*2.3. Bone Segments*

It has been reported that the macroscopic structure of cortex and cancellous bone in the bovine rib is similar to that in the human jaw, and the bone density dependent on its position could be comparable to classes I–IV according to Lekholm & Zarb or D2–D4 according to the Mish classification in the human jaw [3–6]. The distal part of the bovine rib was used and classified as D2/D3 bone. Therefore, 14 ribs were used in this study to simulate the condition in the human jaw. Bone density measurements have not been made because the friction when inserting cylindrical implants arises mainly on the hard cortex bone [7,8] if a sufficiently high torque has been generated there (40 to 60 Ncm in this study). Therefore, when selecting the bone segments, particular care was taken to ensure that the cortical layer thickness was somewhat the same for all bovine bony ribs [7].

#### *2.4. Experimental Settings*

In each experiment, the implant was connected to the screwing unit, ensuring a stable rotational speed during insertion of the implant. In the screwing unit, the driving force was produced by a pneumatic cylinder (C85N10-75, ISO air cylinder series C85, SMC Pneumatik GmbH, Egelsbach, Germany). Then, the motor (DC-servo drive series 3564K024B CS, Faulhaber GmbH & Co KG, Schöneich, Germany) connected to a gearbox (series 30/1, 134:1, Faulhaber GmbH & Co KG, Schöneich, Germany) transmitted the force to the torque sensor (DR-20-2Nm; Lorenz Messtechnik GmbH, Alfdorf, Germany) and the collet holder. The implant was connected to the collet holder by a round-headed Allen key (3 mm) to allow rotation. The Allen key was loosely fitted to the mounting post of the implant so that insertion of the implant could remain centered in the prepared bone cavity guided by implant threads and a cone-shaped apex.

The screwing unit could be driven forward or backward axially, and its movement was recorded by a linear inductive distance sensor (LVP 100, Micro-Epsilon Messtechnik GmbH & Co. KG, Ortenburg, Germany). To simulate the environment of surrounding tissues, the bone segment and the implant were covered in a preheated polycarbonate thermo box (37 ◦C) to prevent heat dissipation. The infrared camera was fixed to the cover of the thermo box. The complete experimental settings are shown in Figure 3.

**Figure 3.** Diagram of the experimental setting. (**1**) Thermo box; (**2**) lens of infrared camera; (**3**) fully inserted ceramic implant; (**4**) transection of the rib segment; (**5**) collet holder (CH); (**6**) bearing unit (BU); (**7**) torque sensor (TS); (**8**) gearbox (GB); (**9**) motor (M); (**10**) linear bearing; (**11**) pneumatic cylinder; (**12**) distance sensor.

#### *2.5. Experimental Protocol*

Fourteen of the zirconia implants (groups 1–3) were included in the whole experiment, and each implant underwent 3 loops as a result of the surface treatments applied. After each implant had been assigned to one bovine rib, each bovine rib (BR(1)–BR(14)) was cut into 3 sequential rib segments (RS1BR(1)–RS3BR(1) to RS1BR(14)–RS3BR(14)). Each implant (I1–I14) would undergo one type of surface modification (machined (m), sandblasted (sb), sandblasted and acid-etched (ae)) before each test program (TP1–TP3). The test sequence of the samples and the corresponding rib segments were organized as follows:

Each implant was inserted into one new rib segment after each surface modification. First, all machined implants (I1(m)–I14(m)) were inserted into 14 rib segments (RS1BR(1)–BR(14)). Then, after sandblasting, implants (I1(sb)–I14(sb)) were inserted into rib segments (RS2BR(1)–BR(14)). Finally, after acid-etched treatment, implants (I1(ae)–I14(ae)) were inserted into rib segments (RS3BR(1)–BR(14)). Surface modification usually took several days. In order to maintain the bone in the same fresh condition in all the tests, bone segments were frozen beforehand in Ringer's solution at −10 ◦C and were defrosted immediately prior to the experiment.

