhard Lehner <sup>2</sup>**


Received: 28 February 2020; Accepted: 25 March 2020; Published: 26 March 2020

**Abstract:** Frequently occurring damage, as well as elevated blood metal ion levels, are reported in relation to a tumor and revision system for total knee arthroplasty (TKA), which applies a rotating hinge coupling with a metal-on-metal (MoM) articulation. As the patient collective for this specific system is small, there is no data on wear generated from the couplings. In this study, wear volume and influencing parameters were investigated at 44 retrieved TKAs with MoM couplings. A scoring system rating frequently occurring abrasive wear between 0 (no wear) and 3 (distinct wear) was established. The wear score was correlated to time in vivo, bone resection length, patient weight and polyethylene inlay damage. Volumetric wear was estimated applying coordinate measurements. An elevated wear score of two or higher was found in 43% of cases. The mean wear rate accounted to 7.8 mm3/year. The main influencing coefficient for the extent of wear is time in vivo. We found a tendency for higher wear scores with higher inlay degradation scores. Patient weight and bone resection length did not impact coupling wear. Assessment of wear damage by a semi-quantitative scoring system has proven to be a reliable option for non-destructive coupling evaluation. The generated wear volume is high.

**Keywords:** metal wear; retrieval study; metal-on-metal articulation; volumetric wear; megaendoprosthesis; total knee arthroplasty; bone tumor

#### **1. Introduction**

Tumors in the distal femur or the proximal tibia often result in the resection of large bone segments. To salvage the limb, the bone and knee joint are usually reconstructed by endoprosthesis systems [1,2]. Modern systems are constructed modularly in order to address different resection lengths and to permit intra-operational flexibility [1]. As together with the bone also soft tissue and ligaments of the leg are removed, a higher degree of constrain in the knee joint is necessary to ensure stability of the joint. For that, tumor and revisions systems usually couple the distal femur and proximal tibia with a hinge or a rotating hinge system. Generally, the outcome of a fully stabilized and hinged prosthesis is worse than that of a less stabilized knee endoprosthesis [3,4]. Revision rates after five years rise from 3.5% for minimally stabilized total knee arthroplasty (TKA) to 8.1% for the coupled systems, with infection being the most common cause [5].

Aside from infections, mechanical complications like aseptic loosening, structural failure and instability are common reasons for a revision [2,6]. The high incidence of mechanical complications is associated with the coupling, as substantial stresses are transmitted via the coupling mechanism [6,7] and—for the hinged systems—these are further transmitted to the implant–bone interface.

As bone tumors often occur between the second and forth decade of life, the patient collective is considerably younger than that of general primary total knee replacement, which has a mean age of 69 years [5]. It has been shown that the probability for a second complication increases after a first complication requiring revision surgery [8]. Bone tumor patients are therefore very likely to have multiple revision surgeries during the course of their life. Not seldom, the affected limb is sacrificed at some point in the patient's life [8,9].

The modular universal tumor and revision system (MUTARS®, Implantcast, Buxtehude, Germany) is a commonly used endoprosthesis system for tumor patients with good functional clinical results [7,9,10]. The femoral and tibial components of this system are joined by a rotating hinge coupling mechanism, sometimes also referred to as bushing. Despite the satisfactory clinical results of this system, structural complications are reported for up to 30% of the patients [7,9–11]. This failure mode includes periprosthetic fracture, breakage of a prosthesis part and wear of the bushing, all requiring a revision operation. The bushing design was changed several times since the introduction of the system in 1995. The latest design uses a metal-on-metal (MoM) coupling mechanism made of cobalt–chromium–molybdenum-alloy. Nevertheless, problems occurring with respect to the coupling mechanism do not seem to be overcome. Bushing wear has still been reported, as well as elevated blood cobalt and chromium levels related to the MoM coupling [9,12]. Even though exchange of the bushing itself is a standard procedure not requiring other parts of the system to be replaced, it affects the patient's life as an additional surgery. Furthermore, the generated metal particles and ions are released into the human body, where they can cause adverse biological reactions that, in turn, can lead to clinical problems such as soft-tissue necrosis, pseudotumors and particle-induced implant loosening [13–15].

In our in-house retrieval registry, we regularly observed (obvious) coupling failures like bushing fracture. Apart from that we noticed marks of abrasive wear on retrieved MUTARS® total knee replacements. As there is no data on wear generated at coupling mechanisms in constrained total knee arthroplasty, we aimed to answer the following research questions in this retrospective study: (1) Is the degree of wear at MoM couplings correlated to implant or patient specific factors? (2) How much wear is actually generated, and can it be quantified?

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

#### *2.1. Retrievals*

We analyzed 44 retrieved MUTARS® prostheses (Implantcast, Buxtehude, Germany) with a metal-on-metal coupling. Retrievals are from our in-house retrieval registry, which is approved by an ethical committee. Implantation of the 44 prostheses was carried out between 1997 and 2018, with explantation between 2001 and 2018. The mean time in vivo was 31.0 ± 38.7 months (median 17.6 months, range 0.4–172.1 months). The coupling is made of cobalt–chromium–molybdenum-alloy (CoCrMo, as specified in DIN ISO 5832-4 [16]), and consists of a sleeve, inside which a piston with a hemispherical head articulates. Among the retrieved couplings we identified four different designs, as displayed in Figure 1a. In the oldest (design A), the piston is fixed within the tibia plateau by a thread. All following designs connect the piston by a stuck bolt to the tibia plateau. Design B is characterized by a bore in the piston stem. Designs C and D are characterized by a bore on top of the hemispherical piston head instead of the one in the stem. Designs C and D further differ in the dimensions of sleeve and piston. In the latest design from our cohort, design D, the piston stem and head are thicker, and the sleeve is shorter as for the designs A, B and C.

**Figure 1.** (**a**) Metal-on-metal (MoM) coupling mechanisms consisting of piston and sleeve. The four design variations from our cohort are displayed. Design A refers to the oldest modification, while design D is the most recent. (**b**) The sleeve after separation of the lid for quantitative wear analysis. The arrow indicates the articulating counter face, which was investigated by a coordinate measuring machine (CMM).

#### *2.2. Semi-Quantitative Wear Analysis by Scoring System*

Failure modes of these prostheses were classified by the Henderson classification. As wear of the coupling occurs independently from the actual failure mechanism of the device, and appears to the naked eye in almost all retrieved samples, we developed a semi-quantitative scoring system to rate coupling wear. The scoring system is based on the Hood and Goldberg scoring systems, which both are well established visual methods to assess damage of retrieved implant parts [17,18]. Arnholt et al. applied a similar scoring system to assess damage of CoCr femoral condyles [19]. As both parts are prone to wear, we rated them individually. Rating was accomplished on a scale between 0 (no wear) and 3 (distinct wear), and the criteria for the point assignment and examples are displayed in Table 1. For the cylindrical sleeve, the external surface was divided into four quadrants, which were rated individually and subsequently averaged. The piston was divided in two zones, the hemispherical head and the stem. Both zones were rated individually from 0 to 3, and subsequently averaged. The score for the whole coupling was determined by averaging the sleeve and piston scores. Scoring was carried out by two independent investigators in order to verify the assessment of wear by the developed scoring scale. All data presented refer to the evaluation of investigator 1.



Couplings were ultrasonically cleaned in two steps prior to score determination. In the first step, cleaning was carried out at 60 ◦C in a detergent-containing solution for one hour. Subsequently, couplings were immersed in Ethanol for 10 min.

Pre-operative radiographs for metering the bone resection lengths of both the tibia and femur, respectively, were available from 35 patients. In case of the resection of the distal femur and proximal tibia, both values were totalized for further analysis.

Damage of polyethylene inlays was assessed on 40 inlays applying a modified Hood score [18]. We rated five damage mechanisms, i.e., delamination, pitting, scratching, burnishing and surface deformation at a scale between 0 (no damage) and 3 (distinct damage) for seven zones of the inlay, as shown in Figure 2. Achieved degradation scores were summed up, which results in a maximal possible score of 105. Determination of the modified Hood score was done by two investigators. The

mean difference of the degradation score of the PE-inlays between the two investigators referred to 4.78 ± 3.2. The data presented refers to the evaluation of investigator 1.

**Figure 2.** Polyethylene inlay with typical traces of wear. For the determination of the Hood score, the inlay was divided into seven zones.

#### *2.3. Quantitative Wear Analysis by Coordinate Measurements*

In order to estimate the material loss over the period of implantation, we assessed the geometry of the articulating surfaces of the coupling mechanism by a coordinate measuring machine (CMM; MS222, Mahr, Göttingen, Germany). For that, the sleeve and piston were separated by cutting the welding seam by which the lid was connected to the sleeve (see Figure 1b). Subsequently, the piston head was cut from the piston stem. The articulating faces of the semispherical piston head and the opposing counter face (Figure 1b) were scanned point wise in spherical coordinates with angle steps of 4◦ and 6◦ for theta (θ, polar angle) and phi (ϕ, azimuthal angle), respectively, as demonstrated for a cut off piston head in Figure 3a. Point clouds were further processed by ImagewareTM (UGS Corporation, Plano, TX, USA). As the original geometry of the measured parts is unknown, we did the analysis applying a best-case scenario. For this, we considered as much as possible of the spherical cap as un-worn and reasonable, and derived the referencing surface by fitting a sphere to these areas. Normal distances between each point and the fitted surface were derived (illustrated in Figure 3b,c) and further processed to determine the volumetric material loss. The latter was done using Matlab (Version 7.10.0, The MathWorks, Inc., Natick, MA, USA).

**Figure 3.** (**a**) Piston head (cut from the piston stem) with point cloud measured by CMM. (**b**) Color plot of linear wear, i.e., normal deviation between measured data and referencing surface. Gray areas refer to linear wear < 3 μm, which can be considered as unworn. (**c**) Overlay of linear wear plot and piston head image illustrates the unworn and most heavily abraded areas, respectively.

#### *2.4. Statistical Analysis*

Correlations of wear score and volumetric material loss, respectively, and time in vivo, bone resections length, patient weight and inlay degradation score, were assessed by Spearman's rank order-correlation. The correlation of derived coupling wear scores from two independent investigators was assessed by the Pearson correlation coefficient. Differences in mean values ± standard deviation were tested for statistical significance by student's t-test on unpaired samples with a level of significance of *p* ≤ 0.05. Statistical analyses were carried out applying SPSS software (IBM Corp., Version 22.0., Armonk, NY, USA).

#### **3. Results**

#### *3.1. Failure Modes of Retrieved Implants*

Figure 4 represents the classification of the reasons for revision according to Henderson et al. [6]. 52% of the implants were revised because of infection (Henderson type IV). Aseptic loosening (Henderson type II) caused revision in 27% of the cases. Structural failure (Henderson type III), like breakage or joint instability due to worn couplings, occurred in 18% of the cases, while 2% were revised because of soft tissue complications (Henderson type I). No revisions were necessary because of local recurrences (Henderson type V). We found one case of piston breakage and four cases of sleeve breakage (see Figure 5a) in our collective. Three out of four explants with broken sleeves were revised for aseptic loosening. In one case, the piston was bent (see Figure 5b).

