Application Study of Interpretable Machine Learning Models for Predicting Postoperative Refracture After Vertebral Augmentation in Osteoporotic Vertebral Compression Fractures

Application Study of Interpretable Machine Learning Models for Predicting Postoperative Refracture After Vertebral Augmentation in Osteoporotic Vertebral Compression Fractures | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Application Study of Interpretable Machine Learning Models for Predicting Postoperative Refracture After Vertebral Augmentation in Osteoporotic Vertebral Compression Fractures zongjie guo, junping bao, lei zhang, rui shi, shu yang, lei zhang, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8863516/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Osteoporotic vertebral compression fractures (OVCFs) are severe osteoporosis complications; vertebral augmentation is the preferred minimally invasive treatment, but postoperative refracture risk exists. Traditional logistic regression fails to accurately assess individual risks, while interpretable machine learning (ML) excels in high-dimensional data processing, with limited relevant studies. Purposes: To develop an interpretable ML prediction model for identifying risk factors of subsequent fractures after vertebral augmentation in patients with OVCFs. Methods A retrospective analysis was conducted on clinical data of 1,502 OVCF patients who underwent vertebral augmentation. Thirty-six characteristic indicators were extracted from electronic medical records and imaging systems. Six ML prediction models were constructed. Prediction performance was comprehensively evaluated using receiver operating characteristic (ROC) curves, accuracy, recall, F1 score, precision, calibration curves, and decision curve analysis. The optimal model was interpreted globally and locally via Shapley Additive exPlanations (SHAP) to analyze the contribution of key features. Results The 2-year post-operative subsequent fracture incidence in the study cohort was 9.65% (145 cases). After data preprocessing and model training, the extreme gradient boosting (XGBoost) model demonstrated the best performance on the test set. Calibration curve and decision curve analyses showed high consistency between predicted results and actual risks, with significant clinical net benefit. SHAP analysis identified nine key risk factors ranked by importance: age, bone cement leakage and types, history of osteoporosis, Previous history of fractures, bone mineral density, thoracolumbar fascitis, types of trauma, duration of surgery, and Braden score. Conclusions The XGBoost model combined with SHAP represents an effective tool for predicting subsequent fracture risk after vertebral augmentation in OVCF patients. Clinical application of this prediction model can assist clinicians in formulating individualized intervention strategies, thereby optimizing treatment protocols and post-operative management to reduce post-operative subsequent fracture incidence. Osteoporotic vertebral compression fractures Vertebral augmentation Refracture Interpretable machine learning Risk prediction model Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Osteoporosis (OP), the most common clinical disease of the skeletal system, is pathologically characterized by decreased bone mass and destruction of the bone microarchitecture, which in turn leads to increased bone fragility and a higher fracture risk [ 1 , 2 ]. With the aging of the global population, osteoporotic fracture poses an increasingly serious threat to the health of middle-aged and elderly people [ 3 ]. Osteoporotic fractures are the most serious complication of osteoporosis [ 4 , 5 ], occurring in the vertebrae, hips, and pelvis, with vertebral fractures being the most frequent. Osteoporotic vertebral compression fractures (OVCFs) have an insidious onset, and their incidence has been on the rise in recent years [ 3 ]. The principles of clinical intervention for OVCFs have been clearly defined, including fracture reduction and fixation, functional rehabilitation, and anti-osteoporotic treatment [ 7 ]. Percutaneous vertebroplasty (PVP) and percutaneous kyphoplasty (PKP) are preferred for strengthening the vertebral column when non-surgical treatments are not effective. Despite the widespread use of vertebral body strengthening surgical treatment, postoperative vertebral re-fracture is still common and can lead to repeated pain and treatment needs [ 8 – 10 ]. Previous studies have used logistic regression and multifactorial analysis to identify risk factors for recurrent fractures and develop predictive models; however, such models are difficult to accurately assess the expected risk of an individual [ 11 ]. Machine learning (ML), as a multidisciplinary technology, has revolutionized the epidemiological research paradigm by modeling and analyzing the complex associations between predictor and response variables [ 12 – 26 ]. ML algorithms powered by Shapley Additive exPlanation (SHAP) can quantify feature contributions and visualize global and local interpretations of models with better accuracy than ordinary linear models [ 27 , 28 ]. However, there are limited studies based on interpretable machine learning algorithms for modeling and predicting the risk of recurrent vertebral fractures after vertebral body strengthening in patients with OVCFs. Suppose such a prediction model can be successfully constructed. In that case, it can not only achieve the early and accurate identification of high-risk groups but also formulate early and personalized intervention strategies based on individual risk factors, which can reduce the risk of recurrent fractures, the consumption of healthcare resources, and conflicts between doctors and patients, as well as reduce the burden on both individuals and the public health system. Methods 1.1 Study population This is a retrospective study based on clinical data, and the flowchart of the study is shown in Fig. 1. The medical records and imaging data of 1502 patients who were admitted for vertebral body enhancement surgery due to OVCFs at the Center for Spine Surgery, CU Hospital, Southeast University, from January 2014 to December 2022 were retrospectively collected. The inclusion criteria were as follows: (1) patients with a precise diagnosis of OVCFs requiring surgical intervention and (2) undergoing vertebral body strengthening surgery for the first time. The exclusion criteria were as follows: (1) non-osteoporotic vertebral compression fractures; (2) pathologic compression fractures of the vertebral body secondary to other factors, such as tumors and infections; (3) patients who did not undergo vertebral body enhancement surgery or in combination with other surgical modalities; (4) patients who had a combination of systemic or localized infections; (5) patients who had been in combination with other surgical modalities; and (6) patients who died within the observation time window (within 2 years) after the first surgery or patients for whom valid follow-up information could not be obtained. 1.2 Calculation of sample size In recent years, Riley et al. [ 29 ] proposed a widely recognized scientific method to determine the sample size in prediction models. In this study, the pmsampsize package for clinical predictive modeling in the R language, version 4.4.0, was used, and the sample size was calculated based on the guidelines provided above. The incidence of recurrent fractures after vertebral body strengthening was approximately 9.6%. A total of 36 predictor variables were included in this study, which was calculated using software to determine a minimum sample size of 1,189 patients required for this study. After combining the inclusion and exclusion criteria, 1502 patients were finally included in the study. 1.3 Inclusion characteristics By searching the hospital's medical record system and imaging system, and integrating general patient information, laboratory test results, surgery-related data, and pre- and post-operative imaging data, a total of 36 characteristic indicators were ultimately included. To ensure consistency in typing criteria, three spine surgery experts discussed and established uniform criteria for the imaging data. The experts were also double-blind to the characteristics and statistical analysis of the study subjects. All surgeries were performed by our senior spine surgical team, which has more than 10 years of surgical experience. The surgeons strictly followed standardized protocols and completed relevant training before the study to ensure consistency and minimize bias. 1.4 Statistical analysis IBM SPSS 29.0 statistical software was used for the statistical analysis of the data in this study. Measurement information was expressed as raw data, and for variables that obeyed normal distribution, a one-sample t-test was used. For variables that do not follow a normal distribution, the rank sum test was used. Comparisons of categorical variables were performed using the chi-square test or Fisher's exact probability test. A two-sided P value < 0.05 was considered statistically significant. 1.5 Data Preprocessing Data preprocessing is the foundation for ensuring data quality and model accuracy. In this study, the IterativeImputer module of the Python Scikit-Learn library is used to perform multiple interpolations to mitigate the impact of missing values on the robustness of the results. The feature scales are then unified through standardization of deviation and normalization to ensure data distribution consistency. The dataset is randomly divided into a training set and a test set with a ratio of 7:3. If there is an imbalance in sample distribution, oversampling or undersampling techniques are used for optimization. Finally, exploratory data analysis is conducted through data visualization to reveal the distribution laws of the features. 1.6 Model Development This study develops machine learning models based on Anaconda3 2023.07 (64-bit) and Python 3.10.0, including XGBoost, Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), and Artificial Neural Networks (ANN). Among them, XGBoost is implemented using a proprietary library that relies on parallel computing and regularization for efficient training. Grid search, combined with 5-fold cross-validation, is used to select the hyperparameters of the training set. The DT, RF, SVM, and LR models are then developed using Scikit-Learn. All models are evaluated on a test set to assess their efficacy in predicting real-world scenarios. Referring to previous studies [ 33 , 34 ], we used accuracy, precision, recall, F1 score, and Area Under the Curve (AUC) as the evaluation metrics in this study; analyzed the difference between predicted and actual results by calibration curves [ 35 ]; and evaluated the net clinical benefit by Decision Curve Analysis (DCA) [ 36 ] to comprehensively screen the best machine learning prediction models. The best machine learning prediction model was selected. 1.7 Model Interpretability Assessment The SHAP technique devised by Lundberg and Lee [ 27 ] significantly improves the interpretability of complex algorithms by mapping machine learning feature SHAP values to clinical variables. The method is based on the Shapley value concept of cooperative game theory, which assigns specific contribution values to each feature to help explain the “black box” mechanism. In this study, the SHAP values of the features are calculated using the SHAP toolkit in Python 3.10.0, and the relative importance of the features is assessed by comparing the difference in prediction results between the presence and absence of the features [ 28 ]. 1.8 Ethical Review This study was ethically reviewed by the Ethics Committee of Zhongda Hospital, affiliated with Southeast University, Ethics Approval No: 2024ZDSYLL187-P01. Results 2.1 Characteristics of the study population The general information and laboratory examination details of the study population are presented in Table 1 . A comparison of surgical and imaging data between the recurrent fracture group after vertebral body strengthening (RF group) and the non-recurrent fracture group (NRF group) is shown in Table 2 . A total of 1,502 patients were included in the present study, of whom 145 (9.65%) had a recurrent fracture after surgery (RF group), and the remaining patients were in the NRF group. Table 1 Demographic characteristics and laboratory examination of study participants Variables NRF RF χ2/t/Z P (n = 1357) (n = 145) Sex (n%) 0.009 0.926 Male 267(19.7) 29(20.0) Female 1090(80.3) 116(80.0) Age (n%) 282.634 < 0.001 <70 years 958(70.6) 0(0.0) ≥ 70 years 399(29.4) 145(100.0) BMI (kg/m²) 23.01(21.64,25.83) 22.73(21.21,25.33) 0.702 0.661 Trauma (n%) 71.399 < 0.001 No 362(26.7) 81(55.9) Concealed trauma 386(28.4) 46(31.7) Obvious trauma 609(44.9) 18(12.4) Hypertension (n%) 0.023 0.878 No 739(54.5) 78(53.8) Yes 618(45.5) 67(46.2) Diabetes (n%) 6.305 0.092 No 1161(85.6) 128(88.3) Yes 196(14.4) 15(11.7) Heart disease (n%) 1.852 0.174 No 1203(88.7) 123(84.8) Yes 154(11.3) 22(15.2) Cranial disease (n%) 6.206 0.073 No 1132(83.4) 109(75.2) Yes 225(16.6) 36(24.8) Osteoporosis (n%) 141.588 < 0.001 No 704(51.9) 0(0.0) Yes 653(48.1) 145(100.0) Previous history of fractures (n%) 498.414 < 0.001 No 1239(91.3) 30(20.7) Yes 118(8.7) 115(79.3) Segment of freshly fractured vertebrae (n%) 30.418 < 0.001 Above T12 level 157(11.6) 21(14.5) T12 289(21.3) 24(16.6) L1 388(28.6) 20(13.8) L2 189(13.9) 22(15.2) L3 63(4.6) 4(2.8) L4 48(3.5) 9(6.2) L5 5(0.4) 0(0.0) Multiple segments 218(16.1) 45(31.0) Number of freshly fractured vertebrae (n) 1(1,1) 1(1,1) -2.837 0.065 Braden score (n) 20(18,21) 18(17,20) -1.219 0.023 Fall risk assessment (n) 3(2,4) 3(2,4) -2.205 0.057 VAS (n) 3(2,3) 4(3,5) -2.852 0.004 Hb (10^12/L) 128(120,136) 122(115.5,129) -5.224 < 0.001 Ca (mmol/L) 2.22(2.16,2.31) 2.23(2.14,2.3) -0.121 0.904 ALB (g/L) 39.7(37.25,43) 40.3(35.85,44.6) -0.033 0.974 BUN (mmol/L) 5.6(4.7,7.1) 6.2(4.55,7.2) -1.011 0.312 Cr (µmol/L) 58(48,71) 60(43.5,73.5) -0.503 0.615 Table 2 Surgery-related and radiographic factors in study participants Variables NRF RF χ2/t/Z P (n = 1357) (n = 145) Duration of surgery (min) 40(31,50) 45(35,60) -2.172 0.030 Surgical methods (n%) 1.961 0.195 PVP 669(49.3) 73(50.3) PKP 688(50.7) 72(49.7) Method of puncture (n%) 1.477 0.224 Unilateral puncture 384(28.3) 48(33.1) Bilateral puncture 973(71.7) 97(66.9) Bone cement dosage (ml) 6(6,9) 6(5.75,8) -1.315 0.189 Bone cement leakage (n%) 280.326 < 0.001 No 1045(77.0) 15(10.3) Yes 312(23.0) 130(89.7) Types of bone cement leakage (n%) 341.644 < 0.001 No 1045(77.0) 15(10.4) I 184(13.6) 45(31.0) II 63(4.6) 31(21.4) III 13(1.0) 10(6.9) IV 23(1.7) 19(13.1) V 29(2.1) 25(17.2) Bone cement distribution type (n%) 6.922 0.265 I 978(72.1) 101(69.7) II 94(6.9) 10(6.9) III 88(6.5) 10(6.9) IV 128(9.4) 14(9.7) V 69(5.1) 10(6.9) Contact with the endplate (n%) 40.842 < 0.001 No 922(67.9) 60(41.4) Yes 435(32.1) 85(58.6) Thoracolumbar fascitis (n%) 176.892 < 0.001 No 1068(78.7) 40(27.6) Yes 289(21.3) 105(72.4) Scoliosis (n%) 17.696 0.158 No 1153(85.0) 113(77.9) Yes 204(15.0) 32(22.1) AVHRR (%) 0.1(0.02,0.22) 0.09(0.04,0.28) -1.157 0.247 Cobb (°) 1(0,5.5) 2(-2,4) -1.154 0.249 PI (°) 50(41,57) 50(43.5,61.5) -1.459 0.144 PT (°) 18(10.5,26) 18(11,27) -1.133 0.257 SS (°) 31(25,38.5) 33(27,41.5) -1.617 0.106 BMD (g/cm^2) -3(-3.8,-2.2) -4(-4.65,-2.85) -8.527 < 0.001 In terms of baseline characteristics, the proportion of patients aged 70 years or older was significantly higher in the RF group (100.0%) than in the NRF group (29.4%, P < 0.001), and the proportion of patients with no obvious history of traumatic injury was higher (55.9% vs. 26.7%, P < 0.001). The proportion of patients with a history of previous fracture was significantly higher in the RF group (79.3%) than in the NRF group (8.7%, P < 0.001), and the Braden score was lower (P = 0.001). The Braden score was lower (P = 0.023), the VAS score was higher (P = 0.004), and the hemoglobin concentration was lower (P < 0.001). Regarding fracture-related characteristics, the proportion of osteoporosis history in the RF group was 100.0%, significantly higher than that in the NRF group (48.1%, P < 0.001); the distribution of fresh fracture segments differed significantly (P < 0.001), with 31.0% of the RF group suffering from multisegmental fractures, whereas the NRF group had a predominance of T12 (21.3%) and L1 (28.6%) segments. Regarding surgical and imaging characteristics, the surgical time was longer in the RF group (P = 0.030); the incidence of cement leakage was significantly higher (89.7% vs. 23.0%, P < 0.001), and there was a significant difference in the type of leakage (P < 0.001), with a predominantly non-leakage in the NRF group (77.0%). In addition, the RF group had a higher proportion of cement contacting the endplate (58.6% vs 32.1%, P < 0.001), a higher proportion of lumbar dorsal musculoskeletal fasciitis (72.4% vs 21.3%, P < 0.001), and a lower BMD value (P < 0.001). 