| Objective: Through the bibliometric analysis of the predictive model of pancreatic fistula after pancreatoduodenectomy,the visualization showed the current research trends and hotspots in this field,providing some help for the future research direction.The risk factors leading to pancreatic fistula were analyzed,and the clinical prediction model of pancreatic fistula was established to further verify the stability and reliability of the model.By intuitively assessing the risk of pancreatic fistula and identifying high-risk patients,we could find pancreatic fistula early and reduce the incidence of pancreatic fistula.Methods: This study was divided into two parts.Part I: This study was based on the Web of Science database and determined the relevant retrieval terms to retrieve the literature on pancreatic fistula from the beginning of the study to 2022.And the exported literatures were saved in TXT file format.Cite Space and VOSvivewer were used to visually analyze the collaboration networks and keyword clusters among countries,institutions,authors and journals.Part II: The risk factors of pancreatic fistula were explored based on the results of the first step and the actual situation of our center.We retrospectively collected 213 patients who underwent PD surgery for malignant tumors in our hospital from January 2015 to December 2019 as the prediction group;From January 2020 to December 2021,107 patients undergoing PD surgery for malignant tumors in our hospital were taken as the validation group.The preoperative,intraoperative and postoperative data were analyzed,and the risk factors were screened by single factor analysis.The consistency of preoperative imaging data and intraoperative actual data was compared.The possible risk factors were included in the multivariate logistic regression analysis to establish the independent risk factors of pancreatic fistula.This study used R language software to draw the clinical nomogram prediction model,calculated the area under the curve to express the model differentiation,and further verified the model calibration curve and clinical practical value.P<0.05 had statistical difference.Results: Part Ⅰ: In the bibliometric analysis of the predictive model of pancreatic fistula,76 articles were included.699 authors from 199 institutions in 21 countries have been published in 37 journals.The largest number of documents were issued in 2021.The countries with the largest number of publications were China(n=21)and the United States(n=16),which were leaders in this field.The University of Pennsylvania(n=6),Mayo Medical Center(n=5),University of Alabama(n=5),Seoul National University(n=4)and Sungkyunkwan University(n=4)were influential institutions in this field.BASSI C,CALLERY M,VOLLMER CJ were key authors in this field.Annals of Surveillance(n=8),HPB(n=6),Journal of Gastrointestinal Surveillance(n=4),Surveillance(n=4),World Journal of Gastroenterology(n=4)were the most prolific organizations in this field.Surgery,replication,pancreatic formula,and pancreaticcoduonectomy were the key keyword link networks in this field,and the keyword clustering presented 8 clusters in total.The first five risk factors of pancreatic fistula mainly include pancreatic duct diameter(n=39),BMI(n=30),pancreatic texture(n=30),pathology(n=14),and amylase level on the first day after operation(n=9).Part II: In this study,there were 213 cases in the prediction group and 107 cases in the verification group,44 cases and 18 cases of pancreatic fistula,respectively.The risk of pancreatic fistula was screened by univariate analysis in the prediction group.It was found that the pancreatic neck-main pancreatic duct diameter(P<0.001),BMI(P=0.001),pancreatic texture(P<0.001),LMR(P=0.001),and pancreatic neck parenchyma-minor axis(P<0.001)were possible risk factors.The consistency analysis of preoperative image measurement and intraoperative actual measurement showed that the ICC value was more than 0.75,P<0.001,indicating that there was a high consistency between preoperative image measurement and intraoperative actual measurement.The inclusion of risk factors in multivariate logistic analysis showed that LMR(P=0.005,OR=0.616,95%CI: 0.439-0.864),BMI(P=0.003,OR=1.245,95%CI: 1.078-1.438),pancreatic texture(P<0.001,OR=-1.465,95%CI: 0.103-0520),pancreatic neck-main pancreatic duct diameter(P=0.001,OR=-0.831,95%CI: 0.266-0.714),and pancreatic neck parenchyma-minor axis(P=0.033,OR=0.163,95%CI: 1.013-1.367)were the independent risk factor for pancreatic fistula.According to the five risk factors,the clinical nomogram prediction model was constructed,which showed that the prediction efficiency was good and had a moderate differentiation(AUC=0.869).The calibration curve was drawn,and the calibration curve points fluctuated around the standard curve,indicating that the calibration of the prediction model was good.At the same time,the ROC curve of the validation group(AUC=0.881)showed that the model had good prediction efficiency.The further drawing of clinical decision curve indicated that the model had good clinical application value.Conclusion:(1)Using the method of bibliometric analysis,this paper analyzed and explained the number of documents in the field of pancreatic fistula model,including the cooperation network and co-occurrence network of national institutions,journals,authors and keywords.It comprehensively and intuitively explained the past research directions and possible research hotspots and development trends in the future,which might provide some help for future clinical diagnosis and treatment activities and scientific research.(2)pancreatic neck-main pancreatic duct diameter,pancreatic neck parenchyma-minor axis,pancreatic texture,BMI and LMR were independent risk factors for pancreatic fistula after pancreaticoduodenectomy.(3)The construction of clinical nomogram prediction model based on five types of risk factors showed that the prediction effect was good,which was helpful to identify high-risk patients with pancreatic fistula in clinical early stage. |
PDF Full Text Request