**The E**ff**ect of Ultraviolet Photofunctionalization on a Titanium Dental Implant with Machined Surface: An In Vitro and In Vivo Study**

**Jun-Beom Lee 1, Ye-Hyeon Jo 2, Jung-Yoo Choi 3, Yang-Jo Seol 1, Yong-Moo Lee 1, Young Ku 1, In-Chul Rhyu 1,\* and In-Sung Luke Yeo 2,\***


Received: 7 June 2019; Accepted: 26 June 2019; Published: 28 June 2019

**Abstract:** Ultraviolet (UV) photofunctionalization has been suggested as an effective method to enhance the osseointegration of titanium surface. In this study, machined surface treated with UV light (M + UV) was compared to sandblasted, large-grit, acid-etched (SLA) surface through in vitro and in vivo studies. Groups of titanium specimens were defined as machined (M), SLA, and M + UV for the disc type, and M + UV and SLA for the implant. The discs and implants were assessed using scanning electron microscopy, confocal laser scanning microscopy, electron spectroscopy for chemical analysis, and the contact angle. Additionally, we evaluated the cell attachment, proliferation assay, and real-time polymerase chain reaction for the MC3T3-E1 cells. In a rabbit tibia model, the implants were examined to evaluate the bone-to-implant contact ratio and the bone area. In the M + UV group, we observed the lower amount of carbon, a 0◦-degree contact angle, and enhanced osteogenic cell activities (*p* < 0.05). The histomorphometric analysis showed that a higher bone-to-implant contact ratio was found in the M + UV implant at 10 days (*p* < 0.05). In conclusion, the UV photofunctionalization of a Ti dental implant with M surface attained earlier osseointegration than SLA.

**Keywords:** dental implants; titanium; osseointegration; photofunctionalization; ultraviolet light; surface treatment

#### **1. Introduction**

Dental implant restorations to replace missing teeth have become a routine practice in dental clinics. Using the implants as a prosthesis helps patients feel more comfortable and these implants are more functional compared to the traditional removable dentures [1–3]. For successful implant restorations, osseointegration must be achieved between the bone and the implant. Osseointegration is the direct structural and functional connection between living bone and the surface of a load-carrying implant [4], and it is an essential factor in achieving a successful implant. Generally, it is necessary to wait for several months after implant placement for osseointegration to be achieved [5]. Unsuccessful osseointegration leads to the early failure of implants, meaning that the implants cannot endure masticatory forces, resulting in implant mobility or pain [6,7]. This includes other time-consuming situations, such as an edentulous area with limited bone quantity, or problems in patients with osteoporosis, diabetes, cancer, irradiation, old age, and heavy smokers [8–10].

The original implant surface was a smooth machined surface, with an approximate Sa value of 0.5 μm [11,12]. This machined surface has certain advantages, including a simple manufacturing

process (turning and polishing) and the ability to maintain a good hygienic state, resulting in a low incidence of peri-implant disease [12,13]. However, implants with a machined surface have shown a low bone-to-implant contact ratio (BIC) and a frequent failure of osseointegration before loading [14]. To enhance the osseointegration process, various surface modifying techniques have since been developed, where a roughened surface has demonstrated the best clinical long-term results. There are various roughening techniques, although the sandblasted, large-grit, acid-etched (SLA) surface is the most widely used and reported technique. The SLA surface sufficiently differentiates the pre-osteoblastic cells, enhances the osseointegration process, and leads to a higher BIC compared to the machined surface [15,16]. However, the roughened surface has been reported to accelerate plaque accumulation, wherein it is more difficult to remove plaque on the roughened surface than on the machined surface [12]. In this regard, reports show a greater incidence of peri-implant disease stemming from the use of the roughened surface compared to the machined surface [17].

Meanwhile, the photofunctionalization of implants using ultraviolet (UV) light has been highlighted as a simple and effective method to stimulate osseointegration in machined surfaces [18–20]. UV photofunctionalization is a phenomenon of changes occurring in titanium (Ti) surfaces after UV treatment. The process was discovered in 1977, where UV treatment transforms the natural hydrophobic properties of Ti surfaces into superhydrophilic properties by altering the physicochemical properties of the Ti. The process has been applied in environmental engineering and microbiology [21,22]. UV treatment creates the hydrophilic phase on the surface structure, thereby transforming the surface into a hydrophilic surface [23]. Photofunctionalization is also reported to enhance biological capabilities [18,19]. Consequently, the purpose of this study was to evaluate the effect of UV photofunctionalization on implants with a machined surface compared to the SLA surface, using an in vitro and in vivo study.

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

#### *2.1. Ti Samples, Surface Analysis, and UV Treatment*

#### 2.1.1. Preparation of the Ti Disc and Implant

In this experiment, commercially pure grade 4 Ti was tested in the shape of a disc (10 mm in diameter and 1 mm in thickness) and a screw-shaped implant (3.3 mm in diameter and 7 mm in length). The surface of the specimen was treated using the following methods: (a) M: The machined surface was turned and polished using sandpaper (600–1000 times); (b) SLA: The surface was sandblasted with alumina (Al2O3) particles, which were 50 μm in size and acid-etched, using hydrochloric acid and sulfuric acids (SLA surface; Point Implant Co., Seoul, Korea); and (c) M + UV: Machined surface treated with ultraviolet (UV) light. For the disc type, all the three surface treatments were examined for a negative control (M), a positive control (SLA), and an experimental group (M + UV). For the implant type sample, two surface treatments, i.e., the control (SLA) and the experimental group (M + UV), were investigated.

#### 2.1.2. Surface Analysis

Three samples were used in each group for each examination. We performed field emission scanning electron microscopy (FE–SEM; Hitachi S-4700, Hitachi, Tokyo, Japan) for a qualitative evaluation of the overall surface image. This was followed by a confocal laser scanning microscope (CLSM; LSM 800, Carl Zeiss AG, Oberkochen, Germany), where the surface roughness was quantitatively measured. The surface roughness parameters—arithmetical mean height (Sa), root mean square height (Sq), and developed interfacial area ratio (Sdr)—were measured at three randomly selected points in each sample. In addition, the chemical composition was analyzed using electron spectroscopy for chemical analysis (Sigma Probe, Thermo VG, East Grinstead, UK). Furthermore, the surface wettability of the Ti discs was examined using the contact angle from the sessile drop method, as measured by a contact angle meter (Pheonix 150, SEO, Kyunggido, Korea). All the procedures were performed under controlled conditions of 20 ◦C temperature and 46% humidity.

#### 2.1.3. UV Light Treatment

UV light treatment was achieved by irradiating the Ti discs in a specially manufactured generator using four 15 W bactericidal lamps (G15T8, Sankyo Denki, Tokyo, Japan), for at least 48 h. The intensity was approximately 5 mW/cm2 (<sup>λ</sup> <sup>=</sup> <sup>254</sup> <sup>±</sup> 20 nm).

#### *2.2. In Vitro Experiment*

#### 2.2.1. Cell Culture

Murine pre-osteoblast MC3T3-E1 cells were purchased from ATCC (American Type Culture Collection; Manassas, VA, USA). The cells were seeded onto the discs (1 <sup>×</sup> 10<sup>4</sup> cells/well) in a 12-well culture plate (Nunc, Roskilde, Denmark), and then cultured in α-minimum essential medium (α-MEM; Thermo Fisher Scientific, Waltham, MA, USA) supplemented with a 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. The cells were incubated at 37 ◦C under a humidified atmosphere of 95% air and 5% CO2. The culture medium was replaced every three days, and the osteogenic medium contained 10 mM β-glycerophosphate and 50 μg/mL ascorbic acid in the α-MEM.

#### 2.2.2. Cell Attachment

At 24 h after being seeded, the cell attachment was dual-stained using fluorescent dyes: 4 ,6-diamidino-2-phenylindole (DAPI; Invitrogen, Carlsbad, CA, USA) and Alexa Fluor 568 phalloidin (Invitrogen, Carlsbad, CA, USA) to detect the nucleus and actin filaments, respectively. Fluorescence was visualized by a CLSM (LSM 800, Carl Zeiss AG, Oberkochen, Germany), and analyzed with the ZEN2010 software (Carl Zeiss, Oberkochen, Germany).

#### 2.2.3. Cell Proliferation

The proliferative activity of cells was measured using a methyl thiazolyl tetrazolium (MTT) assay (Sigma-Aldrich, St. Louis, MO, USA) at 1, 3, and 7 days after being seeded. The culture media was replaced with an MTT solution and incubated for 3 h at 37 ◦C. After removing the MTT solution, 0.5 mL of 10% dimethyl sulfoxide in isopropanol (iDMSO) was added for 30 min at 37 ◦C. Then, the proliferation rate was assessed by its optical density (OD) at 570 nm. The value of the OD was measured using a microplate reader (BioTek, Winooski, VT, USA).

#### 2.2.4. Cell Differentiation

Total RNA in the cell cultures was extracted using the TRIzol method described by Chomczynski at 1, 4, 7, 10, and 14 days after osteoblast differentiation [24]. A reverse transcriptase–polymerase chain reaction (RT–PCR) was performed with primer sets for type I collagen (Col), alkaline phosphatase (Alp), and osteocalcin (Ocn), as described in Table 1. Quantitative real-time PCR was performed using a Takara SYBR premix Ex Taq (Takara Bio, Kusatsu, Japan) on a 7500 real-time PCR system (Applied Biosystems, Foster City, CA, USA). Each primer contained a final concentration of 200 nM, and a quantity of cDNA corresponding to 50 ng of total RNA. The PCR primers were synthesized using Integrated DNA Technology (Coralville, IA, USA). According to the manufacturer's instructions, the PCR cycling conditions comprised 40 cycles at 95 ◦C for 5 s, and 60 ◦C for 34 s after denaturation at 95 ◦C for 30 s. The cycle threshold (Ct) values were acquired using the automated threshold analysis in the Sequence Detection software version 1.4 (Applied Biosystems, Foster, CA USA). Each target mRNA expression was calculated using the comparative cycle threshold method according to the manufacturer's instructions. The relative mRNA expression levels were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The GAPDH mRNA expression levels remained steady during the osteoblast differentiation, showing similar Ct values.


**Table 1.** Primer sequences for the reverse transcriptase–polymerase chain reaction (RT–PCR).

<sup>1</sup> Type I collagen; <sup>2</sup> Alkaline phosphatase; <sup>3</sup> Osteocalcin.

#### *2.3. In Vivo Experiment*

#### 2.3.1. Animals

The rabbit tibia model was used. All the procedures were conducted with the approval of the Ethics Committee of Animal Experimentation of the Institutional Animal Care and Use Committee (CRONEX-IACUC 201803003; Cronex, Hwasung, Korea), according to the guidelines of Animal Research: Reporting In Vivo Experiments (ARRIVE).

Thereafter, four female New Zealand white rabbits (3–4 months old and 2.5–3.0 kg in weight) were anesthetized via a 1 mL intramuscular injection with a dose of 15 mg/kg of tiletamine/zolazepam (Zoletil 50, Vibrac Korea, Seoul, Korea) and 5 mg/kg of xylazine (Rompun, Bayer Korea, Seoul, Korea). Then the tibiae of the rabbits were shaved and disinfected with povidone iodine solution. Local anesthesia was administered in the surgical area with 2% lidocaine containing 1:100,000 epinephrine (Yuhan Co., Seoul, Korea).

#### 2.3.2. Surgical Procedure

A full-thickness flap was made on the medial side of both tibiae, followed by exposure of the underlying bone. In each tibia, two holes for implant placement were drilled bicortically using implant surgical drills under copious sterile saline irrigation. The diameter of the drills was increased sequentially, with a final drill diameter of 2.8 mm. Then, the implant with a diameter of 3.3 mm was inserted into the hole and engaged bicortically with sufficient stability. The SLA and M + UV implants were allocated to each hole based on a 2 × 2 Latin square randomization. Following the implant placement, the periosteum and fascia were sutured with 4-0 resorbable polyglactin material (Vicryl, Ethicon, Somerville, MA, USA), and the skin was sutured with 4-0 monofilament nylon (Blue nylon, Ailee, Busan, Korea). Post-operatively, each rabbit was kept in a separate cage and administered with 5 mg/kg of enrofloxacin (Komibiotril, Komipharm International Co., Siheung, Korea) for seven days.

#### 2.3.3. Sacrifice and Microcomputed Tomography (Micro-CT)

Two experimental animals were sacrificed at 10 days and the other two animals at 28 days after the surgery by an intravenous overdose of potassium chloride. The implants were surgically harvested en bloc with the surrounding bone. Then the implant–bone blocks were immediately immersed in a 10% neutral buffered formalin fixative. Micro-CT imaging was performed using a SkyScan 1275 (Bruker microCT, Kontich, Belgium). The X-ray source was set at an acceleration voltage of 100 kV and a pixel size of 10 μm. Each sample was scanned three times, using 360◦ spiral scanning on the SkyScan 1275 with a scanning time of 2 h. Reconstruction was performed using an NRecon (v. 1.7.3.2, Bruker microCT). The region-of-interest (ROI) was defined as the area within consecutive threads engaged in the upper cortical bone (Figure 1a). The analysis was performed using the CTAn software (v. 1.18.4.0, Bruker microCT; Figure 1b), and it also involved the visualization software DataViewer (v. 1.5.4.0, Bruker microCT) and CTVox (v. 3.3.0, Bruker microCT).

