Facet Violation With Percutaneous Pedicle Screw Placement: Impact of 3D Navigation and Facet Orientation

Abstract

Background: The gold standard for percutaneous pedicle screw placement is 2-dimensional (2D) fluoroscopy. Data are sparse on the accuracy of 3-dimensional (3D) navigation percutaneous screw placement in minimally invasive spine procedures. Objective: We sought to compare a single surgeon’s percutaneous pedicle screw placement accuracy using 2D fluoroscopy versus 3D navigation, as well as to investigate the effect of facet orientation on facet violation when using 2D fluoroscopy. Methods: We conducted a retrospective radiographic study of consecutive cohort of patients who underwent percutaneous lumbar instrumentation using either 2D fluoroscopy or 3D navigation. All procedures were performed by a single surgeon at 2 academic institutions between 2011 and 2018. Radiographic measurement of screw accuracy was assessed using a postoperative computed tomographic scan. The primary outcome was facet violation, and secondary outcomes were endplate/tip breaches, the Gertzbein-Robbins classification for cortical breaches, and the Simplified Screw Accuracy grade. Statistical comparisons were made between screws placed using 2D fluoroscopy versus 3D navigation. Axial facet angles were also measured to correlate with facet violation rates. Results: In the 138 patients included, 376 screws were placed with fluoroscopy and 193 with navigation. Superior (unfused) level facet violation was higher with 2D fluoroscopy than with 3D navigation (9% vs 0.5%), which comprises the main cause for poor screw placement. Axial facet angles exceeding 45° at L4 and 60° at L5 were correlated with facet violations. Conclusion: This retrospective study found that 3D navigation is associated with lower facet violation rates in percutaneous lumbar pedicle screw placement when compared with 2D fluoroscopy. These findings suggest that 3D navigation may be of particular value when facet joints are coronally oriented.

Introduction

There has been a shift toward minimally invasive spine (MIS) surgery in recent years [12,14]. Skeptics have raised concerns about the accuracy of pedicle screws placed percutaneously without direct visualization, and some studies have validated these concerns [20].

The majority of percutaneously placed screws studied have been placed with the guidance of 2-dimensional (2D) fluoroscopy. Despite the advent of novel technologies such as 3-dimensional (3D) navigation, robotic surgery, and augmented reality, 2D fluoroscopy has been the mainstay of image guidance for MIS instrumentation [3,5,13]. This is due to the perception that advanced imaging adjuncts prolong MIS surgeries without providing a clinically meaningful improvement in pedicle screw accuracy. There is, in fact, evidence that 3D navigation improves screw accuracy in open surgery [4,5], but data are limited on its value in MIS procedures. The literature has also focused largely on a specific type of screw misplacement—the medial breach—at the expense of other contributions to poor screw placement. Facet violation, for instance, is a consequential type of screw misplacement, because it is linked to adjacent segment disease (ASD) and is a risk for revision surgery [1,10]. Furthermore, studies suggest that facet violation is more common with percutaneous screw placement [6].

We hypothesized that 3D navigation may be superior to 2D fluoroscopy in screw placement accuracy. To test this hypothesis, we conducted a retrospective radiographic study comparing a single surgeon’s percutaneous pedicle screw placement using 2D fluoroscopy versus 3D navigation. This study also enabled us to investigate the impact of facet orientation (as measured by the axial facet angle) on the likelihood of facet violation when percutaneously placing screws with 2D fluoroscopic guidance.

Methods

This is a retrospective radiographic cohort study involving patients who underwent minimally invasive lumbar spine surgery at 2 academic institutions from 2011 to 2018. Institutional review board approval was obtained for the collection and analysis of data at both institutions. One surgeon (S.Q.) performed all procedures. All patients underwent primary posterior lumbar or lumbosacral pedicle screw instrumentation using either percutaneous 2D fluoroscopic or percutaneous 3D navigation–guided methods for degenerative spinal stenosis, with or without spondylolisthesis and/or degenerative scoliosis. Exclusion criteria included open surgery, traumatic fractures, prior instrumentation at the involved vertebral levels, or the absence of a postoperative computed tomographic (CT) scan. Patient demographics and operative data were extracted from the electronic medical records via Epic Hyperspace (Epic Systems Corp., Verona, WI). All radiographic evaluation was performed using a picture archiving and communication system (PACS; GE Healthcare, Boston, MA).

