The Role of Prophylactic Vertebroplasty at UIV+1 and Cement-augmented Fenestrated Pedicle Screw for Prevention of PJK/PJF in Patients With Osteoporosis: A Propensity Score–matched Analysis

Objective: To elucidate the role of prophylactic vertebroplasty (PV) at UIV+1 and cement-augmented fenestrated pedicle screw (CAFPS) in prevention of PJK and PJF. Background: Cement augmentation at UIV and UIV+1 was found to prevent PJK and PJF. But most studies are retrospective and have a selection bias due to multifactorial etiology of PJF, making it dicult to identify the ecacy of prophylactic cement augmentation. Methods: We enrolled 208 surgically treated adult spinal deformity (ASD) patients who were followed for at least 2 years, to elucidate whether prophylactic cement augmentation was truly minimizing the risk of PJK and PJF in a uniform population from one center database. Patients were classied into two groups. Two comparable groups were propensity-matched with one to one nearest neighbor matching. The main outcome variables including PJK, PJF, pedicle screw loosening and cement leakage were compared. Results (cid:0) After propensity score matching, there were 58 propensity-matched patients in group A and B (n=29 in each group) whose parameters including age, BMI, BMD, number of instrumented vertebrae, SVA, spinal-pelvic parameters and the frequency of UIV and LIV were similar. The incidence of PJK showed no signicantly statistical difference between group A and B (10.3% vs 13.8%, p >0.99). Compared to group A, group B had a higher proportion of patients developing PJF (24.1% vs 0%, p=0.01). Pedicle screw loosening at UIV and (or) LIV showed no difference (24.1% vs 3.4%, p=0.052) between two groups. In group A, cement leakage was detected in 5 cases via uoroscopy, but none of them had neurological decit or pulmonary cement embolism. Conclusion: Combined application of CAFPS and PV could reduce the incidence of PJF, but it could not prevent the development of PJK in the surgical management of ASD with low BMD or


Introduction
Proximal junctional kyphosis (PJK) and proximal junctional failure (PJF) has exerted great challenge on spine surgeons in the management of spinal deformity. Though there is still no consensus on the de nition of PJK, it has been proven that PJK is correlated with up back pain and may progress to PJF 1,2 . PJF may require revision surgery for proximal extension of fusion, which is de ned as vertebral fracture of uppermost instrumented vertebrae (UIV) or (and) vertebrae above UIV (UIV + 1), pull-out of instrumentation at UIV or (and) sagittal subluxation 3 . The incidence of PJK/PJF varies greatly due to multiple contributing factors including selection of the UIV and lowest instrumented vertebrae (LIV), spino-pelvic parameters, fusion levels, older age, body mass index (BMI), bone mineral density (BMD) and other elements 1,2 . In order to prevent the PJK and PJF after surgical correction of adult spinal deformity (ASD), some surgeons have resorted to the hook-only constructs, hybrid constructs or ligament augmentation 4,5 .
Low BMD or osteoporosis has been con rmed as an important risk factor of PJF 6 . In order to reduce the incidence of fracture of UIV/UIV + 1, screw loosening and screw pull-out in osteoporotic spine, cement augmentation at UIV and UIV + 1 have already been used [7][8][9][10][11][12] . However, there still exists controversy about whether cement augmentation can reduce the incidence of the PJK/PJF 5,10,13 . Since most studies concerning cement augmentation for prevention of PJK/PJF have a selection bias due to multifactorial etiology of PJK/PJF and the heterogeneity of ASD population between groups, it is di cult to identify the e cacy of prophylactic cement augmentation at UIV and UIV + 1.
Propensity score matching (PSM) statistical method is commonly used to remove confounding bias when comparing treatment effect in observational cohorts, in which randomization cannot be performed 14 . Therefore, we carried out the present study to elucidate the role of prophylactic vertebroplasty (PV) at UIV + 1 and cement-augmented fenestrated pedicle screw (CAFPS) in the prevention of PJK/PJF in a propensity score-matched patient cohort.

