Long-term outcomes of additional surgical versus nonsurgical treatment after non-curative endoscopic resection for T1 superficial esophageal cancer.

Introduction
Esophageal cancer is among the most prevalent malignant tumors globally. According to the 2020 global cancer statistics, there were approximately 604 000 new cases and 544 000 deaths attributed to esophageal cancer. The high incidence and mortality rates pose a significant threat to human health and impose a substantial disease burden[1]. Early-stage esophageal cancer often lacks distinctive clinical symptoms, leading to delayed diagnosis; approximately 90% of diagnosed patients present with locally advanced or metastatic disease, resulting in poor overall prognosis. The 5-year age-standardized net survival rate ranges from 10% to 30%[2]. Consequently, early diagnosis and timely treatment are crucial for improving patient outcomes and survival rates.
With advancements in endoscopic technology, endoscopic resection (ER) has emerged as a mainstream therapeutic approach due to its minimally invasive nature and efficacy. Studies from Japan, South Korea, and Europe have demonstrated that ER can achieve curative results comparable to surgery for early-stage esophageal cancer confined to the mucosa or submucosa without lymph node metastasis (LNM)[3–6], with a 5-year survival rate reaching up to 95%[7]. However, Post-ER pathological findings show that a small proportion of patients experience non-curative resections, characterized by positive margins, poorly differentiated tumors, lymphovascular invasion, or submucosal involvement >200 μm[3–5,8]. These high-risk groups have a significantly increased risk of LNM and esophageal cancer recurrence. According to the Japanese Classification of Esophageal Cancer, 11th Edition, Clinical T1NxMx denotes superficial esophageal cancer (SEC)[9]. Research shows LNM incidence in T1 SEC increases with tumor depth: m1–m2 (0.4–5.6%), m3 (8.7–18.0%), sm1 (7.7–53.1%), and sm2/3 (36.2–53.9%)[10,11]. Low-risk curative resections post-ER correlate with an LNM incidence of 1–3%[3], while non-curative resections, especially with lymphovascular invasion, exhibit a higher LNM rate of 20–30%. This rate drops to 5% after additional surgery[3,8]. Consequently, Guidelines recommend additional radical esophagectomy or chemoradiotherapy for patients with pT1a non-curative resection or pT1b status[3-6].
There is limited research on the quantified benefits of additional surgical resection (ASR) versus non-ASR treatments in high-risk patients with T1 SEC post-ER. This scarcity is due to the rarity of such cases, the long follow-up period required for RCTs, and challenges in randomizing these patients in real-world settings. Some single-center studies suggest better survival outcomes with ER followed by surgery compared to ER combined with chemoradiotherapy, but these studies have small sample sizes and weak evidence quality. Existing meta-analyses, mostly based on retrospective studies, lack quantitative analyses of specific treatment benefits, particularly disease-specific survival (DSS), which excludes other non-tumor factors of cause of death, and is the most representative indicator of a patient’s tumor-related cause of death. Therefore, we conducted a comprehensive systematic review to quantify the short-term, medium-term, and long-term survival advantages of ASR over non-ASR treatments. This study has been reported in accordance with the TITAN standards[12], and artificial intelligence was not used in the research or writing of this manuscript.

Material and methods

Data sources and search strategy
All data included in the current study are publicly available in the PubMed/MEDlINE, Embase, Web of Science and Cochrane Library databases. Two authors independently conducted computer searches of PubMed/MEDlINE, Embase, Web of Science, and Cochrane Library data from database creation to 24 February 2025. MeSH search terms included esophageal neoplasms; endoscopic mucosal resection (EMR); Esophagectomy/Surgical Procedures, Operative. The detailed search strategy is presented in Supplementary Digital Content Table 1, available at: http://links.lww.com/JS9/F118. Then the retrieved articles are screened according to the exclusion criteria. This systematic review and meta-analysis was reported according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020[13] and AMSTAR (Assessing the methodological quality of systematic reviews) guidelines[14]. The present systematic review and meta-analysis was registered in PROSPERO.

Inclusion and exclusion criteria

Inclusion criteria
(1) The study subjects were high-risk patients with T1-stage SEC after endoscopic non-curative resection.
(2) Treatment includes two different interventions: ASR and non-ASR: Patients undergoing thoracoscopic or open esophagectomy and lymphadenectomy were classified as the ASR group, and patients undergoing radiotherapy and chemotherapy or close follow-up after ER were classified as the non-ASR group. ER was defined as EMR or endoscopic submucosal dissection (ESD).
(3) The outcome indicators in this paper were as follows: overall survival (OS), DSS and recurrence-free survival (RFS) in both groups.
(4) Study design: randomized controlled trials (RCTs) or observational studies including cohort and case–control studies.
(5) The article was published in English.
It should be added that progression-free survival, disease-free survival were classified as RFS, and cause-specific survival were classified as DSS.

