Clinical Cancer Research CTRC-AACR San Antonio Breast Cancer Symposium Infection and Cancer: Biology, Therapeutics, and Prevention
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online
Clinical Cancer Research 13, 4154-4163, July 15, 2007. doi: 10.1158/1078-0432.CCR-07-0173
© 2007 American Association for Cancer Research
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Lee, H. S.
Right arrow Articles by Kim, W. H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Lee, H. S.
Right arrow Articles by Kim, W. H.

Imaging, Diagnosis, Prognosis

Protein Expression Profiling and Molecular Classification of Gastric Cancer by the Tissue Array Method

Hye Seung Lee1, Sung-Bum Cho2, Hee Eun Lee3, Min A Kim3, Ji Hun Kim3, Do Joong Park4, Ju Han Kim2, Han-Kwang Yang4,5, Byung Lan Lee5 and Woo Ho Kim3,5

Authors' Affiliations: 1 Department of Pathology, Seoul National University Bundang Hospital, Gyeonggi, and 2 Seoul National University Biomedical Informatics, Departments of 3 Pathology and 4 Surgery, and 5 Cancer Research Institute, Seoul National University College of Medicine, Seoul, Korea

Requests for reprints: Woo Ho Kim, Department of Pathology, Seoul National University College of Medicine, 28 Yeongeon-dong, Seoul 110-799, Korea. Phone: 82-2740-8269; Fax: 82-2765-5600; E-mail: woohokim{at}snu.ac.kr.


   Abstract
Top
 Abstract
Materials and Methods
 Results
 Discussion
 References
 
Purpose: Gastric cancer is heterogeneous clinically and histologically, and prognosis prediction by tumor grade or type is difficult. Although previous studies have suggested that frozen tissue–based molecular classifications effectively predict prognosis, prognostic classification on formalin-fixed tissue is needed, especially in early gastric cancer.

Experimental Design: We immunostained 659 consecutive gastric cancers using 56 tumor-associated antibodies and the tissue array method. Hierarchical cluster analyses were done before and after feature selection. To optimize classifier number and prediction accuracy for prognosis, a supervised analysis using a support vector machine algorithm was used.

Results: Of 56 gene products, 27 survival-associated proteins were selected (feature selection), and hierarchical clustering identified two clusters: cluster 1 and cluster 2. Cluster 1 cancers were more likely to have intestinal type, earlier stage, and better prognosis than cluster 2 (P < 0.05). In 187 early gastric cancers (pT1), cluster 2 was associated with the presence of metastatic lymph nodes (P = 0.026). Kaplan-Meier survival curves stratified by pathologic tumor-lymph node metastasis revealed that cluster 2 was associated with poor prognosis in stage I or II cancer (P < 0.05). Support vector machines and genetic algorithms selected nine classifiers from the whole data set, another nine classifiers for stage I and II, and eight classifiers for stage III and IV. The prediction accuracies for patient outcome were 73.1%, 88.1%, and 76%, respectively.

Conclusions: Protein expression profiling using the tissue array method provided a useful means for the molecular classification of gastric cancer into survival-predictive subgroups. The molecular classification predicted lymph node metastasis and prognosis in early stage gastric cancer.

The incidence and mortality of gastric cancer have declined steadily over the past several decades. Nonetheless, gastric cancer remains a major public health issue as the fourth most common cancer and the second leading cause of cancer death worldwide (1, 2). Gastric cancer is a heterogeneous disease both histologically and genetically, and patient outcome is difficult to predict using classic histologic and molecular classifications. Many histopathologic classifications, including histologic type and grade by WHO (3), Lauren's classification (4), Ming's classification (5), and Goseki classification (6) have been applied for the prediction of patient survival, but their usefulness remains controversial (3, 7).

