Compare The Efficacy Of Tumor Resection In Patients With Borderline Resectable Primary Pancreatic Cancer And Develop A Related Prediction Model

Background:Pancreatic cancer is a fatal malignant tumor both in our country and around the world.It has the characteristics of "three highs and three lows",including high morbidity,high recurrence and metastasis,high mortality,low early diagnosis rate,low effective treatment efficiency,and low 5-year survival rate.Surgical treatment is the only curative method for pancreatic cancer.However,for borderline resectable primary pancreatic,whether surgical resection should be performed has always been controversial.Method:In this study,a total of 11,568 pancreatic cancer patients from 2004 to 2015 were retrospectively analyzed by using the US National Institutes of Surveillance,Epidemiology,and End Result(SEER)database,including 9048 in the surgical group and 2520 in the non-surgery group.The distribution and relative risk of each sub-variable group in the surgical group and non-surgical group in the population were preliminarily analyzed by cluster analysis sub-forest diagram,and then the factors influencing survival were identified by univariate Cox regression.The clinicopathological characteristics of the patients in the surgical group and non-surgical group were matched by propensity score matching(PSM)in 1:1.Finally,Kaplan-Meier survival curves and log-rank test were performed to compare the overall survival(OS)and Cancer-specific Survival(CSS)between the surgery group and the non-surgery group.In addition,the variables were screened by the least absolute shrinkage and selection operation method(LASSO)regression to construct a CSS prognostic model and presented in the form of a nomogram.The area receiver operating characteristic curve(AUC),Calibration curve and Decision analysis curve(DCA)were used to verify the models respectively in training cohort and validation group.Result:1.KM analysis results in stage T3N0:HR 3.37(CI 2.99-3.8)for OS,HR 3.36(CI 3.2-4.12)for CSS.KM analysis results in in stage T3N1:HR 2.25(CI 1.92-2.64),for OS,HR 2.38(CI 2.02-2.81)for CSS.KM analysis results in in stage T4N0:HR 2.16(CI 1.59-2.94),for OS,HR 2.38(CI 1.94-3.74)for CSS.KM analysis results in in stage T4N1:HR 2.06(CI 1.7-2.94)for OS,HR 2.15(CI 1.77-2.261)for CSS.The P value in each stage was less than 0.001.2.LASSO regression selected the related variables including age,marriage,tumor primary site,differentiation,size,T stage,N stage,surgery,histological type and chemotherapy as the prediction model.3.The nomogram model performed in the AUC(3 years:0.798;5 years:0.819)for the training cohort.And the AUC in validation cohort was(3 years:0.806,5 years:0.820).The nomogram has a favorable performance both in the calibration curve and the DCA curve.Conclusion:1.Surgical resection did extend survival in patients with resectable adenocarcinoma and those with borderline resectable pancreatic cancer,and the difference is statistically significant(P<0.001).However,when there is lymph node metastasis,the postoperative survival time of the patients was significantly reduced.Especially for patients who are borderline resectable,the actual benefit is very limited at stage T4N1.2.Age,marriage,tumor primary site,differentiation,size,T staging,N staging,surgery,histological type are all independent risk factors for locally middle to advanced primary pancreatic cancer.Among them,histological type has the greatest impact on the survival and prognosis of patients,while surgery,radiotherapy and chemotherapy are the protective factors for the patients.3.The establishment of the CSS nomogram for locally middle to advanced primary pancreatic cancer can well evaluate the survival prognosis of patients and provide patients with a personalized survival score scale with clinical value.