This retrospective study evaluated the precision and positional accuracy of different orthognathic procedures following virtual surgical planning in 30 patients. To date, no studies of three-dimensional virtual surgical planning have evaluated the influence of segmentation on positional accuracy and transverse expansion. Furthermore, only a few have evaluated the precision and accuracy of genioplasty in placement of the chin segment. The virtual surgical plan was compared with the postsurgical outcome by using three linear and three rotational measurements. The influence of maxillary segmentation was analyzed in both superior and inferior maxillary repositioning. In addition, transverse surgical expansion was compared with the postsurgical expansion obtained. An overall, high degree of linear accuracy between planned and postsurgical outcomes was found, but with a large standard deviation. Rotational difference showed an increase in pitch, mainly affecting the maxilla. Segmentation had no significant influence on maxillary placement. However, a posterior movement was observed in inferior maxillary repositioning. A lack of transverse expansion was observed in the segmented maxilla independent of the degree of expansion.
Conventional planning of orthognathic surgery is carried out on the basis of an X-ray cephalometric analysis and mock surgery on plaster cast dental models mounted in a semi-adjustable articulator. Hitherto, this has been the gold standard for treatment planning. Unfortunately, cast model planning is complex, and the risk of incorporating errors during transfer of the occlusal plane may lead to errors in the treatment plan. Furthermore, occlusal cant and the asymmetric deformity of the bony skeleton are difficult to diagnose correctly, even with accurate recreation of the occlusal plane. In comparison, virtual surgical planning enables precise analysis of a three-dimensional (3D) model that accurately represents the clinical situation and facilitates diagnosis and treatment plannning.
A previously performed systematic review of the literature found only seven clinical trials validating the virtual surgical planning protocol in orthognathic surgery. A recent study by Zinser et al. compared accuracy and precision between conventional planning, navigation-assisted placement, and computer-aided design and computer-aided manufacture (CAD/CAM) surgical positioning splints. All studies focused on bimaxillary procedures; only three included genioplasty, and none included segmentation of the maxilla.
The main purpose of this study was to retrospectively evaluate the difference between the virtual surgical plan and the postsurgical outcome using 3D virtual surgical planning software. Segmentation of the maxilla and genioplasty of the mandible were included in order to assess the added complexity and subsequent influence on precision and accuracy.
Materials and methods
Virtual surgical planning has been used routinely in the Department of Oral and Maxillofacial Surgery, Odense University Hospital, Denmark since 2008 for planning orthognathic surgery. Currently, more than 200 orthognathic surgical procedures have been planned using virtual surgical planning.
A retrospective, observational study was performed on a cohort of 30 patients who were selected randomly from a population of 72 patients treated consecutively from 1 June 2011 to 1 January 2013. For inclusion, patients had to have undergone bimaxillary orthognathic surgery due to abnormal growth of the jaws and to have attended follow-up 1 week after surgery. All patients were treated with bimaxillary orthognathic surgery with/without segmentation of the maxilla and/or genioplasty due to abnormal growth of the jaws and had attended follow-up 1 week after surgery. Patients who were diagnosed with a temporomandibular joint disorder and those who had undergone single jaw orthognathic surgical procedures were excluded from the study. Randomization was achieved by a blinded, third party to avoid selection bias and give a realistic assessment of precision in routine orthognathic procedures. Seventy-two patients were assigned a number and 30 corresponding numbers were picked. All patients from the pool of 72 had an equal chance of being included in this analysis.
Virtual surgical planning and orthognathic procedures
The standard Houston/computer-aided surgical simulation (CASS) planning protocol (Materialise, Leuven, Belgium) for orthognathic surgery was followed, as described by the manufacturers and in previous investigations. The facial skeleton was digitally recreated from a cone beam computed tomography (CBCT) scan (NewTom 3G; QR s.r.l., Verona, Italy) with standard settings (field of view (FOV) 20 cm × 20 cm; 110 kV; radiation exposure 59 μSV, according to the 2005 International Commission on Radiological Protection tissue weighting factors ). A preoperative CBCT was taken just prior to the surgery, after orthodontic decompensation retained with passive wires. Virtual surgical planning was carried out using Dolphin 3D (Dolphin Imaging and Management, Chatsworth, CA, USA) in an online meeting between the surgeon and a biomedical engineer at 3D Systems–Medical Modeling.
