**Improvement in Fatigue Behavior of Dental Implant Fixtures by Changing Internal Connection Design: An In Vitro Pilot Study**

#### **Nak-Hyun Choi 1,**†**, Hyung-In Yoon 2,**†**, Tae-Hyung Kim <sup>3</sup> and Eun-Jin Park 1,\***


Received: 9 September 2019; Accepted: 4 October 2019; Published: 7 October 2019

**Abstract:** (1) Background: The stability of the dental implant–abutment complex is necessary to minimize mechanical complications. The purpose of this study was to compare the behaviors of two internal connection type fixtures, manufactured by the same company, with different connection designs. (2) Methods: 15 implant–abutment complexes were prepared for each group of Osseospeed® TX (TX) and Osseospeed® EV (EV): 3 for single-load fracture tests and 12 for cyclic-loaded fatigue tests (nominal peak values as 80%, 60%, 50%, and 40% of the maximum breaking load) according to international standards (UNI EN ISO 14801:2013). They were assessed with micro-computed tomography (CT), and failure modes were analyzed by scanning electron microscope (SEM) images. (3) Results: The maximum breaking load [TX: 711 ± 36 N (95% CI; 670–752), EV: 791 ± 58 N (95% CI; 725–857)] and fatigue limit (TX: 285 N, EV: 316 N) were higher in EV than those in TX. There was no statistical difference in the fracture areas (*P* > 0.99). All specimens with 40% nominal peak value survived 5 <sup>×</sup> 106 cycles, while 50% specimens failed before 10<sup>5</sup> cycles. (4) Conclusions: EV has improved mechanical properties compared with TX. A loading regimen with a nominal peak value between 40% and 50% is ideal for future tests of implant cyclic loading.

**Keywords:** dental implants; fracture strength; mechanical stress; fatigue; dental implant–abutment connection; dental implant–abutment design

#### **1. Introduction**

Dental implants have been a fairly reliable and predictable treatment option for edentulous patients since their introduction [1]. In previous systematic reviews, a 5-year survival rate of 95.6–97.2% and 10-year survival rate of 93.1% in implants supporting fixed partial dentures were reported, implying that implants have a high survival rate [2–4]. Although dental implants have been clinically and scientifically studied as a viable treatment option to restore the edentulous area [5–7], complications remain a big concern for clinicians. Complication of dental implants can be mainly classified as either biological or mechanical. Biological complications include early loss of osseointegration, marginal bone loss, and peri-implantitis, eventually leading to the implants failing and falling out. Mechanical complications include loose abutment or screw, veneer or ceramic fracture, loss of retention, sinking down of abutment, and fracture of implant fixture, abutment, or screw [2]. Previous reports have demonstrated incidence rates of implant fixture fracture of 0.2–1.1% and abutment or screw fracture of 0.7–2.3% [2]. In particular, fracture of implant fixture is a catastrophic complication,

which requires extensive surgical treatments. To overcome mechanical failure of dental implants and guarantee long-term clinical success, the stability of the implant–abutment connection to withstand a masticatory load is important [4,8]. The mechanical stability of the connection may be affected by modifying the implant–abutment connection design as well as by improving material properties of the components [9]. Recently, manufacturers of dental implant systems such as Astra Tech Dental have introduced a modified connection design of the dental implant fixture to improve its mechanical properties. To date, however, only a few studies have demonstrated the effect of different connection designs on the mechanical properties of the implant fixtures [10].

Fatigue is the process of localized, permanent structural change of a material under fluctuating stress [11]. Mechanical complications of implants are generally caused by fatigue stress related to mechanical overload [12]. The interpretation of fatigue limit in implants is slightly different from general mechanics. The fatigue limit of dental implants is defined as "the maximum loading value that can withstand 5 <sup>×</sup> <sup>10</sup><sup>6</sup> cycles", contrary to the general definition in mechanics: "the maximum loading value that can withstand infinite cycles" [11,13]. To evaluate fatigue stress in the laboratory, finite element analysis and cyclic loading can be utilized [14–16]. While finite element analysis is considered to simulate fairly reliable results, cyclic loading is used as a method to observe the mechanical properties of actual specimens [16]. To standardize the testing method in the laboratory, ISO 14801 was suggested to simulate a "worst-case scenario" applied on an implant–abutment assembly and consists of sinusoidally curved cyclic loading [13]. These methods can be utilized to substitute in vivo tests, and while a generalized clinical conclusion may not be drawn, a tendency can be observed to provide insight to researchers and clinicians.

The purpose of this study was to compare the mechanical behaviors of two internal connection type dental implant fixtures with different connection designs manufactured by a single manufacturer. By strictly adhering to the procedures considered as the norm, it may provide data on implant systems widely used on the market, and moreover, supply background for additional protocols to the universal standard. The null hypothesis was that the fatigue behavior including the mode of failure of the dental implant–abutment complex is not affected by the modification of the connection design of the dental implant fixture.

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

#### *2.1. Preparation of Specimens*

The fixtures and abutments tested in this study are listed in Table 1 and the flow of the experiment is shown in Figure 1. A total of 30 implant–abutment assemblies were prepared for test and control groups (n = 15 per group): Osseospeed® EV (EV) and Osseospeed® TX (TX) (Astra Tech Dental, Dentsply Sirona Implants, Mölndal, Sweden). Among the 15 specimens, 3 were tested for single-load fracture tests to identify the maximum fracture load and the other 12 specimens were divided into 4 groups (n = 3 each) for fatigue test under cyclic loading. Each specimen was marked with an indelible marker indicating where the load would be applied to analyze the fractured surface with a scanning electron microscope (SEM). Each implant and abutment was connected with a torque of 25 N cm, as recommended by the manufacturer.

#### *2.2. Micro-CT Image Observation*

The implant–abutment assemblies were scanned with micro-computed tomography (CT) scanner (SkyScan 1172, Bruker, Kontich, Belgium) to obtain a series of detailed structural images prior to cyclic loading. The samples were firmly fixed to a full 360◦ rotational inspection jig. The frame rate was four frames per rotational step of 0.5◦, for a total of 2880 images per specimen.


**Table 1.** Materials used in this study. All fixtures and abutments were composed of commercially pure grade 4 titanium.

**Figure 1.** Overall flow of the experiment.

#### *2.3. Single-Load Failure Test and Fatigue*

All mechanical tests were performed according to ISO 14801:2013 (Figure 2). The testing apparatus should impose a force within ± 5% of the maximum error range of the nominal peak value with constant frequency. The testing apparatus should also be able to monitor maximum and minimum load values and to stop when the specimen fractures. A servo-hydraulic test system (MTS Landmark, Minneapolis, MN, USA) under load control was used. Single-load failure tests and fatigue tests were conducted in an atmospheric environment of 20 ◦C ± 5 ◦C. Each implant–abutment assembly was inserted in a custom stainless-steel jig and collet up to the first thread of the implant fixture (approximately 3.0 mm). The collets were then held at a 30◦ off-axis angle and fixed to the jig and testing machine. A hemispherical cap was engaged to the implant–abutment assembly, contacting the flat head of the universal testing machine. Compressive load increasing at a speed of 1mm/min was applied to the implant–abutment assembly until fracture or deformation occurred. Three implants from each group were tested and their maximum fracture load values were recorded. The average value of maximum fracture load of the tested implants served as the nominal peak value for the fatigue test.

For the fatigue testing under cyclic loading, we applied a sinusoidal oscillation with 15 Hz frequency between a nominal peak level (maximum) and a 10% value of the nominal peak level (minimum) to the implant–abutment assembly. The cyclic loading was conducted until the fracture occurred. If fracture did not occur, the cyclic loading was conducted up to a maximum number of <sup>5</sup> <sup>×</sup> <sup>10</sup><sup>6</sup> cycles. The nominal peak levels of 80%, 60%, 50%, and 40% of the maximum fracture load from the previous single-load-to-failure test were selected. Three samples for each nominal peak level group were tested and the number of cycles in which fracture occurred was recorded. If the implant–abutment assembly survived the entire loading cycle, 5 <sup>×</sup> <sup>10</sup><sup>6</sup> cycles were recorded. The results were then plotted on an S/N curve, which is a plot of the magnitude of an alternating stress versus the number of cycles to failure for a given material. The S/N curves were estimated by a logarithmic linear regression model utilizing the least squares method. The fatigue limit of the tested dental implant was defined as the maximum fracture load value which can withstand 5 <sup>×</sup> 106 cycles [13].

**Figure 2.** Schematic diagram of the loading test device according to ISO 14801:2013.

#### *2.4. Failure Modes and Microscopic Observation*

The fractured area of each implant–abutment assembly was microscopically observed and divided into three categories of failures: fixture-level, abutment-level, and screw-level. Two representative specimens were randomly selected before fatigue testing to examine the connection area of the intact implant–abutment assembly using a field emission scanning electron microscope (SEM) (S-4700, Hitachi, Tokyo, Japan). The specimens were inspected one more time after fatigue testing. The frontal and coronal sectional views of fractured specimens were microscopically examined with 15.0 kV accelerating voltage at ×25 and ×30. For the frontal view, specimens were aligned to show the loading direction from left to right. The abutment and fixture cross-sectional views were symmetrically aligned such that the loading direction could be observed from 12 o'clock and 6 o'clock respectively.

#### *2.5. Statistical Analysis*

Mean values and standard deviations of the maximum breaking loads and mean values of the performed cycle from the fatigue tests were calculated. Fisher's exact test was used to analyze the numbers of each type of failure to evaluate the difference of failure modes (fixture-level, abutment-level, and screw-level failure). All statistical analyses were performed using SAS® version 9.4 (SAS Institute, Cary, NC, USA).

#### **3. Results**

#### *3.1. Micro-CT Image Observation*

Frontal and coronal cross-sectional views of micro-CT showed the detailed design of TX and EV (Figure 3). The thinnest areas, excluding the most coronal portion of the fixture, were expected to be the initiation point of the crack; however, the initiation point was the first thread under the microthread, which does not coincide with the thinnest part.

**Figure 3.** Frontal and cross-sectional micro-CT view: (**a**) TX, (**b**) EV; red arrow = location of the thinnest part of the implant fixture.

#### *3.2. Maximum Breaking Load and Fatigue Limit*

The TX samples that underwent single-load failure tests showed a mean maximum breaking load of 711 ± 36 N (95% CI; 670–752), and the EV samples showed an average value of 791 ± 58 N (95% CI; 725–857) (Table 2). The trend of the load and fracture of the specimens was plotted on a time–load diagram, the peak being the point when deformation occurs on the implant–abutment complex (Figure 4). Fatigue testing results are shown in Table 3 and were plotted on an S/N curve with the logarithmic values of the cycles endured on the X-axis and nominal peak level on the Y-axis (Figure 5). All three TX samples of 40% nominal peak level of 285 N endured 5 <sup>×</sup> 10<sup>6</sup> cycles, whereas the other nine specimens failed to resist breaking. The fatigue limit was 285 N to withstand 5 <sup>×</sup> 106 cycles. However, all three EV samples of 40% nominal peak level of 316 N endured 5 <sup>×</sup> 106 cycles, while the other nine samples failed. The fatigue limit was 316 N to withstand 5 <sup>×</sup> 106 cycles.

