Accuracy and safety of C2 pedicle or pars screw placement: a systematic review and meta-analysis

Study design Systematic review and meta-analysis. Aim The purpose of this study was to compare the safety and accuracy of the C2 pedicle versus C2 pars screws placement and free-hand technique versus navigation for upper cervical fusion patients. Methods Databases searched included PubMed, Scopus, Web of Science, and Cochrane Library to identify all papers published up to April 2020 that have evaluated C2 pedicle/pars screws placement accuracy. Two authors individually screened the literature according to the inclusion and exclusion criteria. The accuracy rates associated with C2 pedicle/pars were extracted. The pooled accuracy rate estimated was performed by the CMA software. A funnel plot based on accuracy rate estimate was used to evaluate publication bias. Results From 1123 potentially relevant studies, 142 full-text publications were screened. We analyzed data from 79 studies involving 4431 patients with 6026 C2 pedicle or pars screw placement. We used the Newcastle-Ottawa Scale (NOS) to evaluate the quality of studies included in this review. Overall, funnel plot and Begg’s test did not indicate obvious publication bias. The pooled analysis reveals that the accuracy rates were 93.8% for C2 pedicle screw free-hand, 93.7% for pars screw free-hand, 92.2% for navigated C2 pedicle screw, and 86.2% for navigated C2 pars screw (all, P value < 0.001). No statistically significant differences were observed between the accuracy of placement C2 pedicle versus C2 pars screws with the free-hand technique and the free-hand C2 pedicle group versus the navigated C2 pedicle group (all, P value > 0.05). Conclusion Overall, there was no difference in the safety and accuracy between the free-hand and navigated techniques. Further well-conducted studies with detailed stratification are needed to complement our findings.


Background
Atlantoaxial instability or upper cervical spine instability is defined as excessive mobility as a result of either a bony or ligamentous abnormality [1]. Operative treatment of atlantoaxial instability is performed with a variety of fixation techniques. Spinous process wiring techniques were developed in 1910; laminar wiring techniques were developed in 1939; C1-2 laminar and modified posterior wiring technique were developed in 1991 [2]. These techniques did not provide sufficient biomechanical stability [2]. To address this matter, the C1-C2 transarticular screw fixation technique was introduced in 1992 [3]. However, 22% of cases were not appropriate candidates for transarticular screws because of an increased risk of vertebral artery injury [4]. Some more recently developed methods of C1-C2 fixation, C1 lateral mass screws combined with C2 pedicle/pars/laminar screws, have enhanced the stability of the upper cervical spine fixation techniques [2,5]. C2 pedicle screw placement was first described by Goel et al. in the 1980s [2].
An alternative to the prior mentioned techniques is the pars screw, sometimes referred to as an isthmus screw. C2 screw fixation techniques have been enhanced by the development of poly-axial screws and top-loading rods [2]. Researchers showed that C2 pars and pedicle screw utilization leads to high rates of arthrodesis [5,6]. These techniques are also employed in the subaxial cervical spine [5]. C2 pedicle and pars screws require accurate placement to avoid injury to vital structures, such as the vertebral artery, spinal cord, and nerve roots [2,5].
Overall, navigated and free-hand technique has been reported in detail elsewhere [7]. CT-based intraoperative navigation can be applied to determine a safe trajectory for C2 pedicle and pars screws placement but may be associated with increased time for image acquisition, increased radiation exposure to the patient, and possible registration inaccuracies. On the other hand, the freehand technique minimizes radiation exposure to the surgeon and patient [5].
No systematic reviews to date have compared the accuracy and safety of C2 pedicle and pars screws placed with the free-hand technique to the safety and accuracy of screws placed with the assistance of navigation. Therefore, the purposes of this systematic review and meta-analysis are (1) to assess C2 pedicle and pars screw placement accuracy and (2) to evaluate the difference in C2 pedicle and pars screw placement accuracy between free-hand and navigation techniques based on radiographic malposition.