**Figure 1.** (**a**) The schematic drawing of the region-of-interest (ROI). The ROI was defined as an area within threads engaged in the upper cortical bone (red dot box); (**b**) microcomputed tomography (micro-CT) images for the measurement of the bone-to-implant contact ratio (BIC); (**c**) the bone-to-implant contact ratio (BIC) was calculated by the length of the green line divided by the total length of the well (green and red line); (**d**) the definition of the bone area (BA) was calculated by the area of red color divided by the total area of the well.

#### 2.3.4. Histological Preparation and Histomorphometric Measurement

After μCT scanning, the implant–bone blocks were dehydrated in a series of ethanol with increasing concentrations and then embedded in light-curing resin (Technovit 7200 VLC, Hereaus Kulzer, Hanau, Germany). The embedded blocks were sectioned perpendicular to the longitudinal axis of the implant using the EXAKT system (EXAKT Apparatebau, Norderstedt, Germany), following the method described by Donath and Breuner [25]. The section was then ground to a thickness of 40 μm and stained with hematoxylin and eosin (H&E) for examination using a light microscope. These undecalcified, ground sections were observed under a light microscope (BX51, Olympus, Tokyo, Japan) to measure the BIC and bone area (BA; Figure 1c,d). The region-of-interest (ROI) was defined as the same area that was used in the micro-CT analysis. The measurement was performed under a ×100 magnification, using the SPOT software version 4.0 (Diagnostic Instruments, Sterling Heights, MI, USA) and Image-Pro Plus (Media Cybernetics, Rockville, MD, USA).

#### *2.4. Statistical Analysis*

The Kruskal–Wallis test was used to evaluate statistically significant differences amongst the three groups of discs. If there was a significant difference amongst the three groups, the post-hoc Tukey method was applied. To compare the two groups of implants, the Mann–Whitney U test was performed to determine the statistically significant differences. *p* < 0.05 was set as the statistical significance. All statistical analyses were performed using SPSS 20.0 (IBM Corp., Armonk, NY, USA).

#### **3. Results**

#### *3.1. Surface Characteristics*

In the overall evaluation of the Ti samples via the FE–SEM images, the M and M + UV surfaces showed similar evidence of machine turning (continuous straight traces) with smooth surfaces, although the SLA surface presented a rougher surface with a typical honeycomb appearance (Figures 2a and 3a).

**Figure 2.** (**a**) Scanning electron microscopy (SEM) images of the Ti discs. M: Machined surface (top row); SLA: Sandblasted with large grit and acid etched surface (middle row); M + UV: machined surface treated with ultraviolet light (bottom row). Scale bars: 10 μm. (**b**) Surface roughness (Sa and Sq) and surface area ratio (Sdr) of the Ti discs evaluated by confocal laser scanning microscopy (CLSM) analysis. The values for the M and M + UV surfaces are similar and smaller than the SLA. (**c**) Element content of the surfaces of the Ti disc according to the energy dispersive x-ray spectrometer (XPS). M + UV shows a significantly lower carbon percentage than the M disc, but there is no significant difference between the SLA and M + UV. (**d**) The changes in wettability of the Ti discs. Superhydrophilicity after UV treatment for 48 h was observed. Scale bars: 10 mm. (**e**) The value of the contact angle of the Ti discs. Without UV treatment, the Ti disc was hydrophobic, but became superhydrophilic (zero degree angle) after UV treatment for 48 h. Error bars show the standard deviation. (\*) and (\*\*) represents significance compared with each pair, *p* < 0.05 and *p* < 0.01, respectively.

Surface roughness parameters for the samples are shown in Figures 2b and 3b. In the disc specimen, the SLA surface showed higher Sa, Sq, and Sdr values, and these were significantly different from the M and M + UV surfaces (*p* < 0.01). There was no statistical difference between the M and M + UV surfaces (*p* > 0.05). Similarly, for the implant specimen, the M + UV and SLA surfaces were statistically different across all the parameters (*p* < 0.05).

Chemical compositions from the x-ray spectrometer (XPS) revealed that, compared with the 43.42% ± 0.31% carbon in the M disc, the M + UV disc contained about 32.35% ± 1.50% carbon, which was statistically different (*p* = 0.000). The SLA showed 31.87% ± 1.42% carbon, with no significant difference from the M + UV disc (*p* = 0.051). On the other hand, the M + UV implant showed a significantly lower carbon percentage compared to the SLA implant (*p* = 0.049; Figures 2c and 3c).

**Figure 3.** (**a**) Scanning electron microscopy (SEM) images of the Ti implants. SLA: Sandblasted with large grit and acid etched surface (top row); M + UV: Machined surface treated with ultraviolet light (bottom row). Magnification: ×30, ×1000, ×2000, and ×5000 from the left. (**b**) Surface roughness (Sa and Sq) and surface area ratio (Sdr) of the Ti discs according to the confocal laser scanning microscopy (CLSM) analysis. The values of the M + UV surfaces are smaller than the SLA, which means it is smoother than the SLA. (**c**) Element content of the surfaces of the Ti discs according to the energy dispersive x-ray spectrometer (XPS). The M + UV discs contain half of the carbon percentage of the SLA discs. Error bars show the standard deviation. (\*) represents the significance compared with each pair, *p* < 0.05.

Contact angle measurement was performed for the disc-type specimens. The M and SLA discs showed a hydrophobic status with angles of 63.6 ± 4.7◦ and 68.3 ± 2.5◦, respectively. On the other hand, 48 h after UV treatment, the M + UV discs showed superhydrophilicity at a 0◦ contact angle (*p* = 0.000; Figure 2d,e).

#### *3.2. In Vitro Test*

#### 3.2.1. Cell Attachment

The CLSM images of the cells are shown in Figure 4a. A wider spread of cells was observed in the M + UV groups compared to the other groups. In the SLA discs, the cells were sharp and needle-like shaped, implying that the osteoblasts were not prone to attach to the SLA surfaces.

#### 3.2.2. Cell Proliferation

The MTT assay showed that the amount of cells increased in a time-dependent way on all the surfaces. On the M + UV surface, the cells proliferated more significantly than the other surfaces at days 1, 3, and 7, with a *p*-value of less than 0.01, as shown in Figure 4b. On day 7, the amount of cells on the M + UV surface was two times greater than cells on the SLA surface (0.068 ± 0.0005 vs. 0.040 ± 0.001, *p* = 0.000). In particular, at day 3 and 7, the cells proliferated more on the M surface than on the SLA surface.

#### 3.2.3. Quantitative Assessment of the Osteogenic Markers

Figure 4c shows the relative mRNA expression of Col, Alp, and Ocn. The RT–PCR analysis showed that Col was more significantly expressed on the M + UV and SLA surfaces at days 7, 10, and 14 compared to the M surface, although the M + UV and SLA surfaces were not significantly different. The expression level of Alp on the SLA surface was not different from the M + UV surface at days 1 and 4, but it was significantly higher than the expression levels at days 7, 10, and 14. The M + UV surface expressed the Alp gene more than the M surface at days 4, 7, and 14. The expression level of Ocn on the M + UV surface was significantly higher at day 7, but it was lower at days 10 and 14.

**Figure 4.** (**a**) Confocal microscopic images of the MC3T3-E1 cells, 24 h after being seeded on the Ti discs. The areas in the dotted box are magnified in the bottom row. Scale bars: 50 μm at ×100 and 20 μm at ×200 magnification. (**b**) Evaluation of cell proliferation of the MC3T3-E1 cells by an MTT assay at 1, 3, and 7 days after being seeded on the Ti discs. (**c**) Evaluation of the cell differentiation of MC3T3-E1 cells by real-time PCR at 1, 4, 7, 10, and 14 days after being seeded on the Ti discs. The relative mRNA expression levels were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The osteogenic markers are type I collagen (top), alkaline phosphatase (ALP, middle), and osteocalcin (OCN, bottom). UV photofunctionalization enhanced the osteoblastic gene expression. Error bars show the standard deviation. (\*) and (\*\*) represent the significance compared with each pair, *p* < 0.05 and *p* < 0.01, respectively.

#### *3.3. In Vivo Test*

#### 3.3.1. Histomorphometry

All the implants were successfully osseointegrated at days 10 and 28 (Figure 5a). At day 10, the BIC ratios of the M + UV implants (55.93% ± 6.19%) were significantly higher than that of the SLA implants (43.38% ± 3.20%, *p* = 0.021). However, at day 28, the BIC ratios of the M + UV implants (64.88% ± 5.35%) were not significantly different from that of the SLA implants (59.93% ± 6.44%, *p* = 0.149; Figure 5b).

**Figure 5.** (**a**) Representative histologic sections of the rabbit tibia at 10 and 28 days after the implant placement. In the SLA implant, the osteoblast and organic matrix, which had not mineralized yet, was more observable on the interface between the bone and implant compared to the M + UV implant (red arrow head; magnification ×12.5, ×40, and ×100 from the left, hematoxylin and eosin staining). The scale bars: 1 mm at ×12.5, 200 μm at ×40, and 100 μm at ×100 magnification. (**b**) The bone-to-implant contact ratio (BIC) was evaluated histologically at days 10 and 28. The M + UV implant shows a significantly higher BIC than the SLA at day 10, but there is no significant difference at day 28. (**c**) The bone area ratio (BA) evaluated histologically at 10 and 28 days. The M + UV implants show significantly more BA than the SLA at days 10 and 28. (**d**) The bone-to-implant contact ratio evaluated by micro-CT (3D BIC) at days 10 and 28. The M + UV implants show a significantly higher 3D BIC than the SLA at day 10, but there is no significant difference at day 28. Error bars show the standard deviation. (\*) represents the significance compared with each pair, *p* < 0.05.

In terms of the BA, the M + UV implants were significantly higher compared to the SLA implants (46.55% ± 8.59%) at day 10 (65.09% ± 10.42% vs. 46.55% ± 8.59%, *p* = 0.042) and at day 28 (72.70% ± 5.52% vs. 61.83% ± 4.89%, *p* = 0.043; Figure 5c).

#### 3.3.2. Micro-CT

The three-dimensional BIC was evaluated using micro-CT. The micro-CT analysis revealed that the three-dimensional BIC of the M + UV implants was significantly higher than that of the SLA implants at day 10 (88.87% ± 5.1% vs. 81.6% ± 3.28%, *p* = 0.046), but it was not statistically different at day 28 (91.91% ± 1.55% vs. 87.47% ± 2.93%, *p* = 0.201; Figure 5d).

#### **4. Discussion**

In this study, we found that UV photofunctionalization on a Ti screw-shaped implant with an M + UV surface showed a higher BIC than the SLA surface at day 10, and there was no significant difference at day 28. This was confirmed in both two-dimensional and three-dimensional measurements. The results indicated that the UV photofunctionalization could accelerate the osseointegration process, and achieve firm fixation between the implant and the surrounding bone earlier. These findings are supported by other studies, where Park et al. found that after four months of healing, the UV-treated implants in rabbits showed a higher BIC than the untreated implants. The authors observed that UV treatment decreased both carbon impurities on the surface and water contact angles [26]. Similarly, Aita et al. showed that the UV-treated acid-etched implants at week two had a push-in value equivalent to the untreated acid-etched implant [18]. Pyo et al. measured the removal torque test in UV-treated implants and showed that it was 50% higher than in untreated implants [27]. Hirota et al. retrospectively studied and found that the use of photofunctionalization reduced the risk of early implant failure with an odds ratio of 0.30 (*p* < 0.05) [28]. Soltanzadeh studied the effect of UV photofunctionalization on immediately loaded implants in a rat model. After the placement, the implants were immediately loaded with 0.46 N of static lateral force. The results showed that osseointegration was successful in 100% of photofunctionalized implants, but 28.6% of untreated ones. The value of the push-in test was 2.4 times higher in photofunctionalized implants [29].

