Injury to the inferior alveolar nerve (IAN) during surgery is an important complication of bilateral sagittal split osteotomy. With cone beam computed tomography, the course of the nerve and its relationship to the surrounding structures can be assessed in three dimensions. This study aims to determine whether tomography can predict attachment of the neurovascular bundle to the proximal segment of the mandible during sagittal split osteotomy (SSO). Bilateral linear measurements were taken on cross-sectional tomography images. During osteotomy, it was noted for each patient whether the neurovascular bundle was attached to the proximal segment during the split. If attached, a bone-cutting instrument or a blunt instrument was needed to free the nerve. The nerve was attached at more than one-third of operation sites (170 sites). Of these, over 65% of attached nerves (108 sites) required a bone-cutting intervention to free them from the mandible. After correcting for confounding factors, the linear distances from the buccal cortical margin of the IAN canal to the inner and outer buccal cortical margins of the mandible were important predictors of whether the IAN will be attached to the proximal segment of the mandible during SSO.
Sagittal split osteotomy (SSO) is a basic orthognathic surgical procedure. Different complications have been reported associated with bilateral sagittal split osteotomy (BSSO). Of importance among these complications is injury to the inferior alveolar nerve (IAN). IAN injury during surgery largely results from manipulation of the nerve or structures surrounding the nerve, or from direct injury to the nerve during the operation.
The IAN can be endangered at several levels during the surgical procedure. Factors that can cause damage to the IAN include placement of the retractor posterior to or above the lingula at the ascending ramus, horizontal osteotomy at the ascending ramus, the bone cut at the body of the mandible, the bone cut at the lower border of the mandible, the connecting bone cut between the lower border and the buccal osteotomy of the mandibular body, chiselling during the sagittal split, possible impingement of nerve between bone fragments after the bony movement, damage during placement of the osteosynthesis material, and length of the advancement of the lower jaw. The degree of nerve manipulation during the procedure is also a known risk factor. Whenever the neurovascular bundle or neurovascular canal is attached to the proximal segment, careful dissection is mandatory. It is thought that the neurovascular canal is often attached to the proximal segment in asymmetric mandibles (at the excessive side) and in the presence of an unerupted third molar.
Cone beam computed tomography (CBCT) offers the surgeon the opportunity to locate the neurovascular bundle in three dimensions, allowing for individual modification of the approach of the lower border cut depending on the distance from the neurovascular bundle to the lower border and the buccal plate.
Detachment of the IAN canal from the proximal segment of the mandible is a delicate manoeuvre that frequently involves manipulation and pressure on the neurovascular bundle. In these instances, the IAN is more prone to injury during the SSO surgery and as a result, neurosensory disturbances of the IAN may occur in these patients. Panoramic radiography does not allow the clinician to predict whether the IAN is close to the lingual or to the buccal cortex (BC). With CBCT imaging, the course of the IAN and its relationship to surrounding vital structures can be readily observed ( Fig. 1 a and b ).
Before CBCT was available, surgeons had only an estimate of the vertical position of the mandibular canal with respect to the inferior border of the mandible. CBCT offers the possibility of tracing the course of the IAN from images acquired before operation. These images provide vital information about structures in and around the operating site. As such, the surgeons are able to assess the position of the nerve relative to the inferior border and buccal plate of the mandible ( Fig. 2 ). It is thought that the shorter the distance from the buccal aspect of the IAN canal to the outer buccal cortical margin and the closer the canal to the inferior border of the mandible, the higher the probability of injuring the nerve during the split. This assessment involuntarily creates an expectation of the nerve position during the split. Sometimes this anticipation is not met by the actual findings during surgery.
If CBCT were able to show the exact location of the nerve adequately, and to predict attachment of the nerve to the proximal segment during the procedure, it might be possible to modify the technique or to select a different surgical technique as an alternative in high-risk procedures (e.g. when the neurovascular canal resides in the BC). The question investigated in the present study is whether CBCT is able to predict the attachment of the neurovascular bundle to the proximal segment of the mandible during SSO, and to identify other predictive factors.
Materials and methods
This is a prospective study including 220 consecutive patients who underwent BSSO between January 2010 and October 2011.
During the preoperative examination, CBCT images were taken of all patients. CBCT images were acquired using the Galileos CBCT scanner (Sirona Dental Systems, Bensheim, Germany). Each patient was positioned with the median sagittal plane perpendicular to the horizontal plane, as recommended by the scanner patient positioning protocol reference manual. The vertical laser-positioning guide was used to guide the proper orientation and positioning of each patient’s head. Scans were made at 42 mA s and 85 kV with a scan time of 14 s. On completion, panoramic views were automatically generated and presented using Galaxis software.
Assessment of IAN canal position on CBCT images
The positions of the IAN canal from the mandibular second molar region to the mandibular foramen in the front part of the mandible were identified on the CBCT images of the 220 patients. On the cross-sectional CBCT images, linear measurements were taken at the level of the first and second mandibular molar on the right and left sides by one of the authors who was blinded for the perioperative nerve position. Two measurements were taken at a distance of 3 mm apart on each side ( Fig. 3 ).
