The impact of orthodontic-surgical treatment on facial expressions—a four-dimensional clinical trial

This study was a prospective, single-centre, single-arm trial and investigated facial expressions before (T0) and 4 months after orthognathic surgery (T1). It was approved by the institutional ethics committee (application number 13/2/19) in accordance with the Declaration of Helsinki. All participants gave written informed consent to take part in the study. The trial was registered before recruitment started (DRKS00017206). No changes occurred after trial commencement.

Patients seeking orthodontic-surgical treatment with an orthodontics first approach at the Department of Orthodontics, University Medical Center Goettingen, Germany, were enrolled consecutively. The eligibility criteria for participants were adults over 18 years old, orthodontic therapy with fixed appliance, pronounced malocclusion (skeletal class II: Wits appraisal > 2 mm; skeletal class III: Wits-appraisal <  − 2 mm), and indication for combined orthodontic-surgical treatment (classified as grade 4 or higher using the index of orthognathic functional treatment need [19]). Patients with cleft lip and/or palate, craniofacial syndromes, impaired facial motion, previous orthognathic surgery, Menton deviation > 4 mm, or beard were excluded.

As this trial is an early clinical trial with the aim to implement the novel method of 4D videostereophotogrammetry, sample size calculation was not based on a priori-hypothesis testing. Based on previous treatment numbers, in total 45 patients (1 skeletal class I, 21 skeletal class II, 23 skeletal class III) were operated in an interval similar to the scheduled recruitment phase of this study; a feasible sample size of 20 patients was planned for recruitment. The final sample size was 26 patients (13 skeletal class II, 13 skeletal class III) and a complete case analysis was performed.

All patients underwent virtually planned, splint-based orthognathic surgery performed by an interdisciplinary team of orthodontists and maxillofacial surgeons as described previously [20]. The surgical procedure included the preservation of a pre-surgical defined condylar position using a centric splint [21‐23]. In bimaxillary surgeries, Le Fort I osteotomy was performed first, followed by bilateral sagittal split osteotomy according to well-recognised protocols [24‐27]. All patients were invited to record their facial expressions during smiling and lip purse at baseline (T0, after orthodontic decompensation, 1 to 6 weeks before surgery) and 4 months post-surgical (T1).

Blinding was not applicable.

Facial expressions were recorded as 3D videos using non-invasive 4D stereophotogrammetry with nine cameras (6 greyscaled and 3 coloured) and 60 frames per seconds (see Fig. 1; DI4D PRO System, Dimensional Imaging Ltd., Glasgow, UK). Before each capture, the system was calibrated according to a standardised protocol. Each patient was asked to sit upright approximately 95 cm in front of the cameras, to keep the eyes open, to have the head in natural head position and to move the head as little as possible. Then, the patients performed two reproducible facial expressions: (1) maximum smile and (2) lip purse. Both were demonstrated to the patients and practiced several times prior video acquisition.

All video captures and post-processing were performed by the same trained investigator (J.H.). The 3D videos were cut to analyse the movements from the last frame in rest position (= start frame) to the first frame with the facial expression at its maximum extent (= end frame) (DI4D Setup, Dimensional Imaging Ltd., Glasgow, UK). To track the facial movement, a dummy mesh was applied to the individual patient morphology in the first frame of each video (see Fig. 2). This patient characteristic mesh was automatically tracked throughout the sequence of 3D images and manually corrected if necessary (see Fig. 3; DI4D track, Dimensional Imaging Ltd., Glasgow, UK). The Euclidean distance d between the points of the start frame s and the end frame e in the 3D space were calculated using Python and a self-edited SciPy based code [28].

Two regions of interest were determined (see Fig. 4):

For each facial expression, four parameters were calculated to objectively evaluate the movement (see Table 1 and Fig. 5).

To assess the reliability of the 4D tracking of facial expressions, the same investigator repeated the application of the dummy mesh to the individual patients’ morphology in ten cases. Additionally, a second examiner performed the same task one time. Agreement was excellent for the magnitude of facial expression (ICC = 0.982; 95% CI [0.933, 0.995]) and good to excellent for the shape change in the region of interest lower face (ICC = 0.947; 95% CI [0.814, 0.986]).

The analysis was performed using linear mixed-effects models with factors time, skeletal class, and their interaction. A sensitivity analysis using a paired Wilcoxon-signed rank test was performed. Estimated marginal means are reported with 95% confidence interval and compared between time and skeletal class [29]. Pre- and postoperative differences of 4D tracking variables were correlated within a multivariate linear model. Intra-class-correlation (ICC) was calculated to assess agreement between raters and consistency within raters. Baseline data are described using means and standard deviation. Statistical analysis was done using the statistical programming language R 4.1.3 [30].

