Clinical and Histopathological Features of Early Gastric Cancer With Unclear Lateral Demarcation

Background Magnifying-endoscopy with narrow band imaging (M-NBI) is useful to determine lateral demarcation of early gastric cancers, but determining the lateral demarcation is sometimes dicult. Features related to the unclear lateral demarcation remain unknown. We evaluated the clinical and histopathological features of early gastric cancers with unclear lateral demarcation by M-NBI. This single-center retrospective cohort study analyzed early gastric cancer treated by endoscopic submucosal dissection (ESD) from January 2013 to August 2015. We evaluated clinicopathological and immunohistochemical features using anti-p53, -Ki-67, -MUC5AC, -MUC6, -MUC2, and -CD10 antibody staining. We compared the lateral demarcation between the demarcation clear (DC) and demarcation unclear (DU) lesions by using M-NBI.


Introduction
Various modalities have been developed for the pre-procedural diagnosis of lateral demarcation of early gastric cancers. In particular, magnifying-endoscopy with narrow band imaging (M-NBI) is useful for preprocedural diagnosis, because it provides good visualization for both super cial mucosal structure and blood vessel architecture. [1][2][3] The clear determination of its lateral demarcation is very important for a complete en-bloc resection by endoscopic submucosal dissection (ESD), but poor determination can sometimes occur. It is well known that undifferentiated adenocarcinomas show unclear lateral demarcation compared to differentiated adenocarcinomas. [4,5] We sometimes misdiagnose the lateral demarcation even in differentiated adenocarcinomas. Recently, the Helicobacter pylori eradication therapy has spread worldwide; therefore, early gastric cancers are sometimes detected after successful eradication. [6,7] Some reports show that early gastric cancers that developed after eradication therapy sometimes show an unclear lateral demarcation.
[8] Regarding their features, they are reported to have a normal columnar epithelium on the surface of the tumor, a mucin phenotype marker of gastric or gastrointestinal type, and a low Ki-67 labeling index (LI). However, any features related to the unclear lateral demarcation remain unknown. The aim of this study was to evaluate the clinicopathological and immunohistochemical features of differentiated adenocarcinomas with unclear lateral demarcation by M-NBI.

Study design
We conducted a retrospective cohort study of early gastric cancers treated by ESD from January 2013 to August 2015 in our institution. Well or moderately differentiated adenocarcinomas diagnosed by two pathologists during the period were examined. Adenomas, undifferentiated or undifferentiated-mixed cancers, and H. pylori-unrelated gastric cancers, such as fundic gland type cancers, were excluded. All study participants provided written informed consent, and the study was approved by the institutional review board (IRB) and ethics comittee of Osaka university hospital (IRB number: 16451). No patients were under 20 years old. This study was conducted in accordance with the principles laid down in the Helsinki Declaration.

Tumor classi cation
We performed the endoscopic diagnosis according to the vessel plus surface classi cation system by M-NBI. [3] The demarcation line (DL) separates a lesion from the normal surrounding mucosal area. If DL was present all around the lesion, we classi ed the lesion into the demarcation clear (DC) group. If DL was absent for a part of the lesion, we categorized the lesion into to the demarcation unclear (DU) group. All lesions with still images were evaluated at a conference attended by more than six endoscopic experts and divided into two groups (DC and DU).

Clinical examination
We analyzed the differences in clinical features of early gastric cancers, including size, location, macroscopic type, color, and history of H. pylori eradication, between the two groups. We de ned successful H. pylori eradication therapy, if the patients had a history of H. pylori eradication therapy and received at least one negative test of the following methods: tests for the presence of serum IgG antibodies against H. pylori, the 13 C urea breath test, and stool test for H. pylori antigen. We evaluated the presence of H. pylori at the time of lesion detection. In addition, we compared the short-term outcome of treatment, including the rate of en-bloc or curative resection, procedure time, the rate of complication, such as perforation and delayed bleeding, and the rate of additional surgical resection between the two groups. En-bloc resection was de ned as one-piece resection endoscopically, whereas curative resection was de ned as en-bloc resection histopathologically with the absence of lymphovascular invasion and presence of tumor negative lateral and vertical margins of the resected specimens according to the guideline of the indication criteria for curative resection. [9] Perforation was de ned as a defect of the muscular layer during ESD or presence of free air detected by computed tomography after ESD. Delayed bleeding was de ned as hematochezia after ESD that required endoscopic hemostasis. [10] Histological examination Formalin-xed and para n-embedded specimens from the included lesions were retrieved. All hematoxylin and eosin (HE)-stained sections were evaluated by two pathologists. Histological type, combined ulcer or ulcer scar, and depth of invasion were evaluated. Additionally, we evaluated the histopathological features on the super cial epithelium and divided them into the following three categories: non-neoplastic super cial epithelium, partial non-neoplastic super cial epithelium, and neoplastic super cial epithelium. Partial non-neoplastic super cial epithelium was de ned as the presence of both non-neoplastic and neoplastic epithelia.