The incidence of small and early gastric cancer is high in Asia because the increased usage of upper endoscopy has led to the earlier detection of lesions (8). However, despite curative resection of the primary tumor, some early gastric cancer patients succumb to the disease as a result of local or distant tumor recurrence. Adjuvant chemotherapy benefits some patients with early gastric cancer, but it is not necessary in all patients. Therefore, additional markers are required to identify those patients at risk of recurrence or poor prognosis. Recently, large-scale molecular techniques such as DNA microarrays have contributed to our understanding of the molecular complexity of gastric cancer and prognostic classification according to gene expression profile has been achieved using frozen tissue (9, 10). However, in small and early gastric cancer patients, the detection and sampling of proper cancer tissue using gross examination is difficult. In addition, the cost, complexity, and interpretation of DNA microarrays are currently unsuitable for routine use in standard clinical settings. Therefore, prognostic classification on formalin-fixed paraffin-embedded tissue is required, especially in small and early gastric cancer.

Many candidate gene products for the prediction of patient survival have been reported in gastric cancer (11, 12). Genetic alterations including those of p53, MUC1, CEA, E-cadherin, p16, and CD44 have been reported to play important roles in the development and progression of the disease (13–18). Although much has been learned of the genetic factors predicting survival, only a few genetic alterations have been used for the diagnosis and management of patients with gastric cancer. Recently, large-scale molecular studies on formalin-fixed tissue have become possible by tissue array method using immunohistochemical approach. Such large-scale studies involve a relatively large number of markers in addition to a large number of cases. Moreover, combined or cluster analysis using multiple markers has been reported to be significantly correlated with patient survival (19, 20). Jacquemier et al. identified a set of 21 proteins using supervised analysis in breast cancer, the expressions of which were significantly correlated with metastasis-free survival (20).

In this study, we immunostained 659 consecutive gastric cancers using 56 tumor-associated antibodies and the tissue array method. Gastric cancer was subclassified using classic hierarchical clustering before and after feature selection. To obtain the optimal number of classifiers and prediction accuracy for patient outcome, we did supervised analysis using a support vector machine (SVM) algorithm.


   Materials and Methods
Top
 Abstract
 Materials and Methods
Results
 Discussion
 References
 
Specimens. A total of 659 consecutive, surgically resected cases of primary gastric cancer treated over a period of 1 year were identified in the files of the Department of Pathology, Seoul National University College of Medicine (Seoul, Korea). Age, sex, tumor location, lymphatic invasion, vascular invasion, and pathologic tumor-lymph node metastasis (pTNM) stage (21) were evaluated by reviewing medical charts and pathologic records. The mean age of the 659 patients was 54.7 years, and 93% underwent curative resection (R0 according to the American Joint Committee on Cancer guideline). The study included 438 men and 221 women, and included 436 advanced and 223 early gastric cancers. No patient had received preoperative chemotherapy or radiotherapy. Tissue slides were reviewed for histologic classifications (according to WHO and Lauren's classifications; refs. 3, 4). Patient clinical outcome was followed-up from the date of surgery up to a period of 1 to 72 months (mean, 52 months). The cases lost to follow-up and deaths from any other cause other than gastric cancer were regarded as censored data for the analysis of survival rates. This study was approved by the Institutional Review Board for Human Subject Research at Seoul National University Hospital.

Tissue array methods. Twelve array blocks containing a total of 659 cases were prepared as described previously (Superbiochips Laboratories; refs. 19, 22). Core tissue biopsies (2 mm in diameter) were taken from individual paraffin-embedded gastric tumors (donor blocks) and arranged in recipient paraffin blocks (tissue array blocks) using a trephine. As it has previously been proven that staining results obtained from different intratumoral areas in various tumors agree well (22–25), a core was sampled in each case. An adequate case was defined as a tumor occupying >10% of the core area. Each block contained three internal controls consisting of nonneoplastic gastric mucosa from the body, antrum, and intestinal metaplasia. Four-micrometer-thick sections were cut from each tissue array block, deparaffinized and dehydrated.