**Figure 4.** Distribution of failure modes according to the Henderson classification in our collective of 44 retrievals.

**Figure 5.** Damage occurring at couplings: (**a**) broken coupling sleeve, (**b**) bent piston with area of abrasive wear at piston stem (encirculated area, arrow), (**c**) piston head with zones of planar abrasion, the arrows mark the created edge between the worn and the unworn zones, (**d**) piston without noticeable planar abrasion.

#### *3.2. Evaluation of Wear by Damage Score*

Almost all retrieved MUTARS® coupling mechanisms exhibit abrasive wear to some extent. The outside of the sleeves shows abrasion in the form of highly polished areas. Wear at the piston appears in the form of brush marks at the stem and head, as demonstrated in Figure 5. At the heads, abraded zones can be easily distinguished from unworn areas, as enough material is lost in the articulating zone to create an edge that separates the worn from the unworn zones at the originally semispherical surface (Figure 5c). According to the evaluation by damage score, 43% (n = 19) of the couplings were rated with a score of two or higher; i.e., they showed clear signs of planar abrasion. The mean wear score was 1.79 ± 0.62. The score increased with time in vivo (see Figure 6), with a Spearman's rank-order correlation of rs = 0.65, *p* < 0.01.

**Figure 6.** Coupling wear score increases with time in vivo.

We analyzed the correlation between the wear score of the coupling and patient weight, bone resection length and inlay degradation (scatter plots are displayed in Figure 7). Patient weight or bone resection length did not correlate with the damage score of the coupling (rs = −0.02, *p* = 0.9 and rs = −0.32, *p* = 0.06, respectively). The coupling wear showed a trend to increase with the degradation of the polyethylene inlay, as can be seen in Figure 7c; the Spearman's rank-order correlation coefficient rs amounted to 0.49 (*p* < 0.01).

**Figure 7.** Scatter plots of wear score over (**a**) patient weight, (**b**) bone resection length and (**c**) degradation score of the polyethylene inlays.

Assessment of coupling wear by the two investigators showed a high correlation (Pearson r = 0.89). The mean difference in the coupling wear score of both investigators referred to 0.22 ± 0.19, while the median difference referred to 0.13, the minimum to 0, and the maximum to 0.75.

#### *3.3. Quantification of Wear by Coordinate Measurements*

Quantitative wear analysis of 20 retrievals estimated the volumetric material loss of the articulating surfaces of the coupling to range from 0.5 mm<sup>3</sup> to 30 mm3. The total mean value accounted to 7.9 <sup>±</sup> 8.1 mm3. The mean wear rate was 7.8 <sup>±</sup> 8.6 mm3/year. Volumetric wear increased with time in vivo (Figure 8a), with a Spearman's-rank order correlation of rs = 0.69, *p* < 0.01. Wear score and volumetric material loss are correlated, as illustrated in Figure 8b (rs = 0.867, *p* < 0.01).

**Figure 8.** (**a**) Volumetric material loss increases with time in vivo. (**b**) Relationship between wear score and volumetric material loss.

Cases retrieved for joint infection showed 3.5 <sup>±</sup> 3.2 mm3 of cumulated material loss, while the material loss of aseptic cases amounted to 13.2 <sup>±</sup> 9.2 mm3. This can be attributed to the significantly shorter in-vivo periods of septic revisions, as shown in Table 2. Wear rates, i.e., volumetric material loss per time in vivo of both groups, are similar.


**Table 2.** Total wear volume, time in vivo and wear rates of MoM couplings, subclassified into septic and aseptic revisions.

#### **4. Discussion**

The study investigated damage at a MoM rotating hinge mechanism due to abrasive wear with two different techniques: A semi-quantitative examination by which a coupling wear score is visually determined, and a CMM-based analysis that estimates actual volumetric wear. Quantitative measures on wear volume showed a relevant mean wear rate of 7.8 mm3/year. Both applied methods showed a clear relation between the extent of wear at the coupling and the time of in vivo service.

From Figure 6 it appears that the older designs (A and B) exhibited a better in vivo performance, because they show the longest periods of implantation, and the inclination of the wear score over in vivo service time seems to increase in the order of the design A, B, C and D. The shorter in vivo service time is related to the more recent implantations of the newer designs. Design D is implanted since 2015, design C was implanted between 2011 and 2014, design B between 2010 and 2012, and

the retrievals with design A were implanted between 1997 and 2011. The available data, however, does not allow us to compare the long-term performance of designs B, C and D with the performance of design A, because there are no retrievals with coupling designs B, C or D that were implanted for more than 80 months in the investigated retrieval collective. Also, we do not know the number of well working implants that are currently implanted. A possible explanation for the difference in inclination could be that the highest wear rate occurs in the beginning of the in-vivo service. However, the quantitative analysis rather showed a linear relation, while the semi-quantitative correlation could also be interpreted as a logarithmic relation. We believe that this is mainly related to the fact that the semi-quantitative evaluation grades the extent of damage in finer steps in the beginning of the abrasive process. Brush marks at the piston head can be seen and were rated with a score of 1, while the actual volumetric material loss determined by the CMM evaluation can be still close to zero. As soon as the abrasive process led to the formation of the distinct edge between worn and unworn areas, the piston heads were rated with a score of 3, independent from the severity of the created edge.

The semi-quantitative analysis also rates those zones of the coupling that cannot be assessed by coordinate measurements, such as the outside of the sleeve. Here, abrasion is noticed visually in the form of areas with a highly polished appearance. Another advantage of the semi-quantitative wear analysis is its non-destructiveness. Coordinate measurements as quantitative alternative require dissociation of lid and bushing of the sleeve and of the piston head from stem.

It has been shown in simulator and retrieval studies, that wear and corrosion of metallic parts does occur, even in conventional, i.e., minimally stabilized, total knee arthroplasty. Considerable levels of metal ions are released during cyclic wear simulation [20]. Clinically, this leads to elevated metal ion levels in the patient's blood [21–23]. In addition, the released metal ions accumulate in tissue surrounding the joint, and can lead to adverse local tissue reactions like metallosis [19,24]. This in turn can lead to aseptic loosening, which is always followed by a revision operation. However, quantitative data on metal wear from total knee arthroplasty is rare. Kretzer et al. report the total release of Co- Crand Mo-ions to be about 2.5 mg after 5 million cycles [20]. The wear rate of 7.8 mm3/year determined for the MoM couplings would correspond to a metal release into the human body of about 65 mg/year, which would be at least about 50 times the metal release from the articulating parts. Generally, clinical problems like adverse reactions to metallic debris (ARMD) or pseudotumor formation related to wear of CoCr-articulating surfaces are well known from MoM total hip arthroplasty (THA) [15,16,25]. Volumetric wear of retrieved MoM THAs has been examined in several studies. Lord et al. analyzed 22 retrieved cup-head MoM pairs, whereof the majority of patients had been revised for ARMD. The median of the determined wear rates was 9.4 mm3/year [25]. Glyn-Jones et al. found a significantly increased mean volumetric wear rate of 3.3 and 2.5 mm3/year for femoral heads and acetabular cups, respectively, in retrievals revised for pseudotumors compared to a no pseudotumor control group [26].

Grammatopoulos et al. confirmed these findings, which determined the total mean wear rate in the pseudotumor group to be 5.5 mm3/year, whereas the control group exhibited 0.4 mm3/year [16]. Gascoyne et al. reported a total median rate for metal loss from joint articulation and head-stem taper of 1.5 mm3/year in a retrieval study on 24 MoM hip replacements of a single design, with 16 of the 24 patients revised for MoM-related reasons, such as adverse reaction to metallic debris (ARMD) or high blood metal ion levels. The mean wear rate from articulating parts only referred to 4.6 mm3/year ranging from 0.01 to 66.4 mm3/year [27]. The mean amount of metallic wear of about 7.8 mm3/year created at MoM couplings in the MUTARS prosthesis is thus in the same order of magnitude than wear created from MoM hip TEPs that have been revised for metal-related causes. It is very likely that the debris created in the rotating-hinge system leads to similar metal-related problems, as it has been reported for failing MoM artificial hip joints. This hypothesis is supported by several studies that show considerably elevated metal ion levels related to hinged TKA: Laitinen et al. reported elevated Co- and Cr-ion levels of > 5 ppb in whole-blood samples of 19/22 patients with a MUTARS® prosthesis applying a MoM coupling. In the control group with a MUTARS® prosthesis using a metal-on-polyetheretherketone (PEEK) coupling, only 1 of 12 patients showed a similarly elevated

Co-ion level [12]. Klasan et al. found > 5 ppb Co- and Cr-ions in the blood serum of 16/23 patients with a MoM hinge knee design [28]. Friesenbichler et al. showed elevated Co- and Cr-ion levels in serum blood for two hinged TKA systems (different from the MUTARS®-prosthesis), but not for a standard rotating hinge TKA [29].

A difficulty in the determination of material loss from retrievals is that the original geometry of the worn parts is unknown. Determining volumetric and linear wear therefore always requires estimation of the original geometry by a referencing surface or volume. The MUTARS® rotating hinges are based on a ball joint, so the articulating faces can be represented by a spherical cap. We chose the referencing spherical face in such a way that as much of the measured articulation area as possible could be considered unworn, so we analyzed the best possible case for the coupling in terms of wear. In this way, resulting wear plots usually showed two unworn areas: The unworn, dorsally-oriented face not in contact with the bushing and the opposing ventrally-oriented face. As the latter actually is part of the joint articulation, it is very likely that it experienced abrasive wear to some extent, too. In addition, we determined the quantitative wear volume just from the articulating surfaces. Visually, we detected abrasion also on other parts of the hinge, such as the piston stem, the lid of the sleeve, the outside of the sleeve and on the rims of the openings in the sleeve. The calculated wear volumes therefore represent rather a minimum level of occurring metal abrasion.

In order to estimate how many revisions were caused by worn or broken bushings, we had a closer look on the reasons for revision: Seven prostheses out of 8 revised for 'structural failure' were explicitly revised for mechanical complications related to the coupling, such as instability due to bushing wear and material breakage. From the 12 cases that were revised for aseptic loosening, three cases showed broken coupling sleeves, and in four cases, metallosis was described in the operation report. Two of the metallosis-annotated cases coincided with the broken coupling sleeves. In the case of aseptic loosening with a diagnosed metallosis, we attributed implant loosening to the generated metallic wear. This would sum up to 12 out of 44 revisions (27%), which were probably related to the wear of the rotating hinge mechanism. In addition, five retrievals exhibited bushing or stem breakage, which results in a fracture rate of 11% in the investigated retrieval collective. About half of the investigated retrievals were revised due to an infection, which we did not consider to be related to metal wear. However, there are studies that suggest that metallic debris might be able to promote joint infections, as the tissue damaged by metallic wear products is an optimal environment for bacterial growth [30,31]. It should be noted that after a bone tumor resection at the proximal tibia or distal femur, which requires knee reconstruction by an endoprosthesis, musculature and soft tissue around the joint must be removed, and are therefore missing to stabilize the knee. The tumor endoprostheses systems are therefore subjected to much higher loads than the less stabilized TKA systems. The complications with the rotating hinge mechanism therefore have their reason at least partially in the underlying disease. However, an improved coupling design might be able to diminish occurring coupling wear and damage to some extent.

