胸段食管鳞癌喉返神经旁淋巴结转移风险预测模型的构建

Construction of a risk prediction model for recurrent laryngeal nerve lymph node metastasis in thoracic esophageal squamous cell carcinoma

目的探讨胸段食管鳞癌患者喉返神经旁淋巴结转移的影响因素,建立预测喉返神经旁淋巴结转移的列线图模型。方法449例胸段食管鳞癌患者均行食管癌根治术,依据喉返神经旁淋巴结转移情况分为转移组和无转移组。比较2组性别、年龄、体质量指数、吸烟史、饮酒史;比较2组血清谷丙转氨酶/谷草转氨酶比值,血清白蛋白、珠蛋白水平,肿瘤部位,病理分级,T分期,肿瘤直径及淋巴脉管侵犯、神经侵犯、食管旁淋巴结转移、隆突下淋巴结转移、腹部淋巴结转移情况;Lasso回归筛选胸段食管鳞癌患者喉返神经旁淋巴结转移的影响因素,纳入多因素logistic回归分析模型,分析胸段食管鳞癌患者喉返神经旁淋巴结转移的影响因素;根据多因素logistic回归分析结果,建立预测胸段食管鳞癌患者喉返神经旁淋巴结转移的列线图模型,采用ROC曲线、校准曲线及决策曲线评估该模型的预测价值;根据喉返神经旁淋巴结转移概率将食管鳞癌患者分为不同风险亚组,构建决策树模型。结果449例患者中83例有喉返神经旁淋巴结转移,喉返神经旁淋巴结转移率为18.5%。转移组饮酒(88.0%)、胸上段肿瘤(28.9%)、病理分级G;级(33.8%)、T分期T;期(53.0%)、淋巴脉管侵犯(33.7%)、隆突下淋巴结转移(16.9%)、腹部淋巴结转移(31.3%)比率均高于无转移组(73.8%、13.7%、20.8%、37.7%、17.2%、4.9%、16.9%)(P<0.05),年龄,体质量指数,谷丙转氨酶/谷草转氨酶比值,血清白蛋白、珠蛋白水平,肿瘤直径及男性、吸烟、神经侵犯、食管旁淋巴结转移比率与无转移组比较差异均无统计学意义(P>0.05)。男性(OR=1.974,95%CI:1.076~3.620,P=0.028)、饮酒(OR=2.918,95%CI:1.363~6.248,P=0.006)、胸上段肿瘤(OR=3.663,95%CI:1.878~7.145,P<0.001)、T分期T;期(OR=3.296,95%CI:1.412~7.696,P=0.006)、淋巴脉管侵犯(OR=1.888,95%CI:1.047~3.406,P=0.035)、隆突下淋巴结转移(OR=3.325,95%CI:1.462~7.560,P=0.004)是胸段食管鳞癌患者喉返神经旁淋巴结转移的独立危险因素。ROC曲线显示,列线图模型最佳截断值为0.412时,预测胸段食管鳞癌患者喉返神经旁淋巴结转移的AUC为0.754(95%CI:0.697~0.811,P<0.001),灵敏度为66.3%,特异度为74.9%;校准曲线显示,该模型校准曲线与理想曲线基本接近;决策曲线显示,该模型曲线远离横纵坐标,预测价值较高。低风险组、中低风险组、中风险组、中高风险组、高风险组喉返神经旁淋巴结转移概率分别为6.6%、14.0%、27.0%、39.0%、43.0%。结论男性、饮酒、胸上段肿瘤、T;期、有淋巴脉管侵犯和隆突下淋巴结转移是胸段食管鳞癌患者喉返神经旁淋巴结转移的独立危险因素,构建的列线图模型对胸段食管鳞癌患者喉返神经旁淋巴结转移预测价值较高。

Objective To investigate the influencing factors of recurrent laryngeal nerve lymph node metastasis patients with thoracic esophageal squamous cell carcinoma(ESCC),and to establish a nomogram model for predicting recurrent laryngeal nerve lymph node metastasis in thoracic ESCC.Methods Totally 449 patients with thoracic ESCC underwent radical esophagectomy.According to the presence or absence of recurrent laryngeal nervelymph node metastasis,449 patients were divided into metastasis group and non-metastasis group.The gender,age,body mass index,smoking history,drinking history,serum glutamic-pyruvic transaminase/glutamic-oxaloacetic transaminase,serum albumin,serum globin,tumor location,pathological grade,T stage,tumor diameter,lymphatic vascular invasion,nerve invasion,para-esophageal lymph node metastasis,subcarinal lymph node metastasis and abdominal lymph node metastasis were recorded and compared between two groups.Lasso regression was used to screen the influencing factors of recurrent laryngeal nerve lymph node metastasis in patients with thoracic ESCC.Multivariate logistic regression model was used to analyze the influencing factors of recurrent laryngeal nerve lymph node metastasis.Based on the results of multivariate logistic regression analysis,a nomogram model for predicting recurrent laryngeal nerve lymph node metastasis was constructed.ROC curve,calibration curve and decision curve were used to evaluate the predictive value of the model.The patients with ESCC were divided into different risk subgroups according to the probability of recurrent laryngeal nerve lymph node metastasis,and a decision tree model was constructed.Results In 449patients,83developed erecurrent laryngeal nerve lymph node metastasis(18.5%).The percentages of patients with drinking,upper thoracic tumor,G3,T3stage,lymphatic vascular invasion,subcarinal lymph node metastasis and abdominal lymph node metastasis were higher in metastasis group(88.0%,28.9%,33.8%,53.0%,33.7%,16.9%,31.3%)than those in non-metastasis group(73.8%,13.7%,20.8%,37.7%,17.2%,4.9%,16.9%)(P<0.05),and there were no significant differences in the age,body mass index,glutamic-pyruvic transaminase/glutamic-oxaloacetic transaminase ratio,serum albumin,serum globin,tumor diameter,percentage of male patients,percentages of patients with smoking,nerve invasion and para-esophageal lymph node metastasis between two groups(P>0.05).The male patients(OR=1.974,95%CI:1.076-3.620,P=0.028),drinking(OR=2.918,95%CI:1.363-6.248,P=0.006),upper thoracic tumor(OR=3.663,95%CI:1.878-7.145,P<0.001),T3stage(OR=3.296,95%CI:1.412-7.696,P=0.006),lymphatic vascular invasion(OR=1.888,95%CI:1.047-3.406,P=0.035)and subcarinal lymph node metastasis(OR=3.325,95%CI:1.462-7.560,P=0.004)were the independent risk factors of recurrent laryngeal nerve lymph node metastasis.ROC curve showed that when the optimal cut-off value of the model was 0.412,the AUCfor predicting recurrent laryngeal nerve lymph node metastasis was 0.754(95%CI:0.697-0.811,P<0.001),with a sensitivity of 66.3%,and a specificity of 74.9%.The calibration curve showed that the nomogram model curve was basically identical to the ideal predicting curve.The decision curve showed the curve of model was far away from ordinate,indicating a high predictive value.The metastatic probability of recurrent laryngeal nerve lymph nodes of low-risk group,middle-low-risk group,middle-risk group,middle-high-risk group and high-risk group was 6.6%,14.0%,27.0%,39.0%and 43.0%,respectively.Conclusion Male,drinking,upper thoracic tumor,T3stage,lymphatic vascular invasion and subcarinal lymph node metastasis are the independent risk factors of recurrent laryngeal nerve lymph node metastasis in patients with thoracic ESCC,and the constructed nomogram model is of high value to the prediction of recurrent laryngeal nerve lymph node metastasis.