Introduction

Endoscopic submucosal dissection (ESD) has been widely used for the treatment of early gastric cancer (EGC). ESD is a standard treatment for differentiated-type adenocarcinoma without ulceration, of which the depth of invasion is clinically diagnosed as T1a (tumor invasion up to muscularis mucosa) and the diameter is ≤2 cm (Japanese gastric cancer treatment guidelines 2010 (ver. 3) [1]). Resection of EGC by ESD is judged as curative when all seven of the following conditions are fulfilled: en-block resection, HM0, VM0, tumor size of two cm or smaller, histologically differentiated type, pT1a, and no lymphovascular (LV) invasion (Japanese Gastric Cancer Treatment Guidelines 2010 (ver. 3) [1]). The first three conditions rely heavily on the technical skill of ESD operators. The subsequent two conditions can be determined to some extent by using an endoscopic measuring tool and multiple biopsies before ESD. However, pT1a and the absence of LV invasion can only be confirmed by a pathologist following ESD.

LV invasion is considered an independent risk factor for lymph node (LN) metastases [2–5] and a critical prognostic factor [6] in patients with EGC. Sekiguchi et al. [7] identified submucosal (SM) invasion as an independent risk factor for LV invasion in endoscopically resected gastric cancer. Despite extensive efforts [8, 9], correct diagnosis of the invasion depth, prior to ESD remains a challenge. When SM/LV invasion is detected following careful pathological examination of endoscopically resected specimens, further surgical treatment, such as gastrectomy with LN dissection, is needed. The clinicopathological characteristics of SM/LV invasions in EGC (presenting as an endoscopically good indication for ESD) are still largely unknown. Therefore, we conducted a retrospective analysis to elucidate the risk factors for SM/LV invasion in EGC.

Methods

We reviewed clinical records of patients who underwent ESD at the University of Tsukuba Hospital and Tsukuba Memorial Hospital, between 2007 and 2012, and collected data on EGCs that were ≤2 cm, without ulceration, and possessing a differentiated-type adenocarcinoma.

Resected specimens were pinned to a board and fixed in 10 % formalin overnight; they were then serially cut to produce 2–3 mm thick sections and evaluated by hematoxylin-eosin staining, which was performed as described elsewhere. When LV invasion was suspected, immunohistochemical examination was performed with the D2-40 antibody (mouse monoclonal; Clone D2-40; Nichirei, Tokyo, Japan) and Elastica van Gieson stain to identify lymphatic and blood vessels, respectively. Anti-desmin staining (mouse monoclonal; Clone D33; Dako, Tokyo, Japan) was similarly employed to evaluate SM invasion. All specimens were diagnosed at the Tsukuba Human Tissue Diagnostic Center of University of Tsukuba Hospital, according to the Japanese Classification of Gastric Carcinoma (3rd edition) [10]. Mixed types of differentiated and undifferentiated carcinoma were classified according to the quantitative predominance defined in the Japanese Gastric Cancer Treatment Guidelines 2010 (ver. 3) [1].

Patient records were reviewed retrospectively for age, sex, as well as for the following clinicopathological parameters: cancer location, synchronous or metachronous multifocality, pathological diagnosis of resected specimens, including tumor size (defined as the largest diameter of the lesion), gross morphology (categorized histologically according to the macroscopic classification of EGC), dominant histological differentiation type, presence or absence of LV invasion, and depth of invasion. The primary endpoint of this study is to find the risk factors of SM/LV invasion, which requires gastrectomy.

Statistical analysis was performed using IBM SPSS Statistics for Windows (version 21.0). To elucidate risk factors for SM/LV invasion, univariate analysis was conducted using Fisher’s exact test and the Mann–Whitney U test. Potential risk factors revealed by the univariate analysis with P value of less than 0.2 were then subjected to multivariate logistic regression analysis in order to identify independent risk factors for SM/LV invasion. P values of <0.05 were considered statistically significant.

This study was approved by the Ethics Review Board of the University of Tsukuba Hospital (Study number: H24-154), and was performed according to the Japanese Guidelines for Epidemiological Study.

