Analysis of Potential Risk Factors for Cement Leakage into Paraspinal Veins after Vertebroplasty for Acute Osteoporotic Vertebral Fractures Based on a 3D Reconstruction Technique: A Retrospective Matched Case–Control Study

Abstract

Objective

Pulmonary embolism, cardiac embolism, and even cerebral embolism due to paraspinal vein leakage (PVL) are increasingly reported, and their risk factors need to be adequately investigated for prevention. To this end, this study investigated the correlation of the distribution and morphological characteristics of fracture lines with the occurrence of PVL after percutaneous vertebroplasty (PVP), which has not been previously reported.

Methods

Patients with acute single‐segment thoracolumbar osteoporotic vertebral compression fractures (OVCFs) treated with PVP at our institution from January 2019 to July 2022 were selected for a matched case–control study. The case and control groups included those with and without PVL, respectively, matched at a 1:1 ratio based on general clinical characteristics. Additionally, fracture map and heatmap analysis was performed in both groups. In addition to the general clinical characteristics, the vertebral height ratio, puncture angle, delivery rate, and indexes were assessed via the three‐dimensional CT reconstruction fracture line mapping technique, namely, the distribution of fracture lines, fracture line length, main fracture line shape, location of fracture line involvement, and number of fracture line branches, were compared between the two groups. The Wilcoxon rank‐sum test, t tests, analysis of variance, and conditional logistic regression were used for statistical analysis.

Results

Among 658 patients with OVCFs, 54 who did and 54 who did not develop PVL were included in this study. Significant differences in the puncture angle, fracture line distribution (MR‐1, ML‐2, MM‐2, MR‐2, ML‐3, MM‐3, LL‐1, LM‐1, LL‐2, LM‐2), fracture line involvement of the posterior wall, total fracture line length, and main fracture line length were found between the two groups (p < 0.05). Logistic univariate analysis showed significant differences in the puncture angle, fracture line distribution (MR‐1, ML‐2, MM‐2, MR‐2, ML‐3, MM‐3, LL‐1, LL‐2, LM‐2, LL‐3), total fracture line length, main fracture line length, and fracture line involvement of the posterior wall between the two groups (p < 0.05). Logistic multifactorial analysis showed that the fracture line distribution (UR‐3, ML‐3, LM‐2, LR‐2) and main fracture line length were independent risk factors for the development of PVL in both groups. In addition, the fracture maps and heatmaps showed a greater degree of fracture line encapsulation and more extensive involvement in the middle and lower regions of the vertebral body in the PVL group than in the control group.

Conclusions

Through a three‐dimensional computed tomography reconstruction‐based fracture line mapping technique, this study revealed for the first time that the distribution of fracture lines (UR‐3, ML‐3, LM‐2, LR‐2) and main fracture line length were independent risk factors for PVL after PVP in patients with acute single‐segment thoracolumbar OVCFs. In addition, we hypothesized that the fracture line‐vein traffic branch that may appear within 2 weeks after injury in acute OVCF patients may be one of the mechanisms influencing the above potential independent risk factors associated with PVL.

“Distribution of fracture lines (UR‐3, ML‐3, LM‐2, LR‐2) and length of the main fracture line” are independent risk factors for paraspinal venous leakage after PVP in patients with acute OVCFs. In addition, we hypothesized that the fracture line‐vein traffic branch that may appear within 2 weeks after injury in patients with acute OVCF may be one of the mechanisms influencing the above potential independent risk factors associated with PVL.

Introduction

Osteoporotic vertebral compression fractures (OVCFs) are one of the main culprits affecting the daily life of elderly individuals. ¹ Percutaneous vertebroplasty (PVP) has become the preferred surgical treatment for OVCFs because of its ease of operation and low cost. In addition, the advantages of minimal trauma, significant pain relief, early mobility, and short hospital stay have led to the wide use of PVP in clinical practice. ² However, bone cement leakage remains the greatest challenge for spine surgeons to overcome. There are different routes for bone cement leakage, such as through the basivertebral vein, through the paraspinal vein, and through cortical defects, as well as intradiscal leakage; among them, the two venous leakage routes are the most common. ³ , ⁴

Although paraspinal vein leakage (PVL) does not cause symptoms in most cases, in recent years, D'Errico et al. ⁵ and Barakat et al. ⁶ have reported that OVCF patients undergoing PVP can have severe symptoms such as dyspnea, chest tightness, and hemoptysis, which are ultimately attributed to the spread of bone cement along the segmental veins to the lungs and heart. In addition, Llanos et al. ⁷ reported that in an OVCF patient with an open foramen ovale, bone cement was found to invade the brain after PVP treatment, causing cerebral infarction. In addition to the serious threat to patients' lives, treatments for this complication are limited and rely on multidisciplinary interventions, with a very poor prognosis. Moreover, such complications are difficult to predict early and can only be verified by ancillary tests such as diagnostic imaging examinations, and few patients tend to undergo imaging of the chest and head after PVP. ⁵ , ⁶ , ⁷ Therefore, there is no doubt that avoiding the occurrence of PVL is the optimal solution.

