Serum metabonomic study of patients with pancreatic cancer and chronic pancreatitis
10.3760/cma.j.cn115667-20250117-00011
- VernacularTitle:胰腺癌与慢性胰腺炎患者血清代谢组学研究
- Author:
Ming YANG
1
;
Yisha GAO
;
Diya LYU
;
Fei FENG
;
Can XU
Author Information
1. 海军军医大学第一附属医院消化内科，上海 200433
- Publication Type:Journal Article
- Keywords:
Pancreatic neoplasms;
Chronic pancreatitis;
Metabonomics;
Serum;
High performance liquid chromatography;
Tandem mass spectrometry
- From:
Chinese Journal of Pancreatology
2025;25(2):97-103
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To investigate the serum metabonomics of pancreatic cancer and chronic pancreatitis patients and screen the differential metabolites for differentiating pancreatic cancer from chronic pancreatis.Methods:From June 2021 to June 2022, the clinical data of 54 patients diagnosed with pancreatic ductal adenocarcinoma in the Department of Hepatobiliary, Pancreatic and Splenic Surgery of the First Affiliated Hospital of Naval Medical University were collected. A total of 54 patients with chronic pancreatitis who were admitted at the same time were selected as the control group. UHPLC/Q-TOF MS-based metabonomics techniques were applied to analyze the difference in serum metabolites in the two groups with multivariate and univariate statistical method, and the different metabolites were screened and identified in accordance with the molecular weight, metabolites databases and mass spectrometry (MS)/MS information. Two-thirds of the cases were randomly selected from patients with pancreatic cancer and chronic pancreatitis as the modeling population, and the remaining population was used as the validation population. In the modeled population, the receiver operating characteristic curve (ROC) of the screened differential metabolites was plotted and the area under the curve (AUC) was calculated. Differential metabolite variables with AUC ≥75%, a VIP value ≥1.5 were selected into the logistic multivariate regression analysis model, and the regression equation and the regression coefficients of each selected independent variable were obtained by stepwise regression by backward method. The final selected differential metabolites were evaluated by the formula P=1/{1+Exp[-(β0+β1X1+β1X1+…+βiXi)]} to establish a diagnostic model and predict the clinical application value of the model by evaluating it compared to the CA19-9 ROC curve. At the same time, the non-parametric Bootstrap method was used to verify the diagnostic performance of the model in the validation population. Results:There were 18 kinds of different serum metabolits in the final screening and identification in the two groups. The level of hypoxanthine, L-carnitine, acetylcarnitine, C16 sphinganine, linoleic acid, palmitoylcarnitine, linoleyl carnitine, uracil deoxynucleotide, glycocholic chenodeoxycholic acid in serum of pancreatic cancer patients were higher than that in the chronic pancreatitis patients; Uric acid, tryptophan, indoxylsulfuric acid, 1-palmitoyl lysophosphatidic acid, LPA(18∶2/0∶0), LysoPE (18∶1/0∶0), LysoPC (14∶0), LysoPC(15∶0), LysoPC(16∶1) in the serum were lower in patients with pancreatic cancer compared with that in chronic pancreatitis patients, and all the differences were statistically significant (all P value <0.05). In the modeled population, the ROC curve was established according to the peak intensity of 18 differential metabolites, and the metabolic differentiators with AUC of ≥75% and VIP value of ≥1.5 were selected for logistic multivariate analysis, and finally linoleoleic carnitine and LPA (18∶2/0∶0) were included in the logistic regression model. The prediction model of pancreatic cancer with two serum metabolites linoleyl carnitine and LPA (18∶2/0∶0) was established. The AUC value (95% CI) of the prediction model was 0.91 (0.85-0.97), which was higher than that of CA19-9 (0.85, 0.76-0.94), the sensitivity and specificity were 86.4% and 80.6%, respectively, and the sensitivity was higher than that of CA19-9 (77.3%), but the specificity was lower than that of CA19-9 (91.7%). Internal validation showed than the AUC value (95% CI) of the prediction model was 0.91 (0.79-0.94), which was higher than that of CA19-9 ( P<0.05). Conclusions:The serum metabolites linolein carnitine and LPA(18∶2/0∶0) may be potential diagnostic markers to distinguish pancreatic cancer from patients with chronic pancreatitis.