Immunohistochemistry. Immunohistochemical staining against tumor-associated gene products was done using a streptavidin peroxidase procedure following an antigen retrieval process using microwaves or autoclaves. Commercially available antibodies were tested using a human control slide and a stomach cancer control slide for immunohistochemistry (Superbiochips Laboratories). After the test procedure, 56 antibodies, which were properly stained in each positive and negative control, were selected for this study. Table 1 and Fig. 1 list the antibodies used. The expression status of nonneoplastic gastric glands and altered expression patterns in gastric cancer are described in Table 1. For statistical analysis of collected data, immunostaining results were considered positive when ≥10% of neoplastic cells were stained (19, 22). Immunoreactivity was assessed microscopically by two independent pathologists unaware of the clinical details of individual patients. The case-by-case final consensus result was discussed and determined in a common session.


View this table:
[in this window]
[in a new window]

  		Table 1. Antibodies used for immunohistochemical study

 

Figure 1
View larger version (159K):
[in this window]
[in a new window]

  		Fig. 1. Representative expressions of proteins studied by immunohistochemistry (original magnification, x400).

 
Cluster analyses. Expression data were recorded as follows: –1 was designated negative staining, 1 was designated positive staining, missing data were left blank in the data table. In this study, hierarchical cluster analyses were done using the Cluster program (complete linkage clustering) and results were displayed using TreeView (26). All cases with missing values in >20% of the columns were excluded from the cluster analyses. We could perform an unsupervised analysis on 56 gene expressions in 601 gastric cancer samples (91.2%) of 659 consecutive samples. Of these 56 genes, 27 genes which showed significant correlations with patient outcome using the Kaplan-Meier survival method were selected (feature selection). After feature selection, we did a cluster analysis using these 27 gene expression profiles in 614 samples (93.2%) of 659 consecutive samples.

Supervised analysis using SVM and genetic algorithm. SVM and genetic algorithm were used to identify protein classifier sets associated with survival status (27). SVM is well known for its competency among off-the-shelf classification algorithms and its performance comes from mapping of the original feature space (data space) to the higher dimensional space. The hyperplane obtained in the higher dimensional space generally shows excellent classification performance. In addition to the classification power of the SVM algorithm, selection of the input variables was important to the performance of the classification algorithm. However, it is too computationally intensive to regard all the possible combinations of 56 genes. Therefore, the genetic algorithm was applied to select an optimal subset of tumor-associated genes that yielded the best result for classification of the patient's survival status. The genetic algorithm finds an optimal solution by simulating the natural genetic selection procedure. At each step of the gene selection, prediction accuracy with 10-fold cross-validation was calculated. Cross-validation is a widely used method for assessing the performance of a classification algorithm. It separates the whole data set into a training and test data set. For example, a 10-fold cross-validation assigns 9/10 of the whole data into the training set, and 1/10 of the whole data into the test set. When the training and test sets are assigned, the classification model is extracted from the training set and the performance of the model is examined in the test set. This process is repeated until all instances of the data set are assigned to the test set at once. By combining the SVM and genetic algorithm, the classifier consisting of a subset of 56 tumor-associated genes yielding the best classification performance with a 10-fold cross-validation was extracted. All computations were done using the R program package (e1071 package for SVM and genalg package for genetic algorithm; ref. 28).

Statistical analyses. Either the {chi}2 test or Fisher's exact test (two-sided) was done to determine the correlation between gene expression status and clinicopathologic variables. Survival curves were estimated using the Kaplan-Meier product-limit method, and the significance of the differences between survival curves were determined using a log-rank test. Multivariate survival analyses were done using the Cox proportional hazards model. The association between clustering and regional lymph node metastasis was evaluated by multivariate logistic regression. Results were considered to be statistically significant for P < 0.05. All statistical analyses were conducted using SPSS 12.0 statistical software program (SPSS).


