Introduction

Posterior stabilization of the cervical spine (CS) is increasingly carried out by placing pedicle screws in the respective segment [4, 6, 31, 34].

Better primary stability and repositioning accuracy can be achieved in comparison to lateral mass screws, thereby reducing the number of fused segments [5, 12, 27]. Therefore, pedicle screws are particularly applied in cases of long-distance fusion. Recent studies suggest superior precision of navigated pedicle screws over screws placed under fluoroscopic control in the cervical spine [15, 21, 31]. It is assumed that, especially in less experienced hands, screw placement precision can be significantly increased using navigation [23, 25, 26, 31, 36].

For this reason, navigated posterior instrumentation of the cervical spine has become the gold standard in most clinics. Benefits of this technique include lower radiation exposure for the surgical team as well as for the patient and, above all, a lower risk of screw displacement. This is particularly important due to the potentially serious consequences that can result from the anatomical proximity to the vertebral artery as well as the cervical spinal cord [3, 25, 31, 32].

Therefore, in contrast to lateral mass screws in the upper and subaxial CS, or transarticular Magerl screw fixation, which can both be safely done without navigation, transpedicular instrumentation of the subaxial CS is currently completely covered by the navigated surgical technique [21, 25, 26, 30].

There are several navigation procedures. Most frequently, CT-based or 3D fluoroscopy navigation are used. Both methods have advantages and disadvantages [9, 13, 14, 19, 21, 22, 36]. One advantage of CT-based navigation is better image quality. However, data sets recorded pre-surgically with supine positioning of the patient can differ from images performed during surgery with the patient lying prone, and this can make vertebral height localization difficult. This is why each vertebral body must be verified before instrumentation in a rather time-consuming way by palpation of reference points on the bone surface. 3D fluoroscopy navigation methods combine the advantages of the ability to produce a reference data set after patient positioning on the operating table, and again, if necessary, after reduction of a deformity, very good height localization, and relatively low radiation exposure. A clear drawback is the worse image quality compared to CT, especially at the cervico-thoracic junction.

The purpose of the present study is to investigate the positional accuracy of pedicle screws placed in the cervical spine using 3D fluoroscopy navigation.

Materials and methods

Patient population and data collection

This is a single-center analysis performed in the orthopedic and trauma surgery department of a university hospital. Data were obtained from 64 patients who underwent axial and subaxial screw insertion surgery to the cervical spine between 2007 and 2012. A total of 207 screws were implanted from levels C1 to C7. Prior to the planned posterior stabilization of the cervical spine, CT angiography was routinely performed to demonstrate the exact course of the vertebral artery. With only one exception, the operation was carried out by a team containing two surgeons.

After general anesthesia with muscle relaxation, the patient was placed in a prone position; the head was fixed with a carbon Mayfield clamp, arms pulled inferiorly with strips of adhesive bandage (Fig. 4).

In order to extend the upper body as far as possible beyond the operating table to avoid the interference of metal braces with X-ray imaging, a radiolucent board with a gel cushion was used for positioning the chest. When necessary, closed reduction of displacement was performed under fluoroscopic control in two planes. The posterior cervical spine was approached through a median incision above the spinous process line. After exposure of the operative field, reference reflectors were attached to a spinous process. For vertebral bodies C1–C4, the spinous process of C2 was used, and for C5–C7, the spinous process of C7. The reference terminal with three infrared light-reflecting diodes was attached either via intraosseous pin or via terminal to the spinous process, depending on bone quality. C-arm-based 3D imaging (3D ArcadisOrbic, Siemens, Germany) was used for navigation. During a 180° orbital rotation, a total of 100 radiographs were made and upsampled to a 3D data set. Once the instruments, also equipped with infrared light-reflecting diodes, were detected, this data could be used for navigated spine surgery using VectorVisionfluoro 3D Trauma Software (BrainLAB Inc., Heimstetten, Germany). To this point, the X-ray image intensifier was used only for height control in the lateral beam projection. Once navigational accuracy was verified, the pedicle entry point was approached with a 1.8-mm drill sleeve (Fig. 7). Because the courses of the C3–C6 pedicles converge, separate auxiliary incisions were made for each before advancement of a blunt trocar until bone contact. After transpedicular placement of 1.8 mm K-wires, a 180° scan was performed without navigation to visualize the intraosseous course of the K-wires. The thread for the definitive pedicle screw was then cut over these K-wires with a cannulated screw tap. Two different fixation systems were applied, the Neon system (Ulrich Medical Inc., Ulm, Germany) and the VertexMax system (Medtronic Inc., Basel, Switzerland). The latter offers the advantage of being compatible with thoracic systems implanted with a minimally invasive technique, allowing for long segment cervico-thoracic stabilization. The disadvantage is that an additional screw–rod connection for extensors between longitudinal rod and pedicle screw is often necessary because of the far lateral entry point, reducing the stability of the construct. In addition, the cervical longitudinal bars in this system are comparatively thin.

