INTRODUCTION
Odontoid fractures account for 9%–15% of all spinal fractures, and most patients do not present with neurologic deficits [1,2]. Because of the high instability of this entity, rigid external immobilization may be insufficient for treating type II or rostral shallow type III odontoid fractures [3]. Anterior odontoid screw fixation (AOSF) has advantages over posterior C1–2 fusion in its ability to preserve C1–2 rotational motion, reduce operative morbidity, and avoid the need for bone grafting [4-6].
Despite these advantages, little is known regarding the factors that influence successful fusion in odontoid fracture patients managed with AOSF [7,8]. Our previous studies revealed that a fracture gap greater than 2 mm and delayed operation were risk factors for nonunion [7]. However, similar to our studies, previous research has mainly focused on 2-dimensional (2D) metrics such as the fracture gap and fracture displacement in predicting fusion outcomes. These approaches may overlook the complexity of the fracture geometry and the 3-dimensional (3D) interactions at the fracture site.
In this study, we propose the concept of fracture deficit volume (FDV), a 3D measurement that quantifies the spatial gap between the edges of the fractures. We hypothesize that a reduction in the FDV, which increases the contact area among the fracture edges, will have a positive effect on the success of fusion in patients with AOSF. We aimed to provide a more comprehensive assessment of factors contributing to successful bone fusion after surgery through 3D volumetric analysis.
While traditional risk factor analyses have focused on preoperative parameters, this approach may overlook the impact of surgical technique in modifying certain risk factors. In odontoid fractures, parameters such as fracture gap, fracture displacement, and FDV are not fixed but can be directly altered through surgical techniques. Thus, our study emphasizes the concept of modifiable risk factors, highlighting that intraoperative correction of FDV may have a profound impact on fusion success.
MATERIALS AND METHODS
1. Patient Enrollment
Forty-four patients treated with AOSF at our institute from January 2010 to November 2023 were enrolled in this retrospective study. The study protocol was approved by the Institutional Review Board of Kyungpook National University Hospital (KNUH 2024-12-007). The surgical indications for AOSF were a Grauer type II or a rostral shallow type III odontoid fracture with confirmation of an intact transverse ligament on magnetic resonance imaging [9]. The Grauer classification is a treatment-oriented classification for odontoid fractures. Type IIA fractures are transverse fractures without comminution and a displacement of less than 1 mm. Type IIB fractures are fractures that extend from anterosuperior to posteroinferior or displaced transverse fractures measuring more than 1 mm. Type IIC fractures are fractures that extend from anteroinferior to posterosuperior or those with significant comminution. Patients with disruption of the transverse ligament or less than 12 months of follow-up were excluded.
2. Surgical Technique
We used the standard surgical technique described in detail in our previous studies [7,10,11]. Under general anesthesia, the patient was placed in a supine position. Preoperative positional reduction was applied to achieve correct alignment of the C2 vertebral body and fragment of the odontoid process via 2 C-arm fluoroscopes (open-mouth and lateral views). If correct alignment of the C2 vertebral body and fragment of the odontoid process was not achieved, posterior C1–2 fusion was performed. Therefore, all patients who underwent AOSF had sufficient intraoperative alignment correction to allow an effective screw trajectory.
A transverse skin incision was made at the C3–4 level. After the anteroinferior edge of C2 was exposed, a 2-mm Kirschner wire was inserted from the midline of the anterior-inferior margin of C2 toward the tip of the apical dens. We used a single 4.5-mm diameter Herbert screw, which has a double thread with different pitches on the leading and trailing threads. During insertion of the Herbert screw, the screw angle was drilled toward the middle of the tip of the dens. The fracture gap can be reduced with fenestration of the tip of the apical dens. The surgical outcomes and advantages of AOSF with the Herbert screw have been previously described [11,12].