**Figure 4.** Single-load-to-failure test results with two different implant fixtures: (**a**) TX, (**b**) EV. Compressive load increasing at a speed of 1mm/min was applied. The peak indicates when deformation starts to occur on the implant–abutment assembly, which is the maximum breaking load. The average maximum breaking load of TX = 711 ± 36 N; EV = 791 ± 58 N.


**Table 3.** Values of the Fatigue Tests.

**Table 2.** Values of the maximum breaking loads in single-load failure tests on three specimens each.


0 100 200 300 400 500 600 700 34567 FORCE (N) LOG10(CYCLE) 0 100 200 300 400 500 600 700 34567 FORCE (N) LOG10(CYCLE) (**a**) (**b**) 3x 3x

**Figure 5.** Plotted S/N curves from cyclic loading tests results: (**a**) TX, (**b**) EV. The x-axis represents the logarithmic value of the number of cycles performed. The loading level represents the maximum of the sinusoidal loading level; red arrow = 3 dots overlapped.

#### *3.3. Failure Modes*

Failure modes were observed to speculate the fracture mechanism of the TX and EV samples and are shown in Table 4. For the TX groups, every tested assembly except one, which showed abutment-level failure at the 80% loading level, exhibited failure at the fixture level. For the EV groups, two tested assemblies appeared to have torn-out fixtures at the 80% loading level, which were designated as fixture-level failures. The other assemblies exhibited fixture-level fractures occurring between the first and second threads. All fractures of the specimens were accompanied by screw fractures. There was no statistical difference between fractured areas between the TX and EV groups (*P* > 0.99).


**Table 4.** Fisher's exact test showed no difference between fractured areas (*P* > 0.99).

#### *3.4. Microscopic Observation*

Based on the SEM examination, all the samples, except one TX sample which had an abutment-level fracture, showed a tendency of fixture-level fracture around the first and second threads apical to the microthread area. The thinnest part at the implant–abutment interface and the fractured area did not correspond for TX specimens (Figure 6). For the EV specimens, the 50% loading-level group was characterized with a clean-cut fracture tendency at the first thread level. The other groups showed a tendency to be torn out in a wavy pattern apical and coronal to the first thread. The fractured area was almost at the same level as the thinnest part of the fixture itself (Figure 7). Therefore, from the results, the null hypothesis was accepted.

**Figure 6.** Frontal view of TX samples (×30). Fixtures are aligned to represent a load subjected from left to right.

**Figure 7.** Frontal view of EV samples (×25). Fixtures are aligned to represent a load subjected from left to right.

#### **4. Discussion**

During masticatory function, the dental implant fixture and abutment complex should withstand high axial and lateral force of the jaw [17]. An average value of the axial direction force on a single molar implant restoration was previously reported as 120 N [18]. The reported values of maximum loads ranged from 108 to 299 N in the incisor region and from 216 to 847 N in the molar region [18–21]. In previous research, Park et al. have reported fracture strength under static loading between 799 and 1255 N in the grade 4 titanium implant–abutment assemblies with a diameter close to 4.0 mm [22]. Marchetti et al. have reported fracture strength of 430 N and a fatigue limit of 172 N (i.e., 40% of the maximum breaking load) in a grade 4 titanium implant fixture with a diameter of 3.8 mm [23]. Although a direct comparison between current findings and previous results was impossible due to the difference in the loading conditions between the studies, a similar tendency could be observed. Both TX and EV systems used in this study could overcome the normative requirements, and could be characterized by stable mechanical properties. Furthermore, the calculated fatigue strength proportion between TX and EV in our study was approximately 11%. A study conducted by Johansson and Hellqvist has previously reported that the EV system had 11–20% superior fatigue resistance compared to the TX system, which was consistent with the current findings [24]. The increased strength of EV may be the result of a more apically-located implant–abutment joint area, leading to a better stress distribution, which can be speculated from the micro-CT images. Even with similar chemical compositions, the geometry of the implant–abutment connection can affect the mechanical performance in dental implants. Therefore, the clinician should consider the mechanical properties of implant systems in the treatment planning phase, especially in locations where intraoral conditions may be harsh.

Although ISO 14801 provides a standardized protocol for cyclic loading, it does not provide a loading regimen other than starting at a nominal peak value of 80%, and so one must design the interval between the loading values. This leaves the researcher to guess a nominal peak value that can withstand 5 <sup>×</sup> 106 cycles, which in this case was 40%. However, a 50% nominal peak value seems to be too high a value to accurately estimate the fatigue limit. The 40% groups of TX and EV in this study that endured 5 <sup>×</sup> 106 cycles are equivalent to 20 years of service time in the mouth. Previous studies have shown that humans have an average of 250,000 mastication cycles per year [25,26]. Therefore, it can be assumed that 5 <sup>×</sup> 106 cycles are equivalent to 20 years of service time in the mouth. As the worst-case scenario simulates the harshest environment, it can also be assumed that both specimens can successfully survive intraoral clinical conditions. In contrast, the 50% groups fractured before an average of 50,000 cycles, which is equivalent to less than three months of service time. This "extreme" loading could have affected the failure modes as well as the estimated fatigue limit. Therefore, a loading regimen that includes a nominal peak value between 40% and 50% is recommended for future implant cyclic loading tests. In addition, we speculate that extrapolation to a clinical situation of extreme loading is less applicable for the interpretation of the 50% peak value.

The observation of fractured areas is also important in understanding the fracture mechanism in dental implants. With the advance of technology, micro-CT can be used in observing possible deformations of the dental implant [27]. In this study, the micro-CT images taken before the loading test revealed design differences between the two fixtures. The thinnest part, excluding the coronal portion of the fixture, of TX was located at the microthread area, which was approximately 0.5–1 mm coronal to the first thread of the implant fixture. The thinnest part of EV was located at the first thread of the implant fixture. The thinnest areas of each fixture are shown in Figure 4 and were expected to be the mechanically weakest parts, eventually being the fracture-prone area. However, the fracture lines initiated around the first thread of the fixture in this study. The first thread area was at the same level as the thinnest part in the EV fixture, while they were not at the same level in the TX. This suggested that the weakest part, not necessarily the thinnest part, of the TX and EV fixtures was located around the first thread area. These findings are consistent with a previous study by Shemtov-Yona et al. who tested a conical 13 mm dental implant made of titanium alloy. Three different diameters (3.3, 3.75, and 5 mm) at the implant neck were tested for fatigue performance under cyclic loading. All 5 mm implants fractured at the abutment neck and screw, while all 3.3 and 3.75 mm implants fractured at the implant body. As the implants became thinner, they showed a tendency to fracture more apically than thicker samples [10]. While the results of this study showed no statistical difference between the fracture modes of two groups in the present study (*p* > 0.99), the one sample that fractured at the abutment level may have been affected by the diameter of the implant, and not solely from the design of the connection.

A limitation of this study is the relatively small size of the samples, three specimens for each group. While ISO 14801:2013 states that at least three specimens for each group is required, the small size of samples may not be enough to extract a general conclusion. However, the tendency of the results may provide a surmise on how different implant–abutment complexes react to fatigue. Also, another limitation of this study is that loading conditions such as the number of cycles, loading force, and loading angle were not similar to intraoral masticatory conditions. However, to the best of our knowledge, no testing apparatus or protocol currently can perfectly mimic the function of physiologic mastication. Additional research with large sample size and long-term cyclic loading program is required in the future. Also, a standardized testing protocol with further detail may be a prerequisite to the research.

#### **5. Conclusions**

Within the limitation of this study, we conclude the following:

1. While both implant–abutment complexes are suitable for intraoral use, the EV fixtures in this study performed better than the TX fixtures, which indicates possible differentiation between the two implant–abutment complex designs.

2. Since all specimens with a 40% nominal peak value survived 5 <sup>×</sup> <sup>10</sup><sup>6</sup> cycles and 50% specimens failed before 105 cycles, a loading regimen with nominal peak value between 40% and 50% may be recommended for future testing of cyclic loading for the dental implant fixture.

3. The weakest parts of the tested fixtures were located at the first thread area, which happens to be the area directly coronal to the fixation simulating a 3 mm bone loss, and not necessarily the thinnest part.

From these conclusions, future researchers and implant manufacturers may benefit from starting cycling loading at the loading regimen and by considering the weakest part presented when designing an implant. On the other hand, clinicians should consider the mechanical properties of the implants they plan to use.

**Author Contributions:** Conceptualization, H.-I.Y., T.-H.K., E.-J.P.; methodology, H.-I.Y., T.-H.K., E.-J.P.; validation, H.-I.Y., E.-J.P.; formal analysis, N.-H.C., H.-I.Y., T.-H.K., E.-J.P.; investigation, N.-H.C.; resources, N.-H.C., H.-I.Y.; data curation, N.-H.C.; writing—original draft preparation, N.-H.C.; writing—review and editing, H.-I.Y., T.-H.K., E.-J.P.; visualization, N.-H.C.; supervision, H.-I.Y., E.-J.P.; project administration, E.-J.P.; funding acquisition, H.-I.Y., T.-H.K., E.-J.P.

**Funding:** This study was funded in part by Dentsply Sirona Implants (Mölndal, Sweden) and Yuhan Co. (Seoul, S. Korea) and also supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (NRF-2017R1C1B2007369).

**Acknowledgments:** This study was funded in part by Dentsply Sirona Implants (Mölndal, Sweden) and Yuhan Co. (Seoul, S. Korea). The authors would like to thank Hong-Seok Lim of Dongguk University who helped with the collection of data.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of 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* **Axial Displacements and Removal Torque Changes of Five Di**ff**erent Implant-Abutment Connections under Static Vertical Loading**

#### **Ki-Seong Kim and Young-Jun Lim \***

Department of Prosthodontics and Dental Research Institute, School of Dentistry, Seoul National University, Seoul 03080, Korea; namsang0249@nate.com

**\*** Correspondence: limdds@snu.ac.kr; Tel.: +82-2-2072-2940

Received: 3 January 2020; Accepted: 31 January 2020; Published: 4 February 2020

**Abstract:** The aim of this study was to examine the settling of abutments into implants and the removal torque value under static loading. Five different implant-abutment connections were selected (Ext: external butt joint + two-piece abutment; Int-H2: internal hexagon + two-piece abutment; Int-H1: internal hexagon + one-piece abutment; Int-O2: internal octagon + two-piece abutment; Int-O1: internal octagon + one-piece abutment). Ten implant-abutment assemblies were loaded vertically downward with a 700 N load cell at a displacement rate of 1 mm/min in a universal testing machine. The settling of the abutment was obtained from the change in the total length of the entire implant-abutment unit before and after loading using an electronic digital micrometer. The post-loading removal torque value was compared to the initial torque value with a digital torque gauge. The settling values and removal torque values after 700 N static loading were in the following order, respectively: Ext < Int-H1, Int-H2 < Int-O2 < Int-O1 and Int-O2 < Int-H2 < Ext < Int-H1, Int-O1 (α = 0.05). After 700 N vertical static loading, the removal torque values were statistically different from the initial values, and the post-loading values increased in the Int-O1 group and Int-H1 group (α = 0.05) and decreased in the Ext group, Int-H2 group, and Int-O2 group (α = 0.05). On the basis of the results of this study, it should be taken into consideration that a loss of the preload due to the settling effect can lead to screw loosening during a clinical procedure in the molar region where masticatory force is relatively greater.

