Long-term outcomes of additional surgery versus non-gastrectomy treatment for early gastric cancer after non-curative endoscopic submucosal dissection: a meta-analysis

Abstract

Background: 

Endoscopic resection is increasingly used in the treatment for early gastric cancer (EGC); however, about 15% of endoscopic submucosal dissection (ESD) cases report non-curative resection. The efficacy of different remedial interventions after non-curative ESD for EGC remains controversial. This meta-analysis aimed to compare the long-term outcomes of additional surgery and non-gastrectomy treatment for EGC patients who underwent non-curative ESD.

Methods: 

All relevant studies published up to October 2021 were systematically searched in the PubMed, Web of Science, and Embase databases. The medical subject headings terms “early gastric cancer,” “gastrectomy,” “endoscopic submucosal dissection,” and their related free keywords were used to search relevant articles without restrictions on regions, publication types, or languages. The Newcastle–Ottawa Quality Assessment Scale was used to evaluate the quality of the included studies. Odds ratios (ORs) with 95% confidence intervals (CIs) of 5-year overall survival (OS), disease-specific survival (DSS), disease-free survival (DFS) and hazard ratios (HRs) with 95% CIs of OS were calculated using a random- or fixed-effects model.

Results: 

This meta-analysis included 17 retrospective cohort studies with 5880 patients, of whom 3167 underwent additional surgery and 2713 underwent non-gastrectomy. We found that patients receiving additional gastrectomy had better 5-year OS (OR = 3.63, 95% CI = 3.05–4.31), DSS (OR = 3.22, 95% CI = 2.22–4.66), and DFS (OR = 4.39, 95% CI = 1.78–10.82) outcomes than those receiving non-gastrectomy treatments. The pooled HR also showed that gastrectomy following non-curative ESD significantly improved OS (HR = 0.40, 95% CI = 0.33–0.48). In addition, elderly patients benefited from additional surgery in consideration of the 5-year OS (HR = 0.54, 95% CI = 0.41–0.72).

Conclusions: 

Compared with non-gastrectomy treatments, additional surgery offered better long-term survival outcomes for patients with EGC who underwent non-curative ESD.

Introduction

According to the Globocan2020, gastric cancer is the fifth most frequently occurring malignant tumor and the fourth leading cause of cancer-related deaths worldwide.^[1] Owing to rapid advances in endoscopic screening, the early detection of gastric cancer has recently become more accessible, especially in East Asia.^[2] Early gastric cancer (EGC) refers to lesions confined to the mucosa or submucosa regardless of lymph node metastasis (LNM).^[3] Although the previous mainstream treatment of EGC involved radical gastrectomy with lymph node dissection, endoscopic resection has been commonly regarded as a minimally invasive management option for most EGC cases, particularly in those with a very low risk of LNM.^[4] Currently, there are two available endoscopic resection techniques, endoscopic mucosal resection (EMR) and endoscopic submucosal dissection (ESD), of which the latter is recommended because of its superior efficacy.^[5]

However, with the extensive use of endoscopic resection techniques, approximately 15% of ESD cases report non-curative resection or endoscopic curability C (eCuraC) resection, which correlates with a greater risk of remnant tumor, local recurrence, and distant metastasis.^[4,6-10]

For non-curative or eCuraC resection, open or laparoscopic surgery is highly recommended because of the likelihood of recurrence and further metastasis.^[9] Patients who meet the criteria of eCuraC-1, namely having either piecemeal resection or a positive horizontal margin (HM+) as the only non-curative factor, are provided with alternative options, including repeat ESD, surgery, close observation, and endoscopic coagulation, in consideration of their extremely low LNM rates.^[4,6,11-14] Patients who refuse to undergo additional surgery because of old age or severe comorbidities are more likely to experience recurrent diseases and dismal prognosis.^[15-17]

Although several studies have compared treatments after non-curative ESD in patients with EGC, discussion on long-term outcomes is limited, and their results are inconsistent. Previous meta-analyses by Li et al^[18] and Nie et al^[19] evaluated the prognosis of different remedial interventions after non-curative endoscopic resection. Both studies failed to exclude patients who underwent EMR as the initial therapy, and the study by Nie et al^[19] also included several studies that lacked control groups or 5-year survival data. Additionally, several eligible studies have been published since. Therefore, this meta-analysis was performed to compare the long-term efficacy of additional gastrectomy and non-gastrectomy treatments after non-curative ESD and thus provide guidance for clinical decisions regarding survival outcomes in such situations.

