The custom-made electrical stimulation system comprised an implant, the connecting cables, and a miniature stimulation device including a 3 V battery (Figure 1). The implant resembled a prism of 7.0 mm in length with a round-corner, triangular base of 4.4 mm diameter and was manufactured from Ti6Al4V alloy by primec GmbH (Bentwisch, Germany). An implant with a triangular base was chosen to create a gap between the bone (round drill hole) and the implant’s outer surface, which was supposed to be bridged by the newly synthesized bone, thus creating a defect model. The titanium implants had rough surfaces (surface roughness R_z = 12.9 ± 1.6 µm) to facilitate better adherence of osteoblasts. Additionally, the implant contained a central borehole to insert the connecting cables (copper wires coated with fluorinated ethylene propylene as insulating material) and three lateral boreholes that allowed exposure to the electrodes (Ti6Al4V wire). Each implant was connected to the miniature stimulation device via four cables – three wires carrying an electric current (electrodes) and one ground wire. For most of the length of the connection, the four cables were sheathed in a silicon tube, while the entry point of the cables into the implant was additionally reinforced with a more durable silicon (MasterSil 151Med, Master Bond Inc., Hackensack, NJ, United States). Figure 1 shows an illustration of the implant system.

The developed miniature electrical stimulation device was manufactured by Dorazil GmbH (Berlin, Germany) and contained a 3 V lithium cell as an energy source. The design of the stimulation device was described previously by Plocksties et al. (2020). Briefly, the circular-shaped printed circuit board contained an off-the-shelf MSP430F2274 microcontroller, a 2.4 GHz RF interface based on a TI CC2550 transceiver, and an output unit that approximated an alternating current (AC) sine voltage. With a diameter of 25.4 mm, a height of 1.93 mm, and a weight of 0.7 g at a total volume of 977 mm³, the stimulation device was suitably small for animal studies. The entire device, including the electronics on the circuit board, was encapsulated in a biocompatible polymer (NuSil™ MED1-4213 and MED3-4213, Avantor, VWR, Deutschland). The entry point of the cables into the miniature stimulation device was additionally sealed by several cycles of embedding it in silicone (MasterSil 151Med, Master Bond Inc., Hackensack, NJ, United States). Each stimulation device was able to cater for two implants.

The stimulation parameters were transmitted to the stimulation device over a radio-frequency (RF) interface, enabling a wide range of adjustable parameters. At the same time, feedback control allowed monitoring of whether the stimulation device was still working correctly.

Before animal testing, numerical simulation was used to evaluate the electric field distribution in the animal bone and on the implant’s surface of the implant. The procedure was carried out according to the previously published, detailed protocol to simulate electrical stimulation in rabbit bone (Su et al., 2016). Briefly, the bone model was created in COMSOL (COMSOL Multiphysics® version 4.3b, Comsol AG, Göttingen, Germany) and was integrated into iSIGHT to run the optimization. The modeling in COMSOL was launched through MATLAB batch mode. Frequency and signal waveform were kept constant at 20 Hz and sinus wave. Both the cancellous and cortical bone in the rabbit knee were considered homogenous and isotropic to reduce the complexity of the calculation. Dielectric properties of the rabbit bone, i.e., the initial electric conductivity and permittivity of the tissues, were used as reported by Su et al. (2016). The material properties of the designed implants in the numerical simulation were incorporated into the model based on data sheets from the manufacturer. Table 1 shows the dielectric properties of the adjacent tissues and the implant material used in the electrical simulation at 20 Hz.

A parametric study was carried out to ensure that the majority of the implant surface exhibited the optimum electric field interval of 5–70 V/m. For the parametric study, all parameters were kept constant except for the peak voltage that was varied from 50 mV to 300 mV in 50 mV steps.

The animal experiments were approved by the local review board of the Landesamt fuer Landwirtschaft, Lebensmittelsicherheit und Fischerei M-V (LALLF MV, reference number AZ LALLF M-V/TSD/7221.3-1.1-076/12). The animal study complied with policies and principles established by the Animal Welfare Act and the NIH Guide for the Care and Use of Laboratory Animals. For this purpose, nine female New Zealand White rabbits (Envigo RMS GmbH, Hillcrest, United Kingdom) with an average weight of 3.66 ± 0.30 kg and an age of 6 months were used in this pilot study. New Zealand White rabbits are a well-established experimental model for preliminary research regarding the osseointegration of titanium implants (Götz et al., 2004).