Search strategy
The research strategy was designed around the PICO (Patient, Intervention, Comparison, and Outcome) question format. The present review was performed, based on the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines [8]. Electronic searches were performed using the Scopus, PubMed, Web of Science, and Cochrane Library databases up to April 2020. The literature involving all comparative studies were searched, containing the following search terms: "C2 pedicle," "C2 pars," "atlantoaxial instability," "upper cervical," "spine," "CT-based technique, " "navigated technique," "craniocervical," "freehand technique," "screws," "screws placement," "accuracy rate," and "safety."

Inclusion and exclusion criteria
All identified articles were systematically evaluated against the inclusion and exclusion criteria, independently reviewed by 2 authors, and disagreements were sent to third author for resolution. Any disagreement was resolved by discussion to reach a consensus. The inclusion criteria were as follows: studies presented accuracy rate in pedicle and/or pars C2 screw placement, based on either the free-hand or navigation techniques.
In recent years, different navigation systems such as fluoroscopic navigation, MR-based navigation [9], CTguided navigation, and O-arm-based navigation have been developed for pedicle/pars screw placement guidance. In this study, all of these techniques were considered navigation systems. The free-hand technique is defined by the placement of C2 pedicle or pars screws without the use of any of the aforementioned navigation systems [7]. In addition, screw guide templates and accuracy of preoperative imaging in predicting of trajectory and size of screw were considered free-hand technique.
The exclusion criteria were as follows: (I) duplicate publications; (II) reviews, case reports, commentary, and letters; (III) studies not published in English; (IV) studies which C2 screw sample size < 15; and (V) studies without available data regarding statistical techniques and lack of radiographic malposition reporting; (VI) studies with anterior cervical surgery; (VII) studies regarding cadavers; (VIII) anatomical and biomechanical studies; (IX) studies regarding without detailed information of C2; and (X) studies without separate C2 pedicle and pars screw placement information.

Data extraction
Two authors independently extracted the data from all eligible studies. The following data was extracted using a structured data extraction form from full articles: the first author, year of publication, country, sample size, gender, age, number of patients in C2 pars group in free-hand and navigation approach, number of patients in C2 pedicle group in free-hand and navigation approach, accuracy classification for assessing C2 pedicle/ pars screw placement, and accuracy rate in four subgroups as pedicle, pars free-hand and pedicle, and pars navigation technique based on radiological malposition.

Quality assessment
Identified studies were exported to Endnote version 7, and duplicates were removed. Two independent reviewers performed a full-text quality review. Disagreement between the two reviewers was resolved via discussion and a third author if needed. The NOS [10] was applied to evaluate the quality and risk of bias in included studies. The NOS includes 3 categorical criteria with a maximum score of 9 points: "selection" which accounts a maximum of 4 points, "comparability" which accounts a maximum of 2 points, and "outcome" which accounts a maximum of 3 points. No studies were randomized controlled trials; hence, studies with 7-9 points could be identified as high quality, 5-6 points as moderate quality, and 0-4 as poor quality. A summary of the procedure of quality assessment is presented in Table 1.

Statistical analysis
The raw data were entered into Microsoft Excel. Exact tests were calculated with SPSS. Only mean values were reported for the variables age at surgery and the number of patients; these variables were only semi quantitatively compared. In studies that did not report the age of C2 pedicle/pars screw group, the mean age was considered. In addition, in some of studies, the number of unreported cases was determined by dividing by two the number of the C2 pedicle/pars. Also, in some of studies, overall accuracy rates were considered for subgroups.
The meta-analysis was performed by using the Comprehensive Meta-Analysis version 2 (Biostat, Englewood, NJ). We assumed that the methodology of each study was unique, and the studies were heterogeneous. I-squared statistics were used to evaluate the heterogeneity of pooled accuracy rate estimates. If the I-squared value was > 50% and P value < 0.05, there was significant heterogeneity among the included studies, and a random effects model was applied to estimate the pooled results. Publication bias was estimated using Begg's funnel plot. A 2-tailed P value of less than 0.05 was considered statistically significant for all analyses.