Methods

Search strategy

All relevant studies published up to October 2021 were systematically searched in the PubMed, Web of Science, and Embase databases, without restrictions on regions, publication types, or languages. The following Medical Subject Headings and free text keywords were used: “early gastric cancer”, “early gastric carcinoma”, “early stomach cancer”, “early stage gastric cancer”, “early stage stomach cancer”, “gastrectomy”, “surgery”, “surgical”, “gastrectomy”, “operation”, “Endoscopic Mucosal Resection”, “endoscopic resection”, and “Endoscopic Submucosal Dissection”. More detailed search strategies are provided in the Supplementary Methods, https://links.lww.com/CM9/B435. We also screened the relevant references listed in the retrieved articles to identify additional relevant studies.

Inclusion and exclusion criteria

According to the Japanese Gastric Cancer Association,^[4,10] endoscopic resection is considered non-curative, or eCuraC, when pathological findings do not fulfil the criteria for curative resection, also known as endoscopic curability A and B.

The inclusion criteria were as follows: (1) participants: EGC patients who underwent non-curative or eCuraC (including eCuraC-1 and eCuraC-2) resection of ESD; (2) intervention: additional surgery; (3) comparison: non-surgical treatments, including simple follow-up, repeat endoscopic resection (re-ER), and argon plasma coagulation (APC); (4) outcome: reporting at least one of the endpoints listed in Table 1; and (5) study design: randomized controlled trials, prospective or retrospective cohort, and case-control studies.

Table 1 - Main characteristics of the included studies.

Studies	Country	Language	Study period	Cases (G/N)	Endpoints	NOS score
Katsube et al [22]	Japan	English	2003-2013	17/37	5-year OS	6
Esaki et al [16]	Japan	English	2000-2011	1064/905	5-year OS, 5-year DSS	7
Suzuki et al [23]	Japan	English	1999-2010	356/212	5-year OS, 5-year DSS	6
Kim et al [24]	Korea	English	2005-2016	113/175	5-year OS, 5-year DSS, 5-year DFS	6
Toya et al [25]	Japan	English	2002-2010	45/21	5-year OS, 5-year DSS	7
Hoteya et al [26]	Japan	English	2005-2011	109/56	5-year OS	6
Yamanouchi et al [27]	Japan	English	2001-2012	28/51	5-year OS, 5-year DSS	6
Jeon et la [28]	Korea	English	2005-2014	18/16	5-year OS, 5-year DSS, 5-year DFS	6
Iwai et al [17]	Japan	English	2007-2012	49/46	5-year OS, 5-year DSS	6
Jeon et la [29]	Korea	English	2007-2016	264/248	5-year OS, 5-year DFS	6
Choi et al [30]	Korea	English	2003-2010	28/61	5-year OS, 5-year DFS	7
Harada et al [31]	Japan	Japanese	2007-2016	26/19	5-year OS, 5-year DSS	6
Kawata et al [32]	Japan	English	2002-2012	323/183	5-year OS, 5-year DSS	6
Sumiyoshi et al [33]	Japan	English	2003-2010	15/17	5-year OS	5
Eom et al [34]	Korea	English	2004-2014	126/67	5-year OS, 5-year DSS	6
Suzuki et al [35]	Japan	English	2000-2011	553/553	5-year OS, 5-year DSS	8
Zhou et al [36]	China	Chinese	2011-2018	41/66	5-year OS, 5-year DFS	6

DFS: Disease-free survival; DSS: Disease-specific survival; G: Additional gastrectomy group; N: Non-gastrectomy group; NOS: Newcastle–Ottawa Scale; OS: Overall survival.

