Updated evaluation of additional surgery versus non-gastrectomy treatment for early gastric cancer after noncurative endoscopic resection: a meta-analysis.

Objectives
In general, supplemental surgery with lymph node dissection is primarily suggested for patients who have undergone noncurative resection because of the likelihood of local and further LNM [5, 6]. However, some patients firmly refuse additional surgery, possibly because of the high cost, poor physical condition, or fear, leading to poor prognosis. Additionally, studies have shown that noncurative resection does not always contribute to LNM or poor outcomes; thus, some elderly patients choose to undergo routine endoscopic monitoring as an alternative treatment [3]. Previous reports have compared and analyzed treatments in patients with EGC after noncurative resection; however, the results are debatable. Only a few studies have focused on prognostic indicators [7]. Consequently, this study aimed to review the contrast in long-term noncurative ER effects in patients with EGC who underwent additional surgery versus those who underwent nonsurgical intervention, to provide a theoretical basis for clinicians to develop treatment strategies.

Methods
After the meta-analysis was registered in PROSPERO, our meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.

Search strategy
This meta-analysis included relevant retrospective cohort studies on noncurative ER in patients with EGC published in English until May 2024. Step-by-step literature screening was mainly performed using the following Medical Subject Heading (MeSH) terms, along with the related titles and abstracts: “endoscopic submucosal dissection,” “dissection endoscopic submucosal,” “endoscopic resection,” “endoscopic full-thickness resection,” “endoscopic mucosal resection,” “early-stage gastric cancer,” “early gastric cancer,” “early cancer of the stomach,” “early stomach neoplasm,” “early stage stomach cancer,” “early gastric neoplasm,” “early gastric carcinoma,” “early stomach cancer,” “operative procedures,” “surgery,” “general surgery,” “gastrectomy,” “surgical procedure,” “additional surgery,” and “additional operation,” through the following databases: PubMed, Embase, and Web of Science.

Endoscopic curative resection and noncurative resection
When gastrointestinal tumors are in an early stage and when the risk of LNM is low, clinicians often prioritize ESD for en bloc resection. Compared with patients who undergo surgery, patients with EGC who achieve curative resection of the lesion via ER may exhibit a better prognosis and a higher quality of life [3]. In the evaluation of tumor curability through endoscopic treatment, risk factors associated with complete clearance of the primary lesion and LNM are particularly important. Patients who underwent curative resection and met the criteria of endoscopic curability A (eCura A) or B (eCura B) were generally considered to have a good prognosis.
eCura A is considered: [1] differentiated type-dominant, en bloc resection regardless of the tumor size (or undifferentiated type-dominant, en bloc resection with a tumor size ≤ 2 cm), UL0(-), pT1a, horizontal margin (HM0), vertical margin (VM0), and lymphovascular infiltration (Ly0, V0); [2] differentiated type-dominant, en bloc resection with a tumor size ≤ 3 cm; (UL1), pT1a, HM0, VM0, Ly0, and V0. The following lesion characteristics can be regarded as eCura B: differentiated type-dominant, en bloc resection, pT1b1 (SM1) (distance from the muscularis mucosa less than 500 µm), tumor size ≤ 3 cm, HM0, VM0, Ly0, and V0.
Lesions were classified as endoscopic curability C (eCura C) or noncurative ER through comprehensive pathological and clinical evaluations when it did not meet the criteria of eCura A or eCura B based on the Japanese GC Guidelines [8].

Inclusion and exclusion criteria
The inclusion criteria were as follows: [1] studies in which all patients with EGC met the criteria for noncurative ER based on the Japanese GC treatment guidelines and [2] studies that included groups. The relevant patients included in our meta-analysis were divided into an additional surgery group and a nonsurgery group. Patients with EGC in the additional surgery group who had undergone noncurative ER were further recommended for surgery and lymph node dissection. In contrast, patients who did not undergo surgery were referred mainly for nonsurgical treatment, and almost all of them were monitored closely.
The exclusion criteria included the following studies: (1) without survival endpoints of interest; [2] involving patients with multiple cancers other than EGC; [3] published in languages other than English; [4] with quality scores less than five; [5] lacking the full text; [6] with missing critical data; [7] unrelated to noncurative resection; [8] without clear grouping; and [9] other types, including case reports, animal experiments, reviews, meta-analyses, comments, conference abstracts, and editorials.

