Incidence and Predictive Factors of Distal Pedicle Screw Loosening Followed Posterior Corrective Surgery for Degenerative Lumbar Scoliosis

The aim of this study was to identify the prevalence of distal pedicle screw loosening (DPSL) following posterior decompression and instrumented fusion for DLS, to search for predictable risk factors for DPSL, and hope to provide references in decision making and surgical planning for spine surgeons. One hundred and fty-three consecutive DLS patients surgically 2018 were retrospectively reviewed. The diagnostic criteria for DPSL developed by X-ray include the radiolucent area (thicker than 1 mm) around screw and the “double halo” sign dened as the presence of radiolucent area and radiopaque rim at the same X-ray. According to the occurrence of DPSL at two-year follow-up, patients were divided into two groups: study group (screw loosening) and control group (without screw loosening). To investigate predictive factors for distal pedicle screw loosening, three categorized factors were analyzed statistically.

compression and distraction [3]. Screws loosening is one of the frequently reported typical complication of pedicle screws xation, and usually occurs in the cranial and caudal vertebrae of the long fusion, much more frequent in caudal screws [4]. Screw loosening may lead to pseudarthrosis, nonunion, loss of correction, revision surgery, and therefore need to be paid speci c attention [5]. The frequency of screw loosening has varied widely in different studies, which ranged from less than 1% to 15 % in nonosteoporotic patients and even up to 60 % in osteoporotic spines, depending on the study population and the duration of follow-up [6][7][8]. Knowledge of risk factors is very important for minimizing the occurrence of screw loosening and to allow surgeons to take measures for its prevention when possible.
Bone mineral density (BMD) is supposed to be an important factor in uencing the stability of pedicle screws [9]. Wittenberg et al. proved that early loosening of pedicle screws may be expected at a BMD less than 90 mg/mL, as measured by quantitative computed tomography [10]. The majority of studies were focused on impact of BMD on the occurrence of screws loosening, however, it has been proven that bone quality is not the only contributing factor [8,11,12]. Several other factors affecting the stability of pedicle screws, such as the length, outer diameter, design, tness in the pedicle, elasticity of the cancellous bone, were supposed to potentially in uence the development of screw loosening [13][14][15]. Although previous reports have alerted spine surgeons about screw loosening subsequent to posterior decompression and instrumented fusion, very few published reports describing speci c risk factors that need to be taken into account to avoid this undesirable event for DLS patients.
The aim of this study was therefore to identify the prevalence of distal pedicle screw loosening (DPSL) following posterior decompression and instrumented fusion for DLS, to search for predictable risk factors for DPSL, and hope to provide references in decision making and surgical planning for spine surgeons.

Patients
A retrospective study was conducted, it was approved by the Institutional Review Board of Peking University Third Hospital before data collection and analysis. One hundred and fty-three consecutive DLS patients surgically treated between January 2013 and July 2018 were retrospectively reviewed.
Inclusion criteria: 1. DLS patients treated by posterior instrumented fusion with pedicle screws. 2. followup duration more than two years with radiological data including antero-posterior (A/P) and lateral X-ray at preoperation, early postoperation, and two-year follow-up. Exclusion criteria: 1. patients with spinal deformities derived from idiopathic scoliosis, ankylosing spondylitis, neuromuscular diseases, infections, or scheuermann kyphosis. 2. patients treated with anterior or lateral lumbar fusion surgery, minimally invasive lumbar fusion surgery. 3. posterior instrumentation include iliac screw (IS) or sacral-alar-iliac (S2AI).

Clinical and radiological evaluation
The diagnostic criteria for DPSL developed by X-ray include the radiolucent area (thicker than 1 mm) around screw and the "double halo" sign de ned as the presence of radiolucent area and radiopaque rim at the same X-ray [16,17]. ( gure 1) The rst three authors resolved all ambiguities by consensus.
According to the occurrence of DPSL at two-year follow-up, patients were divided into two groups: study group (screw loosening) and control group (without screw loosening). To investigate predictive factors for distal pedicle screw loosening, three categorized factors were analyzed statistically.
Patient general data including age, sex, body mass index (BMI), BMD, comorbidity (hypertension, diabetes, coronary heart disease) were investigated. BMD was evaluated by Houns eld unit (HU) value on L1 from computed tomography (CT), which was performed via a previously described technique using Philips Easy Vision Picture Archiving and Communication System (PACS) [18]. This protocol records HU on axial images at three levels within the vertebral body: immediately caudal to the superior end plate, mid-body, and cranial to the inferior end plate. ( gure 2) For each measurement, the largest possible elliptical region of interest was drawn, excluding the cortical margins to prevent volume averaging.
Surgical data including surgery time, estimated blood loss, fusion level, lower instrumented vertebrae (LIV), Posterolateral fusion were reviewed.
Radiological data were collected and evaluated at pre-operation, early post-operation, two years postoperation. Cobb's angle was measured using lines projected from the upper border of the most tilted vertebra above and lower border of the most tilted vertebra within the main curve. Lumbosacral coronal angle (LSCA) was measured using lines projected from the upper border of L4 and upper border of S1 on P/A X-ray. Coronal balance distance (CBD) was measured as the distance between C7 plumb line and center sacral vertical line. Lumbar lordosis (LL) was measured from T12 inferior endplate to S1 superior endplate by the Cobb method on lateral X-ray. Thoracolumbar junction (TL) was measured from T10 superior endplate to L2 inferior endplate on lateral X-ray. Pelvic incidence (PI) was de ned as the angle between the line perpendicular to the sacral plate and the line connecting the midpoint of the sacral plate to the bicoxofemoral axis. Sacral slope (SS) was the angle between the S1 superior endplate and a horizontal line. Sagittal vertical axis (SVA) was the distance from C7 plumb line to the perpendicular line drawn from superior posterior endplate of S1. ( gure 3)

