Introduction

According to reports by Gotoda et al. [1] and Hirasawa et al. [2], undifferentiated-type early gastric cancers (UD-type EGCs) have a negligible risk of lymph node metastasis so long as they measure 20 mm or less, are without ulcer scars, and have not invaded the lymph or blood vessels; consequently, they fall into the category of extended indications for endoscopic submucosal dissection (ESD), as proposed by the Japanese Gastric Cancer Association (JGCA) [3]. However, it should be noted that these studies used surgical specimens.

UD-type EGCs mainly comprise poorly differentiated adenocarcinoma (PDA) and signet ring cell carcinoma (SRC). Differences between these carcinomas have occasionally been reported with regard to surgical cases. For instance, comparisons between surgical specimens of mixed PDA and pure SRC have revealed higher frequencies of lymph node metastasis and submucosal invasion in the former [4,5,6]. Treatment outcomes and prognoses of UD-type EGCs after endoscopic treatment have also been studied, and the suitability of endoscopic treatment in these cases has been discussed [7,8,9]. However, mixed PDA and pure SRC have not been studied separately. Although there have been some reports on the endoscopic treatment of UD-type EGCs, none has addressed what differences exist between mixed PDA and pure SRC as histological factors and whether UD-type EGC with mixed PDA or with pure SRC predicts curative or noncurative resection.

Previous reports on surgical cases [4,5,6] suggest that pure SRC manifests a lesser degree of malignancy than mixed PDA, but the results may differ between histological preparations of surgical specimens that contain the full thickness of the gastric wall (sectioned at 5-mm intervals) and those from endoscopic resection, which are limited to the submucosal layer of the stomach (sectioned at 2-mm intervals).

In this study, we compared UD-type EGCs treated with curative resection with those treated with noncurative resection and examined whether histological type would have an influence on curative or noncurative resection, as has been observed in surgical cases. In addition, only tumors conforming to the following criteria specified by the JGCA guidelines [3] were studied: diameter of 20 mm or less, intramucosal carcinoma, negative lymphovascular invasion, negative ulcer scars, negative vertical margin, negative horizontal margin, and R0 resection. These cases were also examined for other factors that might predict curative or noncurative resection. In addition, the lesions were divided into mixed PDAs and pure SRCs, and the difference in the rate of curative resection between these two groups was analyzed.

Patients and methods

Patients

A total of 3419 patients with early gastric cancers were treated with ESD in our hospital between March 2005 and March 2017. Among these patients, there were 268 pure UD-type EGCs in 264 (7.3%) patients, excluding remnant cancers (because of difficulty in technique), differentiated carcinomas, and mixed histological-type carcinomas (including predominantly undifferentiated mixed type). The 268 UD-type EGCs comprised 229 lesions treated with curative resection and 39 treated with noncurative resection. In addition, there were 129 pure SRCs and 139 mixed PDAs. The cases of mixed PDA included 8 cases of pure PDA devoid of SRC. Pure PDAs were included with mixed PDAs in our analysis because the purpose of this study was to compare lesions with and without PDA in accordance with the aforementioned previous studies of surgical cases [4,5,6].

Methods

This was a single-center retrospective study. First, all undifferentiated carcinomas were categorized by patient age, sex, lesion location, macroscopic type, and treatment outcomes of ESD (maximum tumor diameter, tumor depth, histological type, lymphovascular invasion, horizontal margin, vertical margin, ulcerative findings, en bloc resection, R0 resection, rate of curative resection, and adverse events). Following this, cases of curative and noncurative resection were compared. Cases of curative and noncurative resection were examined separately for pure SRC and mixed PDA to identify any differences in the rate of curative resection between the two groups. All patients with UD-type EGCs treated with ESD underwent preoperative conventional endoscopy, dye-spraying endoscopy, and narrow-band imaging (NBI)–combined magnifying endoscopy to diagnose the extent and depth of the tumor. According to the guidelines issued by the JGCA [3], cases with intramucosal carcinomas measuring 20 mm or less and without ulcerative findings were preoperatively determined as meeting the criteria for expanded indications for ESD, which was performed accordingly. For each lesion, biopsies from four sites surrounding the lesion were confirmed to be negative. For each resected ESD specimen, sections were prepared at 2-mm intervals for pathological evaluation. Pathological examination was always performed by pathologists specialized in gastroenterology. In accordance with previous studies in surgical cases [4,5,6], specimens consisting of SRC alone were defined as pure SRCs, and SRC lesions were defined as mixed PDAs if even a minimal PDA component was present. Endpoints were maximum tumor diameter, invasion depth, histological type, ulcerative findings, lymphovascular invasion, horizontal margin, and vertical margin. Curative resection was defined as the presence of intramucosal UD-type EGCs measuring 20 mm or less in diameter without ulcerative findings, and with negative horizontal margins, negative vertical margins, and no lymphovascular invasion, in accordance with the JGCA gastric cancer treatment guidelines [3].