Descriptive statistics
The literature search identified a total of 1320 articles. Figure 1 shows the flow diagram for the selection process for the systematic review. After removing 197 duplicated articles, 1123 remaining records were screened for title and abstract. Of those articles, 981 were excluded. Thus, 142 articles were assessed for eligibility by reading the full text. No randomized controlled trials were identified. Seventynine articles including 67 retrospective studies and 12 prospective studies were included for meta-analysis. The mean age of patients was 49.9 ± 13.3 years, and 57.4% of patients were male. A tabulated summary of the all studies are presented in Table 2 [5,9,.

Study characteristics and quality assessment
The characteristics of each study are shown in Table 2. Fifty-seven studies were conducted in Asian countries, 12 studies in North America, and 10 studies in Europe. Sixty-seven studies were retrospective, and 12 were prospective in design. Sample size ranged from 10 to 328 patients. The reported accuracy rate ranged from 65.2 to 100% for patients after cervical surgery. The NOS for each study can be found in Table 2. All of the studies analyzed in this systematic review scored five or above, which is considered of moderate to high quality studies [10], and 52 of the studies were considered high-quality studies.

Comparability
The accuracy rate screw placement and any additional factors as age, gender, and accurate classification of radiological malposition in different outcome groups are comparable, based on the study design or analysis. Confounding factors are controlled. (if yes, two stars; one star was assigned if one any additional factors was not reported) Outcome assessment 6. Ascertainment of the outcome: clearly defined outcome of accuracy rate (yes, two star for information ascertained by record accuracy rate based on classification of radiological malposition; one star if this information was not reported) 7. Appropriate statistical analysis: The statistical test used to analyze the accuracy rate is clearly described and appropriate for C2 pedicle or pars pedicle (if yes, one star; no star was assigned if the accuracy rate is reported overall)

Meta-analysis
A total of 79 studies, comprising 4431 patients with upper cervical fusion, were included in the metaanalysis. Overall, 6026 C2 pedicel/pars were used as follows: C2 pedicle free-hand (n = 4558), C2 pars free-hand (n = 506), C2 pedicle navigation (n = 941), and C2 pars navigation (n = 21). There were 55 studies indicating the association between the pedicle screw placement and the accuracy rate of upper cervical fusion patients. Since there was significant heterogeneity among the above 55 studies (I-squared value = 79.8% and P value < 0.001), we performed a random effects model to assess the pooled accuracy rate estimate and corresponding 95% CI. As shown in Fig. 2, the accuracy rate of the C2 pedicle screw free-hand technique was 93.8% (P value < 0.001). Forest plot for C2 pars screw placement of freehand technique (15 studies, I-squared value = 0.0%, and P value = 0.599), C2 pedicle screw placement of navigation technique (22 studies, I-squared value = 21.63%, and P value = 0.178 ), and C2 pars screw placement of navigation technique (2 studies, I-squared value = 0.0%, and P value = 0.608 ) are shown in Fig. 3 (a fixed effects model; accuracy rate 93.7%; P value < 0.001), Fig. 4 (a fixed effects model; accuracy rate 92.2%; P value < 0.001 ), and Fig. 5 (accuracy rate 86.2%; P value < 0.001), respectively. In this systematic review study, no statistically significant results were observed between the accuracy of placement C2 pedicle versus C2 pars in free-hand technique and the free-hand C2 pedicle group versus the navigated C2 pedicle group (all, P value > 0.05).

Publication bias
Publication bias was measured by Begg's test. For C2 pedicle screw of free-hand technique, the P value for Begg's test was 0.117, indicating that there was no significant publication bias among the included studies. Also, the P value for Begg's test was 0.766 for the C2 pars screw free-hand technique. Funnel plot and Begg's test did indicate obvious published bias for C2 pedicle screw of navigation technique (P = 0.001). In addition, due to studies, less than 3 Begg's test was not performed for C2 pedicle screw of navigation technique.