The study has several limitations. First of all, it is a retrieval study, meaning that we draw our conclusions only based on explants that clinically failed without any control group. Wear volume was estimated on the basis of a fitted referencing spherical surface, as the actual geometry of the measured parts is unknown. In addition, the spherical fit was done on rather small areas of the measured spherical caps. However, the total wear volume we calculated can be considered a minimum value, as we analyzed the measured parts under the estimation of a best-case scenario. Evaluation of coupling wear by damage score instead of quantitative analysis is subjective, and thus to some extent observer-dependent. Also, it does not scale linearly with the amount of abrasive wear.

#### **5. Conclusions**

This study investigated damage on the rotating hinge of a tumor and revision system (MUTARS®) for total knee replacements. Pronounced wear on the coupling mechanism occurs frequently, and was assessed by two different methods. Applying a scoring system has proven to be a reliable, non-destructive option to rate the degree of wear. It was shown that the extent of damage increases with time in vivo, but not with patient weight or bone resection length. Volumetric wear was estimated using coordinate measurements, and revealed high wear rates of about 8 mm3/year.

**Author Contributions:** Conceptualization, S.J.; methodology, T.B., S.J. and J.P.K.; software, T.B., J.P.K.; validation, T.B.; formal analysis, T.B.; investigation, L.N., L.C. and T.B.; resources, S.J., J.P.K., B.L. and G.O.; data curation, T.B.; writing—original draft preparation, T.B.; writing—review and editing, S.J., J.P.K., G.O., R.B. and B.L.; visualization, T.B.; supervision, B.L., J.P.K. and G.O.; project administration, S.J.; All authors have read and agreed to the published version of the manuscript.

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

**Acknowledgments:** We acknowledge financial support by the Baden-Württemberg Ministry of Science, Research and the Arts and by Ruprecht-Karls-Universität Heidelberg.

**Conflicts of Interest:** The authors declare no conflict of interest related to this study.

#### **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* **Antimicrobial Silver Multilayer Coating for Prevention of Bacterial Colonization of Orthopedic Implants**

**Martin Fabritius 1,\*, Amir Andreas Al-Munajjed 2, Christiane Freytag 3, Henriette Jülke 3, Markus Zehe 4, Thomas Lemarchand <sup>5</sup> , Jacobus J. Arts 6, Detlef Schumann 1, Volker Alt <sup>7</sup> and Katrin Sternberg <sup>1</sup>**


Received: 20 February 2020; Accepted: 12 March 2020; Published: 20 March 2020

**Abstract:** Due to increasing rates of periprosthetic joint infections (PJI), new approaches are needed to minimize the infection risk. The first goal of this study was to modify a well-established infection model to test surface-active antimicrobial systems. The second goal was to evaluate the antimicrobial activity of a silver multilayer (SML) coating. In vitro tests with SML items showed a >4 Log reduction in a proliferation assay and a 2.2 Log reduction in an agar immersion test (7 d). In the in vivo model blank and SML coated K-wires were seeded with ~2 <sup>×</sup> <sup>10</sup><sup>4</sup> CFU of a methicillin-sensitive Staphylococcus epidermidis (MSSE) and inserted into the intramedullary tibial canal of rabbits. After 7 days, the animals were sacrificed and a clinical, microbiological and histological analysis was performed. Microbiology showed a 1.6 Log pathogen reduction on the surface of SML items (*p* = 0.022) and in loosely attached tissue (*p* = 0.012). In the SML group 7 of 12 SML items were completely free of pathogens (cure rate = 58%, *p* = 0.002), while only 1 of 12 blank items were free of pathogens (cure rate = 8%, *p* = 0.110). No silver was detected in the blood or urine of the SML treated animals and only scarcely in the liver or adjacent lymph nodes. In summary, an in vivo infection model to test implants with bacterial pre-incubation was established and the antimicrobial activity of the SML coating was successfully proven.

**Keywords:** periprosthetic joint infections; infection prophylaxis; *Staphylococcus epidermidis*; in vivo osteomyelitis model

#### **1. Introduction**

Periprosthetic joints infection (PJI) is a severe complication for patients undergoing a joint replacement procedure that can lead to a straining month-long treatment process of early implant revision, multiple re-revisions or even an amputation of the infected limb. While infection rates after primary implantation are relatively low (0.5% to 2%) [1–3], they increase dramatically in the case of an

implant revision (>10%) [4,5]. If the implant is revised due to a previous infection, the risk to develop a subsequent infection is even higher (~26%) [6]. Besides the high risk for the patient, PJI is also a tremendous economic burden to the healthcare system. Kasch et al. found that in Germany the direct hospital care costs for the management of a septic revision total knee arthroplasty (TKA) are about twice as high as for an aseptic failure [7]. Other papers conclude that the costs for a septic revision are 3 to 4 times that of an primary implantation [8,9].

Early (<3 month after surgery) or delayed infections (3–24 month after surgery) are caused by exogenic pathogens entering the surgical wound during surgery [4]. While there is a controversial discussion on which pathogens are the most relevant to cause PJI, it is generally agreed that Staphylococci species are predominant [4,10–14]. The main focus of scientific as well as public interest is given to *Staphylococcus aureus* (*S. aureus*) due to its high virulence and an increased awareness of antibiotic resistances that renders methicillin resistant *Staphylococcus aureus* (MRSA) the biggest threat to develop PJI. Nevertheless, there are various papers and clinical case studies that highlight the importance of coagulase-negative Staphylococci (Co-NS) e.g., *Staphylococcus epidermidis* (*S. epi*) to be the predominant pathogens causing PJI [4,11–15].

While early infections are mostly caused by very virulent pathogens like *S. aureus*, *Escherichia Coli or Pseudomonas aeruginosa*, delayed infection are caused by less virulent pathogens like Co-NS or Propionibacterium species [16–19]. Early infections are relatively easy to detect by swelling, increased temperature or pain. In contrast, delayed infections are mostly clinically unobtrusive, with delayed and/or nonspecific signs [17,19,20]. If an infection is detected in less than three weeks after implantation, there is a reasonable chance to treat it with the least invasive treatment option of debridement, antibiotic treatment and implant retention [17]. In the case of delayed or chronic infections, the implant most likely has to be revised. This means that delayed infections caused by Co-NS pathogens like a *Staphylococcus epidermidis* are even more dangerous and harder to treat than early infections caused by *S. aureus* [4,12–14,19,21].

Even though operations are carried out under strict hygiene measures, a perioperative contamination from the air or the patients skin can occur [22–24]. The immune system is well capable to address high loads of over 10<sup>6</sup> pathogens. However, in the presence of a foreign body material like an implant, as little as 100 pathogens suffices to cause a severe infection [25]. This is caused by the biofilm formation of pathogens on the artificial surface, which renders the bacteria practically immune to host immune attacks or antibiotic treatment [22]. Therefore, it is of great interest to prevent the biofilm formation on the implants surface. To minimize the infection risk, the thorough implementation of prevention guidelines is essential [26]. In addition to that, technical solutions that protect the implant surface against bacterial colonization should be developed and transferred into clinical practice.

Many techniques and antimicrobial systems have been reported in literature, ranging from active antibiotic release devices to contact killing surfaces [27–31]. Silver is a long known antimicrobial substance, which is successfully applied in various clinically implemented and currently available implant systems on the market [32].

The MUTARS® tumor prosthesis is galvanically silver coated (m(Ag) = 0.33–2.89 g) and is widely used in Europe, Australia and various Asian countries [33,34]. Hardes et al. Donati et al. and Zajonz et al. reported a successful treatment and a reduced infection rate with the silver coated MUTARS® prosthesis compared to a standard implant [35–38]. The Agluna® technology dopes a titanium surfaces with silver ions by an electrochemical process [39]. In a case control study with 170 patients, Wafa et al. reported lower rates of early PJI when Agluna®-treated tumor implants were used [15]. The PorAg® coating is a dual layer system with a silver base layer (1 μm thick) and a rigid top layer of TiAg20N (0.1 μm thick) [40]. Scoccianti et al. reported the successful use of PorAg® coated tumor prosthesis in 33 patients without negative side-effects like argyria [41]. The AgPROTEX® coating is a Hydroxyapatite/Ag2O system which is applied to metal surfaces by flame spraying (T = 2700 ◦C) [42]. The coating is approved for the use in primary total hip arthroplasty (THA) in Japan [43]. According to

Eto et al. no adverse reactions were detected in the first clinical application of AgPROTEX® coated primary THA implants (m(Ag) = 1.9 to 2.9 mg) with 20 patients [44]. However, all these silver-based systems are only suitable for metal implants and cannot be applied to polymer surfaces. This leaves the Polyethylene (PE) liner unprotected, even though it is known that the PE components are most often affected by PJI and carry the highest bacterial load [45].

The silver multilayer coating (SML) (HyProtectTM, Bio-Gate, Nuernberg, Germany) can be applied to both metal and polymer components. Silver clusters are embedded in a polysiloxane (SiOxCy) matrix and act as a reservoir for the release of silver ions that are anti-microbially active on the coating surface. Therefore, elementary silver itself is not in direct contact with the surrounding bone or tissue. Due to its ultra-thin layer (~90 nm), SML maintains the porosity of nano-structured/porous surfaces intact and does not seal them, which is important for osseointegration. The combination of osteoconductive and osteoinductive biomaterials like calcium phosphate, hydroxyapatite, bisphosphonate and silicates in combination with nanoscale therapeutics like BMP-2 have also shown to support bone regeneration, which proves beneficial for secondary implant stability [46,47].

A paper by Khalilpour et al. reported on various successful tests of the SML coating like the in vitro antimicrobial activity, no cytotoxicity according to ISO 10993-5 and ex vivo antimicrobial activity [48]. In a recent case study, the SML coating was used in a successful knee arthrodesis after recurrent periprosthetic knee infection, and silver levels in the drainage fluid and blood were evaluated. Silver blood concentrations after 48 h remained under the detection limit of 2 ppb, whereas the silver concentrations in the wound drainage fluid reached 170 and 57 ppb 24 and 48 h post-operatively, respectively [49].

