KR102070916B1 - Biomarker for predicting prognosis in patients with pacreatic cancer and use thereof - Google Patents

Abstract

The present invention relates to: a preoperative biomarker which can predict the prognosis after pancreatic resection of a pancreatic cancer patient; and a method using the same. According to a composition for predicting prognosis of the pancreatic cancer patient, a kit thereof, and a prognostic method for predicting prognosis of the pancreatic cancer patient according to an aspect, it is possible to simply diagnose the prognosis after pancreatic resection of the pancreatic cancer patient, thereby being able to reduce side effects of pancreatic cancer resection and perform patient-specific treatment.

Description

Biomarker for predicting prognosis in patients with pacreatic cancer and use approximately}

The present invention relates to a preoperative biomarker and a method using the same for predicting the prognosis after treatment of pancreatic cancer patients.

Pancreatic cancer is one of the most deadly cancers of the gastrointestinal tract, and the survival rate is only 14 months, compared to other cancers. According to the 2008 national cancer registration statistics, pancreatic cancer ranks 9th (8.7 out of 100,000) among all carcinomas, while it is 5th. The overall five-year survival rate of all pancreatic cancer patients in Korea is 7.6%. On the contrary, the survival rate of pancreatic cancer has been almost constant for the last two decades. Did not improve. Pancreatic cancer is predicted to be the second leading cause of cancer-related deaths in 2030.

Although pancreatic resection is considered the most effective single treatment for the treatment of pancreatic cancer, only 15-20% of patients are candidates for resection in the diagnosis of pancreatic cancer. Recurrences and metastases are frequent after surgery, and long-term survival is maintained at less than 25-30%.

Although studies on the prognosis of pancreatic cancer have been conducted, most studies have focused on the early detection of pancreatic cancer, and there have been few successful studies on the long-term prognosis of pancreatic cancer. Therefore, there is a need for a study on biomarkers that can predict the prognosis according to postoperative treatment before operating a pancreatic cancer patient.

A biomarker composition for predicting a patient's prognosis after pancreatic resection from a sample before pancreatic resection, the composition, kit for predicting prognosis of pancreatic cancer comprising an agent measuring a metabolite level of a phosphatidylcholine derivative, and pancreatic cancer using the metabolite A method of providing information on prognostic prediction after pancreatic resection of a patient is provided.

The present invention invented the use of phosphatidylcholine derivatives as markers of metabolites as a result of intensive studies to find markers that can predict the prognosis of pancreatic cancer patients after pancreatic resection in samples isolated from patients before performing pancreatic resection.

One aspect provides a composition for predicting the prognosis after pancreatic resection of a preoperative pancreatic cancer patient comprising an agent that measures the metabolite level of a phosphatidylcholine (PC) derivative.

The metabolite of the PC derivative may be selected from the group consisting of PC.aa.C38: 4, PC.ae.C42: 5, and PC.ae.C38: 6, and may include functional equivalents thereof. . PC.aa.C38: 4 can be used in combination with PC.aa.C38_4, PC.ae.C42: 5 with PC.ae.C42_5, and PC.ae.C38: 6 with PC.ae.C38_6.

The present inventors have found that patients undergoing pancreatic resection can be distinguished into two clusters according to the metabolite expression pattern of phosphatidylcholine derivatives, and before PC.aa.C38: 4, PC.ae.C42: 5, or PC For the first time, the effect of predicting the prognosis of pancreatic cancer patients after pancreatic resection was determined using the concentration of .ae.C38: 6. Patient-specific treatment is possible by predicting the patient's postoperative prognosis before undergoing pancreatic resection.

As used herein, "pancreatic cancer" refers to pancreatic ductal adenocarcinoma, acinar cell carcinoma, neuroendocrine tumor, and serous cystadenoma, which is benign cystic tumor. , Mucinous cystic neoplasms, intraductal papillary mucinous neoplasms (IPMN), solid pseudopapillary tumors, and also classified as stages 1 and 2 Pancreatic cancer according to the stage of cancer progression.

As used herein, "prognosis" refers to determining whether an object for a particular disease or condition, i.e., the subject, will later have the disease, determining the subject's responsiveness to treatment, or therametrics ( Monitoring the condition of the subject, for example, to provide information about the efficacy of the treatment).

The "predictive prognosis after pancreatic resection" refers to determining whether such subjects relapse, metastasis, drug reactivity, resistance, etc. after treatment of pancreatic cancer. For example, the treatment of pancreatic cancer may be pancreatectomy, and predicting disease free survival rate of pancreatic cancer patients after pancreatic resection. Disease free survival may refer to long-term survival, for example, 6 months, 1 year, or 2 years after pancreatic resection.