Statistical analysis
Data were analyzed using Statistical Product and Service Solutions software (version 17; SPSS, Chicago, IL). Continuous variables were measured as mean ± standard deviation, and categorical variables were expressed as frequency or percentages. An independent t test was used to analyze the difference of continuous variables between two groups. An χ2 analysis and Fisher's exact test were used to examine the differences among categorical variables. Binary Logistic regression analysis was used to analyze the assumed predictive factors with backward elimination, in which variables with a signi cance level of >.10 were removed. The con dence interval of the odds ratio (OR) was 95%.

Results
DPSL was detected in 72 of 153 patients (47.1%) at two-year follow up, and were enrolled as study group. Another 81 patients presented no screw loosening at two-year follow up, and were enrolled as control group. ( gure 4 and 5) There was no statistically signi cant difference between the two groups in age at operation, sex, BMI, comorbidity. The HU value on L1 was lower in study group than that in control group. (Table 1) There was no statistically signi cant difference between the two groups in mean surgery time, estimated blood loss. Fusion level was longer in study group than that in control group. The LIV on L5 was less in study group than that in control group. Posterolateral fusion was less in study group than that in control group. (Table 2) The preoperative Cobb angle, postoperative Cobb angle, correction of Cobb angle, preoperative LSCA, correction of LSCA, preoperative TL, postoperative TL were larger in study group than those in control group. There was no statistically signi cant difference in postoperative LSCA, preoperative CBD, postoperative CBD, correction of CBD, preoperative LL, postoperative LL, correction of LL, correction of TL, PI, preoperative SS, postoperative SS, correction of SS, preoperative SVA, postoperative SVA, correction of SVA between the two groups. (Table 3) The