Histologically, the BA was significantly higher at days 10 and 28 in the M + UV compared to the SLA implants, meaning that the M + UV implant had a higher amount of mineralized bones between threads of implants. Pyo et al. evaluated the osteogenic dynamics using fluorescent labeling at four weeks after implant placement; and found that in the UV-treated implant, the interfacial areas between the bone and implant and the areas within the threads were filled with calcein-positive tissues compared to the untreated implants. This meant that UV photofunctionalization could lead to earlier bone deposition [27]. Ueno et al. showed that the UV-treated acid-etched implant had a marked bone formation in a gap healing model without cortical support [20]. Kitajima et al. measured the implant stability quotients (ISQ) for 55 photofunctionalized implants with low and extremely low initial stability at the time of placement and stage-two surgery. Then they calculated the ISQ increase per month, defining the osseointegration speed index (OSI). The OSI ranged 3.9–4.7 substantially higher than the OSIs for untreated implants reported in other literatures (0.36–2.8) [30]. Ijishima et al. evaluated the effect of photofunctionalization on aged rats. The aged rats showed considerably lower biological capabilities (cell attachment, proliferation, and ALP activity) than the young. However, the enhancement of cell attachment and differentiation were observed on the photofunctionalized Ti discs compared with untreated one. Moreover, in the femurs of aged rats, the photofunctionalized mini-implant showed the higher push-in value than untreated one after two weeks of healing. These findings supported that UV photofunctionalization could be also valuable in the compromised sites [31].

Generally, the surface roughness is considered as a main factor for the improvement of osseointegration. However, in this study, UV treatment did not physically change any surface roughness as shown in the SEM and CLSM. Rather, it induced superhydrophilicity (0◦ angle), reduced the percentage of hydrocarbons, and increased the osteoblast proliferation, attachment, and differentiation, as shown in the in vitro study. This indicated that the only physico-chemical changes in the Ti surface could enhance the biological activities. In the XPS analysis, Roy et al. found that UVC photon energy decreased carbon deposition and the amount of H2O on Ti surface, and produced many –OH groups (TiOH) without any changes in surface topography. They explained that, through these chemical changes, the UV photofunctionalization could create the superhydrophilicity of Ti. The improvement of biological capabilities by UV photofunctionalization was supported by other studies [32]. Aita et al. showed that Col and osteopontin (Opn) were more expressed in the UV-treated discs [18]. The RT–PCR analysis performed by Zhang et al. showed that the expression of genes encoding Col, Runx2, BMP, and Opn increased in the UV-treated surface [33]. In contrast, Att et al. assessed the RT–PCR of genes for Opn and Ocn in bone marrow cells derived from the femur of

eight-week-old male Sprague-Dawley rats, and found that there was no significant difference at days 10 and 20. The differences may have been caused by the kinds of cells, time points, and intensity and wavelength of the UV generator. Further research is required to elucidate this aspect.

With regard to plaque accumulation and peri-implant disease, the implant with a smooth surface is considered to be superior to an implant with a rough surface. Berglundh et al. observed that, at five months after the removal of ligature, bone loss accelerated in the SLA implant but not in the polished implant. Histologically, the size of the inflammatory lesion and the area of plaque were larger in the SLA surface [12]. Additionally, Albouy et al. compared the turned and the roughened implant (Ti-Unite), at six months after the ligature removal, and observed a larger amount of bone loss in the Ti-Unite implant compared to the turned implant (1.47 mm vs. 0.3 mm). This meant that spontaneous progression of peri-implantitis had occurred in the implant with a rough surface [17]. However, the machined implant had a definite drawback in that it had a low BIC level, because the osteoblastic differentiation was lower compared to the smooth surface [34,35]. Therefore, the enhancement of osteoblastic differentiation on the Ti with an M surface by UV photofunctionalization is considered to be inspired. Additionally, UV photofunctionalization itself could decrease plaque formation on Ti surface. De Avila et al. found that after 16 h incubation, there were significantly lower oral bacterial attachment on the UV-treated Ti disc compared to the untreated one [36].

The hydrophilicity of the implant surface can be induced by UV photofunctionalization [18–20,27,37] or preservation in a storage medium [38–40]. Both methods are effective and can improve the bone healing process and attain early osseointegration. However, the latter method has been reported to lead to foreign deposition and little elimination of hydrocarbons on the Ti surface. Moreover, the saline storage method is inferior to UV photofunctionalization in osteoblast spreading and adhesion [41]. Considering this point, UV treatment is considered a safer method to modify the implant surface to make it hydrophilic. On the other hand, Att et al. mentioned that, in UV photofunctionalization, the superhydrophilicity is not a significant factor in explaining the higher BIC in the UV-treated Ti discs compared to the acid-etched ones. The elimination of hydrocarbon on the surface was considered to be a significant factor [19]. The aging of the Ti is related to the contamination and accumulation of the hydrocarbon on the Ti surface, and it can suppress cell recruitment and biologic activity [42,43].

The combination of variables such as duration, intensity, and wavelength can create various modes of UV photofunctionalization, noting that the optimal combination is a controversial issue. The exposure time has been used from 12 min to 48 h [44,45]. However, Aita et al. and Att et al. have shown that, between 24 and 48 h, there was an increase in hydrophilicity and biological effects [18,19]. Additionally, treatment with UVC (λ = 240 ± 40 nm) has shown more biological improvements compared to UVA (λ = 360 ± 40 nm). Consequently, in our experiment, to maximize the effects of UV, the mode of UV photofunctionalization was determined as UVC treatment for 48 h.

There are still several questions regarding UV photofunctionalization. Strictly, the UV light treatment used in this study may be called physical photo-activation, rather than functionalization, because no chemical application to the surface, which have been shown in the previous studies, were used for enhanced bone response [11,46]. More obvious concept of the UV surface treatment needs to be established in the physical and chemical aspects. Amongst several factors following UV photofunctionalization, we also still need to identify the main factors for the enhancement of biological activity, superhydrophilicity, and removal of hydrocarbon. If they contribute to the improvement, there is a need to understand the mechanism through which they are inter-connected. Therefore, further studies are needed to fully appreciate the effects of UV photofunctionalization.

#### **5. Conclusions**

Within the limitations of the present study, UV photofunctionalization of a Ti dental implant with an M surface attained an earlier osseointegration compared to an implant with an SLA surface. The enhancement was considered to result from the superhydrophilicity, the elimination of hydrocarbon on the surface, and the improvement of osteoblastic activities.

**Author Contributions:** Conceptualization, J.-B.L., I.-C.R., I.-S.L.Y.; methodology, J.-B.L., Y.-H.C., J.-Y.C., I.-S.L.Y.; software, J.-B.L., Y.-H.C.; validation, I.-C.R., I.-S.L.Y.; formal analysis, J.-B.L., I.-S.L.Y.; investigation, J.-B.L., Y.-H.C., J.-Y.C.; resources, J.-Y.C., Y.-J.S., I.-C.R., I.-S.L.Y.; data curation, J.-B.L., J.-Y.C.; writing—original draft preparation, J.-B.L.; writing—review and editing, Y.-H.C., J.-Y.C., Y.-J.S., Y.-M.L., Y.K., I.-C.R., I.-S.L.Y.; visualization, J.-B.L., J.-Y.C.; supervision, I.-C.R., I.-S.L.Y.; project administration, I.-S.L.Y.; funding acquisition, I.-S.L.Y.

**Funding:** This work was supported by a National Research Foundation of Korea (NRF) grant funded by the Ministry of Science and ICT [NRF-2016R1A2B4014330] and by a Korea Health Industry Development Institute (KHIDI) grant funded by the Ministry of Health & Welfare, Republic of Korea [HI15C1535].

**Acknowledgments:** The authors greatly thank Kyoung-Hwa Kim and Young-Dan Cho for the material preparation and support.

**Conflicts of Interest:** The authors declare no potential conflict of interest related to this study. The funders had no role in the design of the study; in the collection, analyses, or interpretation of the data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2019 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* **A Vitronectin-Derived Bioactive Peptide Improves Bone Healing Capacity of SLA Titanium Surfaces**

**Chang-Bin Cho 1,**†**, Sung Youn Jung 2,**†**, Cho Yeon Park 2, Hyun Ki Kang 2, In-Sung Luke Yeo 1,\* and Byung-Moo Min 2,\***


Received: 13 September 2019; Accepted: 16 October 2019; Published: 17 October 2019

**Abstract:** In this study, we evaluated early bone responses to a vitronectin-derived, minimal core bioactive peptide, RVYFFKGKQYWE motif (VnP-16), both in vitro and in vivo, when the peptide was treated on sandblasted, large-grit, acid-etched (SLA) titanium surfaces. Four surface types of titanium discs and of titanium screw-shaped implants were prepared: control, SLA, scrambled peptide-treated, and VnP-16-treated surfaces. Cellular responses, such as attachment, spreading, migration, and viability of human osteoblast-like HOS and MG63 cells were evaluated in vitro on the titanium discs. Using the rabbit tibia model with the split plot design, the implants were inserted into the tibiae of four New Zealand white rabbits. After two weeks of implant insertion, the rabbits were sacrificed, the undecalcified specimens were prepared for light microscopy, and the histomorphometric data were measured. Analysis of variance tests were used for the quantitative evaluations in this study. VnP-16 was non-cytotoxic and promoted attachment and spreading of the human osteoblast-like cells. The VnP-16-treated SLA implants showed no antigenic activities at the interfaces between the bones and the implants and indicated excellent bone-to-implant contact ratios, the means of which were significantly higher than those in the SP-treated implants. VnP-16 reinforces the osteogenic potential of the SLA titanium dental implant.

**Keywords:** vitronectin; RVYFFKGKQYWE motif; cellular responses; dental implants; osseointegration

#### **1. Introduction**

Attachment of cells is the first step in cell–biomaterial interactions [1]. The transmembrane proteins on the cell membrane recognize and bind biomacromolecules adsorbed on the surface of the biomaterial [1]. These biomacromolecules are from the extracellular matrix (ECM), controlling cellular behaviors such as attachment, spreading, proliferation, and differentiation depending on the interacting transmembrane receptors [2]. Therefore, these ECM biomolecules, if applied to the titanium dental implant surface, are anticipated to have potential in reinforcing osseointegration [3–5]. One candidate molecule is vitronectin.

Vitronectin, one of the ECM proteins, is an abundant multifunctional glycoprotein found in serum, the extracellular matrix, and bone, and is involved in various physiological processes such as cell attachment, spreading, and migration [6–8]. Vitronectin contributes to healing of the bone surrounding a dental implant by promoting the attachment and spreading of the osteogenic cells [9,10]. This ECM protein appears to play a major role in initial bone healing by reorganizing the intracellular

microfilaments and microtubules, which facilitates cell attachment and spreading [11–13]. However, the use of such an original protein has several critical limitations: a high cost for synthesis, antigenicity and instability of the molecule, and steric hindrance of this macromolecule in focal adhesion [5,14,15]. A functional peptide derived from the parent protein is a notable alternative, overcoming these limitations and maintaining the original biological activity [3,5,9,14,16,17]. In addition, bioactive peptides have advantages over larger protein molecules due to their robustness and sterilizability [18]. Recently, a vitronectin-derived functional peptide sequence, RVYFFKGKQYWE (VnP-16), has shown a Janus regulation for bone formation: promotion of osteoblast activity and inhibition of osteoclast activity, which is a desirable effect for osteogenesis that vitronectin does not have [9]. The VnP-16 peptide promoted bone formation by accelerating osteoblast differentiation and activity through direct interaction with β1 integrin followed by focal adhesion kinase (FAK) activation. Concomitantly, the peptide inhibited bone resorption by restraining Janus N-terminal Kinase (JNK)-c-Fos-nuclear factor of activated T cells, cytoplasmic 1 (NFATc1)-induced osteoclast differentiation and αvβ3 integrin-c-Src-proline-rich tyrosine kinase 2 (PYK2)-mediated resorptive function. Moreover, VnP-16 peptide decreased the bone resorbing activity of pre-existing mature osteoclasts without changing their survival rate [9].

The functionalization of a titanium implant via the immobilization of desirable proteins or their bioactive peptides in their native conformations is a promising approach to overcome the bioinertness of the surface, leading to improved osseointegration [18,19]. Several methodologies, including physical adsorption, covalent immobilization via chemical methods, and covalent immobilization via physical methods, have been investigated since the inception of protein functionalization on titanium substrates [19]. Physical adsorption is the simplest method to immobilize proteins on titanium substrates. Chemical immobilization, the most established of the protein functionalization approaches, proved that it was possible to covalently immobilize proteins to titanium substrates, overcoming the unintentional protein release observed in adsorption approaches. Physical covalent immobilization of biomolecules is the most recent approach to dental and orthopedic biomimetic functionalization, and possesses advantages over adsorption and chemical covalent immobilization [19]. Other methods on peptide-functionalization of titanium surfaces are also reported. Control of both peptide orientation and surface concentration is achieved by varying the solution pH or by applying an electric field [18]. In addition, multifunctional coating improves cell adhesion on titanium surfaces by using cooperatively acting peptides [20].

