Introduction

Cervical pedicle screw (CPS) placement was first described for traumatic injuries by Abumi et al. in 19941. Meanwhile, numerous studies have demonstrated clinical feasibility of CPS placement as well as its biomechanical superiority over alternative cervical fixation methods for numerous indications2,3,4,5,6. Nevertheless, compared to pedicle screw placement in the thoracic and lumbar spine, the number of studies on cervical pedicle screws is limited. The clinical application of CPS is rather rare compared to lateral mass screws (LMS) since CPS placement is more challenging due to close proximity of the pedicle to neurovascular structures3,7,8,9,10,11. In addition to the variable anatomy of cervical pedicles, the high mobility of the cervical spine also entails challenging surgical conditions7,12,13,14,15. The pedicle as the screw corridor is only a few millimeters wide, leaving limited room for error1,16. Although not all pedicle wall perforations cause complications and a margin of acceptable error is considered in different classification systems rating CPS accuracy, surgeons should always aim to avoid perforations of the pedicle when placing screws17,18,19. Yet, perforation rates of 3.8–27% have been reported for freehand and fluoroscopy-guided screw placement20,21,22. Since the introduction of intraoperative navigation based on intraoperative computed tomography (iCT) or cone beam computed tomography (CBCT), various studies have compared fluoroscopy-guided and navigation-based techniques and have shown significantly greater accuracy rates and less postoperative complications for navigated screw placement17,23,24,25. In addition, this technique can reduce the time needed for surgery and radiation exposure to staff26,27. However, the number of studies investigating different combinations of 3D imaging devices and navigation systems for navigated CPS placement is scarce. Therefore, the aim of this study was to compare three different combinations of imaging devices and navigation systems regarding accuracy of cervical pedicle screw placement, radiation dose as well as the time needed for screw placement.

Methods

In this experimental study, CPS placement was performed on 24 identical radiopaque artificial cervical spine models (Synbone, Zizers, Switzerland). Three different combinations of intraoperative imaging devices and navigation systems were used. Investigated were a 3D C-arm CBCT (cCBCT; Cios, Siemens Healthineers, Erlangen, Germany) in combination with Nav1 (Pulse, NuVasive, San Diego, California, USA) navigation software (Trial 1). The second combination investigated was a mobile iCT (Airo, Brainlab, Munich, Germany) and Nav2 (Curve Image Guided Surgery, Brainlab, Munich, Germany) (Trial 2). In the third trial, the combination cCBCT and Nav2 was investigated (Trial 3; Fig. 1).

Fig. 1
figure 1

Study design.

Full size image

In each of the 24 models, 12 screws were inserted bilaterally in levels C2 to C7. Depending on the navigation system, the following two respective screw systems, were used: the VuePoint system (NuVasive, San Diego, California, USA) in Trial 1 and the Symphony system (DePuy Synthes, Raynham, Massachusetts, USA) in Trial 2 and 3. All screws placed were 3.5 mm in diameter. Screw length was determined depending on vertebral size of the respective cervical spine level. All screw sizes used were identical for both screw systems, as all models were identical. The screws were placed by two different surgeons, an experienced spine surgeon (ES) prone to navigated procedures and a third-year resident surgeon (RS) with no experience in navigated screw placement or spine surgery in general. In every trial, each surgeon placed screws unilaterally in an open approach. The surgeons aimed for most accurate screw placement possible.

To perform CPS placement, the models were placed in prone position in a spine bed (Synbone, Zizers, Switzerland). The spine bed was then placed on a radiolucent carbon table and fixed to the table with tension belts to avoid movement of the model. The dynamic reference base (DRB) of the navigation system was mounted on the spinous process of C5. A scan of the model was obtained using the respective imaging device and transferred to the navigation system. Using the cCBCT (Trial 1 and 3), the reference scan was acquired in standard quality (200 images, 30 s). In Trial 2 the reference was performed using the standard settings for imaging of the cervical spine (adult cervical spine, 120.0 kV; 95.0 mA). Navigation instruments were then calibrated, and the accuracy of image registration was verified at different anatomical landmarks using the navigated pointer. Subsequently, screw placement was initiated: The navigated drill guide was used to locate the entry point and trajectory of the screw, which was then saved to the navigation system. Trajectory was continuously controlled while electric drilling was performed with a preset depth of 20 mm. Finally, immediately after drilling the screw hole, the screw was placed according to the planned trajectory using the navigated screwdriver. In all models, screw placement was performed from C2 to C7 unilaterally before the contralateral side was instrumented. After every single screw placed the navigation accuracy was checked using the navigated instruments to prevent inaccuracies caused by the torque of previous screw placement.