Discussion
In the current study, 47.1% of the patients (72/153) presented DPSL followed posterior decompression and instrumented fusion for DLS, we found that low BMD, large correction of both main curve and fractional curve, posterolateral fusion were signi cantly and independently associated with the occurrence of DPSL, and can be assessed before surgery, these results were not confounded by other variables potentially affect DPSL.
A critical point for the investigation of screw loosening is the assessment of whether a screw is loosened or not, planar radiographs are most commonly used, whereas a signi cant minority of papers employed clinical CT or low-dose CT [19][20][21]. Potential health risks associated with increased radiation exposure as well as nancial costs limit postoperative CT use to selected patients with persistent symptoms or where there is speci c concern regarding the possibility of implant-related complications [5]. Wu et al. proved that CT might not be superior to X-ray in assessment of screw loosening, especially considering its higher cost and radiation exposure, as X-ray had a sensitivity of 24% and a speci city of 98%, while CT scan had a sensitivity of 22% and a speci city of 95% regarding extraction torque as criterion of screw loosening [22]. Sanden et al. reported that a radiolucent zone around a PS is a good indicator of screw loosening [19]. That is the reason why we chose plain radiographs to evaluate pedicle screw loosening in the present study.
Since its rst description by Watkins in 1953, the technique of posterolateral fusion (PLF) has evolved into the widely adopted method of lumbar arthrodesis for degenerative disease [23]. Tsirikos et al. proved that posterolateral arthrodesis in situ with autologous iliac crest bone without instrumentation achieved a solid fusion and produced excellent clinical outcomes and high patient satisfaction [24]. Moreover, the use of pedicle screw xation to support the PLF could achieve fusion rates as high as 96% [25][26][27]. Addition of interbody fusion via a transforaminal approach (TLIF) has become a popular surgical option for treatment of degenerative lumbar conditions. Although technically more complicated than PLF, it has been suggested that TLIF provides superior immediate stability and protects against early pedicle screw loosening. Kim et al proved that TLIF appears to have a protective effect in reducing rates of early screw loosening by approximately 60% versus PLF [28]. A retrospective study comparing the outcomes between TLIF and PLF found that the PLF construct was more prone to induce instrument breakage and pseudoarthrosis compared with the TLIF technique, indicating the importance of anterior column support in the maintenance of construct stability [29]. In recent years, TLIF has replaced PLF to be the most commonly used lumbar fusion surgery [30]. However, postoperative screw loosening was not completely avoided, especially in patients undergoing long segment pedicle screw xation [5]. Results from the current study suggest that supplementing the TLIF with PLF potentially reduces the rates of DPSL followed corrective surgery for DLS. Following decortication of transverse processes and posterolateral bone, morselized local autograft was placed in the posterolateral intertransverse space in order to create posterolateral arthrodesis, which could reduce the segmental motion of posterior spinal column, and minimize the risk of screw loosening.
It has been widely accepted that osteoporosis is a potential risk factor for pedicle screw loosening. Fixation failure rates due to screw loosening range from 1-15% and can rise in osteoporotic spines for which the strength of xation with pedicle screws decreases along with lower bone density [6][7][8]. In a study of osteoporotic cadavers, pedicle screw pullout strength was found to be highly correlated with BMD [31]. BMD measuring with DEXA scans could not perform differentiation of the cancellous bone from the cortical bone, T-scores can be overestimated due to abdominal vessel wall calci cation, degenerative bony spurs, and facet hypertrophy [32,33]. It has been proved that older patients or patients with degenerative lumbar scoliosis are more likely to have unreliable lumbar T-scores [34]. Measuring HU from computed tomography has been proposed as a useful technique for assessing vertebral bone quality, the reliability and accuracy of HU to determine osteoporosis have been proved in the literature with many reports [35][36][37][38]. In the current study, our ndings are consistent with Jan et al that the determination of bone density by HU value could predict the risk of screw loosening and inform the decision to use cement augmentation to reduce the incidence of screw loosening [39]. Screw augmentation with bone cement as well as expandable screws demonstrated improved screw pullout stability, better xation strength and clinical results in comparison with standard pedicle screws in human cadaver material in vitro [40][41][42][43].
The typical major curve of DLS always lies in the mid-lumbar spine, it is always compensated by a distal curve at the lumbosacral region as the body attempts to maintain coronal balance, named the fractional curve [44,45]. Both pedicle screws compression at the convex side and pedicle screws distraction at the concave side could achieve correction of the main curve and fractional curve [46,47]. No matter which technique is used, the cutting effect of pedicle screw on bone should not be ignored. The more scoliosis correction, the greater the cutting effect of pedicle screw on bone, and the greater the risk of pedicle screw loosening. Based on our previous study, we put forward two suggestions. Firstly, asymmetrical vertebral degeneration in DLS patients is manifested as high HU value within concavity and low HU value within convexity both in major and fractional curve. Concave side of the vertebrae theoretically withstand more cutting forces than convex side based on the HU value measurement, distraction of pedicle screws at concave side should be a priority to correct both main curve and fractional curve [48]. Secondly, inadequate release of the posterior spinal bone elements in PLIF or TLIF technique may hinder the correction of the fractional curve, because the lumbosacral junction tends to be particularly rigid and may already be fused into an abnormal position. Posterior column osteotomy facilitate effective correction of the scoliosis through complete release of dural sac as well as the asymmetrical intervertebral reconstruction by cage is recommended [49].
There are several potential limitations in this study. First, the number of patients is relatively small, and the study may be under powered to detect the signi cance of some risk factors. Second, the study was conducted retrospectively by case selection, and was not randomized and controlled, the subjects selected are all Chinese Han individuals, whether the conclusion is applicable to other ethnic groups needs to be further investigated in the future. Even with these issues, three suggestions are provided for surgical planning based on the ndings in the current study. Firstly, screw augmentation with bone cement or expandable screws is recommended for patients with low BMD measured from preoperative CT scan. Secondly, posterior column osteotomy plus unilateral cage strutting technique on the concavity is recommended for scoliosis correction, screw distraction instead of compression should be preferred. Thirdly, posterolateral fusion is recommended for DLS patients that underwent posterior long segment fusion.

Conclusion
The incidence of DPSL following posterior decompression and instrumented fusion for DLS is 47.1%. Low BMD, large correction of both main curve and fractional curve are predictive factors for DPSL, posterolateral fusion is a protective factor.

Declarations
Ethics approval and consent to participate: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. For this type of study, the informed consent was waived. This study was approved by Peking University Third Hospital Ethics Committee.

Consent for publication: Not applicable
Availability of data: The data used and analyzed during the current study was available from the corresponding author on reasonable request.
Competing interests: The authors declare that they have no competing interests.

Funding: Not applicable
Authors' contributions: WS L conceived the project and analyzed the data. All authors contributed towards the interpretation and the collection of the data. All authors wrote and approved the nal manuscript.