Prediction Of Recurrence Risk And Prognostic Stratification In Patients With Pancreatic Cancer After Curative R0 Resection Using Preoperative Contrast-Enhanced CT

Purpose:To construct a model to predict the risk of recurrence in patients with pancreatic ductal adenocarcinoma(PDAC)after curative R0 resection based on preoperative contrastenhanced computed tomography(CECT)findings.Materials and Methods:From January 2014 to June 2020,126 patients with radically resected PDAC were reviewed and divided into a development cohort(n=90)and a validation cohort(n=36).In the development cohort,clinicopathological parameters and preoperative CECT findings associated with recurrence-free survival(RFS)were identified by using univariate and multivariate analyses.The nomogram was constructed using Cox proportional hazards regression models.The concordance index(C-index)and calibration curves were used to assess the discrimination and calibration of the model respectively.The prognostic performance of the prediction model was assessed using timedependent receiver operating characteristics(ROC)curve analysis.Kaplan-Meier analysis was conducted to verify the preoperative risk stratification performance of the proposed nomogram.Results:Tumor size>4 cm(HR,2.429;95%CI:1.319,4.474),adjacent organs invasion(HR,3.451;95%CI:1.941,6.135),CECT-reported lymph node metastasis(HR,2.279;95%CI:1.345,3.859),and rim enhancement(HR,3.295;95%CI:1.820,5.953)were independently associated with worse RFS(all P values were<0.05).The C-indexes of the proposed nomogram for predicting RFS in the training and validation cohorts were 0.766(95%CI:0.717,0.815)and 0.779(95%CI:0.715,0.843),respectively.Low-risk patients had significantly better RFS than high-risk patients(all P values were<0.001 in both the development and validation cohorts).Conclusion:Nomogram based on preoperative pancreatic CECT findings may predict RFS for PDAC patients after curative resection and facilitate further risk stratification and personalized neoadjuvant treatment.Purpose:The current prognostic assessment of pancreatic ductal adenocarcinoma PDAC)is mainly reliant on surgical clinicopathological data,which is not suitable for preoperative stratification.The aim of this study was to develop a preoperative prognostic stratification model for PDAC after curative resection using preoperative contrast-enhanced computed tomography(CECT)-based radiomics and carbohydrate antigen(CA19-9)levels.Materials and Methods:A total of 142 PD AC patients who underwent curative R0 resection were retrospectively enrolled and allocated randomly to training(n=99)and validation cohorts(n=47).1218 radiomics features were extracted from preoperative CECT portal venous phase images.In the training cohort,radiomics model for predicting overall survival(OS)in PDAC patients after curative resection was constructed using random survival forest(RSF),and the rad-score of patients in the training and validation cohorts were calculated.Multivariate Cox regression analysis was used to construct a preoperative clinical-radiomics model based on preoperative CA19-9 levels and rad-score.The predictive effectiveness of the model was evaluated using time-dependent receiver operating characteristics(ROC)curves and compared with the preoperative radiological model and the postoperative clinicopathological model.Decision curve analysis(DCA)was used to assess the clinical utility of the proposed model.Kaplan-Meier analysis was used to calculate the number of events,median OS and its 95%confidence interval in each subgroup.Survival curves were plotted to verify the preoperative prognostic stratification performance of the proposed model.Results:Increased rad-score(HR,1.049;95%CI:1.033,1.065;P<0.001)and preoperative CA19-9 levels>180 U/mL were the independent factors for the shorter OS.The C-indexes of the preoperative clinical-radiomics model for predicting OS in the training and validation cohorts were 0.745(95%CI:0.686,0.804)and 0.734(95%CI:0.651,0.817),respectively.ROC curves and DCA analysis depicted that the proposed preoperative model outperformed the preoperative radiological model and the postoperative clinicopathological model in predicting postsurgical outcomes of PDAC.Prognostic stratification was performed according to the median model score in the training cohort,and the difference in median OS between the high-risk and low-risk groups was statistically significant(training cohort:44.7 months(95%CI:31.9,NC)vs.17.7 months(95%CI:16.3,20.8),P<0.001;validation cohort:36.8 months(95%CI:27.7,NC)vs.16.7 months(95%CI:15.2,29.4),P<0.001).Conclusion:The preoperative clinical-radiomics model,integrating rad-score and serum CA19-9 levels,aided preoperative prognostic stratification for PD AC and could facilitate clinical decision-making.