**Keywords:** dental implants; implant-abutment connection; settling effect; static loading; removal torque

#### **1. Introduction**

Attempts have been made to understand the factors that could compromise the settling effect of different implant abutment connections [1,2]. Various implant elements including the implant-abutment interface, the types of abutments, the screw characteristics, and the cyclic loading condition have all been shown to influence settling into implants and a loss of preload [3,4].

Loosening of the abutment screws and fixture failure in implant-supported restorations reportedly occur more frequently in the premolar and molar areas than in the incisor region [5,6]. This may result from differences in masticatory force and prosthetic design. Occlusion can be critical for implant longevity due to the nature of the potential load created by tooth contacts. The mechanism and vector of force transferred by posterior teeth differ from those of anterior teeth because posterior teeth have a stronger biting force in the vertical direction. Furthermore, these forces are produced by the action of the masticatory muscles [7].

The various forces that are exerted upon dental implants during function differ in magnitude and direction. In natural dentition, the periodontal ligament has the capacity to absorb stress and allow for tooth movement, but the bone-implant interface has little capacity to allow for the movement of an implant [8,9]. The force is distributed primarily along the crest of the ridge due to the lack of micromovement of implants [10].

Cyclic loading, which simulates functional loading, can significantly influence the overall intimacy of the settling of abutments into implants and their mechanical interlocking at the bone–implant interface [2]. However, cyclic loading is not the only factor that could influence the settling phenomenon in posterior teeth. Cyclic loading and the static loading are two independent conditions, and both can affect the settling of abutments into implants after occlusal loading.

In particular, vertical forces generated on implants in the posterior region are greatest at the implant-abutment interface. This means that vertical masticatory forces can affect settling into implants and a loss of preload after occlusal static loading.

Bruxism or clenching can create destructive lateral stresses and overloading when it transfers force to the supporting bone [11]. Parafunctional movements exert a greater maximum occlusal force than natural mastication. Van Eijden measured the mean magnitudes of a maximal vertical bite force in normal dentition without implants as follows: 469 ± 85 N at the canine region, 583 ± 99 N at the second premolar region, and 723 ± 138 N at the second molar region [12]. These results were comparable to the mean maximum bite force of 738 ± 209 N measured by Braun et al. [13]. In addition, Morneburg and Pröschel investigated vertical masticatory forces in vivo on implant-supported fixed partial dentures and found a mean total masticatory force of 220 N with a maximum of 450 N [14]. On the basis of these findings, the present study evaluated the degree of settling and compared preload loss using the removal torque values before and after 700 N static vertical loading.

The aim of this study was to evaluate the settling of abutments into implants and removal torque values of five different implant-abutment connections that differ significantly in macroscopic geometry after static vertical loading at 700 N.

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

#### *2.1. Implant-Abutment Systems Selection and Study Protocol*

One external and two internal connection implant systems from the Osstem Implant (Osstem Co., Seoul, Korea) were selected for the study. The abutment–implant assemblies were divided into five groups according to the implant connection designs and abutment types (Table 1, Figure 1).

**Figure 1.** Schematic drawing of the test setup. Ext: external hexagon fixture + Cemented abutment; Int-H2: internal hexagon fixture + two-piece abutment; Int-H1: internal hexagon fixture + one-piece abutment; Int-O2: internal octagon fixture + two-piece abutment; Int-O1: internal octagon fixture + one-piece abutment.


**Table 1.** Characteristics of experimental implant-abutment systems.

Ext: external hexagon fixture + Cemented abutment; Int-H2: internal hexagon fixture + two-piece abutment; Int-H1: internal hexagon fixture + one-piece abutment; Int-O2: internal octagon fixture + two-piece abutment; Int-O1: internal octagon fixture + one-piece abutment; Ta: titanium alloy; WC/C Ta: tungsten carbide/carbon-coated titanium alloy; HA: Abutment height; HF: fixture height.


Ten implant-abutment assemblies were constructed for each group (total *n* = 50). Each assembly was held in a vise during the torque tightening procedure. The desired torque was applied to the abutment screw with a digital torque gauge (MGT12, MARK-10 Co., Hicksville, NY, USA).

The schematic diagram of experimental design based on protocol sequence is presented in Figure 2. Each abutment was tightened into the corresponding implant at 30 Ncm torque twice at 10 minute intervals. Ten minutes after the second tightening, the initial removal torque was measured with a digital torque gauge (MGT12E, Mark-10 corp, Hicksville, NY, USA). Each assembly was secured again at 30 Ncm torque, and the total length of the implant-abutment assembly was measured with an electronic digital micrometer (no. 293-561-30, Mitutoyo, Japan). After the initial measurement of the total length, a metal cap fabricated to reproduce the crown was mounted on the abutment of the assembly and the entire unit was fixed in a loading jig (Figure 3). The loading jig was designed to withstand a 700 N vertical static force applied to the implant-abutment assembly. All the specimens were tested in a universal testing machine (Instron 8841, Instron Corp., Mass, Norwood MA, USA) under 700 N vertical static loading, corresponding to the maximum biting force in posterior teeth [12,13].

**Figure 2.** Schematic diagram of experimental design based on protocol sequence.

**Figure 3.** Loading machine and customized jig (Instron 8841, Instron Corp., Mass, Norwood MA, USA).

At the completion of static loading, the total length and removal torque of each implant-abutment specimen were measured in the same manner. The settling value of the abutment was calculated from the changes in the total lengths of the implant-abutment assembly before and after loading. The measurements were accurate up to 0.001 mm (1 μm) and the same operator performed all of the specimen preparations and testing in random order. The details of the experimental protocol and the overall outcomes between the magnitude of applied torque and the axil displacement of abutments into implants in external and internal implant-abutment connections were reported in previous studies [1,2].

#### *2.2. Statistical Analysis*

One-way ANOVA and Tukey's honestly significant difference (HSD) tests were used to analyze settling lengths and removal torque of the five implant-abutment systems before and after 700 N vertical static loading. A paired *t*-test was performed to compare the initial and post-loading removal torques for each implant connection system. *p* < 0.05 was considered to represent a statistically significant difference.

#### **3. Results**

The mean lengths and settling values of the specimen groups after vertical static loading are presented in Tables 2 and 3 and Figure 4. After 700 N static loading, there were statistically significant differences in the settling values in the Ext group (0.8 ± 0.45 μm), Int-H1 group (10.2 ± 0.84 μm), Int-H2 group (11.2 ± 0.84 μm), Int-O2 group (19.2 ± 4.21 μm), and Int-O1 group (25.6 ± 2.97 μm) (α = 0.05). In the internal octagon groups with an 8◦ Morse taper interface, there were greater increases compared with those seen in the other groups. A multiple comparison test by Tukey's HSD exhibited differences in the settling values in each group after 700 N static loading in the following order: Ext < Int-H1, Int-H2 < Int-O2 < Int-O1 (see Tables 2 and 3).


**Table 2.** Mean total lengths and standard deviations of the implant-abutment specimens before and after 700 N static loading.

\* Additional tightening at 30 Ncm after measuring the initial removal torque after the second 30 Ncm tightening. \*\* After 700 N vertical static loading.

**Table 3.** Mean settling values after 700 N static loading in each group and multiple comparisons using Tukey's honestly significant difference (HSD).


† Tukey's HSD method was performed for between group comparisons (*p* < 0.05).

**Figure 4.** Settling of abutments into the implants after static loading (μm).

The mean values of removal torque after loading are presented in Tables 4–6 and Figure 5. After 700 N static loading, the Int-O1 group exhibited the highest removal torque of 39.64 ± 4.28 Ncm. The other groups are shown in the following decreasing order: Int-H1 (36.38 ± 6.25 Ncm), Ext (22.78 ± 0.40 Ncm), Int-H2 (11.62 ± 0.56 Ncm), and Int-O2 (1.14 ± 0.40 Ncm). Using Tukey's HSD, the specific group-wise comparisons in the post-loading removal torque values were as follows: Int-O2 < Int-H2 < Ext < Int-H1, Int-O1.


**Table 4.** Multiple comparisons of mean values of initial removal torque and removal torque after 700 N static loading.

† Tukey's HSD method was performed for between group comparisons (*p* < 0.05).


**Table 5.** Comparison of the mean values of initial and post-loading removal torque in each group.

<sup>a</sup> Removal torque values before loading; b Removal torque values 700 N static loading; † Paired *<sup>t</sup>*-test was performed to compare the removal torque values before and after loading: NS, not significant; \* *p* < 0.01; \*\* *p* < 0.001.

**Table 6.** Comparisons of the mean values of initial removal torque and removal torque after static loading in each group.


\* indicates values that were statistically different (*p* < 0.05).

**Figure 5.** Removal torque (Ncm) after 700 N static loading.

In cases in which one-piece abutments were used for the internal connection system (Int-H1 group and Int-O1 group), the removal torque was increased compared to the initial removal torque. In cases where two-piece abutments were used for the internal connection system (Int-H2 group and Int-O2 group), after 700 N vertical static loading, the removal torque was decreased compared to the initial removal torque to a greater extent. In the Int-O2 group in particular, the abutment screw nearly came loose from the abutment. After 700 N loading, the removal torque value also exhibited a small but significant decrease in the Ext group (Table 6).

#### **4. Discussion**

Along with the expanded indications for implants and the changing clinical protocols, the relationship between implant design and load distribution at the implant–bone interface has become an important issue. The inadequate interaction between these two factors may result in both mechanical and biologic complications such as screw loosening and peri-implant bone loss. Whether an implant prosthesis is placed in function after an undisturbed healing period or immediately after placement, the biomechanical environment is, thereafter, a critical factor that influences implant duration and bone preservation. Loads applied to teeth and implants during physiologic oral functions including chewing, clenching, swallowing, or grinding may vary because the anchorage of natural and artificial abutments in the jaw is not of the same type and quality [15].

Most of the studies related to axial displacement [1–3] are on the magnitude of tightening torque and the duration of cyclic loading, and few studies have applied with static loading. Ko et al. [4] reported that axial displacement and reverse torque loss occurred at significantly low levels after the cyclic and static loading in the case of wide-type implants of 5.0 mm diameter. In addition, the CAD/CAM (Computer Aided Design/Computer Aided Manufacturing) customized abutments, which are currently in the spotlight, may show differences in the fabricating process from the stock abutments produced by manufacturers. Therefore, using implant fixtures and abutments made by the same manufacturer, we wanted to prove that axial displacement could occur even at static loading of 700 N, and the difference comes from different connection types.

For osseointegrated dental implants, previous studies have revealed that occlusal interferences and parafunctional activities may lead to mechanical and biologic complications [16]. Many investigators have attempted to evaluate maximum bite forces. Typical maximum bite force magnitudes exhibited by adults are affected by age, sex, degree of edentulism, bite location, and especially parafunction. In centric occlusion involving swallowing and clenching, forces are transmitted bilaterally, predominantly by molars and premolars. For a single tooth or implant in the molar region, the greatest forces occur along the axial direction [17]. Therefore, the results of this study showed the settling effect in relation to a loss of removal torque after 700 N vertical static loading, corresponding to the maximum masticatory force.

The settling effect after 700 N loading showed a clinical association between screw loosening with a loss of preload and an increase in friction. The results followed a similar pattern with cyclic loading in our previous study [2]. The Ext group showed the lowest settling due to its flat platform interface. Likewise, the internal hexagon and octagon groups had statistically greater settling due to their tapered interface. In particular, the internal octagon group with an 8◦ Morse taper showed the highest settling value compared to the internal hexagon group with an 11◦ taper.