Data extraction
Two authors independently reviewed the abstracts and full texts of all the articles included in our study. The following data were subsequently collected: first author, year of publication, country, study period, sample size, and survival data (5-year overall survival [OS], 8-year OS, five-year disease-specific survival [DSS], 5-year disease-free survival [DFS], 5-year recurrence-free survival [RFS], and 5-year cancer-specific survival [CSS]). Relevant indicators, such as clinical baseline characteristics (sex and age), endoscopic manifestations (tumor location and size), and pathological characteristics (type of lesion differentiation, depth of tumor invasion, lymphatic invasion, vascular invasion, lymphovascular invasion, vertical margin, horizontal margin, and ulceration findings), were extracted separately. When disagreements occurred, a third author intervened to discuss and negotiate a settlement.

Quality assessment
Given that all the studies in this meta-analysis were cohort studies, the Newcastle–Ottawa Scale was used as a quality assessment tool. The scale consists of three sections (selection, comparability, and outcome) with scores ranging from 0 to 9. Studies with scores of less than 5 were considered to be of low quality and were excluded [9].

Outcomes
In this study, we focused on the following prognostic outcomes: 5-year OS, 8-year OS, 5-year DSS, 5-year DFS, 5-year RFS, and 5-year CSS for patients with EGC who underwent noncurative ER in the additional surgery and nonsurgery groups.

Statistical analysis
The meta-analysis was conducted using Reviewer Manager (RevMan, version 5.4). The weighted mean differences (WMDs) and 95% confidence intervals (CIs) were calculated for continuous variables, whereas odds ratios (ORs) with 95% CIs were used for dichotomous variables. For assessing heterogeneity, articles with an I2 value greater than 50% were considered heterogeneous, and the results were analyzed using a random-effects model. Conversely, a fixed-effects model was used when the I2 value was lower. Funnel plots were constructed to assess publication bias. Statistical significance was set at a p value less than 0.05.

Results

PROSPERO registry
Our meta-analysis was registered on PROSPERO, with the assignment number (CRD42024549686).

Study selection
The flowchart for the literature selection process is shown in Fig. 1. First, 7070 records were included after three databases (PubMed, Web of Science, and Embase) were reviewed. After excluding duplicate articles (n = 1453), we screened the remaining 5617 articles by reviewing their titles and abstracts. Through strict application of our screening criteria, 5499 articles were excluded. Finally, we carefully selected 26 articles [10–35] after reading the full texts of all the studies.

Quality evaluation form
The basic characteristics of all the included studies are presented in Table 1. In total, 26 studies included in this review were cohort studies published up to May 2024 in English and originated only from population sources in Korea and Japan. This meta-analysis synthesized data from studies involving 4903 and 4274 patients in the additional surgery and nonsurgical groups, respectively. Table 1 also presents the main prognostic indicators of each study discussed thoroughly in this article (5-year OS, 8-year OS, 5-year DSS, 5-year DFS, 5-year RFS, and 5-year CSS). The Newcastle–Ottawa Scale (Table 2) was used to assess the quality of the studies, and studies with scores less than 5 were excluded. In contrast, the studies with more than 6 points were regarded as high quality.

Basic characteristics, clinical endoscopic features, and pathologic information extracted from all the included studies
Table 3 shows the baseline information, endoscopic features, and pathological characteristics of all lesions analyzed in our meta-analysis, comparing the additional surgery and nonsurgery groups.

Our review suggested that patients who underwent additional surgery were more likely to be men and to be relatively younger, whereas those in the nonsurgical group had more comorbidities, especially respiratory, cardiovascular, and cerebrovascular diseases. Compared with the additional surgery group, the nonsurgical group had more patients with low eCura scores and an increased recurrence rate. With respect to recurrence types, more conditions, such as positive local tumor recurrence, positive regional lymph node, and positive distant metastasis, were observed in the surgical group; however, more patients with synchronous ( +) and metachronous recurrence (+ , GC) were observed in the nonsurgical group. According to mortality analysis, more people died in the nonsurgical group due to chronic liver disease, chronic kidney disease, and diseases other than GC.
In terms of the clinicopathological features of EGC, gross appearances of the lesions were flat or depressed in the nonsurgical treatment group. Histologically, in the additional surgery group, submucosal infiltration ( +) was observed in a relatively large number of patients, as was SM2 invasion ( +), vascular infiltration ( +), lymphatic invasion ( +), vertical margins ( +), and LVI ( +). In contrast, more patients in the nonsurgical group had tissue types of signet-ring cell carcinoma.