Subject enrollment
This was a retrospective analysis of data collected from the Department of Orthopedics in our hospital from January 2010 to January 2018. Study inclusion criteria were as the following. Firstly, the age of patients was above 50 years old at the time of surgery. Secondly, instrumented vertebra was over 5 levels, crossed the thoracolumbar junction and ended proximally in the thoracic spine (T11 or above). Thirdly, patients were diagnosed as adult spinal deformity with at least one of the following radiographic parameters: Cobbs' angle ≥ 20°, sagittal vertical axis (SVA) ≥ 5 cm, or pelvic tilt ≥ 25°. Fourthly, patients underwent posterior spinal fusion with bilateral pedicle screws and rods system (titanium alloy constructs) and were followed up for at least two years.
Exclusion criteria: 1) patients with hook and/or sublaminar wiring usage, 2) long-term use of glucocorticoids or renal dysfunction, 3) neurological disorders affecting balance, 4) the use of antiosteoporotic drugs pre-or post-operatively. According to the use of cement-augmented fenestrated pedicle screw (CAFPS) and prophylactic vertebroplasty (PV), patients were classi ed into two groups (group A: patients underwent CAFPS and PV at UIV + 1, group B: patients underwent screw rod xation system without cement augmentation) (Fig. 1).

Surgical Protocol Of Group A
In group A, the indication for CAFPS and PV was dual-energy X-ray absorptiometry (DEXA) scan T-score < -1 and intraoperative assessment of poor bone quality. T-score for the right femoral neck was used to calculate bone mineral density (BMD) in both groups.
All CAFPS xation with cement augmentation and PV at UIV + 1 was performed by two senior spine surgeons. After con rmation of all screws position via AP and lateral uoroscopy, cement augmentation was performed via cement delivery cannulas. In all cases of group A, the same polymethylmethacrylate (PMMA) bone cement was applied in a standard fashion as described by previous report 15 . UIV and LIV of all patients were xed with CAFPS and other surgical segments were xed using conventional pedicle screws or CAFPS according to intraoperative assessment of poor bone quality. A total of 2 ml high viscosity bone cement was injected into thoracic vertebra bilaterally (1 ml bone cement through per fenestrated pedicle screw). A total of 3 ml high viscosity bone cement was injected into lumbar vertebra bilaterally (1.5 ml bone cement through per fenestrated pedicle screw). PV at UIV + 1 was performed using 3 to 4 ml high viscosity cement after CAFPS xationd. The progress of bone cement injection was visualized with uoroscopic guidance (Fig. 2).

Data Collection And Radiographic Assessment
Preoperative data including age, sex, BMI and BMD were recorded. The pelvic incidence (PI), pelvic tilt (PT), lumbar lordosis (LL), PI minus LL, sagittal vertical axis (SVA) and thoracic kyphosis (TK) were measured via preoperative and early postoperative radiograph. Location of UIV and LIV was con rmed on postoperative radiograph. Radiographic measurements were performed using validated software (Surgimap®). Neurological complication, pulmonary embolism and postoperative PJK/PJF were recorded. At the 24-month follow-up, the spine radiographs were obtained for all patients. Another senior spine surgeon independently evaluated screw loosening, PJK and PJF.
PJK was de ned when Cobbs' angle between the lower endplate of UIV and the upper endplate of two super-adjacent vertebra changed greater than 10° between the immediate post-operative and the 2 years follow-up radiograph 2,3 . PJF, which may require revision surgery for proximal extension of fusion, was de ned as vertebral fracture of UIV or (and) UIV + 1, pull-out of instrumentation at UIV, and/or sagittal subluxation 3 . Pedicle screw loosening was de ned as the clear zone around the pedicle screw exceeded 1 mm on anteroposterior and lateral radiographs 16 .

Statistical Analysis
The propensity score matching procedure was performed using SPSS v24.0 (PS Matching Module). Prior to matching, continuous and categorical variables were compared between two groups using t-test, chisquare test or Fisher's Exact Test, respectively. Then, propensity score was estimated based on baseline and imaging parameter covariates by using logistic regression. Patients in group A and group B were then matched by one to one nearest neighbor matching. Following propensity score matching, balance of observed covariates of baseline and preoperative and early postoperative imaging parameters were reexamined. Differences between group A and B were compared by unpaired t-test, chi-square test or Fisher's Exact Test. The cut-off point of statistical signi cance was de ned as 0.05.