The removal torque values after 700 N vertical static loading may be influenced by the amount of settling and the type and configuration characteristics of the abutment used. When a two-piece abutment, as seen in the Int-H2 and Int-O2 groups, is used, the screw joint connection is based on the tension mechanism, where a screw may become loose due to a loss of preload by settling. Therefore, the settling effect of the Int-H2 and Int-O2 groups produced a significant decrease in the removal torque even to the extent of the loss of the abutment screw in the Int-O2 group. On the other hand, when a one-piece abutment is used, the main retention mechanism is friction. As a result, the settling effect of the one-piece abutment in the Int-H1 and Int-O1 groups created a greater compressive force at the implant-abutment interface, which resulted in the increased post-loading values of removal torque.

The metal cap used in this experimental protocol was inserted into the abutment by friction only, and without dental cement. The simulated crown had a gap between the abutment and the metal cap in order to prevent any forces from being transferred to the abutment during the removal of the crown. However, because the margin of the crown was seated on the fixture in the original internal octagon design, there was no such space. Consequently, this discrepancy may have led to greater settling values than the actual value due to the lack of a vertical stop. In addition, this study could not use the direct method as described by Haack et al., where the change in the preload was evaluated by measuring the length of an elongated screw [18]. Therefore, further studies are warranted to evaluate the actual measurement of an elongated screw as a value of tightening torque.

#### **5. Conclusions**

The current study strived to gain a better understanding of the nature of the implant-abutment screw joint on the basis of the settling effect and removal torque. On the basis of the findings of this study, in the molar region where masticatory force is relatively greater, a loss of preload due to the axial displacement and the possibility of screw loosening should be taken into account in clinical procedures.

The clinical implication of this study is that when the implant fixture of a regular platform with a diameter of 4.0 mm is placed in the posterior molar region, the settling of abutments into implants caused by the vertical force may cause a problem of lowering the occlusion after the prosthesis is mounted.

**Author Contributions:** Conceptualization, writing—Original draft preparation, data curation, K.-S.K.; supervision, visualization, validation, writing—Review and editing, Y.-J.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** The APC was funded by the Dental Research Institute, School of Dentistry, Seoul National University, Seoul, Republic of Korea. This work was supported by grant no. 02-2015-0004 from the Seoul National University Dental Hospital Research Fund, Seoul National University Dental Hospital, Seoul, Korea.

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

#### **References**


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

## *Article* **Dental Implants with Di**ff**erent Neck Design: A Prospective Clinical Comparative Study with 2-Year Follow-Up**

#### **Pietro Montemezzi 1,2,\*, Francesco Ferrini 1,2, Giuseppe Pantaleo 3, Enrico Gherlone 1,2 and Paolo Capparè 1,2**


Received: 24 December 2019; Accepted: 20 February 2020; Published: 25 February 2020

**Abstract:** The present study was conducted to investigate whether a different implant neck design could affect survival rate and peri-implant tissue health in a cohort of disease-free partially edentulous patients in the molar–premolar region. The investigation was conducted on 122 dental implants inserted in 97 patients divided into two groups: Group A (rough wide-neck implants) vs. Group B (rough reduced-neck implants). All patients were monitored through clinical and radiological checkups. Survival rate, probing depth, and marginal bone loss were assessed at 12- and 24-month follow-ups. Patients assigned to Group A received 59 implants, while patients assigned to Group B 63. Dental implants were placed by following a delayed loading protocol, and cemented metal–ceramic crowns were delivered to the patients. The survival rates for both Group A and B were acceptable and similar at the two-year follow-up (96.61% vs. 95.82%). Probing depth and marginal bone loss tended to increase over time (*follow-up*: t1 = 12 vs. t2 = 24 months) in both groups of patients. Probing depth (*p* = 0.015) and bone loss (*p* = 0.001) were significantly lower in Group A (3.01 vs. 3.23 mm and 0.92 vs. 1.06 mm; Group A vs. Group B). Within the limitations of the present study, patients with rough wide-neck implants showed less marginal bone loss and minor probing depth, as compared to rough reduced-neck implants placed in the molar–premolar region. These results might be further replicated through longer-term trials, as well as comparisons between more collar configurations (e.g., straight vs. reduced vs. wide collars).

**Keywords:** dental implants; dental implant neck design; peri-implant bone loss; peri-implant probing depth

#### **1. Introduction**

The scientific debate on dental implant macro-design is a well-known topic in the field of implant dentistry. The ideal fixture design should bring together the most suitable and distinctive characteristics for implant osseointegration, such as type of material (zirconium or titanium), body shape (cylindrical or conical), neck geometry (straight, reduced, or wide), threads depth, width, and pitch, as well as tapered or non-tapered apical portion, body length, and diameter. Although there is no perfect implant design [1,2], nor a best surface treatment [3], scientific evidence has consistently demonstrated that different dental implant macro-designs affect long-term implant success [4,5] and also accelerate the healing process, to allow implant therapy in the population of patients who are more prone to failure [6,7]. Implant collar, being the portion of the implant that connects the fixture with the oral cavity throughout a prosthetic device, is a very important feature related to the peri-implant tissue's health conditions.

Several studies about implant neck design and marginal bone loss can be found in the literature, but the results are controversial. In vivo animal studies reported a greater crestal bone height and thickness of surrounding implant tissue in dental implants with triangular neck designs [8]; smaller crestal bone loss but similar peri-implant tissue thickness in narrow ring extra-shorts implants [9]; and greater bone loss in dental implants with micro-rings on the neck, as compared to open-thread implant collars [10]. Human model studies reported improved biomechanical behavior for stress/strain distribution pattern in dental implants with divergent collar design [11]; no additional bone loss in non-submerged dental implants with a short smooth collar compared to similar but longer implant collar design [12].

Other clinical findings suggest that specific implant neck design might be suitable in anterior areas, where bone loss, even if acceptable, can lead to adverse aesthetic results [13,14].

The purpose of the present study is to compare peri-implant hard- and soft-tissue health conditions in partially edentulous patients who received the same dental implants but with two different implant neck designs, at a two-year follow-up. In this study, the null hypothesis led to the expectation of no differences in survival rate, probing depth, and marginal bone loss among patients who received dental implants with wide or reduced collar morphology.

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

#### *2.1. Patients*

Study participants were selected from patients who attended the Dental Department of IRCCS San Raffaele Hospital, Milan, Italy asking for partial fixed implant-prosthetic rehabilitation. Recruitment occurred from February 2016 to November 2017, and the investigation was conducted following all the ethical regulations related to the institution.

Patients had to meet the following inclusion criteria: (1) hopeless teeth to be extracted at least four months prior to surgery in molar/premolar region; (2) no previous dental implants already in place adjacent to surgical site; (3) natural antagonistic teeth (composite resin restorations allowed); (4) absence of diabetes, periodontitis, bruxism, and smoking; (5) absence of chemotherapy or radiation therapy of head and neck district, as well as anti-resorptive drug therapy (i.e., bisphosphonates); and (6) neither mucosal lesions (lichen planus, epulis fissuratum) nor bone lesions (i.e., simple bone cyst or odontomas). Eligible areas for surgery of edentulous maxilla or mandible were selected to receive 1 to a maximum of 3 dental implants. Participants were verbally informed about the purpose of the study but not assigned to a specific group, as they were randomly chosen either to receive a wide-neck implant (Group A) or a reduced-neck implant (Group B).

Patients were assigned to conditions according to a computer-generated random list, prescribing the use of the reduced vs. wide implant. Clinical measures (i.e., survival rate, peri-implant probing depth, and mean marginal bone loss) were taken at 12 and 24 months. Thus, the design amounted to a 2 (implant: wide vs. reduced) X 2 (time: 12 vs. 24 month follow-up) *mixed* factorial design, following the *Consolidated Standards of Reporting Trials* (CONSORT) guidelines available as supplementary material to this manuscript and on http://www.consort-statement.org/.

Written informed consent was signed before the start of the study; patients were allowed to leave the research at any time, without any consequence.

Implant macrogeometry regarding the two different collar designs used in the present study is shown in Figure 1 (CSR, Sweden & Martina, Due Carrare, Italy).

**Figure 1.** Image shows the CSR full-treatment ZirTi conical dental implant collar with different macro-design. (**a**) Rough wide neck compared with rough reduced neck; (**b**), wide-neck and reduced-neck designs with double conical implant–abutment connection with internal hexagon for prosthetic repositioning; and (**c**) wide-neck and reduced-neck designs with same contact length and tapered angles at the interface.

#### *2.2. Implant Surgery*

The study was based on a single blind design, with patients being unaware of which type of implant neck design (wide or reduced) was used for the therapy.

Local anesthesia was induced with local infiltration of lidocaine 20 mg/mL with 1:50.000 adrenaline (Ecocain, Molteni Dental, Firenze, Italy). A crestal horizontal incision was made, with buccal relieving incisions in the medial and distal portions of the main incision. A full-thickness flap was raised, and dental implants were placed in edentulous sites of 0.5 mm, subcrestally, with a minimum insertion torque of 35 Ncm. Cover screw was positioned, and a periosteal incision was performed in order to allow flap passivation in search for primary intention healing of the wound. Vertical mattress suturing technique was used with a 4-0 coated braided absorbable suture (Vicryl, ETHICON, Johnson & Johnson, New Brunswick, NJ, USA). Sterile dry gauze compression was performed on the wound to control post-operative bleeding. Ice packages were delivered to the patients immediately after surgery, with instruction to apply cold to the surgical area for the following 24 h. Semi-liquid cold diet was recommended for the first 48 h.

At-home pharmacological therapy prescribed was amoxicillin 1 g, every 12 hours, for six days, and non-steroid anti-inflammatory drug ibuprofen 400 mg, every 12 hours, for four days, post-operatively. All implants were loaded after a 4-month healing period, through a delayed loading protocol, with a composite resin temporary restoration, followed by metal–ceramic cemented crowns. Definitive abutments used for both Group A and B were the same and had conical connection with Double Action Tight (DAT), a system that presents a conical interface between the abutment and the implant, plus one more conical interface between the screw and the abutment.

Clinically, abutment screws were tightened at 25 Ncm by using a dental torque wrench.

#### *2.3. Parameters*

Dental implant survival rate was defined as the fixtures being osseointegrated and staying in situ; and capable to guarantee stability for prosthetic support along the 2-year observation period following the surgical placement. Peri-implant probing depth was estimated through a CP12 University of North Carolina color-coded periodontal probe (Hu Friedy, Chicago, IL, USA), in the mesial, distal, buccal, and lingual/palatal surfaces of the fixture. Distance in mm between the mucosal margin and the tip of the probe was considered as pocket depth.

Intraoral radiographs were taken, using extension cone paralleling system (XCP, Dentsply international, RINN), and mean marginal bone loss was calculated, using Digora Optime digital intraoral imaging system (Soredex, Tuusula, Finland).

A line was traced parallel to the long axis of the implant in order to measure in mm the distance between the crestal bone level at the margin of the implant neck and the top of the apical portion of the implant.

#### *2.4. Statistical Analysis*

All analyses were run at the implant level. Peri-implant probing depth and marginal bone loss were submitted to separate 2 (follow-up: t1 = 12 vs. t2 = 24 months)X2(*neck design*: reduced vs. wide) multivariate analyses of variance (MANOVAs), in order to distinguish the effects of follow-up time, implant neck design, and additionally assess any interactive effect(s) of the two factors. Mean values were complemented by standard errors of the mean (*Se*) and 95% confidence intervals (CI).