As used herein, the term "metabolite" refers to a metabolite obtained from a sample of biological origin, and preferably, a sample of biological origin from which the metabolite can be obtained may be whole blood, plasma, serum, platelets, and more preferably May be plasma. The metabolites may include substances produced by metabolism and metabolic processes or substances generated by chemical metabolism by biological enzymes and molecules.

As used herein, the term "marker" is a substance that can predict the prognosis of a pancreatic cancer patient after pancreatic resection, and a metabolite, polypeptide, protein or nucleic acid, gene, lipid, glycolipid, glycoprotein or Organic biomolecules such as sugars and the like. The marker in the present invention is a metabolite of phosphatidylcholine derivative, specifically, may be PC.aa.C38: 4, PC.ae.C42: 5, or PC.ae.C38: 6.

As used herein, "phosphatidylcholine" is a representative phospholipid, and refers to the binding of choline phosphate to diglycerides.

In the present specification, in "PC ae Cx: y" or "PC aa Cx: y", ae represents a diacyl form, ae represents an acyl-alkyl form, x represents a carbon number of a side chain, and y represents a double bond number. . Such matters are obvious to those skilled in the art. For example, PC.aa.C38: 4 is a metabolite of a phosphatidylcholine derivative in the form of diacyl, which means that the number of carbons in the side chain is 38 and the number of double bonds is four. The association between the metabolites of phosphatidylcholine derivatives and the prognostic prognosis of pancreatic cancer is unknown.

As used herein, the term "level measurement" refers to a process for identifying metabolite expression levels of phosphatidylcholine derivatives in a biological sample in order to predict the prognosis of pancreatic cancer. According to an embodiment of the present invention, PC.aa.C38: 4, PC Levels of .ae.C42: 5, or PC.ae.C38: 6 can be identified. Analytical methods for this purpose are chromatographic / mass spectrometry, but are not limited thereto.

The composition may further comprise a sample necessary for predicting the prognosis of pancreatic cancer. For example, the composition may further comprise an agent for measuring the level of serum CA 19-9. Carbohydrate antigen (CA) 19-9 is known as a tumor marker that is sensitive to colorectal cancer, pancreatic cancer, and biliary tract cancer. In one embodiment of the present invention, when compared with the use of serum CA 19-9 as a single biomarker, the use of the metabolite of the phosphatidylcholine derivative as a biomarker significantly increased postoperative prognostic predictive power in patients with pancreatic cancer Confirmed. Therefore, the metabolite of the phosphatidylcholine derivative according to one aspect is superior to the postoperative prognostic effect in patients with pancreatic cancer as compared to the conventional markers.

Another aspect provides a kit for predicting the prognosis after pancreatic resection of a preoperative pancreatic cancer patient comprising an agent that measures the metabolite level of the phosphatidylcholine derivative.

The metabolite of the PC derivative may be selected from the group consisting of PC.aa.C38: 4, PC.ae.C42: 5, and PC.ae.C38: 6.

The kit may further comprise an agent for measuring the level of serum CA 19-9.

Metabolites of the PC derivatives, levels of metabolites, measurements thereof are as described above.

The kit may further comprise a sample for predicting prognosis after pancreatic resection.

In one embodiment, the kit may be, but is not limited to, a chromatography kit.

The kit may include a computer with a built-in agent, apparatus, and algorithm for measuring the metabolite level of the phosphatidylcholine derivative, and through the algorithm correlating the results of the level measurement of the marker with the prognosis of pancreatic cancer, It may be directed to a kit.

Specifically, the kit is a receiver for receiving a variable selected from the group consisting of metabolites of phosphatidylcholine derivatives, eg, PC.aa.C38: 4, PC.ae.C42: 5, and PC.ae.C38: 6 ; A score calculator for calculating a predicted score for the variable; And a probability calculator configured to calculate a survival rate for the pancreatic cancer patient based on the predicted score.

The apparatus may further include an output unit connected to at least one of the receiver, the score calculator, and the probability calculator.

The output unit may output a nomogram capable of inputting the variable.

The device is a device for driving a web page, an application, and the like, and may include, for example, a computing device, a mobile device, a server, and the like. The apparatus may include components of a processor, a storage unit, a memory, an input unit, and an output unit, and a receiver, a score calculator, and a probability calculator may be implemented through components of the device. When implemented as a server, the prognostic predictive apparatus of the pancreatic cancer patient may be driven to transmit the calculated values to another device having an output.

Specifically, the receiver receives variables for each of the PC metabolites, eg, PC.aa.C38: 4, PC.ae.C42: 5, or PC.ae.C38: 6 levels.

An output unit may be connected to the receiving unit, and the output unit may visually output an input monogram for inputting a variable.