Patient population
During this study period, 292 patients who were recruited. 31 cases (group A) underwent CAFPS and PV at UIV + 1 during the period of 2016-2018. 261 cases (group B) underwent screw rod xation system without cement augmentation during the period of 2010-2016. Though some patients in group B had low T-scores, they did not receive cement augmentation because cement augmentation was not commonly used among surgeon before 2015 in our department. Finally, only 208 patients had complete data and were therefore included in this study. Patients were classi ed into group A (31 cases) and group B (177 cases) (Fig. 1). The demographic data, preoperative and early postoperative spino-pelvic parameters, UIV, LIV and the frequence of instrumented vertebrae and three-column osteotomy were illustrated in Table 1. Prior to matching, there were statistically differences between two groups with regards of age (p < 0.001), pre-LL(p < 0.001), BMD (p < 0.001), PT (p < 0.05) and LIV (p < 0.001) ( Table 1).

Propensity Score-matched Two Groups
In this study, there were 58 propensity-matched patients in group A and B with similar parameters. There were 4 males and 25 females in group A (n = 29) and the same sex proportion in group B (n = 29). Age, BMI, BMD, PI, preoperative and early postoperative PT, LL, PI minus LL, SVA and TK showed no signi cant differences between the two groups. In addition, number of instrumented vertebrae and the frequency of UIV and LIV showed no signi cant difference between two groups (p = 0.878, p > 0.99 and p = 0.773 respectively). All these data indicated that propensity score matching had eliminated the confounding factors and thus both two groups were comparable (Table 1).
After propensity score matching, the common characteristic of both groups was low BMD (T-score: -1.7 ± 0.7 vs -1.6 ± 0.7, p = 0.65). The postoperative complications of both groups were summarized in Table 2. The incidence of PJK showed no signi cantly difference between group A and B (PJK 10.3% vs 13.8%, p > 0.99). Compared to group A, the group B had a higher proportion of patients developing proximal junctional failure (24.1% vs 0%, p = 0.01). Pedicle screw loosening at UIV and (or) LIV showed no difference (24.1% vs 3.4%, p = 0.052) between two groups. In group A, cement leakage was detected in ve cases via AP and lateral uoroscopy during surgery. These ve cases with peri-vertebral cement leakage had no neurological de cit or pulmonary cement embolism that was con rmed by postoperative X-ray. One case in group A underwent revision surgery due to the onset of neurological de cit which was caused by thoracic intervertebral disc failure and stenosis between UIV and UIV + 1. Seven cases in group B had developed into PJF while six cases of them underwent revision surgery (4 cases with fracture of UIV + 1, 2 case with fracture of UIV and UIV + 1) (Fig. 3).

