预测骨质疏松性椎体压缩骨折经皮椎体成形术后残余背痛的列线图

A nomogram for predicting residual back pain after percutaneous vertebroplasty for osteoporotic vertebral compression fractures

目的:构建预测骨质疏松性椎体压缩性骨折(osteoporotic vertebral compression fractures,OVCFs)经皮椎体成形术(percutaneous vertebroplasty,PVP)后残余背痛(residual back pain,RBP)的列线图。方法:回顾性分析2020年1月至2022年12月245例接受PVP治疗的OVCFs患者的临床资料,男47例,女198例,年龄65~77(71.47±9.03)岁,根据是否发生RBP分为RBP组与无RBP组。收集患者的性别、年龄、合并症情况、骨折发生节段、身体质量指数(body mass index,BMI)、骨密度(bone mineral density,BMD)、视觉模拟评分(visual analogue scale,VAS)、Oswestry功能障碍指数(Oswestry disability index,ODI)等一般资料;以及术前与术后24 h的影像学参数,包括椎体前缘高度(anterior vertebral height,AVH)、椎体前缘高度比(anterior vertebral height ratio,AVHR)、Cobb角、椎体内真空裂(intravertebral vacuum cleft,IVC)、胸腰筋膜(thoracolumbar fascia,TLF)损伤、椎旁肌脂肪变性、骨水泥注射量、骨水泥渗漏、骨水泥弥散形态、椎体前缘高度恢复比(anterior vertebral height recovery ratio,AVHRR)、Cobb角变化等。对以上因素进行单因素分析,再采用多因素Logistic回归模筛选术后发生RBP的独立危险因素,并完成Nomogram模型的构建与验证,分别采用受试者工作特征(receiver operating characteristic,ROC)曲线和校准曲线进行模型的预测性能和准确性的判定,另采用Hosmer-Lemeshow(H-L)检验进行评估,计算ROC曲线下面积(area under curve,AUC),使用Harrell一致性指数(C指数)评价模型的预测效能,使用决策曲线分析(decision curve analysis,DCA)评价模型的临床实用性。结果:RBP组34例,无RBP组211例。两组性别、年龄、合并症、骨折节段、BMI、BMD、VAS及ODI、AVH、AVHR、Cobb角等比较,差异均无统计学意义(P>0.05)。单因素分析结果显示:RBP组6例出现IVC,无RBP组13例,RBP组IVC比例高于无RBP组(χ2=5.400,P=0.020);RBP组6例出现TLF损伤,无RBP组11例,RBP组TLF损伤比例高于无RBP组(χ2=7.011,P=0.008);RBP组椎旁肌脂肪变性3-4级为18例,无RBP组为41例,RBP组高于无RBP组(χ2=21.618,P<0.001),RBP组骨水泥弥散形态为团块型比例高于无RBP组(χ2=6.836,P=0.009)。多因素Logistic回归分析结果显示,存在IVC(χ2=4.974,P=0.025)、合并TLF损伤(χ2=5.231,P=0.023)、椎旁肌脂肪变性Goutallier分级>2级(χ2=15.124,P<0.001)以及骨水泥弥散形态为团块型(χ2=4.168,P=0.038)为PVP术后发生RBP的独立危险因素。模型ROC曲线得出原始模型AUC为0.816[OR=2.862,95%CI(0.776,0.894),P<0.001],通过200个自举样本进行模型内部验证,得出C指数值为0.936,校准曲线显示预测概率曲线与实际概率曲线接近,H-L拟合优度检验结果为χ2=5.796,P=0.670,DCA分析结果显示当阈值在6%~71%时决策曲线位于None线与All线上方。结论:存在IVC、合并TLF损伤、椎旁肌脂肪变性Goutallier分级>2级以及骨水泥弥散形态为团块型为PVP术后发生RBP的独立危险因素,本研究所构建的PVP术后发生RBP的风险预测模型具有较好的预测性能和较好的临床实用性。

Objective To construct percutaneous vertebroplasty for predicting osteoporotic vertebral compression fractures(OVCFs)nomogram of residual back pain(RBP)after percutaneous vertebroplasty(PVP).Methods Clinical data of 245 OVCFs patients who were performed PVP from January 2020 to December 2022 were retrospectively analyzed,including 47 males and 198 females,aged from 65 to 77 years old with an average of(71.47±9.03)years old,and were divided into RBP group and non-RBP group according to whether RBP occurred.Gender,age,comorbidities,fracture stage,body mass index(BMI),bone mineral density(BMD),visual analogue scale(VAS),Oswestry disability index(ODI)and other general information were collected;anterior vertebral height(AVH),anterior vertebral height ratio(AVH),anterior vertebral height ratio(AVHR),Cobb angle,intravertebral vacuum cleft(IVC),thoracolumbar fascia(TLF)injury,paravertebral muscle steatosis,injection volume and leakage of bone cement,bone cement dispersion pattern,anterior vertebral height recovery ratio(AVHRR),Cobb angle changes,etc.imaging parameters before operation and 24 h after operation were collected.Univariate analysis was performed to analysis above factors,and multivariate Logistic regression model was used to investigate independent risk factors for postoperative RBP,and Nomogram model was constructed and verified;receiver operating characteristic(ROC)curve and calibration curve were used to determine predictive performance and accuracy of the model,and Hosmer-Lemeshow(H-L)test was used for evaluation.The area under curve(AUC)of ROC was calculated,and Harrell consistency index(C index)was used to evaluate the predictive efficiency of model;decision curve analysis(DCA)was used to evaluate clinical practicability of model.Results There were 34 patients in RBP group and 211 patients in non-RBP group.There were no significant differences in gender,age,comorbidities,fracture stage,BMI,BMD,VAS,ODI,AVH,AVHR and Cobb angle between two groups(P>0.05).Univariate analysis showed 6 patients occurred IVC in RBP group and 13 patients in non-RBP,the number of IVC in RBP group was higher than that in non-RBP group(χ2=5.400,P=0.020);6 patients occuured TLF injury in RBP group and 11 patients in non-RBP group,the number of TLF injury in RBP group was higher than that in non-RBP group(χ2=7.011,P=0.008);In RBP group,18 patients with grade 3 to 4 paraptebral steatosis and 41 patients in non-RBP group,RBP group was higher than non-RBP group(χ2=21.618,P<0.001),and the proportion of bone cement mass in RBP group was higher than non-RBP group(χ2=6.836,P=0.009).Multivariate Logistic regression analysis showed IVC(χ2=4.974,P=0.025),TLF injury(χ2=5.231,P=0.023),Goutallier grade of paravertebral steatosis>2(χ2=15.124,P<0.001)and proportion of bone cement(χ2=4.168,P=0.038)were independent risk factors for RBP after PVP.ROC curve of model showed AUC of original model was 0.816[OR=2.862,95%CI(0.776,0.894),P<0.001].The internal verification of model through 200 bootstrap samples showed the value of C index was 0.936,and calibration curve showed predicted probability curve was close to actual probability curve.H-L goodness of fit test results wereχ2=5.796,P=0.670.DCA analysis results showed the decision curve was above None line and All line when the threshold value ranged from 6%to 71%.Conclusion IVC,TLF combined injury,paravertebral muscle steatosis with Goutallier grade>2,and bone cement diffusion with mass type are independent risk factors for RBP after PVP.The risk prediction model for RBP after PVP established has good predictive performance and good clinical practicability.