Results

A flow chart depicting how lesions of EGC resected by ESD were selected is shown in Fig. 1. Of a total of 208 EGC lesions resected by ESD, 65 lesions (>2 cm, 50 lesions; undifferentiated, 3 lesions; Ul+, 12 lesions) were excluded, while 143 lesions in 132 patients were included in this study. Of these 132 patients, 107 were male and 25 were female; the mean age was 71 ± 7.7 years (ranging from 50 to 86 years). Mean tumor size was 1.1 ± 0.5 cm (ranging from 0.2 to 2.0 cm). 0-IIa (or dominant) was the most common gross morphological type (36.4 %), followed by 0-IIc (or dominant) (32.2 %). Well-differentiated tubular adenocarcinoma (tub1) was the main histological type (84.6 %), with the remaining lesions classified as either moderately differentiated tubular adenocarcinoma (tub2) (14.0 %) or papillary adenocarcinoma (pap) (1.4 %). Out of a total 143 tumors, 14 lesions were SM invasive (9.8 %; T1b), and 9 lesions were considered LV invasive (6.3 %; 2 showing lymphatic invasion, 4 showing vascular invasion, and 3 showing both types of invasion). These findings were fairly comparable between the two hospitals.

Fig. 1
figure 1

Flow chart illustrating the selection process for lesions of early gastric cancers resected by endoscopic submucosal dissection (ESD). Lesions of ≤2 cm in tumor size, exhibiting a histologically differentiated type of adenocarcinoma, and without ulceration were used in this study

Full size image

Nine of 16 patients with SM/LV invasive EGC underwent additional radical surgical treatments. LN metastasis was found in one patient with LV invasion, while no residual cancer was detected in the remaining eight patients. Seven patients who did not undergo additional treatments revealed no recurrence of EGC during 10.5 months of median follow-up (ranging from 3 to 43 months).

The clinicopathological data of the lesions, with or without SM/LV invasion, are summarized in Table 1. Univariate analysis revealed significant differences in tumor size (P = 0.034) and dominant histological type (P < 0.001) between EGCs with and those without SM/LV invasion. Mean tumor size was 1.4 cm in the case with SM/LV invasion whereas 1.0 cm without invasions. The cutoff point for tumor size with the highest sensitivity and specificity for SM/LV invasion was set at 1.5 cm, based on ROC curve analysis which revealed that tumors ≥1.5 cm in size were significantly associated with SM/LV invasions (P = 0.030). 50 % (8/16) of SM/LV invasive cases possessed dominant histology of tub2 or pap whereas 89 % (113/127) of cases without SM/LV invasions were tub1 dominant. Additionally 31 % of cases with SL/LV invasion were morphologically 0-IIa + IIc or 0-IIc + IIa whereas 15 % of cases without SM/LV invasions were 0-IIa + IIc or 0-IIc + IIa.

Table 1 Characteristics of submucosal or lymphovascular (SM/LV) invasive early gastric cancers in 143 ESD lesions which were ≤2 cm in tumor size and a histologically differentiated type of adenocarcinoma without ulceration
Full size table

Multivariate analysis was performed on tumor size (≥1.5 vs. <1.5 cm), pathological diagnoses (tub2 or pap vs. tub1), tumor number (solitary vs. multiple tumors), gross type (0-IIa + IIc or IIc + IIa vs. other types), and location (UM vs. L). The results of this analysis demonstrated that dominant histology of tub2 or pap [odds ratio (OR) 11; 95 % confidence interval (CI) 2.9–42; P < 0.001], gross type of 0-IIa + IIc or 0-IIc + IIa (OR 4.8; 95 % CI 1.3–24.0; P = 0.031), and tumor size of ≥1.5 cm (OR 3.6; 95 % CI 1.0–13.0; P = 0.042) were independent risk factors for SM/LV invasion (Table 2). Nine of 19 lesions (47 %) exhibiting two or more risk factors were identified as SM/LV invasive. We could enrich SM/LV invasions using two or more risk factors at an odds ratio of 15 (95 % CI 4.6–49.0; P < 0.0001).

Table 2 Multivariate analysis of potential risk factors for submucosal or lymphovascular invasive early gastric cancers
Full size table

Discussion

Recently, ESD has been undertaken for expanded lesions beyond the conventional indication [1]. The majority of previous reports on the risk factors for SM/LV invasion included all lesions resected by ESD, resulting in differing histological backgrounds among the resected lesions which may have influenced the results and obscured the study goal, i.e., distinguishing lesions that could be curatively resected by ESD. In the present study, we have made an effort to address this problem by restricting analysis to differentiated-type adenocarcinomas that lack ulceration and are ≤2 cm in size, characteristics considered strongly indicative for ESD. Using this approach, three independent risk factors were identified for SM/LV invasion.