The titanium implants underwent identical protocols. Since they were made especially for this study, a higher number of pieces was available at the beginning of the trial. Therefore, with torques outside the interval of 40 to 60 Ncm, new Ti implants were used to repeat the insertion test so that 15 measurements per group could be realized (Table 1). The selected torque window was clinically sufficient for immediate restoration of ceramic implants (>35 Ncm), which are often still used as one-piece implants. Furthermore, smaller torques do not produce sufficiently high temperatures by friction between the implant surface and the hard cortical bone [8] to be able to detect the effect of different thermal conductivities between titanium and ceramic at the bone–implant interface.

Contrary to Ti implants, for this study only 14 ceramic implants were available from the manufacturer. Therefore, when the torque was too low or too high, bone penetration tests with the respective surface structure were no longer repeated in order to mechanically protect the connection of the implant driver with the ceramic implant. In addition, cleaning the surfaces of bone remnants would have been a process in which, above all, the roughened ceramic surfaces could have been structurally changed; thus, the equality of specimens could no longer be ensured. Because of this limitation of available ceramic implants, only 8 measurements in group 1, 11 measurements in group 2, and 11 measurements in group 3 could be realized (Table 1).

When one test started, the selected bone segment would be defrosted in Ringer's solution at 37 ◦C and stored at the same temperature in the oven to simulate the human condition.

The implant bed was prepared using a twist drill (Type N, 118◦ DIN 338 R-N, Gühring KG, Albstadt, Germany) at the speed of 800 rpm. Since the bone was already wetted, additional water cooling was not necessary. To ensure a comparable insertion resistance in the three loops, the diameter of each implant site matched the implant diameter, that is, twist drills with diameters of 4.3, 4.25, and 4.20 mm were used in loops 1, 2, and 3, respectively.

To dynamically record temperature directly at the bone–implant interface, two bone windows perpendicular to the insertion path were made at 3 and 9 mm subcrestally by a twist drill (Figure 4).


 =

 =

 =

 =

**Table 1.** Temperature increase of the 6 groups.

**Figure 4.** Dynamic temperature record at the bone–implant interface. (**a**) Transection of a rib segment; (**b**) overview of a rib segment; (**c**) infrared image at T3 in Figure 5; (**d**) infrared image at T7 in Figure 5. I-C = Implant Cavity 1; B-W 1 = Bone Window 1; B-W 2 = Bone Window 2.

**Figure 5.** *Cont.*

**Figure 5.** Box plots comparing (**A**) insertion torque, (**B**) cortical bone thickness, and (**C**) roughness.

Before each test, the bone was brought to a temperature of 37 ◦C inside an oven with saturated humidity. After the bone segment was immobilized, the implant was prewetted in Ringer's solution used to simulate human blood, and then it was driven with a starting pressure of 5 N and a constant rotational speed of 25 rpm to a depth of 12 mm inside the bone. Meanwhile, the infrared camera continuously recorded the temperature through each bone window (2b & T7 in Figure 5), and the difference between the maximum and the starting temperature (37 ◦C) during insertion would be calculated (ΔT). The change of torque was recorded by torque sensors (DR-20, Lorenz Messtechnik GmbH, Alfdorf, Germany) with the frequency of 100 Hz. Data from groups in which the insertion torque was above 60 Ncm or below 40 Ncm were excluded from the analysis.

After each experiment, the thickness of cortical bone at the insertion point was measured with a caliper gauge. When the temperature, measured at the position 3 mm subcrestally, exceeded 47 ◦C (ΔT >10 ◦C) [9], 50 ◦C (>13 ◦C) [10], or 55 ◦C (>18 ◦C) [11], the corresponding exposure time was calculated.

#### *2.6. Statistical Analysis*

Continuous variables are represented as mean ± standard deviation for each group. Comparisons were performed using the Mann–Whitney *U* test, *t*-test, or paired *t*-test. Comparisons with more than 2 groups were analyzed by ANOVA or the Kruskal–Wallis test, depending on Gaussian distribution. Additionally, comparisons using adjusted linear mixed-effects models (LMEs) were performed. Gaussian distributions of data were assessed with the Shapiro–Wilk test. The level of significance in post hoc tests was corrected for multiple testing. The level of significance was set at α = 0.05, and all tests were two-sided. Statistical analysis was performed using R 3.6.1 with the packages multicomp 1.4-10, nlme 3.1-140, and plotrix 3.6-3. R, and the packages used are available from CRAN at http://CRAN.R-project.org/.