3. Radiological Measurements
All radiological outcomes were assessed by a single radiologist who was blinded to the study. All fractures were preoperatively assessed with initial open-mouth and lateral x-ray, CT, and magnetic resonance imaging. The 2D fracture gap and fracture displacement were analyzed on CT scans obtained with an INFINITT PACS system (INFINITT healthcare Co., Ltd., Korea). The fracture gap was defined as the maximum distance between the superior and inferior fracture lines on the CT scan. Fracture displacement was defined as the distance of the perpendicular line drawn along the posterior margin of the odontoid fracture. 3D measurements of the FDV were performed with Aquarius iNtuition software ver. 4.4.13 (TeraRecon Inc., USA). FDV was calculated from preoperative CT scans and immediate postoperative CT scans obtained within 48 hours after surgery, enabling assessment of volumetric changes resulting directly from surgical technique. For Grauer type II fractures, the FDV was calculated on the basis of 3D reconstructions aligned with sagittal CT images (Fig. 1A). In contrast, for type III fractures, which include at least one of the superior articular facets of C2, the FDV was calculated using axial CT images as the primary reference (Fig. 1B). This approach is necessary because type III fractures can present with ambiguities on sagittal views, and this orientation provides the clearest view of fracture lines on the C2 vertebra. The FDV was determined by tracing the spatial gap between the edges of the fractures on each slice of the sagittal and axial CT images, followed by 3D reconstruction to quantify the volume (Fig. 1C).
4. Follow-up and Fusion Assessment
Radiological follow-up included cervical spine radiography (open-mouth, lateral, flexion and extension views) at 6 weeks, 3 months, 6 months, and 12 months after surgery. CT scans of the cervical spine were obtained at 6 and 12 months after surgery to assess bone fusion. Follow-up was terminated when the fractures were considered both clinically and radiologically stable.
As we previously described [7], successful fusion was defined by the presence of a bony bridge and definite continuity of cortical bone. Fibrous union was considered present when no degree of motion could be observed on dynamic radiographs, but cortical bone discontinuity persisted within the fracture gap on CT scans. Nonunion was defined as a definite fracture gap with abnormal motion of the fracture on dynamic radiographs and CT scans. Fusion failure was defined as fibrous union or nonunion.
5. Statistical Analysis
Statistical analyses were performed by an independent statistician with R ver. 4.4.1 (R foundation for Statistical Computing, Austria). Continuous variables were compared with the t-test, and categorical variables were assessed with the chi-square test or Fisher exact test. Logistic regression was performed to identify factors associated with fusion failure.
To further explore the relationships between preoperative and postoperative differences in fracture parameters and the likelihood of fusion, we used generalized additive models (GAMs) with a significance threshold of p<0.05. GAMs are a flexible extension of generalized linear models that allow the possibility of determining nonlinear relationships between predictor variables and the response variable [13].
To determine optimal preoperative threshold values for predicting fusion failure, receiver operating characteristic (ROC) curve analyses were conducted for age, preoperative fracture displacement, fracture gap, and FDV. Fusion outcomes were classified as binary variables (successful fusion versus fusion failure). Sensitivity, specificity, accuracy, positive predictive value, negative predictive value (NPV), area under the curve (AUC), and 95% confidence intervals (CIs) were calculated. The optimal threshold for each variable was selected based on the maximum sum of sensitivity and specificity. The predictive performance of preoperative fracture gap and preoperative fracture displacement was compared to the preoperative FDV (reference model). This comparison involved calculating the p-value for AUC difference, integrated discrimination improvement (IDI), and continuous net reclassification improvement (NRI).
RESULTS
1. Patient Demographics
The clinical characteristics of the 44 study subjects are summarized in Table 1. A total of 44 patients (29 males and 15 females) with odontoid fractures were included in this retrospective study. The mean age of the patients was 49.84±22.00 years, and the average follow-up period was 20.23±14.35 months. The most common cause of injury was traffic accidents involving cars, accounting for 47.73% (21 patients), followed by falls, accounting for 22.73% (10 patients). According to the Grauer classification, 9 patients (20.45%) had type IIA fractures, 15 patients (34.09%) had type IIB fractures, 4 patients (9.09%) had type IIC fractures, and 16 patients (36.36%) had rostral shallow type III odontoid fractures. The mean time from injury to surgery was 14.36±28.40 days.