Over the past two decades, many investigators have focused on the correlations of the bone cement viscosity and volume, the intravertebral vein distribution, and intravertebral clefts with PVL and have proposed many surgical optimization measures. ⁸ , ⁹ , ¹⁰ , ¹¹ However, these studies have not delved deeper into the associations of the fracture line distribution and morphological features with PVL, and characterization of the fracture line distribution and morphological features based on three‐dimensional (3D) reconstruction imaging data has become a trend in surgical plan exploration in fracture populations. ¹² In 2009, Armitage et al. ¹⁴ first introduced scaphoid fracture maps for visualization of the distribution and morphological characteristics of fracture lines. In subsequent studies, CT reconstruction‐based fracture maps have been widely used in the field of orthopaedics, with a focus on the characterization of extremity fractures. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ Unlike traditional 2D mapping techniques, this 3D reconstruction technique can more accurately show the specific distribution of fracture lines and their morphological characteristics, which is extremely important for the preoperative development of appropriate surgical plans. ¹⁵ , ¹⁶ In recent years, this fracture line mapping technique has been applied to describe spinal fractures. ¹⁷ , ¹⁸ We believe that assessing the distribution and morphological characteristics of fracture lines in patients with PVL may be more favorable for optimizing the puncture protocol. Therefore, in this work, a retrospective analysis was performed with the following objectives: (1) to investigate the correlations of the distribution and morphological characteristics of fracture lines with the occurrence of PVL after PVP in patients with thoracolumbar OVCFs and propose preventive measures; and (2) to explore the underlying mechanisms of potential risk factors.

Materials and Methods

Patient Demographics and Grouping

This retrospective study was approved by the medical ethics committee of our institution (Medical Ethics Committee, China‐Japan Union Hospital of Jilin University, No.2023083003). Written informed consent was obtained from each participant. We enrolled 658 consecutive patients with single‐segment thoracolumbar OVCFs treated with PVP at our institution from January 2019 to July 2022. The inclusion criteria were as follows: ≤14 days after OVCF; age ≥55 years and severe back pain; and clear vertebral bone marrow edema on magnetic resonance imaging (MRI). Patients who did not have available preoperative and postoperative computed tomography (CT) data from our institution, were previously treated with PVP or were diagnosed with a pathological fracture caused by a tumor, infection, etc., were excluded.

PVL was evaluated based on postoperative CT (Toshiba Aquilion One 320 spiral CT scanner) data and lateral X‐ray films. The following two observations were considered to indicate PVL: (1) CT showed that the bone cement was distributed along the paravertebral vein, with an approximately vertical or oblique 2D distribution on the sagittal plane and an approximately horizontal 2D on the axial plane; and (2) lateral X‐ray films showed vertical, strip‐shaped dense shadows outside the vertebral body. The evaluation process was performed by two investigators, and any disagreements were resolved by the corresponding author.

Because of the limited incidence of PVL, a 1:1 matched case–control design was used. The case group included patients who developed PVL, and the control group, named the no bone cement leakage (NBCL) group, included patients who did not develop bone cement leakage. Patients in the NBCL and PVL groups were matched according to the following general clinical characteristics: age (within 5 years); sex (male or female); body mass index (BMI) (within 2 kg/m); bone mineral density (BMD) (within −1 SD); fracture time (within 3 days); and bone cement volume (within 1 ml). The number of cases of PVL was determined first and then matched with the same number of NBCL cases according to the above variables; then, the fracture line distribution and morphological characteristics were compared between the two groups.

Operative Technique

With the patient under local anesthesia in the operating room, vertebroplasty was performed using a transforaminal approach with puncture instruments and a bone cement system provided by Heraeus Medical, Ltd. (low‐viscosity bone cement in all cases). Under C‐arm fluoroscopy, bone cement with a toothpaste‐like consistency was slowly injected into the vertebral body using a screw syringe to achieve symmetrical and satisfactory filling of the patient's diseased vertebra. The bone cement injection was stopped when the first two‐thirds of the vertebral body was filled. In most patients, the bone cement volume was kept in the range of 3–5 ml. All patients were restricted to bed rest for several hours after the procedure until the bone cement reached its final strength and were discharged the next day.

3D Modeling, Fracture Mapping, and Fracture Line Heatmapping

The CT data of the spine extracted from the imaging database were saved in DICOM format and imported into the 3D reconstruction software Mimics Medical 20.0 (Materialize, Belgium) to create target files in STL format. Taking the T12 vertebra as an example (Figure 1), the target vertebra was analyzed and processed simultaneously on three planes, that is, coronal, axial, and sagittal, to generate a 3D model of the spine (Figure 1).

FIGURE 1.

3D reconstruction of CT data using Mimics software.