A micro-roughened surface of commercially pure titanium has been clinically used in the field of implant dentistry [10]. A sandblasted large-grit acid-etched (SLA) surface with approximately 1.5 μm of arithmetic mean deviation is known to accelerate osseointegration at the bone-implant interface, compared to the turned surface with no modification at the micro-level [10,21,22]. The application of VnP-16 to the SLA surface increases clinical relevance in the further enhancement of bone formation and makes use of dental implants extended to patients suffering from bone metabolic weaknesses, like osteoporosis. The VnP-16-treated SLA titanium surface has not been investigated yet.

This study aimed to evaluate the early bone response to the VnP-16-treated SLA titanium surface in vivo. In vitro tests were also performed using osteoblast-like cells. The hypothesis underlying this study was that the application of VnP-16 would further reinforce the osteogenic potential of the SLA surface.

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

#### *2.1. Cells, Peptides, and Reagents*

HOS and MG-63 cells, lines derived from human osteosarcomas, were purchased from the American Type Culture Collection (Rockville, MD, USA) and cultured in Dulbecco's modified Eagle's medium (Gibco BRL, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum. Each peptide was synthesized using the 9-fluorenylmethoxycarbonyl-based solid-phase method with a C-terminal amide using a Pioneer Peptide Synthesizer (Applied Biosystems, Foster City, CA, USA) in the Peptron (Daejeon, Korea). The synthetic peptides used in the study had a purity greater than 95%, as determined using high-performance liquid chromatography. Human plasma vitronectin was obtained from Millipore (Bedford, MA, USA).

#### *2.2. Disc Preparation and Surface Characterization*

Titanium disc specimens, which were 0.5 mm thick and 10 mm in radius, were made of commercially pure grade 4 titanium. The discs, serving as control, were prepared by polishing with #600 and #1200 sandpaper. The other discs were subjected to sandblasting with large alumina particles and etched with a hydrochloric acid solution to generate the SLA surface (Deep Implant System, Seongnam, Korea). The SLA titanium discs were rinsed, ultrasonically washed, and dried. One group of the SLA discs was left untreated, another was treated with a scrambled peptide (SP; 10.5 μg/cm2), and the other was treated with VnP-16 (10.5 μg/cm2).

The surfaces of the four types of disc were imaged using field emission-scanning electron microscopy (FE-SEM; S-4700, Hitachi, Tokyo, Japan). The element composition of each group was analyzed by electron spectroscopy for the chemical analysis (ESCA; Sigma Probe, Thermo Scientific, Waltham, MA, USA). Confocal laser scanning microscopy (CLSM; LSM 800, Carl Zeiss AG, Oberkochen, Germany) calculated two surface parameters for surface topography of the investigated discs; arithmetic mean deviation (Ra) and the developed surface area ratio (Sdr) [23].

#### *2.3. Cell Attachment and Spreading Assays*

The cell attachment assay was performed as described previously [9]. The physical adsorption method was used for the application of peptides. Twenty-four-well culture plates were coated with 0.26 μg/cm2 human plasma vitronectin for 24 h at 4 ◦C or 10.5 μg/cm2 synthetic peptides for 24 h at room temperature, blocked with 1% heat-inactivated bovine serum albumin (BSA) in phosphate-buffered saline (PBS) for 1 h at 37 ◦C, and then washed with PBS. Cells (1 <sup>×</sup> 105 cells/500 <sup>μ</sup>L) were added to each plate and incubated in serum-free culture medium for 1 h at 37 ◦C. After incubation period, unattached cells were removed by rinsing twice with PBS. Attached cells were fixed with 10% formalin for 15 min, stained with 0.5% crystal violet for 1 h, gently washed with distilled water three times, and dissolved with 2% sodium dodecyl sulfate for 5 min. Absorbance was measured at 570 nm using a microplate reader. For cell-spreading assay, cells (7 <sup>×</sup> 104 cells/500 <sup>μ</sup>L) were added to each substrate-coated plate and incubated for 3 h at 37 ◦C. To determine cell spreading, formalin-fixed and crystal violet-stained cell surface area was measured with Image-Pro plus software (Version 4.5; Media Cybernetics, Silver Spring, MD, USA).

#### *2.4. Migration Assay*

Migration assays were performed using a transwell migration chamber (Corning, Pittston, PA, USA) possessing 8 μm pores as described previously [24]. The lower side of each transwell filter was coated with vitronectin (0.26 μg/cm2), or synthetic peptides (10.5 μg/cm2) by drying for 24 h at <sup>4</sup> ◦C (vitronectin) or for 24 h at room temperature (peptides). Cells (2 <sup>×</sup> 104 cells/24-well) were seeded in the upper chamber of a transwell filter and allowed to migrate for 24 h at 37 ◦C. Cells were then fixed with 10% formalin for 15 min and stained with 0.5% crystal violet. Unmigrated cells in the upper side of the transwell filter were removed with a cotton swab, and cell migration was quantified by counting the number of cells that had migrated through the filter. Human placental laminin was used as the positive controls and SP was used as the negative control.

#### *2.5. Cell Viability Assay*

The viabilities of cells were investigated using the EZ-Cytox Cell Viability Assay kit (water-soluble tetrazolium salt method; Daeil Lab Service, Seoul, Korea). A 96-well microplate was coated with VnP-16 peptide (0, 10.6, 21.2, or 42.4 <sup>μ</sup>g/cm2) by drying for 24 h at room temperature. Cells (1.5 <sup>×</sup> 10<sup>4</sup> cells/100 μL) were seeded onto a 96-well microplate and then cultured for 24 h or 48 h at 37 ◦C. The water-soluble tetrazolium salt reagent solution (10 μL) was added to each well, and the plate was incubated for 2 h at 37 ◦C. The absorbance at 450 nm was then measured using a microplate reader.

#### *2.6. In Vivo Experiment*

Sixteen screw-shaped grade 4 titanium implants were made, which were 3.5 mm in major diameter and 11 mm in length (Warantec, Seongnam, Korea). Four implants were used as they were without any surface modification and designated as turned surface (control). The surface of another four implants was SLA (Deep Implant System, Seongnam, Korea). Half of the rest of the eight implants were treated with SP while the other half were treated with VnP-16 (1.0 mg/cm2). Using the physical adsorption method, the Ti implants were placed on 0.2 ml PCR tubes and coated with the synthetic peptides by drying for 7 d in a vacuum at room temperature.

All the animal experiments performed in this study were approved by the Ethics Committee of Animal Experimentation of the Institutional Animal Care and Use Committee (CRONEX-IACUC 201705001; Cronex, Hwasung, Korea). These experiments were conducted following the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines for the care and use of laboratory animals [25]. Four New Zealand white rabbits were used in this study, which were male, approximately five to six months in age and 2.5 to 3.0 kg in weight. The experimental animals were intramuscularly anesthetized with a dose of 15 mg/kg tiletamine hydrochloride and zolazepam hydrochloride (Zoletil, Virbac, Carros, France) and 5 mg/kg xylazine (Rompun, Bayer AG, Leverkusen, Germany). The skin hair was shaved at the tibial area of the rabbits, which was disinfected with aqueous iodine. A full-thickness incision from the skin to the periosteum of the tibiae was made, and the flaps were elevated to expose the medial surfaces of the tibiae. Drilling was conducted for to make the holes on the medial surfaces for implant insertion. The final diameter of the holes was 3.2 mm. Two implants were inserted into each tibia and arranged according to the split plot design. The periosteum and fascia were sutured with 4-0 polyglactin 910 (Vicryl, Ethicon, Somerville, NJ, USA) while the skin was sutured with 4-0 Nylon (Ethilon, Ethicon, Somerville, NJ, USA). Each experimental animal was housed in a separate cage and an antibiotic, enrofloxacin, (Biotril, Komipharm International, Siheung, Korea) was administered to prevent infection.

#### *2.7. Light Microscopic Evaluation*

The rabbits were sacrificed under general anesthesia with the intravenous administration of potassium chloride at 14 days after implant insertion. The implants were removed en bloc with the surrounding bones and fixed in 10% neutral buffered formalin for 2 weeks. After formalin fixation, each implant-bone block was dehydrated with ethanol. Then, the blocks were resin-embedded (Technovit 7200, Heraeus Kulzer, Hanau, Germany) and ground for light microscopy using an EXAKT system (EXAKT Apparatebau, Norderstedt, Germany), according to the methods described in the previous studies [23,26]. Sections of the implant-bone blocks were prepared with a final thickness of approximately 50 μm and modified Goldner's Masson trichrome staining [27]. The interfacial areas between the bones and implants were observed and evaluated from the bone crests to 2 mm in depth for histomorphometry, where bone-to-implant contact (BIC) and bone area (BA) ratios were calculated. The image analyses and the histomorphometric calculations were performed on ×100 magnified images using a light microscope (BX51, Olympus, Tokyo, Japan), SPOT version 4.0 software (Diagnostic Instruments, Sterling Heights, MI, USA) and Image-Pro Plus (Media Cybernetics, Rockville, MD, USA).

#### *2.8. Statistics*

Descriptive statistics for the data were presented as the mean ± standard deviation (SD). The statistical analyses were performed with R software (version 3.6.1, R Foundation for Statistical Computing, Vienna, Austria). All the data obtained in this study were confirmed to be normally

distributed by the Shapiro–Wilk test. Analysis of variance tests were used for comparisons among the groups. When a significant difference was found, Tukey's honestly significant difference test was further applied for pairwise comparison. The level of significance was 0.05 in this study.

#### **3. Results**

#### *3.1. Surface Characteristics*

The FE-SEM images of the specimens showed very different topographical features between the polished and SLA surfaces (Figure 1A). Some grooves on the overall flat surfaces were found for the polished specimens, while a honeycomb-like irregular topography was observed for the SLA specimens. The treatments of SP and VnP-16 had little effect on the surface physical features of the specimens (Figure 1A). There was no significant difference in either the Ra or Sdr among the SLA titanium discs, regardless of the peptide treatment (*p* > 0.05) (Figure 1B). However, in surface chemistry, the treatments of the functional peptides were confirmed from the results of higher nitrogen contents for the SP- and VnP-16-treated surfaces, compared to those for the other groups (polished and SLA titanium surfaces) (*p* < 0.05) (Figure 1C). The highest content element was carbon for every group.

**Figure 1.** Surface characteristics of the titanium specimens investigated in this study. (**A**) Field emission scanning electron microscopy definitely shows different topographical features between the polished and sandblasted, large-grit, acid-etched (SLA) surfaces. (**B**) The mean values of the measured surface parameters indicated that the peptide treatment did not change the surfaces physically at the micro level. Note the significant differences in the surface parameters between the polished and the other SLA surfaces. \*\* *p* < 0.01 vs. the polished surface. (**C**) Electron spectroscopy for chemical analysis detected high nitrogen content on the peptide-treated surfaces. Almost no nitrogen was found in the other groups. \*\* *p* < 0.01 vs. the polished and SLA surfaces (significant differences are marked only for the nitrogen content).

#### *3.2. E*ff*ects of VnP-16 Peptide on Cellular Responses of Human Osteoblast-Like Cells*

To investigate whether a human vitronectin-derived peptide, VnP-16, could mediate cell behavior of osteoblasts, cell attachment, spreading, and migration of human osteoblast-like cells, including HOS and MG-63, were assayed. The attachment of osteoblast-like cells was evaluated using a cell adhesion assay in a serum-free medium. Human plasma vitronectin strongly promoted cell attachment (Figure 2A upper, B) and spreading (Figure 2A lower, C) in osteoblast-like HOS cells. The VnP-16 peptide also promoted greater cell attachment (Figure 2A upper, B) and spreading (Figure 2A lower, C) than the BSA or SP control, and its attachment and spreading activities were comparable to those of

vitronectin (Figure 2A–C). In addition, vehicle and SP did not participate in cell migration in HOS cells. On the other hand, vitronectin and the VnP-16 peptide promoted cell migration in HOS cells, while the VnP-16 peptide was significantly less effective than vitronectin (Figure 2D). The VnP-16 peptide did not affect the proliferation or viability of HOS cells (Figure 2E), indicating that its stimulatory effect on the cell behavior of HOS cells was not due to cytotoxicity or enhanced cell proliferation. These results support that VnP-16 is functionally active in promoting osteoblastic responses.

**Figure 2.** Cell attachment, spreading, and migration of osteoblast-like HOS cells seeded on culture plates treated with vitronectin and synthetic peptides. (**A**) Photographs of osteoblast-like HOS cells adhering (upper panel) and spreading (lower panel) to culture plates treated with 1% bovine serum albumin (BSA), vitronectin (0.26 μg/cm2), scrambled peptide (SP), and VnP-16 peptide (10.5 μg/cm2). Bar = 100 μm. (**B**,**C**) Cell attachment (**B**) and spreading (**C**) to immobilized synthetic peptides. HOS cells were allowed to adhere to peptide-treated plates for1h(**B**) or 3 h (**C**) in serum-free medium. (**D**) Migration of osteoblast-like HOS cells induced by vitronectin and synthetic peptides. HOS cells were seeded into the upper chambers of transwell filters coated with vitronectin (0.26 μg/cm2), SP, or VnP-16 (10.5 μg/cm2) and were incubated for 24 h. ND, not detected. (**E**) The viabilities of osteoblast-like HOS cells treated with VnP-16 for 24 or 48 h. \*\* *p* < 0.01 vs. the SP-treated control group. Data in (**B**–**E**) (*n* = 4) represent the mean ± SD.