The score calculator calculates a predicted score for the variable received by the receiver. In detail, the score calculator may match a predetermined value for each of the variables. For example, the matching may be performed by using a nomogram in which a measurement range of each variable is matched to at least a portion of a score line having a lowest point and a highest point. In one aspect, the nomogram may be the monogram of FIG. 3.

The probability calculator receives a predicted score for each variable matched by the score calculator and calculates a survival rate for the pancreatic cancer patient based on the predicted score. In detail, the probability calculator may add a total score by summing all prediction scores for the variable, and match the total score with a predetermined survival rate. For example, the matching may be performed using a nomogram with a matched total score sum of a predetermined survival rate of the pancreatic cancer patient and a predicted score for each variable.

The nomogram may be determined by a survival rate calculation formula of a pancreatic cancer patient.

Known algorithms may be used in the kit, including but not limited to linear or nonlinear regression algorithms; Prior or nonlinear classification algorithms; ANOVA; Neural network algorithm; Genetic algorithm; Support vector machine algorithms; Hierarchical analysis or clustering algorithms; Hierarchical algorithms using decision trees, or Kernel principal components analysis algorithms; Markov Blanket algorithm; recursive feature elimination or entropy-based recursive feature elimination algorithms; a plurality of algorithms arranged in a committee network; And a forward floating search or a backward floating search algorithm.

Another aspect provides a method of providing information necessary for predicting prognosis after pancreatic resection of a pancreatic cancer patient comprising measuring metabolite levels of phosphatidylcholine derivatives in a biological sample isolated from the pancreatic cancer patient.

The metabolite of the PC derivative may be selected from the group consisting of PC.aa.C38: 4, PC.ae.C42: 5, and PC.ae.C38: 6.

The method may further comprise measuring the level of serum CA 19-9.

Metabolites of the PC derivatives, levels of metabolites, measurements thereof are as described above.

The subject may be an individual suffering from pancreatic cancer or an individual who has not undergone pancreatic resection.

The subject may be a mammal, including a human.

As used herein, the term "biological sample" refers to a sample obtained from an organism. The biological sample may be blood, plasma, platelets, serum, ascites fluid, bone marrow fluid, lymph fluid, saliva, tear fluid, mucosal fluid, amniotic fluid, or a combination thereof. When the biological sample is blood or plasma, blood and plasma, which are easily collected, are used as a sample, so the organ tissue of the individual may not be extracted, and thus the analysis may be performed without burdening the individual.

The measuring may include, but is not limited to, performing chromatography / mass spectrometry.

The method measures the levels of metabolites, PC.aa.C38: 4, PC.ae.C42: 5, and PC.ae.C38: 6, and converts the measured levels into scores, for example. The level of the metabolite of the measured PC derivative can be converted into a unique value. For example, to calculate the eigenvalues, a standardized value can be used that is subtracted from the measured metabolite level and divided by the standard deviation. In addition, the method converts the measured level of each metabolite into a unique value, and gives a point corresponding to the converted unique value, thereby making it totally disease-free through total points that add up the scores of each variable. Predict survival rates.

In one embodiment of the invention, the method comprises a metabolite of a PC derivative of a pancreatic cancer patient, for example PC.aa.C38: 4, PC.ae.C42: 5, and PC.ae.C38: 6 Obtaining a measurement value for the selected variable from the step of calculating a prediction score corresponding to the obtained measurement value, and calculating a survival rate of the pancreatic cancer patient based on the total score of the sum of the prediction scores for each variable. It includes.

In the calculating of the predicted score, a nomogram in which the measured value range of each variable is matched to at least a portion of the score line having the lowest and highest points. The nomogram may be the nomogram of FIG. 3.

In calculating the survival rate of the pancreatic cancer patient, a nomogram matched with a total score obtained by summing the survival rate of the pancreatic cancer patient and a predicted score for each variable may be used.

As used herein, "variable" is a substance that can predict the prognosis of a pancreatic cancer patient after pancreatic resection, and an organic substance such as metabolites, polypeptides, proteins or nucleic acids, genes, lipids, glycolipids, glycoproteins, or sugars, which have different prognosis after pancreatic resection. Biomolecules and clinical variables such as age, sex, tumor location, tumor size, surgical method, and the presence of symptoms at the time of cancer diagnosis. For example, it may include PC.aa.C38: 4, PC.ae.C42: 5, and PC.ae.C38: 6, which are metabolites of PC derivatives.

The method may further use clinical information other than the marker of the patient, in addition to the marker analysis results, to provide information regarding the prediction of pancreatic cancer recurrence prognosis. Such clinical information includes, for example, age, sex, weight, diet, body mass, ultrasound, computed tomography (CT), magnetic resonance imaging (MRI), angiography, endoscopic retrograde pancreatic duct angiography, ultrasound endoscopy, Tumor marker testing, and the like.