2.2 Performance of the model A total of 1,502 patients were included in this study, comprising 145 in the refracture (RF) group and 1,357 in the non-refracture (NRF) group. To address the imbalance in the proportion of original data categories and the limited sample size, the study expanded the dataset through various data enhancement techniques. At the same time, it scientifically divided the training set and the test set according to a 7:3 ratio to ensure the accuracy of model evaluation. The final test set comprises 815 cases, with sample sizes of the two groups essentially balanced. The prediction performance of the six models on the test set is evaluated. The results are shown in Table 3 and Fig. 2. All models show good classification ability in the training stage, among which the XGBoost model is outstanding in recurrent fracture prediction, with various indexes significantly better than the other models, including precision = 0.9926, recall = 0.9951, accuracy = 0.9939, F1 score = 0.9924, and AUC = 0.9996. The confusion matrix results in Fig. 3 further confirm that the XGBoost model has the best classification effect; the calibration curve in Fig. 4 shows that its curve is closer to the 45-degree perfect calibration line; the decision curve analysis (DCA) results in Fig. 5 show that the net benefit of XGBoost is stable in most intervals in the training set, and the validation set shows good prediction accuracy and potential for clinical application, with significant advantages in practical decision-making. In the validation set, the XGBoost model shows good prediction accuracy and potential for clinical application, with substantial advantages in practical decision-making. Overall, the XGBoost model performed the best among the six models and was the most accurate prediction model in this study. Table 3 Performance of six machine learning models for the testing set (n = 815) Model Accuracy Recall Precesion F1-score AUC eXtreme Gradient Boosting 0.9926 0.9951 0.9939 0.9924 0.9996 Decision Tree 0.9901 0.9829 0.9865 0.9828 0.9478 Random Forest 0.9878 0.9878 0.9877 0.9878 0.9988 Logistic Regression 0.9854 0.9902 0.9877 0.9878 0.9991 Support Vector Machine 0.9878 0.9951 0.9914 0.9924 0.9982 Artificial Neural Networks 0.9806 0.9951 0.9877 0.9878 0.9985 2.3 Interpretability of the model The XGBoost model combined with the SHAP method can provide both global and local interpretable results. A total of 36 features were included in the model. The prediction results are presented in Fig. 6A, where rows represent specific features, dots represent samples, and colors distinguish feature values (red for high values and blue for low values). Figure 6B lists the 9 key features in order of importance, including: age, cement leakage and its imaging typology, history of previous osteoporosis, history of prior fracture, bone mineral density level, lumbar dorsal musculoskeletal fasciitis, trauma type, length of surgery, and Braden score. With the help of SHAP force maps, the ML combined with the SHAP approach can further elucidate the mechanism of the influence of each feature on the individual prediction results, and Fig. 7 demonstrates the SHAP force maps of two cases, which quantify the contribution of each feature to the prediction of the XGBoost model through the SHAP value. Case 1 (Fig. 7A) was predicted to have a high risk of recurrent fracture after vertebral body strengthening after analyzing the effects of all factors, which was consistent with the actual postoperative recurrent fracture results of this patient; the prediction results of Case 2 (Fig. 7B) were also consistent with the actual situation. Discussion Postoperative re-fracture of OVCFs is a challenging aspect of clinical treatment, significantly impacting patients' recovery and quality of life [ 8 , 9 ]. Traditional risk prediction studies have limited accuracy and are challenging to meet the demand for precision medicine. Although ML has demonstrated its advantages in the diagnosis and treatment of lumbar degenerative diseases in spine surgery, it is still in the exploratory stage in the prediction of postoperative recurrent fracture risk in OVCFs. In this study, ML was introduced into this field for the first time. Multiple algorithmic models were constructed by collecting multidimensional diagnostic and treatment data, which were then combined with SHAP analysis to improve the interpretability of the models. This approach provided new insights for the early identification of high-risk populations and the formulation of personalized intervention plans, ultimately contributing to the development of effective diagnosis and treatment strategies for OVCFs. XGBoost, a machine learning model integrated with a decision tree, can solve various types of regression, classification, and ranking problems [ 37 , 38 ], and is highly efficient and effective for predicting postoperative fracture risk. XGBoost is an integrated decision tree machine learning model that can solve various regression, classification, and ranking problems [ 37 , 38 ], offering significant advantages in terms of efficiency, accuracy, scalability, flexibility, and stability [ 39 – 43 ]. The SHAP method, as a powerful model interpretation tool, can significantly improve the interpretability of ML algorithms [ 27 , 28 ]. In this study, we combined the two, realized the integration of predictive validity and interpretability, and clarified the key predictive indices affecting re-fracture, which provides support for precise clinical prevention and treatment. This study confirms that advanced age is a closely associated factor for postoperative re-fracture in OVCFs, which is consistent with the findings of several clinical studies [ 44 – 46 ], and the core mechanism lies in the imbalance of bone metabolism in advanced-aged patients [ 47 – 52 ] and the impact of age-related muscle dysfunction on spinal stability [ 53 – 55 ]. Although advanced age is a recognized risk factor, age stratification criteria have not been standardized. In this study, we propose 70 years as the threshold value, which is both scientifically and clinically valuable: People over 70 years of age enter the stage of severe osteoporosis, and the incidence of fragility fracture is 2–3 times higher than that of people over 60 years of age [ 56 ]; people between the ages of 50 and 70 years of age may have early-onset osteoporosis, which is not very relevant to their age [ 57 ]; people over the age of 80 years of age have a high heterogeneity of samples due to the co-occurrence of multiple diseases, and the generalizability of the study is limited. Therefore, osteoporosis in people over 70 years of age more closely resembles age-associated natural degeneration, highlighting the value of age as an independent risk factor. Cement leakage is a common complication after vertebral body augmentation [ 58 ], and its status and type are key risk factors for re-fracture [ 59 , 60 ], especially when the leakage occurs in the intervertebral disc, where the risk of re-fracture of the adjacent vertebrae is significantly higher [ 61 ]. From a biomechanical perspective, the mechanism varies among different leakage sites. Leakage from the intervertebral disc disrupts the elastic cushioning function, leading to abnormal stress distribution [ 62 – 64 ]. Leakage from the spinal canal or neural foramina compresses the nerve roots, forcing a change in the spinal force pattern and disrupting the mechanical equilibrium [ 65 ]. Leakage from the paravertebral veins does not directly affect vertebral mechanics. Still, it may lead to complications, such as pulmonary embolism, and indirectly increases the risk by decreasing activity and accelerating bone mass loss [ 66 , 67 ]. Currently, there is a lack of uniform criteria for classifying bone cement leakage, and previous studies have primarily analyzed it as a binary variable, overlooking its biomechanical heterogeneity. Therefore, large-sample, multicenter studies are urgently needed to establish a standardized system for clarifying the impact of each subtype. In this study, all patients in the re-fracture group were diagnosed with osteoporosis for the first time. In conjunction with previous studies, it has been shown that a history of previous osteoporosis is an independent risk factor for postoperative re-fracture in patients with OVCFs, i.e., the risk of re-fracture in those who were diagnosed before the fracture was significantly higher than that of those who were diagnosed for the first time at the time of the current fracture [ 68 – 70 ]. This seemingly paradoxical phenomenon may be related to multiple factors: on the one hand, early diagnosed patients have a longer disease duration, and long-term bone loss and trabecular destruction lead to a significant decrease in bone strength and toughness [ 71 , 72 ]; on the other hand, although early diagnosed patients receive anti-osteoporosis treatment, 75.8% of them still have the problem of suboptimal bone density or persistent symptoms after two years of treatment [ 73 ], and they may have had preoperative occult minor fractures, whereas initiation of standardized interventions for acute fractures in those diagnosed for the first time reduces the risk [ 74 , 75 ]. Numerous studies have consistently shown that a history of previous fractures is a key risk factor for postoperative re-fracture in patients with OVCFs. A history of previous fracture significantly elevates the risk of future fracture, with the predictive value of a history of previous vertebral fracture being particularly significant [ 59 , 76 ]; specifically, the risk of re-fracture in those with a history of fracture is approximately two times higher than that in those without a history of fracture, whereas in those with a previous vertebral fracture, their risk of re-fracture increases to four times [ 77 , 78 ]. Mechanistically, previous fractures can severely damage the normal structure of bones and decrease the mechanical properties of newly formed bone tissues [ 79 – 81 ]; at the same time, fracture trauma can disrupt the balance of bone metabolism and further exacerbate the condition [ 82 – 84 ]. Therefore, the clinic should inquire in detail about previous fracture history, develop a more aggressive anti-osteoporosis program for individuals with a history of fracture, and enhance postoperative rehabilitation and follow-up. This study confirmed that BMD is a significant risk factor for postoperative re-fracture in OVCFs, consistent with the findings of several previous studies [ 76 , 85 ]. BMD T-value < -2.2 SD effectively predicted subsequent fracture, and the incidence of re-fracture was significantly higher at T-value ≤ -2.5 SD [ 86 ]. In addition, for every 1 SD decrease in BMD, the risk of fracture increased 1.4–1.8 times [ 87 – 89 ]. Bone strength and toughness are significantly reduced in patients with low BMD, which makes it difficult to withstand daily stresses. If the bone metabolic imbalance is not corrected early in the postoperative period, the stress resistance and microinjury repair capacity are reduced, which further increases the risk of re-fracture [ 90 – 94 ]. Therefore, patients with OVCFs should pay attention to BMD monitoring and management after surgery and enhance BMD levels through active anti-osteoporotic treatment to reduce the risk of re-fracture. Lumbar dorsal fasciitis is a high-risk factor for postoperative re-fracture in patients with OVCFs. Paravertebral muscle mass is significantly reduced in patients with refracture after percutaneous vertebral kyphoplasty [ 96 , 97 ], emphasizing its importance in the treatment of OVCFs. As a core structure to maintain spinal stability, the lumbar dorsal muscles maintain spinal mechanical balance through coordinated contraction [ 98 , 99 ]; whereas, lumbar dorsal fasciitis can trigger muscle spasms, leading to spinal mechanical conduction malfunction and stress concentration, which is susceptible to microfractures in the long term [ 100 ]. Subsequent studies can explore the molecular mechanisms in depth and search for early diagnostic indicators and intervention targets to improve lumbar dorsal muscle function, attenuate the inflammatory response, and reduce the risk of postoperative re-fracture. This study found that the type of trauma was closely related to re-fracture. Patients who did not experience significant trauma or low-energy trauma had a significantly higher risk of postoperative re-fracture than those with high-energy trauma [ 101 ]. Clinical studies have confirmed that the risk of re-fracture is significantly higher in the population with low-energy trauma fractures [ 102 , 103 ], and most of them have severe osteoporosis, which can be induced by slight external forces [ 104 ]. In contrast, the fractures in patients with high-energy trauma are mostly due to overloading of external forces, and the underlying bone mass may not be poor. They still retain a certain degree of mechanical support and repair potential after surgery. Therefore, for low-energy trauma patients, the focus needs to be on improving bone quality, strengthening postoperative rehabilitation, and anti-osteoporosis treatment. Surgery is the key treatment for OVCFs [ 105 ], but prolonged surgery is an independent risk factor for postoperative re-fracture. Previous studies have shown that the excessive duration of surgery is also an independent risk factor for cement leakage [ 106 ], suggesting that it may directly or indirectly affect the risk of re-fracture. Prolonged surgery will increase periprosthetic tissue damage and cause abnormal postoperative stress distribution [ 107 ]; simultaneously, it will increase the difficulty of bone cement operation and the likelihood of leakage [ 108 , 109 ]. Therefore, clinical optimization of surgical procedures and techniques is needed to improve efficiency and safety. The Braden score is widely used to predict the risk of pressure ulcers [ 110 ], and its correlation with postoperative refracture in patients with OVCFs has not been confirmed in previous studies. In this study, we hypothesized that some of the indicators in the score may indirectly indicate risk, such as poor nutritional status, which affects bone metabolism and healing and increases the likelihood of re-fracture [ 111 , 112 ]. Although the Braden score does not directly predict re-fracture, it can be used to focus on the risk of pressure ulcers, and preventive measures can be taken to indirectly help patients recover and reduce the potential risk of re-fracture. This study has the following limitations: first, as a retrospective study, there is a potential risk of selection bias and incomplete data collection. Second, the ML model was constructed based on single-center clinical data, and the institution-specific diagnosis and treatment patterns and regional population characteristics may limit the generalization ability of the model; also, the model was only validated by an internal dataset, and it needs to be further extrapolated and tested by an independent external dataset. Third, anti-osteoporosis medication was not included in the study model and analysis due to inconsistencies in the type of anti-osteoporosis medication used by different patients and the lack of documentation for some medication details, which made it difficult to assess the impact of this factor on the results. Conclusions In this study, we focused on predicting the risk of recurrent fracture after vertebral body strengthening in patients with OVCFs. For the first time, we used an interpretable ML method to construct a risk prediction model based on multidimensional clinical characteristics. The results showed that the XGBoost prediction model has excellent ability to predict the risk of postoperative recurrent fracture; meanwhile, key risk factors were identified by SHAP analysis, which provides a quantitative basis for accurate risk assessment. This technology can help clinicians clarify the decision logic of the model, identify high-risk groups and formulate individualized interventions, which is of great significance for optimizing treatment strategies and postoperative management modes, reducing the incidence of re-fracture, and providing a new methodological reference for the practice of precision medicine in OVCFs. Declarations Disclosure of potential conflicts of interest ZongJie Guo, JunPing Bao, Rui Shi, Shu Yang, and Lei Zhang declare that they have no conflict of interest. Ethics approval and consent to participate This study has passed the ethical review of the Ethics Committee of Zhongda Hospital Affiliated to Southeast University, and has been approved to exempt informed consent. The ethics approval number was 2024ZDSYLL187-P01. Consent for publication Not applicable. Clinical trial number: Not applicable. Author contributions ZongJie Guo, JunPing Bao, and Lei Zhang contributed to the study conception and design. Data collection was performed by ZongJie Guo. Analyses were performed by ZongJie Guo, JunPing Bao, and Lei Zhang. The first draft of the manuscript was written by ZongJie Guo. Rui Shi, and Shu Yang evaluated the patients' X-ray, CT and MRI images. Lei Zhang and JunPing Bao commented on the previous versions of the manuscript. All authors have contributed to the manuscript and approved the submitted version. All authors have reviewed the final version of the manuscript and approved it for publication. Funding 2023 Jiangsu Health Development Research Center Open Project (JSHD202312); Jiangsu Province Capability Improvement Project through Science, Technology and Education (ZDXYS202210); Jiangsu Province High-level Hospital Construction Funds; project number: CZXM-GSP-RC53. Author Contribution ZongJie Guo, JunPing Bao, and Lei Zhang contributed to the study conception and design. Data collection was performed by ZongJie Guo. Analyses were performed by ZongJie Guo, JunPing Bao, and Lei Zhang. The first draft of the manuscript was written by ZongJie Guo. Rui Shi, and Shu Yang evaluated the patients' X-ray, CT and MRI images. Lei Zhang and JunPing Bao commented on the previous versions of the manuscript. All authors have contributed to the manuscript and approved the submitted version. All authors have reviewed the final version of the manuscript and approved it for publication. Data Availability The datasets analyzed in this study can be obtained from the corresponding authors with reasonable requirements. References Shubhashree M, Naik R, Doddamani S, et al. An updated review of single herbal drugs in the management of osteoporosis [J]. Int J Complement Altern Med, 2018, 11: 82–86. Sakat B T, Sakhare R B, Suryvanshi U C, et al. Osteoporosis: The brittle bone [J]. Asian Journal of Pharmaceutical Research, 2018, 8(1): 39–43. Si L, Winzenberg T M, Jiang Q, et al. Projection of osteoporosis-related fractures and costs in China: 2010–2050 [J]. Osteoporos Int, 2015, 26(7): 1929–1937. Johnston C C, Longcope C. Premenopausal bone loss–a risk factor for osteoporosis [J]. N Engl J Med, 1990, 323(18): 1271–1273. Khan A A, Slart R H J A, Ali D S, et al. Osteoporotic Fractures: Diagnosis, Evaluation, and Significance From the International Working Group on DXA Best Practices [J]. Mayo Clin Proc, 2024, 99(7): 1127–1141. Siris E S, Adler R, Bilezikian J, et al. The clinical diagnosis of osteoporosis: a position statement from the National Bone Health Alliance Working Group [J]. Osteoporos Int, 2014, 25(5): 1439–1443. Wang G, Yang L, Yang C, et al. Expert Consensus on Bone Repair Strategies for Osteoporotic Vertebral Compression Fractures[J]. Journal of Clinical Surgery, 2024, 32(04): 442–448. Voormolen M H J, Lohle P N M, Juttmann J R, et al. The risk of new osteoporotic vertebral compression fractures in the year after percutaneous vertebroplasty [J]. J Vasc Interv Radiol, 2006, 17(1): 71–76. Klazen C A H, Venmans A, de Vries J, et al. Percutaneous vertebroplasty is not a risk factor for new osteoporotic compression fractures: results from VERTOS