The following linear distances were measured automatically on the EIZO monitor (Eizo Nanao Corporation, Shikawa, Japan) using software (Sidexis, Sirona, Germany) connected to the CBCT scanning system: the linear distance between the alveolar crest and the lower border of the mandible; the linear distance from the buccal cortical margin of the IAN canal to the inner buccal cortical margin of the mandible (nerve to inner BC); the linear distance from the buccal cortical margin of the IAN canal to the outer buccal cortical margin of the mandible (nerve to outer BC); the linear distance from the inferior aspect of the IAN canal to the inner inferior border of the mandible (nerve to inner inferior cortex (IC)); the linear distance from the inferior aspect of the IAN canal to the outer inferior border of the mandible (nerve to outer IC); the width of the mandible, measured from the outer lingual cortical margin to the outer buccal cortical margin; the width of the IAN canal, measured from the lingual cortical margin of the canal to the buccal cortical margin of the canal ( Fig. 4 ).
Clinical assessment of nerve attachment
All procedures were conducted by the same surgeon, using the same approach. The surgeon has carried out more than 1000 BSSO procedures. All patients were treated for orthognathic reasons. In all patients, the unerupted lower third molars, if present, were removed at least 6 months preoperatively. The SSO surgical technique has been described in detail by Falter et al.
A panoramic image of the patient with the position of the IAN annotated on it was used during the operation; but the surgeon was not aware of the preoperative measurements from the scan. The buccal osteotomy of the SSO was started at the lower border and includes part of the lingual cortex, as described by Reyneke. Drilling was stopped as soon as bleeding points were observed during the cut of the lower border. Next, the vertical part of the mandibular body osteotomy was made, connecting the lower border cut to the buccal osteotomy of the mandibular body. The second stage of the split was carried out using unsharp wedge osteotomes. If the neurovascular bundle remained attached to the proximal segment of the mandible during the continuation of split, the split was stopped and the bundle was carefully dissected from the proximal segment, either using blunt instruments (freer, nerve hook) or surgically using an osteotome, a drill, or a piezzo ultrasonic device (Mectron s.p.a., Bois d’Amont, France). A freer or a nerve hook-type instrument was used whenever the nerve was not entirely surrounded by bone. Whenever the nerve seemed entirely surrounded by bone, the surgeon had the option to chisel, drill, or use the piezzo surgical unit to remove the bony resistance, followed by a freer or a nerve hook-like instrument to liberate the nerve further from its canal.
For each patient, it was noted whether the neurovascular bundle was attached to the proximal segment during the split. If attached, the method required to free it was also noted (bone-cutting or blunt technique).
Thereafter, the relationship between measured distances on CBCT images and attachment of the neurovascular bundle to the proximal segment of the mandible was assessed statistically. Sex, age, and site were included as variables. The type of movement (advancement, set-back, rotation) was also assessed, to account for the variable of mandibular asymmetry.
Descriptive statistics were used to describe patient and operation characteristics. To reduce the number of measurements on each side, the minimum of the different distances on the two slides (mandibular height, mandibular width, nerve to inner BC, nerve to outer BC, nerve to inner IC, nerve to outer IC, canal width) were measured.
The predictive ability of the different distances was assessed using a logistic generalized estimating equations (GEE) model that took into account the correlation between the outcomes on both sides from each patient. In addition to the seven measurements, the following confounding factors were added to the model: sex, age, type of movement (advancement, set-back and rotation), and side. To reduce multicollinearity in the multiple regression analysis, four new distances were created. Mandibular width was replaced by the distance between the nerve and the left side of the mandible, and is referred to as ‘nerve to left’. Mandibular height was replaced by the distance between the nerve and the upper side of the mandible, and is referred to as ‘nerve to upper’. The distance from the nerve to the outer BC was replaced by the distance from the nerve to the outer BC minus the distance from the nerve to the inner BC, and is referred to as ‘BC thickness’. The distance from the nerve to the outer IC was replaced by the distance from the nerve to the outer IC minus the distance from the nerve to the inner IC, and is referred to as ‘IC thickness’.
To select the variables for the multivariable logistic GEE model, an automatic backward selection strategy using a standard logistic regression was applied to 1000 bootstrap samples using side as a sampling unit and with a reduced sample size of 256 (instead of 440) to compensate for the design effect. This effect was calculated as 1 + (2 − 1) × Kappa = 1.72. As a standard logistic regression, which was chosen for computational reasons, does not take into account the correlation between the outcomes on both sides, a reduced number of sides must be analysed to avoid selecting some predictors too often. Following Heymans et al., several strategies were applied as a sensitivity analysis. A P -value to stay of 0.5, 0.175, and 0.05 with a selection level in the 1000 samples of 90%, 80%, and 70% was used, respectively. Variables that were selected in all three of these analyses were chosen to be included in the multivariable logistic GEE model. Linearity of the predictors and second order interactions between the different distances were verified. The confounding factors were kept fixed in the model.
The predictive ability of the different measurements is represented by the C -index. To verify the additional value of a multiple regression model on the discriminating power to classify whether or not the nerve is free, the integrated discrimination improvement (IDI) statistic was calculated. The IDI is a measurement that takes into account changes in sensitivity and specificity, particularly how much is gained in mean sensitivity without sacrificing mean specificity. Both absolute and relative IDI were calculated. Finally, a receiver operating characteristic (ROC) curve was produced. All analyses were performed using SAS version 9.22.