Comparative analysis of clinical factors associated with pedicle screw pull-out between intraoperative cone-beam computed tomography and postoperative computed tomography

No studies to date have elucidated the clinical factors associated with pedicle screw pull-out. The aim of this study was to assess the frequency of pedicle screw pull-out by comparing intraoperative scans obtained using cone-beam computed tomography (CBCT) with scans acquired postoperatively using computed tomography (CT). We also sought to determine the incidence of pedicle screw pull-out and identify relevant risk factors. This retrospective analysis of prospectively collected data for 636 in 66 who compared between The associated logistic

Nowadays, many institutions can perform spinal instrumentation surgery using intraoperative cone beam computed tomography (CBCT) in a hybrid operating room [6,7]or an O-arm imaging system (Medtronic, Minneapolis, MN) [8,9]. CBCT is a three-dimensional (3D) imaging tool that reconstructs projection data by a rotational C-arm with a at panel detector [7]. It can visualize low-contrast objects, such as soft tissue or small vessels, as well as high-contrast structures, including enhanced vessels or bone. Several authors have reported the utility of CBCT [6,7], and thus it is being increasingly used in spine surgeries. The recent advent of a 3D-CT-based navigation system for intraoperative CBCT has improved the accuracy of pedicle screw insertion. Further, intraoperative CBCT has made it possible to con rm the presence or absence of screw deviation during surgery.
Long instrumented fusion requires insertion of a large number of pedicle screws, and the abovementioned intraoperative modalities help to ensure surgical safety. The position of the pedicle screws can be con rmed using intraoperative CBCT in a hybrid operating room after pedicle screw insertion before rod connection. We have encountered several cases of intraoperative pedicle screw pull-out as a result of rod connection during posterior thoracic or lumbar spine surgery. In long instrumented fusion especially, pedicle screw pull-out was found to be a risk factor for pedicle screw loosening [10]. It is reported that pedicle screw pull-out can occur during, but not after, surgery [10,11].
A previous study investigated the percutaneous pedicle screw (PPS) pull-out during rod reduction and reported its association with screw loosening. However, the clinical factors associated with pedicle screw pull-out have not been elucidated.
Hence, in this study, we investigated the frequency of pedicle screw pull-out by comparing intraoperative CBCT images and postoperative CT images acquired 2 days after surgery. We also sought to identify and evaluate the relevant risk factors for screw pull-out.

Materials And Methods
This study was approved by the Institutional Ethics Committee (Approval No. R201904-06). In this study, we conducted a retrospective analysis of prospectively collected data from 66 consecutive patients who underwent long posterior spinal instrumented xation at 3 levels or more for thoracic or lumbar spinal injury, spinal metastasis, or pyogenic spondylitis between April 2014 and March 2020. The patients were evaluated using CBCT intraoperatively in a hybrid operating room. Evaluation by CT was performed again on postoperative day 2 before mobilization. Demographic and surgical data collected included age, sex, BMI, operating time, estimated blood loss, underlying disease, number of fused segments, use of a Hook system, PPS insertion, presence of diffuse idiopathic skeletal hyperostosis (DISH), medical history of osteoporosis, smoking, type of health insurance cover, and preoperative comorbidities.
Surgical procedure All patients underwent posterior spinal xation in the prone position with the trunk on a radiolucent operating table in a hybrid operating room. Following anesthesia, standard surgical exposure was performed via a midline skin incision using an open approach. PPS placement was performed via a 1.5cm stab incision made laterally. The targeting needle was inserted into the pedicle at the superolateral border based on the anterior-posterior view under rotational C-arm uoroscopic guidance. The guidewire was inserted through the targeting device and into the pedicle. The pedicle was then tapped using the guidewire. Next, the pedicle screw was inserted over the guidewire. CBCT images were obtained at this timing to con rm the screw position. Finally, rods were installed aligned to each screw placed in the open surgery and were inserted starting from the most cranial skin incision in the PPS surgery. The order in which the set-screws were installed varied among the operators (Fig. 1). A Solera Voyager spinal system (Medtronic, Memphis, TN) or Reline spinal system (NuVasive, San Diego, CA) was used for posterior spinal xation.