According to the composition, kit, and prognostic method for predicting prognosis of pancreatic cancer patient according to one aspect, the prognosis after pancreatic resection of pancreatic cancer patient can be easily diagnosed before surgery, thereby reducing side effects on pancreatic cancer resection and providing customized treatment. It is possible.

1 is a cluster phylogeny according to expression patterns of preoperative serum metabolites in resected pancreatic cancer patients. Cluster 1: right, blue; Cluster 2: left, red; The y axis represents a measure of proximity of individual clusters.
Figure 2 is a graph showing disease-free survival according to preoperative serum metabolite based clustering.
3 is a nomogram capable of predicting the prognosis of patients with preoperatively resected pancreatic cancer.

It will be described in more detail through the following examples. However, these examples are provided to illustrate one or more embodiments illustratively and the scope of the present invention is not limited to these examples.

Example 1 Biomarker Screening for Predicting Prognosis of Pancreatic Cancer

1.1 Subject Patient

The study was performed on 57 patients with pancreatic ductal adenocarcinoma who underwent pancreatectomy from November 2012 to June 2014 with preoperative blood samples and long-term follow-up data. Proceeded. The patient's medical records were reviewed retrospectively, including age, sex, neoadjuvant treatment, jaundice, CA 1-9, tumor size, tumor location, surgical method, pathological findings, AJCC cancer stage, postoperative assistance. Preoperative clinical findings including chemotherapy were investigated. This protocol was approved by the Institutional Review Board of Severance Hospital.

General characteristics of the 57 patients are as follows. Of the 57, 32 patients were male (56.1%) and 25 patients were female (43.9%). Pancreatic head cancer was found in 42 patients (73.7%) and body and tail cancers of the pancreas were found in 15 patients (26.3%). Serum CA 19-9 was 1058.1 (24744) at initial diagnosis. Neoadjuvnat chemo ± radiation therapy was given to 12 patients (21.1%). Pancreaticoduodenectomy was performed in 41 patients, spleen ablation and distal pancreatic resection in 15 patients, and total pancreatectomy in 1 patient. The size of the resected tumor was 2.8 (1.2) cm in diameter and the number of metastatic lymph nodes was 1.8 (2.5).

1.2 Serum in Preoperative Blood Samples of Pancreatic Cancer Patients Before Pancreatic Resection Metabolite (metabolomes) detection

Metabolites were detected from preoperative blood samples collected by patients with pancreatic ductal cell carcinoma who had the above characteristics, and were subjected to high-performance liquid chromatography (HPLC) -tandem mass spectrometry and Absolute. The IDQ ^™ p180 kit (BIOCRATES Life Sciences AG, Innsbruck, Austria) was used to analyze based on a targeted metabolomics approach. Metabolites include 21 amino acids, 21 biogenic amines, 40 acylcarnitines (Cx: y), 93 glycerophospholipids [14 lysophosphatidylcholines (lyso PCx: y) and 79 phosphatidylcholines (PC aa x : y or PC ae x: y)], 15 sphingolipids (SMx: y or SM (OH) x: y) and 1 hexose. Cx: y represents the lipid side chain arrangement, where x represents the number of carbons in the side chain and y represents the number of unsaturated chains. Sample preparation and analysis procedures were performed according to the manufacturer's instructions with simultaneous measurement of three quality control substances. Samples were analyzed with a QTRAP 5500 mass spectrometer (SCIEX, Woodlands Central, Singapore) combined with Agilent 1290 series HPLC. However, ten of the metabolites were not detected in the major ratio and thus were not used for statistical analysis. Finally, 178 metabolites were used for statistical analysis.

1.3 Statistical Analysis: Hierarchical Clustering Analysis

Hierarchical clustering analysis was performed on the preoperative serum metabolites of the resected pancreatic cancer patients detected as described above to derive two groups (clusters 1 and 2) with similar metabolite patterns.

Specifically, in order to convert highly correlated preoperative serum metabolites into variables grouped by hierarchical methods and to see how clusters are formed, Euclidean distance as a distance measure is used as a linkage method. The 'Ward.D2' algorithm was used.

To determine the appropriate number of preoperative serum metabolite-based clustering groups, a hierarchical clustering algorithm based on the silhouette method was performed on the R package factoextra .

Categorical variables are expressed in frequency and percentage by Fisher's accuracy test results. Continuous variables are described as the mean standard deviation when the normality assumption is met, otherwise as the median (quadrant range). When the normality assumptions for the continuous variables were violated, the Wilcoxon rank sum test was performed instead of the Student's t-test.

The p-value obtained in the test for comparing two clusters was adjusted by Bonferroni correction to solve the multiple comparison problem. In order to appropriately reduce the number of variables used in the model using only those having a large difference between clusters among the 178 metabolites used, the p-value was extracted to be adjusted to less than 0.001.