Screw placement time was defined as the time required to place a screw from the first contact of the drill guide with the pedicle of the spine model until the screw was placed to the surgeon’s satisfaction. Screw placement time did not include the time needed for image acquisition and calibration of the navigated instruments. After all screws were placed, a second standard quality scan (200 images, 30 s) was performed, according to the clinical procedure. This scan was considered in the radiation measurement. Subsequently, for the purpose of this study and to allow optimal assessment of screw position, another scan was performed in high quality. Assessment of cervical pedicle screw position was performed by a blinded observer according to the Neo classification and repeated after six weeks28. As suggested by Neo et al., screws with no breach (Grade 0) or a breach of < 2 mm (Grade 1) were deemed acceptable while breaches of ≥ 2 mm (Grade 2–3) were considered inacceptable.

All screws were assessed twice by the observer. If a screw was rated differently in the two evaluations, the worse rating was used.

To compare the radiation dose applied during navigated screw placement between the groups, the dose protocols were retrieved of the two imaging systems used. To allow comparability between the dose area product (DAP; [mGycm2) as reported by the 3D C-arm CBCT and the dose length product (DLP; [mGycm]) as reported by the iCT, the data was converted into effective dose ([mSv]) using conversion factors. For DAP, a conversion factor of 0.27 mSv/mGycm2 was used as suggested by Hwang et al.29. For DLP, a conversion factor of 0.0059 mSv/mGycm was used as suggested by the American Association of Physicists in Medicine for the anatomical region of the neck.

Kolmogorov-Smirnov test was used to check for normal distribution of data. Descriptive statistics for continuous variables are shown as means ± standard deviation (SD) with Gauss distribution, and median ± 95% confidence interval (95% CI) for non-normally distributed data.

The statistical tests were chosen according to the distribution of the data and the assumptions of the tests: Screw accuracy according to the Neo classification was compared between the trials using Kruskal-Wallis test with Dunn’s correction for multiple comparisons. Two-tailed Fisher’s exact test was used to compare screw accuracy rates between surgeons. Screw placement, reported in minutes (min), between the trials was compared using Kruskal-Wallis test with Dunn’s correction for multiple comparisons. Two-tailed Mann-Whitney test was chosen for comparison of time demand between the surgeons. The radiation dose measured was compared between trials using ordinary one-way ANOVA with Tukey’s correction for multiple comparisons.

To calculate intrarater reliability, kappa (κ) was used with its interpretation according to Landis and Koch. To account for close matches of the two assessments, Kappa was weighted (κw).

The screw-to-pedicle diameter ratio was calculated for all levels. The pedicle diameters were measured in an iCT scan performed prior to screw placement. After multiple measurements at each level, the smallest diameter perpendicular to the optimal screw axis was considered the pedicle diameter.

The significance level was set at P < 0.05. Prism 10 (Version 10.1.1, Graphpad Software, San Diego, California, USA) was used for statistical analysis and figure creation.

Results

80 subaxial pedicle screws and 16 C2 pedicle screws were placed in each trial, resulting in a total of 288 screws placed (Fig. 2). The overall screw placement accuracy was 99.1% for Trial 1, 99.1% for Trial 2 and 98.2% for Trial 3 showing no statistically significant differences (P = 0.30; Fig. 3). The comparison of accurately placed screws depending on the surgeon also did not show a significant difference with 97.9% and 99.3% accuracy for the RS and the ES, respectively (P = 0.31). One critical perforation each was found for C3, C4, C6 as well as C7 (Table 1). Except for C7, all critical perforations occurred on the lateral pedicle wall. The screw-to-pedicle diameter ratio for the levels C2 to C7 is shown in Table 2.

Fig. 2
figure 2

Oblique view on the cervical spine model with a lateral Grade 1 perforation of C5.

Full size image
Fig. 3
figure 3

Accuracy of CPS placement according to Neo classification for Trials 1, 2 and 3 (n = 96 each).