Exclusion criteria
(1) Unable to extract the Kaplan–Meier curve of the two groups of patients;
(2) Repeated publications;
(3) Case reports, conference minutes, nonclinical studies, reviews, and meta-analyses;
(4) Literature with incomplete data or difficult data extraction.
(5) If there was duplication of data cited in different literature, the literature with the smaller sample size was eliminated and the one with the larger sample size was selected.
Finally, some references of relevant systematic reviews and meta-analyses were searched to supplement the articles that may have been missed. The specific retrieval process is shown in Figure 1.

Data extraction and processing
Two researchers independently extracted key information from the included literature. The details covered article titles, author names, publication years, study designs, patient characteristics (age, sex, tumor size, stage, histology, tumor location, lymphovascular invasion, and resection margins), treatment regimens, and outcome measures (OS, DSS, and recurrence-free survival [RFS]). Additionally, we compiled the definitions of non-curative resection from each study, as detailed in Supplementary Digital Content Table 2, available at: http://links.lww.com/JS9/F119. Discrepancies were resolved by a third investigator. Risk of bias in observational studies were assessed by two reviewers using the Newcastle–Ottawa Scale[15]. Any doubts were discussed and resolved within the author group.
If hazard ratios (HRs) for death comparing the non-ASR group to the ASR group were not provided, we used Engauge Digitizer software to extract survival data from Kaplan–Meier curves following the previously reported methods[16–19]. We then refitted the survival curves and calculated the HRs and 95% confidence intervals (CIs) using Stata (version 17.0). OS, RFS, and DSS data were aggregated, and an integrated Kaplan–Meier curve was fitted to derive survival rates at different time points. Finally, differences between the two groups were analyzed.

Statistical analysis
We used Stata (version 17.0) for statistical analysis. The Engauge Digitizer 11.3 software was used to extract data from the survival curves, compared study results by HR and 95% CI, and used Cochran Q test and I2 test to assess inter-study heterogeneity, with I2 > 50% indicating moderate to high heterogeneity[20,21]. Fixed-effects models were used to control for potential clinical and statistical heterogeneity across studies. Meta-regression and subgroup analysis were used to explore potential confounding factors in HR, and publication bias was assessed by funnel plots and Egger’s test. P < 0.05 was considered statistically significant.

Results
A total of 17 papers were retrieved through a systematic search with reference to the literature inclusion criteria. However, one of them, Yamamoto et al published a multicenter retrospective cohort study[22] covering 65 hospitals partially overlapped with the data in nine other papers[23–31], and the sample size included in this paper was smaller than the total sample size of these nine papers. Similarly, the data of the two articles published by Suzuki et al in 2018 and 2022 were all from the same hospital[32,33], which partially overlapped. So, in order to avoid the bias of results caused by duplicated data extraction and to maximize the clinical data, it was decided to include literature with relatively large sample sizes[23–31,33] and to discard Suzuki et al’s published in 2018 as well as Yamamoto et al’s published papers[22,32]. Finally a total of 15 retrospective studies were included. Among these studies, 11 of the 15 included studies originated from Japan, 2 from China, 1 from South Korea, and 1 from the United Kingdom. The heterogeneity among the studies was minimal. Given the high prevalence of squamous cell carcinoma in East Asia, coupled with the widespread adoption of ESD techniques in this region, studies from these areas were more numerous and contributed the majority of the data in this research. These studies included 13 on OS, 10 on RFS, and 8 on DSS, involving a total of 1348 patients (424 in the ASR group and 924 in the non-ASR group). Fifteen retrospective studies with a pooled median follow-up of 62 months (IQR, 48–79; maximum, 150 months) met the eligibility criteria. Study characteristics and quality assessments are summarized in Table 1. All studies scored at least 6 points on the Newcastle–Ottawa Scale. Tumor characteristics among patients in the included studies are shown in Table 2.

Overall survival
This analysis included 13 articles[23–30,33–37]. These studies, with varying follow-up periods, showed that patients in the ASR group had a significantly higher OS rate compared to those in the non-ASR group (HR: 2.16, 95% CI: 1.46–3.19; P < 0.001, I2 = 9.97%; n = 13) (Fig. 2A).