Evaluation of screw position
In all cases, the implant position was assessed intraoperatively using CBCT scans obtained using a DynaCT system (Artis Zeego; Siemens Healthcare, Erlangen, Germany), a 3D imaging tool that reconstructs projection data obtained by a rotational C-arm with a at panel detector. The series consisted of 0.616-mm CBCT sections reconstructed at 0.616-mm intervals. Raw data were used to reconstruct axial 2.0 mm-thick CBCT sections every 2.0 mm with a eld of view adequate for visualization of the spine, as well as sagittal and coronal reformat images of the thoracolumbar spine.
The position of the implant was assessed postoperatively on CT scans obtained using a 320-row areadetector CT system (Toshiba Medical Systems, Tokyo, Japan). The series consisted of 0.5-mm-thick CT sections reconstructed at 0.5-mm intervals. Raw data were reconstructed with 2.0 mm-thick axial CT sections every 2.0 mm with a eld of view that was adequate for spine visualization and for sagittal and coronal reconstruction of the thoracolumbar spine.
Screw misplacement was classi ed according to the system devised by Schizas et al. [12]. Screw malposition was categorized as minor (< 3 mm), moderate (3-6 mm), or severe (> 6 mm) and the direction of perforation was classi ed as medial or lateral. Screw pull-out distances measured in the axial and sagittal planes on CT scans obtained 2 days postoperatively were compared with those on intraoperative CBCT images by two independent observers. Screw pull-out was de ned as a distance of more than 4 mm change in the axial and sagittal views on postoperative CT compared with the position on the intraoperative CBCT views (Fig. 2). Data were also compared between the cases with screw pullout and cases without screw pull-out. Risk factors for pedicle screw pull-out were identi ed by multivariate analysis.

Statistical analysis
Statistical analysis was performed using Fisher's exact test and the Mann-Whitney U test, with statistical signi cance set at P < 0.05. Risk factors were determined using logistic regression analysis with a forward stepwise procedure (P < 0.1 for entry). The threshold of alpha and forward selection was chosen to stabilize the statistical model during stepwise covariate selection considering the relatively small number of our outcomes. Inter-observer agreement was measured using kappa coe cient scores. Any discrepancy between the observers was resolved by discussion. All statistical analysis was performed using JMP version 13.1.0 software (SAS Institute Inc., Cary, NC).

Results
A total of 636 pedicle screws were inserted in the thoracic or lumbar spine of the 66 patients. Patient demographics are shown in Table 1. Mean age of the 40 men and 26 women was 68.6 ± 14.8 years (range, 25-93 years) and mean BMI was 22.2 ± 3.3. Mean operating time was 174.9 ± 83.9 min and mean estimated blood loss was 219.5 ± 252.2 g. The diagnosis was spinal injury in 47 cases (71.2%), spinal metastasis in 14 (21.2%), and pyogenic spondylitis in 5 (7.6%). Mean number of fused segments was 4.8 ± 1.3. A hook system was used in 5 cases (7.6%) and the PPS method in 47 cases (71.2%). Eleven patients (16.7%) were smokers. Other diagnoses included DISH (n = 22, 33.3%), osteoporosis (n = 27, 40.9%), rheumatoid arthritis (n = 2, 3.0%), diabetes (n = 13, 19.7%), and asthma (n = 6, 9.1%). Two patients (3.0%) were on dialysis. Twenty-two patients (33.3%) were covered by national health insurance, 10 by employees' pension insurance (15.2%), 25 (37.9%) by the advanced elderly medical service system, 7 (10.6%) by workers' accident compensation insurance, and 2 (3.0%) by livelihood subsidies public assistance.  Table 2. Twenty-eight of the pedicle screws (4.4%) were found to be misplaced on the postoperative CT views. Sub-classi cation analysis revealed minor perforation of 14 screws (2.2%), moderate perforation of 8 screws (1.3%), and severe perforation of 6 screws (0.9%). Both pull-out and misplacement occurred with 2 screws (0.3%). Pull-out screws were observed in the upper instrumented vertebra in 3 cases (17.6%), in the lower instrumented vertebra in 6 (35.3%), and in the inter-levels in 8 (47.1%; Fig. 3). There was substantial inter-observer agreement in judging screw pull-out distance (κ = 0.79). All values indicated substantial agreement. There was no signi cant difference in sex, BMI, operating time, estimated blood loss, underlying disease, surgical procedure, or smoking status between the non-pull-out and pull-out group. Patients in the pull-out group were more likely to be elderly, have more fused segments, have a diagnosis of DISH or osteoporosis, and have advanced elderly medical service system insurance cover (Table 3).  The risk factors for pedicle screw pull-out were evaluated using logistic regression analysis. From univariate analysis, the dependent variable was de ned as the presence of screw pull-out and the independent variables were age, number of fused segments, DISH, osteoporosis, advanced elderly medical service system insurance, estimated blood loss and PPS procedure. As a result, the independent risk factors identi ed were older age (odds ratio 1.15, 95% con dence interval 1.04-1.272, P = 0.007) and DISH (odds ratio 6.10, 95% con dence interval 1.23-30.20, P = 0.027; Table 4).

