The Diagnosis Of Pancreatic Cancer Based On The Score Of Conventional Clinical Index

Objective:Pancreatic cancer(PC)is highly malignant and its morbidity and mortality are on the rise worldwide.Most patients are diagnosed at an advanced stage with a poor prognosis.The clinical diagnosis of PC mainly relies on CA19-9 and imaging,but it is sometimes difficult to differentiate it from benign pancreatic tumors and chronic pancreatitis.In order to provide new simple and effective means for the diagnosis of PC,in this study,we retrospectively collected the conventional clinical indicators of patients with PC,benign pancreatic tumors and chronic pancreatitis,systematically analyzed their diagnostic value for PC,established a diagnostic model and a scoring system for PC based on the conventional clinical indicators and their derivative,and evaluated their diagnostic value.Methods:Patients with PC(case group),benign pancreatic tumors(pancreatic cystaden-oma,pancreatic intraductal papillary mucinous tumor)and chronic pancreatitis(control group)who were hospitalized at the First Affiliated Hospital of Nanchang University from January 2012 to October 2021 were collected retrospectively.Medical records were reviewed to collect clinical information,laboratory tests and other findings before treatment after admission to the hospital.Pancreatic cancer and benign pancreatic tumors were diagnosed based on pathological findings,and chronic pancreatitis was diagnosed according to the diagnostic criteria in the 2018 Guangzhou Guidelines for the Diagnosis and Treatment of Chronic Pancreatitis.Patients with the presence of co-morbidities or who had received relevant treatment that might affect the laboratory findings were excluded.In order to more effectively evaluate the diagnostic value of clinical indicators for PC,the continuous variables were subjected to "optimal discretization",natural logarithm transformation,and grouping according to clinical normal reference values to obtain new categorical variables with suffix "-B",prefix "Ln-" and suffix "-N".Statistical differences in each clinical indicator were compared between the groups.The diagnostic value of each index for pancreatic cancer was evaluated using subject operating characteristic(ROC)curve analysis and with area under the curve(AUROC).Multivariate logistic stepwise regression analysis was used to screen the indicators with independent diagnostic value for PC.The indicators with independent diagnostic value or high diagnostic value were selected to jointly establish a pancreatic cancer diagnostic model,and AUROC was used to evaluate the diagnostic value of the model for PC and to calculate the diagnostic efficacy of the model(sensitivity,specificity,accuracy,positive and negative predictive values,positive and negative likelihood ratios).Calibration curves and decision curves of the diagnostic model were plotted to analyze the stability and clinical usefulness.To simplify the relevant parameters of the diagnostic model,establish a simple and practical PC diagnostic score and evaluate its diagnostic efficacy.Results:The study included a total of 240 patients.The case group is 126 patients,Including 70 males and 56 females with an age of 60.28±10.77 year.There were 55 males and 59 females in the control group with an age of 49.69±14.62 year.There was significant difference in age between the two groups(P<0.001).In this study,the lesion site of pancreatic cancer was commonly found in the head and neck of the pancreas,whereas the majority of benign pancreatic tumors were located in the body and tail,having a significant difference between the two(P <0.003).There were 67.3% of benign pancreatic tumors were Plasmacytoid cystadenoma and 95.2% of pancreatic adenocarcinomas were adenocarcinomas.A total of 67 clinical indexes were analyzed,30 of which were significantly different between the two groups(P < 0.05～0.001),including age and the following laboratory indices: HGB,PLT,MPV,PDW,LYM,LYMP,NEUTP,TBIL,DBIL,ALT,AST,GGT,ALP,UA,GLU,GSP,LDH,TC,INR,Fbg,TT,D-2D,CEA,CA19-9,CA125;and derived indices: NLR,LMR,AFR,PFR.The ROC curve found that 23 individual indicators of diagnostic PC with AUROC≥0.6,including 7 indicators with AUROC>0.7,namely Fbg,D-2D,CEA,CA125,CA19-9,AFR and PFR,and CA19-9 being the largest(AUROC= 0.826).Multivariate logistic regression analysis of similar indicators found 13 indicators with independent diagnostic value(P<0.001),namely WBC,LYM-B,Ln-PDW,GGT-B,Ln-DBIL,LDL-C,Ln-CK-MB,Fbg-B,Ln-D-2D,CA125-B,CA19-9-B,PLR,PFR.Two hundred and forty patients were randomly divided into training group(n=138)and control group(n=102)on a 3:2 basis.Multifactorial logistic forward stepwise regression modeling was performed with data from the training group.Four indicators(age,CA19-9,PLR,PFR)entered the diagnostic model.The model had high diagnostic efficacy in both groups,with accuracy of 84.8% or better,better than CA19-9.Calibration curve results showed good stability of the model and decision curve analysis showed that the net benefit of the model was higher than that of CA19-9,which has good clinical utility.The relevant parameters of the model were simplified and a ACPP score was constructed: ACPP score = 0.65*Age + 0.12*CA19-9 + 0.15*PLR-0.35*PFR.The ACPP score discriminating pancreatic cancer from benign pancreatic tumor group had an AUROC of 0.917(95%CI: 0.873-0.962),discriminating pancreatic cancer from chronic pancreatitis with AUROC of 0.925(95%CI: 0.882-0.968),and discriminating pancreatic cancer from benign pancreatic disease with AUROC of0.921(95%CI: 0.882-0.960).Referring to the results of ROC curve analysis,42 was selected as the diagnostic positive threshold.The ACPP score had a sensitivity of82.1%,specificity of 90.4%-92.3%,and accuracy of 84.9%-86.0% for the diagnosis of pancreatic cancer,all of which were better than CA19-9.The ACPP score also had a higher positive rate than CA19-9 in all stages of pancreatic cancer.Conclusion:1.In addition to CA19-9,some conventional clinical indicators such as age and platelet count also have certain diagnostic value for pancreatic cancer.2.The comprehensive indexes AFR and PFR derived from routine clinical indexes have better diagnostic value for pancreatic cancer than their corresponding single indexes.3.The diagnostic model of pancreatic cancer established by age,CA19-9,PLR and PFR has good diagnostic value for pancreatic cancer,which is better than CA19-9.4.The ACPP score based on the diagnostic model of pancreatic cancer is simple,practical and convenient for the clinical application of the model.