Accuracy and safety of in-house surgeon-designed three-dimensional-printed patient-specific implants for wafer-less Le Fort I osteotomy

The design and fabrication of three-dimensional (3D)-printed patient-specific implants (PSIs) for orthognathic surgery are customarily outsourced to commercial companies. We propose a protocol of designing PSIs and surgical guides by orthognathic surgeons-in-charge instead for wafer-less Le Fort I osteotomy. The aim of this prospective study was to evaluate the accuracy and post-operative complications of PSIs that are designed in-house for Le Fort I osteotomy. The post-operative cone beam computer tomography (CBCT) model of the maxilla was superimposed to the virtual surgical planning to compare the discrepancies of pre-determined landmarks, lines, and principal axes between the two models. Twenty-five patients (12 males, 13 females) were included. The median linear deviations of the post-operative maxilla of the x, y, and z axes were 0.74 mm, 0.75 mm, and 0.72 mm, respectively. The deviations in the principal axes for pitch, yaw, and roll were 1.40°, 0.90°, and 0.60°, respectively. There were no post-operative complications related to the PSIs in the follow-up period. The 3D-printed PSIs designed in-house for wafer-less Le Fort I osteotomy are accurate and safe. Its clinical outcomes and accuracy are comparable to commercial PSIs for orthognathic surgery. Clinical trial registration number: HKUCTR-2113. Date of registration: 29 July 2016.


Introduction
Patients with dentofacial deformities suffer from reduced quality of life and low self-esteem as the consequences of poor function and aesthetic [1,2]. Such deformities are also known to be related to temporomandibular joint pain and obstructive sleep apnea [3,4]. Correction by orthognathic surgery has become a routine procedure in most oral and maxillofacial surgery centers in the world. The traditional method of surgical planning involves transferring the patient's jaw record through mounted stone models of the dentition, and through mock surgery on the stone models to simulate the surgical procedures, and to fabricate surgical wafers that are used to guide the intra-operative surgical movements [5]. However, significant errors could be incorporated in each step of the planning process which could affect the execution and ultimately the outcomes of the orthognathic surgery [6].
With the development of orthognathic planning software and three-dimensional (3D) imaging technology such as cone-beam computed tomography (CBCT) and intraoral scanning systems, virtual surgical planning (VSP) is made possible and has broadened the boundary of orthognathic surgery [7]. VSP allows visualization of the bony structure in a 3D perspective. Advancement in 3D printing further expand the possibilities in this field. The surgical wafers can be designed and printed by 3D printers after VSP, which in theory reduces the potential source of error from the transfer of records and laboratory-related variables [8][9][10]. However, no matter laboratory-made or 3D-printed, the use of a wafer to position the osteotomized maxillary segment in Le Fort I osteotomy relies on the mandible, which is a mobile unit at the temporomandibular joints. It is therefore inevitable to have certain discrepancies from the planned movement when the wafer is used, in particular the vertical dimension of the jaw positioning [9,11].
The concept of wafer-less positioning and fixation of Le Fort I osteotomy was developed to solve the wafer-related inaccuracies. Several studies have shown patient-specific implants (PSIs) planned and fabricated by commercial companies achieved good accuracy [12][13][14]. However, outsourcing the design and production of PSI has several drawbacks-such as long turnaround time (12), high cost (15,16), and the surgeon's lack of control in designing the implant. These disadvantages of commercial PSIs may limit its application and adoption, particularly in lower-income countries or teaching institutes.
With the popularization of computer-aided design/computer-aided manufacturing (CAD-CAM) technology and further advancement of 3D printing, it is predicted that the cost of 3D printing for metal will become more affordable. There is also a push for open source VSP programs and design tools to allow more surgeons to plan and design PSIs for orthognathic surgery and head and neck reconstruction [15][16][17]. The legislation and accreditation of approved PSIs and other implantable medical products in different countries are expected to change when the technology is further matured, so more patients may benefit from the PSI technologies but could not afford the commercial PSIs.
Our center initiated a research project to develop selfdesigned PSIs for head and neck surgeries [17]. For orthognathic procedures, a protocol was set for surgeons of the Le Fort I osteotomy to design PSIs and surgical guides inhouse. The feasibility, accuracy, and safety of such PSIs are yet to be proven and verified. Therefore, this study aimed to evaluate the accuracy and post-operative complications of in-house surgeon-designed surgical guides and PSIs for wafer-less positioning of the maxilla in Le Fort I osteotomy.