II [J]. AJNR Am J Neuroradiol, 2010, 31(8): 1447–1450. Lindsay R, Burge R T, Strauss D M. One year outcomes and costs following a vertebral fracture [J]. Osteoporos Int, 2005, 16(1): 78–85. Chen Z, Yao Z, Wu C, et al. Assessment of clinical, imaging, surgical risk factors for subsequent fracture following vertebral augmentation in osteoporotic patients [J]. Skeletal Radiol, 2022, 51(8): 1623–1630. Jordan M I, Mitchell T M. Machine learning: Trends, perspectives, and prospects [J]. Science, 2015, 349(6245): 255–260. Mishra R K, Reddy G S, Pathak H. The understanding of deep learning: A comprehensive review [J]. Mathematical Problems in Engineering, 2021, 2021(1): 5548884. Abdollah V, Parent E C, Dolatabadi S, et al. Texture analysis in the classification of T2 -weighted magnetic resonance images in persons with and without low back pain [J]. J Orthop Res, 2021, 39(10): 2187–2196. Huber F A, Stutz S, Vittoria de Martini I, et al. Qualitative versus quantitative lumbar spinal stenosis grading by machine learning supported texture analysis-Experience from the LSOS study cohort [J]. Eur J Radiol, 2019, 114: 45–50. Han Z, Wei B, Leung S, et al. Automated Pathogenesis-Based Diagnosis of Lumbar Neural Foraminal Stenosis via Deep Multiscale Multitask Learning [J]. Neuroinformatics, 2018, 16(3–4): 325–337. Levitt J, Edhi M M, Thorpe R V, et al. Pain phenotypes classified by machine learning using electroencephalography features [J]. Neuroimage, 2020, 223: 117256. Oude Nijeweme-d'Hollosy W, van Velsen L, Poel M, et al. Evaluation of three machine learning models for self-referral decision support on low back pain in primary care [J]. Int J Med Inform, 2018, 110: 31–41. Rak M, Steffen J, Meyer A, et al. Combining convolutional neural networks and star convex cuts for fast whole spine vertebra segmentation in MRI [J]. Comput Methods Programs Biomed, 2019, 177: 47–56. Kuang X, Cheung J P, Wu H, et al. MRI-SegFlow: a novel unsupervised deep learning pipeline enabling accurate vertebral segmentation of MRI images [J]. Annu Int Conf IEEE Eng Med Biol Soc, 2020, 2020: 1633–1636. Li X, Dou Q, Chen H, et al. 3D multi-scale FCN with random modality voxel dropout learning for Intervertebral Disc Localization and Segmentation from Multi-modality MR Images [J]. Med Image Anal, 2018, 45: 41–54. Paugam F, Lefeuvre J, Perone C S, et al. Open-source pipeline for multi-class segmentation of the spinal cord with deep learning [J]. Magn Reson Imaging, 2019, 64: 21–27. Wirries A, Geiger F, Hammad A, et al. Artificial intelligence facilitates decision-making in the treatment of lumbar disc herniations [J]. Eur Spine J, 2021, 30(8): 2176–2184. Roller B L, Boutin R D, O'Gara T J, et al. Accurate prediction of lumbar microdecompression level with an automated MRI grading system [J]. Skeletal Radiol, 2021, 50(1): 69–78. Karhade A V, Fogel H A, Cha T D, et al. Development of prediction models for clinically meaningful improvement in PROMIS scores after lumbar decompression [J]. Spine J, 2021, 21(3): 397–404. Guo Z, Wang P, Ye S, et al. Interpretable Machine Learning Models Based on Shapley Additive Explanations for Predicting the Risk of Cerebrospinal Fluid Leakage in Lumbar Fusion Surgery [J]. Spine (Phila Pa 1976), 2024, 49(18): 1281–1293. Chen H, Lundberg S M, Lee S-I. Explaining a series of models by propagating Shapley values [J]. Nat Commun, 2022, 13(1): 4512. Sun J, Sun C-K, Tang Y-X, et al. Application of SHAP for Explainable Machine Learning on Age-Based Subgrouping Mammography Questionnaire Data for Positive Mammography Prediction and Risk Factor Identification [J]. Healthcare (Basel), 2023, 11(14). Riley R D, Ensor J, Snell K I E, et al. Calculating the sample size required for developing a clinical prediction model [J]. BMJ, 2020, 368: m441. Li J, Li K, Wang X, et al. Classification and Prevention of Bone Cement Leakage in Percutaneous Kyphoplasty[J]. Chinese Journal of Osteoporosis and Bone Mineral Research, 2009, 2(01): 50–52. Ni W, Chi Y, Lin Y, et al. Types and Clinical Significance of Bone Cement Leakage Complicating Percutaneous Vertebral Augmentation[J]. Chinese Journal of Surgery, 2006, (04): 231–234. Zhang D, Mao K, Qiang X, et al. Classification and Clinical Significance of Bone Cement Distribution Patterns after Vertebral Augmentation[J]. Chinese Journal of Traumatology, 2018, 34(2): 130–137. Ren G, Yu K, Xie Z, et al. Current Applications of Machine Learning in Spine: From Clinical View [J]. Global Spine J, 2022, 12(8): 1827–1840. Wang P, Liu L, Xie Z, et al. Explainable Machine Learning Models for Prediction of Surgical Site Infection After Posterior Lumbar Fusion Surgery Based on Shapley Additive Explanations [J]. World Neurosurg, 2025, 197: 123942. Fenlon C, O'Grady L, Doherty M L, et al. A discussion of calibration techniques for evaluating binary and categorical predictive models [J]. Prev Vet Med, 2018, 149: 107–114. Zhang Z, Rousson V, Lee W-C, et al. Decision curve analysis: a technical note [J]. Ann Transl Med, 2018, 6(15): 308. Takefuji Y. Beyond XGBoost and SHAP: Unveiling true feature importance [J]. J Hazard Mater, 2025, 488: 137382. Chen T, Guestrin C. Xgboost: A scalable tree boosting system; proceedings of the Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, F, 2016 [C]. Al-Zakhali O A, Zeebaree S, Askar S. Comparative Analysis of XGBoost Performance for Text Classification with CPU Parallel and Non-Parallel Processing [J]. The Indonesian Journal of Computer Science, 2024, 13(2). Chen T, Guestrin C. Xgboost: Reliable large-scale tree boosting system; proceedings of the Proceedings of the 22nd SIGKDD Conference on Knowledge Discovery and Data Mining, F, 2017 [C]. Grinsztajn L, Oyallon E, Varoquaux G. Why do tree-based models still outperform deep learning on typical tabular data? [J]. Advances in neural information processing systems, 2022, 35: 507–520. Ali Z A, Abduljabbar Z H, Tahir H A, et al. eXtreme gradient boosting algorithm with machine learning: A review [J]. Academic Journal of Nawroz University, 2023, 12(2): 320–334. Stokanović S, Đukić D, Miljković N. Robustness of XGBoost algorithm to missing features for binary classification of medical data; proceedings of the 2024 23rd International Symposium INFOTEH-JAHORINA (INFOTEH), F, 2024 [C]. IEEE. Feng L, Feng C, Chen J, et al. The risk factors of vertebral refracture after kyphoplasty in patients with osteoporotic vertebral compression fractures: a study protocol for a prospective cohort study [J]. BMC Musculoskelet Disord, 2018, 19(1): 195. Dai C, Liang G, Zhang Y, et al. Risk factors of vertebral re-fracture after PVP or PKP for osteoporotic vertebral compression fractures, especially in Eastern Asia: a systematic review and meta-analysis [J]. J Orthop Surg Res, 2022, 17(1): 161. Lee B G, Choi J-H, Kim D-Y, et al. Risk factors for newly developed osteoporotic vertebral compression fractures following treatment for osteoporotic vertebral compression fractures [J]. Spine J, 2019, 19(2): 301–305. Jung B H, Jeon M J, Bai S W. Hormone-dependent aging problems in women [J]. Yonsei Med J, 2008, 49(3): 345–351. Majeska R J, Ryaby J T, Einhorn T A. Direct modulation of osteoblastic activity with estrogen [J]. J Bone Joint Surg Am, 1994, 76(5): 713–721. Väänänen H K, Härkönen P L. Estrogen and bone metabolism [J]. Maturitas, 1996, 23: S65-S69. Tehrani S S, Moallem M, Ebrahimi R, et al. Status of circulating bone turnover markers in elderly osteoporosis/osteopenia patients in comparison with healthy subjects [J]. Asian Biomed, 2020, 14(3): 97–106. Szulc P, Garnero P, Munoz F, et al. Cross-sectional evaluation of bone metabolism in men [J]. Journal of Bone and Mineral Research, 2001, 16(9): 1642–1650. Garnero P, Hausherr E, Chapuy M C, et al. Markers of bone resorption predict hip fracture in elderly women: the EPIDOS Prospective Study [J]. Journal of bone and mineral research, 1996, 11(10): 1531–1538. Cheng C, Chen P, Kuo Y, et al. The effects of disc degeneration and muscle dysfunction on cervical spine stability from a biomechanical study [J]. Proceedings of the Institution of Mechanical Engineers, Part H: Journal of Engineering in Medicine, 2011, 225(2): 149–157. Mohamed A, Mohamed E, Mohamed J. Exercises Training Program: Its Effect on Muscle Strength and Activity of Daily Living among Elderly People [J]. Evidence-Based Nursing Research, 2019, 1(4): 24–34. Aagaard P, Suetta C, Caserotti P, et al. Role of the nervous system in sarcopenia and muscle atrophy with aging: strength training as a countermeasure [J]. Scandinavian journal of medicine & science in sports, 2010, 20(1): 49–64. Tsuda T. Epidemiology of fragility fractures and fall prevention in the elderly: a systematic review of the literature [J]. Curr Orthop Pract, 2017, 28(6): 580–585. Lukert B P, Raisz L G. Glucocorticoid-induced osteoporosis: pathogenesis and management [J]. Ann Intern Med, 1990, 112(5): 352–364. Zhu S-Y, Zhong Z-M, Wu Q, et al. Risk factors for bone cement leakage in percutaneous vertebroplasty: a retrospective study of four hundred and eighty five patients [J]. International Orthopaedics, 2016, 40: 1205–1210. Mao W, Dong F, Huang G, et al. Risk factors for secondary fractures to percutaneous vertebroplasty for osteoporotic vertebral compression fractures: a systematic review [J]. J Orthop Surg Res, 2021, 16(1): 644. Lin Y-H, Lin J, Xu J-Y, et al. What Risk Factors Are Associated With Recurrent Osteoporotic Vertebral Compression Fractures After Percutaneous Vertebral Augmentation? A Meta-analysis [J]. Clin Orthop Relat Res, 2025. Ma Y, Lu Q, Wang X, et al. Establishment and validation of a nomogram for predicting new fractures after PKP treatment of for osteoporotic vertebral compression fractures in the elderly individuals [J]. BMC Musculoskelet Disord, 2023, 24(1): 728. Baroud G, Bohner M. Biomechanical impact of vertebroplasty. Postoperative biomechanics of vertebroplasty [J]. Joint Bone Spine, 2006, 73(2): 144–150. Oda K, Shibayama Y, Abe M, et al. Morphogenesis of vertebral deformities in involutional osteoporosis. Age-related, three-dimensional trabecular structure [J]. Spine (Phila Pa 1976), 1998, 23(9). Kopperdahl D L, Pearlman J L, Keaveny T M. Biomechanical consequences of an isolated overload on the human vertebral body [J]. J Orthop Res, 2000, 18(5): 685–690. Laredo J-D, Hamze B. Complications of percutaneous vertebroplasty and their prevention [J]. Semin Ultrasound CT MR, 2005, 26(2): 65–80. Abd El-Rahman A M, Lazzarotti A G, Cosottini M, et al. Pulmonary embolism caused by cement leakage during percutaneous vertebroplasty. A case report of successful conservative management [J]. Neuroradiol J, 2012, 25(4): 481–485. Lim S H, Kim H, Kim H K, et al. Multiple cardiac perforations and pulmonary embolism caused by cement leakage after percutaneous vertebroplasty [J]. Eur J Cardiothorac Surg, 2008, 33(3): 510–512. Yu B, Wang C-Y. Osteoporosis: The Result of an 'Aged' Bone Microenvironment [J]. Trends Mol Med, 2016, 22(8): 641–644. Killinger Z. [Early diagnosis of postmenopausal osteoporosis] [J]. Bratisl Lek Listy, 2000, 101(3): 179–180. Ning L, Zhu J, Tian S, et al. Correlation Analysis Between Basic Diseases and Subsequent Vertebral Fractures After Percutaneous Kyphoplasty (PKP) for Osteoporotic Vertebral Compression Fractures [J]. Pain Physician, 2021, 24(6): E803-E810. Parfitt A M. Trabecular bone architecture in the pathogenesis and prevention of fracture [J]. Am J Med, 1987, 82(1B): 68–72. Garrison J G, Gargac J A, Niebur G L. Shear strength and toughness of trabecular bone are more sensitive to density than damage [J]. J Biomech, 2011, 44(16): 2747–2754. Mellström D, Yang X, Li Z, et al. Proportion and Characteristics of Patients in Sweden Remaining at High Risk of Fracture Despite Prior Treatment [J]. Clin Ther, 2016, 38(7). Rizzoli R. Postmenopausal osteoporosis: Assessment and management [J]. Best Pract Res Clin Endocrinol Metab, 2018, 32(5): 739–757. Lems W F, Raterman H G. Critical issues and current challenges in osteoporosis and fracture prevention. An overview of unmet needs [J]. Ther Adv Musculoskelet Dis, 2017, 9(12): 299–316. Feng S-T, Yang Y, Li X, et al. Risk Factors of New Symptomatic Fractures After Vertebroplasty: A Retrospective Cohort Study of 268 Patients with Painful Osteoporotic Vertebral Compression Fracture [J]. World Neurosurg, 2024, 187: e890-e897. Klotzbuecher C M, Ross P D, Landsman P B, et al. Patients with prior fractures have an increased risk of future fractures: a summary of the literature and statistical synthesis [J]. J Bone Miner Res, 2000, 15(4): 721–739. Gehlbach S, Saag K G, Adachi J D, et al. Previous fractures at multiple sites increase the risk for subsequent fractures: the Global Longitudinal Study of Osteoporosis in Women [J]. J Bone Miner Res, 2012, 27(3): 645–653. Hoerth R M, Seidt B M, Shah M, et al. Mechanical and structural properties of bone in non-critical and critical healing in rat [J]. Acta Biomater, 2014, 10(9): 4009–4019. Kreider J M, Goldstein S A. Trabecular bone mechanical properties in patients with fragility fractures [J]. Clin Orthop Relat Res, 2009, 467(8): 1955–1963. Morgan E F, Unnikrisnan G U, Hussein A I. Bone Mechanical Properties in Healthy and Diseased States [J]. Annu Rev Biomed Eng, 2018, 20: 119–143. Huber-Lang M, Kovtun A, Ignatius A. The role of complement in trauma and fracture healing [J]. Semin Immunol, 2013, 25(1): 73–78. Loi F, Córdova L A, Pajarinen J, et al. Inflammation, fracture and bone repair [J]. Bone, 2016, 86: 119–130. Cuthbertson D P. The disturbance of metabolism produced by bony and non-bony injury, with notes on certain abnormal conditions of bone [J]. Biochem J, 1930, 24(4): 1244–1263. Qian Y, Hu X, Li C, et al. Development of a nomogram model for prediction of new adjacent vertebral compression fractures after vertebroplasty [J]. BMC Surg, 2023, 23(1): 197. Hey H W D, Tan J H, Tan C S, et al. Subsequent Vertebral Fractures Post Cement Augmentation of the Thoracolumbar Spine: Does it Correlate With Level-specific Bone Mineral Density Scores? [J]. Spine (Phila Pa 1976), 2015, 40(24): 1903–1909. Melton L J, Atkinson E J, O'Fallon W M, et al. Long-term fracture prediction by bone mineral assessed at different skeletal sites [J]. J Bone Miner Res, 1993, 8(10): 1227–1233. Cummings S R, Black D M, Nevitt M C, et al. Appendicular bone density and age predict hip fracture in women. The Study of Osteoporotic Fractures Research Group [J]. JAMA, 1990, 263(5): 665–668. Cauley J A, Hochberg M C, Lui L-Y, et al. Long-term risk of incident vertebral fractures [J]. JAMA, 2007, 298(23): 2761–2767. Seeman E. Invited Review: Pathogenesis of osteoporosis [J]. J Appl Physiol (1985), 2003, 95(5): 2142–2151. Iolascon G, Napolano R, Gioia M, et al. The contribution of cortical and trabecular tissues to bone strength: insights from denosumab studies [J]. Clin Cases Miner Bone Metab, 2013, 10(1): 47–51. Natesan V, Kim S-J. Metabolic Bone Diseases and New Drug Developments [J]. Biomol Ther (Seoul), 2022, 30(4): 309–319. Kim J-M, Shin D A, Byun D-H, et al. Effect of bone cement volume and stiffness on occurrences of adjacent vertebral fractures after vertebroplasty [J]. J Korean Neurosurg Soc, 2012, 52(5): 435–440. Chen X-S, Jiang J-M, Sun P-D, et al. How the clinical dosage of bone cement biomechanically affects adjacent vertebrae [J]. J Orthop Surg Res, 2020, 15(1): 370. Simons D G, Travell J G. Myofascial origins of low back pain. 1. Principles of diagnosis and treatment [J]. Postgrad Med, 1983, 73(2). Zhao H, He Y, Yang J-S, et al. Can paraspinal muscle degeneration be a reason for refractures after percutaneous kyphoplasty? A magnetic resonance imaging observation [J]. J Orthop Surg Res, 2021, 16(1): 476. Chen M, Yang C, Cai Z, et al. Lumbar posterior group muscle degeneration: Influencing factors of adjacent vertebral body re-fracture after percutaneous vertebroplasty [J]. Front Med (Lausanne), 2022, 9: 1078403. Gracovetsky S. Is the lumbodorsal fascia necessary? [J]. J Bodyw Mov Ther, 2008, 12(3): 194–197. Aspden R M. Review of the functional anatomy of the spinal ligaments and the lumbar erector spinae muscles [J]. Clinical Anatomy: The Official Journal of the American Association of Clinical Anatomists and the British Association of Clinical Anatomists, 1992, 5(5): 372–387. Wong C, Rasmussen J, Simonsen E, et al. The influence of muscle forces on the stress distribution in the lumbar spine [J]. Open Spine J, 2011, 3: 21–26. Hilaire C S, Johnson A, Loseth C, et al. Facial fractures and associated injuries in high- versus low-energy trauma: all are not created equal [J]. Maxillofac Plast Reconstr Surg, 2020, 42(1): 22. Robinson C M, Royds M, Abraham A, et al. Refractures in patients at least forty-five years old. a prospective analysis of twenty-two thousand and sixty patients [J]. J Bone Joint Surg Am, 2002, 84(9): 1528–1533. Center J R, Bliuc D, Nguyen T V, et al. Risk of subsequent fracture after low-trauma fracture in men and women [J]. JAMA, 2007, 297(4): 387–394. Gorter E A, Reinders C R, Krijnen P, et al. The effect of osteoporosis and its treatment on fracture healing a systematic review of animal and clinical studies [J]. Bone Rep, 2021, 15: 101117. Li J, Xu L, Liu Y, et al. Open Surgical Treatments of Osteoporotic Vertebral Compression Fractures [J]. Orthop Surg, 2023, 15(11): 2743–2748. Li W, Wang H, Dong S, et al. Establishment and validation of a nomogram and web calculator for the risk of new vertebral compression fractures and cement leakage after percutaneous vertebroplasty in patients with osteoporotic vertebral compression fractures [J]. Eur Spine J, 2022, 31(5): 1108–1121. Jeon C-H, Lee Y-S, Youn S-J, et al. Factors affecting postural reduction in posterior surgery for thoracolumbar burst fracture [J]. J Spinal Disord Tech, 2015, 28(4): E225-E230. Wang M, Jin Q. High-viscosity bone cement for vertebral compression fractures: a prospective study on intravertebral diffusion and leakage of bone cement [J]. BMC Musculoskelet Disord, 2020, 21(1): 589. Lai P-L, Chen L-H, Chen W-J, et al. Chemical and physical properties of bone cement for vertebroplasty [J]. Biomed J, 2013, 36(4): 162–167. Kring D L. Reliability and validity of the Braden Scale for predicting pressure ulcer risk [J]. J Wound Ostomy Continence Nurs, 2007, 34(4): 399–406. Meesters D M, Wijnands K A P, Brink P R G, et al. Malnutrition and Fracture Healing: Are Specific Deficiencies in Amino Acids Important in Nonunion Development? [J]. Nutrients, 2018, 10(11). Giganti M G, Tresoldi I, Masuelli L, et al. Fracture healing: from basic science to role of nutrition [J]. Front Biosci (Landmark Ed), 2014, 19(7): 1162–1175. