Development And Validation Of A Nutritional Nomogram For Prediction Of Prognosis In Pancreatic Cancer

Objective:To identify and verify the risk factors for predicting the prognosis of pancreatic cancer,to establish a postoperative survival prediction model through preoperative indicators,and to find ways to improve the prognosis of patients with pancreatic cancer in order to improve the postoperative survival time of patients.Methods:A retrospective analysis of 129 patients with pancreatic ductal adenocarcinoma who underwent radical surgery from April 2012 to July 2017 in the First Affiliated Hospital of Soochow University was analyzed.Twenty-six patients with pancreatic ductal adenocarcinoma undergoing radical surgery were collected from the Second Affiliated Hospital of Soochow University using the same standards and methods as the external validation group.All patients were divided into a training group of 95 cases,an internal validation group of 34 cases,and external validation group of 26 cases.The data were statistically processed with SPSS22.0 software and R software version 3.6.1.In order to determine the best cut-off values for lymphocyte count and PNI,the receiver operating curve analysis were applied.Kaplan-Meier method and log-rank test were used for survival analysis and survival curve drawing.The clinical characteristics of patients were analyzed by univariate Cox regression,and clinically significant factors(P<0.1)were selected for multivariate Cox regression to identify independent prognostic factors.The LASSO regression was used to analyze the clinical data of patients to find the factors that affect the prognosis.According to Cox multivariate analysis and LASSO regression results,pre-operative related indicators were selected to draw a nomogram.In order to verify the predictive ability of the nomogram compared to traditional nutrition indicators(PNI,CONUT),the consistency index(C-index),calibration curve,net reclassification index(NRI),integrated discrimination improvement(IDI),and the decision curve analysis(DCA)were performed in four groups.Results:Univariate Cox regression suggested that age,BMI,tumor diameter,albumin,lymphocyte count,diabetes,CA19-9,chemotherapy,pathological stage and differentiation may be influential factors for survival of patients with pancreatic cancer.The LASSO regression suggested that albumin,lymphocyte count,CA19-9,chemotherapy,diabetes may be the prognostic factors of patients.The results of multivariate Cox regression suggested that albumin,lymphocyte count,CA19-9,chemotherapy and TNM stage are factors influencing the survival of patients with pancreatic cancer.Combined with multivariate analysis and LASSO regression results,we selected four preoperative indicators including albumin,lymphocyte count,CA19-9,and diabetes to construct a nomogram for the prognosis of pancreatic cancer.The C-index in the training group,internal validation,external validation and entire validation were 0.777,0.769,0.759,0.774,which were better than the CONUT and PNI scores in the same group.The time-dependent ROC curves of data in different groups were analyzed,and the AUC of nomograms were 0.861,0.845,0.773,and 0.814,which were all higher than the CONUT and PNI in the same group.In the training group,we calculated the NRI and IDI between nomogram and CONUT and PNI,the NRI was 0.4,0.31,respectively;the IDI was 0.113,0.097;in the internal validation,the NRI was 0.536,0.343,and the IDI was 0.271,0.079;in the external validation,NRI was 0.394,-0.012,IDI was 0.094,0.058;in the entire validation,NRI was 0.297,0.149,and IDI was 0.176,0.054.The DCA curve shows that the nomogram model has higher net benefits than CONUT and PNI.Conclusion:Combined with Cox multivariate analysis and LASSO regression results,the four preoperative indicators of albumin,lymphocyte count,CA19-9,and diabetes were related factors affecting the prognosis of patients with pancreatic cancer.Based on this,a nomogram model of the prognosis of patients with pancreatic cancer was established,and the model's ability to predict the prognosis of patients was verified to be superior to the traditional nutrition index CONUT score and PNI score.