2. Radiologic Outcomes
Among the 2D variables, the mean preoperative fracture displacement was 2.58±1.60 mm, and 56.8% of patients had a displacement of less than 2 mm. Postoperatively, the mean fracture displacement decreased to 1.25±0.84 mm, and 86.4% of patients achieved a displacement below 2 mm. The average preoperative fracture gap was 1.50±0.58 mm, and 72.7% of patients had a gap of less than 2 mm. Postoperatively, the fracture gap remained largely unchanged (1.44±0.79 mm). Three-dimensional analysis revealed a preoperative FDV of 0.25±0.12 cm3, which decreased to 0.21±0.11 cm3 immediate postoperatively (Table 2).
3. Fusion Assessment
In the assessment of continuous variables, the difference in fracture displacement between the preoperative and postoperative measurements was not statistically associated with fusion success (p=0.199). Similarly, the difference in the fracture gap before and after surgery was not significantly associated with fusion success (p=0.160). However, the change in FDV was significantly associated with fusion success (p=0.028). In patients with successful fusion, the FDV decreased by an average of -0.07±0.12 cm3, whereas in patients who experienced fusion failure, the FDV increased by an average of 0.06±0.16 cm3 postoperatively (Table 3).
AOSF resulted in solid bony union in 34 of the 44 patients (77.27%), but fusion failure was observed in 10 patients (22.73%). Further analysis revealed that both preoperative and postoperative fracture gaps of ≥2 mm were significantly associated with higher rates of fusion failure (p=0.002 and p=0.004, respectively). Similarly, a postoperative increase in FDV was significantly associated with a greater risk of nonunion (p=0.006). The incidence of nonunion was significantly greater in individuals aged 65 years or older (p<0.001). Variables such as sex, type of odontoid fractures, time from injury to operation, fracture displacement, and whether the tip of the apical dens was fenestrated were not significantly associated with surgical failure (Table 4).
4. Risk Factors for Fusion Failure After AOSF
The results of logistic regression and estimated odds ratios (ORs) are summarized in Table 5. A preoperative fracture gap ≥2 mm was found to be a significant risk factor for nonunion, with an adjusted OR of 42.73 (95% CI, 1.21–1510.59; p=0.039). Age was also associated with nonunion, with an adjusted OR of 1.20 (95% CI, 1.01–1.42; p=0.038). An increased FDV postoperatively demonstrated significance in crude analysis (OR, 9.00; 95% CI, 1.84–44.02; p=0.007), but was not statistically significant in the adjusted analysis. In contrast, preoperative fracture displacement ≥2 mm was not significant predictor of fusion failure (p=0.345, p=0.827).
5. Generalized Additive Model
The predictive value of the postoperative change in fracture displacement for successful fusion was minimal and was not statistically significant (Fig. 2, adjusted R2=0.0769, p=0.535). Similarly, the predictive value of the postoperative change in the fracture gap was also low and not statistically significant (Fig. 3, adjusted R2=0.0375, p=0.116). However, the postoperative change in the fracture gap showed a linear relationship with the quality of fusion; specifically, the greater the reduction in the fracture gap following surgery was, the better the fusion was. Moreover, a greater reduction in the FDV postoperatively was associated with a higher fusion rate, with the model demonstrating strong statistical significance for this relationship (Fig. 4, adjusted R2=0.186, p=0.018).