In most established in vivo osteomyelitis models, an implant is placed in the tibia medullary canal, and a bacteria suspension is injected afterwards [27,28,42,50–52]. This leads to a localized high concentration of pathogens with no uniform distribution along the tibia canal. In case of drug release systems, this is of minor importance as the released drug is able to target present pathogens in a larger vicinity of the implant. In contrast, surface-active coatings are unable to target these pathogens, which can subsequently grow in the more distant tissues. Once an infection is established, high numbers of pathogens are released into the surrounding tissues, which leads to an overpowering of the surface-active system, and therefore, no antimicrobial activity can be proven with such models.

The objective of this study was (i) to establish a suitable in vivo osteomyelitis model in rabbits and (ii) to evaluate the antimicrobial activity of a silver multilayer coating (SML) under realistic pre-clinical conditions. We hypothesized that the SML coating can significantly reduce the CFU count on the K-wire surface at explantation by a minimum of 1 Log reduction compared to the initial CFU count.

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

#### *2.1. Implant Items*

Gamma-sterilized pure titanium K-wires with a diameter of 2.0 mm (MEDE Technik GmbH, Emmingen-Liptingen, Germany) and a length of 150 mm were used for the study. To provide an inert surface, all K-wires were coated with the Advanced Surface® ceramic multilayer coating (AS®, Aesculap AG, Tuttlingen, Germany) over a length of ~140 mm. [53,54] The test items were coated with the SML coating by Bio-Gate AG (Nuremberg, Germany) (Figure 1).

**Figure 1.** (**a**) Test item: SML-coated AS®/titanium K-wire, (**a'**) shows the two coatings in magnification. SML = bronze, AS® = golden. (**b**) Blank item: AS®-coated titanium K-wire, (**b'**) shows the AS® coating in magnification. The silver color at the blunt end of each K-wire shows the uncoated titanium surface.

#### SML Coating

Test items were coated with the SML coating in a three-step process. In a first step, a SiOxCy base layer was deposited on the respective surface by chemical vapor deposition (CVD). In a second step, silver clusters (~2.7 μg/cm2) were deposited on the base layer by physical vapor deposition (PVD), and in a third step covered with a SiOxCy top layer. This resulted in a coating with a thickness of ~90 nm. The coated items were packed individually, and gamma sterilized (BBF Sterilisationsservice GmbH, Kernen-Rommelshausen, Germany). Previous publications give more detail on the SML coating [48,49].

Every SML coating batch was characterized after production by various test methods on planar surface aluminum coated PET foil (dummy items). The chemical structure of the SML coating was analyzed by FTIR spectrometry (Tensor 27; Bruker Optik GmbH, Ettlingen, Germany). The silver content was determined by inductive coupled plasma-optical emission spectroscopy (ICP-OES) according to EN ISO 11885 (Seibersdorf Labor GmbH, Seibersdorf, Austria). The coating thickness was measured by spectral ellipsometry (IFAM Fraunhofer Institut, Bremen, Germany). Each of these tests were performed in triplicate on non-sterilized dummy items. Additionally, titanium test plates were coated simultaneously and analyzed according to ISO 10993-5 to prove non-cytotoxic behavior.

#### *2.2. In Vitro Antimicrobial Activity*

Prior to the in vivo tests various in vitro tests were performed to re-verify the antimicrobial activity of the SML coating on surfaces relevant to orthopedic implants.

#### 2.2.1. QualiScreen® Tests

The in vitro antimicrobial activity of blank and test items was evaluated using a proliferation assay described previously [55,56]. In brief, four replicates of SML coated test items and 4 blank items were incubated in 10% human plasma for 1 h and subsequently washed in 1 × phosphate buffered saline (PBS) for 10 min. Afterwards, the items were incubated in a cell suspension of 5 <sup>×</sup> <sup>10</sup><sup>6</sup> culture forming units (CFU)/mL MSSE (RKI 10-00621) at 37 ◦C for 1 h to allow bacterial cells to adhere to the item surface. Loosely attached bacteria were then removed by rinsing. Subsequently, the remaining cells were incubated for 18 h at 37 ◦C (challenge time). After removal of the test items, 50 μL of tryptic soy broth (TPS) was added to each well. The bacterial growth of the remaining daughter cells was monitored with a microplate reader over a period of 48 h.

#### 2.2.2. Agar Immersion Test

To mimic the in vivo situation, we subsequently tested the antimicrobial activity in an agar immersion test. SML coated items and blank items (AS® coated titanium) were incubated with MSSE (RKI 10-00621) as described in Section 2.3. The seeded K-wires were immersed in a pre-prepared agar slurry (1% Agar and 0.1% TSB) and incubated for 24 and 72 h as well as 7 days. Afterwards, the K-wires were sonicated (3 min) and vortexed (30 s) in PBS to detach adherent bacteria from the item surface. The number of colonies was determined by agar plate count. Each measurement was performed in triplicate.

#### *2.3. In Vivo Study Design*

To determine the antimicrobial activity of the SML an existing in vivo model published by Alt et al. was adapted and an MSSE was used as contaminant [28]. The modification of this model consisted of the us of pre-incubated implants with bacteria in order to assess the effect of the silver coating on the implant surface compared to the inoculation of bacteria into the intramedullary canal after implantation of the K-wire in the referenced model. SML coated and blank K-wires were used as implants. Both implants were loaded with MSSE before implantation (see below). The studies were approved by the German regional authority of Brandenburg (2347-A-4-10-2014) in compliance with the EU principles for animal care.

#### 2.3.1. Pilot Studies

In order to establish and validate the in vivo model, two pilot studies with 6 SPF New Zealand White Rabbits (Envigo, 5800 Venray, The Netherlands) each with a body weight from 3.6 to 4.0 kg where performed. Three animals were treated with a blank K-wire and three animals with an SML coated K-wire. The operation procedure, microbial contamination and subsequent analysis were identical with the one of the main study described below. The silver analysis in organs, blood and urine was only performed in the main study.

#### 2.3.2. Main Study

The study included 27 SPF New Zealand White Rabbits (Envigo, 5800 Venray, The Netherlands) with a body weight from 3.7 to 4.4 kg. As the SML is only surface-active and has no widespread release of antimicrobial substance, we seeded the implants directly with pathogens, instead of injecting bacteria inoculum into the tibia canal. With this approach, we were able to guarantee a contact of pathogens with SML and simultaneously avoid an uncontrolled distribution of bacteria in the tibia canal, which might lead to false negative results.

The K-wires were contaminated in vitro with ~2 <sup>×</sup> 104 colony-forming units (CFUs) of MSSE RKI 10-00621 and implanted into the intramedullary canal of the tibia in the rabbit. The method to determine the bacterial load on the K-wire surface is described in Section 2.4.2.

The test group of twelve animals received a SML coated K-wire (test item), while the control group received blank K-wires (blank item). Both were contaminated with identical loads of bacteria. The remaining three animals were implanted with test items without any microbial load. These three animals served as control to differentiate between implantation process-related and bacterial contamination-related lesions in the histological examination.

#### *2.4. Bacteria*

#### 2.4.1. Bacterial Strains

MSSE RKI 10-00621 was used as a contaminant in this study. RKI 10-00621 is a clinical isolate from a patient with PJI. It was obtained from the national reference center for Staphylococci and Enterococci of the Robert Koch Institute (RKI) in Wernigerode, Germany.

#### 2.4.2. Bacteria Cultivation and Pre-Incubation of Implants

The bacterium was grown in 20% Tryptic soy broth (TSB) for 4 h at 37 ◦C. Prior to implantation, two SML K-wires (per test tube) were contaminated in a test tube over a length of 9 cm by incubation in a bacterial solution of ~1 <sup>×</sup> 106 CFU/mL for 30 to 60 min under dynamic conditions. In pretests, it was proven that the bacterial load level on the item surface can be preserved for this time period. The bacteria were freshly prepared and in logarithmic phase for the controlled contamination of the implant. No biofilm was present at the time of surgery. Subsequently to incubation, non-adherent bacteria were removed by rinsing the test item in PBS for 10 min. One K-wire was implanted into the tibia canal, and the other K-wire was used to determine the pathogen load on the surface at the time of implantation. The pathogens were removed by sonication and vortexing, and the bacterial count was determined by agar plate count.

#### *2.5. Surgery*

Intramuscular anesthesia was performed using ketamine (40 mg kg−<sup>1</sup> body weight) and xylazine (6 mg kg−<sup>1</sup> body weight). Perioperative analgesia was applied subcutaneously (Butorphanol, 0.5 mg kg−<sup>1</sup> body weight).

Surgery was carried out under aseptic conditions according to the model published by Alt et al. and no systemic antibiotics were given [28]. The left hind leg was shaven, fixated, disinfected and draped in sterile covering. The tibia was approached via an infrapatellar skin incision and subsequent preparation of Ligamentum patellae and Tuberositas tibiae. The Tuberositas tibiae was sharply penetrated and the medullary channel opened with a 2.0 mm K-wire. A template K-wire was inserted along the medial cortical bone until the proximal part of Malleolus medialis for channel preparation. After removal of the template, one implant per animal was applied. The implant was inserted into the prefabricated channel at a length of approx. 10 cm and fixated in the proximal part of the Malleolus medialis. The K-wires were contaminated over a length of 9 cm to make sure that no pathogens enter the knee joint. Table 1 lists the group arrangement. Group 1 (test item) and Group 2 (blank item) were seeded with 2 <sup>×</sup> <sup>10</sup><sup>4</sup> CFU MSSE. Group 3 (test item) served as a negative control for histology to determine the lesions induced by the implantation process. The protruding part of the implant was removed, and the implant site was rinsed with Octenisept® and NaCl 0.9%. Afterwards the wound was closed, and post-operative X-ray control was performed (Figure 2).



**Figure 2.** (**a**) Image of SML coated K-wire implanted in intramedullary canal, (**b**) Post-operative X-ray image of the rabbit tibia with the K-wire placement.

After 7 days post operation, the animals were put under general anesthesia (ketamine 40 mg kg−<sup>1</sup> body weight, xylazine 6 mg kg−<sup>1</sup> body weight) prior to euthanasia (T61 intravenously). The implantation site and the surrounding tissue were examined macroscopically, and samples were harvested. For microbiological examination, the implant and bone marrow were collected. The proximal tibia was transversally cut open for implant removal. Bone marrow was extracted from the central tibial bone using sterile instruments. It was manually mixed and halved for microbiological analysis and silver level measurements. For histological evaluation, the tibial bone was segmented into three parts by two transversal cuts. The proximal cut was set distally to the Tuberositas tibiae, and the distal cut was placed proximal to the Malleoli (Figure S1). Both proximal and distal segments were immersed in neutral buffered formalin for subsequent histopathological examination.

#### *2.6. Sample Collection for Silver Measurement*

For pharmacokinetic examinations, samples of whole blood and urine were collected from groups 1 and 2 at defined time points. Samples for determination of zero levels were gathered 7 days prior to surgery at maximum. Subsequently, samples were taken on days 1 and 7 post operation. Approximately 1.5 mL whole blood was taken using a lithium-heparin-tube. Urine samples were harvested as 24-h-samples prior to surgery and on day 1 post operation, whereas on day 7 post operation, urine was favorable collected as punctate. For extended determination of silver levels, situs associated lymph nodes (Lnn. poplitei and Lnn. inguinales), and liver were collected.