A dominant histology of tub2 or pap compared to tub1 was revealed as the highest risk factor for SM/LV invasion. Sekiguchi et al. [7] reported that the presence of the pap component in endoscopically resected gastric cancer was an independent risk factor for lymphatic invasion, and all lesions with the pap component linked with venous invasion showed deep SM invasion. Several reports [11, 12] have suggested that a mixed histological type of EGC with undifferentiated components is a risk factor for LN metastasis; interestingly, the tub2 component is often associated with the mixed histological type [13]. Five lesions of the mixed type included in the present study were all characterized as tub2-dominant mixed type. Yet, among these lesions, only two had been previously identified as belonging to the mixed type in pre-ESD biopsies, and the remaining lesions contained tub2 without the undifferentiated component. Hence, we need to consider the possible presence of undifferentiated adenocarcinoma, even when biopsies reveal a tub-2 histology, but not the mixed type.

Our identification of the gross type of 0-IIa + IIc or 0-IIc + IIa as the second independent risk factor was unexpected. Previously, only one report suggested that the non-flat gross type was associated with node-positive submucosal gastric cancer [14]. Tumor size, when ≥1.5 cm, was the third independent risk factor for SM/LV invasion, supporting similar findings by others [3].

Employing these three risk factors, we stratified SM/LV high-risk lesions. Taking into account all of the lesions included in the present study, the proportion of lesions positive for SM/LV invasion was 11.2 % (16/143). In contrast, when only considering resected lesions with two or more risk factors, 47 % (9/19) were SM/LV invasive. This value dropped to 5.6 % (7/124) in lesions manifesting less than two of these risk factors. These results have the potential to be helpful in determining further surveillance before ESD. Endoscopic ultrasonography (EUS) is valuable for predicting tumor invasion depth [15] and enhanced computed tomography (eCT) is used for the evaluation of LN metastasis. However there are no guidelines for whom these examinations should be recommended before ESD. We propose a possible strategy on this matter as shown in Fig. 2. Consequently, this approach would be useful when obtaining their informed consent.

Fig. 2
figure 2

Possible strategy for surveillance before ESD. EUS endoscopic ultrasonography, eCT enhanced computed tomography

Full size image

Of a total 143 lesions, 34 lesions (23.8 %) were comprised of multiple cancers, including 24 lesions (16.8 %) of synchronous and 10 lesions (7.0 %) of metachronous multiple cancers (Table 1). Although not statistically significant, SM/LV invasion was absent in synchronous multiple cancers (0/24), and only one lesion with metachronous multiple cancers demonstrated SM/LV invasion (1/10). These data lend support to a report showing that multiple EGCs were less likely to exhibit LV invasion than solitary EGCs [16]. Conversely, others have suggested that synchronous EGC has a similar risk of LN metastasis as solitary EGC [17, 18]. Further studies are therefore needed to corroborate the relationship between multiple EGCs and reduced incidence of SM/LV invasion, especially in our defined cohort.

One of the limitations of this study is a small number of events. Our result should be validated in a prospective way with a larger cohort. We need to improve our methods for endoscopically calculating tumor size and for diagnosing dominant histological type more accurately by using multiple biopsies. However, it is worth noting that biopsy findings do not necessarily reflect the actual dominant histology of ESD-resected specimens. For example, although 36 lesions in our study were identified as containing the tub2 component in biopsies taken prior to treatment, only 14 of them (39 %) were diagnosed as possessing tub2-dominant histology in ESD-resected specimens. Similarly, the diagnostic accuracy of using multiple biopsies is limited by the inherent histological heterogeneity of cancer tissues. One report showed the possibility of preoperative endoscopic diagnosis of papillary adenocarcinoma by magnifying endoscopy with narrow-band imaging [19]. This might be a method to overcome the issue of histology prediction in the biopsy specimens. We therefore need to develop further methods to more accurately predict the likelihood of SM/LV invasion prior to ESD.

In conclusion, we have identified three independent risk factors (pap- or tub2-dominant histology, 0-IIa + IIc or IIc + IIa, and tumor size ≥1.5 cm) for SM/LV invasion of EGC, which might otherwise seem to be an endoscopically good indication for ESD.