All patients undergoing treatment for UD-type EGCs were given a detailed explanation of the advantages and disadvantages of ESD before providing their informed consent. Consensus has already been established in our institution regarding the expanded indication  of ESD to UD-type EGCs.

Statistical analysis

Fisher’s exact test was used for comparison of the two groups. The mean values and standard deviations of age and tumor diameter were analyzed using the t test and the F test. The Mann–Whitney U test was used for unequal variances with median. When comparing noncurative and curative cases, factors that were found to have significant differences on univariate analysis were subjected to multivariate analysis (logistic regression analysis).

Statistical significance was set at p < 0.05. StatView software, version 5.0 (SAS Institute, Cary, NC, USA), was used for statistical processing.

Results

Table 1 shows the overall patient characteristics and treatment results in cases with UD-type EGCs. The overall rate of curative resection was 85.4% (229/268). A comparison of patient characteristics between cases of curative resection and noncurative resection (Table 2) revealed that in cases of curative resection the patients were of significantly younger age. Flat-, depressed-, or elevated-type lesions accounted for macroscopic type in the majority of cases, and the flat type was significantly more common in cases of curative resection. A comparison of the treatment results (Table 3) showed that factors determining curative and noncurative resection, as specified in the JGCA guidelines (i.e., tumor diameter, submucosal invasion, lymphovascular invasion, vertical margin, and horizontal margin), were all lower in cases of curative resection than in cases of noncurative resection, and the rates of en bloc resection and R0 resection were significantly higher in cases of curative resection. With regard to histological type, mixed PDA and pure SRC were significantly more frequent in cases of noncurative and curative resection, respectively. To identify factors that predict noncurative resection, other than those specified by the JGCA guidelines for determining curative and noncurative resection, multivariate analysis was performed for those factors that showed a significant difference on univariate analysis (age, flat-type lesion, number of patients with mixed PDA) (Table 4). Compared with cases of curative resection, cases of noncurative resection showed a significant difference for the number of cases with patient age, and the number of mixed PDAs, with odds ratios of 1.052 (95% CI, 1.017–1.089), and 2.746 (95% CI, 1.162–6.485), respectively. When the actual rates of curative resection for mixed PDA and pure SRC were compared, the rate was 77.7% for mixed PDA whereas it was significantly higher, at 93.8%, for pure SRC (Table 5). Thus, among cases of noncurative resection of UD-type EGC, cases with advanced patient age and mixed PDA were significantly more frequent.

Table 1 Patient characteristics and results of endoscopic treatment in undifferentiated-type gastric cancer cases
Full size table
Table 2 Comparison of patient characteristics between cases of curative and noncurative resection
Full size table
Table 3 Comparison of results of endoscopic treatment between cases of curative and noncurative resection
Full size table
Table 4 Multivariate analysis of predictive factors for noncurative resection in undifferentiated-type adenocarcinoma
Full size table
Table 5 Comparison of age and other factors determining curative resection in cases of mixed poorly differentiated adenocarcinoma and pure signet ring cell carcinoma
Full size table

Discussion

The rate of curative resection with ESD for UD-type EGCs varies widely among different institutions, ranging from 63.9% to 82.5% [7, 8, 10,11,12,13,14,15]. In this study, overall the rate of curative resection of UD-type EGCs was 85.4%. This study compared cases of curative and noncurative resection in relationship to their histological type (pure SRC or mixed PDA) to identify factors predictive of curative or noncurative resection, in addition to known predictive factors (tumor diameter, invasion depth, ulcerative findings, lymphovascular invasion, horizontal margin, vertical margin, and R0 resection).

In a comparison of cases of curative and noncurative resection, multivariate analysis revealed a significant difference in age, flat-type lesions, and presence of mixed PDA. Thus, flat-type lesions were identified as a predictor of curative resection, and advanced age and the presence of PDA as predictors of noncurative resection.