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8863516","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":610153139,"identity":"58c4868e-836e-4625-bb11-0810f2ee4d92","order_by":0,"name":"zongjie guo","email":"","orcid":"","institution":"Yancheng First Hospital, Affiliated Hospital of Nanjing University Medical School","correspondingAuthor":false,"prefix":"","firstName":"zongjie","middleName":"","lastName":"guo","suffix":""},{"id":610153143,"identity":"96b5534d-1c04-42e0-a6f5-b44094970672","order_by":1,"name":"junping bao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAv0lEQVRIiWNgGAWjYHACNjDmZ2Y+/IA0LZLtbGkGpGhhYDA4z6MgQZR6+fYeswc/KvjyjA/zMBgw1NhEE9RicOaMuWHPGbZis8O8Bx4wHEvLbSCoRSLHTIK3jS1x22G+BAPGhsOEtcjPf2Mm+ReoZXMzj4EEUVoYbvCYSYNs2cBMrBaDM2nlxjJAv0gcBgZyAjF+kW8/vO3hm4pjefz9hw8/+FBjQ4TDIOBYAphKIFI5CNSQongUjIJRMApGGgAA7Ps7yKNGOtgAAAAASUVORK5CYII=","orcid":"","institution":"Southeast University Zhongda Hospital","correspondingAuthor":true,"prefix":"","firstName":"junping","middleName":"","lastName":"bao","suffix":""},{"id":610153147,"identity":"b91eecc0-5b4a-4057-8d27-437bac3bb601","order_by":2,"name":"lei zhang","email":"","orcid":"","institution":"NHC Key Laboratory of Contraceptives Vigilance and Fertility Surveillance, Jiangsu Provincial Medical Key Laboratory of Fertility Protection and Health Technology Assessment.","correspondingAuthor":false,"prefix":"","firstName":"lei","middleName":"","lastName":"zhang","suffix":""},{"id":610153152,"identity":"80c4b256-af15-4140-ac62-0e868c5e2892","order_by":3,"name":"rui shi","email":"","orcid":"","institution":"Southeast University Zhongda Hospital","correspondingAuthor":false,"prefix":"","firstName":"rui","middleName":"","lastName":"shi","suffix":""},{"id":610153157,"identity":"1ec63e29-9af8-44d5-8d2d-484de1cbfb49","order_by":4,"name":"shu yang","email":"","orcid":"","institution":"Southeast University Zhongda Hospital","correspondingAuthor":false,"prefix":"","firstName":"shu","middleName":"","lastName":"yang","suffix":""},{"id":610153160,"identity":"aac9384d-d4f1-41b3-a0c3-9c78ab193868","order_by":5,"name":"lei zhang","email":"","orcid":"","institution":"NHC Key Laboratory of Contraceptives Vigilance and Fertility Surveillance, Jiangsu Provincial Medical Key Laboratory of Fertility Protection and Health Technology Assessment.","correspondingAuthor":false,"prefix":"","firstName":"lei","middleName":"","lastName":"zhang","suffix":""},{"id":610153162,"identity":"5a55cec2-f705-4591-8bce-8ce3ce673ad0","order_by":6,"name":"junping bao","email":"","orcid":"","institution":"Southeast University Zhongda Hospital","correspondingAuthor":false,"prefix":"","firstName":"junping","middleName":"","lastName":"bao","suffix":""}],"badges":[],"createdAt":"2026-02-12 15:09:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8863516/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8863516/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105296952,"identity":"27bbb65c-4585-4445-9d96-6eb95d5a95b4","added_by":"auto","created_at":"2026-03-24 13:14:59","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":185077,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart for the development of interpretable machine learning models.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8863516/v1/9c072a8993ce9c6accc9252c.png"},{"id":105296953,"identity":"441435d1-5bdb-45e0-b5c5-acda9e2cd3d7","added_by":"auto","created_at":"2026-03-24 13:14:59","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":178942,"visible":true,"origin":"","legend":"\u003cp\u003eArea under the receiver operating characteristic curve (AUC-ROC) of the six machine learning models. (A) Receiver Operating Characteristic curve of the training set; (B) Receiver Operating Characteristic curve of the testing set. ROC curve, Receiver Operating Characteristic curve; XGBoost, eXtreme Gradient Boosting; DT, Decision Tree; RF, Random Forest; LR, Logistic Regression; SVC, Support Vector Machine; ANN, Artificial Neural Networks.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8863516/v1/51c4ebd9e34bc8644610b3c0.png"},{"id":105564978,"identity":"5a905fcd-9759-4177-9835-66944dbac660","added_by":"auto","created_at":"2026-03-27 12:51:30","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":439051,"visible":true,"origin":"","legend":"\u003cp\u003eConfusion matrix diagram of the classification results of the six machine learning models. XGBoost, eXtreme Gradient Boosting; DT, Decision Tree; RF, Random Forest; LR, Logistic Regression; SVC, Support Vector Machine; ANN, Artificial Neural Networks.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8863516/v1/2f2fafefe0c6591b38bfbc23.png"},{"id":105296958,"identity":"5541a45f-5425-4d1c-ad6c-d43bad2fd0f4","added_by":"auto","created_at":"2026-03-24 13:15:00","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":759011,"visible":true,"origin":"","legend":"\u003cp\u003eCalibration plots of the six machine learning models. XGBoost, eXtreme Gradient Boosting; DT, Decision Tree; RF, Random Forest; LR, Logistic Regression; SVC, Support Vector Machine; ANN, Artificial Neural Networks.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8863516/v1/22b048fee9f8e6c3797e11ef.png"},{"id":105296955,"identity":"9102b669-b0e5-48af-96f9-af3e3f967e4c","added_by":"auto","created_at":"2026-03-24 13:14:59","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":300314,"visible":true,"origin":"","legend":"\u003cp\u003eDecision curve analysis of the six machine learning models. (A) Decision curve analysis of the training set; (B) Decision curve analysis of the testing set. XGBoost, eXtreme Gradient Boosting; DT, Decision Tree; RF, Random Forest; LR, Logistic Regression; SVC, Support Vector Machine; ANN, Artificial Neural Networks.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8863516/v1/922ab200433c78325dbff1fb.png"},{"id":105564886,"identity":"a83c61f7-14de-4c21-8811-b1f80f3c46f2","added_by":"auto","created_at":"2026-03-27 12:51:13","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":711130,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP global explanatory analysis of the XGBoost model. (A) Importance ranking of the top 20 risk factors with respect to stability and interpretation using the optimal model. (B) The importance of features of the XGBoost model based on the mean |SHAP value|. SHAP, SHapley Additive exPlanation; BMD, Bone Mineral Density.\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-8863516/v1/acc18a4eae6f71a5136b4926.png"},{"id":105296956,"identity":"add4967f-882c-49ca-8f03-9f55d6f9b2c1","added_by":"auto","created_at":"2026-03-24 13:14:59","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":543609,"visible":true,"origin":"","legend":"\u003cp\u003eInterpretation of model prediction results using two sample cases. (A) Case 1 has a high risk of re-fracture after surgery. (B) Case 2 has a low risk of re-fracture after surgery. BMD, Bone Mineral Density.\u003c/p\u003e","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-8863516/v1/540f96ca8e6e1c127841a843.png"},{"id":107836331,"identity":"0e6fe415-58db-4bca-a52f-33c4057cce3b","added_by":"auto","created_at":"2026-04-26 16:09:58","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3699469,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8863516/v1/ee443bd1-493b-44c1-a157-5f724ba86f2a.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Application Study of Interpretable Machine Learning Models for Predicting Postoperative Refracture After Vertebral Augmentation in Osteoporotic Vertebral Compression Fractures","fulltext":[{"header":"Introduction","content":"\u003cp\u003eOsteoporosis (OP), the most common clinical disease of the skeletal system, is pathologically characterized by decreased bone mass and destruction of the bone microarchitecture, which in turn leads to increased bone fragility and a higher fracture risk [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. With the aging of the global population, osteoporotic fracture poses an increasingly serious threat to the health of middle-aged and elderly people [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Osteoporotic fractures are the most serious complication of osteoporosis [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], occurring in the vertebrae, hips, and pelvis, with vertebral fractures being the most frequent. Osteoporotic vertebral compression fractures (OVCFs) have an insidious onset, and their incidence has been on the rise in recent years [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. The principles of clinical intervention for OVCFs have been clearly defined, including fracture reduction and fixation, functional rehabilitation, and anti-osteoporotic treatment [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Percutaneous vertebroplasty (PVP) and percutaneous kyphoplasty (PKP) are preferred for strengthening the vertebral column when non-surgical treatments are not effective. Despite the widespread use of vertebral body strengthening surgical treatment, postoperative vertebral re-fracture is still common and can lead to repeated pain and treatment needs [\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003cp\u003ePrevious studies have used logistic regression and multifactorial analysis to identify risk factors for recurrent fractures and develop predictive models; however, such models are difficult to accurately assess the expected risk of an individual [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Machine learning (ML), as a multidisciplinary technology, has revolutionized the epidemiological research paradigm by modeling and analyzing the complex associations between predictor and response variables [\u003cspan additionalcitationids=\"CR13 CR14 CR15 CR16 CR17 CR18 CR19 CR20 CR21 CR22 CR23 CR24 CR25\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. ML algorithms powered by Shapley Additive exPlanation (SHAP) can quantify feature contributions and visualize global and local interpretations of models with better accuracy than ordinary linear models [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eHowever, there are limited studies based on interpretable machine learning algorithms for modeling and predicting the risk of recurrent vertebral fractures after vertebral body strengthening in patients with OVCFs. Suppose such a prediction model can be successfully constructed. In that case, it can not only achieve the early and accurate identification of high-risk groups but also formulate early and personalized intervention strategies based on individual risk factors, which can reduce the risk of recurrent fractures, the consumption of healthcare resources, and conflicts between doctors and patients, as well as reduce the burden on both individuals and the public health system.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e1.1 Study population\u003c/h2\u003e \u003cp\u003eThis is a retrospective study based on clinical data, and the flowchart of the study is shown in Fig.\u0026nbsp;1. The medical records and imaging data of 1502 patients who were admitted for vertebral body enhancement surgery due to OVCFs at the Center for Spine Surgery, CU Hospital, Southeast University, from January 2014 to December 2022 were retrospectively collected. The inclusion criteria were as follows: (1) patients with a precise diagnosis of OVCFs requiring surgical intervention and (2) undergoing vertebral body strengthening surgery for the first time. The exclusion criteria were as follows: (1) non-osteoporotic vertebral compression fractures; (2) pathologic compression fractures of the vertebral body secondary to other factors, such as tumors and infections; (3) patients who did not undergo vertebral body enhancement surgery or in combination with other surgical modalities; (4) patients who had a combination of systemic or localized infections; (5) patients who had been in combination with other surgical modalities; and (6) patients who died within the observation time window (within 2 years) after the first surgery or patients for whom valid follow-up information could not be obtained.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e1.2 Calculation of sample size\u003c/h2\u003e \u003cp\u003eIn recent years, Riley et al. [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e] proposed a widely recognized scientific method to determine the sample size in prediction models. In this study, the pmsampsize package for clinical predictive modeling in the R language, version 4.4.0, was used, and the sample size was calculated based on the guidelines provided above. The incidence of recurrent fractures after vertebral body strengthening was approximately 9.6%. A total of 36 predictor variables were included in this study, which was calculated using software to determine a minimum sample size of 1,189 patients required for this study. After combining the inclusion and exclusion criteria, 1502 patients were finally included in the study.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e1.3 Inclusion characteristics\u003c/h2\u003e \u003cp\u003eBy searching the hospital's medical record system and imaging system, and integrating general patient information, laboratory test results, surgery-related data, and pre- and post-operative imaging data, a total of 36 characteristic indicators were ultimately included. To ensure consistency in typing criteria, three spine surgery experts discussed and established uniform criteria for the imaging data. The experts were also double-blind to the characteristics and statistical analysis of the study subjects. All surgeries were performed by our senior spine surgical team, which has more than 10 years of surgical experience. The surgeons strictly followed standardized protocols and completed relevant training before the study to ensure consistency and minimize bias.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e1.4 Statistical analysis\u003c/h2\u003e \u003cp\u003eIBM SPSS 29.0 statistical software was used for the statistical analysis of the data in this study. Measurement information was expressed as raw data, and for variables that obeyed normal distribution, a one-sample t-test was used. For variables that do not follow a normal distribution, the rank sum test was used. Comparisons of categorical variables were performed using the chi-square test or Fisher's exact probability test. A two-sided P value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e1.5 Data Preprocessing\u003c/h2\u003e \u003cp\u003eData preprocessing is the foundation for ensuring data quality and model accuracy. In this study, the IterativeImputer module of the Python Scikit-Learn library is used to perform multiple interpolations to mitigate the impact of missing values on the robustness of the results. The feature scales are then unified through standardization of deviation and normalization to ensure data distribution consistency. The dataset is randomly divided into a training set and a test set with a ratio of 7:3. If there is an imbalance in sample distribution, oversampling or undersampling techniques are used for optimization. Finally, exploratory data analysis is conducted through data visualization to reveal the distribution laws of the features.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e1.6 Model Development\u003c/h2\u003e \u003cp\u003eThis study develops machine learning models based on Anaconda3 2023.07 (64-bit) and Python 3.10.0, including XGBoost, Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), and Artificial Neural Networks (ANN). Among them, XGBoost is implemented using a proprietary library that relies on parallel computing and regularization for efficient training. Grid search, combined with 5-fold cross-validation, is used to select the hyperparameters of the training set. The DT, RF, SVM, and LR models are then developed using Scikit-Learn. All models are evaluated on a test set to assess their efficacy in predicting real-world scenarios.\u003c/p\u003e \u003cp\u003eReferring to previous studies [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], we used accuracy, precision, recall, F1 score, and Area Under the Curve (AUC) as the evaluation metrics in this study; analyzed the difference between predicted and actual results by calibration curves [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]; and evaluated the net clinical benefit by Decision Curve Analysis (DCA) [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e] to comprehensively screen the best machine learning prediction models. The best machine learning prediction model was selected.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e1.7 Model Interpretability Assessment\u003c/h2\u003e \u003cp\u003eThe SHAP technique devised by Lundberg and Lee [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e] significantly improves the interpretability of complex algorithms by mapping machine learning feature SHAP values to clinical variables. The method is based on the Shapley value concept of cooperative game theory, which assigns specific contribution values to each feature to help explain the \u0026ldquo;black box\u0026rdquo; mechanism. In this study, the SHAP values of the features are calculated using the SHAP toolkit in Python 3.10.0, and the relative importance of the features is assessed by comparing the difference in prediction results between the presence and absence of the features [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e1.8 Ethical Review\u003c/h2\u003e \u003cp\u003e This study was ethically reviewed by the Ethics Committee of Zhongda Hospital, affiliated with Southeast University, Ethics Approval No: 2024ZDSYLL187-P01.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Characteristics of the study population\u003c/h2\u003e \u003cp\u003eThe general information and laboratory examination details of the study population are presented in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e1\u003c/span\u003e. A comparison of surgical and imaging data between the recurrent fracture group after vertebral body strengthening (RF group) and the non-recurrent fracture group (NRF group) is shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e2\u003c/span\u003e. A total of 1,502 patients were included in the present study, of whom 145 (9.65%) had a recurrent fracture after surgery (RF group), and the remaining patients were in the NRF group.