Full size image
Table 1 Neo classification Grade considering trial, surgeon and vertebral level.
Full size table
Table 2 Screw and pedicle diameter as well as the corresponding screw-to-pedicle diameter ratio (SPR) for all levels.
Full size table

Intra-rater agreement was fair for all imaging modalities with exact grading agreement in 70 (72.9%), 65 (65.7%) and 67 cases (69.8%), for Trial 1, Trial 2 and Trial 3, respectively (Fig. 4). The agreement on critical perforations was 95/96 (99.0%) for Trial 1 and Trial 3 and 96/96 (100.0%) for Trial 2.

Fig. 4
figure 4

Weighted intra-rater reliability κ ± 95% confidence interval for the exact grading according to the Neo classification for the different imaging modalities.

Full size image

Screw placement time per screw was significantly longer in Trial 1 (1.97 min; 95% CI [1.80, 2.30]) compared to Trial 2 (1.17 min; 95% CI [1.07, 1.33]; P < 0.0001) and Trial 3 (1.35 min; 95% CI [1.22, 1.48]; P < 0.0001) while the comparison of Trials 2 and 3 showed no significant difference (P = 0.77). Statistical analysis of the screw placement time showed a significant difference between the experienced spine surgeon (ES) and the resident surgeon (RS) with a median of 1.32 min (95% CI [1.37, 1.68]) for the ES compared to a median of 2.10 min (95% CI [1.78, 2.40]) for the RS (P < 0.001). The difference between the surgeons was significant for Trial 1 (P < 0.001) and Trial 3 (P = 0.04), while the difference for Trial 2 did not reach significance (P = 0.05). For the RS, a learning curve was observed during Trial 1 when comparing the time needed for placement of the first 24 screws compared to the following 24 screws with a median time of 3.1 min (95% CI [2.7, 5.5]) compared to 2.0 min (95% CI [1.8, 2.8]; P = 0.006).

The mean radiation dose measured in the different trials is shown in Fig. 5. Statistical analysis showed a significant difference between Trials 1 and 2 (P < 0.0001) as well as Trials 2 and 3 (P < 0.0001), while no significant difference between Trials 1 and 3 was found (P = 0.99).

Fig. 5
figure 5

Mean effective dose per model in each group (n = 8 each).

Full size image

Discussion

The aim of this study was to compare three different combinations of intraoperative imaging devices and navigation systems for CPS placement under standardised experimental conditions. In addition to the accuracy of cervical pedicle screw placement, the screw placement time and the radiation dose applied were measured.

The results of this study show high accuracy rates of over > 98% for navigated cervical pedicle screw placement regardless of the imaging modality or navigation system used despite the high screw-to-pedicle diameter ratio – especially for the levels C3, C4 and C5.

These results add to previous reports on the accuracy of cervical pedicle screw placement using different techniques, such as fluoroscopic guidance and spinal navigation23,24,30,31,32,33. In their systematic review of 1,427 CPS placed under clinical conditions using navigation, Mahmoud et al. reported an average accuracy rate of 91.4%23. In another review, Tarawneh et al. found relevant pedicle perforations to be present in 12.5% of screws24. Using CBCT-based navigation (O-arm and StealthStation S7 navigation system, both Medtronic, Minneapolis, Minnesota, USA), Chachan et al. as well as Ishikawa et al. reported accuracy rates of approximately 95%20,30. Both studies used the Gertzbein-Robbins grading system to assess screw position, which is comparable to the Neo classification used in the present study34. However, none of the above studies compared cervical pedicle screw placement using different imaging and navigation systems under quasi-identical conditions.

The differences between the results presented and those reported by other authors may be due to the limited comparability of the experimental setup with clinical reality and the use of different imaging devices and navigation systems. In addition, the surgeons in this study had complete visual feedback due to the open spine model setup, which may have influenced the accuracy screw placement.

In contrast to our clinical procedure, in which the screws are placed only after all drill holes have been completed and the corresponding screw trajectories have been marked with K-wires, in this study, due to the lack of cannulated screw systems, the screw was placed immediately after drilling. In clinical reality, this technique should be avoided in order to avoid a potential compromise of navigation accuracy due to a change in anatomical configuration. In our experimental setup, the navigated instruments were used to check the navigational accuracy using anatomical landmarks after each screw.