6. ROC Curve Analysis and Comparative Analysis of Predictive Models for Fusion Failure
ROC curve analysis revealed that age showed the highest predictive performance with an AUC of 0.909 (95% CI, 0.824– 0.994), using an optimal threshold of 65 years (Fig. 5, Table 6). Preoperative FDV had an AUC of 0.703 (95% CI, 0.495–0.911), with an optimal threshold of 0.254 cm3. Preoperative fracture displacement demonstrated the lowest predictive value with an AUC of 0.591 (95% CI, 0.371–0.811). When preoperative FDV was set as a reference, comparative analyses showed no significant differences in AUC values between preoperative FDV and preoperative fracture gap (p=0.874) or preoperative fracture displacement (p=0.685) (Table 7). IDI and continuous NRI analyses also indicated no statistically significant differences between preoperative FDV and preoperative fracture gap (IDI=-0.014, p=0.847; NRI=0.459, p=0.173). However, preoperative fracture displacement showed a significant difference in continuous NRI compared with FDV (NRI=-0.694, p=0.037), but no significant difference in IDI (IDI=-0.069, p=0.234).
DISCUSSION
Risk factor analysis for fusion outcomes in patients with odontoid fractures treated with AOSF is a long-studied topic [7,8]. The fracture gap is considered a risk factor for nonunion, implying that the surgery should be performed in a way that prioritizes reduction of the fracture gap. Some studies have reported that a fracture displacement ≥5 mm or 6 mm is a risk factor for surgical failure in AOSF [14,15]. However, both our previous study and this study revealed that fracture displacement was not a risk factor for fusion failure. This may be attributed that all patients who proceeded with AOSF underwent careful intraoperative positional reduction under general anesthesia. If satisfactory alignment of the C2 vertebral body and the odontoid process was not achieved despite positional reduction under general anesthesia, we converted to posterior C1–2 fusion [16]. As a result, the cohort included in this study consisted only of patients with adequate alignment at the time of screw insertion, which diminished the impact of fracture displacement of fusion outcome. Furthermore, unsatisfactory alignment between the C2 vertebral body and the odontoid process may reduce the bone-tobone contact area, increasing the likelihood of fusion failure. This surgical approach may introduce a selection bias that potentially improves radiological outcomes of solid fusion. Nonetheless, these exclusion criteria were ethically necessary to avoid procedures that could adversely affect patient safety.
In our clinical experience, even when the amount of preoperative fracture displacement was severe, patients who underwent positional reduction under general anesthesia achieved improved fusion outcomes (Fig. 6). Certain linear fractures may present with a large fracture gap but small FDV. Conversely, certain communited fractures and rostral shallow type III fractures may display a small fracture gap but contain large FDV. These discrepancies underscore the limitations of 2D measurements alone and highlight why volumetric assessments such as FDV provide critical information for surgical planning and predicting fusion outcomes. Traditional 2D metrics, such as fracture gap and displacement, have limitations in capturing the complex spatial interactions at the sites of fracture. In this study, we introduced the FDV as a novel 3D measurement that includes the concepts of both the fracture gap and fracture displacement to assess fusion outcomes in patients undergoing AOSF. To our knowledge, no previous studies have focused on volume-based assessments as predictors of fusion outcomes.
Grauer type IIC odontoid fractures have been considered contraindications for AOSF due to comminution and fracture instability [9]. However, at our institution, fracture morphology alone does not serve as an absolute contraindication for anterior fixation. In our previous study, we demonstrated that AOSF could successfully achieve fusion even in anterior oblique fractures if fracture orientation angles and fragment angulation were favorable and intraoperative alignment was meticulously achieved through positional reduction [10]. In our study, fusion outcomes were not significantly different among fracture types according to Grauer classification (Table 4). Notably, all 4 patients with type IIC fractures achieved solid fusion. The inclusion of type IIC fractures may have introduced heterogeneity and selection bias in our cohort. Nevertheless, our findings emphasize that meticulous preoperative evaluation and intraoperative alignment correction to maximize bone-to-bone contact is the most important aspect of AOSF.