Discussion
To our knowledge, no previous systematic review, with or without meta-analysis, has been reported with the same purpose and methods. The analysis of the literature reveals that there are many studies fulfilling the        [88] 1990 Grade 0, when a screw was placed inside the bone; grade I, screw perforation of the cortex within 2 mm; grade II, screw perforation from 2 to 4 mm; and grade III, screw perforation of more than 4 mm. In some of articles, this classification was modified [28,56]. Grade 0 is considered the accuracy of in C2 screw placement [28]. [17,28,29,50,52,56,59,63,67,68,75,82,84,87] Laine et al. [89] 2000 Based on CT images, in this classification, screw position was staged as screw inside the pedicle or perforation of the pedicle cortex by up to 2 mm, from 2 to 4 mm, from 4 to 6 mm, or by more than 6 mm. Type I and type II were categorized as acceptable placement. [83] Rao et al. [90] 2002 Each screw position was assigned a grade from 0 to 3, as follows: grade 0 reflected no perforation of the pedicle; grade 1 indicated less than 2 mm of perforation of the pedicle; grade 2 represented 2-4 mm of perforation of the pedicle; and grade 3 reflected perforation greater than 4 mm. Grades 2 and 3 insertions were judged to be major perforations.
Overall, it is considered a perforation of less than 2 mm to be satisfactory. [62] Neo et al. [91] 2005 Screw positions were classified into four grades: grade 0, no perforation, and the screw was completely contained in the pedicle; grade 1, perforation < 2 mm (that is, less than half of the screw diameter); grade 2, perforations ≥ 2 mm but < 4 mm; and grade 3, perforation ≥ 4 mm(complete perforation). The screw was classified as grade 0 be acceptable. [27,34,40,43,45,46,48,53,54 ] Upendra et al. [92]. It was modified by Park et al. [79] 2008 Type I, ideal placement-screw threaded completely within bony cortex; type IIa, acceptable placement-< 50% of the diameter of the screw violating surrounding cortex and screw protrusion of < 1 mm from the anterior cortex for pedicle and pars screws; type IIb, relatively acceptable placement-screw violating < 33% of the diameter of the C2 transverse foramen (TF); type IIc, relatively unacceptable placement-screw violating ≥ 33% of the diameter of the C2 TF or ≥ 50% of diameter of screw violating surrounding cortex; type III, unacceptable placement-clear violation of TF or spinal canal; regardless of clinical neurovascular complications. Overall, types I, IIa, and IIb were categorized as acceptable placement and types IIc and III as unacceptable placement. [33,77,79] Sciubba et al. [19] 2009 It is described by location (lateral, medial, inferior, and superior) and percentage of screw diameter over cortical edge (0 = none; grade I = < 25% of screw diameter; grade II = 26-50%; grade III = 51-75%; and grade IV = 76-100%). Type 0 was categorized as acceptable placement. [5,19,47,73] Yukawa et al. [21] 2009 The accuracy of the placement of the pedicle screws into the medial/ lateral pedicle walls was evaluated on axial CT scans (2 mm slices), whereas superior/inferior pedicle wall screw location was examined on oblique radiographs. Incorrect screw placement was classified as either screw exposure or pedicle perforation. A screw was exposed if it broke the pedicle wall, but more than 50% of the screw diameter remained within the pedicle. A pedicle perforation occurred if a screw breached the pedicle wall, and more than 50% of the screw diameter was outside the pedicle. [21] Wang et al. [85] 2010 This classification was based on axial plane, para-sagittal plane, and coronal plane. The grading has been described elsewhere in detail [85]. [85] Kawaguchi et al. [40] 2012 Grade 0, the screw was completely located in the vertebral pedicle; grade I, the screw penetrated the pedicle bone cortex < 2 mm without complications; grade II, the screw penetrated the pedicle bone cortex > 2 mm without complications; and grade III, complications related to screw placement occurred, such as nerve and vertebral artery injuries. Grade 0 was considered to be the correct location of pedicle screws and safe placement. [70,71] Uehara et al. [49]. 2014 The screw insertion status was classified as grade 1 (no perforation), indicating that the screw was accurately inserted in pedicle; grade 2 (minor perforation), indicating perforation of less than 50% of the screw diameter; and grade 3 (major perforation), indicating perforation of 50% or more of the screw diameter. The screw was classified as grade 1 be acceptable. [49,62] Smith et al. [93] 2016 On postoperative CT scans, type I was defined as ideal placement without cortical violation; type II was an acceptable placement with less than half the diameter of the screw violating the surrounding cortex and less than 1 mm protrusion from the anterior cortex; and type III is an unacceptable placement with clear violation of the transverse foramen or spinal canal.
[80]  Hlubek et al. [7] 2018 Grade A, screw completely confined within cortical surfaces; grade B, transverse foramen violation with the screw obstructing 1-25% of the foramen; grade C, transverse foramen violation with the screw obstructing 26-50% of the foramen; grade D, transverse foramen violation with the screw obstructing 51-75% of the foramen; grade E, transverse foramen violation with the screw obstructing 76-100% of the foramen; grade M, medial breach into the spinal canal. Grades A and B were determined to be acceptable placement, and Grades C-E and M were determined to be unacceptable. That is why the current study can include 79 studies. Statistical analyses showed that the placement accuracy rate for the free-hand C2 pedicle group was comparable to that for the navigated C2 pedicle group and between C2 pedicle and pars screws placement. Overall, the freehand technique was not found to accurate than navigation for C2 pedicle/pars screw placement. In this study, there was no difference in the safety and accuracy between the free-hand and navigated techniques, which could be for the following reasons: (a) Screw guide template studies with the highest precision and accuracy were considered free-hand technique. (b) Experience with navigation system also plays a role in this arena. (c) Less number of navigation system studies compared to free-hand technique due to the lack of popular accessibility and (d) heterogeneity in studies.