Recruitment lasted from September 2020 until August 2021. The last follow-up was completed in December 2021. Of a total of 39 consecutive patients assessed for eligibility, 31 fulfilled the inclusion criteria (see CONSORT flow diagram in Fig. 6). During the observation, 1 patient declined surgery, 1 patient missed his follow-up appointment, 2 patients had surgical complications, and 1 patient was still exceptionally swollen at T1. The final sample consisted of 26 participants. Table 2 shows their baseline demographic data and clinical characteristics.

4D motion capture was feasible using stereophotogrammetry and resulted in 3D videos with a rate of 60 frames per second. Colour and texture were recorded as well as surface geometry (see Supplementary Information 1 for a video example). All movements could be tracked from the start frame (rest position) to the end frame (maximum movement) semi-automatically.

For maximum smile the mean magnitude of movement increased significantly from T0 until T1 by 2 mm. At T1, smiling became more asymmetric. No differences in shape change or time was observed. The mean magnitude of lip purse decreased from T0 until T1 by 1 mm. Shape change, symmetry, and time for the facial expression did not differ significantly between T0 and T1 (Table 3).

The post-surgical increase in the magnitude of maximum smile and the decrease in the magnitude of lip purse were observed in all but 5 and 6 patients, respectively.

Subgroup analysis showed that these results apply to patients with skeletal classes II and III (Table 4).

The results of the multivariate linear model with the difference in the magnitude of maximum smile as dependent variable showed that the pre- and postsurgical differences in the magnitude of smile cannot be explained by the surgical movements, the type of surgical intervention, or the skeletal class (Table 5). The multivariate linear model for the magnitude of lip purse indicated the same—the postsurgical change in the movement of the lips showed no correlation with the surgical intervention or the skeletal class (Table 6).

No harms were observed.

The main strength of this study was the implementation of modern videostereophotogrammetry in orthodontics and its use to capture facial movements in 4D. In contrast to previous attempts analysing facial expressions using markers on the patients’ face [4, 6, 7], static recordings of the movement at its maximal extent [31, 32], two-dimensional videography [33, 34], or the use of the facial acting coding system [8, 35], 4D stereophotogrammetry offers an objective and reliable method to measure facial movements. Static recordings at rest position and maximum extent bear the risk of unnatural freezing of the facial expression. Compared to 3D measurements, 2D analyses using videos underestimate the amplitude of facial motion by as much as 43% [36]. This discrepancy is particularly pronounced in the area of the lower face, which is usually in the focus of orthodontics. Marker-based tracking systems are criticised because the direct positioning of the markers on the patient’s face shows large variances and may affect the natural facial expressions [37]. Videostereophotogrammetry proved to be a feasible objective tool for assessing the impact of surgical interventions on facial movements [15], and demonstrated good to excellent reliability within the sample of this study.

A current disadvantage of videostereophotogrammetry is that the necessary cameras, lightning systems, computers, and software are expensive and bulky. However, portable low-cost 4D cameras already exist and make their way in clinical application [38, 39]. Big companies like Microsoft are working on these technologies trying to facilitate automatic landmark tracking, which will not only reduce costs of videostereophotogrammetry but also emphasise the impact of 4D measurements in the near future [40].