Survival outcomes

5-year OS
All 26 studies included in our meta-analysis were used to assess prognostic outcomes. Among these, 23 studies analyzed the 5-year OS between the additional surgery group and nonsurgical group. The 5-year OS (n = 4170) in the surgical group was better than that in the nonsurgical group (n = 3827), with a total OR of 2.91–3.91 (95% CI), and the p value was significant (p < 0.00001). No heterogeneity was detected in the heterogeneity analysis (I2 = 0%; P = 0.58), indicating the reliability of the results (Fig. 2A).
The HR values for the 5-year survival rate of patients who underwent complementary surgery versus nonsurgical treatment were calculated using software, along with the 95% CI, as these values were not reported in some articles. Finally, the forest plot revealed that the 5-year OS in the additional surgery group was greater than that in the nonsurgical group (HR = 0.51; 95% CI = 0.41–0.64; p < 0.00001), accompanied by weak heterogeneity (I2 = 8%; p = 0.36) (Fig. 2B).

8-year OS
The total 8-year OS rate of patients (n = 350) in the additional surgery group was 1.96 times better than that of patients (n = 592) in the nonsurgical group (95% CI = 1.22–3.16; p = 0.005). Similarly, the heterogeneity analysis revealed no heterogeneity (I2 = 0%, p = 0.83) (Fig. 2C).

5-year DSS
Fourteen studies reported an increased 5-year DSS rate in the additional surgery group compared with the nonsurgical group. This comparative analysis involved 3,174 patients who underwent surgery versus 2,765 patients who did not (OR = 3.08; 95% CI = 2.08–4.55; p < 0.00001). Heterogeneity analysis revealed no heterogeneity (p = 0.58; I2 = 0%) (Fig. 2D).

5-year DFS
Five studies demonstrated significant heterogeneity (I2 = 81%, p = 0.0003) in the 5-year DFS between the two groups. Therefore, a random-effects model was applied, which revealed improved outcomes in the additional surgery group (OR = 4.17; 95% CI = 1.53–11.4; p = 0.005) (Fig. 2E).

5-year RFS
We analyzed three studies to compare the RFS between the surgical and nonsurgical groups (OR = 9.14; 95% CI = 3.63–23.01; p < 0.00001). The results revealed a higher survival rate in the surgical group, with some heterogeneity (p = 0.76; I2 = 0%). Consequently, a random-effects model was used (Fig. 2F).

5-year CSS
We selected three studies for CSS analysis; the results were significantly different (OR = 2.54; 95% CI = 1.32–4.9; p = 0.005), with minimal heterogeneity (I2 = 39%; p = 0.19) (Fig. 2H).

Publication bias
Analysis of the funnel plots for prognostic indicators (5-year OS, 8-year OS, 5-year DSS, 5-year DFS, 5-year RFS, and 5-year CSS) between the additional surgery and nonsurgery groups revealed no asymmetry and no significant publication bias (Fig. 3A–F).

Subgroup analysis (prognostic outcomes)
According to subgroup analysis, six studies focusing on elderly patients (≥ 70 years old) reported markedly higher5-year OS in the second gastrectomy group (OR = 3.09; 95% CI = 2.37–4.02; p < 0.00001). The results revealed minimal heterogeneity (p = 0.13; I2 = 42%) (Fig. 3G) and no prominent publication bias (Fig. 3H).

Subgroup analysis (Baseline information, clinical endoscopic features, and pathologic features)
Statistical analysis between the two groups revealed no significant difference in sex for the elderly population. The number of patients with vascular invasion ( +), SM2 invasion depth ( +), and lymphatic invasion ( +) in the additional surgery group was significantly greater than that in the nonsurgery group. However, more patients with ulcers ( +) did not undergo additional surgery (Table 4).