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDemographic characteristics and laboratory examination of study participants\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNRF\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eχ2/t/Z\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;1357)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;145)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eSex (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.009\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.926\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e267(19.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e29(20.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1090(80.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e116(80.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAge (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e282.634\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026lt;70 years\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e958(70.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0(0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026ge;\u0026thinsp;70 years\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e399(29.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e145(100.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBMI (kg/m\u0026sup2;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23.01(21.64,25.83)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e22.73(21.21,25.33)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.702\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.661\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eTrauma (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e71.399\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e362(26.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e81(55.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eConcealed trauma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e386(28.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e46(31.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObvious trauma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e609(44.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18(12.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eHypertension (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.023\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.878\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e739(54.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e78(53.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e618(45.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e67(46.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eDiabetes (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6.305\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.092\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1161(85.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e128(88.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e196(14.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e15(11.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eHeart disease (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.852\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.174\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1203(88.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e123(84.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e154(11.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e22(15.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eCranial disease (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6.206\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.073\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1132(83.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e109(75.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e225(16.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e36(24.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eOsteoporosis (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e141.588\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e704(51.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0(0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e653(48.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e145(100.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003ePrevious history of fractures (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e498.414\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1239(91.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30(20.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e118(8.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e115(79.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eSegment of freshly fractured vertebrae (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e30.418\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAbove T12 level\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e157(11.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21(14.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e289(21.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e24(16.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e388(28.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20(13.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e189(13.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e22(15.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e63(4.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4(2.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e48(3.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9(6.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5(0.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0(0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMultiple segments\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e218(16.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e45(31.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber of freshly fractured vertebrae (n)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1(1,1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1(1,1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-2.837\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.065\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBraden score (n)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e20(18,21)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18(17,20)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-1.219\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.023\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFall risk assessment (n)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3(2,4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3(2,4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-2.205\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.057\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVAS (n)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3(2,3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4(3,5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-2.852\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.004\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHb (10^12/L)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e128(120,136)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e122(115.5,129)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-5.224\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCa (mmol/L)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.22(2.16,2.31)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.23(2.14,2.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.121\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.904\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eALB (g/L)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e39.7(37.25,43)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e40.3(35.85,44.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.033\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.974\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBUN (mmol/L)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5.6(4.7,7.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6.2(4.55,7.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-1.011\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.312\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCr (\u0026micro;mol/L)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e58(48,71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e60(43.5,73.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.503\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.615\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSurgery-related and radiographic factors in study participants\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNRF\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eχ2/t/Z\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;1357)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;145)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDuration of surgery (min)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e40(31,50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e45(35,60)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-2.172\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.030\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eSurgical methods (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.961\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.195\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePVP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e669(49.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e73(50.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePKP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e688(50.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e72(49.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eMethod of puncture (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.477\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.224\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnilateral puncture\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e384(28.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e48(33.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBilateral puncture\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e973(71.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e97(66.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBone cement dosage (ml)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6(6,9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6(5.75,8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-1.315\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.189\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eBone cement leakage (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e280.326\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1045(77.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e15(10.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e312(23.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e130(89.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eTypes of bone cement leakage (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e341.644\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1045(77.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e15(10.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e184(13.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e45(31.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eII\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e63(4.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e31(21.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIII\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e13(1.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e10(6.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23(1.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19(13.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e29(2.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e25(17.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eBone cement distribution type (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6.922\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.265\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e978(72.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e101(69.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eII\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e94(6.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e10(6.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIII\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e88(6.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e10(6.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e128(9.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14(9.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e69(5.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e10(6.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eContact with the endplate (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e40.842\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e922(67.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e60(41.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e435(32.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e85(58.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eThoracolumbar fascitis (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e176.892\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1068(78.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e40(27.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e289(21.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e105(72.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eScoliosis (n%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e17.696\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.158\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1153(85.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e113(77.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e204(15.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e32(22.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAVHRR (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.1(0.02,0.22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.09(0.04,0.28)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-1.157\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.247\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCobb (\u0026deg;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1(0,5.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2(-2,4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-1.154\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.249\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePI (\u0026deg;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e50(41,57)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e50(43.5,61.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-1.459\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.144\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePT (\u0026deg;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18(10.5,26)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18(11,27)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-1.133\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.257\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSS (\u0026deg;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e31(25,38.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e33(27,41.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-1.617\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.106\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBMD (g/cm^2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-3(-3.8,-2.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-4(-4.65,-2.85)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-8.527\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eIn terms of baseline characteristics, the proportion of patients aged 70 years or older was significantly higher in the RF group (100.0%) than in the NRF group (29.4%, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and the proportion of patients with no obvious history of traumatic injury was higher (55.9% vs. 26.7%, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The proportion of patients with a history of previous fracture was significantly higher in the RF group (79.3%) than in the NRF group (8.7%, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and the Braden score was lower (P\u0026thinsp;=\u0026thinsp;0.001). The Braden score was lower (P\u0026thinsp;=\u0026thinsp;0.023), the VAS score was higher (P\u0026thinsp;=\u0026thinsp;0.004), and the hemoglobin concentration was lower (P\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e \u003cp\u003eRegarding fracture-related characteristics, the proportion of osteoporosis history in the RF group was 100.0%, significantly higher than that in the NRF group (48.1%, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001); the distribution of fresh fracture segments differed significantly (P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), with 31.0% of the RF group suffering from multisegmental fractures, whereas the NRF group had a predominance of T12 (21.3%) and L1 (28.6%) segments.\u003c/p\u003e \u003cp\u003eRegarding surgical and imaging characteristics, the surgical time was longer in the RF group (P\u0026thinsp;=\u0026thinsp;0.030); the incidence of cement leakage was significantly higher (89.7% vs. 23.0%, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and there was a significant difference in the type of leakage (P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), with a predominantly non-leakage in the NRF group (77.0%). In addition, the RF group had a higher proportion of cement contacting the endplate (58.6% vs 32.1%, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), a higher proportion of lumbar dorsal musculoskeletal fasciitis (72.4% vs 21.3%, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and a lower BMD value (P\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Performance of the model\u003c/h2\u003e \u003cp\u003eA total of 1,502 patients were included in this study, comprising 145 in the refracture (RF) group and 1,357 in the non-refracture (NRF) group. To address the imbalance in the proportion of original data categories and the limited sample size, the study expanded the dataset through various data enhancement techniques. At the same time, it scientifically divided the training set and the test set according to a 7:3 ratio to ensure the accuracy of model evaluation. The final test set comprises 815 cases, with sample sizes of the two groups essentially balanced.