Can the Japanese guidelines for endoscopic submucosal dissection be safely applied to Korean gastric cancer patients? A multicenter retrospective study based on the Korean Gastric Cancer Association nationwide survey.

INTRODUCTION
Gastric cancer (GC) is one of the most common malignancies and a major cause of cancer-related deaths worldwide [1]. In Korea, due to the nationwide screening program for GC, the detection rate of early gastric cancer (EGC) has increased [234]. While radical gastrectomy is the standard treatment option for EGC, endoscopic submucosal dissection (ESD) has also been accepted as the standard treatment for EGC with a low risk of lymph node (LN) metastasis [5678]. The incidence of ESD has increased in Korea, and in 2021, 11,286 patients underwent ESD for EGC [9].
To ensure oncologic safety only GC patients with absolute indications for ESD should be selected. As surgical LN dissection is the only curative-intent treatment for LN metastasis of GC, so to prevent ruling it out, the tumor conditions likely to contribute to LN metastasis have been proposed [10111213]. According to the guidelines of the Japanese Gastric Cancer Association (JGCA), a tumor with a less than 1% probability of metastasizing to the LNs is considered an absolute indication for endoscopic resection as standard treatment [6]. For this population, endoscopic resection is expected to have a therapeutic effect equivalent to surgical resection. Tumors that are differentiated, confined to the mucosal layer, have no ulcers, are ≤2 cm in size have a nearly zero percent chance of LN metastasis and are considered “absolute indications” for ESD. In addition, the “expanded indications” presented in Japanese multicenter trials (JCOG0607 and JCOG1009/1010) were reclassified as absolute indications for ESD: (1) differentiatedtype adenocarcinoma without ulcerative findings, in which the depth of invasion is clinically classified as T1a, and the diameter is >2 cm; (2) differentiated-type adenocarcinoma without ulcerative findings, in which the depth of invasion is clinically classified as T1a, and the diameter is ≤3 cm; and (3) undifferentiated-type adenocarcinoma without ulcerative findings, in which the depth of invasion is clinically classified as T1a, and the diameter is ≤2 cm [12131415].
Unfortunately, in Korea, there has been a lack of evidence from well-designed large-scale studies supporting the safety of ESD for absolute indications. There are few reports on the possibility of long-term recurrence after ESD for tumors classified as absolute indications according to the JGCA guidelines [1617181920]. Furthermore, a greater proportion of patients with undifferentiated adenocarcinoma had LN metastasis in recent Chinese reports than in the reports by the JGCA, regardless of ulcerative findings, stage T1a disease, and a tumor diameter ≤2 cm [21].
The aim of this study was to analyze large-scale data from the 2019 Korean nationwide database managed by the Korean Gastric Cancer Association (KGCA) and assess the actual percentage of LN metastasis in patients with absolute indications for ESD according to the guidelines by the JGCA. We also examined the risk factors contributing to LN metastasis in EGC patients whose tumors were indications for ESD.

METHODS

Ethics statement
The study was approved by the Institutional Review Board of College of Medicine, the Catholic University of Korea (No. HC23RCSI0033), and the requirement for informed consent was waived because of the retrospective and multicenter design of the study.

Data collection
In 2019, the Information Committee of the KGCA published a report describing a nationwide survey of clinicopathological, intra- and postoperative data from GC patients treated at Korean institutions [2]. After completion of the 54-question survey at each institution, the surgeons at each institution corrected missing and incorrect data. Finally, the data of 14,076 patients who underwent surgery for gastric adenocarcinoma in 2019 were collected from 68 institutions.
Fig. 1 shows the flowchart of patient selection in this study. We excluded patients who underwent neoadjuvant chemotherapy (n = 587), had distant metastasis (n = 614), or underwent nonradical resection (n = 3,455). After excluding patients with inappropriate clinicopathological data (n = 473), as well as those with stage T1b disease (n = 3,533) and advanced GC (n = 4,294), a total of 4,220 patients were ultimately included in the final analysis.
To reduce potential bias associated with a single-year study design, we additionally analyzed and referenced data from nationwide surveys conducted in 2009 and 2014 [2223]. These surveys included 14,658 patients in 2009 and 15,613 patients in 2014. Patients were selected based on the same inclusion criteria described in Fig. 1, and their data were compared with those from the current study. Unfortunately, information on neoadjuvant chemotherapy was not collected in either year’s survey. A total of 4,081 patients from 2009 and 4,467 patients from 2014 were selected and analyzed.