#### *2.7. Evaluation Methods*

#### 2.7.1. Clinical Assessment for Infection

Before harvesting of the bone, the knee joint and the surrounding of the implantation site were evaluated for clinical signs of inflammation or swelling. Therefore, areas surrounding the insertion site being directly net to the knee joint, adjacent soft tissue as well as the external structure of the tibia were examined regarding swelling, edema, excessive fluids and pus.

After implant removal and transversal opening of the tibia, the bone marrow was assessed for signs of infection. The quality of the bone marrow was defined as follows. Physiological quality: grey-white–pinkish–light red coloration, not washed-out, no watery phase, formed structure. Inflamed quality: reddish–red–bright red–dark red coloration, washed-out, with aqueous phase and unformed structure. In case of multiple findings, the most severe grade was taken into account for evaluation. In order to standardize qualitative description, macroscopic evaluation was performed in a blinded manner by the same investigator in all cases.

#### 2.7.2. Microbiological Assessment for Infection

At explantation, it was noted that varying amounts of bone marrow or tissue adhered to the K-wires. To avoid a distortion of the CFU count, the implants were gently rinsed in phosphate buffered saline (PBS) to remove the attached tissue (rinsing solution = "rinsing sol.") Subsequently the K-wires were sonicated (3 min) and vortexed (30 s) in PBS to detach adherent bacteria from the item surface ("K-wire"). The bone marrow was collected as described in Section 2.5 and was sonicated in PBS to harvest containing pathogens. Bacterial contamination in the bone marrow was standardized to 1 g ("BMstd"). The three bacterial suspensions (K-wire, rinsing sol., BMstd) were diluted (1:1, 1:10), plated out on agar plates (1000 μL) and incubated over night at 37 ◦C. The number of colonies was determined by visual agar plate count.

#### 2.7.3. Histological Evaluation

Tibia parts (proximal + distal) for histological examination were fixed in 4% phosphate-buffered formaldehyde solution for 72 h. The formalin-fixed bones were cut with a band saw sagittally, para-sagittally and longitudinally into a total of five ~3 mm slabs, according to Figure S1, 1 transverse slab and 2 longitudinal slabs for the proximal segment and 2 longitudinal slabs for the distal segment. The slabs were then demineralized for 1 to max 2 weeks in a solution containing formic acid (5% formic acid in distilled water) and dehydrated and embedded in paraffin. Formalin-fixed paraffin-embedded tissues were cut into 3 μm thick tissue sections and stained with haematoxylin and eosin (HE) or Gram stain according to a method derived from Brown and Brenn [57]. In order to maximize the chances to detect bacterial colonies and inflammatory lesions, step sections at least 50 μm apart were prepared from each block. The presence or absence and if present, an evaluation of the extent of osteomyelitis and bacterial colonies at the site of the contaminated K-wire insertion was carried out at the proximal and distal ends of the tibia. A qualitative ordinal scoring approach (ordinal non-continuous categorical response variables) (usually improperly referred to as "semi-quantitative scoring") was used for the histological evaluation following current animal disease model literature and pharmaceutical development pathological safety/efficacy investigations, as well as Annex E of DIN EN ISO 10993-6:2014-12 [58,59]. It was based on the presence or absence of a microscopic change, finding or lesion and a scoring of extent and magnitude using a relative or absolute scale in a 6-category system: none, minimal, slight (mild), moderate, marked and very marked (severe). No morphological change quantification with digital section and image analysis was carried out, as most often the change was no longer present in a majority of sections.

#### 2.7.4. Silver Levels

For silver level analysis, samples collected according to Section 2.6 were used. The urine samples were centrifuged, and an aliquot of each sample was diluted and measured with inductively coupled plasma mass spectrometry (ICP-MS) according to ÖNORM EN ISO 17294-2 [60]. Blood, liver and lymph nodes were digested in an UV-digestion apparatus by using nitric acid and hydrogen peroxide and analyzed as described above. The detection limit of the ICP-MS depends on the sample quantity, which is analyzed. The following detection limits apply: blood (d0, d1, d7) < 6 μg/kg; urine (d0, d1, d7) < 0.6 μg/kg; liver < 3–6 μg/kg. Due to the varying sample quantity, the detection limit for the lymph nodes fluctuated considerably.

#### 2.7.5. Statistical Analysis

Statistical analysis was performed with Minitab R 17 (Minitab LLC, PA, USA). Due to a significant deviation from a normal distribution of the SML group, a Mann–Whitney test was used to evaluate whether there is a difference between the contamination at the beginning (implantation) and the end (explantation) of the experiment. This was done for the three respective bacterial suspensions (K-wire, rinsing sol., BMstd). We also compared the CFU count of SML coated and blank items after explantation. A Mann–Whitney test was used to evaluate whether there is a difference between the bacterial counts in the bacterial suspensions K-wire and rinsing sol. Differences were considered as significant for *p* < 0.05.

#### **3. Results**

#### *3.1. In Vitro Antimicrobial Activity*

Proliferation assay: The SML coated items and blank items were tested against MSSE (RKI 10-00621). The blank items showed a brutto Onset-OD of 10.3 ± 0.8 h, while the SML items showed a brutto Onset-OD of 21.5 ± 11.0 h. This results in an average netto Onset-OD of 11.2 ± 7.3 h, which relates to a >4 Log reduction [55,56] (Figure 3).

**Figure 3.** Results of the proliferation assay. Brutto Onset-OD time and 95% confidence interval for blank items and SML coated items. The netto Onset-OD of 11.2 h relates to a >4 Log reduction.

Agar immersion test: SML and blank items were challenged with MSSE (RKI 10-00621) and incubated for 24, 72 and 168 h in agar slurry. At the respective times the SML coated items showed a CFU reduction of 1.4 ± 0.2, 1.3 ± 0.3 and 2.2 ± 0.2 Log compared to uncoated blank items (Figure 4).

**Figure 4.** Results of the agar immersion test. Mean CFU count and 95% confidence interval for blank and SML coated K-wires at time point t = 0 h (t0), t = 24 h (t24), t = 72 h (t72) and t = 168 h (t168).

#### *3.2. In Vivo Experiments*

#### 3.2.1. Clinical Assessment

In general, the established infection model was considered to be mild, as septic arthritis of the knee joint was not detected by clinical observation in any case (Figure 5a,b). In 4 of 12 animals of the SML group signs of osteomyelitis were found in the bone marrow. This was marked by an increased red and/or washed-out coloration, as well as an unformed structure of the bone marrow and partial presence of an aqueous phase. In the other 8 cases of the SML group, the bone marrow was evaluated as physiological or "cured" (SML–cured 8/12 = 67%). On the contrary, bone marrow of animals from the blank group showed signs of osteomyelitis in 11 of 12 cases and only one animal was documented as physiologic or cured (Blank–cured 1/12 = 8%) (Figure 5).

**Figure 5.** (**a**,**b**) Exemplary images of the knee joint postmortem of animals treated with (**a**) SML item and (**b**) blank item. (**c**,**d**) Exemplary images of bone marrow after explantation of (**c**) SML item and (**d**) blank item. (**c**) Physiological bone marrow was found in 11 of 12 animals. This equals a cure rate = 8%. (**d**) Fragmented and hemorrhagic aspects indicating osteomyelitis were found in 4 of 12 animals. This equals to a cure rate of 67%.

#### 3.2.2. Microbiological Assessment for Infection

In the microbiological evaluation, 7 of 12 SML coated K-wires were free of pathogens and the remaining 5 K-wires showed a distinct CFU reduction compared to implantation. This equals a cure rate of 58% (*p* = 0.002). The mean pathogen count for the whole 12 SML K-wires was 353 ± 529 CFU. In the control group, only 1 of 12 K-wires was free of pathogens and the CFU count on the whole twelve blank K-wires was 9.282 ± 10.585 CFU (Figure 6). This equals a cure rate of 8% (*p* = 0.110).

**Figure 6.** Individual value plot of CFU at d0 (implantation, ~2 <sup>×</sup> <sup>10</sup><sup>4</sup> CFU) and d7 (explantation). (**a**) SML coated K-wires, mean CFU = 353 ± 529 CFU. This equals a cure rate of 58% (**b**) Blank K-wires, mean CFU = 9.282 ± 10.585 CFU. This equals a cure rate of 8%. The black dots represent the mean value of each group with a 95% confidence interval.

The results of the Mann–Whitney test for the three bacterial suspensions (K-wire, rinsing sol., BMstd) are listed in Table 2. A comparison of the CFU count at implantation and explantation showed a decrease in all bacterial suspensions and the reduction is significant for both blank and SML coated items (*p* < 0.05, Table 2). The CFU reduction was always higher for SML coated K-wires than for blank K-wires.


**Table 2.** Results from the Mann–Whitney test of Log (Implantation) compared to Log (Explantation).

Sol. is solution and BMstd is bone marrow standardized.

Comparing the CFU count at explantation for SML and blank group showed a 1.6 Log reduction for the SML coated items on the K-wire surface (*p* = 0.022) and in the rinsing solution (*p* = 0.012, Table 3). The bone marrow of the SML group also exhibited less pathogens than the blank group (0.5 Log), yet the effect was less pronounced (Figure 7).

**Table 3.** Results from the Mann–Whitney test of Blank and SML items after explantation.


Sol. is solution and BMstd is bone marrow standardized.

**Figure 7.** Mean values and 95% confidence interval of the CFU count of the in vivo study. (Implantation) = CFU count at implantation on the K-wire surface. (K-wire) = Bacterial suspension derived from the K-wire surface at explantation. (rinsing sol.) = Bacterial suspension derived from rising the K-wire surface to remove attached tissue. (BMstd) = Bacterial suspension derived from the bone marrow (normalized to 1 g). Light grey = Blank items, dark grey = SML items.

#### 3.2.3. Histological Evaluation

Histology showed less heterophilic infiltration/pus and fibroplasia in the distal tibia of animals with SML items, compared to the distal tibia of animals with blank items (Figures 8 and 9). Similar degrees of inflammation and associated repair were noted in the proximal tibia of SML group and blank group animals. In general, the induced inflammation was very mild and barely above the one induced by the surgery alone. Very importantly, the new bone formation around the implant was very active in this disease model and comparable for test item and blank item. The negative control group (SML item with no bacterial contamination) showed no suppuration and excellent implant stabilization by means of fibrous connective tissue and recent new bone formation (Figure 10).

**Figure 8.** (**a**–**c**) Histology images of distal tibia of blank item animal. (**a**) There was a higher incidence of mild focal ongoing osteomyelitis in the distal tibia of blank K-wire implanted animals (**b**). Green arrows indicate new fibrous tissue (fibroplasia). Osteomyelitis was focal and showed evidence of several days old pus (black arrows) at periphery of implant, in the bone marrow and more recent exudate of intact and degenerated heterophils (red arrows and (**c**)) in the bone marrow immediately adjacent to the implant imprint (\*), indicating active suppurative inflammation.

