Analysis Of Risk Factors For Lymph Node Skip Metastasis And Construction Of The Prediction Model For Papillary Thyroid Carcinoma

Background:Preventive lateral cervical lymph node dissection for papillary thyroid carcinoma(PTC)is controversial.Central cervical lymph node metastasis(CLNM)is a key factor for predicting the state of lateral lymph node metastasis(LLNM).In general,the route of cervical lymph node metastasis in PTC is roughly from the primary lesion to the central lymph node and then to the lateral lymph node.However,in clinical practice,the metastasis route of cervical lymph node in some PTC patients directly from the primary lesion to the lateral lymph node.This phenomenon is defined as "skip metastasis".Furthermore,LLNM is difficult to assess in the absence of CLNM.Methods:A total of 238 patients who received total thyroidectomy and central lymph node dissection plus ipsilateral lateral lymph node dissection were enrolled in this study.Clinicopathological features were analyzed by univariate and multivariate analysis.Based on multivariate analysis,a new prediction model for skip metastasis was established.Results:All patients were LLNM.Of the LLNM,16/144(11.1%)patients suffered skip metastasis.In univariate analysis,contralateral central lymph nodes,total central lymph nodes,numbers of LLNM,primary tumor location in the upper portion were associated with skip metastasis.Multivariate analysis showed that primary tumor location in the upper portion,contralateral central lymph nodes,numbers of LLNM were independent risk factors for skip metastasis.Based on above factors,a new model was established.Risk score=0.008+13.962*tumor location in upper portion+4.262*number of LLN+3685*number of CCLN.The area under the curve(AUC)was 0.872(95%CI,0.78-0.964).Conclusion:Tumor located in the upper portion is the only pre-operation factors for predicting skip metastasis.To improve the accuracy for prediction,number of CCL and LLN should be considered.When the score more than 16,skip metastasis probably have been existed based on this model.