Pathologic results
For analysis of the differentiation of GC, papillary adenocarcinoma, well-differentiated tubular adenocarcinoma (TAC), and moderately differentiated TAC were classified as differentiated types. Poorly differentiated TAC and poorly cohesive carcinoma, including signet ring cell carcinoma, were classified as undifferentiated, in accordance with the guidelines by the Korean Society of Pathologists [24]. Because the pathological results obtained from the survey did not explicitly indicate the presence or absence of an ulcer, we referred to the macroscopic type of the tumor. Type I, IIa, and IIb EGC were regarded as ulcer-free, whereas type IIc and III EGC were regarded as having ulceration. Clinicopathologic data, excluding ulcer-related information, were used as reported in the survey results. Analysis of Lauren’s classification was determined to be unsuitable because of the substantial amount of missing data.

Evaluation of curability (eCura system)
In the Japanese guidelines, the curability of endoscopic resection was evaluated using the eCura system [614]. The postoperative pathological results were not suitable for analysis using the eCura system in this study. However, by matching the pathological characteristics of cancer lesions to those of the eCura system, we evaluated whether the eCura system is suitable for risk stratification of Korean patients. Endoscopic resection is classified as having grade A (eCuraA) curability when all of the following conditions are met, provided that cancer has no signs or symptoms of ulceration (UL0): en bloc resection, histologically differentiated type-dominant, tumor size ≤2 cm in case of histologically undifferentiated typedominant, stage pT1a disease, negative horizontal margin (HM0), negative vertical margin (VM0), and no lymphovascular invasion (LVI; Ly0, V0). When the cancer has ulcerative findings (UL1), the resection is still classified as eCuraA when all of the following conditions are met: en bloc resection, tumor size ≤3 cm, histologically differentiated type-dominant, pT1a, HM0, VM0, Ly0, and V0. Resection is classified as eCuraB for pT1b cancer when all of the following conditions are met: en bloc resection, histologically differentiated type-dominant, pT1b1 (SM1) (<500 µm from the muscularis mucosae), HM0, VM0, Ly0, V0, and tumor size ≤3 cm. Resection is classified as eCuraC if it does not meet eCuraA or eCuraB. Resection is classified as eCuraC-1 when it is a histologically differentiated type that is either not resected en bloc or has HM1 even though it met other criteria to be classified into either eCuraA or eCuraB. All other eCuraC resections are subclassified as eCuraC-2. In addition, a risk-scoring system for predicting LN metastasis after ESD was applied and evaluated in patients from this study cohort [25]. To calculate a risk score, points were allocated to 5 predictive factors in proportion to their regression coefficients: 1 point each for tumor size >30 mm, VM1, venous invasion, and SM2 invasion, and 3 points for lymphatic invasion. Each patient’s total risk score in the development cohort was calculated by summing the points corresponding to their individual risk factors, resulting in a score ranging from 0 to 7. Since the data obtained from the 2019 survey did not differentiate between venous and lymphatic invasion, we treated them collectively as‘LVI,’ assigning it a score of 4 points in our analysis. Based on the total score, patients were classified into 3 risk categories for lymph node metastasis: low (0–1 point), intermediate (2–4 points), and high (5–7 points).

Statistical analysis
The data are expressed as medians and interquartile ranges (25th–75th percentiles) for nonparametric continuous variables, as the means ± standard deviations for parametric continuous variables, and as frequencies with percentages for nominal variables. The chi-square test or Fisher exact test was used for nominal variables. The Mann-Whitney U-test was applied for nonparametric variables, and the Student t-test was applied for parametric, continuous variables. Significance was determined using a two-tailed P-value of 0.05. Variables that were found to be significant in previously reported studies were used for univariate analysis, and variables that were found to be significant (P < 0.05) in univariate analysis were included in the multivariate logistic regression model to identify prognostic variables related to LN metastasis. Statistical analyses were performed with PASW Statistics for Windows, ver. 18.0 (IBM Corp.) and R software ver. 3.4.2 (The R Foundation).

RESULTS

Clinicopathologic characteristics of patients
Table 1 summarizes the characteristics and clinical data of 4,220 patients. The median age of the patients was 61 years, and male patients accounted for 60.0% of the cohort. The median size of the tumors was 2.2 cm. The tumors were predominantly located in the lower third of the stomach (49.9%), followed by the middle third (32.9%) and the upper third or the esophagogastric junction (17.2%). The rate of lymph node metastasis was 5.1% (n = 214).
Most surgeries were performed using minimally invasive techniques, and distal gastrectomy is the most common type of resection, performed in 84.1% of cases. LN dissection of D2 or greater was performed in 46.0% of the patients, with the remainder undergoing less extensive dissection. A total of 55.9% of the patients were classified as having an ulcer. Pathologically, 57.5% of the tumors were undifferentiated, and LVI was observed in 5.6% of the patients, whereas perineural invasion (PNI) was present in 2.7% of the patients.

