1. Introduction
Implant-based therapies have become a well-established treatment approach in modern dentistry. Osseointegration of oral implants is generally considered to be the main key to long-term clinical success. However, the functional and structural connection between implant and bone is not the only critical factor in clinics. Superstructures of oral implants, including the complex of abutments, crowns, and their interfaces, as well as their bonds with bone-anchored implants, affect the mechanical behavior and, therefore, the clinical success [
1,
2,
3]. The influence of material and geometry on physical properties is not only of interest from a biomechanical point of view. Seeking advanced materials and improved geometries could also lead to a better therapeutical option for affected patients in the future if sufficient evidence is provided. An advantageous way to fulfill this task is by testing promising and novel materials in an appropriate environment simulating in vivo conditions as close as possible. Accordingly, our workgroup implemented an ex vivo chewing simulator model, which was able to account for the complex mechanical and thermal conditions existent in the oral cavity [
4,
5]. Moreover, it allowed comparing different experimental groups simultaneously, reducing potential bias.
The absence of periodontal ligaments reduces tactile sensitivity compared to a healthy periodontium, which can lead to increased force peaks on the restoration and, thus, to fractures of the abutment material [
6]. Consequently, superstructure materials should have satisfying mechanical properties to withstand the chewing forces. In addition, implant abutments require improved esthetic features when applied in esthetic areas [
7]. With regard to the aforementioned requirements, the use of abutments made of zirconia has become widespread in recent years, backed by evidence from several clinical studies [
8,
9]. Survival, technical, and biological complication rates of zirconia abutments are comparable to titanium abutments for up to 5 years of exposure in the posterior region, although titanium abutments were reported to have a better mechanical resistance in a recent systematic review [
8,
10]. However, the zirconia material also has some disadvantages such as brittleness and aging in an aqueous environment [
11]. These can reduce stability and increase the risk of material failure [
12]. Therefore, materials having mechanical properties of metals, aesthetic features of ceramics, and tooth-like elasticity are still highly warranted.
Polyether ether ketone (PEEK) has been introduced as a high-performance thermoplastic polymer implant material in the 1990s [
13,
14]. PEEK has many advantages, including good mechanical and esthetic properties and good biocompatibility, as it is considered biologically inert [
15,
16]. Its composite structure can be altered to adjust flexural strength and modulus of elasticity for dental applications [
17]. Although its biological inertness might be the greatest challenge for its application as an oral implant material [
18], its application as an implant abutment is reasonable [
19,
20]. Currently, however, PEEK has so far only been used as a temporary abutment material [
21,
22,
23]. In an in vitro study by Rosentritt et al. [
24], eight PEEK abutments on titanium adhesive bases were restored with zirconia anterior crowns. They were then subjected to artificial masticatory loading of 3.6 million masticatory cycles and 10 N under thermocycling between 5 and 55 °C with zirconia abutments as the control group. No test specimen survived the artificial chewing simulation for more than 1 million cycles. However, the reason for this was not a direct failure of the PEEK abutments but was caused by screw loosening and screw fracture during loading. A similarly designed study used PEEK as a framework material for single molar crowns manually veneered with a composite material [
25]. Subsequently, the test specimens were loaded with 50 N for 1.2 million cycles under thermocycling between 5 and 55 °C. Implant-supported restorations were investigated in two groups, with the crowns of one group placed directly on the implant screw-retained abutments. In the other group, the crowns were provided with a transocclusal canal, bonded to the abutments in the laboratory, and only then screwed to the implant via the canal. This channel was then sealed with composite. No statement was made about the implant–abutment connection type and the abutment material. It could be assumed that the abutments were hybrid abutments with titanium adhesive bases. In the third group, veneered PEEK crowns were cemented to natural teeth. All test specimens survived the cyclic chewing simulation in the two cemented groups. Subsequently, these two groups were subjected to a static load test in a universal testing machine. The mean fracture force values were 921 N (abutment group) and 747 N (tooth group). In the transocclusal screw-retained abutment group, cracks occurred in the veneer of all test specimens during the chewing simulation, which had to be considered a failure [
25]. Failure of the PEEK frameworks was not reported. The aforementioned studies showed that PEEK remained stable after the simulated dynamic chewing simulation. However, the tests only covered a simulation of approximately 2–5 years. In addition, the studies only investigated fracture strength. Other changes, such as micro- and macroscopic surface changes that may have occurred during the chewing simulation, were not investigated in the PEEK abutments. However, it seems reasonable to check whether the surface remains unchanged in integrity after continuous mechanical loading before long-term clinical use [
26].
Considering the promising features of PEEK materials and the lack of available evidence in this field, we sought to evaluate the mechanical properties of PEEK abutments with two different internal geometries on two-piece titanium oral implants and compare them to an established zirconia abutment. To maintain a comparable clinical setting, we tested the aforementioned materials in an ex vivo chewing simulator model.
