A Risk Prediction Model for Distant Metastasis of Follicular Thyroid Cancer Based on SEER Database
Objective This study aims to delineate the prediction of distant metastasis in follicular thyroid cancer.Methods A total of 1,275 patients diagnosed as follicular thyroid cancer in the SEER database from 2004 to 2015 were selected for the study.Various predictive factors were evaluated.Furthermore,the optimal cut-off of age and the duration from diagnosis to treatment (TDT)were obtained by ROC analysis.Statistically significant variables in univariate analysis was preliminarily screened using Lasso regression and ten-fold cross validation.A predictive model was then constructed using multivariate logistic regression.The diagnostic effectiveness of the model using ROC curves and AUC was finally assessed.Results 4.6％ of FTC (59 cases)experienced distant metastasis in our study.Among them,10 variables in univariate analysis reached statistical significance,including gender,age,whether there was micro invasive cancer,T stage,N stage,TDT,LNR,tumor size,primary lesion infiltration and multifocal.Lasso regression and ten-fold cross validation identified 6 variables and the stepwise forward backward method was subsequently used.Age,T stage,TDT,and LNR were selected for the final model.AUCs of this model was 0.8815,with a 95％ CI of 0.8353-0.9276.Conclusion The predictive model constructed with age,T-stage,TDT,and LNR variables is highly predictive which might favor clinical diagnosis and treatment for FTC.The prediction of distant metastasis of FTC by LNR and TDT has not been reported in previous studies,which is a highlight of this article.
Follicular thyroid cancerSEER databasePrediction model
万方数据
维普