**Figure 9.** (**a**–**c**) Histology images of distal tibia of test item animal. No evidence of mild ongoing osteomyelitis along the K-wire in most SML K-wire implanted animal and stabilization was often seen to be more significant in the test item at tip of the K-wire imprint (\*). Integration was by means of fibroplasia (green arrows) and new fibrous bone formation at host-implant interface ((**b**), blue arrows), devoid of any heterophilic infiltration (**c**).

ǻǼȱȮ ȱDZȱȬȱȱ
Ȭ

**Figure 10.** (**a**–**c**): Test item SML K-wire-implanted rabbits with no bacterial contamination, showing lack of suppuration and excellent implant stabilization by means of fibrous connective tissue (green arrows inset (**b**)) and recent new bone formation (see magenta arrows in insets (**b**,**c**)). Very recent new fibrous bone formation at host-implant interface (see short magenta arrows in inset (**b**)), devoid of any heterophilic infiltration, and more mature and older new fibrous bone (see longer magenta arrows in panel (**c**) from adjacent step section in metaphysis region adjacent the epiphysis EP). At the interface of the removed K-wire and the new connective tissue, there is limited red blood cell extravasation or hemorrhage (red arrows in inset (**b**)).

#### 3.2.4. Blood and Urine Analysis

No silver was detected in the blood and urine of all SML group animals at days 0, 1 and 7. In 1 of 12 animals, silver was detected in the liver (5 μg/kg), while in 11 of 12 animals, silver levels were below the detection limit. In Lnn poplitei, silver was detectable in 3 of 12 animals and in Lnn inguinales in 6 of 12 animals while in all other cases silver levels were below the detection limit (Figure 11).

**Figure 11.** Individual value plot of silver levels measured by inductively coupled plasma mass spectrometry (ICP-MS) on day 7 (explantation) in blood, urine, liver and two lymph nodes. <LOD = Silver concentration was below the detection limit; Value = Silver concentration could be measured. The big symbols represent the mean value of each group with a 95% confidence interval.

#### **4. Discussion**

In the development of PJI, the initial step of bacteria adhering to the implant surface is of utmost importance. It initiates the cascade of bacterial proliferation and subsequent biofilm formation, which protects the bacteria from the host defense system. This contributes to inferior results in treatment options, as the susceptibility to antibiotics is dramatically decreased at this location. Therefore, the protection of the implant surface from bacterial colonization is potentially the most important step to prevent PJI.

The objective of this study was (i) to establish a suitable in vivo osteomyelitis model with pre-incubated implants with *Staphyloccocus epidermidis* in rabbits and (ii) to evaluate the antimicrobial activity of a silver multilayer coating (SML) under realistic pre-clinical conditions. Our study has some limitations that need to be taken into account when interpreting the results. Hischebeth et al. showed that MRSE infections are more difficult to cure than MRSA infections [21]. However, in our study, we used MSSE instead of MRSE due to safety aspects and to avoid unnecessary use of MRSE. Aspects such as antibiotic resistance, no differences in several bacterial properties such as proliferation, biofilm development and adherence are to be expected between MSSE and MRSE. The limitation of a relatively short test period of 7 days when compared to the timeframe of delayed infections (3–10 weeks) was accepted mainly for animal-welfare reasons. The test period was sufficient to identify the microbiological significant difference in the infection course. To test for complete clearance of the infection, further research with a longer test interval could be performed.

In literature, various models have been established to determine the antimicrobial activity of antimicrobial coatings most of which use MRSA as contaminant [52]. Recent findings emphasize that in clinical reality its "little brother" methicillin-resistant *Staphyloccocus epidermidis* (MRSE) is even harder to cure and spreads around the globe undetected [21,61]. In our study design, we therefore tested the antimicrobial activity of the SML coating against MSSE, which is associated with a strong biofilm forming capacity [62].

Most of the mentioned animal models are designed to test active release systems that address pathogens in the larger tissue region around the coated implant and therefore could actively treat an infection [27,28,50,51]. For example, Suhardi et al. reported the successful treatment of a PJI with an antibiotic releasing PE [27]. However, this approach has a major drawback for cementless endoprosthesis as they have the necessity of bone ingrowth which is essential to guaranty secondary implant stability. Antimicrobial substances like antibiotics or silver can have a cytotoxic effect on osteoblasts and thus may impair bone ingrowth around the implant [63,64]. Therefore, it is important to minimize the exposure to these substances while still guaranteeing an antimicrobial activity. Our measurements showed no silver in blood and urine and very low silver concentrations in the liver and adjacent lymph nodes, which proves a very limited systemic silver exposure. These findings are in line with the histopathological evaluation, which found very active new bone formation for both test item and blank item.

Another relevant aspect to take into account is the clinical approach currently used to treat PJI. The surgical techniques involve the thorough debridement of infected tissue, systemic antibiotic therapy and wound irrigation to reduce the pathogen load in the wound as much as possible. In clinical practice, an implant is never placed in an infected wound but in an environment with as little pathogens as possible. However, a contamination of the implants' surface with a few pathogens from whichever source can always occur, and as little as 100 bacteria are enough to induce a periprosthetic infection [25]. An antimicrobial coating, which is surface active and capable to kill the bacteria trying to adhere to the implant surface adds an additional safety characteristic to the implant while avoiding the potential negative effects of high dose drug release. The remaining planktonic pathogens in the soft tissue can be attacked by the host immune system.

To address the aspects mentioned above, we deviated from established animal models and adapted ours to enable the testing of surface-active coatings without typical release properties and drug efficacy of antibiotics.

The SML coating tested in this study showed a distinct activity against MSSE with a statistically significant reduction of pathogens on the implants surface. Due to the missing or mild infection signs in the clinical observation and histology as well as no systemic effects, the model was considered to be mild, with only a minimal to mild impact on the animal health. The decrease in CFU count of both blank and SML item shows that the immune system of the animals was able to fight the pathogen concentration of ~2 <sup>×</sup> 104 CFU of MSSE. When compared with the initial pathogen dose, a statistically significant CFU reduction of over 3 Log after explantation on the surface of SML coated items and an over 2 Log reduction in the corresponding rinsing solution and bone marrow was detected. This also significant CFU reduction on the AS® coated blank items was unexpected. Recent papers report similar findings, yet to date, there is no explanation for this effect. This will need to be evaluated further [65].

Comparing the CFU count after explantation, the SML coated items showed a 1.6 Log reduction on both the K-wire surface and the rinsing solution compared to the blank items. In the SML group, 7 of 12 test items were completely free of pathogens, which equals to a cure rate of 58% (*p* = 0.002). On the contrary, only 1 of 12 blank items were free of pathogens, which equals a cure rate of 8% ((*p* = 0.110). These results clearly prove the antimicrobial activity of the SML coating. In addition, the results from the main study are backed up by the results from the two pilot studies, which also showed a clear reduction of pathogens. As the difference in CFU count of blank and SML items was far less pronounced in the bone marrow, this indicates the localized antimicrobial effect of the SML coating.

As no reference values exists that indicate how high the in vivo CFU reduction of an antimicrobial system has to be to prove a certain and clinically relevant antimicrobial activity, it might be hard for the reader to interpret the measured 1.6 Log CFU reduction. To put this into perspective, the clinical reality is to be considered in this context. Each K-wire was implanted with a load of ~20,000 CFU on the K-wire surface. This was done to induce an infection in the animals and subsequently detect a difference in CFU count between SML and blank items. Lower contamination doses were quickly eradicated by the animal's immune system, and therefore, no effect of the antimicrobial coating could be detected. A contamination of over 20,000 CFU per K-wire equals over 3500 CFUs per cm2, which is several orders of magnitud higher than the contamination of a hospital toilet door handle (7.97 <sup>±</sup> 0.68 CFU/cm2) or a hospital washroom floor (20 CFU/cm2) [66,67]. In an orthopedic setting with adequate hygiene

management, such high pathogen loads will enter the wound neither through the orthopedic implant, which is delivered sterile, nor through the patient's skin, which is disinfected and draped before the operation.

At explantation, the mean pathogen count on the surface of a blank K-wire was 9.282 ± 10.585 CFU, and only one of twelve K-wires was pathogen free. On the contrary, on SML coated items, the mean pathogen count was 353 ± 529 CFU, and seven of twelve K-wires were pathogen free, which shows a decrease in pathogen count by the SML coating of roughly 9000 CFU. This is over 3 times the pathogen count that is present at the surface of an PE liner, which is explanted due to PJI (2768 CFUs) [45]. Therefore, the 1.6 Log reduction of pathogens when SML is used is considered to be a substantial safety benefit.

To our knowledge, this is the first established mild osteomyelitis model that works with pre-incubated implants. In this model, we successfully proved an antimicrobial activity of the SML coating.

To date, various orthopedic implants with silver-based surface coatings are approved and used successfully in clinical practice. Within the European Union, to date, these systems are limited to large revision and tumor prosthesis and not used on primary or standard revision implants for total joint arthroplasty. However, as they can only be applied to metal, the PE liner, which is most prone to infection, remains unprotected. A technology that can be applied to both metal and polymer surfaces and that has been proven to reduce adherent bacterial load broadens the possibilities to fight PJI. The SML coating is not designed to be wear resistant, nor has it been tested for wear under continuous loads in this study. In areas with high friction like the joint articulation, the SML coating could be sheared off. It is therefore recommended that for orthopedic implants, the majority of the surfaces be coated to protect against bacterial colonization, but that areas exposed to high wear remain uncoated. Further studies should be conducted to address the osseointegration and wear behavior of the SML coating.

#### **5. Conclusions**

There is a great need for new infection prophylaxis systems that can improve the safety of patients undergoing joint replacement surgery. We have successfully (i) established a mild osteomyelitis model in rabbits with pre-incubated implants and (ii) demonstrated excellent antimicrobial activity of the presented SML coating. The performed in vitro and in vivo experiments both showed a statistically significant CFU reduction in a clinically relevant scale. The local and systemic silver release remained close to detection limits. Its broad applicability renders the SML coating a promising candidate as an infection prophylaxis system for orthopedic applications.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1996-1944/13/6/1415/s1, Figure S1: Schematics of band saw cutting of un-demineralized specimens for block preparation. From each of the 5 blocks depicted, 2 section levels spaced at least 50 μm were cut.

**Author Contributions:** Conceptualization, M.F., A.A.A.-M., C.F., H.J., M.Z. and K.S.; investigation, M.F., A.A.A.-M., C.F., H.J., M.Z. and T.L.; methodology, M.F., A.A.A.-M., C.F., H.J. and M.Z.; visualization M.F., C.F., H.J. and T.L.; supervision D.S., K.S; writing—original draft preparation, M.F. and A.A.A.-M.; writing—review and editing, J.J.A., V.A. and K.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** The study was funded by Aesculap AG (Tuttlingen, Germany) and supported by Bio-Gate AG (Nuremberg, Germany).