Discussion
In surgical correction of ASD, high load bearing requirements for the pedicular screw/bone interface is essential for rigid posterior internal xation system, which is important for spinal fusion. Bone quality is crucial to pedicle screw performance 17 . However, elderly patients with ASD are commonly accompanied with low BMD or osteoprosis in clinical practice. It has been proven that osteoporotic bone can increase the risk of implants failure 17 . Patients with low BMD or osteoporosis are more likely to suffer from PJK/PJF 6,18,19 . A propensity score matching study showed that low BMD was a signi cant risk factor for PJF with a odds ratio of 6.4 6 , which was consistent with our present study. After propensity score matching, the result showed that patients in group B with low BMD or osteoporosis were more inclined to develop into PJF or pedicle screws loosening compared to the patients in group A with "normal bone quality" (whose vertebrae was reinforced with PMMA cement).
In the last two decades, pedicle screw cement augmentation has been used to increase fatigue strength and kyphoplasty or verterbraplasty is commonly used to treat osteoporotic vertebrae fracture 12,20 . In management of ASD with low BMD or osteoporosis, prophylactic verterbraplasty (PV) at UIV + 1 and (or) the UIV and UIV + 1) could prevent PJK and PJF. However, the e cacy of prophylactic cement augmentation in the treatment of ASD was not supported by recent studies 8,13 . These results showed that bone cement reinforcement at UIV and UIV + 1 cannot prevent PJK/PJF. In our current study, the result of both pre-and post-propensity scoring matching showed that combination of CAFPS and PV at UIV + 1 could prevent PJF, but could not reduce the incidence of PJK. It suggested that cement augmenttation could not reduce the risk of disc failure, disruption of the posterior osseo-ligamentous complex and fragile paraspinal muscle, which might cause PJK.
The reasons for the controversy about whether prophylactic cement augmentation could minimize the risk for PJK and PJF were as follows. Firstly, multiple risk factors could be correlated with PJK and PJF, previous studies did not eliminate the heterogeneity of ASD population between groups, such as age, BMD or spino-pelvic parameters. In this study, we carried out the propensity score matching analysis aiming to minimize the heterogeneity of multiple risk factors causing the PJK and PJF and also to avoid selection bias. Our results showed that both two groups were comparable after propensity score matching (Table 2). Secondly, there was still no consensus on the de nition of PJK or PJF, which might be another factor that in uence the evaluation of prophylactic cement augmentation. The previous de nition of PJF is a progressive form of the PJK or the symptomatic PJK that requires surgery 1,2 . However, according to our experience, some cases with acute proximal junctional frature did no present the progressive form of the PJK. Therefore, in our study we adjusted de nition of PJK and PJF based on previous articles 1-3 .
In surgical treatment of patients with osteoporosis, cement-augmented fenestrated pedicle screw (CAFPS) had been safely used 15,21 . CAFPS could provide comparable screw stability to solid pedicle screws augmented with high-viscosity cement 22 . In our study, the CAFPS did not develop into screw pullout from the vertebrae, but one case in group A had pedicle screw loosening at LIV. Our result indicated that CAFPS might not prevent the pedicle screw loosening of ASD patients with osteoporosis treated by long posterior instrumented fusion.
The complication of cement leakage had received high concern. The high risks for cement leakage after vertebroplasty or kyphoplasty included intravertebral cleft, cortical disruption, low cement viscosity and high volume of injected cement 23 . Previous report indicated that the risk for CAFPS relevant symptomatic pulmonary cement embolism included CAFPS used in more than 7 vertebrae or more than 14 fenestrated screws, severe osteoporosis and the total cement volume more than 20 to 25 ml. In this study, no cases used more than 20 ml cement volume in total and PMMA cement volume for per vertebrae was less than the critical value reported previously 15,23 . Besides, the vertebroplasty was prophylactic treatment without the risk of intravertebral cleft or cortical disruption. Therefore, no cases in group A have complications correlated with cement leakage, though peri-vertebral cement leakage in 5 cases was detected. Thus, it could be safe to use CAFPS and PV in the management of ASD with osteoporosis.
There are some limitations in this study. Firstly, the potential weakness of this study was its relatively small sample size (n = 58). Future studies with a larger sample size could yield more accurate results.
Secondly, though this propensity score matched cohort analysis minimized the heterogeneity of ASD, some potential risk factors for PJK or PJF were not commonly assessed, such as prevalence of sarcopenia in middle-aged and elderly women 24 and disruption of the posterior osseo-ligamentous complex.

Conclusion
Surgeons have paid high attention to the prevention of postoperative PJK and PJF in patients with long posterior spinal fusions. This propensity score-matched study indicated that combined application of CAFPS and PV could not prevent the development of PJK, but could reduce the incidence of PJF in the surgical management of ASD with low BMD or osteoporosis.

Declarations
Ethics approval and consent to participate: our research is a retrospective study, which used medical records obtained from patients' imaging data and clinical data. Thus, the ethical approval is exempted from the Ethical Committee of our institutions.

Consent for publication:
Consent for publication was obtained from all participants.
Availability of data and materials: the datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Figure 1 Flow chart of the process for propensity score-matched design. image nding of cement leakage. E. Early postoperative lateral radiograph showed proximal junctional angle was 7°. F. The nal follow-up lateral radiograph showed proximal junctional angle was 18°, proximal junctional kyphosis was detected. This patient was asymptomatic.