As shown in Figure 2, we performed vertebral body regionalization and segmentation to count the 2D distribution of fracture lines. Using the posterior edge of the vertebral body as the boundary, two axial lines were first created to evenly segment the vertebral body into three major regions, namely, the upper, middle, and lower levels. Subsequently, each large region was divided into nine cubes using two sagittal and two coronal line segments, ultimately generating 27 cubes for the vertebral body. Among them, the top nine cubes were named upper left 1 ~ 3 (UL‐1 ~ 3), upper middle 1 ~ 3 (UM‐1 ~ 3), and upper right 1 ~ 3 (UR‐1 ~ 3). The middle nine cubes were named middle left 1 ~ 3 (ML‐1 ~ 3), middle middle 1 ~ 3 (MM‐1 ~ 3), and middle right 1 ~ 3 (MR‐1 ~ 3). Finally, the remaining nine cubes were named lower left 1 ~ 3 (LL‐1 ~ 3), lower middle 1 ~ 3 (LM‐1 ~ 3), and lower right 1 ~ 3 (LR‐1 ~ 3). The extracted 2D fracture lines were drawn in the 3D reconstruction software Mimics Medical 20.0 (Figures 2 and 3).

FIGURE 2.

Numbering of the vertebra after segmentation into 27 regions. (A) Partition lines creating 27 cubes: two axes parallel to the end plates (vertical lines on the sagittal plane that divide the central horizontal line of the vertebral body into three equal parts); two sagittal lines (vertical lines on the axes that divide the central horizontal line of the vertebral body into three equal parts); and two coronal lines (vertical lines on the axes that divide the central vertical line of the vertebral body into three equal parts). (B) Schematic diagram of the upper region of the vertebral body. (C) Schematic diagram of the middle region of the vertebral body. (D) Schematic diagram of the lower region of the vertebral body. LL, lower left; LM, lower middle; LR, lower right; ML, middle left; MM, middle middle; MR, middle right; UL, upper left; UM, upper middle; UR, upper right.

FIGURE 3.

Plot of the 2D fracture line distribution for counting the fracture lines on the coronal, sagittal, and axial planes of the vertebral body.

The previously obtained 3D data of the spine were imported into Geomagic 2017 (Geomagic, USA) software in STL format, and reverse reconstruction processes such as denoising, polishing, smoothing, cropping, and filling were performed (Figure 4A) to create a simulated 3D image. Movements and rotations were performed in UG (Unigraphics NX12.0, Siemens PLM Software, USA) 3D CAD modeling software to simulate fracture repositioning and construct a 3D model of the repositioned spinal fracture (Figure 4B). The repositioned 3D model was imported into E‐3D (V19.12, Hunan Six Dimensions Precision Navigation Digital Technology Co., Ltd.) modeling software, and 3D fracture lines were drawn on the standard vertebral body model according to the distribution of the fracture lines (Figure 4C) to construct a map of the spinal fracture lines (Figure 4).

FIGURE 4.

Process of fracture map production. (A) Reconstruction of a simulated 3D model based on 3D fracture data. (B) Repositioning of the simulated 3D fracture model. (C) Mapping of fracture lines onto the standard model.

Finally, a 3D fracture line heatmap was generated according to the blue–red gradient map.

Study Variables

1. Vertebral height ratio: The height of the most obvious compression of the fractured vertebral body/(height of the upper vertebral body at the same location + height of the lower vertebral body at the same location/2).

2. Puncture angle: The angle between the axial puncture needle and the midline of the vertebral body measured on postoperative CT.

3. Delivery rate: As a staged bone injection method was used, each time 0.5–1 ml of bone cement was injected, it was injected over an interval of 30–60 therefore, the delivery rate was defined as rapid when the injection was completed within 2 min; moderate, 2–4 min; and slow, more than 4 min.

The preoperative CT data of all patients were recorded after 3D reconstruction with the following indicators. All data, including the quantitative measures described above, that is, vertebral body height ratio and puncture angle, were assessed independently by two investigators and analyzed using the average of the two investigators' readings.

1. Distribution of fracture lines: The distribution of fracture lines was compared between groups by recording the number of fracture lines within the 27 vertebral regions.

2. Fracture line length: The starting and ending points of the fracture lines in the reconstructed 3D image of the vertebral body were recorded separately, and the length of the main fracture line (taking the fracture line with the longest extension on the surface of the vertebral body, the most extensive involvement, and the greatest impact) was calculated for each case by making a straight line with the two points. The length of each fracture line branch was calculated similarly, and the total length of the fracture lines for each case was calculated by summing the measured lengths.

3. Shape of the main fracture line: The shape of the main fracture line was recorded and classified as follows: linear (not wrapped around the vertebral body, fewer than two fracture line crests); wavy (not wrapped around the vertebral body, two or more fracture line crests); inverted U‐shaped wrapped (wrapped around the vertebral body, fracture line crests in inverted U‐shaped distribution), linear wrapped (wrapped around the vertebral body, fewer than two fracture line crests), and wavy wrapped (wrapped around the vertebral body, two or more fracture line crests).

4. Location of fracture line involvement: The numbers of fracture lines involving the upper and lower endplates and anterior and posterior walls of the vertebral body were recorded. Additionally, whether the upper and lower endplates were involved was recorded, and one case each of upper and lower endplate involvement was included.

