A Predictive Nomogram For Early Death Of Metastatic Esophagus Cancer: A Retrospective Study In The SEER Database

ObjectivesTo identify relevant risk factors and to construct a predictive nomogram and a Web-based probability calculator for early death in Metastatic Esophagus Cancer(m EC).Materials and methodsPatients with initially diagnosed m EC,who had been diagnosed between 2010 and 2016 in the Surveillance,Epidemiology,and End Results datasets,were eligible for this study.(including Race,Age,Sex,Marital status,Primary Site,Insurance,Grade,Pathological type,T stage,N stage,M stage,Bone metastasis,Liver metastasis,Brain metastasis,Lung metastasis,Tumor size,Chemotherapy,Radiotherapy,Surgery and follow-up data).After screening the data,5984 patients were finally included.In this study,univariate Logistic regression model was used to analyze the associated factors of early death(time of death≤3months),followed by lasso regression screening and multivariate Logistic regression analysis.A predictive nomogram and a web-based probability calculator were developed by using the factors contained in the final model.and then validated by receiver operating characteristics(ROCs)curve and calibration plot.ResultOf the 5984 patients,There were 2,434(40.6%)early deaths(cancer-specific early deaths accounted for 94.6% and non-cancer early deaths accounted for 5.4 %).Seventeen independent variables were recognized for independent risk variables of early death about metastatic esophageal cancer patients by using univariate logistic regression models.The factors of P < 0.05 were analyzed by Lasso Logistic regression model,and 10variables(Marital status,Insurance,Grade,Bone metastasis,Liver metastasis,Lung metastasis,Tumor size,Chemotherapy,Radiotherapy,Surgery)with the least number of variables and excellent performance were obtained.The results of multivariate Logistic regression analysis showed that 9 variables(except Insurance)were independent prognostic factors of early death of metastatic esophageal cancer(P<0.05).By comprising these variables,a predictive Nomogram and a web-based probability calculator were constructed.Then,through the verification of the model,the C index is 0.829,which indicates that the model exhibited good discrimination.Bootstrap validation suggested prediction curves(solid line)of the calibration curve were all closely approximated at the 45° line,indicating a precise prediction of these models.ConclusionIn this study,We constructed and validated a novel nomogram and a Web-based probability calculator was constructed and validated by Surveillance,Epidemiology and End Results.The validation results show that the model has good distinction and fit.The mode can help clinicians to better evaluate the risk of early death of metastatic esophageal cancer.These findings can be applied to early screening and to individualized treatment regimens for Patients with high-risk factors.