#### **3. Results**

A total of 97 patients (56 men and 41 women) aged between 33 and 75 years (mean 58.2 ± 6.22 years) were selected for the present study. None of them withdrew from the research, and 122 fixtures were placed in the molar/premolar region.

Fixtures made of titanium grade 4 had a standard length (≥10 mm) and a diameter of 3.8 and 4.2 mm for wide-neck implants and 4.2 and 5.0 mm for reduced-neck ones. Dental implants received the same subtraction procedure, according to the Zir-Ti full-surface treatment (Zirconium Oxide Sand-Blasted and Acid Etched Titanium). The apical portion was tapered with 50◦ accentuated triangular threads and four longitudinal incisions, to increase penetration ability and anti-rotation features. Fifty patients formed Group A (rough wide-neck design) and received 59 implants. Group B (rough reduced-neck design) was composed of forty-eight patients, who received 63 implants.

The two groups were compared at one-year and two-year follow-ups. Survival rate, probing depth, and marginal bone loss were recorded through clinical and radiological checkups. Radiological records for different dental implants placed in Group A and B patients are shown in Figures 2 and 3.

**Figure 2.** Periapical X-rays showing marginal bone level of CSR dental implant with a reduced neck. (**a**) Pre-operative X-ray; (**b**) post-operative follow-up at 12 months; and (**c**) post-operative follow-up at 24 months.

**Figure 3.** Periapical X-rays showing marginal bone level of CSR dental implant with a wide neck. (**a**) Pre-operative X-ray; (**b**) post-operative follow-up at 12 months; and (**c**) post-operative follow up at 24 months.

The overall survival rate of CSR dental implants at the two-year follow-up was 96.72% (four implant failures out of 122 implants placed). Both groups showed similar outcomes: At 12 months, survival rate was 98.30% for Group A and 98.41% for Group B, while it decreased at 96.61% for Group A and 96.82% for Group B at the 24-month follow-up.

Regarding peri-implant probing depth, a 2 (*follow-up*: t1 = 12 vs. t2 = 24 months) X 2 (*neck design*: reduced vs. wide) multivariate analysis of variance (MANOVA) affirmed a main effect of follow-up, *F* (1, 116) = 10.69, *p* < 0.001, such that probing depth was generally lower at 12 months (3.06 mm ± *Se* = 0.046; 95% CI = 2.96, 3.15) than at 24 months (3.18 mm ± *Se* = 0.050; 95% CI = 3.08, 3.28), independently of type of neck design. Furthermore, the analysis also revealed a main effect of neck design, *F* (1, 116) = 6.28, *p* < 0.015, such that probing depth was generally lower for wide (Group A: 3.01 mm ± *Se* = 0.063; 95% CI = 2.88, 3.13) than for reduced-neck implants (Group B: 3.23 mm ± *Se* = 0.061; 95% CI = 3.11, 3.35), independently of time of follow-up. More specifically, the difference between the two groups, considered at one and two years of follow-up were, respectively, as follows: Group A (one year): 2.93 mm ± *Se* = 0.07; 95% CI = 2.79, 3.07 vs. Group B (one year): 3.18 mm ± *Se* = 0.05; 95% CI = 3.07, 3.28 (*p* = 0.007); and Group A (two years): 3.09 mm ± *Se* = 0.07; 95% CI = 2.95, 3.24 vs. Group B (two years): 3.28 mm ± *Se* = 0.06; 95% CI = 3.15, 3.40 (*p* = 0.061). The interaction *follow-up* (t1 = 12 vs. t2 = 24 months) X *neck design* (reduced vs. wide) was not significant, *F* (1, 116) = 0.58, *p* = 0.45, n.s.

A2(*follow-up*: t1 = 12 vs. t2 = 24 months)X2(*neck design*: reduced vs. wide) multivariate analysis of variance (MANOVA) was also conducted for marginal bone loss and revealed a main effect of follow-up, *F* (1, 116) = 198.85, *p* < 0.001, such that marginal bone loss was generally lower at 12 months (0.89 mm ± *Se* = 0.02; 95% CI = 0.86, 0.93) than at 24 months (1.08 mm ± *Se* = 0.01; 95% CI = 1.06, 1.11), independently of type of neck design. Furthermore, the analysis also revealed a main effect of neck design, *F* (1, 116) = 34.04, *p* < 0.001, such that marginal bone loss was generally lower for wide (Group A: 0.92 mm ± *Se* = 0.02; 95% CI = 0.88, 0.95) than for reduced-neck implants (Group B: 1.06 mm ± *Se* = 0.02; 95% CI = 1.03, 1.10), independently of time of follow-up. More specifically, the difference between the two groups, considered at one and two years of follow-up, were, respectively, as follows: Group A (one year): 0.84 mm ± *Se* = 0.03; 95% CI = 0.78, 0.88 vs. Group B (one year): 0.95 mm ± *Se* = 0.02; 95% CI = 0.91, 0.99 (*p* = 0.001); and Group A (two years): 1.00 mm ± *Se* = 0.02; 95% CI = 0.97, 1.03 vs. Group B (two years): 1.17 mm ± *Se* = 0.02; 95% CI = 1.14, 1.20 (*p* = 0.001). Importantly, the two-way interaction *follow-up* (t1 = 12 vs. t2 = 24 months) X *neck design* (reduced vs. wide) was statistically significant, *F* (1, 116) = 3.91, *p* = 0.05, showing that the increase in bone loss for reduced-neck implants (Group B) was steeper than the increase observed for wide-neck implants.

#### **4. Discussion**

Our study focused on dental implants' macro-design, particularly on the clinical performance of the same type of fixture but with two different rough collar designs in partially edentulous patients, using a delayed loading protocol. Examined parameters were peri-implant probing depth, marginal bone loss, and survival rate at two-year follow-up. Both groups of patients showed an acceptable but almost similar implant survival rate. However, patients who received implants with a wide-neck design presented lower probing depth and minor marginal bone loss compared to reduced neck; thus, the null hypothesis of no differences between dental implants with different neck designs was partially rejected. From a clinical point of view, differences in probing depth and marginal bone loss between Group A and B were not relevant at the two-year follow-up. Since the absence of signs of soft-tissue inflammation and the absence of further additional bone loss following initial healing were found, according to peri-implant health definition by Renvert et al. [15], it can be affirmed that both groups of patients showed peri-implant tissue health conditions.

Implant therapy is a very helpful discipline when it comes to rehabilitating dental patients. Even if bone loss around oral implants is described to be an unavoidable and physiologic foreign-body reaction of bone against titanium [16–18], the key for success resides in the neutralization of risk factors at multiple levels: patient level, implant level, and prosthetic level.

Risk factors such as diabetes, periodontitis, bruxism, smoking, antidepressants intake, bone augmentation procedures, head and neck radiotherapy [19–22] play a principal role in long-term implants' outcome. These factors are found at the patient level, meaning that they are poorly controllable over time, as they can worsen along with local or systemic health conditions. Here, we must recall that patients included in the present study where disease-free individuals.

Other factors that are set at prosthesis level also interfere with the success of implant therapy and should not be underestimated. According to Vazquez-Alvarez et al. [23], the distance between the implant platform and the horizontal component of the prosthesis has a significant influence on peri-implant bone loss, and to be adequate, it should range from 3.3 to 6 mm. According to Lemos et al. [24], the retention system for implant-supported prostheses may lead to a different bone-loss pattern, as cement-retained restorations showed less marginal bone loss than screw-retained restorations, and implant survival rate was in favor of cement-retained prosthesis.

Restorations for the present study were cemented crowns where a minimum distance of 3.5 mm was kept between implant-abutment junction and horizontal prosthetic component, and where extreme attention was payed to remove any cement excess that could be found underneath them.

Accuracy of dental impression used, whether traditionally or digitally taken, may lead to differences in the fit of the definitive restoration [25]. In our case, prosthetic rehabilitations were performed by passing through light and putty consistency polyvinylsiloxane materials.

The type of prosthetic material itself is described to be capable of having an effect on the peri-implant tissues [26]. In this study, the decision for metal–ceramic crowns was supported by appropriate biomechanical properties, as it was demonstrated in the literature [27–29].

Occlusal forces were exerted against natural antagonistic teeth in the molar/premolar region, to standardize the procedure and avoid contact with previously installed dental restorations made of unknown or undefined material properties (e.g., a preexisting zirconium-based bridge in the antagonistic region).

Finally, implant therapy risk factors are also found at the implant level, being the fixture macro-design capable to affect the osseointegration process, as reported by several authors [4,5,30–33]. Fixture micro- and macro-designs can be adequately selected before treatment, and with the ideal concept design, implant success rate would be more predictable.

Starting from the type of material from which implants are manufactured, different osseointegration processes (amount of bone attachment to the surface and strength of the bone-surface interaction) may occur at the bone level.

Recently reported by Taek-Ka et al. [34], a qualitative different osseointegration was found through higher bone-surface interaction in commercially pure titanium grade 2 implants compared to grade 4. Apart from titanium, zirconia has also been proposed as an alternative material for oral fixtures. At the moment, despite its optimal biocompatibility, no definitive decision is available on the clinical performance of such implants [35,36].

Back to implant collar, the manner in which it is configured appears to be of relevant interest: The maximum loading stress distribution in bone is localized at the neck of the implants, as described by Anitua et al. [37] and Huang et al. [38]. Several studies are available in the literature, but no consensus on which collar design is more suitable for osseointegration was agreed on by the authors.

Our study would qualify rough wide-neck implants to reduce bone loss over time, being conscious that a longer follow-up period is necessary to confirm these findings. This may be related to the platform-switching concept, which has been described to be beneficial for osseointegration [39–43]. In fact, even in the case that a platform-matched abutment is used in such implants, a minimal effect of switching platform still exists, being that the neck of the implant is wider in diameter with respect to the main body. Otherwise, reduced-neck implants are less likely to benefit from the platform-switching effect because of their narrower platform.

According to Eshkol-Yogev et al. [44], round neck implants may significantly increase primary stability when compared to triangular neck design. In a paper by Mendoca et al. [45], bone remodeling showed to be of benefit around implants with rough collar design, in mandible but not in maxilla, if compared to machined collar surface implants. In a review by Koodaryan et al. [46], rough-surfaced

micro-threaded neck implants appeared to lose less bone compared to polished and rough-surfaced neck implants.

CSR implants placed in this study had roughened surface collars with no microthreads at the bone cervical region. Presence or absence of microthreads, as well as the amount of surface roughness, may have an effect on bone preservation. Despite that an implant collar with a microthread can help in the maintenance of peri-implant bone against prosthetic loading, [47] this study was focused on conventional rough-surface dental implants, not to add confounding aspects related to numerous available surface topography (e.g., smooth, polished neck vs. machined surface vs. microthread design). Furthermore, CSR implants had a moderate degree of roughness, as no beneficial effect seemed to be associated with an increase in surface roughness. In fact, a 20-year follow-up clinical trial by Donati et al. [48] reported no peri-implant bone preservation related to implants with an increased surface roughness.

Another relevant issue to consider is the implant–abutment connection system. Implants in the present study were provided with DAT connection. Consisting of a double conical interface and internal hexagon for prosthetic repositioning, this type of connection follows the recent literature's outcomes. According to Caricasulo et al. [49], internal connection, particularly conical interfaces seem to better maintain crestal bone level around dental implants.