Next, to determine whether the effects of the VnP-16 peptide on the cell behavior of HOS cells were similar to those of other human osteoblast-like cells, we used human osteoblast-like MG-63 cells. Similar, but not identical, results were obtained for cell behavior in MG-63 cells. Human plasma vitronectin and VnP-16 peptide strongly promoted cell attachment (Figure 3A upper, B) and spreading (Figure 3A lower, C) in MG-63 cells compared to the BSA or SP control. In addition, the cell attachment and spreading activities of VnP-16 peptide were comparable to those of vitronectin (Figure 3A–C). Similarly, VnP-16 peptide did not affect the proliferation or viability of MG-63 cells (Figure 3D). However, the cell migration activities of vitronectin and the VnP-16 peptide were different between HOS and MG-63 cells. In other words, vitronectin and the VnP-16 peptide had no effect on cell migration in MG-63 cells (data not shown). Therefore, the cellular responses of the VnP-16 peptide to the human osteoblast-like cells HOS and MG-63 had different effects on cell migration but had similar effects on cell attachment, spreading, and viability.

**Figure 3.** Cell attachment and spreading of osteoblast-like MG-63 cells seeded on culture plates treated with vitronectin and synthetic peptides. (**A**) Photographs of osteoblast-like MG-63 cells adhering (upper panel) and spreading (lower panel) to culture plates treated with 1% bovine serum albumin (BSA), vitronectin (0.26 μg/cm2), scrambled peptide (SP), and VnP-16 peptide (10.5 μg/cm2). Bar = 100 μm. (**B**–**C**) Cell attachment (**B**) and spreading (**C**) to immobilized synthetic peptides. MG-63 cells were allowed to adhere to peptide-treated plates for1h(**B**) or 3 h (**C**) in serum-free medium. (**D**) The viabilities of osteoblast-like MG-63 cells treated with VnP-16 for 24 or 48 h. \*\* *p* < 0.01 vs. the SP-treated control group. Data in (**B**–**D**) (*n* = 4) represent the mean ± SD.

#### *3.3. Histomorphometry*

Every experimental animal was healthy, and no signs of diseases or pathologic states were found until the sacrifice. There were no special inflammatory or immune cells found in the light microscopic view of the specimens. After 14 days of implant insertion, sufficient mineralization was observed in each section (Figure 4A). The mean value and standard deviation of each group were 47.0% ± 7.5% for the turned surface, 64.4% ± 8.6% for SLA, 42.1% ± 18.1% for SP-treated and 65.0% ± 7.2% for the VnP-16-treated surface. The histomorphometric data showed significant differences in BIC (*p* = 0.027). However, the pairwise comparisons between the groups found no significant differences in BIC (Figure 4B). The mean values and standard deviation in the BA were 58.8% ± 6.0% for the turned group, 56.8% ± 6.4% for SLA, 56.6% ± 8.4% for the SP-treated group, and 61.5% ± 10.6% for the VnP-16-treated group (Figure 4C). There were no significant differences in BA among the groups.

**Figure 4.** The histologic views and histomorphometric data for bone responses to the turned, SLA, SP-treated SLA, and VnP-16-treated SLA titanium implant surfaces. (**A**) The demarcation lines (white arrowheads), difference in stained colors and maturity of the bone (cancellous or cortical) differentiate the new bone from the existing old bone. Here, the new bone is stained more reddish while the old bone is stained more blueish. (**B**) Bone-to-implant contact ratios were measured, which are defined as the percentage of the implant surface in contact with bone to the total implant surface at the region of interest, which was the area ranging from the bone crest to 2 mm in depth in this study (green edged rectangle in (**A**)). (**C**) The ratio of the area filled with bone to the total area of the region of interest (bone area, or BA ratio) was also measured for each implant.

#### **4. Discussion**

The results of these in vitro and in vivo studies indicated that the VnP-16-treated SLA titanium surface augments the initial bone response to a dental implant. The VnP-16 bioactive peptide is expected to accelerate early bone healing further when this peptide is applied to the SLA titanium dental implant. VnP-16 is another candidate biomolecule for stronger osseointegration into a dental implant that is clinically applicable, together with some laminin-derived peptides [3,5,23]. Because VnP-16 has both the upregulation of osteoblast activity and the downregulation of osteoclast activity, different from other peptides, the clinical applicability of this material is considered to be higher [9].

Since VnP-16 downregulates osteoclast activity, this vitronectin-derived peptide is applicable to osteoporotic patients as well as to normal patients. A previous study has already shown the improved bone healing capacity of VnP-16 in an in vivo experiment using ovariectomized rats [9]. The SLA titanium implant is well-known to have long-term clinical performance with high survival

rates (higher than 95%) [28,29]. However, this micro-roughened titanium surface has some limitations in the use for patients with a problem in bone metabolism, including osteoporosis [30]. The cell adhesion molecules can potentiate the bone healing capacities of the modified titanium surfaces used in dental clinics without immune responses, which is shown in this study. The effect of VnP-16 on osseointegration of the SLA dental implant needs to be evaluated for normal and osteoporotic patients.

This study used the physical adsorption method for the VnP-16 application to titanium surface. One of the main advantages of this method is the simplicity, that is, an easy process to functionalize the titanium implant surface while this method has several important disadvantages like low effective peptide concentrations, and denaturation of the three dimensional structures of proteins or peptides [19]. The concentration of VnP-16 was determined from a dose-response curve through the cell attachment assay, as in a previous study (data not shown) [9]. The lowest concentration, showing the maximal effect in cell attachment, was used in this study, which solved the problem of low effective peptide concentration. VnP-16 is a peptide composed of 12 amino acids, which are considered to have no specific three dimensional structure [5]. Furthermore, the other approaches to peptide immobilization require additional reactions and costs [19]. Therefore, physical adsorption was the method of choice in this study although more effective approach needs to be investigated continuously.

In the analysis of surface chemistry of this study, higher nitrogen contents on the peptide-treated surfaces imply the applications of the peptides to the titanium implant surfaces. However, the results of the amounts of carbon detected on the surfaces were hard to interpret, despite that carbon was the most abundant on all the surfaces investigated in this study. These unclear results are considered to be from the phenomenon of hydrocarbon contamination on titanium surface, which is usual when titanium is exposed to air [31,32].

Although the analysis of variance test for BIC in this study showed a *p*-value less than 0.05, the pairwise comparisons found no significant differences between the groups. Perhaps, large standard deviation, especially obtained from the measurements of the turned and SP-treated implants, caused no significant differences in the pairwise comparisons. The high mean BIC ratios in the SLA and VnP-16-treated groups and the low mean BIC in the turned and SP-treated groups were considered to contribute to the significant difference in the analysis of variance test for all the groups. The large standard deviation of the data occurred because of the small sample size in this study. Another reason for the large deviation might be the difficulty in displaying the entire three-dimensional bone-implant interface in light microscopic histology. The selection of one cross-sectional plane for light microscopy is arbitrary, and the data from the cross-section are poorly correlated with the data measured on the whole three dimensional image [33]. In order to obtain the similar data between two- and three-dimensional images, three to four histologic sections for each specimen are needed, which are extremely difficult to prepare from a undecalcified specimen, including a titanium implant [34]. Methodological advancements, like three dimensional imaging analysis by micro-computed tomography, and a technique to make more histologic sections from a hard specimen are needed for more obvious in vivo results.

#### **5. Conclusions**

A human vitronectin-derived peptide, VnP-16 (RVYFFKGKQYWE motif), showed excellent histomorphometric osseointegration data without any special antigen–antibody reaction when this peptide was treated on an SLA titanium dental implant, which has been successfully used in clinics. From the in vitro results of this study, VnP-16 promotes the attachment of osteogenic cells and differentiation into osteoblasts, which may increase the bone healing capacity of the SLA's titanium surface. Considering both the in vitro and in vivo results of this study, VnP-16 reinforces the osteogenic potential of the SLA titanium dental implant when this peptide is applied to the SLA surface. In the future, VnP-16 may expand the clinical indications of SLA titanium dental implants.

**Author Contributions:** Conceptualization, I.-S.L.Y. and B.-M.M.; methodology, I.-S.L.Y. and B.-M.M.; software, C.-B.C., S.Y.J., C.Y.P., and H.K.K.; validation, I.-S.L.Y. and B.-M.M.; formal analysis, C.-B.C. and S.Y.J.; Investigation, C.-B.C., S.Y.J., C.Y.P., and H.K.K.; resources, I.-S.L.Y. and B.-M.M.; data curation, C.-B.C. and S.Y.J.; writing—original draft preparation, C.-B.C. and S.Y.J.; writing—review and editing, C.Y.P., H.K.K., I.-S.L.Y., and B.-M.M.; visualization, C.-B.C. and S.Y.J.; supervision, I.-S.L.Y. and B.-M.M.; project administration, I.-S.L.Y. and B.-M.M.; funding acquisition, I.-S.L.Y. and B.-M.M.

**Funding:** This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (2016R1A2B4014330 and 2019R1F1A1054209).

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

#### **References**


© 2019 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/).

## *Review* **Fracture Resistance of Zirconia Oral Implants In Vitro: A Systematic Review and Meta-Analysis**

#### **Annalena Bethke 1, Stefano Pieralli 1,2, Ralf-Joachim Kohal 2, Felix Burkhardt 1,2, Manja von Stein-Lausnitz 1, Kirstin Vach <sup>3</sup> and Benedikt Christopher Spies 1,2,\***


Received: 20 December 2019; Accepted: 21 January 2020; Published: 24 January 2020

**Abstract:** Various protocols are available to preclinically assess the fracture resistance of zirconia oral implants. The objective of the present review was to determine the impact of different treatments (dynamic loading, hydrothermal aging) and implant features (e.g., material, design or manufacturing) on the fracture resistance of zirconia implants. An electronic screening of two databases (MEDLINE/Pubmed, Embase) was performed. Investigations including > 5 screw-shaped implants providing information to calculate the bending moment at the time point of static loading to fracture were considered. Data was extracted and meta-analyses were conducted using multilevel mixed-effects generalized linear models (GLMs). The Šidák method was used to correct for multiple testing. The initial search resulted in 1864 articles, and finally 19 investigations loading 731 zirconia implants to fracture were analyzed. In general, fracture resistance was affected by the implant design (1-piece > 2-piece, *p* = 0.004), material (alumina-toughened zirconia/ATZ > yttria-stabilized tetragonal zirconia polycrystal/Y-TZP, *p* = 0.002) and abutment preparation (untouched > modified/grinded, *p* < 0.001). In case of 2-piece implants, the amount of dynamic loading cycles prior to static loading (*p* < 0.001) or anatomical crown supply (*p* < 0.001) negatively affected the outcome. No impact was found for hydrothermal aging. Heterogeneous findings of the present review highlight the importance of thoroughly and individually evaluating the fracture resistance of every zirconia implant system prior to market release.

**Keywords:** dental implant; zirconia; ceramics; aging; artificial mouth; fracture load; fatigue; chewing simulation; meta-analysis

#### **1. Introduction**

To date, titanium can be considered the gold standard material in oral implantology [1]. However, due to increasing esthetic standards and a discussed impact of metal/titanium particle release on the pathogenesis of peri-implant bone loss [2,3], a renaissance of ceramic oral implants can be observed in dental media. Nowadays, the market share of zirconia oral implants seems to be increasing, even if still comparatively small compared to conventional titanium implants.

Nonetheless, the superiority of ceramic oral implants regarding esthetics and biocompatibility, or, as an example, the frequently claimed patients' demand for metal-free implantology are still not

soundly scientifically evidenced. Nevertheless, the majority of dental experts are of the opinion that zirconia oral implants will be coexistent with titanium implants in the near future [4].

When zirconium dioxide (zirconia, ZrO2) was introduced as ceramic implant material, research focused to evaluate and improve its osseointegrative potential by creating a microroughened surface topography [5]. In the first instance, parameters like bone-to-implant contact (BIC), push-in values and removal torque were assessed in animal experiments. As a result, zirconia implants with various surface modifications (additive by sintering a porous ceramic layer, subtractive by sandblasting and/or acid-etching or, for example, by texturing the inner surface of a mold in case of an injection-molded implant) can nowadays be considered comparable to titanium implants by means of osseointegration in preclinical studies [6]. This finding was confirmed in clinical trials, however limited to short- and mid-term observation periods and the replacement of up to three adjacent missing teeth (single-tooth restorations and three-unit fixed dental prostheses) using one-piece ceramic implants [7].