Following planning, surgical splints were produced using CAD/CAM printing techniques ( Fig. 1 ), and the treatment plan was finalized by marking the inferior alveolar nerve and the distances of proposed movements between bony segments.
Four types of surgery were planned and performed: bimaxillary surgery, bimaxillary surgery with segmentation of the maxilla, bimaxillary surgery with genioplasty, and bimaxillary surgery with segmentation of the maxilla and genioplasty. Orthognathic surgery was carried out according to the treatment plan using surgical splints and surgical calipers. Surgical procedures followed well-recognized protocols for orthognathic surgery, with bilateral sagittal split and Le Fort I osteotomies. Segmentation of the maxilla was performed as a three-piece segmentation between the lateral incisor and the canine bilaterally. The vertical movement was controlled by external measurements from the right medial canthal ligament and internal measurement relative to the osteotomy according to the planned movement. The sagittal split was fixed with three bicortical screws along Champy’s line bilaterally, and the maxilla was fixed with four L-shaped plates. Bony interferences were removed and interposed in the osteotomies. The surgical splint was fixed for 6 weeks in segmented procedures and for 1 week in non-segmented procedures. All surgeries were performed according to the standardized departmental approach by three surgeons (TT, EA, and PT).
Follow-up 1 week after surgery included a clinical examination and a CBCT scan. The surgical splint was still in place during examination and scanning.
Evaluation of accuracy and precision
The main outcome variable of interest was the difference between the virtual planned movement and the surgical movement obtained. Alignment and placement of reference points and measurement of the planned and postoperative movement were performed by 3D Systems–Medical Modeling and inserted into a spread sheet. The first author performed the data analysis.
The following data were collected retrospectively from the patient journal: patient age, surgeon, type of surgery, date of surgery, complications during surgery, and intraoperative alterations to the plan.
After surgery, CBCT scans were taken of the patients, and the planned outcome was superimposed on the postoperative outcome to determine how well the virtual surgical plan translated to the surgical result. The postoperative CBCT scans were converted to 3D surface files using both automatic and manual segmentation techniques (Mimics; Materialise, Leuven, Belgium). The virtually planned anatomy was aligned to the postoperative anatomy using best-fit surface alignment. The maxillary anatomy was aligned using the non-operated midface as its own reference ( Fig. 2 ). For the mandible (as this is a mobile segment), the anatomy was compared by aligning the condylar head position and the vertical position of the anterior mandible. Additionally, the genioplasty was compared using best-fit alignment of the distal mandible. When the condylar head position was assessed, the maxillary anatomy was aligned as described above and the mandible was brought along in the scanned relationship with the midface.
Reference points were placed on the top of the mesiobuccal cusp on the first molar on each side and in the midline at the edge of the central incisors, both in the maxilla and the mandible ( Fig. 3 ). Mean linear differences between the virtual surgical plan and postoperative outcome were calculated as the mean difference of the three reference points in all three dimensions. The linear difference for the placement of the chin segment was calculated at a single point, as the linear difference in menton. All measurements were recorded in relative numbers according to the corresponding axis to evaluate any systematic differences in direction. The positive values of the axis were left (mediolateral axis), anterior (anteroposterior axis), and superior (superoinferior axis).
Differences in orientation were calculated as the difference between a representative reference point and the mean linear difference of the maxilla or mandible ( Fig. 4 ). Roll was evaluated around the anteroposterior axis as the individual superoinferior difference of the mesiobuccal cusp of the first molar relative to the roll. A positive difference in roll meant superior movement of the left molar and inferior movement of the right molar. Pitch was calculated around the mediolateral axis as the difference in superoinferior direction between the midline of the incisal edge and the mean linear difference. A positive pitch meant an upward movement of the central incisors and a decrease in the occlusal plane angulation. Yaw was calculated around the superoinferior axis as the transverse difference between the midline of the incisal edge and the mean linear difference. A positive difference in yaw meant a rotational displacement of the incisors to the left. The orientation of the chin segment was not evaluated. This method accounts for all six degrees of freedom in a 3D model, as proposed by Xia et al. and Hsu et al.
The transverse expansion of segmented maxillary procedures was calculated between the mesiobuccal cusps on the first molars. The difference between the virtual surgical plan and the postsurgical outcome was evaluated against the planned expansion to visualize correlations.