As stated by Kim et al. [50], transmission of the occlusal load from the restoration to the implant, and then from the implant to the surrounding bone, is essential to stimulate osteoblasts activity. This is to say, to avoid minimum but regular and continuous bone resorption, described to be around 1 mm for the first year and of 0.2 mm per year thereafter [51], bone deposition must be encouraged.

The concept of biocompatibility related to implant-prosthetic rehabilitation can be considered as the ultimate key for success: proper design of the fixture, together with a correct function of the implant–abutment connection, and optimal adaptation of the prosthetic restoration generates a self-defensive mechanism that guarantees long-term survival rates.

Considering multiple and confounding aspects which affect implant failure, with risk factors set at patient, implant, and prosthetic level, it is important to affirm that bone loss in not solely determined by collar morphology. Further studies should be conducted on multiple heterogeneous implant collar design in different populations (e.g., diabetic vs. nondiabetic) and with different prosthetic restorations (e.g., screwed vs. cemented). Longer follow-up periods could highlight the enhancement of the clinical performance of dental implants with specific neck configurations.

#### **5. Conclusions**

Within the limitations of the present prospective clinical comparative study, peri-implant probing depth and marginal bone level around dental implants placed in edentulous sites in molar/premolar region were affected by different neck designs. Patients who received implants with rough wide-neck design presented lower probing depth and minor marginal bone loss compared to patients with rough reduced-neck implants.

Reduced-neck implants showed a tendency to lose comparatively more bone over time if compared with wide-neck implants.

However, dental implants' survival rate was acceptable and satisfactory for both groups of patients and showed no differences at the two-year follow-up.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1996-1944/13/5/1029/s1, Consort Statement and 2010 Checklist.

**Author Contributions:** P.M.: conceptualization, writing-original draft preparation; F.F.: investigation, data curation; G.P.: study design, research methodology, statistical analysis, drafting and final approval of the manuscript; E.G.: supervision, project administration; P.C.: conceptualization, investigation. All authors have read and agreed to published version of the manuscript.

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

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

#### **References**


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

## *Article* **A Hybrid Model for Predicting Bone Healing around Dental Implants**

#### **Pei-Ching Kung, Shih-Shun Chien and Nien-Ti Tsou \***

Department of Materials Science and Engineering, National Chiao Tung University, Ta Hsueh Road, Hsinchu 300, Taiwan; gong1014.mse06g@nctu.edu.tw (P.-C.K.); play\_58032.05g@g2.nctu.edu.tw (S.-S.C.)

**\*** Correspondence: tsounienti@nctu.edu.tw; Tel.: +886-3-5712121 (ext. 55308)

Received: 24 May 2020; Accepted: 23 June 2020; Published: 25 June 2020

**Abstract:** Background: The effect of the short-term bone healing process is typically neglected in numerical models of bone remodeling for dental implants. In this study, a hybrid two-step algorithm was proposed to enable a more accurate prediction for the performance of dental implants. Methods: A mechano-regulation algorithm was firstly used to simulate the tissue differentiation around a dental implant during the short-term bone healing. Then, the result was used as the initial state of the bone remodeling model to simulate the long-term healing of the bones. The algorithm was implemented by a 3D finite element model. Results: The current hybrid model reproduced several features which were discovered in the experiments, such as stress shielding effect, high strength bone connective tissue bands, and marginal bone loss. A reasonable location of bone resorptions and the stability of the dental implant is predicted, compared with those predicted by the conventional bone remodeling model. Conclusions: The hybrid model developed here predicted bone healing processes around dental implants more accurately. It can be used to study bone healing before implantation surgery and assist in the customization of dental implants.

**Keywords:** dental implant; tissue differentiation; bone remodeling; mechano-regulation theory; short-term healing; long-term healing

#### **1. Introduction**

Implant stability is one of the important indexes to determine dental implant survival rates in the clinic [1]. It is dominated by the bone healing around the surgery site. Bone healing is a series of complex physiological processes that involved the regulation of several tissue phenotypes. At the beginning of bone healing, micro-vessels, and new connective tissue form on the surface of the wound, which is collectively referred to as granulation tissue [2–5]. After the formation of granulation tissue, further tissue differentiation initiates. Cells then transfer into fibrous connective tissues, cartilages, and new bones according to biophysical stimulus [6–8]. The final stage of bone healing is referred to as bone remodeling, which is a lifelong process, where the skeletal system maintains a dynamic equilibrium, related to the regulation of osteoclasts and osteoblasts [9–11]. When the balance is disrupted by external forces, a new equilibrium state can be achieved spontaneously. Thus, according to the healing process mentioned above, tissue differentiation and bone remodeling stages have a great impact on the short-term and long-term stability of implants, respectively [12,13].

To efficiently predict the short-term stability of implants in advance, we aim to simulate the tissue differentiation process by using the mechano-regulation algorithm. The origin of the method is based on Pauwels [14] who first specified that distortional stress and hydrostatic compression dominate tissue differentiation. Carter et al. [15] implemented the theory into a finite element model (FEM), revealing the evolution of connective tissues. Prendergast et al. [16] modified the methods by adopting octahedral shear strain and fluid flow as the solid and fluid stimuli. Lacroix further improved the

model by using poroelastic finite elements, which can describe the biological mechanism in bones more accurately [5]. In recent years, many studies have revealed the effect of the mechanical environment and the geometric design of implants on the performance of tissue differentiation [17,18].

Bone adapts itself based on its mechanical environment and loading conditions, greatly affecting the morphology of bone and long-term stability of implants [11,13,19]. Many scientists have developed numerical methods to describe the behavior of bone remodeling [20–24]. Carter et al. [15,25,26] proposed that bone apparent density is dominated by strain energy density and studied the energy transfer in hip stems. The internal changes in bone morphology and the aging of connective tissues affected by the external loads were also predicted by FEM. Huiskes et al. [27] adopted a similar approach and simulated the femoral cortex around intramedullary prostheses to reveal the relationship between stress shielding effect and bone resorption. The bone remodeling algorithm was then extended to predict the variation of bone apparent density after implantation treatment [28–31]. The algorithm was verified by computed tomographic (CT) images, showing a high degree of similarity [32]. Most of the studies assumed a simple initial state of the models, where uniform material properties were assigned around the implants [29,31,33,34], i.e., the short-term bone healing has no effect on bone remodeling results. However, short-term healing is crucial since bone remodeling is an iterative process, where different initial conditions may lead to different bone density distribution around dental implants.

In order to test the null hypothesis of the bone healing process in the conventional model, we proposed a hybrid algorithm that regards the procedure of bone healing as two stages: (1) the short-term stage which was simulated by a tissue differentiation model and (2) the long-term stage which simulated by a bone remodeling model. At the beginning of the tissue differentiation model, it was assumed that the wound was filled with granulation tissue. The mechano-regulation algorithm was then applied to determine the tissue phenotypes for the following time steps. Once a stable tissue differentiation has been reached, the current tissue distribution with the material properties, such as Young's modulus, apparent bone density, and Poisson's ratio, in callus around the implant then served as the initial condition for the bone remodeling model. Then, the resulting long-term distribution of Young's modulus and the remodeling stimulus will be discussed. and compared with the results which were similar to those done by Chou et al. [29], where the effect of short-term tissue differentiation was not considered.

The objective of this study is to develop a hybrid model that can predict the stability of dental implants and the strength of the surrounding bones with consideration of both the short-term and long-term bone healing process. The results of the current work can reveal the effect of bone with different material properties on bone healing, providing useful information for dental clinics. Furthermore, the proposed model can be used to rapidly examine the morphology design of dental implants (such as implant radius, length, thread geometry) and the placement protocol (such as insertion angle and depth) to improve the osseointegration between implants and bones.

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

Figure 1 shows the flowchart of the current hybrid algorithm, including short-term and long-term bone healing models. The distribution of strain, fluid velocity, and stem cell diffusion in the initial model (*t* = 0) was firstly calculated by FEM. Granulation tissues then differentiated into various tissue phenotypes based on the mechano-regulation algorithm. Next, the rule of mixture and smoothing procedure [35] was applied to determine the updated material properties and the detail will be discussed in Section 2.2. After the short-term healing process finished, the distribution of tissue phenotypes and the corresponding material properties around the implants was obtained and assigned to the bone remodeling model at the initial state for simulating the long-term healing process. Where bone remodeling algorithm adjusted the bone apparent density of each element iteratively until the equilibrium state of the remodeling stimulus under the given loading condition was achieved. The procedures will be discussed in more detail in Sections 2.2 and 2.3.

**Figure 1.** Flow chart of bone healing preoperative evaluation.

#### *2.1. Three-Dimensional FEM Model*

The current hybrid algorithm was applied to study the bone healing of the mandibular second molar (back teeth in the upper jaw), where the bone geometry, density, and other material properties were adopted based on Chou et al. [29]. The geometry of the bone structure was obtained by extruding a planar CT image with a thickness of 80 mm, as shown in Figure 2a. It was referred to as the bone-tooth system, consisting of a layer of cortical bone overlying on cancellous bone, and a natural tooth. The physiological stimulus of the healthy state, i.e., bone-tooth system, was served as the objective function (i.e., attractor stimulus) for the calculation of bone remodeling in the bone-implant-prosthesis system, details of the calculation will be introduced in the next section. Next, the bone-implant-prosthesis system replaced the tooth in the bone-tooth system by a prosthesis and a short implant with a size of 5.0 × 5.1 mm; the remaining region, i.e., the extraction socket, was filled with callus, as shown in Figure 2b.

The models of the two systems were built by a commercial finite element package ANSYS 18.0 (ANSYS, Inc., Canonsburg, PA, USA). The implant, tooth, and prosthesis were meshed by the built-in element type, SOLID185; the remaining parts of the tissues were meshed by CPT215, which allows the calculation of poroelastic material properties, such as the fluid velocity and pressure in the pores of the bones. There were approximately 138,000 elements and 94,500 nodes for both systems. To maintain the balance between computation time and accuracy, finer meshes were applied around the interfaces between bone and the tooth/implant, as shown in Figure 2. The interfaces were set to allow sliding with a friction coefficient of 0.3. A symmetry boundary condition was applied to the mesial side of the model. All of the nodes in the distal side were constrained in all degrees of freedom. A displacement of 10.5 μm was applied at nodes on the top of the tooth/prosthesis. This value was equivalent to a biting force of 100 N [36–38]; the angle of the displacement was set according to it used in Chou et al. [29]. Note that loading, setting, and properties of all materials, including bones, prosthesis, tooth, implant, and bone graft, used in the current work were based on Chou et al. [29] for comparison, as shown in Table 1. It is worth mentioning that, in the tissue differentiation process, the material properties of elements in the callus region transformed with iterations, i.e., they evolved according to the corresponding

tissue phenotypes during the iteration process. Details of the mechanism of the mechano-regulation algorithm will be explained in the next section.

**Figure 2.** FEM model of (**a**) bone-tooth and (**b**) bone-implant-prosthesis systems.


**Table 1.** The material properties used in the current work [29].

N/A: not applicable.