**Conflicts of Interest:** Three of the authors (M.F., D.S., K.S.) are employees of Aesculap AG Tuttlingen, a manufacturer of orthopedic implants. One author (A.A.A.-M.) is employee of Bio-Gate AG Nuremberg, a manufacturer of antimicrobial materials. Four authors (C.F., H.J., M.Z., T.L.) are employees of the respective labs (C.F., H.J., FreyTox Gmbh Herzberg; M.Z., QualityLabs Nuremberg; T.L. TPL Path Labs Freiburg). Two of the authors (J.J.A., V.A.) are scientific advisors in Aesculap R&D projects. V.A. is member of the Supervisory Board of Bio-Gate AG (Nuremberg, Germany).

#### **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* **Comparison of Di**ff**erent Locking Mechanisms in Total Hip Arthroplasty: Relative Motion between Cup and Inlay**

#### **Sebastian Jaeger 1,\*,**†**, Maximilian Uhler 1,\*,**† **, Stefan Schroeder <sup>1</sup> , Nicholas A. Beckmann 2,3 and Ste**ff**en Braun <sup>1</sup>**


Received: 21 February 2020; Accepted: 16 March 2020; Published: 19 March 2020

**Abstract:** The resulting inflammatory reaction to polyethylene (PE) wear debris, which may result in osteolysis, is still considered to be a main reason for aseptic loosening. In addition to the primary wear in hip joint replacements caused by head-insert articulation, relative motions between the PE liner and the metal cup may cause additional wear. In order to limit this motion, various locking mechanisms were used. We investigated three different locking mechanisms (Aesculap, DePuy, and Zimmer Biomet) to address the resulting relative motion between the acetabular cup and PE liner and the maximum disassembly force. A standardized setting with increasing load levels was used in combination with optically based three-dimensional measurements. In addition the maximum disassembly forces were evaluated according to the ASTM F1820-13 standard. Our data showed significant differences between the groups, with a maximum relative motion at the maximum load level (3.5 kN) of 86.5 ± 32.7 μm. The maximum axial disassembly force was 473.8 ± 94.6 N. The in vitro study showed that various locking mechanisms may influence cup-inlay stability.

**Keywords:** cup-inlay stability; total hip arthroplasty; disassembly forces; relative motion

#### **1. Introduction**

Total hip arthroplasty (THA) is one of the most successful orthopedic procedures in joint replacement [1,2]. The THA shows survival rates between 82% and 96% after 10 years, depending on the age of the patient and their physical functions [3]. However, the resulting inflammatory reaction to polyethylene (PE) wear debris, which may result in osteolysis [4], is still considered to be a main reason for aseptic loosening [5–7]. Therefore, wear debris production is thought to be the main factor limiting long-term survival of THA. It is known that not only can primary wear processes at the articulation sliding surfaces of hip cup and femoral head play an essential role, but in addition, secondary PE wear also takes also place at the backside of the PE liners (backside wear) due to relative motion between cup and insert [8–11]. A non-conforming fit between cup and PE liner, in combination with increased relative motion, could lead to burnishing, gouging, scratching or third-body wear on the PE backside surface [12]. In order to limit the motion between the metal shell and the liner, various locking mechanisms are used. It is assumed that a firm connection of the components will reduce the relative motion and the particle migration inside the cup system [13,14]. Another rather rare reason for revision is the dislocation between cup and PE liner, which presumably can be attributed to the acetabular locking mechanism.

The aim of this study was to evaluate specific locking mechanisms regarding the resulting relative motion between the acetabular cup and the PE liner. In addition, the maximum dissembling forces were investigated. For this purpose, three implant designs were tested, to determine if different types of locking mechanisms affect the magnitude of relative motion and dissembling force.

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

#### *2.1. Acetabular Hip Cup Systems*

Three acetabular titanium hip cup designs, with different inner designs to fixate the PE liner, were investigated (Figure 1): (A) Allofit®-S Alloclassic® with a cross-linked Durasul®-PE liner (Zimmer Biomet, Warsaw, IN, USA), (B) Pinnacle®-Multihole combinates with a cross-linked Marathon®-PE liner (DePuy Synthes, Warsaw, IN, USA), (C) Plasmafit® Plus7 with a conventional ultra-high-molecular-weight polyethylene (UHMWPE) liner (Aesculap, Tuttlingen, BW, Germany). All hip cup systems were non-cemented with an outer cup diameter of 52 mm and a corresponding 32 mm PE liner. In each group, six acetabular hip cups with corresponding PE liners were used.

**Figure 1.** Acetabular titanium hip cup components. (**a**) Zimmer Biomet, (**b**) DePuy Synthes, (**c**) Aesculap.

The PE liner of the hip cup system Allofit®-S Alloclassic® is fixed by a circular press fit mechanism. In addition, two spikes in the inner polar region of the cup on which the PE liner is inserted improve rotational stability. The inner anchoring surface of the cup is spherical, and has a smooth surface (Ra = 1 μm; Rz = 6–8 μm) [15].

The anchoring of the PE liner of the hip cup system Pinnacle®-Multihole is realised by a central dome region and a taper lock mechanism. Additionally, 12 grooves are placed at the rim of the shell. Corresponding to the grooves, six tabs on the liner were found to improve rotational stability. The inner anchoring surface of the cup showed values of Ra = 0.2 μm and Rz = 1 μm.

The PE liner of the hip cup system Plasmafit® Plus7 is fixed by a conical locking mechanism. The cup shows a rough inner surface (Ra = 4 μm, Rz = 25 μm) [9,15].

An overview of the locking mechanisms and the surface specifications of the different cup systems are shown in Figure 2.

**Figure 2.** Locking mechanisms and surface specifications of the different cup systems. (**a**) Zimmer Biomet, (**b**) DePuy Synthes, (**c**) Aesculap.

#### *2.2. Test Setup*

In order to carry out a standardized PE liner stability test, the cups were fixed according to Braun et al. in polyurethane (RenCast FC 53 A/B, Gößl & Pfaff GmbH, Karlskron, Germany) at an angle of 30 degrees to the vertical load axis [9]. A loading scenario was implemented using a uniaxial servo hydraulic testing machine (Bosch Rexroth, Lohr am Main, Germany). The components were loaded with an increasing dynamic load at a frequency of 1 Hz and a sinus shaped wave form. The initial load level started at 0.5 kN and increased incrementally with each load level by 1 kN, up to the maximum load of 3.5 kN. The cyclic load at each load level was maintained over 1000 cycles. The fixed acetabular cups were integrated into the testing machine and mounted on a rocker (Figure 3b). The rocker was placed on a free linear bearing to prevent transverse forces [9,15].

A rigid connection was generated between femoral head and PE liner. This, in combination with a 2 mm offset (Figure 3b L) between the centre of the implant and the rotational axis (Figure 3b green) of the rocker, initiated torque in the cup-system (Figure 3b) [9,15]. In combination with the axial compression a theoretically torque of 1, 3, 5 and 7 Nm were applied.

**Figure 3.** (**a**) Arrangement of the optical marker; (**b**) Relative motion measurement setup; (**c**) Camera coordinate system.

#### *2.3. Determination of Relative Motion*

The determination of relative motion was performed by placing optical markers (uncoded passive white markers, diameter 0.8 mm, GOM Item Number: 21874; GOM Gmbh, Braunschweig, Germany) along the rim of the hip cup and PE liner (Figure 3a). These markers were detected using a stereo camera system and a triangulation algorithm to calculate the 3D marker position (x-, y-, and z-axes). Based on the measured marker motion in the respective spatial direction (x, y and z direction), the resulting vector was calculated using the equation XYZ = √ X2 + Y<sup>2</sup> + Z2 . In a next step, the cup and PE liner markers were separated and the resultant relative motion XYZ between both components was calculated. This calculation was carried out for every measurement time point. In a further step,

the maximum of the resulting relative motion (XYZ) was determined for each individual load cycle in order to calculate the mean value of the maximum resulting relative motion for each load level (0.5, 1.5, 2.5 and 3.5 kN) [16]. The coordinate system was defined as shown in Figure 3c. The relative motion between the components was measured at cycles 10, 100, 300, 500, 800 and 995 for each load level.

#### *2.4. Disassembly Test*

In addition, the maximum disassembly force of the PE Liner was determined according to the ASTM F1820-13 standard. The PE liners of the investigated cup designs were assembled in the shell with a mean peak load of 2002.2 N ± 1.1 N using a material testing machine (Zwick Roell, Z005, Ulm, Germany). Therefore, a displacement control rate of 0.04 mm/s was used. The force was applied by a femoral head coincident with the polar axis of the liner. The assembled liner shell constructs were placed in a fixture frame, with continuous support of the cup as shown in Figure 4. The axial force was applied through the centre hole of the cup with a 6 mm rod and a rate of 51 mm/min. The maximum disassembly force was recorded [17].

**Figure 4.** Test Setup Disassembly Test.

#### *2.5. Statistical Analysis*

A repeated measures analysis of variance (ANOVA) was conducted to test for significant differences in relative motion depending on the load level for all investigated cup designs. Additionally, we conducted a one-way ANOVA to assess the effect of cup design on the maximum pull out force and relative motion. To identify which particular differences between pairs of means were significant, a post hoc analysis was performed. We used a Bonferroni Test as post hoc analysis to explore differences between the three group means while controlling the experiment wise error rate. Pre-analysis, the normal distribution of the data was evaluated using a Shapiro–Wilk-test and the homogeneity of variance was verified using the Levene-test. The results allowed for the use of the ANOVA test. For the repeated measures ANOVA, a Greenhouse–Geisser adjustment was used to correct for violations of sphericity.

Additionally, the data were evaluated descriptively using the arithmetic mean, standard deviation, minimum and maximum. The data were analysed using SPSS 25 (IBM, Armonk, NY, USA) with a significance level of *p* < 0.05.

#### **3. Results**

#### *3.1. Relative Motion*

The mean resulting maximum relative motion in XYZ-direction for each load level and all investigated cup designs are shown in Table 1 and Figure 5.