A total of 138 patients were included in this study. These patients were selected from 340 consecutive patients who underwent percutaneous lumbar spine instrumentation between February 2011 and October 2018. Of those included in the study, 97 patients underwent fluoroscopic minimally invasive posterior instrumentation (376 screws; a number of patients had only unilateral fusion) and 41 patients underwent 3D-navigated minimally invasive posterior instrumentation (193 screws). All surgeries in the fluoroscopy cohort were performed at institution 1, and all but 2 surgeries in the 3D navigation cohort were performed at institution 2. A postoperative CT scan including the lumbar spine was obtained on average 351 days after surgery—384 days for the fluoroscopy group and 274 days for the 3D navigation group. The patient cohorts demonstrated statistically significant differences in American Society of Anesthesiologists (ASA) score and the rate of pars defects (both were higher for the fluoroscopy group). There was no statistical difference in body mass index (BMI), sex, smoking, or the presence of preoperative spondylolisthesis, central stenosis, or foraminal stenosis (Table 1).

Table 1.

Patient demographics and preoperative radiographic findings.

	Fluoroscopy	Navigation	P value
No. of patients (n)	97	41
No. of screws	376	193
No. of screws per patient
M	3.88	4.71
Median [IQR]	4 [4–4]	4 [4–6]
Range	2–6	2–12
Age	58.5 ± 11.7	60.5 ±11.7	.359
Gender			.060
Male	42 (43.3%)	24 (61.5%)
Female	55 (56.7%)	15 (38.5%)
Body mass index (kg/m2)	29.16 ± 5.76	28.19 ± 4.58	.351
Current smoker (within 1 year)			.182
Current active smoker	4 (4.1%)	2 (5.1%)
Former smoker	44 (45.4%)	11 (28.2%)
Nonsmoker	49 (50.5%)	26 (66.7%)
ASA classification			<.0001
1	11 (11.3%)	2 (5%)
2	49 (50.5%)	36 (92.3%)
3	35 (36.1%)	1 (2.6%)
4	2 (2.1%)	0 (0.0%)
Spondylolisthesis	67 (69.1%)	30 (76.9%)	.408
Central stenosis	83 (85.6%)	28 (71.8%)	.085
Foraminal stenosis	75 (80.6%)	27 (69.2%)	.175
Pars fracture	17 (17.7%)	1 (2.6%)	.023

ASA American Society of Anesthesiologists, IQR interquartile range. Significance for the bold values are P < .05.

Two separate surgical methods were used for this study: 2D fluoroscopic–guided and 3D navigation–guided percutaneous pedicle screw placement. All surgeries were performed on patients placed in the prone position.

Fluoroscopic-guided pedicle screw placement was performed using a previously published protocol [8]. The protocol involves the Jamshidi needle docked at the transverse process and gently migrated medially until the ideal start point is obtained at the lateral border of the pedicle (the lateral-to-medial technique [LMT]). The Jamshidi needle is then gently advanced while working toward the medial border of the pedicle. We ensure that there is a bony end point on insertion of the guidewire. The pedicle is then tapped and the screw is placed over a guidewire under radiographic guidance.

The technique for 3D navigation–guided percutaneous screw placement was also performed using a published protocol [18,19]. It begins with an intraoperative 3D scan using a Ziehm Vision fluoroscope (Ziehm Imaging, Nuremberg, Germany) that is taken with the patient in the prone position. A noninvasive skin-based tracker is placed overlying the surgical site. Once syncing and trajectory planning are complete, a Jamshidi needle is advanced in the correct trajectory of the pedicle. A guidewire is then placed through the Jamshidi needle, ensuring that a bony end point is felt. The pedicle is then tapped and the screw is placed over a guidewire under radiographic guidance. In all cases, final anteroposterior and lateral fluoroscopy or x-ray is taken to ensure that all screws are placed appropriately.