\u003c/p\u003e \u003cp\u003eThe prediction performance of the six models on the test set is evaluated. The results are shown in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Fig.\u0026nbsp;2. All models show good classification ability in the training stage, among which the XGBoost model is outstanding in recurrent fracture prediction, with various indexes significantly better than the other models, including precision\u0026thinsp;=\u0026thinsp;0.9926, recall\u0026thinsp;=\u0026thinsp;0.9951, accuracy\u0026thinsp;=\u0026thinsp;0.9939, F1 score\u0026thinsp;=\u0026thinsp;0.9924, and AUC\u0026thinsp;=\u0026thinsp;0.9996. The confusion matrix results in Fig.\u0026nbsp;3 further confirm that the XGBoost model has the best classification effect; the calibration curve in Fig.\u0026nbsp;4 shows that its curve is closer to the 45-degree perfect calibration line; the decision curve analysis (DCA) results in Fig.\u0026nbsp;5 show that the net benefit of XGBoost is stable in most intervals in the training set, and the validation set shows good prediction accuracy and potential for clinical application, with significant advantages in practical decision-making. In the validation set, the XGBoost model shows good prediction accuracy and potential for clinical application, with substantial advantages in practical decision-making. Overall, the XGBoost model performed the best among the six models and was the most accurate prediction model in this study.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePerformance of six machine learning models for the testing set (n\u0026thinsp;=\u0026thinsp;815)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRecall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePrecesion\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eF1-score\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eeXtreme Gradient Boosting\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9926\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.9951\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9939\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.9924\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.9996\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDecision Tree\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9901\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.9829\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9865\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.9828\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.9478\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRandom Forest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9878\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.9878\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9877\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.9878\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.9988\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLogistic Regression\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9854\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.9902\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9877\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.9878\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.9991\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSupport Vector Machine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9878\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.9951\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9914\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.9924\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.9982\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eArtificial Neural Networks\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9806\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.9951\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9877\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.9878\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.9985\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Interpretability of the model\u003c/h2\u003e \u003cp\u003eThe XGBoost model combined with the SHAP method can provide both global and local interpretable results. A total of 36 features were included in the model. The prediction results are presented in Fig.\u0026nbsp;6A, where rows represent specific features, dots represent samples, and colors distinguish feature values (red for high values and blue for low values). Figure\u0026nbsp;6B lists the 9 key features in order of importance, including: age, cement leakage and its imaging typology, history of previous osteoporosis, history of prior fracture, bone mineral density level, lumbar dorsal musculoskeletal fasciitis, trauma type, length of surgery, and Braden score.\u003c/p\u003e \u003cp\u003eWith the help of SHAP force maps, the ML combined with the SHAP approach can further elucidate the mechanism of the influence of each feature on the individual prediction results, and Fig.\u0026nbsp;7 demonstrates the SHAP force maps of two cases, which quantify the contribution of each feature to the prediction of the XGBoost model through the SHAP value. Case 1 (Fig.\u0026nbsp;7A) was predicted to have a high risk of recurrent fracture after vertebral body strengthening after analyzing the effects of all factors, which was consistent with the actual postoperative recurrent fracture results of this patient; the prediction results of Case 2 (Fig.\u0026nbsp;7B) were also consistent with the actual situation.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003ePostoperative re-fracture of OVCFs is a challenging aspect of clinical treatment, significantly impacting patients' recovery and quality of life [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Traditional risk prediction studies have limited accuracy and are challenging to meet the demand for precision medicine. Although ML has demonstrated its advantages in the diagnosis and treatment of lumbar degenerative diseases in spine surgery, it is still in the exploratory stage in the prediction of postoperative recurrent fracture risk in OVCFs. In this study, ML was introduced into this field for the first time. Multiple algorithmic models were constructed by collecting multidimensional diagnostic and treatment data, which were then combined with SHAP analysis to improve the interpretability of the models. This approach provided new insights for the early identification of high-risk populations and the formulation of personalized intervention plans, ultimately contributing to the development of effective diagnosis and treatment strategies for OVCFs. XGBoost, a machine learning model integrated with a decision tree, can solve various types of regression, classification, and ranking problems [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e], and is highly efficient and effective for predicting postoperative fracture risk. XGBoost is an integrated decision tree machine learning model that can solve various regression, classification, and ranking problems [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e], offering significant advantages in terms of efficiency, accuracy, scalability, flexibility, and stability [\u003cspan additionalcitationids=\"CR40 CR41 CR42\" citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e]. The SHAP method, as a powerful model interpretation tool, can significantly improve the interpretability of ML algorithms [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. In this study, we combined the two, realized the integration of predictive validity and interpretability, and clarified the key predictive indices affecting re-fracture, which provides support for precise clinical prevention and treatment.\u003c/p\u003e \u003cp\u003eThis study confirms that advanced age is a closely associated factor for postoperative re-fracture in OVCFs, which is consistent with the findings of several clinical studies [\u003cspan additionalcitationids=\"CR45\" citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e], and the core mechanism lies in the imbalance of bone metabolism in advanced-aged patients [\u003cspan additionalcitationids=\"CR48 CR49 CR50 CR51\" citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e] and the impact of age-related muscle dysfunction on spinal stability [\u003cspan additionalcitationids=\"CR54\" citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e]. Although advanced age is a recognized risk factor, age stratification criteria have not been standardized. In this study, we propose 70 years as the threshold value, which is both scientifically and clinically valuable: People over 70 years of age enter the stage of severe osteoporosis, and the incidence of fragility fracture is 2\u0026ndash;3 times higher than that of people over 60 years of age [\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e]; people between the ages of 50 and 70 years of age may have early-onset osteoporosis, which is not very relevant to their age [\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e]; people over the age of 80 years of age have a high heterogeneity of samples due to the co-occurrence of multiple diseases, and the generalizability of the study is limited. Therefore, osteoporosis in people over 70 years of age more closely resembles age-associated natural degeneration, highlighting the value of age as an independent risk factor.\u003c/p\u003e \u003cp\u003eCement leakage is a common complication after vertebral body augmentation [\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e], and its status and type are key risk factors for re-fracture [\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e], especially when the leakage occurs in the intervertebral disc, where the risk of re-fracture of the adjacent vertebrae is significantly higher [\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e]. From a biomechanical perspective, the mechanism varies among different leakage sites. Leakage from the intervertebral disc disrupts the elastic cushioning function, leading to abnormal stress distribution [\u003cspan additionalcitationids=\"CR63\" citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e]. Leakage from the spinal canal or neural foramina compresses the nerve roots, forcing a change in the spinal force pattern and disrupting the mechanical equilibrium [\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e]. Leakage from the paravertebral veins does not directly affect vertebral mechanics. Still, it may lead to complications, such as pulmonary embolism, and indirectly increases the risk by decreasing activity and accelerating bone mass loss [\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e]. Currently, there is a lack of uniform criteria for classifying bone cement leakage, and previous studies have primarily analyzed it as a binary variable, overlooking its biomechanical heterogeneity. Therefore, large-sample, multicenter studies are urgently needed to establish a standardized system for clarifying the impact of each subtype.\u003c/p\u003e \u003cp\u003eIn this study, all patients in the re-fracture group were diagnosed with osteoporosis for the first time. In conjunction with previous studies, it has been shown that a history of previous osteoporosis is an independent risk factor for postoperative re-fracture in patients with OVCFs, i.e., the risk of re-fracture in those who were diagnosed before the fracture was significantly higher than that of those who were diagnosed for the first time at the time of the current fracture [\u003cspan additionalcitationids=\"CR69\" citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e]. This seemingly paradoxical phenomenon may be related to multiple factors: on the one hand, early diagnosed patients have a longer disease duration, and long-term bone loss and trabecular destruction lead to a significant decrease in bone strength and toughness [\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e, \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e]; on the other hand, although early diagnosed patients receive anti-osteoporosis treatment, 75.8% of them still have the problem of suboptimal bone density or persistent symptoms after two years of treatment [\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e], and they may have had preoperative occult minor fractures, whereas initiation of standardized interventions for acute fractures in those diagnosed for the first time reduces the risk [\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e, \u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eNumerous studies have consistently shown that a history of previous fractures is a key risk factor for postoperative re-fracture in patients with OVCFs. A history of previous fracture significantly elevates the risk of future fracture, with the predictive value of a history of previous vertebral fracture being particularly significant [\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e, \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e]; specifically, the risk of re-fracture in those with a history of fracture is approximately two times higher than that in those without a history of fracture, whereas in those with a previous vertebral fracture, their risk of re-fracture increases to four times [\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e, \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e]. Mechanistically, previous fractures can severely damage the normal structure of bones and decrease the mechanical properties of newly formed bone tissues [\u003cspan additionalcitationids=\"CR80\" citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e81\u003c/span\u003e]; at the same time, fracture trauma can disrupt the balance of bone metabolism and further exacerbate the condition [\u003cspan additionalcitationids=\"CR83\" citationid=\"CR82\" class=\"CitationRef\"\u003e82\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR84\" class=\"CitationRef\"\u003e84\u003c/span\u003e]. Therefore, the clinic should inquire in detail about previous fracture history, develop a more aggressive anti-osteoporosis program for individuals with a history of fracture, and enhance postoperative rehabilitation and follow-up.\u003c/p\u003e \u003cp\u003eThis study confirmed that BMD is a significant risk factor for postoperative re-fracture in OVCFs, consistent with the findings of several previous studies [\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e, \u003cspan citationid=\"CR85\" class=\"CitationRef\"\u003e85\u003c/span\u003e]. BMD T-value \u0026lt; -2.2 SD effectively predicted subsequent fracture, and the incidence of re-fracture was significantly higher at T-value \u0026le; -2.5 SD [\u003cspan citationid=\"CR86\" class=\"CitationRef\"\u003e86\u003c/span\u003e]. In addition, for every 1 SD decrease in BMD, the risk of fracture increased 1.4\u0026ndash;1.8 times [\u003cspan additionalcitationids=\"CR88\" citationid=\"CR87\" class=\"CitationRef\"\u003e87\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR89\" class=\"CitationRef\"\u003e89\u003c/span\u003e]. Bone strength and toughness are significantly reduced in patients with low BMD, which makes it difficult to withstand daily stresses. If the bone metabolic imbalance is not corrected early in the postoperative period, the stress resistance and microinjury repair capacity are reduced, which further increases the risk of re-fracture [\u003cspan additionalcitationids=\"CR91 CR92 CR93\" citationid=\"CR90\" class=\"CitationRef\"\u003e90\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR94\" class=\"CitationRef\"\u003e94\u003c/span\u003e]. Therefore, patients with OVCFs should pay attention to BMD monitoring and management after surgery and enhance BMD levels through active anti-osteoporotic treatment to reduce the risk of re-fracture.\u003c/p\u003e \u003cp\u003eLumbar dorsal fasciitis is a high-risk factor for postoperative re-fracture in patients with OVCFs. Paravertebral muscle mass is significantly reduced in patients with refracture after percutaneous vertebral kyphoplasty [\u003cspan citationid=\"CR96\" class=\"CitationRef\"\u003e96\u003c/span\u003e, \u003cspan citationid=\"CR97\" class=\"CitationRef\"\u003e97\u003c/span\u003e], emphasizing its importance in the treatment of OVCFs. As a core structure to maintain spinal stability, the lumbar dorsal muscles maintain spinal mechanical balance through coordinated contraction [\u003cspan citationid=\"CR98\" class=\"CitationRef\"\u003e98\u003c/span\u003e, \u003cspan citationid=\"CR99\" class=\"CitationRef\"\u003e99\u003c/span\u003e]; whereas, lumbar dorsal fasciitis can trigger muscle spasms, leading to spinal mechanical conduction malfunction and stress concentration, which is susceptible to microfractures in the long term [\u003cspan citationid=\"CR100\" class=\"CitationRef\"\u003e100\u003c/span\u003e]. Subsequent studies can explore the molecular mechanisms in depth and search for early diagnostic indicators and intervention targets to improve lumbar dorsal muscle function, attenuate the inflammatory response, and reduce the risk of postoperative re-fracture.\u003c/p\u003e \u003cp\u003eThis study found that the type of trauma was closely related to re-fracture. Patients who did not experience significant trauma or low-energy trauma had a significantly higher risk of postoperative re-fracture than those with high-energy trauma [\u003cspan citationid=\"CR101\" class=\"CitationRef\"\u003e101\u003c/span\u003e]. Clinical studies have confirmed that the risk of re-fracture is significantly higher in the population with low-energy trauma fractures [\u003cspan citationid=\"CR102\" class=\"CitationRef\"\u003e102\u003c/span\u003e, \u003cspan citationid=\"CR103\" class=\"CitationRef\"\u003e103\u003c/span\u003e], and most of them have severe osteoporosis, which can be induced by slight external forces [\u003cspan citationid=\"CR104\" class=\"CitationRef\"\u003e104\u003c/span\u003e]. In contrast, the fractures in patients with high-energy trauma are mostly due to overloading of external forces, and the underlying bone mass may not be poor. They still retain a certain degree of mechanical support and repair potential after surgery. Therefore, for low-energy trauma patients, the focus needs to be on improving bone quality, strengthening postoperative rehabilitation, and anti-osteoporosis treatment.