Discussion
Previous reports have shown that pedicle screw loosening after surgery is a serious complication of spinal xation surgery. Screw loosening causes nonunion back pain and sometimes neurological impairment, and it can be an indication for reoperation [13,14]. Ohba et al. reported that pedicle screw pull-out was a risk factor for postoperative screw loosening [10]. In the current study, we investigated the incidence of screw pull-out and attempted to identify relevant risk factors.
In our study, pull-out occurred for a total of 17 pedicle screws in 13 cases, giving an overall pedicle screw pull-out rate of 2.7%, which is relatively low compared with the rate of 16.2% reported previously [10].
Although screw pull-out has been recognized as a distance of > 1 mm in previous studies, we de ned it as ≥ 4 mm in our study. Considering that one pitch of the screw is 4 mm for the Voyager spinal system and 2.7 mm for the Reline spinal system, it seems reasonable to de ne pull-out as 1 or more pitches for the Voyager screw and as 2 or more pitches for the Reline screw. Thus, the inter-observer agreement was high (κ = 0.79) in this study. The difference in the cutoff value may explain why the incidence of screw pull-out was lower in this study than in previous investigations. Pedicle screw pull-out was detected in 19.7% of the patients in this study. Of note, all the postoperative CT images were obtained on postoperative day 2 before the patients started to mobilize, meaning that screw pull-out is most likely to occur during the operation. Other studies also mentioned that pedicle screw pull-out occurs during rod connection [10,11]. Generally, screw pull-out can occur if there is a gap between the shape of the rod and the actual spinal alignment. In the PPS system especially, pull-out may easily occur at the time of inserting set screws because the gap between the rod and the screw head is not visible [10,11]. Furthermore, we found that pedicle screw pull-out could occur at any of the levels including the cranial end, caudal end and interlevels. The force in the direction in which the screw comes off is considered to vary according to the order in which the set screws are placed.
In this study, patients in the pull-out group were more likely to be elderly, to have a diagnosis of DISH or osteoporosis, and to have advanced elderly medical service system insurance cover. Other researchers have also identi ed that older age and osteoporosis are risk factors for pedicle screw loosening [13,15]. Advanced elderly medical service system insurance is available for citizens over 75 years of age only, meaning that our patients with this type of insurance cover were indeed elderly. The number of fused segments was also associated with screw pull-out. In long fusion, force may be applied to the direction in which the rod does not t the screw head and the force on the lever arm of the rod is increased, thereby increasing the risk of pedicle screw pull-out compared with short fusion. Additionally, in long spinal fusion, there are multiple screws to be connected to the rod, making it di cult to achieve appropriate rodbending and to t the rod completely to each screw head.
Logistic regression analysis identi ed advanced age and presence of DISH to be independent risk factors for pedicle screw pull-out. DISH appears on radiographs and CT images as ossi cation along the anterolateral aspect of the vertebral bodies [16,17]. Therefore, movement of the spine becomes limited by spinal ossi cation. Given that DISH restricts the segmental motion of the spine, the screw-rod system applies force in the direction in which the screw comes off if the fused segments were over-corrected by de-kyphotic rod placement. Additionally, although there was no signi cant difference in DISH in this study, DISH is often performed PPS surgery, as it is considered that the screw comes off because the tting between the rod and the screw head cannot be checked directly. Patients with DISH also tend to be elderly and have poor bone quality [18][19][20]. Therefore, posterior segmental fusion extending at least three levels above and below has been recommended in patients with DISH [21]. To our knowledge, there have been no reports showing that the presence of DISH is associated with pedicle screw pull-out, and ours is the rst to clearly demonstrate that DISH is a signi cant risk factor for intraoperative screw pull-out with an odds ratio of 6.10.
The most important factor is to ensure appropriate rod bending and gently connecting it to each screw head. As it is sometimes di cult to create a perfect curve by manual bending, a new automatic rod bending system may be useful to improve the tting of the rod [22]. Also, a new pedicle screw device using cement augmentation has recently become available as a strategy to prevent pull-out [23].
There are several limitations in this study. First, we did not evaluate the screw design, position, or trajectory as factors affecting pull-out. Second, radiation exposure is a risk for patients, although we used intraoperative CBCT, in which the radiation exposure is reduced.

Declarations
The data and materials may be made available upon request through sending e-mail to rst author.