Abstract
Aim: The first aim of this study was to search for new biomarkers to be used in gastric cancer diagnostics. The second aim was to verify the findings presented in literature on a sample of the local population and investigate the risk of gastric cancer in that population using a multivariant statistical analysis. Patients and Methods: We assessed a group of 36 patients with gastric cancer and 69 healthy individuals. We determined carcinoembryonic antigen, cancer antigen 19-9, cancer antigen 72-4, matrix metalloproteinases (-1, -2, -7, -8 and -9), osteoprotegerin, osteopontin, prothrombin induced by vitamin K absence-II, pepsinogen I, pepsinogen II, gastrin and Helicobacter pylori for each sample. Results: The multivariate stepwise logistic regression identified the following biomarkers as the best gastric cancer predictors: CEA, CA72-4, pepsinogen I, Helicobacter pylori presence and MMP7. Conclusion: CEA and CA72-4 remain the best markers for gastric cancer diagnostics. We suggest a mathematical model for the assessment of risk of gastric cancer.
Gastric cancer is a malignant disease with a very poor prognosis. The number of newly diagnosed cases is gradually decreasing each year. Nevertheless, mortality rates remain constant and prognosis is still unfavorable, with a five-year survival range of between 5% and 20% (1). Early diagnosis increases the chance of successful treatment. The first aim of our study was to search for new biomarkers for use in gastric cancer diagnostics. The second aim was to verify the findings presented in literature on a sample of a local population and, by using multivariant statistical analysis, to create a mathematical model for the investigation of the risk of gastric cancer in the human population.
Patients and Methods
Group of patients. This study was conducted between November 2013 and June 2015. We assessed a group of 105 probands:36 patients with gastric cancer and 69 healthy individuals. The age characteristics of the gastric cancer and control groups are shown in Table I. The gastric cancer group consisted of 20% of patients at stage I, 10% at stage II, 20% at stage III and 50% at stage IV of the disease. The control group of healthy individuals consisted of people who underwent regular preventive examinations in our hospital. None of the healthy individuals had had a cancer diagnosis, stomach diseases or disorders of anamnesis. All gastric cancer diagnoses were histologically verified. The study was approved by the local Ethical Committee on 11th July 2012 and all the patients gave their informed consent for this study. We determined values of carcinoembryonic antigen (CEA); cancer antigen 19-9 (CA19-9); cancer antigen 72-4 (CA72-4); matrix metalloproteinases (MMP) 1, 2, 7, 8 and 9; osteoprotegerin (OPG); osteopontin (OPN); prothrombin induced by vitamin K absence-II (PIVKA II); pepsinogen I; pepsinogen II; gastrin and Helicobacter pylori IgG, and we calculated ratio of pepsinogen I to pepsinogen II for each sample.
Plasma samples. Blood samples were collected at the time of diagnosis and prior to surgery or any other form of treatment. Peripheral venous blood was collected using the VACUETTE blood collection system (Greiner Bio-one Company, Kremsmünster, Austria). Serum or plasma was separated by 10 min centrifugation at 1,300 × g, and immediately frozen to −80°C. Samples were thawed only once, just prior to analyses.
Age characteristics of the patient groups.
Analytical methods used. Serum CEA and CA19-9 levels were assayed using a chemiluminescent ACCESS CEA kit (Beckman Coulter, Brea, CA, USA). Serum CA 72-4 levels were assayed using an electrochemiluminescent COBAS CA72-4 kit (F. Hoffmann–La Roche, Basel, Switzerland). Plasma MMP levels were assayed using a Bio-Plex Pro Human MMP assay (Bio-Rad Laboratories, Hercules, CA, USA). OPG and OPN were assayed using a Human Bone Panel Milliplex Map kit (Millipore Corporation, Billerica, MA, USA). Serum PIVKA II levels were assayed using a chemiluminescent PIVKA II kit (Abbott, Libertyville, IL, USA). Pepsinogen I, pepsinogen II, gastrin and Helicobacter pylori IgG were assayed using Biohit enzyme-linked immunosorbent assay (ELISA) kits (Biohit, Helsinki, Finland).
Statistical methods. Statistical Analysis Software release 9.2 (SAS Institute Inc., Cary, NC, USA) was used for all statistical analyses. A summary is presented of statistical findings concerning age and biomarker levels. The Wilcoxon test was used to compare distributions of biomarker values between the groups of patients and the group of healthy persons. Receiver operating characteristic (ROC) curves and the area under the curve (AUC) are presented to aid detailed assessment of the usefulness of selected biomarkers for gastric cancer diagnosis. We calculated the gastric cancer index (GCI) using a multivariant analysis. A multivariate stepwise logistic regression was used whereby the biomarkers (as gastric cancer predictors) were added one-by-one until they achieved statistical significance after addition to the model. The previously added predictors were eliminated if, after the addition of other predictors, they had become statistically non-significant. This ensures that all the predictors which in the final logistic regression model remain statistically significant.
Results
Table II shows a comparison of CEA, CA 19-9, CA 72-4, MMPs, OPG, OPN and PIVKA II levels between the group of healthy individuals and patients with gastric cancer, including the corresponding p-value. We observed a statistically significant increase in the gastric cancer patient group in the following biomarkers: CEA, CA72-4, MMP1, MMP7, OPN and PIVKA-II. There were no statistically significant differences in MMP2, MMP8, and OPG levels between the two groups.
Table III shows the evaluation of the gastric cancer risk factors. We observed a statistically significant decrease in pepsinogen I in the gastric patient group. We observed a statistically significant increase in the gastric patient group for the following two parameters: gastrin and Helicobacter pylori IgG in the group of healthy persons. We observed no statistically significant difference in pepsinogen II level.