#### **3. Results**

#### *3.1. Group-Related Parameters of Specimens and in Vitro Setup*

To ensure that temperature change assessments were comparable with each other, only the implants with insertion torque between 40 and 60 Ncm were included in the analysis. Six machined and three sandblasted implants (groups 1 and 2, respectively) that exceeded an insertion torque of 60 Ncm, and three sandblasted etched implants (group 3) that had an insertion torque under 40 Ncm, were excluded from the analysis. From groups 4 to 6, all 15 specimens were analyzed. The resulting insertion torque was comparable for all 6 groups (group 1: 52.61 ± 5.10 Ncm; group 2: 48.97 ± 6.43

Ncm; group 3: 48.79 ± 5.19 Ncm; group 4: 50.10 ± 8.69 Ncm; group 5: 50.99 ± 7.97 Ncm; group 6: 50.72 ± 7.22 Ncm; *p* > 0.05; Figure 5a.).

The thickness of the cortical plate of the bovine bone rib was similar between the 6 groups (group 1: 2.51 ± 0.25 mm; group 2: 2.75 ± 0.47 mm; group 3: 2.61 ± 0.39 mm; group 4: 2.46 ± 0.12 mm; group 5: 2.46 ± 0.12 mm; group 6: 2.42 ± 0.09 mm; *p* > 0.05; Figure 5b).

The average roughness (Sa) of the machined implants (group 1: 0.93 ±0.16 μm; group 4: 0.37 ± 0.04) was, as expected, significantly lower than that of sandblasted implants (group 2: 1.86 ± 0.38 μm; group 5: 2.15 ± 0.24 μm; all comparisons have *p*-values < 0.001) and sandblasted and acid-etched implants (group 3: 2.14 ± 0.62 μm; group 6: 1.89 ± 0.18 μm; all comparisons have *p*-values < 0.001; Figure 5c). The difference in surface roughness between ceramic and titanium implants after sandblasting (group 2 vs. group 5) and after sandblasting and acid-etching (groups 3 vs. 6) was not statistically significant (*p* > 0.05). The roughness between the machined surfaces (groups 1 and 4) was not statistically significant (*p* > 0.05). In paired comparisons of the surface treatments, the machined and the sandblasted groups showed significant differences (groups 1 and 4 *p* = 1.213 <sup>×</sup> 10−4; groups 2 and 5 *p* = 0.025756). The sandblasted and acid-etched groups (groups 3 and 6) showed no significant differences (*p* > 0.05).

#### *3.2. Maximum Temperature Increase*

#### 3.2.1. Bone Window 1

In bone window 1 (located 3 mm subcrestally), the average temperature increase at the zirconia implants had a high statistically significant difference to that of titanium implants (*p* = 7.163 <sup>×</sup> 10–9, resulting from group-adjusted linear mixed-effects models).

Within the zirconia implants (groups 1: 19.94 ◦C ± 3.28 ◦C; 2: 19.39 ◦C ± 5.73 ◦C; 3: 15.44 ◦C ± 3.63 ◦C) there were significant differences between groups 2 and 3 (*p* = 0.04426) and groups 1 and 3 (*p* = 0.007525; Figure 6). By contrast, there was no significant difference between the comparison of groups 1 and 2 (*p* > 0.05).

**Figure 6.** Box plots showing rise of temperature at bone window 1—results of all 6 groups.

Within the titanium implants (group 4: 9.33 ◦C ± 4.18 ◦C; group 5: 9.20 ◦C ± 3.31 ◦C; group 6: 7.68 ◦C ± 2.68 ◦C), the temperature changes were similar for all groups (*p* > 0.05).

In paired comparisons of the different materials with the identical surface treatments, all groups showed highly significant differences (groups 1 and 4 *<sup>p</sup>* <sup>=</sup> 3.728 <sup>×</sup> <sup>10</sup>–11; groups 2 and 5 *<sup>p</sup>* <sup>=</sup> 7.721 <sup>×</sup> 10–6; groups 3 and 6 *<sup>p</sup>* <sup>=</sup> 1.725 <sup>×</sup> 10–6 Figure 6).