5. Number of fracture line branches: The presence of fracture line branches was recorded in each case, and the number of branches was counted.

Statistical Analysis

SPSS software (version 22.0, SPSS, Inc., Chicago, IL, USA) was used for statistical analysis of the data. Only age and main line length satisfied both normal distribution and chi‐square tests; thus, they are expressed as the mean ± standard deviation and were compared between groups by the t test. Other continuous variables are expressed using the median and interquartile range (IQR)and were compared between groups by the Mann–Whitney U test. Count data are expressed as percentages (%) and were compared between groups by the chi‐square test. Due to the presence of theoretical frequencies less than five in the cross‐tabulations, the following variables were compared by the corrected chi‐squared test: segment, MM‐3, LM‐2, LL‐3, LM‐3, linear wrapped, posterior wall, lower endplate, and number of branches. Quantitative indexes were subjected to logistic univariate analysis, and variables with p < 0.2 were screened for logistic multifactorial analysis. p < 0.05 was considered to indicate a statistically significant difference.

Results

General Patient Characteristics

The case screening process for this study is detailed in Figure 5. The inclusion and exclusion criteria were used to initially include 441 patients with acute OVCFs of a single thoracolumbar vertebra, and 54 patients were included in the PVL group based on imaging assessment criteria. The remaining 387 patients included 280 with no bone cement leakage, 47 with basilar vein leakage, 31 with intradiscal leakage, and 29 with cortical defect leakage. The 280 patients without cement leakage were matched to those in the PVL group at a 1:1 ratio according to general clinical characteristics such as age, sex, BMI, BMD, fracture time, and bone cement volume, and 54 patients were finally included in the NBCL group.

Figure 5.

Case screening flow chart. BMI, body mass index; BMD, bone mineral density; CT, computed tomography; NBCL, no bone cement leakage; PVL, paraspinal vein leakage.

As shown in Table 1, there was no significant difference in sex (p > 0.999), age (p = 0.786), BMI (p = 0.053), BMD (p = 0.204), fracture time (p = 0.667), bone cement volume (p = 0.983) or segment (p = 0.441) between the two groups. There were significant differences in the puncture angle, fracture line distribution (MR‐1 (p = 0.020), ML‐2 (p = 0.007), MM‐2 (p < 0.001), MR‐2 (p = 0.004), ML‐3 (p = 0.004), MM‐3 (p < 0.001), LL‐1 (p = 0.029), LM‐1 (p = 0.040), LL‐2 (p = 0.007), and LM‐2 (p = 0.045)), total fracture line length (p = 0.030), main fracture line length (p = 0.003), and posterior wall fracture line involvement (p<0.001) between the two groups (Table 1).

TABLE 1.

General data of patients in both groups.