The meta-analysis showed no significant difference in 3-year OS between the two groups (HR: 1.72, 95% CI: 0.96–3.09, P = 0.070, I2 = 0.00%, n = 13) (Fig. 2B). However, the non-ASR group had significantly higher mortality risks at 5 years (HR: 1.76, 95% CI: 1.14–2.72, P = 0.010, I2 = 0.76%, n = 13) (Fig. 2C) and 10 years (HR: 2.14, 95% CI: 1.26–3.65, P = 0.010, I2 = 0.00%, n = 4) (Fig. 2D).
We extracted individual patient survival data from the survival curves in the studies and reconstructed a Kaplan–Meier curve (Fig. 3). The results showed that at 3 years, the OS rates were 95.9% (95% CI: 93.3–97.5%) for the ASR group and 88.0% (95% CI: 85.5–90.1%) for the non-ASR group, P < 0.001. At 5 years, the rates were 91.4% (95% CI: 87.8–94.0%) and 78.2% (95% CI: 75.0–81.1%), P < 0.001. At 10 years, they were 86.7% (95% CI: 80.4–91.1%) and 56.9% (95% CI: 51.3–62.1%), P < 0.001. These findings indicate an 8–30% relative improvement in OS for the ASR group compared to the non-ASR group at 3, 5, and 10 years, with the benefit increasing over time.

To account for potential confounding factors, we conducted a multivariate regression analysis incorporating patient age, gender, tumor characteristics (size, depth of invasion, lymphatic and vascular invasion, and margin status) (Table 3). The results indicated that none of these factors significantly influenced the HRs (HRs) between the two groups (all P > 0.05). Additionally, we assessed publication bias using a funnel plot, which revealed no evidence of significant publication bias among the included studies (Supplementary Digital Content Fig. 1, available at: http://links.lww.com/JS9/F117, Egger test P = 0.168).

Recurrence-free survival
A total of 10 studies[23–26,28,30,31,33,36,37] were included in this analysis. A comprehensive pooled analysis of these studies, which encompassed varying follow-up periods, demonstrated that the risk of tumor recurrence was significantly higher in the non-ASR group compared to the ASR group (HR:1.68, 95% CI: 1.16–2.41, P = 0.010, I2 = 0.00%; n = 10) (Fig. 4A).

The meta-analysis results showed a significantly higher risk of tumor recurrence in the non-ASR group compared to the ASR group for 3-year RFS (HR: 2.09, 95% CI: 1.21–3.63, P = 0.010, I2 = 0.00%, n = 10), 5-year RFS (HR: 1.59, 95% CI: 1.08––2.34, P = 0.020, I2 = 0.00%, n = 10), and 10-year RFS (HR: 1.76, 95% CI: 1.13–2.76, P = 0.010, I2 = 0.00%, n = 3) (Fig. 4B-D).
To analyze the RFS data from the literature, we reconstructed a Kaplan–Meier curve (Fig. 5). The results demonstrated that at 3 years, the RFS rates were 93.4% (95% CI: 89.4–95.9%) in the ASR group and 82.9% (95% CI: 79.9–85.5%) in the non-ASR group (P < 0.001). At 5 years, the RFS rates were 83.6% (95% CI: 77.7–88.0%) for the ASR group and 73.8% (95% CI: 70.2–77.0%) for the non-ASR group (P < 0.001). At 10 years, the RFS rates were 74.5% (95% CI: 61.4–83.7%) for the ASR group and 57.5% (95% CI: 52.0–62.5%) for the non-ASR group (P = 0.022). These findings indicate that the differences in RFS between the ASR and non-ASR groups were approximately 10–17% at 3, 5, and 10 years, with the ASR group showing a progressively greater survival benefit as follow-up time increased.

To account for potential confounding factors, we conducted a multivariate regression analysis incorporating patient age, gender, tumor characteristics (size, depth of invasion, lymphatic and vascular invasion, and margin status) (Table 4). The results indicated that none of these factors significantly influenced the HRs between the two groups (all P > 0.05). Additionally, we assessed publication bias using a funnel plot, which revealed no evidence of significant publication bias among the included studies (Supplementary Digital Content Fig 1, available at: http://links.lww.com/JS9/F117. Egger test P = 0.253).

Disease-specific survival
A total of eight studies[23,25,27,29,30,34,37,38] were included in this analysis. A pooled analysis of these studies, which had varying follow-up periods, revealed that patients in the non-ASR group had a significantly higher risk of cancer-related death compared to those in the ASR group (HR: 2.83, 95% CI: 1.07–7.44, P = 0.040, I2 = 0.00%; n = 8) (Fig. 6A).