Study design
This was a prospective study on the accuracy and post-operative complications of patients treated with in-house surgeondesigned PSI for wafer-less Le Fort I osteotomy. Patients with dentofacial deformity under the care of Oral and Maxillofacial Surgery, Faculty of Dentistry, The University of Hong Kong, were recruited. The inclusion criteria were (1) patient age 18 years or older; (2) Le Fort I osteotomy as part or whole of the orthognathic procedure; and (3) in-house surgeon-designed patient-specific implant and surgical guide were planned for the Le Fort I osteotomy. The exclusion criteria were (1) patients who were pregnant; (2)

Virtual surgical planning (VSP)
All patients were assessed preoperatively by the chief surgeon with the same protocol. Clinical measurements, clinical photographs in the neutral head position, and study models were performed for treatment planning. CBCT scans of the whole skull region were also taken. The preoperative CBCT image was imported into Proplan CMF 3.0 software (Materialize, Leuven, Belgium) for 3D model reconstruction and virtual surgery. The images were then segmented to remove artefacts and ensure adequate bony surfaces are exposed for design and fabrication of the PSI. The orientation of the virtual skull image is based on multiple factors: the neutral head position, the facial midline determined during clinical examination, and the Frankfort plane of the 3D cephalometry analysis. Virtual surgery was performed in the program and movements were determined based on the clinical findings. Outcome was verified and predicted with digital simulation.

In-house surgeon-designed PSIs and surgical guide
After completing the VSP, the PSIs and surgical guide were designed on the planned maxilla model by the surgeon-incharge, using Materialise 3-matic 12.0 software (Materialise, Leuven, Belgium). The fixation screws were positioned to avoid any vital structures and tooth roots. The patient-specific fixation plate was designed to fit passively onto the surface of the virtual maxilla ( Fig. 1a-c). By reverse engineering, the planned maxilla was registered back to its original position together with the fixation screws. A patient-specific surgical guide was then designed to incorporate the existing screw holes on the maxilla (Fig. 2a-c). The completed design was sent to printing with selective laser melting (SLM) technology from grade 2 pure titanium by a titanium metal printer (Concept Laser Mlab cusing 200R Metal laser melting system, Germany). Removal of supporting units and smoothening of the edges were performed. Lastly, the PSI and surgical guide were autoclaved and sterilized before the surgery.

Surgical procedures
A standard exposure for the Le Fort I osteotomy was performed. After vestibular incision, the lateral wall of maxilla, infraorbital foramen, and the pyriform aperture were exposed. The patient-specific titanium surgical guide was placed to ensure passive fitting and temporarily secured with titanium screws, which were standard stock miniscrews (Matrix, DePuy Synthes) (Fig. 3a). The screw holes of the surgical guide were drilled, and the osteotomy at the anterior maxillary wall was marked. The guide was then removed. Osteotomies at the lateral nasal wall, the lateral maxillary wall, and the nasal septum were completed with osteotomes. Posterior separation was performed by tuberosity cut. The maxilla was down fractured and mobilized. Bone interferences were removed according to the surgical movement. The patient-specific titanium fixation plate was placed and secured with 6-mm titanium screws using the pre-drilled screw holes (Fig. 3b). The precise positioning of the maxilla was determined by a passive placement of the patient-specific implant, as well as cross-checking with the planned movement and its occlusal relation to the mandible.
After the Le Fort I osteotomy, other orthognathic procedures were performed where indicated. The wounds were closed in layers. Standard medications that include analgesics and antibiotics were prescribed.

Assessment of post-operative accuracy
Postoperative CBCT scans (the same scanner, which was calibrated, was used in all patients for all pre-operative and post-operative scans) were taken 1 month after the surgery. The VSP was superimposed onto the postoperative CBCT model of the maxilla to assess the accuracy of the PSIs. The 3D models were first aligned in Proplan CMF by superimposing an area that was outside the surgically treated region-the frontal part of the skull. The models were then imported into Materialise 3-Matic and aligned again using global registration until the average distance error after 50 iterations was less than 0.01 mm.
Four sets of measurements were made to evaluate accuracy: (1) the linear deviation between the planned and actual postoperative models of the skull along maxilla dental reference points (Fig. 4a); (2) the linear deviation along the x (horizontal), y  (3) the intercanine and inter-molar distances of the planned movement and the actual post-operative models; (4) the angular deviations of the principal axes in terms of pitch, roll, and yaw between the planned and postoperative models. The pitch was measured as the angle between the Frankfurt plane and the occlusal plane; the roll as the angle between the horizontal line and occlusal plane; and the yaw as the angle between the mid-arch and mid-sagittal planes (Fig. 4b). The definitions of all the points, lines, and principal axes used are listed in (Table 1). All the measurements were performed by one assessor who did not participate in the surgery to avoid inter-reviewer heterogeneity. The assessor randomly selected 20% of the cases for remeasurement to test for intra-assessor repeatability with a gap of 1 week to test for intra-assessor reliability. The two measurements were all within 0.05 mm (Fig. 5).

