Predictability of maxillary positioning: a 3D comparison of virtual and conventional orthognathic surgery planning

In orthognathic surgery, virtual planning is gradually taking over in clinical practice. The possibilities range from CAD/CAM manufactured splints to customized titanium plates as cutting guides or splint-less navigation [1, 2]. Besides the improved visualization of craniofacial deformities, such as occlusal canting and asymmetries [3, 4], virtual surgery planning (VSP) is believed to be less time-consuming and less expensive than conventional surgery planning (CSP) [5]. The elimination of face-bow transfer and the mounting of dental casts, which are known as major sources of error in CSP [6‐8], provide further arguments for the use of VSP.

In articulator experiments, rapid-prototyped surgical splints and manually manufactured splints showed similar accuracy [9, 10]. However, the error expected by the laboratory steps in CSP is less based on splint manufacturing itself but more on the transfer of the patient’s individual inclination of the occlusal plane and the orientation of the plaster casts mounted on the articulator [7, 11]. For example, an angular discrepancy between the real and the mounted occlusal plane of 20° may lead to a vertical maxillary displacement of more than 3 mm during a planned sagittal forward movement of 10 mm [11]. This type of error is avoided in VSP.

Several studies have investigated the accuracy and predictability of either conventional [12‐15] or virtual planning procedures [1, 16‐18]. However, only two studies compared both methods within the same team of clinicians in rather small samples [19, 20]. Their analyses were based on linear measurements in CTs [19] or two-dimensional (2D) measurements in cephalograms [21]. The advantages of three-dimensional (3D) imaging have not been fully used thus far. Therefore, there is still a lack of comparative studies for the accuracy of splint-based maxillary positioning in CSP versus VSP [22].

The knowledge of errors in maxillary positioning is essential as the patients usually have high expectations about this elective surgery [23] and are known to be very sensitive to post-surgical facial deviations [24]. In general, positional differences within 2 mm are assumed as clinically irrelevant [17, 19, 20]. For midline deviations, a more stringent threshold of 1 mm should be applied [25]. The aim of the present study was to investigate if the differences between the planned and achieved postoperative outcomes are within these clinically acceptable limits in splint-based VSP and CSP, and to compare whether one method is superior to the other.

This retrospective, comparative study was approved by the Institutional Ethics Committee (no. 7/1/16) in accordance with the Declaration of Helsinki. All patients provided written informed consent to participate in the study.

The accuracy of maxillary positioning in orthognathic surgery was investigated in 52 patients and was compared between virtual and conventional surgery planning (a detailed description of the study sample is displayed in Table 2). There was no association between the type of surgery planning and skeletal deformities, regarding skeletal class (p = 0.192), vertical relation (p = 0.498) or maxillary (p = 0.876) and mandibular asymmetry (p = 1.0). Bland–Altman plots revealed high intra- and interrater agreement for all measurements with average differences < 1 mm/ 1° and small limits of agreement.

VSP and CSP both demonstrated significant discrepancies between the planned and actual achieved positions of the maxilla (Table 3). Regarding maxillary translation, inaccuracies were highest for vertical movement [i.e., U1(z), see Fig. 2]. The positive value of the mean difference (CSP: 1.4 mm; VSP: 2.1 mm) revealed that this discrepancy was caused by an undercorrection of the planned impactions. For rotational movements, the inclination of the occlusal plane around the y-axis was the least predictable variable with over- and underachievement of the planned maxillary inclination. Considering clinical relevance, in vertical direction 73% in the CSP group and 46% in the VSP group achieved the desired maxillary position (Table 4). Interestingly, in most patients the maxillary position was undercorrected while overcorrection was observed in one case only. No significant difference in accuracy between CSP and VSP was observed in any dimension (Table 5).

The scatterplots (Fig. 3) revealed that inaccuracies in vertical translation correlated significantly with discrepancies in sagittal translation (r = − 0.49; p < 0.001) and antero-posterior rotation [OcP(y)] (r = 0.668; p < 0.001). This means that undercorrection of vertical impaction was associated with increased sagittal movement and reduced posterior inclination of the maxilla. Consequently, a correlation between sagittal translation and OcP (y) was found (r = − 0.607; p < 0.001). Regarding transverse movements, undercorrection of occlusal canting (OcP(x)) correlated with an overachievement in midline correction (r = − 0.613; p < 0.001).

No correlations between the magnitude of the planned movements and the inaccuracy in maxillary positioning were observed (Fig. 4).

The comparison of VSP and CSP revealed that the clinically achieved predictability of both methods is similar. VSP and CSP demonstrated significant differences between the planned and postoperative outcomes in all dimensions, with the vertical impaction of the maxilla as the least predictable movement. In all but one cases, the vertical movement was undercorrected, which may result in persistence of vertical maxillary excess, increased lower face height and gummy smile. This is not surprising as errors in maxillary impaction are well-known in orthognathic surgery [6, 14, 19, 20, 25] and deviations between the predicted and the actual vertical movement of 10 mm have been described [14].

