胸段食管癌喉返神经旁淋巴结转移的风险模型构建及预测能力的初步验证

Establishment of a model of paralaryngeal lymph node metastasis in thoracic esophageal cancer and validation of its predictive ability

目的:探讨影响胸段食管癌喉返神经旁淋巴结转移的危险因素,构建模型并检验其预测效能。方法:收集2014年1月至2017年1月我院收治的89例患者为造模组,根据术后病理结果,将患者分为转移组与未转移组,分析可能影响喉返神经旁淋巴结转移发生的相关因素。根据危险因素权重,构建风险模型。同时,连续性收集2017年2月至2018年12月的47例患者资料,验证模型的预测效能。结果:Logistic回归分析发现,肿瘤位置、分化程度、浸润深度、淋巴结SUV值、CT下淋巴结征象是影响喉返神经旁淋巴结转移发生的独立性危险因素(P<0.05)。根据Logistic回归分析,可得回归方程:P预测=-0.754-1.347X1-1.361X 2+1.282X3+3.380X4+1.912X 5,X1:肿瘤位置(1=上段,2=中段,3=下段)、X2:分化程度(1=低分化,2=中分化,3=高分化)、X3:浸润深度(1=T1,2=T2,3=T3,4=T4)、X4:淋巴结SUV值(1=SUVmax≥2.5,2=SUV max<2.5)、X 5:CT下淋巴结征象(1=阴性,2=阳性)。ROC曲线的Youden指数最大值为0.710,截断值为0.662,敏感性为70.9%,特异性为99.1%,曲线下面积为0.918。将验证组患者各因素带入预测模型,检验该模型预测效能,结果发现,ROC曲线下面积为0.79,Hosmer-Lemeshow拟合优度检验显示,χ^2=1.52,P=0.93。模型拟合效度好,预测价值高。结论:影响胸段食管癌发生喉返神经旁淋巴结转移的高危因素多,临床应进行及时有效评估。本研究构建的预测模型有较好的评估效能,有一定的临床应用价值。

Objective:To explore the risk factors of paralaryngeal lymph node metastasis in thoracic esophageal cancer.Methods:89 patients admitted to our hospital were collected as the modeling group.Patients were divided into metastasis group and non-metastasis group according to pathological.Risk factors affecting the occurrence of lymph node metastasis of recurrent laryngeal nerve were analyzed and risk models were established.At the same time,data of 47 patients were collected to verify the predictive effectiveness of the model.Results:Tumor location,differentiation degree,depth of invasion,lymph node SUV value and CT lymph node signs were independent risk factors affecting the occurrence of lymph node metastasis(P<0.05).The regression equation was obtained:P prediction=-0.754-1.347X 1-1.361X 2+1.282X 3+3.380X 4+1.912X 5,X 1:Tumor location,X 2:Differentiation degree,X 3:Invasion depth,X 4:Lymph node SUV value,X 5:CT signs of lymph nodes.Youden index of ROC curve was 0.710.Sensitivity was 70.9%.Specificity was 99.1%,and AUC was 0.918.The factors of validation group were brought into the prediction model to test the prediction efficiency of the model.The results showed that the area under ROC curve was 0.79.Hosmer-Lemeshow goodness-of-fit test showedχ^2=1.52,P=0.93.The model had good fitting validity and high predictive value.Conclusion:The predictive model constructed in this study has better evaluation efficiency and has certain clinical application value.