Postoperatively, multiplanar CT scans were performed and then assessed by both a spine surgeon and an experienced radiologist. Position accuracy of screw placement was evaluated according to the modified classification by Gertzbein and Robbins [10] consisting of 5 grades. Grade 1 describes ideal screw position with pedicle wall perforation of <1 mm (Fig. 1), grade 2 with pedicle wall perforation of <2 mm, grade 3 of <3 mm, and grade 4 of <4 mm. Grade 5 (Fig. 2) represents a cortical breach of >4 mm and/or obstruction of the transverse foramen by more than half a screw diameter. Secure anchoring of the Magerl screws within the C1 lateral mass was additionally evaluated. Analysis of intra- and postoperative surgical complications was based on separate documentation. In the study, The Magerl screw had to be changed because of uncorrect placement (Table 3), so revision was performed via open reduction of the disposition by dismantling the longitudinal rods, removing the inserted Magerl screws, renewing the inclining reposition in the Mayfield clamp (for enhanced screw positioning), and navigated repositioning of the Magerl screws.

Fig. 1
figure 1

Postoperative CT scan in sagittal and axial planes, grade 1 in modified Gertzbein and Robbins classification

Full size image
Fig. 2
figure 2

Postoperative CT scan in sagittal, coronal, and axial planes, grade 5 in modified Gertzbein and Robbins classification

Full size image

Statistical analysis

Patient characteristics were described using summary statistics (continuous data with mean, standard deviation, median, minimum and maximum; binary and categorical data with number and percentage).

Multivariate logistic regression was used to assess the risk for lower accuracy (grades 3–5) according to cervical spine level (upper cervical spine vs. subaxial cervical spine), taking into account relevant possible confounders such as diagnosis (trauma vs. other diagnosis) or surgeon (2 surgeons) as factors, and age as a continuous covariate. The possible within-patient correlation was accounted for using generalized estimating equations. The n = 203 includes all observed cases except those operated by a single surgeon, who operated on only 2 patients (4 screws). Statistical analyses were carried out using SAS V9.3.

Results

Between 2007 and 2012, a total of 64 patients with an average age of 61.3 years (SD 17.4; range 19–89 years) were included (Table 1). Altogether, 21 lateral mass screws in the atlas, 33 Magerl screws, 27 pedicle screws in the axis, 120 pedicle screws in the subaxial cervical spine, and 6 laminar screws were placed.

Table 1 Characteristics of the complete patient collective and classified by the level of screw placement [upper cervical spine (CS), subaxial cervical spine, or both; number of patients and percent with reference to group or statistical parameters, respectively; SD standard deviation]
Full size table

Mean operative time was 262.4 (SD 45.8) min when the entire cervical spine was involved (n = 10), 177.8 (SD 71.7) min for operations solely at the upper cervical spine (n = 28), and 236.1 (SD 84.2) min for the lower cervical spine (n = 26) (Table 1).

With regard to etiology, 36 patients were included in the study due to trauma, 5 due to degenerative changes of the cervical spine, 9 because of tumors, and 13 because of rheumatoid changes; there was also 1 patient with spondylodiscitis (Fig. 3).

Fig. 3
figure 3

Overview of the different diagnoses (a trauma, b rheumatoid instability, c tumor-related instability and d degenerative instability) with sample images

Full size image

Positioning accuracy of all screws is shown in Table 2. Screw position of grade 3 or worse was found in 19.01 % (n = 23/121) of screws implanted in surgeries for trauma. In surgeries for degenerative changes, 18.0 % of screw positions (n = 4/22) showed low accuracy (rated grade 3 or worse), in the field of tumor and rheumatic disease surgeries, low accuracy was found in 6.7 % each (n = 2/30). No low accuracy in screw position (n = 0/4) was observed in the spondylodiscitis case (Fig. 4).