From a technical point of view, such a 1-piece design, comprising the abutment and endosseous part in a single piece, might benefit from increased fracture resistance and reduced susceptibility for low-temperature degradation or so-called "aging" (by exposing a reduced total surface area to aging by inducing oral fluids), compared to 2-piece ceramic implants. Furthermore, 1-piece implants do not have a micro-gap in between the assembled implant and abutment. One might consider the absence of such a micro-gap beneficial, since it is capable in hosting bacteria, potentially resulting in marginal inflammation and consecutive bone resorption [8]. However, no advantage of a monobloc design was found for "seamless", 1-piece implants made from titanium [9]. Moreover, from a practitioner's point of view, a 1-piece implant design is associated with several surgical and prosthodontic shortcomings [10]. As an example, submerged implant healing is hardly possible, since the transmucosal part of a 1-piece implant cannot be detached. If no sufficient primary stability can be attained or guided bone regeneration is necessary, a missing option for wound closure might be considered disadvantageous. Furthermore, there is only a limited potential to compensate for mal-positioned implants with the provisional and final restoration. When trying to remove subsections in case of misaligned implants to support a bridge, intra-oral grinding of the zirconia abutment is necessary [11]. This, however, might have an impact upon the osseointegration (due to potential heat development or the displacement of zirconia particles in surrounding tissues) and fracture resistance of the implant [12]. Therefore, a two-piece design represents the favorable option for daily clinical use. Today, several two-piece zirconia implants are available on the market. In these systems, implant-abutment assembly is mostly realized by either luting the abutment to the implant or by screw-retention [13]. Luting the abutment to the implant seals the micro-gap, and allows for initial but irreversible correction of the implant angulation, but misses flexibility for future restorations of the implant. On the other hand, when going for screw-retention, several ceramic implants are still assembled with a titanium screw, and therefore, still not metal-free in the proper sense.

Even if the market share of zirconia dental implants increases, concerns regarding their fracture resistance are still present, and standardized testing protocols for zirconia implants adequately addressing the aging behavior of the final product are still missing [14]. To overcome this, different treatments were proposed to mimic intraoral conditions to the extent possible for the evaluation of ceramic implants. These treatments included thermal aging (high-temperature conditions or thermal cycling) [15,16] and/or dynamic loading procedures (various exposure times and different applied loading modes) [12,17]. Zirconia implants evaluated regarding their fracture resistance in the literature comprised a heterogeneous range of features like material selection (yttria-stabilized tetragonal zirconia polycrystal, Y-TZP or alumina-toughened zirconia, ATZ) [18], design (1- or 2-piece) [13], manufacturing (subtractive or by ceramic injection molding, CIM) [19], restoration (anatomical crown, hemisphere or no restoration) [20,21], abutment preparation (in the case of 1-piece implants) [22], or assembly (in the case of 2-piece implants) [13].

Therefore, the objective of the present systematic review was to evaluate the influence of the aforementioned treatments and features on the fracture resistance of zirconia oral implants in different preclinical studies. The null hypothesis supposed no distinction between treatments and features in relation to bending moment when statically loading the implant to fracture.

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

#### *2.1. Study Design*

To determine a selection of comparable studies on the question of zirconia implant fracture resistance, the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement of 2009 was applied [23]. Therefore, this report takes the appropriate Enhancing the Quality and Transparency of health Research (EQUATOR) (http://www.equator-network.org) guidelines into account.

#### *2.2. Focused Question*

Is there a variable significantly affecting the fracture resistance of 1- and 2-piece zirconia implants in preclinical in-vitro studies?

#### *2.3. Search Strategy*

Two databases, namely the Medical Literature Analysis and Retrieval System Online (MEDLINE) (PubMed) and Embase (accessed via Ovid), were screened for relevant articles. The database specific search strategies consisted of a combination of subject headings and free text words. Data was extracted from the databases on 3rd December 2019 without applying any time restrictions. Thereafter, references of included articles were screened for further records satisfying the inclusion criteria (cross-referencing). In case of the availability of the full methodological procedures in the literature and accessibility of information regarding the included samples, unpublished data of the authors of the present review was likewise included. The resulting studies were imported and stored in a reference managing program (EndNote X9; Clarivate Analytics, Philadelphia, PA, USA). Articles written in English and the German language were considered.

#### *2.4. Screening Process*

To build up the search terms, three categories addressing the samples (dental implants), materials (zirconia ceramics) and outcome (fracture load) were combined ("AND"). These categories consisted of combinations ("OR") of free text words and indexed vocabulary (MEDLINE: MeSH terms, Embase: Emtree terms). An asterisk was used in combination with some free text words as a truncation symbol (e. g. "ceramic \*") to allow for the so-called "wildcard search".

#### Pubmed search term:

*((((dental implant [MeSH Terms]) OR ((((oral) AND ((implant) OR implants))) OR ((dental) AND ((implant) OR implants))))) AND (((zircon \*) OR ceramic \*) OR ceramics[MeSH Terms])) AND (((((ageing) OR aging) OR artificial mouth) OR fracture resistance) OR load \*)* 

#### Embase search term:

*('tooth implant'/exp OR (oral AND implant) OR (dental AND implant)) AND (zircon \* OR ceramic \* OR 'ceramics'/exp) AND (ageing OR aging OR (artificial AND mouth) OR (fracture AND resistance) OR load \*)* 

#### *2.5. Eligibility Criteria*

Studies to be included in this systematic review needed to fulfill the following inclusion criteria:

#### - Language: English or German


#### *2.6. Selection of Studies*

Concerning the inclusion criteria, both the first author and the senior author of this manuscript (A.B. and B.C.S.) independently screened the titles and abstracts of the extracted data in the reference management program. If sufficient information needed for inclusion or exclusion was not provided within the title or abstract, the corresponding full texts were read. In case of disagreement, a third author (S.P.) was consulted for final decision making.

#### *2.7. Data Extraction*

Besides the total number of samples within one study, the number of implants made from different materials (Y-TZP, ATZ), processing routes (subtractive, injection molding), design (1- and 2-piece) and diameters were retrieved. Further features like restoration mode (anatomical crown, hemisphere or no reconstruction), abutment preparation (yes/no in case of 1-piece implants), implant-abutment connection (screwed/bonded in case of 2-piece implants), thermal aging (thermal cycling, high temperature, no aging) or dynamic loading (yes/no), dynamic loading conditions (exerted load and amount of cycles), crosshead speed during static fracture, and angulation, were likewise extracted. This allowed us to group the implants finally subjected to static loading within the included studies in cohorts. For standardization purposes, the bending moment at the time point of fracture [Ncm] was considered the outcome measure of interest, and the corresponding authors of the articles to be included were contacted by email in case of solely providing fracture load values [N] without mentioning the lever arm. Extracted cohorts were subdivided into groups subjected to comparable treatments:


#### *2.8. Statistical Analysis*

From the included nineteen studies/datasets, two to twelve observations were extracted each. One observation consisted of the mean bending moment and standard deviation (at the time point of fracture) and/or mean fracture load and standard deviation (including additional information allowing us to calculate the bending moment) of a specific cohort of implants (comprising the same type of implant subjected to the same treatment) extracted from one included study. These observations had sample sizes of 2 to 12 implants. To analyze the effect of specific treatments of features (as indicated in 2.7) on the bending moment, a multilevel mixed-effects generalized linear model was used for each outcome, with each investigation as random effect to cluster observations by the respective studies. The Šidák method was used to correct for multiple testing. The level of significance was set at *p* < 0.05.

In order to compare the aforementioned groups (1–6, depending on load and cycles) for heterogeneity of the data, both inter- and intra-standard deviations with 95% confidence intervals (Cis) were computed. In addition, the cohort-specific standard error of the bending moment was used for weighting. Furthermore, box plots were created for visualization of the data. The data were analyzed with STATA 16.1 (StataCorp LLC, Texas, TX, USA).

#### **3. Results**

#### *3.1. Screening Process*/*Included Data*

Screening of two databases using the aforementioned specifically adapted search terms resulted in a total of 1864 records. After the removal of 622 duplicates, another 1202 records were withdrawn for analyses by screening the titles and abstracts. After reading the full texts of the remaining 40 studies, a further 23 manuscripts were excluded (Figure 1). Detailed reasons for exclusion can be found in Table A1. In general, the most frequent reasons for exclusion were the fracture of zirconia abutments assembled with titanium implants (mostly excluded by title and abstract) and the fracture on the restoration level using zirconia one-piece implants as support (mostly excluded during full-text screening). When only the fracture load [N] during static loading was reported, three options allowed for the calculation of the bending moment: (1) embedding was described to fully respect ISO 14801 (prescribing a lever arm of 5.5 mm allowing for the calculation of the bending moment), (2) all details regarding the embedding were provided in the manuscript (e.g., by providing a scheme) or (3) the bending moment and/or lever arm were provided by the authors upon request. As an example, six of the included studies adopted ISO 14801 for embedding [15,17,21,24–26], whereas three provided all necessary information [19,27,28] allowing us to calculate the bending moment (embedding level, angulation, total sample length, point of loading). In the remaining cases the bending moment was reported [13,20,22] or sent by the authors [12,18,29,30]. Finally, 17 full-texts were analyzed in the present systematic review (Table 1). In addition, the datasets of two finalized projects, currently under review and in preparation of the manuscript, were included. Two authors of the present review (R.K. and B.C.S.) were involved in both of these two investigations, and were able to access the full data. The applied materials and methods were already described in detail in precedent publications [21,26]. Since available on the market, the material composition of the included implant systems is likewise available and accessible. In detail, three zirconia implant systems (1-piece: Straumann PURE Ceramic, Straumann AG, Basel, CH; 2-piece: 5s-50-10, Z-Systems AG, Oensingen, CH and Ceralog Hexalobe Implant, Axis biodental, Les Bois, CH) were subjected to identical treatments and fracture load measurements, as described in two of the included studies [21,26]. In the case of Straumann 1-piece (as-received: 609 ± 20 Ncm; loaded/aged: 557 ± 36 Ncm) and Z-Systems 2-piece implants (as-received: 463 ± 21 Ncm; loaded/aged: 443 ± 39 Ncm), aging/loading (as described in [21,26]) did not affect the fracture resistance to a statistically significant level (*p* = 0.171). In contrast, the fracture resistance of 2-piece Ceralog Hexalobe Implants (as-received: 547 ± 89 Ncm; loaded/aged: 413 ± 127 Ncm) was significantly affected (*p* = 0.046) by aging/loading (as described in [21,26]).

**Figure 1.** Flowchart according to the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines.


**Table 1.** A total of 731 one- and two-piece implants made from yttria-stabilized tetragonal zirconia polycrystal (Y-TZP) and alumina-toughened zirconia (ATZ), extracted from 17 studies and two unpublished datasets, subjected to different dynamic loading and thermal aging conditions prior to static loading to fracture, were finally included in meta-analyses.

\* Unpublished data, Ref. = Reference, *n* = total number of included implants, TC = thermal cycling, HT = high temperature.

#### *3.2. Meta-Analyses*

All 17 articles published between 2009 [29] and 2019 [15,16] were included and analyzed in the present meta-analysis. Moreover, unpublished data of two projects currently under review and in preparation of the manuscript were included (Table 1). From the included articles/datasets, 114 observations were extracted or calculated (mean bending moment), comprising different implant features (e.g., diameter, material, crown supply, abutment preparation or implant-abutment-connection) or treatments (e.g., thermal aging or dynamic loading). One observation consisted of the mean bending moment and standard deviation (SD) of up to 12 included implants.

In order to evaluate the impact of different dynamic loading procedures (implants were subjected to prior to fracture loading) on the outcome (bending moment), groups as indicated in Section 2.7 were analyzed for heterogeneity. As a result, standard deviation as a measure of variation within and in between the included studies revealed to be within the same range (Table 2). No heterogeneity of the bending moments for groups 1–6 was found, even if a decreased mean value for group 3 was calculated (*p* = 0.612). This did not change when stratifying the implants according to their design (1-piece: *p* = 0.951; 2-piece: *p* = 0.056).

**Table 2.** Groups 1–6 (as indicated in 2.7) were tested for heterogeneity regarding the outcome.


<sup>1</sup> Mean bending moment [Ncm], <sup>2</sup> Standard deviation/variation within included studies, <sup>3</sup> Standard deviation/variation in between included studies.

#### *3.3. Outcomes*

Outcomes extracted from the 17 included studies and the two unpublished datasets were calculated and stratified for the material selection, manufacturing, implant diameter, anatomical crown supply, abutment preparation (1-piece implants), implant-abutment-connection (IAC; 2-piece implants), thermal aging procedure prior to static loading (none; TC = thermal cycling, mostly in between 5–55 ◦C; HT = high temperature, mostly in between 60–134 ◦C) and/or dynamic loading in a chewing simulation device applying different loads (ranging from 50 to > 500 N) for a different amount of cycles (ranging from 1 to 10 millions). In total, 731 implants were available for analyses, revealing a mean bending moment at the time point of fracture of 386.4 ± 167.6 Ncm. Furthermore, the outcome was stratified for 1- and 2-piece implants. Mean bending moments, standard deviations and the included number of implants are listed in Table 3. Significance (linear mixed models, level of significance *p* < 0.05) calculated for differences regarding the implant design, different covariables and treatments can be found in Table 4.