As a result, it was confirmed that 57 observed patients could be clearly separated into two clusters by using 157 preoperative serum metabolites.

1 is a cluster phylogeny according to expression patterns of preoperative serum metabolites in resected pancreatic cancer patients. Cluster 1: right, blue; Cluster 2: left, red; The y axis represents a measure of proximity of individual clusters.

Of the 157 detected metabolites, the top 15 metabolites that are most distinguished between the two clustering groups are summarized in Table 1 below. The metabolites listed in Table 1 have adjusted p-values to <.001. The R value is the p-value obtained from Bonferroni's calibration and Wilcoxon rank sum test.

Metabolite Cluster 1 (N = 41) Cluster 2 (N = 16) R value PC.ae.C36_4 12054 (10989, 12858) 7225 (6699.8, 8850) 4.40E-05 PC.ae.C38_4 7136 (6450, 8166) 5068.5 (4220, 5402.8) 5.90E-05 PC.ae.C40_5 2574 (2302, 3137) 1727 (1578.2, 2070.8) 5.90E-05 PC.ae.C38_5 12576 (11306, 13565) 8770.5 (7702.5, 9768.2) 7.10E-05 PC.ae.C40_4 1386 (1238, 1630) 957.5 (857, 1120) 1.10E-04 PC.ae.C42_4 468 (435, 543) 353 (318, 411.8) 1.50E-04 PC.ae.C36_5 10046 (8675, 10712) 6391.5 (5230.2, 7379) 1.60E-04 PC.ae.C42_5 1181 (1049, 1301) 872.5 (805.8, 974) 2.10E-04 PC.ae.C44_4 235 (223, 246) 184.5 (171.8, 194.2) 2.50E-04 PC.aa.C40_6 41083 (34973, 49833) 25533.5 (21715.2, 32665) 3.30E-04 PC.aa.C40_4 1941 (1743, 2226) 1314 (1134.2, 1543.8) 4.30E-04 PC.ae.C38_6 7616 (6387, 8615) 5007 (3841.8, 5751.2) 4.70E-04 PC.aa.C38_4 76637 (65875, 87120) 54081 (44843.5, 60592.8) 5.10E-04 PC.ae.C40_1 758 (667, 905) 532 (450.8, 586.5) 7.20E-04 PC.aa.C38_0 3047 (2659, 3710) 2126.5 (1755, 2487.2) 7.90E-04

As shown in Table 1, all metabolites are associated with phosphatidylcholine (PC), which is distinguished by the presence of esters ("a") and ethers ("e") bound to the glycerol moiety. Among them, the "aa" (= diacyl) form of PC was identified in six metabolites (40%), and the "ae" (= acyl-alkyl) form of PC was identified in nine metabolites (60%). . It was also found that the PC derivatives were significantly lower in the cluster 2 patients compared to cluster 1. The clinical and pathological characteristics were compared between the two preoperative metabolite based clustering groups and the results are shown in Table 2.