Table 2 Number of screws inserted and accuracy of screw placement according to the modified Robbins and Gertzbein classification, by cervical spine level (a) and screw type (b)
Full size table
Fig. 4
figure 4

Intraoperative overview of the positioning of the patient, drill guide with reference spheres and navigation unit

Full size image

85.7 % (n = 18/21) of the lateral mass screws inserted in the atlas were assessed as grade 1 on the modified Gertzbein and Robbins scale, 4.8 % (n = 1/21) were classified as grade 2, 4.8 % (n = 1/21) as grade 3, 4.8 % (n = 1/21) as grade 4 or less (Table 2). Regarding the Magerl screws, 81.8 % (n = 27/33) were rated as Gertzbein and Robbins grade 1 and 15.2 % (n = 5/33) as grade 2; however, 3.0 % (n = 1/33) were classified as grade 5. The placement of 147 pedicle screws (axial and subaxial together, see Table 2) was classified as follows: 45.6 % (n = 67/147) grade 1, 36.1 % (n = 53/147) grade 2, 12.9 % (n = 19/147) grade 3, 3.4 % (n = 5/147) grade 4, and 2 % (n = 3/147) grade 5. 83.3 % of the laminar screws (n = 5/6) were classified as grade 1 and 16.7 % (n = 1/6) as grade 3. Regarding subgroups, 27 axial pedicle screws were classified as follows: 63.0 % (n = 17) as grade 1, 33.3 % (n = 9) as grade 2, 3.7 % (n = 1) as grade 3, and 0 % as grade 4 or 5 (Table 2b). Of the 120 subaxial pedicle screws inserted, 41.7 % (n = 50) were classified as grade 1, 36.7 % (n = 44) as grade 2, 15.0 % (n = 18) as grade 3, 4.2 % (n = 5) as grade 4, and 2.5 % (n = 3) as grade 5 (Figs. 5, 6).

Fig. 5
figure 5

Accuracy of screw placement according to cervical spine level

Full size image
Fig. 6
figure 6

Accuracy of screw placement according to screw type

Full size image

Comparing the different types of screws according to the percentage assessed as grades 1 and 2, Magerl screws yielded 96.97 %. 90.48 % of the lateral mass screws to the atlas were rated grade 2 or better, 83.3 % of the laminar screws, and 81.63 % of the pedicle screws (Fig. 7).

Fig. 7
figure 7

Intraoperative screenshot of the navigation unit while positioning

Full size image

In terms of spinal levels, the following differences were identified. Grade 1 precision of screws inserted to C1 or C2 was 85.7 and 73.9 %, respectively (Table 2). Regarding precision of screws inserted into C3 and C4, only 4 of 16 screws (C4) were rated less than grade 2. In contrast, the precision of screws inserted into C5, C6, and C7 was lower. 12–31 % of these screws were rated as grade 3 or higher.

Applying a multiple logistic regression model including level of screw placement, diagnosis, and surgeon as factors (with age as a continuous covariate), and accounting for the possible within-patient correlation, the precision of subaxial cervical pedicle screws was significantly lower than that of screws inserted in the upper cervical spine (P = 0.005) (Table 4). Regardless of screw type, 93.9 % of screws placed in the upper CS (at C1 and C2) were rated grades 1 and 2, compared to only 78.5 % located in the subaxial cervical spine (C3–C7).

Analysis of surgical complications (Table 3) revealed that 178 of the 207 screws (86 %) were implanted without causing clinical symptoms. In three cases the vertebral artery was affected, one case with fatal outcome. Screw revision was only necessary in one case (a malpositioned Magerl screw requiring surgical correction), and there was only a single case of CSF fistula. Spondylodiscitis in the adjacent segment, after anterior–posterior spinal canal decompression requiring anterior–posterior revision (6 affected screws), was diagnosed once, and impaired wound healing requiring surgical revision occurred three times (18 affected screws). Radicular symptoms due to poor screw position did not occur (Table complications according to location) (Table 3).

Table 3 Complications from screw insertion stratified by screw type (number of screws inserted inserted and percentage of all screw insertions)
Full size table

Discussion

Based on the present data, we conclude that axial and subaxial screws can be inserted with a high grade of accuracy using 3D-fluoroscopy-based navigation systems. There is no statistical significance for any of the cervical spine levels regarding screw insertion.

Cervical pedicle screw fixation is considered to be the most stable fixation [1, 2, 14, 23]. Due to a high perioperative perforation rate during the insertion [16], and the risks of injury to the vertebral arteries and neural structures associated with screw malposition, several methods supporting the surgeon or improving cervical screw placement have been previously investigated [36]. Roy-Camille et al. [28] stated that screw placement in the C3–C6 pedicles are accompanied by unacceptable risks. Over the years, technology and experience have continuously improved, so that a best rate of 7 % of misplaced cervical pedicle screws has been reported in the literature [3], even with the use of conventional insertion techniques without navigation. These are excellent results, and not reproducible by less experienced spine surgeons, even today. Currently, rates of screw misplacement of 5–40 % are reported using conventional techniques [10, 11, 17, 25, 33].