The meta-analysis revealed no significant difference in cancer-related mortality risk between the ASR and non-ASR groups for both 3-year (HR: 2.25, 95% CI: 0.75–6.72, P = 0.150, I2 = 0.00%, n = 8) and 5-year (HR: 2.61, 95% CI: 0.99–6.90, P = 0.050, I2 = 0.00%, n = 8) DSS (Fig. 6B, C). That is to say, within the short-term follow-up period of 3–5 years, no statistically significant difference was observed in the net benefit of DSS between the ASR and non-ASR groups.
To analyze the DSS data from the included studies, we reconstructed Kaplan–Meier curves (Fig. 7). The results showed that at 3 years, the DSS rates were 98.4% (95% CI: 95.7–99.4%) for the ASR group and 94.5% (95% CI: 92.3–96.1%) for the non-ASR group (P = 0.021). At 5 years, the rates were 97.8% (95% CI: 94.7–99.1%) and 91.1% (95% CI: 88.3–93.2%), respectively (P = 0.002). At 10 years, the rates were 97.8% (95% CI: 94.7–99.1%) and 87.9% (95% CI: 84.0–90.8%), respectively (P = 0.199). These findings indicate a 3–10% net benefit in DSS for the ASR group compared to the non-ASR group at 3, 5, and 10 years. Notably, within the short-term follow-up (3–5 years), the absolute difference in DSS was modest, ranging from approximately 4 to 6 percentage points.

To account for potential confounding factors, we conducted a multivariate regression analysis incorporating patient age, gender, tumor characteristics (size, depth of invasion, lymphatic and vascular invasion, and margin status) (Table 5). The results indicated that none of these factors exhibited statistical significance and did not influence the HRs between the two groups (all P > 0.05). Additionally, we assessed publication bias using a funnel plot. The analysis revealed significant publication bias among the studies (Supplementary Digital Content Fig. 1, available at: http://links.lww.com/JS9/F117 Egger’s test P = 0.020). After applying the trim and fill method to adjust for this bias, the adjusted results remained consistent with our original conclusions.

Stability and heterogeneity analysis of meta-conclusions
This study strictly followed the PRISMA 2020 and AMSTAR guidelines for systematic reviews and meta-analyses. Literature searches adhered to the PICO (Participants, Interventions, Comparisons, and Outcomes) framework, and studies were screened using predefined inclusion and exclusion criteria. Fifteen retrospective studies were included in the analysis. For statistical analysis, we employed established methods to calculate pooled HRs and used a fixed-effects model for the meta-analysis. This approach minimizes potential biases from individual studies and provides more robust conclusions. Heterogeneity was assessed using the Cochran Q test and I2 statistic, showing minimal heterogeneity for OS(I2 = 9.97%), RFS(I2 = 0.00%), and DSS(I2 = 0.00%) between the two groups. The observed low heterogeneity may stem from the limited number of studies included and their relatively small sample sizes, potentially reducing the sensitivity of the heterogeneity tests. A subsequent sensitivity analysis identified no outliers, suggesting the results are robust and reliable (Supplementary Digital Content Fig. 2, available at: http://links.lww.com/JS9/F117). Overall, considering the relatively low incidence of endpoint events observed in the included studies and the minimal heterogeneity detected among studies in further sensitivity analyses, the inter-study consistency was deemed adequate. Therefore, we opted for a fixed-effects model for the summary analysis in the final systematic meta-analysis.
Regression analyses of potential confounders, including age, gender, tumor size, depth of invasion, lymphatic invasion, vascular invasion, and margin status, did not significantly influence HRs for OS, RFS, or DSS. The primary reasons are (1) Incomplete information prevents obtaining detailed information, making direct analysis impossible; (2) The data are indirect and estimated, analyzed indirectly through compositional methods, and the meta-regression method is also an indirect assessment method. This series of indirect assessments further reduces statistical power. Publication bias was evaluated using Egger’s test, revealing significant bias for DSS (P = 0.020) but not for OS or RFS (P = 0.168, P = 0.253). Given the limited number of included studies, conventional funnel plot-based tests possess limited statistical power. Consequently, any nonsignificant Egger’s results may underestimate the true risk of publication bias. After adjusting for publication bias using the trim-and-fill method, the results remained consistent, confirming the overall conclusions.