Although several studies have shown advantages of navigated over other CPS placement techniques, freehand and fluoroscopically guided techniques are widely used. Jung et al. postulated that the use of a freehand technique without image-guidance may result in screw placement accuracy comparable to image-guided techniques when performed by a well-trained surgeon. The authors concluded that regardless of the technique used, surgeon experience is the determining factor in screw accuracy21. In contrast, for navigated screw placement, the results of the present study show no significant differences in screw placement accuracy between an experienced spine surgeon and an inexperienced resident. However, it should be considered that real-life intraoperative factors did not interfere with CPS placement in this study due to its experimental design and the use of artificial bone models. Nevertheless, the results presented suggest that navigation in general may be a helpful tool for the less experienced surgeon and could be used to improve the accuracy of screw placement. While there was no difference in accuracy, the inexperienced surgeon took significantly more time to place the screws than the experienced spine surgeon. However, this difference may not be clinically relevant as the average difference, although significant, was approximately 30 s per screw. In addition, the inexperienced surgeon showed a learning curve regarding the time required for CPS placement, as reported by Heo et al.35. This may also be the reason for the significantly longer screw placement time in Trial 1 compared to Trials 2 and 3.

However, experience in spine surgery, and in particular cervical spine surgery, and in the use of intraoperative navigation is crucial to minimise the challenges of navigated CPS placement. In particular, the cervical spine has a high degree of intersegmental mobility, which is a major concern when using navigation in this region, as it can lead to a mismatch between the actual and displayed anatomy. A further challenge is the selection of a sufficiently stable position for the Dynamic Reference Base (DRB). This is essential to avoid its interference with the surgical field, which could lead to navigation inaccuracies due to unintentional contact with the array, and also to minimise the relative movements of the instrumented vertebra in relation to the vertebra on which the array is mounted36,37. This risk was unequal between the navigation systems, as the DRB of Nav2 was evidently smaller than that of Nav1. However, considering the results presented, this inequality did not seem to effect accuracy rates in this study.

Although the absolute numbers of radiation doses measured cannot be compared to clinical practice, this experimental study showed significant differences between the imaging devices investigated. The effective doses applied in Trials 1 and 3 were comparable, which was to be expected as the same intraoperative imaging system was used in both trials. In Trial 2, in which the iCT was used, the effective dose was more than 23 times higher compared to Trial 1 and 3. When comparing different imaging systems regarding radiation dosage, previous studies have shown inconsistent results38,39. A recent study performed by Klingler et al. reported the radiation dose for intraoperative imaging of the cervical spine using iCT to be approximately 14 times higher when measured at the thyroid gland and 42 times higher when measured at the eye lens of the patient compared to 3D C-arm CBCT, which is in accordance with the results presented here40.

Tkatschenko et al. have stated that the potential benefits of 3D navigation may outweigh the potentially higher one-time radiation exposure of 3D imaging compared to fluoroscopy; however, minimising radiation exposure to the patient and staff should remain a priority, especially when comparable systems with much lower radiation doses are available41.

A comparison of radiation exposure for lumbar screw placement using the same imaging devices was performed on human specimens by Foster et al. Interestingly, the authors found no significant difference between the iCT and CBCT, which may be due to a different form of data collection and a different spinal region investigated42.

The main limitation of this study is the use of artificial spine models in an experimental setting. Due to the experimental setup and the lack of soft tissue in our setup, the absolute radiation doses measured cannot be compared to doses recorded in a clinical setting. However, creating quasi-identical conditions to allow direct comparability, as seen in this study, in a clinical setting is challenging, if not impossible. Other considerations when using spinal models include the lack of difference between cortical and cancellous bone, which results in the screw perforating the pedicle rather than running along the cortex of the pedicle, which also goes unnoticed due to the lack of haptic feedback36,43. On the other hand, screw placement is simplified by the absence of soft tissue and bleeding, which can hamper the surgeon even in open procedures20,44. While the intra-rater reliability for the exact grading of the screws according to the Neo-classification was only fair for both imaging modalities, the differentiation between acceptable and unacceptable screws seems to be possible with both iCT and CBCT. Possible reasons for the low level of intra-rater agreement include the low contrast of the spine models used in 3D imaging compared to human bone, as well as the screw-to-pedicle diameter ratio of approximately 1 for all levels except C7, which makes it difficult to assess screw perforations with millimeter accuracy.

Conclusions

Under experimental conditions, high accuracy rates of over > 98% were achieved for navigated cervical pedicle screw placement regardless of the imaging modality or navigation system used despite the high screw-to-pedicle diameter ratio. Radiation doses were significantly lower for CBCT-guided navigation compared to iCT-guided navigation. Further studies on human specimen should be performed under comparable conditions to confirm the results presented.