The Construction And Validation Of Predictive Model And Scoring System For Central Lymph Node Metastasis In Differentiated Thyroid Carcinoma Patients Of Children And Adolescents

Objective:Univariate and multivariate analysis was used to explore the risk factors of central lymph node metastasis（CLNM）in children and adolescents with differentiated thyroid carcinoma（DTC）.Logistic regression model and scoring system were established and validated internally and externally to better assist the clinical optimization of treatment methods.Methods:A total of 204 children and adolescents who received surgical treatment and were diagnosed as DTC by postoperative pathology were collected into this study from January 2010 to December 2020 in the second affiliated hospital of Nanchang university.The selected patients were divided into derivation group（160 patients）and validation group（44 patients）according to the admission time.All patients underwent central lymph node dissection,and they were divided into lymph node metastasis group and lymph node non-metastasis group according to the postoperative pathological results.Preoperative general data,biochemical indexes and ultrasonographic data of all patients were collected.Univariate and multivariate analyses were used to determine the independent risk factors for the occurrence of CLNM in children and adolescents with DTC,and a regression model was established.According to the regression coefficientβvalue of each variable in the model,the corresponding score value was calculated and a scoring system was established.Hosmer-Lemeshow（H-L）test and receiver operating characteristic（ROC）test were used to evaluate the regression model and scoring system.Finally,the data of validation group were substituted into the regression model and the scoring system for external verification,and U test was used to compare the difference between the regression model and the scoring system in the area under the curve（AUC）.Results:1.Multivariate analysis showed that:male（OR=2.964;95%CI:1.176～7.474;P=0.021）,microcalcification of the nodule（OR=2.800;95%CI:1.180～6.641;P=0.019）,the nodule is adjacent to the capsule（OR=2.418;95%CI:1.045～5.595;P=0.039）,nodule invades the capsule（OR=6.088;95%CI:1.944～19.070;P=0.002）,thyroglobulin（TG）>31.25ng/ml（OR=3.560;95%CI:1.606～7.890;P=0.002）are independent risk factors for CLNM in children and adolescents with DTC.2.Prediction model was:logit P=-2.828+1.087X₁+1.029X₂+0.883X_3-1+1.806X_3-2+1.270X₄(X₁=male,X₂=microcalcification of the nodule,X_3-1=the nodule is adjacent to the capsule,X_3-2=nodule invades the capsule,X₄=TG>（31.25n g/ml）.The H-L test showed that P>0.05,which shows that the model fits well.The AUC of the model is 0.814（95%CI:0.744～0.884）,which shows that the model has a good degree of differentiation.3.According to the regression coefficientβvalue of each variable in the model,the corresponding score value was calculated and a scoring system was established which ranging from 0 to 5.5 points.Among them were male（1 po int）,microcalcification of the nodule（1 point）,the nodule is adjacent to the caps ule（1 point）,nodule invades the capsule（2 points）,TG>31.25ng/ml（1.5 points）.Th e AUC of the scoring system is 0.802（95%CI:0.731～0.872）,which indicates th at the system has a good prediction accuracy.2.5 points was the best diagnosis threshold.The total score≥2.5 points was classified as high-risk group,while<2.5 points was classified as low-risk group.And the sensitivity,specificity,accuracy was 76.77%、75.41%、76.25%,respectively.4.Data of the validation group showed that:The H-L test of the model showed that P>0.05,which indicates that the model fits well.The AUC of the model is0.867（95%CI:0.756～0.978）,which indicates that the model has a good prediction accuracy.The AUC of the scoring system is 0.881（95%CI:0.756～0.978）,which indicates that the model has a good prediction accuracy.The sensitivity,specificity and accuracy of the scoring system with 2.5 points as the best diagnostic threshold was 75.00%,87.50%and 79.55%,respectively.5.The AUC comparison between logistic regression model and scoring system had no statistically significant difference in the derivation group and validation group（all P>0.05）,which indicates that the scoring system can be used to predict the risk of CLNM in children and adolescents with DTC instead of prediction model.Conclusion:1.Male,TG>31.25ng/ml,the nodule is adjacent to the capsule,nodule invades the capsule,microcalcification of the nodule are independent risk factors for CLNM in children and adolescents with DTC.2.The prediction model was:logit P=-2.828+1.087X₁+1.029X₂+0.883X_3-1+1.806X_3-2+1.270X₄(X₁=male,X₂=microcalcification of the nodule,X_3-1=the nod ule is adjacent to the capsule,X_3-2=nodule invades the capsule,X₄=TG>31.25n g/ml).A scoring system was established based on 5 variables,with a total scor e range of 0～5.5 points.Among them were male（1 point）,microcalcification of the nodule（1 point）,the nodule is adjacent to the capsule（1 point）,nodule invad es the capsule（2 points）,TG>31.25ng/ml（1.5 points）.2.5 points was the best di agnosis threshold.The total score≥2.5 points was classified as high-risk group,while<2.5 points was classified as low-risk group.3.Through internal and external verification,the prediction model and scoring system established in this study have good prediction accuracy,which is convenient for clinical application and has strong practicability.It can be used to predict the risk of CLNM in children and assist in clinical optimization of treatment.