\u003c/p\u003e \u003cp\u003eSurgery is the key treatment for OVCFs [\u003cspan citationid=\"CR105\" class=\"CitationRef\"\u003e105\u003c/span\u003e], but prolonged surgery is an independent risk factor for postoperative re-fracture. Previous studies have shown that the excessive duration of surgery is also an independent risk factor for cement leakage [\u003cspan citationid=\"CR106\" class=\"CitationRef\"\u003e106\u003c/span\u003e], suggesting that it may directly or indirectly affect the risk of re-fracture. Prolonged surgery will increase periprosthetic tissue damage and cause abnormal postoperative stress distribution [\u003cspan citationid=\"CR107\" class=\"CitationRef\"\u003e107\u003c/span\u003e]; simultaneously, it will increase the difficulty of bone cement operation and the likelihood of leakage [\u003cspan citationid=\"CR108\" class=\"CitationRef\"\u003e108\u003c/span\u003e, \u003cspan citationid=\"CR109\" class=\"CitationRef\"\u003e109\u003c/span\u003e]. Therefore, clinical optimization of surgical procedures and techniques is needed to improve efficiency and safety.\u003c/p\u003e \u003cp\u003eThe Braden score is widely used to predict the risk of pressure ulcers [\u003cspan citationid=\"CR110\" class=\"CitationRef\"\u003e110\u003c/span\u003e], and its correlation with postoperative refracture in patients with OVCFs has not been confirmed in previous studies. In this study, we hypothesized that some of the indicators in the score may indirectly indicate risk, such as poor nutritional status, which affects bone metabolism and healing and increases the likelihood of re-fracture [\u003cspan citationid=\"CR111\" class=\"CitationRef\"\u003e111\u003c/span\u003e, \u003cspan citationid=\"CR112\" class=\"CitationRef\"\u003e112\u003c/span\u003e]. Although the Braden score does not directly predict re-fracture, it can be used to focus on the risk of pressure ulcers, and preventive measures can be taken to indirectly help patients recover and reduce the potential risk of re-fracture.\u003c/p\u003e \u003cp\u003eThis study has the following limitations: first, as a retrospective study, there is a potential risk of selection bias and incomplete data collection. Second, the ML model was constructed based on single-center clinical data, and the institution-specific diagnosis and treatment patterns and regional population characteristics may limit the generalization ability of the model; also, the model was only validated by an internal dataset, and it needs to be further extrapolated and tested by an independent external dataset. Third, anti-osteoporosis medication was not included in the study model and analysis due to inconsistencies in the type of anti-osteoporosis medication used by different patients and the lack of documentation for some medication details, which made it difficult to assess the impact of this factor on the results.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn this study, we focused on predicting the risk of recurrent fracture after vertebral body strengthening in patients with OVCFs. For the first time, we used an interpretable ML method to construct a risk prediction model based on multidimensional clinical characteristics. The results showed that the XGBoost prediction model has excellent ability to predict the risk of postoperative recurrent fracture; meanwhile, key risk factors were identified by SHAP analysis, which provides a quantitative basis for accurate risk assessment. This technology can help clinicians clarify the decision logic of the model, identify high-risk groups and formulate individualized interventions, which is of great significance for optimizing treatment strategies and postoperative management modes, reducing the incidence of re-fracture, and providing a new methodological reference for the practice of precision medicine in OVCFs.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003e\u003cstrong\u003eDisclosure of potential conflicts of interest\u003c/strong\u003e\u003c/h2\u003e\n\u003cp\u003eZongJie Guo, JunPing Bao, Rui Shi, Shu Yang, and Lei Zhang declare that they have no conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study has passed the ethical review of the Ethics Committee of Zhongda Hospital Affiliated to Southeast University, and has been approved to exempt informed consent. The ethics approval number was 2024ZDSYLL187-P01.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003ch2\u003eClinical trial number:\u003c/h2\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003ch2\u003eAuthor contributions\u003c/h2\u003e\n\u003cp\u003eZongJie Guo, JunPing Bao, and Lei Zhang contributed to the study conception and design. Data collection was performed by ZongJie Guo. Analyses were performed by ZongJie Guo, JunPing Bao, and Lei Zhang. The first draft of the manuscript was written by ZongJie Guo. Rui Shi, and Shu Yang evaluated the patients' X-ray, CT and MRI images. Lei Zhang and JunPing Bao commented on the previous versions of the manuscript. All authors have contributed to the manuscript and approved the submitted version. All authors have reviewed the final version of the manuscript and approved it for publication.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003e2023 Jiangsu Health Development Research Center Open Project (JSHD202312); Jiangsu Province Capability Improvement Project through Science, Technology and Education (ZDXYS202210); Jiangsu Province High-level Hospital Construction Funds; project number: CZXM-GSP-RC53.\u003c/p\u003e\n\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\n\u003cp\u003eZongJie Guo, JunPing Bao, and Lei Zhang contributed to the study conception and design. Data collection was performed by ZongJie Guo. Analyses were performed by ZongJie Guo, JunPing Bao, and Lei Zhang. The first draft of the manuscript was written by ZongJie Guo. Rui Shi, and Shu Yang evaluated the patients' X-ray, CT and MRI images. Lei Zhang and JunPing Bao commented on the previous versions of the manuscript. All authors have contributed to the manuscript and approved the submitted version. All authors have reviewed the final version of the manuscript and approved it for publication.\u003c/p\u003e\n\u003ch2\u003eData Availability\u003c/h2\u003e\n\u003cp\u003eThe datasets analyzed in this study can be obtained from the corresponding authors with reasonable requirements.\u003c/p\u003e\n\u003ch2\u003e\u0026nbsp;\u003c/h2\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eShubhashree M, Naik R, Doddamani S, et al. An updated review of single herbal drugs in the management of osteoporosis [J]. Int J Complement Altern Med, 2018, 11: 82\u0026ndash;86.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSakat B T, Sakhare R B, Suryvanshi U C, et al. Osteoporosis: The brittle bone [J]. Asian Journal of Pharmaceutical Research, 2018, 8(1): 39\u0026ndash;43.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSi L, Winzenberg T M, Jiang Q, et al. Projection of osteoporosis-related fractures and costs in China: 2010\u0026ndash;2050 [J]. Osteoporos Int, 2015, 26(7): 1929\u0026ndash;1937.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohnston C C, Longcope C. Premenopausal bone loss\u0026ndash;a risk factor for osteoporosis [J]. N Engl J Med, 1990, 323(18): 1271\u0026ndash;1273.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhan A A, Slart R H J A, Ali D S, et al. Osteoporotic Fractures: Diagnosis, Evaluation, and Significance From the International Working Group on DXA Best Practices [J]. Mayo Clin Proc, 2024, 99(7): 1127\u0026ndash;1141.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSiris E S, Adler R, Bilezikian J, et al. The clinical diagnosis of osteoporosis: a position statement from the National Bone Health Alliance Working Group [J]. Osteoporos Int, 2014, 25(5): 1439\u0026ndash;1443.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang G, Yang L, Yang C, et al. Expert Consensus on Bone Repair Strategies for Osteoporotic Vertebral Compression Fractures[J]. Journal of Clinical Surgery, 2024, 32(04): 442\u0026ndash;448.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVoormolen M H J, Lohle P N M, Juttmann J R, et al. The risk of new osteoporotic vertebral compression fractures in the year after percutaneous vertebroplasty [J]. J Vasc Interv Radiol, 2006, 17(1): 71\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKlazen C A H, Venmans A, de Vries J, et al. Percutaneous vertebroplasty is not a risk factor for new osteoporotic compression fractures: results from VERTOS II [J]. AJNR Am J Neuroradiol, 2010, 31(8): 1447\u0026ndash;1450.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLindsay R, Burge R T, Strauss D M. One year outcomes and costs following a vertebral fracture [J]. Osteoporos Int, 2005, 16(1): 78\u0026ndash;85.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen Z, Yao Z, Wu C, et al. Assessment of clinical, imaging, surgical risk factors for subsequent fracture following vertebral augmentation in osteoporotic patients [J]. Skeletal Radiol, 2022, 51(8): 1623\u0026ndash;1630.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJordan M I, Mitchell T M. Machine learning: Trends, perspectives, and prospects [J]. Science, 2015, 349(6245): 255\u0026ndash;260.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMishra R K, Reddy G S, Pathak H. The understanding of deep learning: A comprehensive review [J]. Mathematical Problems in Engineering, 2021, 2021(1): 5548884.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbdollah V, Parent E C, Dolatabadi S, et al. Texture analysis in the classification of T2 -weighted magnetic resonance images in persons with and without low back pain [J]. J Orthop Res, 2021, 39(10): 2187\u0026ndash;2196.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuber F A, Stutz S, Vittoria de Martini I, et al. Qualitative versus quantitative lumbar spinal stenosis grading by machine learning supported texture analysis-Experience from the LSOS study cohort [J]. Eur J Radiol, 2019, 114: 45\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan Z, Wei B, Leung S, et al. Automated Pathogenesis-Based Diagnosis of Lumbar Neural Foraminal Stenosis via Deep Multiscale Multitask Learning [J]. Neuroinformatics, 2018, 16(3\u0026ndash;4): 325\u0026ndash;337.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLevitt J, Edhi M M, Thorpe R V, et al. Pain phenotypes classified by machine learning using electroencephalography features [J]. Neuroimage, 2020, 223: 117256.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOude Nijeweme-d'Hollosy W, van Velsen L, Poel M, et al. Evaluation of three machine learning models for self-referral decision support on low back pain in primary care [J]. Int J Med Inform, 2018, 110: 31\u0026ndash;41.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRak M, Steffen J, Meyer A, et al. Combining convolutional neural networks and star convex cuts for fast whole spine vertebra segmentation in MRI [J]. Comput Methods Programs Biomed, 2019, 177: 47\u0026ndash;56.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKuang X, Cheung J P, Wu H, et al. MRI-SegFlow: a novel unsupervised deep learning pipeline enabling accurate vertebral segmentation of MRI images [J]. Annu Int Conf IEEE Eng Med Biol Soc, 2020, 2020: 1633\u0026ndash;1636.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi X, Dou Q, Chen H, et al. 3D multi-scale FCN with random modality voxel dropout learning for Intervertebral Disc Localization and Segmentation from Multi-modality MR Images [J]. Med Image Anal, 2018, 45: 41\u0026ndash;54.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePaugam F, Lefeuvre J, Perone C S, et al. Open-source pipeline for multi-class segmentation of the spinal cord with deep learning [J]. Magn Reson Imaging, 2019, 64: 21\u0026ndash;27.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWirries A, Geiger F, Hammad A, et al. Artificial intelligence facilitates decision-making in the treatment of lumbar disc herniations [J]. Eur Spine J, 2021, 30(8): 2176\u0026ndash;2184.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoller B L, Boutin R D, O'Gara T J, et al. Accurate prediction of lumbar microdecompression level with an automated MRI grading system [J]. Skeletal Radiol, 2021, 50(1): 69\u0026ndash;78.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarhade A V, Fogel H A, Cha T D, et al. Development of prediction models for clinically meaningful improvement in PROMIS scores after lumbar decompression [J]. Spine J, 2021, 21(3): 397\u0026ndash;404.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuo Z, Wang P, Ye S, et al. Interpretable Machine Learning Models Based on Shapley Additive Explanations for Predicting the Risk of Cerebrospinal Fluid Leakage in Lumbar Fusion Surgery [J]. Spine (Phila Pa 1976), 2024, 49(18): 1281\u0026ndash;1293.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen H, Lundberg S M, Lee S-I. Explaining a series of models by propagating Shapley values [J]. Nat Commun, 2022, 13(1): 4512.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun J, Sun C-K, Tang Y-X, et al. Application of SHAP for Explainable Machine Learning on Age-Based Subgrouping Mammography Questionnaire Data for Positive Mammography Prediction and Risk Factor Identification [J]. Healthcare (Basel), 2023, 11(14).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRiley R D, Ensor J, Snell K I E, et al. Calculating the sample size required for developing a clinical prediction model [J]. BMJ, 2020, 368: m441.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi J, Li K, Wang X, et al. Classification and Prevention of Bone Cement Leakage in Percutaneous Kyphoplasty[J]. Chinese Journal of Osteoporosis and Bone Mineral Research, 2009, 2(01): 50\u0026ndash;52.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNi W, Chi Y, Lin Y, et al. Types and Clinical Significance of Bone Cement Leakage Complicating Percutaneous Vertebral Augmentation[J]. Chinese Journal of Surgery, 2006, (04): 231\u0026ndash;234.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang D, Mao K, Qiang X, et al. Classification and Clinical Significance of Bone Cement Distribution Patterns after Vertebral Augmentation[J]. Chinese Journal of Traumatology, 2018, 34(2): 130\u0026ndash;137.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRen G, Yu K, Xie Z, et al. Current Applications of Machine Learning in Spine: From Clinical View [J]. Global Spine J, 2022, 12(8): 1827\u0026ndash;1840.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang P, Liu L, Xie Z, et al. Explainable Machine Learning Models for Prediction of Surgical Site Infection After Posterior Lumbar Fusion Surgery Based on Shapley Additive Explanations [J]. World Neurosurg, 2025, 197: 123942.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFenlon C, O'Grady L, Doherty M L, et al. A discussion of calibration techniques for evaluating binary and categorical predictive models [J]. Prev Vet Med, 2018, 149: 107\u0026ndash;114.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang Z, Rousson V, Lee W-C, et al. Decision curve analysis: a technical note [J]. Ann Transl Med, 2018, 6(15): 308.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTakefuji Y. Beyond XGBoost and SHAP: Unveiling true feature importance [J]. J Hazard Mater, 2025, 488: 137382.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen T, Guestrin C. Xgboost: A scalable tree boosting system; proceedings of the Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, F, 2016 [C].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAl-Zakhali O A, Zeebaree S, Askar S. Comparative Analysis of XGBoost Performance for Text Classification with CPU Parallel and Non-Parallel Processing [J]. The Indonesian Journal of Computer Science, 2024, 13(2).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen T, Guestrin C. Xgboost: Reliable large-scale tree boosting system; proceedings of the Proceedings of the 22nd SIGKDD Conference on Knowledge Discovery and Data Mining, F, 2017 [C].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrinsztajn L, Oyallon E, Varoquaux G. Why do tree-based models still outperform deep learning on typical tabular data? [J]. Advances in neural information processing systems, 2022, 35: 507\u0026ndash;520.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAli Z A, Abduljabbar Z H, Tahir H A, et al. eXtreme gradient boosting algorithm with machine learning: A review [J]. Academic Journal of Nawroz University, 2023, 12(2): 320\u0026ndash;334.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStokanović S, Đukić D, Miljković N. Robustness of XGBoost algorithm to missing features for binary classification of medical data; proceedings of the 2024 23rd International Symposium INFOTEH-JAHORINA (INFOTEH), F, 2024 [C]. IEEE.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeng L, Feng C, Chen J, et al. The risk factors of vertebral refracture after kyphoplasty in patients with osteoporotic vertebral compression fractures: a study protocol for a prospective cohort study [J]. BMC Musculoskelet Disord, 2018, 19(1): 195.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDai C, Liang G, Zhang Y, et al. Risk factors of vertebral re-fracture after PVP or PKP for osteoporotic vertebral compression fractures, especially in Eastern Asia: a systematic review and meta-analysis [J]. J Orthop Surg Res, 2022, 17(1): 161.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee B G, Choi J-H, Kim D-Y, et al. Risk factors for newly developed osteoporotic vertebral compression fractures following treatment for osteoporotic vertebral compression fractures [J]. Spine J, 2019, 19(2): 301\u0026ndash;305.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJung B H, Jeon M J, Bai S W. Hormone-dependent aging problems in women [J]. Yonsei Med J, 2008, 49(3): 345\u0026ndash;351.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMajeska R J, Ryaby J T, Einhorn T A. Direct modulation of osteoblastic activity with estrogen [J]. J Bone Joint Surg Am, 1994, 76(5): 713\u0026ndash;721.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eV\u0026auml;\u0026auml;n\u0026auml;nen H K, H\u0026auml;rk\u0026ouml;nen P L. Estrogen and bone metabolism [J]. Maturitas, 1996, 23: S65-S69.