Discussion
ER is preferred for early-stage esophageal cancer due to its minimally invasive nature and efficacy. However, non-curative resections, marked by factors like lymphovascular invasion or positive margins, increase the risk of LNM in high-risk patients. Studies show that ASR can significantly reduce metastasis rates in these cases[3,8]. Guidelines recommend additional surgery for T1 SEC patients with non-curative resections to improve prognosis[3–6], but it is unclear if all patients benefit equally. Further investigation is needed to compare the advantages of additional surgery over conservative management.
To date, several studies have examined the risk factors for recurrence and metastasis in patients with T1-stage SEC after ER. Some retrospective analyses have also evaluated long-term survival outcomes following various adjuvant treatments post-ER. These findings suggest that additional surgical treatment yields better outcomes than conservative management or no treatment. However, the specific survival benefits of different adjuvant therapies for high-risk patients with T1 SEC remain unclear. Existing research has not adequately quantified the real-world benefits of additional surgery across short-, medium-, and long-term horizons. DSS provides a more accurate reflection of tumor-related mortality compared to OS and RFS. Given the limited number of studies on the real-world benefits of DSS, we conducted a systematic review and meta-analysis to explore the advantages of additional surgical treatment for high-risk patients and to quantify the survival benefits of ASR versus non-ASR over different periods.
In this study, we adhered to stringent inclusion and exclusion criteria and conducted rigorous screening, ultimately selecting 15 articles covering 1348 patients. Results showed that the ASR group had significant benefits in OS, RFS, and DSS compared to the non-ASR group across short-, medium-, and long-term follow-ups. The survival benefit of ASR became more pronounced as patient survival time increased. This supports current guidelines recommending additional surgery for high-risk patients with T1-stage SEC[3–6]. However, OS is limited by confounding factors like selection bias, where surgically treated patients tend to be younger and healthier, while those choosing nonsurgical options are often older or have underlying conditions that preclude surgery. DSS, which excludes non-tumor-related deaths and mitigating selection bias, provides a more accurate reflection of ASR’s true survival benefits. Therefore, we conducted a quantitative analysis of DSS and found that the difference in DSS rates between the two groups was modest, particularly within the short to medium-term (3–5 years) observation period. Specifically, the net survival benefit of the ASR group over the non-ASR group was approximately 4–6%. The quantified net survival benefits from this study can serve as a valuable reference for clinicians and patients when making personalized treatment decisions, particularly for elderly patients with multiple comorbidities.
From the meta-analysis of OS, RFS, and DSS, ASR generally provides significant benefits to patients. Our quantitative analysis shows that across short-, medium-, and long-term periods, the ASR group demonstrated varying degrees of benefit compared to the non-ASR group. As survival time increased, the benefits of additional surgery became more pronounced. This aligns with previous studies, which reported that the 5-year OS rate after additional esophagectomy following ER ranges from 90% to 100%, while that after additional chemoradiotherapy is between 75% and 85%[39,40]. The relatively poorer survival outcomes observed in patients undergoing chemoradiotherapy may be related to patient selection bias. Patients with better general baseline conditions are more likely to tolerate the trauma of surgical intervention, while those with relatively poorer baseline conditions may be more likely to be assigned to the conservative plus chemoradiotherapy group. These patients may experience lower OS due to their poorer baseline conditions and more complex comorbidities, potentially leading to death from other underlying diseases. For OS, while the ASR group showed higher OS, the difference was not statistically significant at 3 years (HR: 1.72, 95% CI: 0.96–3.09, P = 0.070, I2 = 0.0%, n = 13), possibly due to high early postoperative complications and mortality in the ASR group during the early postoperative period. Regarding tumor recurrence reduction, the ASR group showed approximately a 10% net benefit over the non-ASR group within 5 years. For DSS, no significant differences were observed in cancer-related death risk between the ASR and non-ASR groups at 3 years (HR: 2.25, 95% CI: 0.75–6.72, P = 0.150, I2 = 0.0%, n = 8) or 5 years (HR: 2.61, 95% CI: 0.99–6.90, P = 0.050, I2 = 0.0%, n = 8). Further quantitative analysis revealed that the net survival benefit of the ASR group compared to the non-ASR group was limited, with a difference of approximately 4–10% in DSS rates at 3, 5, and 10 years. The maximum difference in survival rates between the two groups was approximately 10%, and this difference only became apparent after more than 10 years of observation. Based on this findings, while additional surgical intervention may offer some benefit, achieving these significant advantages requires a relatively long observation period. Consequently, the decision to pursue additional surgery should involve a more careful, individualized approach, considering factors such as the patient’s overall condition, comorbidities, and life expectancy. Indiscriminate application of additional surgery may expose a subset of patients to overtreatment.