The procedure of multivariate stepwise logistic regression led to the following biomarkers being chosen as the best predictors of gastric cancer: CEA, CA72-4, pepsinogen I, Helicobacter pylori, MMP7 and the procedure determined the optimal values of the multipliers (weights). We suggest the following calculation for gastric cancer diagnostics:
We plotted ROC curves and calculated the AUC for each individual biomarker chosen as the best predictor by multivariate logistic regression and compared them with the ROC curve and AUC for the GCI (Figure 1, Table IV). GCI had the best ROC curve and the highest AUC (0.9553) in comparison with the individual biomarkers.
Discussion
In spite of the decreasing number of cases, Gastric cancer is still a serious disease with a poor prognosis (2). Only early diagnosis can provide any hope of subsequent successful treatment. The prerequisite for successful treatment is to perform radical surgery. Long-term observation shows that only about 25% of patients diagnosed with gastric cancer are indicated for radical surgery. The low number of patients undergoing radical surgery is the result of delayed diagnosis and advanced disease (3).
Tumor markers CEA, CA19-9 and CA72-4 have long been incorporated into routine practice for gastric cancer diagnostics (4). However, their ability to help in making an early detection is limited due to their low sensitivity (5). In our pilot study, we decided to test the quality of new biomarkers in gastric cancer detection and their ability to contribute to the tumor markers which are already used in the diagnosis of gastric cancer.
MMPs are a family of zinc-dependent extracellular matrix remodeling endopeptidases implicated in pathological processes such as carcinogenesis. In this regard, their activity plays a pivotal role in tumor growth and the multistep processes of invasion and metastasis, migration and angiogenesis (6). We tested the entire panel of MMPs currently available as a Bio-Plex commercial panel. In our evaluation, the differences of MMP1, MMP7 and MMP9 were found to be statistically significant. The AUC of MMP7 was one of the highest AUC's in our panel of biomarkers and MMP7 was chosen as one of the best predictors of gastric cancer by our mathematical model, mentioned in the text below.
Biomarkers in patients with gastric cancer vs. healthy individuals.
Risk factors of gastric cancer.
Area under the curve values for carcinoembryonic antigen (CEA), cancer antigen 72-4 (CA72-4), pepsinogen I, Helicobacter pylori, matrix metalloproteinase 7 and gastric cancer index (GCI).
Receiver operating characteristic curves for carcinoembryonic antigen (CEA), cancer antigen 72-4 (CA72-4), pepsinogen I, Helicobacter pylori, matrix metalloproteinase 7 (MMP7) and gastric cancer index (GCI).
OPG and OPN are known due to their role in bone metabolism (7). OPG is mainly associated with bone metastasis of several types of cancer including gastric cancer (8). Several studies have found that increased OPN levels are related to a poor prognosis of gastric cancer (9, 10). In our group of patients, OPN level was found to be statistically significantly higher in the group of patients with gastric cancer compared to the group of healthy individuals. This is consistent with the findings in the studies mentioned above.
PIVKA-II was first described as a biomarker for the detection of hepatocellular carcinoma (11). The production of PIVKA-II by gastric cancer has also been described (12). We also confirmed PIVKA-II positivity in several cases in our group of patients.
In the next part of our work, we tested the risk factors of gastric cancer according to the findings in literature (13). We confirmed the important role of pepsinogen I, gastrin and Helicobacter pylori as risk factors in gastric cancer development. Pepsinogen I and Helicobacter pylori were also selected as predictors of gastric cancer by our mathematical model.
Currently, we can say that efforts to discover new biomarkers to assist in the early diagnosis of gastric cancer have not produced much success as would be desirable. Our pilot study confirms this fact. The best results (p-value, ROC and AUC) were achieved by the traditional tumor markers CA72-4 and CEA. For some cancer diagnoses there is currently a tendency to use statistical methods to increase the efficiency of the diagnostic process through introduction of the parameters calculated from the values of measured biomarkers. For example, this approach has proven to be effective in the diagnosis of prostate cancer using the prostate health index (PHI) (14) and in the risk of ovarian malignancy algorithm (ROMA) in the case of ovarian cancer (15). Therefore, we decided to apply a new approach to the search for a tool that might be used to help make an early diagnosis of gastric cancer. Using multivariate analysis might lead us more quickly to a solution of the diagnostic problem presented by gastric cancer and increase the chances of successful treatment of these patients. Using multivariate analysis, we selected five biomarkers whose results appear to be most effective for gastric cancer diagnostics and used them to build a mathematical model for the calculation of gastric cancer risk. The GCI is the result of our approach. When using GCI, we achieved the best ROC curve and the highest AUC. In a future study, we shall focus on verifying the contribution of individual parameters and on the confirmation of, and improvements to the algorithm for calculating the GCI.
In conclusion, tumor markers CEA and CA72-4 remain the best individual markers for gastric cancer diagnostics. We identified MMP1, MMP7, OPN and PIVKA-II as good candidates for additional biomarkers for gastric cancer detection. We confirmed lower pepsinogen I levels, higher gastrin levels and the presence of Helicobacter pylori as risk factors of gastric cancer development. We have introduced a mathematical model for the assessment of the risk of gastric cancer in the human population. A further study will be performed to confirm these findings incorporating a larger number of patients.
Acknowledgements
This study was supported by IGA grant project IGA NT 14227-3.
- Received January 25, 2016.
- Revision received March 10, 2016.
- Accepted March 15, 2016.
- Copyright© 2016 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.