Studies were excluded if they met the following criteria: (1) inclusion of patients with other malignancies than EGC; (2) initial management including not only ESD; (3) involving only non-surgical interventions; (4) other publication types, including: animal experiments, case reports, reviews, meta-analyses, comments, editorials, and conference abstracts; (5) duplicate articles: if multiple articles reported the same patient cohort, articles without the endpoints of interests were excluded; and (6) quality score <5 points.

Data extraction

Two authors (Li and Wei) independently screened the included studies and extracted the following relevant data: first author, publication year, country, study design, study period, measures of the experimental and control groups, follow-up period, sample size, patient characteristics (age, sex, and comorbidities), clinicopathological characteristics of lesions (location, depth, size, macroscopic type, histological type, ulceration, lymphatic invasion, vascular invasion, vertical margin, and HM+), and outcomes (local recurrence, overall survival [OS], disease-specific survival [DSS], and disease-free survival [DFS]). For survival data, we extracted the unadjusted and adjusted hazard ratios (HRs) with 95% confidence intervals (CIs). If HRs with 95% CIs were not reported, we extracted data from Kaplan-Meier survival curves using Origin 2019 software (OriginLab Corporation, Northampton, MA, USA) and calculated them according to the methods provided by Tierney.^[20] Disagreements were resolved by discussion with a third author (Tian).

Quality assessment

All the included articles were retrospective cohort studies. Two authors (Li and Wei) independently evaluated the quality of the enrolled studies according to the Newcastle–Ottawa Quality Assessment Scale (NOS), which is structured into three parts (selection, comparability, and outcome).^[21] The full NOS score was 9 points, and studies scoring <5 points were excluded.

Statistical analysis

Data were analyzed using Review Manager software (version 5.3; Cochrane Collaboration, Oxford, UK) and R 4.2.1 software (R Core Team, Vienna, Austria). Odds ratios (ORs) and weighted mean differences with 95% CIs were calculated to analyze dichotomous and continuous variables, respectively. To appropriately examine the survival data, a time-to-event analysis was performed using HRs with 95% CIs. According to the Cochrane review guidelines, we considered I² >50% as high heterogeneity, and in this case, the pooled estimate was calculated using the random-effects model. Otherwise, a fixed effects model was used. All statistical tests were performed two-sided, and the level of significance was defined as P value <0.05. Publication bias was evaluated using funnel plots and verified using Egger's test.

Results

Study selection

Figure 1 illustrates the selection process. A total of 5085 records were identified from PubMed, Web of Science, and Embase databases. After removing duplicates, 3155 records remained. We subsequently reviewed their publication types, titles, and abstracts and excluded 3108 records. After further assessment of the full-text reading, 17 studies were included in the meta-analysis.

Figure 1: Flow diagram of study selection. ESD: Endoscopic submucosal dissection; ER: Endoscopic resection; EMR: endoscopic mucosal resection; EGC: early gastric cancer.

Characteristics of the included studies

The main characteristics of the studies are listed in Table 1.^{[16,17,22-36]} All articles were reported in east Asia (11 in Japan, five in Korea, and one in China) between 2015 and 2021. We included 5880 patients who underwent non-curative ESD for EGC, including 3175 in the additional gastrectomy group and 2733 in the non-gastrectomy group. The results of quality assessment are also listed in Table 1, and the NOS score ranged from 5 to 8 [details are shown in Supplementary Table 1, https://links.lww.com/CM9/B435].

As Table 2 shows, there were significant differences in the baseline characteristics of patients and lesions between the additional surgery and non-surgical treatment groups. The patients who underwent additional gastrectomy were significantly younger and had a lower prevalence of baseline comorbidities. In terms of the clinicopathological characteristics, the additional gastrectomy group had smaller tumors, deeper invasion, a lower proportion of undifferentiated histology type, fewer ulcerations, higher prevalence of lymphatic or venous invasion, and more positive vertical margins.

