| Objective:Pancreaticoduodenectomy is currently recognized as the only treatment modality that has the potential to cure pancreatic head cancer.However,its overall survival is still poor.The purpose of this study is to analyze the factors affecting the tumor-specific survival(PCSS)of patients with pancreatic head cancer after pancreaticoduodenectomy and to develop a predictive model for individualized prognosis of patients with pancreatic head cancer after pancreaticoduodenectomy.Methods:9523 patients with complete data between 2004 and 2015 were extracted from the Surveillance,Epidemiology,and End Results(SEER)database in the United States.The extracted data were randomly assigned to the modeling cohort(n=6667)and the internal validation cohort(n=2856)according to 7:3.Sixty-four patients with pancreatic head cancer who underwent pancreaticoduodenectomy at the Affiliated Hospital of North Sichuan Medical College from January 2015 to March 2021 were collected for the external validation group.One-way COX risk regression was used to analyze the factors influencing PCSS after pancreaticoduodenectomy in patients with pancreatic head cancer,and indicators with statistical significance(P<0.05)were included in Lasso regression and cross-validation method for further screening,and the re-screened indicators were included in multi-factor COX regression analysis to determine the independent risk factors for PCSS after pancreaticoduodenectomy in patients with pancreatic head cancer.Predictive column line plot models were developed using R software on the basis of multifactorial COX regression,and internal validation of the column line plots was performed using consistency index(C-index),time-dependent ROC curves,and calibration curves.Clinical decision curves(DCA)were used to evaluate the clinical application value of the column line chart.Risk stratification was performed according to the prognostic column line plot model,and Kaplan-Meier curves were drawn to compare survival outcomes between risk subgroups.Results:Independent factors influencing PCSS after pancreaticoduodenec-tomy for pancreatic head cancer included age,gender,tumor pathological type,degree of tumor differentiation,SEER stage,T stage,M stage,radiotherapy,chemotherapy,number of cleared lymph nodes,number of positive lymph nodes,and tumor size.The C index of the modeling cohort was 0.686,and its area under the ROC curve at 1,3,and 5 years was 0.754(95%CI=0.741-0.767),0.735(95%CI=0.720-0.749),and 0.739(95%CI=0.719-0.759),respectively.The internal validation cohort C index was:0.692,and its area under the 1-year,3-year,and 5-year ROC curves were 0.750(95%CI=0.730-0.770),0.751(95%CI=0.729-0.774),and 0.776(95%CI=0.748-0.805),respectively,and the established column line graph model was validated by C index,calibration curve,and ROC curves to verify its good predictive effect on PCSS at 1,3,and 5 years,and clinical decision curve analysis suggested that patients could benefit from using this model.The area under the ROC curve at 1,3,and 5 years for the external validation cohort was 0.899(95%CI=0.824-0.975),0.833(95%CI=0.720-0.945),and 0.870(95%CI=0.762-0.978),respectively;risk stratification based on the column line graph model,KM survival analysis showed that the modeling cohort,internal validation cohort and external validation cohort all had significantly better prognosis in the low-risk group than in the high-risk group(P<0.05).Conclusion:1.Age,gender,tumor pathological type,degree of tumor differentiation,SEER stage,T stage,M stage,radiotherapy,chemotherapy,number of cleared lymph nodes,number of positive lymph nodes and tumor size were independent influencing factors of PCSS after pancreaticoduodenectomy for pancreatic head cancer2.The tumor-specific survival model established for patients with pancreatic head cancer undergoing pancreaticoduodenal surgery can be used to predict PCSS at 1,3,and 5 years,and accurate risk stratification can be performed based on the model,which is used in domestic clinical work to guide clinical decision making. |
PDF Full Text Request