The main finding was that orthodontic-surgical treatment increases the magnitude of the facial expression maximum smile by approximately 13% (= 2 mm), while it decreases that of the facial expression lip purse by approximately 11% (= 1 mm). The parameters shape change, time, and symmetry of the movement demonstrated no relevant changes from T0 to T1. The novelty of videostereophotogrammetry makes it difficult to compare these results with existing evidence. There are first attempts to integrate the dynamics of facial motion in orthodontic-surgical research [6, 8, 17, 33]. However, the interpretation of the data is limited by small study samples, methodological limitations, and divergent results. In 13 patients with maxillary hypoplasia and Le Fort I osteotomy, Al-Hiyali and co-workers reported a reduction in the magnitude of facial expressions of approximately 23% [17]. Since the authors summarised the data for three facial expressions (maximum smile, lip purse, and cheek puff), it remains unclear whether the change in the magnitude of movement differed between maximum smile and lip purse like observed in the current sample. Furthermore, in accordance with our observations, they demonstrated no clinically relevant difference in the asymmetry score for patients without facial asymmetry. For three patients with skeletal class III, Nooreyazdan and co-workers reported greater upward movement for cheilion left and right during smiling and reduced lip movement during lip purse [6], which is in agreement with the present results. Johns and co-workers, who also observed an increase in facial movement while smiling in patients with maxillary advancement, postulated that the anterior movement of the maxilla lengthens the facial muscles resulting in increased mobility [33]. This might only explain the increased magnitude in maximum smile in part as we found no interaction between the surgical movements and the parameters of facial expressions. Maybe, the explanation for the increased smile is as simple as the patients prefer smiling and displaying their teeth after surgical correction of the dentofacial deformity more than before. Nevertheless, it can be speculated that the surgical change in the underlying skeletal support affects the mobility of the overlying soft tissues. How good this adaption succeeds and whether the dynamics of facial soft tissues have an impact on skeletal relapse remain to be elucidated and is an aspect for future studies using the novel method of 4D motion capture. However, no significant impact of surgical movements, type of surgical intervention, or skeletal morphology on soft tissue dynamics was observed in this preliminary study. This may be due to the small sample sizes, implying a comparably low statistical power, but indicates that the postsurgical adaptation is highly individual and prediction of postsurgical facial movements is currently not possible. In contrast, Nooreyazdan and co-workers found presurgical differences between patients with skeletal class II, open bite, and class III for the movement lip purse, while there were no differences for the movements smile, mouth opening, cheek puff, eye opening, eye closure, or grimace even though their sample size was even smaller than in the present study [6].

Furthermore, the question arises whether 1 or 2 mm (11–13%) changes in the magnitude of the facial expressions, like found in the present study, are clinically relevant. With regard to social interaction the interpretation of the data gets even more complex as culture and gender shapes our expectations about the intensity of facial expressions in the sense of emotions [41, 42]. In contrast to orthopaedics, where gait measures start to establish as potential disease biomarkers [43, 44], no references for an ideal facial movement exist in orthodontics. Therefore, efforts should be made to collect movement data in class I subjects and the integration of facial dynamics in diagnostics should be driven forward to achieve patient-centred comprehensive treatment plans. Until then, differences of 1–2 mm should be judged as relevant because studies assessing the activity of other muscles of the skull, e.g. the dimension of masticatory muscles or jaw movement, often show pre-/postsurgical differences of 1–2 mm [45, 46].

The participants of the study were not operated on by the same surgeon, but all surgeries were performed at a single centre following the same stringent protocol. Since a multi-centre study revealed that even different surgeons at different centres obtain similar accuracy when following a virtual treatment plan [47], it can be assumed that the surgical results within this study were reproducible and representative for orthodontic-surgical interventions.

It is possible that the patients’ compliance, the intensity of care provided by the orthodontic-surgical treatment team, and subsequently the speed of post-surgical recovery were influenced by the fact that the patients were part of a clinical trial. Because of that Hawthorne effect, the participants in the present study might show increased facial expressiveness.

The observed changes in maximum smile and lip purse were rather small. Due to the novelty of the method, there are no reference values which indicate what amount of change should be considered clinical relevant and what extent of change is detected by orthodontists, surgeons, laypersons, or the patients themselves. Future studies are needed to generate norm values.

The results of this trial were from a single specialist clinic in Germany, which might reduce the applicability to other clinical settings. However, the eligibility criteria include most patients undergoing orthodontic-surgical treatment and the followed surgery protocol is widely used and well established.

This study did not have the purpose to investigate gender differences in orthodontic surgical treatment and gender distribution within the study sample was not equal. This reflects the real clinical situation since women are more likely to accept surgical intervention than men [48‐50]. The results may be generalised to a similar population with indication for orthodontic-surgical treatment and similar gender distribution.

The focus of this study was the integration of videostereophotogrammetry in orthodontic-surgical treatment. To fully understand the effect of orthodontic-surgical treatment on facial expressions, long-term observations and extended study periods are needed. Baseline assessment was performed after orthodontic decompensation with fixed appliances in situ, which might have affected facial movements. Future studies should consider recruitment of patients with severe malocclusions prior orthodontic treatment. The follow-up of 4 months post-surgical in this study was chosen because the swelling has usually subsided by that time [51, 52] and the fixed appliance is still in place. This allowed a comparison of the pre- and post-surgical data without considering an influence of the orthodontic appliance. Since we expected an average post-surgical orthodontic treatment time of 5 to 6 months [53, 54], T1 was set at 4 months post-surgical to reduce dropouts. However, studies on changes of masticatory performance following orthodontic-surgical treatment indicate that an observation period of up to 5 years is necessary to show all effects induced by the intervention since the muscles of mastication need time to regain full strength [55, 56]. Whether this also applies to the facial muscles has not yet been investigated and further studies with an extended follow-up period are required.