Table 2 - Outcomes of baseline characteristics between additional gastrectomy group and non-gastrectomy group.

		Results
Variables	No. of studies	G	N	OR/WMD (95% CI)	P value	I 2 (%)
Characteristics of patients
Male	14	2411/3103 (77.7)	1965/2637 (74.5)	1.16 (1.02, 1.31)	0.019	0.0
Age (years)	11	67.1 ± 9.9	69.4 ± 10.9	–4.42 (–6.55, –2.28)	<0.001	90.9
Comorbidity: CCI ≥ 2	3	58/531 (10.9)	65/325 (20.0)	0.52 (0.35, 0.77)	0.001	0.0
Clinicopathological characteristics of lesions
Diameter (mm)	7	27.6 ± 15.4	30.8 ± 17.8	–1.05 (–1.83, –0.27)	0.009	36.9
Location: lower third	8	490/1219 (40.2)	516/1196 (43.1)	1.02 (0.86, 1.22)	0.810	0.0
Gross appearance: flat and depressed type	8	877/1219 (71.9)	881/1196 (73.7)	0.97 (0.81, 1.17)	0.766	35.0
Histology: undifferentiated type	12	451/2990 (15.1)	500/2548 (19.6)	0.74 (0.55, 0.99)	0.040	57.7
Depth: SM2 invasion	13	1886/3005 (62.8)	1135/2565 (44.2)	2.72 (1.88, 3.95)	<0.001	87.3
Presence of ulcer	8	502/2243 (22.4)	533/2066 (25.8)	0.81 (0.70, 0.93)	0.003	0.0
Positive lymphatic invasion	6	844/2126 (39.7)	462/1858 (24.9)	2.42 (1.25, 4.71)	0.009	90.5
Positive venous invasion	5	431/2081 (20.7)	241/1837 (13.1)	2.34 (1.01, 5.42)	0.048	83.9
Positive horizontal margin	5	100/774 (12.9)	135/679 (19.9)	0.79 (0.42, 1.50)	0.476	53.7
Positive vertical margin	9	556/2796 (19.9)	261/2395 (10.9)	2.52 (1.22, 5.21)	0.012	92.7
Non-R0 resection	3	162/422 (38.4)	151/350 (43.1)	0.92 (0.68, 1.25)	0.606	42.3
Piecemeal resection	4	22/455 (4.8)	29/520 (5.6)	0.73 (0.42, 1.29)	0.282	0.0

Data were shown as n (%), n/N (%) or mean ± SD. CCI: Charlson Comorbidity Index; CI: Confidence interval; G: Additional gastrectomy group; N: Non-gastrectomy group; OR: Odds ratios; SD: Standard deviation; SM2: Deep portion of the submucosa ≥500 μm from the muscularis mucosae; WMD: Weighted mean difference.

Survival rate

As shown in Figure 2, 16 studies^{[16,17,22-28,30-36]} reported 5-year OS data, for which no heterogeneity was observed (I² = 0%, P = 0.51). The pooled OR was 3.63 (95% CI = 3.05–4.31), indicating that patients who underwent additional gastrectomy had a higher 5-year OS rate than patients receiving non-gastrectomy treatment.

Figure 2: Forest plot comparing 5-year overall survival in patients who received additional gastrectomy and non-gastrectomy treatment. CI: Confidence interval; M-H: Mantel-Haenszel.

As shown in Figure 3, 12 studies^{[16,17,23-25,27-29,31,32,34,35]} compared the 5-year DSS data between different interventions. The pooled OR was 3.22 (95% CI = 2.22–4.66) without obvious heterogeneity (I² = 0%, P = 0.87), which demonstrated that the 5-year DSS outcome was better in the additional gastrectomy group.