Anatomically, the diameter of the cervical pedicle is narrower than that of the thoracolumbar pedicle [16, 20, 24, 29, 35], the pedicle axis is more strongly convergent, and there are few anatomic landmarks to identify the exact entry point [8, 16, 18]. And although it is possible to identify the entry points on the basis of landmarks, targeting the exact drill angle remains very difficult [24]. C3 pedicle width averages 4.9 mm in males and 4.5 mm in females, with a minimum width reported as 3.0 mm. C4 pedicle width averages 4.7 mm in males and 4.6 mm in females with a minimum of 3.1 mm [7, 26]. This underscores the need for preoperative CT imaging.

The superiority of 3D navigation over free-handed techniques to improve the accuracy of cervical screw insertion and thus reduce the complication rate has been evaluated and documented in several studies [14, 21, 23, 25, 26, 36]. As a result, navigated posterior instrumentation of the cervical spine has become the standard in most hospitals. Advantages and disadvantages of the method have been well documented [21, 23]. In a multivariate analysis, precision was determined according to operated level and etiology. The risk of lower precision for screw placement in the subaxial cervical spine is increased by a factor of 3.88 compared to the upper cervical spine (Table 4). Diagnosis (Trauma vs. other) and surgeon showed no significant influence on precision.

Table 4 Risk of low accuracy (grade 3–5) according to cervical spine level and additional influencing factors based on a multivariate logistic regression model (cervical spine level, diagnosis, surgeon as factors and age as a continuous covariate). The possible within-patient correlation was taken into account by generalized estimating equations (GEEs; all observed cases excluding patients operated by one surgeon (3), who operated only 2 patients (4 screws). The level of significance was set at 5 %
Full size table

The results of the current study provide evidence for a high level of precision (defined as pedicle perforation of <2 mm, corresponding to grades 1 and 2 according to the modified Gertzbein and Robbins [10] classification).

In addition and regardless of type, 93.9 % of the screws inserted into the upper cervical spine (C1 and C2) showed an accuracy of grade 2 or better; the same applies to 78.51 % of the subaxially inserted screws (C3–C7). This difference can be explained by the fact that the anatomical conditions are more difficult in the lower versus upper cervical spine (smaller pedicle, more convergence requiring a more lateral entry point, inferior image quality due to shadow cast by the shoulder). Based on the current data, screw placement accuracy is better in the upper cervical spine (C1–C2) than in the lower cervical spine (C3–C7). Responsible for this was the partly lacking navigation accuracy in the lower cervical spine, especially the moderate visual quality caused by X-ray based navigation! Subsequently, a K-wire was inserted and a control scan performed. The moderate quality of this scan (metal-related artifacts caused by multi-exposure, so that the wire sometimes appeared fan-shaped, poor radiolucency of the lower cervical spine due to shoulder superimposition and bad image resolution) made it possible for the malposition to remain undetected. A second control scan performed after screw positioning brought similar problems.

Screw displacement is less frequent in the upper cervical spine, but when it occurs, it is fraught with more serious complications. One of three patients with vertebral artery involvement died of this complication.

All of the 3 vertebral artery affections were caused by lateral mass screws in the atlas and not in C2. This is the area of closest anatomical proximity to the vertebral artery—at least when the posterior arc of the atlas is chosen as entry point, as it was performed in 2 of the cases. On the other hand, the C2-nerve root is prevented from damage when this technique is used.

In the third case, the screw was inserted by using the Harms technique, which entails approaching the mass by surgical preparation below the posterior arc. This technique may result in more blood loss and increased danger for the C2 root, but the vertebral artery is kept at a safer distance. In this case an osteoporotic vertebra led to the dislocation of the screw.

Taking into account the modified Gertzbein and Robbins classification, we consider it justifiable to conclude that the accuracy is higher in the upper area of the cervical spine in statistical regards. This statement remains correct irrespective of anatomical conditions.

The level allowing for the most accurate screw insertion in our study was C3; here, 10 screws were implanted, and all met our definition of high precision (4 screws graded 1, 6 graded 2). At C2, 48 of 65 screws were placed with grade 1 accuracy, 14 with grade 2.

This study has some limitations. Although the learning curves of the surgeons were eliminated by previous experience, there were 3 experienced surgeons. Because of this number of surgeons, minor inaccuracies are possible. A single surgeon would have made standardization simpler. Because of the small case number and low complication rates, factors influencing the occurrence of complications could not be further analyzed in our patient collective.

Our results are supported by a meta-analysis conducted by Nakashima et al. [23], who did not identify statistical significance for screw insertion to any of the cervical spine levels in his multivariate analysis based on 390 implanted cervical screws. Nevertheless, his study reached the same conclusion as ours, that spine level should be regarded as an important factor in screw placement accuracy [23].

Conclusion

Axial and subaxial screws can be inserted with a high grade of accuracy using 3D fluoroscopy-based navigation systems. Nevertheless, while this useful technical innovation helps minimize the risks of misplacement, the surgery remains a challenge, as complications, when they arise, remain severe.