In all cases, somatosensory evoked potentials and electromyographic monitoring were used alongside screw stimulation to detect proximity to neural structures. Screws were subsequently adjusted if necessary. We did not routinely perform intraoperative 3D scans to confirm screw placement before case completion in either cohort.

The accuracy of screw placement was assessed via postoperative CT scan, which was routinely obtained for all patients at approximately 1-year follow-up to assess fusion, with blinding to the surgical methods used. This was performed by physicians experienced in the assessment of orthopedic hardware placement on advanced imaging (T.C., A.V.). Screws were assessed for the presence or absence of facet joint violations, endplate breaches, and the Gertzbein-Robbins classification for cortical breaches. (The Gertzbein-Robbins grading is grade 0, no cortical breach; grade I, 0–2 mm breach; grade II, 2–4 mm breach; grade III, 4–6 mm breach; and Grade IV, >6 mm breach [15].) This was assessed using axial CT cuts. Breach distance was assessed at the pedicle for pedicle breach and at the vertebral body for tip breaches. Maximum breach was measured from the edge of the cortex in a perpendicular fashion to the outermost edge of screw threads (Fig. 1). For inferior pedicle breaches, where axial imaging does not adequately identify breach distance, the maximum breach distance was measured using coronal or sagittal reconstruction cuts. Based on all of these data, a grade was assigned to each screw based on the Simplified Screw Accuracy (SSA) grade (good, acceptable, or poor), an assessment the study authors devised a priori based on consensus opinion (Table 2).

Fig. 1.

Graphical representation of the radiographic safe zone as seen on postoperative computed tomographic imaging (a). Examples of a right-sided facet violation (b) and a grade III lateral pedicle cortex breach (c) are shown.

Table 2.

Simplified Screw Accuracy grade.

Score	Description
Good	No tip, endplate, pedicle, or facet breach
Acceptable	Pedicle breach within the radiographic “safe zone” of up to 4 mm superior/lateral or 2 mm inferior/medial, or any distance of tip breach [16,17]
Poor	Facet violation affecting the superior unfused level, any breach distance outside of safe zone, or endplate breach into unfused level

To assess the effect of coronal facet joint orientation on 2D fluoroscopic screw placement accuracy, axial facet angles were measured at each screw-associated facet joint in our 2D fluoroscopy cohort. This analysis was determined a priori to our 2D versus 3D screw accuracy analysis. Axial facet angles were measured, with blinding, using axial cuts of a preoperative CT scan or magnetic resonance image (MRI), by a physician experienced in assessment of orthopedic hardware placement on advanced imaging (T.C.). When preoperative imaging was unavailable, a postoperative CT scan was used, although care was taken to exclude facets that were debrided or remodeled. This was performed at all lumbar levels involved in the instrumentation construct of each patient. Axial facet angle (Fig. 2) was measured against a midsagittal line that was drawn perpendicular to the posterior cortex of the vertebral body and bisecting the vertebral body symmetrically. A tangential line was then drawn against the cortical margins of the facet joint of the superior articular process to form the angle. Care was taken to account for any facet overgrowth, osteophytes, or distorted anatomy of the associated inferior articular process when determining the orientation of the facet joint.

Fig. 2.

Measurement of the axial facet angle (a)—the angle formed between the tangential line to the facet joint and a midsagittal line.