Rate of lymph node metastasis by tumor condition
Using data from 4,220 patients, the incidence of nodal metastasis was analyzed on the basis of tumor depth, ulceration status, tumor size, and histological type (Table 2). Among them, 2,012 patients met the criteria for absolute indication, with lymph node metastasis observed in 2.4% (n = 49). When stratified by pathological tumor conditions, the incidence ranged from 1.6% to 2.9%, which is significantly higher than the rates suggested in the JGCA guidelines. Specifically, for mucosal tumors without ulceration (UL0), differentiated tumors measuring ≤2 cm had a nodal metastasis rate of 1.6% (6 of 387) with a 95% confidence interval (CI) of 1.0%–2.2%. When the tumor size exceeded 2 cm, the nodal metastasis rate increased slightly to 2.1% (10 of 474), with a 95% CI of 1.6%–2.8%. Differentiated tumors ≤3 cm with ulceration (UL1) had a metastasis rate of 2.9% (20 of 689), with a 95% CI of 2.3%–3.5%, whereas tumors >3 cm had a significant increase to 8.6% (21 of 243).
In undifferentiated tumors, the nodal metastasis rate is greater. For UL0 tumors ≤2 cm, the incidence was 2.8% (13 of 462), with a 95% CI of 2.0%–3.6%, whereas for UL0 tumors >2 cm, the incidence increased to 5.4% (29 of 563). Among UL1 tumors, the metastasis rate for tumors ≤2 cm was 3.9% (25 of 646), whereas that for tumors >2 cm was considerably higher, at 11.5% (90 of 783).
Among the 4,081 patients from 2009, 2,103 met the criteria for absolute indication for ESD. Of these, 2.5% (n = 52; 95% CI, 1.9%–3.2%) had LN metastasis following surgery (Supplementary Table 1). Similarly, among the 4,467 patients from 2014, 2,191 met the criteria for absolute indication for ESD, and 2.1% (n= 46; 95% CI, 1.6%–2.8%) had LN metastasis after surgery (Supplementary Table 2).

Rate of lymph node metastasis according to the eCura system
Assuming that all patients’ specimens were resected en bloc with HM0 and VM0, the proportion of patients classified as eCuraA was 45.3% (1,911 of 4,220), and the lymph node metastasis rate in these patients was 1.6% (31 of 1,911), with a 95% CI of 1.1%–2.2%. Since this study included only patients with mucosal tumors, no patients were classified as eCuraB. Similarly, when en bloc resection and negative resection margins were assumed, no patients fell into the eCuraC-1 category. The proportion of patients classified as eCuraC-2 was 54.7% (2,309 of 4,220), among whom the lymph node metastasis rate was 7.9%, with a 95% CI of 6.8%–9.0%. Although the distribution could not be stratified in greater detail due to limitations in the available dataset, analysis using the risk score also demonstrated a trend similar to that observed in previous Japanese studies, with LN metastasis increasing as the risk score rose (Table 3). When patients were grouped by risk category, the LN metastasis rates were 3.5% in the low-risk group, 21.3% in the intermediate-risk group, and 42.6% in the high-risk group (Table 3).

Logistic regression analysis of the risk factors for lymph node metastasis
Univariate logistic regression analysis revealed that a tumor size >2 cm, the presence of ulceration, undifferentiated histology, presence of LVI, PNI, and ESD performed within 90 days before surgery were significant factors (P < 0.05) for LN metastasis (Table 4). In the multivariate analysis of lymph node metastasis, all 5 factors other than ESD within 90 days prior to surgery were found to be statistically significant. The odds ratio (OR) for LVI was 7.56, and those for PNI, tumor size >2 cm, undifferentiated tumors, and the presence of ulceration were 3.02, 2.37, 2.30, and 1.66, respectively.