#### *2.2. Mechano-Regulation Algorithm*

The mechano-regulation algorithm proposed by Lacroix and Prendergast [5,35] was adopted in the current work to predict the distribution of tissue phenotypes. The procedures of the algorithm, including the calculation in each iteration, updates of material properties, and post-processing to the results, were implemented by MATLAB 2019 (The MathWorks, Inc., Natick, MA, USA). At the beginning of the calculation, elements in the callus region were set with the material properties of granulation tissues. According to the theory, tissue differentiation (TD) is induced by the combination of octahedral shear strain (γ) and fluid flow (ν) caused by external loads. It is referred to as biophysical stimulus STD such that:

$$\mathbf{S\_{TD}} = \frac{\mathbf{\hat{y}}}{\mathbf{a}} + \frac{\mathbf{\hat{y}}}{\mathbf{b}} \tag{1}$$

where a = 0.0375 and b = 3 μm/s are empirical constants. Then, the tissue phenotype for the next iteration can be determined based on the value of STD as shown in Table 2. The material properties of the tissue phenotype were updated accordingly.


**Table 2.** The ranges of biophysical stimulus for different tissue phenotypes.

The concentration of mesenchymal stem cells determined the level of the transition from granulation tissue to the other tissue phenotypes. The migration [39] and proliferate [40] of mesenchymal stem cell can be simplified as a classical isotropic diffusion, such that:

$$\frac{dn}{dt} = D\nabla^2 n \tag{2}$$

where *t* is time; *D* is the diffusion coefficient; *n* is the current concentration of mesenchymal stem cell. The migration of stem cells started from the boundary of the extraction socket, which is called cells origin and marked by yellow lines in Figure 2b. At the last iteration, the concentration of the stem cell reached the maximal value. In the current work, *<sup>D</sup>* is set as 8.85 <sup>×</sup> <sup>10</sup>−<sup>14</sup> <sup>m</sup>2/s. Then, the effective material properties of tissues for the next iteration, including Young's modulus, Poisson's ratio, and permeability, can be obtained by a linear combination between the granulation tissue (*Xg*) and the differentiated tissue (*Xd*) as follows:

$$X\_{\rm mix} = \frac{n\_{\rm max} - n}{n\_{\rm max}} X\_{\rm g} + \frac{n}{n\_{\rm max}} X\_{\rm d} \tag{3}$$

where *nmax* is the maximum concentration; *n* is the current concentration of stem cells determined by Equation (2); *Xd* is material properties of the differentiated tissue phenotype shown in Table 1.

To avoid the instability and dramatic change of material properties between iterations, Lacroix and Prendergast [35] suggested a smooth procedure to average the material properties from the previous nine iterations, which can be written as:

$$X\_i = \frac{1}{N} (X\_{\text{mix}} + X\_{i-1} + X\_{i-2} + \dots + X\_{i-(N-1)}) \tag{4}$$

where *N* = 10; *i* is the current iteration number; *Xmix* is the effective material properties calculated from Equation (3). Note that when the iteration number is *i* < 9, smoothing operation is applied to the iteration *i* to the first [35]. The short-term healing time was set as 70 days, which is the average healing period for implantation surgeries [41–43]. The short-term healing process and the differentiated tissue phenotypes can then be predicted.

#### *2.3. Bone Remodeling Algorithm*

After the numerical calculation for the short-term healing, long-term healing, i.e., bone remodeling (BR), occurred to alter the internal structure of bones and reach a new equilibrium state based on the mechanical environment. Huskies et al. [27] proposed a bone remodeling theory assuming the driving force of self-adaptive activity is determined by remodeling stimulus (*SBR*, unit: J/kg), such that:

$$S\_{BR}(\overrightarrow{\boldsymbol{r}},t) = \frac{\mu(\overrightarrow{\boldsymbol{r}},t)}{\overrightarrow{\rho(\boldsymbol{r},t)}}\tag{5}$$

where *<sup>u</sup>* is strain energy density (unit: J/m3); <sup>ρ</sup> is the apparent bone density (kg/m3), *<sup>t</sup>* is time; and *r* is the position vector [27]. When the value of remodeling stimulus *SBR* is greater than the given threshold, bone formation occurred and Young's modulus and bone density increase accordingly. On the contrary, when the remodeling stimulus *SBR* is less than the threshold, bone resorption occurred, and Young's modulus and bone density decrease. In addition, Carter [25] stated that bones maintain a state of homeostasis when the remodeling stimulus is in a certain range, which is referred to as a "lazy zone." Thus, the bone remodeling process can be expressed by nonlinear functions of the remodeling stimulus [27]:

$$\frac{d\underline{\rho}}{dt} = \begin{cases} A\_f \Big[ \mathcal{S}\_{\rm BR} - (1+\mathbf{s}) \mathcal{K}(\stackrel{r}{r}) \Big]^2, & \mathcal{S}\_{\rm BR} \ge (1+\mathbf{s}) \mathcal{K}(\stackrel{r}{r}) & \text{(Formation)}\\ 0, & (1-\mathbf{s}) \mathcal{K}(\stackrel{r}{r}) < \mathcal{S}\_{\rm BR} \le (1+\mathbf{s}) \mathcal{K}(\stackrel{r}{r}) & \text{(Lazy zone)}\\ \quad A\_f [\mathcal{S}\_{\rm BR} - (1-\mathbf{s}) \mathcal{K}(\stackrel{r}{r})]^3, & \mathcal{S}\_{\rm BR} \le (1-\mathbf{s}) \mathcal{K}(\stackrel{r}{r}) & \text{(Resorption)} \end{cases} \tag{6}$$

where *Af* and *Ar* are formation and resorption coefficients; *s* is the threshold of the lazy zone, which is set as 0.75 [44]; *K* - *r* is the attractor stimulus induced by the biting force in the bone-tooth system, which is determined by Equation (5). The value of *K* - *r* in the region of callus was set to 5, which is the average of the overall remodeling stimulus in bone element. It is worth noting that the rate of bone resorption is greater than that of bone formation based on clinical observations, resulting in a greater exponential term of resorption in Equation (6). The apparent density of a bone element m at the jth iteration can be derived by integrating Equation (6) with a forward Euler method [27,45–47], such that:

$$\boldsymbol{\rho}\_{\rm{m}}^{\rm{j}} = \begin{cases} \rho\_{\rm{m}}^{\rm{j}-1} + A\_{\rm{f}} \Delta t \big[ \mathbf{S}\_{\rm{BR}}^{\rm{j-1}} - (1+\mathbf{s}) \mathbf{K}(\mathbf{r}) \big]^{2}, & \mathbf{S}\_{\rm{BR}}^{\rm{j-1}} \ge (1+\mathbf{s}) \mathbf{K}(\mathbf{r}) & \text{(Formation)}\\ \mathbf{0}, & (1-\mathbf{s}) \mathbf{K}(\mathbf{r}) < \mathbf{S}\_{\rm{BR}}^{\rm{j-1}} \le (1+\mathbf{s}) \mathbf{K}(\mathbf{r}) & \text{(Lazzy zone)}\\ \boldsymbol{\rho}\_{\rm{m}}^{\rm{j-1}} + A\_{\rm{r}} \Delta t \big[ \mathbf{S}\_{\rm{BR}}^{\rm{j-1}} - (1-\mathbf{s}) \mathbf{K}(\mathbf{r}) \big]^{3}, & \mathbf{S}\_{\rm{BR}}^{\rm{j-1}} \le (1-\mathbf{s}) \mathbf{K}(\mathbf{r}) & \text{(Respontion)} \end{cases} \tag{7}$$

where Δt is the time increment [21,48]. Young's modulus (*E*, unit: GPa) of bone elements was associated with the corresponding apparent density based on the finding of Carter and Hayes [26], such that:

$$E = \mathbb{C}\rho^3\tag{8}$$

where *C* is constant. Note that the integration coefficients *Af*Δt and *Ar*Δt in Equation (7) were set as <sup>1</sup> <sup>×</sup> <sup>10</sup>−<sup>11</sup> and the constant was set as 3.79, based on the setting used in Chou et al. [29]. The value of Young's modulus indicated the strength of the internal structure of bone, which affected the calculation at the next iteration. The average remodeling stimulus (*Save*) in each iteration was recorded and served as a measure of convergence of the model, such that:

$$S\_{\text{ave}} = \frac{1}{N\_{\text{total}}} \sum\_{k=1}^{N\_{\text{total}}} S\_k \tag{9}$$

where *Ntotal* is the total number of bone elements; *Sk* is remodeling stimulus of the local bone element *k*. It is worth noting that long-term bone healing, i.e., bone remodeling, is a lifelong process and the bone system evolves to reach an equilibrium state according to its current mechanical environment. Thus, the time step used here is, in fact, a computational increment and is not associated with a real-life time scale. The end of the calculation depends on the convergence of *Save* as mentioned above.

#### **3. Results**

#### *3.1. Short-Term Healing and Tissue Di*ff*erentiation*

Tissue differentiation and the evolution of bone ingrowth around the implant were evaluated by the mechano-regulation model. Figure 3 shows the percentage of tissue phenotypes in each day during the short-term healing process. In the early stage of tissue differentiation, granulation tissues still existed in the callus region. This is because tissue differentiation was initiated when the concentration of stem cells above certain levels. At this stage, the inner callus region has relatively low concentration as the stem cells diffused from the boundary of the callus region. Moreover, it was found that the soft tissue with a higher biophysical stimulus (STD > 1) such as cartilage and fibrous tissue decrease with time while the bone tissue increased continuously until the middle of the differentiation process (i.e., around the 35th day). After the 35th day, bones possessed a certain degree of strength, i.e., higher Young's modulus. This resulted in the decreasing values of bone stimuli, and thus, maturate and immature bones gradually became the dominant tissue phenotype in the entire callus region. Then, most of the immature bones transformed into maturate bones as the healing process was closed to the 70th day.

**Figure 3.** Percentage of various tissue phenotype in each day during the short-term bone healing process.

Details of the tissue phenotype at the specific days are shown in Figure 4. On the 4th day, more than half of the callus region remained as granulation tissue, as shown in Figure 4. It is observed that cartilages and fibrous tissues occurred around the threads in the middle part and the bottom of the implant, respectively. It is because the applied load caused the stress concentration around the threads and the bottom, giving higher biophysical stimulus to the elements in that region. Note that, although there were mature and immature bones, the effective material properties, such as Young's modulus, were still close to that of granulation tissue based on Equation (3), due to the low concentration of stem cell on the 4th day. Then, it can be observed that, on the 10th day, granulation tissues gradually transformed into mature and immature bones as the concentration of stem cells increased with time. It is worth noting that cartilages accumulated at tips of the threads since stress concentration was partially released with the increasing maturity of the surrounding bones. On the 30th day, granulation tissues were disappeared entirely. Cartilages were mainly located at the lingual side because of the oblique biting force. Then, maturate and immature bones gradually dominated the entire system around the implant on the 50th to 70th days. It is worth noting that there were very few elements of bone resorption on the 10th day and were not shown in the current cross-section in Figure 4. In this way, the short-term healing pattern around the implant was obtained.

**Figure 4.** The tissue differentiation history predicted by the mechano-regulation algorithm.

#### *3.2. Bone Remodeling*

Now consider two bone remodeling models. The first, i.e., the current model, adopted the result at the end of short-term healing predicted by the mechano-regulation algorithm as the initial state; the second was the conventional model based on bone remodeling algorithm with the assumption that the extraction socket was filled with bone graft and regarded as the initial state where the material properties were set as uniform. The setting of the second model followed the work done by the literature [29]. Both models shared the same distribution of the objective attractor stimulus *K*( *r*) based on the natural tooth, as shown in Figures 5a and 6a. Then, the bone remodeling models altered the bone apparent density of all the bone elements in the following iterations to achieve that objective distribution.

