腰椎间盘突出症患者发生下肢深静脉血栓的危险因素及列线图术前预测模型构建

Risk factors of lower extremity deep vein thrombosis in patients with lumbar disc herniation and construction of preoperative predictive nomogram model

目的分析腰椎间盘突出症患者发生下肢深静脉血栓的影响因素,探讨建立的列线图模型对下肢深静脉血栓的术前预测价值。方法2019年7月—2022年7月河南省洛阳正骨医院拟行手术治疗的腰椎间盘突出症患者862例,入院后均行超声检查下肢深静脉血栓发生情况。将862例患者根据收治病区分为训练集596例,验证集266例。收集训练集和验证集患者临床资料,包括性别,年龄,体质量指数,病程及红细胞计数、白细胞计数、血小板计数、三酰甘油、总胆固醇、D-二聚体水平,椎间盘退变Pfirrmann分级,骨密度,视觉模拟评分等;采用单因素logistic回归和lasso回归分析筛选训练集患者发生下肢深静脉血栓的影响因素,纳入多因素logistic回归分析模型,分析其发生下肢深静脉血栓的危险因素;根据多因素logistic回归分析结果建立预测腰椎间盘突出症患者发生下肢深静脉血栓的列线图模型;采用ROC曲线、校准曲线及决策曲线评估该模型预测训练集和验证集患者发生下肢深静脉血栓的价值,比较列线图模型与D-二聚体预测验证组患者发生下肢深静脉血栓的效能。结果862例患者中发生下肢深静脉血栓170例,其中训练集115例,验证集55例。训练集性别、年龄、体质量指数、病程、视觉模拟评分、合并症、药物服用情况、手术史、实验室检查结果、Pfirrmann分级等与验证集比较差异均无统计学意义(P>0.05)。年龄>48岁(OR=18.724,95%CI:5.386~65.080,P<0.001),D-二聚体>0.5 mg/L(OR=1.367,95%CI:1.051~1.992,P<0.001)、体质量指数>30 kg/m^(2)(OR=2.226,95%CI:1.707~3.829,P=0.004)、总胆固醇>5.2 mmol/L(OR=-1.065,95%CI:-1.649~-1.016,P<0.001)、白细胞计数>10×10^(9)/L(OR=2.134,95%CI:1.673~3.855,P=0.007)、红细胞计数<4.3×10^(12)/L(OR=-0.474,95%CI:-0.995~0.041,P=0.042)是训练集患者发生下肢深静脉血栓的危险因素。列线图模型最佳截断值为1.729时预测训练集患者发生下肢深静脉血栓的AUC为0.822(95%CI:0.784~0.861,P<0.001),灵敏度为90.1%,特异度为79.6%;列线图预测验证集患者发生下肢深静脉血栓的AUC(0.807)大于D-二聚体(0.616)(Z=-3.371,P=0.001);列线图在验证集和训练集校准曲线均与理想曲线基本接近,决策曲线均远离横纵坐标,预测腰椎间盘突出症患者发生下肢深静脉血栓的净收益均高于D-二聚体。结论年龄>48岁、D-二聚体>0.5 mg/L、体质量指数>30 kg/m^(2)、总胆固醇>5.2 mmol/L、白细胞计数>10×10^(9)/L、红细胞计数<4.3×10^(12)/L是腰椎间盘突出症患者发生下肢深静脉血栓的危险因素,构建的列线图模型术前预测腰椎间盘突出症患者下肢深静脉血栓的价值较高。

Objective To analyze the influencing factors of lower extremity deep vein thrombosis(DVT)in patients with lumbar disc herniation(LDH),and to investigate the value of construction of nomogram model to the prediction of lower extremity DVT.Methods A total of 862 patients with LDH were scheduled to perform operation in Luoyang Orthopedic-Traumatological of Henan Province from July 2019 to July 2022.All patients received ultrasound to observe lower extremity DVT after admission.These 862 patients were divided into training set(n=596)and validation set(n=266)according to the department they admitted.The clinical data were collected,including gender,age,body mass index(BMI),LDH disease course,red blood cell count,white blood cell count,platelet count,triacylglycerol,total cholesterol,D-dimer level,Pfirrmann grade of intervertebral disc degeneration,bone mineral density,and visual analogue scale(VAS)score.Univariate logistic and lasso regression analyses were done to screen the influencing factors of DVT in the training set.Multivariate logistic regression analysis model was used to analyze the risk factors of lower extremity DVT in patients with LDH.A nomogram model for predicting lower extremity DVT in patients with LDH was constructed based on the results of multivariate logistic regression analysis.The predictive values of the nomogram model in two sets was evaluated by ROC curve,calibration curve and decision curve,and the efficiencies of nomogram and D-dimer on predicting lower extremity DVT were compared in the validation set.Results In 862 LDH patients,lower extremity DVT occurred in 170 patients,including 115 patients in the training set and 55 patients in the validation set.There were no significant differences in the gender,age,body mass index,LDH course,VAS score,comorbidities,drug use,surgical history,laboratory test results and Pfirrmann grade between two sets(P>0.05).Age >48 years old(OR=18.724,95%CI:5.386-65.080,P<0.001),D-dimer >0.5 mg/L(OR=1.367,95%CI:1.051-1.992,P<0.001),body mass index >30 kg/m^(2)(OR=2.226,9594 CI:1.707-3.829,P=0.004),total cholesterol>5.2 mmol/L(OR=-1.065,95%CI:-1.649 to -1.016,P<0.001),white blood cell count 10×10^(9)(OR=2.134,95%CI:1.673-3.855,P=0.007)and red blood cell count >4.3×10^(12)/L(OR=-0.474,95%CI:-0.995 to 0.041,P=0.042)were the risk factors of lower extremity DVT in the training set.When the optimal cut-off value of the nomogram model was 1.729,the AUC for predicting lower extremity DVT in the training set was 0.822(95%CI:0.784-0.861,P<0.001),the sensitivity was 90.1%,and the specificity was 79.6%.The AUC of nomogram model for predicting lower extremity DVT(0.807)was higher than that of D-dimer(0.616)in the validation set(Z=-3.371,P=0.001).The calibration curves in the validation set and training set were basically close to the ideal curves,and the decision curves were far away from the horizontal and vertical coordinates.The net benefit of predicting lower extremity DVT in LDH patients was higher than that of D-dimer.Conclusion Age>48 years old,D-dimer>0.5 mg/L,body mass index>30 kg/m^(2),total cholesterol>5.2 mmol/L,white blood cell count>10×10^(9) and red blood cell count>4.3×10^(12)/L are the risk factors of lower extremity DVT in LDH patients,and the constructed nomogram model has a high value to the preoperative prediction of lower extremity DVT in LDH patients.