Study consistency
Of the 79 articles, only 12 fully reported on patients' recruitment or the source of prospective data. No randomized trial was found. Learning curve and size of screws  were not consistently reported, resulting in a potential bias. The surgical approach was described in nearly all studies, while new entry point and trajectory, which could indicate a potential for screw malposition, were not consistently reported. For accuracy assessment of C2 pars/pedicle screw placement, a variety of grading criterion are reported in the literature. Comparison between accuracy rates was limited by the presence of twelve different definitions of accuracy rate and twenty-five studies (31.6%; 25/79) not presenting any definition. In addition, 14 articles (17.7%; 14/79) used the Gertzbein and Robbins grading system for evaluation of accuracy of screw placement. In a review study of C2 pedicle screw placement, Elliott et al. [94] showed that the incidence of malposition, confirmed by CT scan, varied from 1.1 to 44% in cases with fluoroscopic guidance. However, in this systematic review, the reported accuracy rate ranged from 65.2 to 100%. This wide range could be a result of varying classification method of screw displacement among studies.

Study quality
Only 59.4% (47/79) of studies used a clearly defined accuracy rate classification definition. Most studies were small with an average study group size of 44 patients dropping to 31 when removing the eight studies with over 100 patients. The method of screw insertion was well defined, or a pre-defined method was cited. In some of studies, the type and size of screws was not specified. Only two studies [52,84] assessed the accuracy rate of navigated C2 pars screw malposition, and data were limited for comparison. Therefore, further research with large sample sizes comparing accuracy rates of navigation with free-hand methods is warranted.
Studies included heterogeneous populations with varying pathological types. However, accuracy of either procedure should not have been affected by pathology. Furthermore, more complex pathology or anatomy was not reason for choosing navigation over free-hand technique or vice versa [7]. Also, here was considerable regarding the length of C2 pars/pedicle screw, navigated technique, surgeon's experience, and grading criteria of accuracy, which can affect results. A standardized assessment process, moving forward, would greatly assist in future analyses in this arena. According to this 20-year study , over the past 20 years, numerous navigation systems such as MR-based navigation, CTguided navigation, and O-arm-based navigation have been developed. Each of these systems has strengths and weaknesses concerning yield, cost, speed, and learning carve. Hence, it may cause heterogeneity to put all navigation systems in the same group. Albeit, it could be evaluated separately in the future.
Until now, a few studies have compared the accuracy of C2 pedicle and pars screw placement for atlantoaxial fusion [7,84]. Lee et al. showed that O-arm navigation slightly improved the accuracy rate of C2 pedicle screw positioning, compared to the free-hand technique, though statistically meaningful results were not reported [84]. A C2 screw accuracy rate was reported to be 100% by Wu et al. [9]. They used 3D model simulation software for better evaluation of anatomy and then applied this to the navigation process [9]. Contrary to their study, Hlubek et al. found that the free-hand technique was significantly more accurate than CT-based navigation for C2 pedicle/pars screw placement [7]. Hence, illustrating the ongoing challenge associated with data analysis.