Characteristic	NBCL, N = 54	PVL, N = 54	t/Z/χ 2	p
Gender			0.000	>0.999
Female	39 (72.2%)	39 (72.2%)
Male	15 (27.8%)	15 (27.8%)
Age (years)	70.06 ± 7.99	70.46 ± 7.55	−0.272	0.786
BMI (kg/m2)	22.47 (21.90, 23.31)	23.25 (22.20, 23.79)	−1.932	0.053
Bone density (SD)	−2.78 (−2.95, −2.63)	−2.69 (−2.87, −2.58)	−1.269	0.204
Fracture time (d)	5.00 (3.00, 6.00)	4.00 (3.00, 7.00)	−0.431	0.667
Cement volume (ml)	4.10 (3.60, 4.50)	4.30 (3.30, 4.80)	0.022	0.983
Puncture angle (°)	16.48 (14.33, 18.21)	12.50 (11.50, 15.58)	3.931	<0.001
Vertebral height ratio	0.87 (0.80, 0.92)	0.86 (0.78, 0.89)	1.416	0.157
Segment			3.746	0.441
L1	18 (33.3%)	21 (38.9%)
L2	9 (16.7%)	6 (11.1%)
T10	0 (0.0%)	1 (1.9%)
T11	11 (20.4%)	6 (11.1%)
T12	16 (29.6%)	20 (37.0%)
Delivery rate			0.502	0.778
Fast	5 (9.3%)	6 (11.1%)
Medium speed	43 (79.6%)	44 (81.5%)
Slow	6 (11.1%)	4 (7.4%)
UL‐1	46 (85.2%)	39 (72.2%)	2.707	0.100
UM‐1	46 (85.2%)	42 (77.8%)	0.982	0.322
UR‐1	42 (77.8%)	36 (66.7%)	1.662	0.197
UL‐2	44 (81.5%)	35 (64.8%)	3.818	0.051
UM‐2	32 (59.3%)	28 (51.9%)	0.600	0.439
UR‐2	35 (64.8%)	36 (66.7%)	0.041	0.839
UL‐3	30 (55.6%)	24 (44.4%)	1.333	0.248
UM‐3	14 (25.9%)	19 (35.2%)	1.091	0.296
UR‐3	31 (57.4%)	24 (44.4%)	1.815	0.178
ML‐1	29 (53.7%)	35 (64.8%)	1.381	0.240
MM‐1	32 (59.3%)	40 (74.1%)	2.667	0.102
MR‐1	25 (46.3%)	37 (68.5%)	5.453	0.020
ML‐2	17 (31.5%)	31 (57.4%)	7.350	0.007
MM‐2	13 (24.1%)	35 (64.8%)	18.150	<0.001
MR‐2	17 (31.5%)	32 (59.3%)	8.405	0.004
ML‐3	7 (13.0%)	20 (37.0%)	8.346	0.004
MM‐3	4 (7.4%)	21 (38.9%)	15.042	<0.001
MR‐3	7 (13.0%)	11 (20.4%)	1.067	0.302
LL‐1	6 (11.1%)	15 (27.8%)	4.788	0.029
LM‐1	8 (14.8%)	17 (31.5%)	4.216	0.040
LR‐1	8 (14.8%)	13 (24.1%)	1.478	0.224
LL‐2	5 (9.3%)	16 (29.6%)	7.153	0.007
LM‐2	3 (5.6%)	11 (20.4%)	4.021	0.045
LR‐2	5 (9.3%)	12 (22.2%)	3.421	0.064
LL‐3	4 (7.4%)	12 (22.2%)	3.595	0.058
LM‐3	1 (1.9%)	6 (11.1%)	2.444	0.118
LR‐3	6 (11.1%)	10 (18.5%)	1.174	0.279
Total length (mm)	94.01 (67.77, 121.80)	118.29 (83.91, 158.29)	−2.169	0.030
Main line (mm)	66.48 ± 19.48	80.41 ± 26.64	−3.101	0.003
Branch line 1 (mm)	23.45 (0.00, 38.69)	24.04 (0.00, 38.47)	−0.233	0.815
Branch line 2 (mm)	0.00 (0.00, 14.14)	0.00 (0.00, 18.89)	−0.656	0.512
Branch line 3 (mm)	0.00 (0.00, 0.00)	0.00 (0.00, 0.00)	−0.079	0.937
Branch line 4 (mm)			3.010	0.390
0	53 (98.1%)	52 (96.3%)
14.24	1 (1.9%)	0 (0.0%)
16.36	0 (0.0%)	1 (1.9%)
17.45	0 (0.0%)	1 (1.9%)
Branch line 5 (mm)				>0.999
0	54 (100.0%)	53 (98.1%)
19.59	0 (0.0%)	1 (1.9%)
Branch line AVG	6.99 (0.00, 11.37)	7.54 (0.00, 13.08)	−0.657	0.511
Inverted U‐shaped	8 (14.8%)	5 (9.3%)	0.787	0.375
Linear‐like	12 (22.2%)	10 (18.5%)	0.228	0.633
Wave	6 (11.1%)	10 (18.5%)	1.174	0.279
Linear wrapped	1 (1.9%)	5 (9.3%)	1.588	0.208
Wavy wrapped	29 (53.7%)	25 (46.3%)	0.593	0.441
Anterior wall	47 (87.0%)	48 (88.9%)	0.087	0.767
Posterior wall	2 (3.7%)	18 (33.3%)	15.709	<0.001
Upper endplate	35 (64.8%)	36 (66.7%)	0.041	0.839
Lower endplate	4 (7.4%)	8 (14.8%)	0.844	0.358
Number of branches			1.370	0.928
0	17 (31.5%)	15 (27.8%)
1	18 (33.3%)	18 (33.3%)
2	12 (22.2%)	14 (25.9%)
3	6 (11.1%)	5 (9.3%)
4	1 (1.9%)	1 (1.9%)
5	0 (0.0%)	1 (1.9%)

Abbreviations: LL, lower left; LM, lower middle; LR, lower right; ML, middle left; MM, middle of the middle; MR, middle right; NBCL, no bone cement leakage; PVL, paraspinal venous leakage; UL, upper left; UM, upper middle; UR, upper right.

Univariate Analysis of Factors Affecting PVL

As shown in Table 2, univariate analysis using conditional logistic regression with a paired design showed a significant difference in the puncture angle, fracture line distribution (MR‐1 (p = 0.024), ML‐2 (p = 0.014), MM‐2 (p < 0.001), MR‐2 (p = 0.007), ML‐3 (p = 0.008), MM‐3 (p = 0.002), LL‐1 (p = 0.048), LL‐2 (p = 0.015), LM‐2 (p = 0.038), LL‐3 (p = 0.046)), total fracture line length (p = 0.035), main fracture line length (p = 0.004), and posterior wall fracture line involvement (p = 0.003) between the two groups of patients (Table 2).

TABLE 2.

Logistic univariate analysis of quantitative indicators.