Post-operative complications relating to PSIs
Post-operative complications of the PSIs that include infection or exposure of the PSIs, loosening of screws, or incidences that led to removal of the PSI post-operatively were recorded at post-operative 6 weeks and 6 months.

Statistical analysis
All statistical analyses were performed using the software SPSS version 25.0 (SPSS Inc., Chicago, IL, USA). Kolmogorov-Smirnov and Shapiro-Wilk test were used for normality test. Descriptive statistics were used to report the accuracy of the maxilla position in terms of translation and rotation. Absolute values were used for description and subsequent analyses. Mann-Whitney U Test was used to investigate the difference in deviation between one-piece Le Fort I and two-piece Le Fort I. Wilcoxon signed-rank tests were used in comparison between deviation in pitch, yaw and roll. A 5% probability level was taken as the cut-off for statistical significance. Adjustment of p value with Bonferroni Test were done in multiple group comparisons to reduce the possibility of Type I errors.

Results
Twenty-five patients (12 females and 13 males) with a mean age of 26.4 years (SD 5.4 years, ranged 18-41 years) were recruited in the study. The majority of the patient presented Fig. 3 a The patient-specific titanium surgical guide was placed to ensure passive fitting and secured with titanium screws temporarily. The screw holes were then pre-drilled; b after mobilization of the maxillary segment and removal of bony obstacles, the patient-specific titanium fixation plate was tried in and secured with 6 mm titanium screws using the pre-drilled screw holes  The most inferior and mesial point of the incisal edge of maxillary left central incisor 21i The most inferior and mesial point of the incisal edge of maxillary right central incisor 13 The cusp tip of the right maxillary canine 23 The cusp tip of the left maxillary canine 16 The mesio-buccal cusp tip of the right maxillary first molar 26 The with maxillary hypoplasia (80%) with or without maxillary canting. Sixteen patients had Le Fort I osteotomy in one piece and 9 patients received a two-piece Le Fort I osteotomy. The demographic information, diagnoses and treatment received are presented in (Table 2).

Accuracy of the linear measurements
The absolute median deviation from planned position of maxilla were 0.74 mm on the x axis, 0.75 mm on the y axis, and 0.72 mm on the z axis. The median translation deviation was 1.75 mm (xyz axis). The mean deviation of the actual versus planned maxillary position at the x, y, and z axes were 0.91 mm (S.D. 0.72 mm), 1.12 mm (S.D. 1.09 mm) and 1.08 mm (S.D. 0.96 mm), respectively. The absolute median deviation of inter-canine and inter-molar distance were 0.54 mm and 0.98 mm, respectively (Table 3). Boxplots were produced for absolute linear deviation at different landmarks of maxilla, inter-canine, and inter-molar distance (Figs. 6 and 7).

Accuracy of the principal axes angles
The deviations of the principal axes between the post-operative maxilla models and the planned models were measured. The median deviation of the pitch, yaw, and roll planes were 1.40°, 0.90°, and 0.60°, respectively (Table 3). Wilcoxon signed-rank tests were used to investigate any differences between pitch, yaw, and roll. Statistically significant differences were found between pitch versus yaw (p = 0.010), and pitch versus roll (p = 0.001) ( Table 4). Boxplot was produced for absolute angular deviation from the planned position of the maxilla (Fig. 7).

Comparison between one-piece Le Fort I and two-piece Le Fort I
Mann-Whitney U tests were used to compare the difference in linear and angular deviation between one-piece and twopiece Le Fort I groups. No statistically significant difference was found in all landmarks, inter-canine, inter-molar distance, and principal axes (p > 0.05) ( Table 5).

Post-operative complications
No cases were presented with infection or exposure of the PSIs, loosening of screws, or incidences that led to removal of the PSIs up to post-operative 6 months.