DISCUSSION
The pathological outcomes of EGC patients who were considered suitable for ESD on the basis of their indications were retrieved from a large-scale nationwide dataset collected in Korea in 2019 [2]. To the authors’ knowledge, this is the first large-scale study conducted on the Korean population. Analysis of data from a total of 4,220 patients revealed that the rate of LN metastasis in patients with absolute indications according to the criteria outlined in the Japanese treatment guidelines—previously regarded as indications for safe ESD—ranged from 1.6% to 2.9% [614]. Furthermore, through this large dataset, the risk factors for LN metastasis were analyzed, reaffirming that previously known factors, such as large tumor size, poor differentiation, the presence of ulceration, LVI, and the PNI, are significant risk factors [262728].
According to the guidelines by the JGCA, the absolute indications for ESD are tumors with a less than 1% likelihood of LN metastasis. For such populations, endoscopic resection is expected to have a therapeutic effect equivalent to that of surgical resection. Initially, the criteria included differentiated tumors confined to the mucosa (T1a), a UL0 status, and ≤2 cm in size, but these indications have been expanded in subsequent studies. However, as demonstrated in our study, certain subsets of mucosal tumors present LN metastasis rates exceeding 2%. Furthermore, the 95% CI for the rate of LN metastasis for differentiated tumors considered UL0, measuring ≤2 cm in size, and confined to the mucosa (T1a) had a lower bound of 1.0%, and the 95% CIs for other patient groups all exceeded 1%. This trend was similarly observed in the Korean cohort of 2009 and 2014 (Supplementary Tables 1, 2). Moreover, the rate of LN metastasis among patients categorized as eCuraA, excluding those with LVI, was determined to be 1.6%. According to the Japanese guidelines, patients classified as eCuraA are recommended to undergo observation without additional treatment. However, the findings revealed in this study suggest that the absolute indications documented for Japanese patients may not be suitable for Korean patients. Although the data used in this study were revised on the basis of feedback, large-scale survey data may inherently contain missing data. Additionally, as this study involved over 50 centers, there is inevitable heterogeneity in the evaluation and reporting of pathological results, possibly leading to data distortion. Moreover, the absence of preoperative, intraoperative, and long-term postoperative data is a limitation that precludes more detailed analyses. However, a recently published study from China revealed LN metastasis rates exceeding 1% in patients with absolute indications, with rates reaching as high as 7.8% in cases of undifferentiated cancer [21]. These findings suggest that directly applying Japanese research results to the Korean population without careful consideration may pose significant risks. Even for patients with absolute indications, careful assessment of preoperative imaging findings, such as CT or PET-CT, is strongly advised for any suspicion of LN metastasis, and thorough evaluations with endoscopic ultrasound or other modalities should be conducted in patients with identified risk factors to better exclude LN metastasis before considering ESD.
The lack of a clear description regarding ulceration and the assessment of its presence solely on the basis of the tumor’s morphological type are limitations of this study. Ulceration is defined as a full-thickness disruption of the muscularis mucosae, including both active and scarring ulcers, and is determined by histological findings rather than endoscopic observations [242930]. According to the Korean pathology guidelines, there is an approximately 5% discrepancy between ulcerations assessed grossly and those confirmed through pathological findings [243132]. In this study, 55.9% of patients had ulceration, which is higher than the 28.6% reported in a Chinese study and 41.7%–44.0% reported in Japanese studies [121521]. The OR for LN metastasis associated with ulceration in this study was 1.66, which was slightly greater than the value of 1.375 reported in a Chinese study but lower than the value of 1.683 reported in a Japanese study. However, it is worth noting that T1a tumors smaller than 3 cm with good differentiation are included in the absolute indication regardless of the presence of ulceration. Given that this study revealed that LN metastasis rates exceeded 2%, even within this patient group, further research is needed to explore the interaction between ulceration and other risk factors contributing to LN metastasis.
PNI is identified when cancer cells are observed within or surrounding nerves [24]. In this study, the incidence of PNI was 2.7%, whereas a meta-analysis revealed a positive rate of 35.9% in GC patients, ranging from 6.9% to 75.6% [33]. Compared with other well-established factors, such as LVI, tumor differentiation, and tumor size, the impact of PNI on LN metastasis has been less frequently studied. In this study, the correlation between the PNI and LVI was analyzed, revealing a weak correlation with a coefficient of –0.0437. However, the multivariate analysis revealed that the PNI significantly influenced LN metastasis, with an OR of 3.02, which is similar to the OR of 3.60 reported in the previously mentioned meta-analysis. These findings suggest that the PNI, as an independent factor distinct from LVI, is an important predictor of lymph node metastasis and highlight the need for further research on this association.
The major limitation of this study is that it contained a considerable amount of missing data. As the study was conducted using data collected through surveys, it was not possible to revalidate cases with uncertainties. The inclusion of preoperative findings, such as endoscopic and CT results, would have allowed for more detailed analyses. In cases of submucosal invasion, the lack of detailed data on invasion depth prevented an assessment of the risks of ESD for SM1 tumors, which is a notable limitation. Additionally, the absence of long-term follow-up data made it impossible to analyze the relationship between LN metastasis and recurrence. Future well-designed studies are recommended to address these limitations and provide further insights. This study included patients who underwent ESD within 90 days prior to surgery. Although incorporating information on preoperative ESD could offer meaningful insights, the nationwide database only records whether ESD was performed within this timeframe, without providing further details on its direct association with the gastrectomy (whether cancer-related or not) or the indication for additional surgery. Due to this limitation, a detailed analysis was not feasible. Therefore, we did not exclude these patients from the study cohort. Instead, we presented this variable in Table 1 and included it in the univariate and multivariate analyses of lymph node metastasis (Table 4).
In conclusion, this study revealed a relatively high proportion of LN metastasis among Korean GC patients with absolute indications. These results suggest that the Japanese guidelines for ESD could be applied to Korean GC patients if done carefully.