4. Discussion
The present study investigated the physico-mechanical characteristics of PEEK versus zirconia abutments on two-piece titanium-based implants. An additional research focus was set on the impact of PEEK abutment geometries on outcomes assessed with the presented ex vivo chewing simulator model.
Dynamic loading resulted in a 94% survival rate for the PEEK groups and 100% for the zirconia group. When assuming a masticatory rate of 250,000 contacts/year [
27], the present ex vivo data translates to satisfying survival rates after a simulation of approximately 20 years of clinical function for both examined materials. Fracture load is an important biomechanical parameter for the long-term performance of implant–abutment constructions. All 48 test specimens in this investigation were loaded to fracture in a static load test in order to be able to classify the fracture load of the PEEK abutments in comparison with zirconia abutments. The inclusion of an unloaded control group further allowed us to evaluate the impact of dynamic loading on fracture resistance. The evaluation of the fracture pattern showed the implant shoulder and the titanium base to be the bottleneck of the systems. When comparing the fracture loads of the study groups, the PEEK groups performed better than the zirconia abutments. The PEEK groups had mean fracture load values of up to 1101 N, whereas the zirconia group reached 772 N. The large difference in elasticity between PEEK and zirconia was probably the main influencing factor for the differences in fracture load. One might assume that the deformation of the PEEK material diverted the acting forces and that the forces acting on the implant shoulder were, thereby, reduced in this process. Stress maxima which arose could, therefore, be absorbed more safely. However, it would require future in-depth research, including biomechanical analysis, such as finite element analysis, to evaluate the material’s behavior and stress distributions during loading.
Preis et al., investigated the long-term resistance of CAD/CAM fabricated implant-supported PEEK frameworks in an artificial chewing simulation and subjected the test specimens to a static loading test. The average fracture load was 921 N, whereas that of the control group (teeth) was 747 N [
25]. In addition to PEEK crowns, six other systems made of zirconia and composite were tested. Overall, PEEK exhibited the lowest average fracture loads. The mean fracture load for zirconia was approximately five-fold higher (4817 N) than PEEK. However, compared to the present work, the test specimens were loaded axially and not at an angle of 30°. As a result, there was no leverage subjected to the abutment during loading. The presented values might, therefore, not be comparable with those of the present study. In summary, the authors reported that PEEK abutments veneered with composite performed worse than other examined specimens. In contrast, based on ex vivo testings, the workgroup of Rosentritt concluded that PEEK could be proven up to 0.4 × 106 loading cycles, which translated to 1.5–2 years of clinical application [
25]. However, they considered further tests to be necessary and stated that PEEK might be particularly appropriate for application in the anterior zone. With minimum fracture load values of 801 N in the static load test, the fracture load of the PEEK test specimens in our study was above most of the reference values found [
23,
24,
25,
46] and above the maximum masticatory forces of 595 N typically found in vivo [
47].
Regarding the influence of material and geometry on microgap values, it could be shown that ZA was generally associated with lower microgap values compared to the microgaps detected for PEEK abutments. Gehrke et al., measured TAIMs of zirconia hybrid abutments and reported mean values of 45.6 μm [
48]. These were higher than the values found for the ZA test group in the present work. The difference in the marginal gap size between PEEK and zirconia could be due to a poorer fit of the PEEK abutments or to the different manufacturing processes of test specimens. Notably, many of the PEEK test specimens showed foreign bodies in TAIMs after the chewing simulation. The marginal gap of the ZA comparison group was closed and filled with an adhesive bond due to the adhesive bonding process. Thus, no particles could be deposited here during the chewing simulation. This suggests that the manufacturing process of PEEK abutments may not allow the marginal gap to be adequately sealed and, thus, entails the risk of a foreign body reservoir during clinical application.
A comparison of PEEK groups showed that slight differences existed for the pre-loading situation, whereas the type of abutment connection did not influence microgap value differences in the post-loading comparison. However, the linear regression model adjusted for timepoint of measurement revealed that conical test specimens generally had 5.4 μm lower IACM values than parallel implant–abutment geometries. A possible explanation for this could be the different dynamics of the assembling process. When screwing the parallel abutments, the contact surfaces are pressed against each other under tension, but both components can easily separate when the retaining screw is loosened. Baj et al. [
49] reported marginal gap values of 1 μm to 4 μm for conical connections, whereas Gehrke et al., found values of 0.3 μm to 9.0 μm [
49]. Conical test specimens’ IACM values in our experiments ranged from 6.9 μm to 9.2 μm which was in accordance with the data provided by Gehrke et al. [
48]. Overall, when ranking our results in the range of the available literature IACM values (0.3–14.3 µm), PBJ IACM values ranked at the upper end (median: 10.1–15.7 μm), followed by PC (median: 6.9–9.2 μm) and ZA (median: 1.7–1.8 μm).