Demographic data
(Demographics) all
(N = 57) Cluster 1
(N = 41, 71.9%) Cluster 2
(N = 16, 28.1%) p-value age 67 (58, 72) 67 (57, 73) 67 (60.75, 69.25) 0.742 gender female 25 (43.9) 20 (48.8) 5 (31.2) male 32 (56.1) 21 (51.2) 11 (68.8) Neoadjuvnat chemo ± radiation therapy 0.287 No 45 (78.9) 34 (82.9) 11 (68.8) Yes 12 (21.1) 7 (17.1) 5 (31.2) jaundice 0.771 No 34 (59.6) 25 (61) 9 (56.2) Yes 23 (40.4) 16 (39) 7 (43.8) Glucose 129 (110, 206) 122 (107, 180) 180.5 (131.75, 324.5) 0.035 Total bilirubin 6.9 (6.4, 7) 7 (6.4, 7.1) 6.8 (6.38, 7) 0.252 protein 4 (3.8, 4.2) 4 (3.8, 4.1) 4 (3.8, 4.1) 0.703 albumin 1.2 (0.6, 8.1) 12 (05, 8) 1.25 (0.95, 8.67) 0.742 CA 19-9 157 (32, 555.1) 98.1 (26.2, 524.3) 178 (49.6, 625.33) 0.689 Tumor size 2.5 (2, 3.3) 2.3 (2, 3) 3 (2.25, 3.85) 0.092 Tumor location > .999 head 42 (73.7) 30 (73.2) 12 (75 Body + Tail 15 (26.3) 11 (26.8 4 (25) Surgical Method > .999 Pancreaticoduodenectomy
: PD) 3 (5.3) 2 (4.9) 1 (6.2) Pylorus-preserving pancreaticoduodenectomy (PPPD) 38 (66.7) 27 (65.9) 11 (68.8) Distal pancreatectomy and splenectomy 15 (263) 11 (26.8) 4 (25) Total pancreatectomy (TP) 1 (1.8) 1 (2.4) 0 (0) Distribution 0.662 Well & Moderate 50 (89.3) 35 (87.5) 15 (93.8) Poor 6 (10.7) 5 (12.5) 1 (6.2) Lymphatic invasion 0.752 No 39 (68.4) 27 (65.9) 12 (75) Yes 18 (31.6) 14 (34.1) 4 (25) Vascular infiltration 0.236 No 35 (61.4) 23 (56.1) 12 (75) Yes 22 (38.6) 18 (43.9) 4 (25) Peripheral Infiltration 0.74 No 13 (22.8) 10 (24.4) 3 (18.8) Yes 44 (77.2) 31 (75.6) 13 (81.2) Margin > .999 Macroscopic residual cancer at the margin 47 (82.5) 34 (82.9) 13 (81.2) R1 (microscopic residual cancer at the margin) 10 (17.5) 7 (17.1) 3 (18.8) AJCC 8th_T stage 0.669 T1 16 (28.1) 13 (31.7) 3 (18.8) T2 31 (54.4) 21 (51.2) 10 (62.5) T3 10 (17.5) 7 (17.1) 3 (18.8) AJCC8 th_N stage 0.666 NO 27 (47.4) 21 (51.2) 6 (37.5) N1 21 (36.8) 14 (34.1) 7 (43.8) N2 9 (15.8) 6 (14.6) 3 (18.8) #Retrieved LNs 16.93 ± 9.3 17.32 ± 9.86 15.94 ± 7.9 0.585 #Positive LNs 1 (0, 3) 0 (0, 3) 1 (0, 2) 0.858 LNR 0.05 (0, 0.15) 0 (0, 0.15) 0.06 (0, 0.13) 0.749 Postoperative adj-CTx No 9 (15.7) 7 (17.0) 2 (12.5) Yes 48 (84.3) 34 (83.0) 14 (87.5) 0.720

As shown in Table 2, other clinicopathological features such as age, sex, CA 19-9, preoperative chemoradiotherapy, tumor characteristics, and pathologic findings between the two clusters were found to be statistically insignificant. p> 0.05). However, among the clinicopathological features, patients in Cluster 2 showed statistically significantly higher preoperative serum glucose (122 (107,180) vs. 180.5 (131.75, 324.5), p = 0.035).

1.4 Biomarker Screening Associated with 1-Year Disease-Free Survival in Pancreatic Cancer Patients

First, in order to screen important prognostic factors for predicting 1-year disease-free survival, covariates based on age, sex, prior chemoradiotherapy, tumor size, preoperative CA 19-9, jaundice and tumor location were selected. Univariate Cox proportional hazard model was used. A multivariate Cox proportional hazard model was designed to add metabolites to distinguish prognostic factors as well as clustering groups. Finally, the best model was finally designed to include metabolites under constraints where normalization of the metabolite data and the last setting of covariates was less than 10 with a variation inflation factor (VIF). R package was implemented in My.stepwise . Proportional principle assumptions for the Cox model were also identified. Heagerty's integrated time-dependent area under the curve (iAUC) for each of the Harrells concordance (C) -index and 1000 bootstrap samples was used to assess the performance between established relapse-predictive models.

As a result, cluster 2 (HR = 2.839, P = 0.015), tumor size (HR = 1.433, p = 0.015), and surgery were used as independent markers for predicting 1 year in patients with resected pancreatic carcinoma after pancreatic ductal carcinoma. All CA 19-9 (HR = 1.0001, P = 0.043) were identified and the relevant data is shown in Table 3 below.

1 year disease free survival forecast Univariate Multivariate HR (95% confidence interval) p-value HR (95% confidence interval) p-value Cluster 2 (by: Cluster 1) 2.874 (1.293, 6.389) 0.01 2.839 (1.227, 6.571) 0.015 Tumor size 1.558 (1.18, 2.057) 0.002 1.433 (1.073, 1.912) 0.015 CA 19-9 1.0001 (1, 1.0002) 0.054 1.0001 (1, 1.0003) 0.043

Among other preoperative variables, age (HR = 0.985, P = 0.483), gender (HR = 2.1774 P = 0.07), prior chemoradiotherapy (HR = 0.6097, P = 0.365). Jaundice (HR = 1.3349, P = 0.474), and tumor location (HR = 0.5803, P = 0.277) were found to be not important prognostic factors for predicting 1 year disease free survival. Therefore, according to the expression of the metabolites detected in the serum metabolites prior to pancreatic resection, the patients can be divided into two clusters, and cluster 2 was found to have a significant correlation with the disease-free survival prediction of the resected pancreatic cancer patients.