**Table 3.** Calculated mean bending moment (in Ncm) and standard deviation depending on the implant design, several covariables and treatments.


*n* = number of included implants, SD = standard deviation, <sup>1</sup> 1- and 2-piece implants pooled together, <sup>2</sup> the authors of one included study could not provide the manufacturing mode for all included implants [28].


**Table 4.** Significance (linear mixed models (LMMs), level of significance *p* < 0.05) was calculated for differences regarding the implant design, different covariables and treatments.

<sup>1</sup> 1- and 2-piece implants pooled together, TC = thermal cycling, HT = high temperature.

#### 3.3.1. Implant Design

Eight studies [12,15,17,18,20,22,25,29] focused on 1-piece zirconia implants, whereas six studies solely included 2-piece implants [16,19,21,24,26,30]. The remaining investigations evaluated a mixture of both 1- and 2-piece implants [13,27,28]. Regardless of all other variables, 1-piece implants (431.9 ± 151.0 Ncm) were found to be more fracture resistant than 2-piece implants (291.7 ± 162.4 Ncm, *p* = 0.004; Figure 2).

**Figure 2.** Boxplot showing the bending moment at the time point of fracture for 1- and 2-piece zirconia implants. Whiskers are used to represent all samples lying within 1.5 times the interquartile range (IQR). Dots represent outliers. Detailed data can be found in Tables 3 and 4.

#### 3.3.2. Material

Material selection of the included studies is listed in Table 1. Of the included implants, 577 were made from Y-TZP, whereas 154 were manufactured from ATZ [13,18,19,22,24,26]. When pooling the outcome for 1- and 2-piece zirconia implants, the bending moment at the time point of implant fracture was significantly affected by the material (*p* = 0.002; Table 4). In detail, implants made from alumina-toughened zirconia (ATZ, 418.7 ± 106.0 Ncm) were more fracture-resistant compared to implants made from yttria-stabilized tetragonal zirconia polycrystals (Y-TZP, 378.7 ± 160.1 Ncm, *p* = 0.002). When stratifying the outcome for 1- and 2-piece implants, however, material selection only affected 1-piece implants (*p* = 0.001, Figure 3a), whereas 2-piece implants performed the same, regardless of the material selection (*p* = 0.282, Figure 3b).

**Figure 3.** Boxplots showing the bending moment at the time point of fracture depending on the material selection for 1- (**a**) and 2-piece (**b**) zirconia implants. Whiskers are used to represent all samples lying within 1.5 times the interquartile range. Dots represent outliers. Detailed data can be found in Tables 3 and 4.

#### 3.3.3. Manufacturing

Manufacturing was mostly subtractive (*n* = 591 implants), but ceramic injection-molding (CIM) was likewise used for the production (*n* = 120 implants) [15,19,21,25]. There was no statistically significant difference in the fracture resistance of implants when manufacturing method (subtractive: 397.5 ± 177.4 Ncm, CIM: 364.8 ± 116.7 Ncm) was regarded (*p* > 0.095). Boxplots can be seen in Figure 4.

**Figure 4.** Boxplots showing the bending moment at the time point of fracture depending on the manufacturing method for 1- (**a**) and 2-piece (**b**) zirconia implants. Whiskers are used to represent all samples lying within 1.5 times the interquartile range. Detailed data can be found in Tables 3 and 4.

#### 3.3.4. Implant Diameter

No statistically significant difference could be calculated for the bending moment at the time point of fracture regarding the implant diameter ranging from 3 to 5 mm (*p* = 0.327). This did not change when stratifying the outcome for 1- (*p* = 0.273) and 2-piece (*p* = 0.191) implants. However, the included studies evaluated only very few implants in the range of 3 mm (range: 3.0–3.3 mm; *n* = 15, 207.2 ± 14.3 Ncm) [24,27] and 5 mm (range: 4.5–5.0 mm; *n* = 41, 349.4 ± 125.4 Ncm) [15,19,27,28], whereas the majority of implants had a diameter in the range of 4 mm (range: 3.8–4.4 mm; *n* = 675, 394.9 ± 170.4 Ncm). Boxplots can be seen in Figure 5.

**Figure 5.** Boxplots showing the bending moment at the time point of fracture depending on the implant diameter for 1- (**a**) and 2-piece (**b**) zirconia implants. Whiskers are used to represent all samples lying within 1.5 times the interquartile range. Dots represent outliers. Detailed data can be found in Tables 3 and 4.

#### 3.3.5. Anatomical Crown Supply

Of the included 731 implants, 209 were restored with an anatomically shaped crown, mostly made from ceramic materials. Most of the crowns were designed to replace maxillary central incisors but also some premolar reconstructions were included. The remaining 522 implants did not receive any reconstruction and were directly loaded to the abutment or were equipped with a non-anatomical stainless-steel hemisphere according to ISO 14801. When pooling the data for 1- and 2-piece implants, anatomical crown supply (237.5 ± 96.6 Ncm) negatively affected the outcome compared to implants with no crowns or equipped with a hemisphere (455.2 ± 147.7 Ncm, *p* < 0.0001). When stratifying for 1- and 2-piece implants (Figure 6), statistical significance was only reached for the group of 2-piece implants (*p* < 0.0001), likewise revealing an inferior outcome for implants restored with anatomical crowns. Fracture resistance of 1-piece implants was not affected by crown supply (*p* = 0.080).

**Figure 6.** Boxplots showing the bending moment at the time point of fracture depending on the crown supply for 1- (**a**) and 2-piece (**b**) zirconia implants. Whiskers are used to represent all samples lying within 1.5 times the interquartile range. Dots represent outliers. Detailed data can be found in Tables 3 and 4.

#### 3.3.6. Abutment Preparation and Implant-Abutment-Connection (IAC)

Of the 1-piece implants (*n* = 495), 112 abutments were prepared/modified by grinding [12,18,22,29], whereas 383 abutments remained untouched until fracture. In most cases, abutment preparation should simulate a clinically relevant situation of a 1-piece implant installed in anterior regions of the mouth. In both groups, some implants were restored with anatomically shaped incisor crowns, and some did not receive any reconstruction. Grinding of the abutment (411.3 ± 126.2 Ncm) resulted in a significantly reduced bending moment at the time point of fracture compared to non-grinded implants (436.5 ± 156.5 Ncm, *p* < 0.0001; Figure 7a).

Of the two-piece implants included in the present review (*n* = 236), 159 abutments were assembled by screw retention [13,16,19,21,24,26]. Most screws were made from titanium, but also gold and polyetheretherketone (PEEK; in one study, carbon-fiber-reinforced [26]) were used. The remaining 77 two-piece implants were irreversibly assembled by adhesive bonding [13,19,27,28,30]. The type of abutment retention (screw-retained: 327.5 ± 179.0 Ncm, bonded: 217.0 ± 86.0 Ncm) did not affect the fracture resistance (*p* = 0.584; Figure 7b).

**Figure 7.** Boxplots showing the bending moment at the time point of fracture depending on the abutment preparation for 1-piece (**a**) and depending on the implant-abutment-connection (IAC) for 2-piece (**b**) zirconia implants. Whiskers are used to represent all samples lying within 1.5 times the interquartile range. Dots represent outliers. Detailed data can be found in Tables 3 and 4.

#### 3.3.7. Thermal Aging

Regardless of the implant design, in 297 implants, no aging was induced prior to static loading to fracture, whereas 124 implants were subjected to a high temperature (HT) treatment in a humid environment, ranging from 60 up to 134 ◦C for different time periods lasting from 5–30 h (134 ◦C) [15,25] to 60 days (85 ◦C) [21,26]. High temperature treatment was applied in combination or during dynamic loading or alone. The remaining 310 implants were subjected to a thermal cycling (TC) procedure, exposing the samples to a changing water bath set at 5 and 55 ◦C [12,16,18–20,22,27–30]. The latter was mostly performed during dynamic loading in a chewing simulation device. Compared to untreated implants (406.2 ± 180.4 Ncm), neither HT treatment (392.9 ± 115.9 Ncm) nor TC (355.5 ± 171.7 Ncm) did affect the fracture resistance (*p* = 0.446). This did not change when calculating the outcome for 1- (*p* = 0.538) and 2-piece implants (*p* = 0.776) separately (Figure 8).

**Figure 8.** Boxplots showing the bending moment at the time point of fracture, depending on the thermal aging conditions (none, HT = high temperature, TC = thermal cycling) for 1- (**a**) and 2-piece (**b**) zirconia implants. Whiskers are used to represent all samples lying within 1.5 times the interquartile range. Dots represent outliers. Detailed data can be found in Tables 3 and 4.

#### 3.3.8. Dynamic Loading

The effect of dynamic loading was evaluated from different perspectives. The simplest one assigned the included implants to two categories subjected to either no dynamic loading procedure ("No") or those being subjected to dynamic loading ("Yes"; Figure 9, Table 4). Furthermore, the effect of dynamic loading was evaluated regarding the dynamically "applied load", ranging from 45 [30] up to more than 500 N [17], or regarding the "amount of cycles" ranging from 1.2 [12,16,18,22,28–30] to 10 million [13,20,21,26] loading cycles. Finally, a combination of "applied load" and "amount of cycles" was used to from six groups, as mentioned in Section 2.7 (Figure 10).

When pooling the extracted data for 1- and 2-piece implants, dynamic loading did not affect the fracture resistance (dynamically-loaded implants showed a mean bending moment at the time point of fracture of 389.4 ± 169.2 Ncm compared to 383.2 ± 166.3 Ncm calculated for non-loaded implants (*p* = 0.410)). This did not change when evaluating 1- and 2-piece implants separately (*p* > 0.474). Solely the category "applied load" was close to statistical significance (*p* = 0.05). However, none of the multiple pairwise comparisons comparing different dynamically applied loads showed a statistically significant difference (*p* > 0.07). When solely evaluating 2-piece implants, "amount of cycles" significantly affected the fracture resistance (*p* < 0.0001), whereas "applied load" (*p* = 0.202) and groups 1–6 respecting the applied load and the amount of cycles (*p* = 0.056) did not affect the outcome.

**Figure 9.** Boxplots showing the bending moment at the time point of fracture depending on dynamic loading (Yes: Implants were subjected to dynamic loading, No: Implants were not dynamically loaded) for 1- (**a**) and 2-piece (**b**) zirconia implants. Whiskers are used to represent all samples lying within 1.5 times the interquartile range. Dots represent outliers. Detailed data can be found in Tables 3 and 4.

**Figure 10.** Boxplots showing the bending moment at the time point of fracture depending on dynamic loading conditions respecting the applied load and amount of cycles (as categorized in Section 2.7) for 1- (**a**) and 2-piece (**b**) zirconia implants. Whiskers are used to represent all samples lying within 1.5 times the interquartile range. Dots represent outliers. Detailed data can be found in Tables 3 and 4. No 2-piece implants were allocated to group 5.

#### **4. Discussion**

The present systematic review and meta-analysis included the data of 17 studies and two unpublished datasets. To be finally able to compare the outcomes of the included data, it was necessary to extract or calculate the bending moment at the time point of implant fracture [Ncm], since the mostly reported fracture load values [N] do not respect the leverage (length of the lever arm) and are therefore, if not considering a rigorously standardized embedding procedure as described in ISO 14801, not comparable to each other. Of the included 19 investigations/datasets, three studies reported the bending moment individually calculated for each included implant [13,20,22], whereas six studies [15,17,21,24–26] and the two included unpublished datasets fully respected ISO 14801 for embedding. Fully respecting this ISO implies the fixation of the endosseous part in a rigid clamping device or embedding in a material with a modulus of elasticity higher than 3 GPa. Moreover, the embedding/clamping level should respect a distance of 3.0 ± 0.5 mm apically from the nominal bone level, as specified in the manufacturer's instructions for use. Furthermore, implant abutments need to be equipped with a non-anatomical hemisphere designed to realize a distance of *l* = 11.0 ± 0.5 mm from the center of the hemisphere to the embedding level (Figure 11).

(**a**) Hemisphere (**b**) Anatomical crown (**c**) No crown

**Figure 11.** Exemplary schemes of embedded implants according to ISO 14801 (**a**) [21], equipped with an anatomically shaped incisor crown (**b**) or without any restorative supply (**c**) [20]. When embedding the samples according to ISO 14801, the lever arm measures 5.5 mm. In the latter two cases, the lever arm needs to be individually calculated and reported.