**Table 1.** Mean Relative Motion XYZ ± (μm).

**Figure 5.** Resulting maximum relative motion, \* showed significant difference.

For the load level of 0.5 kN the Bonferroni-adjusted post-hoc analysis revealed a significant difference (*p* = 0.001) for mean resulting maximum relative motion of the Zimmer Biomet and the Aesculap group (52.67, 95%-CI [82.85, 22.49]) and also between the Zimmer Biomet and DePuy group (*p* = 0.004) (43.61, 95%-CI [13.43, 73.79]). There was no significant difference (*p* = 1.0) for the relative motion between the Aesculap and DePuy groups. The load level of 1.5 kN showed a significant difference (*p* < 0.001) for the relative motion between the Zimmer Biomet and the Aesculap group (61.75, 95%-CI [28.31, 95.18]). There was no significant difference between the Zimmer Biomet and DePuy groups (*p* = 0.082) or between the Aesculap and DePuy groups (*p* = 0.070). The load level of 2.5 kN showed a significant difference (*p* = 0.001) between the Zimmer Biomet and Aesculap groups (56.85, 95%-CI [22.52, 91.18]) and between the Aesculap and DePuy groups (*p* = 0.017) (41.23, 95%-CI [6.90, 75.56]). There was no statistically significant difference for the relative motion between the Zimmer Biomet and DePuy groups (*p* = 0.718). For the load level of 3.5 kN, the Bonferroni-adjusted post-hoc analysis revealed a significant difference (*p* = 0.006) for the mean resulting maximum relative motion between the Zimmer Biomet and Aesculap groups (48.28, 95%-CI [13.41, 83.14]) and between the Aesculap and DePuy groups (*p* = 0.014) (42.92, 95%-CI [8.06, 77.79]). There was no statistically significant difference for the relative motion between the Zimmer Biomet and DePuy groups (*p* = 1.0). The repeated measures ANOVA with a Greenhouse–Geisser correction determined that the load levels showed a significant influence on the relative motion, *F* (1.264, 18.964) = 309.886, *p* < 0.001.

#### *3.2. Disassembly Test*

The mean maximum axial disassembling forces showed the highest values for the Aesculap group at 473.7 ± 94.6 N, followed by the Zimmer Biomet group at 294.8 ± 48.2 N and the DePuy group at 146.8 ± 49.8 N (Figure 6). The mean maximum force differed statistically significantly for the investigated cup designs, *F* (2, 15) = 29.25, *p* < 0.001.

The Bonferroni-adjusted post-hoc analysis revealed a significant difference for the mean maximum disassembling force between the Aesculap and Zimmer Biomet groups (*p* = 0.002), (178.90, 95%-CI [63.60, 294.20]) and between the Aesculap and DePuy groups (*p* < 0.001), (326.90, 95%-CI [211.60, 442.20]). There was also a significant difference between the Zimmer Biomet and DePuy groups (*p* = 0.011), (148.00, 95%-CI [32.70, 263.30]).

**Figure 6.** Resulting maximum axial disassembling forces.

#### **4. Discussion**

Osteolysis-induced bone destruction and the following aseptic loosening is one of the major complications of prosthetic hip replacement [5–7]. PE wear particles can cause osteolysis due to an inflammatory reaction. Existing publications proved that, additionally to the PE wear due to the main articulation, wear could be a consequence of the backside wear of the convex side of the PE liner [18]. The design and material could have an impact on the damage of the backside of polyethylene in modern modular acetabular cups [19]. The motion between the PE liner and the cup of a modular acetabular component exacerbated the polyethylene wear as well. This kind of micromotion, respectively, relative motion between cup and liner, is influenced by several factors. Factors could include a reduced conformity, screw holes, and different locking mechanisms and geometries [12,14,20,21].

Kurtz et al. determined the relative motion between cup and liner using a three-dimensional finite element model [12]. The focus was on considering the influence of nonconforming between liner and metal cup. Furthermore, the effects of rim and equatorial restraints were considered. The result of the study stated that backside nonconformity and locking restraints have a relevant influence on backside relative motion. Systems with a conforming cup exhibited an axial motion between 8.5 and 12.8 μm [12]. Nonconforming models offered up to 63% higher motion [12]. An in vitro study was performed by Williams et al. to determine the fixation of PE liners in metal cups [14]. Therefore, the rotational and axial motion was measured using linearly variable differential transducers. Axial loads from 0.272 kN to 2.720 kN and torsional loads from ± 7.5 Nm over 10 million cycles were applied. In the evaluation of the results, it could be derived that the rim, rotational and dome micromotion decreased as the cycles increased. Additionally an influence of the different locking mechanisms was identified. Williams et al. achieved micromotions between 3.35 ± 2.50 μm and 164.7 ± 112.5 μm after 1 million cycles [14].

Our data also showed considerable differences between cup and PE liner. Regardless of the implant design and loading situation, the range was between 6.2 ± 1.1 μm and 86.5 ± 32.7 μm. The Zimmer Biomet hip cup system with a circular snap-fit mechanism, in combination with two spikes in the inner polar region of the cup, showed relative motion between 58.8 ± 32.6 μm and 86.5 ± 32.7 μm (0.5 and 3.5 kN). The DePuy system with a central dome region and a taper lock mechanism plus grooves at the rim of the shell showed relative motion between 15.2 ± 8.3 μm and 81.1 ± 18.5 μm (0.5 and 3.5 kN). For the Aesculap system with a conical locking mechanism and a locking free contact with the base of the cup we found relative motion between 6.2 ± 1.1 μm and 38.2 ± 9.9 μm (0.5 and 3.5 kN).

For the initial load level of 0.5 kN, the DePuy and the Aesculap cups showed no significant differences (*p* = 1.0). The Zimmer Biomet cup showed significantly higher values compared to the Aesculap and DePuy cups (*p* = 0.001, *p* = 0.004). At the highest load level of 3.5 kN, the Zimmer Biomet group and DePuy system no longer showed any significant difference (*p* = 1.0). However, the Aesculap system still showed significantly lower values compared to the Zimmer Biomet and DePuy cups (*p* = 0.006, *p* = 0.014). The locking mechanism of the Aesculap system showed the lowest relative motion between cup and PE liner for the investigated load levels compared to the Zimmer Biomet and DePuy cup. A retrieval study conducted by Wasielewski et al. evaluated the backside of 55 polyethylene liners [22]. At this juncture, a distinction was made between micromotion caused by rotational instability of the insert in the shell and micromotion induced through elastic or plastic deformation. A total of 52 of the 55 inserts showed polyethylene wear on the backside in varying degrees. The most severe wear occurred in the posterior-superior and anterior-superior areas. In addition, five of six patients with acetabular osteolysis showed a high grade of backside wear, which was mainly attributed to elastic or plastic deformation related micromotion [22].

Another aspect besides the relative motion between cup and PE liner was the maximum disassembly force with regard to the risk of dislocation. Therefore, the maximum disassembly force was determined according to ASTM F1820-13 standard [17]. The maximum disassembling force showed the highest values for the Aesculap group with 473.7 ± 94.6 N, followed by the Zimmer Biomet with 294.8 ± 48.2 N and DePuy with 146.8 ± 49.8 N. In comparison, the Aesculap group showed significantly higher forces compared to the Zimmer Biomet and DePuy groups (*p* = 0.002, *p* = 0.001). There was also a significant difference between the Zimmer Biomet and DePuy groups (*p* = 0.011). In a comparative study of eight different cup systems, Tradonsky et al. determined a disassembly force between 129 N and 2950 N [23]. Our data are in the lower third compared to Tradosky et al. However, acetabular liner dissociation is a rare complication following [24].

The results of our investigation showed that a tapered locking mechanism and non-locking contact with the bottom of the cup (Aesculap) showed a significant improvement in relative motion and disassembly force compared to a system with a central dome region and a taper lock mechanism (smooth surface) plus grooves at the rim (DePuy) and a hip cup system with a circular snap-fit mechanism in combination with two spikes in the inner polar region of the cup (Zimmer Biomet).

This relative motion between cup and PE liner could be a predictor of backside wear. However, features such as the surface structure, specifically roughness of the inner geometry of the acetabular cup, could also have a major impact on backside wear behaviour. Braun et al. showed 45% higher backside wear particles for the Aesculap system (Plasmafit® Plus7) compared to the Zimmer Biomet system (Allofit®-S Alloclassic®) using an identical experimental setup [9].

In addition, Braun et al. verified that backside wear particles can occur in clinically established cup systems [15]. Furthermore, Reyna et al. was able to detect backside wear caused by micromotion and poor conformity between cup and liner in an in vitro study [25]. As already mentioned, these particles could increase the potential of a resulting inflammatory reaction, which may result in osteolysis. Compared to the in vitro data from Braun et al. and Reyna et al., similar effects could be demonstrated in an in vivo study from Long et al. [26]. In this study, a series of early aseptic loosenings of acetabular cups were identified, an analysis of x-ray images was realized and an optical analysis of the inlays was carried out. The osteolysis that occurred behind the cup could be attributed to backside wear particles in conjunction with optical wear marks on the backside of the PE inlay [26]. Due to the fact that backside wear particles are also significantly smaller than articulating wear particles, they may additionally cause stronger biological responses [9].

However, experimental and clinical studies can show that backside wear in combination with relative motion between the cup and PE liner may have an impact on the osseointegration of the acetabular cup. Furthermore, subsequent studies have to show the characteristics (particle size, roundness and aspect ratio) of the PE particles that may arise and prove their influence on the human body.

From these findings, and the measured results in our study, it could be concluded that there is not a one dimensional relation between a tight fit of the inlay inside the cup, which is reflected in a low relative motion and increased disassembly forces, and the amount of backside wear. In addition, a closer look will be needed at the inner structure of the cup. It is necessary to distinguish between a smooth and a rough surface. A smooth surface with a high rate of relative motion could generate fewer backside wear particles than a rough surface with less motion.

#### *Limitations*

Initially, it should be noted that the fixation of the cups by fixating them in polyurethane does not correspond to the physiological fixation in vivo. However, the advantage of this connection was standardization of the fixation and the focus on the locking mechanisms of the cup systems. Other influencing factors could be excluded. There was a defined impaction force applied on the inlay, but no peak force as it occurs at the impaction by the surgeon.

In the analysis of motion, this work deals with relative motion. The relative motion between cup and inlay, however, could be a combination of plastic and elastic deformation, migration, rotation, creeping and many other factors.

#### **5. Conclusions**

This experimental study showed significant differences in relative motion at the cup–liner interface and in the disassembly forces for the three investigated locking mechanisms. How relative motion and assembly force affect backside wear in combination with different roughnesses should be analyzed in more detail in a further investigation.

**Author Contributions:** Conceptualization, S.J., M.U. and S.B.; Data curation, S.J. and M.U.; Formal analysis, S.J., M.U., S.S. and N.A.B.; Methodology, S.B.; Project administration, S.J., M.U. and S.B.; Writing—original draft, S.J. and M.U.; Writing—review & editing, S.J., M.U., S.S., N.A.B. and S.B. All authors have read and agreed to the published version of the manuscript.

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

**Acknowledgments:** We acknowledge financial support for publication by the Baden-Wuerttemberg Ministry of Science, Research and the Arts and by Ruprecht-Karls-Universitaet Heidelberg.

**Conflicts of Interest:** The authors declare no pertinent conflict of interest. S.J. report grants from B Braun Aesculap, Johnson & Johnson Depuy Synthes, Heraeus Medical, Waldemar Link, Peter Brehm and Zimmer Biomet that are not related to the current study. M.U., S.S. and S.B. report no conflict of interest. N.A.B. reports research grants from Johnson & Johnson DePuy Synthes that are not related to the current study.

#### **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*