Figure 3: Forest plot comparing 5-year disease-specific survival in patients who received additional gastrectomy and non-gastrectomy treatment. CI: Confidence interval; M-H: Mantel-Haenszel.

Furthermore, five studies^{[24,28-30,36]} evaluated the 5-year DFS data. Heterogeneity was detected (I² = 64%, P = 0.03); therefore, we used a random effects model. The pooled OR was 4.39 (95% CI = 1.78–10.82), showing that patients who underwent additional gastrectomy had better 5-year DFS [Supplementary Figure 1, https://links.lww.com/CM9/B435].

In summary, patients who underwent additional gastrectomy had higher survival rates than those who did not undergo gastrectomy.

HR on OS

To analyze the survival data thoroughly, we also extracted and calculated HRs comparing OS between the two groups from 14 studies.^{[16,17,22-27,30,32-36]} As shown in Figure 4, no obvious heterogeneity was observed (I² = 0%, P = 0.60), and similar to the results of ORs, patients with additional gastrectomy had better OS (HR = 0.40, 95% CI = 0.33–0.48).

Figure 4: Forest plot showing the pooled hazard ratio of overall survival. IV: Instrumental variables; CI: Confidence interval; SE: Standard error.

Furthermore, two studies^[16,33] specifically investigated the survival outcomes in the elderly group. No heterogeneity was detected (I² = 0%, P = 0.99), and the pooled HR was 0.54 (95% CI = 0.41–0.72) [Supplementary Figure 2, https://links.lww.com/CM9/B435].

Publication bias

Funnel plots of the ORs and HR were symmetrical [Supplementary Figure 3, https://links.lww.com/CM9/B435]. In accordance with these results, Egger's tests showed no evidence of publication bias in the OR of 5-year OS (P = 0.55), 5-year DSS (P = 0.52), and HR of OS (P = 0.48).

Discussion

The current meta-analysis clearly showed that additional gastrectomy had better long-term efficacy for EGC patients undergoing non-curative ESD, as additional surgery significantly prolonged the 5-year OS, DSS, and DFS compared with non-surgical treatments. Previous meta-analyses that included both EMR and ESD as initial treatments for EGC had similar results. Li et al^[18] included 10 studies and found that additional surgery provided better 5-year OS (OR = 3.50, 95% CI = 2.89–4.24) and DSS (OR = 3.99, 95% CI = 2.50–6.36). In addition, Nie et al^[19] included 21 articles, and reported that patients who underwent additional gastrectomy had a longer 5-year OS (HR = 0.34, P < 0.001), DFS (HR = 0.52, P = 0.001), and DFS (HR = 0.50, P < 0.001). However, this study had several strengths. First, our study focused on EGC patients who underwent non-curative ESD and their long-term outcomes after different treatments. Second, we included several more recently published articles and more patients, which resulted in more reliable cumulative effects. In addition, we used both OR and HR to evaluate the survival data. Since few studies directly provided HR values, we extracted data from Kaplan-Meier curves to calculate HRs with 95% CIs, thus obtaining more comprehensive results.

In elderly patients, the 5-year OS showed that this cohort could benefit from additional gastrectomy. However, we included only two studies with 1190 patients that evaluated survival outcomes in the elderly group, which was insufficient to obtain credible results. In clinical practice, the selection of a therapeutic strategy is influenced by efficacy and risk of recurrence, and by age, baseline physical condition, postoperative living quality, wishes of patients or their family members, and economic condition. Although surgery is supposed to prolong elderly patients’ survival time [Table 2], patients who underwent radical surgery were significantly younger, indicating that age was an important factor in clinical decision making after non-curative ESD. In addition, several articles have stated that high-risk comorbidities were rated more strongly in treatment decisions from the perspective of patients. A retrospective study showed that high-risk comorbidity (Charlson Comorbidity Index [CCI] ≥ 3), but not advanced age, was an independent prognostic factor.^[17] Shimada et al^[37] reported that although CCI ≥ 3 was an independent predictor of short-term survival in patients who did not undergo additional surgery, gastric cancer was not the cause of most deaths, suggesting that follow-up may be a more appropriate choice for patients in poor situations. Similarly, a multicenter study suggested that no subsequent treatment was a suitable option for elderly patients with a CCI of not <2 points.^[38] In summary, for elderly patients with EGC undergoing non-curative ESD, CCI should be measured before making therapeutic decisions. Additional surgery is recommended for elderly patients with better physical conditions, and close follow-up is probably more appropriate for high-risk patients, especially those with a CCI ≥ 2 or 3.