Evaluation of histopathological structure
The randomly selected visual eld (100 ⋅) in both the lesion and the background mucosa from one tissue was assessed. One of the authors evaluated the width of the interfoveolare, length of duct, and width of pit to compare the difference in the histopathological structure between the lesion and the surrounding mucosa in both groups.

Evaluation of immunohistochemical staining
We selected one tissue section with the largest cancerous area for each resected lesion. The randomly selected visual eld (100⋅) in each lesion was assessed and scored. A brown stain was regarded as positive. The expressions of p53 and Ki-67 were quanti ed using image processing software (WinROOF, ver.5.7.2; Mitani Corp., Tokyo, Japan). This software measures the color intensity and transforms the color selected into a percentage in each eld. We calculated the intensity of staining at the threshold of color and the extension of staining at the percentage of positive cells against total cancer cells. The p53 immunostaining was analyzed semi-quantitatively using a scoring system for both intensity (0, no staining; 1, weak staining; 2, moderate staining; 3, strong staining) and extension (0, no unclear staining; 1, < 10% staining; 2, 10%-50% staining; 3, > 50% staining). The score for intensity and extension was summed for each lesion. We classi ed a score of > 4 as marked staining for p53 according to a previous report. [12] Ki-67 LI was calculated as the percentage of positive cells against total cancer cells. We classi ed the lesions into a high group if Ki-67 LI was greater than the median value, which was also according to a previous report. [13] The mucin expression of tumor cells was examined with gastric phenotype markers, such as MUC5AC and MUC6, and intestinal phenotype markers, such as MUC2 and CD10. The results of each phenotype marker were de ned as positive if > 10% of the tumor cells were stained. Tumors were classi ed by phenotype as gastric (G type), intestinal (I type), gastrointestinal (GI type), and null (N type); G type tumors expressed only a gastric phenotype marker, whereas I type tumors expressed only an intestinal phenotype marker. GI type tumors expressed both gastric and intestinal phenotype markers, whereas N type tumors were negative for all markers. [14] We evaluated the expression of each mucin phenotype marker not only in the tumor but also in the surrounding mucosa.

Statistical methods
Categorical variables were presented as frequencies and proportions. They were compared between the two groups by using Fisher's exact test. Continuous variables were presented as medians (interquartile range [IQR]). They were compared between the two groups by using Wilcoxon rank sum test. A p-value < 0.05 was considered signi cant. To reveal the factors associated with unclear lateral demarcation by M-NBI, univariate and multivariate logistic regression analyses were performed. All statistical analyses were performed with JMP statistical software ver. 14 (SAS Institute Inc, Cary, NC).

Clinicopathological features
A total of 265 tumors were treated by ESD from January 2013 to August 2015 at our institution. Among them, 41 lesions, including 15 adenomas, 22 undifferentiated or mixed cancers and four H. pyloriunrelated gastric cancers, including three fundic gland type and one pyloric gland type lesion, were excluded. A total of 224 differentiated adenocarcinomas were analyzed in this study. As a result of the evaluation, 18 and 206 lesions were divided into the DU and DC groups, respectively (Fig. 1).
We compared the clinical characteristics between the DU and DC groups (Table 1). There were no differences in size, location, macroscopic type, color, histological type, combined ulcer or ulcer scar, and depth of invasion. The history of successful H. pylori eradication was signi cantly more frequent in the DU group than in the DC group (p = 0.001). In the short-term clinical outcome of the treatment, the en-bloc resection rate was lower in the DU group than in the DC group because of the positive horizontal margin (p = 0.033). There were no differences in the procedure time and complication rate between the two groups. To reveal the cause of unclear lateral demarcation of early gastric cancers in the DU group, we evaluated the histopathological structures between the cancerous lesion and the surrounding mucosa. The DU group had no differences in width of interfoveolare (p = 0.987), length of duct (p = 0.351), and width of pit (p = 0.296) between the lesion and the surrounding mucosa, whereas the DC group had signi cant differences in width of interfoveolare (p < 0.01), length of duct (p < 0.01), and width of pit (p = 0.018) between them (Fig. 2a-c).