1.5 Establishment of Nomogram for Predicting Long-Term Disease-free Survival in Pancreatic Cancer Patients

Based on the above results, in order to establish a nomogram for predicting long-term disease-free survival rate in patients with resected pancreatic ductal cell carcinoma, the most predictive Cox model was considered. All statistical hypothesis tests were two-sided tests with a significance level of 0.05, and all statistical analyzes were implemented using R package version 3.4.0.

First, the results of long-term disease-free survival according to the clustering group based on preoperative serum metabolites were confirmed. Figure 2 is a graph showing disease-free survival according to preoperative serum metabolite based clustering.

As shown in FIG. 2, long-term disease-free survival was significantly different according to the preoperative serum metabolite based clustering group. Since patients with cluster 2 have lower initial survival compared to patients with cluster 1, it can be seen that patients with cluster 2 exhibit early tumor recurrence. (Median 8 months [95% confidence interval: 5.521-10.479] vs. median 15 months [95% confidence interval: 8.375-21.625], p = 0.034). However, there was no statistical difference in disease specific survival between the two groups (p = 0.312). This result is consistent with the 1-year disease-free survival rate in patients with resected pancreatic cancer.

Next, tumor size among the factors identified in Example 1.4 was excluded from the nomogram development because the pathologist is expected to determine. That is, preoperative serum CA 19-9 which is a measurable variable before surgery; Three important metabolites of serum metabolites, PC.aa.C38_4, PC.ae.C42_5 and PC.ae.C38_6, were selected as final factors.

3 is a nomogram capable of predicting the prognosis of patients with resected pancreatic cancer.

A nomogram is a result that makes it easy to visually represent and interpret statistical prediction models based on clinical data. By estimating the estimates of the variables used in the model, we can estimate the risk corresponding to the corresponding values, and calculate the predicted final survival probability by adding the scores corresponding to all the variables. In FIG. 3, 'Points' represents the entire range of scores that variables can have. Preoperative serum CA 19-9, preoperative PC.aa.C38_4, PC.ae.C42_5, PC.ae.C38_6 are variables. For each variable, we use the standardized value subtracted from the mean and divided by the standard deviation.

In the nomogram, the names of the variables (preoperative serum CA 19-9, preoperative PC.aa.C38_4, PC.ae.C42_5, PC.ae.C38_6) and the model show the intervals of the points of each variable. Indicated. Total Points represents the sum of the scores corresponding to the variables, and is changed to the predicted values (Linear Predictor) and 6-month and 12-month survival probabilities (6-month Survival Probability), respectively. The range of values and their numerical values are shown.

In addition, the 6- and 12-month survival probabilities using the estimates and equations estimated from the predictive model are:

번째 대상자에 대한

시점에 대한 생존확률로 정의된다.

는 기저생존함수,

모형에 포함된 변수, 각 변수의 효과는 모수

를 통해 추정된다. 생존확률은 지수함수인

를 통해서 표현된다.Survival chance

For the first

It is defined as the survival probability for a point in time.

Is the basis survival function,

Variables included in the model, the effect of each variable is a parameter

Is estimated through. Survival probability is an exponential function

Is expressed through

Predicted 6-month disease-free survival = exp {-0.144521} ^ exp (0.00007899923 * (ca199a-ca199a mean) -1.273374 * ((PC.ae.C38_4-mean) / standard deviation-standardized variable mean) + 0.8878073 * (( PC.ae.C42_5-average) / standard deviation-standardized variable mean)-0.5831592 * ((PC.ae.C38_6-average) / standard deviation-standardized variable mean)

Example 12 months disease-free survival = exp {-0.545664} ^ exp (0.00007899923 * (ca199a-ca199a mean) -1.273374 * ((PC.ae.C38_4-mean) / standard deviation-standardized variable mean) + 0.8878073 * (( PC.ae.C42_5-average) / standard deviation-standardized variable mean)-0.5831592 * ((PC.ae.C38_6-average) / standard deviation-standardized variable mean)

Where ca199a is preoperative serum CA 19-9.

Experimental Example  1. Predicting the Prognosis of Preoperatively Resected Pancreatic Cancer Patients Nomogram  Verification

Model performance was compared to three models to test for the accuracy of predicting early disease free survival in resected pancreatic cancer. Model-1 is preoperative CA 19-9, Model-2 is Model-1 + Cluster 2, Model-3 is Model-1 + 3 metabolites (PC.aa.C38_4, PC.ae.C42_5, PC.ae .C38_6), Model-4 was established based on three metabolites (PC.aa.C38_4, PC.ae.C42_5, PC.ae.C38_6). The evaluation results of the prognostic predictive powers for the three models are shown in Tables 4 and 5 below.

