Estabishment Of A Scoring System For Estimating Clinical Diagnosis Of Pancreatic Cancer And The Research On Metabonomics Of Serum From Patients With Pancreatic Cancer

Pancreatic cancer is one of the most common cancers in human, whoseprevalence and mortality has been increasing year by year. Most imaging ofpancreatic cancer performed as the solid mass lesions of pancreas, but a small numberof low invasive malignant and benign lesions also showed solid mass. Benign andmalignant pancreatic mass lesion significantly differed in treatment options anddiagnosis, so it is need to stratificate the risk of pancreatic mass lesions in patients inorder to consider further surgical exploration or invasive diagnosis. With the lack ofearly detection, the patients had been on the metastatic stage once diagnosed withpancreatic cancer. Therefore, it is the key to improve the efficiency in early diagnosisof pancreatic cancer. As a new metabolite biomarkers, metabonomics may be aneffective tool for the early diagnosis of tumor. Recent studies on metabonomics hadfound that there were significant differences in metabolites of serum of pancreaticcancer and benign hepatobiliary diseases and chronic pancreatitis patients.The first series of our studies retrospectively analyzed clinical data of patientswith pathological diagnosis of pancreatic mass lesions to develop a prediction ruleand scoring system for estimating clinical diagnosis of the benign and malignantpancreatic mass lesions and estimate its diagnostic efficiency. Secondly, we comparedthe difference in the serum metabonomics of pancreatic cancer, chronic pancreatitispatients and health controls with the application of UHPLC/Q-TOF MS-basedmetabonomics techniques to looking for a specific marker for early diagnosis ofpancreatic cancer. Then we established the prediction rule of pancreatic cancer withthe different metabolites in serum of pancreatic cancer, chronic pancreatitis patientsand health controls with evaluation and validation of its diagnostic efficacy.1．Estabishment of a scoring system for estimating clinical diagnosis ofpancreatic cancerObjective: To develop a prediction rule and scoring system for estimating clinicaldiagnosis of the benign and malignant pancreatic mass lesions and estimate itsdiagnostic efficiency.Methods: We retrospective analyzed medical records of patients with a fnaldiagnosis of pancreatic mass lesions confrmed histologically from Changhai Hospitalbetween November2008and May2013. Two thirds of included patients wasrandomly selected as the derivation cohort, the rest1/3as the validation cohort. A prediction rule was developed from a logistic regression model by using a regressioncoeffcient-based scoring method, and then was externally validated. The diagnosticefficiency of the scoring system was valued by its calibration, discrimination andaccuracy.Results: A total of1000eligible patients were included in our research. The scoringsystem, which was scored from0to14points, comprised4variables: age, anorexia,diabetes and CA19-9.The system had good calibration(P=0.13) and gooddiscrimination(area under the receiver operating characteristic curve=0.82,95%confidence interval:0.79-0.86,P＜0.001). Score2was used as the predictive cut-offvalue. The sensitivity, specificity, accuracy rate, positive predictive value, negativepredictive value, positive likelihood ratio and negative likelihood ratio wererespectively81.46%,66.88%,77.86%,88.26%,54.12%,2.46,0.28.The risk(88.3%) ofmalignant pancreatic mass lesions among patients with high-risk scores(＞2) wasdistinguished higher than that(45.9%) among patients low-scores (≤2)(P＜0.001).The scoring system had good discrimination (area under the ROC curve=0.81,95%CI:0.76-0.86,P＜0.001). and calibration (P=0.72) in external validation.Conclusion: The scoring system can provide available risk stratifcation for patientswith histological diagnosis of benign or malignant pancreatic mass lesions, whichmay provide initial proof to making clinical decision of the benign and malignantpancreatic mass lesions.2. The research on metabonomics of serum from patients with pancreaticcancer.Objective: To investigate the overall serum metabolic differences between pancreaticcancer, chronic pancreatitis patients and normal groups, screen biomarkers that canidentify patients with pancreatic cancer and chronic pancreatitis and heath controlsand establish the prediction rule of pancreatic cancer with the different metabolites inserum by evaluating and validating its diagnostic efficacy.Methods: A case-control based on hospital and community health population wasconducted with