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTehrani S S, Moallem M, Ebrahimi R, et al. Status of circulating bone turnover markers in elderly osteoporosis/osteopenia patients in comparison with healthy subjects [J]. Asian Biomed, 2020, 14(3): 97\u0026ndash;106.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSzulc P, Garnero P, Munoz F, et al. Cross-sectional evaluation of bone metabolism in men [J]. Journal of Bone and Mineral Research, 2001, 16(9): 1642\u0026ndash;1650.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarnero P, Hausherr E, Chapuy M C, et al. Markers of bone resorption predict hip fracture in elderly women: the EPIDOS Prospective Study [J]. Journal of bone and mineral research, 1996, 11(10): 1531\u0026ndash;1538.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng C, Chen P, Kuo Y, et al. The effects of disc degeneration and muscle dysfunction on cervical spine stability from a biomechanical study [J]. Proceedings of the Institution of Mechanical Engineers, Part H: Journal of Engineering in Medicine, 2011, 225(2): 149\u0026ndash;157.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMohamed A, Mohamed E, Mohamed J. Exercises Training Program: Its Effect on Muscle Strength and Activity of Daily Living among Elderly People [J]. Evidence-Based Nursing Research, 2019, 1(4): 24\u0026ndash;34.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAagaard P, Suetta C, Caserotti P, et al. Role of the nervous system in sarcopenia and muscle atrophy with aging: strength training as a countermeasure [J]. Scandinavian journal of medicine \u0026amp; science in sports, 2010, 20(1): 49\u0026ndash;64.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTsuda T. Epidemiology of fragility fractures and fall prevention in the elderly: a systematic review of the literature [J]. Curr Orthop Pract, 2017, 28(6): 580\u0026ndash;585.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLukert B P, Raisz L G. Glucocorticoid-induced osteoporosis: pathogenesis and management [J]. Ann Intern Med, 1990, 112(5): 352\u0026ndash;364.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhu S-Y, Zhong Z-M, Wu Q, et al. Risk factors for bone cement leakage in percutaneous vertebroplasty: a retrospective study of four hundred and eighty five patients [J]. International Orthopaedics, 2016, 40: 1205\u0026ndash;1210.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMao W, Dong F, Huang G, et al. Risk factors for secondary fractures to percutaneous vertebroplasty for osteoporotic vertebral compression fractures: a systematic review [J]. J Orthop Surg Res, 2021, 16(1): 644.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin Y-H, Lin J, Xu J-Y, et al. What Risk Factors Are Associated With Recurrent Osteoporotic Vertebral Compression Fractures After Percutaneous Vertebral Augmentation? A Meta-analysis [J]. Clin Orthop Relat Res, 2025.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa Y, Lu Q, Wang X, et al. Establishment and validation of a nomogram for predicting new fractures after PKP treatment of for osteoporotic vertebral compression fractures in the elderly individuals [J]. BMC Musculoskelet Disord, 2023, 24(1): 728.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaroud G, Bohner M. Biomechanical impact of vertebroplasty. Postoperative biomechanics of vertebroplasty [J]. Joint Bone Spine, 2006, 73(2): 144\u0026ndash;150.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOda K, Shibayama Y, Abe M, et al. Morphogenesis of vertebral deformities in involutional osteoporosis. Age-related, three-dimensional trabecular structure [J]. Spine (Phila Pa 1976), 1998, 23(9).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKopperdahl D L, Pearlman J L, Keaveny T M. Biomechanical consequences of an isolated overload on the human vertebral body [J]. J Orthop Res, 2000, 18(5): 685\u0026ndash;690.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLaredo J-D, Hamze B. Complications of percutaneous vertebroplasty and their prevention [J]. Semin Ultrasound CT MR, 2005, 26(2): 65\u0026ndash;80.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbd El-Rahman A M, Lazzarotti A G, Cosottini M, et al. Pulmonary embolism caused by cement leakage during percutaneous vertebroplasty. A case report of successful conservative management [J]. Neuroradiol J, 2012, 25(4): 481\u0026ndash;485.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLim S H, Kim H, Kim H K, et al. Multiple cardiac perforations and pulmonary embolism caused by cement leakage after percutaneous vertebroplasty [J]. Eur J Cardiothorac Surg, 2008, 33(3): 510\u0026ndash;512.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu B, Wang C-Y. Osteoporosis: The Result of an 'Aged' Bone Microenvironment [J]. Trends Mol Med, 2016, 22(8): 641\u0026ndash;644.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKillinger Z. [Early diagnosis of postmenopausal osteoporosis] [J]. Bratisl Lek Listy, 2000, 101(3): 179\u0026ndash;180.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNing L, Zhu J, Tian S, et al. Correlation Analysis Between Basic Diseases and Subsequent Vertebral Fractures After Percutaneous Kyphoplasty (PKP) for Osteoporotic Vertebral Compression Fractures [J]. Pain Physician, 2021, 24(6): E803-E810.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eParfitt A M. Trabecular bone architecture in the pathogenesis and prevention of fracture [J]. Am J Med, 1987, 82(1B): 68\u0026ndash;72.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarrison J G, Gargac J A, Niebur G L. Shear strength and toughness of trabecular bone are more sensitive to density than damage [J]. J Biomech, 2011, 44(16): 2747\u0026ndash;2754.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMellstr\u0026ouml;m D, Yang X, Li Z, et al. Proportion and Characteristics of Patients in Sweden Remaining at High Risk of Fracture Despite Prior Treatment [J]. Clin Ther, 2016, 38(7).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRizzoli R. Postmenopausal osteoporosis: Assessment and management [J]. Best Pract Res Clin Endocrinol Metab, 2018, 32(5): 739\u0026ndash;757.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLems W F, Raterman H G. Critical issues and current challenges in osteoporosis and fracture prevention. An overview of unmet needs [J]. Ther Adv Musculoskelet Dis, 2017, 9(12): 299\u0026ndash;316.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeng S-T, Yang Y, Li X, et al. Risk Factors of New Symptomatic Fractures After Vertebroplasty: A Retrospective Cohort Study of 268 Patients with Painful Osteoporotic Vertebral Compression Fracture [J]. World Neurosurg, 2024, 187: e890-e897.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKlotzbuecher C M, Ross P D, Landsman P B, et al. Patients with prior fractures have an increased risk of future fractures: a summary of the literature and statistical synthesis [J]. J Bone Miner Res, 2000, 15(4): 721\u0026ndash;739.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGehlbach S, Saag K G, Adachi J D, et al. Previous fractures at multiple sites increase the risk for subsequent fractures: the Global Longitudinal Study of Osteoporosis in Women [J]. J Bone Miner Res, 2012, 27(3): 645\u0026ndash;653.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoerth R M, Seidt B M, Shah M, et al. Mechanical and structural properties of bone in non-critical and critical healing in rat [J]. Acta Biomater, 2014, 10(9): 4009\u0026ndash;4019.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKreider J M, Goldstein S A. Trabecular bone mechanical properties in patients with fragility fractures [J]. Clin Orthop Relat Res, 2009, 467(8): 1955\u0026ndash;1963.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorgan E F, Unnikrisnan G U, Hussein A I. Bone Mechanical Properties in Healthy and Diseased States [J]. Annu Rev Biomed Eng, 2018, 20: 119\u0026ndash;143.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuber-Lang M, Kovtun A, Ignatius A. The role of complement in trauma and fracture healing [J]. Semin Immunol, 2013, 25(1): 73\u0026ndash;78.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLoi F, C\u0026oacute;rdova L A, Pajarinen J, et al. Inflammation, fracture and bone repair [J]. Bone, 2016, 86: 119\u0026ndash;130.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCuthbertson D P. The disturbance of metabolism produced by bony and non-bony injury, with notes on certain abnormal conditions of bone [J]. Biochem J, 1930, 24(4): 1244\u0026ndash;1263.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQian Y, Hu X, Li C, et al. Development of a nomogram model for prediction of new adjacent vertebral compression fractures after vertebroplasty [J]. BMC Surg, 2023, 23(1): 197.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHey H W D, Tan J H, Tan C S, et al. Subsequent Vertebral Fractures Post Cement Augmentation of the Thoracolumbar Spine: Does it Correlate With Level-specific Bone Mineral Density Scores? [J]. Spine (Phila Pa 1976), 2015, 40(24): 1903\u0026ndash;1909.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMelton L J, Atkinson E J, O'Fallon W M, et al. Long-term fracture prediction by bone mineral assessed at different skeletal sites [J]. J Bone Miner Res, 1993, 8(10): 1227\u0026ndash;1233.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCummings S R, Black D M, Nevitt M C, et al. Appendicular bone density and age predict hip fracture in women. The Study of Osteoporotic Fractures Research Group [J]. JAMA, 1990, 263(5): 665\u0026ndash;668.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCauley J A, Hochberg M C, Lui L-Y, et al. Long-term risk of incident vertebral fractures [J]. JAMA, 2007, 298(23): 2761\u0026ndash;2767.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSeeman E. Invited Review: Pathogenesis of osteoporosis [J]. J Appl Physiol (1985), 2003, 95(5): 2142\u0026ndash;2151.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIolascon G, Napolano R, Gioia M, et al. The contribution of cortical and trabecular tissues to bone strength: insights from denosumab studies [J]. Clin Cases Miner Bone Metab, 2013, 10(1): 47\u0026ndash;51.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNatesan V, Kim S-J. Metabolic Bone Diseases and New Drug Developments [J]. Biomol Ther (Seoul), 2022, 30(4): 309\u0026ndash;319.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim J-M, Shin D A, Byun D-H, et al. Effect of bone cement volume and stiffness on occurrences of adjacent vertebral fractures after vertebroplasty [J]. J Korean Neurosurg Soc, 2012, 52(5): 435\u0026ndash;440.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen X-S, Jiang J-M, Sun P-D, et al. How the clinical dosage of bone cement biomechanically affects adjacent vertebrae [J]. J Orthop Surg Res, 2020, 15(1): 370.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSimons D G, Travell J G. Myofascial origins of low back pain. 1. Principles of diagnosis and treatment [J]. Postgrad Med, 1983, 73(2).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao H, He Y, Yang J-S, et al. Can paraspinal muscle degeneration be a reason for refractures after percutaneous kyphoplasty? A magnetic resonance imaging observation [J]. J Orthop Surg Res, 2021, 16(1): 476.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen M, Yang C, Cai Z, et al. Lumbar posterior group muscle degeneration: Influencing factors of adjacent vertebral body re-fracture after percutaneous vertebroplasty [J]. Front Med (Lausanne), 2022, 9: 1078403.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGracovetsky S. Is the lumbodorsal fascia necessary? [J]. J Bodyw Mov Ther, 2008, 12(3): 194\u0026ndash;197.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAspden R M. Review of the functional anatomy of the spinal ligaments and the lumbar erector spinae muscles [J]. Clinical Anatomy: The Official Journal of the American Association of Clinical Anatomists and the British Association of Clinical Anatomists, 1992, 5(5): 372\u0026ndash;387.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWong C, Rasmussen J, Simonsen E, et al. The influence of muscle forces on the stress distribution in the lumbar spine [J]. Open Spine J, 2011, 3: 21\u0026ndash;26.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHilaire C S, Johnson A, Loseth C, et al. Facial fractures and associated injuries in high- versus low-energy trauma: all are not created equal [J]. Maxillofac Plast Reconstr Surg, 2020, 42(1): 22.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRobinson C M, Royds M, Abraham A, et al. Refractures in patients at least forty-five years old. a prospective analysis of twenty-two thousand and sixty patients [J]. J Bone Joint Surg Am, 2002, 84(9): 1528\u0026ndash;1533.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCenter J R, Bliuc D, Nguyen T V, et al. Risk of subsequent fracture after low-trauma fracture in men and women [J]. JAMA, 2007, 297(4): 387\u0026ndash;394.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGorter E A, Reinders C R, Krijnen P, et al. The effect of osteoporosis and its treatment on fracture healing a systematic review of animal and clinical studies [J]. Bone Rep, 2021, 15: 101117.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi J, Xu L, Liu Y, et al. Open Surgical Treatments of Osteoporotic Vertebral Compression Fractures [J]. Orthop Surg, 2023, 15(11): 2743\u0026ndash;2748.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi W, Wang H, Dong S, et al. Establishment and validation of a nomogram and web calculator for the risk of new vertebral compression fractures and cement leakage after percutaneous vertebroplasty in patients with osteoporotic vertebral compression fractures [J]. Eur Spine J, 2022, 31(5): 1108\u0026ndash;1121.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJeon C-H, Lee Y-S, Youn S-J, et al. Factors affecting postural reduction in posterior surgery for thoracolumbar burst fracture [J]. J Spinal Disord Tech, 2015, 28(4): E225-E230.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang M, Jin Q. High-viscosity bone cement for vertebral compression fractures: a prospective study on intravertebral diffusion and leakage of bone cement [J]. BMC Musculoskelet Disord, 2020, 21(1): 589.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLai P-L, Chen L-H, Chen W-J, et al. Chemical and physical properties of bone cement for vertebroplasty [J]. Biomed J, 2013, 36(4): 162\u0026ndash;167.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKring D L. Reliability and validity of the Braden Scale for predicting pressure ulcer risk [J]. J Wound Ostomy Continence Nurs, 2007, 34(4): 399\u0026ndash;406.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeesters D M, Wijnands K A P, Brink P R G, et al. Malnutrition and Fracture Healing: Are Specific Deficiencies in Amino Acids Important in Nonunion Development? [J]. Nutrients, 2018, 10(11).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGiganti M G, Tresoldi I, Masuelli L, et al. Fracture healing: from basic science to role of nutrition [J]. Front Biosci (Landmark Ed), 2014, 19(7): 1162\u0026ndash;1175.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Osteoporotic vertebral compression fractures, Vertebral augmentation, Refracture, Interpretable machine learning, Risk prediction model","lastPublishedDoi":"10.21203/rs.3.rs-8863516/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8863516/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eOsteoporotic vertebral compression fractures (OVCFs) are severe osteoporosis complications; vertebral augmentation is the preferred minimally invasive treatment, but postoperative refracture risk exists. Traditional logistic regression fails to accurately assess individual risks, while interpretable machine learning (ML) excels in high-dimensional data processing, with limited relevant studies.\u003c/p\u003e\u003ch2\u003ePurposes:\u003c/h2\u003e \u003cp\u003eTo develop an interpretable ML prediction model for identifying risk factors of subsequent fractures after vertebral augmentation in patients with OVCFs.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eA retrospective analysis was conducted on clinical data of 1,502 OVCF patients who underwent vertebral augmentation. Thirty-six characteristic indicators were extracted from electronic medical records and imaging systems. Six ML prediction models were constructed. Prediction performance was comprehensively evaluated using receiver operating characteristic (ROC) curves, accuracy, recall, F1 score, precision, calibration curves, and decision curve analysis. The optimal model was interpreted globally and locally via Shapley Additive exPlanations (SHAP) to analyze the contribution of key features.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe 2-year post-operative subsequent fracture incidence in the study cohort was 9.65% (145 cases). After data preprocessing and model training, the extreme gradient boosting (XGBoost) model demonstrated the best performance on the test set. Calibration curve and decision curve analyses showed high consistency between predicted results and actual risks, with significant clinical net benefit. SHAP analysis identified nine key risk factors ranked by importance: age, bone cement leakage and types, history of osteoporosis, Previous history of fractures, bone mineral density, thoracolumbar fascitis, types of trauma, duration of surgery, and Braden score.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThe XGBoost model combined with SHAP represents an effective tool for predicting subsequent fracture risk after vertebral augmentation in OVCF patients. Clinical application of this prediction model can assist clinicians in formulating individualized intervention strategies, thereby optimizing treatment protocols and post-operative management to reduce post-operative subsequent fracture incidence.\u003c/p\u003e","manuscriptTitle":"Application Study of Interpretable Machine Learning Models for Predicting Postoperative Refracture After Vertebral Augmentation in Osteoporotic Vertebral Compression Fractures","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-24 13:14:55","doi":"10.21203/rs.3.rs-8863516/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8b111546-79b8-4784-9a6b-6cbf41cb66b7","owner":[],"postedDate":"March 24th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-26T16:09:28+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-24 13:14:55","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8863516","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8863516","identity":"rs-8863516","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}