While the average vertical discrepancy in CSP was within the clinically acceptable limit of 2 mm, the VSP group exceeded this limit by 0.1 mm. However, this difference between both methods was statistically not significant. Nonetheless, this outcome was against our expectations as CSP and VSP result in the same design of surgical splints, which have no control on the vertical maxillary movement. Therefore, the vertical position has to be validated intraoperatively by the use of internal or external reference points [15], what was done by the surgeons in our study in both groups. One possible explanation for this discrepancy may be the differing splint material between VSP and CSP. The VSP material had a lower flexural strength compared to the CSP material, which may have affected the resistance to deformation under the load during surgery. Due to the manufacturing process, the splints in CSP were thicker than in VSP, which possibly had a further effect on splint stability. Additionally, thin splints mask the tendency of the surgeon to remove less bone than required at the superior part of the maxilla. In contrast, a thick, bulky splint reveals a lack of vertical bone removal, especially in the posterior area, more obviously by showing a bony gap in the anterior contact area and encourages the surgeon to remove this interference [31, 32]. However, this contrasts with the expectation that thinner splints are more favorable as they reduce the magnitude of mandibular autorotation and the displacement of the condylar position as seen for example in occlusal registrations, in which accuracy is of tremendous importance [33]. Further studies regarding splint thickness in VSP are required to elucidate this issue. Splint-less surgeries with customized titanium plates and cutting guides eliminate the discussion on splint thickness and provide an interesting option to avoid this vertical error [1].

The direction of surgical movement is believed to affect the outcome of maxillary positioning [34]. However, we observed no correlation between the magnitude of the planned maxillary displacement and the positional discrepancies. Instead, the discrepancies in the sagittal, vertical and transverse dimensions correlated with each other. For example, an undercorrection in the posterior maxillary impaction (i.e., an error in the inclination of the occlusal plane) coincided with increased anterior positioning of the upper central incisors. This interaction between the inclination of the occlusal plane and the antero-posterior position of the incisors is based on geometric reasons and is also observed in errors caused by face-bow transfer in CSP [6, 11]. Against our expectations, this positioning error was also observed in the VSP group. Therefore, we assume that this correlation is caused intraoperatively.

The purpose of this study was not to criticize surgery planning in general, but to emphasize the problems facing the surgeon and the orthodontist: to correct a vertical maxillary excess, occlusal canting or increased gingival display by orthognathic surgery is very demanding, and the type of planning method or splint production does not change that. However, a precautionary overcorrection in surgery planning cannot be recommended, as we observed over- and underachievement in all movements.

The applied analyses of accuracy were chosen very strictly to detect even small positioning errors, and, compared to previous studies [16, 18], avoided reporting only the mean discrepancies. By referring to those summary statistics, similar values on either side of zero cancel each other out and mask the true discrepancies. Instead of evaluating the overall maxillary position [17, 18], our measurements focused on the upper central incisors and the inclination of the occlusal plane. Compared to surface to surface measurements, inaccuracies by this method appear larger [22]. As the patient assesses his/her postoperative smile based on the incisor position, it is important to evaluate the success of maxillary positioning at the incisors.

A further strength of this study lies in the fact that the CSP results were digitized and analyzed with the same method as VSP, which was not done in previous studies. The planned and achieved outcomes were superimposed by a voxel-based algorithm and compared in the same coordinate system. The surgery protocol was consistent for all patients: The condyles were repositioned, and the maxillary surgery was performed first. The patients were not operated on by the same surgeon, even though the same surgeon supervised all surgeries. However, the accuracy between different surgeons in different centers obtained similar results when the same stringent surgery protocol was used [17].

It has to be pointed out that it was beyond the scope of this study to compare the whole planning process between VSP and CSP and that we focused on the errors caused by orientation of the maxillary casts and manufacturing of the splints. For example, unsatisfactory results based on undetected facial asymmetries or incorrect cephalometric diagnosis due to missing 3D information were not part of our investigation. Moreover, the accuracy of mandibular positioning was not investigated because the postoperative CBCTs were taken within 2 weeks of surgery. During this early postoperative period, the possibility that swelling and edema would affect the mandibular position is high, and no reliable analyses could be provided with the intended precision. As a 1 mm error in occlusion might cause more clinical problems than a 1 mm malpositioned maxilla [35], further research investigating the accuracy of mandibular positioning is required.

In summary, splint-based VSP and CSP result in clinical acceptable accuracy for maxillary positioning. The advantages of VSP, such as the visualization of the preoperative condylar position or the collision of the proximal and distal mandibular segments, as well as the improved communication possibilities between surgeons and orthodontists, provide arguments for the use of VSP. Nevertheless, it should be kept in mind that vertical positioning is challenging in both methods and VSP combined with splint-less procedures should be driven forward to reduce this error.