Statistical Analysis

Descriptive statistics were summarized as “mean + SD” and “median [interquartile range]” for normally and non-normally distributed continuous variables, respectively, and as “n (%)” for categorical variables. Independent-samples t test was used for comparison of independent means, Mann-Whitney U test for comparison of multiple medians of nonparametric continuous variables, and χ² test for categorical variables. To ascertain whether factors such as demographics, preoperative radiographic diagnoses, level of surgery, and type of screw guidance (fluoroscopic or 3D navigation) affect screw placement, ordinal logistic regression was conducted with the above factors as independent variables and screw placement grade (good, acceptable, and poor) as the dependent variable. A significance level of .05 was used. Analyses were performed using the IBM Statistical Package for the Social Sciences (SPSS) version 22 (IBM Corp., Armonk, NY).

Results

Screw placement analysis revealed a significant difference in the rate of facet violation: 9% for the fluoroscopy group and 0.5% for the 3D navigation group (P < .0001). Good screw placement by SSA classification (no tip, endplate, pedicle, or facet breach) was found in 72.3% of fluoroscopic-guided screws and 75.6% of 3D-navigated screws (Table 3). Screws were fully contained in pedicles in 89.6% of cases in fluoroscopic-guided screws and in 87.6% in 3D-navigated screws. Only 1 instance of endplate violation was found in the 2 cohorts. No instances of superior pedicle breach were found. There was no statistical difference in the rate and grade of lateral breach, inferomedial breach, or tip breach. Interestingly, although there were no cases in which the immediate postoperative lateral lumbar spine radiograph or fluoroscopic image demonstrated any amount of tip breach, 62 total tip breaches were identified on postoperative CT. This represents an 11% false-negative rate in identifying tip breaches using postoperative lateral x-ray or fluoroscopy.

Table 3.

Screw placement accuracy.

	Fluoroscopy	Navigation	P value
No. of screws	376	193
Level of screw placement
L1	0 (0.0%)	2 (1.0%)
L2	0 (0.0%)	10 (5.2%)
L3	4 (1.1%)	12 (6.2%)
L4	117 (31.1%)	57 (29.5%)
L5	172 (45.7%)	73 (37.8%)
S1	83 (22.1%)	39 (20.2%)
Lateral breach			.627
None	347 (92.3%)	175 (90.7%)
Grade I	18 (4.8%)	12 (6.2%)
Grade II	7 (1.9%)	2 (1.0%)
Grade III	3 (0.8%)	2 (1.0%)
Grade IV	1 (0.3%)	2 (1.0%)
Inferomedial breach			.215
None	366 (97.3%)	187 (96.9%)
Grade I	9 (2.4%)	4 (2.1%)
Grade II	0 (0.0%)	2 (1.0%)
Grade III	0 (0.0%)	0 (0.0%)
Grade IV	1 (0.3%)	0 (0.0%)
Tip breach			.229
None	337 (89.6%)	170 (88.1%)
Grade I	13 (3.5%)	5 (2.6%)
Grade II	15 (4.0%)	6 (3.1%)
Grade III	8 (2.1%)	11 (5.7%)
Grade IV	3 (0.8%)	1 (0.5%)
Endplate violation	1 (0.3%)	0 (0.0%)
Facet violation	34 (9.0%)	1 (0.5%)	<.0001
SSA grade			.022
Good	272 (72.3%)	146 (75.6%)
Acceptable	66 (17.6%)	40 (20.7%)
Poor	38 (10.1%)	7 (3.6%)

SSA simplified screw accuracy. Significance for the bold values are P < 0.05.

An ordinal logistic regression was performed to identify whether demographic and radiographic variables significantly influenced screw placement accuracy, as graded by the SSA. Of all the variables examined, only S1 screw placement (with L5 as reference) and 3D navigation (with fluoroscopy as reference) were independently correlated with better screw placement (Supplemental Table 1). Regression analysis showed that compared with 3D navigation, fluoroscopy had twice the odds of incurring poor screw placement, as defined by the SSA.