   Results
Top
 Abstract
 Materials and Methods
 Results
Discussion
 References
 
Expression profiling of 56 proteins and hierarchical cluster analysis. The staining results of the 56 antibodies are summarized in Table 2 . By Kaplan-Meier analyses, the expression status of 27 proteins were found to be significantly associated with patient survival (P < 0.05). The overall expression patterns for 601 samples of gastric cancer were analyzed by hierarchical clustering after excluding those with values missing in >20% of the columns. The combined protein expression patterns defined two clusters: cluster A (24 cases) and cluster B (577 cases). Cluster B was subdivided into three clusters, cluster B1 (57 cases), cluster B21 (140 cases), and cluster B22 (380 cases; Fig. 2A and B ). Cluster B22 cases tended to have better survival than clusters A, B1, or B21 (P < 0.0001; Fig. 2C). Multivariate analysis, including pTNM stage (II-IV versus I) and the molecular classification (cluster B22 versus clusters A + B21 + B1), showed that the molecular classification was an independent prognostic indicator of survival (hazard ratio, 0.654; 95% confidence interval, 0.498-0.858; P = 0.002).


View this table:
[in this window]
[in a new window]

  		Table 2. Expression rates of the 56 proteins in gastric cancer and 5-y survival rates

 

Figure 2
View larger version (34K):
[in this window]
[in a new window]

  		Fig. 2. Classification of 601 gastric cancers based on the expression of the 56 proteins. A, matrix format presenting the whole data. In data matrix (upper), a row corresponds to a single tumor, and each column corresponds to a single protein biomarker. Colors in columns represent expression levels: negative expression (green), positive expression (red), and missing data (gray). Horizontal bars (lower) correspond to the clusters. B, protein dendrogram. C, univariate survival analysis by Kaplan-Meier method.

 
Hierarchical cluster analysis of gastric cancer after feature selection. Twenty-seven survival-associated proteins were selected after univariate survival analyses of the 56-protein expression. A total of 614 gastric cancer cases were analyzed by hierarchical clustering analysis after excluding those with values missing in >20% of the columns. Tumors were separated into two main branches (Fig. 3A ). The right branch, i.e., cluster 1, consisted of 356 cases and was characterized by intestinal type (according to Lauren's classification; P = 0.002), lower lymph node metastasis (P < 0.001), and left pTNM stage (P < 0.001) compared with the lower branch, i.e., cluster 2, which consisted of 256 cases (Table 3 ). Of the 614 gastric cancers, 187 cancers were limited to the mucosal or submucosal levels (pT1). In the 187 early gastric cancers, cluster 2 cancers were more likely to have metastatic lymph nodes (P = 0.026). By multivariate logistic regression analysis, cluster 2 was significantly correlated with lymph node metastasis independently of lymphatic invasion and depth of invasion (P = 0.027; Table 4 ).


Figure 3
View larger version (33K):
[in this window]
[in a new window]

  		Fig. 3. Classification of 614 gastric cancers based on the expression of the 27 survival-associated proteins after feature selection. A, matrix format presenting the data. Gastric cancer was divided into two subgroups, clusters 1 and 2. B, protein dendrogram. C, univariate survival analysis by Kaplan-Meier method. D, Kaplan-Meier survival curve in stage I cancers. E, Kaplan-Meier survival curve in stage II cancers.

 

View this table:
[in this window]
[in a new window]

  		Table 3. Clinicopathologic characteristics of the two clusters by hierarchical cluster analysis with 27 survival-associated proteins

 

View this table:
[in this window]
[in a new window]

  		Table 4. Prediction of lymph node metastasis in 187 early gastric cancers

 
Using a protein dendrogram, 27 survival-associated proteins were classified into three groups (Fig. 3B). Group 1 consisted of proteins overexpressed in gastric cancer, e.g., CD24, hexokinase II, smad7, MUC1, S100A4, and CEA, whereas group 3 consisted of suppressor proteins, e.g., APC, smad4, KAI-1, MGMT, PML, PTEN, and E-cadherin.

To determine whether these two clusters might represent clinically distinct subgroups of patients, univariate survival analysis was done. Cluster 2 was associated with poorer prognosis compared with cluster 1 (P < 0.0001; Fig. 3C), and multivariate analysis, including pTNM stage and the above cluster-based molecular classification, showed that the classification was an independent prognostic indicator of survival (P = 0.013; Table 5 ). Kaplan-Meier survival curves stratified by pTNM stage (stages I-IV) revealed that cluster 2 was associated with a probability of lower survival for patients with stage I or II cancer (P = 0.0005 and 0.0020, respectively; Fig. 3D and E).