Characteristic	OR	95% CI	p
Puncture angle	0.892	0.801, 0.993	0.037
Vertebral height ratio	0.025	0.000, 3.243	0.137
Delivery rate			0.703
Slow	—	—
Medium	1.937	0.349, 10.762	0.450
Fast	2.363	0.281, 19.841	0.428
UL‐1	0.500	0.202, 1.239	0.134
UM‐1	0.636	0.247, 1.642	0.350
UR‐1	0.600	0.263, 1.371	0.226
UL‐2	0.471	0.203, 1.090	0.079
UM‐2	0.692	0.296, 1.620	0.396
UR‐2	1.067	0.527, 2.157	0.857
UL‐3	0.684	0.338, 1.385	0.292
UM‐3	1.556	0.673, 3.594	0.301
UR‐3	0.500	0.202, 1.239	0.134
ML‐1	1.750	0.734, 4.172	0.207
MM‐1	2.000	0.856, 4.673	0.109
MR‐1	2.714	1.141, 6.457	0.024
ML‐2	2.750	1.224, 6.177	0.014
MM‐2	4.667	1.932, 11.270	<0.001
MR‐2	3.500	1.413, 8.672	0.007
ML‐3	5.333	1.554, 18.304	0.008
MM‐3	6.667	1.981, 22.435	0.002
MR‐3	1.667	0.606, 4.586	0.323
LL‐1	2.800	1.009, 7.774	0.048
LM‐1	2.500	0.970, 6.443	0.058
LR‐1	2.000	0.684, 5.851	0.206
LL‐2	4.667	1.341, 16.239	0.015
LM‐2	5.000	1.096, 22.820	0.038
LR‐2	2.750	0.876, 8.636	0.083
LL‐3	3.667	1.023, 13.143	0.046
LM‐3	6.000	0.722, 49.837	0.097
LR‐3	1.800	0.603, 5.371	0.292
Total length	1.010	1.001, 1.020	0.035
Main line	1.031	1.010, 1.054	0.004
Branch line 1	1.001	0.983, 1.018	0.951
Branch line 2	1.017	0.986, 1.048	0.282
Branch line 3	1.013	0.960, 1.068	0.641
Branch line 4	1.055	0.904, 1.231	0.496
Branch line 5	2.532	0.000, Inf	0.998
Branch line AVG	1.021	0.963, 1.083	0.485
Inverted U‐shaped	0.571	0.167, 1.952	0.372
Linear‐like	0.800	0.316, 2.027	0.638
Wave	1.667	0.606, 4.586	0.323
Linear wrapped	218,630,744.441	0.000, Inf	0.998
Wavy wrapped	0.714	0.317, 1.608	0.416
Anterior wall	1.250	0.336, 4.655	0.739
Posterior wall	9.000	2.088, 38.787	0.003
Upper endplate	1.091	0.481, 2.472	0.835
Lower endplate	2.333	0.603, 9.023	0.220
Number of branches	1.126	0.761, 1.665	0.554

Abbreviations: LL, lower left; LM, lower middle; LR, lower right; ML, middle left; MM, middle of the middle; MR, middle right; NBCL, no bone cement leakage; PVL, paraspinal venous leakage; UL, upper left; UM, upper middle; UR, upper right.

Multifactorial Analysis of Factors Affecting PVL

Factors with p < 0.2 were screened from the univariate analysis and included in the conditional logistic regression model for multifactorial analysis. As shown in Table 3, the fracture line distribution (UR‐3 (p = 0.018), ML‐3 (p = 0.028), LM‐2 (p = 0.013), LR‐2 (p = 0.048)) and main fracture line length (p = 0.008) were independent risk factors for the development of PVL in both groups (Table 3).

TABLE 3.

Logistic multifactorial analysis of quantitative indicators.

Characteristic	OR	95% CI	p
UR‐3	0.172	0.040, 0.741	0.018
ML‐3	5.778	1.212, 27.547	0.028
LM‐2	37.919	2.137, 672.745	0.013
LR‐2	0.095	0.009, 0.977	0.048
Main line	1.045	1.012, 1.080	0.008

Abbreviations: LM, lower middle; LR, lower right; ML, middle left; UR, upper right.

PVL‐Related Fracture Maps and Fracture Line Distribution Heatmaps

Figure 6 shows the fracture maps and fracture line distribution heatmaps for the two groups. Intuitively, a common feature in both groups is that the fracture line distribution is concentrated in the upper and anterior third of the vertebral body with circular wrapping. However, overall, the fracture lines in the PVL group showed a greater degree of wrapping, were densely distributed in the upper and middle parts of the vertebral body and were more densely distributed in the inferior region of the vertebral body. In contrast, the fracture line distribution in the NBCL group was more orderly, concentrated in the upper third of the vertebral body, and more sparsely distributed in the middle and lower regions. In addition, from a top view, the fracture lines in the PVL group were distributed in a complete fan shape at the anterior edge of the vertebral body, while those in the NBCL group were distributed in a semi‐fan shape (Figure 6).

FIGURE 6.

Fracture maps and distribution heatmaps for both groups of patients, showing posterior, anterior, top, bottom, right lateral, and left lateral views. The red line in the upper fracture map is the reconstructed fracture line. The lower fracture line distribution heatmap is presented as a red and blue gradient, with the color representing the density of the fracture line distribution. The color distribution map from the bottom to the top represents the increasing density of the fracture line distribution, for example, the red color represents a higher fracture line distribution density. NBCL, no bone cement leakage; PVL, paraspinal vein leakage.