The patients with noncurative resection were significantly older. Patients of advanced age often have more underlying diseases than younger patients and are in poorer general condition [16, 17]. It has also been reported that the 5-year survival rate after surgical gastrectomy is lower in patients aged 75 years or older than in young or middle-aged patients [18]. Therefore, it is sometimes undesirable to proceed with aggressive surgical treatment in elderly patients, and it is suggested that complete endoscopic biopsy be initially attempted as a means of diagnostic therapy, particularly when it is difficult to make a definitive judgment in diagnosing the depth of invasion. This detail may explain why patients with noncurative resection were of significantly higher ages.

As for the significantly higher percentage of mixed PDA, previous studies comparing pure SRCs and mixed PDAs in surgical specimens have shown that the rates of lymph node metastasis and biological malignancy were higher in mixed PDAs [4,5,6]. In contrast, studies using surgical specimens of early gastric cancer have found that SRC has a better prognosis [6, 19,20,21,22,23]. It has also been reported that SRC progresses from intramucosal carcinoma to PDA at the site of invasion [24]. A recent report has also documented that submucosal invasion was more frequent in PDAs than in SRCs treated with endoscopy [25]. A higher frequency of positive vertical margins in PDAs has also been reported [26]. However, in these reports of endoscopic treatment, there was no detailed analysis concerning whether SRCs were pure SRCs or whether they were SRCs with mixed PDA components. In this study, we defined the SRC group as pure SRCs alone, and cases that were predominantly SRC but contained a mixed PDA component were included in the mixed PDA group. Our results suggested that the presence of a PDA component was generally a predictor of noncurative resection. When the actual rates of curative resection were compared, the percentage was 93.8% among pure SRCs, whereas for mixed PDAs it was significantly lower at 77.7%. Among the factors determining noncurative resection, tumor diameter was significantly greater and submucosal invasion and positive vertical margin were more frequent, consistent with findings from previous reports [25, 26]. Additionally, ulcerative findings were also more common. Noncurative resection is defined as the risk of local recurrence or lymph node recurrence at a later stage, unless additional surgical resection is performed. Therefore, specimens from endoscopic resection also predicted the high biological malignancy of mixed PDA.

In this study, there were 129 pure SRCs, accounting for 48.1% of all UD-type cancers, and the rate of curative resection was 93.8%. The rate of curative resection is reported to be 91.1% among cases of differentiated-type early gastric cancer measuring 20 mm or less [27], which is higher than the reported rates of curative resection of UD-type EGCs (63.9–82.5%) [7, 8, 10,11,12,13,14,15]. Preoperative biopsy involves obtaining a small portion of the lesion. Therefore, even if the preoperative biopsy specimen is indicative of SRC, this does not mean the whole tumor is SRC. However, if a preoperative biopsy specimen is indicative of mixed PDA, the lesion is classified as mixed PDA. Mixed PDA has a lower curative resection rate than SRC.

In the future, it will be helpful to distinguish between mixed PDA and SRC before endoscopic treatment for UD-type EGCs. It will be important to explain the differences to patients to allow them to choose endoscopic treatment based on a sufficient understanding of the possibility of additional surgical resections.

However, the results of this study suggest that the rate of curative resection of pure SRCs is comparable to that of differentiated cancer. To date, there have been no reports stating that pure SRC is distinguishable from mixed PDA before endoscopic treatment. Further studies are necessary to investigate differences in endoscopic findings between pure SRCs and mixed PDAs.

The limitations of this study are its retrospective nature and single-nstitution setting. However, the subjects of this study represent an accumulation of clinical cases treated in a cancer-specialized hospital over a period of approximately 12 years, enabling a sufficiently meaningful investigation. Because our institution is a cancer-specialized hospital, there might have been a selection bias when comparing with the general prevalence rate, taking into consideration the frequency of patient referral. However, as the study included all patients who underwent endoscopic treatment, the bias is likely to be minimal. When patients are referred to our hospital, a preoperative biopsy will have already been taken at their previous hospital. If cancer is diagnosed based on this biopsy, we will not conduct a biopsy from the lesion again. As a negative biopsy, only a surrounding biopsy is taken, and endoscopic treatment is performed. Lesions found at our hospital underwent biopsy and were diagnosed by a pathologist who specializes in the digestive tract. Therefore, the condition of the preoperative pathology diagnosis differs according to the case. The cases aggregated this time are all cases who presented at our hospital during the study period. Because the preoperative biopsy diagnosis is not uniform, a strong bias is likely. Therefore, the inclusion of the preoperative biopsy results was considered inappropriate and was not considered for analysis.

In conclusion, advanced age and mixed PDA were predictors of noncurative resection. Compared to mixed PDA, the rate of curative resection for pure SRC was found to be as high as that of differentiated-type early gastric cancer.