经皮椎体后凸成形术后邻近椎体骨折的列线图预测模型构建

Construction of a nomogram prediction model for adjacent vertebral fractures after percutaneous kyphoplasty

中文关键词:  骨质疏松椎体压缩骨折  椎体后凸成形术  邻近椎体骨折  危险因素  预测模型

英文关键词:Osteoporosis vertebral compression fracture  Percutaneous kyphoplasty  Adjacent vertebral body fracture  Risk factor  Prediction model

中文摘要:

  【摘要】 目的：分析骨质疏松性胸腰椎压缩性骨折（osteoporotic thoracolumbar compression fracture，OVCF）患者经皮椎体后凸成形术（percutaneous kyphoplasty，PKP）后邻近椎体新发骨折的危险因素，并构建列线图预测模型。方法：回顾性分析2014年3月～2019年3月本院收治的行PKP治疗的OVCF患者的临床资料。根据随访过程中邻近椎体是否新发骨折，将患者分成骨折组和非骨折组（对照组），记录两组患者的性别、年龄、骨折椎体位置、骨水泥注射途径（单/双）、注射剂量、注射后形态、骨水泥渗漏情况、椎旁肌横截面积（cross-sectional area，CSA）、椎体CSA、手术前后及随访时的伤椎前缘高度和Cobb角等临床资料。应用单因素及多因素回归分析PKP术后邻近椎体骨折的危险因素，并利用R软件建立其列线图预测模型，运用受试者工作特征（receiver operating characteristic，ROC）曲线、校准图以及决策曲线验证该模型的预测效能、准确度以及临床价值。结果：总共纳入224例患者，其中男性42例，女性182例，平均年龄65.19±8.62岁。术后36例患者（共45个椎体）发生相邻椎体骨折，发生率为16.07%。单因素分析结果显示，年龄、椎旁肌CSA/椎体CSA值、骨折椎体位置、骨水泥注射剂量、骨水泥渗漏、注射后形态在骨折组与非骨折组之间存在显著性差异（P＜0.05）。进一步多元Logistic回归分析显示，高龄、椎旁肌CSA/椎体CSA值降低、胸腰段骨折、骨水泥注射剂量、骨水泥渗漏、注射后呈团块状为PKP术后邻近椎体新发骨折的高危因素。构建列线图模型，所建模型ROC曲线的曲线下面积（area under the curve，AUC）为0.988；校准图显示模型校准曲线与标准曲线较为接近；决策曲线提示模型在1%～98%范围内，模型预测净获益值较高，表明该模型具有较高的预测效能。结论：OVCF患者PKP术后邻近椎体骨折与年龄、椎旁肌退化程度、骨折椎体位置、骨水泥注射剂量、骨水泥渗漏、注射后形态等因素相关，据此构建的PKP术后邻近椎体骨折列线图预测模型具有良好的预测效能和准确度。

英文摘要:

  【Abstract】 Objectives: To analyze the risk factors of new fractures of adjacent vertebrae after percutaneous kyphoplasty(PKP) in patients with osteoporotic thoracolumbar compression fractures(OVCF), and to construct a nomogram prediction model. Methods: Clinical data of patients with OVCF who underwent PKP in our hospital from March 2014 to March 2019 were retrospectively analyzed. The patients were divided into fracture group and non-fracture group(control group) according to whether a new fracture in the adjacent vertebrae occurred or not during follow-up. Clinical data of gender, age, location of fractured vertebral body, paravertebral muscle cross-sectional area(CSA), vertebral CSA, injection route(unilateral and bilateral), injection dose, post-injection morphology(lumps or spongy), bone cement leakage, operation time, the anterior height ratio of injured vertebrae and the Cobb angle of kyphosis before and after operation and during follow-up of both groups were collected. Single-factor and multi-factor regression were used to analyze the risk factors of adjacent vertebral fractures after PKP, and a nomogram prediction model was established by R software. Receiver operating characteristic(ROC) curve, calibration diagram and decision curve were used to verify the prediction efficiency, accuracy and clinical value of the model. Results: A total of 224 patients were enrolled, including 42 males and 182 females, with an average age of 65.19±8.62 years. Postoperative fractures of adjacent vertebrae occurred in 36 patients(45 vertebrae in total), with an incidence of 16.07%. Univariate analysis showed that there were significant differences in age, CSA value of paravertebral muscle/vertebral CSA value, location of fractured vertebral body, injection dose of bone cement, bone cement leakage and post-injection morphology between the fracture group and non-fracture group(P<0.05). Further multivariate logistic regression analysis showed that advanced age, decreased CSA value of paraspinal muscle/vertebral CSA, thoracolumbar fracture, injection dose of bone cement, bone cement leakage and agglomerate after injection were the risk factors for new fracture of adjacent vertebral body after PKP. The nomogram prediction model was constructed, and the AUC of the ROC curve of the model was 0.988. The calibration diagram showed that the calibration curve of the model was close to the standard curve. The decision curve indicated that the model was in the range of 1%-98%, and the predicted net benefit value of the model was high, indicating a high prediction efficiency of the model. Conclusions: The fracture of adjacent vertebrae after PKP in patients with OVCF is correlated with age, degree of paraspinal muscle degeneration, position of fractured vertebrae, dosage of bone cement injection, bone cement leakage, shape after injection and other factors. Based on this, the risk prediction model of adjacent vertebral fractures in patients with OVCF after PKP has good prediction efficiency and accuracy.