Normality of the overall distribution was evaluated visually with a box-and-whisker plot ( Fig. 5 ). Non-parametric statistical hypothesis testing was performed due to the low number of participants in subgroups. The evaluation of heterogeneity with regard to surgeon was performed using Kruskal–Wallis analysis for the maxilla, mandible, and menton in all three dimensions. If no difference existed, the data were to be pooled and analyzed as one database; if differences existed, the results were to be investigated individually. The mean linear difference was used as a measure of accuracy and the standard deviation as a measure of precision. Additional statistical evaluation was provided using 95% confidence intervals. Analysis of the differences between the virtual plan and the postoperative outcome was performed by Wilcoxon signed-rank test, while the Mann–Whitney U -test was used to analyze differences between the dependent groups when comparing precision between the independent subgroups. Multiple tests were performed, and therefore chance findings may occur. The statistical significance level in all tests was set at P ≤ 0.05. The clinical success criterion was set at a difference of less than 2 mm between the virtual surgical plan and the actual surgical outcome, as proposed in previous studies.
A random sample of 30 patients (10 males and 20 females) was analyzed. All had undergone surgery between 1 June 2011 and 1 January 2013. The mean age of the patients was 23.1 ± 6.8 years, with a median of 21 years and a range of 18–42 years.
Thirteen bimaxillary procedures, four bimaxillary procedures combined with genioplasty, 11 bimaxillary procedures combined with three-piece segmented maxilla, and two bimaxillary procedures with both segmented maxilla and genioplasty were performed. The genioplasty was positioned by a guide in three patients and placed by hand in the remaining three patients.
Intraoperative complications occurred and were resolved during surgery as close to the planned outcome as possible. In one surgery, the intermediate surgical splint did not fit. The final splint was mounted and the segments had to be placed by hand according to the planned position. To date, this is the only intermediate surgical splint that did not fit. However, one intermediate and one final surgical splint were judged by the surgeons to have a poor fit, but the fit did not interfere clinically with the planned outcome.
Moreover, an unfavourable split (or bad split) occurred during two surgeries, both on the left side during the sagittal split. In one case, the split was repaired using two bicortical screws between the proximal segment and the free segment at Champy’s line, and two bicortical screws between the free segment and the dental segment at the base of the mandible. In the other case, the intermediate segment was fixed to the dental segment with an internal fixation plate, but the condylar segment could not be fixed and was left floating. When the condyle was left floating, the split did not have a negative impact on the lateral placement of either maxilla or mandible. However, the maxilla was positioned 2.7 mm posterior compared to the virtual surgical plan.
Whenever intraoperative complications occurred, the data were compared with mean differences and standard deviations. Intraoperative complications were rectified within one standard deviation of the mean; therefore all procedures were included in the final analysis.
Accuracy and precision
Variation in differences within and between surgeons was calculated by Kruskal–Wallis analysis, but no statistical significance was observed. The P -values were evenly distributed and ranged from P = 0.07 to P = 0.91. No heterogeneity between surgeons was noted. Therefore, the results of the procedures were pooled and analyzed as one group.
All mean accuracies, measured as the mean linear difference for the maxilla, mandible, and chin segment in all three planes, were within 0.5 mm ( Table 1 ). The mean precision, measured as the standard deviation, had the smallest deviation superoinferiorly, followed closely by mediolateral deviation, and finally the largest deviation was found anteroposteriorly. Figure 5 shows the distribution of the mean linear differences between the planned and postsurgical outcomes as a box-and-whisker plot.
|Segment||Axis||Linear difference||SD||95% CI Upper limit||95% CI Lower limit||P -value a Planned vs. outcome|
|Maxilla||Mediolateral||−0.31 mm||1.41 mm||0.20 mm||−0.81 mm||0.284|
|Anteroposterior||−0.39 mm||2.38 mm||0.46 mm||−1.24 mm||0.237|
|Superoinferior||−0.05 mm||1.37 mm||0.45 mm||−0.54 mm||0.644|
|Mandible||Mediolateral||−0.14 mm||1.40 mm||0.36 mm||−0.64 mm||0.910|
|Anteroposterior||0.28 mm||2.08 mm||1.02 mm||−0.47 mm||0.658|
|Superoinferior||0.13 mm||0.73 mm||0.39 mm||−0.14 mm||0.614|
|Menton||Mediolateral||−0.16 mm||1.89 mm||0.56 mm||−0.89 mm||0.951|
|Anteroposterior||−0.16 mm||3.52 mm||1.20 mm||−1.51 mm||0.977|
|Superoinferior||−0.41 mm||1.07 mm||0.01 mm||−0.82 mm||0.110|