Because screw accuracy results suggest that facet joint violation represents a significant portion of screws in the “poor placement” category, we performed a post hoc subgroup analysis by repeating the comparative statistics after excluding facet breach screws from both cohorts. Doing so nullified all statistical differences in screw accuracy between the 2 experimental groups (Supplemental Table 2). In fact, with facet violations removed, there were only 11 (1.9%) screws with poor SSA grade in a total of 569 screws examined. These comprise all cases of severe pedicle breach in our study.

We investigated facet violations by studying facet joint orientation in the 2D fluoroscopy cohort. The mean facet angles for these patients were calculated to be 36.8° at L4, 45.8° at L5, and 50.5° at S1. Of the 353 screws, 35 (9.9%) were categorized as poor placement by SSA. Note that fewer screws were included compared with the total of 376 screws, as some adjacent facet joint axial angles could not be measured (due to reasons such as surgical debridement or deformity). The mean axial facet angles associated with poorly placed screws, defined by SSA, were 42.7°at L4 and 51.4° at L5. These differences from the overall means were statistically significant for L5 (P = .013) but not L4 (P = .063). Given this finding, and the findings from our subgroup analysis above, another subgroup analysis was performed to identify the relationship between facet angle and the rate of facet violation. The mean axial facet angles associated with facet violation were 44.0° at L4 and 53.4° at L5. These means were also significantly higher than the overall means at L4 (P = .027) and L5 (P = .021). No poor screw placement was found at the S1 level, and thus S1 was excluded from statistical analysis. Overall facet violation rates were 12.2% at L4, 10.6% at L5, and none at S1.

To identify a useful threshold axial facet angle that correlates to an increased risk of facet breach, we performed a receiver operating characteristic (ROC) analysis at L4 and L5. This was performed under the assumption that a higher axial facet angle predisposes to a higher risk of facet breach. L4 and L5 were analyzed separately to control for differences in anatomy. The ROC analysis for L4 identified an ideal cutoff (Youden index) for the axial facet angle of 45° as a clinically useful value, with a specificity of 83% and a sensitivity of 63% for detecting a facet violation (P = .004; Fig. 3). The ROC analysis for L5 identified a cutoff for the axial facet angle of 60°, with a specificity of 92% and a sensitivity of 67% for detecting a facet violation (P = .048).

Fig. 3.

L4 and L5 receiver operating characteristic (ROC) curves, demonstrating the sensitivity and specificity in identifying facet breach by use of cutoffs (black arrows) in axial facet angle. For L4, P = .004, area under the curve = 0.736 (SE = 0.074), and point of optimum cutoff for facet angle (Youden index) for facet breach is 45° (Sp = 0.83, Sn = 0.63). For L5, P = .019, area under the curve = 0.671 (SE = 0.0.067), and point of optimum cutoff for facet angle (Youden index) for facet breach is 56° (Sp = 0.85, Sn = 0.44). Sp specificity; Sn sensitivity

Finally, a post hoc review of clinical outcomes was also performed for all cases of severe pedicle breach (outside of the defined radiographic safe zone) and for all cases of inferomedial pedicle breach (regardless of the breach distance). This was performed to assess the clinical severity of pedicle breach in MIS screw placement. Electronic charts were reviewed to obtain an anecdotal account of clinical progress and revision surgery after the index procedure. An assessment was made as to whether postoperative clinical symptoms correlate with the anatomical pedicle breach. Of the 19 patients (24 screws; Supplemental Table 3) identified using these criteria, only 4 patients demonstrated postoperative persistent radiculopathy that correlate with the anatomical location of the breach. Of these cases, 2 were from the fluoroscopy cohort and 2 were from the navigation cohort. One 2D fluoroscopy-guided case underwent reoperation on postoperative day 7 for unilateral hardware removal for radiculitis. One 3D navigation case underwent inferior extension of fusion 5 months later, due to ASD and hardware removal on the side with a known facet violation. Both cases of reoperation resulted in improvement in symptoms. The remaining 2 cases had mild symptoms, so the patients did not undergo revision surgery.

