Development And Validation Of Scoring System For Predicting Recurrent Laryngeal Nerve Lymph Node Metastases In Clinical Stage T1-3N0M0 Esophageal Squamous Cell Carcinoma

Objectives: Recurrent laryngeal nerve lymph node(RLN LN) dissection isdifficult for esophageal squamous cell carcinoma(ESCC) and usually causesrecurrent laryngeal nerve injury. This study aims to develop a simple scoringsystem to predict RLN LN metastases in clinical stage T1-3N0M0 ESCC andexplore indications for selective RLN LN dissection.Methods: Patients with clinical stage T1-3N0M0 ESCC at the Fujian MedicalUniversity Union Hospital from January 2010 to December 2014 were analyzed.Patients were divided into modeling group and validation group. Patient andpathologic characteristics were compared between patients in modeling group withRLN LN metastases and those without. Univariate and multivariate analyses wereperformed to establish a scoring system that estimates risk for RLN LN metastases.The scoring system is developing by weighing beta coefficient of the final model.The final model performance is evaluated by the area under the ROC curve ofmodel group and validation group. Patients were divided into low-risk,moderate-risk and high-risk groups according to the scoring system and theindications for RLN LN dissection in various groups were validated by RLN LNmetastases rate, survival rate, QOL and complications.Results: A total of 218 cases selected from 974 ESCC patients were includedfor analysis and randomly divided into 9 groups including 169 patients(7 groups)in modelling group and 49 patients(2 groups) in validation group. The rate of RLNLN metastases in modelling group was 31.3%(53/169). Tumor location, depth oftumor invasion and lymphovascular invasion(LVI) were independent predictors ofRLN LN metastases. A weighted scoring system was devised from the finalmultivariate model and included location(+2 for upper thoracic, +1 formid-thoracic), depth of invasion(+2 for T3, +1 for T2), and LVI(+1 if present).The area under the ROC curve of model group and validation group were 0.756 and 0.791 respectively. The total number of points estimated the probability ofRLN LN metastases(low-risk [0-1 point], 2.5%; moderate-risk [2-3 points], 34.3%;and high-risk [4-5 points], 62.9%). Low-risk group had more complications, worseQOL and similar survival rate when compared with control group.Conclusions: We developed a simple scoring system that accurately estimatesthe risk of RLN LN metastases in patients with clinical stage T1-3N0M0 ESCC.RLN LN dissection is not recommended for patients in the low-risk group withlower thoracic T1-2N0M0 or mid-thoracic T1N0M0 ESCC without LVI.