objectives of patients with pathological diagnosis of pancreatic ductaladenocarcinoma as case group, health population with physical examination andpatients with chronic pancreatitis as control groups. We compared with the differencein the serum metabonomics of pancreatic cancer, chronic pancreatitis patients andhealth controls (age and sex-matched) with the application of UHPLC/Q-TOF MS -based metabonomics techniques，and analized the difference in serum metabolites inthe three groups with multivariate and univariate statistical method. The differentmetabolites were screened and identified in accordance with the molecular,metabolites databases and MS/MS information. Two thirds of patients withpancreatic cancer and controls was respectively and randomly selected as the trainingcohort, the rest1/3as the validation cohort. A prediction rule of pancreatic cancerbased on the serum metabolites was developed from a logistic regression model byusing a regression coeffcient-based scoring method in the training cohort, and thenwas internally validated by Bootstrap. The diagnostic efficiency of the prediction rulewas valued by its discrimination and accuracy with ROC curve.Results: In this study, there were54patients with pathological diagnosis of pancreaticductal adenocarcinoma,54patients with chronic pancreatitis and54health controlswith match of age and sex. There were22kinds of different serum metabolits in thefinal screening and identification in the three groups. The hydroxy-arachidonic acidand phenylalanine decreased in the serum of pancreatic cancer, chronic pancreatitispatients and normal controls successively, LysoPC (18:0) and LysoPC (P-16:0) aresequentially increased. The level of hypoxanthine, L-carnitine, acetylcarnitine, C16sphinganine, linoleic acid, palmitoylcarnitine, linoleyl carnitine, and uracildeoxynucleotide and glycocholic chenodeoxycholic acid in serum of pancreaticcancer patients were higher than that in the chronic pancreatitis patients and normalcontrols (p<0.05,respectively).Uric acid, tryptophan, indoxylsulfuric acid,1-PalmitoylLysophosphatidic acid, LPA(18:2/0:0), LysoPE (18:1/0:0), LysoPC (14:0),LysoPC(5:0), LysoPC(16:1)in the serum were lower in patients with pancreatic cancercompared with that in chronic pancreatitis patients and normal controls (p<0.05,respectively).Then we established the prediction rule of pancreatic cancer with twometabolites, linoleyl carnitine and LPA (18:2/0:0), in serum of pancreatic cancer,chronic pancreatitis patients,which had good discrimination(area under the receiveroperating characteristic curve=0.91（95%confidence interval:0.85,0.97, P＜0.001)compared to that of CA19-9[0.85(0.76,0.94)]（P＜0.05）. Its sensitivity andspecificity were86.4%、80.6%（cut-off=0.46）respectively, while those of CA19-9were77.34%、91.7%（cut-off=42.33U/ml）.The diagnostic model had gooddiscrimination (area under the ROC=0.91) compared to CA19-9(AUC=0.85)（P＜0.05）in internal validation.Its sensitivity were higher than that those of CA19-9in internal validation.We established the prediction rule of pancreatic cancer with two metabolites,linoleyl carnitine and LPA (18:2/0:0), in serum of pancreatic cancer and non-pancreatic cancer patients(chronic pancreatitis patients and heath controls), which hadgood discrimination(AUC=0.93,95%CI:0.88-0.97,P＜0.001) compared to that ofCA19-9[0.87(0.78,0.96)]（P＜0.05）. Its sensitivity and specificity were84.1%、85.4%（cut-off=0.34）respectively, while those of CA19-9were77.3%、96.3%（cut-off=42.33U/ml）.The diagnostic model had good discrimination (AUC=0.93)compared to CA19-9（AUC=0.87）（P＜0.05）in internal validation.Its sensitivity werehigher than those of CA19-9in external validation and its sensitivity and specificitywere over80%which indicated the model had good accuracy.Conclusion: The metabonomics in serum of patients with pancreatic cancer andchronic pancreatitis and health controls differed significantly and there were22different metabolites screened and identified from the three groups finally. In patientswith pancreatic cancer and chronic pancreatitis, pancreatic cancer patients andnon-pancreatic cancer cohort, the predict rule of pancreatic cancer established by thecombination of serum metabolites linoleyl carnitine and LPA (18:2/0:0), had gooddiscrimination and accuracy in both derivation and validation cohort, which evenmore excellent that CA19-9in some extent. It was suggesting that the differences ofmetabolites, linoleyl carnitine and LPA (18:2/0:0) in serum, had a high value in thediagnosis on pancreatic cancer.They might play an important role in development andprocess of pancreatic cancer.