In the current model, the distribution of Young's modulus of the initial state is shown in Figure 5d. It can be observed that certain regions in the callus in Figure 5d were in dark and light blue colors giving low and high Young's modulus, ranging from 0.001 to 6 GPa based on Table 1, due to the non-uniform stem cell concentration and the presence of different tissue phenotypes. It is worth noting that the average value of Young's modulus in that region was around 2 GPa, having a good agreement with the value used in the conventional model where uniform Young's modulus was assumed, as shown in Figure 6d. The corresponding bone apparent density of each element in the callus region can then be obtained by Equation (8). The resulting model with the updated bone apparent density was then subjected to the biting force, giving the distribution of the remodeling stimulus (*SBR*) for the 1st iteration, as shown in Figure 5b. The values are shown in Figure 5a,b were substituted into Equation (7) to obtain the updated bone apparent density for the next iteration. The corresponding distribution of Young's modulus in the 1st iteration was shown in Figure 5e. It can be observed that bone regions with high Young's modulus (above 12.18 GPa, colored in red) fully covered around the implant; most of the bones attached to the surface of the implant were also with non-uniform Young's modulus (ranging from 3.04 to 6.08 GPa). The calculation continued until the 100th iteration was achieved. It is worth noting that the average remodeling stimulus (*Save*) of the current model quickly converged around the 10th iteration, showing great stability. The converged values *Save* for both the current and conventional models were identical. The distribution of *SBR* of the 100th iteration is shown in Figure 5c. It can be observed that most of the regions had the value of *SBR* closed to those of the target, i.e., the objective attractor stimulus *K* - *r* . The final state of bone remodeling gave the distribution of Young's modulus, shown in Figure 5f.

The average Young's modulus of the entire system predicted by the current model was around 4.77 GPa. It is worth noting that bone resorptions (colored in gray) occurred around the threads toward the lingual side and around the neck of the implant. The total volume fraction of bone resorption was 0.042%. Similar to the 1st iteration, bone regions with high Young's modulus were still fully covered around the implant. Two additional high strength bone tissue bands connected to cortical bones were formed in the bottom right (the lingual side) and top left (the buccal side) of the implant, which provided extra supports and enhanced the stability of the implant.

**Figure 5.** The results generated by the current model. (**a**) The target distribution of the attractor stimulus *K*( *r* ) based on the natural tooth. (**b**,**c**) are bone remodeling stimulus (*SBR* ) at the 1st and 100th iterations. (**d**–**f**) are the corresponding distribution of Young's modulus during the bone remodeling process at the initial state, 1st, and 100th iteration.

**Figure 6.** The results generated by the conventional model. (**a**) The target distribution of the attractor stimulus *K*( *r* ) based on the natural tooth. (**b**,**c**) are bone remodeling stimulus (*SBR )* at the 1st and 100th iterations. (**d**–**f**) are the corresponding distribution of Young's modulus during the bone remodeling process at the initial state, 1st, and 100th iteration.

Next, Figure 6d shows the distribution of Young's modulus of the initial state adopted in the conventional model. It can be observed that constant Young's modulus of 2 GPa in the callus in Figure 6d was assumed without considering the result of differentiating tissue phenotypes during the short-term healing. Then, the distribution of the remodeling stimulus (*SBR*) for the 1st iteration can be determined as shown in Figure 6b. In the 1st iteration, the distribution of *SBR* was similar to that generated by the current model, apart from there was no high stimulus bands around the implant. However, such a small difference resulted in a very different distribution of Young's modulus shown in Figure 6e compared with that in Figure 5e, wherein, the high strength bones partially covered around the implant; most of the bones attached to the surface of the implant remained a constant Young's modulus of 2 GPa. Then, similar to the current model, the average remodeling stimulus (*Save*) quickly converged. Finally, the 100th iteration was achieved, giving the distribution of *SBR* as shown in Figure 6c. The resulting Young's modulus distribution is shown in Figure 6f. The volume fraction of bone resorption was at the value of 0.044%, which was very similar to the current model. A significant difference between the results of the two models was the location of bone resorption. It was found that bone resorption (colored in grey) occurred in the buccal side in the conventional model while it appeared on the lingual side in the current model. This will be discussed in more detail in the next section. Another obvious difference between the two models was the distribution of high Young's modulus bands. The high Young's modulus band was absent in the buccal side, and thus, no supporting connection between the surface of the implant and cortical bone. The average Young's modulus at the 100th iteration was around 3.65 GPa which was relatively lower than that generated by the current model.

#### **4. Discussion**

The results mentioned above show that the initial state of bone remodeling can greatly affect the distribution of Young's modulus at the final state. In the current model, the initial state of bone remodeling was the result of the mechano-regulation model (short-term healing process), giving the top and bottom regions in the callus with lower Young's modulus. This leads to a higher strain energy density and a low corresponding bone apparent density base on Equation (8). Then, these regions had higher remodeling stimulus based on Equation (5), promoting the formation of bones, i.e., Young's modulus and apparent density increased. This non-uniform Young's modulus at the initial state leads to a dramatic change of Young's modulus in the entire callus region in the 1st iteration. On the contrary, in the conventional model, Young's modulus was assumed uniform, resulting in the change of Young's modulus around the implant only in the 1st iteration.

In the 100th iteration, it can be observed that the stability of the implants was greatly influenced by the initial states. In the current model, there were two high strength bone tissue bands connected to cortical bones, while there was only one connected bone tissue band in the result generated in the conventional model. In addition, the average Young's modulus in the current model was higher than it predicted by the current model. Thus, the stability of the implant was underestimated in the conventional model where uniform Young's modulus in the callus region at the initial state was assumed. This indicated that short-term healing can greatly affect the results of bone remodeling and cannot be neglected.

In the results of short-term healing generated by the mechano-regulation model, soft tissues occurred in the early stage and then replaced by bone tissues due to the decrease of biophysical stimulus in the later stage. This was in accordance with both the experimental [14] and computational [5] works in the literature. Where the literature reported that bone tissue forms after the formation of soft tissues (i.e., fibrous tissue and cartilage), and then bone began to differentiate, giving the increase of fluid flow since woven bone is more permeable. Furthermore, three additional features can be found in the remodeling result in the 100th iteration predicted by both the current and conventional models, which were in accordance with the clinical observations. Firstly, both models predicted a region of bone resorption at the top surface in the lingual (right) side. This phenomenon was known as marginal bone loss [49,50], which was an important factor for implant stability. Secondly, both models generated a bone resorption region in the middle of the bone-implant interface. This is the so-called stress shielding effect [27,47], which was the result of the occurrence of high strength bone tissue (colored in red) at the region around the first thread. However, the locations of the bone resorption region predicted by the two models were different, where the region predicted by the current model was in the lingual (right) side and that predicted by the conventional model was in the buccal (left) side. According to the literature [51], bone resorption occurred on the right-hand side when the loading direction was similar to that in the current work, i.e., the load was applied from the top right to the bottom left. The current model successfully captured this feature, reproduce the bone resorption region at the right-hand side at the bone-implant interface, while the conventional model predicted the location of bone resorption at the opposite side. This shows that the current model can give a more accurate result of bone remodeling procedure in the bone-implant-prosthesis system. Third, the result in the 100th iteration predicted by the current model shows that around 70% of the surface of the implant was covered by bone tissues (i.e., the elements with Young's modulus greater than that of immature bone, 2 GPa). This value was similar to that reported by Lian et al. [31], where they suggested around 60% contact between bone and implant when an equilibrium of bone remodeling is reached. Based on the features mentioned above, we conclude the rejection of the null hypothesis that short-term bone healing has no effect on bone remodeling results.

Since the current hybrid model was implemented by the finite element method, which can perform virtual tests on a wide variety of people characteristics and dental materials by simply changing the geometry of the model, boundary conditions, and material properties of bones. Thus, the current model can potentially provide estimated information and may even provide optimized dental implants for dentist clinics. Next, to demonstrate the applicability of the current model and reveal the effect of individual differences of the patients, we adopted the case of middle-aged male adults with higher strength material property of bone, referred to as to higher bone strength case. Where Young's modulus of cortical, cancellous, immature, and mature bone was 1.4 times [52] than it in Table 1, while the properties of the remaining tissues, such as fibrous tissue and cartilage, stayed unchanged. Then, the short- and long-term bone healing processes were evaluated by the current model. The corresponding results are shown in Figures 7 and 8, respectively. It was found that the trend of tissue differentiation in the short-term healing process was almost identical to it the standard case as shown in Figure 4. The only difference between the two cases was the averaged Young's modulus in the high strength case was about two times higher than it was in the standard case, which can be seen in Figure 8a, where the material properties in the callus region were assigned according to the tissue differentiation result. In the 100th iteration, there were several significant differences between the two cases. Firstly, there was merely no bone resorption in the higher bone strength case. The volume fraction of bone resorption was at a value of 0.0068%. Secondly, the connective tissue bands in the higher bone strength case were thicker than in the standard case. The two features indicate that the bone system in the higher bone strength case can provide excellent supports and enhance the stability of the implant. This result also has good agreement with the clinical observation, where the dental implant failure rate for the middle-aged male adults was relatively low [53].

Although the current model considered both the short-term and long-term healing process and reproduced many features that were discovered in the experiments, the model can be further improved by considering the physiological mechanism listed as follows. For example, a more complex diffusion mechanism of stem cell migration in the short-term healing process, such as the growth of vessels which can be implemented by the random-walk model [54]; periodontal ligaments (PDL), which play a crucial role in bone remodeling, can be simulated by taken the anisotropic and nonlinear elastic stress-strain behavior into account [55–57]. It is expected a more accurate prediction can be achieved if these factors were considered.

**Figure 7.** The tissue differentiation history predicted by the mechano-regulation algorithm in higher bone strength case.

**Figure 8.** The results generated by the current model in higher bone strength case. (**a**–**c**) are the corresponding distribution of Young's modulus during the bone remodeling process at the initial state, 1st, and 100th iteration.

#### **5. Conclusions**

In this study, a hybrid numerical bone healing algorithm was developed to predict the morphology of bone around dental implants with the consideration of both short-term and long-term bone healing. The results showed that the effect of short-term bone healing should not be ignored, and the assumption of uniform material properties for the initial state in the bone remodeling model is inappropriate. The current hybrid model can reveal many bone healing features having a very good agreement with the literature. It can be extended to simulate different implant geometries, applied loads, and bone properties of patients, enabling an early prediction of the performance of clinical treatments.

**Author Contributions:** Conceptualization, S.-S.C. and N.-T.T.; methodology, P.-C.K. and S.-S.C.; software, P.-C.K. and S.-S.C.; validation, P.-C.K. and N.-T.T.; writing—original draft preparation, P.-C.K.; writing—review and editing, N.-T.T.; visualization, P.-C.K.; supervision, N.-T.T.; project administration, N.-T.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by Ministry of Science and Technology Taiwan, Grant No. MOST 106-2218-E-009-003 and Grant No. MOST 107-2218-E-080-001.

**Acknowledgments:** The authors would like to thank Ming-Jun Li for his comments and assistance to this work. The authors wish to acknowledge the National Center for High-Performance Computing, Taiwan, for providing the computational platform.

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

#### **References**


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

*Article*