Discussion
A subset of patients with early gastric tumors who receive noncurative ER may refuse to undergo complementary surgery for the following reasons: age, complex comorbidities, economic conditions, lack of understanding and willingness to undergo surgery, and the doctor’s choice of treatment strategy [3, 13]. However, previous studies have revealed that patients who undergo complementary operations exhibit improved prognostic survival rates [6, 36], raising questions about whether supplemental surgery is needed for all patients with EGC following noncurative ER.
The findings of this study revealed that compared with the nonsurgical group, the additional surgery group had better OS, DSS, DFS, RFS, and CSS outcomes. HR calculations for OS revealed that the elderly subgroup had the same prognostic trend in the complementary surgery group, with significant p values and low heterogeneity. In accordance with our conclusions, Li et al. reported higher survival rates in patients who underwent noncurative ER and additional surgery than in those in the follow-up group [36]. Jiao et al. revealed that additional surgery can be more beneficial for patients who underwent noncurative surgery, even if the target population is older than 70 years [2]. Nie et al. demonstrated that patients who underwent noncurative ER of EGC who underwent secondary gastrectomy with lymph node dissection may have higher 5-year OS, DSS, and DFS rates [3]. Li et al. suggested that additional surgery may result in increased survival time for patients with EGC who undergo noncurative ER [7].
Our research stands out in several ways. First, our study included the most recent and relevant studies published in 2023. Second, we included more relevant studies than previous meta-analyses related to this topic did, especially in the subgroup analysis, leading to more convincing results. Finally, we evaluated numerous prognostic indicators, including 8-year OS, 5-year DFS, RFS, and CSS, and 5-year OS and DSS.
Several studies have revealed that the final diagnosis of EGC relies on a comprehensive analysis using white light endoscopy, magnifying endoscopy, endoscopic ultrasound, and imaging [37]. Clinical endoscopic features and pathological evaluations of lesions are unavailable until ESD specimens are obtained. Even if the lesions meet absolute or expanded indications for ESD, they may still be classified as noncurative ER after a thorough clinicopathological assessment. Therefore, a detailed histological evaluation followed by a complete ER is crucial in guiding subsequent treatment [38, 39].
In our study, histopathological markers, mainly related to LNM, such as submucosal invasion ( +), depth of invasion ( +), SM2 invasion ( +), vertical margin ( +), lymphatic invasion ( +), venous invasion ( +), and LVI ( +), were more extensive in patients in the supplemental surgery group than in those in the nonsurgery group. Studies have reported that the incidence of LNM is generally between 6.3% and 12.7% and can reach 14.0% in the secondary surgery group compared with that in the nonsurgical group; thus, LNM may be an important reference factor for complementary surgery [30, 40–42]. Therefore, the risk factors for LNM are important for patients with EGC after unhealed ER. In line with our findings, similar conclusions have been reported by studies conducted at different centers: Makimoto et al. reported that the two most important risk factors for LNM in the additional surgery group were positive lymphatic invasion and positive vertical invasion [30]. Additionally, several studies have reported that LVI is a key indicator of LNM [21, 38, 43]. Studies have shown that positive deep submucosal and lymphovascular invasion is a high-risk indicator of LNM or distant metastasis [32]. A positive vertical margin has been suggested to indicate deeper invasion of the malignancy; therefore, pathologists must collect standardized specimens and evaluate them in detail [6, 16]. Suzuki et al. suggested that several clinicopathological indicators, such as lymphatic or vascular invasion ( +), SM2 invasion ( +), a vertical margin ( +), and a tumor size > 3 cm (14), predict the possibility of LNM. Yang et al. reported that the more serious risk factors for LNM are positive venous or SM2 invasion (11). Zhao et al. reported similar conclusions that some histological markers, such as LVI ( +), SM2 ( +), and vertical margin ( +), may be key points for clinicians to determine whether to perform surgery [6].
En bloc resection involves only one whole piece and not a fragmented excision via ESD. Complete resection (R0 resection) involves an entire en bloc resection specimen that is negative on pathological evaluation of the horizontal and vertical margins at length, with no lymphovascular metastasis (33). In this study, concerning the two factors mentioned above, no significant differences were detected between specimens from secondary surgery patients and nonsurgical patients, which is consistent with the findings of previous studies [15, 21, 24, 27, 36]. Kim et al. reported that other common risk factors for LNM and distant metastasis are tumor size and differentiation type [27]. The comparative analysis between the two groups in our meta-analysis revealed no differences in the tumor size, presence of ulcers on the surface of the lesion, or tumor differentiation morphology, which may be risk factors for LNM. First, these factors might not be the most important determinants of complementary surgery in patients who undergo noncurative resection. Second, our findings must be reanalyzed after more high-quality studies are available because of the insufficient sample size of the studies included in this review, leading to limited results [6]. Chika et al. grouped patients with EGC after noncurative ER into high-risk and low-risk groups on the basis of the risk of LNM. Patients in the high-risk group were defined as having the following characteristics: “positive lymphoid or/and venous invasion” or “deep submucosal (SM2) invasion.” In contrast, patients in the low-risk group were considered to have fewer risk factors than those in the high-risk group: intramucosal carcinoma larger than 30 mm with the appearance of ulcers on the lesion surface and submucosal carcinoma (SM1 > 30 mm). Patients with EGC who underwent noncurative ER and exhibited high-risk factors for LNM were more likely to undergo additional surgery [10].