The influence of dynamic loading on IACMs was heterogeneous among experimental groups. For PBJ, IACMs decreased due to the loading procedure from 15.9 ± 4.5 μm to 10.6 ± 5.1, whereas IACMs increased from 6.8 ± 1.5 µm to 9.3 ± 1.7 µm in the PC group. One might assume that this probably occurred due to loading at an angle of 30° and different internal geometries forwarding the applied force differently for both specimens. For parallel abutments, the marginal gap was found in a vertical plane, whereas microgaps appeared in a horizontal plane for conical abutments. This could be considered an explanation for vertical height loss in PBJ and horizontal widening of the implant shoulder in PC. Interestingly, no influence of the chewing simulation on the marginal gaps in the parallel zirconia group compared to PBJ could be found. This might be explained due to higher initial marginal gap widths in PBJ (15.9 μm) compared to the small pre-loading microgap values of the ZA group (2.0 µm). The clinical relevance of the marginal gap might be seen in the potential risk of developing peri-implantitis [
50]. It is reasonable to keep the marginal gap width small in order to offer anaerobic bacteria as small a germ reservoir as possible. Overall, further improvements are needed to maintain smaller microgap values for PEEK implant abutments.
Furthermore, Zirconia was associated with 0.5 µm (95% CI: 0.472–9.571) higher Ra values and 7.6 µm (95% CI: 6.916–8.211 µm) higher Rz values compared to PEEK implant abutments in the present ex vivo model, indicating a favorable PEEK surface property in oral cavities. Additionally, PC showed a significant decrease in surface roughness Ra in 2/4 measurement positions after the chewing simulation. In the analysis of Rz, 4/4 measurement positions showed a significant decline in surface roughness after the chewing simulation. The decrease ranged between 1.6 μm and 8.6 μm. The surface change was possibly caused by constant rinsing with water during thermal cycling, as well as by the cyclical compression of the superstructures during the dynamic chewing load. However, it is notable that the PEEK group PBJ did not show a significant impact of the chewing simulation on Ra and Rz values. Although the abutment geometries could theoretically represent an explanation for the different roughness changes, abutments were generally constructed similarly and came from the same production batch. Therefore, this finding requires further exploration. Notably, there was a significant source of heterogeneity when measuring surface roughness. The main reason lies in different measuring devices with varying levels of precision [
51,
52]. Overall, although surface roughness measurements may have highly varied in the literature, PEEK seems to perform at least as well as its zirconia comparator when it comes to wear-related surface changes.
The present study is associated with strengths and limitations. To the best of the knowledge of the presenting authors, this is the first ex vivo study on the application of PEEK as a novel definitive implant abutment material. Utilizing the presented ex vivo chewing simulator model allowed us to compare two different PEEK abutment geometries with well-established zirconia abutments. The validity and reproducibility of experiments were great advantages of the applied model, as parts of the experiment were standardized according to ISO standard 14801 (DIN EN ISO 14801). Compliance with standards allowed for the comparison of multiple study outcomes compared to non-standardized ex vivo testing approaches [
53]. All applied loading settings in the present study were within the required tolerance range stated in ISO 14801. In order to interpret variable values specified under ISO standard 14801 (modulus of elasticity, bone loss, and loading angle), they should be related to clinical findings. The modulus of elasticity of the PEEK sample holder used in this work was 8 GPa and, therefore, adjusted in the range of cortical bone (7–22 Gpa) [
54,
55]. Although the simulation of 3 mm bone recession in our approach may have appeared high compared to clinical situations, it took into account more unfavorable clinical conditions [
56]. The dynamic load of 49 N selected in this study was based on values from clinical studies reporting mean physiological chewing forces of 24 N to 75 N for the anterior region [
57,
58]. Furthermore, 49 N is a typical loading protocol allowing future between-study comparability [
59,
60]. Our approach included static and dynamic loading mimicking the in vivo situation as closely as possible compared to studies performing solely static loading. Availability of multiple testing chambers allowed to test different experimental groups simultaneously with the same experimental condition, thus, reducing potential experimental bias. Another advantage was the inclusion of a control group that underwent solely static loading, allowing the evaluation of the impact of artificial loading/aging on the examined materials. Limitations of the chosen approach included the lack of standardized failure measurements. Failures had to be measured twice a day by one examiner. Thus, exact timepoint determination and documentation of failure occurrence could not be performed. Future sensor- and video-based documentation of failure mechanisms are warranted [
61]. Moreover, biomechanical analysis techniques such as 3D finite element analysis could be applied to gain a more in-depth knowledge of stress distributions and failure characteristics within examined materials. Finally, it needs to be mentioned that the PEEK surface is exposed to additional factors (biofilm, discoloration, acids, abrasive particles, etc.) in clinical situations, which could not be taken into account in the present study. Consequently, one cannot conclude the possible degenerative influences of these factors. It would make sense to examine the material stress on the PEEK surface in further experimental models in this context. The wear resistance of PEEK, for example, was reported to be only half that of titanium [
62]. A visible material erosion impairing the stability is, therefore, conceivable. Likewise, intrinsic discolorations that interfere with aesthetics cannot be ruled out with examination despite inert surfaces.