Prediction Study Of Papillary Thyroid Macrocarcinoma With Cervical Lymph Node Metastasis Based On ACR TI-RADS

Objective:To evaluate the application effect of papillary thyroid macrocarcinoma（PTMC）and predict the risk of cervical lymph node metastasis of PTMC under the guideline of Thyroid Imaging Report and Data System（TI-RADS）by American College of Radiology（ACR）in 2017 and screen the risk factors.And to construct a risk scoring system for predicting cervical lymph node metastasis to provide an objective basis for ultrasonography in detecting cervical metastatic lymph nodes.Methods:A total of 270 PTMC patients diagnosed by pathological results from January 2019 to June 2020 were continuously enrolled by the cluster sampling method.The first 2/3 were used as development cohort and the last 1/3 were used as validation cohort.The PTMC were classified and scored according to the ACR TI-RADS（TIRADS1、TIRADS 2、TIRADS 3、TIRADS 4、TIRADS 5/TR1、TR2、TR3、TR4、TR5）,and the score composition table and histogram were used to describe the distribution of PTMC’s TI-RADS score.A univariate logistic analysis was performed on all variables,and variables with P<0.05 were selected for multicollinearity test.Multivariate logistic regression analysis was performed on variables without collinearity,and the forward:LR method was used to include each variable.The dummy variables were assigned to categorical variables.Variables with P<0.05 were independent risk factors.According to the results of multivariate analysis,theβvalue of the variable were obtained to establish a logistic prediction model.The regression coefficientβi was obtained of each variable according to the logistic regression model.The reference value W_ijof each group of variables was specified and one of the categorical variables was selected,setting the reference value to 0,and assigning the other variable to 1.The basic risk reference value W_iREFof each variable was determined.The reference value 0 was set as the basic risk reference value in the categorical variables.The distance D was calculated between the grouping of each variable and the basic risk reference value,D=(W_ij-W_iREF)×β,and setting the constant B to represent one point in the scoring tool.In this study,B was used to reflect the risk of lymph node metastasis when wider than tall,B=1×β_shape;Points_ij=D/B=(W_ij-W_{i REF})×β_i/B for each category of variables.The total score was calculated and the risk prediction probability value was calculated corresponding to each score according to the multivariate logistic regression equation.According to the risk prediction probability corresponding to each score,a critical score was determined,which represent a greater possibility of metastasis.And the sensitivity,specificity,accuracy,and positive predictive value and negative predictive value of the scoring system were calculated in the modeling sample and the verification sample based on the score.Results:Almost all PTMC were distributed in TR5,showing a normal distribution,with a score of 9-10,accounting for 51.6%of all nodule scores.The results of univariate analysis showed that ACR TI-RADS score was not a risk factor for lymph node metastasis.The results of multivariate analysis showed that male,wider than tall,marginal lobulation or irregularity,and extra-thyroidal extension were risk factors for cervical lymph node metastasis.For the risk of lymph node metastasis,male was 4.566times than that of female,wider than tall was 2.331 times than that of taller than wide,margin lobulation or irregularity was 3.024 times than that of smooth or ill-defined margin,and extra-thyroidal extension was 8.117 times than that of smooth or ill-defined margin.According to the results of multivariate analysis,a scoring system was constructed,assigning 2 points for male,1 point for wider than tall,1 point for margin lobulation or irregularity,2 points for ETE,with a total score of 0-5 points corresponding to the risk of cervical lymph node metastasis,which were 10%,15%,30%,55%,70%,85%respectively.A score of 3 points or more was considered to have a greater risk of metastasis.The sensitivity,specificity,accuracy,positive predictive value,and negative predictive value of the prediction effect in the development cohort were 57.33%,78.10%,69.44%,65.15%,and 71.93%respectively.The sensitivity,specificity,accuracy,positive predictive value,and negative predictive value of the prediction effect in the validation cohort were 46.34%,67.35%,57.78%,54.29%and60.00%respectively.Conclusion:ACR TI-RADS score and classification could be used for the diagnosis of PTMC.PTMC was almost all distributed in TR5,with 9-10 divided into main ones.Based on the ACR TI-RADS classification,male,wider than tall,lobulated or irregular margins,and ETE were risk factors for PTMC with cervical lymph node metastasis.And the total score greater than or equal to 3 points was considered to have a greater risk of metastasis.