Animals underwent the surgical procedure under general anesthesia, including ketamine (50 mg/kg of body weight, intramuscular; bela-pharmGmbH & Co. KG, Vechta, Germany) and xylazine (5 mg/kg of body weight, intramuscular; Bayer AG, Leverkusen, Germany). After general anesthesia was allowed to work for 10 min, metamizole at 40 mg/kg body weight and enrofloxacin (Baytril^® 40 mg/kg of body weight, Bayer AG, Leverkusen, Germany) were injected into the gluteal muscle. Both hind legs were then shaved and disinfected before the surgical site was numbed by subcutaneous administration of lidocaine hydrochloride (Xylocitin^® 2%; MIBE GmbH, Brehna, Germany). Stab incisions were made medially to the proximal tibial metaphysis to facilitate the positioning of the drill. Using a circular drill (diameter: 4.3 mm), bone defects of 8 mm length were prepared in the proximal tibia. Additionally, a lateral incision in the area of the proximal femur was used to form a skin pocket in the lateral pelvic region, and the miniature stimulation device was slid into the pocket (Figure 2). The implants were advanced under the skin towards the knee joint and inserted into the tibial defects by press-fit while taking care to avoid tensile stress on the connecting cables. The wounds were sutured with absorbable suture material and sealed with silver spray. The total duration of the surgery was 50–60 min.

In the first 5 days after surgery, the animals received subcutaneous injections of carprofen (5 mg/kg of body weight) and enrofloxacin (Baytril^® 40 mg/kg of body weight, Bayer AG, Leverkusen, Germany) and were supplemented with metamizole and a mixture of vitamins via the drinking water. All animals were weighed once a day and examined twice daily for signs of pain and wound-healing disorders. The first operated animal had to be euthanized early due to an implant infection. As a consequence, the sealing of the stimulation device was improved, and no further peri-implant infections were observed in the subsequently operated animals.

In these animals, electrical stimulation via the implants was started on the third postoperative day. Of the implants, one was electrically stimulated, while the other remained unstimulated and served as a negative control. The stimulated implants in the animals were subject to a stimulation protocol at 20 Hz of 3 × 45 min with 150 mV peak voltage daily, evenly distributed over a period of 12 h during daytime. Postoperatively and after 6 weeks, the animals were X-rayed to confirm the correct location of the implants and the stimulation device. Electrical stimulation was carried out for a total of 12 weeks, whereupon the animals were sacrificed. Subsequently, the tibiae were carefully dissected, freed of soft tissue, and fixed in a buffered formalin solution (4%). The osseointegration of the implant was evaluated by determining the bone-to-implant contact (BIC) with μCT and histomorphometry.

The histomorphometric analysis was performed as previously described by Gabler et al. (2015). Briefly, the fixed bone specimens, including the implant, were dehydrated in a graded series of alcohol and embedded in polymethylmethacrylate before being cut with a cutting/grinding system with 0.1 mm diamond bands according to the thin-section technique by Donath and Breuner (Donath and Breuner, 1982). For each implant, one slice parallel to the base was prepared from the middle of the implant at exactly 3.5 mm from the edge. Subsequently, a microgrinding system (Exact 400CS; Exakt Apparatebau, Norderstedt, Germany) was used to reduce sections to a thickness of approximately 20 μm using 1200-grit sandpaper and to polish these afterwards with 4000-grit polishing paper. The bone sections were then stained with toluidine blue.

For histological evaluation, images were taken at 42.3× magnification under a digital microscope (VHX-6000, Keyence Deutschland GmbH, Neu-Isenburg, Germany). The histomorphometric images were analyzed with ImageJ (Schneider et al., 2012).

Bone-to-implant contact (BIC) in histomorphometry was quantified as the total length in micrometers (BIC length) that resulted from the sum of all segments with direct contact between toluidine blue stained bone and the implant surface. To calculate the percentage of BIC in relation to the total circumferential length of the implant, a polygonal frame was drawn around the implant corresponding to the maximum circumference. The recesses for the electrodes were subtracted from the length of the frame to determine the maximum length available for BIC. The percentage was calculated as the actual total length divided by the maximum available length of BIC.