Meanwhile, perioperative risk associated with esophagectomy is substantial: multicenter studies and systematic reviews report postoperative pulmonary complications in 25–40% of patients[41,42], anastomotic-leak rates of 10–33%[41,42], atrial fibrillation in 16–25%[43], and 30- and 90-day mortality of 1–3%[44,45] and 4–9%[42,44,45], respectively. When severe pneumonia or an anastomotic leak occurs, the risk of death can rise to 7–15%[41,45], accompanied by markedly higher hospital costs and readmission rates[43]. In other words, the potential long-term benefit of surgery is may offset by the high incidence of perioperative adverse events. For high-risk elderly patients with T1 squamous cell carcinoma, poor baseline health and limited life expectancy make the “risk–benefit inversion” particularly clear: radical resection, while theoretically able to eliminate early-stage disease, is unlikely to extend survival due to limited cardiopulmonary reserve and a high risk of complications. It may instead cause functional decline, lower quality of life, or even early death[41–45]. Consequently, surgical indications in this population should be weighed with greater caution, fully considering nonsurgical options and patient preferences to achieve truly patient-centered, precision decision-making.
While our study findings are robust and reliable, several limitations must be acknowledged. First, due to the lack of original data such as HRs in most studies, we were unable to directly obtain raw data for more precise and in-depth analyses, so we had to extract data indirectly from Kaplan–Meier curves to estimate HR values. This process may introduce errors and biases, affecting result quality. Additionally, the reconstructed Kaplan–Meier curves were based on predicted rather than actual statistical values, and the timing of events in these curves may not accurately reflect the true event occurrence times, potentially weakening evidence quality. Second, the meta-analysis had a relatively small sample size and insufficient follow-up duration, limiting its scope and depth. Conducting large-scale, long-term studies is challenging and costly, and relevant data are limited. Third, due to inconsistent reporting of certain baseline characteristics (e.g., baseline performance status, comorbidity burden, and pulmonary function) across the 15 included studies, quantitative adjustment was not feasible. Consequently, the potential influence of residual confounding should be considered when interpreting survival differences between the groups. Moreover, all included studies were retrospective, leading to lower evidence quality. Ethical considerations also make RCTs difficult for cancer patients, as clinical decisions for cancer patients are often made by the patients themselves, impacting research quality. Given that the majority of studies included in this analysis were conducted in East Asia and primarily focused on esophageal squamous cell carcinoma (ESCC), with insufficient representation of adenocarcinoma, future research should incorporate more comprehensive data on adenocarcinoma from diverse geographical regions. This approach will enhance the generalizability and robustness of subsequent analytical investigations. Finally, given the retrospective nature of our study, we only analyzed OS, DSS, and RFS in the nonsurgical and surgical patient groups. In future studies, we need large-sample, multicenter data to further explore the relevant influencing factors of surgical selection and to investigate factors associated with recurrence and mortality. We also aim to establish relevant predictive models, as demonstrated in the study by Vanstraelen et al[46]. Currently, several studies[47,48] have proven the feasibility of translating clinical knowledge into predictive models, which, to some extent, can guide patients in making personalized clinical treatment decisions. This suggests that future research should focus on factors related to patients’ treatment decisions regarding ASR or non-ASR approaches. Despite these limitations, our study has comprehensively analyzed available data, maximizing the extraction and synthesis of information. The overall conclusions remain consistent, and the quantitative analysis results are within a similar range. Notably, a phase III RCT comparing additional esophagectomy with chemoradiotherapy for patients with cN0-pT1b ESCC post-ESD has been conducted[49], providing further insights into whether conservative treatment (chemoradiotherapy) is more beneficial for such patients. Most included studies are from single Asian centers, potentially introducing selection bias. We encourage large-scale, multicenter studies to address this clinical issue and look forward to open-source data platforms for sharing original data and more RCTs to provide higher-quality evidence-based medical evidence.

Conclusion
In summary, for high-risk patients with T1-stage SEC, the ASR group shows limited net benefit (4–6%) in DSS compared to the non-ASR group within the first 5 years. Nonsurgical treatment, balancing efficacy and low invasiveness while maintaining postoperative quality of life, may be suitable for special populations like elderly patients with comorbidities or limited life expectancy. This study aims to provide valuable reference for clinicians treating these patients.

Supplementary Material