View this table:
[in this window]
[in a new window]

  		Table 5. Multivariate analysis for factors predictive of survival (Cox proportional hazards model)

 
Supervised analysis using the SVM algorithm. After 1,000 iterations, the genetic algorithm selected nine protein classifiers (CEA, c-fos, caspase-1, c-kit, cytokeratin6, cytokeratin19, S100A4, HDAC1, and DNA-PKcs) which showed the best classification performance compared with other combinations of tumor-associated genes in the whole data set. The prediction accuracy of these classifiers for patient survival was 73.1% (sensitivity, 36.7%; specificity, 94.7%). When 421 samples of stage I and II tumors were analyzed, the prediction accuracy was 88.1% for nine classifiers (MUC2, MUC6, smad4, PTEN, MGMT, TCF4, rad9, cytokeratin8, and S100A6). The sensitivity of these classifiers in stage I and II tumors decreased (30%), whereas the specificity increased (99.3%). The molecular classifiers for the data set of stage III and IV tumors (n = 220) were p53, p16, KAI1, TCF4, IRS-1, cytokeratin6, GST-p, and osteonectin. The prediction accuracy for patient survival was reduced to 76% compared with that of the classifiers in stage I and II tumors. Sensitivity and specificity for prediction of patient survival showed different features. Sensitivity increased sharply (95.5%) and specificity decreased (33.8%). Results are listed in Table 6 .


View this table:
[in this window]
[in a new window]

  		Table 6. Prediction for patient outcome of molecular classifications by supervised analysis using the SVM algorithm

 

   Discussion
Top
 Abstract
 Materials and Methods
 Results
 Discussion
References
 
Gastric cancer is a heterogeneous disease histologically and genetically. Histologically, it is subdivided into intestinal and diffuse types by Lauren classification (4), but some gastric cancers cannot be easily classified in this way because these two types are frequently admixed within single tumors. Moreover, it is controversial as to whether Lauren's classification is an independent prognostic factor. Genetically, many earlier studies have reported genetic alterations during gastric carcinogenesis and progression, e.g., p53 mutations (29), microsatellite instability (30), EBV infection (31), CpG island methylation (32), and chromosomal instability (33). Thus, gastric cancer is associated with various genetic alterations and no single genetic marker can predict gastric cancer biology or prognosis. The potential use of combinations of biomarkers instead of a single marker or histologic feature has been previously commented upon (19, 22, 34). Moreover, high-throughput analysis based on formalin-fixed tissues is possible using the tissue array method. Recent studies have applied tissue array methods to the molecular classifications of various cancers such as breast and brain tumors (20, 35), and protein expression profiling has been found to be clinically useful for the prognostic classification of neoplasms. The present study shows that a molecular classification of gastric cancer can be accomplished based on hierarchical cluster analysis of the immunohistochemical profiles of tumor-associated biomarkers using tissue array sections.

The tissue array method enabled us to immunostain 659 formalin-fixed gastric cancer specimens with 56 tumor-associated antibodies. Moreover, using the large amount of data generated, various hierarchical cluster analyses could be done. The overall protein expression patterns defined four survival-associated clusters, i.e., cluster A (24 cases), cluster B1 (57 cases), cluster B21 (140 cases), and cluster B22 (380 cases). Patients with cluster B22 cancers had the best prognosis among the four clusters (P < 0.0001). However, 56 markers are too many for practical use. To more accurately predict patient survival, and to reduce the number of markers, we did a feature selection using Kaplan-Meier survival analyses. Twenty-seven survival-associated proteins were selected as described in Materials and Methods. Tumors were separated into two main branches by hierarchical clustering, i.e., cluster 1 (356 cases) and cluster 2 (256 cases), and these two clusters were found to have distinct clinicopathologic features and patient outcomes.