According to the latest version of the Japanese gastric cancer treatment guidelines,^[4] when resection is classified as eCuraC-1, secondary treatment options include simple follow-up, re-ER, APC, and gastrectomy. However, the non-gastrectomy groups included in our analysis had limited interventions, except for follow-up and close surveillance. Only one article by Jeon et al^[29] reported 50 cases of re-ER for non-surgical treatments, which found that additional surgery significantly provided better 5-years DSS and DFS than both re-ER and observation. Further analysis showed that although there was no significant difference in the 5-year DFS between the re-ER and observation groups, the re-ER group had a higher DFS rate in patients with HM+ as the only non-curative factor, which meets the criteria of eCuraC-1. Additionally, Kim et al^[13] reported that recurrence or residual tumor rates after re-ER and APC in patients with HM+ alone were relatively low, suggesting that re-ER or APC may be substitutes for surgery in such circumstances. In contrast, Kim et al^[39] clarified that re-ER is a better additive treatment for APC, since APC alone cannot guarantee complete ablation of remnant lesions histologically. To accurately evaluate the risk of residual or recurrent tumors in patients with only HM+, Hwang et al^[40] established a scoring system that included the number of involved directions, total length, and rate of HM+, and undifferentiated type. The results suggested that patients with grades <3 points should undergo close endoscopy surveillance, and the others benefit from re-ER or surgery. Taken together, the therapeutic strategy for patients with positive HM+ alone after non-curative ESD is debated. Close follow-up is acceptable for patients classified as low risk, and re-ER is likely to be the most effective management for patients at high risk, as this technique eliminates residual cancer and enhances patient recovery. However, the sample size in these studies were insufficient to obtain robust results, and the long-term survival data of the different treatments were also limited. Furthermore, few studies have focused on the therapeutic strategy for patients with piecemeal resection as the only non-curative factor and other criteria of eCuraC-1. Therefore, it remains unclear whether re-ER shows notable efficacy in resections for patients classified as eCuraC-1, and further long-term studies with larger sample sizes are needed to determine the most effective form of therapy and the optimal time of performance.

This meta-analysis had several limitations. First, all the included studies were retrospective cohorts, resulting in potential patient selection bias. Second, all studies were performed in Japan, Korea, and China, where endoscopic technology is well developed, which makes it challenging to extrapolate the results to other countries. Third, unadjusted HRs, adjusted HRs, and HRs indirectly calculated from Kaplan-Meier curves were combined for the pooled effects of OS, which may cause slight calculation errors. Fourth, we did not explore the impact of clinicopathological characteristics on long-term survival outcomes since only three of the included studies reported relevant data, leading to limitations and unreliability in the results. Finally, as mentioned before, the eCura system was recommended in the latest version of gastric cancer treatment guidelines; however, none of the studies evaluated patients with eCuraC-1 and 2.

In conclusion, our meta-analysis showed that additional gastrectomy offers better long-term survival outcomes than non-gastrectomy treatments after non-curative ESD for EGC, by significantly prolonging the 5-year OS, DSS, and DFS. In addition, elderly patients can benefit from additional surgery considering the 5-year OS. Furthermore, prospective, randomized, and multicenter studies are warranted to confirm these findings.

Funding

This study was supported by a grant from the National Natural Science Foundation of China (No. 82070653).

Conflicts of interest

None.