The corridor for C2 pedicle and pars screw placement is often narrow. Hence, it would seem that navigation techniques would present a natural solution to this corridor definition challenge in anatomically complex cases. There are several advantages of using an intraoperative image guidance for cervical surgery, including multi planar CT images of different operative levels in a single sequence can be achieved to increase accuracy of surgery, decreased radiation exposure to the surgeon and patient, and screw positions can be tested in the surgical field, which will reduce the failure rates [84]. On the other hand, surgical landmarks and fluoroscopy have been applied routinely for pedicle screw insertion, but a number of studies disclose inaccuracies in placement using these conventional techniques. Moreover, the free-hand technique is safe and accurate when it is in the hands of an experienced surgeon [95]. Then, it could be argued that the use of the navigation for C2 pars and pedicle placement is better than free-hand technique. However, there are many probable sources of error with the navigated Fig. 5 Point estimates with 95% confidence intervals and forest plot of studies reporting on accuracy rates of fusion following posterior atlantoaxial fusions with C2 pars screw and navigation technique method that resulted in less accurate screw placement. The CT image may be distorted because of metal artifacts from prior implant placement and the extra time required to set up the navigation system [84]. Also, the motion of C2 relative to the reference frame may introduce error. In addition, registration inaccuracies could be related to lack of correspondence between the pre-operative CT image, obtained in the standard supine position, and the intraoperative prone position, especially in patients with cervical instability. Other sources of inaccuracies include accidental displacement or reference frames [7]. Hence, in order to correct the source of error, further research is required to provide evidence of the precise cause of inaccuracy with navigated C2 pedicle and pars screw placement.

Strengths and limitations
The strengths of this review include the broad search strategy in four major databases and high sensitivity of the abstract search. This study has several limitations, though. First, this is a meta-analysis carried out at study level, meaning that different confounding factors from the patient level were not evaluated and included in the analysis. Second, the search was limited to English publications. Potentially relevant studies could have been missed. Third, although it seems that the CT-based navigation could be useful in C2 pedicle screw placement, this intraoperative CT navigation is not universally available. Moreover, it is mandatory to consider the radiation exposure for operative staff, which is significantly higher with CTbased navigated than with standard techniques. Fourth, all studies were performed retrospectively. To the best of our knowledge, no prior prospective randomized control studies have been performed to compare the safety and accuracy of the free-hand technique versus navigation for the placement of C2 pedicle and pars screws; hence, a high level of evidence was lacking in our review. Finally, the main limitation of the study was the high level of heterogeneity in the methods used among the included trials. In particular, there were heterogeneities in (1) variety in surgical technique and screw guide templates, (2) variety in navigation systems, (3) the screw placement accuracy measures applied, (4) length and size of screw (presently, there are no criteria on the size of C2 pedicle screws that maximizes the C2 accuracy rate placement), (5) the learning curve associated with using free-hand techniques and navigation systems, (6) costs from acquiring guidance technology, and (7) radiation exposure. These items were not discussed in the included articles, but it would be of interest in future prospective studies.

Conclusion
The C2 pedicle/pars placement accuracy rate for the free-hand group was comparable to that for the navigated group. Further randomized controlled trials with large sample sizes comparing accuracy rates of navigated with free-hand methods are warranted to complement the existing evidence.
Abbreviations NOS: Newcastle-Ottawa Scale; CMA: Comprehensive Meta-Analysis Software