Additionally, the area of newly formed bone in the proximity of the implant was assessed by drawing a circle around the implant with a distance of 100 µm to each of the three corners of the triangular base and measuring the area of toluidine blue stained bone (Figure 3). The thus drawn circle covered a slightly bigger area than the initial borehole. The area of the implant was subtracted from the total area of the circle to calculate the percentage of newly formed bone in the proximity of the implant by dividing the blue stained areas by the total area of surrounding tissue in the circle.

The acquisition of μCT image data of the specimens was performed with a microcomputer-tomograph Nanotom 180 nF (Phoenix nanotom, GE Measurement and Control solutions, phoenix|X-ray, Wunstorf, Germany). Voltage and current were set to 70 kV and 135 μA to reach the optimum contrast. The used μCT was a cone-beam computer tomograph with vertical specimen alignment. The samples rotated 360° in 0.75 steps. At each step, three 2D images were recorded; altogether, 1,440 2D images were acquired per sample. The reconstruction of CT data (composing X-ray 2D images to a 3D volume) was performed with the software datos|X-reconstruction (GE, Wunstorf, Germany). A beam hardening correction of 6.7 was used to compensate for the inhomogeneous reconstructed volume of the implant since titanium implants are harder to penetrate by X-rays than bone.

For further processing, the transformation of the volume data in the DICOM data was required. Each DICOM dataset had a maximum voxel edge resolution of 4–5 μm. The segmentation algorithm was executed on a workstation (Intel Quad Core Q9400 2.66 GHz, 2 GB RAM) with the segmentation software Amira 5.4.1 (FEI Visualization Sciences Group, Hillsboro, OR, United States) and consisted of the following processing steps: image data preprocessing, segmentation, and surface postprocessing. The segmentation algorithm comprised threshold detection of the histogram-oriented greyscale intensity of voxels, followed by inspection and interactive editing of the segmented areas in each slice of a dataset.

Two regions of interest (ROIs) for the analysis of the reconstructed three-dimensional images were set within the range of the implant surface with approximately 1 mm surrounding bone tissue, as shown in Figure 4A. The first ROI was placed at 1 mm from the bony side of the implantation and was also 1 mm wide. ROI 1 anatomically indicated the cortical bone. The second ROI was located 3.5 mm from the edge of the implant and anatomically corresponded to cancellous bone and bone marrow. The starting points for both ROI were determined by selecting the sectional images in which the implant was visible, taking into account the Z-value, i.e., the sectional thickness of the image. Based on this information, the starting point of the ROI 1 was calculated as follows: 1   mm Thickness   o f   slice   in   mm = Number  o f the image at  1  mm f rom the edge

For each ROI, a circular area was placed around the implant base surface, which was then extended to a width of 1 mm to create a cylinder (3D) around the implant (Figure 4B). The bone volume was determined using the homogeneous Hounsfield unit (HU) of the bone. Within the ROIs, the titanium implant was marked and excluded from the calculation of the possible bone volume. The plastic coating of the electrode could only be segmented from the surrounding bone tissue by manual identification and subsequent correction, as automatic identification failed due to similar absorption values. Following this, the segmentation surface models were computed with a voxel-sustaining algorithm matching the exact voxel boundaries of each segmented material. Finally, the marked bone volume in mm³ was calculated using the Amira 5.4.1 software.

Data analysis and illustration were performed by GraphPad PRISM v.7.02 (GraphPad Inc., San Diego, CA, United States). Results are shown as box plots. Boxes depict interquartile ranges, horizontal lines within boxes depict medians, and whiskers depict maximum and minimum values. Data were analyzed regarding normal distribution with the Shapiro-Wilk test. Comparisons of original as well as percentage data of BIC length and the area of newly formed bone from the histomorphometric analyses (2D) between electrically stimulated and unstimulated samples were performed by t-test or Mann-Whitney test depending on the normal distribution of data. These tests were also used to compare µCT data (3D) of stimulated to unstimulated tibia samples. The analyzed µCT data included the BIC area and the bone volume in the cylinder around the entire implant (Figure 4C) as well as in the cylinders defined as ROI1 and ROI2. Since µCT data were scored blinded by two independent researchers, each score was incorporated as a single value in the statistical analysis. Significance was set to p-values less than 0.05. Further details of statistical tests are indicated in the results section and the figure legends.