Recently, the incidence, and thus the importance, of small and early gastric cancers have increased because of the increased use of upper gastrointestinal endoscopy (8). Minimal resection and surgery such as endoscopic mucosal resection and laparoscopic surgery have been increasingly used to treat early gastric cancer (36). Moreover, the prediction of lymph node metastasis in endoscopic mucosal resection specimens is clinically important and the prediction of patient survival in early gastric cancer specimens would allow accurate decision-making concerning postoperative management. In the present study, Kaplan-Meier survival curves that were stratified according to pTNM revealed that cluster 2 was significantly associated with poor survival for stage I and II cancers (P = 0.0005 and 0.0020, respectively). Our study contained 187 cancers that were limited to the mucosa or submucosa (pT1). Cluster 2 cancers were also more associated with the presence of metastatic lymph nodes in 187 early gastric cancers (pT1; P = 0.026), independently of lymphatic invasion (P = 0.035), indicating that the molecular classification may be applied to the prediction of lymph node metastasis or prognosis in early gastric cancer.

Using the protein dendrogram, 27 survival-associated proteins were classified into three groups. Whereas group 1 consisted of proteins overexpressed in gastric cancer, group 3 consisted of proteins lost in gastric cancer. In a previous study of breast cancer (20), a protein dendrogram revealed four major protein clusters, i.e., estrogen receptor–associated proteins, a differentiation cluster, a mitosis cluster, and a proliferation cluster. However, in the present study, the protein groups were not well defined, which is probably due to the more heterogeneous nature of gastric cancer.

In this study, the classifications based on the 27 selected proteins were significantly associated with clinicopathologic features and prognosis. However, 27 proteins are too many for routine use. To obtain an optimal number of protein classifiers for the accurate prediction of patient outcome, 1,000 iterations were done using the genetic algorithm, and as a result, nine classifiers were chosen in the whole data set with a prediction accuracy for patient outcome of 73.1%, sensitivity of 36.7%, and specificity of 94.7%. When 421 stage I and II samples were analyzed, another nine classifiers were chosen and prediction accuracy was 88.1%, sensitivity was 30%, and specificity was 99.3%. The identified classifiers for the stage III and IV samples (n = 220) were p53, p16, KAI1, TCF4, IRS-1, cytokeratin6, GST-p, and osteonectin, and these had a prediction accuracy of 76%, sensitivity of 95.5%, and a specificity of 33.8%. These classifier numbers were found to be optimal for practical use in pathology laboratories. There was a large gap between the sensitivity and specificity of the classification analysis. This result was due to our analysis scheme. In this analysis, we used the genetic algorithm to select an optimal subset of tumor-associated genes that yielded the highest prediction accuracy with a 10-fold cross-validation. During selection, only the prediction accuracy was considered. The selected classifier genes therefore generated the optimal prediction accuracy with the sacrifice of sensitivity or specificity. Even if the sensitivity was low, the classifier genes from stage I and II data could be applied as indicators for good prognosis because the specificity is high. In cases with classifiers from stage III and IV, they could be used as the screening test for unfavorable outcome because the sensitivity is high. Further validation is needed concerning the use of the above classifiers for routine pathologic diagnoses.

In summary, we immunostained 659 consecutive gastric cancers with 56 tumor-associated antibodies using the tissue array method. With 27 survival-associated gene products, the hierarchical cluster analysis identified two clusters with different clinicopathologic features and prognoses. Moreover, the molecular classification predicted lymph node metastasis and prognosis in early stage gastric cancer. To optimize classifier numbers and accurately predict patient outcome, we did the supervised analysis using the SVM algorithm. The genetic algorithm selected nine classifiers in the whole data set, another nine classifiers in stages I and II, and eight classifiers in stages III and IV. Prediction accuracies for the patient outcomes of these classifiers were 73.1%, 88.1%, and 76%, respectively.


   Acknowledgments
 
We thank S.P. Kim, S.H. Chung, and Superbiochip Laboratories (Seoul, Korea) for their technical assistance.


   Footnotes
 
Grant support: 21C Frontier Functional Human Genome Project grant FG06-11-03 from the Ministry of Science and Technology of Korea.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received 1/21/07; revised 3/23/07; accepted 4/ 6/07.