The risk of tumor invasion manifesting at the lateral and vertical margins may be related to residual tumor [38, 39], which is seen as a remnant tumor at the ESD cicatricial site. Another study reported that remnant tumors were associated with positive horizontal margins but not with positive vertical margins [1]. Researchers have reported that the occurrence of LNM is low as the horizontal margin becomes the single noncurative risk factor, possibly because of a weaker electrocautery effect in the horizontal direction than in the vertical direction [27, 31]. In this study, no significant difference in positivity at the horizontal margin assessment was observed between the additional surgical and nonsurgical groups. These findings may explain why a positive horizontal margin may be a key risk factor for residual tumors, which is not the decisive factor for complementary surgery, and additional ESD therapy may be an alternative treatment modality [8].
The eCura system, proposed by the results of previous multicenter studies in Japan, is a scoring system used to assess the risk of lymph node metastasis (LNM) in patients with EGC after ESD and is suitable for patient classification after noncurative ER on the basis of the risk of LNM. Five risk factors were included in this scoring system: tumor size > 3 cm, venous invasion, vertical margins, SM2 (submucosal invasion depth ≥ 500 μm), and lymphatic invasion (1 point was assigned for each positive item for the first four items, and 3 points were assigned for the fifth positive item). On the basis of the total scores after the score accumulation calculation for each risk factor, patients were classified as low-risk (0–1 score), intermediate-risk (2–4 score), or high-risk (5–7 score). On the basis of the grouping reported in the literature, the risk probabilities of LNM are 2.5%, 6.7%, and 22.7%, respectively [44]. Patients in the high-risk group are advised to undergo additional surgery because of their increased risk of LNM, accompanied by more severe tumor-related mortality and tumor recurrence [8, 44]. Yang et al. changed and upgraded the scoring system mentioned above to meet the clinical needs of patients after noncurative ER of poorly differentiated EGC. The rules of this modified scoring system differ; however, patients in the high-risk group are strongly recommended to undergo additional surgery [45]. Additionally, simple follow-up may be the priority for patients in the low-risk group who underwent noncurative resection on the basis of the eCura system [8, 29, 44]. This observation also explains why, in our study, more patients were at low risk on the basis of the risk grouping of the eCura score system.
In accordance with the latest guidelines for patients with EGC who underwent noncurative ER, a tumor that is differentiated by less than 3 cm, has a positive horizontal margin, and is accompanied by one of the three factors (UL1, pT1a [M], or pT1b1 [SM1]), but lacks high-risk factors (lymphatic invasion, venous invasion, and vertical margins), can be classified as eCura 1. The following options are available: repeat ESD, additional surgical resection, and close observation, owing to the low risk of LNM [8, 30, 46]. However, the prognostic benefits need to be confirmed in long-term studies. In summary, in the future, our research will provide clarity for clinicians and patients to facilitate better informed treatment decisions for patients with ESD who have undergone noncurative resection.
The outcomes of this meta-analysis (5-year OS, 8-year OS, 5-year DSS, 5-year RFS, and 5-year CSS) showed no or weak heterogeneity (p > 0.05), whereas the 5-year DFS exhibited significant heterogeneity (I2 = 81%, p = 0.0003) and was analyzed using a random-effects model. Overall, the results are reliable.
Our study has several limitations. First, some degree of selection bias may exist owing to its retrospective cohort nature, geographic restriction to Japan and Korea, and language restriction to English. For example, the possibility that differences in baseline characteristics such as age and comorbidities between the surgical and nonsurgical groups were evident, which may be the result of “differences in patient background and patient selection” rather than “the effectiveness of surgery.” That is, part of the observed survival difference may reflect patient selection rather than the pure effect of treatment. Second, during the analysis, the scarcity of studies on certain prognostic indicators also contributed to this heterogeneity. Considerable heterogeneity may have led to the lack of some data for subgroup analysis. Some indicators of clinical features or baseline information could not be analyzed because of the small sample sizes of the included studies. For example, there was a certain degree of heterogeneity in terms of DFS (I2 = 81%); therefore, we ultimately used a random-effects model for the scientific analysis. Given that the definitions of disease-free survival (DFS) and event criteria may vary across studies and that the number of included studies was limited, further research is needed to improve clarity. Furthermore, some of the heterogeneity is because the nonsurgical group likely included patients who were managed primarily with observation and possibly some who underwent additional endoscopic treatment. Third, calculation errors might have occurred in the forest plot analysis of HR because most HR values for OS were computed rather than directly provided by the authors of the individual studies.

Conclusions
This review demonstrates that additional surgery combined with lymph node dissection results in better prognostic outcomes in the additional surgery group than in the nonsurgical group, even for elderly patients over 70 years old. Various factors influence treatment options; however, the risk factors for LNM are the most prominent elements, and in some cases, secondary endoscopic therapy may be an acceptable treatment alternative. As medical and endoscopic technologies advance, our understanding of EGC also increases. Our meta-analysis has several limitations; however, future studies [47] should verify our findings and provide a more robust theoretical basis for overcoming the uncertainties faced by clinicians.