Discussion

This retrospective cohort study was designed to identify differences in facet violation rates, as well as overall accuracy, between 2D fluoroscopy and 3D navigation–guided percutaneous pedicle screws placed in the lumbosacral spine. To our knowledge, this is one of the first attempts to compare 3D navigation with 2D fluoroscopic guidance, specifically in the context of MIS. Our findings suggest that 3D navigation significantly and independently protects against facet violation. This is consistent with prior studies that suggest 2D percutaneous placement may result in rates of facet violation as high as 58% [7,11,17]. Our study found that of the 35 (9.9%) fluoroscopically placed pedicle screws that were categorized as poor placement, 32 were due to facet violation.

Limitations of this study include the following. First, although this study benefits from the consistency of being a single-surgeon series, this also limits the study’s generalizability to other MIS surgical protocols or navigation methods. Second, from 2011 to 2017, adoption of 3D navigation occurred over time, with all fluoroscopic cases performed early in this period at one academic institution and almost all 3D navigation cases performed later, at a different academic institution. This systematic difference could explain the differences between the cohorts (such as ASA status) but may also introduce systematic error from the sequential nature of the study. Third, it is possible that experience gained from years of performing fluoroscopic percutaneous screw placement could have positively influenced the accuracy of pedicle screws placed by 3D navigation. To control for this, both cohorts in our study included only those patients who underwent procedures after the operating surgeon had sufficiently surpassed the learning curve of the respective surgical method. Finally, in establishing the SSA, although literature exists supporting the radiographic safe zone defined in this study, anatomical variation alone should disallow any notion of a “safe” pedicle breach. As surgeons, we should aim to have no breach at all in the use of pedicle screws. In the face of maturing navigation technology and as a stepping stone toward the continual improvement of screw accuracy, the SSA grade can help to compartmentalize data and allow for meaningful comparisons.

The rate of poor screw placement in this study is similar to that of previous reports [2,17]. Baird et al suggested that the most common form of poor pedicle screw placement is superior articular facet violation [2]. Removal of cases of facet violation from our analysis accordingly removed the statistical difference in screw placement accuracy between 2D fluoroscopy and 3D navigation. Specifically, pedicle breach rates between the 2 groups in our study were comparable, despite prior studies demonstrating that 3D navigation may improve pedicle breach rates [5].

In the literature, facet violation has been associated with increased BMI [9], right-sided screws, and—most relevant to our work—facet angles greater than 45° [17]. Our results also show that an axial facet angle greater than 45° at L4 and 60° at L5 corresponds to an increased risk of facet breach based on ROC analysis. It should be noted that despite use of the safer lateral-to-medial technique for fluoroscopy-guided screw placement [16], the much larger screw diameter compared with that of the Jamshidi needle may risk facet violation regardless. We found that facet violation in our 2D fluoroscopy cohort was associated with higher axial facet angles: 44.0° versus 36.8° at L4, and 53.4° versus 45.8° at L5. Therefore, the advantages afforded by 3D navigation appear particularly relevant for highly angled (coronally oriented) facets.

We also performed a limited clinical review of select cases that suffered severe pedicle breach or inferomedial pedicle breach. Of a total of 19 patients examined (24 screws), only 4 cases were identified in which a pedicle breach anatomically correlated with subsequent radicular symptoms. Of these, 2 patients underwent reoperation (1 in the fluoroscopy group and 1 in the navigation group) and had symptomatic hardware removed. Only 1 of the 4 patients had a persistent mild motor deficit at final follow-up. Although this finding attests to the safety of MIS instrumentation, a long-term clinical study would be better suited to make this determination.

In conclusion, this retrospective cohort study suggests that as an intraoperative technology, 3D navigation may improve percutaneous pedicle screw trajectories so as to avoid facet violation, when compared with 2D fluoroscopic guidance. It also suggests that highly angled lumbar facets (greater than 45° at L4 and 60° at L5) may be a risk factor for facet joint violation, and may advocate for using 3D over 2D screw placement methods in MIS surgery. More rigorous study is warranted.