Discussion
This study aimed to evaluate the accuracy of in-house surgeon-design PSIs and cutting guides for wafer-less positioning of the maxilla in Le Fort I osteotomy.
The key findings of this study were (1) the median deviations of the actual maxilla when compared to the planned position in the x, y, and z axes, and the inter-canine and inter-molar lines were less than 1 mm, which indicate an accurate positioning of the maxilla by the cutting guides and PSIs; (2) the deviations of the principal axes in pitch, yaw, and roll were 1.40°, 0.90° and 0.60°, respectively, with pitch significantly less accurate than the other two orientations; (3) there were no difference in accuracy of using PSIs in Le Fort I osteotomy in one-piece or two-piece; and (4) there were no post-operative complications of the PSIs for the Le Fort I osteotomy during the 6-monthsreview period. Traditional orthognathic surgery planning and execution relies on facebow transfer, stone model surgery, and laboratory-made wafers for positioning of the osteotomized jaws. The inevitable errors of manual record taking and transfer as well as the inadequacy of the articulator to simulate the mandibular movement have proved to be the causes of inaccurate surgical execution [6,18]. The advancement in virtual planning is a significant breakthrough in orthognathic surgery as it allows surgeons to visualize the bony structure pre-operatively and to plan the surgical movement with precision. Combined with 3D printing, surgical wafers can now be constructed without the need of laboratory transfer of records, thus reducing the potential errors during the process [8]. However, these 3D-printed wafers still led to discrepancies from planned movement due to their reliance on the mandible for the fixation of the maxillary segment. [19]. Imperfect fitting of the surgical wafer to the occlusion and the difference in the condyle position during the scanning process are also possible errors of a wafer-guided orthognathic surgery [20,21].
In order to avoid the errors from using a surgical wafer, the concept of wafer-less orthognathic surgery has blossomed. Wafer-less surgery with PSI enables accurate placement of the maxilla independent of the condyle or mandible, and without the need for extraoral reference points [22]. Several studies on commercial company-designed and   fabricated PSIs have demonstrated great potentials. It was shown that these PSIs enabled precise positioning and fixation in maxillary osteotomy [13] with adequate strength and stability [23]. Furthermore, wafer-less PSIs improved the accuracy, predictability, and surgical time of orthognathic surgery [24]. However, there are drawbacks of PSI made by commercial companies: (1) the high cost, (2) the long turnaround time that spans from 4 to 6 weeks; and (3) and the surgeon's lack of control in designing the implant. We discovered that by having our surgeons design the PSI in-house, the disadvantages of contracting commercial PSI can be overcome.
The accuracy of PSIs is one of the most important factors in determining their adoption. Accuracy is measured by comparing the difference between the planned model and the actual postoperative jaw position in linear and angular dimensions. It was shown that the discrepancy in linear deviation for the 3D wafer approach between the planned and actually movement could be as high as 2 mm for any variable [25]. In a systematic review on the accuracy of VSP in orthognathic surgery, it was agreed that linear deviations of less than 2 mm (xyz axis) between the planned and actual post-operative models were considered acceptable [24]. It was also stated in the review that the angular accuracy of pitch, yaw and roll of VSP were up to 2.75°, 1.7°, and 1.1°, respectively [24]. Previous studies on the accuracy of commercial PSIs showed the difference between planned and actual models were around 1 mm [12]. In this study, it is shown that in-house surgeon-designed PSIs could match commercial PSIs in terms of accuracy in the linear angulations. In addition, the deviations of the median angular measurements in pitch, roll, and yaw were extremely low (0.6-1.4°), indicating the execution of the orthognathic planning was precise.
Another novel finding of this study was that there were no differences in accuracy of PSIs in one-piece or two-piece Le Fort I osteotomy. The cases that require multiple segmentation of Le Fort I (3 or 4 pieces) were not analyzed as these procedures are not common in many centers around the world. Moreover, multi-segmentation of Le Fort I may be better fixed with the use of a surgical wafer and customized arch bar, instead of the sole application of PSIs.
One major advantage of designing PSIs in-house is the time saved. In our center, surgeons receive trainings to design PSIs. On average, a trained surgeon requires only 3 to 4 h to conduct VSP and design the surgical guides and PSIs for a case. The designs are then sent out for printing, and the products are typically ready in 7 working days. This in-house process is significantly more time-efficient than outsourcing the PSIs to be designed and made. With less time required to prepare for surgery, patient waiting time for operation is shortened and a turnover of cases is higher. Another advantage of designing in-house is that the surgeons can implement specific design features into the PSIs and surgical guides according to their personal preferences. It is also particularly beneficial for junior surgeons to understand the challenges of a case through the planning and designing process.
Safety, as reflected in the incidence of surgical complications, is an important factor to consider when implementing a new treatment modality. Rückschloß et al. suggested that VSP using PSIs can help reduce the risk of post-operative complications such as nerve injury or damaged dental roots because surgeons are able to mark out these structures and plan for the depth of penetration into bone [26]. In this study, there were no cases that presented with infection or exposure of the PSIs, loosening of screws, or incidences that led to removal of the PSIs in the early postoperative period. It proves that the in-house surgeon-designed PSIs are safe for use. An extended follow-up is required to observe if there are any long-term complications.
In conclusion, this study shows that in-house surgeondesigned 3D-printed PSIs for wafer-less Le Fort I osteotomy are accurate in the execution of VSP, with minimal discrepancies in all linear and angular measurements. The pitch is relatively less accurate when compared to the yaw and the roll. There were no complications observed in the early postoperative period. In-house surgeon-designed 3D-printed PSIs can be an alternative to commercial PSIs for waferless Le Fort I osteotomy as it shortens the preparation time of surgery, and potentially reduces the surgical cost of the patient or the institute.