Model-1 Model-2 Model-3 Model-4 HR
(95% Cl) p-value HR
(95% Cl) p-value HR
(95% Cl) p-value HR
(95% Cl) p-value CA 19-9 1.0001 (
1, 1.0002) 0.054 1.0001
(1, 1.0003) 0.018 1.0001
(0.9999, 1.0002) 0.205 Cluster 2 (Base. Cluster 1) 3.303
(1.443, 7.556) 0.005 PC.aa.C38_4 0.28 (0.134, 0.587) <.001 0.271
(0.126, 0.58) 0.001 PC.ae.C42_5 2.43 (1.245, 4.743) 0.009 2.768
(1.433, 5.348) 0.002 PC.ae.C38_6 0.558 (0.324, 0.963) 0.036 0.538
(0.315, 0.918) 0.023

Harrells' C-index (95% Cl) iAUC Model-1 0.619 (0.487, 0.733) 0.573 (0.507, 0.653) Model-2 0.695 (0.591, 0.794) 0.684 (0.586, 0.782) Model-3 0.823 (0.75, 0.891) 0.816 (0.736, 0.893) Model-4 0.811 (0.735, 0.88) 0.816 (0.736, 0.893) Model-1 vs. Model-2 -0.076 (-0.209, 0.021) -0.111 (-0.207, -0.022) Model-1 vs. Model-3 -0.204 (-0.349, -0.08) -0.243 (-0.346, -0.15) Model-1 vs. Model-4 -0.193 (-0.348, -0.061) -0.243 (-0.346, -0.15) Model-2 vs. Model-3 -0.128 (-0.238, -0.04) -0.132 (-0.227, -0.045) Model-2 vs. Model-4 -0.116 (-0.232, -0.014) -0.132 (-0.227, -0.045) Model-3 vs. Model-4 0.012 (-0.013, 0.06) 0 (0, 0)

As shown in Tables 4 and 5, Model-3 or Model-4 using preoperative serum metabolites (PC.aa.C38_4, PC.ae.C42_5, PC.ae.C38_6) was 6 months after pancreatic resection or It can be seen that it is a powerful preoperative clinical model for predicting the 1-year cancer free survival rate. Harrells' C-index is 0.823 [95% CI: 0.750, 0.891], iAUC is 0.816 [95% CI: 0.736, 0.893], and C-index of 0.6 or higher is generally estimated to be predictive. By confirming the expression of the serum metabolites PC.aa.C38_4, PC.ae.C42_5, PC.ae.C38_6, the disease-free survival rate of patients after pancreatic resection can be effectively predicted and used to provide information on patient-specific treatment. In particular, the metabolites showed better prognostic ability than serum CA 19-9, which is known as a factor for prognostic prediction of pancreatic cancer.

Claims (9)

A composition for predicting the prognosis after pancreatic resection in a patient with preoperative pancreatic cancer, the composition comprising an agent for measuring the metabolite level of a phosphatidylcholine (PC) derivative, wherein the metabolite of the PC derivative is PC.aa.C38: 4 , PC.ae.C42: 5, and PC.ae.C38: 6, wherein the prognosis relates to disease-free survival of pancreatic cancer patients after pancreatic resection. The composition of claim 1, further comprising an agent for measuring the level of CA 19-9. A kit for predicting prognosis after pancreatic resection in a patient with preoperative pancreatic cancer, the kit comprising an agent for measuring the metabolite level of a phosphatidylcholine (PC) derivative, wherein the metabolite of the PC derivative is PC.aa.C38: 4 , PC.ae.C42: 5, and PC.ae.C38: 6, wherein the prognosis is for disease-free survival of pancreatic cancer patients after pancreatic resection. The kit of claim 3, further comprising an agent for measuring the level of CA 19-9. A method of providing information necessary for predicting prognosis after resection of pancreatic cancer.
Measuring metabolite levels in a biological sample isolated from a pancreatic cancer patient, said metabolite group consisting of PC.aa.C38: 4, PC.ae.C42: 5, and PC.ae.C38: 6 Wherein the prognosis relates to disease-free survival in pancreatic cancer patients following pancreatic resection. The method of claim 5, wherein the biological sample is a sample isolated from a pancreatic cancer patient prior to undergoing pancreatic resection. The method of claim 5, wherein the biological sample is blood, plasma, platelets, serum, or a combination thereof. The method of claim 5, further comprising measuring serum CA 19-91 levels in the biological sample isolated from the pancreatic cancer patient. The method of claim 5, wherein measuring the level of the metabolite is performed using chromatography / mass spectrometry.

Images