When the fusion rate was analyzed according to categorizing variables, nonunion was significantly more likely in patients over 65 years of age (p<0.001) and was significantly associated with fusion failure after AOSF (OR, 1.2). Fusion failure in elderly patients remains a controversial issue [17,18]. Our previous study revealed that patient age was not a risk factor for fusion failure, so we proposed that AOSF could be performed in selected elderly patients after considering their general condition and bone quality [7]. However, in the current study, 62.50% of elderly patients aged 65 years or older had nonunion. Among the ROC curve analysis, age emerged as a powerful predictor of fusion failure, displaying a high AUC value of 0.909 with an optimal threshold of 65 years (Fig. 5). Therefore, AOSF may not be suitable for patients aged 65 years or older.
Although whether the tip of the apical dens was fenestrated was not significantly associated with the successful fusion rate (p=0.074), the rate of nonunion was nevertheless greater in the group in which the apical dens tip was not fenestrated (36.8%) than in the group in which the tip was fenestrated (12.0%). Because the tip of the apical dens is close to the ventral dura of the spinal cord, fenestration of the apical dens tip is a delicate procedure from the surgeon’s perspective. Our previous study reported that the safe margin beyond the apical dens tip to the ventral dura was greater than the safe margin beyond the posterior end of the dens tip to the ventral dura [19]. If the trajectory of the AOSF is targeted toward the apical dens tip, the fenestration can be safely extended several millimeters beyond the dens tip. In addition, the fracture gap can be effectively reduced when the apical dens tip is fenestrated while a cannulated lag screw or Herbert screw is inserted [20]. As previously mentioned, effective reduction of fracture displacement can be achieved through meticulous alignment of the C2 vertebral body and odontoid process. Additionally, fenestration of the apical dens tip using a Herbert screw reduces the fracture gap. These 2 techniques may enhance bone-to-bone contact, thereby significantly reducing the FDV.
Factors such as sex, patient age, and injury-to-operation interval are nonmodifiable risk factors [20]. However, other risk factors, including fracture gap, fracture displacement and FDV, can be modified by selecting appropriate surgical techniques. Therefore, it is important for surgeons to perform the operation in a manner that seeks to reduce these modifiable risk factors and increase the bone-to-bone contact area. In our study, the modifiable risk factors included the fracture gap and FDV. Importantly, our study highlights that some risk factors for fusion failure are not fixed characteristics, but modifiable through appropriate surgical technique. By using immediate postoperative CT scans, we confirmed the actual geometric correction achieved intraoperatively. If the fracture gap and FDV can be reduced by the surgeon, the rate of successful fusion can be improved.
We predicted the possibility of fusion on the basis of the postoperative changes in fracture displacement, fracture gap, and FDV using a nonlinear model called the GAM [21-23]. The GAM is expressed as a sum of several functions, each of which represents the result of the nonlinear transformation of an independent variable. The results of the analysis of the GAM can subsequently explain fusion as a function of the postoperative changes in the values of different variables (fracture displacement, fracture gap, and FDV) via binary logistic regression. According to the GAM, as the postoperative change in the FDV decreases, the rate of successful fusion increases linearly and there is no clear threshold value for FDV (Fig. 4). Moreover, the fit to a linear model was statistically significant (p=0.018), suggesting that the GAM well explained the relationship between the variables and is reliable (adjusted R2=0.186). The analysis supports the importance of reducing the FDV as a modifiable risk factor during surgery to increase successful fusion.
However, it is equally important to analyze risk factors for fusion failure using only preoperative parameters rather than postoperative changes. Identifying a specific threshold for preoperative variables can guide surgeons in deciding between AOSF and alternative surgical techniques such as posterior C1–2 fusion. As demonstrated in Tables 4 and 5, age, fracture gap, and FDV were significant predictors of fusion failure. Hence, we conducted ROC curve analyses to establish clinically relevant thresholds. Age was the most powerful variable in predicting fusion failure in ROC curve analysis, but it is a nonmodifiable risk factor (Table 6). Analysis of modifiable risk factors aimed to evaluate their predictive capabilities and to suggest practical thresholds for clinical decision-making. Preoperative FDV and preoperative fracture gap demonstrated high AUC value and high NPV, suggesting their utility in preoperative assessment and surgical planning. In contrast, preoperative fracture displacement alone showed limited predictive accuracy and should be combined with other variables rather than used independently.