Evaluation of histopathological features on super cial epithelium
We considered that a history of successful H. pylori eradication might in uence the unclear lateral demarcation by M-NBI. We randomly selected 36 lesions with a history of successful H. pylori eradication and 18 lesions without such history in the DC group with reference to the frequency of a history of successful H. pylori eradication in the DU group. A total of 72 lesions, including 18 lesions in the DU group and 54 lesions in the DC group, underwent histopathological and immunohistochemical examinations.
Then, we evaluated the features of the super cial epithelium on the cancerous lesions between the DU and DC groups. Representative images of the partial and non-neoplastic super cial epithelia are shown in Fig. 3a. The DU group was more likely to have a non-neoplastic super cial epithelium than the DC group (p = 0.0058) (Fig. 3b). There was no difference in the frequency of partial non-neoplastic super cial epithelium between the two groups.

Evaluation of mucin phenotype
Next, we performed the immunohistochemical staining regarding the mucin phenotype marker between the DU and DC groups (Fig. 4a). We classi ed the mucin phenotype marker into the following two groups: gastric phenotype marker, including G and GI type, and non-gastric phenotype marker, including I and N type. The DU group had a higher expression of gastric phenotype marker in the cancerous lesion than the DC group (p = 0.023) (Fig. 4b). In the surrounding mucosa, there was no signi cant difference in the mucin phenotype marker between the two groups.

Evaluation of p53 positivity
We performed p53 staining to evaluate the difference in the oncogenic potential. The proportion of the lesions with low p53 score was signi cantly higher in the DU group than in the DC group (p = 0.0002) (Fig. 5a, b).

Evaluation of proliferation activity
We examined the Ki-67 LI to evaluate the difference in the proliferation activity between the two groups. The proportion of the lesions with low Ki-67 LI (p = 0.0293) was higher in the DU group than in the DC group (Fig. 5c, d).

Factors associated with unclear lateral demarcation by M-NBI
Multivariate analysis revealed that the presence of non-neoplastic super cial epithelium and low p53 score were independent variables associated with unclear lateral demarcation by M-NBI (Table 2). often present with a crypt epithelium and MUC5AC positivity, and goblet cells with MUC2 positivity in the mucin phenotype marker rarely appeared. In fact, there were no tumor cells on the super cial surface, and non-neoplastic epithelium with a gastric phenotype marker was replaced. Therefore, we considered this phenomenon was a regeneration after anti-in ammation.
The non-neoplastic super cial epithelium, gastric phenotype marker, low p53 score, and low Ki-67 LI were signi cant factors associated with unclear lateral demarcation by M-NBI in our histopathological and immunohistochemical analyses. In particular, non-neoplastic super cial epithelium and low p53 score were found to be independent and signi cant factors in the multivariate logistic regression analyses. We considered that these factors interfere in obtaining accurate endoscopic ndings. We considered that the DU group has a small difference in the histological feature between the lesion and the surrounding mucosa because of the wide interfoveolare, long tumor duct, and open pit. We presumed that these ndings are due to the fact that there are few tumor cells on the super cial surface, and tumor cells develop in the deep layers of the mucosal lamina propria.
In our study, 72 lesions, including 18 lesions in the DU group and 54 lesions in the DC group, underwent immunohistochemical examinations. The number of lesions in the DU group was small; thus, we randomly selected 54 lesions in the DC group with reference to the frequency of the history of successful H. pylori eradication in the DU group. We could not retrospectively evaluate the reason for undergoing H. pylori eradication; thus, we selected the objective lesions using a random number list, instead of using a propensity score.
The limitations of this study were its retrospective and single-center design and the small number of lesions in the DU group.

Conclusions
In conclusion, our study demonstrated that unclear lateral demarcation by M-NBI occurs when the tumor cells developed in the deep layers of the mucosal lamina propria. In addition, these lesions are characterized by the presence of a gastric phenotype marker, low p53 score, and low Ki-67 LI. Nonneoplastic super cial epithelium and low p53 score were associated with the di cultly in determining lateral demarcation in early gastric cancers by M-NBI. The evaluation of clinicopathological and immunohistochemical features associated with unclear lateral demarcation can facilitate the performance of appropriate treatment for these lesions.    Difference in the mucin phenotype markers (MUC5AC, MUC6, MUC2, CD10) between the DU and DC groups. a. Comparison of the mucin phenotype marker in the lesion and the surrounding mucosa between the DU and DC groups. b. Comparison of gastric type in the lesion and the surrounding mucosa between the DU and DC groups. G, gastric type; GI, gastrointestinal type; I, intestinal type; N, null type; DU, demarcation unclear; DC, demarcation clear