   References
Top
 Abstract
 Materials and Methods
 Results
 Discussion
 References
 

  1. Crew KD, Neugut AI. Epidemiology of gastric cancer. World J Gastroenterol 2006;12:354–62.[Medline]
  2. Landis SH, Murray T, Bolden S, Wingo PA. Cancer statistics, 1998. CA Cancer J Clin 1998;48:6–29.[Abstract]
  3. Fenoglio-Preiser C, Carneiro F, Correa P, et al. International Agency for Research on Cancer (IARC). World Health Organization classification of tumors; pathology and genetics of tumors of the digestive system. Lyon (France): IARC Press; 2000. p. 37–52.
  4. Lauren P. The two histological main types of gastric carcinoma: diffuse and so-called intestinal-type carcinoma. An attempt at a histo-clinical classification. Acta Pathol Microbiol Scand 1965;64:31–49.[Medline]
  5. Ming SC. Gastric carcinoma. A pathobiological classification. Cancer 1977;39:2475–85.[CrossRef][Medline]
  6. Goseki N, Takizawa T, Koike M. Differences in the mode of the extension of gastric cancer classified by histological type: new histological classification of gastric carcinoma. Gut 1992;33:606–12.[Abstract/Free Full Text]
  7. Lewin KJ, Appleman HD. Carcinoma of the stomach. Tumors of the esophagus and stomach. Atlas of Tumor Pathology, 3rd series, Fascicle 18. Washington DC, Armed Forces Institute of Pathology, 1996. p. 245–330.
  8. Noguchi Y, Yoshikawa T, Tsuburaya A, Motohashi H, Karpeh MS, Brennan MF. Is gastric carcinoma different between Japan and the United States?: A comparison of patient survival among three institutions. Cancer 2000;89:2237–46.[CrossRef][Medline]
  9. Chen CN, Lin JJ, Chen JJ, et al. Gene expression profile predicts patient survival of gastric cancer after surgical resection. J Clin Oncol 2005;23:7286–95.[Abstract/Free Full Text]
  10. Hippo Y, Taniguchi H, Tsutsumi S, et al. Global gene expression analysis of gastric cancer by oligonucleotide microarrays. Cancer Res 2002;62:233–40.[Abstract/Free Full Text]
  11. Fuchs CS, Mayer RJ. Gastric carcinoma. N Eng J Med 1995;333:32–41.[Free Full Text]
  12. Tahara E. Genetic alterations in human gastrointestinal cancers. The application to molecular diagnosis. Cancer 1995;75:1410–7.[CrossRef][Medline]
  13. Gabbert HE, Muller W, Schneiders A, Meier S, Hommel G. The relationship of p53 expression to the prognosis of 418 patients with gastric carcinoma. Cancer 1995;76:720–6.[CrossRef][Medline]
  14. Utsunomiya T, Yonezawa S, Sakamoto H, et al. Expression of MUC1 and MUC2 mucins in gastric carcinomas: its relationship with the prognosis of the patients. Clin Cancer Res 1998;4:2605–14.[Abstract]
  15. Kim DY, Kim HR, Shim JH, Park CS, Kim SK, Kim YJ. Significance of serum and tissue carcinoembryonic antigen for the prognosis of gastric carcinoma patients. J Surg Oncol 2000;74:185–92.[CrossRef][Medline]
  16. Yonemura Y, Nojima N, Kaji M, et al. E-cadherin and urokinase-type plasminogen activator tissue status in gastric carcinoma. Cancer 1995;76:941–53.[CrossRef][Medline]
  17. Feakins RM, Nickols CD, Bidd H, Walton SJ. Abnormal expression of pRb, p16, and cyclin D1 in gastric adenocarcinoma and its lymph node metastases: relationship with pathological features and survival. Hum Pathol 2003;34:1276–82.[CrossRef][Medline]
  18. Washington K, Gottfried MR, Telen MJ. Expression of the cell adhesion molecule CD44 in gastric adenocarcinomas. Hum Pathol 1994;25:1043–9.[CrossRef][Medline]