Discussion

The statistical analysis of this study showed that patients in the PVL group had a greater distribution of fracture lines in the UR‐3, ML‐3, LM‐2, and LR‐2 regions than patients in the NBCL group; additionally, the length of the main fracture line was longer in patients in the PVL group, suggesting that the distribution of fracture lines in the UR‐3, ML‐3, LM‐2, and LR‐2 regions and the main fracture line length are potentially independent risk factors for the occurrence of PVL. In addition, the fracture maps and fracture line distribution heatmaps of the two groups showed that more fracture lines were distributed in the middle and lower vertebral regions in the PVL group than in the NBCL group and that the fracture lines in the PVL group were longer, similar to the results of the conditional logistic regression analysis described above.

Preliminary Mechanistic Exploration of Fracture Line Distribution and Main Fracture Line Length in Relation to PVL

The results of the logistic multifactorial analysis in this study indicated that the location of fracture lines in the UR‐3, ML‐3, LM‐2, and LR‐2 regions was a potential independent risk factor for the development of PVL. The vertebral venous system consists of an interconnected intra‐ and extravertebral venous plexus and the vertebral basilar veins, which form a valveless plexiform venous network. ¹⁹ Since the extravertebral venous plexus is most closely connected to the superior and inferior vena cava systems, it dominates the diffusion of bone cement along the paraspinal veins to the lungs, heart, and other vital organs. ¹¹ , ¹⁹ The presence of fracture lines directly disrupts the vertebral venous system. ²⁰ Previous studies have suggested that the intravertebral venous system tends to be concentrated in the middle region. ¹¹ , ²¹ We hypothesize that the intersection of the lateral extravertebral venous plexus and the intravertebral veins ¹⁰ (some branch veins can be interpreted as common accessory branches of both) is more densely distributed in the UR‐3, ML‐3, LM‐2, and LR‐2 regions of the vertebral body. When the fracture line only involves this area, many small “fracture line‐vein traffic branches” will be formed, and the presence of these traffic branches will cause the dense venous plexus to form new intersection channels. When unconsolidated bone cement diffuses toward the intravertebral fracture lines, it will inevitably enter these traffic branches, resulting in PVL. In addition, the repaired vein wall is often more fragile than normal, ¹⁰ and the distribution of bone cement into the cavity formed by the fracture line may cause high local pressure and increase the subsequent infusion pressure of bone cement ²² ; coupled with the thermal effect of bone cement, this may allow the bone cement to easily break through the repaired vein wall and leak into the vein. However, it has been suggested that disruption of the vertebral venous system after fracture can lead to occlusion of the injured vein and prevent the development of PVL, ²³ , ²⁴ which may be due to the inability to effectively repair long‐term damage (>2 weeks) to the venous system. ²⁵ In contrast, the patients we included were all acute OVCF patients treated less than 2 weeks after injury. Therefore, traffic branches between the fracture lines and the vertebral venous system are possible early after the onset of acute OVCFs, and this could be a potential mechanism underlying the association of the fracture line distribution with PVL. In addition to these two mechanisms, considering that all the included patients suffered from severe osteoporosis, one has to wonder whether the presence of dilated neighboring paravertebral veins caused by trabecular bone loss and destruction has some influence on this outcome. ¹⁰

Our results also show that the main fracture line length is a potential independent risk factor for PVL. This result seems to suggest that the length of the main fracture line indicates the difference in the damage to the vertebral venous system associated with the fracture line ²⁶ , ²⁷ ; that is, a wider fracture line indicates more damage to the vertebral venous system, resulting in more fracture line‐vein communicating branches and increasing the incidence of PVL. Of course, this can also explain the differences between the fracture maps and fracture line distribution heatmaps.

Nevertheless, it is undeniable that there is a certain degree of variation in deconstruction and in the vertebral venous system among individuals, ²² , ²⁸ which may interfere with the postulated mechanism. Due to technical limitations, it is difficult to describe the specific distribution of paraspinal veins in each patient. Therefore, much more research is still needed.