Comparative analysis using preoperative FDV as a reference indicated that FDV was the strongest predictor of fusion failure among modifiable risk factors (Table 7). Although the predictive accuracy of FDV was slightly superior to that of the preoperative fracture gap, the difference was not statistically significant. In contrast, preoperative fracture displacement showed statistically significant lower predictive capability in the continuous NRI analysis. Therefore, preoperative FDV and preoperative fracture gap can both be considered clinically valuable predictors for fusion failure. While fracture gap is already a widely acknowledged risk factor for AOSF, our study introduces FDV as a novel 3D measurement, emphasizing the need for further research incorporating this 3D volumetric parameter.
In fact, the precise FDV at the time of fixation may differ from preoperative values due to intraoperative positional reduction. Ideally, intraoperative CT imaging immediately prior to screw fixation would offer the most accurate assessment of FDV during surgery. Despite this limitation, our findings highlight the clinical importance of meticulous intraoperative alignment of the C2 vertebral body and odontoid process, combined with careful fenestration of the apical dens tip, to effectively reduce FDV. Future studies incorporating intraoperative imaging may further clarify the relationship between immediate prefixation FDV and fusion success.
Our analysis of continuous variables revealed that the change in the FDV from the preoperative to postoperative values was the only variable that was significantly correlated with the fusion rate (Table 3) (p=0.028). Although the FDV was not the only variable to achieve statistical significance in the categorical analysis of the fusion rate, the group with an increased FDV postoperatively had a greater rate of fusion failure (Table 4) (p=0.006). Additionally, GAM analysis revealed that the FDV was the only variable that demonstrated a significant association with the fusion rate. Given these results, we conclude that FDV has the greatest impact among modifiable risk factors on fusion success in patients with odontoid fractures who undergo AOSF.
There are several limitations of the present study. First, this retrospective study was performed at a single institution, which introduces selection bias and limits the external validity and generalizability of our findings. To overcome these limitations, multi-center prospective studies are required to further validate the clinical utility of FDV. Second, this study had a relatively small sample size (n=44). Therefore, the statistical power of our findings may be limited. Future studies with larger cohorts are needed to confirm the validity and reliability of FDV as a predictor of fusion outcomes. Third, owing to limitations in the imaging protocol, we were unable to use thin-section analysis during 3D reconstruction, which may have introduced measurement errors in the 3D measurement. Future prospective studies are needed to incorporate thin-section CT scans and achieve more accurate 3D measurements. Fourth, all radiological measurements were performed by a single radiologist. Therefore, inter-observer reproducibility could not be evaluated. To minimize potential measurement errors from single observer variability, each radiological measurement was performed 3 times independently by the same radiologist, and the average of these 3 measurements was utilized for analysis. Nevertheless, future studies should incorporate assessments of inter- and intraobserver reproducibility to validate reliability. Fifth, the inclusion of Grauer type IIC fractures and exclusion of patients with unsatisfactory alignment may have introduced selection bias. To address this limitation, future studies should consider performing subgroup analyses based on fracture type and alignment status.
CONCLUSION
Although the optimal treatment for Grauer type II and type III fractures remains controversial, we achieved successful bony fusion using AOSF in 77.3% of patients in this retrospective study. Our analysis revealed that the risk factors for fusion failure include older age (≥65 years), a fracture gap greater than 2 mm, and increased FDV postoperatively. Given these findings, alternative surgical techniques such as posterior C1–2 fusion should be considered for patients aged 65 years or older. Among the modifiable risk factors, FDV had the greatest impact on fusion success when performing AOSF in patients with odontoid fracture.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Full text via PMC
Download Citation
Print