  19. Lee HS, Lee HK, Kim HS, Yang HK, Kim WH. Tumour suppressor gene expression correlates with gastric cancer prognosis. J Pathol 2003;200:39–46.[CrossRef][Medline]
  20. Jacquemier J, Ginestier C, Rougemont J, et al. Protein expression profiling identifies subclasses of breast cancer and predicts prognosis. Cancer Res 2005;65:767–79.[Abstract/Free Full Text]
  21. American Joint Committee on Cancer. AJCC cancer staging manual 5th ed. Philadelphia: Lippincott-Raven; 1997.
  22. Lee HS, Lee HK, Kim HS, Yang HK, Kim YI, Kim WH. MUC1, MUC2, MUC5AC, and MUC6 expressions in gastric carcinomas: their roles as prognostic indicators. Cancer 2001;92:1427–34.[CrossRef][Medline]
  23. Nocito A, Bubendorf L, Tinner EM, et al. Microarrays of bladder cancer tissue are highly representative of proliferation index and histological grade. J Pathol 2001;194:349–57.[CrossRef][Medline]
  24. Torhorst J, Bucher C, Kononen J, et al. Tissue microarrays for rapid linking of molecular changes to clinical endpoints. Am J Pathol 2001;159:2249–56.[Abstract/Free Full Text]
  25. Zhang D, Salto-Tellez M, Putti TC, Do E, Koay ES. Reliability of tissue microarrays in detecting protein expression and gene amplification in breast cancer. Mod Pathol 2003;16:79–84.[CrossRef][Medline]
  26. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 1998;95:14863–8.[Abstract/Free Full Text]
  27. Duda RO, Hart PE, Stork DG, Pattern classification (hardcover). 2nd Ed. Wiley Interscience; 2000.
  28. Comprehensive R archives network (CRAN). Available from: http://cran.r-project.org/.
  29. Kim JH, Takahashi T, Chiba I, et al. Occurrence of p53 gene abnormalities in gastric carcinoma tumors and cell lines. J Natl Cancer Inst 1991;83:938–43.[Abstract/Free Full Text]
  30. Keller G, Rudelius M, Vogelsang H, et al. Microsatellite instability and loss of heterozygosity in gastric carcinoma in comparison to family history. Am J Pathol 1998;152:1281–9.[Abstract]
  31. Shibata D, Tokunaga M, Uemura Y, Sato E, Tanaka S, Weiss LM. Association of Epstein-Barr virus with undifferentiated gastric carcinomas with intense lymphoid infiltration. Lymphoepithelioma-like carcinoma. Am J Pathol 1991;139:469–74.[Abstract]
  32. Toyota M, Ahuja N, Suzuki H, et al. Aberrant methylation in gastric cancer associated with the CpG island methylator phenotype. Cancer Res 1999;59:5438–42.[Abstract/Free Full Text]
  33. Furuya T, Uchiyama T, Murakami T, et al. Relationship between chromosomal instability and intratumoral regional DNA ploidy heterogeneity in primary gastric cancers. Clin Cancer Res 2000;6:2815–20.[Abstract/Free Full Text]
  34. Beenken SW, Grizzle WE, Crowe DR, et al. Molecular biomarkers for breast cancer prognosis: coexpression of c-erbB-2 and p53. Ann Surg 2001;233:630–8.[CrossRef][Medline]
  35. Ikota H, Kinjo S, Yokoo H, Nakazato Y. Systematic immunohistochemical profiling of 378 brain tumors with 37 antibodies using tissue microarray technology. Acta Neuropathol (Berl) 2006;111:475–82.[CrossRef][Medline]
  36. Ono H, Kondo H, Gotoda T, et al. Endoscopic mucosal resection for treatment of early gastric cancer. Gut 2001;48:225–9.[Abstract/Free Full Text]




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Lee, H. S.
Right arrow Articles by Kim, W. H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Lee, H. S.
Right arrow Articles by Kim, W. H.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online