When loading such samples with an angle of α = 30◦ to the vertical, the lever arm (*y*) or bending moment (*M*) for this configuration can be calculated with the reported fracture load (*F*) by using Equation (1).

$$M = y \cdot F = \sin \alpha \cdot l \cdot F \tag{1}$$

This results in *y* = 0.55 cm when embedding according to ISO 14801. For the aforementioned publications/datasets fully respecting ISO 14801 for embedding and reporting the fracture load values [N], the bending moment was therefore calculated by multiplying the fracture load with 0.55. Interestingly, some of the included investigations reported embedding according to ISO 14801, but solely adopted the embedding level (simulation of a bony recession of 3 mm), and sometimes the angulation (30◦), but did not use a loading hemisphere, finally resulting in a lever arm different to 5.5 mm, as proposed by the ISO standard [16,19,27,28]. In most cases, anatomical crowns (maxillary premolars or incisors) made from ceramic materials were used instead of the hemisphere, finally resulting in altered lever arms and loading conditions. In the investigations of one group, the crown design and embedding procedure were described in detail (*l*, α and *F* were reported), allowing us to calculate *y* and *M* [19,27,28]. To calculate the bending moment for the remaining studies, authors needed to provide the necessary data upon request or standardized photographs provided in the publications, or by the authors needed to allow the approximation of the lever arm by using an image analysis software (ImageJ, National Institutes of Health, Bethesda, MD, USA) [12,16,18,29,30]. In order to be able to compare the outcome of preclinical studies evaluating the fracture resistance of dental implants, it is therefore recommended to either fully adopt an ISO standard for the embedding procedure or to provide the bending moment additionally to the fracture load. Considering different lever arms due to different embedding procedures for the implants included in this systematic review and meta-analysis, one needs to keep in mind that dynamic loading prior to static loading to fracture can result in altered fatigue, even if the applied load was the same.

The heterogeneity of the included samples comprising a mixture of market-available products (finally sterilized and incorporating a micro-roughened surface) [15,16,22,24,26] but also prototype implants (e.g., with or without any surface post-processing) [13,19,21,25,28,30] represents a major limitation of the present systematic review and meta-analysis. However, it was shown that, for example, surface modifications like micro-roughening to enhance osseointegration or steam-sterilization can significantly compromise fracture strength and ageing kinetics [31,32].

Another shortcoming of this systematic review presents the fact that of the 19 included datasets, more than half (nine published and two unpublished studies) were at least partially authored by the collaborates of the current paper. This might be considered a reasonable risk of bias. However, the present review was conducted according to standardized guidelines, and the available literature was systematically screened on the basis of predefined search terms and inclusion criteria. Modifying the search strategy, outcome measure or inclusion criteria in consequence of unexpected or homogeneously authored findings would likewise present a source of bias.

Regarding the treatments, the included samples have been subjected to prior to loading, and six groups (representing different categories of loading conditions as indicated in Section 2.7) have been evaluated for heterogeneity of the outcome. As a result, no heterogeneity of the bending moments for groups 1–6 was found (*p* = 0.612). This did not change when stratifying the implants according to their design (1-piece: *p* = 0.951; 2-piece: *p* = 0.056). Therefore, it was decided to pool the data of all groups for any further calculations, and yet still, one can hardly generalize the present findings and apply them to a specific zirconia implant system.

No statistically significant influence of hydrothermal aging on the fracture resistance of zirconia implants was calculated in the present review. It is important to note that aging or so-called low-temperature degradation (LTD) can, depending upon the sample quality and surface conditions, result in both increased [21,25] and decreased [33] fracture load. This might be explained by the following: Assuming a zirconia sample surface with various process-related defects/impurities, the largest defects/impurities are thought to act as "locus minoris resistentiae", and can thereby be

considered representative for the fracture resistance of this sample. Increased fracture load of such zirconia samples after a hydrothermal aging procedure is thought to be attributed to a transformed layer at the sample surface, inducing a compressive stress on the surface, tending to close a potential advancing crack at such existing defects/impurities located on the surface. This phenomenon is liable to cause an increase in the strength of the material, and was described for the first time three decades ago [34]. On the other side, at some point when the degradation process penetrates deeper into the material, the contribution from the aging may instead cause the strength of the same sample to be decreased, since once transformed to the monoclinic, zirconia grains cannot exhibit stress-induced phase transformation toughening anymore [33]. As an example, in the included investigation of Monzavi and co-workers [15] the effect of artificial aging on the mechanical resistance and micromechanical properties of commercially- and noncommercially-available zirconia dental implants was evaluated. In this study, the bending moment was significantly increased after aging for three of six groups, whereas two groups showed no influence of the aging procedure, and one group was negatively affected in terms of fracture resistance by the treatment [15]. When pooling the outcomes of the included studies showing positive, negative or no effects of LTD on the fracture resistance of zirconia implants in one dataset, as happened in the present meta-analyses, no effect of hydrothermal aging on the bending moment at the time point of fracture was calculated (*p* > 0.446). This, however, might be misleading, since several of the included studies indeed showed that aging can significantly affect the fracture resistance. However, due to the explanation given at the beginning of this paragraph, both in a negative or positive way. Therefore, missing significance, as calculated for pooled data in this review, should not be interpreted as an argument to refrain from aging tests of a zirconia implant system prior to market release. Therefore, pooling the data from different studies using the different conditions of thermal aging needs to be considered a limitation of the present review. It is discussed in the literature that the present amount of transformation to the monoclinic on the surface of as-delivered zirconia implants can be decisive for the ongoing fracture resistance after further hydrothermal aging procedures. In detail, implants showing no or very limited transformation to the monoclinic when released to the market (e.g., due to final temperature annealing [35] or manufacturing by ceramic injection-molding [21,25]) were observed to be less fracture-resistant in the original as-delivered state, but significantly gained fracture resistance due to increasing compressive stress at the sample surface after transformation to the monoclinic occurred. In contrast, samples already revealing a transformed layer of several micrometers (e.g., due to subtractive manufacturing or post-processing steps like sandblasting in order to roughen the surface to enhance osseointegration [26]) mostly do not benefit from further aging by means of an increased fracture resistance. Besides the amount of already transformed grains, implant surface topography showed to have a significant impact on aging susceptibility and its impact on fracture resistance [32,36]. As an example, implants structured with porous or alveolar surfaces were more likely to be negatively affected by aging procedures due to interconnected porosities in the surface layer, offering a path for the transformation to start at every surface accessible by water [25]. Finally, a layer structured in this way can be transformed in a shorter period of time.

Of the implants included in the present investigation, 209 of 731 were restored with anatomically-shaped crowns [16,19,20,27–30]. Most of these crowns were designed as maxillary central incisors, and were manufactured from: lithium disilicate [20], veneered [29] or monolithic [19,27,28] zirconia, or porcelain fused to metal [30]. Another included study restored the implants with maxillary first premolar restorations made from lithium disilicate [16], whereas Joda and collaborates restored the implants with non-anatomical hemispheres likewise made from lithium disilicate [24].

Most of the included studies not restoring the implants with anatomically-shaped crowns were conducted by adopting ISO 14801. According to this standard, the loading force shall be applied to the hemispherical loading surface, by a loading device with a plane surface normal to the loading direction of the machine, without additional horizontal loading forces. In contrast, especially incisor crowns present an inclined plane when loaded during the dynamic and finally static loading procedure, resulting in an increased shear force. Additionally, some investigations applied horizontal forces

during the dynamic loading procedure (as it happens in the oral cavity), causing further fatigue of the sample [20,29,30]. Therefore, not the restoration itself, but the altered investigational setup, resulting in increased shear forces and fatigue during static loading, and in some cases, precedent chewing simulation might be considered responsible for decreased fracture resistance. Nonetheless, this finding should be taken into account when drafting international standards in order to guarantee clinical safety, since the anatomical reconstruction of zirconia oral implants and horizontal shear forces during loading represent clinical reality. Regarding the nature or location of failure, 1-piece implants mostly fractured at the embedding level or slightly below, with crack initiation on the tensile side of the implant. As described in the included studies, it seems that the fracture mode was not affected by crown supply. In 2-piece implants, fracture modes were generally observed to be highly heterogeneous, depending on the mode of assembly and the materials used.

When it comes to clinical reality, the fracture resistance of a zirconia implant should finally withstand the maximum voluntary bite forces of the patients. Nonetheless, one cannot find the definition of any indication specific (e.g., for implants installed in anterior or posterior regions) minimum value for the fracture strength of a zirconia implant in ISO 14801. This, as an example, is provided in detail in ISO 6872 for ceramic materials used for reconstructions (e.g., crowns, bridges) in dentistry [37]. Taking the highest bending moment measured in vivo (95 Ncm) with the help of strain gauge abutments into account [38], and applying a safety buffer of 100%, one might consider a minimum fracture resistance of 200 Ncm sufficient to guarantee clinical safety. When applying this requirement to the included studies, mostly 2-piece prototype implants and implants with a reduced diameter (≤ 3.3 mm) did not meet this demand [19,24,27,28,30].

Of the zirconia implants included in the present investigation, 577 were manufactured from Y-TZP and 154 from ATZ. Overall, implant stability was significantly affected by the material, in favor of ATZ (*p* = 0.002). When evaluating 1- and 2-piece implants separately, however, only 1-piece implants made from ATZ performed better (*p* = 0.001), whereas 2-piece implants performed the same, regardless of the material selection (*p* = 0.282). This might be explained by the fact that 1-piece zirconia implants or even, as an example, 2-piece titanium implants are mostly made from one single material (in the case of titanium: the implant, the abutment and the abutment screw are mostly fabricated from titanium). In contrast, most of the available 2-piece zirconia systems represent a multi-material complex comprising at least two or sometimes even three different materials. In some cases, only the implant body is manufactured from zirconia, whereas the screw (e.g., titanium or PEEK) and/or abutment (e.g., glass-fiber or polyetherketoneketone/PEKK) might be manufactured from different materials revealing different aging or degradation behavior during treatments (hydrothermal aging, dynamic loading), precedent to final static loading to fracture. To date, sound correlations to approximate intraoral aging conditions in an accelerated way in the dental laboratory are mostly available for zirconia ceramics, but missing for screw and abutment materials prone to degradation in aqueous environments, like e.g., polyetherketones [39,40]. In consequence, no standardized testing procedures were proposed to the present date, sufficiently evaluating multi-material, 2-piece implants regarding their fracture resistance, and individually respecting the degradation behavior of several included components. Regrettably, the sample size and heterogeneity of the extracted data gathered from 2-piece implants included in the present review did not allow for the statistical evaluation of a potential impact of the screw or abutment material on the fracture resistance of 2-piece zirconia implants. In one of the included studies, the aim was to measure the abutment rotation and fracture load of 2-piece zirconia implants screwed with three different abutment screw materials [16]. Implants and abutments of the included system were assembled with screws made from gold, titanium and PEEK.

As a result, no significant differences were found for these three materials, even if PEEK screws showed inferior results. When choosing PEEK as an abutment screw material, the incorporation of continuous carbon fibers proved to positively affect the maximum tensile strength of the screw [41]. However, a strengthening effect on the entire implant-abutment complex in case of zirconia implants still needs to be evidenced. In one of the included studies [26], a 2-piece ATZ implant system assembled with

a carbon-fiber-reinforced abutment screw showed to be non-inferior compared to a market-established 2-piece titanium implant of a highly comparable design regarding its fracture resistance.

#### **5. Conclusions**

The null hypothesis of the present review, supposing no distinction between treatments and features in relation to bending moment when statically loading a zirconia implant to fracture, needs to be partially rejected. The focused question can be answered as follows: In general, 1-piece implants can be considered more fracture resistant than 2-piece implants, even if some of the included studies showed very promising results for 2-piece zirconia implants. When focusing on 1-piece implants, implants made from ATZ are more fracture resistant than implants made from Y-TZP. Due to its negative impact on fracture resistance, abutment preparation of 1-piece zirconia implants should be avoided. When drafting international standards to guarantee clinical safety, one should keep in mind that the loading of anatomically shaped crowns might result in the decreased fracture resistance of zirconia implants compared to non-anatomical loading hemispheres, as mentioned in ISO 14801. Further research is needed to define adequate hydrothermal aging and dynamic loading conditions for 2-piece ceramic implants, nowadays mostly comprising a multi-material complex.

**Author Contributions:** Conceptualization, B.C.S.; methodology, B.C.S. and A.B.; validation, S.P., M.v.S.-L., F.B. and K.V.; statistical analysis, K.V.; writing—original draft preparation, B.C.S. and A.B.; writing—review and editing, S.P., K.V., R.-J.K., F.B. and M.v.S.-L. All authors have read and agreed to the published version of the manuscript.

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

**Acknowledgments:** We acknowledge support from the German Research Foundation (DFG) and the Open Access Publication Fund of Charité—Universitätsmedizin Berlin.

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

#### **Appendix A**



#### **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/).

*Review*