The parametric study, that was conducted within the range of 50–300 mV, showed that 150 mV peak voltage had to be applied to this particular implant design to achieve the optimal electric field of 5–70 V/m on the surface of the implants. This particular electric field strength resulted in the optimum effect on bone growth (Kraus, 1984). The electric field distribution around the implant in the hind limbs of the animals was calculated in the early phase of osseointegration when it is thought that the defect in the bone is filled with blood (Figures 5A,C) as well as in the later phase when the blood (or bone marrow) in the defect is replaced by newly formed bone (Figures 5B,D). Numerical simulation showed that not only on the implant´s surface, but also within the adjacent bone tissue the strength of the electric field ranged from 5 to 70 V/m. When gaps between bone and implant were considered as blood, the electric field on the surface of the chosen implant covered almost the entire implant electrode, except the wire electrode and the round edges of the implant. Due to its direct contact with the adjacent bone tissue, the round edge is of less importance in the defect model. When gaps were considered as cancellous bone, the electric field was distributed equally on nearly the entire surface of the implant electrode rather than the surface of the wire electrode.

The iso-surfaces of the electric field distribution in the gaps between bone and implant considering gaps as blood or cancellous bone are depicted in Figures 5C,D, respectively. In both cases the electric field within the stimulated tissue was homogenously distributed throughout the gaps around the implant.

During the 84 days of stimulation, it became apparent that some of the electric circuits were unresponsive to the transmission of stimulation parameters by the radio transceiver. The days when a certain electrode stopped working and was no longer visible in the transmission software were recorded to determine the actual duration of electrical stimulation. Explantation of the stimulation devices after the sacrifice of the animals revealed that the failure was caused by a complete discharge of the battery. This was probably due to insufficient leak-proof sealing of the silicon sheathing. In total, two implants were electrically stimulated over the entire duration of 84 days, while three further implants that were stimulated for at least 3 weeks stopped working after 19, 31, and 43 days, respectively. All implants, including the five unstimulated control implants, remained in the rabbit knees for the entire period of 84 days.

In the histomorphometric images, the length of BIC (mm) as well as the area of newly formed bone (mm²) within the bone defect created by the borehole was determined for tibia samples stimulated with 150 mV (3 × 45 min/day) and the unstimulated controls. When comparing the five stimulated and five unstimulated samples, there was no effect of stimulation on the length of BIC (p = 0.3095, Mann-Whitney test, Figure 6A), but a trend for an increase in bone area in the defect after electrical stimulation (p = 0.0600, unpaired t-test, Figure 6B). However, since the cut for the image slice was not always set exactly parallel to the base of the implant, there were distortions regarding the length and area between the images. To address this bias, the ratios of actual length and area to possible length and area as determined in the respective image were calculated and expressed as percentages. The comparison of the percentage values showed a significant effect of electrical stimulation on the area of newly formed bone (p = 0.0418, unpaired t-test, Figure 6D). On average, 27.7% of the defect area was filled with newly formed bone after thrice daily stimulation for 45 min with 150 mV, while without electrical stimulation only 13.0% of the defect area represented new bone. However, there was no influence on ratios of BIC length (p = 0.2629, unpaired t-test, Figure 6C).

The results of the µCT analyses are depicted in Figure 7. When analyzing the BIC area for the entire implant as well as the volume of newly formed bone in the entire defect space, there were no significant effects of electrical stimulation despite the observation of higher mean values for both measures after stimulation with 150 mV compared to unstimulated control (BIC area: 20.2 mm² vs. 17.6 mm²; Bone volume: 8.6 mm³ vs. 7.2 mm³ for thrice daily stimulation for 45 min with 150 mV vs. unstimulated control, respectively). The detailed analysis of the two different ROIs showed a significant increase of BIC area in the region of cortical bone (ROI 1, p = 0.0190, unpaired t-test, Figure 7B) but not in the cancellous bone or intramedullary (ROI 2). No differences were observed in bone volume in the ROIs (Figures 7E,F). It is interesting to note that there were significant differences in the volume of newly formed bone between the two ROIs, with higher bone volume in the cortical region compared to the cancellous region. This was observed for the electrically stimulated (p < 0.0001, paired t-test) as well as unstimulated bone samples (p = 0.0112, paired t-test).