Patients and lesions
“Crawling-type” adenocarcinomas were collected from a total of 691 cases of pT1 primary gastric cancer in patients without any preceding treatments. Patients included in this study underwent surgical intervention at Tokyo Medical and Dental University Hospital from January 2002 to July 2011, or underwent endoscopic submucosal dissection at Kudanzaka Hospital from October 2003 to July 2011. We examined 606 surgically resected and 270 endoscopically resected specimens, including simultaneous or metachronous multiple lesions, from these patients. After formalin fixation and paraffin embedding, specimens from the surgically resected samples were cut into long strips 5 mm wide and those from endoscopically resected samples were cut into long strips 2 mm wide. Some sections (3-μm-thick) were stained with hematoxylin–eosin and elastica van Gieson to evaluate histological findings, and other sections were used for immunohistochemistry (IHC). Histological diagnosis of CTAC was made according to the criteria proposed by Takizawa et al. [10, 11]. Adenocarcinomas including at least some of the CTAC components were obtained for this study. Two independent observers (N.O. and H.K.) evaluated the histopathological findings and IHC results. In any disagreements between the two observers, the observers re-evaluated the specimens and reached a consensus after discussion.
This study was performed in accordance with the World Medical Association’s Declaration of Helsinki. The ethics committees of Tokyo Medical and Dental University and Kudanzaka Hospital approved the protocols.
Clinicopathological findings and histological evaluation of CTAC
Patients’ age, sex, therapeutic method, tumor location, macroscopic type including the presence of ulcerative changes, and LNM status were obtained from pathology records in each hospital. For M-CTACs, tumor size was evaluated as the maximum diameter, whereas for SM-CTACs the maximum diameters of both the intramucosal area and the submucosal invasive area were evaluated.
All lesions were classified as M- or SM-cancer based on the depth of invasion. Further, in SM-cancers, the vertical distances from the muscularis mucosa to the invasive front were measured. Lesions less than 500 μm away from the muscularis mucosa were subclassified as SM1, and lesions 500 μm or more away from the muscularis mucosa were subclassified as SM2, according to the Japanese classification of gastric carcinoma . Lymphatic and venous permeations were graded as positive or negative in hematoxylin–eosin-stained sections and in elastica van Gieson-stained sections, respectively.
The stromal volume and tumor infiltration pattern were evaluated according to the Japanese classification of gastric carcinoma  as follows. The stromal volume is classified into three categories: medullary-type (med), which has scanty stroma; scirrhous-type (sci), which has abundant stroma; and intermediate-type (int), which is intermediate between med and sci. The tumor infiltration pattern was also classified into three categories: INFa, in which tumors display expanding growth with a distinct border separating them from the surrounding tissue; INFc, in which tumors display an infiltrative growth pattern with no distinct border; and INFb, in which tumors show an intermediate pattern between INFa and INFc.
Condition of the surrounding mucosa
The mucosa surrounding the lesions was examined histologically. Both proximal and distal sides were divided into fundic gland mucosa, pyloric gland mucosa, and intermediate zone mucosa. The degree of intestinal metaplasia (IM) was classified as diffuse IM, partial IM, or without IM.
Sections from one or two representative paraffin blocks of each case were prepared for immunohistochemical staining. We used monoclonal antibodies to human gastric mucin (HGM; clone 45M1, Novocastra, Newcastle upon Tyne, UK, diluted 1:50), MUC6 glycoprotein (MUC6; clone CLH5, Novocastra, diluted 1 : 50), MUC2 glycoprotein (MUC2; clone CCP58, Novocastra, diluted 1:100), CD10 protein (CD10; clone 56C6, Novocastra, diluted 1:100), Cdx2 protein (Cdx2; clone AMT28, Novocastra, diluted 1:50), p53 protein (p53; clone DO-7, Novocastra, diluted 1:1000), and Ki-67 antigen (Ki-67; clone MIB-1, Dako, Glostrup, Denmark, diluted 1:800). All sections were incubated with a primary antibody at its working dilution for 60 min at room temperature (HGM and Cdx2) or for 24 h at 4 °C (MUC6, MUC2, CD10, p53, and Ki-67), subsequent to antigen retrieval. Antigen retrieval was performed using a microwave at 97 °C for 40 min (HGM and Cdx2), or an autoclave at 121 °C for 20 min (MUC6, MUC2, CD10, p53, and Ki-67). Sections were stained to detect each antigen using a Vectastain ABC Immunoperoxidase kit (Vector Laboratories, Burlington, CA, USA) or an EnVision+ System (Dako, Glostrup, Denmark). The sections were incubated in 3,3′-diaminobenzidine for 10 min, which stained the antigen brown, and then counterstained with Mayer’s hematoxylin.
Phenotypic classification based on the presence of mucin, brush border, and Cdx2 expression
We classified lesions as having one of four phenotypes: gastric phenotype (G-phenotype), in which 10 % or more of the cancer cells were immunoreactive for HGM and/or MUC6, fewer cells were stained for MUC2 and CD10, and cancer cells were weakly positive or negative for Cdx2; intestinal phenotype (I-phenotype), in which 10 % or more of the cancer cells were immunoreactive for MUC2 and/or CD10, or strongly positive for Cdx2, and also in which fewer cells were stained for HGM and MUC6; mixed phenotype, in which 10 % of the cancer cells were immunoreactive for HGM and/or MUC6, and were also immunoreactive for at least one of MUC2, CD10, and Cdx2; and null phenotype (N-phenotype), in which fewer than 10 % of the cancer cells were immunoreactive for any of the five antigens. Further, the mixed phenotype was classified into two subgroups as follows: mixed gastric phenotype (MG-phenotype), in which cancer cells were predominantly immunoreactive for HGM and/or MUC6; and mixed intestinal phenotype (MI-phenotype), in which cancer cells were predominantly immunoreactive for MUC2 and/or CD10, and/or Cdx2. These definitions are based on the definitions of Egashira et al. , and according to the results of Mizoshita et al. [15, 16].
Quantitative evaluation of intramucosal poorly differentiated component (PDC)
The PDC was defined as an intramucosal component composed of signet-ring cell carcinoma and/or poorly differentiated adenocarcinoma, without tubular or papillary structures. The maximum diameter of the PDC was measured and used for quantitative evaluation.
Distribution pattern of Ki-67-positivity
The lesions were divided into two groups, as follows, based on the distribution pattern of the Ki-67-positive cancer cells in the intramucosal area: limited pattern, Ki-67-positive cancer cells showed zonal distribution and occupied 50 % or less of the mucosa; and expanded pattern, Ki-67-positive cells showed zonal distribution and occupied more than 50 % of the mucosa, or showed diffuse distribution.
The p53 overexpression was defined as positive when intensely stained cancer cells were distributed diffusely. Weak and scattered staining was considered negative. The definition was based on the definitions of Fukunaga et al.  and Gabbert et al. .
Fisher’s exact probability test, or the Wilcoxon signed-ranks test were used for analysis of the differences between groups. Differences with P values of less than 0.05 were considered to be statistically significant. StatView software (version 5.0; SAS Institute, Cary, NC, USA) was used for the statistical analyses.