Avoiding PVL on the Basis of 3D Reconstruction

Considering that the fracture line distribution is an independent risk factor, the following measures could be effective in avoiding PVL after PVP: (1) choose a puncture point below the traditional puncture point (at the 2 or 10 o'clock position on the orthogonal vertebral X‐ray projection); (2) perform the puncture obliquely upward, avoiding passing through the upper right and mid‐left posterior part of the vertebral body during the initial entry to ensure that the puncture ends at the upper third of the vertebral body; and (3) avoid premature contact of the puncture needle with the main fracture line when entering the vertebral body (Figure 7). Of course, if there is excessive collapse of the upper endplate and a large degree of vertebral compression, percutaneous kyphoplasty (PKP) or even other techniques may be required to compensate for the puncture risk. In addition, this method of puncture is not exempt from the question of whether it ensures effective bone cement dispersion within the fracture and re‐establishes biomechanical stability. Although there is controversy in recent studies that have explored the effect of the degree of bone cement dispersion in the fracture on the prognosis at different times, ²⁹ in practice, most physicians are still accustomed to puncturing in the direction of the fracture line to enable the bone cement to effectively fill the fracture, thereby fixing the fractured trabeculae and improving the early stability of the injured vertebra. ³⁰ We believe that this method of puncture is flawed. Because of the high viscosity of the bone cement and low pressure in the early stage of injection, the cement easily disperses along the cavity, and the injection of an appropriate volume can ensure effective filling of the fracture line. ³¹ , ³² In addition, our results show that there may be a certain correlation between the puncture angle and PVL, with a smaller puncture angle potentially leading to PVL more easily, which is similar to the results reported by Liu et al. ³³ Compared with the fracture line distribution, the main fracture line length is more likely to be influenced by factors such as the timing and volume of cement injection, which is difficult to effectively address by adjusting the puncture direction and angle. Therefore, these findings also suggest that attention should be given to other factors in addition to examining the influence of the fracture line on PVL. ³¹ , ³² , ³³

FIGURE 7.

Schematic diagram of an optimized puncture protocol. LM, lower middle; LR, lower right; ML, middle left; UR, upper right.

With the advancement of technology, robot‐assisted surgery has shown promising applications in spine surgery. ³⁴ In 2021, Shi et al. ³⁴ applied the ZOEZEN robot for PVP in 30 patients with thoracolumbar fractures and achieved satisfactory clinical results. Despite the need for more tedious procedures and certain learning curves, the greatest advantage of such techniques is the increased accuracy of the puncture. Therefore, combining robotic systems with preoperative 3D reconstruction techniques has potential for future applications in clinical treatment.

Limitations and Prospect of Clinical Application

We acknowledge the limitations of this study. First, this was a single‐center retrospective study. Second, the limited number of patients who could be included in the PVL group was due to a lack of imaging data due to multiple factors, such as economic factors and preoperative imaging being performed at other hospitals. Many patients did not undergo systematic CT examinations at our institution; thus, a large sample size is required for follow‐up studies. Third, this study only included patients with single‐segment thoracolumbar OVCFs, a common condition that we believe to be representative. However, we cannot exclude the impact of fractures affecting other or multiple segments on the results. Finally, because we used low‐viscosity bone cement, we cannot rule out the potential effect of different bone cement viscosities on the quantitative indexes.

Nevertheless, to our knowledge, the present study is the first to discuss the correlation between the distribution and morphologic characteristics of fracture lines and the occurrence of PVL based on 3D reconstruction techniques. In addition, we proposed a possible puncture protocol to clinically reduce the occurrence of PVL, which, unlike previous studies, places additional emphasis on the distribution and morphological characteristics of the fracture line within the vertebral body, and by analyzing the risk areas for the occurrence of PVL, indicates that a safe puncture path may lead to better clinical outcomes, and to a certain extent reflects the need for precision medical care in spine surgery. In addition, with the gradual development of precise digital medical technology, it is the future trend to emphasize the application of CT 3D reconstruction technology in the field of spinal fracture, which can provide clinicians with more comprehensive and accurate information, so as to rationally plan the surgical plan.

Conclusion

In this paper, the potential association of the fracture line distribution (UR‐3, ML‐3, LM‐2, LR‐2) and main fracture line length with PVL after PVP was first revealed by a 3D CT reconstruction‐based fracture line mapping technique. In addition, we proposed the existence of fracture line‐vein traffic branches in patients with acute OVCFs within 2 weeks after injury, and confirmation of the existence of this structure may be a subsequent research direction. In conclusion, we believe that the application of this 3D reconstruction technique should not be limited to extremity fractures but should also be emphasized in spinal fractures and that further studies will lead to better treatment strategies, such as the optimized puncture protocol proposed in this study. Nevertheless, it must be acknowledged that the current process of using this technology is complex and will inevitably affect its widespread use in clinical treatment; thus, subsequent updating and simplification of the technology is also extremely necessary.

Conflict of Interest Statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Ethics Statement

This retrospective study was approved by the Medical Ethics Committee of the China‐Japan Union Hospital of Jilin University (No. 2023083003). Written informed consent was obtained from each participant.

Author Contributions

Fan Yang, Boyin Zhang, and Qingsan Zhu contributed to the study design. Fan Yang and Boyin Zhang contributed to drafting the manuscript. Fan Yang, Zhengang Liu, and Pengfu Li contributed to the data extraction and statistical analysis. Qinwan He and Yuling Liang contributed to the data extraction. Boyin Zhang and Qingsan Zhu contributed to the supervision. Fan Yang and Boyin Zhang contributed to the response to reviewers. All authors contributed to the review and revision of the manuscript. All authors have read and approved the final manuscript.

Funding Information

This study was funded by the Bethune Project of Jilin University (2023B22) and the Jilin Province High Quality Education Project (JGJX2023C3).

Consent to Participate

Informed consent was obtained from all individual participants included in the study.

Data Availability Statement

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.